Basic Number Theory

Andre Weil was born May 6, 1906 in Paris.
After
studying mathematics at the Ecole Normale
Superieure and receiving a doctoral degree from
the University of Paris in 1928, he held professorial
positions in India, France, the United States and
Brazil before being appointed to the Institute
for Advanced Study, Princeton in 1958 where he
is now Professor Emeritus.
Andre Weil's work laid the foundation for abstract

algebraic geometry and the modern theory of
abelian varieties. A great deal of his work was
directed towards establishing the links between
number theory and algebraic geometry and
devising modern methods in analytic number
theory. Weil was one of the founders, around 1934,
of the group that published, under the collective
name of N. Bourbaki, the highly influential multi-
volume treatise Elements de mathematique.
Classics in Mathematics
Andre Weil Basic Number Theory
Andre Weil
Basic Number Theory

Reprint of the 1974 Edition
Springer
Andre Wei!
Institute for Advanced Study
Princeton. NJ 08540
USA
Originally published as Vol. 144 of the

Grundlehren der mathematischen Wissenschaften
Mathematics Subject Classification (1991): llR
ISBN-13: 978-3-540-58655-5 e-ISBN-13: 978-3-642-61945-8

DOl: 10.1007/978-3-642-61945-8
Photograph by kind permission of Lucien Gillet
elP data applied for
This work is subject to copyright. All rights are reserved, whether the whole or part of the material is
concerned. specifically the rights of translation. reprinting. reuse of illustration, recitation. broadcasting.
reproduction on microfilm or in any other way. and storase in data banks. Duplication of this publication
or parts thereof is permitted only under the provision of the German Copyright Law of September 9. 1965.
in its current version. and permission for use must always be obtained from Springer-Verlag. Violations are
liable for prosecution under tbe German Copyright Law.
@Springer-VerlasBerIinHeideiberg 1995
The use of general descriptive names, registered names. trademarks. etc. in this publication does not imply.
even in the absence of a specific statement, that such names are exempt from the relevant protective laws
and regulations and therefore free for general use.
SPIN 10701941 4113111 - 5 4 3 21 - Printed on acid-free paper

Andre Weil
Basic Number Theory
Third Edition 1974
Springer-Verlag Berlin • Heidelberg • New York

1974
Professor Andre Weil
The Institute for Advanced Study, Princeton, N. J. 08540
All rights, especially that of translation into foreign languages, reserved. It. is also
forbidden to reproduce this book, either whole or in part, by photomechanical
means (photostat, microfilm and/or microcard) or by other procedure without
written permission from Springer-Verlag
© 1973 by Springer-Verlag New York Inc.
Library of Congress Catalog Card Number 73-75018
Foreword
)tPI(}jlOV, e~oxov (10CPUljlr1.'CWV
Aiux., llpop.. .dsup..
The first part of this volume is based on a course taught at Princeton

University in 1961-62; at that time, an excellent set of notes was prepared
by David Cantor, and it was originally my intention to make these notes
available to the mathematical public with only quite minor changes.
Then, among some old papers of mine, I accidentally came across a
long-forgotten manuscript by Chevalley, of pre-war vintage (forgotten,
that is to say, both by me and by its author) which, to my taste at least,
seemed to have aged very well. It contained a brief but essentially com-
plete account of the main features of classfield theory, both local and
global; and it soon became obvious that the usefulness of the intended
volume would be greatly enhanced if I included such a treatment of this
topic. It had to be expanded, in accordance with my own plans, but its
outline could be preserved without much change. In fact, I have adhered
to it rather closely at some critical points.
To improve upon Hecke, in a treatment along c1assicallines of the
theory of algebraic numbers, would be a futile and impossible task. As
will become apparent from the first pages of this book, I have rather
tried to draw the conclusions from the developments of the last thirty
years, whereby locally compact groups, measure and integration have
been seen to play an increasingly important role in classical number-
theory. In the days of Dirichlet and Hermite, and even of Minkowski,
the appeal to "continuous variables" in arithmetical questions may well
have seemed to come out of soine magician's bag of tricks. In retrospect,
we see now that the real numbers appear there as one of the infinitely
many completions of the prime field, one which is neither more nor less
interesting to the arithmetician than its p-adic companions, and that
there is at least one language and one technique, that of the adeles, for
bringing them all together under one roof and making them cooperate
for a common purpose. It is needless here to go into the history of these
developments; suffice it to mention such names as Hensel, Hasse,
Chevalley, Artin; everyone of these, and more recently Iwasawa, Tate,
Tamagawa, helped to make some significant step forward along this
road. Once the presence of the real field, albeit at infinite distance, ceases
to be regarded as a necessary ingredient in the arithmetician's brew, it
VI Foreword
goes without saying that the function-fields over finite fields must be
granted a fully simultaneous treatment with number-fields, instead of
the segregated status, and at best the separate but equal facilities, which
hitherto have been their lot. That, far from losing by such treatment,
both races stand to gain by it, is one fact which will, I hope, clearly emerge
from this book.
It will be pointed out to me that many important facts and valuable
results about local fields can be proved in a fully algebraic context,
without any use being made of local compacity, and can thus be shown
to preserve their validity under far more general conditions. May I be
allowed to suggest that I am not unaware of this circumstance, nor of
the possibility of similarly extending the scope of even such global results
as the theorem ofRiemann-Roch? We are dealing here with mathematics,
not with theology. Some mathematicians may think that they can gain
full insight into God's own way of viewing their favorite topic; to me,
this has always seemed a fruitless and a frivolous approach. My intentions
in this book are more modest. I have tried to show that, from the point
of view which I have adopted, one could give a coherent treatment,
logically and aesthetically satisfying, of the topics I was dealing with.
I shall be amply rewarded if I am found to have been even moderately
successful in this attempt.
Some of my readers may be surprised to find no explicit mention of
cohomology in my account of classfield theory. In this sense, while my
approach to number-theory may be called a "modern" one in the first
half of this book, it may well be described as thoroughly "unmodern" in
the second part. The sophisticated reader will of course perceive that a
certain amount of cohomology, and in fact no more and no less than is
required for the purposes of classfield theory, hides itself in the theory
of simple algebras. For anyone familiar with the language of "Galois
cohomology", it will be an easy and not unprofitable exercise to translate
into it some of the definitions and results of our Chapters IX, XII and
XIII; in one or two places (the most conspicuous case being that ofthe
"transfer theorem" in Chapter XII, § 5), this even makes it possible to
substitute more satisfactory proofs for ours. For me to develop such an
approach systematically would have meant loading a great deal of
unnecessary machinery on a ship which seemed well equipped for this
particular voyage; instead of making it more seaworthy, it might have
sunk it.
In charting my course, I have been careful to steer clear of the arith-
metical theory of algebraic groups; this is a topic of deep interest, but
obviously not yet ripe for book treatment. Partly for this reason, I have
refrained from discussing zeta-functions of simple algebras beyond what
was needed for the sake of classfield theory. Artin's non-abelian L-func-
Foreword VII
tions have also been excluded; the reader of this book will find it easy
to proceed to the study of Artin's beautiful papers on this subject and
will find himself well prepared to enjoy them, provided he has some
knowledge of the representation theory of finite groups.
It remains for me to discharge the pleasant duty of expressing my
thanks to David Cantor, who prepared from my lectures at Princeton
University the set of notes which reappears here as Chapters I to VII
of this book (in many places with no change at all), and to Chevalley,
who generously allowed me to make use of the above-mentioned manus-
cript and expand it into Chapters XII and XIII. My thanks are also
due to Iwasawa and Lazard. who read the book in manuscript and offered
many suggestions for its improvement; to H. Pogorzelski, for his assis-
tance in proofreading; to B. Eckmann, for the interest he took in its
publication; and to the staff of the Springer Verlag, and that of the
Zechnersche Buchdruckerei, for their expert cooperation and their
invaluable help in the process of bringing out this volume.
Princeton, May 1967. ANDRE WElL
Foreword to the second edition
The text of the first edition has been left unchanged. A few correc-
tions, references, and some brief remarks, have been added as Notes at
the end of the book; the corresponding places in the text have been
marked by a * in the margin. Somewhat more substantial additions will
be found in the Appendices, originally prepared for the Russian edition
(M. I. R., Moscow 1971). The reader's attention should be drawn to the
collective volume: J. W. S. Cassels and A. Frohlich (edd.), Algebraic
Number Theory, Acad. Press 1967, which covers roughly the same
ground as the present book, but with far greater emphasis on the cohomo-
logical aspects.
Princeton, December 1971. ANDRE WElL

Contents
Chronological table. . . . XII
Prerequisites and notations. XIII
Table of notations . . . . . XVII
PART I. ELEMENTARY THEORY
Chapter I. Locally compact fields. . . . . 1

§ 1. Finite fields . . . . . . . . . . . 1
§ 2. The module in a locally compact field 3
§ 3. Classification of locally compact fields . 8
§ 4. Structure of p.fields. . . . . . . . . 12
Chapter II. Lattices and duality over local fields 24

§ 1. Norms . . . . . . . . . . . . . 24
§ 2. Lattices . . . . . . . . . . . . . 27
§ 3. Multiplicative structure of local fields 31
§ 4. Lattices over R. . . . 35
§ 5. Duality over local fields . . . . . . 38
Chapter III. Places of A·fields . . . . . . . 43

§ 1. A-fields and their completions . . . . 43
§ 2. Tensor-products of commutative fields. 48
§ 3. Traces and norms . . . . . . . . . . 52
§ 4. Tensor-products of A-fields and local fields. 56
Chapter IV. Adeles . . . 59

§ 1. Adeles of A-fields . 59
§ 2. The main theorems 64
§ 3. Ideles. . . . . 71
§ 4. Ideles of A-fields . 75
x Contents
Chapter V. Algebraic number-fields. . 80

§ 1. Orders in algebras over Q . . . 80
§ 2. Lattices over algebraic number-fields. 81
§ 3. Ideals . . . . . . 85
§ 4. Fundamental sets. . . . . . . . . 89
Chapter VI. The theorem of Riemann-Roch 96
Chapter VII. Zeta-functions of A-fields . . 102

§ 1. Convergence of Euler products . . . 102
§ 2. Fourier transforms and standard functions. 104
§ 3. Quasicharacters. . . . . . 114
§ 4. Quasicharacters of A-fields. 118
§ 5. The functional equation . . 120
§ 6. The Dedekind zeta-function 127
§ 7. L-functions . . . . . . . 130
§ 8. The coefficients of the L-series 134
Chapter VIII. Traces and norms . . 139

§ 1. Traces and norms in local fields . 139
§ 2. Calculation of the different . 143
§ 3. Ramification theory. . . . . . 147
§ 4. Traces and norms in A-fields . . 153
§ 5. Splitting places in separable extensions. 158
§ 6. An application to inseparable extensions . 159
PART II. CLASSFIELD THEORY
Chapter IX. Simple algebras. . . . . . . . 162

§ 1. Structure of simple algebras . . . . . 162
§ 2. The representations of a simple algebra. 168
§ 3. Factor-sets and the Brauer group 170
§ 4. Cyclic factor-sets . . . 180
§ 5. Special cyclic factor-sets . . . . 185
Chapter X. Simple algebras over local fields 188

§ 1. Orders and lattices . . . . . 188
§ 2. Traces and norms. . . . . . 193
§ 3. Computation of some integrals 195
Contents XI
Chapter XI. Simple algebras over A-fields 202

§ 1. Ramification. . . . . . . . . . 202
§ 2. The zeta-function of a simple algebra 203
§ 3. Norms in simple algebras . . . . ' . 206
§ 4. Simple algebras over algebraic number-fields 210
Chapter XII. Local classfield theory. . . 213

§ 1. The formalism of classfield theory. 213
§ 2. The Brauer group of a local field . 220
§ 3. The canonical morphism. . . . . 226
§ 4. Ramification of abelian extensions. 230
§ 5. The transfer . . . . . . . . . 240
Chapter XIII. Global classfield theory. 244

§ I. The canonical pairing . . . 244
§ 2. An elementary lemma . . . 250
§ 3. Hasse's "law of reciprocity" 252
§ 4. Classfield theory for Q. . . 257
§ 5. The Hilbert symbol . . . . 260
§ 6. The Brauer group of an A-field 264
§ 7. The Hilbertp-symbol . . . . 267
§ 8. The kernel of the canonical morphism 271
§ 9. The main theorems . . . . . . . . 275
§ 10. Local behavior of abelian extensions 277
§ II. "Classical" classfield theory . 281
§ 12. "Coronidis loco". . . . . . . . . 288
Notes to the text. . . . . . . . . . 292

Appendix I. The transfer theorem . . 295
Appendix 11. W-groups for local fields. 298
Appendix III. Shafarevitch's theorem . 301
Appendix IV. The Herbrand distribution 308
Index of definitions. . . . . . . . . .
Chronological table
(In imitation of Hecke's "ZeittaJer' at the end of his "Theorie der
algebraischen Zahlen", and as a partial substitute for a historical survey,
we give here a chronological list of the mathematicians who seem to
have made the most significant contributions to the topics treated in
this volume.)
Fermat (1601-1665) Riemann (1826-1866)

Euler (1707-1783) Dedekind (1831-1916)
Lagrange (1736-1813) H. VVeber(1842-1913)
Legendre (1752-1833) Hensel (1861-1941)
Gauss (1777-1855) Hilbert (1862-1943)
Dirichlet (1805-1859) Takagi (1875-1960)
Kummer (1810-1893) Hecke (1887-1947)
Hermite (1822-1901) Artin (1898-1962)
Eisenstein (1823-1852) Hasse (1898- )
Kronecker (1823-1891) Chevalley (1909-
Prerequisites and notations
No knowledge of number-theory is presupposed in this book, except
for the most elementary facts about rational integers; it is useful but not
necessary to have some superficial acquaintance with the p-adic valua-
tions of the field Q of rational numbers and with the completions Qp
of Q defined by these valuations. On the other hand, the reader who
wishes to acquire some historical perspective on the topics treated in the
first part of this volume cannot do better than take up Hecke's unsur-
passed Theorie der algebraischen Zahlen, and, if he wishes to go further
back, the Zahlentheorie of Dirichlet-Dedekind (either in its 4th and final
edition of 1894, or in the 3rd edition of 1879), with special reference
to Dedekind's famous "eleventh Supplement". For similar purposes, the
student of the second part of this volume may be referred to Hasse's
Klassenkorperbericht (J. D. M. V., Part I, 1926; Part II, 1930).
The reader is expected to possess the basic vocabulary of algebra
(groups, rings, fields) and of linear algebra (vector-spaces, tensor-
products). Except at a few specific places, which may be skipped in a
first reading, Galois theory plays no role in the first part (Chapters I
to VIII). A knowledge of the main facts of Galois theory for finite and
for infinite extensions is an indispensable requirement in the second
part (Chapters IX to XIII).
Already in Chapter I, and throughout the book, essential use is made
of the basic properties of locally compact commutative groups, including
the existence and unicity of the Haar measure; the reader is expected to
have acquired some familiarity with this topic before taking up the
present book. The Haar measure for non-commutative locally compact
groups is used in Chapters X and XI (but nowhere else). The basic facts
from the duality theory of locally compact commutative groups are
briefly recalled in Chapter II, § 5, and those about Fourier transforms
in Chapter VII, § 2, and play an essential role thereafter.
As to our basic vocabulary and notations, they usually agree with
the usage of Bourbaki. In particular, this applies to N (the set of the
"finite cardinals" or "natural integers" 0,1,2, ...), Z (the ring ofrational
integers), Q (the field of rational numbers), R (the field of real numbers),
C (the field of complex numbers), H (the field of "classical", "ordinary"
or "Hamiltonian" quatemions). If p is any rational prime, we write Fp
for the prime field with p elements, Qp for the field of p-adic numbers
(the completion of Q with respect to the p-adic valuation; cf. Chapter I,
XIV Prerequisites and notations
§ 3), Zp for the ring of p-adic integers (i. e. the closure of Z in Qp). The
fields R, C, H, Qp are always understood to be provided with their usual
(or "natural") topology; so are all finite-dimensional vector-spaces over
these fields. By Fq we understand the finite field with q elements when
there is one, i. e. when q is of the form pn, p being a rational prime and
n an integer ~ 1 (cf. Chapter I, § 1). We write R+ for the set of all real
numbers ~O.
All rings are assumed to have a unit. If R is a ring, its unit is written
l R , or 1 when there is no risk of confusion; we write R x for the multi-
plicative group of the invertible elements of R; in particular, when K is
a field (commutative or not), K X denotes the multiplicative group of
the non-zero elements of K. We write R~ for the multiplicative group
of real numbers >0. If R is any ring, we write MiR) for the ring of
matrices with n rows and n columns whose elements belong to R, and
we write In for the unit in this ring, i.e. the matrix (OJ) with oij= lR or 0
according as i = j or i =f. j. We write I X for the transpose of any matrix
X E M n(R), and tr(X) for its trace, i. e. the sum of its diagonal elements;
if R is commutative, we write det(X) for its determinant. Occasionally
we write Mm.n(R) for the set of the matrices over R with m rows and n
columns.
If R is a commutative ring, and T is an indeterminate, we write R[T]
for the ring of polynomials in T with coefficients in R; such a polynomial
is called monic if its highest coefficient is 1. If S is a ring containing R,
and x an element of S commuting with all elements of R, we write R[x]
for the subring of S generated by R and x; it consists of the elements of
S of the form F(x), with FER[TJ. If K is a commutative field, L a field
(commutative or not) containing K, and x an element of L commuting
with all elements of K, we write K(x) for the subfield of L generated by
K and x; it is commutative. We do not speak of a field L as being an
"extension" of a field K unless both are commutative; usually this word
is reserved for the case when L is of finite degree over K, and then we
write [L: K] for this degree, i. e. for the dimension of L when L is regarded
as a vector-space over K (the index of a group g' in a group g is also
denoted by [g:g'] when it is finite; this causes no confusion).
All topologies should be understood to be Hausdorff topologies,
i. e. satisfying the Hausdorff "separation" axiom ("separated" in the sense
of Bourbaki). The word "homomorphism", for groups, rings, modules,
vector-spaces, should be understood with the following restrictions:
(a) when topologies are involved, all homomorphisms are understood to
be continuous; (b) homomorphisms of rings are understood to be .. uni-
tary"; this means that a homomorphism of a ring R into a ring S is assumed
to map lR onto Is. On the other hand, in the case of groups, homo-
morphisms are not assumed to be open mappings (i. e. to map open sets
Prerequisites and notations xv
onto open sets); when necessary, one will speak of an "open homo-
morphism". The word "morphism" is used as a shorter synonym for
"homomorphism"; the word "representation" is used occasionally, as a
synonym for "homomorphism", in certain situations, e. g. when the
homomorphism is one of a group into C x , or for certain homomorphisms
of simple algebras (cf. Chapter IX, § 2). By a character of a group G, com-
mutative or not, we understand as usual a homomorphism (or "represen-
tation") of G into the subgroup of C x defined by z z= 1; as explained
above, this should be understood to be continuous when G is given as
a topological group. The words "endomorphism", "automorphism",
"isomorphism" are subject to the same restrictions (a), (b) as "homo-
morphism"; for "automorphism" and "isomorphism", this implies, in
the topological case, that the mapping in question is bijective and bi-
continuous. Occasionally, when a mapping f of a set A into a set B,
both with certain structures (usually fields), determines an isomorphism
of A onto its image in B, we speak of it by "abuse of language" as an
"isomorphism" of A into B.
In a group G, an element x is said to be of order n if n is the smallest
integer ~ 1 such that x"=e, e being the neutral element of G. If K is a
field, an element of K x of finite order is called a root of 1 in K; in
accordance with a long-standing tradition, any root of 1 of order divid-
ing n is called an n-th root of 1 in K; it is called a primitive n-th root of
1 if its order is n. Thus the n-th roots of 1 in K are the roots of the
equation X" = 1 in K.
If a, b are in Z, (a,b) denotes their g.c.d., i.e. the element d ofN such
that dZ=aZ+ bZ. If R is any ring, the mapping n-+n'l R of Z into R
maps Z onto the subring Z·lR of R, known as "the prime ring" in R;
the kernel of the morphism n-+n-lR of Z onto Z'lR is a subgroup of Z,
hence of the form m' Z with meN; if R is not {O} and has no zero-divisor,
m is either 0 or a rational prime and is known as the characteristic of R.
If m=O, n-+n'lR is an isomorphism of Z onto Z·lR' by means of which
Z' 1R will frequently be identified with Z. If the characteristic of R is a
prime p> 1, the prime ring Z·1 R is isomorphic to the prime field Fp.
We shall consider left modules and right modules over non-commu-
tative rings, and fix notations as follows. Let R be a ring; let M and N
be two left modules over R. Then morphisms of Minto N, for their
structures as left R-modules, will be written as right operators on M; in
other words, if IX is such a morphism, we write it as m-+mIX, where meM;
thus the property of being a morphism, apart from the additivity, is
expressed by r(mIX)=(rm)IX for all reR and all meMo This applies in
particular to endomorphisms of M. Morphisms of right R-modules are
similarly written as left operators. This notation will be consistently
used, in particular in Chapter IX.
XVI Prerequisites and notations
As morphisms of fields into one another are assumed to be "unitary"

(as explained above), such morphisms are always injective; as we have
said, we sometimes refer to a morphism of a field K into a field L as an
"isomorphism", or also as an embedding, of K into L. In part of this
book, we use for such mappings the "functional" notation; beginning
with Chapter VIII, § 3, where the role of Galois theory becomes essential,
we shall use for them the "exponential" notation. This means that such
a mapping A is written in the former case as X-+A(X) and in the latter
case as x-+xA. If L is a Galois extension of K, and A, Il are two auto-
morphisms of Lover K, we define the law of composition (A,Il)-+AIl in
the Galois group g of Lover K as being identical with the law (A,Il)-+Aoll
in the former case, and as its opposite in the latter case; in other words,
it is defined in the former case by (AIl)X=A(llx), and in the latter case
by xAJl = (XA)Jl. For instance, if K' is a field between K and L, and ~ is the
corresponding subgroup of g, consisting of the automorphisms which
leave fixed all the elements of K', the automorphisms of Lover K which
coincide on K' with a given one A make up the right coset A~ when the
functional notation is used, and the left coset ~A when the exponential
notation is used.
When A, B, C are three additively written commutative groups
(usually with some additional structures) and a "distributive" (or "bi-
additive", or "bilinear") mapping (a, b)-+ab of Ax B into C is given,
and when X, Yare respectively subgroups of A and of B, it is customary
to denote by X· Y, not the image of X x Y under that mapping, but the
subgroup of C generated by that image, i. e. the group consisting of the
finite sums LXiYi with XiEX and YiE Y for all i. This notation will be
used occasionally, e.g. in Chapter V.
For typographical reasons, we frequently write exp(z) instead of eZ ,
and e(z) instead ofexp(2niz)=e 27tiZ , for ZEC; ordinarily e(z) occurs only
for ZER.
Finally we must explain the method followed for cross-references;
these have been inserted quite generously, with a view to helping the
inexperienced reader; the reader is advised to follow them up only when
the argument is not otherwise clear. Theorems have been numbered
continuously throughout each chapter; the same is true for propositions,
for lemmas, for definitions, for the numbered formulas. Each theorem
and each proposition may be followed by one or several corollaries.
Generally speaking, theorems are to be regarded as more important
than propositions, but the distinction between them would hardly stand
a close scrutiny. Lemmas are merely auxiliary results. Not all new con-
cepts are the object of a numbered definition; all concepts, except those
which are assumed to be known, are listed in the index at the end of the
book, with proper references. Formulas are numbered only for purposes
Prerequisites and notations XVII
of quotation, and not as an indication of their importance. When a

reference is given thus: "by prop. 2", "by corollary 1 of tho 3", etc., it
refers to a result in the same §; when thus: "by prop. 2 of § 2", "by tho 3
of § 3", etc., it refers to another § of the same chapter; when thus: "by
prop. 2 of Chap.IV-2", it refers to proposition 2 of Chapter IV, § 2.
Numbers of Chapter and § are given at the top of every page. A table of
the most frequently used notations is given below, in the order of their
first appearance.
Table of notations
Chapter I.
§ 2: modG' mod y , mod K •
§ 3: Ixlp, Ixl oo ' Qoo=R, lxiv, Qv (v = rational prime or 00).
§ 4: K (any p-field), R, P, n, q, ordK , ord, M", M.
Chapter II.
§ 3: 1 + P" (as subgroup of K" for n ~ 1).
§ 5: (g,g*)G' (g,g*), G*, H*, V*, L*, Y, [v,v']y, [v,v'], X, ord(x).
Chapter III.
§ 1: (for a place v of an A-field k) lxiv, kv, rv, Pv (for qv, see Chap. VII-I);
00 (as a place ofQ), wlv, Ev=E®kkv, ev, d v, (Xv.
§ 3: End(E), TrJilI'" NJilI'" Tr"'I'" N"'I"·
Chapter IV.
§ 1: P, P00' kA(P), kA' X, Xv, EA(P,e), E A, d A' d A(P,(X), (k'/k)A' (E/k)A.
§ 3: Aut(E), d;, d A(P, (x)" , lalA.
§ 4: kl, M, Q(P) = kA(P) x ,Ol(P), E(P).
Chapter V.
§ 2: koo' E oo , t, Lv.
§ 3: Vv, I(k), idea), P(k), h, 91(a).
§ 4: Idx /\ dxl, R, c".
Chapter VI.
deg(a), a ~ b, div(a), D(k), P(k), Do(k), g, div(x).
XVIII Table of notations
Chapter VII.
§ 1: qv, Ck(cf. § 6).
§ 2: n n
<P*, <Pv , (XV'
§ 3: Q(G), Q1' w•.
§ 4: n
Gk=kAx /k x , Q(GJ, w 1, w., G~, Q1' M, N, w v , w v , Z(w,<P).
§ 6: G1(s), G2 (s), Ck (cf. Chap. V-4), Gw(s), Ck(s), Zk(S).
§ 7: n
f(v), Sv' A, B, N v, <Pm, K= Kv' a=(a v), b=(bv), Gw , l(v), n v, U.s,w),
f,A (s,w). .
§ 8: Gp , l(P), D(P).
Chapter VIII.
§ 1: K, K', n, q, R, P, n, q', R', P', n', J, e, Tr, N, 91, d, D(K'/K), D, r'.
§ 2: 11.
§ 3: v(l), 9".
§ 4: n, r, 91k 'lk' 91, :no
Chapter IX.
§ 1: A L , A®B, AO.
§ 2: T, v.
§ 3: CI(A), B(K), K, K sep , (D,~, K', K', K;ep, (D', p, H(K).
§ 4: {X,O}, [L/K; X,OJ.
§ 5: Xn,~, g,O}n' Xp,~, g,O}p-
ChapterX.
§ 1: Hom(V, W), Hom(V,L; W,M), End(V,L), Aut(V,L).
§ 3: l:, l:', l:"; l:, U.
Chapter XII.
§ 1: Kab> (D(1), m:, X g, p, Gg, (X,g)K' n, Gi, U g, X 0' m: o, K o·
§ 2: h(A), 11, (x,O)g (for K =R,C); rol, Ko, ~o, K n, CPo, X o, cP, 11, (X,O)g,
n, h(A).
§ 3: Ug, ~o.
Chapter XIII.
§ 1: K", ksep, k",sep, kab' kv,ab' (D, m:, (D", ~'" p", Xk, x,,, (x",Z)v, (x,Z)k'
k,
n, k::' +, F, q, ko, ~o, X 0, kn , m: o, CPo, cp, Q, e, ~m' g, ~.
§ 3: hv(A), Uk'
§ 5: (X,Y)n,g, (z,z')n, Q(P) (cf. Chap. IV-4), Q'(P).
§ 7: (X,Z)p,K' <P, Q'(m,K), (x,z)p, Q'(m).
§ 9: ~(L), N(L).
§ 10: k', g,~, U,~, U", ~'" 'Y, 'Y", f(w), :no
§ 11: Gp, G'p, L p, lp, pr, Up, J(U,P).
Chapter I
Locally compact fields

§ 1. Finite fields. Let F be a finite field (commutative or not) with the
unit-element 1. Its characteristic must clearly be a prime p> 1, and the
prime ring in F is isomorphic to the prime field Fp=Z/pZ, with which
we may identify it. Then F may be regarded as a vector-space over F p;
as such, it has an obviously finite dimension J, and the number of its
elements is q = pl. If F is a subfield of a field F' with q' = pI' elements, F'
may also be regarded e. g. as a left vector-space over F; if its dimension
as such is d, we have f' = df and q' = qd = pdI . *
THEOREM 1. All finite fields are commutative.
o
This theorem is due to Wedderburn, and we will reproduce Witt's
modification of Wedderburn's original proof. Let F be a finite field of
characteristic p, Z its center, q = pI the number of elements of Z; if n is
the dimension of F as a vector-space over Z, F has q" elements. The
multiplicative group F X of the non-zero elements of F can be partitioned
into classes of "conjugate" elements, two elements x,x' of F x being called
conjugate if there is YEF x such that x'=y-1xy. For each xEF x , call
N(x) the set of the elements of F which commute witli x; this is a sub-
field of F containing Z; if b(x) is its dimension over Z, it has qd(x) elements.
As we have seen above, n is a multiple of b(x), and we have b(x) < n unless
x E Z. As the number of elements of F x conjugate to x is clearly the index
of N(xr in F i.e. (q"-I)/(q"(X)-l), we have
X
,
(1) q"-I=q-l +L q" - 1 ,

x qd(x)_1
where the sum is taken over a full set of representatives of the classes of
non-central conjugate elements of FX. Now assume that n> 1, and call
P the "cyclotomic" polynomial n
(T - 0, where the product is taken
over all the primitive n-th roots of 1 in the field C of complex numbers.
By a well-known elementary theorem (easily proved by induction on n),
this has integral rational coefficients; clearly it divides (1"'-1)/(1""-1)
whenever b is a divisor of n other than n. Therefore, in (1), all the terms
except q -1 are multiples of P(q), so that P(q) must divide q -1. On the
n
other hand, each factor in the product P(q) = (q - 0 has an absolute
value > q -1. This is a contradiction, so that we must have n = 1 and
F=Z.
2 Locally compact fields
We can now apply to every finite field the following elementary result:
LEMMA 1. If K is a commutative field, every finite subgroup of K x

is cyclic.
In fact, let r be such a group, or, what amounts to the same, a finite
subgroup of the group of all roots of 1 in K. For every n ~ 1, there are at
most n roots of xn = 1 in K, hence in r; we will show that every finite
commutative group with that property is cyclic. Let IX be an element of r
of maximal order N. Let Pbe any element of r, and call n its order. If n
does not divide N, there is a prime p and a power q=pv of p such that q
divides n and not N. Then one verifies at once that the order ofIXpn/q is
the l.c.m. of Nand q, so that it is > N, which contradicts the definition of
N. Therefore n divides N. Now x n= 1 has the n distinct roots lX iN/ n in r,
with O:::;i<n; as p is a root of x n = 1, it must be one of these. This shows
that IX generates r.
THEOREM 2. Let K be an algebraically closed field of characteristic
p> 1. Then, for every f~ 1, K contains one and only one field F = Fq
with q = pI elements; F consists of the roots of x q = X in K; F x consists
of the roots of x q - 1 = 1 in K and is a cyclic group of order q-1.
If F is any field with q elements, lemma 1 shows that F x is a cyclic
group of order q -1. Thus, if K contains such a field F, F x must consist
of the roots of X q - 1 = 1, hence F of the roots of x q - X = 0, so that both
are uniquely determined. Conversely, if q = pi, x -+ xq is an automorphism
of K, so that the elements of K which are fixed under it make up a field F
consisting of the roots of x q - X = 0; as it is clear that X q - X has only
simple roots in K, F is a field with q elements.
COROLLARY 1. Up to isomorphisms, there is one and only one field
with q = pI elements.
This follows at once from theorem 2 and the fact that all algebraic
closures of the prime field Fp are isomorphic. It justifies the notation Fq
for the field in question.
COROLLARY 2. Put q = pI, q' = pI', with ~ 1, f ~ 1. Then Fq • contains
a field Fq with q elements if and only if f divides /'; when that is so, Fq.
is a cyclic extension of Fq of degree /'/ f, and its Galois group over Fq is
generated by the automorphism x-+xq•
We have already said that, if F q• contains F q , it must have a finite
degree dover F q , and then q'=qd and f=df Conversely, assume that
t.
f = df, hence q' = and call K an algebraic closure of Fq'; by theorem 2,
the fields F q , F q., contained in K, consist ofthe elements of K respectively
invariant under the automorphisms IX. Pof K given by x .... xq• x .... xq' ; as
P=lXd• F q , contains F q • Clearly IX maps F q , onto itself; if cp is the auto-
morphism of F q' induced by IX. F q consists of the elements of F q' invariant
under CPo hence under the group of automorphisms of F q' generated by cp;
this group is finite. since cpd is the identity; therefore. by Galois theory.
it is the Galois group of F q' over F q and is of order d.
COROLLARY 3. Notations being as in corollary 2. assume that f' = df.
Then. for every n~ 1. the elements of Fq ,. invariant under x .... x q ". make
up the subfield of Fq , with qr elements. where r=(d,n).
Let K be as in the proof of corollary 2; the elements of K, il!va-
riant under x .... x q", make up the subfield F' of K with qn elements; then
F' n F q' is the largest field contained both in F' and Fq ,; as it contains
Fq , the number of its elements must be of the form qr, and corollary 2
shows that r must be (d,n).
§ 2. The module in a locally compact field. An arbitrary field. provided
with the discrete topology, becomes locally compact; thus the question
of determining and studying locally compact fields becomes significant
only if one adds the condition that the field should not be discrete.
We recall the definition of the "module" of an automorphism, which
is basic in what follows. For our purposes. it will be enough to consider
automorphisms of locally compact commutative groups. Let G be such
a group (written additively), A an automorphism of G, and IX a Haar
measure on G. As the Haar measure is unique up to a constant factor, A.
transforms IX into CIX, with cER~; the constant factor c, which is clearly
independent of the choice of IX. is called the module of A and is denoted by
modG(A). In other words, this is defined by one of the equivalent formulas
(2) oc(A.(X») = modG(A.)oc(X), Jf(A -l(x»)dlX(X)= modG(A) Jf(x)dlX(x),
where X is any measurable set. f any integrable function. and
O<IX(X)< + 00, JfdlX fO; the second formula may be written symbolical-
ly as dlX(A(x»)=mod G(A)dlX(x).IfG is discrete or compact, the first formula
(applied to X={O}, X=G, respectively) shows that the module is
1. Obviously, if A, A' are two automorphisms of G, the module of AOA'
is the product of those of A and A.'. We shall need the following lemma:
LEMMA 2. Let G' be a closed subgroup of G, and A an automorphism
of G which induces on G' an automorphism A' of G'. Put G" = G/G', and
call A" the automorphism of G" determined by A modulo G'. Then:
modG(A) = modG,(A')modG·(A").
In fact, it is well-known that one can choose Haar measures IX,IX',IX"
on G, G', G" so as to have, for every continuous function f with compact
support on G:
J f(x)d rx(x) = d·r (JG' f(x + y)d rx'(y»)d rx"(x);

G
here x denotes the image of x in G", and the function f(x+ y)drx'(y), J
which is written as a function of xe G, but is constant on the classes
modulo G' in G, is to be understood as a function ofX on Gil in the obvious
manner. Applying A. to both sides, one gets the conclusion of the lemma.
Now, if K is any topological field, and aeK x , x-+ax and x-+xa
are automorphisms of the additive group of K; if K is locally compact,
we may consider their modules. Similarly, if V is a topological left vector-
space over K, v-+av is an automorphism of V for every aeKx; if V is
locally compact, we may consider the module of this automorphism;
this will be denoted by mody(a); we also define mody(O) to be O. In other
words, if JI. is a Haar measure on V and X any measurable subset of V
with o<JI.(X) < + 00 (e.g. any compact neighborhood of 0 in V), mody(a)
is defined, for all ae K, by
JI.(aX)
mody(a) = JI.(X) .
In particular, for any locally compact field K, we define modK(a) to be

the module of x -+ a x in K if a -1= 0, and 0 if a = O. It will be seen later that
the module of x-+xa is always the same as that of x-+ax. Clearly, if
K =R,CorH,modK(a)isequal,respectively, to lal, lal 2 =aaorlaI 4 =(aa)2.
In the rest of this section, K will denote, once for all, a nondiscrete
locally compact field (commutative or not), and rx a Haar measure on
the additive group of K.
PROPOSITION 1. The function mod Kis continuous on K, and modK(ab) =
=modK(a)modK(b) for all aeK, beK.
The latter assertion is obvious. Now let X be a compact neighborhood
of 0 in K. For any aeK and any 8>0, there is an open neighborhood U
of the compact set aX such that rx(U)~rx(aX)+8; let W be a neighbor-
hood of a such that WX c U. Then, for all xe W, we have
modK(x) ~modK(a) + rx(X)-18.
This shows that mod K is upper semicontinuous. In particular, it is con-
tinuous at O. As modK(x)=modK(x-1)-1 for x -1=0, it is also lower semi-
continuous everywhere on K x , hence continuous on K x •
As K is not discrete, prop. 1 shows that there is, for every 8>0, aeK
such that 0<modK(a)~8, hence also, for every M>O, beK such that
modK(b)~M. As modK is not bounded, K cannot be compact.
PROPOSITION 2. For all m> 0, the set Bm of the elements x of K such

that modK(x)~m is compact.
Let V be a compact neighborhood of 0 in K; let W be a neighborhood

of 0 such that WV c V. As above, we can choose re V11 W such that
0< modK(r) < 1; by induction on n, we have r"e V for all n ~ 1. If r' is any
limit point of the sequence {r"}"~I' modK(r') must be 0, since modK(r")
has the limit 0 for n~ + 00. Therefore that sequence can have no other
limit point than 0; as it is contained in the compact set V, it has the limit O.
Now take m>O and aeBm; as r"a tends to 0, there is a smallest integer
v~O such that rVaeV; if a is not in V, then rv - 1 a$V, hence rVaeV-(rJl).
Call X the closure of V - (r V); clearly X is compact, and 0 is not in X;
therefore, if we put Jl=infxexmodK(x), we have Jl>O. Let N be an integer
such that modK(r)N :!!;'Jl/m. Then, if aeBm, a$ V, and v is defined as above,
we have
modK(r)N m:!!;'Jl :!!;,modK(rVa) = modK(rtmodK(a):!!;,modK(r)Vm,
hence v:!!;,N. This proves that Bm is contained in the union of the com-
pact sets V, r- 1 V, ... , r- NV. As prop. 1 shows that Bm is closed, this
completes the proof.
COROLLARY 1. The sets Bm, for m>O, make up a fundamental system
of neighborhoods of 0 in K.
Let V be any compact neighborhood of 0 in K; take
m>suPxeymodK(x), so that Bm::J V; call X the closure of Bm- V, and put
m' =infxexmodK(x). Then O$X and X cBm , so that, by prop. 2, X is
compact; therefore O<m':!!;,m. Take O<Jl<m'; then BJlcBm, BJl 11 X = 0,
hence BJlc V.
COROLLARY 2. For aeK, lim,,-++ooa"=O if and only if modK(a) < 1.
COROLLARY 3. A discrete subfield of K is finite.
Let L be such a field. If aeL, we must have modK(a):!!;, 1, since other-
wise, by corollary 2, the sequence {a-"},,~o would be contained in Land
not discrete. Therefore L is a discrete subset of the compact set B 1 ,
hence finite. Of course this cannot happen if K is of characteristic O.
THEOREM 3. Let V be a topological left vector-space over K, and
let V' be a finite-dimensional subspace of V, with a basis {VI' ... , v,,}. Then
the mapping
(Xl'''''X,,)~ L" XjVj
j=1
of K" onto V' is an isomorphism for the structures of K" and V' as topological
left vector-spaces; V' is closed in V and locally compact.
Let f be the mapping defined above; it is bijective, K-linear, and
continuous by the definition of a topological vector-space. In order to
show that it is an isomorphism, it is enough to prove that f- 1 is conti-

nuous, i.e. that f is an open mapping; in view of corollary 1 of prop. 2
and of the linearity of f, we need only show that the image of (Bm)" by f
contains an neighborhood of 0 in V' for every m > O. Call S the subset
of K" defined by
supjmodK(xj)= 1.
Then O$S; by prop. 1, S is closed; it is contained in (B 1)", hence compact
by prop. 2. Therefore O$f(S), and f(S) is compact. Hence there exists a
neighborhood W of 0 in V, and a neighborhood of 0 in K which we may
assume to be of the form Be with e > 0, such that Be We V - f(S), i. e.
yWnf(S)=0 whenever modK(y)~e. Now take m>O, and take aeK
such that O<modK(a)~me. Let v= ~:XjVj be any point in Vn a W, other
than 0, and take h such that supjmodK(xj)=modK(xh); then Xh=FO. Put
X;=X,;-l Xj for 1 ~i~n, and
As (x~, ... ,x~) is in S, we have v'ef(S); as veaW, we have v'eyW with

y=x,;-la; by the definition of Wand e, this implies modK(y»e, hence
modK(xh)<e-1modK(a)~m. Therefore (Xl' ... , X") is in (Bm )", and v is
in the image ofthat set by f We have thus shown that this image contains
Vn a W, which is a neighborhood of 0 in V. Let now w be in the closure
of V' in V, and apply what we have proved to the finite-dimensional
subspace V' of V generated by V and w; we see then that V must be
closed in V'. As this implies that we V, it completes the proof of the
theorem.
COROLLARY 1. Every finite-dimensional left vector-space over K can
be provided with one and only one structure of topological lefi vector-
space over K.
In fact, if V is of dimension n, one can define such a structure on V by
means of any K-linear bijective mapping of K" onto V; the unicity is an
immediate consequence of tho 3 applied to V. From now on, every such
vector-space will tacitly be assumed to carry the structure defined by
this corollary.
COROLLARY 2. If V is a locally compact topological left vector-space
over K, then V has a finite dimension dover K, and modv(a)=modK(a)d
for every aeK.
The latter assertion, for a space of dimension d, is an immediate
consequence of Fubini's theorem and of the fact that such a space is
isomorphic to Kd by corollary 1. Now assume merely that V is locally
compact, and take ae K such that 0 < modx(a) < 1. Then, by corollary 2
of prop. 2, lima"=O, hence mody(a) < 1. Let V' be a subspace of V of
finite dimension (); by tho 3, it is closed in V; put V" = VIV'. By lemma 2
we have then
mody(a) = mody,(a)mody.(a) = modx(a)"mody.(a),
and therefore, since mody.(a) also must be < 1 if V" 4= {O}, and is 1 if
V"={O}:
mody(a) ~ modx(a)".
This gives an upper bound for fJ. valid for the dimension of all finite-
dimensional subspaces of V; therefore V itself has a finite dimension.
If V is a left vector-space over K, of finite dimension n, topologized
as we have said above, Fubini's theorem shows at once that every sub-
space of V of dimension n' <n is of measure O. Now let A be any K-linear
mapping of V into V; if it is of rank n, it is an automorphism of V also in
the topological sense, and we may consider its module mody(A). If it is
of rank n' < n, it maps V onto a subset of V of measure 0, and we define
mody(A) to be O.
COROLLARY 3. Let A be an endomorphism of a left vector-space Vof
finite dimension over K. If K is commutative, then mody(A) = modx(detA).
Call n the dimension of V. If A is of rank < n, the assertion is clear.
If not, identify V with K" by choosing a basis for V. It is well-known that
every automorphism of K" can be written as a product of automorphisms
of the following three types: (a) permutations of the coordinates;
(b) mappings of the type
with aeKx; (c) mappings ofthe type
(Xl'X 2 , ... , X,,)-+{Xl

\:
+.f a x
.=2
j j ,X 2 , •.• , x,,\.
.)
For type (a), the assertion is obvious; for types (b) and (c), it follows
from a straightforward application of Fubini's theorem, just as in classi-
cal analysis (where one proves the theorem for the case K = R).
PROPOSITION 3. The function modx induces on K x an open homomor-
phism of K x onto a closed subgroup r of R ~ .
Call r. r' the images of K X and of K under the mapping modK ;
clearly r r
is a subgroup of R ~, and r' = u {O}. For every m> 0, the
intersection of r' with the closed interval [O,m] is the image of Bill under
modx; by prop. 1 and 2, this is compact; therefore r' is closed in R+,
and r is so in R~. Now call U the kernel ofmodK in K X , i.e. the set
{xEKlmodK(x) = I}. Let V be any neighborhood of 1 in K X , and V'its
image under modK; in order to prove the openness of the homomorphism
modK of K x onto r, we have to show that V' is a neighborhood of 1 in r.
Assume that this is not so; then there is a sequence (YII) in r - V' such
that limYII= 1. For each n, let a"EK x be such that YII=modK(a ll ). By
prop. 2, the sequence (an) has at least one limit point a; clearly modK(a) = 1,
i.e. aE U. But UV is a neighborhood of U, and so there must be some n
such that ailE UV, hence YIIE V'. This contradicts the assumption.
THEOREM 4. There is a constant A> 0 such that
(3)
for all xEK, YEK. If (3) is valid for A= 1, then the image r of K X
under
modK is discrete in R~. Morever, (3) is valid for
A = sUPxeK. modK(x)" 1 modK(1 + x),
and this is the smallest value of A for which it is valid.
Define A by the last formula; clearly I~A< +00. For x=y=O, (3)
is obvious; otherwise we may, after interchanging x and y if necessary,
assumethatx+OandmodK(y)~modK(x), Putz= yx-l;thenmodK(z)~ 1,
hence mod K(1 +z)~A, and therefore
modK(x+ y)=modK(1 +z)modK(x)~AmodK(x).
This proves (3). Also, taking y= 1 and xEB l in (3), with Bl as in prop. 2,
we see that the value we have chosen for A is the smallest for which (3)
can be valid. Now assume A= 1. Then the image of 1 +Bl by mod K is
contained in the interval [0, 1]; as this, by prop. 2 and 3, must contain a
neighborhood of 1 in r, r must be discrete.
COROLLARY. If (3) is valid with A=I, then modK(x+y)=modK(x)
whenever modK(y) < modK(x).
As (_1)2=1, we have modK(-1)=1, hence modK(-y)=modK(y).
As x = (x + y) + (- y), our assumptions imply
modK(x)~sup(modK(x+y),modK(y»)~modK(x),
hence the conclusion.

DEFINITION 1. The inequality (3) with A = 1 is called the ultrametric
inequality; if this is valid, then mod K , and K itself, are said to have the
ultrametric property; or to be ultrametric.
§ 3. Classification of locally compact fields. Here we shall need the
following elementary lemma:
§ 3. Classification of locally compact fields 9
LEMMA 3. Let F be a function on the set N of natural integers, with

values in R+. Assume that F(mn)=F(m)F(n) for all m, n, and that there is
A> 0 such that
F(m + n) ~ A sup(F(m),F(n»)
for all m, n. Then either F(m)~ 1 for all m, or there is A>O such that
F(m)=mA. for all m.
The first assumption on F implies, for m=O, that F(O)=O unless F
is the constant 1, and, for m= 1, that F(l)= 1 unless F is the constant 0;
it also implies that F(m~=F(m)" for all integers k~ 1. Leaving aside the
trivial cases where F is the constant 0 or 1, we may assume that F(O)=O
and F(mk)=F(m)" for all integers k~O. Put f(m)=sup(O,logF(m»), this
being understood to mean in particular that f(m) = 0 whenever F(m)=O.
Our lemma amounts now to saying that f(m)=Alogm for all m~2, with
some constant A~O. Put a=sup(O,logA); then we have, for all m, n, k:
f(mk)=kf(m), f(mn)~f(m)+ f(n), f(m+n)~a+sup(f(m),J(n»).
The last relation gives, by induction on r:
(4)
Now let m, n be integers ~2; m may be expressed in the form
r
m= i..J
~ e·nI
i
,
i=O
b=sup(f(O),J(l), ... ,J(1l-1»).

Then we have, for every i:
f(eini)~b+if(n),
and therefore, in view of (4):

f(m)~ra+b+r f(n).
As nr~m, i.e. r logn~logm, this gives

f(m) ~ a+ f(n) + _b_
logm"" logn logm·
In this inequality. replace m by m"; this does not change the left-hand side,
and, for k-+ + 00, we get
f(m) ~ a+ f(n)
logm"" logn .
Now replace n by nk; for k-+ + 00, we get

f(m) ~ f(n)
logm ""'" logn .
Interchanging m and n, we see that f(m)/logm is constant for m;3:2; as
we have observed above, this proves the lemma.
Now we consider again a non-discrete locally compact field K. For
greater clarity, in the rest of this section, we shall denote by lK (not by 1)
the unit-element of K; then the prime ring in K consists of the elements
m ·I K with meZ; if K is of characteristic p> 1, then p ·I K =O. For meN,
we write F(m)=modK(m' l K); then, for every meZ and every xeK, we
have modK(mx) = FOml)modK(x).
LEMMA 4. Assume that F is bounded, i.e. that mod K is bounded on

the prime ring in K. Then F ~ 1, and mod K is ultrametric on K.
Since F(mn)=F(m)F(n), the first assertion is obvious. Now let A
be as in tho 4 of § 2; take n;3: 1, put N = 2", and let Xl' ... , XN be N elements
of K. By induction on n, one gets the inequality
mOdK(tl x0 ~A"suPi(modK(xi»)'
Replacing some of the Xi by 0, one sees that this same inequality remains
valid whenever N ~ 2". Applying this to the relation
(x+yf"= I
;=0
(2~)Xiy2"-i,
I
we get
modK(x+ y)2"~A"+1 SUPi (F (e")) mOdK(X)imodK(y)2"-i).

Assume for instance that modK(y)~modK(x); as F~ 1, we get:
modK(x+ y)2" ~A"+ 1 mod K(x)2".
This is so for all n;3: 1; for n -+ + 00, we get
modK(x+ y)~modK(x),
i. e. the ultrametric inequality.
Next we recall the definition of the usual "valuations" on the field Q
of rational numbers. Let first p be a rational prime. Every xeQ x can be
written in one and only one way in the form x=p"a/b, where n, a, bare
integers, b>O, and a and b are relatively prime to each other and to p;
when that is so, put Ixlp=p-"; also, put 10Ip=0. The function x-+lxlp
§3. Classification of locally compact fields 11
defined in this way on Q is known as the p-adic valuation on Q; clearly it

satisfies the ultrametric inequality; it determines a topology on Q, viz.
the one defined by the distance function
l5(x,y) = Ix - yip'
The completion of Q for this metric is the field of p-adic numbers and is
denoted by Qp; the closure of Z in that field is the ring of p-adic integers
and is denoted by Zp. Clearly the p-adic valuation on Q can be extended
by continuity to Qp and remains ultrametric on Qp; this extension is
still denoted by Ixl,. It is easily seen that Zp is compact (the reason for
°
this may be expressed by saying that Zp is the "projective limit" of the
finite groups Zip" Z for n-+ + (0); as it is a neighborhood of in Qp, Qp
is locally compact; clearly it is not discrete.
On the other hand, we shall write Ixl <Xl whenever convenient, instead
of lxi, for the "ordinary" absolute value on Q and on R. As R is nothing
else than the completion of Q for the distance function Ix - yl <Xl' we shall
sometimes write Q<Xl for R. Thus the symbol Qu, where v may be either 00
or a rational prime, denotes anyone of the completions Q <Xl = Rand
QpofQ.
THEOREM 5. Let K be a non-discrete locall.v compact field; put
F(m)=modK(m '1 K) for meN. Then: either (a) K is of characteristic
p> 1, and then F(m)=O for m==O (mod. p) and F(m) = 1 for (m,p) = 1; or
(b) K is a division algebra of finite dimension 15 over a field Qu, and then
F(m)=lml~·
By prop. 1 and tho 4 of § 2, F satisfies the assumptions in lemma 3;
hence, by that lemma, it is of the form m -+ mA with A. > 0, or it is ~ 1.
Assume that we are in the latter case; with Bl as in prop. 2 of § 2, this
means that the sequence (m·l K ), for meN, is contained in B 1 ; as Bl is
compact, it must have at least one limit point a. Then, by corollary 1 of
prop. 2, there are, for every 8>0, infinitely many meN such that
modK(m' 1K-a)~8. Let m, m' be two such integers, with m<m'. Since
F ~ 1 implies, by lemma 4, that modK is ultrametric, we have then
modK(m' '1 K-m '1K)~8,
i. e. F(m' - m) ~ 8. In particular, this shows that there are integers n ~ 1
such that F(n) < 1; let p be the smallest of such integers. Since F(mn) =
=F(m)F(n) for all m, n, clearly p must be a prime. For any xeN, we have
F(P x) < 1, hence F(1 + Px) = 1 by the corollary of tho 4, § 2. For any
integer m ~ 1, prime to p, we have mP - 1 == 1 (P), hence F(m P - 1 ) = 1 by
what we have just proved, and therefore F(m) = 1. If K is of characteristic
p' > 1, then F(P') = 0, so that p' can be no other than p; then F is as stated
in case (a) of our theorem. If K is of characteristic 0, F(P) cannot be 0,
and we may put F(P)=p-A with A>O; then F(m)=lml; for all m, as one
sees at once by writing m=p"m' with (m',p) = 1. Accordingly, whenever
K is of characteristic 0, F must be of the form m-Iml~ with A>O. The
mapping n -+ n . 1K of Z onto the prime ring Z . 1K of K is then an algebraic
(not necessarily a topological) isomorphism, which can be extended to
an isomorphism of Q onto the prime field in K; to simplify the language,
identify the latter with Q by means of that isomorphism. From what we
have found about F, it follows at once that mod K induces the function
x -+ Ixl! on Q; therefore, by corollary 1 of prop. 2, § 2, the topological
group structure induced on Q by that of K is the one determined by the
distance function Ix- Ylv. As the closure of Q in K is locally compact,
hence complete for that structure, it follows that this closure is isomorphic
to the completion Qv of Q for the valuation v. As the prime ring, hence
also the prime field, are clearly contained in the center of K, the same is
true of Qv. Now K can be regarded as a vector-space over Qv; as such,
by corollary 2 ofth. 3, § 2, it must have a finite dimension J, and we have,
for every XEQv, modK(x) = modQ.(x)". To complete the proof, it only
remains to be shown, in the case V= 00, that modR(m)=m for mEN,
which is clear, and, in the case v = p, that modQp (P) = p -1; this follows
at once from the fact that Zp is a compact neighborhood of 0 in Qp, and
that its image p. Zp, under x-px, is a compact subgroup of Zp of
index p, so that its measure, for any Haar measure IX on Qp, is p - 1 IX (Zp).
It will be convenient to formulate separately what has just been proved:
COROLLARY. In the case (b) of theorem 5, modK(x)=lxl~ for XEQv·
DEFINITION 2. A non-discrete locally compact field K will be called
a p-field if p is a prime and modK(p· 1K) < 1, and an R-field if it is an algebra
over R.
By lemmas 3 and 4 and tho 4 of § 2, the image r of K x under mod Kis
discrete when K is a p-field, so that such a field cannot be connected; this
shows that a topological field is an R-field if and only ifit is connected and
locally compact. It is well known that there are no such fields except R,
C and the field H of "ordinary" (or "classical") quaternions; a proof for
this will be included in Chap. IX-4.
§ 4. Structure of p-fields. In this section, p will be a prime and K will

be a p-field with the unit element 1.
THEOREM 6. Let K be a p-field; call R, R x and P the subsets of K

respectively given by
R={xEKlmodK(x)~l}, R ={xEKlmod K(x)=l},
X
P={xEKlmodK(x)< 1}.
§4. Structure of p-fields 13
Then K is ultrametric; R is the unique maximal compact subring of K;

R x is the group of invertible elements of R; P is the unique maximal left,
right or two-sided ideal of R, and there is neP such that P=nR=Rn.
Moreover, the residual field k = R/P is a finite field of characteristic p;
if q is the number of its elements, the image r of K X in R~ under mod K
is the subgroup of R~ generated by q; and modK(n)=q-l.
The set R is the same as the one previously denoted by B 1; it is com-
pact, and so is R x • By tho 5 of § 3, modK is ~ 1 on the prime ring of K;
therefore, by lemma 4 of § 3, K is ultrametric. This, by tho 4 of § 2, is the
same as to say that R+R=R; as R is obviously closed under multipli-
cation, it is a ring. Clearly every relatively compact subset of K which is
closed under multiplication is contained in R; therefore R is the maximal
compact subring of K. The invertible elements of R are those of R x •
By tho 4 of § 2, r is a discrete subgroup of R ~ ; let ybe the largest element
of r' which is < 1, and let neK x be such that modK(n)=y. Clearly y
generates r; therefore, for every xeK x , there is one and only one neZ
such that modK(x) = yn; then xn- n and n-n x are in R x. It is clear that
P = n R = R n; this implies that P is compact. As R - P = R x , P has the
maximal properties stated in our theorem. As R is a neighborhood of 0,
and R=R+R, R is open; so is P; as R is compact, k=R/P is finite. As
p. 1eP, the image of p. 1 in k is 0, so that k is of characteristic p; if it
has q elements, q is the index of P = n R in the additive group of R. There-
fore, ih is a Haar measure on K, a(R)=qa(nR), hence modK(n)=q-l.
This completes the proof.
DEFINITION 3. With the notations of theorem 6, q will be called the

module of K; any element n of K such that P=nR=Rn will be called
X
a prime element of K. For any xeKx, the integern such that modK(x) = q-n
will be denoted by ordK(x). For each neZ, one writes pn=nn R=Rnn.
We will write ord(x), instead of ordK(x), when there is no danger of
confusion. We also put ord(O)= + 00; then pn is the set of the elements
x of K such that ord(x)~n. With these notations, we can state as follows
some corollaries of theorem 6:
COROLLARY 1. Let (xo,x 1 , ...) be any sequence with the limit 0 in K.
+00
Then the series Lo Xi is commutatively convergent in K.
For each ne N, put
en = SUPi > n modK(xi )·
Our assumption means that limen=O. Let now S, S' be two finite sums
L
of terms in the series Xi' both containing the terms xo, Xl' ... , Xn and
possibly some others. The ultrametric inequality gives modK(S-S')~en'

The conclusion follows from this at once (the "filter" of finite sums of
the series LXiis a "Cauchy filter" for the distance-function modK(x- y»).
COROLLARY 2. Let, be an element of P, other than 0; put n=ord(e),
and let A be a full set of representatives of the classes modulo pn in R.
Then, for all veZ, every xepnv can be expressed in one and only one way
in the form
+00
X= L ai,i
i="
with aie A for all i ~ v.
,v
Writing x=x' with x'eR, we see that it is enough to deal with the
case v = O. Then one sees at once, by using induction on N, that one can
determine the aie A in one and only one way by the condition
x= Lai,i
N
(p"(N+l»
i=O
for N = 0,1, ... This is equivalent with the assertion in our corollary.
COROLLARY 3. Every automorphism of K (as a topological field)
maps R onto R, Ponto P, and has the module 1 when it is viewed as an auto-
morphism of the additive group of K.
COROLLARY 4. For every aeKX, the automorphisms x-+ax and
x -+ x a of the additive group of K have the same module.
This follows at once from corollary 3, applied to the automorphism
x-+a- 1 xa. As the same fact is easily verified for the field H of "ordinary"
quaternions, it holds for all locally compact fields.
COROLLARY 5. Let K be a commutative p-jield, and K' a division

algebra over K. Then K' is a p-field; every automorphism of K' over K in
the algebraic sense is a topological automorphism; and, if R and R' are
the maximal compact subrings ofKand of K', and P and P' are the maximal
ideals in R and in R', then R=KnR' and P=KnP'.
Regarding K' as a finite-dimensional vector-space over K, we provide
it with its "natural" topology according to corollary 1 of tho 3, § 2. As
this is unique, it is invariant under all K-linear mappings of K' onto
itself, and in particular under all automorphisms of K' over K. Identifying
K, as usu&l, with the subfield K 'l K, of K', we see that K' is not discrete.
By corollary 2 ofth. 3, § 2, and tho 5 of§ 3, it is a p-field. The rest is obvious.
COROLLARY 6. Assumptions and notations being as in corollary 5,
call q and q' the modules of K and of K', respectively; let 1t be a prime
§ 4. Structure of p-fields 15
element of K, and put e=orddn). Then q' =qt, where f is an integer ~ 1,

and the dimension of K' over K is ef
Put k=R/P and k' =R'/P'; in view of the last assertion of corollary 5,
we may identify k with the image of R in k' = R'/ P'; if then f is the degree
of k' over k, we have q' = qt. Now apply corollary 2 ofth. 3, § 2, to modK(n)
and to moddn); we get the result stated above.
The last corollary shows in particular that ordK,(n) is ~ 1 and is
independent of the choice of the prime element n in K. This justifies the
following definition:
DEFINITION 4. Let assumptions and notations be as in corollaries 5

and 6 of theorem 6. Then e is called the order of ramification of K' over
K, and f the modular degree of K' over K; K' is said to be unramified
over K if e = 1, and to be fully ramified over K if f = 1.
PROPOSITION 4. Let K be a commutative p-field; let K' be a fully
ramified division algebra of finite dimension over K; let R, R' be the
maximal compact subrings of K and of K', respectively, and let n' be a
prime element of K'. Then K'=K(1t'), R'=R[1t'], and K' is commutative.
Let P, P' be the maximal ideals in R and in R', respectively, and
let A be a full set of representatives of the classes modulo P in R. As
K' is fully ramified over K, corollaries 5 and 6 of theorem 6 show at once
that A is also a full set of representatives of the classes modulo P' in R'.
Applying corollary 2 of tho 6 to K', R', P' and A, and to e=1t', we see
e-l
that the elements of R' of the form L aj1t,j, with ajeA for O~i~e-1,
j=O
make up a full set of representatives A' of the classes modulo p,e in R'.
Take now a prime element 1t of K, and put e=ordd1t); e is the order
of ramification of K' over K, hence also the dimension of K' over K,
by corollary 6 of tho 6. Applying now corollary 2 of tho 6 to K', R', p,e, A'
e
and to = 1t, we see that every element of p,ev can be written in one
+00
and only one way in the form L aj1tj, with ajeA' for allj~v. As K is
contained in the center of K', 1tj~~mmutes with 1t'; therefore, in view of
the definition of A', every such element can be written as
with ajje A for 0 ~ i ~ e -1, j ~ v, or, what amounts to the same in view
e-l
of corollary 2 ofth. 6, as L (ljn,j with (ljer for O~i~e-1. This shows
j=O
that K' =K(1t'), and, for v=O, it shows that R' =R[1t']. As K is contained
in the center of K', n' commutes with all elements of K; therefore K' is
commutative.
COROLLARY 1. Let K be a commutative p-field of characteristic p;
call KP its image under the endomorphism x-+xP, and let n be a prime
element of K. Then K is a fully ramified extension of KP of degree p, and
K=KP(n).
Put K'=KP; x-+x P is an isomorphism of K onto K', which we may
use to transfer to K' the topology of K; K may then be regarded as a
topological vector-space over K'; as such, by corollary 2 of tho 3, § 2, it
must have a finite dir-lension. This shows that K is of finite degree over
K'. As K and K' are isomorphic, they have the same module, so that
the modular degree of Kover K' is 1. By proposition 4, this implies
that K=K'(n); as nPeK', the degree of Kover K' must be p or 1. As
ordK(n) = 1, n is not in KP, so that K =1= K'. Therefore K is of degree p
over K'.
COROLLARY 2. Let K be as in corollary 1, and let K be an algebraic
closure of K. Then, for every n;;:::O, K contains one and only one purely
inseparable extension of K of degree p"; this is the image Kp- n of K
under the automorphism x-+xrn oiK.
It follows at once from corollary 1 that Kr 1 is of degree p over K;
by induction on n, one sees then that Kp- n is of degree p" over K. On
the other hand, it is well-known, and easily proved, that, if K' is purely
inseparable of degree ~p" over K, it must be contained in Kr". Our
conclusion follows from this at once.
THEOREM 7. Let K be a p-field; call q its module, R its maximal com-
pact subring and P the maximal ideal of R. Then K x has at least one
subgroup of order q - 1; every such subgroup is cyclic; if M x is such a
subgroup, the set M =Mx u{O} is a full set of representatives of the
classes modulo P in R, and there is a prime element n of K such that
nMx =Mx n. If K is commutative, there is only one such group M X ;
it is the group of the roots of 1 of order prime to p in K.
The construction of M x depends upon the following lemma:
LEMMA 5. For all n;;:::O, (1 +P)p nc 1 +p,,+1.
This can be immediately verified by induction on n. It amounts to

saying that, ifx == 1 (P), xpn == 1 (P" + 1).
Now call p the canonical homomorphism of R onto k=R/P. By tho 2
of § 1, k x is cyclic of order q -1. In particular, for all xe R x, we have
p(X)q-1 = 1, i. e. x q- 1== 1 (Pl. If q = pI, lemma 5 shows now that
x(q-1)qn == 1 (pI" + 1); this can also be written as
X
qn+ 1 =-xqn (pfn+ 1) .
Applying now corollary 1 of tho 6 to the series

x+(xq-x)+(x q2 -x q)+,,·,
we see that it is convergent for all xe R x , so that we may write
w(x) = lim" .... +00 x q"
for xeR x, and of course also for xeP, hence for all xeR. Clearly
w(xy)=w(x)w(y) whenever xy= yx; in particular, we have w(XV)=w(xt
for all x e R x , ve Z. As the above series for w(x) shows, we have w(x) == x (P)
for all xeR; obviously w(x)=O for xeP, and lemma 5 shows that w(x) = 1
for xe1+P. Therefore w-l(O)=P and w- l (1)=1+P. As x Q - l e1+P
for all xeR x, we have w(x)q-l=1 for xeRx. Take a representative Xl
in R x of a generator of the cyclic group k x = (RIP) x, and put J.ll = w(x l );
for neZ, we have J.l~ = 1 if and only if w(x~)= 1; as this is equivalent to
x~ == 1 (P), hence to n==O (q-1) in view of our choice of Xl' this shows
that J.ll generates a cyclic subgroup of R x of order q -1. Conversely,
let r be any finite subgroup of K x of order n prime to p; clearly it is a
subgroup of R x • The image of q, in the multiplicative group (ZlnZ) x of
the integers prime to n modulo n, must have a finite order N; then
qN == 1 (n). As zn = 1 for every zer, we get now zqNv =z for all v ~O and all
zer, hence w(z)=z, so that z==1 (P) implies z=1. This shows that the
morphism of r into k X = (RIP) X induced by p is injective, and therefore
that r is cyclic, that its order divides q -1, and that, if it is of order q - 1,
ru{O} is a full set of representatives of RIP in R. In particular, if K is
commutative, we see that w induces on R x a morphism of R x onto the
group M x of the (q -1 )-th roots of 1 in K, that it maps R onto
M = M x u {O}, and that it determines a bijection of RIP onto M; moreover,
every subgroup r of K x of order prime to p is then contained in M x; in
particular, M x contains all the roots of 1 of order prime to p in K. As to
the existence of a prime element of K with the property stated in our
theorem, it is trivial if K is commutative. Assume that this is not so,
and take any prime element n of K. For every aeK x , x-+axa- l is an
automorphism of K; by corollary 3 of tho 6, it maps R onto R, Ponto P,
so that it determines an automorphism A(a) of k = RIP; clearly a -+ A(a)
is a homomorphism of K x into the group of automorphisms of k. For
aeR x , A(a) is ~-+p(a)~p(a)-l, which is the identity since k is commu-
tative. Therefore, if a is any element of K X , and ord(a)=n, A(a)=).(n)n.
By corollary 2 of tho 2, § 1, applied to k and to the prime field in k, A(n)
must be of the form ~-+~pr; this means that we have, for every xeR:
nxn-l==x pr (P),
or, what amounts to the same:
n X == xpr n (P 2 ).
Take now M x as above, and put

n' = - }' Jl.pr n JI. - 1.
"lliix
In view of the above congruences, each one of the q - 1 terms in the sum
in the right-hand side is == n modulo p2. Since q . 1 E P, this gives
n' == (1- q)n == n (P 2 ),
which implies that n' is a prime element of K. At the same time, the
definition of n' gives
for all Jl.EM x , and therefore n' M X =M x n'. This completes the proof.
One could show, by a similar argument, that, if M x and N x are two
subgroups of K x of order q -1, there is a prime element n of K such
that nMx =Nxn.
COROLLARY 1. If K and M are as in theorem 7, and K is of charac-
teristic p, then M is a subfield of K. If at the same time K is commutative,
M is the algebraic closure of the prime field in K.
Let ko be the prime field in K, and let JI. be a generator of the group
MX. Then ko(JJ.) is a commutative field of characteristic p in which the
equation x q - X = 0 has q roots, viz., the elements of M; therefore, by
tho 2 of § 1, M is a field. If K is commutative, every element, other than 0,
of the algebraic closure of ko in K is a root of 1 of order prime to p,
again by tho 2 of§ 1; therefore, by theorem 7, it must be in M.
COROLLARY 2. Let K be a commutative p-field, q its module, and K'
an extension of K of finite degree, generated by roots of 1 of order-
prime to p. Then K' is unramified and cyclic over K, and its Galois group
over K is generated by an automorphism qJ which induces the permutation
JI.~Jl.q on the group of roots of 1 of order prime to pin K'.
By corollary 5 of tho 6, K' is a p-field. Let R, P, q, k, p, M x be as in

theorem 7 and its proof, and let R', P', q', k', p', M'x be similarly defined
for K'. By theorem 7, K' is generated over K by M' x, i. e. by the roots of
Xq'-l = 1; therefore it is a Galois extension of K, and an automorphism
of K' over K is uniquely determined by the permutation it induces on
M'x. By corollary 5 of tho 6, we have R=KnR', P=KnP'; we may
therefore identify k with a subfield of k', and then p is the mapping
induced by p' on R. Let IX be an automorphism of K' over K; as it maps
R' onto R' , P' onto P', and leaves fixed every element of R, it determines
an automorphism A.(oc) of k' over k. Then A., i.e. the mapping OC~A.(IX), is
a morphism of the Galois group of K' over K into that of k' over k. By
corollary 2 of tho 2, § 1, l(a) must be of the form e eqs. Therefore we

-+
have, for allllE M' x :
p'(a(Il)} = p'(Il)qs = p'{JlqS).
As p' induces on M' x, by theorem 7, an isomorphism of M' x onto k' x ,
this implies that a{Jl) = Il qs. In particular, if s = 0, i. e. if l(a) is the identity,
a is the identity; this shows that l is injective; therefore, if n is the degree
of K' over K, and f that of k' over k, we have n ~f As q' = qf, corollary 6
of tho 6 shows now that K' is unramified over K and that n = f, so that l
is an isomorphism of the Galois group of K' over K onto that of k' over k.
In view of corollary 2 of tho 2, § 1, this completes our proof.
COROLLARY 3. Let K and q be as in corollary 2; then a division
algebra of finite dimension over K is un ramified if and only if it is commu-
tative and can be generated over K by roots of 1 of order prime to p. For
every f~ 1, K has one and (up to an isomorphism) only one unramified
extension of degree f; this is the extension generated over K by a primitive
(qf -l)-th root of 1.
Let K' be an unramified division algebra of dimension f over K; let
q, q' be the modules of K and of K', respectively; then q' = qf, by corollary 6
of tho 6. Take a subgroup M' x of K' x of order q' -1; by theorem 7, it is
cyclic; take a generator Il of M' x , and put K" = K{Jl). Clearly K" is commu-
tative; as it contains M'x, its module is at least q', so that, by corollary 6
of tho 6, its degree over K is at least f; therefore K" = K', which, together
with corollary 2, proves the first part of our corollary. Now take any
f~ 1 ; put q' = qf, and call K' the extension of K generated by a primitive
(q' -l)-th root of 1, or, what amounts to the same, by the set M' x of all
the roots of X q ' -1 = 1; by theorem 7, its module is at least q', so that, by
corollary 6 of tho 6, its degree over K is at least f On the other hand, by
corollary 2, it is unramified and cyclic over K, and its Galois group over
K is generated by the automorphism cp defined there; as cpf induces the
identity on M' x , it is the identity, so that the degree of K' over K is at
most f Therefore it is f As the foregoing results show that every unrami-
fied extension of K of degree f must contain an extension isomorphic
to K', this completes our proof.
COROLLARY 4. Let K' be a finite extension of a commutative p-

field K; call f its modular degree over K, and e its order of ramification
over K. Then there is a unique maximal unramified extension K1 of K,
contained in K' ; it is of degree f over K, and K' is fully ramified of degree
e over K 1 .
This follows at once from the foregoing results, K 1 being generated
by the roots of 1 of order prime to p in K'.
DEFINITION 5. Let K be a commutative p-field, and K' an unramified

extension of K; the generator qJ of the Galois group of K' over K which
is defined by corollary 2 of theorem 7 is called the Frobenius automorphism
of K' over K.
e
In corollary 2 of theorem 6, one can take for a prime element 1t
of K, and then take for A the set M defmed in theorem 7. For commu-
tative fields of characteristic p, this gives the following:
THEOREM 8. Every commutative p-field of characteristic p is iso-
morphic to a field of formal power-series in one indeterminate with coeffi-
cients in a finite field.
Take notations as in theorem 7; corollary 1 of tho 7 shows that M is
e
a field with q elements. Taking = 1t and A = M in corollary 2 of tho 6,
we get for every xeK with ord(x)~n a unique series expansion
+00
x= LJli1ti,
i=n
with JlieM for all i~n. One verifies at once that the rules for the addition
and multiplication of such series are the usual ones for formal power-
series in algebra (or for convergent power-series in classical analysis).
Moreover, this is an isomorphism also in the topological sense if the
field of formal power-series is provided with its usual topology, that for
which the ring Ro of "integral" power-series (those containing no power
of the indeterminate with an exponent < 0), and the ideals generated
in it by the powers of the indeterminate, make up a fundamental system
of neighborhoods of O. We recall that, for this topology, the ring Ro of
integral formal power-series in one indeterminate over any finite field F
is compact, since the additive group of Ro is clearly isomorphic to the
product of enumerably many groups isomorphic to F; therefore the
corresponding field is locally compact. Thus theorem 8 shows that the
commutative p-fields of characteristic p are all of that type, so that (up
to an isomorphism) they are in a one-to-one correspondence with the
fmite fields Fq , with q=p", n~1.
By a local field, we will understand a commutative non-discrete
locally compact field. We have thus obtained a complete list of the local
fields of characteristic p> 1, while those of characteristic 0 are given by
theorem 5 of § 3; they are R, C and the finite algebraic extensions of the
fields Qp, for all p.
Using the same idea as in the proof of theorem 8, we give now one
more result for the non-commutative case.
PROPOSITION 5. Let K be a p-field, commutative or not, with the
maximal compact subring R. Then the center Ko of K is a p-field; if d
is the modular degree of Kover K o, its order of ramification over Ko is

also d, and its dimension over Ko is d2 ; it contains a maximal commutative
subfield K I which is unramified and of degree dover Ko. Moreover, if K I
is such, and if RI is its maximal compact subring, K has a prime element 1t
with the following properties: (a) 1td is a prime element of Ko;
(b) {l, ... , 1td - I} is a basis of K as a left vector-space over K I , and generates
R as a left R I-module; (c) the inner automorphism x -+ 1t - I X1t of K induces
on Klan automorphism lX which generates the Galois group of K lover Ko.
Let notations be as in theorems 6 and 7; choose M and 1t as in theo-
rem 7, and apply corollary 2 ofth. 6 to 1t and M; this shows that, for every
nEZ, each XEpn can be uniquely written as
+00
(5) X= L lLi1t i
i=n
with lLiE M for all i ~ n. Therefore an element of K is in the center Ko of K

if and only if it commutes with 1t and with every element of M (or, what
amounts to the same, with some generator of the cyclic group M)(). As
x-+1t- 1X1t induces a permutation on M, some power of it must induce
the identity on M; this amounts to saying that there is v> 0 such that
1t V commutes with every element of M. Then Ko contains 1tvn for all
nEZ; this proves that it is not discrete; as it is clearly closed in K, it is
locally compact; if now we consider K as a vector-space, hence an algebra,
over K o, we see, by corollary 2 of tho 3, § 2, that it has a finite dimension
over Ko; corollary 5 of tho 6 shows then that Ko is a p-field. Call q the
module of K o, d the modular degree of Kover K o, and Kl the field gen-
erated over Ko by M, or, what amounts to the same, by any generator of
the cyclic group M)(; as M)( is of order qd_l, such a generator is a
primitive (qd -1)-th root of 1, so that, by corollary 3 ofth. 7, K 1 is unrami-
fied of degree dover Ko. As x-+1t- 1 x1t induces a permutation on M,
and the identity on K o, it induces on Kl an automorphism lX of Kl
over Ko. An element of Kl commutes with all the elements of M; it
commutes with 1t if and only if it is invariant under lX; in other words, the
elements of K 1 which are invariant under lX are those of K o, so that lX
generates the Galois group of K lover Ko; it is therefore of order d, so
that, as we have seen above, ~ is in K o, and 1t is not in Ko unless v is a
V
multiple of d. Now take XE K and ILE M)(; write x in the form (5). Then we
have
+00
1L-1XIL= LlLi 1ti
i=n
where we have put

In this last formula, the last factor on the right-hand side belongs to
M)(, so that Il; is in M. In view of the unicity of the expansion (5) for
xe K, this shows that x = 1l- 1 X Il, i. e. that x commutes with Il, if and only
if Il; = Ili for all i. Now clearly, for each i, Il; = Ili ifand only if either Ili = 0
or ni commutes with Il. Consequently, x commutes with all elements of
M)( if and only if ni does so whenever III =/= O. In view of what has been
proved above, this is so if and only if lli=O whenever i is not a multiple
of d; we have then
As 1I:"eKo, x is then in the closure of K 1, hence in Kl itself, which is

therefore a maximal commutative subfield of K. It is also clear now, in
view of (5) and of the unicity of (5), that {1, n, ... , 11:" - l} is a basis of K as a
left vector-space over K 1 , that it generates R as a left Rcmodule, and
that nd is a prime element of K 1, hence also of Ko since it lies in Ko. As
this implies that the order of ramification of Kover Ko is d, it completes
the proof.
Notations being as in proposition 5, let qJ be the Frobenius auto-
morphism of K lover Ko; as this also generates the Galois group of K 1
over K o, we must have qJ = rx,r, with r prime to d and uniquely determined
modulo d. It will be shown in Chapter XII that, when Ko is given, d and
r may be chosen arbitrarily, subject to these conditions, and characterize
the structure of the division algebra K uniquely; in other words, two
division algebras of finite dimension over K o, with the center K o, are
isomorphic if and only if they have the same dimension d2 over K o, and
the integer r has the same value modulo d for both.
We conclude this Chapter with a result about the maximal compact
subrings in p-fields. We recall that, if R is any commutative ring, and x
an element of a ring containing R, x is called integral over R if and only
if it is a root of some monic polynomial over R, i. e. of some polynomial
with coefficients in R and the highest coefficient equal to 1.
PROPOSITION 6. Let K be a p-field and Ko a p-field contained in the
center of K; let R, Ro be the maximal compact subrings of K and of Ko.
Then R consists of the elements of K which are integral over R o.
Let x be in K and integral over Ro; this means that it satisfies an
equation
x"+a1x,,-1 + ... +a,,=O.
with aieRO for 1 ~i~n. Assume that x is not in R, i.e. that ordK(x)<O.
Then x =/= 0, and we have
1 = -a1x- 1- ... -a"x-";
§ 4. Structure of p- fields 23
here all the terms in the right-hand side are in the maximal ideal P of
R, so that lEP, which is absurd. Conversely, let x be any element of R.
By corollary 2 of tho 3, § 2, K has a finite dimension over Ko; therefore,
if we put K' = Ko(x), this is a commutative field and a finite extension of
Ko. Call F the irreducible monic polynomial, with coefficients in K o, such
that F(x)=O; in some algebraic closure of K', call K" the field generated
over Ko by all the roots of F, so that F splits into linear factors in K". As
K', K" are finite extensions of K o, they are p-fields; call R', R" their maxi-
mal compact subrings. Then R' = K' nR = K' nR"; as x is in R, it is in R'
and in R". As F is irreducible, every root x' of Fin K" is the image of x
under some automorphism of K" over Ko; as such an automorphism
maps R" onto R", all such roots are in R". Therefore all the coefficients
of F are in R"; as they are in K o, they are in Ro. This completes the proof.
If K is commutative, proposition 6 may be expressed by saying that
R is the integral closure of Ro in K.
Chapter II
Lattices and duality over local fields

§ 1. Norms. In this § and the next one, K will be a p-field, commu-
tative or not. We shall mostly discuss only left vector-spaces over K;
everything will apply in an obvious way to right vector-spaces. Only
vector-spaces of finite dimension will occur; it is understood that these
are always provided with their "natural topology" according to corollary 1
ofth. 3, Chap. 1-2. By tho 3 of Chap. 1-2, every subspace ofsuch a space V
is closed in V. Taking coordinates, one sees that all linear mappings of
such spaces into one another are continuous; in particular, linear forms
are continuous. Similarly, every injective linear mapping of such a space
V into another is an isomorphism of V onto its image. As K is not com-
pact, no subspace of V can be compact, except {O}.
DEFINITION 1. Let V be a left vector-space over the p-:field K. By
a K -norm on V, we understand a function N on V, with values in R +, such
that: (i) N(v)=O if and only if v=O; (ii) N(xv)=modK(x)N(v) for all
x E K and all v E V; (iii) N satisfies the ultrametric inequality
(1) N(v+w)~ sup(N(v),N(w»)
for all v, w in V.
On Kn, one defines a K-norm No by putting N o(x) =
SUPt~i~n (modK(xi») for all x=(Xt, ... ,xn) in Kn. As every vector-space
of finite dimension over K is isomorphic to a space K n, this shows that
there are K-norms on all such spaces.
One can obviously use any K-norm on V in order to topologize V, by
taking N(v - w) as distance-function.
PROPOSITION 1. Let V be a left vector-space of finite dimension over
the p-:field K. Then every K-norm N on V defines the natural topology
on V. In particular, every such norm N is continuous, and the subsets Lr
of V defined by N(v) ~ r are compact neighborhoods of 0 for all r>O.
As to the first assertion, in view of corollary 1 ofth. 3, Chap. 1-2, we
need only show that the topology defined by N on V makes V into a
topological vector-space over K. This follows at once from the inequality
N(x' v' -xv) ~ sup(modK(x')N(v' - v), modK(x' - x)N(v»)
which is an immediate consequence of def. 1. Therefore N is continuous,
and the sets Lr make up a fundamental system of closed neighborhoods
§l. Norms 25
of 0; in particular, Lr must be compact for some r>O. Now, for any

s>O, take aEK x such that modK(a) ~ r/s; then, as one sees at once, L.
is contained in a-I Lr; therefore it is compact.
COROLLARY 1. There is a compact subset A of V - {O} which con-
tains some scalar multiple of every v in V - {O}.
Call q the module of K, and take a K-norm N in V. If n is a prime
element of K, we have modK(n)=q-l, by tho 6 of Chap. 1-4, hence
N(nnv) = q-n N(v) for all nEZ and all VE V. Let A be the subset of V defined
by q-I ~ N(v) ~ 1; by proposition 1, it is compact; and, for every v fO,
one can choose nEZ so that nnvEA.
Corollary 1 implies the fact that the "projective space" attached
to V is compact.
COROLLARY 2. Let cp be any continuous function on V - {O}, with
values in R, such that cp(av) = cp(v) for all aEK x and all VEV-{O}. Then
cp reaches its maximum at some point v I of V - {O}.
In fact, this will be so if we take A as in corollary 1 and take for VI
the point of A where cp reaches its maximum on A.
°
COROLLARY 3. Let f be any linear form on V, and N a K-norm on V.
Then there is VI f in V, such that
(2) N(V)-I modK(f(v)) ~ N(VI)-I modK(f(v l ))
for all v fO in V.
This is a special case of corollary 2, that corollary being applied to
the left-hand side of (2). If one denotes by N*(f) the right-hand side of (2),
then N*(f) is the smallest positive number such that
modK(f(v)) ~ N*(f) . N(v)
for all VE V, and f --. N*(f) is a K-norm on the dual space of V, i.e. on the
right vector-space made up of the linear forms on V (where the addition
is the obvious one, and the scalar multiplication is defined by putting
(fa)(v) =/(v) a when I is such a form, and aEK).
By a hyperplane in V, one understands a subspace of V of codimen-
sion 1, i.e. any subset H of V defined by an equation l(v)=O, where I
is a linear form other than 0; when H is given, I is uniquely determined
up to a scalar factor other than 0. Now, if (2) is valid for all v f 0, and for a
given norm N, a given linear form f fO and a given VI fO, it remains so
°
if one replaces f by la, with aEK x , and VI by bV I with bEKx; in other
words, its validity for all v f depends only upon the hyperplane H
defined by 1=0 and the subspace W of V generated by vI; when it
holds for all v f 0, we shall say that Hand Ware N -orthogonal to each
other.
26 Lattices and duality over local fields II
PROPOSITION 2. A hyperplane H and a subspace W of V of dimen-

sion 1 are N-orthogonal if and only if V is the direct sum of Hand W, and
N(h+w)=sup(N(h),N(w)) for all heH and weW.
Let H be defined by f(v)=O, and assume first that Hand Ware
N-orthogonal. Then (2) is satisfied if one replaces Vl in it by any weW
other than 0. This implies that f(w) is not 0, for otherwise f would be 0;
therefore V is the direct sum of Hand W. Now replace v in (2) by h+w
with heH; as f(h+w)= f(w) =1=0, (2) gives N(h+w)~N(w). Applying
the ultrametric inequality (1) to h=(h+w)+( -w), we get N(h)~N(h+w);
applying it to h + w, we get the formula in our proposition, for w =1= 0; as
it is trivial for w = 0, this proves the necessity of the condition stated there.
Now suppose that V is the direct sum of Hand W; take any v =1=0, and
write it as V= h + w with he H and we W, so that f(v) = f(w). If w =1= and
N(h+w) ~ N(w), then we have
°
N(V)-l modK(f(v)) ~ N(W)-l modK(f(w)).
As the right-hand side does not change if we replace w by any generator
Vl of W, this shows that (2) holds for any such Vl' and any v not in H. For
veH, i.e. w=O, it holds trivially. This completes the proof.
Accordingly, we shall also say that two subspaces V, V' of V are
N -orthogonal to each other whenever V is the direct sum of V and V;
and N(v' + v") = sup(N(v'),N(v")) for all v' e V and all v" e V'.
PROPOSITION 3. Let V be of dimension n over K, and let N be a
K-norm on V. Then there is a decomposition V= Vl + ... + v,. of V into
a direct sum of subspaces V; of dimension 1, such that N(~>i)=SUPiN(Vi)
whenever Vie V; for 1 ~ i ~ n. Moreover, if Wl = V, W2 , ... , w,. is a sequence
of subspaces of V such that W; is a subspace of W; _ 1 of codimension 1 for
2 ~ i ~ n, then the V; may be so chosen that W; = V; + ... + v,. for all i.
This is clear for n= 1. For n> 1, use induction on n. By corollary 3 of
prop. 1, we may choose Vl so that the space Vl generated by V l is N-
orthogonal to W2 ; then, by prop. 2, N(V'l +w2)=sup(N(v~),N(W2)) when-
ever V'l e Vl , W2 e ~. Applying the induction assumption to the K -norm
induced by N on W2 , and to the sequence W2 , ... , w,., we get our result.
COROLLARY. To every subspace W of V, there is a subspace W' which
is N -orthogonal to W.
Take a sequence Wl , .•. , w,., as in proposition 3, such that W is one
of the spaces in that sequence, say W;. Take the V; as in proposition 3.
Then the space W = Vl + ... + V; _1 is N -orthogonal to W.
PROPOSITION 4. Let N, N' be two K-norms in V. Then there is a
decomposition V = Vl + ... + v,. of V into a direct sum of subspaces V; of
§ 2. Lattices 27
dimension 1, such that N(~:Vi)=SUPiN(Vi) and N'(~:Vi)=SUPiN'(Vi)

whenever ViE "i for 1 ~ i ~ n.
For n=l, this is clear. For n>l, use induction on n. Applying co-
rollary 2 of prop. 1 to qJ=NjN', we get a vector VI +0 such that
N(v)N'(v)-1 ~ N(vl)N'(VI)-1
+
for all V 0; call VI the space generated by vI . By the corollary of prop. 3,
there is a hyperplane W which is N -orthogonal to VI; then, if f = 0 is
an equation for W, we have
N(V)-I modK(f(v») ~ N(VI)-I modK(f(vl»)
for all v +o. Multiplication of these two inequalities gives
N'(V)-I modK(f(v») ~ N'(VI)-I modK(f(v l »),
which means that W is N'-orthogonal to VI. Applying now prop. 2 to N,
VI and W, and also to N', VI and W, and applying the induction assump-
tion to the norms induced by Nand N' on W, we get the announced
result.
One should notice the close analogy between propositions 3 and 4, *
and their proofs, and the corresponding results and proofs for norms
defined by positive-definite quadratic forms in vector-spaces over R,
or hermitian forms in vector-spaces over C or H. For instance, prop. 4
corresponds to the simultaneous reduction of two quadratic or hermitian
forms to "diagonal form".
§ 2. Lattices. In this sectiOJ;l, K will again be a p-field, and we shall
use the notations introduced in Chapter I. In particular, we write R
for the maximal compact subring of K, P for the maximal ideal in R,
q for the module of K, and 11: for a prime element of K. For neZ, we
write P" = 11:" R = R 11:".
We shall be concerned with R-modules in left vector-spaces of finite
dimension over K; if V is such a space, an R-module in V is a subgroup
M of V such that R·M =M.
PROPOSITION 5. Let V be a left vector-space of finite dimension over K.
Let M be an R-module in V, and call W the subspace of V generated
by Mover K. Then M is open and closed in W; it is compact if and only if
it is finitely generated as an R-module.
Let ml , ... , mr be a maximal set of linearly independent elements
over K in M; they make up a basis of W over K. By tho 3 of Chap. 1-2,
the set R mI + ... + R mr is an open subgroup of W; as both M and W - M
are unions of cosets with respect to that subgroup, they are both open.
If M is compact, it is the union of finitely many such cosets and therefore
finitely generated; the converse is obvious.
On the other hand, in view of corollary 2 of tho 6, Chap. 1-4, a closed

subgroup X of V satisfies R· X =X if and only if 1tX eX and aX eX
for every a in a full set A of representatives of RIP in R. In particular, if
q=p, i.e. if RIP is the prime field, we may take A={0,1, ... ,p-1}, and
then aX eX for all aeA, so that X is then an R-module if and only if
1t X e X. In the case K = Qp, we may take 1t = p, and then every closed
subgroup of V is a Zp-module.
In K itself, viewed as a left vector-space over K, every R-module,
if not reduced to {O}, is a union of sets P", and thus is either K or one of
these sets.
DEFINITION 2. Bya K-Iattice in a left vector-space V of finite dimen-
sion over K, we understand a compact and open R-module in V.
When no confusion can occur, we say "lattice" instead of K-Iattice.
If L is a p-field contained in K, every K-Iattice is an L-Iattice; the converse
is not true unless L = K.
Clearly, if L is a lattice in V, and W is a subspace of V, L n W is a
lattice in W; similarly, if f is an injective linear mapping of a space V
into V, f- 1 (L) is a lattice in V; if f is a surjective linear mapping of V
onto a space V', f(L) is a lattice in V'.
If N is a K-norm in V, the subset Lr of V defined by N(v)::::;; r is a
K-Iattice in V for every r>O. In fact, (iii), in def. 1 of§ 1, together with (ii)
applied to X= -1, shows that it is a subgroup of V; then (ii) shows that
it is an R-module, and prop. 1 of § 1 shows that it is a compact neighbor-
hood of 0 in V, hence open since it is a subgroup of V. This has a converse;
more generally, we prove:
PROPOSITION 6. Let M be an open R-module in V; for every ve V, put

N M(v)=infxeKx ,xveMmodK(x)-I.
Then the function N Mon V satisfies conditions (ii) and (iii) in definition 1
of § 1, and M is the subset of V defined by N M(V) ::::;; 1; N M is a K-norm if
and only if M is a K-lattice in V.
For aeK x , we have X· aveM if and only if x=ya- 1 with yveM;

this gives NM(av)=modK(a)NM(v); as NM(O)=O, this is also true for
a=O. Therefore NM satisfies (ii) ofdef. 1. For each veV, call Mv the set
ofthe elements x of K such that xveM; as this is an open R-module in K,
it is either K or a set P" with some neZ. If Mv=K, N M(V)=O; if Mv=P",
we have xveM ifand only ifmodK(x)::::;; q-", so that N M(V)=q". In parti-
cular, we have N M(V)::::;; 1 ifand only if Mv~R, hence if and only ifveM.
Let v, w be in V and such that NM(v) ~NM(W); then MvcMw, so that
xveM implies xweM, hence also x(v+w)eM; therefore Mv+w~Mv'
§ 2. Lattices 29
hence N M(V + w) ::::; N M(V); this proves (iii) of def. 1. Finally, M is a

K-Iattice if and only if it is compact, and N M is a K-norm if and only if
N M(V»O for all v-=foO, i.e. if and only if Mv-=foK for v-=foO. By prop. 1 of§ 1,
if N Mis a K-norm, M is compact. Conversely, assume that M is compact,
and take v-=foO; then Mv is the subset of K corresponding to (Kv)nM
under the isomorphism x ~ x v of K onto K v; therefore M v is compact
and cannot be K. This completes our proof.
COROLLARY 1. An open R-module M in V is a K-Iattice if and only

if it contains no subspace of V other than O.
It has been shown above that, if M is not compact, N M cannot be a
K-norm, so that there is v-=foO in V such that N M(V)=O, hence Mv=K, i.e.
KvcM. Conversely, as every subspace of V, other than 0, is closed in V
and not compact, no such subspace can be contained in M if M is compact.
COROLLARY 2. Let M be an open R-module in V; let W be a maximal

subspace of V contained in M, and let W be any supplementary subspace
to W in V. Then Mn W is a K-Iattice in W, and M=(MnW)+ W.
The first assertion is a special case of corollary 1; the second one is

obvious.
Proposition 6 shows that every K-Iattice in V may be defined by an
inequality N(v)::::;l, where N is a K-norm; this was our chief motive in
discussing norms in § 1. For a given K-Iattice M, the norm N M defined
in prop. 6 may be characterized, among all the norms N such that M is
the set N(v)::::; 1, as the one which takes its values in the set of values
taken by mod Kon K, i.e. in the set {O}U{qn}nEz.
PROPOSITION 7. If V has the dimension lover K, and if L is a K-Iattice

in V, then V has a generator v such that L=Rv.
Take any generator w of V; the subset Lw of K defined by XWEL
must be of the form pn; taking v=nn w, we get L=Rv.
THEOREM 1. Let L be a K-Iattice in a left vector-space V of dimen-

sion n over K. Then there is a basis {v i , ... ,vn} of V such that L= IRv i.
Moreover, if Wi = V, W2 , ••• , w" is any sequence of subspaces of V such
that Wi is a subspace of Wi-i of codimension 1 for 2~i~n, the Vi may
be so chosen that, for each i, {vi, ... ,vn} is a basis of Wi.
Take a K-norm N such that L is defined by N(v)~ 1. Choose sub-
spaces Vu ... , Vn of V as in prop. 3 of § 1; then L= I(LnV;). Applying
prop. 7 to V; and LnV; for each i, we get the basis (vJ
Theorem 1 may be applied for instance whenever K' is a p-field con-

taining K, and R' is the maximal compact subring of K'. Clearly, if K'
is viewed as a left vector-space over K, R' is a K-Iattice in K'. Therefore
there is a basis {Yt , ... ,Yn} of K' over K such that R' = IRYj; then, if
we write, for any yeR', YYj= IajjYj, with aijeK for 1 ~i,j~n, all the
ajj must be in R. In particular, if K is commutative, these relations, which
hold in the commutative field K{y), imply det{y ·1 n -A)=O, where 1n
is the unit matrix and A=(a jj), so that we get an alternative proof for
the second part of prop. 6, Chap. 1-4.
THEOREM 2. Let L, L' be two K -lattices in a left vector-space V of
finite dimension over K. Then there is a basis {Vt, ... , vn} of V, and a
sequence of integers (vt, ... ,vn), such that L= L:Rvj and L'= L:r; Vj.
Take K-norms N, N' such that L is defined by N(v)~ 1 and L' by
N'(v)~1. Choose subspaces Vt , ... , v.. of V as in prop.4 of § 1; then
L=L:(Ln~) and L'=L:(L'n~). For each i, apply prop. 7 to ~ and
Ln~, and also to ~ and L'n~; this gives Vi such that Ln~=Rvi and
v; such that L'n ~=Rv;. Writing v;=xjVj with xjeKx, and putting
vj=ord(xj), we get integers Vj with the required property.
COROLLARY 1. Let V and L be as in theorems 1 and 2, and let M be
an R-module in V. Then there is a basis {vt> ... ,vn } of V over K, and there
are integers r, sand Vt,. •. ,V" such that O~r~s~n, L=IRvj and
, s
M= L:rJvj +
j=t
L:
,,=,+t
Kv".
Let W be the subspace of V generated by M, and W' the maximal

subspace contained in M; call s the dimension of W, and r the codimen-
w..
sion of W' in W. In tho 1, choose the sequence Wt , ... , so that it includes
Wand W'.Then tho 1 gives us a basis {Wt, ... , wn} of V which generates L
as an R-module and contains bases for Wand for W'; renumbering this
basis in an obvious manner, we may assume that {wt, ... ,ws } is a basis
for Wand that {w,+t, ... ,ws } is one for W'. Call W" the subspace of V
with the basis {w 1> ••• ' w,}. By prop. 5, M is open in W; therefore, by
corollary 2 of prop. 6, we have M = M' + W', where M' = M n W" is a
K-Iattice in W". Applying now th.2 to M' and to L'=LnW", we get a
basis {Vt , ... ,V,} for W", and integers V t , ... , v" such that 1:= IRvj and
M' = L:ri Vj. Taking Vj=Wj for i>r, we get the basis required by our
corollary.
COROLLARY 2. Every finitely generated R-module IDl is the direct
sum of finitely many summands, each of which is isomorphic either to R
or to a module R/r with v>O. Moreover, the number of summands of
type R, and, for each v, the number of summands of type R/r, are uniquely
determined when IDl is given.
Let IDl be generated by elements ml, ... ,mn' Take a vector-space V

of dimension n over K, with a basis {vl, ... ,vn}; put L=IRv j • Then
the formula
where the Xj are taken in R for 1 ~ i ~ n, defines a morphism of L onto IDl;

therefore IDl is isomorphic to LIM, where M is the kernel of that mor-
phism. Apply now corollary 1 to Land M; as MeL, we have vj ~ 0 for
1 ~j~r, and r=s. Our first assertion follows from this at once. As to
the second one, put IDlj = njIDl for all i~O; as these are R-modules, their
quotients 9lj=IDlftJli+l are R-modules; as nn=O for all ne91j, 9lj may
be regarded as a module, i.e. as a vector-space, over the field k=RIP;
as such, it has a dimension nj, which depends only upon IDl and i. Write
now IDl as a direct sum of modules Rand RIPY, in numbers respectively
equal to No and Ny; then one sees at once that nj=N o+ INy. There-
v>i
fore No=nj for i large enough, and Ny=ny_l-ny.
COROLLARY 3. In corollary 1, the integers r,s, Vi"'" vr depend only

upon Land M.
As s is the dimension of the subspace W generated by M, and s - r is
the dimension of the maximal subspace contained in M, they depend
only upon M. Now put Ll =LnW, and take i~O such that n jLl eM;
our assertion follows now at once from the application of corollary 2
to the R-module MI(n j Ll)'
In corollary 2, the number of summands of IDl isomorphic to R is
called the rank ofIDl; with this definition, we have:
COROLLARY 4. Let IDl be a finitely generated R-module, and IDl' a sub-

module of IDl. Then the rank of IDl is the sum of those of IDl' and of IDljIDl' .
As in the proof of corollary 2, write IDl as LIM, where L is the lattice

IRvj in the vector-space V with the basis {Vl""'V n}, and M is an
R-module. Then the inverse image of IDl' in L is an R-module L; and the
three modules in our corollary are respectively isomorphic to LIM,
L,/ M and LIL: Let W, V' be the subspaces of V respectively generated
by M and by L'; then, as corollary 1 shows at once, the ranks of LIM,
L,/M and LIL' are respectively the codimensions of W in V, of W in V'
and of V' in V.
§ 3. Multiplicative structure of local fields. Let notations be as above;

then, for each integer n ~ 1, the set 1 + pn of the elements x of R which
are == 1 (P") is clearly an open and compact subgroup of R x , and these
subgroups make up a fundamental system of neighborhoods of 1 in R x .
Moreover, th.7 of Chap. 1-4 shows that R x = M x • (1 + P) if M x is any

subgroup of order q -1 of R x, and tho 6 of Chap. 1-4 shows that
K x = II· R x if II is the discrete subgroup of K x, isomorphic to Z,
which is generated by any prime element 1t of K. In these formulas, the
products are "semidirect".
From now on, until the end of this §, it will be assumed that K is
a commutative p-field; then the above products are direct products, so
that we may write K X = IIxRx and R X =M x x(l+P); moreover, by
tho 7 of Chap. 1-4, M x is now the group of roots of 1 of order prime to p
in K. Consequently, the investigation of the structure of K x amounts
to that of 1 + P.
Take any xe 1 + P; then, for every ae Z, x" is in 1 + P, and the map-
ping a--+x" is a homomorphism of the additive group Z into the multi-
plicative group 1 + P; as lemma 5 in the proof of tho 7, Chap. 1-4, shows
that x"el +p"+ 1 whenever a=O (p"), i.e.lalp~p-II, this homomorphism
is continuous when Z is provided with the p-adic topology, i.e. that
induced on Z by Qp; as 1 + P is compact, it can therefore be uniquely
extended to a continuous homomorphism, which we again denote by
a-+ x", of the additive group Zp into the multiplicative group 1 + P. If
xel +P", x" is in 1 +P" for all aeZ, hence for all aeZp- From this, using
the formula yh(X"t 1 =(yx-1)bXb-lI, one concludes at once, in the usual
manner, that the mapping (a,x)-+x" of Zpx(l+P) into l+P is con-
tinuous. One verifies then immediately that this mapping defines, on
the group 1 + P, a structure of Zp-module (the "addition" of vectors
being written multiplicatively, and the "scalar multiplication" by elements
of Zp being written exponentially).
PROPOSITION 8. If n is any integer prime to p, and v any integer ~ 1,

x-+ x" induces on 1 + r an automorphism of 1 + r; (K X)" is an open sub-
group of K X , of index n· (n,q-l) in K X ; if n divides q-l, that index is n 2 •
The first assertion is a special case of the fact that X-+X" is an auto-
morphism of 1 + pv whenever a is an invertible element of Zp; it implies
that (K X)" is open in K x • Moreover, as we have seen above, K x is the
direct product of the group II, which is isomorphic to Z, of the cyclic
group M X of order q-l, and of l+P; therefore the index of (KX)" in
K X is the product of the similar indices for II, M X and 1 +P; clearly,
these are respectively equal to n, to the g.c.d. (n,q-l) of nand q-l,
and to 1. This proves our proposition.
We will now determine the structure of the Zp-module 1 +P; this
depends upon the characteristic of K. If K is of characteristic 0, it is
a finite algebraic extension of Qp, so that, as we have observed, its
maximal compact subring may be regarded as a Qp-Iattice in K; tho 1
of § 2 shows then that it is the direct product of factors, all isomorphic

to Zp, whose number is equal to the degree of Kover Qp.
PROPOSITION 9. Let K be a commutative p-field of characteristic 0,

with the maximal compact subring R. Then there is an integer m~O such
that 1 + P, as a (multiplicatively written) Zp-module, is isomorphic to the
(additively written) Zp-module R x (ZJpm Zp); m is then the largest integer
such that K contains a primitive pm-th root of 1.
For any XER and aEN, the binomial formula may be written as:
(l+x)Q=I+ax+ ax t (~=l)
1=2
xi-lli.
For i ~ 2, call ph the largest power of p dividing i; if h = 0, i-I> h; if

h > 0, then, as i ~ ph, one verifies at once that i-I> h except for i = P = 2,
so that 2(i-l»h in all cases. Therefore, in the above formula, the sum
in the last term in the right-hand side is in pR whenever XEp2 R. This
gives, for XEp2 R, aEN:
(3) (1 +x)Q= 1 +ax (paxR),
which must remain valid, by continuity, for all XEp2 Rand aEZ p, since
N is dense in Zp- Now call d the degree of Kover Qp; by tho 1 of § 2,
we can find a basis {Vl , ... ,Vd} of Kover Qp such that R= IZpVi. By (3),
we have now, for 1 ~i~d, v~ 1, a;EZp:
(pv+2 R),
and therefore:
d
f1 (l+p2 Vi )pV-I Q; =1+pv+l L a;vi
d
(4) (pv+2R).
;= 1 i= 1
It follows from this that, if x 1 E p2 R, we can define by induction a sequence

(X l ,X2, ... ), with XvEpv+l R for all v~l, by putting, for each v:
with aviEZp for 1 ~i~d, and then

l+xv+l =(I+x v)n(l+p2 vi ) -p'-la\,;.
;
It is now clear that we have
(5) 1+ Xl = n(l + p
i
2 V;)b i
where the bi are given, for 1 ~i~d, by

+00
bi = L pV'-lavi .
v=l
This shows that, as a multiplicative Zp-module, the group 1 + p2 R is
generated by the d elements 1 + p2 Vi; as it is an open subgroup of the
compact group 1 + P, hence of finite index in 1 + P, and as 1 + P, as a
Zp-module, is generated by the elements 1 + p2 Vi and by a full set of
representatives of the classes modulo 1 + p2 R in 1 + P, this implies that
1+P is finitely generated. Now assume that (5) can hold with Xl =0
while the bi are not all 0; then, taking for v -1 the smallest of the orders
of the bi in Qp, we can write bi=pV-l aj with v~ 1, ajEZ for 1 ~i~d,
and the aj not all in pZp. Then (4) gives Lajvj=O (PR), i.e. ~(p-l aj)vjER,
which contradicts the defmition of the Vj' This shows that 1 + p2 R, as
a Zp-module, is the free module generated by the 1+ p2 Vi> so that it is
isomorphic to (Zpt We can now apply corollary 4 of tho 2, § 2, to the
Zp-modules 1 + P and 1 + p2 R. As their quotient is finite, it is of rank 0;
as 1 + p2 R is isomorphic to (Z/' it is of rank d. Therefore 1 + P is of
rank d, hence, by corollary 2 of tho 2, § 2, the direct product of d factors
isomorphic to Zp and of finitely many factors, each isomorphic to a
module ZJpv Zp. As the latter are finite groups, their product is the
group of all elements of finite order in 1 + P and is itself a finite group,
whose order is a power of p; it is therefore the group of all roots of 1
in 1 +P; by lemma 1 of Chap. 1-1, if pm is the largest of the orders of
its elements, it is cyclic of order pm, hence, as a Zp-module, isomorphic
to Z"Ipmzp. Finally, writing K as the direct product of n, M X and
X
1 + P, we see that any root of 1 in K whose order is a power of p must

be in 1 + P. This completes the proof.
COROLLARY. Let K be as in proposition 9. Then, for every integer
n ~ 1, (K x)n is an open subgroup of K x , of finite index in K x , and that
index is n' (n,r)' modK(n)-l if r is the order of the group of all roots
of 1 in K.
Clearly the latter group is the direct product of M x and of the
group of roots of 1 in 1 +P, which is of order pm; therefore it is cyclic
of order r=(q-1)pm, and K X is the direct product of n, of that group,
and of a Zp-module isomorphic to R. Now nR is an open subgroup of
the additive group R, whose index in R, by the definition of modK, is
modK(n)-l. The conclusion follows from this at once, by the same
argument as in the proof of prop. 8.
PROPOSITION 10. Let K be a commutative p-field of characteristic p.
Then 1 + P, as a Zp-module, is the direct product of a countably i1ifinite
family of modules isomorphic to Zp.
§ 4. Lattices over R 35
By tho 8 of Chap. 1-4, we may regard K as the field of formal power-

series in one indeterminate 1t, with coefficients in the field F q with q = pI
elements. Here it is easy to give explicitly a family of free generators
for the Zp-module 1 +P. In fact, take a basis {1X 1 , ... ,IX/ } for Fq over the
prime field F p" As generators of 1 + P, we take the elements 1 + IX/X",
where 1 ~ i ~ f, n running through the set of all integers > 0, prime to p.
For any N>O, put N =np', with v~O and n prime to p. For any integers
ai~O (1 ~i~f), we have
n
j=l
(l +lX j 1t"t iPV = n(1+pj~)aj==1 +a:ajpa~
j I
(p N + 1)
with pj=lXf. As x-+xpv is an automorphism of Fq over F p, the pj also

make up a basis of Fq over Fp; thus, for any given lXeFq, one may, in
one and only one way, choose integers aj such that O~aj<p and that
Lajpj=lX. Now take any Xl eP; we define inductively a sequence
(X 1'X 2' ...), with xNepN for all N~ 1, as follows. For each N, putting
N = n p' with n prime to p as above, we choose the integers aj so that
O~aj<pfor 1~i~fand that
YN = n(1 +
j
IXj1t"tjPV == 1 +XN (p N + 1),
which can be done in one and only one way in view of the foregoing
remarks, and put then
1 +X N + 1 =(1 +XN)y;l.
One sees at once, putting these formulas together, that they give for
1 + X 1 an expression as a convergent infinite product of factors of the
form (1 +lXj1t"t, with 1 ~i~f, n prime to p, and beZp • Moreover, the
above calculations show also that this expression is unique, which proves
our assertions.
§ 4. Lattices over R. The concept of lattice, as developed for p-fields

in §§ 1-2, cannot be applied to R-fields. The appropriate concept is here
as follows:
DEFINITION 3. By an R-lattice in a vector-space V of finite dimen-
sion over an R:field, we understand a discrete subgroup L of V such that
VILis compact.
We have to recall here some elementary facts about discrete sub-

groups. Let G be a topological group, r a discrete subgroup of G, and
<p the canonical mapping of G onto Glr. Then, if U is a neighborhood
of the neutral element e in G, such that U- 1 • U contains no element
of r other than e, lP induces, on each set of the form g U with geG, a

homeomorphism of that set onto its image in Gjr; one expresses this by
saying that lP is a "local homeomorphism"; one may say that it is a
"local isomorphism" in U if r is normal in G, since in that case it maps
the group law in G onto the group law in Gjr. Assume that G is locally
compact, and let a right-invariant measure 0( be given on G. Then it is
easily seen that there is one and only one measure 0(' on Gjr such that,
whenever X is a measurable subset of G which is mapped by lP in a one-
to-one manner onto its image X'=lP(X) in Gjr, O('(X') is equal to O(X);
in particular, this will hold for every measurable subset of every set g U,
where U is as above. Then, if f is any continuous function with compact
support in G, we have
(6) fj(g)dO(g)= J( L f(gY»)dO(I(g);
~ GfT YET"
here we have put g= lP(g), and the integrand in the right-hand side,
which is written as a function of g but is constant on cosets g r, is to
be understood as a function of g. This, in fact, is clear ifthe support of f
is contained in any set g U, and the general case follows from this at
once; also, as well known in integration theory, the validity of (6) for
continuous functions with compact support implies its validity for all
integrable functions, and for all measurable functions with values in R+.
Clearly, 0(' is invariant under the action of G on Gjr if and only if 0( is
left-invariant; this will be so, in particular, whenever Gjr is compact,
since then Gjr is a set of finite measure which is invariant under the
action of G. Then, if at the same time r is normal in G, (1.' is a Haar measure
on Gjr.
Things being as above, 0(' will be called the image of 0( in Gjr; we will
denote this image simply by 0( when no confusion is likely. The following
lemma (which takes the place of what was known as Minkowski's theo-
rem in classical number-theory) is now obvious:
LEMMA 1. Let G be a locally compact group with a Haar measure 0(;
let r be a discrete subgroup of G, such that Gjr is compact; let X be a
measurable subset of G such that O(X)> O(Gjr). Then there are two
distinct elements x, x' of X such that x - 1 x' e r.
One should only note that, since Gjr is compact, any right-invariant
measure on G is also left-invariant; therefore the Haar measure 0( is
bi-invariant, and its image in Gjr is well-defined.
LEMMA 2. Let G, 0( and r be as in lemma 1, and let r l be a discrete
subgroup of G, containing r. Then r has a finite index [rl :r] in r l , and
this is given by
§ 4. Lattices over R 37
As G/r is compact, there is a continuous function fo ~ 0 with compact

support on G, such that
for all ge G. Then the function f = fo/f1 is continuous on G, has the same
support as fo, and is such that Lf(gy), where the sum is extended to
all yer, is 1 for all g; this implies that the similar sum, extended to all
yer1, has the constant value [r1 :r]. If now we apply (6) to G, f and r,
and also to G, f and r 1 , we get the result in our lemma.
From these facts, we now deduce the following classical result about
R-Iattices:
PROPOSITION 11. Let L be a subgroup of a vector-space V of dimen-

sion n over R. Then the following three statements are equivalent: (i) Lis
an R-lattice in V; (ii) L is discrete in V, finitely generated, and contains
a basis jor V over R; (iii) there is a basis {V1' ... , vn } of V over R which
generates the group L.
Assume (iii), and consider the isomorphism
(7)
of Rn onto V; L is the image of zn under that isomorphism; therefore it

is discrete in V, and V/L is isomorphic to (R/Z)", hence compact. There-
fore (iii) implies (i) and (ii). Now assume (i); let W be the subspace of V
generated by L, and call W a supplementary subspace to W in V. Then
V, as a locally compact group, is the direct product of Wand W, and L
is a discrete subgroup of W; therefore V/L is isomorphic to the direct
product of W/L and W. This cannot be compact unless W is so; then
W' must be {O}, and W = V, so that L contains a basis of V over R. Now
let ex be the Haar measure on V which is such that ex(V/L) = 1. For every
basis B= {V1' ... , vn } of V, contained in L, call ({JB the isomorphism of
Rn onto V defined by (7); this maps Z" onto the sublattice LB of Lgener-
ated by B, and maps the Lebesgue measure A. on Rn onto some scalar
multiple mii 1 ex of ex. As A.(Rn/zn) = 1, we have mii 1 ex(V/L B ) = 1; by lemma 2,
this shows that mB is the index of LB in L. Now choose B so that this
index has the smallest possible value; it will be shown that then LB = L.
In fact, assume that L contains a vector w, not in LB , and write W= Laivi
with coefficients ai in R; as w is not in LB , at least one of the ai' say a1'
is not in Z; replacing then w by w-mv 1 with meZ, m<a 1<m+l, we
may assume that O<a 1< 1. Now put V'l =w, v;=v i for 2~i~n, and
B' = {V'l' ... , v~}; clearly B' is a basis for V, contained in L. A trivial cal-
culation shows that If>ii 10 If>B' is the automorphism of Rn given by

(Xl' ... , xn) - (a 1X 1,X 2 +a 2 x 1, .. ·, xn+ anx 1),
whose module is a1 (cf. corollary 3 ofth. 3, Chap. 1-2). Take a measurable
set X in Rn; put Y = If>B(X) and Y = If>B'(X), By the definition of mB, mB"
we have a(Y)=mBA,(X), a(Y)=mB'A,(X); therefore If>B'0lf>ii1, which maps
Y onto Y, has the module mB,/mB. Since If>B'0lf>ii 1 can be written as
If>B,O(lf>ii 10 If>B')0lf>ii, 1, it has the same module as If>ii 10 If>B'' Thus we get
mB,/mB=a 1 < 1, which contradicts the definition of B. This completes
the proof, as it shows that (i) implies (ii) and (iii).
§ S. Duality over local fields. Among the most important properties

of commutative locally compact groups are those which make up the
content of the "duality theory". We recall that, if G is such a group, a
character of G (in the sense of that theory) is a continuous representation
of G into the multiplicative group of complex numbers of absolute value 1.
Ifg* is such a character, its value at a point g of G will frequently be written
as <g,g*)G' for which we write <g,g*) if there is no danger of confusion.
We shall write the group law on G additively; and, on the set G* of the
characters of G, we put a commutative group structure, also written
additively, by writing
<g,g! +gDG= <g,g!>G' <g,gDG;
one should note that the neutral element of G*, which is denoted by 0
in this additive notation, corresponds to the "trivial" character of G
with the constant value 1 on G. One topologizes G* by assigning to it
the topology of uniform convergence on compact subsets of G; this
makes it into a locally compact group, called the topological dual of G,
or simply its dual if there is no danger of confusion. Conversely, the
characters of G* are the functions g*-<g,g*)G' for all gEG, and this
determines an isomorphism between G and the dual of G*. In other
words, G may be identified with the dual of G* by writing
<g*,g)G* = <g,g*)G'
and it will always be tacitly assumed that they are so identified. The
group G is compact if and only if G* is discrete; therefore G is discrete if
and only if G* is compact.
If H is any closed subgroup of G, the characters of G which induce
the trivial character on H make up a closed subgroup of G*, which will
be denoted by H * and is said to be associated with H by duality; it is
isomorphic to the dual of G/H. When G is regarded as the dual of G*,
the subgroup of G associated with H * is then H itself, which is therefore
§ 5. Duality over local fields 39
isomorphic to the dual of G*IH*. As H is open in G if and only if GIH

is discrete, we see that it is so if and only if H* is compact; consequently,
H * is open in G* if and only if H is compact. Similarly, H is discrete if
and only if G*IH* is compact, and GIH is compact if and only if H* is
discrete.
All this may be applied to the additive group of any left vector-
space V of finite dimension over a non-discrete locally compact field K
(commutative or not). In that case, if V* is the topological dual of V, and
ifv*eV*, then, for every aeK, the function v-(av,v*)v on Vis clearly
again a character of V, which we will denote by v* a; one verifies at once,
by going back to the definitions, that this makes V* into a right ve.ctor-
space over K; by corollary 2 of tho 3, Chap. 1-2, its dimension must be
finite. In other words, the structure of V* as a right vector-space over K
is defined by the formula
(8) (av,v*)v=(v,v*a)v.
Conversely, if V and V* are dual groups, and V* has a structure of right
vector-space over K, (8) may be used in order to define V as a left vector-
space over K. Thus we may still identify V with the dual of V* when their
structures as vector-spaces over K are taken into account. If L is any
closed subgroup of V, the subgroup L* of V* associated with L by
duality consists of the elements v* of V* such that (v,v*)v= 1 for all
veL; in view of (8), this implies that, if L is a left module for some subring
of K, L* is a right module for the same subring, and conversely. In parti-
cular, if K is a p-field and R is the maximal compact subring of K, L is a
left R-module if and only if L* is a right R-module. As we have seen that
L is compact and open in V if and only if L* is so in V*, we see that L
is a K-Iattice if and only if L* is one. When that is so, we say that the
K -lattices Land L. are dual to each other; then a Land L* a - 1 are dual
to each other for every aeKx. On the other hand, if K is R, C or H,
then, clearly, L is an R-Iattice if and only if L. is one.
On the other hand, if V is as above, we may consider its algebraic
dual V', which is the space of K-linear forms on V; as well-known, if we
denote by [v, v']v the value of the linear form v' on V at the point v of V,
we can give to· V' a "natural" structure of right vector-space over K by
means of the formula
[av,v' b]t =a[v,v']v b,
valid for all ve V, v' e V' and all a, b in K. If Xis any character of the additive
group of K, then, for every v' e V', there is an element v* of the topological
dual V* such that (v,v*)v=X([v,v']v) for all veV. We shall use this
operation in order to establish the relation between the algebraic and
the topological dual.
THEOREM 3. Let K be a non-discrete locally compact field, and V

a left vector-space of finite dimension n over K; let X be a non-trivial
character of the additive group of K. Then the topological dual V* of V is
a right vector-space of dimension n over K; the formula
(v,v*)v=x([v,v']v) for all veV
defines a bijective mapping v' -+ v* of the algebraic dual V' of V onto V*;
if x(x y) = x(y x) for all x, y in K, this mapping is an isomorphism for the
structures of V', V* as right vector-spaces over K.
Let X K be the topological dual of K. The structure of K as a left

vector-space of dimension 1 over itself determines on X K a structure of
right vector-space over K; as such, it has a certain finite dimension d.
Similarly, the structure of K as a right vector-space over K determines
on X K a structure of left vector-space of a certain dimension d' over K.
Let V be as in theorem 3; by taking a basis of V over K, V can be written
as the direct sum of n subspaces of dimension 1; therefore its dual V*,
as a right vector-space, is isomorphic to the direct sum of n spaces iso-
morphic to X K , and has therefore the dimension nd. Similarly, the dual
of V* is a left vector-space of dimension ndd'; as it is isomorphic to V,
with which we have in fact agreed to identify it, we get ndd'=n, hence
d = d' = 1. Now let X be as in theorem 3; this defines an element c* i= 0 in
the additively written group X K, so that we have x(t) = (t,C*)K for all
teK. As d' = 1, every element of X K can be uniquely written as xc*, with
xeK; as d= 1, every element of X K can be uniquely written as c·y, with
yeK. Therefore the relation xc*=c*y determines a bijection IX of K
onto itself, and one verifies at once that this is an automorphism of K.
In view of (8), c* y is the character t -+ X(y t) of K, and similarly xc * is
t -+ X(t x). Therefore X(t x) = X(IX(X) t) for all x, tin K, and this determines IX
uniquely; in particular, IX induces the identity on the center of K, and it
is the identity if and only if X(t x) = X(x t) for all x, t in K. Now consider
the mapping v' -+ v* of V' into V* which is defined in theorem 3; take
x e K, put w' = v' x, and assume that the mapping in question maps w'
onto w*. We have
x([v, w']v) = x([v,v']v x)= X(IX(X) [v,v']v) = X([IX(X) v, v'] v).
In view of the definition of v· and w*, this gives
(v, w*)v= (IX(X)V,V*)v= (V,V*IX(X)v
for all v, hence w· = V*IX(X). It is customary to express this by saying that

the mapping v' -+ v* is IX-semilinear. At the same time, it is clearly injective;
§ 5. Duality over local fields 41
for v* =0 means that X([v,v']v) is I for all VE V, hence also that X(x[v,v']v)
is I for all XEK and all VEY; as X is not trivial, this implies [v,v']v=O for
all v, hence v' =0. As V' and V* have the same dimension n over K, an
IX-semilinear mapping of V' into V* cannot be injective without being
bijective; this completes the proof. For purposes of reference, we formu-
late separately the result about the characters of K:
COROLLARY. Let K and X be as in theorem 3; then every character

of K can be uniquely written as t--"X(tx), with xEK, and also as t--"X(yt),
with YEK.
A more "intrinsic" way of formulating theorem 3 would be to say

that there is a canonical isomorphism, given by the formula in tho 3,
between V* and the tensor-product V'® xXx (and similarly between
V* and X x ® K V' if V is given as a right vector-space); this will be left
to the reader. One may also note that there is always a non-trivial
character X of K for which X(x y) = X(y x) for all x, y; for instance, one
may take X= XOOt, where t is the "reduced trace" in K over its center Ko
(cf. Chap. IX-2), and Xo is a non-trivial character of Ko; the same result
could be deduced from the fact that, in view of the Skolem-Noether
theorem (which will be proved as prop. 4 of Chap. IX-I), IX in the proof
of tho 3 must be an inner automorphism of K. Of course the distinction
between right and left becomes entirely superfluous if one considers only
commutative fields.
It is frequently convenient, having chosen once for all a character
X of K with the properties described in theorem 3, to identify the topo-
logical and algebraic duals of every vector-space over K by means of
the isomorphism described in that theorem; when doing this, one will
refer to X as "the basic character". In particular, K will then be identified
with its own topological dual, as shown in the corollary of tho 3. When
this is done for a p-field K, the subgroup of K associated by duality with
each subgroup of the form pn must be of the same form, since in general
the dual of a K-Iattice is a K-Iattice. In order to give a more explicit
statement, we set up a definition:
DEFINITION 4. Let K be a p-field, R its maximal compact subring

and P the maximal ideal of R. Then the order of a non-trivial character
X of K is the largest integer v E Z such that X is I on P - v; it will be denoted
by ord(x).
In other words, P- V is the dual K-Iattice to R when K is identified

with its dual by means of x; this shows that v is finite.
PROPOSITION 12. Let K be a p-field and X a non-trivial character

of K of order v. Then, for any nEZ, X(xt)= 1 for all tEpn if and only if
XEP-n-v.
This is obvious, and amounts to saying that the dual K-Iattice to

P" is p-n-v when K is identified with its dual by means of X.
As to the explicit construction of characters for local fields, the case
of R is well-known; there one may take as basic character the one given
by Xo(x) = e(x)= e21ti\ in Cor H, one may then take as basic any character
XO 0 f, where f is an R-linear form other than 0 (e.g. the trace over R).
If K is a local field of characteristic p, one may write it as a field of formal
L
power-series x = ai T with coefficients in F q' and take as basic the
character of order 0 given by X(x)= "'(a-I)' where", is a non-trivial
character of the additive group of F q • For Qp, an explicit construction
will be given in Chap. IV-2, as part of the proof of theorem 3 of that
Chapter.
Chapter III
Places of A-fields
§ 1. A-fields and their completions. By an algebraic number-field, it
is customary to understand a finite algebraic extension of Q. One main
object of this book, and of number-theory in general, is to study algebraic
number-fields by means of their embeddings into local fields. In the last
century, however, it was discovered that the methods by which this can
be done may be applied with very little change to certain fields of charac-
teristic p> 1; and the simultaneous study of these two types of fields
throws much additional light on both of them. With this in mind, we
introduce as follows the fields which will be considered from now on:
DEFINITION 1. A field will be called an A-field if it is either a,finite

algebraic extension of Q ora finitely generated extension of a finite
prime field Fp, of degree of transcendency 1 over Fp.
Thus, if k is an A-field of characteristic p> 1, it must contain a trans-

cendental element t over F p' and it is then a finite algebraic extension of
F p(t). Therefore, if once for all we denote by T an indeterminate, so that
F p(T) is the field of rational functions in T with coefficients in F p' an
A-field of characteristic p is one which is isomorphic to a fmite algebraic
extension of F p(T). One should note that such a field always contains
infinitely many fields isomorphic to F p(T).
We shall study A-fields by means of their embeddings into local
fields. In view of theorems 5 and 8 of Chap. I, it is permissible, up to
isomorphism, to speak of the set of all local fields. In fact, for a given
p> 1, the local fields of characteristic p are, up to isomorphism, in a
one-to-one correspondence with the finite fields F q with q = p" elements,
while the local p-fields of characteristic 0 are isomorphic to the subfields
of an algebraic closure of Qp which are of finite degree over Qp- It will
be seen later (as a consequence of lemma 1, Chap. XI-3) that there are
only enumerably many fields of the latter type; this will not be needed
here. It is now legitimate to speak of the set of places of an A-field ac-
cording to the following definition:
DEFINITION 2. Let A. be an isomorphic embedding of an A-field k

into a local field K; then (A.,K) will be called a completion of kif A.(k) is
dense in K. '1Wo completions (A.,K), (A.',K') of k will be called equivalent
44 Places of A-fields III
if there is an isomorphism p of K onto K' such that A.' = pOA. By a place

of k, we shall understand an equivalence class of completions of k.
DEFINITION 3. A place of an A-field k, determined by a completion
(A, K) of k, will be called real if K is isomorphic to R, imaginary if K is
isomorphic to C, infinite in both of these cases, and finite in all other cases.
Let v be a place of k, as above; clearly, for all completions (A,K) of k

belonging to v, the function modKoA on k is the same; this will be written
x-+lxl v. Ifv is imaginary, modK (x-y)1/2 is a distance-function on K;
in all other cases, modK(x- y) is such a function. Therefore we can
always obtain a completion of k, belonging to v, by taking the completion
of k with respect to the distance-function Ix - YI~ with (I; = 1/2 if v is imag-
inary and (I; = 1 otherwise. This completion will be denoted by kv and
will be called the completion of k at v; for all xEkv, we shall write
Ixlv=moddx). If v is a finite place, we write rv for the maximal compact
subring of kv, and Pv for the maximal ideal in rv; these are the subsets of
kv defined respectively by Ixlv~ 1 and by Ixl v< 1.
As shown by tho 5 of Chap. 1-2, Q has one infinite place, corresponding
to the embedding of Q into R=Qoo; this place will be denoted by 00.
The same theorem shows that the finite places of Q are in a one-to-one
correspondence with the rational primes, with which they will usually
be identified, the place p corresponding to the embedding of Q into Qp.
The knowledge of the places of Q provides us with a starting point
for determining the places of algebraic number-fields, considered as
finite algebraic extensions of Q. In order to proceed in the same way
for A-fields of characteristic p> 1, we have to know the places of F p{T).
Before determining them, we first give some general results about places
of algebraic extensions.
PROPOSITION 1. Let k be any field, ko an infinite subfield of k, and A

an isomorphic embedding of k into a local field K. Then the closure
Ko of A(k o) in K is a local field, and the closure of A(k) in K is the field
generated by A(k) over K o.
The first assertion follows at once from corollary 3 of prop. 2, Chap. 1-2.
Then, by corollary 2 of tho 3, Chap. 1-2, K must have a finite degree over
K o, so that, by tho 3, Chap. 1-2, every vector-space over Ko in K is closed
in K. The field Kl generated by A(k) over Ko is such a vector-space; on
the other hand, the closure of A(k) in K is clearly a field, and it contains
A(ko), hence K o, and A(k); therefore it i8 K 1.
COROLLARY. Let k be an A-field, k' a finite algebraic extension of k,

and w a place of k'. Let A be the natural injection of k' into its completion
§1. A-fields and their completions 45
k~ at w. Then k~ is a finite algebraic extension of the closure of A(k) in

k~, and the injection of k into that closure, induced on k by A, determines a
place v of k.
In view of our definitions, this is a special case of prop. 1; it enables
us to set up the following definition:
DEFINITION 4. If k, k', wand v are as in the corollary of prop. 1, we

say that v is the place of k which lies below w, and that w lies above v; and
we write wlv.
When that is so, we shall usually identify kv with the closure of k

in k~.
THEOREM 1. Let k be an A-field, k' a finite algebraic extension of k,
and v a place of k. Then there is a place of k' which lies above v, and there
are only finitely many such places.
Let K be an algebraic closure of kv, and kIf the algebraic closure of k
in K; as kIf is algebraically closed, there is at least one isomorphism A
of k' into kIf over k. Call K .. the field generated by A(k') over kv; this is a
finite algebraic extension of kv, so that, by corollary 1 of tho 3, Chap. 1-2,
we can give it its topological structure as a vector-space of finite dimen-
sion over kv ; this makes it into a local field. Then, by prop. 1, (A,K ..) is
a completion of k', and it determines a place of k' which clearly lies
above V. Conversely, let w be any place of k' above V. Then, by the corollary
of prop. 1, k~ is algebraic over kv, so that there is at least one isomorphism
q> of k~ into Kover kv; let A be the isomorphism of k' into K induced on
k' by q>; clearly A maps k' into kIf. By prop. 1, k~ is generated by k' over
kv, so that q>(k~) is the same as the field denoted above by K .. ; moreover,
again by corollary 1 of tho 3, Chap. 1-2, q> is a topological isomorphism
of k'.., onto K .. , so that w is the same as the place of k' determined by the
completion (A, K ..) of k'. Thus there are at most as many places of k' above
v as there are distinct isomorphisms A of k' into kIf over k. As k' is a finite
algebraic extension of k, it is well-known (and easily proved) that there
are only finitely many such isomorphisms.
COROLLARY. An A-field has at most a finite number of infinite places;
it has at least one if it is of characteristic 0, and none otherwise.
The last assertion is obvious; the others are a special case of tho 1,
since clearly a place of an A-field of characteristic 0 is infmite if and
only if it lies above the place 00 of Q.
Now we proceed to the determination of the places of F p(T); more
generally, we will determine those of Fq(T), where Fq is any finite field.
It will be convenient to say that a polynomial x in Fq[T] is prime if

it is monic and irreducible in Fq[T] and if its degree is >0.
THEOREM 2. The field k=Fq(T) has one and only one place v for which
ITlv > 1; for this place, T - 1 is a prime element of kv, and the module
of kv is q. For each prime polynomial x in Fq[T], k has one and only one
place v such that Ixl v< 1; for this place, x is a prime element of kv, and
the module of kv is qli if ~ is the degree of x. All these places are distinct,
and k has no other place.
Let v be a place of k. Assume first that ITl v:s:;;1; then Fq[T] is con-
tained in r V. Call p the canonical homomorphism of r v onto the finite
field rv/pv; it induces on Fq[T] a homomorphism of F'l[T] onto its
image, whose kernel PvnFq[T] is clearly a prime ideal in Fq[T]. As
Fq[T] is infinite, and rv/pv is finite, this ideal cannot be {O}; therefore
it is the ideal x . Fq[T] generated in Fq[T] by some prime polynomialx.
Then Ixl v< 1, and we have Iexl v = 1 for every polynomial ex prime to x
in Fq[T]. Every eek x can be written in the form e=x"ex/ex' with neZ
and a., ex' in Fq[T] and prime to x; when e is so written, we have lelv=lxl:;
in particular, e is in r v if and only if n ~ 0, i. e. if and only if it can be written
as e = PIex with ex, Pin Fq[T] and ex prime to x. As F q( T) is dense in kv,
the range of values taken by Ixl v is the same on Fq(T) as on kv; this im-
plies that x is a prime element of kv. Now let ~ be the degree of x. The
image of Fq[T] in rv/pv is isomorphic to Fq[T]/x· Fq[T], which is an
extension of Fq of degree ~, hence a field with q6 elements; clearly the
image of every element of rvnFq(T) must then be in that same field,
which is therefore no other than rv/pv, since Fq(T) is dense in kv; this
shows that q6 is the module of kv, and we have Ixl v=q-6. Consequently,
the function lelv on k is uniquely determined by x, so that, when x is
given, there can be at most one place v of k with the properties we have
described. Assume now that ITlv > 1; then IT- 11v < 1, and we may proceed
exactly as before, substituting the ring Fq[T-1] for Fq[T], and T- 1
for x; then it is easily seen that, if e = Plex with polynomials ex,p in Fq[T],
other than 0, of respective degrees a,b, we have lelv=qb-a. It is now
clear that, if x is any prime polynomial, Ixl v cannot be < 1 except for the
place v described above, if there is such a place, and that the same holds
for T - 1. In order to show the existence of those places, take first the
case X= T; then the ring Fq[T] can be embedded in an obvious manner
<Xl
into the ring offormal power-series rai Ti with coefficients in Fq; clearly,
o
if we extend this to the corresponding fields, we get a place of k, corre-
sponding to x = T. Exchanging T with T- 1, we get the same result for
T- 1 • Now take a prime polynomial x of degree ~; then FiT) contains
§l. A-fields and their completions 47
the field Fin) and is algebraic over it; its degree d over Fin) is ~b. As
we have just proved, there is a place w of Fq(n) for which Inlw=q-l.
By tho 1, Fq(T) has a place v lying above w. By corollary 2 ofth. 3, Chap. 1-2,
we have then Inlv = Inl~ = q-d. This completes our proof, and shows also,
incidentally, that d = b.
COROLLARY. Notations being as in theorem 2, let v be the place

of k corresponding to the prime polynomialn of degree b. Then the poly-
nomials of degree <b in Fq[T] make up a full set of representatives of
the classes in rv modulo Pv.
This follows at once from what has been proved above and from
the fact that these polynomials make up a full set of representatives of
the classes in Fq[T] modulo n.
From now on, it will be convenient to say that a property, involving

a place of an A-field k, holds for almost all places of k (or, if no confusion
is likely, that it holds almost everywhere) if it holds for all except a finite
number of such places. This will be of use, for instance, in formulating
our next result.
THEOREM 3. Let k be an A-field and e any element of k. Then lelv~ 1

for almost all places v of k.
e
This is clear for k = Q, since we can then write = alb with a, b in Z
and b+O, and lelp~1 for all the primes p which do not divide b. Now
let k be an A-field of characteristic 0, i. e. an algebraic number-field.
e
Then satisfies an equation
with coefficients ai in Q. Let P be the finite set consisting of 00 and of

all the primes which occur in the denominators of the a i • By tho 1, the
set P' of the places of k which lie above the places of Q belonging to P
is finite. Take any place v of k, not in P'; then the place p of Q which
e
lies below v is not in P, so that lailp ~ 1 for 1 ~ i ~ n; therefore is integral
e
over Zp. By prop. 6 of Chap. 1-4, this implies that is in r v , i.e. that
lelv~ 1. For an A-field k of characteristic p> 1, one could give a
e
similar proof; one may also proceed as follows. If is algebraic over
the prime field, we have lel v= 1 or 0 for all v, according as e+0 or e=0.
If not, k is algebraic over F p( e). Let v be a place of k, and let w be the place
of Fp(e) lying below it. By corollary 2 of tho 3, Chap. 1-2, lelv> 1 if and
only if lelw> 1. By th.2, Fp(e) has only one place w with that property.
In view of tho 1, this completes the proof.
48 Places of A-fields, III
COROLLARY 1. Let E be a finite-dimensional vector-space over an

A-field k. Let e,e' be two finite subsets of E, both containing bases of E
over k. For each finite place v of k, put Ev=E®kk v, and call ev,e~ the
rv-modules respectively generated by e and bye' in Ev. Then, for almost
all v, ev= e~.
Here, as on all similar occasions from now on, it is understood that E
is regarded as embedded in Ev by means of the injection e-+e®1 kv .
Put e={el, ... ,er } and e'={e'l, ... ,e~}. As e contains a basis for E over k,
we may write (perhaps not uniquely) ej= LCjiei for 1 ~j~s, with coeffi-
cients cji in k. Then e~c:ev whenever all the Icjil v are ~ 1, hence for almost
all v. Interchanging e and e', we get the assertion in our corollary.
COROLLARY 2. Let .91 be a finite-dimensional algebra over an
A-field k. Let rx be a finite subset of .91, containing a basis of dover k.
For each finite place v of k, put dv=d®kk v, and call rxv the rv-module
generated by rx in .91 v. Then, for almost all v, rxv is a compact subring
of .91 v.
Put rx={al, ... ,ar } and rx'={1,a l , ... ,ar }. As rx contains a basis of .91
over k, we may write aia j = Lcijhah for 1 ~i,j~r, with coefficients Cijh
in k. Then rx~ is a subring of .91 v whenever all the Icijhlv are ~ 1, hence
for almost all v; obviously, it is compact; and rxv=rx~ for almost all v.
§ 2. Tensor-products of commutative fields. If k is an A-field and

k' a finite algebraic extension of k, the proof of theorem 1 gives a construc-
tion for the places of k' which lie above a given place of k. This will now
be replaced by another one, based on the consideration of the tensor-
product k'®kk v. To simplify matters, we shall deal only with the case
where k' is separable over k; this is adequate for our purposes because
of the following lemma:
LEMMA 1. Every A-field of characteristic p> 1 is isomorphic to a
separably algebraic extension of Fp(T) of finite degree.
Let k be such a field; write it as Fp(xl, ... ,XN ), where at least one of
the Xi' say Xl' has to be transcendental over Fp- We will prove, by induc-
tion on N, that there is an Xi such that k is separable over Fixi). This
is clear if N = 1, and also if X2, .•. ,XN are all algebraic over Fp, since in
that case, by th.2 of Chap. 1-1, they are separable over Fp, so that k is
separable over Fp(x l ). If that is not so, then, by the induction assumption,
Fp (x 2 , ... ,XN) is separable over Fixi) for some i~2, say over Fp (X2)' so
that k itself is separable over Fp(x l , X2). As k has the degree of trans-
cendency 1 over Fp, there is an irreducible polynomial If> in Fp[ Xl' X 2]
such that If>(x l , x 2)=O. Then If> is not oftheform If>'P with If>' in Fp[X I'X 2];
as every element rx of Fp satisfies rx P= rx, this is the same as to say that If>
§ 2. Tensor-products of commutative fields 49
contains at least one term ex Xf X~ where ex =/= 0 and a or b is prime

to p. If for instance a is prime to p, Xl is separable over Fp(x 2 ), so that
also k is separable over Fp(x 2).
In the rest of this §, we shall be concerned with the purely algebraic
properties of tensor-products of the form k'®kK, where k is any field,
k' a separably algebraic extension of k of finite degree, and K is any
field containing k; in § 4, this will be applied to the case where k is an
A-field and K a completion of k. We dispose first of a side-issue.
LEMMA 2. If a commutative ring B can be written as a direct sum

of fields, it can be so written in only one way; and a homomorphism of B
into a field must be 0 on all except one of the summands of B.
Let B be the direct sum of the fields KI, ... ,Kr ; put ei= l K;. Then
Ki=eiB, and B has the unit-element IB= Lei' Clearly the solutions of
the equation X 2 = X in B (the "idempotents" of B) are the partial sums
of the sum Lei; consequently the ei are uniquely characterized as those
among the solutions of X 2 =X in B which cannot be written as e+e',
where e, e' are solutions of X 2 = X, other than O. If f is a homomorphism
of B into a field K', it must map each ei onto a solution of X 2 =X in K',
hence onto 1 or O. If f(ei) = 1, then f(ej)=O for all j=/=i, since eiej=O
for i =/= j; this implies that f is 0 on K j •
PROPOSITION 2. Let k be a field and k' =k(~) a separable extension

e
of k generated by a root of an i"educible monic polynomial F of degree n
in k[X]. Let K be a field containing k; let FI, ... ,Fr be the i"educible
monic polynomials in K[X] such that F=FI ... F., and, for each i, let i e
be a root of Fi in some extension of K. Then the algebra A=k'®kK
over K is isomorphic to the direct sum of the fields K(ei)'
As k' is separable over k, F is without multiple roots in all extensions
of k, so that the Fi are all distinct. Call p the k-linear homomorphism
of the ring k[X] onto k', with the kernel F· k[X], which maps X onto e;
this can be uniquely extended to a K-linear homomorphism p' of K[X]
onto A, which has then the kernel F· K [X] and determines an iso-
morphism of A'=K[X]/F'K[X] onto A. We will now show that A'
is isomorphic to the direct sum B of the algebras Bi=K[X]/Fi' K[X]
over K; as these are respectively isomorphic to the fields K(e i ) in our
proposition, our proof will then be complete. Let f be any element
i
of K[X]; call] its image in A', and its image in Bi for every i. Clearly
each i is uniquely determined by], so that ] --+(]l'''' ,]r) is a homo-
morphism cp of A' into B. As the Fi are mutually prime, it is well-known
(and easy to prove, by induction on r) that there are polynomials PI' .... 'Pr
in K[X] such that F-l=LPjF,-l; this implies, for all i and allN=i:
(1) pjFj- l F:1 (Fj); pjFj-lF:O (Fj ).
Take r polynomials fl""'/' in K[X]; for each i, call ]; the image of
J; in Bj; put f = LP,Fj- 1 F J;, and call 1 the image of f in A'; then 1 is
uniquely determined by the ];, so that (]1,"" 1.)-+ 1 is a mapping '"
of B into A'. Clearly "'ocp is the identity on A', and (1) shows that cpo",
is the identity on B. Therefore cp is an isomorphism of A' onto B.
Let k, k' and K be as in proposition 2. Clearly an isomorphism A of
k' into an extension K' of K induces the identity on k if and only if it
is k-linear. Such an isomorphism will be called proper above K if K' is
generated by A(k') over K;, then (A, K') will be called a proper embedding
of k' above K. Two such embeddings (A,K'), (A.',K") will be called equi-
valent if there is a K-linear isomorphism p of K' onto K" such that
A.' = pOA. One will notice that these are the algebraic concepts under-
lying definition 2 and proposition 1 of § 1.
PROPOSITION 3. Let k be a field, k' a separably algebraic extension

of k of finite degree n, and K a field containing k; put A=k'®A;K. Then,
up to equivalence, there are only finitely many proper embeddings (Ai> K j )
(1 ~ i ~ r) of k' above K; the sum of the degrees of the K j over K is n. The
mapping (Al, ... ,A,) of k' into the direct sum B of the fields K j is a k-linear
isomorphism of k' into B, and its K-linear extension cp to A is an iso-
morphism of A onto B.
We may write k' = k( e), and then, calling F the irreducible monic
e,
polynomial in k[X] with the root apply prop. 2 to k, k', F and K; e,
this shows that there is a K-linear isomorphism cp of A onto the direct
sum B of certain fields K j. For each i, call pj the projection from B to
K j; then J.lj= Pjocp is a K-linear isomorphism of A onto K j, and J.lj induces
on k' a k-linear isomorphism Aj of k' into K j • Clearly J.lj is the K-linear
extension of Aj to A, so that cp, which is,the same as (J.l1""'J.l,), is the
K-linear extension of (Al, ... ,A,) to A. If Aj was not proper above K,
there would be a field K"';= K j , between K and K j , such that Aj would
map k' into K"; then J.lj would map A into K", and not onto K j• Now
let A be any k-linear isomorphism of k' into a field K' containing K, and
call J.l the K-linear extension of A to A; J.l is then a homomorphism of A
into K', so that J.l0cp-l is a homomorphism of B into K'. By lemma 2,
this is 0 on all except one of the summands K j of B, so that we can write
it as (10Pj, where (1 is a K-linear homomorphism of K j into K'; as these
are fields, and as (1 is not 0, (1 must be an isomorphism of K j onto its
image K; in K'. This gives P=(10Pj, hence 1I.=(1oll.j; if KifK', 11., which
maps k' into K;, is not proper; therefore, if A is proper, (1 is an iso-
§ 2. Tensor-products of commutative fields 51
morphism of Ki onto K', so that (A, K') is equivalent to (Ai' KJ Finally,

if at the same time we had A= a' 0 Ai with j =1= i, a' being an isomorphism
of K i into K', this would imply Jl = a' 0 Jl i' hence JlO cp - 1 = a' 0 Pi' and
JlO cp - 1 would not be 0 on Ki' In particular, if A is proper, (A, K') is not
equivalent to more than one of the embeddings (Ai' K i); this shows that
the latter are all inequivalent, which completes our proof.
COROLLARY 1. Notations being as above, let A be any k-linear iso-

morphism of k' into a field K' containing K. Then there is a unique i,
and a unique isomorphism a of Ki into K', such that A=aoAi.
This was proved above; it is also an immediate consequence of pro-

position 3 and of the fact that, if K" is the subfield of K' generated by
A(k') over K, (A, K") is a proper embedding of k' above K, so that it must
be equivalent to one of the (Ai' Ki).
COROLLARY 2. Notations being as above, assume also that k' is a
Galois extension of k, with the Galois group G. Let (A, K') be any proper
embedding of k' above K. Then K' is a Galois extension of K; to every
automorphism P of K' over K, there is a unique aeG such that pOA=Aoa,
and p-+a is an isomorphism of the Galois group of K' over K onto a
subgroup H of G. The proper embeddings of k'above K, up to equivalence,
are all of the form (Aoa,K') with deG; if a, a' are in G, (Aoa',K') is equi-
valent to (Aoa,K') if and only if a' e H a.
Clearly A(k') is a Galois extension of k; as K' is generated by A(k')

over K, this implies that K' is a Galois extension of K, and that the
restriction to A(k') of the automorphisms of K' over K defines an injective
morphism of the Galois group H 1 of K' over K into that of A(k') over k;
this is equivalent to the first part of our corollary. For aEG, (Aoa, K')
is obviously a proper embedding of k' above K; if a,a' are in G, (AO a', K')
is equivalent to (Aoa, K') if and only if there is an automorphism p of K'
over K such that Aocr'=POAocr, i.e. POA=AO(cr'ocr- 1); this is so if and
only if a' 0 cr- 1 is in H. Therefore the number of inequivalent proper
embeddings of that form is equal to the index of H in G, i.e. to n/n' if
n, n' are the degrees of k' over k, and of K' over K, respectively. By pro-
position 3, the sum of the degrees of the fields Ki over K, in any set of
inequivalent proper embeddings (Ai> K i ) of k' above K, must be ~n;
therefore, up to equivalence, there can be none except those of the form
(AO a, K').
A useful special case of corollary 2 is that in which k' is a subfield
of K', generating K' over K; one may then take for A the identity; the
proper embeddings of k' above K can all be written in the form (a,K'),
with aeG, and the morphism p-+a of the Galois group of K' over K
into that of k' over k is the restriction to k' of the automorphisms of

K' over K.
COROLLARY 3. Let k and k' be as in proposition 3, and let K be an

algebraically closed or separably algebraically closed field containing k.
Then there are n, and no more than n, distinct k-linear isomorphisms
A1 , ... ,An of k' into K; they are linearly independent over K; if A, A.' are
any two of them, and K is an algebraic closure of k, there is an auto-
morphism IX of K such that A.' = IXO A.
A field K is said to be separably algebraically closed if it has no

separably algebraic extension, other than itself. The first assertion in
our corollary, which is obvious, is inserted here for the sake of reference,
and as an illustration of proposition 3, of which it is a special case;
in fact, if K is as in our corollary, all the Ki in that proposition must be
the same as K. The second assertion (a well-known theorem, due to
Dedekind, and easily proved directly) can be deduced as follows from
proposition 3. Assume that ~::CiAi=O, i.e. that LCiAi@=O for all eek',
with Cie K for 1 ::s:;; i::S:;; n. The Jl.i and Pi being as in the proof of proposition 3,
this implies LCiJl.i=O, hence LCiPi=O, which is clearly impossible unless
all the Ci are O. The last assertion, also inserted here for the sake of re-
ference, follows at once from the unicity, up to an isomorphism, of the
algebraic closure of k, which implies that each Ai can be extended to an
isomorphism of an algebraic closure k of k' onto K.
COROLLARY 4. Assumptions and notations being as in corollary 3,
assume also that k' is a Galois extension of k. Then all the Ai map k' onto
the same subfield of K.
This follows at once from corollary 2.
§ 3. Traces and norms. We first recall the concept of "polynomial
mapping". Let E,E' be two vector-spaces of finite dimension over a field k
with infinitely many elements; let 8 = {e 1'" ., en} and 8' = {ei, ... , e~} be
bases for these spaces over k. Then a mapping f of E into E' is called
a polynomial mapping if there are polynomials Pj in k[X 1, ... ,XnJ such
that
for all values of the Xi in k. This is clearly independent of the choice of

the bases 8,8'; moreover, since k has infinitely many elements, the poly-
nomials Pj are uniquely determined by f,8 and 8'. If E' = k, f is called
a polynomial function; the degree of the corresponding polynomial P
is then independent of 8 and is called the degree of f. If K is any field
§ 3, Traces and norms 53
containing k, put EK = E Q9 kK and E~ = E' Q9 k K; then there is one and

only one polynomial mapping of EK into E~ which coincides with f
on E; this will be called the extension of f to EK and E~ (or more briefly
to K) and will again be denoted by f; with respect to the bases e, e' of
EK , E~ over K, it is given by the same polynomials Pj as before,
If E is as above, we write End(E) for the ring of endomorphisms of E,
considered as an algebra over k. If aEEnd(E), we write tr(a) and det(a)
for the trace and the determinant of a; the former is a linear form, and
the latter is a polynomial function of degree equal to the dimension
of E, on End(E) considered as a vector-space over k.
Now let d be an algebra of finite dimension over k; as always, it is
tacitly assumed to have a unit element 1. For every aEd, call p(a) the
endomorphism x~ax of d when d is viewed as a vector-space over k;
writing End(d) for the algebra of all endomorphisms of that vector-
space, one may thus consider p as a homomorphism of d into End(d);
it is known as the regular representation of d; as d has a unit, it is an
isomorphism of d onto a subalgebra of End (d). The trace and the
determinant of p are known as the regular trace and the regular norm,
taken in dover k, and are denoted by Tr .r;I/k and N .r;I/k, or (when there
can be no confusion) by Tr and by N, respectively; the former is a linear
form on d viewed as a vector-space over k, and the latter is a poly-
nomial function, of degree equal to the dimension of dover k. If K is
a field containing k, and d is extended to the algebra dK=d Q9kK
over K, the regular trace and the regular norm in d Kover K are the
extensions of Tr .r;I/k and N.r;I/k to d K • and will still be denoted by Tr .r;I/k
and N.r;I/k' When d is a field k' of finite degree over k, one drops the
word "regular" and calls Trk'/k' Nk'/k the trace and the norm in k' over k.
These concepts will now be applied to the situation described in § 2.
PROPOSITION 4. Let k be a field, k' a separably algebraic extension

of k of finite degree n, and K a field containing k. Put A=k'Q9kK; let
(A;,K;)l,,;;,,;r be a maximal set of inequivalent proper embeddings of k'
above K, and let Il;, for each i, be the K-linear extension of A; to A. Then,
for all aEA:
n NK;/K(Il;(a)).
r r
Trk'/k(a)= L TrKi/K(Il;(a)),
;= 1
Nk'/k(a) =
;= 1
In fact, let notations be the same as in prop. 3 of § 2 and its proof,

and put b=q>(a). For every i, b has the projection Pi(b)=lli(a) on K i.
Then Trk'/k(a) and Nk'/k(a) are the trace and the determinant of y~by
regarded as an endomorphism of B. Taking for B a basis consisting of
the union of bases for the Ki over K, we get the formula in proposition 4.
COROLLARY 1. If k and k' are as in proposition 4, the k-linear form

Tr k' Ik on k' is not O.
In proposition 4, take for K an algebraically closed field contain-

ing k; then Ki=K for all i, and proposition 4 gives Trk'lk(a) = LJli(a).
With the same notations as before, put b=lp(a), hence Pi(b)=Jli(a); as
the projections Pi (b) of b on the summands of B can be chosen arbitrarily,
we can choose them so that Trk'lk(a) is not O. As Trk'ik on A is the extension
to A of the k-linear form Trk'ik on k', and the former is not 0, the latter
is not O.
COROLLARY 2. Notations and assumptions being as in proposition 4,

we have, for all x E k' :
Trk'lk(X) = L TrKi/K(Ai(X)), Nk'lk(X) = nNKi/K(Ai(X)).

i i
COROLLARY 3. Let k, k' be as in proposition 4; let K be an algehrl/-

ically closed field containing k, and call Al' ... , An the distinct k-linear
isomorphisms of k' into K. Then,for all xEk':
Trk'lk(X) = LAi(X), Nk'lk(X) =
i
nAi(X),
i
This follows at once from proposition 4 and corollary 3 of prop. 3, § 2.

COROLLARY 4. Let k and k' be as in proposition 4, and let k" be a
separably algebraic extension of k' of finite degree. Then:
Trk"lk= Trk'ik o Trk"lk" Nk"lk = Nk'lkONk"lk"
Take for K an algebraic closure of k"; define the Ai as in corollary 3;
similarly, call n' the degree of k" over k', and call Aj, for 1 ~j ~ n', the
distinct k'-linear isomorphisms of k" into K. Each Ai can be extended to
an automorphism lpi of K. Put A;j = lpioAj for 1 ~ i ~ n, 1 ~j ~ n'; these are
k-linear isomorphisms of k" into K. Clearly A;j = A~, implies i = h, since
A;j induces Ai on k', and j=l, since lpi-loA;j=Aj. Moreover, if A" is any
k-linear isomorphism of k" into K, it must induce on k' one of the iso-
morphisms Ai' and then lpi-loA" is k'-linear and must be one of the Aj,
so that A" =A~j' Now corollary 3 gives, for xEk";
Trk'lk(X) = LA;j(X) = Llpi(LAj(X))
i,j i j
= Llpi( Trk"lk'(X)) = LAi( Trk"lk' (x)) = Trk'lk( Trk"lk'(X)),

i i
This proves our first assertion. The formula for the norm can be proved
in exactly the same manner.
§ 3. Traces and norms 55
For the sake of completeness, we will also deal briefly with the
trace and the norm for inseparable extensions. Let k' be any algebraic
extension of k of finite degree; it is well known that it contains a unique
maximal separable extension ko of k, and that it is purely inseparable
over it; let q = pm be the degree of k' over ko, p being the characteristic;
e
it is easily seen that x'lEko for all xEk'. Take a basis {el' ... , q} of k'
over ko; take aEk'. Then k', as a vector-space over k, is the direct sum of
the subspaces eikO for 1 ~i~q, and these are invariant under x-+aqx
since aqEk o. Therefore we have
N",/,,(a~= N"o'l,,(aq)q,
which obviously implies
N k'/k(a) = N "o'/daq).
Call no the degree of ko over k, so that the degree of k' over k is n=noq.
If K is an algebraically closed field containing k, each k-linear iso-
morphism of ko into K can be uniquely extended to one of k' into K;
therefore, by corollary 3 of prop. 3, § 2, there are no such isomorphisms
A.i (1 ~i~no), and the above formula for N",/", together with corollary 3
of prop. 4 applied to ko and k, gives, for all xEk':
N",/,,(x) = n
i
A.i(X)n/no.
Now let k" be any finite extension of k'. Proceeding exactly as in the
proof of corollary 4 of prop. 4, we get again
N""/"=N",/"oN,,"/,,,,
which is therefore valid, whether k' and k" are separable over k or not.
As to the trace, the elementary properties of the determinant, and
the definition of the trace and the norm, show that, if d is any algebra
over k, 1t.w'/"(x), as a linear form on d, is the sum ofthe terms of degree 1
in the polynomial function N .w'/,,(1 + x) when the latter is expressed as a
polynomial in the coordinates of XEd with respect to some basis of d
over k. This, applied to the present situation, shows that 1t",/,,(x) is the
sum of the terms of degree 1 in N"'/A:(l+x). As the latter is equal to
N"b/A:(1 +xq), it contains only terms whose degree is a multiple of q.
This shows that 1t",/"=0 if q> 1, and therefore, in view of corollary 1 of
prop. 4, that 1t"'/k=l=O if and only if k' is separable over k.
PROPOSITION 5. Let k' be a separably algebraic extension of k of
degree n, and let {a 1 , ••. , an} be a basis of k' over k. Then the determinant
of the matrix
is not O.
In view of corollary 1 of prop. 4, this is contained in the following

lemma, which will also be useful later :
LEMMA 3. Let k' be any extension of k of degree n; let E be the
vector-space over k underlying k', and let A be any linear form on E, other
than 0. Then (x, y) -+ A(X y) is a non-degenerate bilinear form on E x E; one
can identify E with its algebraic dual E' by putting [x, y] = A(X y); and, if
ai' ... , an is a basis of k' over k, the determinant of the matrix (A(ajaj))
is not 0.
+
As Ais not 0, there is aek' such that A(a) 0. For each yek', define a
k-linear form A, on k' by A,(X) = A(XY) for all xek'. Then y-+A, is a mor-
phism of E into its dual E'. This has the kernel 0, since y+o implies
A,(ay-l)+O, hence A,+O. As E and E' have the same dimension over k,
this shows that y-+A, is an isomorphism of E onto E'; identifying E and
E' by means of that isomorphism, we get [X,y]=A(XY). By definition,
this is the same as to say that (X,Y)-+A(XY) is non-degenerate. Finally,
if the matrix (A(aja)) had the determinant 0, one could find Yl' ... , Yn
in k, not all 0, so that LA(ajaj)Yj=O, hence, putting y= LajYj, A,(aj)=O
j j
for all i, and therefore A,=O, which contradicts what has been proved
above.
§ 4. Tensor-products of A-fields and local fields. Let k be an A-field

and k' a separable extension of k; let v be a place of k, and kv the completion
of k at v. Then, by prop. 1 of § 1 and its corollary, the completions (A, K')
of k' which induce on k its natural injection into kv are the same as the
"proper embeddings" of k' above kv as defined in § 2. We may therefore
use propositions 2 and 3 of § 2 in order to determine the places of k'
above v; this will be done now.
THEOREM 4. Let k be an A-field, k' a separably algebraic extension

of k of finite degree n, and /X a basis of k' over k. For every place v of k,
let kv be the completion of k at v, and put Av=k' ®kkv; for every finite
place v of k, call rv the maximal compact subring of kv, and /Xv the rv-module .
generated by /X in Av. Let WI' ••• , Wr be the places of k' which lie above v;
for each i, call k; the completion of k' at Wj, Aj the natural injection of k'
into k; and Jl.i the kv-linear extension of Ai to Av. Then the mapping
tPv= (Jl.l, ... ,Jl.r) is an isomorphism of Av onto the direct sum Bv of the fields
k;, and, for almost all v, it maps /Xv onto the sum of the maximal compact
subrings r; of the fields k;.
The first assertion is just a special case of prop. 3 of § 2, obtained by

taking K = kv in that proposition; more briefly, but less accurately, it
§ 4. Tensor-products of A-fields and local fields 57
can be expressed by saying that the completions ki of k' at the places of

k' which lie above v are the summands of k'®kk v when this is written as a
direct sum of fields. Now take for v any finite place of k; clearly the sum
of the ri is the maximal compact subring of Bv; therefore its image Pv
under lP;; 1 is the maximal compact subring of A v , and we have to show
that this is the same as (Xv for almost all v. As each of the ri contains rv'
Pv is a kv-Iattice in Av; by tho 1 of Chap. 11-2, we can find a basis
{u v, 1, ... , uv,n} of Av over kv such that Pv is the rv-module generated by
that basis. For almost all v, by corollary 2 of tho 3, § 1, (Xv is a compact
subring of A v, hence contained in Pv; call P the finite set of places of k
for which this is not so. Put (X= {a1' ... , an}; for v not in P, (Xv is contained
in Pv, so that we can write ai= LCv,ijUv,j with cv,ijErv for l~i,j~n; the
matrix Cv=(Cv,i) is then in Mn(r v), and we have (Xv=pv if and only if Cv
is invertible in M n(rv), i. e. if and only if its determinant is invertible in r v.
Now, writing 7r for the trace Trk,/ko call A the determinant of the matrix
M =(Tr(aia))1.,.i,j"'n;
A is in k, and, by prop. 5 of § 3, it is not O. Applying tho 3 of § 1 to A and to
A -1, we see that IA Iv = 1 for almost all V. On the other hand, if U is ~ny
element of A v, Tr(u) is the trace of x~ux in Av; writing u· uv,i= Ldijuv,j
with dijEk v for 1 ~i,j~n, we get Tr(u) = Ldii • As Pv is a ring, all the dij
are in rv if UEpv; this shows that Tr maps Pv into rv' Therefore, if we write
N v for the matrix (Tr(Uv,iUv)), N v is in Mn(rv). Substituting now
LCv,ijUv,j for ai in the matrix M, we get M=CvNv'C v, hence A =
=det(N v) det(Cy. Here N v is in Mn(r v), and so is Cv if v is not in P;
and IAlv= 1 for almost all V. Clearly this implies that Idet(Cv)lv = 1 for
almost all v, as was to be proved.
In Chap. VIII, it will be shown that theorem 4 remains valid even
if k' is not assumed to be separable over k.
COROLLARY 1. Assumptions and notations being as in theorem 4, the

sum of the degrees over kv of the completions ki of k' at the places Wi
of k' which lie above v is equal to the degree n of k' over k.
In fact, this sum is the dimension of Bv over kv, while that of Av

over kv is n.
COROLLARY 2. Let k be an algebraic extension of Q of degree n;

call r1 the number of the real places of k, and r2 the number of its imaginary
places. Then r 1+2r2=n.
We get this by replacing k, k', v by Q, k, 00 in corollary 1.

S8 Places of A-fields III
COROLLARY 3. Assumptions and notations being as in theorem 4, the

extensions of 'Irk'I" and N k'ik to Av are given by
'fr"'I"(X) = L'fr"UdJ.Li(X»),
i
N"'I"(x) = nN"udJ.Li(X»),
i
This follows at once from the application of prop. 4 of § 3 to the

situation described in theorem 4.
COROLLARY 4. Assumptions and notations being as in theorem 4,

assume also that k' is a Galois extension of k, with the Galois group G.
Let w be one of the places Wi of k'. Then the completion k~ of k' at w is
a Galois extension of kv; the restriction to k' of the Galois group H of k~
over kv determines an isomorphism of H onto the subgroup of G, consisting
of the automorphisms of k' over k which leave w invariant; the Wi are the
images of wunder G, and all the k; are isomorphic to k~.
Let l be any isomorphic embedding of k' into a local field K, such
that l(k') is dense in K; then, by definition, this determines a place of k',
and the image of that place by an automorphism u of k' is to be under-
stood as the place determined by the embedding lou of k' into K. That
being so, we get our corollary by combining theorem 4 with corollary 2
of prop. 3, § 2, the latter being applied to the natural injection of k'
into k~.
Chapter IV
Adeles
§ 1. Adeles of A-fields. Throughout this Chapter, k will denote an A-

field; if v is a place of k, kv will denote the completion of k at v; if v is a
finite place of k, we write r v for the maximal compact subring of kv and
Pv for the maximal ideal of r v' these being the subsets of kv respectively
defined by Ixl v :::; 1 and by Ixl v < 1. We write P 00 for the set of the infinite
places of k, and P for any finite set of places of k, containing P 00. For
any such set P, put
(1) kA(P) = n kv x n rv,

veP v,P
where the second product is taken over all the places of k, not in P.
With the usual product topology, this is locally compact, since the kv
are so and the rv are compact. On kA(P), we put a ring structure by
defining addition and multiplication componentwise; clearly this makes
kA(P) into a topological ring. Set-theoretically, kA(P) could be defined
n
as the subset of the product kv consisting of the elements x = (xv) of
that product such that IXvl v:::; 1 for all v not in P. If P' is also a finite set
of places of k, and P'~P, then kA(P) is contained in kA(P'); moreover,
its topology and its ring structure are those induced by those of kA(P'),
and kA(P) is an open subset of kA(P').
Now we define a locally compact topological ring k A , the "ring of
adeles" of k. Set-theoretically, this is to be the union of all the sets kA(P);
in other words, it consists of the elements x = (xv) of the product kv n
which satisfy IXvl v:::; 1 for almost all v. The topological ring structure of
kA will be defined by prescribing that each kA(P) is to be an open sub ring
of kA- This means firstly that, if x=(x v) and y=(Yv) are in k A, then
x+ y=(xv+ Yv) and xy=(xvYv); it is clear, in fact, that these are both in
k A • Secondly, we get a fundamental system of neighborhoods of 0 in the
additive group of kA by taking such a system in anyone of the kA(P), for
instance inkA(P 00) which is the smallest one ofthe sets kA(P); equivalently,
we get such a system by taking all the sets of the form n U v' where U v
is a neighborhood of 0 in kv for all v, and U v= r v for almost all v.
60 Adeles IV
DEFINITION 1. By the adele ring kA of the A-field k, we understand

the union of the sets kA(P) defined by (1), when one takes for P all the
finite sets of places of k which contain the set of all infinite places. The
topological ring structure of kA is that for which each kA(P) is an open
subring of kA •
The elements of kA will be called the adeles of k.
Take a place v of k; when P contains v, one can write kA(P) as the
product of kv with an infinite product; denoting the latter by k~(P,v),
we may proceed with the products k~(P,v) just as we have done for the
products kA(P), taking now for P all the finite sets of places of k which
contain Pa) and v. The union of all the k~(P,v) is then a locally compact
ring k~(v), and kA is obviously isomorphic to the product kv x k~(v);
by means of this isomorphism, the first factor kv of the latter product is
obviously mapped onto the set of the adeles x = (xv) for which Xw = 0 at
all places w =1= v; this set will be called the quasi/actor of k A belonging to v,
and will always be identified with kv' The mapping (xv)-+xv of kA onto
kv, which corresponds to the projection from the product kv x k~(v)
onto its first factor, will be called the projection from kA onto the quasi-
factor kv; it is obviously continuous. Clearly, too, instead of one place v
of k, one could start with any finite set Po of such places so as to write
kA as the product of the fields kv for vePo and of one more factor.
Take any character X of the additive group of kA ; it induces on
kA(P), for every P, a character Xp of kA(P), and on the quasifactor kv,
for every v, a character Xv of kv' It is well-known that a character of an
infinite product of compact groups must induce the trivial character 1
on almost all the factors; this, applied to the character induced by XP
on the product nrv in (1), shows that Xv is trivial on rv for almost all v;
then we have, for all x=(x v) in kA :
(2)
v
the product here is taken over all the places v of k; for each x=(xv) in kA'
almost all the factors are equal to 1.
e
Let be an element of k. In view ofth. 3 of Chap. III-I, we define an
e
adele x = (xv) by putting Xv = for all v; we write this qJ(e), and call qJ the
canonical injection of k into kA ; we will frequently identify k with its
image in kA by means of qJ when there is no danger of confusion.
Let E be a vector-space of finite dimension n over k. For each place
v of k, we will write Ev=E®tkv; as usual, we take E to be "naturally"
embedded in Ev by the injection e-+e®l tv ' On the other hand, since k
has been embedded in kA by the canonical injection qJ defined above, we
may consider the tensor-product EA =E®tkA' and regard E as being
"naturally" embedded in it by the mapping e-+e®qJ(l). We define the
§l. Adeles of A-fields 61
topology of EA as the coarsest one for which the extensions to kA of the

linear forms on E are continuous. Equivalently, take a basis e of E
over k; this determines an isomorphism of k" onto E, hence an isomorphism
of (k A)" onto E A; the topology of EA is that which is obtained by trans-
ferring to EA the topology of (kAt by means of that isomorphism; it
would be easy to verify directly that this does not depend upon e.
Let E and E' be vector-spaces of finite dimension over k, and let f be
a polynomial mapping of E into E'; then f can be extended in an obvious
manner to a mapping of EAinto EA, viz., the one which is defined by the
same polynomial equations if E, E' are identified with spaces k", km, and
consequently E A, EA with (kAt, (kA)m by the choice of bases for E, E'
over k. This extension of f will again be denoted by f; it is clearly conti-
nuous, since addition and multiplication are continuous in k A .
PROPOSITION 1. Let E be a vector-space of finite dimension n over k.

Let e be a finite subset of E, containing a basis of E over k. For each
finite place v of k, call ev the rv-module generated by e in Ev' For each
finite set P of places of k, containing Pem write
EA(P, e) = nEv x nev'

veP v,P
Then each EA(P,e) is an open subgroup of E A, and EA is the union of these

subgroups.
This should be understood in the sense that each product EA(P,e) is

endowed with its product-topology, and that the latter coincides with
the one induced by that of EA' Clearly ev is a kv-Iattice in Ev, hence open
and compact in E v, for all finite places v. Therefore EA(P,e) is an open
subgroup of EA(P',e) whenever PcP'. Take a basis e' of E over k, and
use it to define an isomorphism of k" onto E, hence one of (kAt onto E A;
then our definitions show at once that EA is the union of the sets EA(P,e'),
and that these are open in EA' By corollary 1 ofth. 3, Chap. III-1, there
is a finite set Po of places of k, containing P00' such that ev= e~ when v is
not in Po. This shows that EA is the union of the sets EA(P,e), and also,
for P'=>PuPo, that EA(p'e) is open in EA(P',e'), hence in EA' Of course
one could use proposition 1 to define directly the topology of EA,just as
the topology ofkA has been defined above; corollary 1 ofth. 3, Chap. III-1,
would then show this to be independent of e.
COROLLARY 1. Assumptions and notations being as in proposition 1,

let C be a compact subset of EA' Then there is a finite set P of places of k,
such that CcEA(P,e).
62 Adeles IV
As C is contained in the union of the open sets EA(P,Il), it must be

contained in the union of finitely many such sets EA(P;,Il), hence in
EA(P,Il) for P= UPi'
If .91 is any algebra of finite dimension over k, we will denote by .91A
the topological ring obtained by extending the multiplication law of .91
to the spaced A in the manner explained above. Clearly this may be
regarded as an algebra over kA' and kA·1 '" is a closed subspace and a
subring of d A , isomorphic to kA-
COROLLARY 2. Let .91 be an algebra of finite dimension over k, and

cx a finite subset of .91, containing a basis ofd over k. For each finite place
v of k, call CX v the rv-module generated by cx in .91v' For each finite set P of
places of k, containing P00' write
dA(P,cx) = ndvx ncxv'
veP v+P
Then there is such a set Po with the property that dA(P,cx) is an open
subring of .91A whenever P => Po; and dAis the union of these subrings.
This follows at once from corollary 2 ofth. 3, Chap. 111-1, and from
proposition 1.
Take now an algebraic extension k' of k, of fmite degree. As k' is an
A-field, we may apply to it our general construction, obtaining thus its
adele ring k A. On the other hand, we may regard k' as an algebra over k
and apply to this algebra the construction given above; this gives a ring
which we write as (k'/k)A; as we have seen, it is an algebra over kA' and
contains the closed subring kA ·l k ., which we identify with kA in the
obvious manner. It is a central fact in the theory of adeles that the rings
kA' (k'/k)A defined in this way are canonically isomorphic; this will be
proved now, but only for the case where k' is separable over k. The
inseparable case will be treated in Chap. VIII-6.
THEOREM 1. Let k be an A-field and k' a separably algebraic exten-

sion of k of finite degree. Then there is a unique isomorphism 4J of (k'/k)A
onto kA with the following properties: (i) 4J induces the identity on k'
when k' is naturally embedded both in (k'/k)A and in kA;(ii) on each quasi-
factor (k'/k)v of (k'/k)A' 4J induces a kv-linear isomorphism 4J v of (k'/k)v
onto the product of the quasifactors k~ of kA corresponding to the places
w of k' which lie above v.
Write .91 for the algebra k'/k, i.e. for k' considered as an algebra
over k. Then d A , in the notation explained above, is the same as (k'/k)A,
and d v the same as (k'/k)v, i.e. as the algebra k'®tkv over kv which was
studied in Chap. 111-4. For a finite number of summands, a "direct sum"
§l. Adeles of A-fields 63
is the same as a product; we may therefore interpret tho 4 of Chap. 111-4

n
as defining an isomorphism cP v of (k'lk)v onto the product k~ of the
fields k~ for the places w lying above v; this is kv-linear and maps every
n
eek' onto the element (e, ... , e) of k~, and it is uniquely characterized
by these properties. Similarly, if we take a basis oc of k' over k, the same
theorem shows that, for almost all v, cP v maps OC v onto the product r~ n
of the maximal compact subrings ofthe fields k~; let Po be a finite set of
places of k, containing P00' such that cP v has that property for all v not
in Po. For each place w of k', call f(w) the place of k lying below it. Then,
for P-::JPo, the mappings cP v determine in an obvious manner an iso-
morphism cPp of dA(P,oc) onto k~(f-l(P»), where dA(P,oc) is the open
subring of d A= (k'lk)A defined as in corollary 2 of prop. 1. As every set
f - 1 (P) is finite, and every finite set P' of places of k' is contained in
r 1 (p) for P=f(P'), k~ is the union of the sets kA,(f-l(P») for p-::Jp .
o
As cP PI coincides with cPp on the domain of definition of cPp whenever
P 1 -::J P, there is an isomorphism cP of d A onto k~ which coincides with
cPp on that domain whenever p-::Jpo. It is now clear that cP has the pro-
perties stated in our theorem and that it is uniquely characterized by
these properties.

call f(w), for every place w of k', the place of k lying below W. Then, if
x=(x v) is in kA' cP(x) is the element y=(Yw) of k~ such that yw=xf(w) for
every place w of k'.
This follows at once from the fact that cP(1)= 1 and that cP v is kv-linear
for every V.
From now on, kA will usually be identified with its image in k~ by
means of the isomorphism, induced on kA by cP, which is described in
corollary 1. Clearly kA is thus a closed subring of k~.
COROLLARY 2. Let k and k' be as in theorem 1; let Elk' be a vector-

space of finite dimension over k', and call Elk the underlying vector-space
over k. Then the identity mapping of Elk onto Elk' can be uniquely extended
to a kA-linear mapping of (Elk)A into (Elk')Ao and this is an isomorphism
of (Elk)A onto (Elk')A-
In view of corollary 1, this is merely a restatement of theorem 1 if
E = k'; the case E = k'" follows from this immediately, hence also the
general case, since E can always be identified with a space k'" by the
choice of a basis.
According to the definitions given above, the k-linear form 'Irk'ik and
the polynomial function Nk'I" on the space k', considered as a vector-
space over k, may be extended to mappings 'Ir"'/lo N"'I" of (k'lk)A into k A;
64 Adeles IV
then 'Irk'lkOCP-1 and Nk'lkOCP-1 are mappings of kA into kA • We will

simplify the formulation of the next corollary by identifying in it (k'/k)A
with kA by means of CP, so that the latter mappings may be written
simply as 'Irk'ik and Nk'ik.
COROLLARY 3. Let x' =(x~) be any element of k A; put y= 'Irk'lk(X')
and z=Nk'lk(X'). Then y, z are the elements (Yv), (zv) of kA respectively
given by
L
Yv= Trk'wlk v (x~),
wlv
Zv = n
N "wlk v (x~)
wlv
for every place v of k, the sum and the product being taken over all the
places w of k' which lie above v.
This is an immediate consequence of prop. 4, Chap. 111-3, and th.l.
§ 2. The main theorems. In view of lemma 1 of Chap. 111-2, every
A-field is a separably algebraic extension of one of the fields Q and
Fp(T). Theorem 1 of § 1 enables us now to prove properties of adele
spaces by dealing first with the special cases k=Q and k=Fp(T). This
method will presently yield some important results; in stating them, we
simplify notations by identifying A-fields, and vector-spaces over such
fields, with their natural images in the corresponding adele spaces, as
explained in § 1; in the proofs, we shall again use cp to denote the canonical
injection of an A-field k into kA-
THEOREM 2. Let k be an A-field and E a vector-space of finite dimen-

sion over k. Then E is discrete in EM and EAlE is compact.
In view of corollary 2 of th.1, § 1, and of lemma 1 of Chap. 111-2, it

is enough to prove this for k=Q and k=Fp(T). If n is the dimension
of E, E is isomorphic to k", so that, if the theorem is proved for E = k,
it must be true in general. Thus we need only treat the cases E = k = Q
and E=k=Fp(T). We begin with Q.
e
For each prime p, call Q(p) the set of the elements of Q such that
Iel p ' :s:;; 1 for all the primes p' other than p. Clearly this is a subring of Q,
consisting of the numbers of the form p-n a with neN and aeZ.
LEMMA 1. For every prime p, we haveQp=Q(P)+ Zp andQ(P)nZp=Z.
The first assertion follows at once from corollary 2 ofth. 6, Chap. 1-4,
applied to Qp, to the prime element p and to the set of representatives
{O, 1, ... ,p-l}. The second one is obvious.
LEMMA 2. Put A", = R x n

ZP' and call cp the canonical injection
of Q into QA. Then QA =cp(Q)+A~ and cp(Q)nA", =cp(Z).
With the notation of (1), § 1, Aoo is the same as QA({ 00 }); it is there-
fore an open subring of QA' The second assertion in the lemma is obvious.
Now take any x = (xv) in QA; call P the set ofthe primes p such that xp
is not in Zp; it is a finite set. For each PEP, the first part of lemma 1
shows that we may write xp=ep+x~ with ~EQ(P) and X~EZp- For p not
in P, put ep=O and x~=xp- Put now e= l.3p, the sum being extended
to all p, and y=x-qJ<e). If y=(yv), we have, for every prime p:
yp=xp-ep- L ep'=x~- L ep"
P''i' p P''i'P
By the definition of Q(p), all the terms in the right-hand side are in Zp.
This shows that y is in A oo ' hence x in qJ(Q)+Aoo.
We can now prove our theorem for E=k=Q. As Aoo is open in QA'
the first assertion will be proved if we show that qJ(Q)nAoo, i.e. qJ(Z),
is discrete in Aoo; this is clear, since its projection onto the factor R of
the product Aoo is Z, which is discrete in R. Now callI the closed interval
n
[ -1/2, 1/2] in R, and put C = I x Zp- Clearly Aoo = qJ(Z) + C, hence
QA = qJ(Q) + c. As C is compact, this completes the proof.
For E=k=Fp(T), the proof is similar but simpler. For each place v
of k, call k(v) the set of the elements e of k such that lelw~ 1 for all the
places w of k, other than v.
LEMMA 3. For every place v of k, kv=k(v)+ rv and k(V)nrv=Fp-
The last assertion is obvious in view of the dermition of the func-

tions lelv on k which was given in the proof of tho 2, Chap. 111-1. As to
the first one, it is enough to consider a place attached to a prime poly-
nomial 7t of Fp[T], since otherwise we merely interchange T and T- 1 •
Then it follows at once from corollary 2 ofth. 6, Chap. 1-4, applied to kv,
to the prime element 7t and to the set of representatives supplied by the
corollary ofth. 2, Chap. III-I.
LEMMA 4. Put Ao = nrv' Then kA = qJ(k) + Ao and qJ(k)nAo = qJ(Fp).

With the notation of (1), § 1, Ao is the same as kA(~); it is a compact
open subring of k A. The last assertion is again obvious. Now take x = (xv)
in kA' For every v for which IXvl v> 1, lemma 3 shows that we may write
xv=ev+x~ with evEk(v) and x~Erv' For all other places v, put ev=O and
L
x~ = xv. Put e = ev and y = x - qJ<e). Just as in the proof of lemma 2,
we get yEAo.
The theorem is now obvious for E=k=Fp(T), since Ao is compact
and open in kA and Fp is finite. This completes the proof.
Now we consider a vector-space E over an A-field k, its algebraic
dual E', and the corresponding adele spaces EA,EA. We write [e,e'] for
66 Adeles IV
the value at a point e of E of the linear form determined by a point e' of E',
and we use the same notation for the extension of this bilinear form to
EA x E;'. As the additive group of EA is a locally compact commutative
group, we may consider its topological dual, which we denote by E1;
and we write (e,e*) for the value at eeEA of the character determined
by e* e E1. With these notations:
THEOREM 3. Let k be an A-field and X a non-trivial character of kA'
trivial on k. Let E be a vector-space of finite dimension over k; let E'
be its algebraic dual, and E1 the topological dual of EA' Then the formula
(e,e*)=x([e,e']) for all eeEA (e'eE;',e*eE1)
determines an isomorphism e' ~e* of E;' onto E1- Moreover, if e' is such
that x([e,e']) = 1 for all eeE, then e'eE'.
The last statement amounts to saying that the isomorphism e' ~e*
of E;' onto E1 defined in our theorem maps E' onto the subgroup of E1
associated by duality with the discrete subgroup E of EA'
We begin by treating the case E=k=Q. Use again the same nota-
tions as in the first part of the proof of tho 2. In view of lemma 2, every
character of A oo , trivial on q>(Z), can be uniquely extended to a character
ofQA' trivial on q>(Q). We get such a character X by putting x(x)=e( -xoo)
for x=(xv)eAoo (we recall that we write e(t) = e2 "it for teR). If we extend
this to a character X of QA' trivial on q>(Q), and call Xv, for every place v
of Q, the character induced by X on the quasifactor Qv of QA' then X
is obviously characterized by the following facts: it is trivial on q>(Q),
Xp is trivial on Zp for every prime p, and Xoo(x)=e( -x) for xeR. In order
to calculate XP' consider again the group Q(p) defined in the proof of
tho 2, and take any ~ e Q(p). Then ~ e Zp' for all primes p' =1= p, so that we
have, by (2) of § 1 :
1 = X(q>@)= Xoo(~)Xp(~)=e( - ~)Xp(~)

and therefore Xp(~)=e(~). By lemma 1, Xp is completely determined by
this and by the fact that it is trivial on Zp, and its kernel is Zp; it is there-
fore of order 0 in the sense of def. 4 of Chap. 11-5.
Now let x' be any character of QA; for every place v of Q, call X~
the character induced by x' on the quasifactor Qv of QA' By the corollary
of tho 3, Chap. 11-5, we can write X~ uniquely in the form X~(x) = Xv(avx)
with aveQv' As we observed when writing formula (2) of § 1, X~ must be
trivial on Zp for almost all p if X is to be continuous on QA; this implies
Xp(a p) = 1, hence apeZ p for almost all p; therefore a=(a v) is in QA' so
that, by (2) of § 1, x' is the character Xa of QA given by Xa(x) = x(ax) for
all XEQA- We have thus shown that the mapping a~Xa of QA into the
topological dual G = QX of QA is surjective. One sees at once that it

is continuous and injective, so that it is a bijective morphism of QA onto
its dual G. Call r the subgroup of G associated by duality with <p(Q),
i. e. consisting of the characters of QA' trivial on <p(Q); as X has that
property, the same is true of Xa for all ae<p(Q), so that a-+Xa maps <p(Q)
n
into r. Conversely, let b be such that Xber. As in the proof of th.2
for Q, put C =1 x Zp with 1 = [ -1/2,1/2]' We have shown there that
QA =<p(Q)+C; therefore we may write b=<p@+c with ~eQ, ceC, and
then Xeer. Writing c=(cv), we have now, since cpeZ p for all p:
1 = Xc(<p(1)) = X(c) = Xoo(coo)=e( -coo),
° n
hence Coo = since Coo e 1. Therefore Xc is trivial on Aoo = R x Zp; as
it is trivial on <p(Q), lemma 2 shows that it is trivial on QA' so that c=O,
hence be<p(Q). Therefore a-+Xa maps <p(Q) onto r. Finally, as <p(Q) is
discrete in QA' and QAi<p(Q) is compact, the duality theory shows that r
is discrete in G and that G/r is compact. Consequently a-+Xa determines
a bijective morphism of the compact group QAi<p(Q) onto the compact
group G/r; it is well-known that this must be an isomorphism. As G is
"locally isomorphic" to G/r, and QA to QAi<p(Q), this implies that
a-+Xa is bicontinuous, so that iris an isomorphism. This completes the
prooffor E=k=Q.
Now take E=k=Fp(T). In analogy with Q, call 00 the place of k
for which T- I is a prime element (although this is of course not an
infinite place). Then IT-lloo=p-l. We may now apply corollary2 of
th.6, Chap. 1-4, to koo' to the prime element T- I and to the set of re-
presentatives Fp , and therefore identify koo with the field of the formal
power-series
+00
(3) X= L ajT- i
i=n
where neZ and ajeFp for all i~n. Call I/J the character of the additive
group ofFq given by 1/J(1)=e(1/p); call XOO the character of koo defined by
putting Xoo(x) = I/J (- al ) when x is given by (3); for xeFp[T], we have
a l =0, hence Xoo(x) = 1. Now put Aoo =koo x nrv, the product being taken
over all the places v of k other than 00; with the notation of (1), § 1,
this is kA({ oo}); it is an open subring of kA and contains the set Ao defmed
in lemma 4, so that, by that lemma, kA=<P(k)+Aoo. When ~ek, <p@ is
in Aoo if and only if 1~lv~ 1 for all the places v of k attached to prime
polynomials in Fp[T], hence if and only if ~ is in Fp[T]. This means
that <p(k)nAoo = <p(Fp[T]). Accordingly, every character of A oo ' trivial on
<p(Fp[T]), can be uniquely extended to one of kA' trivial on <p(k). Applying
this to the character X of A oo' given by X(x)=Xoo(x oo ) for x=(xv)eAoo'
68 Adeles IV
we get a character X of kA which can be characterized by the following

facts: X is trivial on qJ(k); for every vi:- 00, the character Xv induced by X
on kv is trivial on rv, and X induces on kOC) the character XOC) defined above.
In order to calculate Xv for a place v attached to a prime polynomial 1t
of degree b in Fp[T], call k!» the set of the elements of k such that e
lel w ::S:;1 for all the places w of k other than v, and leI0C)<1; the same
proof which was given for lemma 3 shows now that kv is the direct sum
of k!» and rv; as Xv is trivial on rv, it is therefore completely determined
by its values on k!». Take eek!»; this can be written as e=1t- nIX, where
neN and IX is a polynomial of degree <nb in Fp[TJ. Call a l the coef-
ficient of Tn~- 1 in IX. As 1t is monic, it can be written as Til (0, where (0
is in Fp[T-l] and has the constant term 1. This gives
e=1t-nIX=(O-nT-n~IX=al T- l (T- 2 )
in the ring rOC), hence XOC)(e) = tfr( -al) by the definition of XOC). Now we
have, by (2) of § 1 :
1 = X(qJ<e)) = XOC)(e) Xv (e) = tfr (-al)Xv<e)
and therefore Xv (e) = tfr(a l ), which completes the determination of Xv.
e
Furthermore, if is as above and not 0, call d the degree of the poly-
nomiallX, and a the coefficient of Td in IX; then Xv(e T n6 - l - d) has the value
e
tfr (a), which is not 1 since a i:- O. This shows that, if is in k!» and not 0,
Xv(et) cannot be 1 for all terv. As Xv is trivial on rv, and as kv=k!»+r v,
we conclude now from prop. 12 of Chap. 11-5 that the character Xv is
of order 0 in the sense of def. 4 of Chap. 11-5. In other words, if x is in kv
and such that Xv(xt)= 1 for all terv, x must be in rv.
Now we can proceed just as in the case of Q. Let x' be any character
of kA. For each place v of k, the character X~ induced by X on kv can be
written as X~(x) = Xv(avx) with avekv; then, from the fact that X~ must
be trivial on rv for almost all v, we conclude that a = (a v) must be in kA'
so that X' is the character Xa defined by Xa(x) = x(ax). As before, we see
that a-Xa is a bijective morphism of kA onto the topological dual
G=kl of kA' and that it maps qJ(k) into the subgroup r of G associated
by duality with qJ(k). Assume that Xber for some bekA; by lemma 4, we
may write b=qJ(e)+c with eek, ceAo; then Xc is trivial on qJ(k). Put
c=(c v), so that cvErv for all v; then there is yeFp such that COC)=y (T- l );
e
replacing by e+y and c by c-qJ(Y), we get cOC)=O (T- l ). We have now
1 = XAqJ(1») = X(c) = XOC)(cOC)'
which implies, in view of the definition of XOC)' that is in T- 2 r and
COC) OC)'
therefore that XOC)(cOC) t) = 1 for all ter Consequently, Xc is trivial on A o,

OC).
hence on kA by lemma 4. This gives c=O, hence beqJ(k). The proof can
now be completed just as in the case of Q.
We can now complete the proof of our theorem by a purely formal

argument. Denote by T(E/k,X) the statement in theorem 3. What we
have proved above can be expressed by saying that, for each one of the
fields k=Q and k=Fp(T), there is a character X of kA for which T(k/k,X)
is true. Obviously this implies that T(k"/k,X) is true for every n, so that
T(E/k,X) is true for every vector-space E over k. In particular, take a
finite algebraic extension k' of k; as in lemma 3 of Chap. III-3, write E
for the underlying vector-space over k; choose a k-linear form A. on E,
other than 0, and identify E with its algebraic dual E' by putting
[x,y]=A.(xy). We can then extend A. to a mapping of EA into kA' the
identification between E and E' to one between EA and E~, and then we
have again [x,y]=A.(xy) for x,y in EA=(k'/k)A' If we write X'=X0A., this
is clearly a non-trivial character of EA , trivial on E. If now we assume
that k' is separable over k, we can identify EA with k~ by means of the
isomorphism cP described in tho 1 of § 1. When this is done, x' becomes
a non-trivial character on k~, trivial on k', and the statement T(E/k,X)
becomes exactly T(k'/k', X'). As we can take for k' any A-field, taking for k
either Q or Fp(T), we see that, for every A-field k, theorem 3 is true for
at least one choice of X. Now assume T(k/k,X) for such a field, and let Xl
be another character with the properties stated in theorem 3; T(k/k,X)
implies that Xl is of the form Xl(x)=x(ax) with aEk and afO. Then
the mapping e' -+e* defined as in theorem 3, but by means of Xl' is
composed of the similar mapping defined by X and of the mapping
e' -+ae' of E~ into itself. As the latter is clearly an automorphism of E~,
mapping E' onto itself, we see that T(E/k,X) is equivalent with T(E/k,Xl)'
This completes the proof.
COROLLARY 1. Let X be as in theorem 3, and call Xv, for every place v

of k, the character induced by X on the quasifactor kv of k. Then, for
every v, Xv is non-trivial, and, for almost all finite places v of k, Xv is
of order 0 in the sense of def. 4, Chap. 11-5.
For each aEk A, call Xa the character of kA defined by Xa(x) = x(ax).

If X was trivial on the quasifactor k", that quasifactor would be in the
kernel ofthe morphism a-+ Xa of kA into its topological dual; as theorem 3
says that this is an isomorphism, this would be a contradiction. In parti-
cular, for every finite place v of k, we may put v(v)=ord(xv) in the sense
of def. 4, Chap. 11-5. For each mapping v-+n(v) of the set of finite places
of k into Z, call G(n) the group of the elements x = (xv) of kA such that
ord(xv)~n(v) for all finite places v, and H(n) the subgroup of G(n) con-
sisting of the elements x=(x v) of G(n) such that xw=O for all infinite
places w of k. In view of the definition of the topology in kA in § 1, it is
obvious that G(n) is open in kA if and only if n(v)~O for almost all V.
70 Adeles IV
It is also clear that H(n) is compact ifn(v)~O for almost all V; conversely,
by corollary 1 of prop. 1, § 1, and with the notation of (1), § 1, every
compact subset of kA is contained in one of the sets kA(P), so that H(n)
cannot be compact unless n(v)~O for almost all V; therefore this is
necessary and sufficient for the compacity of H(n). Now prop. 12 of
Chap. 11-5, combined with the fact that Xw is not trivial for any infinite
place of k, shows that the set of elements x of kA such that X(x y) = 1 for
all yeG(O) is H(-v), and that the set of elements x such that X(xy)= 1
for all yeH(O) is G( -v). If we identify kA with its topological dual by
means of the isomorphism described in theorem 3, this means that
H( - v) and G( - v) are the subgroups of kA respectively associated by
duality with G(O) and H(O). As G(O) is open and H(O) is compact, duality
theory shows that H( - v) must be compact and G( - v) open. As we have
seen, this implies that -v(v)~O for almost all V and that -v(v)~O for
almost all v.
COROLLARY 2. Let E be a vector-space of finite dimension over k,
and let v be any place of k. Then E+Ev is dense in EA'
If this is true for E = k, it is clearly true for E = kn and therefore for
every E. If k + kv were not dense in kA' there would be a non-trivial
character of kA which would be trivial both on k and on k v ; this contra-
dicts corollary 1.
As in the case of local fields, it is frequently convenient, having chosen
once for all a "basic character" X with the properties described in theo-
rem 3, to identify the topological dual of EA with the space E~ by means
of the isomorphism in that theorem, for all vector-spaces E of finite
dimension over k. For every quasifactor k" of kA' one will then take as
"basic character" the character Xv induced by X on k", and use this to
identify the topological and algebraic duals of vector-spaces over kv as
explained in Chap. 11-5. This being understood, we have:
COROLLARY 3. Let assumptions and notations be as in propo~tion 1

of § 1. Let E be a vector-space over k, and E' its algebraic dual. Let e, e' be
finite subsets of E and of E', respectively, containing bases of these spaces
over k. For each place v of k, identify E~ with the topological dual of Ev
as explained above. Then, for almost all finite places v of k, e~ is the dual
kv-Iattice to ev.
For E = E' = k and e = e' = {1 }, this is just a restatement of corollary 1;

it is an immediate consequence of that corollary if e = {e 1"'" en} is a
basis of E and e' = {e~, ... , e~} is the dual basis to e, determined by
[ej, ej] = 1 if i = j and 0 if i -+ j. The general case follows from this at once
by corollary 1 ofth. 3, Chap. 111-1.
§ 3. Ideles 71
§ 3. Ideles. As before (cf. Chap. 111-3), if E is a vector-space of finite

dimension over any field k, we write End(E) for the ring of endomorphisms
of E, considered as an algebra over k. We will write Aut(E) for the group
of automorphisms of E; this is the same as the group End(Er of in-
vertible elements of End(E), and it is the subset of End(E) determined
by det(a)+O; therefore, if k is a topological field, Aut(E) is an open
subset of End(E); clearly it is a topological group for the topology in-
duced on it by that of End(E). If K is a field containing k, End(EK) is the
same as End(E)K=End(E)®/cK, and the determinant in End(EK) is the
extension to that space of the determinant in End(E).
Let d be an algebra of finite dimension over k; call p its regular
representation into End(d), as defined in Chap. 111-3, and write d X , as
usual, for the group of invertible elements of d. Take any aed; then
p(a) is the endomorphism x--.ax of the vector-space underlying d; if
it is an automorphism, it is surjective, so that there is bed such that
a b = 1W; then b = a - 1, and a e d x. As the converse is obvious, this
shows that d is the subset of d determined by Nw/k(a)+O. Therefore,
X
if k is a topological field, d x is open in d; moreover, p is then a topologi-

cal isomorphism of d onto a subalgebra of End(d), which maps d x
onto p(d)nAut(d); this implies that d x is then a topological group
for the topology induced on it by that of d.
Now, d being an algebra of finite dimension over an A-field k,
consider the group d ~ of invertible elements of the ring d A' The
simplest examples, e. g. d = k, show that x --. x - 1 is not continuous on
that group for the topology induced on it by that of d A' We will give
it the coarsest topology for which the injection into d A and x --. x - 1
are both continuous; this is more conveniently stated as follows:
DEFINITION 2. Let d be an algebra of finite dimension over the
A-field k. Then we denote by .s;I~ the group of invertible elements of d A
with the topology for which x--.(x,x- 1) is a homeomorphism of d ~ onto
its image in d A X d A'
It is customary (particularly in the case d = k) to call d ~, with this
topology, the idele group of d, and to call its elements the ideles of d.
Obviously (x,y)--.xy and x--.x- 1 are continuous on d~, so that our
definition does make it into a topological group. At the same time, if
we call f the mapping (x, y)--. x y of d x d into d and its natural exten-
sion to d A X d A, our definition says that d ~ is homeomorphic to the
subset f-l({l}) of the latter space; as f is continuous, this is a closed
set, so that d ~ is locally compact. It is also clear that d x is canonically
embedded in d~; as x--.(x,x- 1) maps it onto the intersection off-1({1}}
with the discrete subset d x d of d A x d A, it is a discrete subgroup
of d~.
72 Adeles IV
One can give an alternative definition of the idele group of .91, equi-
valent to definition 2, by using corollary 2 of th.3, Chap.III-1, and
corollary 2 of prop. 1, § 1. As in these results, take a finite subset IX of .91,
containing a basis of .91 over k, and call lXv, for each finite place v of k,
the rv-module generated by IX in d v • By corollary 2 ofth.3, Chap. III-1,
there is a finite set Po of places of k, containing P00' such that, for all v
not in Po, IXv is a compact subring of d v (containing the unit element).
For each v, as we have seen, .91: is an open subset of d v, and x-+x- 1
is continuous on it; therefore x-+(x,x- 1 ) maps it homeomorphically
onto its image in d v x d v • For v not in Po, IX: is the set ofthe elements
of .91: which are mapped into IX" x IXv by x-+(x,x- 1 ); therefore it is an
open compact subgroup of .91: and an open compact subset of IXv. We
shall now prove the following result, analogous to corollary 2 of prop. 1,
§ 1:
PROPOSITION 2. Let .91, IX, IXv and Po be as explained above. Let P
be any finite set of places of k, containing Po. Then the group
(4) dA(P, IX) = n .91: x nIX:
x
VfliP v,p
is an open subgroup of d ~; the topologies induced on it by those of .91 ~

and of .91A are both the same as the product topology for the right-hand
side of (4); and d ~ is the union of these groups.
Let dA(P,IX) be defined as in corollary 2 of prop. 1, § 1. The topology
induced on .91A(P,IX) x by that of dAis the same as that induced by that
of dA(P,IX), hence the same as the product topology for the right-hand
side of (4). For each v, .91: is open in .91v' and x-+x- 1 is continuous on
it; therefore x-+ x - 1 is continuous on .91A(P, IXr for that product topology.
This implies that x-+(x,x- 1 ) is a homeomorphism of dA(P,IX)X onto
its image in .91AX .91A; therefore the product topology on that set is
also that induced by d~. Furthermore, dA(p,lXr is the subset of d~
which is mapped by x-+(x,x- 1) into dA(P,IX) x dA(P,IX); as the latter
set is open in .91AX .91A' and as .91AX dAis the union of sets of that form,
this completes the proof.
COROLLARY. An element a=(a v) of kA is in k~ if and only if av=/=O
for all v and lavl v= 1 for almost all v. For every finite set P of places of k,
containing P00' the group
kA(Pr = n k: x n r:
VfliP vtP
is an open subgroup of k~, and k~ is the union of these groups.

The first statement is obvious; the rest is a special case of propo-
sition 2.
§ 3. Ideles 73
For every element a = (a v) of k;, we will write
the product being taken over all the places v of k; in view of the corollary
of prop. 2, almost all the factors of that product are equal to 1 whenever
a is in k~ . Usually, when there is no danger of confusion about the field
of reference, we will write lalA instead of lalk. for this product; it is some-
times called the module of a.
PROPOSITION 3. Let E be a vector-space of finite dimension n over k.

Put .lII=End(E), and let a=(a v) be an element of .lilA' Then the follow-
ing assertions are equivalent: (i) a is in d ~ ; (ii) det(a) is in k; ; (iii) e--+a e
is an automorphism of EA' When that is so, the module of the latter auto-
morphism is Idet(a)IA' Moreover, the mappings a--+det(a) and a--+ Idet(a)I A
are morphisms of .lII~ into k~ and into R~, respectively.
Take a basis 8 for E over k; we will use it to identify E with k" and
.lII with M,,(k). Then a basis (X for A over k is given by the "matrix units"
a..I' for 1 ::::;:;A.,J-t::::;:;n, where a..I' is the matrix (xij) given by x"1'= 1 and
Xjj=O for (i,j) =1= (A.,J-t). For every place v of k, an element av of Mikv) is
invertible in M,,(kv) if and only if det(a v) =#=0; for every finite place v of k,
an element av of M,,(rv) is invertible in M,,(rv) if and only if det(a v) is
invertible in rv, i.e. if and only if Idet(av)lv= 1. With the notations of
prop. 2 and its corollary, this amounts to saying that a is in .lIIA(p,(Xr
if and only if det(a) is in kA(Pr; clearly this implies the equivalence of
(i) and (ti) in our proposition, and it also shows that the mapping
a--+det(a) of .lII~ into k~ is continuous on .lIIA(P,(X) x for every P, hence
on .lII;'. As it is clear that the mapping z--+ IzlA of k;' into R~ is continu-
ous on kA(P) x for every P, hence also on k')., a--+ Idet(a) IA is a continuous
morphism of.lll'). into R~. If a is in .lII')., it has an inverse a-I in .lilA,
and then the endomorphism e--+ae of EA has the inverse e--+a- 1 e, so
that it is an automorphism. Conversely, take any a = (a v) in.lll A; prop. 1
of§ 1, applied to.lll and (x, shows that av is in Mikv) for all v and in M,,(rv)
for almost all v. The same proposition, applied to E and 8, shows that a
fundamental system of neighborhoods of 0 in EA is given by the sets
n
U = Uv' where Uv is a neighborhood of 0 in Ev = (k v)" for all v, and
Uv= (r v)" for almost all v. If e --+ ae is an automorphism of EA' it must map
every neighborhood of 0 onto a neighborhood of 0; this implies that av
is invertible in M,,(kv) for, all v, and that, for almost all v, the image of
(r v)" under av contains (rv)", i. e. that a;; 1 is in M ,,(rv) for almost all v. As
we have observed above, this is the same as to say that a is in .lII').. Let
then P be a finite set of places of k, containing P <Xl' such that av is in M lI(rv) x
74 Adeles IV
for all v not in P. As the set EA(P,e) is open in EA and invariant under
e-+ae, the module of e-+ae in EA is the same as its module in that set;
this, in view of the definition of that set in prop. 1 of § 1, is the product
of the modules of the automorphisms ev-+ave v of its factors; these, by
corollary 3 of tho 3, Chap. 1-2, are respectively equal to Idet(av)lv, which
completes our proof.
COROLLARY. Let .91 be an algebra of finite dimension over k, and
let a be an element of .91A' Then the following assertions are equivalent:
(i) a is in d~; (ii) N Jt'lk(a) is in k~; (iii) x-+ax is an automorphism of the
additive group of .91A' When that is so, the module of that automorphism is
IN Jt'lk(a)IA' Moreover, a-+NJt'lk(a) and a-+IN Jt'lk(a) IA are morphisms of d~
into k~ and into R~, respectively.
As we are always assuming that .91 contains a unit, (iii) implies (i).
All our other assertions follow at once from proposition 3, applied to
the underlying vector-space E of .91 over k and to the embedding of .91
into End(E) given by the regular representation p.
Of course all that has been said about the endomorphisms x -+ a x of
an algebra .91 applies equally well to the endomorphisms x-+xa; the
determinant N'(a) of the latter, sometimes called the "coregular norm"
on .91, is again a polynomial function, of degree equal to the dimension
ofd over k, and the module of the automorphism x-+xa of d A, for
aEd~, is equal to IN'(a)IA' Obviously N' = Nrllk when .91 is commutative;
the same is known to be true for all semisimple algebras and will be
proved in Chap. IX for simple algebras and in particular for division
algebras; this will not be needed here.
THEOREM 4. Let D be a division algebra of finite dimension over k.
For every real number Jl~ 1, call DI' the set of the elements d of D~ such
that the modules of the automorphisms x-+dx and x-+xd of DA are
respectively ~ Jl and ~ Jl- 1. Then DI' is a closed subset of D~ whose image
in D~/Dx is compact.
* Write N for the regular norm N Dlk' and N' for the "coregular norm"
as defined above; by the corollary of prop. 3, d-+IN(d)IA is continuous
on D~, and the same is true of d-+ IN'(d)IA for similar reasons; in view
of that same corollary, this implies that DI' is closed. By tho 2 of § 2, D is
discrete in D A, and D AiD is compact; therefore there is a Haar measure
a on DAsuch that a(D AiD) = 1, this being defined in the manner explained
in Chap. 11-4. As DA is not compact, we can choose a compact subset C
of D A such that a( C) > Jl. Call C' the image of C x C under the mapping
(x,y) -+ x - y of D A X D A into D A, and C" the image of C' x C' under the
mapping (x,y) -+ x y of DAx D A into DA; as these mappings are continuous,
§ 4. Ideles of A-fields 75
C' and C" are compact. Take any deD,.; as the module of x--+xd is
;;?; p,- 1, it maps C onto a set Cd whose measure is > 1; therefore, by lemma 1
of Chap. II -4, there are two elements x, y of C such that x d - y d is in D
and is not 0, i.e. such that it is in DX. Write Cl =x- y and b 1 =c1d; then
C 1 eC' and b 1 eDx. Similarly, x--+d- 1 x, being the inverse ofx--+dx, has a
module ;;?; p,-1, so that it maps C onto a set d- 1C of measure > 1; as
before, we conclude that there is C2eC' such that b2 =d- 1 C2 is in DX.
Then b 1 b2 = c 1 C2' so that b 1 b2 is in D x' nC", which is a finite set since D is
discrete and C" compact in DA' Call Yh ... , YN all the distinct elements of
DXnC"; C 1 C2 is equal to one of these, say Yi' so that Yi 1 c 1 C2 =1. This
shows that C2 is invertible in DA and has the inverse c;1=Yi 1 c 1 • As
db 2 =C2' we see that db 2 belongs to the set X of the elements x of D~
whose image under the mapping x --+ (x, x - 1) is in the union of the sets
C' x (Yi 1 C') for 1 ~ i ~ N. In view of deC. 2, X is a compact subset of D~;
as D,. eX' D X, the image of D" in D~/D x is contained in that of X,
which proves our theorem.
§ 4. Ideles of A-fields. We will now consider more in detail the case

d=k.
THEOREM 5. Let k be any A-field; then the morphism z--+lzlA of k:
into R ~ induces the constant 1 on k x •
If eekX, x--+ex is an automorphism of kA which maps k onto itself.

By tho 2 of § 2, k is discrete in kA' and kAik is compact. Therefore the
module of x--+ex, which is lelA by prop. 3 of § 3 (if one takes E=k in
that proposition), is equal to 1, e.g. by lemma 2 of Chap. 1-2.
Theorem 5 is known as "Artin's product formula". From now on, we
will write kl for the kernel ofthe morphism z--+lzIA' i.e. for the SUhl!TOUp
of k~ given by IzlA = 1; by theorem 5, this contains kX.
COROLLARY 1. If k is of characteristic p> 1, k~ is the direct product

of kl and of a discrete subgroup isomorphic to Z.
For every place v of k, k" is of characteristic p, so that lxiv, for every
xek:, is in the subgroup of R~ generated by p; therefore the same is
true of IzlA for every zek~. This is the same as to say that the image of
k~ under the morphism z--+ IzlA is a subgroup ofthe group in question;
as it is clearly not reduced to {1}, it is generated by some integer Q = pN,
where N is an integer ;;?;1. Take zlek~ such that IZ1IA=Q; then k~
is the direct product of kl and of the subgroup generated by z h which is
clearly discrete and isomorphic to Z.
COROLLARY 2. Assume that k is of characteristic 0; for each AeR~,
call Z(A) the ide Ie (zv) such that Zv = 1 for every finite place v and Zw = A
76 Adeles IV
for every infinite place w of k. Then A-H(A) is an isomorphism of R~

onto a closed subgroup M of k~, and k~ is the direct product of kl and of M.
With the notation of the corollary of prop. 2, § 3, it is clear that
A-+Z(A) is an isomorphism of R~ onto a subgroup M of kA(P"Y. The
definition of IzIA' together with corollary 2 of tho 4, Chap. 111-4, shows
that IZ(A)IA = Aft, n being the degree of k over Q. The last assertion is now
obvious.
THEOREM 6. Let kl be the subgroup of k~ defined by IzlA = 1. Then
k X is a discrete subgroup of kl; the factor-group kllk x is compact; and
k~/k x is the direct product of that compact group and of a group isomorphic
to R~ or to Z according as k is of characteristic 0 or not.
The first assertion is contained in tho 5; the second one is the special
case D = k, J.l = 1 ofth. 4 of§ 3; the others follow at once from the corollaries
ofth.5.
We will now investigate more closely the structure of various sub-
groups of k~ and of k x and of some of their factor-groups. It will be
convenient to write D(P) for the group denoted by k A (P) x in the corollary
of prop. 2, § 3. In other words, we will write, from now on:
(5) D(P)= Ok; x Or;.
veP v,P
As always, P is assumed to be a finite set of places of k, containing the

set P<Xl of the infinite places; it may be empty, but only if k is not of
characteristic O. We recall that D(P) is always an open subgroup of
k~; clearly it is compact if and only if P is empty. We will also write:
D 1 (P)=D(P)nkl;
here we may take P = 0 if k is of characteristic p> 1, and then we have
D 1 (fi)=D(fi).
THEOREM 7. If P is not empty, the group k~/kx D(P) is finite. When k

is of characteristic p> 1, kllk x D(ftl) is finite, and k~/kx D(ftl) is the direct
product of that group and of a group isomorphic to Z.
In all cases, kllk x D 1 (P) is isomorphic to the quotient of kl!k" by

the image of D 1 (P) in kl!k x . As D 1 (P) is open in kl, that image is open;
as kl!k x is compact by tho 6, the quotient in question is fmite. If k is of
characteristic 0, D(P) contains the group M defined in corollary 2 of
tho 5; that corollary shows then that D(P) is the direct product of D 1 (P)
and of M, so that k~/e D(P) may be identified with kllk x Dl(P),
Assume now that k is of characteristic p> 1. As Q(0) = Q 1 (0), corollary 1
of tho 5 shows that k~/kx Q(ftl) is the direct product of the fmite group
g = kl/k x .0(0) and of a group y isomorphic to Z. If P =1= 0, .Q(P) contains

.0(0) and is not contained in kl; therfore k~/e .Q(P) is the quotient
of k~/kx .0(0), i. e. of g x y, by the image of k X .Q(P) in that group, and that
image is not contained in the image g of kl; it is then obvious that this
quotient is a finite group.
COROLLARY. Notations being as in theorem 7, one can choose P so
that k~ = k .Q(P).
X
Take any non-empty pi, and take a full set of representatives Z 1' •.. , Z N
for the classes in k~ modulo k X .Q(P'). As k~ is the union of all the
groups .Q(P), one can choose P::::> pi so that all the' Zj are in .Q(P). Then P
has the required property.
In the case when k is an algebraic number-field, and P = P00' theorem 7,
as will be seen in the next Chapter, is in substance the classical theorem of
the finiteness of the number of ideal-classes in k.
THEOREM 8. Let F be the set of the elements e of k such that lelv~ 1

for all places v of k, and put E = F - {O}. Then E is a finite cyclic group
consisting of all the roots of 1 in k.
The set F is the intersection of k and of the set of the elements (xv)
of kA such that IXvlv~ 1 for all v; clearly the latter set is compact, and, by
tho 2 of § 2, k is discrete in kA; therefore F is finite. If eeE, tho 5 shows
that we must have lelv= 1 for all v; therefore E is a subgroup of k X of
finite order, hence cyclic by lemma 1 of Chap. 1-1. Conversely, it is obvious
that every root of 1 in k must be in E.
COROLLARY. If k is of characteristic p > 1, the set F defined in theorem S-
is a finite field, the algebraic closure of the prime field in k.
Here the definition of F can be written as F = kn(n r v), where the
product is taken over all the places v of k; this shows that F is a ring;
as E = F - {O} is a group, F is a field. By tho 2 of Chap. 1-1, if an element
of k, other than 0, is algebraic over the prime field, it is a root of 1, so that,
by tho 8, it is in E.
When k is of characteristic p> 1, the finite field F defined in the
corollary of tho 8 is called the field of constants of k.
Now, the set P being as before, we define a subgroup E(P) of k x by
putting
E(p)=e n.Q(P)=kx n(n k: x n r:).
veP VfP
This consists of the elements e of k X such that lelv= 1 for all v not in P.
Obviously E(P) contains the group E defined in theorem 8. As k x is
78 Adeles IV
discrete in k~, E(P) is a discrete subgroup of Q(P), and also, in view of

tho 5, of Q 1 (P). One may also describe E(P) as the group k(P) x of invertible
elements (or, as one says traditionally, of "units") of the subring k(P)
of k given by
k(P)=kn (n kv x nrv) ,
veP v*P
and consisting of the elements e of k such that lelv ~ 1 for all v not in P.
In order to determine the structure of E(P), we need an elementary
lemma:
LEMMA 5. Let G be a group, isomorphic to Rr x zs+ l-r, with s~r~O.
If r>O, let A be a morphism of G into R, non-trivial on Rr; otherwise
let A be a non-trivial morphism of G into Z. Let Gibe the kernel of A,
and let r be a discrete subgroup of G1 , such that Gt/r is compact. Then
r is isomorphic to ZS.
We may assume that G=Rr x zs+l-r; then every element x of G
can be written as (xo, ... , x s), with XjE R for 0 ~ i< rand XjE Z for i ~ r,
and A can be written as
X=(Xo, ... , Xs)-tA(X)= L ajX j

j=O
with ajER for all i, if r>O, and ajEZ for all i, if r=O; in both cases, in
view of our assumptions about A, we may assume that ao f 0, and in the
former case we may assume that ao= 1. Consider G as embedded in the
obvious manner in the vector-space V=R s + 1 over R; then the above
formula defines A as a linear form on V; let VI be the subspace of V defined
by l(x)=O, so that G1 =GnVl • For l~j~s, call ej the point (Xj) in V
given by Xo= -aj' xj=a o, and xj=O for if 0 and ifj. As {e 1 , ... , es}
is a basis for VI, it generates an R-Iattice H in VI, so that V1/H is compact;
as He Gl' and G 1 is closed in VI' this implies that Vt/G 1 is compact.
Consequently, if r is as in the lemma, vt/r is compact, so that r is an
R-Iattice in VI' hence isomorphic to zs by prop. 11 of Chap. 11-4.
THEOREM 9. Let P be any finite set of places of k, containing Pro;
e
let E(P) be the subgroup of k x consisting of the elements of k x such
that lelv= 1 for all v not in P. Then E(P) is the direct product of the group E
of all roots of 1 in k, and of a group isomorphic to ZS, with s=O if Pis
empty, and s = card(P) -1 otherwise.
If P is empty, this is contained in tho 8; therefore we may assume
Pf0. Call v the morphism ofQ(P) into R~ induced by z-+ IzIA; its kernel
is Ql(P) and is open in kl. The canonical morphism of kl onto kl/e
induces on Q 1 (P) a morphism of Q 1 (P) onto its image in k)jk x , with the
kerneIE(P)sincekxn 0 1 (P)is the same as kXn O(P). Therefore 0 1 (P}/E(P)

is isomorphic to an open subgroup of k}Jkx, hence compact by tho 6.
On the other hand, for each place v of k, call U., the compact subgroup
of k; defined by Ixl v = 1, this being the same as r~ when v is a finite
n
place; put U = Uv' the product being taken over all the places of k;
this is a compact subgroup of O(P) and of 0 1 (P). Put G = O(P)/U ; clearly
this is isomorphic to the product ofthe groups k;/Uv for VEP; as k;/Uv
is isomorphic to R~, or, what amounts to the same, to R, when v is an
infinite place, and to Z otherwise, G is isomorphic to R' x Z'+ 1-" where r
is the number of infinite places of k, and s is as defined in our theorem.
As U is contained in the kernel 0 1 (P) of v in O(P), v determines on G a
morphism of G into R~, or, what amounts to the same, a morphism A.
of G into R, which is clearly non-trivial on each one of the factors k; /Uv
of G, and in particular on those which are isomorphic to R if there are
such factors, i. e. if r > O. On the other hand, if r = 0, we know, by corollary 1
of tho 5, that IzlA takes its values in a group isomorphic to Z, so that, up
to an isomorphism, A. maps G into Z. Therefore G and A. satisfy the
assumptions in lemma 5; the kernel G1 of A. is here the image of 01(P)
in G, i. e. 0 1 (P)/U. Call now r the image of E(P) in G.1f W is any compact
neighborhood of 1 in O(P), WU is compact and has therefore a finite
intersection with E(P). As the image of that intersection in G is the inter-
section of r with the image of WU in G, and as the latter is a neighborhood
of 1 in G, this shows that r is discrete in G. The factor-group Gtlr is
isomorphic to 01(P)/E(P)U, hence to a factor-group of the compact
group 0 1 (P)/E(P), and is therefore compact. We can now apply lemma 5
to G, A. and r; it shows that r is isomorphic to Z'. As E(P)nU =E, the
morphism of E(P) onto r, induced by the canonical morphism of O(P)
onto G, has the kernel E. Let now e1 , ... , e. be representatives in E(P)
of a set of s free generators of r; obviously they generate a subgroup of
E(P), isomorphic to Z', and E(P) is the direct product of E and of that
group. This proves our theorem; we have also proved the following:
COROLLARY. Assume that P is not empty; let E(P) be as in theorem 9;
put .Ql(p)=.Q(P)nkl and G 1 =.Q1(P)/U, where U is the group of the
r
elements (zv) of k~ such that IZvlv= 1 for all V. Then the image of E(P)
in G 1 is discrete in G 1 , and G 1/r is compact.
In the case when k is an algebraic number-field, and P = P00' theorem 9,

as will be seen in the next Chapter, is Dirichlet's famous "theorem of the
units".
Chapter V
Algebraic number-fields
§ 1. Orders in algebras over Q. We shall need some elementary results
about vector-spaces over Q, involving the following concept:
DEFINITION 1. Let E be a vector-space of finite dirrension over Q.

By a Q-Iattice in E, we understand a finitely generated subgroup of E
which contains a basis of E over Q.
PROPOSITION 1. Let E be a vector-space of finite dimension over Q;
let L, 1: be two Q-lattices in E. Then there is an integer m > 0 such
that mLc1:.
Let {el' ... , e,} and {e~, ... , e~} be finite sets of generators for Land
for L', respectively. As the latter must contain a basis for E over Q, we
can write (perhaps not uniquely) ej = Laijej for 1 ::::;;i::::;;r, with coeffi-
cients aijEQ. Take for m an integer >0 such that maijEZ for all i,j.
Then mLc1:.
COROLLARY 1. Let E be as in proposition 1. Then every Q-Iattice L

in E has a set of generators which is a basis of E over Q.
Let p be a basis of E over Q, contained in L; let L' be the Q-Iattice

generated by p; by proposition 1, there is an integer m > 0 such that
mLc1:. Consider E as embedded in E.=E®QR. By prop. 11 of
Chap. 11-4, 1: is an R-Iattice in Ea; as L is contained in m- 1 L: the same
proposition shows, firstly, that L is also an R-Iattice in Ea, and secondly
that it is generated by a basis of Ea over R; as this basis is contained
in E, it is clearly a basis of E over Q.
COROLLARY 2. Let E and L be as in corollary 1. Then every subgroup

1: of L which contains a basis of E over Q is a Q-lattice in E.
Let pi be a -basis of E over Q, contained in L'; let L" be the Q-Iattice

generated by p'. By proposition 1, there is an integer m>O such that
mLcL". Then m- 1 L"-=:JL-=:JL'-=:JL". Clearly, if n is the dimension of E
over Q, L" has the index m" in m- 1 LN. Therefore L" is of finite index
in L'; as [) is generated by P' and any full set of representatives of the
classes modulo L" in L', this proves our corollary.
§ 2. Lattices over algebraic number-fields 81
DEFINITION 2. Let .91 be an algebra of finite dimension over Q. A sub-

ring of .91 will be called an order of .91 if it is a Q-Iattice in .91 when .91 is
viewed as a vector-space over Q.
Here, as always, a subring of .91 is understood to contain the unit
ofd.
PROPOSITION 2. Every algebra .91 of finite dimension over Q con-
tains at least one order.
Let {a 1 , ... ,aN } be a finite subset of .91, containing a basis of .91
over Q; then we can write a;aj = LC;jhah for all i,j, with coefficients C;jh
in Q. Let m be an integer >0, such that mC;jheZ for all i,j,h. Then the
Q-Iattice generated by 1,ma 1 , ... ,maN is an order.
Take for instance d=Q. By corollary 1 of prop. 1, every Q-Iattice
in Q is of the form aZ, with aeQx. If this is an order, we must have
a2 eaZ, hence aeZ, and 1eaZ, hence a- 1 eZ; this gives a= ±1, which
shows that Z is the only order in Q.
PROPOSITION 3. Let a be any element of an order in an algebra .91
of finite dimension over Q. Then a is integral over Z, and Trd/k(a) and
N d/k(a) are in Z.
Let R be an order containing a, and let {a 1 , ... ,aN} be a finite set
of generators for R. Then we can write a' a;= LCijaj for 1 ~i~N, with
coefficients c;jeZ; this can be written as L(<5ija-c;j)aj=O, where (<5;)
is the unit matrix IN' Write D(T) for the determinant of the matrix
(<5;j T - C;j), where T is an indeterminate, and Dij(T) for its minors for
1 ~ i~j ~ N; these are polynomials in Z[T], and we have
LD;h(T)' (<5;j T - cij) = <5 hj D(T)
;
for 1 ~h,j~N. Substitute a for T, multiply to the right with aj' and
sum over j for 1 ~j~N; we get D(a)ah=O for all h, hence D(a)x=O for
all X; for X= 1, this gives D(a)=O, which proves our first assertion since
D(T) is monic. By corollary 1 of prop. 1, we may assume that we have
taken for {al, ... ,aN} a basis ofd over Q; then Trd/k(a) and Nd/k(a) are
the trace and the determinant of the matrix (c;j), so that they are integers.
§ 2. Lattices over algebraic number-fields. From now on, until the end
of this Chapter, k will denote an algebraic number-field. We keep the
notations explained in Chapter IV. In particular, if v is any place of k,
kv is the completion of k at v; if v is a finite place, rv is the maximal
compact subring of kv, and Pv the maximal ideal ofrv' We write kA for the
adele ring of k, and q> for the canonical injection of k into kA • We will write
q>E for the canonical injection of any finite-dimensional vector-space E
82 Algebraic number-fields v
over k into its adele space EA, this being defined by e-+e®<p(1) as
explained in Chap. IV-I.
Consider now the algebra k ®QR over R; this is the same as (kIQ)oo
in the notation ofth.1, Chap. IV-1, and it has an isomorphism fPoo onto
n
the direct product kw of the completions of k at its infinite places w,
this being fully characterized by the properties stated in tho 4 of
Chap. 111-4. We will simplify notations by identifying (kIQ)oo with that
product by means of fP 00' and by writing koo for both. Similarly, if E is
any fmite-dimensional vector-space over k, we will write Eoo for E ®QR,
which is the same as (EIQ)oo in the notation of coronary 2 of tho 1,
Chap. IV-1; as this is also the same as E ®kkoo, we identify it with the
product nEw taken over the infinite places w of k.
With this notation, the open subgroup kA(p00) of kA' given by for-
mula (1) of Chap. IV-1, can be written as koo x (n
rv), where the latter
product is taken over all the finite places v of k and is compact. Here,
and in similar situations, the following group-theoretic lemma will be
found useful:
LEMMA 1. Let G be a locally compact group with an open subgroup
G1 of the form G1 =G' x G", where G' is locally compact and G" is com-
pact. Let r be a discrete subgroup of G such that Glr is compact, and
call F' the projection of rnG 1 onto G'. Then F' is discrete in G', and
G'IF' is compact.
Let W be a compact neighborhood of the neutral element in G' (we
need not assume that G, G', G" are commutative, although only this case
will be used). As W x G" is compact, its intersection with r is rmite;
as the projection of that intersection onto G' is WnF', this shows that F'
is discrete. As G1 is open in G, G1 r and G-G 1 r are open, since they
are unions ofteft cosets for G1; therefore the image of G1 in GIr is open
and closed there, hence compact. As it is isomorphic to G1/r1 with
r1 =rnG t , this implies that there is a compact subset C of G1 such
that G t = C . r t • Then, if C' is the projection of C onto G', G' = C' . F',
which shows that G'IF' is compact.
THEOREM 1. Let k be an algebraic number-field; put t= C) (knrv),
v
where v runs thr~ugh all the finite places of k. Then t is an order of k;
it is the unique maximal order of k, and it is the integral closure of Z in k.
As explained above, write kA (P00) as koo x (n

rv). Clearly an element e
of k is in t if and only if <p@ is in that product; when that is so, write
n
<poo(e) and r/J@ for the projections of <p@ onto koo and onto rv' respec-
tively. Obviously t is a subring of k. Now apply lemma 1 to G=kA,
Gt=kA(Poo), G'=koo, G"=nrv, r=<p(k); then, with the notations of
§ 2. Lattices over algebraic number-fields 83
that lemma, r' is cpoo(r), and the lemma shows that this is an R-Iattice in koo .
As CPoo is also the same as the injection induced on r by the natural
injection of k into koo = k ®Q R, this implies that r is a Q-Iattice in k,
hence an order. Let r' be any subring of k whose additive group is finitely
generated; clearly the rv-module generated by r' in kv is a compact sub-
ring of k v ; it contains rv , since r' contains 1; therefore it is rv , so that
r' c rV. As this is true for all v, we get r' c r. By prop. 3 of § 1, r is con-
tained in the algebraic closure of Z in k. Conversely, if an element of k
is integral over Z, prop. 6 of Chap. 1-4 shows that it is in rv for all v,
hence in r.
The mapping IjJ of r into TI r v' defined in the proof of theorem 1,
will be called the canonical injection of r into TI rv; it maps every E r e
e
onto the element (xv) of that product given by Xv = for all v. It is a ring-
isomorphism of r onto ljJ(r), addition and multiplication in TI r v being
defined coordinatewise. With this notation, we have:
COROLLARY 1. Let k, rand IjJ be as above defined. Then IjJ (r) is dense
in TI r,,, and its projection onto every partial product of that product
is dense there. In particular, rv is the closure of r in kv.
Let G, GI , G', G", r be as in the proof of theorem 1. By corollary 2

of tho 3, Chap. IV-2, koo + cp(k), which is the same as G'r in that notation,
is dense in G = kA' so that its intersection with GI must be dense in GI;
as that intersection is koo + cp(r), this implies that its projection onto
G" = TI r v , which is the same as the projection IjJ (r) of cp(r) onto G", is
dense there. The other statements in our corollary follow trivially from this.
COROLLARY 2. If k' is a finite algebraic extension of k, the maximal
order of k' is the integral closure of r in k'.
This follows again from prop. 6 of Chap. 1-4, just as in the proof of
theorem 1.
DEFINITION 3. Let k be an algebraic number-field, r its maximal order,
and E a vector-space of finite dimension over k. An r-module in E will
be called a k-Iattice in E if it is finitely generated and contains a basis
of E over k.
If k' is a finite algebraic extension of k, r' its maximal order, and E
a vector-space of finite dimension over k', i~ is clear that an r' -module
in E is a k'-lattice if and only if it is a k-Iattice when E is viewed as a
vector-space over k.
Let E be a vector-space of finite dimension over k. Let L be a k-Iattice
in E; let e be a finite subset of E such that L is the r-module generated
bye in E. Then, for every finite place v of k, the rv-module ev generated
by e is the same as the rv-module L" generated by L, and prop. 1 of
Chap. IV-l shows that EA(Poo ' e) is the same as Eoo x nLv and is an open
subgroup of EA' For every eeL, we can define an element (e v) of nLv
by putting ev=e for all V; if we call this element t/lL(e), t/lL will be called
n
the canonical injection of L into Lv' Then:
PROPOSITION 4. Let E be a vector-space of finite dimension over k.
Let L be a k-Iattice in E; for every finite place v of k, let Lv be the rv-
module generated by L in Ev; and let t/I L be the canonical injection of
n n
L into Lv' Then t/I dL) is dense in Lv; its projection onto every partial
n
product of Lv is dense there; in particular, for every v, Lv is the closure
of Lin Ev'
Let e= {e 1"'" eN} be a finite subset of L such that L is the r-module

n
generated bye. Take any element (e v) of Lv; then, for every v, we can
n
write ev= Lx~)ei with coefficients x~) in rv' Put Xi=(X~») for l~i~N;
the Xi are elements of r v' By corollary 1 of tho 1, we can find elements ~i
of r such that, for every i, t/I (~i) is arbitrarily close to Xj; clearly, then,
t/lL(L~jei) can be made to be arbitrarily close to (e v)'
THEOREM 2. Let k be an algebraic number-field, E a vector-space

of finite dimension over k, and La k-Iattice in E. For each finite place v
of k, let Lv be the closure of L in Ev and Mv any kv-Iattice in Ev' Then
there is a k-Iattice M in E whose closure in Ev is M v for every v if and
only if M v= Lv for almost all v; when that is so, there is only one such
k-lattice, and it is given by M = nv
(EnMv)'
Assume that there is such a k-Iattice M; in view of prop. 4, the fact

that then M v = Lv for almost all v is merely a restatement of corollary 1
of tho 3, Chap. 111-1. Now assume that Mv=Lv for almost all v; in view
n
of prop. 1 of Chap. IV-I, this implies that Eoo x Mv is open in EA' We
n
can therefore apply lemma 1 to G = EA, G' = Eoo , Gil = M vand r = Cf'E(E),
where Cf'E is the canonical injection of E into EA' Clearly, if we put
M=n(EnM v), Cf'E(M) is the same as Cf'E(E)nG 1 with G1 =G'xG";
lemma 1 shows now that M is an R-Iattice in Eoo ' hence a Q-Iattice
in E; as it is obviously an r-module, it is a k-Iattice. By corollary 2 of
th.3, Chap.IV-2, Eoo + Cf'E(E) is dense in EA; therefore its intersection
Eoo + Cf'E(M) with G1 is dense in Gl' This is the same as to say that the
n
projection of Cf'E(M) onto Gil = M v is dense there, and implies that M
is dense in Mv for every V. As above, call t/lM the canonical injection
n
of Minto M v. Assume now that there is another k-Iattice M' in E
with the closure M v in Ev for every v; clearly M' is contained in M;
n
moreover, by prop.4, t/lM(M') is dense in M v, hence also in t/lM(M).
§ 3. Ideals 85
By prop. 1, there is an integer m>O such that M'~mM. Call Gm the

image of G1 under the automorphism e-+IP(m)~ of EA ; this can be written
n
a.s Gm = G' x G~ with G~= (mMv); clearly mMv=Mv for almost all v
(viz. for all the finite places of k which do not lie above some prime
divisor of min Z), and G~ is an open subgroup of G". Then IPE(mM) is
the same as IPE(E)nG m, hence also the same as IPE(M)nG m , and is con-
tained in IPE(M'); this is the same as to say that t/lM(M)nG~ is contained
in t/lM(M'). Now take any J1.EM; as t/lM(M') is dense in t/lM(M), there
is J1.' EM' such that t/lM(J1. - J1.') is in G~; then it must be in t/lM(M'), so
that J1.-J1.'EM' and J1.EM'. This shows that M=M', which completes
the proof.
COROLLARY. Let L, L' be two k-Iattices in E. Then L+ L' and LnL'

are k-Iattices in E, and, for every finite place v of k, their closures in Ev
are given in terms of the closures Lv, L~ of L, L' by
The assertions about L+ L' follow at once from prop. 4. As to LnL:

put Mv=LvnL~ for every v; for every v, this is a kv-Iattice in Ev, and it is
the same as Lv for almost all v. Therefore there is a k-Iattice M in E
n
with the closure Mv in Ev for every v, and it is given by M = (EnMv);
in view of tho 2, this is the same as L n L'.
§ 3. Ideals. In this §, k will denote an algebraic number-field and

t its maximal order; the results of § 2 will be applied to the case E = k.
Clearly an t-module other than {OJ in k is a k-Iattice if and only if it is
finitely generated. By prop. 1 of § 1, if a is a k-Iattice in k, there is an
integer m>O such that ma is contained in t; then, clearly, ma is an ideal
in the ring t. Conversely, by corollary 2 of prop. 1, § 1, every ideal in l,
other than {OJ, is a k-Iattice. This shows that a subset of k is a k-Iattice
if and only if it is of the form ~ a, where a is an ideal in l, other than {O},
and ~Ekx.
DEFINITION 4. Any k-Iattice in k will be called a fractional ideal in k;

a fractional ideal in k is said to be integral if it is contained in l.
Accordingly, {OJ is not a fractional ideal.

Let a be a fractional ideal in k, and let L be a k-Iattice in a vector-
space E of finite dimension over k. By a L, one understands the sub-
group of E generated by the elements ~e with ~Ea, eEL; this is clearly
a k-Iattice in E. Let v be any finite place of k; as before, write av for the
closure of a in kv, and Lv, (aL)v for those of L, aL in Ev; by prop. 4 of
§ 2, these are the same as the r v-modules generated respectively by a, L
and aL; this makes it clear that (aL)v is the same as the subgroup avLv
of Ev generated by the elements IXe with IXea v, eeLv'
In particular, if a, b are two fractional ideals in k, ab is the subgroup
of the additive group of k generated by the elements IXP with IXea, peb;
it is a fractional ideal, and, for every finite place v of k, we have (ab)v=avb".
If Pv is the maximal ideal in rv, every kv-Iattice in kv is of the form p~
with neZ; in particular, we can write av=p~, bv=pt with aeZ, beZ,
and then it is obvious that avb v= p~+b.
THEOREM 3. Let k be an algebraic number-field and t its maximal

order. For every finite place v of k, put pv=tnpv' Then v--+Pv is a bijection
of the set of finite places of k onto the set of the prime ideals in t, other
than {O}. For the law (a,b)--+ab, the set of the fractional ideals in k is a
group with the neutral element t; it is the free abelian group generated
by the prime ideals in t; the ideals in t, other than {O}, make up the monoid
generated by these prime ideals.
For every fractional ideal a in k, we can define a mapping v--+a(v)

of the set of finite places of k into Z by writing av=p~(V). For a=t, all
the a(v) are O. Theorem 2 of § 2 shows now that a given mapping v--+a(v)
belongs to a fractional ideal a if and only if a(v)=O for almost all v, and
a= n
that it determines a uniquely when that is so, a being then given by
(knp~(v»). If b corresponds similarly to v--+b(v), we have seen
above that a b corresponds to v--+a(v)+ b(v); it is also clear that a c b if
and only if a(v)~b(v) for all v; in particular, a is integral if and only if
a(v)~O for all v. For any given v, put a(v) = 1 and a(v')=O for all v'=!=v;
if we call Pv the corresponding ideal, we have pv=tnpv, and it is clear
that the fractional ideals make up the free abelian group generated by
the PV' As Pv is prime in rv, Pv is prime in t. As to the converse, take any a
in t, so that a(v)~O for all v; if it is neither t nor anyone of the Pv' we
can write it as a' a", where a', a" are ideals in t, other than t. Then a'
contains a and is not a, so that a' - a is not empty; the same is true
of a" - a. Take IX' e a' - a and IX" e a" - a. Then IX' IX" is in a, while neither
IX' nor IX" is in a, so that a is not prime. This completes the proof.
COROLLARY 1. Let a, b be two fractional ideals in k; for each v, call

a(v) and b(v) the exponents of Pv in a and in b when these are expressed
as products of powers of prime ideals of t. Then a + b and an b are fractional
ideals in k, and when they are similarly expressed, the exponents of Pv in
them are min (a (v), b(v)) and max (a (v), b(v)), respectively.
This follows at once from tho 3 and the corollary of tho 2, § 2.

As usual, two ideals a, b in r are called mutually prime if a + b = r.
§ 3. Ideals 87
COROLLARY 2. Every fractional ideal n in k can be written in one

and only one way in the form be-I, where band c are mutually prime
ideals in r.
This follows at once from tho 3 and corollary 1. By analogy with the
case of Q, the ideals b, c in corollary 2 are called the numerator and the
denominator of n, respectively.
We will denote by J(k) the group of fractional ideals of k. If a=(a v)
is any element of k~, then, by the corollary of prop. 2, Chap. IV-3, we
have lavl v= 1, hence avrv=rv, for almost all finite places v of k; therefore,
by tho 3, there is one and only one fractional ideal n of k such that nv = av r v
for all finite places v; we will write n=id(a) for this ideal. Clearly the
mapping a-+id(a) of k~ into J(k) is surjective; we will write D"" for its
kernel, which is obviously k~ x (nr:), i.e. kA(p""r in the notation of
the corollary of prop. 2, Chap. IV-3, and D(P",,) in the notation of for-
mula (5), Chap. IV-4; as this is an open subgroup of k~, a-+id(a) is a
morphism of k; onto J(k) if J(k) is provided with the discrete topology.
We may then identify J(k) with k~ ID"".
In particular, for every ~ekx, we have id(e)=~r; this is the r-module
generated by ~ in k, and is frequently denoted by (e); its numerator and
denominator, as defined above, are called the numerator and the denomi-
nator of~. A fractional ideal is called principal if it is ofthe form ~r with
~ekx; such ideals make up a subgroup P(k) of J(k), which is the image
of k X under the morphism induced by a-+id(a). Identifying J(k) with
k~ ID"", we see that P(k) is the image of k x in the latter group; therefore
we may identify J(k)IP(k) with k~ Ik x D"", which is a finite group by
tho 7 of Chap. IV-4. The elements of J(k)IP(k), or in other words the classes
modulo P(k) in J(k), are known as the ideal-classes of k. The number of
such classes, i.e. the index of P(k) in I(k), will be denoted by h.
THEOREM 4. Let k be an algebraic number-field, E a vector-space of
finite dimension over k, and L, M two k-Iattices in E such that L;:) M.
For every finite place v of k, call Lv, Mv the closures of L, M in Ev, and
call A. v the natural homomorphism of LIM into LvlMv. Then x-+(A.ix»),
where v runs through all finite places of k, is an isomorphism of LIM onto
n
v
(Lvi M v) for their structures as r-modules, r being the maximal order of k.
Call A. that mapping; it is obviously a homomorphism of r-modules.
Let x be any element of LIM, and e a representative of x in L. If A.(x)=O,
e must be in Mv for all v; by tho 2 of § 2, this implies that ee M and x = O.
Therefore A. is injective. Now take any element y=(Yv) of n(LvlMv);
for every v, take a representative ev of Yv in Lv, and put e=(e v)' As
Mv= Lv for almost all v, n n
Mv is open in Lv; therefore prop. 4 of
§ 2 shows that there is an element eo of L such that t/lL(eO)-e is in
nMv, '" L being the canonical injection of L into n

Lv. This is the same
as to say that A(X O) = y if Xo is the image of eo in L/M, which proves
that A is surjective.

we have [L:M] = n[Lv:Mv].
This is obvious. One should observe that 4 = M v for almost all v,
and that, for all v, M v is an open subgroup of the compact group 4,
so that [Lv:Mv] is always finite and almost always 1. The fact that
[L:M] is finite is implicit in prop. 1 of§ 1, or also in lemma 2 of Chap. 11-4.
COROLLARY 2. Let v be a finite place of k; let Vv=tnpv be the prime

ideal in the maximal order t of k, corresponding to v. Then the natural
homomorphism of t/Vv into r v/Pv is an isomorphism of t/Vv onto the residual
fi('ld rv/pv of rv·
COROLLARY 3. Let 0, b be two fractional ideals in k, such that 0:::> b.
n
Let 0 - 1 b = V=(v) be the expression of 0 - 1 b as a product of prime
ideals oft. Then [o:b] = n[t:Vv]ft(V).
By corollary 1, [o:b] is the product of the indices [ov:bv] for all v.

For a given v, we can write ov=P~, bv=p:, and then we have b-a=n(v).
Corollary 2 ofth. 6, Chap. 1-4, shows that [p~:p:] =qb-a with q= [rv:pv].
Our conclusion follows at once from this and corollary 2.
DEFINITION 5. Let k be an algebraic number-field and r its maximal
order. Let 0~91(0) be the homomorphism of the group of fractional ideals
of k into Q x which is such that 91 (V ) = [r: V] for every prime ideal V in r.
Then 91(0) is called the norm of the fractional ideal 0 in k.
Corollary 3 of theorem 5 can now be expressed by saying that, if 0
and b are fractional ideals in k, and 0:::> b, then [o:b] is equal to 91(b)/91(o).
In particular, if 0 is integral, [r:o] =91(0).
PROPOSITION 5. Let a = (a v) be any element of k~. Then 91(id(a))
is equal to the product nlavl; 1, taken over all the finite places v of k.
In view of def. 5, it is enough to verify this for the case when id (a) is a
prime ideal of t; this is so if and only if, for some finite place v of k,
av is a prime element of kv, and lav,lv' = 1 for all finite places v' =F v. Then
it is obvious.
COROLLARY 1. For each eek x , we have NkIQ(e)=( -1)P91(id@),
p being the number of the real places w of k such that the image of in e
kw is <:0.
§ 4. Fundamental sets 89
Combining proposition 5 with tho 5 of Chap. IV-4, we see at once that

9l(id(e)) is equal to the product nlewlw, taken over the infinite places
w ofk. For each real place w ofk, and each xEk:, we have x=(sgnx)'lxlw;
for each imaginary place w of k, and each xEk~, we have Nkw/R(X)=xx=
= Ixlw' Our conclusion follows now at once from corollary 3 of tho 4,
Chap. 111-4, applied to k, Q and the place 00 of Q.
COROLLARY 2. An element eof the maximal order r of k is invertible
in r if and only if Nk/Q(e) = ± 1.
Clearly, it is invertible in r ifand only if er=r; as er is the same as
id(e), our conclusion follows now at once from corollary 1, combined
with the fact that [r: (l] = 9l(a) for every ideal (l in the ring r.
Traditionally, the elements of r X , i.e. the invertible elements of r,
are known as "the units" of k. In the notation of Chap. IV-4, r x is the
same as the group E(Pex,), as defined in theorem 9 of Chap. IV-4; its
structure is given by that theorem; if r+ 1 is the number of the infinite
places of k, it is isomorphic to the direct product of the cyclic group E
of the roots of 1 in k, and of a group isomorphic to Z'. This is Dirichlet's
"unit-theorem".
§ 4. Fundamental sets. Let r be a discrete subgroup of a locally

compact group G; by a "fundamental set" of G modulo r, one under-
stands traditionally a full set X of representatives of the cosets modulo r
in G, which at the same time is measurable, and which is usually expected
to have some additional properties, e. g. to be a Borel set, etc. Then
formula (6) of Chap. 11-4, applied to G, r, to a Haar measure oc on G,
and to the characteristic function of X, shows that oc(X)=oc(Gjr); thus
the calculation of a(G/r) may sometimes be effected by constructing a
convenient fundamental set. More generally, let us say that a measurable
subset X of G is fundamental of order v modulo r if it has exactly v
points in common with every coset modulo r; then the same formula
gives oc(X)=voc(G/r). This will now be applied to kA and to k~.
Let k and r be as before; call n the degree of k over Q. As r is a Q-
lattice in k when k is viewed as a vector-space over Q, prop. 11 of Chap. 11-4
shows that it has a set of generators {e 1, ... , en} which is a basis of k
over Q. Then this is also a basis of koo = k®QR over R; therefore, if we
write, for u=(u 1 , ••• , un) in Rn, O(u) = LUiei, this defines an isomorphism
oof Rn onto k oo •
PROPOSITION
of R. Then O(I") x n• r
6. Let k, rand 0 be as above; call I the interval 0 ~ t < 1
v' where the product is taken over all the finite
places v of k, is a fundamental set modulo k in kA •
Call that set X; it is obviously measurable; we have to show that
every element x of kA can be written in one and only one way as Xo + e
e
with Xo e X and e k. By corollary 2 ofth. 3, Chap. IV-2, koo + k is dense
in k A ; as koo x nrv is open, this shows that, for a given xekA , there is
'1ek such that X-'1 is in koo x nrv, and the definition oft shows that an
n
element 'I' of k has the same property if and only if 'I' - 'I e t. Write y = x - 'I,
and call Yoo the projection of y onto koo in the product koo x r v; then we
can write Yoo =O(u) with u=(u 1 , ... , un) in R". For each i, take ajeZ such
that aj~uj<aj+l, i.e. uj-aje1; put e='1-Lajej and xo=x-e. As
nrv; moreover, the projection of Xo onto koo is
j
e - 'I is in t, Xo is in koo x
Yoo - Lajej= L(uj-aj)ej
i i
and is therefore in 0(1"). It is also clear that the latter condition could not
have been fulfilled by any other choice ofthe integers aj. This proves our
assertion.
This will now be applied to the calculation of a.(k,Jk) for an explicitly
given Haar measure a. on kA • Such measures can be constructed as follows.
For each place v of k, choose a Haar measure a. v on k v; if a.v(rv) = 1 for
almost all v, the product measure n
a. v is well defined and is a Haar
measure on each one of the open subgroups kA(P) of kA given by for-
mula (1) of Chap. IV-I; clearly there is one and only one Haar measure
on kA which coincides with these measures wherever they are defined;
n n
this will be denoted by a. v' In particular, we will write P= Pv for the
Haar measure obtained by taking pv(rv) = 1 for all finite places v of k,
and proceeding as follows at the infinite places. If w is a real place, we
have kw=R and we take dPw(x)=dx, so that Pw is the Lebesgue measure
on R. Ifw is an imaginary place, we have kw=C, and we take dPw(x) =
=ldxJ\dxl; by this we mean that, if we put x=u+iv with u, v in R, so
that dXJ\dx= -2i(duJ\dv), Pw is the measure corresponding to the
differential form 2du J\ dv; in other words, Pw/2 is the Lebesgue measure
in the (u, v)-plane.
In order to calculate P(k,Jk), we need another definition. Notations
being as above, consider the matrix
(1) M = (1t k/Q(eje))l ""j,j"""'
and call D its determinant. By prop. 5 of Chap. III-3, D:j=O; by prop. 3
of§ I,MisinMnCZ), so that DeZ. Ifk=Q, wehavet=Z, so that we have
to take el = ± 1, hence D= 1. If {'II' ... , 'In} is another set of generators
for t, and N the matrix obtained by substituting the '1j for the ej in (1),
we can write '1j= Lajjej with ajjeZ for all i,j; then we have N =AM 'A,
where A is the matrix (a jj). Similarly we can write ej= Lbjj'lj with
bijEZ for all i, j; calling B the matrix (b j ) , we have AB= In' hence
det(A)det(B) = 1; as det(A) and det(B) are in Z, this gives det(A)= ±1,
hence det(N) = det(M). In other words, the determinant D of M does not
depend upon the choice of the basis (~J This justifies the following:
DEFINITION 6. Let k and g 1, ... , ~n} be as above. Then the deter-

minant D of the matrix M given by (1) is called the discriminant of k.
PROPOSITION 7. Let P= npv be the Haar measure on kA obtained

by taking pv(rv) = 1 for all finite places v, dPw(x)=dx for all real places w,
and dPw(x) = Idx A dxl for all imaginary placeswof k. Then P(k.Jk)=IDI1/2,
where D is the discriminant of k.
n
Call Poo the measure npw on koo = kw, the products being taken
over the infinite places of k. By prop. 6, P(k.Jk) is the same as Poo((}(l"));
therefore our proposition will be proved if we show that
dPoo((}(u))=IDI 1/2du 1 .. · dUn.
Call r 1, r 2 the numbers of real and of imaginary places of k, respectively;
put r=r 1+r 2-1; let WO'''''Wr be the infinite places of k, ordered so
that Wj is real for i<r 1 and imaginary for i~r1' For each i, write kj for
n
the completion of k at Wj, Aj for the natural injection of k into kj, and J.lj
for the R-linear extension of Aj to k oo ; if we identify koo with kj as above,
tho 4 of Chap. III-4 shows that J.lj is the projection from koo onto kj.
By corollary 1 of prop. 3, Chap. III-2, every isomorphic embedding A.'
of k into C is of the form O'OA j , where 0' is an R-linear isomorphism of kj
into C; obviously 0' is the natural injection of k j into C if kj=R, i.e. if
i < r1, and it is one of the two mappings x -+ x, x -+ x of C onto C if kj = C,
i. e. j ~ T l ' Therefore, if we put A; = A; for 0 ~ i ~ T. and A~z + i = 1; for
T 1 ~ j ~ T, the A~ for 0 ~ h ~ n -1 are all the distinct isomorphisms
of k into C. Writing now J.l~ for the R-linear extension of A;' to koo' we have,
for u=(u 1, ... , un)ERn and O~h~n-1:
n
;=1
Call N = (A;'(~j)) the matrix of the coefficients in the right-hand sides. By

corollary 3 of prop. 4, Chap. III-3, we have, for all ~Ek, 1rk/Q @= ~);.(~),
and therefore, since the A;' are isomorphisms:
M =(LA;'(~j)A;'(~j))=tN' N,
h
hence D=det(N)2. At the same time, we have, in the exterior algebra of

differential forms on Rn:
(2) n dJl;'(O(u») = ± n dJlj(O(u») A n (dJlAO(U)) A djiAO(U»))

h O~i<rl rl~j:!6r
= ±det(N)du1 A ••• A dUn.

In view of the definition of the measures Pw, this completes the proof.
At the same time, one may note that one gets a real differential form on
Rn by multiplying (2) with i'2; therefore i'2det(N) is real, which is the
same as to say that (_1)'2 D>O.
COROLLARY 1. If k =1= Q, IDI > 1.

Notations being as above, choose cjER~ for O~i~r, and call Y(c)
the set of the elements y = (y v) of kA such that IYvl v~ 1 for all finite places v
of k, and IYw;lw;~cJ2 for O~i~r. For each infinite place w, and each
cER~, the subset of kw given by Ixlw~c/2is an interval of length C if
Wis real, and a circle of pw-measure nc if w is imaginary. In view of the
definition of p, this gives P(Y(c»)=n'2ncj.lfthis is >IDI1/2, lemma 1 of
Chap. 11-4, combined with proposition 7, shows that there are y, y' in
Y(c) such that 11= y' - y is in kX. Then we have 111lv~ 1 for all finite places
v of k, 111lw; ~ Cj if Wj is a real place, and, as one sees at once, 1'1lw; ~ 2cj if
Wj is imaginary; in view of tho 5 of Chap. IV-4, this implies 2'2 n Cj ~ 1.
Therefore, if r 2 >0, we get a contradiction if we assume that IDI = 1 and
choose the Cj so that nCj is >n-'2 and <2-'2. Now assume that r 2 =0,
hence r 1=n, and IDI = 1. Then, for every choice of the Cj such that n cj> 1,
there is 11 Ek x with the properties stated above. Clearly the set of elements
x=(x v) of kA which satisfy IXvlv~ 1 for all finite places v, and IXwlw~2 for
all infinite places w, is compact and therefore contains only finitely many
elements '11' ... , '1N of k; therefore we can choose c' > 1 such that none of
these satisfies 1 < l'1vlwo < c'. Choose now the Cj so that n Cj > 1, 1 < Co < c',
Co <2, and Cj< 1 for 1 ~ i ~n-l; then there is 11EP such that 1'1lv~ 1 for
all finite places, and 1'1lw; ~ Cj for 0 ~ i ~ n -1. In view of the definition
of c' and of our assumptions about the Cj, this implies 1'1lw; < 1 for i> 0,
and 111l wo~ 1. This contradicts tho 5 of Chap. IV-4, unless n = 1.
COROLLARY 2. There are only finitely many algebraic number-fields
k of given degree n over Q and given discriminant D.
As this will not be of any further use to us, we merely sketch the proof.
Proceeding just as above, one sees that there is 11Ekx such that 111lv~ 1
for all finite places v of k, 111lw < 1 for all infinite places Wexcept one such
place wo, and the image ,1.0('1) of 11 in kwo is in the intervallxl ~ 21Dll/2 if
Wo is real, and in the rectangle given by x=u+iv, lul~l, Ivl~IDI1/2 if
Wo is imaginary. As then we must have 111lwo> 1, the latter condition
implies that ,1.0(11) is not real ifwo is imaginary. This implies that k=Q('1);
for, if not, call u the place ofQ('1) lying below Wo; then 111lw> 1 for all the
Fundamental sets 93
places W of k above u, if there is more than one such place, and AO('1)
must be real if u is real and Wo imaginary; as this is not so, corollary 1 of
tho 4, Chap. III-4, shows that the degree of k over Q('1) cannot be > 1.
This implies that the A~('1), for O~h~n-l, are all distinct, so that
n(X -A~('1)) is the irreducible monic polynomial in Q[X] with the root
'1; its coefficients are obviously bounded in terms of IDI; they are all in Z
since 1'1lv~ 1 for all finite places v of k, which is the same as to say that '1 is
in t, i.e. integral over Z. Therefore the polynomial in question, hence
also '1, can take only finitely many values when D is given.
Now we will treat the corresponding problems for k~/kx . As above,

we write Qoo for the kernel of a--+id(a) in kA' this bein~ the group
k~ x nr;, i.e. the same as Q(P 00) in the notation of Chap. IV-4. We will
write Ql' instead of Ql(Poo), for Qoonkl. As in Chap. IV-4, we write U
for the group of the elements (zv) of k~ such that Izlv= 1 for all v, [mite or
not; this is a compact subgroup of Q l' As we have observed in § 3, t x is
the same as the group denoted by E(P 00) in the notation of Chap. IV-4.
We again write E for the cyclic group of the roots of 1 in k. Call again
Wo, ... , w, the infinite places of k, in any ordering. For each z=(zv) in
Qoo,put
(3)
The mapping I of Qoo into R'+ 1, defined by (3), is obviously a morphism

of the (multiplicatively written) group Qoo onto the (additively written)
group R'+ 1, with the kernel U. Let A be the linear form on R'+ 1 given by
A(X) = LXi for x=(x o, ... , x,). Then, for zEQoo' we have 10g(lzIA)=A(/(z));
therefore, if H is the hyperplane defined by A(X)=O in R,+l, the set
1-1(H), which is the kernel of ).0/, is the same as D 1 , and I induces on it
a morphism ofD1 onto H with the kernel U, which we can use to identify
the group G 1 =QdU with the vector-space H. Put r=/(rX); by the
corollary ofth. 9, Chap. IV-4, this is a discrete subgroup of H, and H/r is
compact; in other words, it is an R-Iattice in H. It is then obvious (just
as in the proof of tho 9, Chap. IV-4) that, if we take r elements e1 , ... , e,
of r x , these will be free generators of a subgroup of r x if and only if their
images I(e;) in R'+ 1 make up a basis for H, andthat rX will then be the
direct product of E and of that subgroup if and only if these images
generate r; when that is so, we will say that the ei make up a set of free
generators for r x modulo E. Assume now that they have been so chosen.
For 0 ~ i ~ r, call (j i the degree of kWi over R; this is 1 or 2, according as
Wi is real or imaginary, and, by corollary 2 ofth. 4, Chap. III-4, we have
L (ji = n, i. e. A((j) = n if we write (j for the vector «(jo, ... , (j,) in R' +1. This
i
implies that ~, together with the vectors l(e j ) for 1 ~i~r, makes up a
basis for R,+l, so that we can define an automorphism F ofR,+l by:
(4)
j=l
We have then l(F(t») = to, and also, for flEE and (n1' ... , n,) in Z':
(5) 1('1 neil)=F(O,n1, ... , n,).
PROPOSITION 8. Put 0 00 = k~ x nr:, and let I be the morphism of
0 00 onto R'+ 1 given by (3); let {ai' ... , all} be a full set of representatives
for the cosets modulo k X 0 00 in k~. Let E be the group of the roots of 1
in k; call e its order, {e 1, ... , e,} a set of free generators for rX modulo E,
and F the automorphism of R'+ 1 given by (4). Then, if I is the interval
O~t< 1 in R, the union of the sets ajl-1(F(R x r») for 1 ~i~h is a fun-
damental set of order e modulo k x in k~.
Take any Z = (zv) in k~; there is one and only one i such that ai 1 Z is
e e
in k x 0 00 , and then we can write z = aj z' with E k x , Z' E 0 00 ; moreover,
z' is uniquelydeterminedmodulok x nOoo ' i.e. modulo rX. PutF- 1 (I(z'»)=
=(to, ... , t,); for 1 ~i~r, take njEZ such that nj~tj<nj+ 1; put e= neil,
z"=e- 1z' and e'=ee. Then we have z=e' ajz", and, in view of (4) and (5),
l(z")EF(R x r). Moreover, it is clear that z" is uniquely determined
modulo E by these conditions. This proves our proposition.
As we have seen in § 3, the morphism z-+id(z) of k~ onto I(k) deter-

mines an isomorphism of k~/kx 0 00 onto the group I(k)/P(k) of ideal-
classes of k; therefore the number h, occurring in proposition 8, is the
order of that group, and the ideles aj in that proposition may also be
characterized by saying that the fractional ideals id(a j ) are representatives
of the ideal-classes of k.
Now we define a Haar measure y on k~. Just as in the case of kA'
this may be done by choosing, for each v, a Haar measure Yv on k:, in
such a way that yv(r:) = 1 for almost all v; then we define y by prescribing
that it should coincide with nyv on everyone ofthe groups kA(Pf, and
we write y= nyv for this. As in the case of kA' we need a definition:
DEFINITION 7. Notations being as above, call L the matrix whose

rows are the vectors n-1~, l(e 1), ... , l(e,). Then R=ldet(L)1 is called the
regulator of k.
As L is the matrix of the automorphism F of R'+ 1 given by (4), F has
then the determinant ±R. Our definition would have to be justified by
showing that R is independent ofthe choice of the ej ; this could be done
easily by applying the same argument which we used for the discriminant.
As the same fact will emerge presently as a consequence of proposition 9,
we leave it aside for the moment.
PROPOSITION9. Let y= nyv be the Haar measure on k~ obtained
by taking yv(r~)= 1 for all finite places v of k, dYw(x)=lxl-1dx for each
real place w, and dYw(x)=(XX)-lldxAdxl for each imaginary place w.
For each m> 1 in R, call C(m) the image in k~/kx of the subset of k~
defined by 1 ~lzlA ~m. Then we have y(C(m)) = cklog(m), with Ck given by
ck=2"(21t)'2hR/e.
Here, as before, rl and r2 are the numbers ofreal and of imaginary
places of k, respectively; h is the number of ideal-classes; R is the regulator,
as defined above, and e is the order of the group E of roots of 1 in k, this
being always an even integer since ± 1 are in k. Clearly e = 2 if r1 > 0, since
R contains no root of 1 except ± 1.
We begin by modifying the representatives aj of the cosets modulo
e U oo in k~, introduced in prop. 8, by replacing, for each i, aj by ajbi 1
with bjEU oo and Ibjl A= lajlA; once this is done, we have lajlA = 1 for
1 ~i~h, and prop. 8 shows that ey(C(m)) = hy(X), where X is the inter-
section of 1-1(F(R x 1')) with the set l~lzIA~m in k~. As we have seen
above, if ZEU oo and F-1(l(z))=(t o, ... , t,), we have
10g(lzIA) = 2(/(z)) = 2(F(t)) = to.
Therefore the set X can be written as 1- 1 (F(J x 1')), where J is the interval
o~ t ~ log(m) of R. Now I is a morphism of Uoo onto R'+ 1 with the com-
pact kernel U; therefore, if Y is any compact subset of R'+ 1, r l(y) is a
compact subset of Uoo ' and Y --+y(l-l(y)) is a Haar measure on R'+ 1,
hence a multiple c tX( Y) of the Lebesgue measure tX on R' + 1, with some
constant c>O. This gives y(X) = ctX(F(J x 1')). By the definition of the
regulator R, it is the module of the automorphism F of R'+ 1; therefore
we get:
y(X) = ctX(F(J x 1'))=cRtX(J x l')=cRlog(m).
It only remains for us to determine c. Take Y=J'+\ so that tX(Y)=

= (log m)' + 1. Then 1- 1 (y) is the set of the elements (zv) of Uoo such that
1 ~Izlw~m for all infinite places w of k. In view of the definition of y, we
have then y(I- 1 (y))=a"b'\ with a, b given by
m
a=2 Jx- 1 dx=210g(m), b= H (xx)-1Idxl\dxl=21tlog(m).

1 ~xx~m
This gives c = 2"(21t)'2, which completes the proof. Our conclusion shows
that R is independent of the choice of the Gj, as had been stated above.
Chapter VI
The theorem of Riemann-Roch
The classical theory of algebraic number-fields, as described above

in Chapter V, rests upon the fact that such fields have a non-empty set
of places, the infinite ones, singled out by intrinsic properties. It would
be possible to develop an analogous theory for A-fields of characteristic
p> 1 by arbitrarily setting apart a finite number of places; this was the
point of view adopted by Dedekind and Weber in the early stages of
the theory. Whichever method is followed, the study of such fields leads
very soon to results which cannot be properly understood without the
use of concepts belonging to algebraic geometry; this lies outside the
scope of this book. The results to be given here should be regarded
chiefly as an illustration for the methods developed above and as an
introduction to a more general theory.
From now on, in this Chapter, k will be an A-field of characteristic
p> 1. In the corollary of th.8, Chap.IV-4, we have defined a finite
field F, which we have called the field of constants of k; this is the algebraic
closure of the prime field in k, and may consequently also be described
as the maximal finite field contained in k; from now on, the number of
its elements will be denoted by q, and F will be identified with Fq • Then~
for every place v of k, the completion kv of k at v contains Fq; in view of
corollary 1 of tho 7, Chap. 1-4, and of corollary 2 of tho 2, Chap. 1-1, this
implies that the module qv of kv is of the form qd, where d is an integer ~ 1
which is called the degree of v and is denoted by deg(v).
By the divisors of k, one understands the elements of the free abelian
group D(k) generated by the places of k; this being written additively,
it consists of the formal sums La(v)· v, where a(v)eZ for every place v
v
of k, and a(v)=O for almost all v. If a = La(v)· v is such a divisor, we
will write 0>-0 when a(v)~O for all v; ifa,b are two divisors, we write
a:>-b for a-b>-O. For every divisor a= La(v)·v, we write deg(a)=
= La(v) deg(v), and call this the degree of a. Clearly a-.deg(a) is a non-
trivial morphism of D(k) into Z; in Chap. VII-5, it will be shown that
it is surjective; its kernel, i.e. the group of the divisors of k of degree 0,
will be denoted by Do (k). Obviously a >- 0 implies deg(o);;?:O, and even
deg(a»O unless a=O, anda>-b implies deg(a)~deg(b).
The theorem of Riemann-Roch 97
Let a = (a v) be any element of k~; for each v, we can write avrv= p~(v)
with a(v) = ordv(av); for almost all v, we have lavl v= 1, hence a(v)=O, so
that La(v)· v is a divisor of k; this divisor will be denoted by div(a).
Clearly a--+div(a) is a surjective morphism of k~ onto D(k), whose kernel
is nr; and is the same as the group denoted by Q(~) in Chap. IV-4;
v
we may therefore use this morphism to identify D(k) with k~ /Q(~). The
definition of lal A shows at once that, if aEk~ and a=div(a), then
lalA =q-deg(a); therefore Do(k) is the image of kl in D(k) under the mor-
phism a--+div(a); in particular, the image P(k) of k x in D(k) under that
morphism is contained in Do(k). The group P(k) is known as the group
of the principal divisors. Clearly the morphism a--+div(a) determines iso-
morphisms of the groups kl!Q(~), kl!k x Q(~) and k~/kx Q(~) onto Do(k),
Do(k)/P(k) and D(k)/P(k), respectively; D(k)/P(k) is known as the group
of the divisor-classes of k, and Do(k)/P(k) as the group of the divisor-
classes of degree 0; tho 7 of Chap. IV-4 shows that the latter is finite, and
that the former is the direct product of the latter and of a group iso-
morphic to Z.
Now we consider vector-spaces over k; we have the following result,
a special case of which occurred already in Chapter IV:
PROPOSITION 1. Let E be a vector-space of finite dimension over k.

Let e be a basis of E over k; for each place v of k, let ev be the rv-module
n
generated by e in E v, and let Lv be any kv-lattice in Ev. Then Lv is an
open and compact subgroup of EA if and only if Lv=ev for almost all v.
If P is a finite set of places such that Lvcev for all v not in P, nLv is
a compact subgroup of EA(P,e), hence of EA ; the converse follows at
once from corollary 1 of prop. 1, Chap. IV-I. Now assume that this is so.
Then nLv is a subgroup of EA ; it is open if and only if it contains a
neighborhood of 0; prop. 1 of Chap. IV-l shows that this is so ifand only
if Lv:::) ev for almost all v, which completes the proof.
With the notations of proposition 1, put L = (4); this will be called a

coherent system of kv-Iattices, or more briefly a coherent system, belonging
n
to E, if 4=e v for almost all v. When that is so, we will write U(L) = 4
and A(L)=EnU(L). By prop. 1, U(L) is open and compact; it is also a
n
module over the open and compact subring r v of kA" As to A(L), it is a
finite subgroup of E, since E is discrete and U(L) compact in EA; it is
also a module over the ring kn(n rv); as this ring, by tho 8 of Chap. IV-4
and its corollary, is the field of constants Fq of k, this shows that A(L)
is a vector-space over F q , whose dimension will be denoted by A(L).
Then A(L) has l·(L) elements.
98 The theorem of Riemann-Roch VI
PROPOSITION 2. Put d = End(E), and let L=(Lv), M=(Mv) be two

coherent systems beloning to E. Then there is a = (a v) in d~ such that
Mv=avLvfor all V; moreover, the divisor div(det(a») is uniquely determined
by L and M.
For each v, by tho 1 of Chap. 11-2, there are bases lXv, Pv of Ev over kv
such that 4, Mv are the rv-modules respectively generated by lXv and by
Pv· Calla v the automorphism of Ev which maps lXv onto Pv; then Mv=a v4·
Put dv=det(a v); if Jl.v is any Haar measure on E v, we have
Idvlv=Jl.v(Mv)/Jl.v(Lv), by corollary 3 of tho 3, Chap. 1-2, so that Idvl v is
independent of the choice of the bases lXv, Pv. Moreover, we have Lv = M v'
hence Idvl v= 1, for almost all V. By prop. 3 of Chap. IV-3, this shows that
a=(a v) is in d~ and d=(dv)=det(a) in k~. As Idvl v depends only upon Lv
and M v , we see that div(d) depends only upon Land M.
We will write M =aL when L, M and a are as in proposition 2.
COROLLAR Y 1. Let e be a basis of E over k; put Lo = (e v), and let L be

any coherent system belonging to E. Then there is aEd~ such that
L=aLo; the divisor b=div(det(a») depends only upon Land e, and its
class and degree depend only upon L.
Only the last assertion needs a proof. Replace e by another basis
e'; put L'o = (e~), and calllX the automorphism of E over k which maps e'
onto e. Then Lo = lx I.:o, hence L = a lx £'0, so that b has to be replaced by
b+div(det(lX)); the second term in the latter sum is a principal divisor, so
that its degree is O.
COROLLARY 2. There is a Haar measure Jl. on EA such that Jl.(n ev)= 1
for every basis e of E over k; for this measure, if Land b are as in corollary 1,
n
we have Jl.(U(L»)=q-~(L) with U(L) = Lv and b(L) = deg(b).
Choose one basis e, and take Jl. such that Jl.(n ev) = 1. If a is as in
n
corollary 1, U(L) is the image of U(Lo) = ev under e-+ae. Therefore
Jl.(U(L») is equal to the module of that automorphism, which is Idet(a)IA
by prop. 3 of Chap. IV-3; in view of our definitions, this is q-6(L), as
stated in our corollary. By corollary 1, this does not depend upon e;
therefore, replacing e by another basis e', we get a measure Jl.' such that
Jl.'(U(L») is the same as Jl.(U(L»); this gives Jl.' = Jl., so that Jl.(n e~) = 1.
As in Chap. IV-2, choose now a non-trivial character X of kA' trivial
on k, and call Xv the character induced by X on kv> which is non-trivial
for every v, by corollary 1 of tho 3, Chap. IV-2. Let E be as above, and
call E' its algebraic dual. As explained in Chap. IV-2, we use X to identify
E~ with the topological dual of EA by means of the isomorphism described
in tho 3 of Chap. IV-2, and, for each v, we use XI) to identify E;, with the
topological dual of Ev by means of the isomorphism described in tho 3 of

Chap. 11-5. Let L=(4) be a coherent system belonging to E; for every v,
call L~ the dual lattice to 4. In view of the identifications which have just
n
been made, L~ is a kv-Iattice in E~, and L~ is the subgroup of E~ associated
~ duality with the subgroup U(L) = n 4 of EA' As U(L) is compact,
n
11 L'v is open; as U(L) is open, L;, is compact; by prop. 1, this shows
that L' =(L;,) is a coherent system belonging to E' (a fact which is also
implied by corollary 3 ofth. 3, Chap. IV-2); we call it the dual system to L.
THEOREM 1. To every A-field k of characteristic p> 1, there is an
integer g ~ 0 with the following property. Let E be any vector-space of
finite dimension n over k; let L be any coherent system belonging to E,
and let L' be the dual system to L. Then:
A(L)=A(L')-(j(L)-n(g-1).
Put U = U(L), U' = U(L'); as we have just seen, U' is the subgroup
of E~ associated by duality with the subgroup U of EA' By definition,
A(L) and A(L') are the dimensions of the vector-spaces A = E n U and
A'=E'nU', respectively, over the field of constants Fq of k. By tho 3 of
Chap. IV-2, the subgroup of EA associated by duality with the subgroup
E of EA is E'. Therefore the subgroup of EA associated by duality with
E + U is A', so that E,J(E + U) is the dual group to A' and has the same
number of elements c/(L') as A'. Clearly E,J(E + U) is isomorphic to
(E,JE)j(E + UjE). Take the Haar measure J.l on EA defined by corollary
2 of prop. 2, and write again p. for its image in E,JE, as explained in
Chap. 11-4. As c/(L') is the index of (E + U)jE in E,JE, we have
J.l(E,JE) = c/(L') J.l(E + U j E).
The canonical morphism of EA onto E,JE maps U onto (E + U)jE, with
the finite kernel A = En U; as A has c/(L) elements, this gives, e. g. by
lemma 2 of Chap. II -4:
J.l(U)=c/(L) J.l(E+ UjE).
Combining these formulas with corollary 2 of prop. 2, which gives
p.(U) = q-6(L), we get:
p.(E,JE) = q)'(L')-)'(L)-cJ(L).
This shows that J.l(E,JE) is of the form q' with reZ. In particular, if we
apply corollary 2 of prop. 2 to E = k and to the basis. B = {1 }, we get a
Haar measure J.ll on kA' such that J.ll(n rv) = 1, and we see that we can
write J.ll(kAjk)=qY with yeZ. Now identify our space E with k" by means
of a basis B of E over k; it is.clear that the measure J.l in EA, defined by
corollary 2 of prop. 2, is the product (J.ll)" of the measures J.ll for the n
factors of the product EA=(k A)", and then that q'=(qY)", i.e. r=yn. This
100 The theorem of Riemann-Roch VI
proves the formula in our theorem, with g = y + 1; it only remains for us

to show that g ~ o. As to this, apply that formula to the case E = k, 4 = r v
for all v. Then A=Fq , A(L) = 1, and clearly £5(L)=O; this gives g=A(I:),
which is ~ 0 by definition.
COROLLARY 1. Let Jl. be the Haar measure in EA defined by corollary 2

of proposition 2; then Jl.(EAiE)=q,,(Y-I). In particular, if Jl.I is the Haar
measure in kA for which Jl.I(O rv)= 1, we have Jl.I (kAik)=qY-I.
This was proved above.
COROLLARY 2. Notations being as in theorem 1, we have EA=E+U
if and only if A(I:) =0.
This is a special case of what has been proved above.
DEFINITION 1. The integer g defined by theorem 1 is called the genus ofk.
The results obtained above will now be made more explicit in the
case E=k. Then a coherent system L=(Lv) is given by taking Lv=p;a(v)
for all v, with a(v)=O for almost all v; such systems are therefore in a
one-to-one correspondence with the divisors of k. Accordingly, if
0= La(v)· v is such a divisor, we will write L(o) for the coherent system
(p;a(v»); L(O) being then the coherent system (rv), we see that L(o) is the
coherent system a-l L(O) when aEkl and o=div(a). For L=L(o), we
will also write U(a), A(a), A(a), £5(a) instead of U(L), A (L), A(L), £5(L);
obviously we have £5(a)= -deg(a). The definition of A(a) shows that it
can be written as n(knp;a(v); in other words, it consists of 0 and of
v
the elements ~ of k X such that ordv(~)~ -a(v) for all v, or, what amounts
to the same, such that div@» - o. As the degree of div@ is 0 for all
~Ekx, this shows that A(o)= {O}, hence A(O)=O, whenever deg(a) <0.
Now let the "basic" character X of kA be chosen as above; for each
place v of k, call v(v) the order of the character Xv induced by X on kv,
this being as defined in def.4 of Chap. 11-5. By corollary 1 of tho 3,
Chap. IV-2, we have v(v)=O for almost all v, so that c= Lv(v)· v is a
divisor of k; we call this the divisor of x, and denote it by div(X). If Xl is
another such character, then, by th.3 of Chap. IV-2, it can be written
as x-+X(~ x) with ~Ekx, and one sees at once that div(xl)=div(X)+div(~).
Thus, when one takes for X all the non-trivial characters of kA> trivial
on k, the divisors div(X) make up a class of divisors modulo the group
P(k) of principal divisors of k. This is known as the canonical class, and
its elements as the canonical divisors.
As before, identify kA with its topological dual by means of X, and
put c=div(x). Using prop. 12 of Chap. II-5, one sees at once that the dual
system to L(o) is L(c-o). Theorem 1 gives now:
THEOREM 2. Let c be a canonical divisor of k. Then, for every divisor 0

of k, we have:
;,(0) = A(C - o)+deg(o) - 9 + 1.
COROLLARY 1. If Cis as above, deg(c)=2g-2 and A(C)=g.
We get the first relation by replacing 0 by c- 0 in theorem 2, and
the second one by taking 0 = 0.
COROLLARY 2. If 0 is a divisor of degree> 2g - 2, A(o)=deg(o) -g + 1.
In fact, we have then deg(c- 0)<0, and, as we have observed above,
this implies A( C- 0) = 0, hence our conclusion, by theorem 2.
COROLLARY 3. Let 0 = La(v)· v be a divisor of degree> 2g - 2. Then
kA =k+(TIp;a(v)).
v
This is the special case E=k, L=L(o) of corollary 2 of tho 1, since
in this case, as shown above, we have L'=L(c-o) and A(L') =0.
Theorem 2 is the "theorem of Riemann-Roch" for a "function-field"
k when the field of constants is finite. A proof for the general case can be
obtained on quite similar lines; for the concept of compacity, one has
to substitute the concept of "linear compacity" for vector-spaces over *
an arbitrary field K, K itself being discretely topologized; instead of a
Haar measure, one has to use a "relative dimension" for compact and
open subspaces of locally linearly compact vector-spaces over K. This
will not he considered here.
Another point of some importance will merely be mentioned. Instead
of identifying the topological dual G of kA with kA by means of a "basic"
character, consider it as a kA-module by writing, for every X*EG and
every aEk A , <x,ax*)=<ax,x*) for all xEk A. Call r the subgroup of G
associated by duality with k. Then tho 3 of Chap. IV-2 can be expressed
as follows: if y is any element of r, other than 0, x ~ x y is an isomorphism
of kA onto G which maps k onto r. In particular, r has an "intrinsic"
structure of vector-space of dimension 1 over k. It is now possible to
define "canonically" a differentiation of k into r, i. e. a mapping x ~ d x
of k into r such that d(xy)=x·dy+ y·dx for all x, y in k, and that r may
thus be identified with the k-module of all formal sums LYidxi, where
the Xi' Yi are in k. This remains true for every separably algebraic extension
of finite degree of any field K(T), where T is an indeterminate over the
groundfield K. Even for the case studied here, that of a finite field of con-
stants, this topic can hardly be dealt with properly except by enlarging
the groundfield to its algebraic closure, and we will not pursue it any
further.
Chapter VII
Zeta-functions of A-fields
§ 1. Convergence of Euler products. From now on, k will be an A-field
of any characteristic, either 0 or p> 1. Notations will be as before; if v
is a place of k, kv is the completion of k at v; if v is a finite place, rv is the
maximal compact subring of kv, and Pv the maximal ideal in r v. Moreover,
in the latter case, we will agree once for all to denote by qv the module
of the field kv and by 7t v a prime element of kv, so that, by tho 6 of Chap. 1-4,
r v/Pv is a field with qv elements, and l7t v lv = q; 1. If k is of characteristic
p> 1, we will denote by q the number of elements of the field of con-
stants of k and identify that field with Fq; then, according to the defini-
tions in Chap. VI, we have qv = teg(v) for every place v.
By an Euler product belonging to k, we will understand any product
of the form
where SEC, OvEC and IOvl::s:; 1 for all v, the product being taken over all
or almost all the finite places of k. The same name is in use for more
general types of products, but these will not occur here. The basic result
on the convergence of such products is the following:
PROPOStTION 1. Let k be any A-field. Then the product

(k(CT) = n(1-q;,,)-1,
v
where CTE R and v runs through all the finite places of k, is convergent for
CT> 1, and tends to the limit 1 for CT tending to + 00.
Assume first that k is of characteristic 0, and call n its degree over Q.

By corollary 1 ofth. 4, Chap. 111-4, there are at most n places v of k above
any given place p of Q; for each of these, kv is a p-field, so that qv is of
the form pV with v ~ 1 and is therefore ~ p. This gives, for CT > 0:
1 < (k(CT)::S:; n(1-p-,,)-n
where the product is taken over all rational primes p. Now write:
(CT) = n(1- p-,,)-1 = n(1 +p-" +p-2" + ...).
§1. Convergence of Euler products 103
Expanding the last product, we get
v=l
since every integer v ~ 1 can be uniquely expressed as a product of

powers of rational primes. Furthermore, we have, for (1 > 1 :
I
+ 00
J
v
l<C«(1)<l+ ~ t- adt=l+ t- adt=1+«(1-1t 1 ,

v-l 1
which shows that C«(1) is convergent for (1) 1 and tends to 1 for (1--+ + 00.
This proves our proposition when k is a number-field.
Now assume that k is of characteristic p> 1; then, by lemma 1 of
Chap. III-2, we may write it as a separably algebraic extension of Fp(n
of finite degree n. By tho 2 of Chap. III-I, Fp(n has one place 00 cor-
responding to the prime element T - 1, while its other places are in a
one-to-one correspondence with the prime polynomials 1t in Fp[T]. It
will clearly be enough if we prove the assertion in our proposition, not
for the product '1«(1), but for the similar product "«(1) taken over the
places v of k which do not lie above the place 00 of Fp(n. Then, just as
in the case of characteristic 0, we see that 1 <"«(1)~'i(1)n, where Cp«(1)
denotes the product
'p«(1) = TI (1- p-de g (n)a)-l = TI (1 + p-de g(7t)a + p-2de g(7t)a + ...)
taken over all the prime polynomials 1t in Fp[TJ. As every monic poly-
nomial in Fp[TJ can be uniquely written as a product of powers of prime
polynomials, this gives
'p(U) = LP-deg(I')a
where the sum is taken over all the monic polynomials Jl. in Fp[T]. As
there are p/J monic polynomials of degree (j for every (j ~ 0, we get
+00
Cp«(1)= L pll(l-a)=(l_ pl-a)- 1,
/J=O
which completes the proof in the present case.

COROLLARY 1. Let P be a finite set of places of k, containing P00; for
every v not in P, take OvEC such that IOvl ~ 1. For SEC, put:
E(s) = TI(1-0 v q;S)-1.
Then E(s) is absolutely convergent, holomorphic in s, and =1= 0, for Re(s) > 1,
and it tends to 1, uniformly with respect to Im(s), for Re(s) tending to + 00.
104 Zeta-functions of A-fields VII
In fact, for cr=Re(s), the series 10gE(s) is majorized by the series

log'k(cr). Our conclusion follows now at once from proposition 1 and
the well known elementary theorems on uniformly convergent series of
holomorphic functions.
COROLLARY 2. Let ko be an A-field contained in k; let M be a set of
finite places of k such that, for almost all veM, the modular degree f(v)
of kv over the closure of ko in kv is > 1. Then the product
p(M,cr)= n(1-Q';'T 1
veM
is absolutely convergent for cr> 1/2.

If k is of characteristic 0, both k and ko are of finite degree over Q;
if it is of characteristic p> 1, and T is any element of ko, not algebraic
over the prime field Fp, k and ko are of finite degree over Fp(T); in both
cases, k has a finite degree n over k o. Let v be a finite place of k, and u the
place of ko lying below v; then the closure of ko in kv is (ko)u, and kv is
generated over it by k; therefore the degree of kv over (ko)u is ~ n, so that
1 ~f(v)~n. This shows that M is the union of the sets M 1 , ••• , M",
consisting respectively of the places veM for which f(v) = J, with
1 ~f~ n. Our assumption about M means that M 1 is finite, so that it
is enough to prove our assertion for each one of the sets M I with f?;. 2.
By corollary 1 ofth. 4, Chap. 111-4, there are at most n/fplaces veMI over
each finite place u of ko. Therefore the product p(M I'cr) is majorized by
'ko(fcr)"II; by proposition 1, this is absolutely convergent for cr> 1/f
COROLLARY 3. Let M be as in corollary 2; for every veM, take
evec such that levi ~ 1; then the product
n(1- e Q,;·)-1
veM
v
is absolutely convergent, holomorphic in s, and =1= 0, for Re(s) > 1/2.

In view of corollary 2, the proof is similar to that of corollary 1.
§ 2. Fourier transforms and standard functions. The theory of zeta-
functions depends essentially on the concept of Fourier transforms, ap-
plied to the groups kv, kA attached to an A-field k. We begin by recalling
the results to be used here.
As in Chap. 11-5, let G be a commutative locally compact group, G*
its dual, and let (g,g*) be as defined there. Let tP be a continuous function
on G, integrable for a Haar measure a given on G. Then the function tP*
defined on G* by
tP*(g*) = JtP(g) (g,g*) da(g)
§ 2. Fourier transforms and standard functions 105
is called the Fourier transform of ~ with respect to ex; one verifies at once
that it is continuous on G*. Clearly, if one replaces ex by cex, with cER~,
this replaces ~* by c ~* .
LEMMA 1. Let g-+Ag be an automorphism of G, with the module

modG(A). Let g*-+g* A* be the automorphism of G* such that <Ag,g*) =
= <g,g* A*) for all gEG, g*EG*. Then, if ~* is the Fourier transform of~,
that of g-+~(A -1 g) is g*-+modG(A)~*(g* A*).
In the integral which defines the Fourier transform of ~(A -1 g),

substitute Ag for g; the conclusion follows at once.
By the theory of Fourier transforms, there is a Haar measure ex* on
G*, such that, whenever the function ~* defined as above is integrable
on G*, ~ is given by "Fourier's inversion formula"
~(g)= J.~*(g*)< -g,g*)dex*(g*).
Then we say that ~ is the inverse Fourier transform of ~*. The measure
ex* is called the dual measure to ex. Clearly, for cER~, the dual measure
to cex is c- 1 ex*. In particular, assume that G* has been identified with G
by means of some isomorphism of G onto G*; then ex*=mex with some
mER~, and, as the dual of cex is c- 1 mex, there is one and only one Haar
measure on G, viz., ml/2 ex, which coincides with its own dual for the given
identification of G and G*; this is then called the self-dual Hoor measure
onG.
If G is compact, G* is discrete. Then, by taking ~ = 1, one sees at
once that the dual of the Haar measure ex given by ex(G) = 1 on G is the
one given by ex*({O})= 1 on G*.
A function ~ on G will be called admissible for G if it is continuous,
integrable, and if its Fourier transform ~* is integrable on G*. Now let
r be a discrete subgroup of G, such that G/r is compact. Let r* be the
subgroup of G* associated by duality with r; as G/r is compact, r* is
discrete; as r is discrete, G*/r* is compact. Take for ex the Haar measure
on G determined by ex (G/r) = 1. The function ~ on G will be called
admissible for (G, r) if it is admissible for G and if the two series
L ~(g + y), L ~*(g* + y*)
yeT y·eT.
are absolutely convergent, uniformly on each compact subset with

respect to the parameters g, g*. The first one of these series defines then a
continuous function F on G, constant on cosets modulo r; this may be
regarded in an obvious manner as a function on G/r. As r* is the dual
group to G/r, F has then the Fourier transform
y*-+ f (L tP(g + y») (g,y*)d(X(g),

Glf yef
where, as usual, g is the image of gin Gjr under the canonical homo-
morphism of G onto Gjr, and the integrand, which is written as a func-
tion of 9 but is constant on cosets modulo r, is regarded as a function
of g. According to formula (6) of Chap. 11-4, this integral has then the
value tP*(y*), so that the Fourier transform of F, when F is regarded as
a function on Gjr, is the function induced by tP* on r*. Since tP has been
assumed to be admissible for (G,r), this is integrable on r*, so that we
get, by Fourier's inversion formula for Gjr and r*:
F(g)= L tP(g+y)= L tP*(y*) ( -g,y*).
yef y*ef.
For g=O, this gives:

(1) LtP(y)= L tP*(y*).
yeT y·ef.
This is known as Poisson's summation formula, which is thus shown to

be valid whenever tP is admissible for (G,r), and (X is such that (X (Gjr) = 1.
Assume that there are admissible functions tP for (G,r) for which both
sides of (1) are not 0; this assumption (an easy consequence of the general
theory of Fourier transforms) will be verified by an explicit construction
in the only case in which we are interested, viz., the case G=EA , r=E
when E is a vector-space of finite dimension over an A-field. Call then (X*
the dual measure to (X; put (X*(G*jr*)=c, and interchange the roles of G,
G* in the above calculation, starting wifih tP* and taking its inverse
Fourier transform by means of the Haar measure c - 1 (X* on G*; as this
is c- 1 tP, we find as end-result the same formula as (1), except that tP has
been replaced by c-1tP. A comparison with (1) gives now c=1. This
shows that the Haar measures (x, (X* given on G, G* by (X (Gjr) = 1,
(X*(G*jr*) = 1 are dual to each other. In particular, ifthere is an isomor-
phism of G onto G* which maps r onto r*, and this is used to identify G
and G*, the self-dual measure on G is the one given by (X (Gjr) = 1.
Now we construct special types of admissible functions for the groups
in which we are interested; these will be called "standard functions".
On any space, a function is called locally constant if every point has a
neighborhood where the function is constant. If f is such, f-l({a}) is
open for every a; it is also closed, since its complement is the union of
the open sets f-l({b}) for b=foa. In a connected space, e.g. any vector-
space over R, only the constant functions are locally constant.
DEFINITION 1. Let E be a vector-space of finite dimension over a

p-field K. By a standard junction on E, we understand a complex-valued
locally constant function with compact support on E.
It will be enough to consider the case when K is commutative. Let E*

be the "topological dual" of E, i.e. the dual of E when E is regarded as a
locally compact group. On E*, we put a structure of vector-space over K
in the manner described in Chap. 11-5; as we proved there, E* has the
same dimension as E over K. With these notations, we have:
PROPOSITION 2. A function 11> on E is standard if and only if there are

K -lattices L, M in E such that L;:) M and that 11> is 0 outside L and constant
on cosets modulo M in L. Then, if L* and M * are the dual K -lattices to
Land M, we have M*;:) L*, and the Fourier transform 11>* of 11> is 0 outside
M * and constant on co sets modulo L* in M *.
If 11>, L, M have the properties stated in our proposition, it is clear

that 11> is standard. Conversely, assume that it is such. Take a K-norm N
on E, and call J.l an upper bound for N on the support of 11>; then, as we
have seen in Chap. 11-2, the set L defined by N(e)~J.l is a K-lattice, and it
contains the support of 11>. As the sets 11> - 1 ({ a}), for aE C, are open, and
L is compact, L is contained in the union of finitely many such sets; in
other words, 11> takes only finitely many distinct values al, ... , an on L.
Take e > 0 such that la; - a) > ewhenever i =1= j. As 11> is uniformly continuous
on L, there is £5>0 such that N(e-e')~£5, for e and e' in L, implies
II1>(e)-I1>(e')I~e. Then the set M defined by N(e)~£5 is a K-Iattice, con-
tained in L if we have taken £5~J.l, and 11> is constant on cosets modulo
Min L. Now consider the Fourier transform
11>*(e*)= J11> (e) <e,e*) dlX(e),
E
where IX is any Haar measure on E. As 11> is 0 outside L, this integral is

not changed by taking it over L. Replace e* by e* + eT with eT E L*; by
definition, the latter assumption means that <e, et) = 1 for all eEL, so
that the integral is not changed; therefore 11>* is constant on cosets
modulo L* in E*. On the other hand, as M is an open subgroup of the
compact group L, L is the union of finitely many cosets e; + M. As 11> is
constant on each one of these, we have
(2) Ll1>(e;) J<e; + e,e*) d lX(e) = Ll1>(e;) <e;,e*) J<e,e*) dlX(e).
11>* (e*) =
M M
As the last integral is clearly 0 unless the character e --+ <e, e*) is trivial
on M, i.e. unless e*EM*, we see that 11>* is 0 outside M*.
COROLLARY 1. If 11> is the characteristic function of the K-lattice Lin

E, IX(L)-l 11>* is the characteristic function of the K-lattice L* dual to
Lin E*, and IX*(L*)=IX(L)-l if IX* is the dual measure to IX.
The first assertion follows at once from (2) for L = M, IP(O) = 1. It

implies that the inverse Fourier transform of IP* is IX*(L*)IX(L) IP; as this
must be IP, we get the last assertion.
COROLLARY 2. Every standard function on E is admissible for E.
This is an obvious consequence of proposition 2 and the definitions.
In the next corollary, we identify K with its topological dual by means
of a character X of K in the manner explained in Chap. 11-5, i.e. by
writing (x,Y)=X(xy) for x, y in K; for this identification, we may then
speak of a self-dual measure on K.
COROLLARY 3. Let R be the maximal compact subring of K, and cp the
characteristic function of R. Let X be a non-trivial character of K, of order
v, and let IX be the self-dual Haar measure on K for the identification of K
with its dual, based on X. Let aEK x be such that ordK(a)=v. Then
IX(R)= mod K(a)1/2, and the Fourier transform of cp is y-+mod K(a)1/2 cp(ay).
Apply corollary 1 to E=K, L=R; then, by prop. 12 of Chap. 11-5,
L*=P-', i.e. L*=a- 1R if a is as defined above; then the characteristic
function of L* is cp (a y), and we have IX(L*)=mod K(a)-l lX (R). Our
assertions follow now at once from corollary 1.
DEFINITION 2. Let E be a vector-space of finite dimension over R.

By a standard function on E, we understand any function of the form
e-+p(e)exp( -q(e)), where p is a complex-valued polynomial function on
E and q a real-valued positive-definite quadratic form on E.
PROPOSITION 3. Let E be as in definition 2; then every standard function

on E has a Fourier transform which is a standard function, and is admissible
for (E,L) if Lis any R-lattice in E.
Choose a basis for E over R, such that, when E is identified with Rn

by means of that basis, the quadratic form q is given by q(x)=nLx;.
It is clearly enough to prove our first assertion for a function
M(x)exp( -q(x)), where M(x) is a monomial in the x •. Byth. 3 of Chap. 11-5,
we may identify R n with its dual by putting (x,y) =e(Lx.y.); then we
see that it is enough to deal with the case n = 1, i. e. to show that the
Fourier transform of xmexp( -nx 2 ) is standard on R for every integer
m~O. The Fourier transform of exp( -nx 2) is exp( _ny2), as shown by
the well-known formula
J
exp( _ny2)= exp( -nx 2+ 2nixy)dx.
Differentiating both sides m times with respect to y, one sees at once,
by induction on m, that the left-hand side is of the form Pm(}')exp( _ny2),
where Pm is a polynomial of degree m, and that the differentiation may

be carried out inside the integral in the right-hand side. This gives
Pm(y)exp( -ny2) = f(2nixtexp( -nx 2 + 2nixy)dx,
which proves our first assertion. Now let L be an R-Iattice in E. By
prop. 11 of Chap. 11-4, there is a basis of E over R which generates the
group L; in other words, identifying E with Rn by means of that basis,
we may assume that E = Rn and L= Z". In order to prove that standard
functions in Rn are admissible for (R",zn), it is now enough to show that,
if /f> is such a function, L:I
/f>(x + v)l, taken over all VE zn, is uniformly
convergent on every compact subset C of Rn. Put /f>(x) = p(x)exp( -q(x))
and r(x) = LX;;take f> > 0 such that the quadratic form q - f> r is positive-
definite; this will be so provided f> < J.l, if we call J.l the lower bound of q
on the sphere r= 1. Then the function y-+/f>(y) exp(f>r(y-x)) tends to 0,
uniformly in x for XEC, when r(y) tends to + 00. This implies that this
function is bounded for XEC and all YER", and therefore, replacing y
by x + u, that, for a suitable A> 0, we have
1/f>(x+u)1 ~Aexp( -f>r(u))
for all XEC. This gives
~1/f>(x+v)1 ~A ~exp( -f>~vr)=A ex", exp( -f>v2)) n,
which completes our proof.

We will also need a more explicit statement for some special cases
of prop. 3, corresponding to E = R or C; in each case we choose a "basic"
character X, and identify R (resp. C) with its topological dual by means
of that character, just as we have done above for p-fields, according to
Chap. 11-5. The self-dual measures to be considered now are taken with
reference to that identification.
PROPOSITION 4. On R, the self-dual Haar measure, with reference to
the basic character x(x)=e( -ax) with aE RX, is doc(x)=lal l / 2dx. If
(fJA(x)=xAexp( -nx 2 ) with A=O or 1, the Fourier transform of (fJA is
(fJ~(y) = i- Alal l / 2 (fJA(a y).
Put doc(x) = c . dx with CE R~; then (fJ~ is given by

(fJ~(y)= c Sexp( -nx 2 - 2niax y)dx.
R
As recalled above, this is equal to c(fJo{ay). Applying now Fourier's

inversion formula and lemma 1, we get c = lal l / 2 • Differentiating both
sides of the above formula for (fJ~(y) with respect to y, we get the Fourier
transform of (fJl'
PROPOSITION 5. On C, the self-dual H aar measure, with reference

to the basic character x(x)=e( -ax-ax), with aeC x , is dcx(x)=
=(a2i)1/2Idx 1\ dxl. If qJA(x)=xAexp( -2nxx), A being any integer ~O,
the Fourier transform of qJA is i-A(a2i)1/2 qJA(ay), and that of qJA is
i-'A(a2i)1/2qJA(ay).
The proof of the assertions about cx and about the Fourier transform
of 11'0 is quite similar to that in prop. 4. Differentiating A times, with
respect to y, the formula for the Fourier transform of 11'0' we get that
of qJA; that of qJA follows from this at once.
DEFINITION 3. Let E be a vector-space of finite dimension over an

A-field k. Let Il be a basis of E over k; for each finite place v of k, let Ilv
be the rv-module generated by Il in Ev. By a standard function on EA, we
understand any function of the form
e=(ev)-+IP(e) = nIPv(ev)
v
where IP v is, for every place v of k, a standard function on Ev, and, for
almost every v, the characteristic function of Ilv.
Corollary 1 of tho 3, Chap. 111-1, shows that the latter condition is

independent of the choice of Il. The formula which defines IP, for which
n
we will write more briefly IP= IPv, is justified by prop. 1 of Chap. IV-1,
which shows that almost all the factors in the right-hand side are equal
to 1 whenever e is in EA ; the same proposition shows also that IP is 0
outside EA(P,Il) for a suitable P, and that it is continuous.
Just as in the case of kA in Chap. V-4, a Haar measure on EA can
be defined by choosing a Haar measure CX v on Ev for each v, so that
CXv(Il,,)= 1 for almost all v; when the CX v satisfy the latter condition, we
will say that they are coherent. Then there is a unique measure cx on EA
which coincides with the product measure nCX v on everyone of the
n
open subgroups EA(P,Il) of EA; this will be written as cx= CX v. Clearly,
if a Haar measure cx is given on EM one can find coherent measures CX v
n
such that cx = CX v by choosing any set of coherent measures on the
spaces Ev and suitably modifying one of them.
From now on, we also choose, once for all, a "basic" character X
of kA' i.e. a non-trivial character of kA' trivial on k; we denote by Xv
the character induced by X on k." which is non-trivial by corollary 1 of
th.3, Chap.IV-2. If E is any vector-space of finite dimension over k,
we call E' its algebraic dual, and we use X and Xv for identifying the
topological dual of EA with E~, and that of Ev with E~ for each v, in the
manner described in Chap. IV-2, i.e. by applying tho 3 of Chap. IV-2 to
the former space and tho 3 of Chap. 11-5 to the latter.
§ 2. Fourier transforms and standard functions III
THEOREM 1. Let E be a vector-space of finite dimension over the

A-field k. Let the (Xv be coherent Haar measures on the spaces Ev, and
n
let ~= ~v be a standard function on EA' Then the Fourier transform
n
of ~, with respect to the measure (X = (Xv on EA , is a standard function
n
on E~, given by ~' = ~~, where ~~, for every v, is the Fourier transform
of ~v with respect to (Xv' Moreover, ~ is admissible for (EA,E).
Let ll,ll' be bases for E and for E' over k; for each finite place v, let
llv be as before, and let e~ be similarly defined for E~. By corollary 3 of
th.3, Chap. IV-2, there is a finite set P of places of k, containing Poo ,
such that ll~ is the dual kv-Iattice to llv when v is not in P; in view of our
assumption on the (Xv, we may also assume that P has been so chosen
that (Xv(llv) = 1 for v not in P. Then, by corollary 1 of prop. 2, the Fourier
transform of the characteristic function of llv is the characteristic function
of ll~, and the dual measure (X~ to (Xv is given by (X~(e~)= 1, for all v not
n
in P. Now let ~= ~v be a standard function on EA ; for each v, call ~~
the Fourier transform of ~ v with respect to (Xv' From what has just been
said, and from propositions 2 and 3, it follows that ~' = ~~ is a n
standard function on E~; we will show that it is the Fourier transform
of ~. Replacing P if necessary by some larger set, we may assume that ~ v
is the characteristic function of llv for v not in P; in particular, the support
of ~ is contained in EA(P,Il), so that the Fourier transform ~" of ~ is
given by the integral
J
~"(e')= ~(e)x([e,e'])d(X(e)
taken on EA(P,Il). In view of our definitions, the integrand here, for

e=(e v), e' =(e~), is given by
lP(e) x([ e,e']) = n(IPv(ev)Xv([ev, e~]));

v
moreover, when e' is given, the factor in the right-hand side corresponding
to v has, for almost all v, the constant value 1 on llv' In view of the defini-
tion of EA(P,Il) in prop. 1 of Chap. IV-I, this implies that ~"(e') is the
same as ~'(e').
Now, in order to prove that ~ is admissible for (EA,E), it is enough
to show that, for each compact subset C of EA , the series
(3)
'leE
L Inv ~v(ev+")1
L 1~(e+")1 = 'leE
is uniformly convergent for eEC. By corollary 1 of prop. 1, Chap. IV-I,
C is contained in some set EA(P,Il); take P such that this is so and that
EA(P,Il) also contains the support of ~. For each place v in P, call Cv
the projection of C onto Ev; for each finite place VEP, call C~ the support
n
of cP,,; for v not in P, put C"=C~=B,,. As C" is compact and contains C,
n
it will be enough if we prove our assertion for C = C". Assume first
that k is of characteristic p> 1. Then cP is 0 outside the compact set
n
C' = C~, so that all the terms of (3) are 0 for eeC except those corres-
ponding to "eEnC", where C" is the image of C x C' under the mapping
(e,e')-e' -e; as C" is compact, EnC" is finite, and the assertion becomes
obvious. Now let k be of characteristic O. For each finite veP, take a
k,,-norm N" on E", and call L" the k,,-lattice given by N,,(e,,)~JI., where JI.
is an upper bound for the values of N" on the compact set C"uC~. For v
n
not in P, put L"=B,,. Put L= (EnL,,), where v runs through all the
finite places of k; by tho 2 of Chap. V-2, this is the k-Iattice in E with the
closure L" in E" for all finite v. Clearly, for e=(e,,) in C, cP,,(e,,+,,) is 0
unless" is in EnL", so that cP(e +,,) is 0 unless" is in L. Furthermore,
for almost all v; putting A

majorized by the series
=n
if A" is the upper bound of IcP"I for each finite place v of k, we have A" = 1
A", we see now that (3), for ee C, is
A Lin cPw(ew+,,)1
"eL w
where the product is taken over the infinite places w of k. As explained

in Chap. V-2, put Eoo=E®QR, and identify this with the product nEw
taken over the infinite places of k. It is then obvious that the function cP00
on Eoo ' defined for e oo = (e w ) by
cPoo(e oo )= ncPw(ew),
w
is standard. As L is a k-Iattice in E, it is a Q-Iattice in E regarded as a

vector-space over Q, hence an R-Iattice in Eoo. Our assertion is now
contained in prop. 3.
COROLLARY 1. If the ex" are coherent measures on the spaces E",
n n
their duals ex~ are coherent; the dual of ex = ex" is ex' = ex~; if ex(EAiE) =
= 1, then ex'(E;JE') = 1.
The first assertion has been proved above; the second one follows
at once from theorem 1 and the definitions. As to the last one, we know,
by tho 3 of Chap. IV-2, that E' is the subgroup of EA associated by duality
with the subgroup E of EA; therefore, as we have seen, our assertion
follows from Poisson's formula provided we can exhibit a function cP,
admissible for (EA,E), for which the left-hand side of (1) is not 0; by
theorem 1, any standard function CP~O such that cP(O»O has these
properties.
An important special case is that in which E=E'=k, [x,y]=xy;
then we identify kA and k" with their topological duals by means of x,x",
as explained before, and we have:
COROLLARY 2. Let a,a v be the self-dual measures on kA,kv' Then the

n
av are coherent, a= av, and a(kJ.ik) = 1.
Take any coherent measures {3v on the groups kv; by corollary 1,
their duals {3~ are coherent, which implies that {3v = {3~ for almost all v;
in other words, /3v coincides with the self-dual measure av for almost
all v; this implies that the av are coherent. Our other assertions follow
now at once from corollary 1.
Notations being as in corollary 1, the measure a on EA for which
a(EJ.iE) = 1 is known as the Tamagawa measure on EA; corollary 1 shows
that its dual is the Tamagawa measure on E~. In particular, on kA' the
Tamagawa measure and the self-dual measure are the same.
Now, for each finite place v of k, call v(v) the order of X", which is
for almost all v by corollary 1 of tho 3, Chap. IV-2, and choose avek~
°
such that ordv(av)=v(v). On the other hand, for each real place v of k,
apply to x-+e( -x) the corollary of tho 3, Chap. 11-5; it shows that there
is one and only one avek~ such that Xv(x)=e( -avx) for all xek v.
Similarly, for each imaginary place v, there is one and only one avek~
such that Xv(x)=e( -avx-avx) for all xek v. As v(v)=O for almost all v,
(a v) is in k~.
DEFINITION 4. Let X be a non-trivial character of kA' trivial on k,

inducing Xv on kJor every V. An idele a=(a v) of k will be called a difJerental
idele attached to X ~f ordv(av) is equal to the order v(v) of x" for every
finite place v of k, Xv(x)=e( -avx) for every real place v, and Xv(x)=
=e( -avx-avx) for every imaginary place v of k.
Clearly, when Xis given, the differental idele a is uniquely determined
modulo n r~ , the latter product being taken over all the finite places v
of k. If Xl is another character such as X, then, by tho 3 of Chap. IV-2,
it can be written as Xl(X)=X(ex) with eekx; if a is as above, ea is then
a differental idele attached to Xl ' Consequently, the set of all differental
n
ideles is a coset modulo k x r~ in k~. If k is of characteristic p> 1,
a is a differental idele attached to X if and only if div(a) = div(X), in the
sense explained in Chap. VI; this implies that div(a) belongs to the
canonical class.
PROPOSITION 6. Let a be a difJerental idele. Then, if k is of charac-

teristic 0, laIA=IDr l , where D is the discriminant of k; if k is of charac-
teristic p> 1, and ifFq is its field of constants and g its genus, lalA =q2-2 u.
The latter statement is equivalent to deg(div(a»)=2g-2; as div(a)

is a canonical divisor, this is corollary 1 of tho 2, Chap. VI. In the case
of characteristic 0, let a, av be the self-dual measures in kA,kv, so that
(X = n n
(Xv by corollary 2 of tho 1; let P= Pv be as in prop. 7 of Chap. V-4.
Applying corollary 3 of prop. 2, and propositions 4 and 5, we get
(X = lal1'2 p. As (X(k.Jk) = 1 by corollary 2 of tho 1, and P(k.Jk) = ID11/2 by
prop. 7 of Chap. V-4, we get lalA = IDI- 1.
§ 3. Quasicharacters. We first insert here some auxiliary results. As

before, if zeC, we denote by Re(z), Im(z) its real and imaginary parts,
and we put Izl =(ZZ)1/2, Izlc., =moddz)=zz.
LEMMA 2. A character w of a group G is trivial if Re(w(g))>O for
all geG.
If zeC, Izl=l, z=Fl and Re(z»O, we can write z=e(t) with teR,
0<ltl<I/4. Call n the smallest integer such that nltl>I/4; then
(n -1 )Itl ~ 1/4, hence 1/4 < nltl < 1/2 and Re(z") < O. Therefore the subset
ofC determined by Izl = 1, Re(z) > 0 contains no subgroup ofC x except {I}.
LEMMA 3. Every homomorphism w of a compact group G into C x is a
character of G.
In fact, g-+lw(g)1 must map G onto a compact subgroup of R~, and
there is none except {I}.
A group G is called totally disconnected if there is a fundamental
system of neighborhoods of the neutral element in G, consisting of sub-
groups of G. For instance, irK is a p-field, with the maximal compact
subring R, and the maximal ideal P in R, the groups K and K x are
totally disconnected, since the subgroups pn in K, and the subgroups
1 + P" in K x , for n ~ 1, make up such fundamental systems.
LEMMA 4. Let the group G be locally compact and totally disconnected;
then every representation of G into C x is locally constant. If G is compact,
every such representation is a character of G of finite order. Conversely,
if G is a compact commutative group, and if every character of G is of
finite order, G is totally disconnected.
If G is locally compact and totally disconnected, lemmas 2 and 3
show that every representation of G into C x is trivial on some open
subgroup of G, hence locally constant. If G is compact, any open sub-
group of G is of finite index, hence the second assertion. If G is commuta-
tive and compact, its dual G* is discrete. As G may be identified with the
dual of G*, there is then a fundamental system of neighborhoods of 0
in G, consisting of sets defined by conditions of the form IWj(g)-11 ~8
(1 ~ i ~ N), where the Wj are characters of G. If all the Wj are of finite
order, we can take 8 such that these inequalities imply Wj(g) = 1 for
1 ~i~N; then the neighborhood which is so defined is a subgroup of G.
§ 3. Quasicharacters 115
From now on, we shall be chiefly concerned with representations

into C x of groups of the form K X. where K is a local field, and k ~/k x ,
where k is an A-field. All these groups have the property stated in the
following definition:
DEFINITION 5. A group G will be called quasicompact if it is the direct
product of a compact commutative group G1 and of a group isomorphic
to R or to Z; a representation of G into C x will then be called a quasi-
character of G.
It would be easy to show that a group G is quasicompact if and only
if it is commutative and locally compact, and its dual G* is locally
isomorphic to R, i. e. if it has an open subgroup isomorphic to R or to
R/Z; the latter condition may even be replaced by the weaker requirement
that G* should have a neighborhood of 0, homeomorphic to R. From
this, one concludes easily that G is quasicompact if and only if it has a
compact subgroup G1 such that G/G 1 is isomorphic to R or Z. These
facts will not be needed in the sequel. It is clear that, if G has the property
described in definition 5, G1 is its unique maximal compact subgroup.
DEFINITION 6. If G is a quasicompact group, a quasicharacter of G
will be called principal if it is trivial on the maximal compact subgroup
G1 0fG.
The quasicharacters of a quasicompact group G make up a group
in an obvious manner; this will be denoted by O(G) and written multi-
plicatively. In other words, if w, w' are in O(G), we write row' for the
quasicharacter g-+w(g)w'(g) of G. Clearly the principal quasicharacters
of G make up a subgroup 0 1 of O(G).
PROPOSITION 7. Let G be a quasicompact group and G 1 its maximal

compact subgroup. Then G has non-trivial representations into R~; if w 1
is such a representation, its kernel is Gl' and every representation of G
into R ~ can be written in one and only one way in the form g -+ W 1(g)"
with UER.
Put G = G1 X N, with N isomorphic to R or Z. By lemma 3, every

representation w of G into R ~ must be trivial on G1; writing elements
of Gas (g1,n) with g1 eG 1 , neN, we see that w must then be ofthe form
(g1,n)-+q>(n), where q> is a representation of N into R~. Identify N with
R or with Z, as the case may be. In the former case, the condition for q>
amounts to saying that n-+logq>(n) is an endomorphism ofR, hence of
the form n-+an with aeR, so that q>(n) = exp(an). For N=Z, q> is
obviously of the form q>(n) = b" with be R~ and may still be written as
q>(n)=exp(an) with a=logb. In both cases, q> is non-trivial if a=l=O.
Therefore, if Wl is as in our proposition, it can be written as W1 (g 1, n) =

=exp(a 1 n) with a1 =1= O. This has obviously the kernel G1 ; moreover, if w,
qJ and a are as above, we have w=(w 1 ) .. with (1= ala 1 , and (1 is uniquely
determined by this.
COROLLARY 1. Let G, G 1 and W1 be as in proposition 7. Then the

group Q1 of the principal quasicharacters of G is isomorphic to C or to C x
according as GIG 1 is isomorphic to R or to Z; every such quasicharacter is
of the form
g-+W.(g)=W 1 (g'f
with SEC; and s-+w. is a morphism of C onto Ql' whose kernel is {O} or
of the form i a Z with aE R~, according as GIG 1 is isomorphic to R or to Z.
Let W be any quasicharacter of G; with the above notation, proposi-
tion 7, applied to g -+ Iw(g)l, shows that Iwl = w.. with (1E R; then w' = w; 1 W
is a character of G. If w is trivial on G l' so is w'; with the same notations
as in the proof of prop. 7, we may then write w'(gl,n)=I/1(n), where 1/1
is a character of N. As in that proof, identify N with R or with Z, as the
case may be, WI being given by w 1 (gt>n)=exp(a 1 n) in both cases. Every
character of N can be written as l/1(n)=e(1:n) with 1:ER; this is obvious
for N = Z and is well known (and a special case of tho 3, Chap. 11-5) for
N=R. That being so, we get W=W., with s=(1+2nirlal' Moreover, (1
and 1/1 are uniquely determined by w; 1: is uniquely determined by 1/1 if
N = R, and uniquely determined modulo Z if N = Z. This shows that
s -+ w. is an isomorphism of C onto a 1 if N = R; if N = Z, we have
w(gl,n)=u" with u=exp(als), and u-+w is an isomorphism ofC x onto
Q l ' This completes the proof.
COROLLARY 2. Let G be a quasicompact group, the direct product of

the compact group Gland of a group N, isomorphic to R or Z. Then the
group Q(G) of quasicharacters of G is the direct product of the group Q1
considered in corollary 1, and of the group of the characters of G, trivial
on N; the latter is isomorphic to the dual of G1 •
We have already noted above that every quasicharacter w of G can

be uniquely written as w.. l/1, where 1/1 is a character of G, and (1ER.
Clearly 1/1 can be uniquely written as 1/111/12' with 1/11 trivial on Gland
1/12 trivial on N; then W= (W .. 1/11)1/12' and W.. 1/11 is in Ql' The last assertion
in our corollary is obvious.
So far we have refrained from mentioning any topology on Q(G). We
will put on ai' not only the topology, but also the complex structure
determined by the morphism s -+ w. of C onto Q 1 defined in corollary 1
§ 3. Quasicharacters 117
of proposition 7; we define the topology on .o(G) by prescribing that .0 1

shall be an open subgroup of .o(G), and we define a complex structure
on .o(G) by putting, on every coset modulo .0 1 in .o(G), the complex
structure deduced from that of .0 1 by translation. Then .o(G)/.o 1 is
discrete, hence isomorphic to the dual of G1 also as a topological group
since that dual is discrete. The connected components of .o(G) are the
cosets modulo .0 1; they are all isomorphic to e or to ex, as the case
maybe.
Clearly the above concepts and results can be applied to G = K x if
K is any local field, with w 1 (x)=mod K (x); we can take for N the sub-
group R~ of K x if K is R or e, and the group generated by any prime
element n of Kif K is a p-field. In the latter case, this gives:
PROPOSITION 8. Let K be a p-field and n a prime element of K. Then

the principal quasicharacters of K X are those of the form x--+modK(x)S
with see; the group .o(K x ) of quasicharacters of K x is the direct product
of the group of principal quasicharacters and of the group of the characters
I/! of K x such that I/!(n)= 1.
By lemma 4, every quasicharacter of K x is locally constant. If R

and P have their usual meaning, the groups R x and 1+ pn for n ~ 1 are
open in K x and make up a fundamental system of neighborhoods of 1.
This justifies the following definition:
DEFINITION 7. Let K be a p-ji.eld, R its maximal compact subring and

p the maximal ideal of R. Let w be a quasicharacter of K x ; let f be the
smallest integer ~O such that w(x) = 1 for xeR X , x-lepl. Then pI
is called the conductor of w.
Obviously w is principal if and only if /=0, i.e. if and only if its
conductor is R; when that is so, we will also say that w is unramified.
For K = R or e, we have the following result:
PROPOSITION9. Every quasicharacter of RX can be written in one and

only one way as x--+x-Alxl s with A=O or 1, and seC. Every quasicharacter
of ex can be written in one and only one way as x--+X-AX-B(XX)S, where
A and B are integers, inf(A,B) =0, and seC.
For R X , this is an immediate consequence of prop. 7 and its corollaries,

since here G1 = { ± 1}. For G = C X, G1 is the group determined by
xx=l; as this is the dual of Z, its characters are the functions x--+xn
with neZ; this can be written as x--+(X/lxl)-A with A= -n~O if n~O,
and as x--+(X/lxl)-B with B=n~O ifn~O. Our assertions follow at once
from this and prop. 7.
§ 4. Quasicharacters of A-fields. By tho 6 of Chap. IV-4, if k is an

A-field, k~/kx is quasicompact. From now on, we will write Gk=k~/kx ;
this group is known as the "idele-class group" of k. We write Q(G k ) for
the group of quasicharacters of Gk , provided with its topology and its
complex structure according to our definitions in § 3. The quasicharacters
of Gk will be identified in an obvious manner with the representations of
k~ into ex, trivial on kX.
As z--tlzl A is a non-trivial representation of k~ into R~, trivial
on e, it determines a non-trivial representation of Gk into R~, which
will be denoted by WI' and to which we can apply prop. 7 of § 3 and its
corollaries, writing again w s =(w 1)" for seC. In particular, the kernel
G~=kl/e of WI is the maximal compact subgroup of C k ; s--tws is a
morphism of e onto the group Q 1 of principal quasicharacters of Gk ;
if W is any quasicharacter of Gk , there is one and only one (Fe R such
that Iwi = wa •
If k is of characteristic 0, corollary 2 of tho 5, Chap. IV-4, shows that
Gk is the direct product of G~ and of the image N in Gk of the group M
defined in that corollary. On the other hand, if k is of characteristic
p> 1, we choose an element Zl of k~ among those for which IzlA has its
smallest value Q> 1; as we have seen in Chap. VI that the values of IzlA
are all of the form q" with neZ if Fq is the field of constants of k, we have
Q= qV with v ~ 1; it will be seen later that v = 1, Q= q (this is corollary 6
of tho 2, § 5). Then we call M the subgroup of k~ generated by Zl' and
N its image in Gk • In all cases, we will identify N with its image in R ~
under WI' so that WI may be regarded as the projection from the product
Gk = G: x N onto the factor N. Thus N = R ~ if k is of characteristic 0;
otherwise it is the subgroup of R~ generated by Q; this implies that in
the latter case the morphism s --t Ws of e onto Q 1 has the same kernel as
the morphism s --t QS of e onto ex, i. e. 2n i(log Q) - 1 Z.
Let W be any quasicharacter of Gk ; regarding it as a representation
of k~ into ex, trivial on k X , we will, for every place v of k, denote by
Wv the quasicharacter ofk:induced on k: by W. As the groups kA(Pr
defined in the corollary of prop. 2, Chap. IV-3, are open in k~, every
neighborhood of 1 in k~ contains a subgroup of the form nVEpr~;
therefore, by lemma 2 of § 3, W must be trivial on some such group, which
is the same as to say that Wv is unramified for almost all V. Consequently,
for all z=(zv) in k~ we have W(z) = nwv(zv), the product being taken
over all the places v of k; for each z, almost all the factors in that product
have the value 1. We will write this more briefly as W = n W v•
The chief purpose ofthis Chapter can now be stated; it is to investigate
the integrals of the form
(4) Z(w, cfJ) = JcfJ(j(z»)w(z)djl(z),
k~
§ 4. Quasicharacters of A-fields 119
where the notations have the following meaning. For Jl., we take a Haar
measure on k~; for w, we take a quasicharacter of Gk = k~/kx, regarded
as above as a function on k~. For q>, we take a standard function on k A •
By j, we denote the natural bijection of k1 onto the set of invertible ele-
ments of kA' which is a continuous mapping of k~ into kA' by prop. 2 of
Chap. IV -3. By abuse of notation, we will usually write q>(z) instead of
cPU(z)) in the future.
As to Jl., it has already been observed in Chap. V-4, in the case of
characteristic 0, that such a measure can be defined by choosing, for
every v, a Haar measure Jl." on k:, in such a way that Jl.v(r:) = 1 for
almost all v. Then we write Jl.= nJl.v for the measure on k~ which
coincides with the product measure nJl.v on everyone of the subgroups
kA (P) x • The construction of the measures Jl.v is contained in the following:
LEMMA 5. Let K be a local field and ex a Haar measure on K. Then the
formula dJl.(x)=mod K(x)-ldex(x) defines a Haar measure Jl. on KX;
moreover, if K is a p-field, q its module, and R its maximal compact subring,
then Jl.(R X)=(l-q-l)ex(R).
By the definition of modK , x-+ax leaves Jl. invariant for aeKx; this
proves the first assertion. The second one follows at once from tho 6 of
Chap. 1-4.
PROPOSITION 10. Let cP= n
cP v be a standard function on kA' W= Wv n
a quasicharacter of G" = k~/kX, and Jl. = nJl.v a H aar measure on k~.
Assume that Iwl = w" with (T > 1. Then the integral Z(w, cP) in (4) is absolutely
convergent, and its value is also given by the absolutely convergent product
(5) Z(w,cP) = n(J cPv(x)wv(x)dJl.v(x)).

v k~
For each finite place v of k, put I['v=lcPvl; for each infinite place w
of k, choose a standard function I['w on kw such that Y'w~lcPwl; then,
clearly, I[' = n I[' v is a standard function on kA' majorizing IcPl, and
Z(w,cP) is majorized by Z(w", Y'). Call [(P), J(P) the integrals of cPwdJl.
an~ of I[' w"dJl., respectively, on kA{P)x, Call I", J " the integrals ofcPl.wl,dJl."
and of I[' vlwvld Jl.v, respectively, on k~, and, for every finite v, call [~, J~
r:
the same integrals taken on instead of k:; Iv is the factor correspond-
ing to v in the right-hand side of (5). For almost all finite places v of k,
cP v is the characteristic function of rv' Wv is unramified, and Jl.v(r:) = 1 ;
let Po be a finite set of places, containing P00' such that this is so for v
not in Po. Then, for v not in Po, [~=J~= 1. This implies that we have, for
all P=>Po:
veP veP
Therefore Z(w"" 1[1) is < + 00 provided all the integrals J v' and the infInite
product nJ
v , are convergent; moreover, if we show that this is so, it
will imply that Z(w, IP), the integrals I v and the product n
I v are all
absolutely convergent, and that Z(w, IP) is equal to that product, which
is what we have to prove. For any v, take a Haar measure IXv on k,,;
then, by lemma 5, dJlv(x)=mvlxl;l dlXv(x) with some mvER~. This gives
Jv=mvJ I[Iv(x)lxl~-ldlXv(X).
k~
In view of our defInition of a standard function, one sees at once that

this is convergent for (1 ~ 1; it would still be so even for (1 > 0, but this is
not needed here. On the other hand, for v not in Po, we have, since k~ n rv
is the disjoint union of the sets uv = 1t~r: for v ~ 0:
+00 +00
J v= L Jlxl~dJlv(x)= L q;v"'=(1_q;",)-1.
v=o Uv v=o
n
Prop. 1 of § 1 shows now that J v is convergent, which completes the
proof.
The method of calculation which we have just given for J v can be
applied to Iv; we formulate this as follows:
PROPOSITION 11. Let K be a p-field, q its module, R its maximal
compact subring, and Jl the H aar measure on K x such that Jl(R X) = 1.
Call qJ the characteristic function of R. Then, for Re(s»O:
JqJ(x)modK(x)"djl(x)=(l-q-·)-l.
K"
In fact, we can write K x nR as the disjoint union of the sets U v =

= 1t V R x
= pv _ pv+ 1 for v ~ O. Then our integral can be written as
+00 +00
L JmodK(x)"dJl(x) = L q-v"
v=o Uv v=O
which is absolutely convergent for Re(s»O and has the value stated
above.
§ s. The functional equation. We will fIrst choose a Haar measure

on k~. On the compact group Gt, take the Haar measure Jll given by
Jll(Gt)= 1. On the group N, take the measure v given by dv(n)=n- 1 dn
if N=R~ and by v{{1}}=1 otherwise. On G,,=G~ xN, we take the
measure Jl=Jlt xv. Finally, on k~, as explained in Chap. 11-4, we choose
as jl the measure whose image in G,,=k~/kx is the one we have just
defIned.
§ 5. The functional equation 121
LEMMA 6. Let Fl be a measurable function on N such that O~Fl ~ 1;

a.s.sume also that there is a compact interval [t o,t 1] in R~ such that
Fl (n) = 1 for nE N, n < to, and Fl (n) = 0 for nEN, n> t 1 . Then the integral
A(S) = In' Fl (n)d v(n)
N
is absolutely convergent for Re(s»O. The function A(S) can be continued

analytically in the whole s-plane as a meromorphic function. If we put
AO(S)=S-l if N=R~, and Ao(s)=1(l+Q-')(l-Q-')-l if N={QV}veZ'
then A-AO is an entire function of s. Finally, if F1 (n)+F1 (n- 1 )=1 for
all nEN, then A(S)+A( -s)=O.
Take first for Fl the function fl given by fl (n) = 1 for n < 1,

fl(1)= 1/2, fl(n)=O for n> 1. Then A becomes, for N =R~, the integral
1 + <Xl
In s - 1 dn, and, for N={QV}, the series!+ IQ-vs; in both cases it
o 1
is absolutely convergent for Re(s) >0, and equal to AO(S). This gives, for
any F 1 :
A(S)-AO(S)= I nS(Fl(n)- fl(n))dv(n).
N
As Fl - fl is a bounded measurable function with compact support on N,

the last integral is absolutely convergent for all s, uniformly on every
compact subset of the s-plane; this implies that it is an entire function
of s. Assume now that F1 (n)+F1 (n- 1 )= 1; as fl has the same property,
the function F2=F1 -fl satisfies F2(n- 1 )= -F2(n). Replacing n by n- 1
in the last integral, and observing that AO( -s)= -Ao(S), we get
A( -s)= -A(S).
Lemma 6 implies that A has at s = 0 a residue equal to 1 if N = R ~
and to (logQ)-1 if N = {QV}. Here, and also in the next results, it is
understood that residues are taken with respect to the variable s; in other
words, if a function f(s) of s has a simple pole at s = So, its residue there
is the limit of (s - so) f(s) for s-+so.
THEOREM 2. Let qJ be a standard function on kA- Then the function
w-+ Z(w, qJ)defined by formula (4) of§ 4 when the integral in (4) is absolutely
convergent can be continued analytically as a meromorphic function on
the whole of the complex manifold Q(G k). It satisfies the equation
Z(w, qJ) = Z(w 1 W- 1, qJ'),
where qJ' is the Fourier transform of qJ with respect to the Tamagawa
measure on k A • Moreover, Z(w,4» is holomorphic everywhere on Q(G k)
except for simple poles at Wo and w l ' with the residues - p 4>(0) at Wo and
p4>'(O) at w 1 , where p= 1 if N =R~ and p=(logQ)-1 if N = {QV}.
* On R:, choose two continuous functions F 0' F 1 with the following

properties: (i) Fo ~ 0, Fl ~ 0, Fo+ Fl = 1; (ii) there is a compact interval
[t o,t 1 ] in R:such that Fo(t)=O for O<t<t o, and Fl(t)=O for t>t 1 •
Take any B> 1. Then, for aER, a~B, tER~, we have ttl Fo(t)~to-BtB.
Write now, for i=0,1:
As before, put 1£01=£0" with aER; by prop. 10 of §4, Zo and ZI are

absolutely convergent for a> 1. On the other hand, if a ~ B, Zo is major-
ized by the integral
f 14>(z) 1·lzl~Fo(lzIA)dJl(z)~to-B f 14>(z)1 ·lzl!dJl(z),

~ ~
which is convergent by prop. 10 of § 4. In particular, Zo(w s w,4» is ab-

solutely convergent for all SEC, and one verifies easily that this is so
uniformly with respect to S on every compact subset of C. As the quasi-
characters wsw, for SEC, make up the connected component of £0 in
Q(G k ), with the complex structure determined by the variable s, this
shows that w-+Z o(w,4» is holomorphic on the whole of Q(G k ).
Now apply formula (6) of Chap. 11-4 to the group k~, the discrete
subgroup k x and the integrals Zo, Z 1. This gives:
Zi= f ( L 4>(z~)) w(z)Fi(lzIA)dJl(z),

Gk ~Ek'
where z is the image of z in Gk = k ~/k x , and the integrand is to be under-

stood as a function of z. Here the integrals for Zo, Z 1 are absolutely
convergent whenever the original integrals for Zo, Z 1 are so, i. e. for
a> 1 in the case of Z 1 and for all a in the case of Z 0.
For each zEk~, we may apply lemma 1 of § 2 to the automorphism
x-+z- 1 X of kA ; applying then Poisson's formula, i.e. (1) of § 2, to the
function x-+4>(z x), we get:
4>(0)+ L 4>(z~)=lzIAI (4)'(0)+ L 4>'(~Z-I)),

~Ek' ~Ek'
and therefore:
ZI = J (L 4>'(~z-I)+4>'(O)-lzIA4>(O))
Gk ;Ek~
IzIA 1W (z)F1 (lzIA)dJl(z).
On the other hand, what we have proved above for Zo remains valid if
we replace £0 by £0 1£0- 1,4> by 4>' and Fo by the function t-+F1(t-I).
Calling Z~ the result of this substitution, we get:

Z~ = J4>'(z)lzIA W(Z)-l Fl (lZIA 1 )dJl(Z);
k~
therefore this is always absolutely convergent, and holomorphic on the

whole of Q(Gk ). In this integral, replace z by Z-l; this changes the Haar
measure Jl into a Haar measure C Jl, where c2 = 1 since it is a homeo-
morphism of order 2 of k~ onto itself, hence c = 1. After this change of
variable, apply again formula (6) of Chap. 11-4 to k ~ and k x • This gives:
Z~= f (L 4>'(C lZ - l )) IzIAlW(z)Fl(izIA)dJl(z),

Gk ~Ek'
this again being always absolutely convergent. As e-+e- 1 is a bijection

of k x onto itself, we get now:
Z 1 - Z~ = J(4)'(0) -lzlA 4>(O))lzIA 1 w(z)Fl (lzIA)dJl(Z),
Gk
this being absolutely convergent for u> 1, since Zl and Z~ are so. By
corollary 2 of prop. 7, § 3, we can write w = Ws t/J, where t/J is a character
of Gk , trivial on N. In view of our definition of Jl as the measure Jll x v
on Gk = Gl x N, our last formula can now be written:
Zl-Z~= (L t/JdJll)
k
. U(4)'(O)-n4>(O))n S - l Fl(n)dv(n)) .
The first factor in the right-hand side is 1 or 0 according as t/J is trivial

or not, i. e. according as w is principal or not; write ()", for this factor.
The second one can be evaluated at once by lemma 6. If A.(s) is as defined
in that lemma, this gives:
Zl -Z~={),,,(4)'(0)A.(s-1)-4>(0)A.(s)).
As Z(w,4»=ZO+Zl' this proves that Z(w,4» can be continued every-
where on Q( Gk ) outside the connected component Q 1 of Wo = 1 as a
holomorphic function, and on that component as a meromorphic func-
tion having at most the same poles as A.(s-1) and A.(s); as to the latter
poles and their residues, they are given by lemma 6 and are as stated
in our theorem. Finally, assume that we have chosen Fo, Fl so that
Fo(t)=Fl(t-l) for all t; this can be done by taking for Fl a continuous
function for t~ 1, such that O~Fl(t)~ 1 for all t ~1, Fl(l)= 1/2, and
Fl(t)=O for t~tl' and then putting Fl (t)=1-F l (t-l) for 0<t<1, and
Fo = 1- Fl. That being so, we have Z~ = ZO(WI W-1,4>'), and therefore
Z(w, 4» = Zo(w, 4>)+ ZO(WI W-1, 4>') + (),,,(4)'(O)A.(s -1) - 4>(O)A.(s)).
In this formula, replace W by WI W -1 and 4> by 4>'. In view of Fourier's

inversion formula, this replaces 4>' by the function 4>/1 given by
cP"(x)=cP(-x). As w is trivial on kX, we have w(-I)=I, hence

w( -z)=w(z) for all z, so that Zo(w,cP") is the same as Zo(w,cP); there-
fore this substitution merely interchanges the first two terms in the right-
hand side of our formula; as it changes s into 1- s, lemma 6 shows that
it does not change the last term. This completes the proof of the "func-
tional equation" in theorem 2.
COROLLARY 1. Let P be a finite set of places of k, containing P 00;

then the product
p(k,P,s) = n(l_q;·)-l
v,P
is absolutely convergent for Re(s) > 1, and (s-l)p(k,P,s) tends to a finite

limit > 0 when s tends to 1.
The first assertion is contained in corollary 1 of prop. 1, § 1. Now
take Haar measures avon kv, Pv on k ~ , as explained above; by lemma 5
of § 4, we have, for every v, dPv(x)=mvlxl;ldOCv(x), with some mv>O.
Take the standard function cP so that cPv is the characteristic function
of rv for all v not in P, and that cPv;?;O and cPiO»O for all v. Apply
prop. 10 of § 4 to Z(w.,cP) for Re(s) > 1; the factor Iv corresponding to v,
in the right-hand side of the formula in that proposition, can now be
written as
For v not in P, by prop. 11 of § 4, this differs from (l-qv-')-l only by

the scalar factor Pv(r ~), which is always > 0, and which is 1 for almost
all v. For veP, one can verify at once that Iv is continuous for Re(s);?; 1
(one could easily show, in fact, that it is holomorphic for Re(s) >0, and,
in the next §, one will obtain a much more precise result for a specific
choice of cP, but this is not needed now); for s tending to 1, it tends
to mvJ 4>"da v , which is >0. This shows that Z(w., cP) differs from the
product p(k,P,s) in our corollary by a factor which tends to a finite
limit > 0 when s tends to 1. On the other hand, theorem 2 shows that
Z(w.,cP) has a simple pole at s= 1, with the residue pcP'(O), and p>O;
J
as cP'(O) = cPda, and as this is obviously >0, this completes the proof.
COROLLARY 2. Let P be as above; let w be a non-trivial character
of k~, trivial on k X, such that Wv is unramified for all v not in P; for v
not in P, put A(V)=Wv(1tv), where 1tv is a prime element of kv' Then the
product
n
p(k,P,w,s) = (1-A(V)qv-·)-1
v,P
is absolutely convergent for Re(s) > 1 and tends to a finite limit when s
tends to 1; if w 2 is not trivial, this limit is not O.
As W is a character, we have IA(V)I = 1 for all v not in P, so that the

first assertion is again contained in corollary 1 of prop. 1, § 1. Take
lXv' I1v as before, and take <I> so that, for v not in P, <l>v is the characteristic
function of rv. Apply prop. 10 of § 4 to Z(wsw, <1» for Re(s) > 1; the fac-
tor Iv is now
Iv = mv J<l>v(x)wv(x) Ixl~- 1 dlXv(X).
k~
For v not in P, WV is unramified and may be written as wv(x)=lxl~v,

where Sv can be determined by A(V) = qv- Sv ; then prop. 11 of § 4 shows
that Iv differs from (1- A(V)qv-st 1 only by the scalar factor I1v(rvX ), which
is 1 for almost all v. For v E P, we observe, as before, that I v is continuous
for Re(s) ~ 1; taking prop. 9 into account when v is an infinite place, one
sees easily that, for each VEP, <l>v may be so chosen that Iv is not 0 for
s = 1, and we will assume that it has been so chosen (for specific choices
of <l>v, Iv will be computed explicitly in § 7). We see now that Z(wsw, <1»
differs from the product p(k,p,w,s) in our corollary by a factor which
tends to a finite limit, other than 0, when s tends to 1. In view of theorem 2,
this proves the second assertion in our corollary. As to the last one,
we need a lemma: *
LEMMA7. For tEC, AEC, put cp(A,t)=(l-t)3(I-At)4(I-A 2 t). Then
Icp(A,t)I<1 for tER, O<t<l, AX=1.
In fact, we have then
10glcp(A, tW = 10g(cp(A, t)cp(X, t))
=- I
n= 1
~
n
(6+4A n+4Xn+ A2n+X2n)
tn
L -
00
=- (2+A"+Xn)2<O.
n= 1 n
If now cp(A, t) is defined as in the lemma, we have
p(k,p,S)3 p(k,p,W,S)4 p(k,P,w 2 ,s)= n cp(A(V),qv-st
v,P
1.
By the lemma, this has an absolute value > 1 for sER, s> 1, so that it
°
cannot tend to for s tending to 1. For s tending to 1, as shown above,
p(k,P,w 2 ,s) tends to a finite limit if w 2 is not trivial, and p(k,P,w,s) is
the product of a factor, tending to a finite limit other than 0, and of
Z(wsw, <1», which is holomorphic in a neighborhood of s = 1; therefore,
if p(k,w,P,s) tends to 0, it must be of the form F(s)(s -1), with F bounded.
In view of corollary 1, this implies that the left-hand side of the last
formula tends to 0 for s tending to 1. This completes our proof. It is
an important fact that the conclusion of our corollary remains true
* even if oi = 1; the proof for this, which requires quite different methods,
will be given in Chap. XIII-12.
COROLLARY 3. Let ko be an A-field contained in k; let V be a set of
finite places of k, such that, for almost all the finite places v of k, not
in V, the modular degree of kv over the closure of ko in kv is > 1. Then
the product
veY
is absolutely convergent for Re(s) > 1, and (s-l)q(k, V,s) tends to a finite
limit > 0 when s tends to 1.
In fact, with the notation of corollary 1, p(k, P00' s) is the product of

q(k, V,s) and of the similar product, taken over the set M of all the finite
places of k, not in V; applying corollary 3 of prop. 1, § 1, to the latter
product, and corollary 1 to p(k, P00' s), we get our conclusion at once.
Of course our corollary implies that V cannot be a finite set, or in other
words that there are infinitely many places v of k for which the modular
degree in question is 1.
COROLLARY 4. Let ko and V be as in corollary 3; let k' be a separably

algebraic extension of k of finite degree n, and assume that there are n
distinct places of k' above every place ve V. Then k' =k.
Call V' the set of the places of k! lying above those of V. By corollary 1
of tho 4, Chap. III -4, if ve V, and w lies above v, we have J<., = kv, hence
q~=qv. For any place v of k, and any place w of k' above v, the modular
degree of J<., over the closure of ko in kv is at least equal to that of kv
over that closure; therefore, for almost all v, not in V, or, what amounts
to the same, for almost all w, not in V', that degree is > 1. We can now
apply corollary 3 to the products q(k, V,s) and q(k', V',s); as the latter is
equal to q(k, V,st, this gives n= 1.
COROLLARY 5. Let k be an A-field of characteristic p>l, and let P
be a finite set of places of k. Then there is a divisor m= Lm(v)· v of k
of degree 1 such that m(v)=O for all veP.
Call v the g. c. d. of the degrees of all the places v, not in P; we have
to show that v= 1. Let F=Fq be the field of constants of k; by th.2 of
Chap. 1-1, there is, in an algebraic closure of k, a field F' with qV elements,
and it is separable over F. Call k' the compositum of k and F', and n
its degree over k; k' is separable over k. Let v be any place of k, not
in P; let w be a place of k' above v; by prop. 1 of Chap. III-I, J<., is
generated over k,; by k', hence by F'. By the definition of v, the module
of kv is of the form qvr, where r is an integer; therefore, by corollary 1
of tho 7, Chap. 1-4, combined with corollary 2 of tho 2,Chap. 1-1, kv con-
tains a subfield with qV elements. By tho 2 of Chap. 1-1, 1<. cannot contain
more than one field with qV elements; therefore F' c kv, hence 1<. = kv'
Corollary 1 of th.4, Chap. 111-4, shows now that there are n distinct
places of k' above each place v of k, not in P. Taking ko = k in corollary 4,
and taking for V the complement of P, we get k' = k, hence F' c F,
i.e. v= 1.
COROLLARY 6. Let k be as in corollary 5, and let Fq be its field of
constants. Then the value-group N of IzlA on k~ is generated by q.
As we have seen in § 4, N is generated by the value-groups of Ixl v

on k~ for all v, hence by the modules qv=qdeg(v), so that it has the
generator Q=q., where v is the g.c.d. of all the degrees deg(v). By
corollary 5, v = 1.
Taking corollary 6 into account, we can reformulate the last asser-
tion of theorem 2, in the case of characteristic p> 1, as follows:
COROLLARY 7. Let k and Fq be as in corollary 6; let notations be as
in theorem 2. Then Z(Ws' 4»+ 4>(O)(1-q-S)-1 is holomorphic at s=O.
This follows at once from the results we have just mentioned and
from the fact that (1_q-S)-1 has the residue (logq)-1 at s=O.
§ 6. The Dedekind zeta-function. Special choices of 4> in Z(w,4»

lead to the definition of important functions on the connected compo-
nents of Q(Gk ); these will now be investigated more in detail. We begin
with the consideration of the connected component Q1 of Wo = 1 in
Q( Gk), i. e. of the group of the principal quasicharacters of Gk> choosing 4>
as follows. Whenever v is a finite place of k, we take for 4>v the characteristic
function of rv' When v is real, i.e. kv=R, we take <Pix)=exp( -nx 2 ).
When v is imaginary, i.e. kv=C, we take 4>v(x)=exp( -2nxx). We have
now to calculate the factors in the product (5) for Z(w,4», for this choice
of 4> and for W=Ws; when v is a finite place, these are given by prop. 11
of § 4, up to a scalar factor depending on Jl. For the infinite places, they
are as follows:
LEMMA 8. Let G1 ,G2 be defined, for all SEC, by the formulas
G1(s)=n- S/2 r(s/2), G2 (s)=(2n)1-S r(s).
Then we have, for Re(s»O:
Jexp( _nx 2)lxI S- 1 dx = G1 (S),
R'
Jexp( - 2nxx)(xxr Ildx 1\ dxl = G2 (s).

c·
This can be verified at once by obvious changes of variables, viz.,

Ixl = t 1/ 2
in the first integral and x = t 1 / 2 e(u) in the second one, with
teR~, ueR, 0~u<1; in the latter case, Idx/\dxl=2ndtdu.
n
Now consider the measure y = Yv on k~, given by taking yv(r:) = 1
for every finite place v, dYv(x)=lxl-1dx when v is real, and
dYv(x)=(XX)-lldx /\ dxl when v is imaginary; when k is of characteristic
0, this is the measure occurring in prop. 9 of Chap. V-4. The relation
between y and the measure JI. introduced at the beginning of § 5 is as
follows:
PROPOSITION 12. Let JI. be as in § 5, and y as above. If k is of characteri-

stic 0, we have y=C"JI., where Ck is as defined in proposition 9 of Chapter
V-4. If k is of characteristic p> 1, with the field of constants Fq , and if h
is the number of divisor-classes of degree 0 of k, then y=C"JI. with
c,,=h/(q-1).
In view of our definition of JI., the first assertion is merely a restatement

of prop. 9, Chap. V-4. Now let k be of characteristic p> 1, and put
U= nr: ; this is the same as Q(0) in the notation of Chap. IV-4, and it
is an open subgroup of k~; by definition, we have y(U)= 1. As explained
in Chap. 11-4, we will also write y for the image of the measure y in
G" = k~/k" ; G~ being, as before, the image of kl in G", JI. is defmed by
JI.(GD= 1, so that we have y=C"JI. with c,,=y(GD. Call U' the image of U
in G,,; by tho 8 of Chap. IV-4 and its corollary, the kernel of the morphism
of U onto U', induced by the canonical morphism of k~ onto G", is
F;, so that we can compute y(U') by taking G = U, r 1 = F;, r = {1} in
lemma 2 of Chap. 11-4; this gives y(U')=(q_l)-l. Clearly the index of
U' in G~ is equal to that of k" U in ki ; as we have seen in Chap. VI that
kl!k x U may be identified with the group Do(k)/P(k) of the divisor-
classes of degree 0 of k, that index is h. Therefore y(GD= h/(q -1).
Now, for each infinite place w of k, put Gw =G 1 or Gw =G 2 according

as w is real or imaginary. Combining prop. 10 of § 4, prop. 11 of § 4,
lemma 8, and prop. 12, we get for Re(s) > 1, tP being chosen as explained
above:
(6) Z(WS,tP)=C;l n Gw(s) n (1-q;S)-t,
w€p~ v.p~
with c" as in prop. 12. By tho 2 of § 5, the left-hand side can be continued
analytically as a meromorphic function over the whole s-plane; as the
same is true of the factors Gw , it is also true of the last product in the
right-hand side. This justifies the following definition:
DEFINITION 8. The meromorphic function (k in the s-plane, given for

Re(s) > 1 by the product
taken over all the finite places v of k, is called the Dedekind zeta-function
ofk.
When tP is as above, its Fourier transform tP' is immediately given
by tho 1 of § 2 and its corollary 2, combined with corollary 3 of prop. 2,
§ 2, and propositions 4 and 5 of § 2. This gives
tP'(Y) = lali'2 tP(a y)
where a is a difTerental idele attached to the basic character X. In view of
the definition of Z(w, tP) by formula (4) of § 4, we have now:
Z(w,tP')=lali'2 w (a)-1 Z(w,tP),
hence in particular, for W=W., i.e. w(x)=lxIA:
(7) lali'2 -s Z(ws' tP);
Z(ws' tP') =
moreover, the value of lal A is that given in prop. 6 of § 2.
We are now ready to formulate our final results on the zeta-function.
THEOREM 3. Let k be an algebraic number-field with r 1 real places and

r2 imaginary places. Call (k its zeta-function, and write
Zk(S) = G1(S)'I G 2 (S)'2(k(S).
Then Zk is a meromorphic function in the s-plane, holomorphic except for

simple poles at s = 0 and s = 1, and satisfies the functional equation
1
Zk(S) = IDlz -s Zk(l-s)
where D is the discriminant of k. Its residues at s = 0 and s = 1 are respec-
tively - Ck and IDI- 1/2 Ck' with Ck given by
Ck= 2rl(2n)'2 hR/e,
where h is the number of ideal-classes of k, R its regulator, and e the
number of roots of 1 in k.
This follows immediately from (6), (7), prop. 12, prop. 6 of § 2, and
from tho 2 of § 5.
COROLLARY. The Dedekind zeta-function (k(S) has the residue IDI- 1/2 Ck
at s= 1.
This follows from tho 3 and the well-known fact that G 1 (1)= G2 (1)= 1.
THEOREM 4. Let k be an A-field of characteristic p> 1; let Fq be its

field of constants and g its genus. Then its zetajunction can be written in
the form
P(q-,)
C,,(s) = (1_q-,)(1_ q 1-S)
where P is a polynomial of degree 2g with coefficients in Z, such that
(8) P(u) = q'u 2 , P(ljqu).
Moreover, P(O) = 1, and P(l) is equal to the number h of divisor-classes of
degree 0 of k.
In fact, corollary 6 of tho 2, § 5, shows at once that S-+Ws has the
same kernel as s-+q-S, so that C,,(s) may be written as R(q-S), where R
is a meromorphic function in C)(, with simple poles at 1 and at q - 1.
Moreover, corollary 1 of prop. 1, § 1, shows that R(u) tends to 1 for u
tending to 0, so that R is holomorphic there, and that R(O) = 1. We may
therefore write R(u)=P(u)j(l-u)(l-qu), where P is an entire function in
the u-plane, with P(O) = 1. Now (7), combined with (6) and with prop. 6
of § 2, gives formula (8) of our theorem; clearly this implies that P is a
polynomial of degree 2g. Finally, corollary 7 of tho 2, § 5, combined
with prop. 12, gives P(l)=h.
§ 7. L-functions. We will now extend the above results to arbitrary

quasicharacters of G,,; in order to do this, we adopt the following nota-
tions. Let W be any quasicharacter of G,,; as we have seen in §§ 3--4, we
may write Iwi =w", with O'eR. For every v, we write Wv for the quasi-
character of k: induced on k: by W. For every [mite place v, we write
pt;(v) for the conductor of wv;f(v) is 0 if and only if Wv is unramified, hence,
as we have seen in § 4, at almost all finite places of k; when that is so, we
write wv(x)=lxl~v with sveC; clearly we have then Re(sv) = 0'. At the
infinite places of k, we can apply prop. 9 of § 3; this shows that Wv may be
written as wv(x)=x-Alxl sv if v is real, with A =0 or 1 and sveC, and as
wv(x)=X-AX-B(XX)Sv if v is imaginary, with inf(A,B)=O and sveC; in
the former case we put Nv=A, and we have Re(sv)=Nv+O', and in the
latter case we put Nv=sup(A,B), and we have Re(sv) = (NJ2) +0'. As the
connected component of w, in the group Q(G,.) of the quasicharacters
of Gk> consists of the quasicharacters WsW for seC, the integers f(v),
N v have the same values for all the quasicharacters in that component.
They are all 0 if w is principal, or, more generally, if w is trivial on the
group U of the ideles (zv) such that IZvlv= 1 for all places v of k; the struc-
ture of the group of the quasicharacters with that property can easily
be determined by the method used in the proof of tho 9, Chap. IV-4.
§ 7. L-functions 131
Furthermore, with the same notations as above, we attach to w a

n
standard function cPOl= cP v on kA' as follows. For each finite place v
where f(v) =0, i.e. where Wv is unramified, we take for cP v, as before, the
characteristic function of rv' For each finite place v where f(v)~ 1, we
r:
take cPv equal to w;; 1 on and to 0 outside r: .
At each infmite place v,
we take cPv(x)=xAexp( _1lX2) if v is real, and cPv(x) = xAxBexp( -21lxx)
if v is imaginary, the integers A, B being as explained above. Then cPOl
will be called the standard function attached to W; it is clear that it does
not change if w is replaced by w.w, with any seC, and also that the
function attached in this manner to ro, or to w - 1 = W _ 2 .. ro, or to
, -1 . ;;;-
W =w 1 W ,IS 'YOl'
We need to know the Fourier transform of cPOl' or, what amounts to
the same in view of tho 1 of § 2, those of the functions cP v defined above.
The latter are given by our earlier results except when v is a finite place
where Wv is ramified. For that case, we have:
PROPOSITION 13. Let K be a p-field; let R be its maximal compact

subring, P the maximal ideal of R, and w a quasicharacter of K x with
the conductor pI, where f~ 1. Let X be a character of K of order v, ex the
self-dual measure on K with reference to x, and let beK x be such that
ordK(b) = v +f Let qJ be the function on K, equal to w - 1 on R x and to 0
outside R x. Then the Fourier transform of qJ is
qJ'(y) = Kmod K(b)I/2 qJ(b y),
where K is such that KK= 1 and is given by
K=mod K(b)-1/2 JW(X)-1 X(b- 1x)dex(x).
RX
By prop. 12 of Chap. 11-5, the dual of the K-Iattice pI in K is p-I-v;

as qJ is constant on classes modulo pI in K, prop. 2 of § 2 shows that qJ'
is 0 outside P-I-v=b- 1R. The definition of qJ gives
(9) qJ'(y)= JW(X)-1 X(xy)dex(x).
RX
Obviously the measure induced by ex on R x is a Haar measure on R x
(this may also be regarded as a consequence of lemma 5, § 4). Take y
such that ordK(y)~ -f-v + 1; then, by prop. 12 of Chap. 11-5, x--+X(xy)
is constant on classes modulo pI-I. Assume first thatf= 1; then X(xy) = 1
on R, so that (9) is the integral of w - 1 d ex on R x , which is 0 since w is a
non-trivial character of the compact group R x • Assume now f> 1 ; then
(9) is the sum of the similar integrals taken over the classes modulo
pI - 1 contained in R x , which are the same as the cosets of the subgroup
1+ pI-l in R \ since the definition of the conductor implies that w is
non-trivial on 1 + pI-" the same argument as before gives again
/'P'(y)=0 in this case. Now take y=b-1u in (9), with ueRx; substituting
u-1x for x, we get /'P'(b- 1 u)=w(u)/'P'(b- 1). This proves that /'P' is of the
form c /'P(by) with ceC x • Applying to this Fourier's inversion formula
and lemma 1 of§2, we get cc=modg(b). As c=/'P'(b- 1), we get for"
the formula in our proposition. It would be easy to verify directly that
"K= 1 when" is defined by that formula; moreover, as the integrand
there is constant on classes modulo pI in R, we can rewrite the integral
as a sum over Rlpl; sums of that type are known as "Gaussian sums".
PROPOSITION 14. Let W be a quasicharacter of G... and «P", the standard
function attached to w. Then the Fourier transform of «P"" with reference
to the basic character X of kA' is given by
«P'(y) = "lbli'2 «P",(b y) = "lbli'2 «Pc;,(b y)
where "=n,,v, "veC and "vKv=l for all v, b=(bv)ek~, and "v' bv are
as follows. Let a=(a v) be a differental idele attached to X; then bv=av at
each infinite place v, and, for each finite place v of k, ordv(bva; 1) = f(v).
At every infinite place v of k, "v=i- Nv ; at every finite place v where
f(v)=O, "v= 1; at all other places:
"v= Ibvl; 1/2 Jwv(xt 1 xv(b; 1 x)dcxv(X),
r~
where CX v is the self-dual Baar measure on kv with reference to Xv.

This follows at once from prop. 13, propositions 4 and 5 of § 2, and
corollary 3 of prop. 2, § 2.
COROLLARY. Let w be as in proposition 14, and put w' = W 1 W -1. Then
Z(w, «P",) = "lblA" 1/2 w(b)Z(w', «P",.).
For all w, by tho 2 of § 5, Z(w, «P",) is equal to Z(w', «P'), where «P' is
as in prop. 14. Express Z(w',«P') by the integral in (4), §4, under the
assumption that it is convergent, which, as one sees at once, amounts to
q < O. Expressing «P' by proposition 14, and making the change of variable
z -+ b- 1 Z in that integral, one gets the right-hand side of the formula in
our corollary. By tho 2 of § 5, both sides can then be continued analytically
over the whole of the connected component of win D(Gk ), so that the
result is always true.
Now apply prop. 10 of § 4 to Z(w, «P",); for q> 1, this gives an infinite
product whose factors are all known to us except those corresponding to
the finite places v of k wheref(v»O; as to these, our choice of«Pvmakes it
obvious that they are respectively equal to Ilv(r~). As in § 6, put Gw = G1
when w is a real place, and Gw = G2 when it is an imaginary place. Taking
into account prop. 11 of § 4, lemma 8 of § 6, and prop. 12 of § 6, we get,
for q> 1:
§ 7. L-functions 133
(10) Z(W,4>co)=Ck 1 n Gw(Sw) n(l_q;Sv)-l,

WEP~ v,P
where P is the set consisting of the infinite places and of the finite places
whereJ(v»O.
For every place v of k, not in the set P which we have just defined, put
A(V)=q;Sv; these are the finite places where Wv is unramified, and the
definition of Sf' for such places shows that we can also write this as
A(V)=Wv(:n: v), where :n:v is a prime element of kv, or even as A(V)=W(:n:v)
if k; is considered as embedded as a quasifactor in k~. Clearly we have
IA(V)I = q; tr. .
In (10), replace now w by WsW, with seC; as observed above, this

does not change 4>co; it replaces the right-hand side of (10) by a product
which is absolutely convergent for Re(s) > 1- u. As tho 2 of § 5 shows
that this can be continued analytically over the whole s-plane (as a
holomorphic function if w is not principal), and as the same is true of
the factors Gw when they occur, we may now introduce a meromorphic
function L(s,w), given, for Re(s) > 1- u, by the product
(11) L(s,w) = n(1- A(V)q;st

v
1
taken over all the finite places v where Wv is unramified.

In order to formulate our final result in the case of characteristic 0,
n
we introduce the ideal in t given by f= lfv(v), which is called the con-
ductoroJ W.
THEOREM 5. Let k be an algebraic number-field, and w a non-principal
quasicharacter oj Gk=k~/kx, with the conductor f. Then
A(s,w)= n Gw(s+sw)'L(s,w)
wePco
is an entire Junction oj s, and satisfies the Junctional equation

1
A(s,W)= Kw(b)(IDI91(f)) 1: -s A(1-s,w- 1 ),
where K and b are as in proposition 14.
This is an immediate consequence of the corollary of prop. 14, when
one replaces w in it by WsW, taking into account the definitions of a, b
and f and the fact that lalA = IDI- 1 . As it is well-known that r(s)-l is an
entire function, the same is true of the functions Gw(s + sw) - 1; therefore
theorem 5 implies that L(s,w) is an entire function of S.
According to their definition, the above functions do not depend
essentially upon the choice of w in a given connected component of
Q(G k ); more precisely, they are independent of that choice, up to a trans-
lation in the s-plane, since, for every teC, L(s,wtw) is the same as
L(s + t,OJ), this being also true for A(s, OJ). In view of corollary 2 of prop. 7,
§ 3, one may therefore always assume, after replacing OJ by OJ_tOJ with a
suitable tEe if necessary, that OJ is a character of k~, trivial on k x and
also on the group M defined in corollary 2 of tho 5, Chap. IV-4. The
latter assumption can be written as ~)ovsv-Nv)=O, where the sum is
taken over the infinite places of k, Sv and N v are as above, and ov= lor 2
according as kv is R or C. Since this implies that OJ is a character, we have
then (1=0.
On the other hand, if k is of characteristic p> 1, we introduce the
divisor f= L:f(v)' v, and call this the conductor of OJ. Then:
THEOREM 6. Let k be an A-field of characteristic p > 1; let Fq be its

field of constants, g its genus, and OJ a non-principal quasicharacter of
Gk=k~/e with the conductor f. Then one can write L(s,OJ)=P(q-s,OJ),
where P(u,OJ) is a polynomial of degree 2g - 2+ deg(f) in U; and we have
P(u, OJ) = KOJ(b) .(q1 /2U)2 g - 2 +deg(f). P(I/qu,OJ -1)
where K and b are as in proposition 14.

The fact that we can write L(s, OJ) = P(q-S, OJ), where P(u,OJ) is holo-
morphic in the whole u-plane, is proved just as the corresponding fact in
theorem 4. The last formula in our theorem is then an immediate conse-
quence of the corollary of prop. 14 when one replaces OJ by OJsOJ there,
provided one takes into account the definitions of a, band f and the fact
that lal A =q2-2 g• Then that formula shows that P(u,OJ) is a polynomial
whose degree is as stated.
Here again one will observe that, for tEe, P(u,OJ,OJ) is the same as
P(q-tu,OJ). In this case, we have written k~ =kl x M, where (if one takes
corollary 6 of tho 2, § 5, into account) M is the subgroup of k~ generated
by an element Zl such that IZllA =q, i.e. such that div(zl) has the degree
-1. Then corollary 2 of prop. 7, § 3, shows that, after replacing OJ by
OJ_tOJ with a suitable tEe, if necessary, one may assume that OJ(Zl) = 1;
the corollary in question shows also that OJ is then a character of k~, i. e.
that (1 = 0; furthermore, if one combines it with lemma 4 of § 3, and with
the obvious fact that in the present case the group k~, hence also the
groups k,l., Gk , G~ are totally disconnected, it shows that OJ is then a
character of finite order of k ~ .
§ 8. The coefficients of the L-series. When an Euler product such as

the right-hand side of (11) is given, the question arises whether it can be
derived from a quasicharacter OJ of k~/e. The answer to this, and to
a somewhat more general problem which will be stated presently, depends
on the following result:
PROPOSITION 15. Let P be a finite set of places of k, containing P00;

let Gp be the subgroup of k~, consisting of the ideles (zv) such that Zv= 1
for all VEP. Then k Gp is dense in k~.
X
Put kp= 0 kv, the product being taken over the places VEP; write
Ap for the subgroup of kA consisting of the adeles (xv) such that x,.=O
for all VEP; then kA=kpxAp and k~=k;xGp, and our assertion
amounts to saying that the projection from k~ onto k; maps kX onto
a dense subgroup of k~. In fact, k; is an open subset of kp, and its
topology is the one induced by that of k p ; our assertion follows now at
once from corollary 2 of tho 3, Chap. IV-2, which shows that the pro-
jection from kA onto kp maps k onto a dense subset of kp.
From prop. 15, it follows at once that a continuous representation w
of k~ into any group r, trivial on k is uniquely determined when its
X
k:
,
values on the groups are known for almost all V. In particular, if

r = ex, or more generally if r is such that every morphism of k~ into
r:
r is trivial on for almost all v, w is uniquely determined when the
w(1tv) are given for almost all V. Clearly every flnite group r has that
property, since the kernel of every morphism of k~ into a fmite group
is open in k~ and therefore contains Or:
for some P; the same is true
vfp
of every group r without arbitrarily small subgroups, for the same
reason for which it is true for r = ex. Another case of interest is given by
the following:
PROPOSITION 16. Let K be a p-field, and assume that k is not of charac-
teristic p. Then every morphism w of k~ into K is trivial on r~ for almost
X
all v, and is locally constant on k: whenever kv is not a p.Jield.

As k is not of characteristic p, we have Iplv= 1 for almost all v, and
then kv is not a p-field. As every morphism of a connected group into a
totally disconnected one must obviously be trivial, w is trivial on k:
when kv = C, and on R: when kv = R. Call R the maximal compact
subring of K, and P its maximal ideal. Let v be any flnite place of k such
that kv is not a p-fleld; let m ~ 1 be such that w maps 1+ p~ into 1 + P. For
every n ~ 0, by prop. 8 of Chap. II -3, every z E 1+ p~ can be written as
z'P" with Z' E 1+ p~; therefore w(z) is in (1 + P)P", hence in 1 + pn+ 1 by
lemma 5 of Chap. 1-4; as n is arbitrary, this shows that w is trivial on
1 + p~, hence locally constant on k:. By tho 7 of Chap. 1-4 if K is of
characteristic p, and by that theorem and prop. 9 of Chap. 11-3 if it is of
characteristic 0, there are only flnitely many roots of 1 in K, and we can
choose v> 0 so that there is no root of 1, other than 1, in 1+ P·. Take a
neighborhood of 1 in k~ which is mapped into 1 + r by w; as this con-
tains r; for almost all v, we see now that, for almost all v, w is trivial on
1+ Pv and also on the group of all roots of 1 in kv, and therefore on r~ .
For every finite set P of places of k, containing P00' we will write

G~ = nr; ;
",p
this is an open subgroup of the group Gp defined in prop. 15;
it consists of the ideles (zv) such that Zv= 1 for VEP, and zvEr;, i. e.
IZvlv= 1, for v not in P; if r is any group with the property described
above, and (J) is any morphism of k~ into r, there is a set P such that (J)
is trivial on Gpand therefore determines a morphism qJ of Gp/G~ into r;
if at the same time (J) is trivial on k X , prop. 15 shows that (J) is uniquely
determined by qJ. We will discuss now the conditions on qJ for such a
morphism (J) to exist.
If k is an algebraic number-field, and P is as above, we will say that a
fractional ideal of k is prime to P if no prime ideal Pv, corresponding to
a place VE P, occurs in it with an exponent =1= O. Similarly, if k is of charac-
teristic p> 1, we say that a divisor is prime to P if no place VEP occurs
in it with a coefficient =1= O. We will write I(P) (resp. D(P)) for the group
of the fractional ideals of k (resp. of the divisors of k) prime to P. Clearly
the morphism z--+id(z) of k~ onto I(k). (resp. the morphism z--+div(z)
of k~ onto D(k)) determines an isomorphism of Gp/G~ onto I(P) (resp.
D(P)), which may be used to identify these groups with each other, or,
what amounts to the same, with the free abelian group generated by
the places of k, not in P. In particular, every mapping v--+,1,(v) of the set
ofthese places into a commutative group r can be uniquely extended to a
morphism qJ of I(P) (resp. D(P)) into r; then qJo(id) (resp. qJo(div)) is a
morphism of Gp into r, trivial on Gp.
PROPOSITION 17. Let qJ be a morphism of I(P) (resp. D(P)) into a

commutative group r; for each VEP, let gv be an open subgroup of k;,
contained in r; whenever v is finite. Then the morphism qJo(id) (resp.
qJo(div)) of Gp into r can be extended to a morphism (J) of k~ into r,
trivial on kX, if and only if one can find,Jor every VEP, a morphism !/Iv
of gv into r, so that qJ(id(e)) (resp. qJ(div<e))) is equal to n!/Iv(e) for all
eEn 1
(kX ngv). When that is so, (J) is unique and induces !/I;- on gv for
every VEP.
Put g = ngv; this being considered as a subgroup of k ~ in the ob-

veP
vious manner, g . Gp is an open subgroup of k ~ and is the direct product
of g and Gp. Then kXg·Gp is an open subgroup of k~, so that, in view
of prop. 15, it is k~. It is now obvious that a morphism of g. Gpinto r
can be extended to one of k ~ = k x g . Gp , trivial on k x, if and only if
unique. Clearly y is the same as the group n

it is trivial on the group y=k x n(g· Gp ), and that the extension is then
(k ng,,) in our proposition.
X
As z--+id(z) (resp. z--+div(z)) is trivial on g, it maps g·Gp=g x Gp onto

I (P) (resp_ D(P»; therefore if we write cp 1 for the morphism cpo(id) (resp.
cpo(div» of Gp into r, its extensions to g' Gp are the morphisms of the
form 1/1-1 CPl' where 1/1 is any morphism of g into r. Writing I/Iv for the
morphism induced by 1/1 on gv, we get our conclusion.
Obviously, if the condition in proposition 17 is satisfied for some
choice of the groups gv and of the morphisms l/lpo it remains so when
one substitutes, for each gv' any open subgroup g~ of gv, and then for
I/Iv the morphism induced by I/Iv on g~. For instance, one may always
take gv = R: when kv = R, and take for gv one of the groups 1+ p': with
m ~ 1 when v is a finite place. The same idea gives the following:
COROLLARY. In proposition 17, assume that r is (a) discrete, or (b)

the group ex, or (c) the group K x, where K is a local p-field. Then the
extension (J) exists if and only if groups gv and morphisms I/Iv can be found
with the properties stated in proposition 17 and the following additional
one: in case (a), 1/1 v= 1 for all VEP; in case (b), I/Iv=1 for all the finite
places vE P; in case (c), 1/1 v= 1 for all the places v E P for which kv is not
a p-field.
In fact, assume that the conditions in proposition 17 are fulfilled for

some choice of the groups gv and of the morphisms I/Iv' Then, in case (a),
we can replace gv by the kernel g~ of I/Iv for each VEP, since this is an
open subgroup of gv, and then I/Iv by 1. In case (b), we can do this for
every finite place VEP, by lemma 4 of § 3; this can also be done, for
similar reasons, whenever r is a group without arbitrarily small sub-
groups. Case (c) can be treated similarly, with the help of prop. 16.
group y = n
Instead of verifying the condition in proposition 17 for all ~ in the
(k x n gv), it is clearly enough to verify it for a set of gener-
ators of y; in this connection, the following result is occasionally useful:
PROPOSITION 18. Notations being as in proposition 17, assume that k
n
is an algebraic number-field, and call t its maximal order. Then the group
y = (k x ngv) is generated by ynt.
Take any ~EY, and write ~t=ba-t, where a,b are two ideals in t,
e
prime to each other. For every finite place VEP, is in r~, so that Pv is
not a prime factor of a or of b. Apply corollary 1 of tho 1, Chap. V-2,
n
to the projection of k onto the product rv taken over the finite places
v of k which either belong to P or correspond to the prime ideals divid-
ing a; it shows that there is OCEt such that OCEgv for every finite VEP,
OCEa, and (X+O; then (X2 satisfies the same conditions and is in gv for every
infinite place v, so that it is in y, hence in ynt. That being so, also ~ (X2
is in ynt; this proves our proposition.
In particular, assume that gv = 1+ p:;a(v), with m(v) ~ 1, for every finite

n
place VE P; put m = ,,:;a(v); let V1'"'' Vp be all the real places of k for
which gll=R~. Then one sees at once that the set ynr, in prop. 17,
consists of the elements of r which are == 1 (m) and whose image in
k"j is >0 for 1 ~ i ~p.
Chapter VIII
Traces and norms

§ 1. Traces and norms in local fields. In §§ 1-3, we will consider
exclusively local fields (assumed to be commutative). We denote by K a
local field and by K' an algebraic extension of K of finite degree n over K.
If K is an R-field and K' =/= K, we must have K = R, K' = C, n = 2; then, by
corollary 3 of prop. 4, Chap. 111-3, 'Irc/R(x)=x+x and NC/R(x)=xx;
TrC/R maps C onto R, and NC/R maps ex onto R:, which is a subgroup
of R x of index 2.
From now on, until the end of § 3, we assume K to be a p-field and
adopt our usual notations for such fields, denoting by q the module of K,
by R its maximal compact subring, by P the maximal ideal in R, and by
n a prime element of K. The field K' being as stated above, we adopt
similar notations, viz., q', R', P', n', for K'. We write f for the modular
degree of K' over K and e for the order of ramification of K' over K, as
defined in def.4 of Chap. 1-4; then q' = qf and n= ef, by corollary 6 ofth. 6,
Chap. 1-4. As e=orddn), the R'-module generated in K' by PV=n V R, for
any ve Z, is p,ev; for this, we will write I (PV).
By corollary 1 of prop. 4, Chap. 111-3, and the remarks following that
proposition, 'IrK'IK is +0 if and only if K' is separable over K; then, being
K -linear, it maps K' onto K. By the definition of the norm, and by corol-
lary 3 ofth. 3, Chap. 1-2, we have, for all x'eK':
(1)
In view of tho 6 of Chap. 1-4, this implies that x' e R' if and only if
N K'IK(x')eR, and x'eR' x if and only if N K'IK(x')eR x. As mod K(n)=q-l
and moddn')=q- f, (1) may also be written as follows, for x'=/= 0:
(2) ordK(NK'IK(x'»)=j-orddx').
From now on, we will write Tr, N instead of 'IrK'IK, N K'IK' except when
there are more fields to be considered than K and K'. For every veZ, we
will write m(p'V) = pfV; by (2), this is the R-module generated in K by the
image of p'v under N.
PROPOSITION 1. Let K' be separable over K. Then, if x' eR', Tr(x')eR;

if x'eP', Tr(x')eP and N(1 +x')= 1 + 'Ir(x') + y with yeRnx'2 R'.
140 Traces and norms VIII
Let K be an algebraic closure of K' ; call Ai' ... , An the distinct K -linear
isomorphisms of K' into K; then, by corollary 3 of prop. 4, Chap. 111-3,
we have
(3) 7r(x') = ~);(x'), N(1 + x') = n(1 + A;(X'»).
; ;
Call K" the compositum of the fields A;(K'), which is the smallest Galois
extension of KinK, containing K'; define R", P" for K" as R, Pare
defined for K. By corollary 5 ofth. 6, Chap. 1-4, we have A;(R')cR" and
A;(P') c P" for all i, so that 7r(x') is in R" if x' e R', and in P" if x' e P' ; as
the same corollary shows that R=KnR" and P=KnP", this proves our
assertions concerning 7r. Now assume x'eR', x'+O, and put
y=N(1 + x')-I-7r(x');
by (3), this is a sum of monomials of degree ~ 2 in the A;(X'). As one of the
A; is the identity, and as the Ai' by corollary 2 of prop. 3, Chap. 111-2, differ
from one another only by automorphisms of K" over K, all the A;(X')
have the same order as x' in K", so that yx,-2 is in R" if x' is in R'. As
R' = K' nR", this proves our last assertion. In view of the fact that 7r =
if K' is inseparable over K, and of the remarks about that case in Chap.
°
111-3, our proposition is still valid (but uninteresting) in the inseparable
case.
COROLLARY. If x'ep,-e+l, 7r(x')eR.
By definition, e = ordK,(n); therefore our assumption amounts to

nx'eP', which implies 7r(nx')eP by prop. 1, hence 7r(x')eR since 7r is
K-linear.
DEFINITION 1. Let K' be separable over K; let d be the largest integer
such that fr(x')eR for all x'ep,-d. Then p,d is called the different of K'
over K, and d its differental exponent.
For the different, we will write D(K'/K), or simply D. If K' is insepa-

rable over K, 7r is 0, so that it maps P' - Y into R for all v; in that case we
put d= + 00, D(K'/K)=O.
By the corollary of prop. 1, we have d ~ e -1. In particular, if d = 0,
e = 1, so that K' is unramified over K. The converse is also true; this will
be a consequence of the following results:
PROPOSITION 2. Let K' be unramified over K; call p, p' the canonical

homomorphisms of R onto k=R/P, and of R' onto k'=R'/P', respectively.
Then,for x'eR', we have
p(7r(x'») = 7rk'/k(P'(X'»), p(N(x'») = Nk'/k(P'(X'»).
§l. Traces and norms in local fields 141
As in tho 7 of Chap. 1-4 and its corollaries, call M' x the group of
roots of 1 of order prime to p in K' ; by corollary 2 of that theorem, K' is
cyclic of degree f over K, and its Galois group is generated by the Fro-
benius automorphism, which induces on M' x the permutation Jl. -+ Jl. q.
In view of corollary 2 of tho 2, Chap. 1-1, this amounts to saying that the
automorphisms of K' over K determine on k' = R'/P' the automorphisms
which make up its Galois group over k. Our conclusion follows at once
L n
from this, the formulas 'Ir(x') = Aj(X'), N (x') = Aj(X') and the similar
ones for k and k', i.e. from corollary 3 of prop. 4, Chap. III-3, applied
first to K and K', and then to k and k'.
PROPOSITION 3. Let K' be unramified over K. Then 'Ir maps P'v surjec-
tively onto Pv for every VE Z, and N maps R' x surjectively onto R x •
Let k, k' be as in prop. 2. As k' is separable over k, 'Irk'ik is not 0; the

first formula in prop. 2 shows then that the image 'Ir(R') of R' under 'Ir is
not contained in P; as it is contained in R by prop. 1, and as it is an R-
module since R' is an R-module and 'Ir is K-linear, it is R. As K' is un-
ramified, a prime element 1t of K is also a prime element of K' ; therefore,
for VEZ, p,v=1t vR'. As 'Ir is K-linear, we get
As to the norm, put Go=Rx, G~=R'x, Gv=l+pv and G~=I+P'v for

all v ~ 1. The last assertion in prop. 1 shows that, for every v ~ 1, N maps
G~ into Gv , and also, in view of what we have just proved about the trace,
that it determines on G~/G~+ 1 a surjective morphism of that group onto
GJG v+ l' On the other hand, call qJ the Frobenius automorphism of K'
over K, and Jl. a generator of the group M'x of the roots of 1 of order
prime to p in K'; then p. is of order q' -1, i. e. t/ -1, and its norm is given by
N(Jl.)= n !1"'= n Jl.q'=p.1+q+"'+ql-I=Jl.(ql-l)/(q-l);

/-1
j=O
/-1
j=O
clearly this is a root of 1 of order q - 1, hence a generator of the group M x

of roots of 1 of order prime to p in K. As M x is a full set of representatives
of cosets modulo G1 = 1+ P in Go = R x, this shows that N determines on
G~/G'1 a surjective morphism of that group onto GO/G 1 • Now, for every
x~ ER x, we can determine inductively two sequences (xv), (x~) such that,
for all v~O, xvEGv, x~EG~, N(X~)ExvGv+l and Xv+ 1 =N(X~)-1 xV' Then,
for y~ = x~ X'I ... x~ _ l' we have N(y~) = Xo X; 1. Clearly the sequence
(y~) tends to a J;mit y' ER' x, and N (y') = xo.
COROLLARY. Let K' be any extension of K of finite degree. Then the

different of K' over K is R', i. e. d = 0, if and only if K' is unramified over K.
Proposition 3 shows that d = 0 if K' is unramified over K. Conversely,

if d=O, K' is separable over K, and then, as we have already observed
above, corollary 1 of prop. 1 gives e = 1.
PROPOSITION 4. Let K' be separable over K, and let p,d be its different
over K. Then,for every VEZ, the image of p'v under Tr is pll, where Ji is
such that eJi~ v+d <e(Ji+ 1).
As Tr is K-linear and not 0, it maps every K-Iattice in K', and in parti-
cular every set P", onto a K-Iattice in K, i.e. onto a set of the form pll.
If Ji is as stated in Ol'r proposition, then, since ordx-(n)=e, p'v is con-
tained in nil p'-d and contains nil + 1 p,-d-1. In view of the definition of d
and of the K-linearity of Tr, this implies that Tr(P'V) is contained in
nil R=PIl and not in nil + 1 R=PIl+1. This completes the proof.
COROLLARY 1. For every x' E K' x , we have:
orddTr(x'))=e.ordK(Tr(x'))~ ordx-(x') +d -e + 1.
In fact, if we put v = ordx-(x'), and if we define Ji as in proposition 4,

the left-hand side of the inequality in our corollary is ~ e Ji by that propo-
sition, and the definition of Ji shows that this is > v + d - e.
COROLLARY 2. Tr(R') = R if and only if d = e-1.

In fact, by proposition 4, Ji = v = 0 implies d < e. As d ~ e - 1 by the
corollary of prop. 1, we get d = e-1.
If d = e -1, one says that K' is tamely ramified over K.
COROLLARY 3. Let X be a character of K of order Ji; then XO Tr is a

character of K' of order d+ eJi.
Our assumption means that Xis trivial on P-Il and not on P-Il-1. Put
v=d+ eJi; proposition 4 shows that Tr(P'-V)=P-1l and that Tr(P'-v-1)=
=P-Il-1. Therefore xoTr is trivial on P'-V and not on P'-v-t, which is
what we had to prove.
In the next corollary, we introduce an algebraic extension K" of K'
of finite degree; R", P" will have the same meaning for K" as R, P have
for K. For every VEZ, we will write 1'(P'V) for the R"-module generated in
K" by P'v, which is p"e'v if e' = ordK,,(n') is the order of ramification of K"
over K'. With these notations, we have:
COROLLARY 4. Let K, K', K" be as above; let D=P'd, D' = p"d', D" = p"d"
be the differents of K' over K, of K" over K' and of K" over K, respectively.
Then D" = 1'(D)·D' and d" =e' d+d', where e' is the order of ramification of
K" over K'.
§ 2. Calculation of the different 143
This is trivially so if K" is inseparable over K, since then D" = 0 and

either D or D' must be 0; we may therefore assume that K" is separable
over K, and, putting c5=e' d +d', we have to prove that d" =c5. In fact, by
proposition 4, IrK"IK' maps P"-~ onto p,-d and p"-~-l onto p'-d-l,
and IrK'IK maps p'-d onto Rand p,-d-1 onto p-1. Our assertion follows
at once from this and from the "transitivity of traces", i. e. corollary 4 of
prop. 4, Chap. III-3.
COROLLARY 5. Let K and K' be as above, and let K 1 be the maximal
unramified extension of K, contained in K'. Then K' has the same different
over K as over K 1.
For the definition of K 1 , cf. corollary 4 ofth. 7, Chap. 1-4. Our asser-
tion follows then at once from corollary 4, combined with the corollary
of prop. 3.
PROPOSITION 5. Let K, K' be as in proposition 4; then the norm N
determines an open morphism of K' x onto an open subgroup of K x .
As before, call p~d the different of K' over K, and put Gv= 1+ r,
G~=l+P'v for v~1. Take any J1.>2d, and put v=eJ1.-d. By prop. 4,
Ir(P'V)=P/.I; moreover, we have e(J1.-1)~2d, hence 2v~e(J1.+1), hence
p,2v cn /.l+1 R', and therefore Knp,2v c P/.I+1. That being so, the last part
of prop. 1 shows, firstly, that N maps G~ into G/.I' and secondly that it deter-
mines a surjective morphism of G~ onto G,JG/.1+ l ' Take now any xoEG/.I;
we can choose inductively two sequences (x j ), (x;), so that, for all i~O,
xjEG/.I+j, xiEG~+ej, N(xi)ExjG/.I+j+ 1 and Xj+1 =N(xi)-l Xj' Then, putting
y;=x~x'l ... xi, we have N(y;)=x O Xj+1 -1. Clearly the sequence (y;)
converges to a limit Y'EG~, and N(y')=x~. This shows that N maps G~
onto G/.I' which proves our proposition, since the groups G/.I' G~, for
fl > 2 d, v = e fl- d, make up fundamental systems of neighborhoods of 1 in
K x and in K' x, respectively. By using corollary 2 of prop. 4, Chap. 1-4, and
the results of Chap. 111-3, it would be easy to show that the conclusion of
our proposition remains valid for any extension K' of K of finite degree,
separable or not. Obviously it is also valid for R-fields.
§ 2. Calculation of the different. Let assumptions and notations be

as in § 1. When K' is regarded as a vector-space of dimension n over K,
R' is a K-Iattice, to which we can apply tho 1 of Chap. 11-2. This shows
that there is a basis {a 1 , ... ,an } of K' over K, such that R' = IRaj.
Now assume that K' is separable over K, so that Ir is not 0; then, by
lemma 3 of Chap. 111-3, we may identify K', as a vector-space over K, with
its algebraic dual, by putting [x',y'] = Ir(x' y'); the dual basis {P1'''' ,Pn}
to {a 1'"'' an} is then the one given by Ir( aj Pj) = c5ij for 1 ~ i,j ~ n.
PROPOSITION 6. Let K' be separable over K; call D = p'd its different.

Let {cx l,' .. , cxn} be a basis of K' over K such that R' = L R CXi' and let
{Pl, ... ,Pn} be the basis of K' over K given by Tr(cxiPj) =biJor 1 ~i,j~n.
Then D- l =p,-d= LRPi'
In fact, take any x' ER', any Y'EK', and write x' = LXiCXi and y' = LYiPi
with XiER and YiEK for 1 ~i~n. Then 'It(X'y') = LXiYi; this shows that
'It(x' Y')E R for all x' E R', i. e. that Tr maps R' y' into R, if and only if YiE R
for all i. By the definition of the different, this means that y' is in p'-d if
L
and only if it is in R Pi' as was to be proved.
COROLLARY. Let assumptions be as in proposition 6, and call L1 the

determinant of the matrix
M = ('It(CXiCXj))l .;; i, j';;n'
Then ord K(L1)=fd, and L1R = 9l(D).
Write CXi= Laijpj, with aijEK for 1 ~i,j~n. Multiplying both sides
with CXj and taking the trace, we get 'It(cxiCXj)=aij, hence M = (aij)' There-
fore the automorphism of the vector-space K' over K which maps
{Pl' ... , Pn} onto {CXl' ... , cxn}, hence the K-latticeD- l onto R', is represent-
ed by the matrix (aij) with respect to the first one of these bases, and its
module, by corollary 3 ofth. 3, Chap. 1-2, has the value mod K (L1). As the
mapping x'-+n'd x' also maps D- l =p'-d onto R', its module modK,(n,d)
must be the same as mod K (L1). This givesfd=ord K(L1), hence 9l(D)=L1R.
One will note that our corollary remains valid in the inseparable case,
since then 'It=O and D=O. Clearly our result implies that ordK (L1) is
independent of the choice of CX l , ... ,CXn ; this could easily be verified directly,
and justifies the following definition:
DEFINITION 2. Let L1 be as in the corollary of proposition 6; then the

ideal L1 R in R is called the'discriminant of K' over K.
Still assuming K' to be separable of degree n over K, call K an alge-

braic closure of K'. As in § 1, let Al, ... ,An be the n distinct K-linear iso-
morphisms of K' into K; as the identity is one of them, we may assume
e e e)
that it is Al . Take any E K', and put i = Ai( for 1 ~ i ~ n, hence in parti-
e e.
cular 1 = If v is the degree of K' over K (e), there are v distinct K (e)-
linear isomorphisms of K' into K, hence v and no more than v distinct
e
ones among the Ai which map onto itself. This shows that K(e)=K' if
e
and only if ed=- for all i =1= 1.
Take now an indeterminate X over K. We can, in the manner described
in Chap. III-3, extend the K-linear mapping 'It of K' into K, and the
§ 2. Calculation of the different 145
polynomial mapping N of K' into K, to mappings of K'[ X] = K' ®K K [X]

into K[X], which we again denote by Tr and N. Put then:
n (X -ei)=Xn+i=1L ai Xn - i.
n n
(4) F(X)=N(X -e)=

i=1
This is a monic polynomial in K[X]; calling F' its formal derivative, we
have
n (e-eJ
n
F'@=
i=2
In particular, in view of what has been proved above, we have K@=K'
ifand only if F'(e) +0. It is well known, and easily verified, that F(X)-1
has in K(X) the "partial fraction decomposition" given by
1 n 1
F(Xt t1 F'(ei)(X -ei)·
Considering the field K(X) as embedded in the obvious manner in the
field offormal power-series in X - 1 with coefficients in K, we get from this:
n n + ao
X-n(l +LaiX- i)-1 = L F'(e i)-1 L er X- v- 1
i= 1 i=1 v=o
which may also be written as
x- v~ Ct1 aix-')v= v~ Tr(F,@-1e )X-

n V V
-
1.
Equating coefficients on both sides, we get

(5)
for V~O, where Pv(a) is, for all v, a polynomial in Z[a1, ... ,an], with
Pv=O for O~v<n-l, and Pn- 1= l.
PROPOSITION 7. Let K' be separable of degree n over K, and call D
its different. For any ~EK', let F be the polynomial defined by (4). Then
all the coefficients aj of F are in R if ~ER', and in P if ~EP'; moreover, if
~ER', F'(~)D-1 is contained inR[~], and it is the largest R'-module con-
tained in R [ ~].
The assertions about the aj are proved exactly as the assertions about
the trace in prop. 1. In fact, if the ei=Ai@ are defined as above, the as-
sumption eER' (resp. ~EP') implies that, for every i, ei is in Ai(R') (resp.
in Ai(P')), hence in the maximal compact subring R" ofthe compositum K"
of the fields Ai(K') (resp. in the maximal ideal P" of R"); (4) shows then
that all the a j are in R", hence in R = K nR" (resp. in P", hence in P = K nP").
As to the assertions about F'(~}, assume first that F'<e}=O; as we have

+
seen, this is so if and only if K(~} K'; then K<e}, hence also R[ ~], cannot
contain any R'-module other than {O}, which proves our assertion in this
case. Assume now that F'<e}+O; then K'=K(~}, so that {1,~, ... ,~n-1}
is a basis of K' over K. As F is monic and in R[X], and F(~}=O, a well
n-1
known elementary argument shows that R[~] is the R-module L R~i.
i=O
n-1
Take now any x'eK'; write F'(~}x'= L
Xi~i with xieK for O~i~n-1.
i=O
Multiplying this with F'<e}-l ~v and taking the traces of both sides, we get,
in view of (5):
n-1
(6) 1r(x' ~v}= L XiPv+i(a}
i=O
for all v ~ 0, hence in particular, for 0 ~ v ~ n - 1 :
n-1
(7) xn _ v _ 1 =1r(x'e V
}- L XiPv+i(a}.
i=n-v
Assume first that x' e D - 1; then (7) shows, by induction on v for 0 ~ v ~ n -1,
that all Xi are in R, i.e. that F'<e}x' is in R[e], so that F'(e)D- 1 c:R[e].
On the other hand, assume that XieR for 0~i~n-1, i.e. that
F'(e)x'eR[e]; then (6), for v=O, shows that 1r(x')eR. Replacing X' by
x'y' with y'eR', we see that, if X' is such that F'<e}x'R'c:R[~], then
x' e D - 1. This proves our last assertion.
COROLLARY 1. Assumptions and notations being as in proposition 7, we
have D=F'(e)R' if and only if R' =R[e].
This follows at once from the second part of proposition 7.

COROLLARY 2. Let assumptions and notations be as in proposition 7;
assume also that K' is fully ramified over K; put
n
F(X) = N(X -n,}=xn+ L aiXn- i,
i=1
where n' is any prime element of K'. Then ordK(ai)~ 1 for 1 ~i~n,
ordK(an)= 1, and D=F'(n')R'.
e
Taking = n' in proposition 7, we get the first assertion; the second
one is obvious in view of formula (2) of § 1, since an = N ( -n'); the
last one follows at once from corollary 1, combined with prop. 4 of
Chap. 1-4.
§ 3. Ramification theory 147
COROLLARY 3. Let assumptions and notations be as in corollary 2;

then K' is tamely ramified if and only if n is prime to p.
As K' is fully ramified, we have, in our usual notation, f = 1 and
n=e. By corollary 2, we have d=ordK,(F'(n'»), and all the terms in F'(n')
except the first one nntn - 1 are of order ~e=ordK,(n) in K'. Therefore
d=e-l, i.e. K' is tamely ramified, if and only if orddn)=O, i.e. if and
only if n is prime to p.
A polynomial F satisfying the conditions in corollary 2, i.e. a monic
II
polynomial X" + I aiX n - i in K[X] such that ordK(ai)~ 1 for all i and
i= 1
ordK(all) = 1, is called an Eisenstein polynomial over K.
PROPOSITION 8. Let F be an Eisenstein polynomial over K. Then F is
irreducible in K[X], and, if n' is a root of F in any extension of K, K(n')
is a fully ramified extension of K, having n' as a prime element.
Assume that F = G H, with G and H in K[ Xl Let a, b be the smallest
integers such that G 1 = 7t" G and H 1 = nb H are in R[ X], and put F1 =
=na+b F, so that F1 =G 1H 1. Put k=RIP, and call Fo, Go, Ho the poly-
nomials in k[X] obtained by replacing each coefficient in F1 , G1 , H 1,
respectively, by its image in RIP under the canonical homomorphism of
R onto RIP. By the definition of a and b, Go and Ho are not 0, so that
Fo=foO; this implies that a+b=O, F1 =F, and Fo = X"; consequently
there is v such that Go=XV, Ho=X"- v. Then the degrees of G1, H1
are at least v, n-v; as F1 =G 1H 1, they are v, n-v. If v>O, n-v>O,
call g, h the constant terms in G1, H 1; as Go=Xv and Ho=X"- v,
g and h are both in P; as the constant term of F is now g h, it is in p2,
which contradicts the definition of an Eisenstein polynomial. Now
let n' be a root of F in an extension of K, which we may assume to be
algebraically closed; as F is irreducible, the distinct K-linear isomor-
phisms of K' =K(n') into that extension map n' onto all the distinct roots
of F, so that F(X)=N(X -n'), hence, by the definition of an Eisenstein
polynomial, ord K(N(n'») = 1. By formula (2), § 1, this implies that f = 1
and that n' is a prime element of K'.
§ 3. Ramification theory. In this §, it will be convenient to write

isomorphisms and automorphisms of fields exponentially, i.e. as X--+XA,
etc. Furthermore, K being as before, it is convenient to extend ordKto all
algebraic extensions of K as follows. Let x' be any element of such an
extension; let K' be any extension of K of fmite degree, containing x';
n being as before a prime element of K, put
ordK(x') = ordK,(x')/orddn);
here, if we replace K' by any similar field K" containing K', orddx')
and ordK,(x) are both multiplied with the order of ramification of K"
over K', so that our definition of ordK(x') is independent of the choice of
K' ; of course one could take K' = K (x') in that definition. That being so,
ordKcoincides on K x with the mapping ord Kof K x into Z, as previously
defined, and determines a mapping of every algebraic extension of K into
Qu{ + oo}, with ordK(x') = + 00 if and only if x' =0.
As before, let K' be an extension of K of degree n, which we assume to
be separable; let notations be as in §§ 1-2; in particular, let D=p,d be the
different of K' over K. Call Kl the maximal unramified extension of K
contained in K', this being uniquely defined by corollary 4 ofth. 7, Chap.
1-4. Then K' has the degree e over K 1 , and, by corollary 5 ofprop.4, § 1,
it has the different Dover K l' Put
F(X)=NK'/K,(X -x');
by corollary 2 of prop. 7, § 2, this is an Eisenstein polynomial over K t,

and D=F'(x')R'.
Let L be any Galois extension of K of finite degree, containing K'; for
instance, one may take for L the compositum of the images of K' under all
the distinct K-linear isomorphisms of K' into some algebraic closure of
K'. For every K-linear isomorphism X'-+X,A of K' into L, put
v(A)=minx'eR,orddx'-x'A)=minx'eR,ordL(x' -x'A)/ordL(x').
Since orddx' - X'A) is an integer ~ 0 or + 00, this is well defined; it is + 00
if and only if A is the identity; the identity, i. e. the natural injection of K'
into L, will be denoted bye. By tho 7 of Chap. 1-4 and its corollaries 3 and
4, Kl is generated over K by the roots of 1 of order prime to p in K', and
these, together with 0, make up a full set of representatives for R'/P' in R';
therefore, if A does not induce the identity on K 1, there is such a root C
for which CA =1= C, and then C- CA is in K' and not in P', so that, taking
x' = C, we get V(A)=O. Now assume that A induces the identity on K t .
As K' is fully ramified over K 1, prop. 4 of Chap. 1-4 shows that R' = R 1 [x'],
R 1 being the maximal compact subring of K 1, so that every x' e R' can be
written as G(x') with GeRt[X]' This gives:
x' - X,A =G(x') - G(X'A) = (x' _X'A) H(x', X,A)
with H eRl [X, Y]. As we have already observed in the proof of prop.
1, X,A has the same order as x' in L; this implies that orddx'A) = ordK,(x') =
= 1, so that we have
ordx-{x' - X,A) ~ ordK,(x' _X'A) ~ 1
and therefore, whenever 1 induces the identity on K 1 :

(8)
which implies that orddn'-n').) does not depend upon the choice of n'.
Now, F being as defined above, we have, by formula (4) of § 2:
F(X) = n(X -n').),

).
where the product is taken over all the distinct K 1-linear isomorphisms
A of K' into L, and therefore
F'(n') = n (n'-n').)
).1'-
where the product is now taken over the same isomorphisms except the
identity. This gives
d=ordK,(F'(n'»= L v().),
).1'-
where the sum is taken over those same isomorphisms, and also, since the
number of such isomorphisms is e - 1 :
d-e+l= L (v().)-l).
).1'-
As V(A)=O when A does not induce the identity on K 1 , it amounts to the

same to write:
(9) d= L V(A), d-e+l= L (v().)-l)+,
where the sums are now taken over all the distinct K-linear isomorphisms
of K' into L, other than the identity; moreover, the number of terms > 0
in the latter sum is ::;;; e - 1.
If K' is itself a Galois extension of K, we may take L=K', and the
isomorphisms). are the automorphisms of K' over K; they make up the
Galois group 9 of K' over K. The definition of V(A) shows that it is now an
integer or + 00; if A=F s, V(A) is the largest of the integers v such that A
determines the identity on the ring R/r. For every v~O, the automor-
phisms ). of K' over K for which v().) ~ v make up a subgroup 9v of 9; we
have 90 = 9, and the groups 9v for v ~ 1 are known as "the higher ramifica-
tion groups" of K' over K. As we have seen above, 91' which is traditio-
nally known as "the group ofinertia" of K', consists ofthe automorphisms
of K' which induce the identity on K 1 ; in other words, it is the subgroup
of 90 = 9 attached to K 1 in the sense of Galois theory; it is of order e, and
90/91 may be identified with the Galois group of K1 over K, which, as we
know, is cyclic of order f and generated by the Frobenius automorphism
of Kl over K.
Still assuming K' to be a Galois extension of K, call gy the order of the

group gy for each v ~ O. Then gy -1 is the number of the elements A. of g,
other than e, for which v(A.)~ v. We can therefore rewrite (9) as follows:
+00 +00
(10)
y= 1 y=2
PROPOSITION 9. Let K' be a Galois extension of K with the Galois

group 9 = go; let the gy, for v ~ 1, be its higher ramification groups. Put
Go = R' x and G~ = 1+ P,y for v ~ 1. Then,for each v ~ 1, gy consists of the
elements A. of g1 such that X,A X,-1 is in G~-1; when that is so, the image
y(A.) of X'A X,-1 in the group ry=G~_dG~ is independent of the choice of
the prime element x' of K i , and A. - y(A.) is a morphism of gy into r ywith the
kernel gy+ 1.
The first assertion follows at once from (8) and the definitions. Re-
place x' by another prime element of K' ; this can be written as x' u with
ueR'x ; for A.eg y , this modifies X,A X,-1 by the factor uAu-l, which, by the
definition of gy, is in 1 + P,y, i. e. in G~; this shows that y(A.) is independent
of the choice of x'. If A., p. are in gy, put U=X'A x'-l, v=X'''X,-1. Then
X,A"X'-1=(U"U- 1)uv; as ueR'x, u"u- 1 is in G~; this shows that A.-y(A.)
is a morphism. It is then obvious that its kernel is gy+1.
COROLLARY 1. For every v~O, gv/gy+1 is commutative; for v=O, it is

cyclic of order f; for v = 1, it is cyclic, and its order eo divides q' -1,
q' being the module of K'; for v ~ 2, it is isomorphic to a subgroup of
the additive group of R'jP', and its order divides q'.
For v=O, this was proved above. Now put k'=R'/P'; this is a field
with q' elements. The canonical morphism of R' onto k' induces on Go a
morphism of Go onto k' x with the kernel G~, so that r1 is cyclic of order
q' -1. Similarly, for v ~ 2, the mapping x' -1 + x,y - 1 x' of R' onto G~ - 1
determines an isomorphism of R' jP' onto ry. Our assertions for v ~ 1 are
immediate consequences of these facts and of proposition 9.
COROLLARY 2. Assumptions and notations being as in corollary 1, we

have e=eopN with N~O and eo prime to p.
This is obvious in view of corollary 1, since g1 is of order e.
COROLLARyl If v(A.) has the same value v for all A.+e in g, 9 is com-
mutative, with an order dividing q - 1 if v = 1 and q if v ~ 2.
In fact, we have then gy=g, gy+1={B}; moreover, ifv~1, we have
e=n, hence f=1 and q=q'.
Finally, the numbers V(A) have important "transitivity properties". As

above, let K' be a separable extension of K of finite degree n, but not
necessarily a Galois extension; let K" be a separable extension of K' of
finite degree; take for L a Galois extension of K of finite degree, con-
taining K". Notations for K and K' being as before, let K2 be the maximal
unramified extension of K, contained in K" ; call K~ the compositum of
K' and K 2. Call e' the order of ramification of K" over K', and I' its
modular degree over K'. As K' has the same module q' as K i , and K" and
K~ have the same module as K 2, K2 is the unramified extension of
Ki of degree 1', and K~ is the maximal unramified extension of K'
contained in K" and is of degree I' over K'. As K' is of degree e over K 1,
this implies that K~ is of degree el' over K i , hence of degree e over K 2.
Each K 2 -linear isomorphism (I of K~ into L induces on K' a Ki-linear
isomorphism A of K' into L; as K~ is the compositum of K' and K 2, two
sQch isomorphisms (I, (I' cannot coincide on K' unless (I = (I'; as there
are e such isomorphisms, and the same number of K i-linear isomorphisms
of K' into L, (I-+A is a bijection of the former onto the latter; in particular,
each isomorphism A of K' into L, inducing the identity on K 1, can be
uniquely extended to an isomorphism (I of K~ into L, inducing the iden-
tity on K 2 •
Now, calling Tt" a prime element of K", put:
e'
G(X) = NK"/K'z(X -Tt")=Xe' + L a.ixe'-i~
i=i
By corollary 2 of prop. 7, § 2, this is an Eisenstein polynomial over K~;
in particular, a.e, is a prime element of K~; so is Tt', since K~ is unramified
over K'. Let Abe any isomorphism of K' into L, other than the identity,
inducing the identity on K i ; as we have seen above, this can be uniquely
extended to an isomorphism (I of K~ into L, inducing the identity on K 2 •
Write Ga for the polynomial obtained by applying (I to each coefficient of
G; we have
e'-i
G(X)-Ga(x)=a.e,-a.:, + L (a.i-a.f)Xe'-i.
i=i
As a.e , and Tt' are prime elements of K~, and K~ is unramified over K', we
have, by what we have proved above:
ordK'z(a.i - a.f)~ ordK'z(Tt' -Tt'a) = V(A) (1 ~ i ~ e'),

and therefore:
orddG(Tt") - Ga(Tt")) = V(A).
We have G(n")=O. On the other hand, G" is the monic polynomial whose
roots are the images n'" of n" under the distinct isomorphisms r of K"
into L which coincide with (T on K~. In other words, we have
G"(n") = TI (n" - n"'),
where the product is taken over all the distinct isomorphisms r of K"
into L which induce A on K' and the identity on K 2 • Let now Y'(r) be de-
fined for K, K" and r, just as Y(A) has been defined for K, K' and A; in
other words, we put Y'(r) =0 if r does not induce the identity on K 2 , and
if it does, we put
Since ord K " = e'· ord K " we get now, by comparing the above formulas:
(11) e'v(A)=Lv'(r),
t
where the sum may be taken over all the isomorphisms r of K" into L
which coincide with A on K', since those which do not induce the identity
on K 2 make no contribution to the right-hand side; for a similar reason,
(11) remains valid when A is an isomorphism of K' into L which does not
induce the identity on K l ' Combining formulas (9) and (11), one gets
another proof for corollary 4 of prop. 4, § 1.
Let now L be a Galois extension of K, not necessarily of finite degree.
Call m its Galois group, topologized in the usual manner, i. e. by taking,
as a fundamental system of neighborhoods of the identity, all the sub-
groups of m attached to extensions of K of finite degree, contained in
L. Then m is compact, and (11) and (9), together with corollary 4 of
prop. 4, § 1, may be interpreted by saying that there is a finitely additive
function H, on the family of all open and closed subsets of m, with the
following property. Let K' be any extension of K of finite degree, con-
tained in L; let e be its order of ramification over K, and d its differental
exponent over K; call ~ the open and closed subgroup ofm, consisting
of the elements of m which induce the identity on K'. Then H(~) = die,
and, for every coset ~A of ~ in m, other than ~, we have H(~A) = - v(A)le,
where V(A) is as defined above. From this, we derive a linear form I -+ H(f),
i. e. a "distribution", on the space of all locally constant functions I on m,
by putting H(f)= H(~A) whenever I is the characteristic function of
~ A, where A. is any element of m, and ~ is as above; as all locally constant
functions on m can be written as finite linear combinations of such
characteristic functions, this determines H uniquely. We will call H the
* Herbrand distribution on m. In view of the foregoing results, it is clear
that its knowledge implies the full knowledge of the ramification pro-
§ 4. Traces and norms in A-fields 153
perties of K" over K' whenever K ' , Kif are of finite degree over K, and
KcK' cK" cL.
§ 4. Traces and norms in A-fields. In this §, we consider an A-field k

and a separably algebraic extension k' of k, of finite degree n over k.
Notations will be as explained in Chap. IV.
THEOREM 1. Let k be an A-field and k' a separable extension of k of
finite degree. Then, for almost all finite places w of k', k~ is unramified
over the closure kv of k in k~.
Let Xbe a "basic character" for k, i.e. a non-trivial character of kA' tri-
vial on k. Put X' = XO 'Jrk'/k; this is a character of kA' trivial on k'. As
'Jrk'/k is not 0, and as it is k-linear on k', there is eEk' such that 'Jrk'/k(e) = 1.
As the extension of 'Jrk'/k to kA is kA-linear, this implies that it maps kA
surjectively onto kA' so that X' is not trivial on k).. Let w be a finite place
of k', and v the place of k lying below w; call XV' X~ the characters respec-
tively induced by X on kv and by X' on k~. By corollary 3 of tho 1, Chap.
° °
IV-I, we have X~ = Xv o 'Jrk;"/k v ' By corollary 1 ofth. 3, Chap. IV-2, Xv is of
order for almost all v, and X~ is of order for almost all w; our conclu-
sion follows now immediately from this and from corollary 3 of prop. 4,
§1.
COROLLARY. Let assumptions be as in theorem 1; then Nk'/k is an open
morphism of k'; onto an open subgroup of kl.
By corollary 3 ofth. 1, Chap. IV-I, Nk'/k induces Nk;"/k v on k~x for all
places w of k'. By prop. 5 of § 1, this is, for all w, including the infinite
places, an open morphism of k~x onto an open subgroup of by k: ;
theorem 1, combined with prop. 3 of § 1, it maps r~x onto r: for almost all
w. In view of the corollary of prop. 2, Chap. IV -3, our assertion follows
immediately from these facts.
If kv and k~ are as above, k~, being generated over kv by k', is separable
°
over kv, so that, if v and hence ware finite places, its different over kv is
not and may be written as p~(w), with d(w)~O. Thisjustifies the following.
definition:
DEFINITION 3. Let k, k' be as in theorem 1; for every finite place w of k',
let p~(w) be the different of k~ over the closure kv of k in k~. Then, by the
n
different of k' over k, we understand the ideal p~(w) of k' if k, k' are of
characteristic 0, and the divisor Id(w)'w of k' if they are of characteristic
p> 1; it will be denoted by bk'/b or by b if no confusion can arise.
We will now consider separately the cases of characteristic
characteristic p> 1.
° and of
PROPOSITION 10. Let k be an algebraic number-field, k' a finite alge-

braic extension of k, t and t' their maximal orders, and b the different of k'
over k. Then b- 1 is the set of the elements" e k' such that 'Ir( e,,) is in t for
all eet'.
Take first any eet' and any "eb- 1 ; then e"eb-l, which means, by
definition, that e"ek' and e"ep~-d(w) for all finite places w of k'. This
implies that 'Ir1:~/de,,)erv for all such places, and therefore, by corollary
3 ofth.1, Chap. IV-1, that 'Irk'/l:(e,,) is in knrv for all v, hence in t. Conver-
sely, assume" to be such that this is so for all eet'; take x' =(x~)ek~, and
put Z= 'Ir"'lk(X',,). Then, by corollary 3 ofth. 1, Chap. IV-1, z=(zv) is given
by
Zv= L 'Irk~/dx~,,).
wlv
Take a finite place v of k; by corollary 1 of tho 1, Chap. V-2, the projection

n
of t' on the product r~, taken over the places w lying above v, is dense
there. As Zv is in rv , by our assumption, whenever x' is in t', and depends
continuously upon x', it is therefore in rv whenever x~er~ for all wabove
v. This implies that 'Irk~/kv maps "r~ into rv, hence, by the definition of the
different, that" is in p:;d(w); as this is so for all w, " must be in b -1.
COROLLARY. If a' is any fractional ideal of k', the set of the elements
" of k' such that Tr", Ik( e ,,) is in t for all e e a' is the fractional ideal a' - 1 b - 1.
In fact, in view of proposition 10, this set consists of all the" such that
"a' cb- 1 •
Now we introduce two morphisms I, m
of the groups I(k), I(k') of
fractional ideals of k and of k' into each other, as follows. Consider again
the morphism a-+id(a) of k~ onto I(k), with the kernel Qoo=kA(P",»(,
which was defined in Chap. V-3; as pointed out there, we may use it to
identify I(k) with k~/Doo; we recall that Q oo is the group k~xnr:
consisting of the ideles (zv) such that IZvl v= 1 for all finite places v of k. If
the group Q:" is similarly defined for k', we may also identify I(k') with
k~)(/D:". Write now 1 for the natural embedding of k~ into k~)( ; by corol-
lary 1 ofth. 1, Chap. IV-1, this maps every z=(zv) in k~ onto the element
I(Z)=(Z~) of k~)( such that z~=zv whenever w lies above v; then IZvlv= 1
implies Iz~lw= 1, so that I(Z) is in Q:" ifand only if zeQ oo • This shows that I
determines an injective morphism of I(k) into I(k'), which we will call the
natural embedding of I(k) into I(k'), and which we will also denote by I;
with this notation, we have (id)ol=lo(id); this may be regarded as de-
fining the injection I of I(k) into I(k'). Clearly, if k" is an extension of k' of
finite degree, and if the morphisms z' of k~)( into k~)( and z" of kI into k~)(
§ 4. Traces and norms in A-fields 155
are defined just as I was defined for k~ and k~x , we have I" = I' ° I ; therefore
the corresponding relation holds true for the natural embeddings of I(k)
into I(k"), of I(k') into I(k") and of I(k) into I(k'). On the other hand, corol-
lary 3 ofth. 1, Chap. IV-I, combined with formula (1) of§ 1, shows that Nk'ik
maps Q'a, into Q oo ; therefore it determines a morphism of I(k') into I(k),
also known as the norm, which we will denote by 9lk'lk; we have(id)oNk'lk
=9lk'lko(id), and this may be regarded as defining 9lk'lk' If k" is as above,
we have 9l k "/k = 9lk'lko 9l k"lk" as an immediate consequence of the corre-
sponding relation for ordinary norms. Furthermore, if n is the degree of
k' over k, we have Nk'lk(X)=X n for all xEk, as an immediate consequence
of the definition of Nk'/k; this implies at once the corresponding relation
for the extension of Nk'ik to kA • For zEk~, we can write it as Nk'lk(/(Z)) = zn,
which implies that we have 9l k'lk(/(a)) = an for all aEI(k).
By tho 3 of Chap. V-3, I(k) and I(k') are the free groups respectively
generated by the prime ideals PV' p~ of r, r'. We will now describe the
morphisms I, 9l k'lk in terms of these generators.
PROPOSITION 11. For each finite place v of k, and each place w of k'
lying above v, call e(w) the order of ramification and f(w) the modular
degree of k~ over kv' Then we have;
l(Pv)= TI p~(w>, 9lk'lk(P~)=P[(W), Le(w)f(w)=n,
wlv wlv
where the product in the first formula, and the sum in the last one, are taken
over all the places w of k' lying above V.
The first formula follows at once from the definitions, and the second
one from the definitions, corollary 3 ofth. 1, Chap. IV-I, and formula (1)
of § 1. As to the last formula, since the degree of k~ over kv is e(w) f(w), it
is nothing else than corollary 1 ofth. 4, Chap. I1I-4; it is also an immediate
consequence of the first two formulas and of 9lk ,/k(1 (Pv)) = p~.
COROLLARY. Let k be an algebraic number-field, and a a fractional ideal
of k. Then 9l k /Q(a) is the fractional ideal9l(a) Z of Q, where 9l is the norm
as dejined in definition 5 of Chap. V-3.
This follows at once from the latter definition and from the second
formula in proposition 11, applied to the fields k and Q.
As every ideal in the ring Z is of the form mZ with mEN, every
fractional ideal of Q can be written in one and only one way as r Z with
rEQ, r>O; one may therefore identify the group I(Q) of fractional ideals
of Q with Q ~ = QX n R~, by means of the isomorphism r-+r Z of the
latter onto the former. Then the norm 9l of definition 5, Chap. V-3, be-
comes the same as 9lklQ as defined above.
PROPOSITION 12. Let X be the character of QA, trivial on Q, such that

XOO (x) = e( - x); let k be an algebraic number-field, and put x' = XO Tr k/Q ;
let a=(a v) be a differental idelefor k, attached to X'. Then av = 1 for every
infinite place v of k, andid(a) is the different bk/Q of k over Q.
The character X is the same which has been introduced in the first
part of the proof ofth. 3, Chap. IV-2; it was shown there that it is uniquely
determined by the condition stated above, and that Xp is of order 0 for
every place p of Q. Our first assertion is now an immediate consequence
of the definition of difTerental ideles in Chap. VII-2, combined with
corollary 3 ofth. 1, Chap. IV-1. Our last assertion is an immediate conse-
quence of the same results, combined with corollary 3 of prop. 4, § 1.
COROLLARY. Let k be as in proposition 12, and let D be its discriminant.
Then IDI = 9l(bk/Q )·
If a is as in proposition 12, we have lalA= IDI- 1, by prop. 6 of Chap.
VII-2. On the other hand, since av = 1 for all infinite places of k, the defi-
nition of 91 shows at once that lalA = 9l(i d(a)f 1; in view of proposition 12,
this proves our assertion.
Now we generalize the definition of the discriminant, i. e. definition 6
of Chap. V-4, as follows:
DEFINITION 4. Let k be an algebraic number-field, k' a finite extension
of k, and b the different of k' over k. Then the ideal '!l=91k'/k(b), in the
maximal order r of k, is called the discriminant of k' over k.
One should note that, according to this, the discriminant of k over
Q is not D, but the ideal D Z = IDI Z in Z. When the latter is given, D is
determined by D=( -1)'2IDI, as follows from the remark at the end of the
proof of prop. 7, Chap. V-4.
PROPOSITION 13. Let k, k', kIf be algebraic number-fields such that

k c: k' c: k"; let band 1), b' and '!l', b" and '!)" be the differents and the dis-
criminants of k' over k, of kIf over k', and of kIf over k, respectively. Then:
b" = l'(b)b', '!l" = !)n' 9l k'/k(!)'),
where I' is the natural embedding of J(k') into J(k"), and n' is the degree of
kIf over k'.
The first formula follows at once from the corresponding local result,
i. e. corollary 4 of prop. 4, § 1. The second one follows from this and defi-
nition 4, combined with the transitivity property of norms.
Now let k be an A-field of characteristic p> 1, and k' a separable
extension ofk of finite degree n. As a-div(a) is a morphism of k~ onto the
Traces and norms in A-fields 157
TI
group D(k) of divisors of k, with the kernel r~, we see, just as in the case
of number-fields, that the natural embedding of k ~ into k~x determines an
injective morphism IOf D(k) into D(k'), which we call the natural embedding:
of D(k) into D(k'). Similarly, the norm mapping Nk'ik of k:: into k~ deter-
mines a morphism of D(k') into D(k), which we denote by 6 k 'lk (the
notation 91 would be undesirable here, since the groups of divisors are
written additively). The properties of 1 and 6 are quite similar to those of
1 and 91 in the case of number-fields. In particular, we have 6 k'lk(I(0))=
= n 0 for every divisor 0 of k, and, with the same notations as in proposi-
tion 11:
l(V)= L e(w)'w, 6 k 'lk(W) = f(w)'v, L e(w)f(w)=n,
wlv wlv
the proof being the same as there. Let F q' F q' be the fields of constants of k
and of k', and let fo be the degree of the latter over the former. Then the
definition of f(w), and that of the degree of a place, give fodeg(w) =
f(w)deg(v), and consequently, at first for places, and then for arbitrary
divisors:
(12) deg (6k'lk(o'))=fodeg(o'), deg(l(o))=(n! fo)deg(o)
where 0' is any divisor of k', and 0 any divisor of k.

Ifb is the different of k' over k, we define the discriminant of k' over k as
being the divisor 6 k'lk(b) of k. With notations similar to those in prop. 13,
we have:
b" = I'(b)+ b', !)" =n'!) + 6 k'lk(!)').
PROPOSITION 14. Let k and k' be as above; let b be the different of k'
over k, and let c be a canonical divisor of k. Then the divisor 1 (c)+ b is a
canonical divisor of k'.
By the definition of a canonical divisor, there is a "basic character" X
for k, such that c =div(X). Then corollary 3 of prop. 4, § 1, combined with
corollary 3 of tho 1, Chap. IV-I, and with the definitions, shows at once
that the divisor of XO Trk'ik is I (c) + b.
COROLLARY. Let k, k' and b be as in proposition 14: let g be the genus of
k; let n be the degree of k' over k, and fo the degree of the field of constants
of k' over that of k. Then the genus g' of k' is given by
2g' -2=(n!fo)(2g- 2) +deg(b).
This follows at once from proposition 14, corollary 1 of tho 2, Chap. VI,
and the second formula (12). It implies that the degree of the different is
always an even integer; a more precise result will be proved in Chap.
XIII-12.
§ S. Splitting places in separable extensions. Assumptions and nota-

tions being as in theorem 1 of § 4, one can express that theorem by
saying that, for almost all places w of k', the degree of k~ over kv is equal
to its modular degree over kv' Therefore corollaries 2 and 3 of prop. 1,
Chap. VII-I, and corollaries 3 and 4 of tho 2, Chap. VII-5, are valid if
"degree" is substituted for "modular degree", provided one adds there the
assumption that k is separable over k o. We will now consider some
consequences of these results.
As before, let k be an A-field, k' a separable extension of k of finite
e
degree n, and v a place of k. We can write k' = k( e), where is a root of an
irreducible monic polynomial F of degree n in k[X]. Combining tho 4 of
Chap. 111-4 with prop. 2 of Chap. 111-2, we see that the places w of /.;'
which lie above v are in a one-to-one correspondence with the irredu-
cible monic polynomials dividing F in kv[ X]; if, for each such place W.
we call Fw the corresponding polynomial, the degree of k~ over kv is
equal to the degree of Fw; by tho 1 of § 4, that degree, for almost all v, is
equal to the modular degree of k~ over kv' We also see that the places w,
lying above v, for which k~ = kv are in a one-to-one correspondence with
the roots of Fin kv' By corollary 1 ofth. 4, Chap. 111-4, there are n distinct
places of k' lying above v if and only if k~ = kv for every such place w;
when that is so, one says that v splits fully in k' ; it does so if and only if F
has n distinct roots in kv' If L is a Galois extension of k, then, by corollary
4 ofth. 4, Chap. 111-4, the completions of L at the places of L lying above
v are all isomorphic; therefore, if Lu= kv for one such place u, v splits fully
in L. Let k' = k( e) be a field between k and L; then, if F is defined as above,
it splits into linear factors in L[ X], and the smallest Galois extension L'
of k, contained in L and containing k', is the subfield of L generated over k
by the roots of F in L. If now t is a place of 1: lying above v, 1:1 is generated
over kv by the roots of F, so that 1:, = kv if and only if v splits fully in k'; in
that case, as we have seen, it also splits fully in L'.
PROPOSITION 15. Let k', kIf be two extensions of k, both contained in a

separable extension L of k of finite degree. Let X be the set of the places v
of k such that k~ = kv for at least one place w of k' lying above V. If almost
all the places veX split fully in kIf, kIf is contained in k'.
We may assume that L is the compositum of k' and kIf. Call W the
set of the places w of k' such that the place v of k which lies below w
splits fully in kIf and that kv = k~. Let u be a place of Labove w, and t
the place of kIf below u; Lu is generated over kv by L, hence by k~ and
k7; therefore, if weW, Lu=kv; this shows that all the places in W split
fully in L. Now take a place w of k', not in W; call v the place of k below
W. If then kv=k~, v is in X, so that it must be in the finite subset of X,
§ 6. An application to inseparable extensions 159
consisting of the places in X which do not split fully in k". If kv =1= k~,
the degree of k~ over kv is > 1; by tho 1 of § 4, this is the same as the
modular degree, except for finitely many places. We have thus shown
that the modular degree of k~ over kv is > 1 for almost all the places w
of k', not in W. Applying now corollary 4 of th.2, Chap. VII-5, to k,
k' and L (instead of ko, k and k' of that corollary), we get k' = L, i. e.
k"ck'.
COROLLARY. Let k', k" be two Galois extensions of k, contained in
some extension of k of finite degree. Let S', S" be the sets of the places
of k which split fully in k' and in k", respectively. Then k' contains k"
if and only if S" contains almost all the places veS'.
If k' => k", it is obvious that a place of k which splits fully in k' does
the same in k". Conversely, as k' is a Galois extension, S' is the same
as the set X in proposition 15; our conclusion is now a special case of
that proposition. In particular, we see that k' must be the same as k"
if S', S" differ by no more than finitely many elements.
§ 6. An application to inseparable extensions. It will now be shown

that one of our main results, the isomorphism between kj. and (k'jk)A'
which was proved for separable extensions as theorem 1 of Chap.IV-1,
is still valid without the assumption of separability. For this, we need
a lemma:
LEMMA 1. Let k be an A-field of characteristic p> 1; then k is purely
inseparable of degree p over its image k P under the endomorphism x~xp.
By lemma 1 of Chap. III-2, we may write k as k=Fp(xo, ... ,XN)'

where Xo is transcendental over F P' and Xi is separably algebraic over
Fixo) for l~i~N. Then kP=Fp(xg, ... ,x~). Put k'=kP(xo)=
Fp(xo,xr, ... ,x~). As each Xi is purely inseparable over Fixf) and sepa-
rable over F p(xo), k is at the same time purely inseparable and separable
over k', so that k = k'; this implies that k is purely inseparable of degree
1 or p over kP. If k was the same as kP, it would contain an element y
such that yP = Xo. Clearly y cannot be in F p(x o), so that it is purely
inseparable over F p(x o); this contradicts the assumption that k is sep-
arable over F p(x o).
Now, in order to extend theorem 1 of Chap. IV-1 to the case of an
inseparable extension k' of k, it is clearly enough to show the validity
of th.4, Chap. III-4, in that case, since the latter alone is involved in
the proof of the former. We will first do this for a purely inseparable
extension of k of degree p. Let k' be such an extension; for any x'ek',
there must then be an integer n~O such that x'P"ek, and, if n is the
smallest such integer, the degree of x' over k is p"; as this must be ~p,
n is 0 or 1. This shows that k'Pckck', hence, in view of lemma 1, that
k=k'p. For that case, we prove the following:
PROPOSITION 16. Let k' be an A-field of characteristic p> 1; put
k=k'p. Then, above each place v of k, there lies one and only one place
w of k'; it is the image of v under the isomorphism X-+Xl/p of k onto k';
we have kv = (k~)P, and the kv-linear extension ~ v of the natural injection
of k' into k~ to Av=k'®kkv is an isomorphism of Av onto k~. Moreover, if
(X is a basis of k' over k, and (Xv' for each v, is the rv-module generated by
(X in A v, then, for almost all v, ~ v maps (Xv onto the maximal compact
subring r~ of k~.
Let v be a place of k, and w a place of k' lying above v. By the corollary

of prop. 1, Chap. 111-1, k~ is generated over kv by k', hence purely in-
separable of degree 1 or p over kv' In the former case, every element of
k must be a p-th power in kv; this is impossible, since k is dense in kv
and therefore contains at least one prime element of kv' Therefore, by
corollary 2 of prop. 4, Chap. 1-4, k~ is uniquely determined, up to an
isomorphism, and y-+ yP is an isomorphism of k~ onto kv' Let A be the
natural injection of k' into k~; this must induce on k the natural in-
jection Ao of k into kv; therefore, for every eEk', we have Ao(ep)=A(e)p;
as this determines Am uniquely, we see that w is uniquely determined
by v, and also that it is the image of v under X-+Xl/p. If now ~v is as in
our proposition, it is clearly a surjective homomorphism of Av onto
k~; as both of these spaces have the dimension p over kv, it is an iso-
morphism. Finally, let (X be a basis of k' over k. In view of corollary 1
of th.3, Chap. 111-1, and of lemma 1 of Chap. 111-2, we may assume
that (X contains an element a such that k' is separably algebraic over
F p(a). Let then v and w be as above, and let u be the place of the field
ko=Fp(a) which lies below w. By th.1 of §4, for almost all w, k~ is
unramified over (k o)Il' Take w such that this is so; since tho 2 of Chap. 111-1
shows that ko has just one place u for which lal.,> 1, we may also assume
that w does not lie above that place. Then, by that theorem, there is a
polynomial nE F pET] such that n(a) is a prime element of (k o)Il' hence
also of k~ since k~ is unramified over (k o)Il' Now, by corollary 2 of tho 3,
Chap. 111-1, (Xv is' a compact subring of Av for almost all v; this implies
that it contains 1. hence r,,' 1. As it contains a, it contains n(a), hence
the ring rv[n(a)]; by prop. 4 of Chap. 1-4 and its corollary 1, this is
the same as r~.
Clearly proposition 16 implies the validity of th.4, Chap. 111-4,
when k = k'p. Now take for k' an arbitrary extension of k, of finite degree.
Call ko the maximal separably algebraic extension of k, contained in
An application to inseparable extensions 161
k'; let pm be the degree of k' over k~. If x' is any element of k', there is
n;;::O such that x,pnEk~, and, if n is the smallest such integer, x' is of
degree pR over k~, so that n~m; this shows that k'~k~~k,pm. Applying
lemma 1 to the sequence of fields k', k'p, ... ,k'pm, we see that each is of
degree p over the next one, so that k' is of degree pm over k'pm, which is
therefore the same as k~. Proceeding now by induction on m, we may
assume that theorem 4 of Chap. III-4 is valid for the extension k'P of
k, and we have to show that it is also valid for the extension k' of k. Put
kif = k'P; let v be a place of k; call w~, ... , w~ the places of kif lying above
v, and, for each i, call ki' the completion of kif at wi. By prop. 16, there
is, for each i, one and only one place Wi of k', lying above wi, and the
completion ki of k' at Wi may be identified with k'®k"ki'. By the in-
duction assumption, we have an isomorphism <P~ of A~=klf®kkv onto
the direct sum of the fields ki', with the properties stated in our theorem.
By the properties of tensor-products, the tensor-product Av = k' ®kkv
is canonically isomorphic, in an obvious manner, to k'®k"A~, hence to
the direct sum of the products k' ®k"ki' and therefore to the direct sum
of the fields ki; it is then easily seen that the isomorphism <Pv of Av onto
the latter sum which has been so defined has the properties required
by our theorem. As to the last part, it can be deduced in the same manner
from the induction assumption and prop. 16, by taking a basis a,' of kif
over k, a basis p of k' over kif, and taking for k' over k the basis a, con-
sisting of all the products a' b of an element a' of a,' and an element b of p.
Chapter IX
Simple algebras
§ 1. Structure of simple algebras. This Chapter will be purely algebraic
in nature; this means that we will operate over a groundfield, subject
to no restriction except commutativity, and carrying no additional
structure. All fields are understood to be commutative. All algebras are
understood to have a unit, to be of finite dimension over their ground-
field, and to be central over that field (an algebra A over K is called
central if K is its center). If A, B are algebras over K with these properties,
so is A ® K B; if A is an algebra over K with these properties, and L is a
field containing K, then AL = A ® K L is an algebra over L with the same
properties. Tensor-products will be understood to be taken over the
groundfield ; thus we write A ® B instead of A ® K B when A, B are algebras
over K, and A ® L or A L, instead of A ® K L, when A is an algebra over K
and L a field containing K, AL being always considered as an algebra
over L.
Let A be an algebra over K, with the unit 1A; all modules over A
will be understood to be unitary (this means, e.g. for a left module M,
that 1A • m = m for all mE M) and of finite dimension over K, when regarded
as vector-spaces over K by putting, e. g. for a left module M, ~ m = (~·1 A) m
for all ~ EK and mE M. IfM' is a subset of a left A-module M, the annihilator
of M' in A is the set of all XEA such that xm=O for all mEM'; this is a
left ideal in A. The annihilator of M in A is a two-sided ideal in A; if it is
{O}, M is called faithful.
DEFINITION 1. Let A be an algebra over K. An A-module is called simple
if it is not {OJ and has no submodule except itself and {OJ. The algebra A
is called simple if it has no two-sided ideal except itself and {O}.
For a given A, there are always simple left A-modules; for instance,
any left ideal of A, other than {OJ, with the smallest dimension over K,
will be such a module.
PROPOSITION 1. Let A be an algebra over K, with a faithful simple
left A-module M. Then every left A-module is a direct sum of modules, all
isomorphic to M.
We first prove our assertion for A itself, considered as a left A-module.
In M, there are finite subsets with the annihilator {OJ in A (e.g. any basis
§l. Structure of simple algebras 163
of Mover K); take any minimal set {m l , ... , mn} with that property.
For O~i~n, call Ai the annihilator of {m i+ l'"'' mn} in A; for i~ 1, put
Mi=Aim i. Clearly Ao={O}, An=A; for i~1, Ai~Ai-l' and Ai=fAi-l'
since otherwise xmj=O for j>i would imply xmi=O, and mi could be
omitted from {ml' ... , mn}. For i~ 1, Ai is a left ideal, Mi is a submodule
of M, and x ~ x mi induces on Ai a morphism of Ai onto Mi with the kernel
A i- l , so that it determines an isomorphism of AJA i - l onto Mi for their
structures as left A-modules. As Ai=fA i - l , Mi is not {O}; therefore it
is M. By induction on i for 0 ~ i ~ n, one sees now at once that
x~(xml' ... , xmi) induces on Ai a bijective mapping of Ai onto the
product Mi = M x .,. x M of i modules, all equal to M; this is obviously
an isomorphism for the structure ofleft A-module. For i=n, this proves
our assertion for A. Now take any left A-module M', and a finite set
{m'l' ... , m~} generating M' (e.g. any basis of M' over K). Then the
mapping of A' into M', given by (Xi)l"'i""~ L.xim;, is a surjective mor-
phism of left A-modules; as we have just proved that A, as such, is iso-
morphic to M n for some n, this shows that there is a surjective morphism
of Mnr onto M', or, what amounts to the same, a surjective morphism F,
onto M', of a direct sum of s=nr modules M;, all isomorphic to M.
Call N the kernel of F, and take a maximal subset {Mh , ... , MiJ of
{M l' ... , M.} such that the sum N' = N + L. M i" is direct; after renumber-
ing the Mi if necessary, we may assume that this subset is {Ml , ... , M h }.
Then, for j>h, the sum N'+ M j is not direct, so that N'nMj is not {O};
as it is a submodule of M j , which is isomorphic to M, it is M j • This shows
that MjeN' for allj>h. Therefore F maps N' onto M'; as its kernel is N,
h
it determines an isomorphism of L. Mi onto N'.
i= 1
PROPOSITION 2. Let A and M be as in proposition 1, and let D be the

ring of endomorphisms of M. Then D is a division algebra over K, and A
is isomorphic to M "(D) for some n ~ 1.
We recall that here, as explained on p. XV, D should be understood as

a ring of right operators on M, the multiplication in it being defined
accordingly. As D is a subspace of the ring of endomorphisms of the
underlying vector-space of Mover K, it is a vector-space of finite dimen-
sion over K. Every element of D maps M onto a submodule of M, hence
onto M or {O}; therefore, if it is not 0, it is an automorphism, hence
invertible. This shows that D is a division algebra over a center which is
of finite dimension over K. By prop. 1, there is, for some n ~ 1, an iso-
morphism of A, regarded as a left A-module, onto M"; this must deter-
mine an isomorphism between the rings of endomorphisms of these two
164 Simple algebras IX
left A-modules. Clearly that of M n consists of the mappings

(mj) 1 :!OJ:!O n-+(L mj d jj )1 :!OJ:!On
with dijE D for 1 ~ i,j ~ n, and may therefore be identified with the
ring Mn(D) of the matrices (dij) over D. On the other hand, an endo-
morphism of A regarded as a left A-module is a mapping f such that
f(xy)=xf(y) for all x, y in A; for y= 1..4" this shows thatfcan be written
as x-+xa with a= f(1..4,); the ring of such endomorphisms may now be
identified with A, which is therefore isomorphic to MiD). As the center
of MiD) is clearly isomorphic to that of D, this implies that the latter
is K, which completes the proof.
THEOREM 1. An algebra A over K is simple if and only if it is iso-
morphic to an algebra Mn(D), where D is a division algebra over K; when
A is given, n is uniquely determined, and so is D up to an isomorphism.
Let A be simple; take any simple left A-module M; as the annihila-

tor of M in A is a two-sided ideal in A and is not A, it is {O}; therefore M
is faithful, and we can apply prop. 2 to A and M; it shows that A is iso-
morphic to an algebra M n(D). Conversely, take A = Mn(D). For 1 ~ h, k ~ n,
call ehk the matrix (Xj) given by Xhk = 1, xij=O for (i,j) 1= (h,k). If a=(ajj)
is any matrix in Mn(D), we have ejjaehk = ajhejk for all i,j, h, k; this shows
that, if a1=O, the two-sided ideal generated by a in A contains all the ejk;
therefore it is A, so that A is simple. Let now M be the left ideal generated
by ell in A; it consists of the matrices (aij) such that ajj=O for j'~2;
if a is such a matrix, we have eija=ajlejl' which shows that, if a1=O,
the left ideal generated by a is M, which is therefore a minimal left ideal
and a simple left A-module. Let now f be an endomorphism of M
regarded as a left A-module, and put f(ell)=a with a=(aij), aij=O for
j~2. Writing that f(eijell)=eija, we get, for j~2, ajl =0; then, for
x=(xij) with xij=O for j~2, we get f(x)=f(xell)=xa=(xijall)' This
shows that the ring of endomorphisms of M is isomorphic to D. As
prop. 1 shows that all simple left A-modules are isomorphic to M, this
shows that D is uniquely determined by A up to an isomorphism. As
the dimension of A over K is n2 times that of D, n also is uniquely
determined.
We recall now that the inverse of an algebra A over K is the algebra A °
with the same underlying vector-space over K as A, but with the multi-
plication law changed from (x,y)-+xy to (x,y)-+yx.
°
PROPOSITION 3. Let A be an algebra over K; call A its inverse, and
put C=A(g)Ao. For all a, b in A, call f(a, b) the endomorphism x-+axb
of the underlying vector-space of A; let F be the K-linear mapping of C
§ I. Structure of simple algebras 165
into EndK(A) such that F(a®b)=f(a,b) for all a, b. Then A is simple if

and only if F maps C surjectively onto EndK(A); when that is so, F is
an isomorphism of C onto EndK(A).
One verifies at once that F is a homomorphism of C into EndK(A).
If N is the dimension of A over K, both C and EndK(A) have the dimen-
sion N 2 over K; therefore F is an isomorphism of C onto EndK(A) if
and only if it is surjective, and if and only if it is injective. Assume that A
is not simple, i.e. that it has a two-sided ideal I other than {O} and A.
Then, for all a, b,f(a, b) maps I into I; therefore the same is true of F(c)
for all CE C, so that the image of C under F is not the whole of EndK(A).
Assume now that A is simple, and call M the underlying vector-space
of A over K, regarded as a left C-module for the law (c,x)-+F(c)x. Any
submodule M' of M is then mapped into itself by x-+a x b for all a, b, so
that it is a two-sided ideal in A; as A is simple, this shows that M is
simple. An endomorphism <p of M is a mapping <p such that <p(axb)=
= a <p(x)b for all a, x, b in A; for x = b = 1A' this gives <p(a) = a <p(1 A)' hence
axb<p(1 A)=ax<p(lA)b, so that <p{1 A) must be in the center K of A; in
other words, <p is of the form x-+~x with ~EK. Call C' the annihilator
of M in C, which is the same as the kernel of F. We can now apply
prop. 2 to the algebra C/C', to its center Z, and to the module M; as D *
is then K, it shows that C/C' is isomorphic to some Mn{K), hence Z
to K; but then, as has been seen in the proof of tho 1, M must have the
dimension n over K, so that n=N. As C/C' has then the same dimension
N 2 over K as C, we get C' = {O}, which completes the proof.
COROLLARY 1. Let L be a field containing K. Then the algebra
AL =A®L over L is simple if and only if A is so.
In fact, let CL, FL be defined for AL just as C, F are defined for A
in proposition 3; one sees at once that CL=C®L, and that FL is the
L-linear extension of F to CL. Our assertion follows now from pro-
position 3.
COROLLARY 2. Let L be an algebraically closed field containing K.
Then A is simple if and only if AL is isomorphic to some Mn{L).
If D is a division algebra over a field K, the extension of K generated
in D by any ~ E D - K is an algebraic extension of K, other than K. In
particular, if L is algebraically closed, there is no division algebra over L,
other than L. Therefore, by tho 1, an algebra over L is simple if and only
if it is isomorphic to some Mn{L). Our assertion follows now from
corollary 1.
COROLLARY 3. The dimension of a simple algebra A over K is of
the form n 2•
In fact, by corollary 2, AL is isomorphic to some M II (L) if L is an

algebraic closure of K; its dimension over L is then n2 , and it is the
same as that of A over K.
COROLLARY 4. Let A, B be two simple algebras over K; then A®B
is simple over K.
Take an algebraic closure L of K; (A®B)L is the same as AL®BL.
Since clearly MII(K)®Mm(K) is isomorphic to M lIm (K) for all m, n, and
all fields K, our conclusion follows from corollary 2.
COROLLARY 5. Let A be a simple algebra of dimension n2 over K.
Let L be a field containing K, and let F be a K-linear homomorphism
of A into MII(L). Then the L-linear extension FL of F to AL is an iso-
morphism of AL onto MII(L).
Clearly FL is a homomorphism of AL into M II (L), so that its kernel
is a two-sided ideal in A L. As AL is simple by corollary 1, and as FL
is not 0, this kernel is {O}, i.e. FL is injective. As AL and M II (L) have
the same dimension n2 over L, this implies that it is bijective, so that
it is an isomorphism of AL onto MII(L).
COROLLARY 6. Let L be an extension of K of degree n; let A be a
simple algebra of dimension n2 over K, containing a subfield isomorphic
to L. Then AL is isomorphic to MiL).
We may assume that A contains L. Then (x,e)--+xe, for xeA, eeL,
defines on A a structure of vector-space over L; call V that vector-
space, which is clearly of dimension n over L. For every aeA, the map-
ping x--+ax may be regarded as an endomorphism of V, which, if we
choose a basis for V over L, is given by a matrix F(a) in M II (L). Our
assertion follows now from corollary 5.
PROPOSITION 4. Let A be a simple algebra over K. Then every auto-
morphism IX of A over K is of the form x --+a - 1 X a with ae A x •
Take a basis {al> ... ,aN} of A over K. Then every element of A®Ao
can be written in one and only one way as Laj®b j, with bjeA O for
1 ~i~N. By prop. 3, IX can therefore be written as x--+ Lajxb j. Writing
that IX(X y) = IX(X)IX(y) for all x, y, we get
0= Lajxybj- LajxbjlX(y)= Lajx(ybj-bjlX(y»).
For each yeA, this is so for all x; by prop. 3, we must therefore have
ybj=bjlX(y). In particular, since this gives y(bjz)=bjlX(y)z for all y and z
in A, bjA is a two-sided ideal in A, hence A or {O}, for all i, so that bj is
either 0 or invertible in A. As IX is an automorphism, the bj cannot all
be 0; taking a=bd=O, we get the announced result.
§l. Structure of simple algebras 167
COROLLARY. Let IX and a be as in proposition 4, and let a' E A be such

that a'lX(x)=xa' for all XEA. Then a'=~a with ~EK.
In fact, the assumption can be written as a' a - 1 X= Xa' a - 1 for all x;
this means that a' a- 1 is in the center K of A.
Proposition 4 is generally known as "the theorem of Skolem-
Noether" (although that name is sometimes reserved for a more com-
plete statement involving a simple subalgebra of A). One can prove,
quite similarly, that every derivation of A is of the form x-+xa-ax, with
aEA.
We will also need a stronger result than corollary 2 of prop. 3; this
will appear as a corollary of the following:
PROPOSITION 5. Let D be a division algebra over K, other than K.
Then D contains a separably algebraic extension of K, other than K.
We reproduce Artin's proof. In D, considered as a vector-space
over K, take a supplementary subspace E to K = K . 1D' and call cp the
projection from D=EffiK· 1D onto E. Then, for every integer m~ 1,
x-+cp(xm) is a polynomial mapping of D into E, whose extension to DL
and EL, if L is any field containing K, is again given by x-+cp(xm), where cp
denotes again the L-linear extension of cp to DL and EL. Now call N
the dimension of Dover K. Clearly every ~ED, not in K, generates
over K an extension K@ of degree> 1 and ::;;"N; moreover, if this is
not purely inseparable over K, it contains a separable extension of K,
other than K. Assume now that our proposition is not true for D. Then
K has inseparable extensions, which implies that it is of characteristic
p> 1 and that it is not a finite field; moreover, every ~ ED must be
purely inseparable over K, hence must satisfy an equation ~p"=xEK,
where p" is its degree over K. As this degree is ::;;,.N, it divides the highest
power q of p which is ::;;,.N, so that ~qEK. Then, if E and cp are as above
defined, the polynomial mapping x-+cp(xq) maps D onto O. As K is an
infinite field, this implies that the s~me holds true for the extension of
that mapping to DL and EL, when L is any field containing K. In other
words, for all L, x-+ x q maps DL into its center L· 1D • This is palpably
false when L is algebraically closed, for then DL is isomorphic to an
algebra M n(L), and taking e. g. x = ell in the notation of the proof of
tho 1, we have x q=e11' and this is not in the center of Mn(L).
COROLLARY. Let A be a simple algebra over K, and L a separably
algebraically closed field containing K. Then AL is isomorphic to an
algebra Mn(L).
The assumption means that L has no separably algebraic extension
other than itself. Then proposition 5 shows that there is no division
algebra over L, other than L. Our conclusion follows now at once from
tho 1, combined with corollary 1 of prop. 3.
§ 2. The representations of a simple algebra. Let A be a simple algebra

over K; by corollary 3 of prop. 3, § 1, its dimension N over K may be
written as N =n 2 • For any field L containing K, call rolL the space of
the K-linear mappings of A into Mn(L); every such mapping F can be
uniquely extended to an L-linear mapping FL of AL into MiL). If one
takes a basis tx = {a I"'" aN} of A over K, F is uniquely determined by
the N matrices Xj=F(aj), so that, by the choice of this basis, rolL is
identified with the space of the sets (Xj)1 :O:;j:O:;N of N matrices in Mn(L),
which is obviously of dimension N 2 over L.
By corollary 5 of prop. 3, § 1, a mapping Ferol L is an isomorphism
of A into Mn(L), and its extension FL to AL is an isomorphism of AL
onto Mn(L), if and only if F is a homomorphism, i.e. if and only if
F(1 A )= 1n and F(ab)=F(a)F(b) for all a, b in A, or, what amounts to
the same, for all a, b in the basis tx. When that is so, we say that F is an
L-representation of A; if we write K(F) for the field generated over K
by the coefficients of the matrices F(a) for all aeA, or, what amounts
to the same, for all aetx, then F is also a K(F)-representation of A.
If L is suitably chosen (for instance, by corollary 2· of prop. 3, § 1,
if it is algebraically closed, or even, by the corollary of prop. 5, § 1, if
it is separably algebraically closed), the set of L-representations of A
is not empty. Moreover, if F and F' are in that set, then F;'oFL I is an
automorphism of MiL), hence, by prop. 4 of§ 1, of the form X ~ y- I X Y
with YeMiLr; this can be written as F;.(FLl(X»)= y- 1 X Y; for
aeA, X =F(a), it implies F'(a) = y- I F(a) Y; we express this by writing
F' = y- I F Y. Moreover, when F and F' are given, the corollary of
prop. 4, § 1, shows that Y is uniquely determined up to a factor in the
center LX of Mn(Lr.
PROPOSITION6. Let A be a simple algebra of dimension n2 over K.
Then there is a K-linear form 7:=/=0 and a K-valued function v on A, such
that, if L is any field containing K, and F any L-representation of A,
7:(a)=tr(F(a») and v(a) = det(F(a») for all aeA; if K is an infinite field,
v is a polynomial function of degree n on A.
Put N =n 2 , and take a basis {al, ... ,aN} of A over K. Take first for L
a "separable algebraic closure" of K, i.e. the union of all separably
algebraic extensions of K in some algebraically elosed field containing K ;
this is always an infinite field. By the corol1ary of prop. 5, § 1, there is
an L-representation F of A, and then, as we have seen above, all such
representations can be written as F' = y- 1 F Y with YeMn(L) X. Clearly
a~tr(Fda») is an L-linear form 7: on AL, and a~det(Fda») is a poly-
§ 2. The representations of a simple algebra 169
nomial function v of degree n on AL; as FL is an isomorphism of AL

onto M n(L), l' is not 0; neither l' nor v is changed if F is replaced by
F'=y-l FY. Writing a=Lxia i with XiEL for l:(i:(N, we can write l'
and v as a linear form and as a homogeneous polynomial of degree n,
respectively, in the Xi' with coefficients in L. If (J' is any automorphism
of Lover K, we will write 1''', v" for the polynomials in the Xi' respec-
tively derived from 1', v by substituting for each coefficient its image
under (J'. Similarly, we write F" for the L-representation of A such that,
for each a in the basis {a 1 , ••• ,aN}' F"(a) is the image F(a)" of F(a) under (J',
i. e. the matrix whose coefficients are respectively the images of those
of F(a). Then, clearly, for all aEAv 1'''(a) and v"(a) are respectively the
trace and the determinant of F"(a); as we have seen above, they must
therefore be equal to 1'(a), v(a) for all aE AL. This implies that all the
coefficients in l' and v, when these are written as polynomials in the Xi'
are invariant under all automorphisms of Lover K, hence that they
are in K. This proves our assertion, so far as only L-representations
are concerned, with L chosen as above. Obviously it remains true for
L'-representations if L' is any field containing L. As every field containing
K is isomorphic over K to a subfield of such a field L', this completes the
proof.
The functions 1', v defined in proposition 6 are called the reduced
trace and the reduced norm in A. Clearlp(x y)= 1'(yx) and v(xy)= v(x)v(y)
for all X, y in A; in particular, v determines a morphism of AX into K x •
COROLLARY 1. Let A and v be as in proposition 6. Then, for every
aEA, the endomorph isms x~ax, x~xa of the underlying vector-space
of A over K have both the determinant NA/K(a)=v(a)n.
It is clearly enough to verify this for AL with a suitable L; taking L
such that AL is isomorphic to Mn(L), we see that it is enough to verify
it for an algebra Mn(L) over L; but then it is obvious. This is the result
announced in the remarks preceding th.4 of Chap. IV-3.
COROLLARY 2. Let D be a division algebra over K; let 1'o, Vo be the
reduced trace and the reduced norm in D. For any m~l, put A=Mm(D),
and call 1', v the reduced trace and the reduced norm in A. Then, for every
x=(xij) in A, 1'(X) = L1'O(X ii ); !f the matrix x=(x i) in A is triangular,
i
i.e. !f xij=O for 1 :(j<i:(m, v(x)= f1 VO(Xii)'
i
Take L such that D has an L-representation F. Then the mapping

which, to every matrix X = (xij) in M m(D), assigns the matrix obtained
by substituting the matrix F(x i) for each coefficient xij in X is an
L-representation of A. Using this for defining l' and v, we get at once
the conclusion of our corollary.
COROLLARY 3. Let assumptions and notations be as in corollary 2.

Then v(A X)=vo(DX).
We may regard A as the ring of endomorphisms of the space V = D m
considered as a left vector-space over D, and consequently A x as the
group of automorphisms of that space. By an elementary result (already
used in the proof of corollary 3 of tho 3, Chap. 1-2, but only for a vector-
space over a commutative field), every automorphism of V can be
written as a product of automorphisms, each of which is either a permu-
tation of the coordinates or of the form
(Xl, ... ,Xm)-+(LXjaj,X2,··.,Xm)

j
with a 1 ED x and ajED for 2~i~m. By corollary 2, the latter auto-

morphism has the reduced norm vo(a 1). As to a permutation of coordi-
nates, the same L-representation of A which was used in the proof of
corollary 2 shows at once that it has the reduced norm 1 if the dimen-
sion d2 of Dover K is even, and ± 1 if it is odd. As vo( -1 D )=( _1)4,
we have thus shown that v(A X) contains vo(D X) and is contained in it.
§ 3. Factor-sets and the Brauer group. Up to an isomorphism, the al-

gebras over a given field K may be regarded as making up a set, since
the algebra structures that one can put on a given vector-space over K
clearly make up a set, and every such space is isomorphic to K" for
some n.
From now on, we will consider only simple algebras over K; it is
still understood that they are of finite dimension and central over K.
Consider two such algebras A, A'; by tho 1 of § 1, they are isomorphic
to algebras M,,(D), M".(D'), where D, D' are division algebras over K
which are uniquely determined, up to an isomorphism, by A, A'. One
says then that A and A' are similar, and that they belong to the same
class, if D and D' are isomorphic over K. Clearly, in each class of simple
algebras, there is, up to an isomorphism, one and only one division
algebra, and there is at most one algebra of given dimension over K.
An algebra will be called trivial over K if it is similar to K, i. e. isomor-
phic to M,,(K) for some n. We will write CI(A) for the class of simple
algebras similar to a given one A.
Let A, A' be two simple algebras, respectively isomorphic to MiD)
and to M".(D'), where D, D' are division algebras over K. By corollary 4
of prop. 3, § 1, D®D' is simple, hence isomorphic to an algebra Mm(D"),
where D" is a division algebra over K which is uniquely determined,
up to an isomorphism, by D and D', hence also by A and A'. By the
§ 3, Factor-sets and the Brauer group 171
associativity of tensor-products, A ® A' is isomorphic to M nn'm(D"). This

shows that the class of A ® A' is uniquely determined by those of A
and A'. Write now:
Cl(A ® A') = CI(A) . CI(A'),
and consider this as a law of composition in the set of classes of simple

algebras over K. It is clearly associative and commutative;' it has a
neutral element, viz., the class CI(K) of trivial algebras over K. More-
over, if A 0 is the inverse algebra to A, prop. 3 of § 1 shows that A ® A 0
is trivial, so that CI(A 0) is the inverse of CI(A) for our law of composition.
Therefore, for this law, the classes of simple algebras over K make up
a group; this is known as the Brauer group of K; we will denote it by
B(K). If K' is any field containing K, and A a simple algebra over K,
it is obvious that the class of A K , is determined uniquely by that of A,
and that the mapping CI(A)-tCI(Ad is a morphism of B(K) into B(K'),
which will be called the natural morphism of B(K) into B(K').
It will now be shown that the Brauer group can be defined in another
way, by means of "factor-sets"; this will require some preliminary defi-
nitions. We choose once for all an algebraic closure K for K; we will
denote by Ksep the maximal separable extension of K in K, i.e. the
union of all separable extensions of K of finite degree, contained in K.
We will denote by (fj the Galois group of Ksep over K, topologized as
usual by taking, as a fundamental system of neighborhoods of the iden-
tity B, all the subgroups of (fj attached to separable extensions of K of
finite degree. Clearly this makes (fj into a totally disconnected compact
group. As K is purely inseparable over Ksep, each automorphism of
Ksep can be uniquely extended to one of K, so that (fj may be identified
with the group of all automorphisms of Kover K.
DEFINITION 2. Let (fj(m) be the product (fj x ... x (fj of m factors equal
to (fj; let ~ be an open subgroup of (fj. Then a mapping f of (fj(m) into
any set S will be called ~-regular if it is constant on left cosets in (fj(m)
with respect to ~(m).
This amounts to saying that f((11, ... ,(1m) depends only upon the left
cosets ~(11, ... ,~(1m determined by the (Ji in (fj. When that is so, f is
locally constant, or, what amounts to the same, it is continuous when S
is provided with the discrete topology. Conversely, let f be a mapping
of (fj(m) into S; if it is locally constant, it is continuous if S is topologized
discretely, hence uniformly continuous since (fj is compact; this implies
that there is an open subgroup ~ of (fj such that f is ~-regular.
DEFINITION 3. Let (fj(m) be as in dt;finition 2. Then a mapping f of (fj(m)

into Mn(Ksep), for any n~ 1, will be called covariant if it is locally constant
and sati~fies the condition
f(rr 1 A, ... ,rrm A)= f(rr 1, ... ,rrm).l.
for all rr 1' ... ,rrm, A in (fj.
LEMMA 1. Let ~ be an open subgroup of (fj; let L be the subfield of
K sep , consisting of the elements invariant under ~. Then an ~-regular
mapping of (fj into Ksep is covariant if and only if it is of the form rr--+e 17,
with eEL.
Let x, i.e. rr--+x(rr), be a mapping of (fj into K sep , and put e=x(e).
e e
If x is covariant, we have x(rr) = 17 for all rr; if this is ~-regular, must
be in L. The converse is obvious.
LEMMA 2. Let ~ be an open subgroup of (fj. Call X m the space of
~-regular covariant mappings of (fj(m) into K sep ,regarded as a vector-
space over K; call X~ the space of all ~-regular mappings of (fj(m) into
K sep , regarded as a vector-space over Ksep. Then X~ = X m ® K K sep , and
the dimension of X mover K, and of X~ over K sep , is nm, if n is the index
of ~ in (fj.
Let L be as in lemma 1 ; it has the degree n over K. Take a full set
a = {(Xl' ... ' (Xn} of representatives of the cosets ~ (X of ~ in (fj; then the
isomorphisms Al, ... ,An , respectively induced on L by the (Xi' are the n
distinct K-linear isomorphisms of L into Ksep. Any mapping XEX~ is
uniquely determined by its values on a x ... x a, and these can be chosen
arbitrarily; therefore X~ has the dimension nm over Ksep, and every
linear form L on X~ can be written as
L(x) = Lai, ... imX((Xil,···,(XiJ
(i)
with coefficients a(i) in Ksep. Now we proceed by induction on m. For

m= 1, lemma 1 shows that Xl, as a vector-space over K, is isomorphic
to L, hence of dimension n, so that we need only show that Xl generates
X~ as a vector-space over Ksep. If not, there would be a linear form
L on X~, other than 0, which would be 0 on Xl ; writing L as above,
and making use oflemma 1, we get 0= Laie i for all eEL; this contra-
dicts the linear independence of the Ai over Ksep, i. e. corollary 3 of
prop. 3, Chap. 111-2. Now, for any m, consider the tensor-product
Ym = Xl ® .... ® Xl, taken over K, of m factors equal to Xl' and the
similar product Y~ = X~ ® ... ® X~ taken over Ksep; as we have just
shown that X~ is the same as Xl ®KKsep, we may, in an obvious manner,
identify Y~ with Ym ®KKsep" Call q> the Ksep-linear mapping of Y~ into
X~ which, to every element Xl ®···®x m of Y~, assigns the mapping
(0" 1, .. ·, O"m)--+ X1 (0" 1)'" Xm(O"m)

of (fj(m) into Ksw This is surjective; for, if a linear form L on X~ is 0 on
qJ(Y~), we must have, for all X 1""'X m in X't:
0= Lait' .. imxl(lXiJ"xm(lXiJ,
(i)
which clearly implies that all the ali) are O. As Y~ has the same dimen-
sion nm as X~, this shows that qJ is an isomorphism of Y~ onto X~. Now
take a basis {.fl"" ,f,,} of X lover K. Then the nm elements /;1 ® ... ® .hm
make up a basis of Ym over K, hence also of Y~ over K sep , so that their
images under qJ make up a basis of X~ over Ksep. This amounts to
saying that every element of X~ can be written uniquely in the form
(0" 1, ... , O"m)--+ Lx it ·· 'i m/;1 (0" 1)'" /;JO"m)
(i)
with coefficients Xli) in Ksw Writing now that this is in X m, Le. that
it is covariant, we see that this is so if and only if all the Xli) are invariant
under (fj, i. e. if and only if they are all in K. Therefore qJ maps Ym onto
X m • This completes the proof.
Let no~ K' be any field containing K, and let K~ K~ep, (fj' be defined
for K' as K, K sep , (fj have been defined for K. As K is determined only
up to an isomorphism, we will always assume, in such a situation, that
we have taken for K the algebraic closure of Kin K'. It is obvious that
Ksep is then contained in K~w Every automorphism 0"' of K' over K'
induces on K an automorphism 0" of Kover K (more precisely, over
KflK'); clearly the mapping 0"'--+0" is a continuous morphism p of (fj'
into (fj; this will be called the restriction morphism; it is injective if K'
is algebraic over K, since then K' = K; in that case one will usually
identify (fj' with its image in (fj, which is always a closed subgroup
of (fj, and is open in (fj when K' is of finite degree over K. If D is any
open subgroup of (fj, and L is the corresponding subfield of K sep , Le.
the one consisting of the elements invariant under D, the subgroup
D' = p - 1 (D) of (fj' is open, and the corresponding subfield of K~.p is the
one generated by Lover K'.
Let notations be as above, and let f be as in definition 2, i. e. a map-
ping of (fj(m) into some set S. We will write fo p for the mapping
(O"~, ... , O"~)--+ f(P(O"'I)"" ,p(O"~))
of (fj,(m) into S. This is obviously continuous, i.e. locally constant, if f
is so; if f is D-regular, it is D'-regular, with D' = P-l(D); if S =M,,(Ksep),
and f is covariant, fop is covariant. If K' is algebraic over K, (fj' is a
subgroup of (fj, and p is its natural injection into (fj; then fop is the
restriction of f to (fj,(m).
After these preliminaries, we can now go back to our main topic.
THEOREM 2. Let A be a simple algebra of dimension n2 over K. Let D

be an open subgroup of (f;, L the corresponding subfleld of K sep , and F
an L-representation of A. Then there is an D-regular covariant mapping Y
r,
of (f; x (f; into M II (Ksep such that F~ = Y(p,(1)-1 FP Y(p,(1) for all p, (1
in (f;; if Y is such, there is an D-regular covariant mapping f of (f; x (f; x (f;
into K."ep such that, for all p, (1, -r in (f;:
(1) f(P, (1, -r) Y(p, -r) = Y(p, (1) Y((1, -r),
and this satisfies the condition
(2) f(p,(1, -r) f(v,p, -r)= f(v,(1, -r) f(v,p, (1)
for all v,p, (1, -r in (f;.
For every Ae(f;, FA is a Ksep-representation of A, hence of the form

r.
Z(A)-1 F Z(A), with Z(A)eMII (Kscp As FA depends only upon the left
coset DA, we may, to begin with, assume that A-+Z(A) is D-regular; it
would then be easy to verify that Y(p,(1)=Z((1p-l)p satisfies all the
conditions of the first part of our theorem, except possibly that of the
D-regularity. To obtain this, we refine our construction as follows. Take
a full set A of representatives of the double cosets D AD in (f; with respect
to D. For each AeA, F and FA are both L'-representations, if L' is the
compositum L· LA of L and its image LAunder A; choose then Z(A) in
M (L')X, so that F A= Z(A)-1 F Z(A). Each pe<D can be written as P=CXAP,
II
with a uniquely determined Ae A and with a, Pin D. If at the same time

we have p=a' AP' with a',p' in D, then:
P' p- 1=,\. -1(a'-1 a)'\',
so that, if we put y = P' p- 1, Y is both in D and in A-I DA, which implies
that it leaves fixed all the elements of L and of LA, hence L· LA and Z(A).
Therefore, if we put Z(p)=Z(A)P, this depends only upon p, not upon
the choice of a, p, subject to the conditions stated above. It is then easy
to check that Y(p,(1) = Z((1 p -1)p satisfies all the conditions stated in
our theorem. Now, for all p,(1,-r:
F' = Y((1, -r)-1 F~ Y((1, -r)= Y((1, -r)-1 Y(p,(1)-1 FP Y(p, (1) Y((1, -r).
At the same time, we have F'=Y(p,-r)-1 FPY(p,-r). As we have observed

above, this implies, by the corollary of prop. 4, § 1, that Y(p,(1) Y((1,-r)
differs from Y(p, -r) only by a scalar factor f(P,(1, -r), which proves (1).
One can then verify (2) by a straightforward calculation, and the re-
maining assertions are obvious.
COROLLARY. Let assumptions and notations be as in theorem 2; let

K' be a field containing K; 6)' being as before, let p be the restriction
morphism of 6)' into 6). Let FK , be the K'-linear extension of F to AK ,·
Then Yo p and fop are related to FK' in the same manner as Y and f to F.
This is obvious. When Y and f are related to an L-representation F

of A in the manner described in theorem 2, we will say that they belong
to A.
DEFINITION 4. A covariant mapping f of 6) x 6) x 6) into K:ep is
called a factor-set of K if it satisfies (2) for all v, p, a, r in 6).
Clearly the factor-sets of K make up a group (K) under multipli-

cation. If K', 6)' and p are as above, f --+ fop is obviously a morphism
of (K) into (K').
Let z be any covariant mapping of 6) x 6) into K:w Obviously the
mapping
(3) (p,a, r)--+z(p,a)z(a, r)z(p, r)-1
is covariant, and one verifies immediately that it is a factor-set.
DEFINITION 5. The factor-set defined by (3) will be called the co-

boundary of z ; a factor-set of K will be called trivial if it is the coboundary
of a covariant mapping of 6) x 6) into K:w
The trivial factor-sets make up a subgroup P(K) of the group (K)
of all factor-sets of K. The quotient (K)/P(K) will be denoted by H(K),
and its elements, i.e. the classes modulo P(K) in (K), will be called
the factor-classes of K. If K' and p are again as before, it is obvious
that f --+ fop maps coboundaries into coboundaries, so that p deter-
mines a morphism of H(K) into H(K'), which we again denote by p.
PROPOSITION 7. The factor-sets belonging to a simple algebra A over

K make up a factor-class of K.
Let i), L, F, Y and f be as in tho 2; let z be any covariant mapping

of 6) x 6) into K: ep ; let i)' be an open subgroup of i) such that z is
i)' -regular; let L' be the subfield of Ksep corresponding to i)'. Then F
is also an L~representation; Y' = z Y is related to F in the same manner
as Y, and it determines the factor-set f' = fof, where fo is the coboundary
of Z. This shows that all the factor-sets in the class determined by f
belong to A. On the other hand, let i)', L; F', Y', f' be related to A in
the same manner as i), L, F, Y and f. Put i)" = i)ni)', and call L" the
corresponding subfield of Ksep, which is the compositum of Land L:
Then there is ZeM,,(L1 X such that F' =Z-l F Z. A trivial calculation

gives now F'C1=W- 1 F'PW with W=(Zp)-lY(p,u)za, so that Y'(p,u)
can differ from W only by a scalar factor. If we write z(p,u) for this
factor, we have now
Y'(p,u)=z(p,U)(Zp)-l Y(p,u)za,
which implies that z is ~"-regular and covariant. Then I' f - 1 is the
coboundary of z, which completes the proof.
COROLLARY. Let K' be a field containing K; then the factor-class
of K' determined by AI' is the image of the factor-class of K determined
by A under the restriction morphism p of 6)' into 6).
This is obvious in view of the corollary of tho 2.
If A is a simple algebra over K, the factor-class of K, consisting
of the factor-sets belonging to A, will be said to belong to A or to be
attached to A.
THEOREM 3. The mapping which, to every simple algebra A over K,
assigns the factor-class of K attached to A, is constant on classes of
simple algebras over K and determines an isomorphism of the group B(K)
of such classes onto the group H(K) of factor-classes of K.
Take first two simple algebras A, A' over K; call n2, n,2 their dimen-
sions over K. Let L, F, Y and f be defined for A as in th.2, and let
L', F', Y', I' be similarly related to A'; call L" the compositum of L and
L'. We may identify M"(L")®M,,.(L") with M"".(L"). Then, if we put
A"=A®A', and if we write F"=F®F' for the K-linear mapping of A"
into M"".(L") given by F"(a®a')=F(a)®F'(a') for allaeA and a'eA',
F" is an L"-representation of A", and one sees at once that Y"= Y® Y'
and f" = fl' are related to A" and F" as in th.2. This shows that the
factor-class attached to A" is the product of those attached to A and
to A'. If A=M,,(K), one can take for F the identity mapping of A onto
M"(K) and then take Y = 1, hence f = 1; therefore the factor-class at-
tached to a trivial algebra is the trivial one, and the factor-classes at-
tached to A' and to M,,(A') are the same. This proves the first assertion
in our theorem and shows that the mapping J.l of B(K) into H(K) which
is thus defined is a morphism. It will now be shown, firstly that J.l is
injective, and then that it is surjective; this will be done in several steps,
which we formulate as lemmas.
LEMMA 3. Let ~ be an open subgroup of 6), and L the subfield of Ksep
corresponding to ~. Let Y be an ~-regular covariant mapping, of 6) x 6)
into M,,(K.ep)X, such that Y(p,r) = Y(p,O') Y(O', 1:) for all p, 0', 1: in <D. Then
there is ZeM,,(L)X such that Y(p,u)=(ZP)-l za for all p,u in 6).
§3. Factor-sets and the Brauer group 177
Take a full set a of representatives of the co sets ~Cl of ~ in ffi; as

we have observed in the proof of lemma 2, they induce on L all the
distinct K-linear isomorphisms of L into K sep , and these are linearly
independent over K sep , as has been shown in corollary 3 of prop. 3,
Chap. III-2. Let Mn(Ksep) operate on the right, by matrix multiplication,
on the space M1,n(K sep ) of row vectors over K sep , and similarly on the
left on column vectors. For each ueM1,n(L), put
z = L u" Y(Cl,e).
"EO
For any peffi, ap is again a full set of representatives of the cosets of ~

in ffi; as Y is covariant, we have then
zP= LU"P Y(Clp,p) = LU" Y(Cl,p);
" "
for pe~, this shows that z is invariant under ~, i.e. that it is in M1,n(L).
Therefore, if we write cp for the mapping u-+z defined above, cp maps
Ml n(L) into itself. Now we show that there are n vectors U1'''',U n in
Ml n(L), such that the vectors CP(Ui) are linearly independent over L.
In fact, if this were not so, there would be a column vector v in Mn 1(L),
other than 0, such that cp(u)v=O for all ueM1. n(L). This can be written
as LU"(Y(Cl,e)v)=O, which, in view of the linear independence of the Cl
on L, implies Y(Cl,e) v =0 for all Cl, hence v=O. Choose now n vectors Ui
such that the cp(u i) are linearly independent over L; call U the matrix
in Mn(L) whose rows are the Ui, and put Z= LU"Y(Cl,e). As the rows
of Z are the CP(Ui), Z is invertible in Mn(L). Just as above, we have, for
all p,u:
Z" = L u" Y(Cl,C1),
"
and therefore Z" = ZP Y(p, u) in view of the assumption on Y. This
shows that Z has the property stated in our lemma.
It is now easy to show that the morphism Jl. of B(K) into H(K) which
has been defined above is injective. In fact, assume a simple algebra A
over K to have a trivial factor-set; in view of prop. 7, this implies that
we can choose ~, L, F and Yas in th.2, so that (1) holds with f = 1.
Let now Z be as in lemma 3, and put F' = Z F Z - 1; then one sees at
once that F'" = F'P for all p, u in ffi. This means that F' is a K-represen-
tation of A, i. e. an isomorphism of A onto M II(K), so that A is trivial.
Finally, the surjectivity of Jl. is contained in the following more
precise result:
LEMMA 4. Let ~ and L be as in lemma 3; call n the degree of Lover K.
Let f be an ~-regular factor-set of K. Then one can choose A, F and Y
with the properties described in theorem 2, so that the factor-set defined
by (1) is the given one and that A contains a subfield isomorphic to L.
* This will be proved by an explicit construction, due to R. Brauer.

We first observe that, if we take v = p = fJ in the formula (2) of tho 2 which
defines factor-sets, we get f(p, p,'r)= f(p,p,p); as f is covariant, this
gives f(P,p;r)=a P with a= f(8,8,8). Now apply lemma 2 to the case
m=2; this gives two spaces X 2 ,X2 of dimension n2 , over K and over
Ksep respectively, and Xl = X 2 ®x Ksep. Take a full set tl of representa-
tives of the cosets f> 0( of f> in 6>. For any x, y in Xl' and any p, fJ in 6>, put
z(p,fJ) = :Lf(fJ,O(,p)x(p,O()y(O(,fJ).
",eo
Clearly z, i.e. the mapping (p,fJ)-+z(p,fJ), is in Xl' and it is in X 2 if X, Y

are in X2 ; more precisely, (x,y)-+z is a bilinear mapping of Xl x Xl
into Xl which induces on X 2 X X 2 a bilinear mapping of X 2 x X 2 into X 2'
It will now be shown that, if we write this as (x,y)-+xy, it makes X 2
into an algebra A with the required properties. In fact, for each PE6>,
and each x E Xl' put
cPp(x) = (f(P, 0(, p)x(O(, P))"" /lea'
After choosing an ordering on the set tl, we may identify mappings of
tlx tl into Ksep with the matrices in Mn(Ksep); then each cPp may be
regarded as a mapping of Xl into M n(K sep); as such, it is obviously
K sep-Iinear and bijective. Using formula (2), one verifies at once that
cPixy)= cPp(x) cPp(y) for all x, y in Xl' Call e the element of X; given
by e(p,fJ)=(aP)-l, with a= f(8, 8, 8), whenever fJ is in the same
coset f> p as p, and e(p, fJ) = 0 otherwise. Clearly e is in X 2; since
f(O(,O(,p)=a'" for all O(,p, we have cPp(e) = In. It is now obvious that, for
each p,cPp maps Xl' with the multiplication (x,y)-+xy, isomorphically
onto the algebra Mn(Ksep), the unit of the former being e. As Xl =
=X2 ®XKsep, this implies, by corollary 1 of prop. 3, § 1, that this multi-
plication makes X 2 into a simple algebra A over K, with 1A =e. For
any eEL, and any xEA, write ex for the element of X 2 given by
(p,fJ)-+e px(p,fJ); this defines on X 2 a structure of left vector-space
over L; moreover, it is clear that (ex)y=e(xy) for all eEL and all x, y
in A; therefore e-+ee is an isomorphism of L into A, and ex=(ee)x
for all eEL and all XEA.
We will now construct F and Y with the properties stated in our
lemma. For all p, fJ in 6>, call D(p,fJ) the diagonal matrix given by
D(p, fJ) = (b"'/I f(O(,p, fJ))"" /lea
with b"'/I = 1 or 0 according as 0( = P or not. Using (2), one verifies at

once that one has, for all p,fJ, and all XEX;:
(4) D(p, fJ) cPa (x) = cPp(x)D(P,fJ).
Now choose a basis {e!> ... ,ell} of Lover K, and call {'h, ... ,'7I1} the dual
basis to this when L is identified with its own dual by putting [e, '7] =
='IrL/K(e'7). As the elements IX of Q induce on L the n distinct K-linear
isomorphisms of L into K scp , we have, for every eEL, 'IrL/K(e) = Le'\
.
so that the definition of the '7i may be written as
c5ij= 'Ir L/K(ei'7j) = L ei'7j·

.. EO
Therefore we may put:
Write now, for each p, Fp=X~pX-l; this gives
Fp(x)=( L ei f(JJ,IX'P)X(IX,/3)'7~)l~i,j~II'
..,/lea
Assume that x EA, i. e. that it is covariant; as f is covariant, and as Q A,

for every A, is a full set of representatives of the cosets of ~ in (D, we have
Fp(x)). = ( L ei f(JJ,IX,pA) x(IX,/3)'7n 1 ~i,j~1I = Fp;'(x),

..,/lea
In particular, if we put F=F., we have, for every p, Fp=FP. By the defini-

tion of Fp, this gives FP=F for PE~, i.e. F(x)P = F(x) for every XEA and
every PE~; in other words, F maps A into MII(L), so that it is an L-
representation of A. For all pE (D, we have FP = X ~ pX- l • In view of (4),
this gives F" = Y(p, or1 FP Y(p,o), where we have put
Y(p,u)=X D(p,u)X- 1 = (L ei f(lX,p,U)'7j) 1 ~i,j~II'

Ilea
One can now verify at once that this, together with A and F, is as re-
quired by our lemma. We also note for future use that the reduced trace
r and the reduced norm v in A can be calculated by means of anyone of
the Kscp-representations ~p of A, e.g. from ~.; this gives, for all xEA,
r(x)=tr(~.(x)), v(x)=det(~.(x)), and in particular:
r(e ·l A )= TrL/K(e), v(e ·l A )=NL/K<e)

for all eEL.
With lemma 4, the proof of theorem 3 is now complete.
COROLLARY 1. Let K' be a field containing K, and p the restriction
morphism of (D' into (D. Let A be a simple algebra over K, and f a factor-set
belonging to A. Then AK , is trivial if and only if fop is so.
This follows at once from theorem 3 and the corollary of tho 2.
COROLLARY 2. Let ~ be an open subgroup of 0), and f an ~-regular

factor-set of K. Then f is trivial if and only if it is the coboundary of an
~-regular covariant mapping of 0) x 0) into K:ep •
Assume that f is trivial; construct A, F and Y as in lemma 4. By

theorem 3, A is trivial, so that there is an isomorphism F' of A onto
MiK). Then F=Z- l F'Z, with some ZEMn(L)X, hence F"=
= Y'(p,O')-l FP Y'(p,O') with Y'(p,O')=(ZP)-l Z". This implies Y(p,O')=
= z(p, 0') Y'(p,O'), where z is ~-regular and covariant. Then f is the co-
boundary of z.
COROLLARY 3. Let L be a separable extension of K of degree n. Let A
be a simple algebra over K. Then AL is trivial if and only if there is an
algebra A' of dimension n2 over K, similar to A, containing a subfield
isomorphic to L; when A' exists, it is unique, up to an isomorphism.
The last assertion is obvious. By corollary 6 of prop. 3, § 1, the
existence of A' implies the triviality of AI., hence that of AL • Conversely,
assume that there is an isomorphism of AL onto a matrix algebra Mn(L);
this induces on A an lrrepresentation F. By tho 2, we can construct an
~-regular factor-set f belonging to A. Then, by lemma 4, we can construct
an algebra A' such as required our corollary.
It is frequently convenient to identify the groups B(K) and H(K) by
means of the isomorphism J.l described in theorem 3. If this is done, and
if K' is any field containing K, the corollaries of tho 2 and of prop. 7
show that the natural morphism of B(K) into B(K'), which maps the
class of every. simple algebra A over K onto the class of A K " coincides
with the restriction morphism p of H(K) into H(K').
§ 4. Cyclic factor-sets. We will now discuss in greater detail a type

of factor-sets of particular importance, attached to the cyclic extensions
of the groundfield K. Here, as always, we understand "cyclic" as meaning
a Galois extension (hence, by definition, a separable one) with a finite
cyclic Galois group. With the same notations as in § 3, the cyclic exten-
sions of K are the subfields L of K corresponding to the open subgroups
~ of 0) with cyclic factor-group. If L and ~ are such, and if n is the degree
of Lover K, O)/~ is isomorphic to the group of the n-th roots of 1 in C;
any isomorphism of O)/~ onto the latter group may be regarded as a
character X of 0), with the kernel ~; such a character, which is of order n,
will be said to be attached to L. If a is a representative in 0) of a generator
ofo)/~, there is one and only one character Xof 0), attached to L, such that
x(a) = e(1/n).
Conversely, let X be any homomorphism of (fj into C x ; by lemmas 3
and 4 of Chap. VII -3, it is a character of (fj of finite order n; its kernel ~
§ 4. Cyclic factor-sets 181
is then an open subgroup of (fj, with a cyclic factor-group of order n,

and the subfield L of Ksep corresponding to ~ is cyclic of degree n over
K; we will then say that L is attached to X.
Let notations be as above; as X is locally constant on (fj, one can
choose, in infinitely many ways, a locally constant mapping tP of (fj into
R such that x(O')=e(tP(O'») for all O'e(fj. For instance, one may choose tP
so that 0 ~ tP(O') < 1 for all 0'; if tP is chosen according to this condition,
it is determined uniquely, and it is ~-regular, since X is so. In any case,
since X is of order n, tP maps (fj into (l/n)Z. Consider now the mapping
(5) (p,O', t)--+ e(p, 0', t)= tP(O' p-l)+ tP(tO'-l) - tP(t p -1)
of (fj x (fj x (fj into R; as tP is locally constant, this is so; as Xis a character,
one sees at once that e maps (fj x (fj x (fj into Z. Put now, for any (JeK x :
(6) f(p,O', t) = ee(p,a,t);
obviously, f is a covariant mapping of (fj x (fj x (fj into Kse; (more

precisely, into K X ), and one verifies at once that it satisfies condition (2)
in tho 2, § 3, i. e. that it is a factor-set. Any factor-set f defined in this
manner will be called a cyclic factor-set. Let tP' be another locally con-
stant mapping of (fj into R such that x(O')=e(tP'(O'») for all 0'; let f' be the
factor-set defined by tP' and e, just as f has been defined by tP and e.
Put 'P = tP' - tP; clearly 'P maps (fj into Z; putting z(p,O') = e'l'(a p- 1 ),
one sees at once that f'f - 1 is the coboundary of Z. This shows that the
class of the factor-set J, modulo the group P(K) of trivial factor-sets, is
uniquely determined by X and e; it will be denoted by {X, e}, and every
such factor-class will be called cyclic.
PROPOSITION 8. For each ee K x , X--+ {X, e} is a morphism of the group
of characters of (fj into the group H(K) of factor-classes of K; for each
character X of (fj, e--+ {X, e} is a morphism of K x into H (K).
This is obvious in view of our definitions.
Let K' be a field containing K ; as in § 3, we assume that K is contained
in K', and we denote by p the restriction morphism of (fj' into m, as well
as the morphisms for factor-sets and factor-classes derived from this in
the manner explained in § 3. If Xis any character of (fj, x' = xo p is a charac-
ter of (fj' ; if X is of order n, the order n' of x' divides n; if ~ is the kernel
of X, the kernel of X' is ~'= p - 1 (~), and p determines an injective mor-
phism of (fj'/~' into (fj/~; if L is the cyclic extension of K attached to X,
the cyclic extension of K' attached to X' is the compositum of Land K' ;
it is cyclic of degree n'. Then, for every eeK x , we have:
(7) {x,e}op = {xop, e}.
PROPOSITION 9. Let X be a character of <D, L the cyclic extension of K

attached to X, and A a simple algebra over K. Then AL is trivial if and only
if the factor-class attached to A can be written in the form {X, £I} with
8eK".
Call D the kernel of X; it is the subgroup of <D corresponding to L.
If the factor-class attached to A is {X, £I}, the one attached to AL is given
by (7) when one takes for p the restriction morphism of D into <D; then
Xop, being the character induced by X on D, is trivial, so that AL is trivial.
Conversely, assume that AL is trivial; then, if n is the degree of Lover K,
corollary 3 ofth. 3, § 3, shows that, after replacing A by an algebra similar
to A if necessary, we may assume A to be of dimension n2 over K. Let F be
an L-representation of A, induced on A by an isomorphism of AL onto
MII(L). As Xis of order n, we may choose ae<D such that x(a) = e(lln); then
<DID is generated by the image of a in that group. There is X e M II(L) " such
that F It =X- 1 F X, hence, by induction on i, F lti =Xi 1 F Xi if we put
Xi=XXIt ... X lti - 1
for all i > O. Take i = n ; as a" induces the identity on L, Fit" = F ; therefore
X /I must be ofthe form £I ·1/1 with £Ie If. Applying a to both sides of the
formula defining XII' we get X:=X- 1 XliX, hence 81t =fJ, so that £I is in
K". Take any ieZ and write it as i=nv+j with v,j in Z and 1 ~j~n; if
then we write X i =8vXi' one verifies easily that, for i>O, this coincides
with Xi as above defined, that X/lv=8 v·1 11 for veZ, and that Xi+i=XiXf
for all i, j in Z. Take now a locally constant function tP on <D such that
x(a)=e(tP(a»), hence ntP(u)eZ, for all u, and put Y(p,u)= (XII~(arl))P
for all p,a in <D. One verifies easily that Y, in relation to A and F, has the
properties required by tho 2 of § 3, and that the factor-set f determined
in terms of Y by (1) ofth. 2 is the one given by (5) and (6).
PROPOSITION 10. Let X and L be as in proposition 9. Then the kernel of

the morphism fJ-+{x,fJ} of K" into H(K) is NL1K(C).
In the proof of proposition 9, take A = M II(K); then we may take for

F the identity, and, as Fit = F, we may take X = e ·1/1 with any eeL".
Then 8=NL1K (e), and {X,8} is trivial, since A is so. Conversely, assume
8eK x to be such that {X,8} is trivial. Take tP such that x(a)=e(tP(a»)
and O~tP(a)< 1 for all a; then tP(ai)=iln for 0~i~n-1, and tP is D-
regular, so that, if we define f by (5) and (6), f is an D-regular factor-set.
As f is trivial, corollary 2 of tho 3, § 3, shows that it is the coboundary of
an D-regular covariant mapping Z of <D x <D into K:ep • As D is a normal
subgroup of <D, left cosets and right cosets of D in <D are the same; this
implies that, for all p, a in <D, z(p,u) is invariant under all AeD and is
therefore in LX. For all ae<D, put w(a)=z(B,a), and put Wi = w(a;i) for all i.
§4. Cyclic factor-sets 183
Then z(p, 0") = W(O" p-l)P. Write now that f(p,O",r), as given by (6), is equal
to the coboundary of z, as given by (3) of § 3, for p = e, 0" = ai, r = ai+ 1 ;
for 0~i~n-2, we get l=Wi(Wl)"IWi~L and, for i=n-l, we get
9=wn- 1 (w 1 )"n-l Wo 1. Therefore 9=N L/K(W 1 ), which completes the proof.
Let. X and L be as in propositions 9 and 10, ~ and a as in the proofs
of these propositions. If cP is chosen as in the proof of prop. 10, we have
f(a- j,a- i,e)=l or 9 according as i~j or i>j, and in particular
f(e, e, e) = 1. We now apply to this factor-set the construction described
in the proof oflemma 4, § 3, and define the algebra A as has been explained
there. As indicated above, the fact that here ~ is a normal subgroup of (fj
implies that every ~-regular covariant mapping of (fj x (fj into Ksep
maps (fj x (fj into L. For iEZ, define Ui as the mapping of (fj x (fj into L
given by Ui(p, 0") = lor 0 according as 0" p-l is in ~a-i or not; clearly UiEA
and Un + i = Ui for all i, and Uo is the same as the unit e = 1A of A. One finds
at once that, forO~i,j~n-l, UiUj=Ui+ j when i+ j~n-l and UiUj=9ui+ j
when i+ j~n. As in the proof of lemma 4, define ex, for eEL, XEA, as
e
given by the mapping (p, 0") ~ p x(p, 0") ; one finds at once that ex = (e·1A)X.
Similarly, define x e as given by the mapping (p, 0") ~ x(p, 0") ell"; then
x e = x( e . 1A)' Clearly e Ui = Ui e,,1 for all eEL and all i. As A has the dimen-
sion n2 over K, it has the dimension n over L when considered either as a
left vector-space, by (e,x)~ex, or as a right vector-space, by (e,x)~xe.
Moreover, {UO,Ul,""U n- 1 } is a basis for both of these spaces; in fact,
if we put X=IeiUi with eiEL for O~i~n-l, we have x(e,a-i)=ei' so
that x=o implies ei=O for all i, and a similar proof holds for A as a
right vector-space. Finally, as has been observed at the end of the proof
of lemma 4, § 3, one can use the isomorphism CPe of A into Mn(Ksep)
which was defined there, and which is now an L-representation of A,
for the calculation of the reduced trace r and of the reduced norm v in A.
Taking {e,a-I, ... , a- n+I} as the full set a of representatives of (fj/f) in G>
used in the definition of cP e' we get at once, for all eEL:
(8) r(e'l A )= 1r L/K(e); r(eui)=O (1 ~i~n-1),
(9) v(e'I A )= N L/K(e); v(u i)=( _1)i(n- il 9i (1 ~ i ~n-l).
PROPOSITION 11. Let L be a cyclic extension of K of degree n, and a

a generator of its Galois group over K. Let X be a left vector-space of
dimension n over L, with the basis {uo,u 1 , .•• , Un-I}' Then, for each 9EK x ,
there is one and only one K-bilinear and associative mapping (x,y)~xy
of X x X into X such that: (i) for all eEL and all XEX, ex=(euo)x and
XU o=x; (ii) ui=(ud for 1 ~i ~n-l; (iii) (u 1 )n= 9u o ; (iv) eU I =u 1 (e"u o).
This makes X into a simple algebra A over K, in which the reduced trace r
and the reduced norm v satisfy (8) and (9), and the factor-class of K attached
to A is {X,9} if Xis the character of (fj attached to L, such that x(a)=e(l/n).
As all this has been proved above for the algebra A which we con-
structed there, it only remains for us to show that the conditions (i) to (iv),
together with the associativity, determine the multiplication uniquely.
In fact, by induction on i, (iv) gives ~Ui=UiWiUO) for O~i~n-1. Then,
using (i) and the associativity of the multiplication, we get, for
o~ i, j ~ n -1 and for all ~, f/ in L:
(~Ui)(f/U) = (~UO)(Ui(f/UO»)Uj = (~UO)(f/",-1 Ui)Uj
=(~UO)(f/",-i uo)uiUj= ~f/",-I UiUj.
By (ti), UiUj=Ui+ j if i+ j~n-1; by (ii) and (iii), uiUj=OUi+ j-II if i+ j~n.

This shows that, using (i) to (iv) and the associativity, one can write
uniquely (~Ui)(f/U) in the form CUk with CeL, wluch completes the proof
of our proposition.
DEFINITION 6. Assumptions and notations being as in proposition 11,
the algebra A defined there will be called the cyclic algebra [L/K ;X,O].
An illustration for the above concepts, which will be considered more

closely in the following chapters, is provided by the division algebras
over a commutative p-field I(. In fact, prop. 5 of Chap. 1-4 may now be
interpreted as saying precisely that every such algebra D can be written
as a cyclic algebra [K tIK ; X, 7t], where K 1 is an unramified extension
of K, X a character attached to K l' and 7t a suitable prime element
of K. But now we can say more; prop. 10, combined with prop. 3
of Chap. VIII-1, shows that {X,~} is trivial for ~eR", so that {X,7t} is
independent of the choice of 7t ; so is [K tIK ; X, 7t], since there can be only
one algebra of given dimension over K in a given class, up to an isomor-
phism.
As a further illustration for the above theory, we will apply it to the
field K = R. We may then identify K with C, and (f) has only two elements,
z,
the identity e and the automorphism q of C given by z ..... and only one
non-trivial character X, given by X(q)= -1. The cyclic extension of R,
attached to X, is C. Combining now corollary 2 of prop. 3, § 1, corollary 3
of tho 3, § 3, and propositions 9 and 11, we see that every class of simple
algebras over R contains a cyclic algebra [C/R;x,O]. As the group
NC/R(C") is R~ and is of index 2 in R" , prop. 10 shows that there are
exactly two such algebras, up to isomorphism, viz., a trivial one and the
algebra H = [C/R ;x, - 1]. The latter is a division algebra; in fact, it is of
dimension 4 over R; writing it as Mil (D), where D is a division algebra
over R, and calling d2 the dimension of Dover R, we get nd = 2, hence
n = 1 since H is not trivial. Writing H in the manner described in prop. 11,
§ 5. Special cyclic factor-sets 185
we see that it has a basis over C consisting of two elements Uo= 1 and UI ,
hence a basis over R consisting of 1, i,j=UI and k=iu l ; it is then trivial
to verify that the multiplication table for 1, i, j, k is the well-known one
for the "quaternion units" in the algebra of "classical" quaternions.
§ 5. Special cyclic factor-sets. Now we apply the results of §4 to the

characters attached to "Kummer extensions" and to "Artin-Schreier
extensions" of K.
In the first place, let n be such that K contains n distinct n-th roots
of 1 ; then these make up a cyclic group E of order n; of course, if K is of
characteristic p> 1, our assumption implies that n is prime to p. Let '"
be an isomorphism of E onto the group of n-th roots of 1 in C; this will
be determined uniquely if we choose a generator B1 of E and prescribe
that", (B I )=e(1/n). Take any ,eK x, and let x be anyone of the roots of
the equation xn =, in K; then x is in K:ep , and the equation xn =, has
the n distinct roots BX, with BeE. In particular, for each ue(D, x" must be
one ofthese roots, so that x"x- 1 is in E. Now put
Xn,~(u) = '" (x" X-I);
as E c K, the right-hand side does not change if we replace x by ex with
BeE and is therefore independent ofthe choice ofa root x for X n =,. For
a similar reason, we have, for all p, u in (D:
x P" x- 1 =(xP X-I)"(X"X-I)=(XPX-I)(X"X-I),
and therefore
Xn,~(Pu)= Xn,~(P)Xn.~(u),
which shows that Xn,~ is a character of (D. Take now any '1eKx, and
call y a root of xn = '1; then x y is a root of xn = ~ fI, and we have, for all
ue(D:
and therefore
which shows that, --+ Xn, ~ is a morphism of K x into the group of char-
acters of (D. It is obvious that Xn,~ is trivial if and only if the equation
X n=, has one root, hence all its roots, in K, i.e. if ,e(Kx)n; in other
words, (K7 is the kernel of ,--+ Xn,~. It would be easy to show that the
image of K x under that morphism consists of all the characters of (D
whose order divides n, but this will not be needed.
e
Now we put, for, and in K:
{Xn,~, e} = {" e}n;
this is known as "Hilbert's symbol"; one should note that it depends

upon the choice of 1/1, or, what amounts to the same, ofthe generator 6 1
of the group E of n-th roots of 1 in K. By prop. 8, § 4, we have:
(10) g e',O},,= g,O},,' {e',O}", g,OO'},,= g,O},,' g,O'}"
for all e,e',O,O' in KX.
Call again x a root of X" = e; clearly the kernel ~ of X", ~ consists of
the elements (1 of ffi such that Xli = X, so that the corresponding subfield
of K sep, which is the cyclic extension of K attached to X",~, is L=K(x).
Call d the order of X",~; then Xn,~ determines an isomorphism of ffi/~
onto the group of d-th roots of 1 in C; d divides n, and it is also the
degree of Lover K. Therefore the distinct conjugates of x over K, i.e.
its images under the d distinct automorphisms of Lover K, are the ele-
ments 6X, where 6 runs through the group E' of d-th roots of 1 in K.
Write e=n/d, and, for any CeK:
e-l
W= n(C-6'1X),
v=o
where 6 1 , as before, is a generator of E; as the 6'1 for 0 ~ v ~ e -1 are a full
set of representatives of the cosets modulo E' in E, we have:
N L/K(W) = n(C - 6X) =en - e.
seE
For C=O and C=I, this shows that -e and l-e are in NL/K(L). By
prop. 10 of§4, this gives:
(11)
these formulas being valid whenever they are meaningful, i. e. the first
one for all e e K x , and the second one for all e =1= 0, 1 in K. In the first one,
replace e bye" with e,,, in KX, and apply (10); we get:
g, - e},,' {e,,,},,' {", - e},,' {",,,},,= 1.
Here, by (11), the first factor is 1, and the last one is equal to {", -I},,;
applying (10) again, we get
(12)
which is known as "the law of reciprocity" for the symbol {e,,,},,. The
same could be proved by the explicit construction of the simple algebra
= ="
corresponding to the latter factor-class; we merely sketch the proof in
the case when the equations X" e, X" are both irreducible over K.
That being assumed, put L = K (x), where x is a root of X" = e; let A be
the cyclic algebra [L/K; X".~,"]. By prop. 11 of§4, where we write now y
instead of Ul' A has a basis over L consisting of the yi for 0 ~j ~ n -1, hence
§ 5. Special cyclic factor-sets 187
a basis over K consisting of the Xi yi with 0 ~ i,j ~ n -1, with the relations
xn = e, e
y" = '1, x Y = 8 1 Y X. If we exchange and '1, and also x and y, A is
clearly replaced by its inverse AO; this implies (12).
Now let K be of characteristic p> 1; identify the prime field in K
with Fp, and call t/I the character of the additive group of Fp given by
t/I(1)='e(1/p). Take any eeK, and let x be anyone of the roots of the
e;
equation X - XP = then x is in K sep , and that equation has the p distinct
roots x+a with aeFp • In particular, for each O"e~,x'" must be one of
these roots, so that x'" - x is in Fp. Now put
Xp,~(O")= t/I(x'" -x);
as the right-hand side does not change if we replace x by x + a with
ae Fp , it is independent of the choice of x. A calculation, quite similar to
the one given above for Xn,~, shows that Xp , ~ is a character of ~, and that
e ~ Xp , ~ is a morphism of the additive group of K into the (multiplicative)
group of characters of ~ ; the kernel of that morphism is the image of K
under the mapping ~ e e- e p of K into itself, and it would again be easy
to show that the image of that morphism consists of X= 1 and of the

characters of ~ of order p. Put now, for all eeK and all lJeKx :
{Xp,~, lJ} = { e, lJ} po
Then we have:
for all e, e'

in K and all lJ, lJ' in K x. Assume now that x is not in K;
then L=K(x) is the cyclic extension of K attached to Xp,~, and it is of
e
degree p over K; XP - X + = 0 must then be the irreducible equation
e e.
for x over K, so that N L/K(X) = ( -1)P = - By prop. 10 of § 4, this gives
(14) {e,-e}p=1,
which is therefore valid whenever x is not in K. If xeK, Xp,~ is trivial, so
e
that (14) is still valid provided it is meaningful, i.e. provided =1=0. There-
fore (14) is valid for all eeKx.
Chapter X
Simple algebras over local fields
§ 1. Orders and lattices. Let D be a division algebra of finite dimension
over any field K; we will consider left vector-spaces over D, whose
dimension will always be assumed finite and > O. If V and Ware such
spaces, we write Hom(V, W) for the space of homomorphisms of V into
W, and let it operate on the right on V; in other words, if oc is such a homo-
morphism, and ve V, we write voc for the image of v under oc. We consider
Hom(V, W), in an obvious manner, as a vector-space over K; as such, it
has a finite dimension, since it is a subspace of the space of K-linear
mappings of V into W. As usual, we write End(V) for Hom(V, V).
If V, V', V" are left vector-spaces over D, and oce Hom(V, V') and
peHom(V', V"), we write ocp for the morphism v-+(voc)P of V into V".
For V= V' = V", this makes End(V) into a ring; as before, we write
Aut(V) for End(V»(, this being the group of automorphisms of V. For
V= V', V" = W, we get for Hom(V, W) a structure of left End(V)-module;
for V' = V" = W, we get for Hom(V, W) a structure of right End(W)-
module.
Let D and V be as above; let d2 be the dimension of Dover K, and m
that of V over D. Take a basis {Vi' ... , vm} of V over D;foreach ~eEnd(V),
write Vj~= ~>ijVj' with xjjeD for 1 ~i,j~m; this defines a mapping
j
~-+(xij)of End(V) into Mm(D), which is obviously an isomorphism of
End(V) onto Mm(D); in particular, this shows that End(V) is a simple
algebra of dimension m2 d2 over K. Obviously V, considered as a right
End(V)-module, is simple; therefore, by prop. 1 of Chap. IX-l, every
such module is a direct sum of modules isomorphic to V.
Let Vand W be left vector-spaces over D; call m, n their dimensions;
put A = End(V), B = End(W), H = Hom(V, W). As H is a right B-module,
it is a direct sum of modules isomorphic to W; comparing dimensions
over K, one sees at once that it is the direct sum of m such modules.
Similarly, as a left A-module, H is the direct sum of n modules isomor-
phic to the dual space V' = Hom(V, D) of V, this being a simple left
A-module and a right vector-space of dimension mover D. One could
easily see that every endomorphism of H for its structure as a left
A-module is of the form ).-+).p with peB, and that every endomorphism
of H for its structure as a right B-module is of the form ).-+oc). with oceA.
§l. Orders and lattices 189
Let D, V and A = End(V) be as above, and let v be the reduced norm

in A; by corollary 1 of prop. 6, Chap. IX-2, the determinant of the endo-
morphisms x-+exx and x-+xti of the underlying vector-space of A
over K, for any exE A, is v(ex)md; in particular, ex is in A x if and only if
v(ex)fO.
If K is a local field, all the above spaces, being vector-spaces of
finite dimension over K, can be topologized as such in one and only
one way, according to corollary 1 of th.3, Chap. 1-2; conversely, by
corollary 2 of the same theorem, the requirement of finite dimensional-
ity over K could everywhere be replaced by that of local compactness.
If we write again A = End(V), the group A x = Aut(V) is the open subset
of A determined by v(ex)fO; as such, it is a locally compact group. More-
over, the Haar measure in it is right-invariant as well as left-invariant;
this is contained in the following lemma, which generalizes lemma 5
of Chap. VII-4:
LEMMA 1. Let K be a local field; let D, V and A=End(V) be as above,
and let ex be a Haar measure on A. Then the measure JI. on AX, given by
dJl.(x) = modK(NAIK(X»)-l dex(x) = modK(v(x)t md dex(x)
is both left-invariant and right-invariant on A x •
This follows at once from corollary 1 of prop. 6, Chap. IX-2, com-
bined with corollary 3 of tho 3, Chap.I-2.
In the rest of this §, we assume that K is a commutative p-field;
D being a division algebra over K, hence also a p-field (a non-commuta-
tive one, unless d = 1), we write Rand RD for the maximal compact
subrings of K and of D, and P and PD for the maximal ideals in Rand
in RD , respectively.
Let V and W be as above; let L be a D-Iattice in V, and MaD-lattice
in W; then we write Hom(V,L; W,M) for the set of all morphisms of
V into W which map L into M. Choose bases {v1, ... ,vm }, {w1, ... ,wn }
for V and W according to th.l of Chap. 11-2, i.e. so that L= LRDvj
and M=LRDwj . For each AEHom(V,W), we can write VjA=LxijWj
with xjjED for 1 ~i~m, 1 ~j~n, and this determines a bijection A-+(Xjj)
of Hom(V, W) onto the space Mm,iD) of the matrices with m rows and n
columns over D; clearly A is then in Hom(V,L; W,M) if and only if the
matrix (xij) is in Mm,n(RD). In particular, this shows that Hom(V,L; W,M)
is a K -lattice in the space Hom(V, W) considered as a vector-space
over K, and also that it can be identified with the space of all morphisms
of L into M for their structures as RD-modules. We write End(V,L) for
Hom(V,L;V,L); this is a K-Iattice and an open compact subring of
End(V), which may be identified with End (L). We write Aut(V,L) for
End(V,Lt; it is the group of automorphisms of L.
190 Simple algebras over local fields x
PROPOSITION 1. Let K be a p-field, D a division algebra over K, V a
left vector-space over D, and LaD-lattice in V. Let v be the reduced norm
in the algebra A=End(V) over K. Then Aut(V,L) consists of the ele-
ments ~ of End(V,L) such that modK(v(~))= 1.
Take ~EA; it is in AX if and only if v(~)+O. If m, d are as before,

the module of the automorphism x-+x~ of A is modK(v(~)td; as A, for
its structure as a right A-module, is the direct sum of m modules iso-
morphic to V, this implies that the module of the automorphism v-+v ~
of V is modK(v(~))d. Now assume that ~ is in End(V,L); then it maps L
onto aD-lattice L' = L ~ contained in L, so that the module of v-+v ~
is equal to [L :L']-1. This shows that L=L' if and only ifmodK(v(~))= 1,
which proves our proposition.
COROLLARY. Notations being as in proposition 1, Aut(V,L) is a com-

pact open subset of End(V,L) and of End(V) and a compact open sub-
group of Aut(V).
This is now obvious.
PROPOSITION 2. Let V be as in proposition 1; let X be a multipli-

catively closed subset of End(V). Then X is relatively compact in End(V)
if and only if there is aD-lattice L in V such that Xc End(V,L).
Let X be relatively compact in End(V); we may assume that it is
compact, since otherwise we replace it by its closure. Let L be any
D-Iattice in V. Call I.; the set of the vectors VEL such that v~EL for all
~EX; clearly this is an RD-module, hence closed, by prop. 5 of Chap. 11-2;
being contained in L, it is compact. As X is compact and L is open,
I.; is open. Therefore I.; is a D-Iattice. As X is multiplicatively closed,
v~ is in I.; for all VEI.; and all ~EX, so that X cEnd(V,L'). The converse
is obvious.
PROPOSITION 3. Let V be as above, and let L, L' be two D-lattices in V.

Then either Aut(V,L) is not contained in End(V,L'), or there is xED x
such that L' = x L.
By tho 2 of Chap. 11-2, there is a basis {V1""'Vm } of V, and there are
integers Vi' such that L= IRDvi and L'= Ip~iVi' Every permutation of
the Vi determines an automorphism of V which belongs to Aut(V,L);
if all these are in End(V,L'), all the Vi must be equal; if V is their common
value, we have L'='lt1L for any prime element 'ltD of D.
THEOREM 1. Let D be a division algebra over a p-field K, and let V

be a left vector-space over D. Then the maJ(imal compact subrings of the
§l. Orders and lattices 191
algebra A = End(V) are the rings End(V,L), and the maximal compact
subgroups of AX are the groups Aut(V,L), when one takes for L all the
D-Iattices in V.
By prop. 2, a compact subring of End(V) must be contained in some

End(V,L), hence must be equal to it if it is maximal. Now assume that,
for some L, End(V,L) is contained in a compact subring X of End(V);
this, in tum, must be contained in some End(V,L). Then, by prop. 3,
L'=xL with some xeDx; this gives End(V,L:) = End(V,L), hence
X = End(V,L). Similarly, a compact subgroup of AX = Aut (V) must be
contained in some End(V,L), hence in End(V,LY, i.e. in Aut(V,L). If
this is contained in a compact subgroup X of Aut(V), X must be con-
tained in some End(V,L'J, and we get L'=xL and X = Aut(V,L), just as
~efore. In the conclusion of theorem 1, one might take for L, instead
of all the D-Iattices in V, a full set of representatives for the equivalence
relation among D-Iattices defined by L'=xL, xeDx.
Compact open subrings of a simple algebra over a p-field are also
called orders; thus, the first part of theorem 1 states the existence of
maximal orders in the algebra A = End (V), viz., all the rings End(V,L).
As we have seen above, these are all isomorphic to Mm(RD) if m is the
dimension of V over D; clearly they are the transforms of one another
under the automorphisms of V, since any basis of V over D can be trans-
formed into any other basis by such an automorphism. It amounts to
the same to say that they are the transforms of one another under the
inner automorphisms of A.
PROPOSITION 4. Let D be as above, and let V, W be two left vector-

spaces over D. Let M, M' be compact open subgroups of Hom(V, W), and
let X be the set of the elements ~ of End(V) such that ~ M eM'. Then X
is a compact open subgroup ofEnd(V); if M =M', it is a subring of End(V).
Obviously X is a subgroup of End(V), and a subring if M =M'.

As M is compact and M' is open, X is open. Now put H = Hom(V, W).
As M is open, it contains a basis {J.tl' ... 'p,.} for H regarded as a vector-
space over K. If now we regard H as a left End(V)-module, the annihila-
tor of that basis in End(V) is the same as that of H, hence {O} since
e e e
End(V) is simple and W is not {O}. Therefore the mapping -+( J.I.t, .•• , p,.)
of End(V) into W = H x ... x H is injective, hence an isomorphism of
End(V) onto its image in W, for their structures as vector-spaces over K,
hence also for their topological structures. This implies that the set X'
e
of the elements of End(V) such that eJl.ieM' for 1 ~i~r is compact.
As 'X is a subgroup of X' and is open in End(V), it is an open subgroup
of X', hence closed in X', hence compact.
For M =M', the ring X defined by proposition 4 is called the left
order of M. Exchanging right and left, we see that the set Y of the ele-
ments t'/ of End(W) such that Mt'/cM is a compact open subring of
End(W); this is called the right order of M. Now we show that, if one
of these orders is maximal, the other is also maximal. This is contained
in the following:
THEOREM 2. Let K and D be as in theorem 1; let V, W be two left

vector-spaces over D, and let L be a D-Iattice in V. Let N be a compact
open subgroup of Hom(V, W) such that eNcN for all e in End(V,L).
Then there is aD-lattice M in W such that N=Hom(V,L;W,M), and
the left and right orders of N are End(V,L) and End(W,M), respectively.
By th.1 of Chap. 11-2, we can choose a basis {vlo""vm } of V over D,

so that L= LRDvj; then, as explained above, we can use this basis for
identifying End(V) with Mm(D) and End(V,L) with Mm(R D), by assigning
to each element e of End(V) the matrix (xij) given by Vje = LXijVj' Now
consider the mapping IX-+(V1IX, ... ,VmIX) of Hom(V,W) into the direct
sum wm of m spaces isomorphic to W; clearly it is a bijection of
Hom(V, W) onto wm; call it cp, and put N' = cp(N), where N is the set
in theorem 2. If IX is in Hom(V, W) and cp(IX)=(W1, ... , wm), and if and e
(xij) are as above, then cp(eIX)=(W~, ... , w~), with w;= LXjjWj for 1 ~i~m;
by our assumption on N, this must be in N' whenever (WI"'" wm) is in N'
and all the Xjj are in RD' Writing eij for the "matrix units" in Mm(D),
as defined in the proof of tho 1, Chap. IX-I, take first for (xij) the matrix
unit e",,; then we see that, if(wl , ..• , wm) is in N', every one ofthe elements
(0, ... ,0, Wh' 0, ... ,0), for 1 ~ h ~ m, must also be in N'. This is the same
as to say that, if we call WI"'" Wm the m summands of W m, and if we
put N;' = N' n w,. for 1 ~ h ~ m, we have N' = LN;'. Similarly, taking for
(x jj) the matrix unit ehk' we see now that N;' = Ni for all hand k; put
M=N;' for any h. Finally, taking for (xij) the matrix x'l m with xERD,
we see that M is an RD-module. As N is open and compact in Hom(V, W),
N' is so in W m, hence N;' in w,., and M in W; therefore M is aD-lattice
in W. Now we see that an element IX of Hom(V, W) is in N if and only if
vjIXis inM for 1 ~ i ~m; this is the same as to say that N = Hom(V,L; W,M).
Then the left order and the right order of N contain End(V,L) and
End(W,M), respectively; as the latter are maximal orders, this com-
pletes our proof.
COROLLARY 1. Let A be a simple algebra over K, R.. a maximal com-

pact subring of A, and I a left ideal in R... Then I is open in A if and only
if it can be written as 1= RA IX with IXE R.. n A )(.
§2. Traces and norms 193
We may assume that A=End(V), where Vis as in theorem 2; then,

by tho 1, we may assume that RA = End(V,L), where L is as in theorem 2.
If I is open, we may apply theorem 2 to it, taking W = V and N = I ;
this gives I =Hom(V,L; V,M), where M is a D-Iattice in V. Take bases
{v1, ... ,vm}, {Wl""'Wm} of V, so that L=LRDvi and M=LRDwi , and
e
call the automorphism of V which maps the former basis onto the
e e; e
latter one. Then M = L and I = ~ as is in I, it is in RA • The con-
verse is obvious.
COROLLARY 2. Let A and RA be as in corollary 1, and let J be a com-
pact two-sided RA-module in A, other than {O}; then J is open in A. If
A = End(V) and RA = End(V, L), with V and L as in theorem 2, then J
can be written as J=Hom(V,L;V,1t;'L), where veZ and 1tD is a prime
element of D.
As RA is a K-Iattice in A, we can choose a basis {tl1, ••. ,tlN} of A
e
over K, consisting of elements of RA • If is in A and not 0, the two-sided
e
ideal generated by in A is A, since A is simple; therefore the elements
tlietlj, for 1 ~i, j~N, generate A as a vector-space over K, so that the
R-module they generate in A is a K-Iattice, hence open. This implies
that the set J in our corollary must be open; then, by theorem 2, we
can write it as J = Hom(V,L; V,M), where M is a D-Iattice in V such
that End(V,M) contains End(V,L). By prop. 3, this gives M =xL with
xeDx; we have then M =1t;'L for v=ordD(x).
If V and Ware as in theorem 2, any set N with the properties des-
cribed there, i.e. any set which can be written as N = Hom(V,L; W,M)
for suitably chosen D-Iattices L in V and M in W, will be called a normal
lattice in Hom(V, W).
§ 2. Traces and norms. As before, we consider a local field K, a

division algebra D of dimension d2 over K, and a simple algebra A
over K, isomorphic to Mm(D) for some m~ 1; we denote by 1" and tD
the reduced traces in A and in D, and by v and VD the reduced norms
in A and in D, respectively. We begin by considering the case of a p-field.
PROPOSITION 5. Let K be a p-jield; let D be a division algebra of
dimension d 2 over K; let RD be the maximal compact subring of D, and
let 1tD be a prime element of D. For any m~ 1, put A=Mm(D) and
RA = Mm(R D), and let t be the reduced trace in A. Then the set of the
elements x of A, such that t(xy)eR for all yeRA' is wR A=RAw, with
w=1t1- d ·l m·
Consider first the case m = 1, A = D, RA = RD' As has already been

observed in Chap. IX4, we may use prop. 5 of Chap. 14 to describe D
as a cyclic algebra [KtlK;x,n] over K, where Kl is an unramified
extension of K of degree d, X a character attached to that extension,
and 1t a prime element of K; then a comparison of that proposition
with the definition of a cyclic algebra in prop. 11 of Chap. IX-4 shows
that u1 , in the latter proposition, is a prime element of D. We may now
substitute U1 for 1tD' as this does not affect the statement in our proposi-
tion, and write therefore 1tD = U1 , hence, with the notations of prop. 11
of Chap. IX-4, u;=1th for O~i~d-l and 1t=~. Call R1 the maximal
compact subring of K 1 ; by (b) in prop. 5 of Chap. 1-4, RD is the left
R1-module generated by Uo, ... ,Ud- 1; therefore any xeD has the pro-
perty stated in our proposition if and only if rD(x'1nb)eR for all '1eR1
and for 0 ~j ~ d -1. Again by (b) in prop. 5 of Chap. 1-4, we may write
X=Lei1th with eieK1 for O~i~d-1. Using formula (8) of Chap. IX-4
i
for the reduced trace rD in D, we get, for j=O, 'ITK.lK(eo'1)eR for all
'1eR1; by prop. 3 of Chap. VIII-I, this is so if and only if eOeR 1. Similar-
ly, for l~j~d-l, our condition can be written as TrK,/K(ed_j'7'1t)eR
with '1'='1 fJ , p=a j - d; as every automorphism P of K1 maps R1 onto
itself, this must be so for all '1' e R1 , and this, just as before, is equivalent
to ed_ j e1t- 1 R1. Therefore the set defined in our proposition is the
R1-module generated by 1,1t- 11tD , ••• ,1t- 11ti-1, i.e. by the elements
w D1th for O~ i~d-l if w D=1tA -d. In view of (b) in prop. 5, Chap. 1-4, this
completes our pro offor the case m= 1. For m> 1, our conclusion follows
immediately from this and from corollary 2 of prop. 6, Chap. IX-2,
which says that r(x) = L rD(Xjj) for x = (xij) in A.
let X be a character of K of order 0, and identify A with its topological
dual by putting (x,y)=x(r(xy»). Then the dual K-lattice to RA is wRA.
In fact, this dual lattice is defined as the set of the elements x of A
such that x(r(xy») = 1 for all yeRA. As r is K-linear, this is the same
as to say that x(r(xy)z) = 1 for all yeRA and all zeR, hence, by prop. 12
of Chap. 11-5, the same as r(xy)eR for all yeRA.
COROLLARY 2. Let A be a simple algebra over K; let r be the reduced
trace in A, X a character of K of order 0, and identify A with its topologi-
cal dual by (x,y)=x(r(xy». Let M and M' be two K-Iattices in A, dual
to each other in A. Assume that both M and M' are subrings of A. Then A
is trivial over K, M is a maximal compact subring of A, and M = M'.
By tho 1 of § 1, M is contained in some maximal compact subring RA

of A, and we may, by using a suitable isomorphism of A with an algebra
of the form Mm(D), identify A with Mm{D) and RA with Mm(R D), where
notations are as in proposition 5. As M eRA' corollary 1 shows now
that M':::JwRA:::JRA, hence, by th.l of § 1, M'=wRA=RA- Clearly this
implies that d= 1, i.e. that A is trivial, and then, again by corollary 1,

that M=RA •
PROPOSITION 6. Let K be a p-field, A a simple algebra over K, and v
the reduced norm in A. Then v(A X) = K X•
By corollary 3 of prop. 6, Chap. IX-2, it is enough to consider the
cas~ A = D; then, just as above, we can write D as a cyclic algebra and
use for VD the formula (9) in Chap. IX-4. With the same notations as
above in the proof of prop. 5, this shows, firstly, that vD(D X ) contains
N K,IK(R~), which is the same as R x by prop. 3 of Chap. VIII-1, and
secondly that it contains VD(U 1)=1t. As K X is generated by R X and 1t,
this proves our proposition.
In the case of R-fields, the conclusion of proposition 6 is of course
valid for A=Mm(K) with K=R or C, but not for K=R and A = Mm(H)·
In fact, as we have seen in Chap. IX-4, the algebra H of "classical"
quaternions has a basis over R, consisting of the "quaternion units"
1, i, j, k, with the relations i 2 = -1, l = -1, k = ij = - ji, which imply
k 2 =-1, i=jk=-kj,j=ki=-ik. Clearly the R-linear bijection x-+x
ofH onto itself which maps 1, i,j, k onto 1, - i, - j, - k is an antiautomor-
phism, i.e. it maps xy onto yx for all x, y. In order to determine the
reduced trace t and the reduced norm v in H, one needs a C-represen-
tation F of H. By applying some of the results in Chap. IX, or by a direct
computation, one finds that such a representation is given by
F(1)= 12 , . (i 0)
F(l)= 0 -i ' F(k)= (~ ~).
Then one finds at once, for x=t+ui+vj+wk, with t, u, v, w in R:
t(x)=x+x=2t,
This shows that v maps H X onto R~ ; by corollary 3 of prop. 6, Chap. IX-2,
the same is therefore true for A = Mm(H) for any m ~ 1.
§ 3. Computation of some integrals. Here, in preparation for the

computation of the zeta-function of a simple algebra in Chapter XI, we
carry out some local calculations, generalizing the results of prop. 11,
Chap. VII-4, and oflemma 8, Chap. VII-6.
Take first a p-field K and a division algebra Dover K. Call RD the
maximal compact subring of D, PD the maximal ideal in RD , and 1tD a
prime element of D. For each e~O, choose a full set A(e) of represen-
tatives of the classes modulo PDin RD' Now, for a given m~ 1, we define
three subsets l:, l:', l:" of Mm(D)X, as follows. By l:, we understand the
group of "triangular" matrices in Mm(Dr, consisting of the matrices
t=(tjj) such that tjj=O for 1~j<i~m, and tjjfO for 1~i~m. By l:',
we understand the subset of l:, consisting of the matrices t=(tjj) in l:,
such that tijE RD for all i, j, and that each tjj, for 1 ~ i ~ m, is of the form
1r:jJ with ej~O. By l:", we understand the subset of l:', consisting of the
matrices t=(tjj) in l:', such that tijEA(ej) for 1~i<j~m, ej being given
by t jj =1r:'jJ. With these notations, we have:
LEMMA 2. Let V be a left vector-space of dimension mover D; let L
be a D-Iattice in V, and let {V1"'" vm} be a basis of V such that L= L RD Vi'
Let I.: be a D-Iattice in V, contained in L. Then there is one and only one
basis {v~, ... , v~} of V, such that I: = L RD v; and that v; = L tjj vdor 1 ~ i ~ m,
j
with a matrix t = (tij) belonging to the set l:".
For 1 ~i~m, call W; the subspace of V generated by Vj, ... ,Vm. Let
{W1"'" wm} be any basis of V, and write it as Wj= LXijVj' Clearly the
matrix x=(xij) is in l:, i.e. it is triangular, if and only if, for each i,
{Wj"'" wm} is a basis of W;. By th.1 of Chap. 11-2, one can choose such
a basis for which I.:= LRDwj; then, since I:cL, all the xij are in RD'
Write Xjj=y j1r:jJ, with yjER~ and ejEZ, for 1~i~m, and replace the Wj
by the vectors Yi 1 Wj, which obviously have the same properties; after
that is done, the matrix (xij) is in l:'. Assume now that there are vectors
v~, ... ,v~ such as required by the lemma, and write v;= LZij~; clearly
the matrix (zij) must then be triangular, and, as I.:= LRDV; = LRD Wi' it
must be in Mm(RDr ; for a triangular matrix (zij), the latter condition
is fulfilled if and only if ZiiER;; and zljERD for all i,j. Then the coefficient
of Vi in vi is Zjj1r:jJ, and, as this must be of the form 1r:il, we must have
Zjj= 1. Now, for 1 ~i<j~m, the coefficient tij of Vj in v; is given by
j-1
tij=xij+ L ZiI,Xllj +Zij1r::J,

II=j+1
and the proof of the lemma will be complete if one shows that the zij'
for 1~i<j~m, can be uniquely chosen in RD so that tjjEA(ej) for
1 ~ i <j ~ m. For each value of i, this can be verified at once by induc-
tion on j for i + 1 ~j ~ m.
LEMMA 3. The set l:" is a full set of representatives for those left
cosets of Mm(RDr in Mm(Dr which are contained in Mm(RD)'
Take a left vector-space V of dimension mover D, aD-lattice L

in V, and a basis {v1, ... ,vm} of V such that L= LRDvj; as before, identify
End(V) with Mm(D) by assigning, to each 'EEnd(V), the matrix (Xjj)
given by Vj'= LXijVj: put A=End(V) and RA=End(V,L); then R A=
=Mm(R D), and R~, i.e. Mm(RD) x , consists of the automorphisms of e

e
V such that L = L. Therefore two elements tX, {J of AX belong to the
same left coset of R~ if and only if LtX=L{J; that left coset is contained
in R. A if and only if L tX c L. At the same time, by lemma 2, every D-Iattice
L' of V, contained in L, can be written in one and only one way as LRDV;
with v;= Lt;jVj and (t;)el:"; this is the same as to say that it can be
written in one and only one way as Lt with tel:", which proves our
lemma.
PROPOSITION 7. Let K be a p-field with the module q; let D be a division
algebra of dimension d2 over K. Let A be a simple algebra over K, isomorphic
to Mm(D). Let R Abe a maximal order in A, and cp its characteristic function.
Let v be the reduced norm in A, and let Jl. be the Haar measure in A x such
that Jl.(R~) = 1. Then the integral
J(s)= Jcp(x) modK(v(x)rd Jl.(x),
AX
where seC, is absolutely convergent for Re(s»d(m-1) and has then the
value
m-l
;=0
As before, identify A with End(V) and RA with End(V,L), where V is

a left vector-space of dimension mover D, and LaD-lattice in V. By
prop. 1 of§ 1, the integrand in J(s) is constant on left cosets of R~ in AX;
in view of the definition of Jl., this gives:
J(s) = LmodK(v(r)r,
the sum being taken over any full set of representatives of those left
cosets of R~ in A x which are contained in R A , and for instance over
the set l:" supplied by lemma 3. If we identify now A with Mm(D) and
RA with Mm(RD) as before, l:" consists of the triangular matrices t=(t;j)
such that t;;=nri and tijeA(ej) for all i, j, the e; being integers ~O. By
corollary 2 of prop. 6, Chap. IX-2, we have then v(t) = nVD(tii)= vD(nD)E
with E = Le;, VD being the reduced norm in D; as we have seen in § 2,
vD(nD) is a prime element of Kif nD has been suitably chosen, and this
implies that the same is true for any choice of nD • This gives modK(v(t») =
=q-E. On the other hand, prop. 5 of Chap. 1-4 shows that the module of
D is qd, so that,for each e~O, the set A(e) consists of qde elements. Therefore,
for a given set of integers e1 , .'" em, there are qdN matrices tel:", with
N = L(i-1)e;. Thus we get:
;
Clearly this is absolutely convergent for Re(s»d(m-l) and has then
the value stated in our proposition.
COROLLARY. Let l(s) be as in proposition 7, and let 10(s) be similarly

defined for the algebra Ao=MiK) with n=dm. Then we have, for
Re(s»n-l:
This follows at once from proposition 7.

We will also need the corresponding results for algebras over R-
fields. Here we have either K=R, and D=R or H, or K=D=C. In all
three cases, x ..... X is an antiautomorphism of D such that xx>O for all
xeDx; it is the identity if D=R, the non-trivial automorphism of C
over R if D = C, and it is as defined at the end of § 2 if D = H. As usual, if
x = (xij) is any matrix in M m(D), we write IX for its transpose, and x for
the matrix (Xjj); then x ..... IX is an antiautomorphism of M m(D). We will
writel: for the set of all triangular matrices (tij) in Mm(D) such that
tjj e R ~ for 1 ~ i ~ m; clearly this is a subgroup of M m(Dt . Now let V be
a left vector-space of dimension mover D. For the sake of brevity, and
although this does not quite agree with the established usage, we will
say that a mapping f of V x V into D is a hermitian form on V if there is a
basis {Vl' ... , vm} of V such that, for all Xj, Yj in D:
f(~>jVj, LYjVj)= LX/Y/;
every basis of V with that property will then be called orthonormal for
f. One sees at once that a basis {w 1, ... , wm} is orthonormal for f if and
only if f(wjo Wj) = ~ij for all i, j, or even if this is so merely for 1 ~ i ~j ~ m.
We topologize the space of all hermitian forms on V by the topology of
"uniform convergence on compact subsets" of V x V; in other words, for
each compact subset C of V x V, and each e > 0, the set of the hermitian
forms f' such that modD(f' - f) ~ e on C is to be a neighborhood of f,
and these make up a fundamental system of neighborhoods of f in the
space of hermitian forms.
LEMMA 4. For D=R,H or C, let V be a left vector-space over D, with

the basis {v l' ... , vm}, and let f be a hermitian form on V. Then there is one
and only one orthonormal basis {v~, ... , v~} for f such that v;= LtijVj
with (tjj)el:, and it depends continuously upon f
The proof is straightforward, and so well known that it may be

omitted here.
Let V be as above; let fbe a hermitian form on V, with the orthonormal

basis {VI' ... , vm}. Let {WI' ... , wm} be a basis of V, given by Wi= LUijVj
with u=(Uij)EMm(D); a trivial calculation shows at once that this is
orthonormal for f if and only if u·1u= 1m; clearly the matrices U with
that property make up a compact subgroup of M m(D) x, which we will
denote by U. Now let (X be any automorphism of V; we will write /"
for the transform of f by (x, i.e. for the mapping defined by /"(v,w)=
= f(V(X-l, W(X-l) for all V, Win V; this is a hermitian form, with the ortho-
normal basis {VI (X, ... , vm(X}. Clearly, when we identify A=End(V) with
Mm(D), as before, by means of the basis {vl, ... ,vm}, U is the subgroup of
A x = M m(D) x consisting of the automorphisms ~ of V such that f~:::; f.
LEMMA 5. The subgroups l: and U of AX:::; Mm(D) x being as defined
above, the mapping (u,t)--+ut is a homeomorphism of U x l: onto A x.
Let V, f and the orthonormal basis {v 1, ... , vm} be as above; use that
basis again to identify A = End(V) with Mm(D), hence A x with Mm(D) x .
Let (x, p be in A x ; we have /" = jP if and only if f = jP"- \ i. e. if and only
if PIX-lEU, or PEU(X. Now, for anY(xEA x, apply lemma 4 to/,,; it shows
that there is one and only one matrix (tij)El: such that the vectors
vi= LtijVj make up an orthonormal basis for /". This is the same as to
say that the automorphism t of V which corresponds to that matrix, i. e.
which maps {VI' ... , Vm} onto {V'I' ... , v~}, transforms f into ft=/".
Moreover, by lemma 4, the matrix (t i ) depends continuously upon /",
hence upon (x. Expressing this in terms of the matrices x, u, t in Mm(D) x
which correspond respectively to (x, (xt- l and t, we get the assertion in
our lemma.
LEMMA 6. Notations being as in lemma 5, let Ji. be a Haar measure
on A x. For every continuous function F with compact support on l:, call
F' the function on A x such that F'(ut)=F(t) for all UE U and all tE~.
Then F' is continuous with compact support on A x, and there is a right-
invariant measure () on l: such that JF' dJi. = JF d(} for all F.
The first assertion follows at once from lemma 5 and the compacity
of the group U. Then, as Ji. is right-invariant on A x by lemma 1, § 1,
J
it is obvious that the mapping F --+ F' dJi. is invariant under right trans-
lations in l:. By the theory of the Haar measure, this shows that () is
the image of a Haar measure, i. e. of a left-invariant measure on l:,
under the homeomorphism t--+t- l of l: onto itself.
LEMMA 7. Let (X be a H aar measure on D; put 15 = [K : R], and write,
for t=(ti)El::
m
d(}(t) = Il (tii 4dl (i- l l - l dtu)' Il d(X(t i)·
l~i<j~m
Then this defines a right-invariant measure on l:.

As ~d2 is the dimension of Dover R, corollary 2 of tho 3, Chap.I-2,
shows that, for every aER~, the module of the automorphism x-+xa
of D is a~d2. A straightforward computation shows then at once, firstly
that the measure 8 in lemma 7 is invariant under t-+t t' for every diagonal
matrix tiEl:, and secondly that it is invariant under t-+tt" for every
matrix t" = (tij)E l: such that t;i = 1 for 1 ~ i ~ m. As every matrix in l:
can be written in the form t' t", this proves our lemma.
PROPOSITION 8. Take K=R and D=R or H, or K=D=C; call ~ the
dimension of Kover R, and d 2 that of Dover K. Call 't" the reduced trace
and v the reduced norm in the algebra A = Mm(D) over K. Let J.I. be a Haar
measure in A x. Then the integral
1(s) = f exp{ -n~'t"Cx· x))modK{v(x))SdJ.l.(x)
AX
is absolutely convergent for Re(s»d(m-1), and, for a suitable choice of

J.I., it has then the value
n
m-l
1(s)=(n~d)-m~dS/2 r(~d(s-di)/2).
i=O
Clearly the first factor in the integrand of 1(s) is constant on left

cosets of U in A x. Now, for any UEU, put z=v(u). If D=K, this means
that z=det(u), so that UE U implies zz= 1, hence modK(z) = 1. If K =R
and D=H, we have, for all xEA, v(x)=det{F(x»), where F is an iso-
morphism of A into M 2m(C); but then x-+t FCx) is also such an iso-
morphism, so that v('x)=v(x); this implies that, for UEU, v(u)2=1, and
therefore v(u) = 1 since we have seen in § 2 that v maps Mm(Hr into
R~. Therefore, in all cases, the second factor in the integrand of 1(s)
is also constant on left cosets of U in A x. That being so, lemmas 6 and
7 show that, for a suitable choice of J.I., 1(s) is the same as the integral
with the same integrand, but taken on l: with the measure dlJ(t). The
reduced trace 't"D in D is given by 't"D(X)=X if D=K, and 't"D(X)=X+ x
if K = Rand D = H; in view of corollary 2 of prop. 6, Chap. IX-2, we
have now, for t=(tij)El::
't"('t·t)=d· L Tijtij' v(t) = n (t;;)d.

l~i~m
This gives:
1(S)=i\ (+1 eXp(-nbdt2)tSi-ldt). (1 exp(_nbdTt)dlX(t»)m<m

2-1)
with s; = ~ d(s - di + d) for 1 ~ i ~ m. The last factor is independent of

s and is > o. The other factors can be transformed into the usual integral
for the gamma function by an obvious change of variable. Up to a
constant factor >0, which can be rendered equal to 1 by changing the

Haar measure J.l, the result is that stated in our proposition.
COROLLARY. Assume, in proposition 8, that K = Rand D = H; let
l(s) be as defined there, and let 10(s) be the similarly defined integral for
the algebra Ao = M n(R) with n = 2m. Then we have, for Re(s) > n -1 :
l(s)lo(s)-l = y n (s-h)
O<h<n
h ;FO(2)
with a constant y > o.

This is an immediate consequence of proposition 8 and of the iden-
tities
r(s+1)=sr(s), r(s)=n- 1/ 2 2s - 1 rG) re;1)
from the theory of the gamma function.

Chapter XI
Simple algebras over A-fields

§ 1. Ramification. In this Chapter, k will be an A-field; we use all
the notations introduced for such fields in earlier Chapters, such as
kA' kv, rv, etc. We shall be principally concerned with a simple algebra
A over k; as stipulated in Chapter IX, it is always understood that A
is central, i. e. that its center is k, and that it has a finite dimension over
k; by corollary 3 of prop. 3, Chap. IX-I, this dimension can then be
written as n2 , where n is an integer ~ 1. We use A v, as explained in Chap-
ters III and IV, for the algebra Av=A@kv over kv, where, in agreement
with Chapter IX, it is understood that the tensor-product is taken over
k. By corollary 1 of prop. 3, Chap. IX-I, this is a simple algebra over
kv; therefore, by tho 1 of Chap. IX-I, it is isomorphic to an algebra
Mm(v)(D(v)), where D(v) is a division algebra over kv; the dimension of
D(v) over kv can then be written as d(V)2, and we have m(v)d(v)=n;
the algebra D(v) is uniquely determined up to an isomorphism, and
m(v) and d(v) are uniquely determined. One says that A is unramified or
ramified at v according as Av is trivial over kv or not, i. e. according as
d(v)= lor d(v) > 1.
THEOREM 1. Let A be a simple algebra over an A-field k; let (X be a finite
subset of A, containing a basis of A over k. For each finite place v of k,
call (Xv the r v-module generated by (X in Av. Then, for almost all v, Av is
trivial over kv, and (Xv is a maximal compact subring of Av'
* By corollary 1 ofth. 3, Chap. 111-1, we may assume that (X is a basis of

A over k, and that 1A belongs to it. Call -r the reduced trace in A; by prop. 6
of Chap. IX-2, it is not 0, and its kv-linear extension to Av is the reduced
trace in Av. By lemma 3 of Chap. 111-3, we may identify the underlying
vector-space of A over k with its algebraic dual by putting [x,yJ = -r(x y).
Now, as in tho 3 of Chap. IV-2, take a "basic character" X of kA- By corol-
lary 1 of that theorem, Xv is of order 0 for almost all v; by corollary 3 of
the same theorem, the kv-Iattice (Xv is its own dual for almost all v, when Av
is identified with its topological dual by putting (x,Y)=Xv(-r(xy)). By
corollary 2 ofth. 3, Chap. 111-1, (Xv is a compact subring of Av for almost
all V. Therefore, at almost all places v of k, the assumptions of corollary 2
of prop. 5, Chap. X-2, are valid, the conclusion being as stated in our
theorem.
The first part of theorem 1 can be expressed by saying that A is unra-

mified at almost all places of k. The object of § 2 will be to show that it
cannot be unramified at all places of k unless it is trivial.
§ 2. The zeta-function of a simple algebra. Let all notations be as

in § 1, and let !X be a basis of A over k. By tho 1 of § 1, !Xv is a maximal com-
pact subring of Av for almost all v; therefore we may, for each finite place
v of k, choose a maximal compact subring Rv of A v, in such a way that
Rv = !Xv for almost all v; that being done, call tP v the characteristic function
of Rv' For each infinite place v of k, choose an isomorphism of Av with
Mm(v)(D(v»), where D(v) is R, H or C, as the case may be; identifying Av
with the latter algebra by means of that isomorphism, define tPv on Av
by putting, for all xEA v, tPv(x)=exp(-1tlhex'x»), where notations are
n
the same as in prop. 8 of Chap. X-3. Then tP= tPv is a standard function
on AA' Taking now a Haar measure JI. on A~, we have:
PROPOSITION 1. The integral
ZA(S)= f tP(z)lv(z)lldJl.(z)
A~
is absolutely convergent for Re(s) > n and is then given by the formula
n n
n-1
ZA(s)=CnZk(s-i)·n ( (1-q~-'»)'( (s-h»)p,
i=O v O<h<n O<h<n
h 1£ O(d(v)) h 1£0(2)
where Zk is the function defined in theorem 3 of Chap. VII -6, or the zeta-
function of k, according as k is of characteristic 0 or not, where p is the
number of real places v of k for which D(v)=H, and C is a constant >0.
For each v, choose a Haar measure Jl.v on A~, so that Jl.v(R:)= 1 for
all finite places v of k; we may then assume that we have taken JJ.= nJJ.v,
in the same sense as has been explained in Chap. VII-4 for the case
A =k. By following step by step the proof of prop. 10, Chap. VII-4, one
finds that the integral ZA(S) is absolutely convergent, and equal to the
infinite product
whenever the factors in that product, and the product itself, are absolutely
convergent. Those factors have been calculated in propositions 7 and 8
of Chap. X-3; the absolute convergence of Z A(S) for Re(s) > n is then an
immediate consequence of the latter results, combined with prop. 1 of
Chap. VII-1. The same results, combined with the definitions in Chap.
VII -6, give now the final formula in our proposition for the case A = Mn(k);
combining this with the corollaries of propositions 7 and 8 of Chap. X-3,
one obtains at once the general case of the same formula.
204 Simple algebras over A-fields XI
One should note that the middle product, in the formula for Z .is) in
proposition 1, is a finite one by th.1 of § 1, since that theorem shows that
d(v) = 1 for almost all the places of k. It should also be pointed out that
the computation of the constant C in that formula, for an explicitly given
Haar measure J.l on AA, offers no difficulty, and that it is important for
some purposes, e.g. for the determination of the "Tamagawa number" of
the subgroup A(l) of AX given by v(x) = 1. As this lies beyond the scope of
the present volume, it will not be pursued any further.
PROPOSITION 2. Let D be a division algebra of dimension d2 over k,

and let ZD(S) be defined as in proposition 1. Then, if k is of characteristic 0,
ZD(S) has no other pole than s=o and s=d; if k is of characteristic p> 1
and has the field of constants F q , ZD(S) has no other pole than the zeros
of (l_q-S)(l_ qd-S).
This will be proved by following step by step the proof of tho 2, Chap.
VII-5. In analogy with that proof, it will be convenient to adopt the follow-
ing notations. For zeD1 and seC, write ws(z)=lv(z)LL W l is then a
morphism of DA into R~. For eeD x , v<e) is in k X ; therefore, by tho 5 of
Chap. IV-4, W 1 is trivial on DX. Consider first the case where k is of
characteristic p> 1 ; by prop. 6 of Chap. X-2, v maps D: onto k: for all v,
so that W 1 maps D; onto the subgroup ofR~ generated by qv; by corol-
lary 6 of tho 2, Chap. VII-5, this implies that W 1 maps D Aonto the sub-
group ofR~ generated by q, ifFq is the field of constants of k. In that case,
take Zl eDA such that W1(Zl)=q, and call M the subgroup of DA gen-
erated by z1; then DA is the product of M and of the kernel Dl of Wl.
On the other hand, if k is of characteristic 0, call M the subgroup of kA
defined in corollary 2 of tho 5, Chap. IV-4; k being identified with the
center of D, kA is to be considered as a subgroup of D ~ ; as v(z) = Zd for
ze k, the corollary we have just quoted shows that W l maps M onto R ~, so
that DA is again the product of M and of the kernel Dl of W 1 • In both
cases, tho 4 of Chap. IV-3 shows that DJ./D x is compact.
As in the proof of tho 1, § 1, take a basic character X of kA ; identify DA
with its topological dual by (x,y)=x(r(xy»), and, for each v, identify Dv
with its topological dual by (x,y) = xlr(x y»). For each v, call1Xv the self-
n
dual Haar measure on Dv; then, by corollary 1 ofth. 1, Chap. VII-2, the
measures IXv are coherent, and IX = IXv is the Tamagawa measure on DA.
n
Let again ~= ~v be the standard function on DA which was used
n
above in the construction of Z D(S), and let If' = If'v be any standard
function on D A • Call Z(If',s) the integral obtained by replacing ~ by If'in
the definition of ZD(S); with the present notations, this can be written:
Z(Y', s) = J Y'(z)ws(z)d Jl(z).
D~
It will now be shown that this is absolutely convergent for Re(s) > d and
that it can be continued as a meromorphic function in the whole s-plane,
with no other poles than those mentioned in proposition 2 ; this will then
contain that proposition as a special case. As to the convergence, we have
'l'v = <P v for almost all v, by the definition of a standard function. For every
finite place v of k, the support Rv of <P v is an open subgroup of Dv' and the
support Sv of 'l'v is compact; then, choosing avEk~ such that avSvcR v,
and putting yv=sup I'l'vl, we have 1'l'v(x) 1:S;;Yv<Pv(avx) for all xEDv.
Similarly, the definition of standard functions shows at once that, for
any infinite place v, one can find Bv and Yv in R~ such that l'l'v(x)l:s;;
:S;;Yv<Pv(BvX) for all xEDv. This shows that there is aEk~ and YER~
such that 1'l'(x)1 :S;;y<P(ax) for all XEDA- Therefore the integral Z('l',s),
for Re(s) = a, is majorized by
y J<P(az)w..(z)d J.l(z) = yw .. (a-1)Z D(a),
D~
which, by prop. 1, is convergent for a> d.

Now take the same two functions Fo, FI as in the proof of tho 2 of
Chap. VII-5; Z('l',s) is then the sum of the two integrals
Zi= J<P(z)ws(z)Fi(W I (z))dJ.l(z).

D~
Exactly as in that proof(but taking now B > d), we see that Zo is absolutely
convergent for all s and defines therefore an entire function of s, and that
the same is true for the integral Z~ obtained by replacing 'l' by its Fourier
transform 'l", s by d-s and Fo by t -+Fl(C I ) in the definition of Z00 Just
as there, one can also apply Poisson's summation formula (i. <;. formula (1)
of Chap. VII-2), in combination with lemma 1 of Chap. VII-2, to the
function x-+ 'l'(z x) on DA; in applying the latter lemma, one has to use the
fact that the module of the automorphism x-+z- 1 X of DA , for zED~, is
Then, proceeding exactly as in the proof in question, one finds that

Z('l', s) is the sum of the entire function Zo + Z~ and of the integral
f(s) = J ('l"(0) -wiz) 'l'(0)) ws-iz) FI (WI (z))d J.l(Z).

MID
Here the integnmd is constant on the cosets of the compact subgrollP

G I =D!!D X of the group G= D~/Dx. As G I is the kernel of the morphism
of G into R~ determined by WI' we may identify GIG! with the image
N of G in R ~ under that morphism, which is R ~ or the group generated by
q, according to the characteristic of k. In view of lemma 6 of Chap. VII -5,

and taking for v the measure occurring in that lemma, we have therefore
(up to a constant factor which may be made equal to 1 by a suitable
choice of J1) :
f(s) = J(tp'(O)-n d tp(O))ns - d F1(n)dv(n)= tp'(O)A(s-d)- tp(O)A(s)
N
where A is as defined there. In view of the statement about the poles of A

in that lemma, this completes our proof.
Now the comparison between propositions 1 and 2 will give us the
main result of this Chapter.
THEOREM 2. A simple algebra A over an A~field k is trivial (f and only if

it is everywhere un ramified, i. e. if and only if Av is trivial over kv for every
place v of k.
It is clearly enough to prove this for a division algebra D. If D,. is
trivial for all v, prop. 1 shows that its zeta-function Z D(S), up to a constant
factor, is given by
n Zk(s-i).
d-l
ZD(S)=
;=0
I n view of theorems 3 and 4 of Chap. VII -6, this has poles of order 2 at
s = 1,2, ... , d - 1 if d> 1. By prop. 2, this cannot be. Therefore d = 1, and
D=k.
Actually the combination of propositions 1 and 2 allows one to dra w
stronger conclusions than theorem 2; for instance, it shows at once that,
if d> 1, D must be ramified at least at two places of k. This need not be
pursued any further now, since much stronger results will be obtained in
Chapter XIII.
§ 3. Norms in simple algebras. As a first application of theorem 7..

* we will now reproduce Eichler's proof for the following:
PROPOSITION 3. Let A be a simple algebra over an Ajzeld k, and let v be
the reduced norm in A. Then v(A X) is the subgroup y of k x ,consisting of the
elements whose image in kv is > 0 for every real place v of k where A is
ramified.
That proof depends upon the following lemmas.
LEMMA 1. Let K be a commutative pjzeld, L=K(~) a separably alge-
braic extension of K of degree n, and put
n
F(X)=N L/K(X -~)=X·+ L a;X·- i•
;= 1
§ 3. Norms in simple algebras 207
n
Let G(X)= LbiXn - i be a polynomial of degree n-l in K[X]. Then,
i= 1
if all the coefficients of G are sufficiently close to 0 in K, the polynomial
F + G is irreducible over K and has a root in L.
It will be convenient to extend modg to a mapping x--+lxl of an
algebraic closure K of L into R+ by putting Ixl = moddx)l/v whenever
K(x)cK'cK., and K' has the finite degree v over K; by corollary 2
ofth. 3, Chap. 1-2, this is independent ofthe choice of K' when x is given
in K. TakeAER~ such that lail~Ai for l~i~n, and assume, for some
B<A, that Ib;l ~Bi for 1 ~i~n. Let" be any root of F+ G in K; then we
have
,,"=- n
L(ai+bi)",,-i
i= 1
and therefore
hence 1,,1 ~A. Call el, ... ,e" the roots of F; they are all distinct, since L is
e
separable over K, and they are the images of under the automorphisms
of Kover K. What we have proved for" can be applied to the ev, by taking
G=O, so that levi ~A for 1 ~v~n. Now put
we have O<IX~A. Assume now that we have taken B<A(IX/A)". As " is

a root of F+ G, we have
" II
D(,,-ev)=- Lhj"n-i
v= 1 ;- 1
and therefore
so that there is v such that I" - evl < IX. Clearly this implies that I" - ell I ~ IX
for all Jl. =1= v. Let (J be an automorphism of Kover K, mapping ev onto e.
After replacing" by.,,", which is also a root of F+ G, we see that I,,-el <
<IX and 1,,-evl~1X for all ev=l=e. Assume that L is not contained in K(,,);
then there is an automorphism t of Kover K(,,) such that (=1= as this e;
must leave I" - el invariant, we get a contradiction. Therefore K(,,) ~L;
as" is at most of degree n over K, this implies that K(,,)=L and that
F+ G is irreducible.
Incidentally, since every extension of K of degree n can obviously be
generated by a root of a monic polynomial F of degree n with coefficients
in the maximal compact subring of K, lemma 1 shows that K has at most
* finitely many separable extensions of given degree, hence also (by corol-
lary 2 of prop. 4, Chap. 1-4) finitely many algebraic extensions of given
degree.
LEMMA 2. Let K be a commutative p-field, R its maximal compact sub-
ring, and L an unramified extension of K. Then, for every xeR x, there is
yeL.such thatNL/K(y)=x and that K(y)=L.
Call n the degree of Lover K; call (j the number of divisors of n ; since

L is cyclic over K, (j is also the number of distinct fields between K and L.
We will first constr:.:ct eeL such that N L/K(e)=1 and that L=K(ei)
for 1:::; i ~(j. Take a common multiple D of the integers 1,2, ... , (j, e. g.
D = (j!. Call (X a generator of the Galois group of Lover K. For 1 ~ h:i:; n - 1,
consider the mapping
of L into itself. This is a polynomial mapping, when L is regarded as a

vector-space over K, as one sees at once by choosing a basis for Lover K
e
and expressing in terms of that basis. Taking again for K an algebraic
closure of L, we can extend the mappings Ph to the algebra .!l' = LQ!h R
over K. Now apply prop. 3 of Chap. III-2 to that algebra and to the n
distinct isomorphisms (Xi of L into K, for 0 ~ i ~ n - 1. As in that propo-
sition, call Jij the K-linear extension of (Xi to .!l', and put lfJ = (Jio,··· ,Jin-I);
that proposition shows that cp is an isomorphism of fI' onto Kn. Then the
mapping JiooPhocp-1 of KR into K is given by
(xo, ... ,xn-I)--+(Xh+ I XO)D -(XhXI)D,
where it should be understood, for h = n - 1, that Xn = Xo' As this is not O.
and as K is an infinite field, we see now that none of the Ph is 0 and that
one can choose eeL such that Ph(e)+O for 1 ~h~n-l. Let be so e
chosen, and put e = e" e- I. Then N L/K(e) = 1, and the images (e"h)D of eD,
under the automorphisms (Xh with 1 ~ h ~ n -1, are all +
eD, so that
L=K(e D). As D is a mUltiple of i for l~i~(j, we have, for each such i,
eD=(ei)D/i, hence K(eD)cK(e i), so that L=K(ei). Now take any xeRx;
by prop. 3 of Chap. VIII-I, we can write it in the form x=NL/K(YI) with
YI eC. Consider the infinite sequence of fields Ki= K(e iYI)' for all i~O.
At most (j of them can be distinct; therefore there are pairs (i,j) of integers
such that O~i<j and Ki=K j, and, if we take such a pair for whichj-i
has the smallest value, we have O<j-i~(j. As eiYI and ejYI are both in
K i, ej - i is in K i. In view of our choice of e, this implies Ki=L. Thus
y=eiYI satisfies the requirements in our lemma.
We can now proceed to prove proposition 3. Call n 2 the dimension of
the given algebra A over k, and Roo the set of the infinite places of k where
§ 3. Norms in simple algebras 209
it is ramified. If v is in Roo, it must be real, and Av must be isomorphic to

an algebra M m(H); as this implies n = 2 m, this can only happen if n is
even, and, of course, if k is of characteristic O. We have seen in Chap. X-2
that the reduced norm v maps Mm(Hr onto R~; therefore v(AX) is
contained in the group y defined in proposition 3. Choose now a non-
empty finite set R' of finite places of k, containing all the finite places of k
where A is ramified, and put R=R'uR oo . Take any veR', and a prime
element 1tv of kv; by prop. 6 of Chap. X-2, there is x,.e Av such that v(xv) =
=1tv' Apply corollary 2 ofth. 3, Chap. IV-2, to A and to some place Vo ofk,
not in R'; it shows that we can choose (XeA so that its image in Av is
arbitrarily close to Xv, and that, for all w=l= v in R', its image in Aw is arbi-
trarily close to 1. In view of the continuity of v, this can be done so that
the image of v«(X) in kv is so close to 1tv as to be a prime element of km and
r: ;
that its image in Ie,. , for every w=l= v in R', is so close to 1 as to be in then
(XeA x, since v«(X) =1= O. For each veR', choose an element (Xv of AX in this
manner. Now take any ~ in the subgroup 'l' of k x defined in our proposi-
tion; we have to show that it is in v(A X). For each veR', put n(v)=
n
= ord v(~)' and put (X = (X:<V); after replacing ~ by ~ V«(X)-1, we see that
it is enough to prove our assertion under the additional assumption that
ordv(~)= 0 for all ve R'. For each place veR', take an unramified extension
k~ of kv, of degree n over kv' By lemma 2, there is Yvek~ such that ~ =
=Nk~/dYv) and k~=kv(yv)' As Yv is then of degree n over kv, it is the root
of an irreducible polynomial Fv of degree n over kv, given by:
,,-1
Fv(X)=Nk~/dX -Yv)=X"+ L ai.vX,,-i+( -1)"~,
i=1
with ai.vek v for 1 ~i~n-l. For each veR oo ' put ai.V=O for 1 ~i~n-1,
and consequently, since the existence of such a place implies that n is even:
Fv(X)=X"+( -1)" ~ =X"+ ~;
then, because of our assumption ~ e y, Fv has no root in kv = R, so that the
same is true of every monic polynomial of degree n over R whose coeffi-
cients are close enough to those of F v' Applying corollary 2 of tho 3, Chap.
IV-2, to k and to some place Vo of k, not in R, we see that we can choose
wie k, for 1 ~ i ~ n -1, so that its image in kv is arbitrarily close to ai. v for
every veR. In view oflemma 1 and what has just been said, this can be
done so that the polynomial
11-1
F(X)=X"+ L WiX i + (-1)"~
n-
i= 1
has the following properties: (a) for every veR', F is irreducible over kv
and has a root in k~; (b) for every veR oo ' F has no root in kv=R. As R' is
not empty, (a) implies that F is irreducible over k and has no mUltiple
roots. Call Ca root of F in some algebraic closure of k, and put k' =k(C).
Take any veR', and a place w of k', lying above v; as the completion of k'
at w must then be generated over kv by a root of F, this completion, by (a),
is isomorphic to k~, with which we can identify it; as it is of degree n over
kv, corollary 1 of tho 4, Chap. 111-4, shows that w is the only place of k'
lying above v, and that theorem shows then that we may identify k~ = k~
with k'®"k v. Similarly, (b) shows that, if v is in Roo, all the places of k'
lying above v are imaginary.
Now consider the algebra A' = A", over k'. Take any place w of k', and
call v the place of k lying below w. By the elementary properties oftensor-
products, A~, which is the algebra A' ®It' k~ over k', may be identified in an
obvious manner with Av®k.k~. As Av is trivial over kv for v not in R,
this shows that also A~ must then be trivial. If v is in Roo, w is imaginary,
so that k~=C and that A~ is trivial. Finally, let v be in R', and write Av as
Mm(v)(D(v)), where D(v) is a division algebra over kv; if its dimension over
kv is d(V)2, we have n=m(v)d(v), so that d(v) divides n. Then k~, which is
unramified, hence cyclic, and of degree n over kv, contains a field k"
which is of degree d(v) over kv, and is of course unramified over kv. By
prop. 5 of Chap. 1-4, D(v) contains a field isomorphic to k" ; therefore, by
corollary 6 of prop. 3, Chap. IX-l, D(V)k" is trivial over k"; obviously this
implies that (A v)". is trivial over k", hence that A~ = (A v)";" is so over k~.
Having thus shown that A' is unramified at all places of k', we can
conclude, by tho 2 of § 2, that it is trivial over k', which is the same as
to say that A has a k'-representation into Mn(k'). Therefore, by tho 2 of
Chap. IX-3, A has an t;-regular factor-set, if t; is the Galois group over
k' of the separable algebraic closure kscp of k'. Then, by lemma 4 of
Chap. IX-3, we can construct an algebra of dimension n2 over k, con-
taining a field isomorphic to k', with the same factor-set as A; as this
implies that it is similar to A, and as it has the same dimension as A
over k, it is isomorphic to A and may be identified with it. As shown
there, we have then V(C·lA)=N"",,(C)=e.
§ 4. Simple algebras over algebraic number-fields. We will now

combine the results of § 1 with some of those of Chapter V in order to
obtain a few basic results in the theory of ideals in simple algebras over
algebraic number-fields.
In this §, k will be an algebraic number-field, t its maximal order,
and all algebras will be simple algebras over k. We recall that, by propA
of Chap. V-2, if L is any k-Iattice in a vector-space E over k, and if v
is a finite place of k, the closure Lv of Lin Ev is the rv-module generated
by Lin Ev.
§ 4. Simple algebras over algebraic number-fields 211
Let D be a division algebra over k. As in Chap. X-1, let V, V', V"

be left vector-spaces of finite dimension over D, other than {O}; put
H = Hom(V, V'), H' = Hom(V', V"), H" = Hom(V, V"). If X, X' are sub-
groups of the additive groups of H and of H', respectively, we write
X X', as usual, for the subgroup of H" generated by the elements ,,'
for ,eX, e'eX'; it is easily seen, e. g. by taking bases for V, V', V" over
D, that HH' =H". Now let L, 1: be k-Iattices, in H and in H', respectively,
when these are regarded as vector-spaces over k; then L1: is obviously
a finitely generated t-module in H", and, as H" = H H', it is a k-Iattice
in H".
PROPOSITION 4.
Let A be a simple algebra over k. Then there are
maximal orders in A; these are k-Iattices in A, and a k-Iattice R in A is
a maximal order if and only if its closure Rv in Av is a maximal order
in Av for every finite place v of k. Every order in A is contained in a
maximal order.
If R is any order in A, the t-module generated by R in A is also
an order, and it is a k-Iattice; this shows that, unless R is a k-Iattice,
it cannot be maximal. Let X be any k-Iattice in A; the last part of tho 1,
§ 1, may be expressed by saying that X v is a maximal order in Av for
almost all V. Then tho 2 of Chap. V-2 shows that there is a one-to-one
correspondence between the orders R in A which are k-Iattices, and the
possible choices of an order Rv in Av for every finite place v of k, subject
to the condition that Rv be a kv-Iattice for all v, and Rv=Xv for almost
all v; if R is given, Rv is the closure of R in A v, and, if the Rv are
given, R is defined by R=n(A('\R v)' In view ofth. 1, Chap. X-1, all our
assertions are now obvious.
PROPOSITION 5. Let D be a division algebra over k. Let V, W be two
left vector-spaces of finite dimension over D; put H = Hom(V, W) and
A = End(V). Let M, M' be two k-Iattices in H. Then the set X of the
elements , of A such that ,M e M' is a k-Iattice in A, whose closure in
A v, for every finite place v of k, is the set X v of the elements x of Av
such thai xMveM~. If M =M', X is an order of A.
For every v, by prop. 4 of Chap. X-t, Xv is a kv-Iattice in A v, and it
is an order if M v = M~. Let L be any k-Iattice in A; as we have seen
above, LM is a k-Iattice in H, whose closure in HI» for every v, is clearly
LvMv' Therefore, for almost all v, we have LvMv=Mv=M~; this
implies that, for almost all v, X v is an order and contains Lv' As Lv is
a maximal order in Av for almost all v, we see that X v= Lv for almost
n
all v. Therefore, by tho 2 of Chap. V-2, there is a k-Iattice X' = (A('\X v)
in A with the closure Xv for all V. Clearly X eX'; conversely, since
X'MeM~ for every v, and M'=n (H(')M~), we have X'eX. This

completes the proof, except for the last assertion, which is now obvious.
If notations and assumptions are as in proposition 5, and if M = M',
the set X is called the left order of M. Exchanging right and left, we see
that the set of the elements" of B=End(W), such that M"eM, is an
order of B; this is called the right order of M.
PROPOSITION 6. Let V, Wand M be as in proposition 5; assume that

there is a maximal order R of A = End(V), such that M is a left R-module.
Then R is the left order of M, and its right order is a maximal order of
B=End(W).
This is an immediate consequence of prop. 5, combined with tho 2

of Chap. X-l.
With the same notations and assumptions as in propositions 5 and
6, a k-Iattice M in Hom(V, W) with the left order R and the right order
S will be called an (R, S)-lattice; it is called a normal lattice if R or S,
and consequently both Rand S, are maximal orders. If V = W, hence
H = A = B, a normal lattice is also known as a "normal fractional ideal".
Clearly, in that case, the three relations M· M e M, MeR, Me S are
equivalent; when they hold, M is a left ideal in the ring R and a right
ideal in the ring S; it is then called a normal ideal and an (R, S)-ideal.
By using the above results and those of Chap. X, one can see at once
that, if Rand S are any two maximal orders in A, there are always
(R,S)-ideals. Furthermore, if a normal (R,S)-lattice M is a maximal left
ideal in R, i. e. if MeR, M =1= R, and if there is no left ideal other than
Rand M between Rand M, then it is a maximal right ideal in S, in the
same sense; when that is so, one must have Mv=Rv=Sv for all finite
places v of k except one. If the multiplication law (M,M')-+MM' is
restricted to those pairs (M,M') of normal lattices in A for which the
right order of M is the same as the left order of M', the normal lattices,
for this law, make up a so-called "groupoid" whose units are the maximal
orders of A. It is also easily seen that, for this law, every normal ideal
can be written, although in general not uniquely, as a product of maximal
ideals. For two-sided ideals and (R,R)-lattices, one has a more precise
result:
PROPOSITION 7. Let R be a maximal order in A. Then, for the law

(M,M')-+MM', the (R,R)-lattices in A make up a commutative group; it
is the free group generated by the maximal two-sided ideals in R; for every
prime ideal p in t, there is one such ideal, and only one, between Rand pRo
This follows in a quite straightforward manner from the above re-
sults and corollary 2 of tho 2, Chap. X-I.
Chapter XII
Local classfield theory

§ 1. The formalism of c1assfield theory. The purpose of classfield
theory is to give a description of the abelian extensions of the types of
fields studied in this book, viz., local fields and A-fields. Here we
assemble part of the formal machinery common to both types.
LEMMA 1. Let G = G 1 X N be a quasicompact group, G 1 being compact

and N isomorphic to R or Z; let H be an open subgroup of G. Then, if
H is contained in G 1 (i.e. if it is compact), N is isomorphic to Z, and H
is of finite index in G1; otherwise it is of finite index in G.
Put HI =HnG 1; as this is open in G1, and G1 is compact, it is of

finite index in G1; this proves the first assertion. As H nN is an open
subgroup of N, it is N if N is isomorphic to R; therefore H = HI X N in
that case, and G/ H is isomorphic to G1/H 1. If N is isomorphic to Z, let
n1 be a generator of N; if H is not contained in G1, it has an element
of the form gln~ with glEG1, IlEZ, 1l1-0. As GdHl is finite, there is
v1- 0 such that g~ E H 1. Thennr is in H, so that H contains the group
H' generated by HI and nr. As H' is obviously of finite index in G,
this proves the lemma. TheoreJ'""_ I of Chap. IV-4 may be regarded as
the special case where G = k~/k x, H being the image of Q(P) in k~/k x •
LEMMA 2. Let G = G 1 X N, G' = G'l X N' be quasicompact groups, G 1

and G'l being compact and N, N' isomorphic to R or Z. Let F be a morphism
of G' into G but not into G1. Then F- 1(G 1)=G'l; the kernel of F is com-
pact; F(G') is closed in G, and G/F(G') is compact.
As G1 is the maximal compact subgroup of G, F(G~) is contained in

G 1. For n'EN', call f(n') the projection of F(n') onto N in G; f is then
a non-trivial morphism of N' into N, hence, obviously, an isomorphism
of N' onto a closed subgroup of N with compact quotient; our first and
second assertions follow from this at once. We also see now that F
induces on N' an isomorphism of N' onto F(N'), and that F(N')nG 1 = {I};
therefore F(G') is the direct product of F(G'l) and F(N') and is closed.
Finally, G/G 1F(N') is clearly isomorphic to N/f(N'), hence compact; as
the kernel of the obvious morphism of G/F(G') onto G/G 1F(N') is the
image of G1 in G/F(G'), hence compact, G/F(G') must also be compact.
214 Local classfield theory XII
From now on, in this §, we will consider a field K; later on, this will
be either a local field or an A-field. As in Chapter IX, we write K for an
algebraic closure of K, K.ep for the union of all separable extensions of K
contained in K, and (fj for the Galois group of Ksep over K, topologized
as usual. We will write Kab for the maximal abelian extension of K con-
tained in K; this is the same as the union of all abelian extensions of K
of finite degree, contained in K, i. e. of all the Galois extensions of K
of finite degree, contained in K, whose Galois group is commutative;
by definition, this is contained in Ksep. We denote by (fj(l) the subgroup
of (fj corresponding to K ab ; this is the smallest closed normal subgroup
of (fj such that (fj/(fj(1) is commutative; it is therefore the same as the
"topological commutator-group" of (fj, i. e. the closure of the subgroup
of (fj generated by the commutators of elements of (fj. We write m: for
the Galois group of Kab over K; this may be identified with (fj/(fj(l); it
is a compact commutative group. Let X be any character of (fj; as in
Chap. IX-4, call ~ its kernel and L the subfield of Ksep corresponding
to ~, which is the cyclic extension of K attached to X; clearly L c: Kab
and ~::::> (fj(l), so that we may identify X with a character of m, for which
we will also write X. Conversely, every character of m: determines in an
obvious manner a character of (fj, with which we identify it. Thus the
group of characters of (fj, for which we will write X g, is identified with
the group of characters of m:; the latter is the same as the dual m:* of
m:, except that we will always write the group X g multiplicatively; we
put on X x the discrete topology, this being in agreement with the fact
that the dual of a compact commutative group is always discrete. By
the duality theory, the intersection of the kernels of all the characters
of m: is the neutral element; this is the same as to say that the intersection
of the kernels ~ of all the characters X of (fj is (fj(l), or also that Kab is
generated by all the cyclic extensions L of K; this is of course well-known.
Let K' be any field containing K; as in Chap. IX-3, we take an
algebraic closure K' of K' and assume at the same time that we have
taken for K the algebraic closure of K in K'; then, as we have seen
there, Ksep is contained in K~ep, and, if 6)' is the Galois group of K'sep
over K', the restriction morphism p of (fj' into (fj is the one which maps
every automorphism of K~ep over K' onto its restriction to Ksep. Obviously
p maps 6)'(1) into (fj(1), so that it determines a morphism of m' = (fj'/(fj'(1)
into m: = (fj/(fj(l), which we also denote by p and call the restriction
morphism of m' into m. It amounts to the same to say that Kab is con-
tained in K~b' and that p maps an element (I.' ofm:', i.e. an automorphism
of K~b over K', onto its restriction to K ab . Correspondingly, X-Xop is
a morphism of Xx into XX"
In classfield theory, one defines a "pairing" of the group X K of the
characters of 6) (or, what amounts to the same, of m:) with a locally
§l. The formalism of classfield theory 215
compact commutative group GK , invariantly associated with K. In this

Chapter, where K will be a local field, we will take GK = K x ; in the next
one, K will be an A-field, and we will take at first GK=K~ and later on
GK= K~/K x. This pairing, which will be called the canonical pairing, is a
mapping of X K x GK into C X , whose value, for XEX K and gEG K, will be
written as (X,g)K; to begin with, we assume that it satisfies the following
condition:
[I] (i) For all X, X' in XK, and all g, g' in GK:
(X X',g)K = (X,g)K'(x',g)K' (X, gg')K= (X, g)K'(x,g')K;
(ii) (X,g)-+(X,g)K is a continuous mapping of X K x GK into C X •
As X K is discrete, the pairing is continuous, i.e. [I(ii)] is satisfied, if

and only if g-+(X,g)K is a continuous mapping of GK into C X for every
XE X K; then [I (i)] implies that it is a character of GK , of an order dividing
that of X. Consequently, if [I(i)] is assumed, [I(ii)] is equivalent to the
following:
[I(ii')] For every XEXK' the kernel of g-+(X,g)K is an open subgroup of
GK •
Assume that such a pairing has been given. Then, for each gEG K,
X-+(X,g)K is a character of X K. As X K is the same as the dual of 2l, the
duality theory shows that this can be uniquely written as X-+X(C() with
c(E2l. We will write a, or, when necessary, aK for the mapping g-+C( of GK
into 2l determined in this manner. Obviously we have a(g g') = a(g) a(g')
for all g, g' in GK , and the continuity of our pairing, i. e. condition [I(ii)] ,
implies at once that a is continuous. Thus a is a morphism of GK into 2l,
determined by the relation
(1) (X,g)K = X(a(g»),
which is valid for all XEXK and all gEG K. We will call a the canonical
morphism of GK into 2l.
Now we assume the additional condition:
[II] If (X,g)K= 1 for all gEG K, thell x= 1.
In view of [I], this is clearly equivalent to either one of the following
conditions:
[II'] X-+xoa is an injective morphism of X K into the group of characters
of GK •
[II"] The image a(GK) of GK by a is dense in 2l.
We also add the assumption that GK should be quasicompact, or rather
a more precise one, which is as follows:
[III] Either (a) GKis the direct product of a compact group Gi and of a
group N isomorphic to R, or (b) GK is the direct product of a compact group
Gi and of a group N isomorphic to Z, and there is, for each integer n ~ 1,
a character XeXK of order n such that (X,g)K= 1 for all geGi.
The two cases in [III] will be referred to as case [11I(a)] and case
[III (b)] , respectively. In both cases, as has been observed in Chap. VII-3,
Gi may be characterized as the unique maximal compact subgroup of GK •
From now on, we will write UK for the kernel of the canonical mor-
phism a ofG Kinto~; it is the intersection ofthe kernels of the characters
xoa of GK, i.e. of the characters g--+(X,g)K' for all XeXK'
PROPOSITION 1. In case [1I1(a)], the canonical morphism a determines an
isomorphism of GK/UK onto ~; every character of GK, trivial on UK' can
be uniquely written as Xoa with XeXK; and x--+ xoa is an injective morphism
of X K into the group of characters of finite order of GK.
As every character Xof~ is offinite order, the last assertion is no more
than a restatement of [II']. For every Xe X K, XO a induces on the subgroup
N of GK a character of N offinite order; as N is isomorphic to R, there is no
such character except the trivial one. Therefore N c UK; if we put Ui =
= UKnGi, we have UK = ui x N, and GKiUK may be identified with
Gi/Ui; as this is compact, a determines an isomorphism of that group
onto a closed subgroup of~, hence onto ~ itself, by [II"]. Then, by the
duality theory, x--+ xoa is the "dual" or "transpose" of a, hence an iso-
morphism of X K onto the subgroup ofthe group of characters of GK which
is associated by duality with UK; this subgroup consists of the characters
of GK, trivial on UK'
COROLLARY. In case [III (a)] , every character of Gi, trivial on Ui=
= UKnGi, can be uniquely extended to a character of GK of the form
xoa.
In fact, it can be uniquely extended to a character of GK , trivial on N!;
this is then trivial on UK and is as required.
PROPOSITION 2. In case [11I(b)], call Xo the subgroup of X K con-
sisting of the characters X such that (X,g)K= 1 for all geGi; call n 1 a
generator of the subgroup N of GK. Then x--+(x,n 1 )K is an isomorphism of
X 0 onto the group of all roots of 1 in C.
As every XeXK is of finite order, (x,g)K is always a root of 1, for all X
and all g. As GK is generated by Gi and n1 , a character of GK which is
trivial on Gi is uniquely determined by its value at n1 ; in view of [II'],
this shows that x--+(x,n 1 )K is an injective morphism of Xo into the group
of roots of 1 in C; in particular, it maps every character X of order n,
belonging to X 0, onto a primitive n-th root of 1 in C. By [III (b)] , there are
such characters for every n ~ 1; therefore the image of X 0 by that mor-
phism contains all the roots of 1 in C.
§l. The formalism of c1assfie1d theory 217

G} is the set of the elements g of GK such that (X,g)x= 1 for all xeX o.
Let v be any integer other than 0; by proposition 2, there is XeXo
such that (x,nI)Kfl, hence (x,n1g)Kf 1 for all geGk:. As GK is the union
of the co sets n1. Gk: for all ve Z, this proves our assertion.
COROLLARY 2. In case [III(b)], the kernel UK of the canonical morphism
a is contained in Gk:; a determines an isomorphism of Gk:/u Konto the inter-
section m:o of the kernels in m: of the characters Xe X 0; and a -l(m: O) = Gk:.
The first and last assertion follow at once from corollary 1. Put
~ = a( Gk:); clearly ~ is compact, and a determines an isomorphism of
Gk:/U K onto ~; moreover, by the definition of Xo, a character X of m:
belongs to X 0 if and only if it is trivial on ~, so that ~ = m:o.
COROLLARY 3. In case [III(b)], every character of Gk:, trivial on UK,
is of finite order and can be extended to a character of GK of the form
xoa, where X is a character of m:.
By corollary 2, every character of Gi, trivial on UK' can be written as
Xl oa 1 , where Xl is a character of m:o and a 1 is the morphism of Gk: onto
m:o induced by a. As X1 can be extended (although not uniquely) to a
character Xofm:, and as every character ofm: is of finite order, this proves
our assertions.
COROLLARY 4. In case [III(b)], the mapping X~ xoa is a bijective
morphism of X K onto the group of the characters of GK of finite order,
trivial on UK; it maps X 0 onto the group of the characters of GKof finite
order, trivial on Gk:.
All we need show is that the mappings in question are surjective. Take
first a character IjJ of GK of finite order, trivial on Gk:. As IjJ (n l ) is then a
root of 1 in C, proposition 2 shows that there is XeXo such that (X,n1)K=
= IjJ (n 1); then xoa coincides with IjJ on Gk: and at nl> hence on GK •
Now take any character t/I of GK of finite order, trivial on UK; by corol-
lary 3, we can find XeXK such that IjJ coincides with xoa on Gk:; then
1jJ' = 1jJ. (Xo a) - 1 is trivial on Gk: and of finite order, so that, by what we
have just proved, it can be written as X' 0 a. This completes our proof.
PROPOSITION 3. Assume [I], [II] and [III], and call a the canonical
morphism of GK into m:. For every extension L of K of finite degree,
contained in K ab , call ~(L) the subgroup of m: corresponding to L, and put
N(L) = a - I(~(L)). Then ~(L) is the closure of a(N(L)) in m:; L consists of
the elements of Kab which are invariant under a(g) for all geN(L); a
determines an isomorphism of GK/N(L) onto the Galois group of Lover K;
and L~ N(L) is a one-to-one correspondence between subfields L of Kab
of finite degree over K, and the open subgroups of GK of finite index in

GK, containing UK'
As ~(L) is open in~, N(L) is open in GK • By [II"], a(GK ) is dense in~;

this implies that a(N(L») is dense in ~(L) and that a determines an iso-
morphism of GKiN(L) onto ~/~(L), which is the same as the Galois
group of Lover K. As the operation of ~ on Kab is continuous, every
element of Kab which is invariant under a(N(L») is invariant under its
closure ~(L), so that it is in L. Finally, let H be any open subgroup of GK
of finite index n, containing UK; call I/Ii' for l~i~n, all the distinct
characters of GK , trivial on H; then H is the intersection oftheir kernels.
By prop. 1 in case [III(a)] and by corollary 4 of prop. 2 in case [III(b)],
we can write I/Ii = XiO a for 1 ~ i ~ n, the Xi being characters of ~; by [II'],
the Xi are uniquely determined and make up a finite subgroup of X K,
since the I/Ii make up a finite subgroup of the group of characters of GK •
Call ~ the intersection of the kernels of the Xi in ~; it is an open subgroup
of~, of index n; therefore the subfield L of K ab , corresponding to ~, is
of degree n over K. Clearly H = a- 1 (~), hence H = N (L). This completes
our proof.
COROLLARY. In case [III(b)], call Ko the subfield of Kab corresponding

to the subgroup ~o=a(Gi) of~. Then, for each integer v~ 1, Ko contains
one and only one extension Ky of K of degree v; this is the cyclic extension
of K, attached to anyone of the characters of order v, belonging to X 0; and
N(Ky) is the subgroup of GK generated by Gi and ni.
By corollary 2 of prop. 2, we have Gi=a-l(~o); therefore, if Land

N(L) are as in proposition 3, we have Lc.Ko if and only if N(L):::JGi;
this implies that N(L) is generated by Gi and n1if v is the index of N(L) in
GK • Then, by proposition 3, L is cyclic of degree v over K, and, if X is a
character of~ attached to L, N(L) is the kernel of xoa, so that X belongs
to Xo and has the order v. Conversely, if X is such, the kernel ofxoa is
generated by Gi and n1, so that the cyclic extension attached to X is L.
Now we consider a cyclic extension K' of K, contained in Ksep. We

use the notations (fj', (fj,(l), ~' = (fj'/(fj,(l) as explained above, and write p
for the restriction morphism of (fj' into (fj and also for that of~' into m.
As K' is cyclic over K, (fj' is an open normal subgroup of (fj, with cyclic
factor-group; consequently, we have (fj:::J (fj':::J (fj(1):::J (fj,(l), and (fj'(l) is a
normal subgroup of (fj. For every AE(fj, the inner automorphism
0''''''' A0' A-1 induces on (fj' an automorphism of (fj'; therefore, if X' is any
character of (fj', we can define a character X';' of (fj' by putting, for every
0" E(fj', X'),(O") = x'(AU' A-1). Clearly x'),= x' if AE(fj', so that X' ...... X'), deter-
§l. The formalism of classfield theory 219
mines an operation of the Galois group ffi/6), of K' over K on the group
X K' of the characters of ffi'. ,
Furthermore, we assume that we have been given canonical pairings
(X,g)K' (x',g')K' of X Kwith a group GK, and of X K, with a group GK" both
of them satisfying [I], [II], [III]; to simplify notations, put G= GK ,
G' = GK" G 1 = Gi, G'l = Gi,; call a, a' the canonical morphisms of G into
~, and of G' into ~', respectively defined by these pairings. Assume also
that ffi operates on G', the action of any Ae ffi on G' being written as g' -+g';'
and satisfying the following condition:
[IV] (i) For Ae ffi', g' -+g';' is the identity on G'; (ii) For each Ae ffi,
g'-+g';' is an automorphism of G', and g,;'g'-l e Gi for all g'eG'; (iii) For all
x'eXK" g'eG' and Aeffi, we have:
(X,;',g';')K' = (X',g')K"
Finally, assume that we have been given a morphism F of G' into G, .
satisfying the following condition:
[V] (i) For all g'eG' and all Aeffi, we have F(g';')=F(g'); (ii) For all
XeXK and all g'eG', we have:
(xoP,g')K' = (X,F(g'»)K'
Clearly [V(ii)] may also be written as poa' = aoF.
PROPOSITION 4. Let K' be a cyclic extension of K; let a, a' be the canon-
ical morphisms respectively defined by canonical pairings of X K with a
group G, and of X K, with a group G', both satisfying [I], [II], [III]. Assume
that the Galois group ffi of Ksep acts on G', that F is a morphism of G' into
G, and that [IV] and [V] are satisfied. Then UnF(Gi)=F(U'nG1),
where U, U' are the kernels of a and of a'; moreover, we have UnF(G')=
=F(U') if G' satisfies [III(a)], or if G and G' satisfy [III(b)] and F does not
map G' into G1 •
By [V(ii)], we have poa'=aoF; therefore F(U') is contained in U,
hence in UnF(G'), and, if we put Ui = U'nG1, F(U1) is contained in
U nF(G'l)' Let '" be a character of G, trivial on F(U 1); then", of is a
character of G', trivial on U'l' Apply now the corollary of prop. 1, in the
case [III(a)], and corollary 3 of prop. 2, in the case [III(b)], to the character
induced on G'l by '" of; this shows that", of coincides on G'l with a
character ofthe form x'oa', with x'eX K ,. In other words, we have, for all
g'eG'l:
'" (F(g'») = (x',g')K"
By [IV (ii)], this must hold if we substitute g';' g' - 1 for g', with any g' e G'
and any Aeffi. In view of [V(i)], this gives
1=(X',g,;'g'-l)K' =(x',g';')K"(x',g')K,l,
and therefore, by [IV(iii)]:

(x',g')K' = (x',g")K' =(t~ -',g')K"
By [II], this shows that X' is invariant under A. for every A.e (fj; more expli-
citly, it is invariant under all the automorphisms of (fj' induced on (fj' by
inner automorphisms of (fj. Therefore the same must be true of the
kernel i>' of x', so that i>', which is an open subgroup of (fj' with cyclic
factor-group, is a normal subgroup of (fj. Let IX be a representative in (fj of
a generator of the cyclic group (fj/(fj'; let p be a representative in (fj' of a
generator of (fj'/i>' ; then (fj' is generated by i>' and p, and (fj is generated by
(fj' and IX, hence by i>', Pand IX. Consequently (fj/i>' is generated by the
images IX', p' of IX,P in (fj/i>'. As x' is invariant under (1' --+1X(1'1X- 1, we get,
for (1'=P,x'(P)=X'(IXPIX- 1 ). This shows that IXPIX- 1 P- 1 is in the kernel
i>' of X', so that IX' commutes with P' in (fj/i>'. Consequently (fj/i>' is com-
mutative. Therefore the character of (fj'/i>' determined by x' can be ex-
tended to a character of (fj/i>'. This is the same as to say that X' can
be extended to a character X of (fj, so that we have X' =xop. In view of
[V(ii)], the definition of X' gives now, for all g' eG'l :
1/1 (F(g'))=(xoP,g')K' = (X,F(g'))K'
This is the same as to say that 1/1 coincides with Xoa on F(G~), so that it
is trivial on U nF(G'l)' As F(U~) is a compact subgroup of G, and as we
have proved that every character 1/1 of G, trivial on F(U~), is trivial on
UnF(G~), we see that F(U~)=:J UnF(G~); in view of what we had proved
before, this completes the proof of the first part of our proposition. If G
and G' satisfy [III(b», we have U'cG'l and UcG 1 , by corollary 2 of
prop. 2; if F does not map G' into G1 , we have F- 1(G 1 )=G'l' by lemma 2;
U nF(G~) is then the same as U nF(G'), which completes the proof of the
second part in that case. Now assume that G' = G~ x N' with N' isomor-
phic to R. As we have seen before, for every x'eXK" the character of N'
induced on N' by x' 0 a', being of finite order, is trivial, so that N' c U',
hence U' = U~ X N' ; the same argument, applied to the character induced
on N' by xoaoF for XeXK, gives now F(N')c U, and therefore:
UnF(G')=(UnF(G~))·F(N')=F(U~)F(N')=F(U').
§ 2. The Brauer group of a local field. From now on, K will be a

local field. As in Chapter IX, we write B(K) for its Brauer group, H(K) for
the group of its factor-classes, and we identify these groups with each
other by means of tho 3, Chap. IX-3. In Chap. IX-4, we have already
determined these groups in the cases K =R, K = C, and we begin by
recalling the results found there, and introducing some additional nota-
tions which will be useful in the next Chapter. As B(R) has two elements, it
§ 2. The Brauer group of a local field 221
has a unique isomorphism '1 onto the subgroup {± 1} of C x ; for any

simple algebra A over R, we write h(A) = '1(CI(A»), and call this the
Hasse invariant of A; it is + 1 or -1 according as A is trivial or not. As
B(C) has only one element, we write '1 for the mapping which maps it
onto {+1}, and, for every simple algebra A over C, we write h(A)=
= '1 (CI(A») = + 1 and call this the Hasse invariant of A. For K =R, the
Galois group (f; of Ksep over K consists of the identity e and of the auto-
morphism x--+ x ofC overR; for K =C, (f;= {e}. For every character Xof
(f), and every (}eK\ we have defined in Chap. IX-4 the factor-class
{X, (}}; identifying H(K) with B(K) as we have said, we may now write, for
K=R or C:
(X, (})K = '1( {X, (}}).
Clearly this is 1 if K = C, or if K = R and X is the trivial character of (f;;

if K =R and X is the non-trivial character of (f;, our results of Chap. IX-4
show that it is + 1 or - 1 according as () > 0 or () < O. One verifies imme-
diately that this is a canonical pairing of X K with K x , in the sense of § 1,
and that it satisfies conditions [I], [II], [III(a)]; the kernel UK of the
canonical morphism is C x if K = C, and R ~ if K = R.
From now on, K will always denote a commutative p-field, except that
occasionally we will point out the validity of some of our results for K.=R
or C. As usual, we write R for the maximal compact subring of K, q for
its module, P for the maximal ideal of R, and 11: for a prime element of K.
We use the notations K, K sep , (f;, K ab , ~, as in § 1. Write IDl for the set of
all roots of 1 of order prime to p in K; this is clearly a subgroup of K;ep'
Put Ko = K(IDl), and call f>o the closed subgroup of (f; corresponding to
K o, i. e. consisting of the automorphisms of Ksep over K which leave
invariant all the elements of K o, or, what amounts to the same, all those of
IDl. By corollary 2 ofth. 7, Chap. 1-4, every finite subset ofIDl generates over
K an unramified extension of K. Conversely, every extension of K, con-
tained in an unramified extension, is itself unramified, so that, by corol-
lary 3 ofth. 7, Chap. 1-4, it is generated by a finite subset ofIDl; moreover,
by the same corollary, there is one and only one such extension KII of
degree n over K, for every n ~ 1. Consequently, Ko is the union of the
fields KII for all n~ 1. Again by corollary 2 of the same theorem, the map-
ping Jl. --+ Jl. q of IDl into itself is an automorphism of IDl, and, for every
n ~ 1, there is one and only one automorphism of KII over K, viz., the
Frobenius automorphism, which coincides with that mapping on IDlnKII .
Clearly this implies that there is one and only one automorphism ({Jo of
Ko over K which induces Jl.--+Jl.q on IDl; this will be called the Frobenius
automorphism of Ko over K, and every automorphism ({J of Ksep over K
which induces ({Jo on Ko will be called a Frobenius automorphism of Ksep
over K; then the Frobenius automorphisms of Ksep over K make up the

coset ~o q> in (fj.
DEFINITION 1. A character X of (fj will be called unramified if the cyclic
extension of K attached to X is unramified; we will write X °for the set of
all unramified characters of (fj.
In view of what has been said above, it is clear that X is unramified if

and only ifthe cyclic extension attached to Xis contained in K o, or, what
amounts to the same, if and only if X is trivial on the subgroup ~o of (fj
°
corresponding to Ko; therefore X is a subgroup of the group X K of all
characters of (fj.
PROPOSITION 5. Let q> be a Frobenius automorphism of Ksep over K.

Then X-+X(q» is an isomorphism of the group Xo of the unramified char-
acters of (fj onto the group of all roots of I in C; it is independent of the
choice of q>.
Clearly that mapping is a morphism of X 0 into the group of the
roots of I in C. With the notations explained above, the cyclic extension
of K attached to an unramified character Xof order n is Kn. As q> induces on
Kn the Frobenius automorphism of Kn over K, and this generates the
Galois group of Kn over K, X( q» is a primitive n-tb root of I ; therefore the
morphism in our proposition is both injective and surjective. The last
assertion follows from the fact that two Frobenius automorphisms can
differ only by an element of ~o, and every unramified character is trivial
on~o·
THEOREM 1. Let K be a commutative p-field, and 71: a prime element of
K. Let X °be the group of the unramified characters of (fj. Then x-+ {X, 7I:}
is an isomorphism of X ° onto the group H(K) of factor-classes of K; it is
illdependent of the choice of 71:.
We can identify H(K) with the Brauer group B(K) of K. Every element
of B(K), i. e. every class of simple algebras over K, contains one and only
one division algebra over K. As has already been pointed out in Chap.
IX-4, and again in Chap. X-2, prop. 5 of Chap. 1-4 shows that such an
algebra, if it is of dimension 112 over K, can be written as [K,.JK;X,7I:1J,
where X is a character attached to Kn and 71:1 is a suitable prime element of
K; therefore the factor-class belonging to that algebra is {X, 71: 1}' Combin-
ing prop.IO, Chap. IX~4, with prop. 3 of Chap. VIII-I, we see that this
is independent of 71: 1, so that it is the same as {X,7I:}. Consequently,
°
x-+ {X, 7I:} is a surjective morphism of X onto H (K). As Kn is unramified of
degree n over K, its modular degree over K is n; therefore 71: cannot be in
N KnIK(K:) unless n = I; again by prop. 10, Chap. IX-4, this shows that,
if X is attached to K n, {X, n} =F 1 unless n = 1, i. e. unless X= 1. This com-

pletes our proof.
COROLLARY 1. Let K and n be as in theorem 1; let Kn be the unramified
extension of K of degree n, and let X be a character attached to Kn. Then
[K,,/ K; X, n] is a division algebra over K.
At any rate, it is of the form Mm(D), D being a division algebra over K;

if d2 is the dimension of Dover K, D can be written as [KiK;x',n],
where X' is a character attached to K d • By theorem 1, this implies X' = X,
hence n=d and m= 1.
COROLLARY 2. Let <p be a Frobenius automorphism of Ksep over K.

There is one and only one isomorphism 11 of H(K) onto the group of all roots
of 1 in C, such that ~({x,n})=x(<p) for all xEXo; it is independent of
the choice of n and of <po
This follows at once from theorem 1, combined with prop. 5.
COROLLARY 3. Notations being as above, let X K be the group of all
characters of (f); for all XEXK and all (}EK put X
,
(X, (})K =11( {X, (}}).

Then this defines a pairing between X K and K which satisfies conditions
X
[I] and [III (b)] of § 1.

By prop. 8 of Chap. IX-4, it satisfies [I(i)). By prop. 10 of Chap. IX-4
and prop. 5 of Chap. VIII-I, it satisfies [I(ii')). As to [III(b)], we have
here to take GK = K x , Gi = R x , and we can take for N the subgroup of
K x generated by n. Then [III (b)] is satisfied by taking for Xany character
attached to the unramified extension Kn of K of degree n, as follows at
once from prop. 10 of Chap. IX-4 and prop. 3 of Chap. VIII-I.
COROLLARY 4. For all XEXo, and all (}EKX, we have (X, (})K = X(<p)Ord(6);
if n is any prime element of K, (X, n)K = x(<p)·
The latter assertion is a restatement of corollary 2. Then the former
holds for (} = n, and also, as proved in the proof of corollary 3, for (}E R x;
the general case follows from this at once.
COROLLARY 5. Let Kl be a field isomorphic to K; let Kl be an al-

gebraic closure of K 1, and A. an isomorphism of K onto K 1 , mapping K
onto K 1 • For every character X of (f), write t for its transform by A., i.e.
for the character of the Galois group (f)1 of (K 1)sep over Kl given by
t(CT1)=X(A.CTlr1) for all CT 1 E(f)1' Then (X'(})K=(t'(}).)K for all XEX K 1
and all (} E K x •
This follows at once from corollary 2, since obviously A maps a

prime element of K onto a prime element of K 1, and transforms a
Frobenius automorphism of Ksep over K into one of (Kl)sep over K l .
From now on, the pairing of X K with K x, defined in corollary 3,
will be called the canonical pairing for K. As explained in § 1, we derive
from this a morphism a of K x into the Galois group 21 of Kab over K
which will be called the canonical morphism for K; it is defined by
(X,O)K= x(a(O»), this being valid for all XEX K and all OEK x. Corollary 4
of tho 1 shows that a(n) induces on Ko the Frobenius automorphism of
Ko over K whenever n is a prime element of K.
As we identify the Brauer group B(K) with the group H(K) considered
in theorem 1 and its corollaries, we may consider the mapping 11 defined
in corollary 2 of tho 1 as an isomorphism of B(K) onto the group of roots
of1 in C; for every simple algebra A over K, we will write h(A) = 11 (CI(A»),
and will call this the Hasse invariant of A; it is 1 if and only if A is trivial.
THEOREM 2. Let K' be an extension of K of finite degree, contained

in K; let (D, (D' be the Galois groups of Ksep over K, and of K~ep over K',
respectively, and let P be the restriction morphism of (D' into (D. Then.
for every XE X K' and every 0' E K' x , we have:
(2)
Let f be the modular degree of K' over K; then the module of K' is
qf, and, if qJ, qJ' are Frobenius automorphisms of Ksep over K, and of
K~ep over K', respectively, qJ' coincides with qJl on the group IDl of the
roots of 1 of order prime to p in K, hence on Ko = K(IDl), so that p(qJ')qJ-1
is in the subgroup ~o of (D which corresponds to Ko. Now assume first
that the character X in (2) is unramified, hence trivial on ~o; this implies
that X(p(qJ'») = X(qJ)/. As we have observed in Chap.IX-4, the cyclic
extension of K' attached to XO P is the compositum of K' and of the
cyclic extension of K attached to X; as the latter is unramified, hence
generated by elements of IDl, the same is true of the former, so that
xop is unramified. We can now apply corollary 4 of th.1 to both sides
of (2); it shows that the left-hand side is X(p(qJ'»' with r=orddO'), and
that the right-hand side is X(qJ)S with s=ordK(NK'IK(O'»), hence s=fr by
formula (2) of Chap. VIII-l. This proves (2) when X is unramified. In
the general case, call n the order of X; as neither side of (2) is changed if
we replace 0' by 0'11'" with 11'EK'x, we may assume that r=ordK,(O')fO.
As we have just shown, if Xl is any unramified character of (D, (Xl 0P,O')K'
is equal to Xl(qJ)fr; in view of prop. 5, we can choose Xl so that this is
equal to any given root of 1 in C, and in particular to the left-hand side
of (2); as (2) has already been proved for unramified characters, it will
therefore be enough, after replacing X by XX 11, to prove our result under

the additional assumption that the left-hand side has the value 1. That
being now assumed, call L the cyclic extension of K attached to X; the
cyclic extension of K' attached to XO p is then the compositum L: of K'
and L. As the left-hand side of (2) is 1, prop. 10 of Chap. IX-4 shows
that there is r(eL: such that ()' = N L'/drt'). This gives, by Chap. III-3:
NK'IK((),)=N L'IK(rt')=N LIK(N L'IL(rt')); the same proposition shows then
that the right-hand side of (2) is 1, which completes the proof.
COROLLARY 1. If a, a' are the canonical morphisms for K and for
K', respectively, we have poa'=aoNK'IK.
In view of our definitions, this is just another way of writing (2).
COROLLARY 2. Let K and K' be as in theorem 2; call n the degree of

K' over K. Then, for every simple algebra A over K, we have h(AK,)=h(A)n.
By th.l, the factor-class belonging to A can be written as {X,x}. By

formula (7) of Chap. IX-4, the restriction morphism of H(K) into H(K')
maps the class {X,()} onto the class {xop,()} for every XeXK and every
()eK X; moreover, for ()eKX, we have N K'IK«())=()". By tho 2, this gives:
h(A K,) = (Xo p, X)K' = (X, Xn)K = h(A)".
COROLLARY 3. If X is a non-trivial character of 6>, ()--+(X'())K is a non-

trivial character of K x •
Call nand d the orders of these two characters; clearly d divides n.
Call L the cyclic extension of K attached to X; call Xl an unramified
character of 6> of order n, Kn the unramified extension of K of degree n,
and put D=[KJK;Xl'XJ. By corollary 2 ofth.l, we have h(D)=Xl(q»,
so that h(D) is a primitive n-th root of 1. By corollary 2, we have then
h(DL) = h(D)n = 1, so that DL is trivial; this is the same as to say that D
has an L-representation into Mn(L); by prop. 9 of Chap. IX-4, the factor-
class attached to D can then be written in the form {X,()}, with some
()eKX, and we have h(D)=(X'())K. Therefore d=n. This shows that our
canonical pairing satisfies condition [II] of § 1.
COROLLARY 4. If L is any cyclic extension of K of degree n, NL1K(L X)
is an open subgroup of K x of index n.
In fact, by prop. 10 of Chap. IX-4, it is the kernel of 9--+(X,9)K' where

X is a character of 6> attached to L, and we have just proved that this is
of order n.
If K' = K, or if X= 1, the conclusion (2) of theorem 2 is trivial; if K'
is the cyclic extension of K attached to X, (2) is equivalent to prop. 10
of Chap. IX-4, since XOp= 1 in that case. No other case than these can
occur if K is an R-field, as one sees at once. Therefore theorem 2 remains
valid for K=R or C; so do its corollaries.
PROPOSITION 6. A character Xof (fj is unramified if and only if (x,()K = 1
for all OeR x.
Call Xo the group of the characters with the latter property; as
before, we call X 0 the group of the unramified characters of (fj. By
corollary 4 ofth. 1, X 0 c Xo.
By prop. 2 of§ 1, X-(x, 1t)K is an isomorphism
of Xo
onto the group of all roots of 1 in C; by tho 1, combined with
prop. 5, this induces on X 0 an isomorphism of X 0 onto the same group.
Therefore Xo = X O.
COROLLARY. A cyclic extension L of K is unramified if and only if
N LIK(L X) contains R x •
In view of prop. 10, Chap.IX-4, this follows at once from the appli-
cation of proposition 6 to a character of (fj attached to L.
§ 3. The canonical morphism. We have now verified conditions

[I], [II], [III (b)] of §1 for the canonical pairing (x,O)K' and we have also
shown that the subgroup X 0 of X K defined by means of such a pairing
in § 1 is here the same as the group X 0 of the unramified characters of
(fj. As in § 1, we will now call UK the kernel of the canonical morphism
a of K x into m:; our main result in this § will be that UK = {I}. In applying
the results of § 1, we have to keep in mind that here Gi must be replaced
by R x , n 1 by a prime element 1t of K, and N by the subgroup of K x
generated by 1t. Corollary 2 of prop. 2, § 1, shows that UK is contained in
R x , and that a determines a morphism of R x onto the intersection m:o
of the kernels of the characters XeXo, when these are considered as
characters of m:. Here, by prop. 6 of § 2, X 0 consists of the characters of
(fj which are trivial on the subgroup ~o of (fj corresponding to the union
Ko of all unramified extensions of K. Therefore m:o is the image of ~o
in m:, i.e. the subgroup of m: corresponding to the subfield Ko of K ab ,
or in other words the Galois group of Kab over Ko.
PROPOSITION 7. Let Ko be the union of all the unramified extensions
of K, contained in Ksep; let CPo be the Frobenius automorphism of Ko
over K, and let a be the canonical morphism of K x into the Galois group
m: of Kab over K. Then, for every OeK\ a(O) induces on Ko the auto-
morphism CPo with r = ord(O).
In fact, corollary 4 of tho 1, § 2, can be expressed by saying that
x(a(O»=x(cp)' for every XeXo, ifcp is an automorphism of Ksep over K
which induces CPo on Ko. This is the same as to say that, if cp induces cp'
§ 3. The canonical morphism 227
on K ab , a( 0) q)' - r is in the intersection of the kernels of all the characters

xeX o, or again, in view of the definition of ~o and Ko, that a(O)q>,-r
induces the identity on K o, as was to be proved.
COROLLARY. Notations being as in proposition 7, call 11" an auto-

morphism of Kab over K, inducing 11'0 on Ko. Then a maps R x onto ~o;
it maps K X onto the union of the cosets ~oq>/ft for neZ, and this union
is dense in ~.
This follows at once from proposition 7 and from [II"] in § 1.
Now we consider the kernel UK of a. By definition, it is the intersection
of the kernels of the characters O-+(X,O)K of K X, when one takes for X
all the characters of 6>. By prop. 10 of Chap. IX-4, this is the same as to
say that it is the intersection of the groups N LlK( C) when one takes for
L all the cyclic extensions of K.
PROPOSITION 8. Let K' be an abelian extension of K of finite degree.

Then UK=NK'/K(Ud.
Assume first that K' is cyclic over K. Then we can apply prop. 4
of § 1, by taking F=NK'/K; in fact, [IV(i)] and [IV(ii)] are obviously
satisfied by the automorphisms x-+X" of K'x, for all Ae6>; so is [IV(iii)],
by corollary 5 of tho 1, § 2; [V(i)] is obviously satisfied, and so is [V(ii)],
by tho 2 of § 2. In the conclusion of prop. 4, U and U' are here the same,
respectively, as UK and UK'; moreover, as we have seen, UK is contained
in NK'/K(K'X), which, in the notation of prop. 4, is the same as F(G').
This proves our assertion when K' is cyclic over K. Otherwise we can
find a sequence K, K 1 , ... ,Km =K' of fields between K and K', such that
each one is cyclic over the preceding one. If we use induction on m, the
induction assumption gives UK. =NK'/K.(Ud, and what we have proved
gives UK = N K./K(U K.); putting these together, we get our conclusion.
The same proof would be valid for any solvable extension, but this will
not be needed.
PROPOSITION 9. Assume that K contains n distinct n-th roots of 1.

Then the intersection of the kernels of the characters O-+(Xn,~,O)K of K X,
for all eeK x, is (Kx)".
Here the assumption on K implies that n is not a multiple of the

characteristic of K, and Xn,~ is as defined in Chap. IX-5. By definition,
the set in question is the intersection of the kernels of all the morphisms
0-+ {e, O} n of K x into H(K). By formula (12) of Chap. IX-5 (the "reciprocity
law"), this consists of the elements 0 of K X such that {O,e}n= 1, i.e.
{Xn,8,e} = 1, i.e. (Xn.8,e)K=1, for all eeKx. By corollary 3 ofth.2, §2,
this is equivalent to X",9= 1; as we have observed in Chap.IX-5, this

is so if and only if Oe(KX )".
COROLLARY. Let K be any p-field; if n is not a mUltiple of the char-

acteristic of K, UK C (K X)".
The assumption on n implies that there are n distinct n-th roots of
1 in Ksep; then they generate an abelian extension K' of K. By proposition
9, we have UK' c (K' X )". By prop. 8, this gives
PROPOSITION 10. Assume that K is of characteristic p. Then the

intersection of the kernels of the characters O-+(Xp,~,O)K of K X , for all
eeK, is (KX)p.
Call Z that intersection; as all the characters Xp,~ are of order p or 1,

Z is a subgroup of K X , containing (K)()P; as Xp,~= 1 for e=O, Z may be
defined as consisting of the elements 0 of K)( such that {e, O} P = 1 for
all eeK x , or, what amounts to the same, such that go,O}p= 1 for all
eeK)(. By formulas (13) and (14) of Chap. IX-5, we have, for all eeK x ,
OeK x :
1 = {eO, -eO}p= {eO, -e}p' {eO,O}p,
so that Z is also the set of the elements 0 of K)( such that {e 0, - e} p = 1

for all ~EKx. Then, by the first formula (13) of Chap. IX-5, Zu{O}
is an additive subgroup of K. As Z is a subgroup of K)(, containing
(K)()P, we see now that Z u {O} is a subfield of K, containing KP; therefore,
by corollary 1 of prop. 4, Chap. 1-4, it is either K or KP. If it was K, all
the characters of the form Xp,~ would be trivial. As we have observed in
e
Chap. IX-5, the kernel of the morphism -+ Xp,~ is the image of K under
the mapping x-+x-x P ; in view of tho 8 of Chap. 1-4, one sees at once
that this image cannot contain 1l- 1, if 1l is any prime element of K;
therefore Xp,~ is not trivial for e=1l- 1 • This proves that Zu{O}=KP,
hence Z = (K)()P.
COROLLARY. If K is of characteristic p, UK C (UK)P,
By proposition 10, UK c (K)( )P, so that, if 0 e UK' it can be written

as tIP with "e K)(. Take any cyclic extension L of K; by prop. 8,
UK=NL/K(UL), and, by prop. 10, ULc(L")P; therefore we can write 0
as N L/K«(P) with (e L)( . This gives tIP = N L/K«()P; as P is the characteristic,
this implies ,,=NL / K «(). We have thus shown that" is in the intersection
Jf the groups N L/K(L )() for all cyclic extensions L of K; as this intersection
is UK' this proves our corollary.
§3. The canonical morphism 229
THEOREM 3. The mapping X-+xoo is a bijective morphism of the group

X Kof characters of m: onto the group of the characters of finite order
of KX.
Take any integer n ~ 1. If K is not of characteristic p, we have
UK c (K x t, by the corollary of prop. 9. If K is of characteristic p, write
n = n' pi with n' prime to p, and i ~ 0, and take any (J E UK; by the same
corollary, we can write (J = e"' e
with E K x. By the corollary of prop.
10, and using induction on i, we see at once that UKC(U K)P\ so that we
can write (J = 11P; with 11 E UK' Take integers a, b such that n' a + pi b = 1;
then (J=W11 a)". This shows that, in all cases, UK c(K7, so that every
character of K x , of order dividing n, is trivial on UK' As this is so for all n,
our conclusion follows now at once from corollary 4 of prop. 2, § 1.
COROLLARY 1. The canonical morphism 0 of K x into the Galois group
m: of Kab over K is injective.
By lemma 2 of § 1, applied to the endomorphism x -+ x" of K x ,
(K X)" is a closed subgroup of K x for every n ~ 1; this implies that it is
the intersection of the kernels of all the characters of K x whose order
divides n; that being so, theorem 3 shows that the kernel UK of 0 is the
t
same as the intersection U' of the groups (K x for all n ~ 1. Clearly U'
is contained in R x • As it is obvious that the compact group R x is totally
disconnected, lemma 4 of Chap. VII-3 shows that all its characters are
of finite order. If n is a prime element of K, every character of R x can be
uniquely extended to a character w of K x such that w(n) = 1, which then
must also be of finite order. This implies that U' is contained in the kernel
of all the characters of R x , so that it is {I}.
COROLLARY 2. The canonical morphism 0 induces on R x an isomorphism

of R x onto the Galois group m:o of Kab over the union Ko of all unramified
extensions of K in K.
This is now obvious, by corollary 1 and the corollary of prop. 7.

THEOREM 4. Let K' be an extension of K of finite degree, contained in
K; put L=K' nKab. Then, for (JEK x , 0«(J) induces the identity on L if and
only if (J is in N K'/K(K' X).
Call p the restriction morphism of m:' into m:, and put ~ = p(m:').
An element of K ab is invariant under ~ if and only if it is in K'; then it is
in L; therefore ~ is the subgroup ofm: corresponding to L. Put X = 0 -1 (~)
and X'=NK'/K(K'X); what we have to prove is that X=X'. By lemma 2
of§ 1, X' is closed in KX. Ifnis the degree of K' over K, we have NK'/K«(J) =
= (J" for (JE K x , so that X' => (K X)"; therefore, if t/I is a character of K x ,
trivial on X', it is trivial on (KX)", hence of a finite order dividing n, so
230 Local c1assfield theory XII
that, by tho 3, it can be written as xoa with XEXK. Then xoao NK'/K is
trivial on K' x ; by corollary 1 of tho 2, § 2, it is the same as XO po a', so that
Xop must be trivial on ~', hence X on p(~')=~, hence'" on X. This
shows that X':::::JX. Conversely, if O=NK'/K(O') with O'EK'X, corollary 1
ofth. 2, § 2, gives a(O) = p(a'(O'»); as this is in~, we see that X' eX, which
COROLLARY 1. Assumptions and notations being as in theorem 4, call
~ the subgroup of ~ corresponding to L. Then NL/K(r)=NK'/K(K'X)=
=a-l(~).
The latter equality is just a restatement of theorem 4. Applying

theorem 4 to K' = L, we get N L/K(r) = a - 1 (~).
COROLLARY 2. For every extension L of K of finite degree, contained in
K ab , call the subgroup of ~ corresponding to L, and put N(L) =
~(L)
NL/K(r). Then N(L)=a-l(~(L»); ~(L) is the closure of a(N(L») in ~;
L consists of the elements of K ab , invariant under a(O) for allOEN(L), and
a determines an isomorphism of K x /N(L) onto the Galois group ~/~(L) of
Lover K. Moreover, L-+N(L) maps the subfields of K ab , of finite degree
over K, bijectively onto the open subgroups of K x , of finite index in K x •
All this is a restatement of prop. 3 of § 1, once theorems 3 and 4 are
taken into account. Traditionally, when Land N(L) are as in our corollary,
one says that L is "the classfield" to the subgroup N(L) of K x • In applying
our corollary, it is frequently useful to keep in mind that, by lemma 1 of
§ 1, an open subgroup of K X is of finite index in K X if and only ifit is not
contained in R x •
COROLLARY 3. Let K and K' be as in theorem 4; let M be a subfield of
K ab , of finite degree over K, and call M' its compositum with K'. Then
N M'/K,(M'X)=N K'/K l(NM/K(M X»).
By corollary 2, N M/K(M X), which is the same as N(M), consists of the
elements 0 of K X such that a(O) leaves every element of M invariant.
Similarly, NM'/K,(M'X) consists ofthe elements 0' of K'x such that a'(O')
leaves every element of M' invariant; the latter condition is fulfilled if and
only if p(a'(O'» leaves every element of M invariant; in view of corollary
1 ofth. 2, § 2, this is the same as to say that a(NK'/K(O'») leaves every element
of M invariant, i.e. that N K'/K(O') is in N M/K(M X ).
It is easily seen that theorem 4 and its corollaries retain their validity
for R-fields; so does theorem 3.
§ 4. Ramification of abelian extensions. The above theory would be

incomplete without the knowledge of the ramification properties of the
abelian extensions of K, and in particular of their differents and discrimi-
§ 4. Ramification of abelian extensions 231
nants. As shown in Chap. VIII-3, these properties can be fully expressed

by a description of the Herbrand distribution on the Galois group ~
of Kab over K. We begin with some preliminary results, the first one of
which has no reference to abelian extensions and may be regarded as
supplementing Chap. VIII-3. We adopt the same notations as there,
e. g. in prop. 9 of that Chapter, calling K' a Galois extension of K of
degree n with the Galois group 9 = go, and calling 9", for v ~ 1, the higher
ramification groups of K' over K. We also call R, R' the maximal com-
pact subrings of K, K', and P, P' the maximal ideals of R, R', respectively.
We denote by e the neutral element of g.
PROPOSITION 11. Let e be the order of ramification of K' over K, and
let p,d be its different. Take h~ 1, ZEP'h and put:
NK'IK(X -z)=X"+a1X,,-1+ ... + all'
where X is an indeterminate.
(i) If v(A)~h + 1 for all Ai=e, then, for 1 ~i~n:
e'ordK(aj)~h+d-e +1.
(ii) If v(A)~h for all Ai=e, then, for 2~i~n:
e'ordK(aj) > h+d-e + 1.
(iii) If V(A) ~ h + 1 for all Ai= e, then, for 1 ~ i ~ n:
ordK(aj)~h.
(iv) If v(A)~h+ 2 for all Ai=e, then, for 1 ~i~n-l:

ordK(aj)> h.
As - a 1 = 1TK'IK(Z), the inequality in (i), for i = 1, is nothing else than
corollary 1 of prop. 4, Chap. VIII-I, and does not depend upon the as-
sumption about V(A) in (i). In any case, we have
(3)
where the sum is taken over all combinations of i distinct elements of
g, or, what amounts to the same, over all subsets 5={Al, ... ,Aj} of 9 of
cardinal number i. For each such subset 5, write:
Z(5)=~1~2 ••• ~i.
Take such a subset 5; for each O"Eg, write 50" for the image ofs under the
translation A-+AO" in g; call ~ the subgroup of g, consisting of the ele-
ments 0" such that 50"=5; calli the order of~, and take a full set {Pl, ... ,Pr}
of representatives of the left cosets ~P of~ in g. Clearly 5 is a disjoint union
of right cosets Jl~ of~ in g; take a full set m= {Jll""'JL",} of representatives
for these cosets, so that s is the disjoint union of the cosets III ~, ... , Ilm ~;
we have i = m 1. Put W = z(m), and call K" the subfield of K' corresponding
to the subgroup ~ of g; we have now:
z(s) = flw/J=NK'/K"(w),
/Jell
In view of the definition of~, the sets s P1, ... ,5 P, are all distinct; as they
have the same cardinal number i as s, all the terms z(sp), for 1 ~j~r,
occur in the right-hand side of (3). As the Pj induce on K" all the distinct
isomorphisms of K" into K, the sum of these terms can be written as
(4) LZ(SPj) = Lz(sY'J = LNK'/K-(WY'J = 1rK"/K(NK'/K-(W»).

j j j
Consequently, the right-hand side of (3) can be written as a sum of terms,

each of which has the form shown in the right-hand side of (4); moreover,
for each one ofthese terms, we have ml=i, where 1 is the order of~, i.e.
the degree of K' over K". All we need do now is to prove the inequalities in
our proposition for each term of that form, with orddw)~mh in view of
the assumption on z and ofthe definition ofw. Call e' the order oframifi-
cation and f' the modular degree of K' over K", so that 1= e' f' by corol-
lary 6 ofth. 6, Chap. 1-4; then, by (2) of Chap. VIII-I, the order of N K'/K"(W)
in K" is ~ f'mh. Call e" the order of ramification and d" the difTerental
exponent of K" over K. If Q) is the order in K of the right-hand side of (4),
we have, by corollary 1 of prop. 4, Chap. VIII -1 :
e" Q)~ f'mh+d" -e" + 1.
As e = e' e", i = m 1, and e'f' = I, this gives

eQ)~ih+e' d" -e+e'.
If now we call d' the difTerental exponent of K' over K", corollary 4 of
prop. 4, Chap. VIII-I, gives d=e'd"+d', so that our last inequality can
be written as
eQ)~ih+(d-e+ 1)-(d' -e' + 1).
Now formula (9) of Chap. VIII-3, applied to K' and K", gives:
d' -e' + 1 = L(v(A.)-I)+,

l
where the sum is taken over all A. =1= B in ~; moreover, as pointed out there,
the number of terms > 0 in that sum is ~ e' -1. If the assumption in (i) is
satisfied, everyone ofthese terms is ~h; this gives
d' -e' + 1 ~(e' -1)h,
§4. Ramification of abelian extensions 233
and therefore
ew-(h+d-e + 1)~(i-e')h;
as e' ~ 1~ i, this proves part (i) of our proposition. If the assumption in (ii)
is satisfied, we get in the same way:
d'-e' + 1 ~(e'-1) (h-1)
and therefore
ew-(h+d-e+ 1)~i-1 + (i-e')(h-1);
this cannot be 0 unless i = 1, which proves (ii). On the other hand, if we
apply formula (9) of Chap. VIII-3 to K' and K, we get
(d-e +1)-(d'-e' + 1)= ~)v(A)-1)+

where the sum is now taken over all Aeg-~ and consists of n-I terms,
n
so that it is ~ (n - h if the assumption in (iii) is satisfied. Then we get:
e(w - h) ~ (i + n -1- e)h,
which proves (iii), since 1~ i and e ~ n. Similarly, the assumption in (iv)
gives:
e(w-h)~(i +n-l-e)h+n-l;
as I ~ i, e ~ n, 1~ n, the right-hand side cannot be 0 unless 1= i, e = n, 1= n,

hence i = n. This proves (iv).
COROLLARY 1. Notations being as in proposition 11, take again h~ 1,

zep,h. That being so:
(i) If V(A) ~ h +1 for all A=/= e, then:
e'ordK(NK'/K(l + z)-l)~h+d-e +1.
(ii) If v(A)~h for all k=/=e, and h=pe-(d-e+1) with peZ, then:
N K'/K(1+z)=1 (P'), N K'/K(1+z)=1+7rK'/K(Z) (p'+1).
(iii) If V(A) ~ h + 1 for all A=/= e, then:
N K'/K(1+Z)=1 (Ph).
(iv) If v(A)~h+2 for all A=/=e, then:
N K'/K(1 + z)= 1+ N K'/K(Z) (PH 1).
In fact, with the notations of proposition 11, we have:

n
N K'/K(1+z)=1+ L(-1)i ai ;
i=l
the four assertions in our corollary follow now immediately from the
corresponding ones in proposition 11.
COROLLARY 2. Assumptions being as in corollary 1 (ii), we have

N K'/K(l + P'") = 1 + J'P.
As d - e +1 ~ 0, these assumptions imply h ~ p e, hence p ~ 1. By
corollary 1 (ii), N K'/K(1 + P''') is contained in 1 + J'P. Conversely, take any
xoeJ'P; then we can define by induction two sequences (XO,x h ".) and
(ZO,Zl"")' with XiePi+p and Ziep,ie+1l for all i~O, by taking, for each
i~O, Ziep,ie+1l such that 'lrK'/K(Zi)=X i, as may be done by prop. 4,
Chap. VIII-1, and then putting
x i+ 1 =(1 + xi)NK'/K(1 + zr 1_1;
corollary 1 (ii) shows at once that this is in pi+ 1 +P, as it should. Then,
obviously, 1+xo=NK'/K(y) with y given by the convergent product
y= n+
00
i=O
(1 Zi); as y is in 1 + p'", this proves our corollary.
PROPOSITION 12. Let K and K' be as above, and assume that V(A) has
the same value i~2 for all A=Fe in g. Then, for 1~h~i, N K'/K(1+ P''') is
contained in 1+ ph; and N K'/K( 1+ p,i -1) is contained in 1+ pi if and only if
the degree n of K' over K is equal to the module q of K.
Here the higher groups of ramification of K' over K are given by
g,,=g for v~i and g,,= {e} for v~i + 1. As gl =g, we have e=n; K' has the
modular degree f = lover K and has the same module q as K. By formula
(9) of Chap. VIII-3, we have d=(n-1)i. Taking h=i in corollary l(i) of
prop. 11, we get our first assertion for that case; in the case h < i, it follows
at once from corollary 1 (iii) of the same proposition. By corollary 3 of
prop. 9, Chap. VIII-3, the degree n of K' over K divides q; by that propo-
sition, if we take a prime element x' of K' and put YA =X'),X,-l for all
Aeg, the mapping A-+ y), maps 9 onto a set Y of elements of 1 + p,i-1
which are all incongruent to each other modulo p,i. In particular, Y
makes up a full set of representatives of the cosets of 1 + p,i in 1 + p,i-1
ifand only ifn=q. Since obviously N K'/K(Y),) = 1 for all A, this shows that,
if n=q, N K'/K(1 + p,i-1) is the same as N K'/K(1 + p,i), hence contained in
1 + pi. In order to prove the converse, take ze K'x such that orddz) = i-1.
As in prop. 11, write
n
NK'/K(X -z)=xn+ L ajXn- j.

j= 1
Then an=NK'/K(-Z), so that, by (2) of Chap. VIII-1, ordK(a n)=i-1.

Taking h = i -1 in prop. 11 (i), we get ordK(aj ) ~ i -1 for 1 ~j ~ n, so that,
if we put bj=aian' all the bj are in R. Now take any ye1+p,i-l; as
(1- y)/z is in R', and as K' has the same module as K, there is aeR such
that (1- y)/z=a (P'), or, what amounts to the same, y=(1-az)u with
ue 1 + p,i. Then N K'IK(U) is in 1 + pi, so that we have:
" "-1
N K'IK(y)=NK'IK(1-az)=1+ Lajaj =1+a"(a"+ Lbja j) (pi).
j= 1 j= 1
For 1 ~j~n-l, call hj the image ofbjin the field RIP, under the canonical
homomorphism of R onto that field; then the above formula shows that
N K' IK(Y) is in 1 + pi if and only if the image of a in the same field is a root
ofthe polynomial T"+ Lhj Tj; in particular, if this is so for all y, all the
elements of RIP must be roots of that polynomial, so that n~q. This
PROPOSITION 13. Let 'It be a prime element of K; for each v~ 1, call N v

the subgroup of K x generated by 'It and 1 + PV, and. call Kv the subfield of
Kab such that N (Kv) = N v in the sense of corollary 2 of theorem 4, § 3. Call
g(v) the Galois group of Kv over K, and a" the morphism of r onto g(v), with
the kernel N v, determined by the canonical morphism Q of K. Call glv), for
i ~ 1, the higher groups of ramification of Kv over K. Then g~) = g(v); for
1 ~p~v and if- 1<i~if, glv) = a,,(Np).
Choose some v ~ 1, and then, to simplify notations, write N instead of

N v, L instead of Kv, 9 instead of g(v), gi instead of glv); as in Chap. VIII-3,
call gi the number of elements of gi' for all i ~ 1; then gi divides 9 j for
i~j. As 9 is isomorphic to K X IN, the degree of Lover K is the index of N
in K X , whi<fh is n=(q _1)qv-1. By the corollary of prop. 6,§ 2, the maximal
unramified extension [, of K, contained in L, is the one for which N([' X) =
=R x N; as R x N =K x , we get [,=K; in other words, L is fully ramified
over K, its order oframification is e=n, and we have 91 =9; moreover, L
has the same module q as K, and the same must then be true of all fields
between K and L, so that, if K c K' c K" c L, K" is fully ramified over K'.
By corollary 1 of prop. 9, Chap. VIII-3, gilgi+ 1 divides q for i~2; therefore
gdg2 =q-l. Put r 1=0 and ri =(g2+'" + gi)/n for all i~2; for each in-
teger p~O, call i(P) the largest ofthe integers i such that ri~p. Assume
ri<p<ri+ 1 for any p~O and any i~ 1; then 0<pn-(g2 + ... +gi)<gi+1;
this is clearly a contradiction, since n, g2"'" gi are multiples of
g/+1' Therefore, for all p, we have ri(p)=p. We have i(O) = 1, and i(p) > 1
for p>O. Take now any p such that O~p<v; put i=i(P); call K', K" the
subfields of L consisting ofthe elements invariant under gi and under gi+ 1>
respectively; the Galois group g" of K" over K' may then be identified
with gJgi+ l ' If p=O, i= 1, and the degree of K" over K' is gdg2 =q-l.
From now on, assume that 1 ~ P < v; we will show that then K" is of de-
gree q over K'. We first observe that the higher ramification groups gj
of K" over K' are given by applying formula (11) of Chap. VIII-3 to K',
K" and L; as K" is fully ramified over K', its order of ramification is the
same as its degree over K', and then that formula shows at once that
v(O''') = i for all 0''' e g" except the identity, so that gJ = g" for j ~ i, and
gJ={e} for j~i +1. Similarly, writ.e g'=g/gi for the Galois group of K'
over K, and call gj its higher ramification groups; in exactly the same
manner, we find that gj is the image gig, of gj in g', for j <i, and that it
is {e} for j~i. Call R', R" the maximal compact subrings of K', K", and
P', P" the maximal ideals in R', R", respectively. As K' is fully ramified
over K, its order of ramification e' is the same as its degree n'=n/g i;
then, if p,d' is the different of K' over K, formula (10) of Chap. VIII-3, gives
i
d' -e' + 1 = L (gigi)-I)=r in/gi -i + 1=pn'-i + 1.
j=2
Take any zeL such that orddz)~i-l; put v=NL/K"(Z), so that vep"i-l.
Applying corollary 1 (iv) of prop. 11 to K" and L, with h = i-I, we get
N L/r(1 + z) == 1+ v (P"i).
Define now weK' by writing
1+ w=NL/d1 + z)=NK"/dN L/r(1 + z»).
Applying to K', K" the first assertion in prop. 12 with h = i, we get

l+w==NK"/K,(I+v) (p'i);
the case h=i-l of the same assertion in prop. 12 gives then wep,i-l.
Now, taking h = i-I in corollary 1 (ii) of prop. 11, we can apply it to K
and K' ; this gives
N L/K(1 + z)=NK'/K(1 + w)== 1 + 'ltK'/K(W) (pp+ 1).
By the definition of K v, and corollary 2 ofth. 4, § 3, this must be in Nv=N,

hence in NvnRx, i.e. in I+P; as p<v, this implies that 'ltK'/K(W) is in
pP+ 1. In view of the values found above for e' and d' - e' + 1, prop. 4 of
Chap. VIII-l shows that 'ltK'/K maps p,i-l surjectively onto pP, and p,i
onto pP+ 1; in particular, there is w' ep,i-l such that 'ltK'/K(W') is not in
pp+l. Then, if we had orddw)=i-l, w'w- 1 would be in R', so that we
could write w'w- 1 ==cx (P') with cxeR, since K' has the same module
as K; this can be written w' == cx w (p'i), which implies
'ltK'/K(W')==CX'ltK'/K(W)==O (pp+l),
against our assumption. This shows that w is in p'i, or in other words
that N K"/K,(I+v) is in l+P,i whenever v=NL/K,,(Z) with zeL, ordL(z)~
~ i -1. Now choose ZoE LX so that ordL(zo) = i-I; put Vo = N LIK"(zO)' so

that ordK.(vo) = i -1; call M x the group of (q -l)-th roots of 1 in K, and
take Z=J.lZo with J.lEMx. As the degree of L over Kif is gi+1, and as
gigj+1 divides q for allj~2, we have v=J.lQvo, where Q=gi+1 divides a
power of q and is therefore prime to q-I, so that J.l-+J.lQ is an automor-
phism of MX. Consequently, when J.l runs through the set M =M x u{O},
1 + v runs through a full set of representatives of the cosets modulo
I+P"i in I+p"i-1. As NK"ldI+v) is in I+P'i for all these elements v,
prop. 12 shows, firstly, that N K"ldl + p lfi - 1) is contained in 1+ p'i, and
then that the degree of Kif over K' must therefore be q.
In other words, we have shown that gilp!gilp) + 1 is equal to q for 1:::::; p:::::;
:::::; v-I; we had already found that it has the value q - 1 for p = O. As we
have
ao
n=(q-I)qY-1= n(gJgi+1)'
i=l
this implies that gi=gi+1 whenever i is not one of the integers i(O) = 1,
i(l), ... ,i(v -1); therefore gj= gilp) + 1 for i(p) <j:::::; i(p + 1), so that gilp+ 1) is
of index q in gilp) for 1:::::; p < v, while this index is q -1 for p = O. By induc-
tion on p, we see at once that gilp)=qY-P for l:::::;p :::::;v. The definition of the
integers r i gives now:
rilp+1)-rilp)=(gilp)+1 + ... + gilp+1»)/n=(i(p+1)-i(p»)qY-P-1 n -1,
for 0:::::; p < v. As r ilp ) = p, the left-hand side is 1; this gives
i(p + 1)- i(p)=(q-I)t(,
and therefore i(p)=qP by induction on p.

To complete the proof, observe that, in view of the values found above
for e' and d' - e' + 1, we may apply corollary 2 of prop. 11, with h = i-I, to
K and the same field K' as above; it shows that 1 + pP is the same as
N K'IK(I+P'i-1) and is therefore contained in the group N'=NK'IK(K'X)
associated with K' according to corollary 2 ofth. 4, § 3. As N' contains the
group N =Ny associated with L=Ky , it contains 1t, hence the group Np
generated by 1t and I+PP. Let a" be the morphism of K X onto g=g(v),
with the kernel Ny, defined in our proposition; by corollary 2 of th.4,
§ 3, this maps N' onto the subgroup gilp) of 9 corresponding to K';
therefore gilp) contains a,,(Np) for l:::::;p <v. In view of our definitions, the
same is obviously true for p = 0 if we define No by No = K x , and also for
p=v. Now we prove by induction on p that gilp) = a,,(Np) for O:::::;p:::::;v.1t
is true for p=O. Assume gilp-1)=a,,(Np- 1), and let N' be as above; N'
contains N p , as we have seen, and it is contained in N p - 1 , since gilp) is
contained in gilp-1); its index-irr Np- 1 is the same as that of gilp) in gilp-1),
238 Local c1assfield theory XII
which is q -1 if p = 1 and q if p > 1. As this is the same as the index of Np in

N p - 1 , we get N' =Np • In view of what has already been proved above,
this completes the proof of our proposition.
COROLLARY. Notations being as in proposition 13, the order of ramifi-

cation of K. over K is the same as its degree and is given bye. = (q -1 )q'- 1 ;
if d. is the differental exponent of K. over K, we have d./e. = v - (q -1) - 1.
The value of e. has already been given above; that of d. can be ob-
tained at once by applying formula (10) of Chap. VIII-3 to the results
stated in proposition 13; this gives the formula in our corollary.
As will presently be seen, proposition 13 contains in substance the

determination of the Herbrand distribution on the Galois group m of
Kab over K, which was our main object in this §. We recall that this has
been defined in Chap. VIII-3 as a certain linear form f - t H(f) on the
space of all locally constant functions on m. As explained there, if X is
any open and closed subset of m, its characteristic function fx is locally
constant, and then we write H(X) instead of H(fx); X -tH(X) is thus a
finitely additive function of X.
LEMMA 3. Let H be the Herbrand distribution on m. Then there is a

unique distribution Ho on mo such that H(f)=Ho(fo) whenever f is a
locally constant function on m, and fo is the function induced on mo by f
Let ~ be any open subgroup of m; let L be the subfield of Kab corre-

sponding to ~. Let Ko be as in § 2, i. e. the union of allunramified exten-
sions of K, so that mo is the subgroup ofm corresponding to Ko. Then the
maximal unramified extension Lo of K, contained in L, is KonL and
corresponds to the subgroup ~ mo of m. If ~ (X is any coset of ~ in m,
other than~, and (X induces on L the automorphism A, H(~(X) is by defi-
nition equal to - v(A)/e, where e is the order of ramification of Lover K;
this is 0 if A does not induce the identity on L o, i.e. if~(X is not contained
in ~mo, or in other words if ~(Xnmo =f). As H is finitely additive, this
implies that H(X)=O whenever Xnmo=f), and H(f)=O whenever the
locally constant function f is 0 on mo. On the other hand, take any
locally constant function fo on mo. As mo is compact, fo is uniformly
continuous, so that there is an open subgroup ~ of m such that f 0 is con-
stant on the co sets of~nmo in mo. Then fo can be uniquely continued to
a function f on m, constant on the co sets of~ in m and 0 outside ~mo.
If then we put H o(fo) = H(f), it is clear that H 0 is as required by our
lemma. Except for obvious notational changes, the lemma and its proof
remain valid for the Herbrand distribution on the Galois group of any
Galois extension of K, abelian or not. This will not be needed.
As the canonical morphism a of K x into m: maps R x isomorphically

onto m: o, we can transport to R x the distribution H 0 oflemma 3 by means
of the inverse to that isomorphism. This defines a distribution H R on R x,
which we extend to a distribution HK on K X by prescribing that
H K(X) = HR (X nR X) for every open and closed subset X of K x • We will
call HKthe H erbrand distribution on K x ; in an obvious sense, its support
is contained in R x • In view of the definition of H 0' we have H(f) = HK(foa)
for every locally constant function f on m:, and H(X)=H K(a- 1 (X»)
for every open and closed subset X of m:. The distribution HK is given by
the following theorem:
THEOREM 5. Let HK be the Herbrand distribution on KX. Then its
support is RX; HK(RX)=O; HK(1+P")=v-(q-1)-1 for all v~1; if
O~p<v, eeR x , and ord K(1-e)=p, then
HK(1 +r)e) = -qP+ 1-·(q_1)-I.
By the defmition of the Herbrand distribution, we have H(m:)=O;
this gives H K(K X) = 0, hence HK(R X) = O. Let K., N., d., e. be as in prop. 13
and its corollary; if ~. is the subgroup of m: corresponding to K., we
have H(~.)=d./e. by the definition of the Herbrand distribution; as
N.=a-l(~.) by corollary 2 ofth.4, §3, and as N.nR x =1+P", this,
together with the corollary of prop. 13, gives for HK (1 + P") the value
e
given in our theorem. Finally, let be as in our theorem; call Athe auto-
morphism of K. induced by am. By the definition of the Herbrand
distribution, H(~.am) is -v(A)/e., or, what amounts to the same, it
is - i/e. if i is the largest integer such that AeS\·). By prop. 13, this is
i=qP if p is the largest integer such that Aea.(Np), or, what amounts
to the same, such that eeNp ; this is given by p=ordK(i -e). On the
other hand, H(~.a(e» is the same as HK(N.e) and as HK(1+p")e).
This completes our proof.
COROLLARY 1. Let X be a character of m:, and pi the conductor of

the character xoa of KX. Then f =H(X)=HK(xoa).
Put 00= xoa. Iff = 0,00 is trivial on R x, so that HK(oo) = HK(R X)=O
by theorem 5. Assume now that f ~ 1; call (fJo the characteristic function
of R x, and (fJi that of 1 + pi for 1 ~i~f. We have:
1-1
HK(oo)= L HK((fJi-(fJi+ 1)00)+ HK«(fJ/ OO ).
i=O
By the definition of the conductor, 00 is trivial on 1+ pi, so that the

last term is equal to HK(1+pl), hence to f-(q-l)-1 by theorem 5.
Also by theorem 5, and in view of the fact that 00 is constant on cosets
modulo 1+ pi in R x , we have, for 0 ~ i ~f -1 :
HK((<Pi-<Pi+ 1)W)= _qi+ 1- I(q _1)-1(Si- Si+ 1)'

where Si is the sum Lw(e) taken over a full set of representatives of
the cosets modulo 1 + pI in R x if i=O and in 1+ pi if i~ 1. By the
definition of the conductor, w is not trivial on R x , nor on 1+ pi for any
i<f; therefore Si = 0 for i<f, and SI = 1. Our conclusion follows from
this at once.
COROLLARY 2. Let L be an abelian extension of K of finite degree.
Let W1' ... 'Wn be all the distinct characters of K X, trivial on the subgroup
N(L) = N LIK(L X) of K x associated with L; for each i, call pI, the conductor
of Wi. Then the discriminant of Lover K is p~ with b= 'Lk
i
Call !B the closure of a(N(L)) in ~; by corollary 2 of tho 4, § 3, it is

the subgroup of ~ corresponding to L, and a determines an isomorphism
of K X /N(L) onto ~/!B. Therefore we can write, for each i, Wi= Xioa,
where Xi is a character of ~, trivial on !B, and the Xi' for 1::::;; i ::::;; n, are
then all the characters of ~, trivial on !B, so that the characteristic
function of!B on 2l can be written as n - 1 LXi. Call e, f and d the order
i
of ramification, the modular degree and the differental exponent of L
over K; then n=ef, and, by the corollary of prop. 6, Chap. VIII-2, the
discriminant of Lover K is pfd. By the definition of the Herbrand
distribution, we have H(!B) = d/e; this can be written as
d/e=H(n-1~Xi) =n-1~H(Xi)'
I ,
hence f d = Lfi' in view of corollary 1. This completes the proof.

i
§ 5. The transfer. Notations being as before, let K' be an extension

of K of finite degree; call a, a' the canonical morphisms of K x into ~,
and of K' x into ~', respectively. As a is injective, there is a mapping t
of the image a(KX) of K X in ~ into the image a'(K'X) of K'x in ~',
defined by t(a(O))=a'(O) for every OEKx, or in other words by toa=a'oj
if j is the natural injection of K x into K' x. The question arises whether
this can be characterized in group-theoretical terms, and extended by
continuity to a morphism t of ~ into ~'; this will now be answered
affirmatively. For simplicity, we will assume K' to be separable over K;
a consideration of the general case would complicate our statements
without adding to them anything of value.
Consequently, let K' be a subfield of K sep ' of finite degree n over K.
As before, we call (})' the subgroup of(}) corresponding to K', and identify
~ with (})/(})(1) and ~' with (})'/(})'(l). It will be shown that the morphism
we are looking for is none other than the so-called "transfer homo-
§ 5. The transfer 241
morphism" t of m: into m:'; we recall that this is defined as follows. Take

a full set {O' 1'''''0'n} of representatives of the right cosets 0' (fj' of (fj' in
(fj. For every 0' E(fj, the mapping 0' j(fj' -+ 0' 0' j(fj' is a permutation of these
cosets, so that one may write, for each i, O'O'j(fj'=O'j(j)(fj', where i-+j(i)
is a permutation of {l,,,.,n}; this can be written as O'O'j=O'j(j)'rj with
'rjE(fj'.1t is now easily seen that the image of'r 1 ,,·tn in m:'=(fj'/(fj'(1),
under the canonical morphism of (fj' onto m:', depends neither upon the
choice of the representatives O'j nor upon their ordering, so that, if we
call that image tX(O'), the mapping O'-+tX(O') of (fj into m:' depends only
upon (fj and (fj'. One sees then at once that tX(O'O")=tX(O')a(O") for all
0', 0" in (fj. Moreover, the subgroup (fj" of (fj, consisting of the elements 0'
such that O'O'jEO'j(fj' for all i, is the intersection of the open subgroups
O'j (fj' 0' i 1 for 1 ~ i ~ n, hence itself an open subgroup of (fj; as it is obvious
that O'-+a(O') is continuous on (fj", it is continuous on (fj, hence a morphism
of (fj into m:'. As m:' is commutative, the kernel of this morphism must
contain (fj(1), so that it determines a morphism t of m: = (fj/(fj(1) into m:';
by definition, this is the transfer homomorphism of m: into m:'.
THEOREM 6. Let K' be an extension of K of finite degree, contained
in K sep; let a, a' be the canonical morphisms of K x into m:, and of K' x
into m:', respectively. Let t be the transfer homomorphism of m: into m:',
andj the natural injection of K X into K'x. Then toa='=a'oj.
Let (fj, (fj' be as above; let L be any Galois extension of K of finite *
degree, containing K' and contained in K sep , and let ~ be the subgroup
of (fj corresponding to L. Then ~ is a normal open subgroup of (fj,
contained in (fj'; the Galois group of Lover K is g = (fj/~, and the sub-
group of g corresponding to K' is g' = (fj'/~. Let K" be any field between
K and L; let (fj" and g" = (fj"/~ be the subgroups of (fj and of g, respec-
tively, corresponding to K". The canonical morphism a" for K" is then
a morphism of K"x into m:"= (fj"/(fj"(1), which, to every eEK"x, assigns
an automorphism a"(e) of K~b over K"; we will write b(K";e) for the
automorphism of LnK~b over K" induced on that field by a"(e). As the
subgroup of (fj corresponding to LnK~b is ~(fj"(1), e-+b(K" ;e) is a mor-
phism of K"x into the group (fj"/~(fj"(1); clearly the latter group may be
identified with g"/g"(1), where g,,(1) is the commutator subgroup of g".
In particular, O-+b(K;O) is a morphism of K X into g/g(1). We will denote
by to the transfer homomorphism defined for g and g' just as t has been
defined above for (fj and (fj'; it is a morphism of g/g(1) into g'/g'(l). Our
theorem will be proved if we show that, for all OEK x, b(K' ;0)= to(b(K;O));
for this implies that a'(O) can differ from t(a(O)) only by an element of
the image of ~(fj'(1) in (fj'/(fj'(1), i.e. by an element which is arbitrarily
close to the identity, since we can take for ~ an arbitrarily small open
subgroup of (fj', normal in (fj.
We will denote by h, h' the canonical morphisms of 9 onto g/g(1)

and of g' onto g'/g'(1), respectively; h is the same as the "restriction
morphism" which, to every automorphism of Lover K, assigns its
restriction to LnKab , and h' can be similarly interpreted. If now K" is
a field between K and L, corresponding to the subgroup g" of g, K" n Kab
is the subfield of L corresponding to the subgroup g" g(1) of g, or, what
amounts to the same, it is the subfield of LnKab corresponding to the
subgroup h(g") of the Galois group g/g(1) of LnKab over K. Conse-
quently, by tho 4 of § 3, and in view of our definition of b(K;(}), the sub-
group NK"/K(K"X) of K X consists of the elements (} of K X such that
b(K;O) is in h(g"). Now assume that g" is commutative, so that ~-+b(K";~)
maps K" x into g" ; then we see in a similar manner, by applying corollary 1
of tho 2, § 2, to K and K", that we have, for all ~ EK" x :
(5) h(b(K" ;~))= b(K;NK/I/K(~))'
If K" ::JK', i.e. if g" c: g', we have a similar formula with K', h' replacing K,h.
Now, for a given OEK x , we can choose a cyclic subgroup r of 9
such that b(K ;0) is in h(r); for instance, we may take for r the group
generated by any yEg such that h(y)=b(K;O). Then, as we have seen
above, if Z is the subfield of L corresponding to r, 0 may be written as
NZ/K(O with (EZ x • Take a full set of representatives {A'1, ... ,Ar } for the
double cosets rAg' of rand g' in g, and call Y1 a generator of r. For
each i, r Ai g' is a union ofright co sets y Ai g' of g', with y Er. If y, y' are in r,
YAig' is the same as y' Aig' if and onlyify-1 y' is in the group I;=rnAig' Ai- l .
Call di the index of Ii in r; then Ii is generated by y1', and di is also the
smallest of the integers d such that Ai 1Y1 Ai is in g'. That being so, r Aig'
is the disjoint union of the cosets y{Aig' for O~j<di' Consequently, the
elements y{ Ai' for 1 ~ i ~ r, 0 ~j < di, make up a full set of representatives
of the right co sets of g' in g, and we can use it for computing the transfer
to(Y) of any element y of r. Taking at first y =Y1, we find at once, in that
case:
(6)
this being true for y=Y1, it is obvious that it remains so for y=y{ for allj,
or in other words for all y Er.
For each i, put Zi=ZA" and call Z; the compositum of Zi and K';
obviously (Ai,Z;) is a proper embedding of Z above K' in the sense of
Chap. III-2. Let (A,Z') be any such embedding; after replacing it if
necessary by an equivalent one, we may assume that Z' is contained in
Ksep, hence in L, so that the isomorphism Aof Z onto Z' can be extended
to an automorphism A of Lover K. Then (A,Z') is equivalent to (Ai'Z;)
§ 5. The transfer 243
if and only if there is a K'-linear isomorphism of Z' onto Z;, which we

can then extend to an automorphism (I of Lover K', such that Acoincides
with Ai(l on Z. Then A=YAi(l with YEr and (lEg'. Consequently (A,Z')
is equivalent to one and only one of the embeddings (Ai'Z;), Now prop.4
of Chap. 111-3 gives:
i=l
As we may apply (5) to K' and to K" =Z; for each i, we get:
(7)
i= 1
Put Y= b(Z; 0; by the definition of b, this is in r. By corollary 5 of tho 1,

§ 2, we have then:
b(Zi; CA,) = Ai 1 YAi'
We may apply (5) to the fields Zi'Z; instead of K, K", replacing at the
same time h by the identity since the Galois group of Lover Zi is the
commutative group Ai- 1 r Ai' The Galois group of Lover Z; is the inter-
section of the latter group with g'; with the same notations as before,
this is Ai 1 r; Ai; it is of index di in Ai 1 r Ai' so that di is the degree of Z; over
Zi' As CA, is in Zi' we have then NZilZ,W')=W,)d,. Therefore (5), applied
to Zi' Z; and CA', gives:
b(Z;; CA,) = b(Zi; (CA,)d,) = (Ai 1 yAl'.
In view of (6) and (7), our conclusion follows from this at once.
Chapter XIII
Global classfield theory

§ 1. The canonical .,airing. In this Chapter, k will be an A-field; we
use the same notations as in earlier chapters, e.g. kv, rv, qv, kA , etc. We
choose an algebraic closure k of k, and, for each place v of k, an algebraic
closure Kv of kv, containing k. We write k sep , kv,sep for the maximal se-
parable extensions of kin k, and of kv in Kv, respectively. We write kab ,
kv,ab for the maximal abelian extensions of kin k sep , and of kv in kv,sep,
respectively. One could easily deduce from lemma 1, Chap. XI-3, that
kv,sep is generated over kv by k sep , and therefore Kv by k, and we shall see
in § 9 of this Chapter that kv,ab is generated over kv by kab ; no use will be
made of these facts. We write ffi and m: = ffi/ffi(l) for the Galois groups of
ksep and of kab , respectively, over k; we write ffiv and m: v= ffivlffi~l) for
those of kv,sep and of kv,ab, respectively, over kv' We write Pv for the re-
striction morphism of ffiv into ffi, and also, as explained in Chap.XII-l,
for that of m: v into m:. We write X k for the group of characters of ffi, or,
what amounts to the same, ofm:; for each XEX k, we write Xv = XO Pv; this is
a character of ffiv, or, what amounts to the same, of m: v'
PROPOSITION 1. Take any XEX k ; call L the cyclic extension of k
attached to X. Let v be any place of k; let L' be the cyclic extension of kv
attached to Xv = xo Pv, and let w be any place of L lying above v. Then there
is a kv-linear isomorphism of L' onto Lw'
As observed in Chap. IX-4, L' is the compositum of Land kv in Kv'
As it is of finite degree over kv, it is a local field, and prop. 1 of Chap.
111-1 shows that L is dense in it; therefore it is the completion of L at a
place lying above v. Our conclusion follows now from corollary 4 ofth. 4,
Chap. I11-4.
COROLLARY. Notations being as in proposition 1, Xv is unramified for
almost all v; if Xv is trivial for almost all v, X is trivial.
The first assertion follows at once from tho 1 of Chap. VIII -4, combined
with proposition 1; the second one follows similarly from corollary 4 of
tho 2, Chap. VII-5, when one takes there for V the set of all the finite places
of k where Xv is trivial.
We can now apply to kv and Xv the definitions and results of Chap.
XII-2. For any zEk:, we will write (Xv,z)v instead of(Xv,z)kv ' The canonical
§l. The canonical pairing 245
morphism of k: into ID: v will be denoted by av; then we have (Xv,z)v=

= Xv(av(z)) for all zEk:. For every zEk:, pv(av(z)) is the automorphism of
kab over k induced on kab by the automorphism av(z) of kv,ab over kv'
Take now z = (Zv)E k1. For almost all v, Zv is in r:, and, by the corol-
lary of prop. 1, Xv is unramified, so that we have (Xv,zv)v= 1 by corollary 4
ofth. 1, Chap. XII-2. Therefore, in the product
(1)
taken over all the places of k, almost all the factors are equal to 1, so that
the product is well defined. The continuity of zv~(Xv,zv)v for each v,
together with the facts mentioned above, implies that Z~(x,Z)k is conti-
nuous on k1; therefore it is a character of k1, whose order is finite since
it divides that of X. The pairing of X" with k1, given by (1), will be called
the canonical pairing for k; it is clear that it satisfies condition [I] of
Chap. XII-I. As to condition [II], assume that z~(X,z)" is trivial on k~;
then Zv ~(Xv' zv)v must be trivial for every v. As [II] is satisfied for local
fields, this implies that all Xv are trivial, hence that Xis so, by the corollary
of prop. 1. This proves [II] for the pairing (1).
As explained in Chap. XII-I, we can now define the canonical mor-
phism a of k1 into ID: by writing, for all XE X k and all Z= (Zv)E k~:
(2)
v
Then, by [II"] of Chap. XII-I, a maps k~ onto a dense subgroup of ID:.
PROPOSITION 2. Let jv be the natural injection of ~ into k~, mapping

k: onto the quasi/actor k: of k~. Then aojv=pvoav'
In fact, if ZvE k:, z = iv(zv) is the idele whose coordinates are all 1
except the one corresponding to v which is Zv' Put ot = av(zv); (2) gives here:
x(a(z)) =(Xv,zv)v = Xv(ot) = X(Pv(ot)).
As this is so for all characters X ofID:, it implies a(z)=pv(ot), as was to be
proved.
THEOREM 1. Let k' be an extension of k of finite degree, contained in Ii;
let (f), (f)' be the Galois groups of ksep over k, and of k~ep over k', respectively,
and let p be the restriction morphism of (f)' into (f). Then, for every charac-
ter XEX", and for every z'Ek'; , we have:
(xop,z')", = (X,N"'I"(z'))".
In view of our definitions, this is an immediate consequence of tho 2
of Chap. XII-2, combined with corollary 3 ofth. 1, Chap.-IV-1.
246 Global classfield theory XIII
COROLLARY 1. If a, a' are the canonical morphisms for k and for k',
respectively, we have po a' = ao N k' Ik'
This is just a restatement of theorem 1.

Nk'lk(k';) is contained in the kernel of xoa if and only if k' contains the
cyclic extension of k attached to X.
In fact, theorem 1 shows that it is contained in that kernel if and only

if xopoa' is trivial, hence, by [II], if and only if Xop is trivial. Let L be
the cyclic extension of k attached to x; then the cyclic extension of k'
attached to XO p is the compositum L' of k' and L, and XO p is trivial if and
only if 1.: = k', i. e. k' => L.
Our main business in this chapter will be to determine the kernel
of the canonical morphism a. For the time being, we merely observe that
it must contain the kernel of av for every v; this is {I} if v is finite, but it is
R~ if v is real, and C x if v is imaginary. We will write k~ + for the product
of the latter kernels in k~, i.e. for the group of the ideles (zv) such that
Zv = 1 for every finite place v, and zv> 0 for every real place v; then this
group is contained in the kernel of a; of course it is {I} if k is of charac-
teristic p > 1.
We will now give explicit formulas for (x,Z)k in some special cases, and
begin by considering a field k of characteristic p > 1. Let then F be the
field of constants of k; call q the number of elements of F, and F the alge-
braic closure of Fin k. By tho 2 of Chap. I-I, EX is the group of the roots
of 1 in k, and all these roots have an order prime to p. We will call ko the
compositum of k and F, f)o the subgroup of (fj corresponding to ko, and
Xo the subgroup of X k consisting of the characters of (fj, trivial on f)o.
Clearly every extension of k of finite degree, contained in ko, is generated
over k by finitely many elements of F, hence by an extension F' of F of
finite degree. More precisely, we have the following:
LEMMA 1. Let F' be the extension of F of degree n, contained in k; then

the compositum k' of k and F' is cyclic of degree n over k; its field of con-
stants is F'; and the restriction morphism of the Galois group of k' over k
into that of F' over F is an isomorphism of the former onto the latter group.
Call F" the field of constants of k', n' the degree of k' over k, and n" that
of F" over F; clearly n' ~n~n". Take CeF" such that F" =F(C), and call P
the irreducible monic polynomial in F[ X] with the root C. If Q is a monic
polynomial in k[X], dividing P in k[X], all its roots are in F, so that its
coefficients are in Frlk, i.e. in F. Therefore P is irreducible in k[X], so
§l. The canonical pairing 247
that n'~n", hence n' =n=n". The assertion about the Galois groups may
now be regarded as a special case of corollary 1 of prop. 3, Chap. III-2, or
also as following from k' = F' ® F k, which is an immediate consequence of
prop. 2, Chap. 111-2.
Whenever k and k' are as in lemma 1, we will say that k' is a constant-
field extension of k. In view of tho 2, Chap. 1-1, there is, for every integer
n ~ 1, one and only one such extension of k of degree n; this will be denoted
by k". Then ko is the union of the cyclic extensions k" for all n ~ 1; in parti-
cular, it is contained in kab ; we will denote by ~o the subgroup of ~
corresponding to ko, i.e. the group of automorphisms of kab over k o. We
may then consider X 0 as being the group of characters of~, trivial on ~o;
a character xeX k belongs to Xo if and only if the cyclic extension of k
attached to X is contained in ko, hence if and only if it is one of the fields
k".
By corollary 2 ofth. 2, Chap. 1-1, combined with lemma 1, there is, for
every n ~ 1, one and only one automorphism of k" over k, inducing on the
field of constants F"=Fnk,, of k" the automorphism x--+xq, where q, as
before, is the number of elements of F ; moreover, this generates the Galois
group of k" over k. Consequently there is one and only one automorphism
CPo of ko over k, inducing on F the automorphism x--+xq; this will be
called the Frobenius automorphism of ko over k. Every automorphism
cP of ksep over k, inducing CPo on ko, will be called a Frobenius automorphism
of ksep over k; then the Frobenius automorphisms of ksep over k make up
the coset i)o cp in 6>.
PROPOSITION 3. Let k be an A-field of characteristic p> 1 with the field

of constants F = Fq. Let X be a character of 6> belonging to X 0' i. e. such
that the cyclic extension of k attached to Xis a constant-field extension of k.
Let cp be any Frobenius automorphism of ksep over k. Take zEk~, and put
IzlA =q-r. Then (X,Z)k= X(cpy.
Put z = (zv). Let v be a place of k of degree d, i. e. such that the module of
kv is qv = qd. Let L be the cyclic extension of k attached to x; this is gener-
ated over k by some extension F' of F, hence by roots of 1 of order prime
to p. Therefore the extension of kv attached to Xv, being generated by F'
over kv, is unramified, so that Xv is unramified. Moreover, a Frobenius
automorphism of kv,sep over kv ind~ces on F the automorphism X--+xqd
and therefore coincides with cpd on F, hence on k o. By corollary 4 ofth.l,
Chap. XII-2, this gives (xv,zv)v=X(cpd)v with v=ordv(zv)' As IZvlv=q-dv
and IzlA = nlzvlv, our conclusion follows from this at once.
COROLLARY 1. Assumptions and notations being as in proposition 3, and

a being the canonical morphism for k, a(z) coincides with cpr on k o.
In fact, they can differ only by an element belonging to the kernels of

all the characters XeXo, and the intersection of these kernels is t)o,
which is the group leaving ko invariant.
COROLLARY 2. If xeXo and ()ek X , (X'())k= 1.
This follows at once from proposition 3 and from the fact that I()IA = 1
by tho 5 of Chap. IV -4.
COROLLARY 3. A character Xof (fj belongs to Xo if and only if (X,Z)k= 1
for all zek;'.
If XeXo, proposition 3 shows that X has the latter property. Now
assume that X has that property. By corollary 6 of tho 2, Chap. VII-5,
there is Z1ek~ such that Iz 11A = q, and then k~ is generated by k;' and z 1.
Let n be the order of x; then (X,Z1)k is a primitive n-th root of 1 in C. As ((J
induces on kn a generator of the Galois group of k n over k, there is a
character x' attached to kn' such that X'«((J)=(X,Z1)k. By proposition 3, we
have then (X',Z1)k= X'«((J)-l, hence (XX', Z1)k= 1, and therefore, in view of
proposition 3 and our assumption on X, (X X', Z)k = 1 for all zek~. This
implies that X=X'-1, so that XeXo.
When k is of characteristic 0, there is no such convenient tool as the
one supplied by the constant-field extensions in the case of characteristic
p> 1; the nearest substitute is provided by the "cyclotomic" extensions;
here we merely consider the case k = Q; then (fj is the Galois group of
Q = Qsep over Q. For m ~ 1, let e be a primitive m-th root of 1 in Q; call tim
the subgroup of (fj corresponding to Q (e), so that the Galois group of
Q(e) over Q is g = (fj/t)m. It is well known that g consists of the automor-
phisms determined by e~ex, when one takes for x all the integers prime to
m modulo m; it may thus be identified with (Z/mZr, i.e. with the multi-
plicative group of the ring Z/mZ. Let X be any character of g, with the
kernel 1); we identify this in an obvious manner with a character of (fj,
which we also call x; this has a kernel t)::) t)m. On the other hand, when we
identify 9 with (Z/m Z) x , we also identify X with a function on the latter
group and therefore with a function on the set of all integers prime to m,
which we also call X, and which is then such that x(ab)=x(a)x(b) when-
ever a, b are two such integers. This can then be uniquely extended to a
character of the subgroup of Q x consisting of the fractions alb, with
a, b in Z and prime to m; also the latter character will be denoted by x.
With these notations, we have:
PROPOSITION 4. Let X be as above, and let Z be the cyclic extension of Q
attached to x. Then, for every rational prime p not dividing m, Xp is unrami-
fled, and, for every zeQ;, (x p ,z)p=x(lzl;1); for every zeR\ (xcx"Z)oo=
=x(sgnz).
§1. The canonical pairing 249
Let p be any prime not dividing m; let w be a place of Z, lying above p,

and let u be a place of the field L = Q(e), lying above w. By prop. 1 of
Chap. 111-1, Lu is generated over Qp bye; as this is of order m prime to p,
Lu is unramified, and so is Zw, hence also XP' by prop. 1. A Frobenius
automorphism cp, over Qp, of the algebraic closure of Qp induces on L u ,
hence on L, the automorphism determined by e--+eP ; therefore x(cp),
according to the notations explained above, is the same as X(p). In view of
this, our assertion about (Xp' z)p is an immediate consequence of corollary
4 ofth. 1, Chap. XII-2, and oflzl p= p-Ord(Z). Similarly, let w be a place of Z
lying above the place 00 ofQ, and u a place of L lying above w. If m= 1 or 2,
X is trivial, and our last assertion is obvious. Ifm>2, Lu=R(e)=C has
the non-trivial automorphism x--+x over R; this is the one determined
by e--+e-1, so that, if 9 and ~ are as explained above, it induces on Zw
the automorphism corresponding to the image of - 1 in gfl>. If X( -1) = 1,
-1 is in~, Zw=R, and XOO is trivial; if X( -1)= -1, -1 is not in~, Zw=C,
and Xoo is non-trivial. The last assertion in proposition 4 follows at once
from this and from the results stated at the beginning of Chap. XII -2.
COROLLARY 1. Assumptions and notations being as above, let w be a
place of z. If w lies above a rational prime p, not dividing m, the degree
of Zw over Qp is the order of X(p) in the group C if w lies above 00,
X
;
the degree of Zw over R is the order of x( -1) in C X •

The latter assertion was proved above; as to the former, prop. 1
shows that the degree in question is equal to the order of the character
z--+(Xp,z)p ofQ p; proposition 4 shows that this is as stated in our corollary.
COROLLARY 2. The character X being as above, take z=(zv)eQ~ such
that, for every prime p dividing m, ordp(zp)=O and (Xp,zp)p= 1. Then
(X,z)Q = x(r(z)), with r(z) given by
r(z)=sgn(zoo)nlzpl;l.
p
In the latter formula, the product is taken over all the rational
primes, or (what amounts to the same, in view of the assumption on z)
over all the primes not dividing m; then x(r(z)) is well defined. Our
assertion follows now at once from proposition 4 and the definitions.
COROLLARY 3. The character X being as above, one can choose, for
every prime p dividing m, an open subgroup gp of Q; such that (x, e)Q = 1
for allee n (Qx ng p ).
For each p dividing m, let p" be the highest power of p dividing m;
then 1+mZp is the same as the subgroup 1+pllZp ofQ;. Take now
for gp' for each p dividing m, the intersection of 1 + mZp with the kernel
of z~(Xp,z)p in Q;. e
Then, if is as in our corollary, corollary 2 shows
that (x,e)Q is equal to x(r(e)); by th.5 of Chap. IV-4, r(e) is equal to
sgn(eHeloo' i.e. to e. Write e=a/b, with a, b in Z and (a, b) = 1. If pis
any prime dividing m, e is in Zp, so that b is prime to p; if pll is as above,
m
e is 1 + pll Zp, so that a == b (Pll). Therefore a and b are prime to m, and
a==b (m); this implies x(a) = X(b), hence, in view of our definitions, X(e)= 1,
which completes the proof.
§ 2. An elementary lemma. As above, let X be a character of (Z/mZY;

considering it again as a function on the set of all integers prime to m,
we now associate with it the function t/I on Z such that t/I(x) = X(x)
whenever x is prime to m, and t/I(x) = 0 otherwise. It is customary, by
abuse of language, to call such a function t/I "a multiplicative character
modulo m", or, more briefly, "a character modulo m" on Z. Obviously
a function t/I on Z is such a character if and only if t/I(x + m) = t/I(x) for
all xeZ, t/I(x)=O for (x,m)f 1, t/I(1) = 1, and t/I(ab)=t/I(a)t/I(b) for all
a, bin· Z; it will be called trivial if t/I(x) takes no other values than 0
and 1, and of order n if t/I" is trivial. If t/I, t/I' are such characters, modulo
m and modulo m' respectively, t/I t/I' is a character modulo mm'.
The object of this § is to prove lemma 3; the lemma and the proof
are due to van der Waerden. We begin with a special case.
LEMMA 2. Let I be a rational prime, n an integer ~ 1, and a1, ... ,a,

integers > 1. Then there is a multiplicative character t/I on Z, such that,
for every i, t/I (a i) is a root of 1 whose order is a multiple of l"; moreover,
there is such a t/I whose order is a power of I, and, if 1= 2, there is such a
t/I for which t/I (-1)= -1.
Clearly the order of t/I(ai) is a multiple of that of t/I(af). If 1=2, we
replace each ai by af; after doing this, we may therefore assume, in that
case, that ai==O or 1 (4) for all i. For each i, we will now derme a sequence
of primes Pi,v(V=O, 1, ...) as follows. If aj¢ 1 (I), we take for Pj,v any prime
divisor of the integer
,v+l 1
aj - = 1+a~v + ... + a~V('-l)
(3) 'v
aj - 1 I I '
Clearly, as r+ 1 == 1 (1-1), the numerator ofthe left-hand side is ==aj-1 (I),

hence not a multiple of I; therefore Pj,v f I. On the other hand, if aj == 1 (I),
write ar in the form ar = 1 + I!l.b with ex~ 1 and b¢O (I); if 1=2, we have
ex ~ 2, in view of our assumption on the aj in that case. Then:
§ 2. An elementary lemma 251
This shows that the left-hand side of (3) is then a multiple of I and not
of 12; as the right-hand side shows, it is > 1; we take then for Pi. any
prime divisor of that left-hand side, other than I. Now we show that,
in all cases, Pi,. cannot divide the denominator of the left-hand side of
(3). In fact, assume that it does; then all the terms in the right-hand side
are == 1 (Pi,.), so that the right-hand side itself is == I (Pi,.); as it is a multiple
of Pi, .. and as Pi,. '1= I, this cannot be. This shows that the image of ai in
the group (Z/Pi,.Z)X is exactly of order [V+ 1; in particular, for each i, all
the Pi,. are distinct. Therefore, choosing an integer p such that IP > r, we
may, for each i, choose an integer vi~n+p-2 such that the prime
Pi=Pi,'i does not divide any of the integers a1 , ••• ,ar • For each i, the
group (Z/Pi Z) x is cyclic of order Pi -1, and the image of ai in that group
has the order [Vi + 1; call Xi a generator of the group of characters of
e·
that group; call i the highest power of 1 dividing Pi -1; put
,u=Ai -v i +n+p-2, m=I-I'(pi-1) and x;=xi; clearly Ai>V i, so that
,u>O; then X; is a character of order II', and it is easily verified that x;(a i)
is a root of 1 of order ,n +P - 1. For each i, extend X; to a multiplicative
character !/Ii modulo Pi on Z, as explained above; let M be an integer
such that 1M is a mUltiple of the order of all the X;, hence of all the !/Ii'
As each Pi is prime to all the aj' we can then write
!/Ii(aj) = eW M bij)
with bijEZ, for 1 ~i,j~r; moreover, for each i, the highest power of 1
dividing bii is IM -n-p+l. Now consider the lMr characters
WX = O(!/Ii)XI
i=l
with 0 ~ Xi < 1M for 1 ~ i ~ r; of course they need not all be distinct. For
eachj, we have
This is a root of 1, of order dividing 1M; that order is a multiple of l"

unless it divides 1"- 1, i. e. unless we have
bjjx j == - L bijX
itj
i
For a given j, and for each set of values for the Xi for i '1= j, this congruence
has either no solution Xj at all, or exactly lM - p solutions modulo 1M;
therefore, for each j, there are at most IMr-p sets of values for the Xi'
satisfying 0 ~ Xi < 1M for 1 ~ i ~ r, such that wx(aj) has an order dividing
1"-1. Consequently there are at most rlMr - p such sets for which at least
one of the (J)",(aj) has an order dividing 1,,-1; as r< IP, the number of such
sets is < 1M '. This proves that one can choose x so that the order of
(J)",(aj) is a multiple of I" for allj. Then r/I = (J)", is a solution of our problem,
except perhaps in the case 1=2, since then we also want r/I to be such
that 1/1 ( -1) = -1. In that case, if (J)",( -1) = -1, we take 1/1 = (J)",. If not,
take a prime Po dividing 4a 1a2 ••• a,-I, and == -1 (4); clearly there is
at least one such prime. Then the group (Z/Po Z) x is cyclic of order
2 mo with mo = (Po -1 )/2 == 1 (2), so that it has exactly one character XO
of order 2; this satisfies Xo( -1)= -1. Extending XO to a multiplicative
character r/lo modulo Po on Z, one sees at once that r/I = r/lo (J)", is a solu-
tion of our problem, provided one has taken n ~ 2, as may of course
always be assumed. This completes the proof.
LEMMA 3. Let a1> ... ,a" n1, ... ,n, be integers > 1. Then there is a
multiplicative character 1/1 on Z such that r/I( -1)= -1 and that, for
every i, 1/1 (a j ) is a root of 1 whose order is a multiple of nj •
Put N =2nn j ; for every prime I dividing N, let 1" be the highest
power of I dividing N, and let 1/1, be chosen according to lemma 2, so
that its order is a power of I, the order of r/I,(a j ) is a multiple of 1" for
every i, and 1/1,( -1)= -1 if 1=2. When I is odd, r/I,( -1), being ± 1 and
n
of odd order, is 1. That being so, it is clear that 1/1 = 1/1, solves the
problem.
§ 3. Hasse's "law of reciprocity". As in Chap. XI, if A is a simple

algebra over k, and v any place of k, we write Av for the algebra A®kk v
over k v ; we have seen in Chap. IX-3 that the mapping CI(A)-CI(Av) is
then a morphism of the Brauer group B(k) of k into the Brauer group
B(kv) of kv. It has been shown in Chap. XII-2 that the Hasse invariant h
determines an isomorphism of B(kv) onto a group H v consisting of all
the roots of 1 in C if v is a finite place, of ± 1 if v is real and of 1 if v is
imaginary. From now on, for any simple algebra A over k, we will write
hv(A)=h(Av); this will be called the Hasse invariant of A at v. By tho 1
of Chap. XI-I, we have hv(A) = 1 for almost all v; therefore the mapping
A -(hv(A») determines a morphism h of B(k) into the "direct sum" of
the groups Hv for all v, i.e. into the subset H of nHv consisting of the
v
elements (11v) of that product such that "Iv = 1 for almost all v. By th.2
of Chap. XI -2, the kernel of h is the class of trivial algebras over k, so
that h is injective. It will be shown in § 6 that h(B(k») consists of the
elements (11v) of H such that n'1v= 1. In this §, we show that nhv(A) = 1
v v
for every simple algebra A over k.
As in § 1, let X be a character of <D, and L the cyclic extension of k

attached to X; for any oee,consider the cyclic algebra A=[L/k;X,O],
corresponding to the factor-class {X,O}. As we have seen in Chap. IX-4,
the restriction morphism maps the factor-class {X,O} of k onto the factor-
class {Xv,O} of kv, so that Av belongs to the latter class. Therefore, by the
definition of the Hasse invariant, we have, for all v:
(4)
v
On the other hand, let k' be an extension of k of finite degree; A being

any simple algebra over k, put A' = A ®k k' ; let w be a place of k', and v the
place of k lying below w. The transitivity properties of tensor-products
show at once that the algebra (A')w=A'®k,k~ over k~ may be identified
with AV®kvk~; therefore, by corollary 2 of tho 2, Chap. XII-2, we have
hw(A') = hv(At(W) if n(w) is the degree of k~ over kv' In particular, in view
of what has been said above, A' is trivial if and only if hv(At(w) = 1 for
every place w of k'.
PROPOSITION 5. For any XeXt> let L be the cyclic extension of k
attached to X. Let A be a simple algebra over k. Then the following asser-
tions are equivalent: (i) AL is trivial; (ii) for every place v of k, and every
place w of L above v, the degree of Lw over kv is a multiple of the order of
hv(A) in the group ex ; (iii) A is similar to a cyclic algebra [L/k;X,O] with
some Oek x ; (iv) there is z=(zv) in k1, such that hv(A)=(Xv, zv)v for every
place v of k. Moreover, if 0 is as in (iii) and z as in (iv), 0- 1 z is in N Llk(L~).
The equivalence of (i) and (ii) is a special case of what has just b~en
proved above; that of (i) and (iii) is contained in prop. 9 of Chap. IX-4.
Assume (iii); then, by (4), (iv) is satisfied if we take z = O. Now assume
(iv); then the order of hv(A) divides that of Xv, which, by prop. 1 of § 1,
is equal to the degree of Lw over kv for every place w of L lying above v, so
that (ii) is satisfied. Finally, let 0 be as in (iii), z as in (iv), and put z' = 0- 1 Z;
by (4), we have then (xv,?~)v= 1 for all v; by prop. 10 of Chap. IX-4, and
prop. 1 of § 1, this implies that, if w is any place of L above v, z~ is in
N Lw/dL:). For each place v of k, choose tweL:, for all the places w of
L lying above v, so that z'v = N Lw/dtw) for one of these places, and tw = 1
for all the others; as Iz~lv= 1 for almost all v, this implies that Itwl w = 1 for
almost all w, so that t = (tw) is in L1 ; then we have z' = N Llk(t).
We will now use proposition 5 in order to show that every simple
algebra A over k is similar to one of a very special type. For this, we re-
quire two lemmas.
LEMMA 4. Let k be of characteristic p> 1. For every place v of k, let
v(v) be an integer ~ 1, such that v(v) = 1 for almost all V. Then there is a
constant-field extension k' of k such that, if v is any place of k and w a

place of k' above v, the degree of k~ over kv is a mUltiple of v(v).
Let F = Fq be the field of constants of k. For a place v of k of degree
d(v), the module of kv is qd(v); therefore, by corollary 3 of tho 7, Chap. 1-4,
if k~ contains a primitive root of 1 of order qd(V)f - 1, its degree over kv
must be a multiple of f. The condition in lemma 4 will therefore be satis-
fied if we take for k' a constant-field extension of k whose degree over k is
a multiple of all the integers d(v)v(v) corresponding to the finitely many
places v where v(v) > 1.
LEMMA 5. Let k be of characteristic O. For every place v of k, let v(v)
be an integer ~ 1, such that v(v) = 1 for almost all v, v(v) = 1 or 2 whenever v
is real, and v(v) = 1 whenever v is imaginary. Then there is an integer m~ 1
and a cyclic extension Z of Q, contained in the extension Q(e) generated by
a primitive m-th root e of 1, with the following properties: (a) if v is any
place of k, and w a place, lying above v, of the compositum k' of k and Z,
the degree of k~ over kv is a multiple of v(v); (b) Imlv=1 whenever v is
a finite place of k and v(v) > 1.
To begin with, let Z be any extension of Q, and let k' be its composi-
tum with k. Let v be any place of k, w a place of k' lying above v, u the
place of Z lying below w, and t the place of Q lying below u. Then kv,
Zu and Qt are respectively the closures of k, Z and Q in k~, so that t also
lies below v. We have:
therefore, if we put
v'(v) = v(v)· [kv:Qt],
and if [Zu:Qt] is a multiple of v' (v), [k~:kv] will be a mUltiple of v(v).
Now, for every finite place t of Q above which there lies some place v of
k where v(v» 1, call net) some common multiple of the integers v'(v) for
all the places v of k above t; for all other finite places t ofQ, put net) = 1; put
n(oo)=2, this being a multiple of v' (v) for every infinite place v of k, as
one sees at once. Then m and Z, in our lemma, will satisfy our require-
ments if [Zu: Qt] is always a multiple of net) and if 1~lt = 1 when t'=l= 00,
net) > 1; in other words, it is enough to prove our lemma for k = Q.
Call thenpl, ... ,Pr the rational primes p for which n(p» 1; apply lemma 3
of § 2 to the integers aj = Pj, nj = n(pj); we get a multiplicative character '"
on Z, modulo some integer m, such that", ( -1) = -1 and that, for each i,
"'(pj) is a root of 1 whose order is a multiple ofn(pj). As "'(x)=O when x
is not prime to m, m is then prime to all the Pj, which is the same as to say
that Imlp= 1 when n(p» 1. Let then X be the character of (Z/mZr deter-
mined by 1/1; consider this as a character of the Galois group of Q(e)

over Q, e being a primitive m-th root of 1, and call Z the cyclic extension
of Q attached to x; then corollary 1 of prop. 4, § 1, shows that m and Z
satisfy all the requirements in our lemma.
THEOREM 2. If A is any simple algebra over k, we have

the product being taken over all the places v of k. v
hv(A) = 1, n
If k is of characteristic p> 1, prop. 5 and lemma 4 show that A is simi-
lar to a cyclic algebra [k'/k; X,O], where k' is a constant-field extension of
k, X a character attached to k', and Oek". Then X is in X o, where Xo is
as defined in § 1, and our conclusion follows at once from (4) and corollary
2 of prop. 3, § 1. If k is of characteristic 0, we apply prop. 5 and lemma 5,
taking for v(v), in the latter lemma, the order of hv(A) in C"; this shows
that A is similar to a cyclic algebra [k'/k;X',O], where k' is as in lemma 5,
x' is any character attached to k', and Oek". By (4), what we have to prove
is that (X', O)k = 1. Let m and Z be as in lemma 5; then we can take x' = xo p,
where p is the restriction morphism of the Galois group ofQ·over k into
that of Q over Q, and X is a character of the latter group attached to Z.
Call VI'" ., VM all the places of k lying above some rational prime dividing
m; for each i, choose a place Wj of k' lying above Vj; call W'I"'" WN all the
places of k', other than the Wj, lying above some Vj; for each i, call kj, k; the
completions of kat Vj, and of k' at Wj, respectively; for eachj, call k'J the
completion of k' at wj. By condition (b) in lemma 5, and in view of our
choice ofthe v(v), we have hvj(A) = 1 for all i; by (4), prop. 1 of§ 1, and prop.
10 of Chap. IX-4, this implies that, for each i, we can write O=Nkl/kj(Zj)
with zjeki". By corollary 2 of tho 3, Chap. IV-2, there is an element, of
k' whose image in k;, for 1 :::;;i:::;;M, is arbitrarily close to Zj, and whose
image in k'J, for l:::;;j:::;; N, is arbitrarily close to 1. In view of corollary 3
of tho 1, Chap. IV-I, this implies that we can choose Cek'" so that the
image of 0 1 = 0 N k' /k(O - 1 in k j is arbitrarily close to 1 for 1:::;; i :::;; M. By
prop. 10 of Chap. IX-4, one does not change the factor-class {x',0} if one
replaces 0 by 01 ; consequently, this does not change the invariants
hv(A)=(X~,e)v of A. Therefore it is enough if we prove our assertion
(X',O)k= 1 under the additional assumption that the image of 0 in kj,
for 1:::;; i:::;; M, is in a prescribed neighborhood of 1. By corollary 3 of tho 1,
Chap. IV-I, these neighborhoods can be so chosen that the image of
Nk/Q(O) in Qp, for every prime p dividing m, is arbitrarily close to 1. As
x' = Xop, and as we have, by tho 1 of § 1:
our assertion follows now from corollary 3 of prop. 4, § 1.

COROLLARY. For every XeX", and every (}ek\ we have (x, (})" = 1.
This follows at once from (4) and theorem 2.
The corollary of theorem 2 is known as "Artin's law of reciprocity"

because Artin discovered it (in substance) and pointed out that the "laws
of reciprocity" of classical number-theory can easily be derived from it
* and from purely local considerations. Theorem 2 is due to Hasse; its
close connection with "Artin's law" accounts for the name of "Hasse's
law of reciprocity" which is usually given to it.
The corollary of theorem 2 may be expressed by saying that, for every
XeX", the character z--'(X,z)" of k~ is trivial on k X , or that k X is contained
in the kernel of the canonical morphism. Consequently, we may now
regard (X,z)" as defining a pairing between X" and the "idele-class group"
G,,=k~/kx of k; in order not to complicate notations, we do not intro-
duce any new symbol for this pairing, but we will apply to it the results of
Chap. XII-1 in an obvious manner. It is clear that it satisfies conditions
[I] and [II] of Chap. XII-1, since these have been verified in § 1 for (X,z)"
considered as a pairing of X" and k~. By tho 6 of Chap. IV-4, G" is quasi-
compact, and we have G~ = kl/k x. If k is of characteristic 0, condition
[III(a)] of Chap. XII-1 is satisfied. If k is of characteristic p > 1, corollary 3
of prop. 3, § 1, together with lemma 1 of § 1, shows that condition [III(b)]
of Chap. XII -1 is satisfied by taking for X, in that condition, a character
attached to the constant-field extension of degree n of k; it also shows
that the group denoted by X 0 in § 1, and consisting of the characters
attached to the constant-field extensions of k, is now the same as the
group which was so denoted in Chap. XII-l. In that case, we can now
apply corollary 2 of prop. 2, Chap. XII-1, which shows that the canonical
morphism a maps kl onto the subgroup 210 of 21 corresponding to the
union ko of the constant-field extensions of k; similarly, if k is of charac-
teristic 0, prop. 1 of Chap. XII-l shows that a maps k~ onto 21. If we call
again U" the kernel of a, it contains k x , and, if k is of characteristic p> 1,
°
corollary 2 of prop. 2, Chap. XII-l, shows now that U"ckl; on the other
hand, if k is of characteristic and if the subgroup k~ + of k~ is defined as
in § 1, U" contains the closure of k x k~ +. In §8, it will be shown that U" is
that closure if k is of characteristic 0, and that otherwise U" = k x •
We will now reformulate prop.4 of Chap. XII-1 for the pairing
between X" and Gk defined above. Let k' be a cyclic extension of k. Then
e--.eJ., for every Ae<f>, and e--.N"'I,,(e) are polynomial mappings of k', into
k' and into k respectively, when k' is regarded as a vector-space over k,
and we have N"'I,,(eJ.)=N"'I,,(e) for all e. As explained in Chap. IV-1,
N "'I'" as a mapping of k~ into kA' is the extension to these spaces of the
polynomial mapping N""" of k' into k, and we now extend to k~ the
§ 4. Classfield theory for Q 257
k-linear mapping ~-+~). of k' onto k' in the same manner. Then we have
Nk'lk(X').) = Nk'lk(X') for all x'Ek~, and in particular for all x'Ek~x. At the
same time, by the corollary of prop. 3, Chap. IV-3, we have Iz'IA =
= IN k'lk(z')IA for all z' Ek~x , hence Iz').IA = Iz'IA for all z' Ek~x and all AEG>.
As the morphisms z' -+ z' \ z' -+ N k' Ik(Z') of k'; onto k'; and of k'; into k~
map k'x onto klX and into k x respectively, they determine morphisms
of Gk, onto Gk, and into Gk respectively; we take these as the mappings
g'-+g'\ g'-+F(g'), in prop. 4, Chap. XII-I. It is now clear that these map-
pings satisfy conditions [IV(iHii)] and [V(i)]; [IV(iii)] is an immediate
consequence of the definitions and of corollary 5 of tho 1, Chap. XII-2 (it
admits of an obvious generalization, quite similar to the latter corollary),
while [V(ii)] is here the same as the assertion in tho 1, § 1. All conditions for
the application of prop. 4, Chap. XII-I, being thus fulfilled, we conclude
that we have, in our present notation:
(5)
where k is any A-field, k' any cyclic extension of k, and Uk' Uk' are the
kernels of the canonical morphisms for k and for k', respectively.
§ 4. Classfield theory for Q. The results already obtained make it

easy to conclude our investigation in the special case k = Q; this is due to
the fact expressed in the following lemma:
LEMMA 6. We have the direct product decomposition:
Q~=Qx xR: xflZ;
p
where the last product is taken over all the rational primes p.
Here R: and the Z;
are to be understood as subgroups of the quasi-
factors R x = Q:O and Q; of Q~. As in corollary 2 of prop. 4, § 1, define a
morphism r of Q~ into Q x by putting, for z = (zv) in Q~ :
r(z) = sgn(z<x,) fllZpl; 1;
p
as we have already observed (in the proof of corollary 3 of prop. 4, § 1), r

induces the identity on Q x, as follows at once from tho 5 of Chap. IV -4.
Therefore, if R is the kernel of r, r determines a direct product decomposi-
tion Q~ = Q x x R and is the projection from that product onto its first
factor. Clearly R = R ~ x fl Z;. This proves the lemma.
p
Clearly the subgroup fl Z;

of Q~ is totally disconnected, so that,
by lemma 4 of Chap. VII-3, all its characters are of finite order. It follows
now at once from lemma 6, combined with corollary 2 of prop. 7, Chap.
VII-3, that every quasicharacter of Q~, trivial on QX, is of the form w. tjJ,
where tjJ is trivial on Q x x R ~ and where w., as in Chap. VII, denotes
n
the quasicharacter z~lzlA and is trivial on Q x and on Z;; tjJ is then a
character of finite order. We recall that, if w is any quasicharacter of Q~,
trivial on Q x, the conductor of w, according to the definition given in
Chap. VII-7 for an arbitrary number-field, is the ideal npf(p) of Z,
where, for each rational prime p, pf(p) is the conductor of the quasi-
character wp induced by w on Q;; here, as usual, we identify a non-zero
ideal in Z with the integer > 0 which generates it.
As explained in § 1, if t: is a primitive m-th root of 1 in Q, we identify
the Galois group 9 of Q(t:) over Q with (Z/mZr, and every character X
m
of 9 with a character ofthe Galois group of Qover Q, or, what amounts
to the same, with a character of the Galois group mof Qab over Q. Of
course Q(t:)CQab for all m.
THEOREM 3. For any m> 1, let t: be a primitive m-th root of 1 in Q,

and let g=(Z/mZr be the Galois group of Q(t:) over Q. Then x~xoa
is an isomorphism of the group of the characters X of 9 onto the group of
the characters of Q~, trivial on Q x x R:, whose conductor divides m.
Call r the latter group. Call P the set consisting of 00 and of the
primes p dividing m; for each prime pE P, put 9p = 1+ pll ZP' where pll
is the highest power of p dividing m; call H the subgroup of Q~ con-
sisting of the ideles (zv) such that Zoo >0, zpEgp for every prime PEP,
r
and ZpEZ; for p not in P. Then is the group of the characters w of
Q~ which are trivial on QX and on H. Put goo=Rx and g=Dgv, the
latter product being taken over all VEP; as in Chap. VII-8, call Gp the
subgroup of Q~ consisting of the ideles (zv) such that zv=l for all VEP;
as 9 x Gp is an open subgroup of Q~, and as Q x Gp is dense in Q~ by
prop. 15, Chap. VII-8, we have Q~ = Q x • (g x Gp ). The morphism r of
Q~ onto Q x defined in the proof of lemma 6 maps 9 x Gp onto the
subgroup Q(m) of Q x consisting of the fractions alb, where a, b are in Z
and are prime to m; the kernel of the morphism of 9 x Gp onto Q(m)
induced by r is the group H defined above. As every character in r is
trivial on H, this implies that, for any WEr, there is a character X of
Q(m) such that Xor coincides with W on 9 x Gp . Then, if aEZ and a= 1 (m),
we have aEg x Gp and r(a)=a, hence x(a)=w(a)= 1. Therefore X deter-
mines a character of (Z/m Z) x ; this being also denoted by X, and being
regarded as a character of g, hence of m, corollary 2 of prop. 4, § 1,
shows that (X,z)Q = x(r(z)) for all zEg' x Gp , if g' is a suitable open sub-
group of g. This means that xoa coincides with xor, hence with w, on
g' x Gp • As Q': =Q x .(g' x Gp ) by prop. 15, Chap. VII-8, and as xoa and
ware both trivial on QX, this proves that xoa=w. Conversely, let now
§ 4. Classfield theory for Q 259
x be any character of g=(Z/mZ)X; as in § 1, consider this as a function

on the set of all integers prime to m, and extend this to a character X of
Q(m); then xor is a character of 9 x Gp . Take ~EQx n(g x Gp ); then
r(~) = ~, and one sees at once, as in the proof of corollary 3 of prop. 4, § 1,
that ~EQ(m) and X(~)=I. Therefore xor is trivial on QXn(gxG p ), so
that it can be uniquely extended to a character w of Q~ =Qx .(g x Gp ),
trivial on Q x ; as r is trivial on H, w is also trivial on H, so that it belongs
to r. As above, corollary 2 of prop. 4, § 1, shows now that XO a coincides
with xor, hence with w, on g' x Gp , if g' is a suitable open subgroup of
g; as above, this gives xoa=w, which completes our proof. We see also
that xoa coincides with xor, not only on g' x Gp , but even on 9 x Gp ; in
other words, the conclusion of corollary 2 of prop. 4, § 1, is valid provided
zpEgp for every prime PEP; we will not formulate this as a separate
result, but will use it in the proof of our next corollary.
COROLLARY 1. Let £ be as in theorem 3; take any z=(z,.) in f]Z; and

put ex = a(z) - 1. Then there is an integer a such that a E Z p+ m Z p jar every
prime p, and, for every such a, we have e~ = ea.
The condition on a can also be written as a=.zp (PI') for every prime
p dividing m, pI' being the highest power of p dividing m; it is well known
that these congruences have a unique solution modulo m (this may also
be regarded as a special case of corollary 1 ofth. 1, Chap. V-2). As ZpEZ;
for all p, a is then prime to m; in particular, it is not O. Put then z' = a - 1 z;
then z~Egp for all primes PEP; therefore, as shown at the end of the
proof of theorem 3, we have x(a(z')) = x(r(z')). As a is trivial on Q x ,
a(z')=a(z)=ex- 1 ; as r(a)=a and r(z) = 1, we get X(ex)=x(a). As this is so
for all characters X of g, it shows that the automorphism of Q(e) induced
by ex is the one determined by £--+e a •
COROLLARY 2. The kernel of the canonical morphism a for Q is

Q x x R:, and a determines an isomorphism of f]Z; onto the Galois
group ~ of Q.b over Q.
In fact, we already knew that the kernel of a contains Q x x R:, and
theorem 3 shows that it is contained in it. The last assertion follows now
at once from lemma 6, and prop. 1 of Chap. XII-I.
COROLLARY 3. Q.b is generated over Q by the roots of 1 in the al-
gebraic closure Qof Q.
Let K be the extension of Q generated by these roots, which is the
same as the union of the fields Q(e) for all m> 1, where e is as in theo-
rem 3. Let 513 be the subgroup of ~ corresponding to K. Then, if X is as
in theorem 3, it is trivial on 513, so that xo a is trivial on a - 1 (513). By theo-
rem 3, 0-1(!13) must therefore be contained in QX xR~; as this, by

corollary 2, is the kernel of 0, we must have !l3 = {I}, hence K = Q abo
§ 5. The Hilbert symbol. The determination of the kernel of the

canonical morphism in the general case depends on two results, corre-
sponding to propositions 9 and 10 ofChap.XII-3. In this §, we deal with
the former one; this will require some preparations.
By n, we will understand any integer > 1.
LEMMA 7. Let G be a quasicompact group. Let y be a group of char-

acters of G, all of order dividing n, and let X be the intersection of their
kernels. Then every character of G, trivial on X, is in y.
By lemma 2 of Chap. XII-I, applied to the endomorphism x-+xn of
G, Gn is a closed subgroup of G, and GIGn is compact; therefore the
subgroup of the dual of G, associated by duality with Gn, is discrete;
it consists of all the characters of G which are trivial on Gn, i. e. whose
order divides n. Consequently y is discrete, hence closed, in the dual of
G. Our assertion follows now from the duality theory.
PROPOSITION 6. Let K be a local field containing n distinct n-th roots
of 1. For x, Y in K x , put (x,Y)n K= (Xn x,Y)K' Then
(y,X)n,K = (X,Y)n,K -1
for all x, Y in K x ; (K x t is the set of the elements Y of K x such that
(X,Y)n,K= 1 for all xEKx; if modK(n) = 1, and if R is the maximal compact
subring of K, the set of the elements Y of K x such that (x, Y)n,K = 1 for all
XER X is (KXtRx.
In view of our definitions in Chap. IX-5 and in Chap. XII-2, (X,Y).,K
is the same as '1( {x, y} n), where '1 is as defined in corollary 2 of tho 1,
Chap. XII-2; our first assertion is then nothing else than formula (12)
of Chap. IX-5. The second one is identical with prop. 9 of Chap. XII-3
if K is a p-field; it is trivial if K = C; it can be verified at once if K = R,
since in that case our assumption, about the n-th roots of 1 being in K,
implies that n = 2. As to the last assertion, the assumption modK(n) = 1
implies that K is a p-field, with p prime to n. In view of our first formula,
and of prop. 6 of Chap. XII-2, our assertion amounts to saying that
Xn,y is unramified if and only if Y is in (K x)n R x. Call q the module of K;
our assumption about the n-th roots of 1 implies that n divides q -1. In
an algebraic closure K of K, take a primitive root, of 1 of order n(q -1).
For any f~ 1, let K f be the unramified extension of K of degree f,
contained in K; then ( is in K f if and only if n(q -1) divides qf -1, i. e.
if and only if 1 +q + ... + qf- 1 =0 (n); as q= 1 (n), this is so if and only
§ 5. The Hilbert symbol 261
if f=.O (n). This shows that K{O=K n. Put e=C n; as this is a primitive
(q-1)-th root of 1, it is in K. In view of the definitions of Chap. IX-5,
we have thus shown that Xn. is an unramified character of order n,
attached to Kn; therefore, by prop. 5 of Chap. XII-2, it generates the
group of the unramified characters of order dividing n. In particular,
for yeK X , Xn,y is unramified if and only if it is equal to (Xn,.)" for some
veZ, i.e. if ye- V is in the kernel of the morphism x-+Xn,x; as we have
seen in Chap. IX-5, that kernel is (K x)". Consequently Xn,y is unramified
if and only if y is in the subgroup of K x generated by (K x)n and e. By
prop. 8 of Chap. 11-3, (Kx)n contains 1 + P; as R x is generated by 1+ P
and e, our assertion is now obvious.
COROLLARY. For every local field K containing n distinct n-th roots

of 1, (x,y)n K defines a locally constant mapping of K x x K x into the
group of the n-th roots of 1 in C.
This is obvious if K =R or C; if K is a p-field, it is an immediate

consequence of proposition 6, and of the fact (contained in prop. 8 of
Chap. 11-3 if K is of characteristic p, since then n must be prime to p,
and otherwise in the corollary of prop. 9, Chap. 11-3) that (Kx)" is an
open subgroup of K X, of finite index in KX. The symbol (X,Y)n,K may be
said to determine a duality between the finite group K x/(K x)" and
itself, by means of which that group can be identified with its own dual.
PROPOSITION 7. Let k be an A:field containing n distinct n-th roots of 1.
Then, for all z=(zv), z' =(z~) in k~, almost all factors of the product
(z,z')n = n(zv,Z~)n,kv
v
taken over all the places v of k, are equal to 1; it defines a locally constant
mapping of k~ x k~ into the group of the n-th roots of 1 in C, and satisfies
(z,z')n=(z',z),;-l for all z, z'. Moreover, (k~)" is the set of the elements z
of k~ such that (z, z'}n = 1 for all z' e k~.
If k is of characteristic p> 1, our assumption about k implies that n
is prime to p; consequently, in all cases, we have Inlv= 1 for almost all v.
As Zv, z~ are in r~ for almost all v, our first assertion follows now at once
from prop. 6; the same facts, combined with the corollary of prop. 6,
show that (z, z')n is locally constant. By prop. 6, if z is in the kernel of all
the characters z-+(z,z')n, we must have zve (k~)" for all v; then, if we write
Zv = t~ with tvek~, the fact that Zv is in r~ for almost all v implies the same
for tv, so that t = (tv) is in k~ and that z = tn.
COROLLARY 1. For every finite set P of places of k, containing all the
places v for which Inl v'" 1, put
262 Global c1assfield theory XIII
Q(P)= n k: x n r:,
VEP v,p
Q'(P) = n
VEP
(k~)n x nr;.
v+P
Then these are open subgroups oj k~, and the set oj the elements Z oj k~
such that (z,z')n= 1 Jor all z'EQ(P) (resp. Jor all z'EQ'(P)) is (k~)n Q'(P)
(resp. (k~)n Q(P)).
Concerning the definition of P, one should observe that Inlv> 1 for
every infinite place of k, so that P contains all these places. Then Q(P)
is the same as the open subgroup of k~ which was so denoted in Chap.
IV -4; as we have seen above, (k~)" is open in k; for all v, so that Q' (P)
is open in Q(P). The first set considered in our corollary consists of the
ideles (zv) such that (zv,Z~)n,kv = 1 for all Zv E k~ if VEP, and for all z~Er~
if v is not in P. Our assertion follows now at once from prop. 6. The other
set can be treated in the same manner.
COROLLARY 2. Let P be as in corollary 1, and assume also that kA =

=kxQ(P). Then (k7=kx n(kDnQ'(p).
In this last relation, (k x)n is clearly contained in the right-hand side.
Conversely, let ~ be an element of this right-hand side. Then, by corollary
1, (~,z)n= 1 for all zEQ(P); by definition, this is the same as to say that
Q(P) is in the kernel of the character z-+(Xn,~,zh of k~. As that kernel
contains e, by the corollary of tho 2, § 3, and as k~ = kXQ(P), this
implies that Xn,~ is trivial, hence that ~E(p)n.
The symbol (z,z')n defined in prop. 7 may be called the Hilbert symbol
for k. As the last assertion of prop. 7 implies that (k~)" is a closed sub-
group of k~, the main content of that proposition may be expressed by
saying that the Hilbert symbol determines a duality between the group
k~/(k~)" and itself, by means of which it can be identified with its own
dual. As observed above, we have, for ~Ek x, zEk~:
(~,z)n = (Xn,~, zh = Xn, ~(a(z))
and therefore, by the corollary of tho 2, § 3, (~, rOn = 1 for all ~, 1'/ in P.
PROPOSITION 8. Let k contain n distinct n-th roots oj 1. Then k X(k~)"

is the set oj the elements z oj k~ such that (~,z)n= 1 Jor all ~Ekx, and it
* contains the kernel Uk oj a.
Call Xn the set in question; it may also be described as the intersection
of the kernels of the characters Xn. ~o a of k~ for all ~ E k x; clearly it con-
tains Uk' As before, put Gk=k~/kx; applying lemma 2 of Chap. XII-1
to Gk and to the endomorphism x -+ xn of Gk , we see that k x (k~)" is
a closed subgroup of k~ with compact factor-group. Applying lemma 7
to Gk> and to the group of the characters of Gk determined by characters
§ 5. The Hilbert symbol 263
of k~ of the form X".~oQ with ~ekx, we see that every character of k~,
trivial on X"' is of that form. Clearly X n contains k x (k~)"; as they are
both closed in k~, our proposition will be proved if we show that there
are arbitrarily small neighborhoods U of 1 in k~ such that X n is con-
tained in k x (k~t U; we will choose U as follows. Let Po be a finite set of
places of k, containing all the places v where Inlv =1= 1, and satisfying the
condition in the corollary of tho 7, Chap. IV-4, i. e. such that k~ = kXD(P 0);
then every finite set of places P::::> Po has these same properties. Take any
such set P; take U = n Uv' where Uv is an arbitrary neighborhood of 1 in
(k~)" for ve P, and U v = r~ for v not in P; clearly U is a neighborhood
of 1 in k~ and can be made arbitrarily small by suitable choices of P
and the neighborhoods Uv for veP. One sees at once that (k1t U is the
same as (k: t Q'(P), where D'(P) is as defined in corollary 1 of prop. 7.
What we have to prove is that X" is contained in the group W(P)=
kX(k:tD'(P), or in other words that Xn W(P) = W(P). By lemma 1 of
Chap. XII-I, applied to Gk=k~/kx and to the image ofW(P) in Gk, we
see that W(P) has a finite index in k~; it will thus be enough to show that
W(P) and Xn W(P) have the same index in k~.
The index of X" W(P) in k~ is equal to the number of distinct
characters of k~, trivial on Xn and on W(P). Being trivial on X n, such a
character must be of the form Xn.~OQ with ~ee. As Xn contains k X (k:)",
this is trivial on W (P) if and only if it is trivial on D'(P), hence, by corol-
lary 1 of prop. 7, ifand only if ~ is in (k:)" D(P). In view of our assumptions
on P, we can write
(k~)" D(P) = (k x D(P»)" D(P) = (k? D(P).
As in Chap. IV-4, put E(p)=e nD(P); we see now that the characters
in question are those of the form X".~OQ with ee(kXtE(P), and we must
compute the number of distinct ones among these, which is the index in
(kx)nE(p) of the kernel of the morphism ~-+Xn.~OQ. That kernel is the
same as that of ~ -+ Xn.~' which is (k x)n; consequently that index is the
same as that of E(P)" in E(P). In view of tho 9 of Chap. IV-4, and ofthe
fact that n divides the order of the group of the roots of 1 in k, that index
is n with c = card(P).
C
Now we have to compute the index of W(P) in k~. Consider the

groups G = k x x D(P), G' = k x x D' (P), and the morphism J of G into k~
given by J(~,u)=~u for ~ekX, ueD(P). Call H the kernel of J; this
consists of the elements (~,C 1) of G with eeE(P). In view of our assump-
t ion on P, J maps G onto k~; it maps G' onto W (P), as appears from the
formula
This gives f-1(W(P))=HG', and therefore:

[k~: W(P)] = [G:HG'] = [G:G']·[HG':G']-l.
Here [G: G'] is given by
[G:G']= [D(P):D'(P)] = n
VEP
[k~:(k~)"].
In the right-hand side, each factor corresponding to an imaginary place v

is equal to 1, which can also be written as n2Inl;;-1, since in that case
Inl v=n 2 • If v is real, n must be 2, since k,,=R must contain a primitive
n-th root of 1; then the corresponding factor is 2, which can again be
written as n2Inl;;-1. The factors corresponding to the finite places VEP
are given by the corollary of prop. 9, Chap. 11-3, if k is of characteristic 0,
and by prop. 8 of Chap. 11-3 otherwise; here one has to take into account
the fact that n divides the order of the group of roots of 1 in k, hence
also in kv, and that consequently it is prime to p if k is of characteristic p.
Then one sees that the factors in question are again respectively equal
to n2Inl;;-1. This gives
[G:G'] = n (n2Inl;;-1)=n2cnlnl;;-1 =n 2c
VEP
since Inlv= 1 for all places v not in P.

Our proof will now be complete if we show that [H G' : G'] = nCo This
is the same as the index of H nG' in H, or, in view of the definition of H
and G', as that of E(P)nD'(P) in E(P). By corollary 2 of prop. 7,
E(P)nD'(P) is contained in E(P)n(k X )n, i. e. in E(p)n, and it is obvious
that it contains E(P)". Therefore the index is question is that of E(p)n
in E(P); we have already found above that this is nC ; this completes
our proof.
§ 6. The Brauer group of an A-field. In § 3, we have seen that a class

of simple algebras A over k is uniquely determined by its local invariants
hv(A), with hv(A) = 1 for almost all v, hv(A) = 1 for all imaginary places,
and hv(A) = 1 or 2 for all real places; and we have proved that n hv(A) = 1.
Therefore the Brauer group H(k) will be known if we prove the following:
THEOREM 4. Let k be an A-field. For each place v of k, let '1v be a root
of 1 in C. Assume that '1v= 1 for almost all v, '1v= 1 for every imaginary
n
place v, '1 v= 1 or 2 for every real place v, and that '1 v= 1. Then there
v
is a simple algebra A over k with the invariants hv (A)='1v.
The proof of this, for a field k of characteristic 0, will be postponed
until the end of this §; we proceed to prove it now for a field k of charac-
teristic p> 1. As in Chap. VI, write D(k) for the group of divisors of k,
§ 6. The Brauer group of an A-field 265
Do(k) for the group of divisors of degree 0, and P(k) for the group of
principal divisors; call h the number of divisor-classes of degree 0, i. e.
the index of P(k) in Do(k). Let VI"'" VN be all the places of k where '1v -+ 1;
taking for n an integer ~ 1 such that ('1vl = 1 for all i, we can write
'1v;=e(aJn) with aieZ for l~i~N. As n'1v=l, we have Lai=na with
n
aeZ; after replacing a l by a l -na, we may assume that Lai=O. For
each i, call di the degree of the place Vi; put d = di and m = L(aid/di)Vi'
Then deg(m)= Laid=O, m is in Do(k), hence hm in P(k), so that there
is Oek x such that div(O)=hm, i.e. ordv;(O)=haid/di for 1 ~i~N, and
ordv(O)=O when '1v= 1. Now consider the constant-field extension k'
of k of degree hnd; let cp be the Frobenius automorphism of k' over k,
and X the character of the Galois group of k' over k such that X(CP)=
=e(l/hnd). Just as in the proof of prop. 3 of § 1, one sees at once, by
applying corollary 4 of tho 1, Chap. XIl-2, that, if V is any place of k,
and () its degree, (Xv, O)v = X( cp"t with v = ord v(0). In view of our choice
of 0 and X, this gives (Xv, O)v = '1v for all V. Therefore, by formula (4) of
§ 3, the cyclic algebra A = [k'/k; X,O] solves our problem.
PROPOSITION 9. For each Xe X k> call U(X) the kernel of the character
xoa of k~. Then: (a) if k' is the cyclic extension of k attached to X, U(x) =
=k XNk'/k(k~X); (b) for every n~ 1, prime to p if k is of characteristic p> 1,
the intersection Un of the kernels U(x), for all the characters XeXk of
order dividing n, is k (k~)n.
X
Put U'(X)=k x Nk'lk(k~X), when X and k' are as above, and put
U~ = k X (k1 In. Applying lemma 2 of Chap. XII-l to the endomorphism
x - xn of the group Gk= k~/k x, we see that U~ is closed in k~; applying
the same lemma to the morphism of Gk , into Gk determined by the
morphism N k'/k of k~x into kA' we see that U' (X) is also closed in kA.
If X is of order n, and k' is as in (a), n is the degree of k' over k, so that
Nk'/k(Z)=Zn for zek, hence also for zekA ; therefore we have then
U'(X)~ U~. A character of k~ is trivial on (k~)n if and only if its order
divides n; therefore a character of k~, trivial on k X , has an order divid-
ing n if and only if it is trivial on U~. As before, call Uk the kernel of a;
we know that it contains kX. If k is of characteristic 0, apply prop. 1 of
Chap. XII-l to the pairing between X k and Gk determined by (X,Z)k;
otherwise apply corollary 4 of prop. 2, Chap. XII-2; in both cases we see
that every character of kA of finite order, trivial on Uk' can be uniquely
written as xoa with XeXk' This implies that Un is the intersection of the
characters of k A; trivial on U~ and on Uk; therefore it is the closure of
U~Uk' and we have Un= U~ if and only if U~~ Uk' We also see now that
every character of kA' trivial on U'(X) and on Uk, must be of the form
x'oa with x'eX k ; by corollary 2 of th.l, § 1, this is trivial on U'(X) if
and only if the cyclic extension of k attached to x' is contained in k',

i.e. if and only if x' = XV with some veZ; as the intersection of the kernels
U(XV) for veZ is obviously U(X), this shows that U(x) is the closure
of U'(X) U", and that we have U(X)= U'(X) if and only if U'(X)::::> Uk. Now
consider first the case of characteristic o. Proceeding by induction on n,
we assume that, for all fields k of characteristic 0, (a) holds for every X
of order <no Then U"ckxN",/,,(k'n whenever k is such a field and k'
is a cyclic extension of k of degree <n; by formula (5) at the end of
§ 3, this implies U"~kxN",/,,(U,,,). Let L be the extension of k generated
by a primitive n-th root of 1; as this is abelian of degree < n over k,
we can find a sequence ko = L, k1 , ••• , Ie, = k of fields between Land k,
such that, for 1:::;; i:::;; r, kj _ 1 is cyclic of degree < n over kj ; therefore we
have, for 1 :::;;i:::;;r:
By induction on i for 1 :::;;i:::;;r, we see now at once that U"i is contained

in kt«ki)~)"' since this is so for i=O by prop. 8. For i=r, we get U"c U~;
as we have seen above, this proves (b), and it implies U"c U'(X) for
every X of order n, which proves (a) for such characters and completes
the induction. Now let k be of charact~ristic p> 1. Take Xand k' a~ in (a);
take ZEU(x), so that (x,Z)" = 1; then, lfwe put z=(zv) and '1v=(xv,zv)v,
these satisfy the conditions in tho 4. As tho 4 has been proved for charac-
teristic p> 1, we conclude that there is a simple algebra A over k with
the invariants hv(A) = '1v; as this is condition (iv) in prop. 5 of § 3, we can
apply that proposition, the last assertion in which shows now that
ZE U'(X). This proves (a). Just as above, we can now conclude, by apply-
ing formula (5) at the end of § 3, that U" = k X N m(U",) for all cyclic
extensions k' of k. Assuming that n is prime to p, take for k' the constant-
field extension of k generated by a primitive n-th root of 1. Then prop. 8
gives U",ck'X(kAX)"; therefore the same is true for k.
We can now prove theorem 4 in the case of characteristic o. For
every place v of k, call v(v) the order of'1v in ex. One can construct a
character X such that, for every v, the order of Xv is a multiple of v(v);
for instance, this will be so if we take for X a character attached to the
cyclic extension k' of k described in lemma 5 of § 3. Then, for every v,
z-+(Xv,z)v is a character of k; whose order, being equal to that of Xv,
is a mUltiple of v(v); therefore we can choose ZvEk; so that (Xv,zv)v='1v.
If, in doing so, we take Zv= 1 whenever '1v= 1, z=(zv) is in k~, and the
n
assumption '1v= 1 implies that Z is in the kernel of XOQ; therefore, by
prop. 9, it is in k XN",/,,(kA
X
), k' being the cyclic extension of k attached
to x. Writing z=(JN",/,,(z') with z'ekAX, one sees at once, by combining

prop. 10 of Chap. IX-4 with corollary 3 of th.l, Chap. IV-I, and with
§ 7. The Hilbert p-symbol 267
prop. 1 of § 1, that the cyclic algebra A = [k'/k; X,O] has the required
local invariants hv(A) = "v, so that it solves our problem.
§ 7. The Hilbert p-symbol. By now, as will be seen in the next §, our

investigation is essentially complete, so far as only algebraic number-
fields are concerned. For the case of characteristic p> 1, we still need
a symbol, similar to the Hilbert symbol studied in § 5 but based on the
factor-classes g,O}p of Chap. IX-5.
In any field K of characteristic p> 1, we will denote by iP the endo-
morphism x --+ x - x P of the additive group of K; its kernel is the prime
field F;,. We begin by considering a local p-field K of characteristic p;
as usual, we write R for its maximal compact subring, P for the maximal
ideal of R, and q for the module of K. Obviously, iP maps R into R,
Pinto P, and, if ord(x)=v<O, we have ord(iP(x))=pv<O, so that
iP- 1 (R)=R.
PROPOSITION 10. Let K, R, P and iP be as above; for x e K, Z e K x,

put (X,Z)p,K=(Xp,x,Z)K' Then iP(K) contains P and not R; the set of the
elements x of K, such that (X,Z)p,K= 1 for all zeK x (resp. for all zeRX)
is iP(K) (resp, R + iP(K)); the set of the elements Z of K such that
X
(X,Z)p,K = 1 for all xe K (resp. for all xeR) is (K X)P (resp. (K X)P R X).
co
For xeP, put 'l'(x)= L xpn; clearly this is convergent and defines
n=O
an endomorphism of P, and one sees at once that both iP 0 'l' and 'l' 0 iP
induce the identity on P. Therefore iP induces on P an automorphism
of the additive group of P, so that Pc iP(K). Call F the algebraic closure
of the prime field Fp in K; by th.7 of Chap. 1-4, F is a field with q ele-
ments, and R = F + P, so that ([>(R) = ([>(F) + P. As the endomorphism
induced by iP on the finite field F has the kernel F;" it is not surjective;
therefore iP (R) =1= R; as iP - 1 (R) = R, this shows that R is not contained
iniP(K). If (x,z)p,K=l for all zeKX, Xp,x must be trivial; as we have
seen in Chap. IX-5, this is so ifand only ifxeiP(K). Take any xeR+iP(K);
as R=F+ P and PciP(K), we can write x=a+iP(u) with aeF, ueK;
then Xp,x is the same as Xp,a and is a character attached to the cyclic
extension of K generated by any root Q( of X -XP=a. As Q( is algebraic
over F, it is 0 or a root of 1 of order prime to p, so that K(Q(), hence also
Xp,x, are unramified over K; therefore (X,Z)p,K= 1 for all zeRx .. Now
take a root ( of X - xq = 1 in some algebraic closure of K, and put
e = ( - (P. The Galois group of K«() over K is generated by the Frobenius
automorphism; as ( is algebraic over F, this maps ( onto (q = (-1, so
that it leaves e invariant; therefore e is in F, K«() is the unramified ex-
tension of K of degree p, and Xp ,. is a character attached to that extension.
Then the unramified characters over K, of order p or 1, are those of the

form (X p, £)" with vE Z; consequently a character Xp , x is unramified if and
only if it can be so written, i.e. if and only if x=ve+4>(u) with uEK;
then x is in R+ 4> (K). As to the last two assertions, one is nothing else
than prop. 10 of Chap. XII-3. Finally, if e is as above, the kernel of
z-(e,z)p,K is the subgroup of K X of index p containing RX; this is
(KX)P RX; for every xER, we have seen that Xp,x is unramified, and it
is of order p or 1, so that the kernel of z-(X,Z)p,K contains (KX)P RX.
This completes our proof.
COROLLARY. For each integer m~O, call Q'(m,K) the set of the ele-
ments Z of K X such that (x, z)p,K = 1 for all XEP-m. Then this is an open
subgroup of K X, containing (KX)P; its index in K X is p.qm-m' if m' is the
largest integer ~m/p. For every neighborhood U of 1 in K X, there is
m~O such that Q'(m,K)c(KX)PU. Moreover, Q'(O,K)=(KX)PR X; and,
for every m~O, the set of the elements x of K such that (x, z)p,K = 1 for
all zEQ'(m,K) is p-m+4>(K).
Let the finite field F be as above, and let 1[ be a prime element of K;
by tho 8 of Chap. 1-4, K may be identified with the field offormal power-
series in 1[ with coefficients in F; therefore, if we call Vm the space of
polynomials of degree ~m in 1[-1 with coefficients in F, p-m is the direct
sum of Vm and P, and Vm is a vector-space of dimension m + 1 over F;
moreover, one verifies at once that Vmn4>(K)=4>(Vm')' with m' as in our
corollary. By prop. 10, Q'(m,K) is the intersection of the kernels of the
characters z-(X,Z)p,K of K X for XE Vm; therefore it is open and contains
(Kx)P, and, by lemma 7 of § 5, all the characters of K X
trivial on
,
Q'(m,K), are of that form. This implies that the index of Q'(m,K) in K X
is equal to the number of distinct characters of that form, which is the
index of Vmn4>(K) in Vm; as Vm, Vm, have respectively qm+l and qm'+l
elements, and as the morphism 4> of Vm, onto Vmn 4>(K) has the kernel Fp,
that index is p.qm-m'. By lemma 2 of Chap. XII-I, and lemma 7 of § 5,
the group G' =K x /(KX)P is compact, and its characters are those deter-
mined by the characters z-(x,z)p,Kof K X for xEK. Therefore the inter-
sections of the kernels of finitely many such characters make up a funda-
mental system of neighborhoods of 1 in G'. This is the same as to say
that, if U is any neighborhood of 1 in K x, one can find finitely many
characters z-(xj,Z)p,K such that the intersection W of their kernels is
contained in (K X)P U; then W contains Q'(m,K) if we take m ~ 0 such that
-m~ord(xj) for all i. By prop. 10, Q'(O,K)=(KX)PR x • Finally, if
z-(Y,Z)p,K is trivial on Q'(m,K), it must coincide with a character
z-(X,Z)p,K with some XE Vm; by prop. 10, this is the same as to say
that YEVm+4>(K); as 4>(K)~P, and Vm+P=P-"', this proves the last
assertion in our corollary.
§ 7. The Hilbert p-symbol 269
From now on, in this §, k will be an A-field of characteristic p. We

will need the following lemma:
LEMMA 8. If v is any place of k,kn(kv)P = kP.
Clearly kn(kv)P is a field between k and k P; by lemma 1 of Chap.
VIII-5, it must be k or kP. As (kv)P contains no prime element of kv, it
is not dense in kv; as k is dense in kv,(kv)P cannot contain k.
PROPOSITION 11. For all x=(x v) in kA> and all z=(zv) in k1, almost
all the factors of the product
(x,z)p= n(xv,ZV)P,kv
v
are equal to 1; it defines a locally constant mapping of kA x kl into the
group of the p-th roots of 1 in C; the set of the elements x of kA (resp.
of the elements Z of k1) such that (x,z)p= 1 for all zek1 (resp. for all
xekA ) is tP(kA) (resp. (k~)P).
All this follows at once from prop. 10.
COROLLARY 1. For every divisor m= Lm(v)'v»O of k, put
Q'(m) = nQ'(m(v),kv)'
v
Then this is an open subgroup of k1, containing (kl)P. For every neigh-
borhood U of 1 in kl, there is a divisor m such that Q'(m)c(kl)PU. The
set of the elements x of kA such that (x,z)p= 1 for all zeQ'(m) is
(np;m(v») + tP(kA)'
v
For all v, Q'(m(v),k v) is an open subgroup of k:, containing (k~ )P,

and, for almost all v, m(v)=O, so that Q'(m(v),k v) contains r:; this proves
the first assertion. In the second assertion, it is enough to consider a
n
neighborhood U = Uv, where Uv is a neighborhood of 1 in k; for all v,
and Uv=r: for almost all v; then our assertion follows at once from
the corollary of prop. 10. Assume that x = (xv) is as in the last assertion;
then, by the same corollary, we can write xv=Yv+tP(uv) with Yvep-;,m(v),
uvekv for all v, and Yv=X v, uv=O whenever m(v)=O and xver v, hence
for almost all v. Then y=(Yv) and u=(u v) are in kA' and x=y+tP(u),
yenp;m(v).
COROLLARY 2. Notations being as in corollary 1, assume also that
deg(m»2g-2. Then (kX)P=k x nQ'(m).
Clearly the right-hand side of this last formula contains (kx)P. Now
take eekx; if eeQ'(m), we have (x,e)p= 1 for all xe np;m(v) and also
for all xek, hence for all xekAby corollary 3 ofth. 2, Chap. VI. By pro-
position 11 this implies ee(k~)P, hence ee(kx)p by lemma 8.
COROLLARY 3. Notations being as in corollary 1, assume that

n= Ln(v)'v is a divisor of k, of degree >2g-2, such that m>-pn. Then
the set of the elements x of kA such that (x,z)p= 1 for all ze.Q'(m) is
(np;m(v») + lP(k).
v
Again by corollary 3 of tho 2, Chap. VI, we can write
lP(kA) = lP(k)+ lP(np;n(v») .
v
Our assumptions imply that the second term in the right-hand side is
a subgroup of np;m(v). Our assertion follows now from the last one in
corollary 1.
PROPOSITION 12. The set of the elements z of k~ such that (~,z)p= 1

for all ~ek is k X (k~)P.
Call that set X p; it is the intersection of the kernels of the charac-
ters Xp,~OQ of k~ for all ~ek; it contains k X (k~)p. By lemma 2 of Chap.
XII-1, applied to Gk = k~/kX, k X (k~)P is a closed subgroup of k~ with
compact factor-group; by lemma 7 of § 5, every character of k~, trivial
on X P' must be of the form Xp , ~o Q with ~ e k. Choose a divisor n>- 0 of k,
of degree > 2g - 2. In view of corollary 1 of prop. 11, it will be enough
for us to show that Xp is contained in W(m)=kx .Q'(m) for all divisors
m= Lm(v)·v>-pn. By lemma 1 of Chap. XII-1, W(m) is of finite index
in k~; therefore it will be enough to show that Xp W(m) and W(m) have
the same index in k~.
The index N of X p W(m) is equal to the number of distinct characters
of k~ of the form Xp,~oQ, or, what amounts to the same, of the form
z-+(~,z)p, with ~ek, which are trivial on .Q'(m). By corollary 3 of prop. 11,
the latter character is trivial on .Q'(m) if and only if ~e U(m)+IP(k), with
U(m) = np;m(v). As in Chap. VI, put A (m)=knU(m). Then we see that N
is the index of lP(k) in A(m) + lP(k), or, what amounts to the same, of
A(m)nlP(k) in A(m). Put m' = Lm'(v)'v, where m'(v), for each v, is the
largest integer ~m(v)/p; callm, m', n the degrees ofm, m', n,respectively;
then m~m'~n>2g-2. Clearly A(m)nlP(k) is the same as IP(A(m'));
corollary 2 of tho 2, Chap. VI, shows that A (m), A (m') are vector~spaces,
of dimension m - g + 1 and m' - g + 1 respectively, over the field of con-
stants F of k; as IP maps A(m') onto A(m) nlP(k) with the kernel Fp, we
see that the latter group has p-lqm'-g+l elements while A(m) has
qm- g +l elements. This gives N=p·qm-m'.
Now we have to compute the index of W(m) in k~. Take a finite
set P of places of k, containing all the places v for which m(v»O, and
satisfying the condition in the corollary of tho 7, Chap. IV-4, i.e. such
that k~ =kx .Q(P). Put:
Q" = nQ'(m(v),kv) vfPnr;.

veP
X
Clearly this contains Q(P)1', and we have Q'(m)=(k~)PQ", hence:

W(m)=k x (k~)PQ" =e (e Q(P})PQ" =k x Q".
Put now G=k x x Q(P), G' =kx x Q"; call f the morphism of G into k~
given by f(e,u)=eu for eEk x, uEQ(P), and call H the kernel of f. Then f
maps G onto k~, G' onto W(m), and H consists of the elements 1) (e,e-
e
with in E(p)=e nQ(P). We have now:
[k~: W(m)] = [G:H G'] = [G: G']' [H G' :G']-I.
Here, in view of the corollary of prop. 10, [G:G'] is given by
[G:G']=[Q(P):Q"]= n [k~ :Q'(m(v),kv)]=pcqm-m'
veP
with c=card(P). Finally, [HG':G'] is the same as the index of HnG' in
H, i.e. as that of E(P)nQ" in E(P). Clearly E(P)nQ" is contained in
k X nQ'(m), which is (kX)P by corollary 2 of prop. 11, and it contains
E(P)P; as E(P)n(kX)P is the same as E(P)P, we see that E(P)nQ" is
E(P)P, and it follows at once from tho 9 of Chap. IV-4 that its index in
E(P) is pc-I. This completes the proof.
COROLLARY. If k is as above, and U" is the kernel of the canonical
morphism a, we have U" c k x (U ,,)P.
By proposition 12, U" is contained in k X (k~)P, so that, if u is any
element of Uk' it can be written as u=ev p with eEkx, vEk~. Take any
XEX,,; call k' the cyclic extension of k attached to X. By prop. 9 of §6, the
kernel U (X) of XO a is k x N It'/,,(k~X ); as U" c U(X), this implies, by formula
(5) at the end of § 3, that U,,=k x N",/"(U,,,). Again by proposition 12, U",
is contained in k' x (k~x)1', so that U" is contained in k x N "'/,,(k~x)1'; there-
fore, ifu is as above, we can write u="N"'/"(w)P with "Ek x, wEk~x. This
gives e"-I=v- PN,,,/,,(w)p. As e,,-l is in k X and in (k~)P, it is in (kX)P,
by lemma 8; writing it as (P with (Ek X , we get v=C 1 N",/,,(w), since p
is the characteristic. This shows that v is in U (X); as this is so for all
XEX", it is in U", which completes the proof.
§ 8. The kernel of the canonical morphism. We are now able to deter-

mine U" in all cases.
THEOREM 5. Let k be an A~field, and a the canonical morphism of k~
into the Galois group m: of kab over k. Then X--+ xoa is a bijective morphism
of the group X" of characters Qf m: onto the group of the characters of k~
of finite order, trivial on k x •
Every character of k~ of order n, trivial on k X, is trivial on kX(k~)n,

hence on Uk, by prop. 9 of § 6, if k is of characteristic 0; in that case, our
conclusion follows from this at once by applying prop. 1 of Chap. XII-1
to Gk=k~/kx. Now let k be of characteristic p> 1, and letw be a character
of k~ of order n, trivial on kX. Write n = n' pi with n' prime to p and i ~ 0;
taking integers a, b such that n' a + pi b = 1, we have w = w' w" with w' = W pib
of order n', and OJ" = wn'Q of order pi, both being trivial on k x • Just as
above, we conclude from prop. 9 of § 6 that w' is trivial on Uk' On the
other hand, one concludes at once from the corollary of prop. 12 of § 7,
by induction on i, that Uk is contained in kX(UJP\ hence in k X(k~)P"
and then, just as above, that OJ" is trivial on Uk' This shows that w is
trivial on Uk; our conclusion follows from this at once by applying
corollary 4 of prop. 2, Chap. XII-1, to Gk=k~/kx.
COROLLARY 1. The kernel Uk of a is the intersection of the closed

subgroups kX(k~r of k~ for all n~ 1.
In the proof of prop. 9, § 6, we have already seen that these are closed
subgroups; clearly, then, k X(k~)n is the intersection of the kernels of
all the characters of k~, trivial on k x , whose order divides n. Our assertion
follows now at once from theorem 5.
COROLLARY 2. If k is of characteristic p> 1, Uk = k x •
Write Gk = G~ x N, with G~ = kl/k and N isomorphic to Z. As G~

is compact, and as it is obvious that it is totally disconnected, lemma 4 of
Chap. VII-3 shows that all its characters are of rmite order; every such
character can be uniquely extended to one of Gk> trivial on N, which is
then also of finite order, hence, by theorem 5, trivial on the image of
Uk in Gk • As corollary 2 of prop. 2,Chap. XII-1, shows that this image is
contained in GL it must therefore be {1}, which is the same as to say
that U,,=k x .
COROLLARY 3. If k is of characteristic 0, U" is the closure of k Xk~ +
in k~, k~ + being the group of the ideles (zv) such that Zv > 0 for all real
places and Zv = 1 for all finite places v of k.
Write Gk= G~ x N, with G~ = kl/k X, N being the image in Gk of the
group M defined in corollary 2 of tho 5, Chap. IV-4. Call U' the closure of
e k~ + in k1, U" its image in Gk> and put U~ = U" n G~. Obviously
k~/k~+ is totally disconnected, so that the same is true of k~/U', hence
of GJU". As M is contained in U', N is contained in U", so that U" =
= U~ x N and that GJU" may be identified with GVU~ and therefore
is compact. This shows that every character of Gk> trivial on ut, or, what
amounts to the same, every character of k~, trivial on U', is of rmite
order. Consequently Uk is contained in U'; as we already knew that it

contains U', we see that it is U'.
In order to obtain more precise results about Uk in the case of

characteristic 0, one needs an algebraic lemma:
LEMMA 9. Let I be a rational prime, K a field, not of characteristic I,
and K an algebraic closure of K. For each n ~ 0, call Kn the extension of K
generated by a primitive l"-th root of 1 in K. Then, if I =/=2, or if 1=2 and
K=K 2, KXn (K:),"=(K x )," for all n. If 1=2, K=/=K 2, 2~m<n and
K2 =/=Km+ 1, then K x n (K; )2" c (K x )2"-m.
Take aEK x n(K~t; assume at first that a is not in (K~),", and *
let i be the smallest integer such that a is in (K~+ d" and not in (K7)'".
Then l~i<n, K i =/=K i+1 , and we can write a=x'" with x in K i+1 and
not in K i • Call 1] a primitive li+l_th root of 1 in K i + 1 , and put 6=1]',
(=1],i; 6, ( are roots of 1 of order Ii and I, respectively, and are in K i.
We have K i+1 = K i(1]), 1]'=6, and 6 is in Ki and is not 1; therefore K i+1
is cyclic of degree lover K i , with a Galois group generated by the auto-
morphism (J given by 1]"=(1]. Put (}=x"x- 1 ; then (}EK i+1 and (}'"=1,
so that () is a root of 1 of some order l" dividing In; therefore ()" must be
of the form (}S with s E Z; moreover, if v ~ i, () is in K i , so that we can take
s = 1, while, if v> i, we have 1] = ()' with some rE Z, hence 1]" = 1]s, (= 1]s-1,
and therefore s = 1 + Ii (Ii + 1). By induction on h, one sees at once that
x"" = x(}1+s+"'+S"- '; for h= I, this gives 1 + s + ... + S'-1 =0 (1"). If v~ i,
we have s = 1, so that the latter congruence implies v ~ 1. If v > i, we have
s = 1+ a Ii with a = 1 (I); if I =/= 2, or if I = 2 and i ~ 2, this implies s' = 1 + b Ii + 1
=
with b 1 (I), which shows that (s' - 1)/(s - 1) cannot be a multiple of 12 ,
hence also not of l"; as this contradicts what we have found above, we
conclude that v ~ 1 except possibly for 1=2, i = 1. Therefore, except in
that case, we can write (}=C with tEZ; writing then x'=1]-t x , we have
x'" =x', so that x' is in K i, and a=x"", which contradicts the definition
of i. This proves that a is in (K~)'" if 1=/=2, and, for 1=2, it proves that,
if it is not in (K~ )2", it is in (K; )2". In the former case, write a = y'" with
yE K I' As the Galois group of K lover K is a subgroup of (Z/I Z) x • the
degree d of K lover K divides 1- 1 and is prime to I; write 1 = de + Inf; we
have ad = b'" with b = N KdK(Y), hence a = (af be)'". If 1=2, we have K 1 = K;
if then a is not in (K x )2", we must have K =/= K 2, and we can apply what
we have found above, with i = 1; if at the same time K 2 =/= Km + l' the order
2" of () cannot be a multiple of 2m + I, so that it divides 2m ; putting then
b=x 2m , we have bet =b, so that b is in KI =K, and, if n>m, a=b 2"-m.
In the case 1= 2, K =/= K 2, one can easily show, by using similar arguments,
that (K x )2" is a subgroup of index 2 of K x n (K: )2", the latter group
being generated by (K x )2" and (1 + ro)2" if ro is a generator of the group
274 Global cIassfield theory XIII
of the roots u of 1 in K 2 , of order dividing 2", such that N K,/K(U) = 1.

These facts will not be used here.
PROPOSITION 13. Let P be a finite set of places of k, containing P00;

put H = nk:. Then U"nH is k;' + if k is of characteristic 0, and {1}
veP
if it is of characteristic p> 1.
The latter assertion is obvious, since in that case U,,=k x ; we may
therefore assume that k is of characteristic O. By corollary 1 of tho 5,
U"nH is the same as the intersection of the groups e(k~tnH for all
N ~ 1. An element of k~ belongs to the latter group if and only if it can be
written as ez N with eek x, z=(zv)ek~ and e=z;;N for all v not in P.
Take N = 1", where I is a rational prime; let k' be the extension of k
generated by a primitive N-th root of 1 in k, and k" the extension of k'
e
generated by any root of XN = in k. Clearly, for all places w of k' which
do not lie above a place veP, we have ee(k~xt; therefore, by corollary 4
ofth. 2, Chap. VII-5, we have k" =k', so that e is in (k'X)~ By lemma 9,
this implies ee(kX)N if 1=/=2. If 1=2, call k2 the extension of k generated
t
by a primitive 4-th root of 1; if k2 = k, we have again e e(k x If k2 =/= k,
call 2m the highest power of 2 dividing the order of the group of roots of 1
in k2 ; then lemma 9 gives ee(kx)N' with N'=2- m N, provided n>m.
TakingN=2m +" in the latter case, and otherwise N=I", we see that
kX (k~l nH is then contained in (k~r nH, which is the same as H'P.
This shows that U"nH is contained in H'P for all primes I and all jl>O.
Take any integer M> 1 ; for every prime I dividing M, let I" be the highest
power of I dividing M; we can find integers a(l) such that 11M = l:l-"a(l).
Take any he U"nH, and, for each I, write h=(hl)'P. with h,eH; then h=h,M
with h'= n(h,)a('); therefore U"nHcH M for all M>1. In corollary 1
ofth. 3, Chap. XII-3, we have shown that the intersection of all the groups
(k:)M, for a given fmite place v of k, is {1}; this same intersection is
obviously ex if kv=C, and R~ if kv=R. Therefore the intersection of
all the groupsHM~S k;'+, so that U"nH is contained in k;'+: as it
obviously contains it, this completes our proof.
COROLLARY. For every place v of k, kv•ab is generated over kv by kab •
This is trivial if kv = C, and it is obvious if kv = R, since then kv ab is C
and is generated by a primitive 4-th root of 1 in k. Assume now' that v
is a finite place. The union kv.o of all unramified extensions of kv is
generated over kv by roots of 1; therefore, if k' is the subfield of kV• ab
generated over kv by kab' it contains k v.o. As in § 1, let mv be the Galois
group of kv •ab over kv, and let Pv be the restriction morphism ofmv into m.
An automorphism Q( of kV• ab over kv induces the identity on k' if and only
if it induces the identity on kab' i.e. if and only if Pv(Q() is the identity.
Assume that this is so; then, as kv,ock', a is in the Galois group of

kv,ab over kv, 0, so that, by corollary 2 ofth. 3, Chap. XII-3, it can be written
as a=av(z) with zEr~. Then, by prop. 2 of§1, we have pv(a)=aUv(z)),
where jv is the natural injection of k~ into k~; if pv(a) is the identity,
Mz) must be in Uk; taking for P, in proposition 13, a set containing v,
we see now that a itself must then be the identity. This proves our corollary.
As an example for the above corollary, we may apply it to the case
k = Q; then, in combination with corollary 3 of tho 3, § 4, it shows that,
for every rational prime p, the maximal abelian extension of Qp, in an
algebraic closure of Qp, is generated by all the roots of 1. This could
also, of course, have been derived directly from the results of Chap. XII.
§ 9. The main theorems. The main results of classfield theory are

either immediate consequences of those found above, or can be derived
from them by following exactly the proofs given for the corresponding
theorems in Chap. XII.
THEOREM 6. If k is of characteristic 0, the canonical morphism a
determines an isomorphism of k~/Uk onto the Galois group m: of kab
over k, Uk being the closure of e k ~ + in k~; if k is of characteristic
p> 1, a determines a bijective morphism of k~/kx onto a dense subgroup
of m:, and an isomorphism of kl!k x onto the Galois group m:o of kab over
the union ko of all constant-field extensions of k.
The first assertion merely repeats part of prop. 1, Chap. XII-1,
corollary 3 of tho 5, § 8, being taken into account. The other assertions
repeat part of corollary 2 of prop. 2, Chap. XII -1, and [II"] of Chap. XII -1,
taking into account the fact that Uk = k x and that m:o has been deter-
mined in § 1.
THEOREM 7. Let k' be an extension of k of finite degree, contained in
k; put L=k' nkab. Then, for zEk~, a(z) induces the identity on L if and
only if z is in k x N k'lk(k~X).
The proof is identical to that of tho 4, Chap. XII-3, except that of
course one must now make use ofth. 5 of§ 8, instead ofth. 3 of Chap. XII-3,
and corollary 1 of tho 1, § 1, instead of corollary 1 of tho 2, Chap. XII-2.
COROLLARY 1. Assumptions and notations being as in theorem 7, call
~ the subgroup of m: corresponding to L. Then:
e NLlk(L~)=e Nk'lk(k:n=a-l(~).
The latter equality is a restatement oftheorem 7. Applying theorem 7

to k'=L, we get e
NLlk(L~)=a-l(~).
COROLLARY 2. For every extension L of k of finite degree, contained

m
in kab' call ~(L) the subgroup of corresponding to L, and put N(L)=
= k X NL/k(L~). Then N(L)=a-l(~(L»); ~(L) is the closure of a(N(L»)
inm; L consists of the elements of kab' invariant under a(z) for all ZE N (L),
and a determines an isomorphism of k~/N(L) onto the Galois group
m/~(L) of Lover k. Moreover, L-+N(L) maps the subfields of kab' of finite
degree over k, bijectively onto the open subgroups of k~, of finite index
in k~ and containing e.
All this merely repeats prop. 3 of Chap. XII-l, the corollaries of
tho 5, § 8, being taken into account; one should notice here that, when k
is of characteristic 0, the group k~ +, being a product of finitely many
factors isomorphic to R ~ or to ex, is generated by every neighborhood
of 1 in that group, and is therefore contained in every open subgroup
ofk~.
COROLLARY 3. Notations being as in corollary 2, let r be the group

of the characters of m, trivial on ~ (L). Then the subgroup N(L) of k~
associated with L is the intersection of the kernels of the characters
w=xoa of k~ for xEr, and X-+xoa is an isomorphism of r onto the
group y of the characters of k~, trivial on N(L).
The first assertion is merely a restatement, in other terms, of the

equality N(L)=a-l(~(L»); similarly, the second one is a restatement of
the fact that a determines an isomorphism of k~/N(L) onto '21/~(L).
COROLLARY 4. Let X be any character of m; then, if L is the cyclic
extension of k attached to X, the subgroup N(L) associated with L is the
kernel of the character w = XO a of kA .
This is a special case of corollary 3, since here the group r of that

corollary is the one generated by X.
COROLLARY 5. Let k and k' be as in theorem 7; let M be a subfield of

kab' of finite degree over k, and call M' its compositum with k'. Let U =
=k x NM/,,(M~), U'=k x NMI/k.(M'D be the open subgroups of k~ and
of k~x associated with the abelian extensions M of k, and M' of k', respec-
tively, by corollary 2. Then U' = N;.MU).
The proof is identical to that of corollary 3 of tho 4, Chap. XII-3.
THEOREM 8. Let k' be an extension of k of finite degree, contained
in ksep ; let a, a' be the canonical morphisms of k~ into m, and of k~x into
the Galois group m' of k~b over k', respectively. Let t be the transfer homo-
morphism of minto m', and j the natural injection of kA into kAx, Then
toa=a'oj.
The proof is identical to that of tho 6, Chap. XII-5, except that here,
of course, one must use tho 7, instead of tho 4 of Chap. XII-3.
§ 10. Local behavior of abelian extensions. Let k be as above; let v

be any place of k; as in § 1, we choose an algebraic closure Kv of kv,
containing the algebraic closure ii of k. If k' is any extension of k of
finite degree, contained in ii, prop. 1 of Chap. 111-1 shows that the sub-
field of Kv generated by k' over kv may be identified with the completion
k~ of k' at one of the places w lying above V. If k' is a Galois extension
of k, with the Galois group g, we can apply corollary 4 of tho 4, Chap. III -4,
as we have already done in similar cases on earlier occasions. This shows
that k~ is a Galois extension of kv; if ~ is its Galois group over kv, the
restriction morphism of ~ into 9 is injective and may be used to identify ~
with a subgroup of g; then the completions of k' at the places of k' lying
above v are in a one-to-one correspondence with the cosets of ~ in 9
and are all isomorphic to k~.
We now apply this to the case when k' is abelian over k. Then, by
corollary 2 of tho 7, § 9, its Galois group 9 is isomorphic to k~/U with
U = N (k') = e N k' /k(k AX), U being then an open subgroup of k ~ of
finite index. More precisely, if $ is the subgroup of the Galois group ~
of kab over k, corresponding to k', the canonical morphism a determines
an isomorphism of k~/U onto g=~/$. On the other hand, if kv, k~ are
as above, k~ is an abelian extension of kv, with which corollary 2 of
tho 4, Chap. XlI-3, associates the open subgroup Uv=Nk:,,/dk~X) of
k:. Call ~v' as before, the Galois group of k v•ab over kv; call $v the
subgroup of ~v corresponding to k~; then the same corollary shows
that the canonical morphism av of k~ into ~v determines an isomorphism
of k: IV v onto ~ = 2lv/~v' The relation between these various groups is
given by the following:
PROPOSITION 14. Assumptions and notations being as above, the
subgroup U v of k:, associated with k~, is given by Uv=k~ nV. If 9 is
identified with k~/U by means of a, and ~ with kvlV v by means of av,
the restriction morphism of ~ into 9 is the same as the morphism of k; I U v
into k~/U determined by the natural injection jv of k~ into k~, and the
places of k' which lie above v are in a one-to-one correspondence with the
cosets of k~ U in k~.
Take any zvEk~, and put ex= av(zv)' By prop. 2 of § 1, the automor-
phism of kab induced by ex is pv(ex)=a(z) with z=jv(zv)' As k~ is generated
by k' over kv, ex induces the identity on k~ if and only if Pv(ex) induces the
identity on k';in view of corollary 2 ofth. 4, Chap. XII-3, and of corollary 2
of tho 7, § 9, this amounts to saying that Zv is in Uv if and only if jv(zv) is
in U, which we express by Uv=k~ n U. The second assertion in our

proposition follows at once from the same facts; they also imply that
the image of ~ in 9 can be identified with that of k~ in k~/U, which is
k: UIU, and that g/~ can be identified with k~/k~ U. As we have
recalled above, the places of k' above v correspond bijectively to the
cosets of ~ in g, hence also to those of k: U in k~; this completes the
proof. Our proposition and its proof remain valid when v is an infinite
place, since theorem 4 of Chap. XII-3, and its corollaries, remain valid
for Rand C, as has been observed at the time. The relations between the
various groups and morphisms considered above are illustrated by the
following diagram.
av
k vX I ~v
~ k:IU v ' ,~/

jv
j j Pv
k~/U , ' 9
/ ~
k~ I~
a
COROLLARY 1. Let y be the group of the characters of k~, trivial

on U; let Yv be the group of the characters of k:, trivial on U v. Then the
mapping which, to every wey, assigns the character Wv induced by w on
k~, is a surjective morphism of y onto Yv, and the order of its kernel is
equal to the number of places of k' lying above v.
Clearly W~Wv determines a morphism of y into Yv. Every character

of k~, trivial on Uv, can be uniquely extended to one of k: U, trivial on
U, and this can be extended to one of k~, which then belongs to y;
therefore the morphism in question is surjective. Its kernel consists of
the characters of k~, trivial on k: U; this is the dual group to k~/k~ U;
in view of the last assertion in proposition 14, its order is therefore as
stated in our corollary.
COROLLARY 2. Assumptions being as above, assume also that v is a
finite place of k. Then the modular degree f, and the order of ramification
e, of k~ over kv are given by
By the corollary of prop. 6, Chap. XII-2, and corollary 2 of th.4,

Chap. XII-3, the maximal unramified extension of kv contained in 1<., is
the one ass~ciated with the subgroup r~ U v of k:; the first part of our
corollary follows from this at once; the second part is an immediate
consequence of the first.
COROLLARY 3. Assumptions being as in corollary 2, k~ is unramified

over kv if and only if U ~ r: ; when that is so, the automorphism of k'
over k, induced by the Frobenius automorphism of 1<., over kv, is the image
in 9=k~/U of any prime element 1tv of kv, and it is an element of 9 of
order f.
To say that 1<., is unramified over kv is to say that e= 1, so that the
first assertion is a special case of corollary 2. The second one follows
at once from proposition 14, combined with corollary 4 of th.1,
Chap. XII-2, which says that (lv(1tv) is here the Frobenius automorphism
of 1<., over kv'
Notations being as in corollaries 2 and 3, we know from the corol-
lary of prop. 3, Chap. VIII-1, that 1<., is unramified over kv if and only
if its different over kv is r~. In view of the definitions of the different
and of the discriminant in Chap. VIII-4, and of the fact that the com-
pletions of k', at the places of k' lying above v, are all isomorphic to 1<."
it amounts to the same to say that 1<., is unramified over kv if and only
if v does not occur in the discriminant of k' over k. By corollary 3 of
prop. 14, this is so if and only if U ~ r; .
This qualitative result can be
refined into a more precise one, as follows:
THEOREM 9. Let k' be an extension of k of finite degree, contained
in kab ; let U=kxNk'lk(kAX ) be the subgroup of k~ associated with k', and
call y the group of the characters of k~, trivial on U. For each COEY, call
f(co) the conductor of co. Then the discriminant ~ of k' over k is given by
~= nf(co), or by ~ = L f(co), according as k is of characteristic 0 or not.
Let notations be the same as in corollary 2 of prop. 14; let Pv be the

maximal ideal in the maximal compact subring rv of kv ; call p~ the dis-
criminant of 1<., over kv, and v the number of places of k' lying above v.
As the completions of k' at these places are all isomorphic to 1<." they
all make the same contribution to the discriminant ~, so that their
total contribution is p~v (resp. t5v·v). Let Yv be defined as in corollary 1
of prop. 14; call co;, for 1 ~i~d, the distinct elements of Yv, and, for
each i, call p[(i) the conductor of w;; by corollary 2 of tho 5, Chap. XII-4,
we have [)= If(i). By corollary 1 of prop. 14, each w; is induced on k~
by exactly v characters WEY. Our assertion is now obvious.
THEOREM to. Assumptions and notations being as in theorem 9, the

Dedekind zeta{unction of k' is given by (k'(S) = n
L(s,w).
roey
It is enough to prove this for Re(s) > 1, when the infinite products
for these functions are absolutely convergent; and then it is enough to
show that, for each finite place v of k, the contribution of the places
of k' above v to (k'(S) is equal to the product of the contributions of v
to the products L(s,w). If f is as in corollary 2 of prop. 14, the contri-
bution of w to the product (ds) is (1 - q';-Is) - \ that is also the contri-
bution of each one of the places of k' above v, so that, if v is their number,
their total contribution is (l_q,;-Is)-v. On the other hand, for WEY,
the contribution of v to L(s,w) is 1 unless Wv is unramified, and
(1-w v(1I: v)q,;-st 1 if it is unramified. In view of corollary 1 of prop. 14,
their product is equal to n(1-w'(1I: v)q,;-st v, where the latter product
is taken over all the distinct characters w' of k~ , trivial on Uv and on r~ ,
i.e. trivial on r~ Uv • By corollary 2 of prop. 14, the group k~ /r: Uv is
of order f; clearly it is generated by the image of 1I:v in it, hence cyclic;
therefore there are f characters w', and the values they take at 1I:v ate
the fth roots of 1 in e. This implies that the product n(1-w'(1I:v)t),
for every tEe, is equal to 1-tI , which completes our proof.
COROLLARY. Assumptions and notations being as in theorems 9

and 10, assume also that k is an algebraic number-field. Then Zk'(S)=
n
1I:np/ 2 A (s,w), where n is the degree of k' over k, and p is the number
roey
of real places of k such that the places of k' above them are imaginary.
Here Zk'(S) and A(s,w) are the functions defined in theorem 3 of

Chap. VII-6, aQd in theorem 5 of Chap. VII-7, respectively. In view of
theorem to, what we have to show is that each infinite place v of k contri-
butes the same G-factors to both sides of the formula in our corollary.
Define v as above; then the total contribution to Zds) of the v places
of k' above v is G1 (s)V or G2 (s)', according as w is real or not. The con-
tribution ofv toA(s,w) is G1 (s+sv) or G2 (s+sv) according as v is real
or not, Sv depending upon Wv in the manner described in Chap. VII-7.
Here Wv has to be trivial on Uv, which, being an open subgroup of k~,
is ex if kv = e, and either R x or R~ if kv = R. If Wv is trivial on k:, we
must put sv=O; if not, we must have kv=R, Uv=R~ and w v(x)=x- 1 Ixl,
hence Sv= 1. As the degree of k~ over kv is [k~ : Uv], it is 2 in the latter
case, and otherwise 1. Taking now corollary 1 of prop. 14 into account,
§ II. "Classical" ciassfield theory 281
we see that the contribution of v to nA(s,w) is G1 (s)" if v and ware

real, G2 (s)" if they are both imaginary, and G1 (S)vG 1 (s+ 1)" if v is real
and w imaginary. In the latter case, we have v = n12, e. g. by corollary 1
of tho 4, Chap. 111-4. Our corollary follows now at once from these facts
and from the identity G2 (S)=1tG 1 (s)G t (s + 1), which is the same as the
identity between gamma functions already quoted at the end of Chap. X.
§ 11. "Classical" c1assfield theory. The reinterpretation of our results,

in the traditional language of this theory, depends upon the following
facts:
(a) Let U be the set of all the open subgroups of k~. containing kX;
let U' be the set of those which are of finite index in k~, and U" the set
of those which are contained in kl and of finite index in kl. Lemma 1
of Chap. XII -1 shows that U = U' and U" = ~ if k is an algebraic number-
field, and that U = U' u U" if k is of characteristic p> 1.
(b) Let 5\ be the set of all the fields between k and k..b' of finite degree
over k; when k is of characteristic p> 1, let 5\0 be the set of all the fields
between ko and kab • of finite degree over k o. Then corollary 2 of tho 7,
§ 9, defines a one-to-one correspondence between U' and 5\, while, by
the last assertion in tho 6, § 9, and Galois theory, there is a one-to-one
correspondence between U" and 5\0 when k is of characteristic p> 1.
(c) As the open subgroups of any group are the kernels of its mor-
phisms onto discrete groups, we may regard the open subgroups of k~
in (a) as kernels of such morphisms, and describe these morphisms in
terms of morphisms of the groups I(P), D(P), in the manner explained
in Chap. VII-8. In order to reinterpret the results of Chap. VII-8 more
conveniently for our present purposes, we will modify its notations as
follows.
As in Chap. VII-8, when P is any finite set of places of k, contain-
ing poo • we write Gp for the group of the ideles (zv) of k such that Zv= 1
for all VEP, and G~ for the group of the ideles (zv) such that Zv= 1 for
VEP, and zvEr:, i.e. IZvlv=l, for v not in P. We will now write 4> for
the free group generated by the places v not in P, that group being
written multiplicatively; this may be identified in an obvious manner
with the group I(P) or D(P) of Chap. VII-8, according to the charac-
teristic of k. We write lp for the morphism of Gp onto L p, with the
n
kernel G~, given by (zv)--+ vr(v) with r(v)=ordv(zv); moreover, for every
vjP
eEk x such that eEr: for all finite places VEP, we write pr(e) = n
v¢P
vp(v)
with p(v)=ordv(e) for v not in P.
DEFINITION 1. A subgroup J of L p will be called a congruence group

if one can find,for every veP, an open subgroup gv of k~, contained in r~
when v is finite, such that pr(~) eJ for every ee n(kX ngv); the group
g= n
gv will then be called a defining group for J.
veP
Clearly it would make no difference in this definition if the groups gv

were restricted to be of the form l+p:;" with m~1 for every finite VEP.
PROPOSITION 15. Notations being as above, call U(P) the set of the
open subgroups of k~ containing k X and containing r~ for all v not in P.
Then, for each U eU(P), the formula U nGp= I; l(J) defines a congruence
n
subgroup J =J(U,P) of Lp; a group g= gv, where the gv are as in de-
finition 1, is a defining group for J if and only if it is contained in U;
U is the closure of k XI; l(J) in k~, and the canonical homomorphism of
k~ onto k~/U determines an isomorphism of Lp!J onto k~/U. Moreover,
U -+J(U,P) maps U(P) bijectively onto the set of all congruence subgroups
of Lp.
Take U e U(P); call w the canonical homomorphism of k~ onto the
discrete group r= k~/U; as the morphism of Gp into r induced by w
is trivial on G~, it can be written as cpolp, where cp is a morphism of Lp
into r; clearly the kernel of cp is J. By the corollary of prop. 17, Chap.
VII-S, this implies that J is a congruence subgroup of Lp; then, by
prop. 17, Chap. VII-S, w is the unique extension of cpolp to k~, trivial on
k x , and it is trivial on g if g is a group of definition for J, so that g c U
when that is so. By prop. 15 of Chap. VII-S, k x Gp is dense in k~; this
implies that cpo lp maps Gp surjectively onto r, so that cp(Lp) = r, and also
that U n(kX Gp)is dense in U; this is the same as k X. (U nGp), i.e. k XI; 1(J).
Conversely, let J be any congruence subgroup of L p , and call cp the
canonical homomorphism of Lp onto the discrete group r=Lp!J; again
by prop. 17 of Chap. VII-S, cpolp can be uniquely extended to a mor-
phism w of k~ into r, trivial on k if then U is the kernel of w, we have
X
;
U eU(P) and J =J(U,P). Finally, if the groups gv are as in def. 1, and if

n n
g= gv' every ee (kX ngv) is in g x Gp, so that, if gc U, the projec-
tion of e onto Gp is in U nGp, and the image of that projection in L p,
which is the same as pr(e), is in J; thus g is then a defining group for J.
COROLLARY 1. Notations being as in proposition 15, let P' be a finite
set of places of k, containing P. Then, if J is any congruence subgroup
of L p, J'=JnL p, is a congruence subgroup of Lp'; if J=J(U,P) with
U eU(P), J' =J(U,P').
Here it is understood that Lp' is to be regarded as a subgroup of L p,
in the obvious manner, for P':::J P. Clearly, then, U(P) c U(P'). If now
UeU(P) and UnGp=l;l(J), it is obvious that UnGp'=I;,l(J') with
J'=JnLp'; our corollary follows at once from this and proposition 15.
§ II. "Classical" c1assfield theory 283
COROLLARY 2. Let P, P' be two finite sets of places of k, containing P00;

let J, J' be congruence subgroups of Lp and of L r , respectively. Then
e 1; l(J) and k 1;,l(J') have the same closure V in k~ if and only if there
X
is a finite set P", containing P and P', such that J nLp" = J' nLr ,; when
that is so, the same is true for all finite sets P" containing P and P', and
V is in U(PnP').
Call V, V' the closures of the two sets in question; then, by pro-
position 15, J = J(V,P) and J' =J(V',P'). If V = V', it follows at once
from proposition 15 that V is in U(PnP'); therefore, by corollary 1, if
P"::JPuP', JnLp" and J'nLp" are both the same as J(V,P"). On the
other hand, if there is P" and J" such that p"::J PuP' and J" = J nLp" =
=J'nLp", corollary 1 gives J"=J(V,P")=J(V',P"), hence V=V' by
proposition 15.
When two congruence groups J, J' are as in corollary 2, one says
that they are equivalent. Since every open subgroup V of kl, contain-
ing k X , belongs to U(P) when P is suitably chosen, it is now clear that
there is a one-to-one correspondence between the set U of all such
groups and the set of equivalence classes of congruence groups. There-
fore the one-to-one correspondence between U and ~ (resp. ~u~o)
mentioned above under (b) determines a similar correspondence between
~ (resp. ~u~o) and the equivalence classes of congruence groups. This
will now be described more in detail.
To begin with, it is obvious, from proposition 15 and its corollaries,
that, when an equivalence class of congruence groups is given, there is
a smallest set P such that this class contains a congruence subgroup J
of Lp; in fact, if V is the open subgroup of k1 corresponding to that
class, P consists of the infinite places, and of the finite places v such
that r~ is not contained in U; if we write U v = U n k~ for all v, this is
the same as to say that r~ is not contained in Vv. Similarly, there is
then a largest defining group for J; this is n
gv' where gv= V v for every
veP
infinite place, and gv= Uvnr: for every finite VEP. When one considers
only defining groups for which gv is of the form 1+ p:;' with m ~ 1 when v
is finite, one must then take, for each such VEP, the smallest integer
m(v) ~ 1 such that 1+ p:;'<V) is contained in Uv. If k is of characteristic
p>1, the divisor Lm(v)'v is then called "the conductor" of U and of
every congruence group equivalent to J. If k is of characteristic 0, one
puts m(v) = 0 or 1, for each real place v of k, according as V v is R x or R~ ;
one puts m(v)=O for all imaginary places v of k; attaching then a sym-
bol Pv' called an "infinite prime", to each infinite place v of k, one calls
the symbol n
veP
p~'(IJ "the conductor" of U, of J, and of the congruence
groups equivalent to J.
284 Global cIassfield theory XIII
In the case of characteristic p> 1, it is obvious that a congruence

subgroup J of L corresponds to an open subgroup U of kl if and only
if it consists of divisors of degree 0 when Lp is identified with the
group D(P) of divisors prime to P. From now on, this case will be ex-
cluded; in other words, when the characteristic is not 0, we consider
exclusively open subgroups of k~ of finite index in k~, abelian extensions
of k of finite degree, and congruence groups which contain at least one
divisor of degree =1= O. This being understood, we can make use of prop. 14
of § 10 and its corollaries. In particular, if k' is the abelian extension
of k corresponding to the open subgroup U of k ~, corollary 4 of that
proposition shows that U contains r~ if and only if J<" is unramified
over kv for all w above v, i. e. if and only if v does not occur in the dis-
criminant 1) of k' over k. We will write ,1 for the set consisting of the
infinite places of k and of those occurring in the discriminant 1); then
there is a congruence subgroup J of L p , corresponding to U, if and only
if P~ ,1. As to the conductor of U, if we leave aside the infinite places,
it is, in an obvious sense, SUPWEY (f(w)) if notations are as in tho 9 of § 10;
as to the infinite places, the proof of the corollary of tho 10, § 10, shows
that such a place occurs in the conductor if and only if it is real and the
places of k' lying above it are imaginary.
Before discussing the relation between the congruence groups asso-
ciated with U and the Frobenius automorphisms, we introduce some
definitions, valid for an arbitrary Galois extension k' of k of finite degree.
Call 9 the Galois group of k' over k; let v be any place of k, and w a place
of k' lying above V. By corollary 4 of tho 4, Chap. III-4, we can identify
the Galois group 1) of k~ over kv with a subgroup of 9 by means of the
restriction morphism of 1) into g. If v is a finite place, and J<" is unramified
over kv, 1) is cyclic and generated by the Frobenius automorphism ((>w
of k~ over kv; after 1) has been identified with its image in g, ((>w may
be regarded as an element of g; this is called the Frobenius automorphism
of k' over k at W. If w' is another place of k' above v, the same corollary
shows that there is a kv-linear isomorphism of k~ onto J<"" determined
by an automorphism (J of k' over k; then the Frobenius automorphism
of k' over k at w' is (J-l ((>w(J. Clearly ((>w is the identity if and only if v
splits fully in k'. In particular, let k, k' be algebraic number-fields; let r, r'
be their maximal orders; let Pv, p~ be the prime ideals, in r and in r'
respectively, corresponding to v and to w; then r/pv, r'/p~ are finite fields,
with q = qv and q' = q~ elements, respectively, and ((>w is the automor-
phism of k' over k which determines on r'/p~ the automorphism x-tx q •
This may also be defined as the automorphism ((> of k' over k for which
~tp == ~q (p~) for every ~ E r'.
If, in addition to the above assumptions, we also assume k' to be
abelian over k, i. e. 9 to be commutative, ((>w is the same for all the places w
§ II. "Classical" c1assfield theory 285
above k; in this case, the only one with which we are concerned here,
CPw is called the Frobenius automorphism of k' over k at v; we will denote
it by (k'jklv), or (k'lv) when there is no risk of confusion. We may now
reinterpret corollary 3 of prop. 14, § 10, as follows. As in that corollary,
call U the open subgroup of k;' associated with k', and identify the
Galois group 9 of k' over k with k~jU by means of the canonical mor-
phism. Take P ~.1, with A defined as above. The canonical homomor-
phism of k~ onto 9 = k~jU is trivial on r;
for every v not in A, so that
it induces on Gp a morphism of Gp into g, trivial on G~, which deter-
mines a morphism cp of Lp=GpjG~ into g. Corollary 3 of prop. 14, § 10,
says now that, for every v not in P, cp(v) is the Frobenius automorphism
CPv=(k'lv) of k' over k at v, as defined above. This morphism cp of Lp
into g, defined for P~A, will be denoted by m--+(k'jklm); one writes
(k'lm) instead of (k'jklm) when there is no risk of confusion, and calls
this "the Artin symbol". It may be characterized as the morphism of Lp
(or, what amounts to the same, ofthe group of ideals [(P), or of the group
of divisors D(P), according to the characteristic) into 9 which maps
every place of k, not in P, onto the Frobenius automorphism of k' over k
at that place. In view of prop. 15, we have thus proved that this mor-
phism is surjective and that its kernel J =J(U,P) is a congruence sub-
group of Lp. When one takes for P all the finite sets of places contain-
ing A, the kernels J(U,P) make up an equivalence class of congruence
groups; they are all contained in J(k')=J(U,A).
The above results show also that a finite place v of k splits fully in k'
if and only if it belongs to J(k'). It follows now from prop. 15 of Chap.
VIII-5 that, if k" is a separable extension of k contained in k, and if
almost all the places of k belonging to J(k') split fully in k", k" is
contained in k'. Obviously this implies that there are infinitely many
places of k belonging to J(k'); it will be seen in § 12 that the same is true
for all the cosets of J(k') in L d • The corollary of prop. 15, Chap. VIII-5,
shows also that, if k" is a Galois extension of k, it contains k' if and only
if almost all the places of k which split fully in k" are in J(k'). From this,
it follows that, if k' and k" are two abelian extensions of k contained
in k, k" contains k' if and only if there is a set P for which J(k")nL p is
contained in J(k')nLp; this may also be considered as a consequence
of the results of § 9, combined with prop. 15 of this §. In particular, k' is
uniquely determined by the equivalence class of congruence groups
determined by J(k'); this, too, is an immediate consequence of the results
of § 9 and of prop. 15 of this §. Traditionally, one says that k' is "the
classfield" for that class of congruence groups or for any group belong-
ing to that class.
The above characterization of the class of congruence groups for
which k' is "the classfield" is based solely on the "Artin symbol"; another
one will now be derived from the fact that U =k x N"'I,,(k~X). More ge-
nerally, if we take for k' any extension of k of finite degree, tho 7 of § 9,
and its corollaries, show that the group U =k x Nk'lk(k~X) is the open
subgroup of finite index of k~ associated with the maximal abelian ex-
tension L of k, contained in k'. Take any finite set P of places of k, con-
taining Poo ; for each veP, take an open subgroup gv of k~, contained
n
in r~ when v is finite; put g= gv, Ug=k x gG~ and Jg=J(Ug,P); then,
veP
in the notation of def. 1, J g is the subgroup of Lp consisting of the ele-
ments pr<e) for ee n (k X ngv)' As UgU contains G~, it determines a con-
gruence group J =J(UgU,P), given by l;l(J)= UgUnG p. Call Hp the
group of the ideles (z~) of k' such that z~ = 1 for every place w of k' lying
above a place veP, and H~ the group of the ideles (z~) of k' such that
z~= 1 when w lies above a place veP, and Iz~lw= 1 otherwise; then
L'p=Hp/H~ is the free group generated by the places of k' which do not
lie above P. As N"'lk maps Hp into Gp and H~ into G~, it determines a
morphism 91 of l}p into L p, which is the same as the morphism 9l"'lk
(resp. 6"'lk) of Chap. VIII-4 when l}p, Lp are interpreted as groups of
ideals (resp. of divisors) of k' and of k. By prop. 15 of Chap. VII -8, k' x H p
is dense in k~x, so that k X Nk'lk(H p ) is dense in U. As Ug is open in k~,
this implies that we have
From this, one concludes immediately that J is the subgroup of Lp

generated by J 9 and 9l(I:p ). Call kIf the classfield for the congruence
group J; this is the abelian extension of k associated with the open
subgroup U 9 U of k~, so that it is contained in the abelian extension L
of k associated with U. Call n, no the degrees of k' and of L, respectively,
over k; it is now clear that the index of J in L p , which is equal to that
of U9 U in k~ and to the degree of kIf over k, is ~ no, and that it is equal
to no if and only if U9 c. U, hence kIf = L; this will be the case when P is
taken large enough, and g small enough. We see at the same time that the
index of J in Lp is always ~ n, and that it is equal to n if and only if k'
is abelian over k and is the classfield for J. In other words, when a con-
gruence subgroup J of Lp is given, an extension k' of k of finite degree is
abelian and is the classfield for J if and only if J contains 9l(I:p) and has
an index in Lp equal to the degree of k' over k.
Finally, we can reinterpret corollary 5 ofth. 7, § 9, as follows. As above,
let k' be an extension of k of finite degree. Let M be an abelian extension
of k contained in some extension of k', and call M' the compositum of M
and k'. Assume that M is the classfield for a congruence subgroup J of
Lp. Let v be a place of k, w a place of k' above v, u' a place of M' above
§ 11. "Classical" classfield theory 287
W, and u the place of M below u'; M~, is the compositum of k~ and M',
hence of k~ and M, hence of k~ and M U. If v is not in P, M u is unramified
over kv; this implies that M~, is then unramified over k~. Therefore M'
is the classfield for some congruence subgroup J' of 1:p • Now let U, U'
be the open subgroups of k~, kAx , respectively associated with M and
with M'; by corollary 5 ofth. 7, §9, U'=N;.MU). By prop. 15, J, J' can
be defined by l;;l(J)=UnG p and by the similar formula for J', U';
therefore an element m' of 1:p is in J' if and only if it is the image of an
element z' of Hp such that z'eU', i.e. Nk'lk(z')eU; as Nt'lk maps Hp
into Gp , this is equivalent to Nk'lk(z')e U nG p , hence to 91(m')eJ. There-
fore we have J' = 91- 1 (J).
As an illustration, we will now apply the above considerations to
the case k = Q, which has been treated from another point of view in § 4.
Take k' = Q(e), where e is a primitive m-th root of 1; as before, identify its
Galois group 9 with (Z/mZr by assigning to the automorphism e-+8",
with xeZ, (x,m) = 1, the image of x in (Z/mZr. As we have observed
before, it is obvious that, for every rational prime p, not dividing m, and
for every place w of k' above p, k~ is unramified over Qp, and that the
Frobenius automorphism of k' over Q at p is the one given by e-+eP,
i.e. the image of p in (Z/mZr. Consequently, only primes dividing m
can occur in the discriminant of k' over Q, and k' is the classfield for some
congruence subgroup J of the group Lm of the fractional ideals of Q,
prime to m; Lm can be identified in an obvious manner with the group
of the fractions r = alb, where a, b are two integers > 0, both prime to m.
Moreover, the Artin symbol r-+(k'/Qlr) is the morphism of Lm into
(Z/mZr which maps every prime p, not dividing m, onto its image in
(Z/mZr; clearly this maps every integer a>O, prime to m, onto its
image in (Z/m Z) x, and its kernel J consists of the elements alb of Lm
for which a == b (m). It can easily be verified that the "conductor" for this
group J is 1 if m = 1 or 2, that it is Vex> (m/2) if m is even and m/2 is odd, and
that it is Vex> m in all other cases. Except in the trivial cases m = 1 or 2,
when k' = Q, one may express this by saying that the conductor is Vex> m',
where m' is the smallest integer such that Q(e) is generated over Q by a
primitive m'-th root of 1. As we have seen, this implies that the primes
occurring in the discriminant of Q(e) over Q are those which divide m';
it would be easy now to compute that discriminant itself, by means of
tho 9 of § 10. It is also a consequence of what we have seen above that,
if k is any algebraic number-field, and e is again a primitive m-th root of 1,
k(e) is the classfield for the congruence subgroup J' of the group 1:m of
fractional ideal" of k, prime to m, consisting of the fractional ideals m
such that 91(m)eJ, where J is as defined above.
§ 12. "Coronidis loco". The results of § 10 give the answer to a

question which could not be settled in Chap. VII-5.
THEOREM 11. Let W be any non-trivial character of k~, trivial on k x •
* Then L(I,w)=#=O.
Except for the case w2 = 1, this is contained in corollary 2 of tho 2,
Chap. VII-5. Assume now that w is of order 2; call U its kernel, which is
an open subgroup of k~ of index 2, containing k x. By corollary 2 of
tho 7, § 9, there is a quadratic extension k' of k associated with U. By tho 10
of § 10, we have
If k is of characteristic 0, by the corollary of tho 3, Chap. VII-6, both

Ck and Ck' have a simple pole at s = 1, and their residues there, whose
values are given by that corollary, are >0. The same is true when k is
of characteristic p> 1, by tho 4 of Chap. VII-6. Therefore L(I,w»O.
One should observe that the above proof can be extended in an
obvious manner to any non-trivial character w of k~ of finite order,
trivial on k X , by applying tho 10 of § 10 to the cyclic extension k' of k
associated with the kernel U of w; so far as the conclusion of theorem 11
is concerned, this adds nothing new to what has already been proved
by a different method in corollary 2 of tho 2, Chap. VII-5, but it supplies
some important relations between the class-numbers of k and k' and the
values of the corresponding L-functions at s = 1; more generally, tho 10
of§ 10 shows at once that similar relations hold for all abelian extensions
of k of finite degree. One should also note that, if w,' for SEC, has the
same meaning as in Chap. VII, and if one replaces w by witw in theorem 11,
* one finds that L(1 + it,w)=#=O for all tER.
COROLLARY. Let ko be an A-field contained in k; let V be a set of finite
places of k, such that, for almost all the finite places v of k, not in V, the
closure of ko in kv is not kv' Let w be a non-trivial character of k~, trivial
on k x , such that Wv is unramified at all the places VE V. Then the product
q(k, V,w,s)= n(l-w v( v)q;'t
ve-y
1t 1
is absolutely convergent for Re(s) > 1 and tends to a finite limit, other
than 0, when s tends to 1.
For almost all v, by tho 1 of Chap. VIII-4, kv is unramified over the
closure (ko)u of ko in kv, so that its modular degree over (ko)u is equal to
its degree over the same field. In view of this, the assumption made
above about V is identical with that made in corollary 3 of tho 2, Chap.
VII-5. That being so, the proof of the latter corollary can be applied here;
§ 12. "Coronidis loco" 289
when that is done, one sees that our assertion is an immediate conse-
quence of theorem 11, combined with corollary 3 of prop. 1, Chap. VII-l.
THEOREM 12. Let L be an A-field, ko an A-field contained in L, and ex
an automorphism of Lover ko. Then there are infinitely many places w
of L such that Lw is unramified over the closure of ko in Lw and that the
Frobenius automorphism of Lw over that closure induces ex on L.
Call k the subfield of L consisting of the elements of L, fixed under ex;
as ko eke L, L has a finite degree dover k; by Galois theory, this implies
that L is cyclic over k, its Galois group 9 over k being the one generated
by ex. For each place v of k, call u the place of ko which lies below v, and
let w be any place of L above v; then the closure of ko in Lw is (k o),,' By
tho 1 of Chap. VIII-4, there is a finite set P of places of k, containing P00'
such that, when v is not in P, kv is unramified over (k o)", and Lw over kv,
hence also over (k o),,' Call then cp the Frobenius automorphism of Lw
over (k o),,; as this generates the Galois group of Lw over (k o)", it leaves no
element of Lw fixed except those of (k o),,; therefore, if it induces ex on L,
we must have kc(ko)", hence kv=(ko)", and then, in view of our defini-
tions in § 11, ex is the Frobenius automorphism of Lover k at V. Call M 0
the set of the places v of k, not in P, such that kv=l= (k o),,; for every place v
of k, not in Pu M 0' call CPv the Frobenius automorphism of Lover k
at v; call M 1 the set ofthe places v of k, not in Pu M 0, for which CPv = ex,
and call V the complement of Pu M 0 u M 1 in the set of all places of k.
Clearly the assertion in our theorem amounts to saying that M 1 is not a
finite set, and M 1 is finite if and only if V has the property described in
the corollary ofth. 11. Assuming now that V has that property, we will
derive a contradiction from this assumption. With our usual notations,
call X a character of m: attached to the cyclic extension L of k; here, of
course, m: is the Galois group of kab over k, and L is regarded as a subfield
of k ab • Let ~ be the subgroup of m: corresponding to L; then we may write
9 = m:/~, and the group of the characters of 9 consists of the characters
Xi for 0 ~ i < d. Put co = XO It; then, by corollary 3 of prop. 14, § 10, CO v is
unramified if and only if Lw is unramified over kv, and then the Frobenius
automorphism CPv of Lover k at v is the image of Xv in 9 under the mor-
phism of k~ onto 9 determined by It. This gives now, with the notation
of the corollary of tho 11:
q(k, V, coi ,s)= n(1- Xi(cpv) q; st
vev
1.
For brevity, call this qi(S); we have now

+00
logqi(s)= L L Xi(cpvtq;ns/n,
veV n= 1
this being absolutely convergent for Re(s) > 1. This gives:

4-1
L Xi(oc- 1 )logqi(S)=
;=0
= v~ Ct: 1
l(oc- q>v») q;s+ v~v n~ :t>i(OC- 1q>:)q;ns/n.
In the right-hand side, all the coefficients in the first series are 0, since
q>v=/=oc for veV. On the other hand, qv~2 for all v, so that, for each v
and for Re(s) > 1, we have
+00 +00
L Iq;nsl/n~t L q;n~q;2.
n=2 n=2
Therefore the second series in the right-hand side of the above formula
is majorized by dLq;2, which is convergent by prop. 1 of Chap. VII-1.
v
We have thus shown that the left-hand side remains bounded for Re(s) > 1.
On the other hand, the corollary of tho 11 shows that, for 1 ~ i < d,
logqi(s) remains bounded when s tends to 1, and corollary 3 of th.2,
Chap. VII-5, shows that logqo(s) does not. This is a contradiction.
COROLLARY. Notations being as in definition 1 of § 11, let J be a con-
gruence subgroup of Lp; if k is of characteristic p> 1, assume that J con-
tains divisors of degree =/= 0. Then there are infinitely many places of k in
every coset of J in Lp.
In fact, let k' be the "classfield" for J, as explained in § 11; call 9
its Galois group over k. It has been shown in § 11 that the places v of k,
in a given coset of J in L p , are those places, not in P, where the Frobenius
automorphism of k' over k is a given one. Our assertion is now a special
case of theorem 12.
As an illustration for theorem 12, take ko = Q, and take for L the field
generated by a primitive m-th root of 1. Then our theorem says that,
if a is any integer prime to m, there are infinitely many rational primes
congruent to a modulo m. This is Dirichlet's "theorem of the arithmetic
progression", and the proof given above for theorem 12 is directly
modelled on Dirichlet's original proof for his theorem.
Finally, let w, k and k' be again as in the proof oftheorem 11, so that
we have
,,,,(s) = ,,,(s) L(s, w).
If k is of characteristic 0, we have also, by the corollary of tho 10, § 10:
Z",(s)=n P Za,{s) A (s,w),
§ 12. "Coronidis loco" 291
where p is as explained in that corollary. Now write that the functions

in these formulas satisfy the functional equations contained in theorems 3
and 4 of Chap. VII-6 and theorems 5 and 6 of Chap. VII-7. Writing that
the exponential factors must be the same in the functional equations for
both sides, one gets nothing new; the relation obtained in this manner is
an immediate consequence of tho 9 of § 9. Writing that the constant
factors are the same on both sides, one gets Kw(b) = 1, with K and b
defined as in theorems 5 and 6 of Chap. VII-7. This will now be applied
to a special case. Assume that we have taken for w a character of k~ of
order 2, trivial on k x Q(P<Xl), or, what amounts to the same, trivial on e ,
on k; whenever v is an infinite place, and on r; whenever v is a finite
place. According to prop. 14 of Chap. V11-7, we have then K v = 1 for all v,
hence K= 1, and the idele b is the same as the differental idele a. Therefore,
for every such character w, we have w(a) = 1. Here, if k is an algebraic
number-field, a may be assumed to have been chosen as in prop. 12 of
Chap. VIII-4, i.e. so that id(a) is the different h of k over Q; ifk is ofcharac-
teristic p> 1, we know, by the definition of a differental idele in Chap.
VII-2, that c= div(a) is a divisor belonging to the canonical class. On the
other hand, the conditions imposed on w amount to saying that it is
trivial on e (k1)2 Q(P<Xl); therefore a is in that group. As k1/k x Q(P<Xl)
may be identified with the group I(k)/P(k) of the ideal-classes of k, if k
is an algebraic number-field, and with the group D(k)/P(k) ofthe divisor-
classes of kif k is of characteristic p > 1, we have thus proved the following
theorem (due to Hecke in the case of algebraic number-fields):
THEOREM 13. If k is an algebraic number~field, there is an ideal-class
of k whose square is the class defined by the different of k over Q. If k
is of characteristic p> 1, there is a divisor-class of k whose square is the
canonical class of k.
Notes to the text
(The places in the text to which these notes belong have been marked by a * in the
margin.)
P.I: Cf. E. Witt, Hamb. Abhandl. 8 (1931),413.

P. 27: The analogy in the text can be pursued much further. Let K and V
be as in definition 1; call two norms N, N' on V equivalent if N'/N is
constant on V. Then the quotient of the set of all K -norms on V by this
equivalence relation can be identified with the so-called "building"
associated by F. Bruhat and J. Tits (cf. Pub!. Math. IHES, n° 41, 1971)
with the group Aut(V), i.e. with GL(n, K) if V =K"; this corresponds to
the "Riemannian symmetric space" associated with GL(n, K) for
K = R, C or H in the classical theory. An "apartment" of that building
consists of the points determined by norms of the form given by pro-
position 3 for a fixed decomposition V = J-1 + ... + v,. of V. The" buildings"
associated with the other "classical groups" over K can also be inter-
preted by means of norms in the spaces on which these groups operate.
P. 74: The proof of theorem 4 given in the text is the one due to G. Fuji-
saki (J. Fac. Sc. Tokyo (I) VII (1958), 567-604). It is in this proof that the
"Minkowski argument" (which appears here in the form of lemma 1,
Chap. 11-4) plays a decisive role, just as it did at the corresponding place
in the classical theory.
P.IOI: For a treatment (due to C. Chevalley) of the topic of "linear
compacity", cf. Chapter II, §§ 27-33, of S. Lefschetz, Algebraic Topology,
A. M. S. 1942. In a locally linearly compact vector-space V over a
(discretely topologized) field K, one can attach, to each linearly compact
open subspace W, an integer d(W) so that, if W::J W', d(W)-d(W') is
the dimension of J-WW' over K; this takes the place of the Haar measure
in the theory of locally compact groups.
P.122: The proof given here is Tate's (cf. J. Tate, Thesis, Princeton 1950=
Chapter XV of Cassels-Frohlich, Algebraic Number Theory, Acad. Press
1967).
P.llS: The proof given here, based on lemma 7, is the classical one, due
to Hadamard (Bull. Soc. Math. 24 (1896), 199-220), with the improve-
ments due to F. Mertens (Sitz.-ber. Ak. Wiss., Wien (Math.-nat. K!.), 107
(1898), 1429-1434).
Notes to the text 293
P.126: In fact, it will be seen (cf. proof of tho 11, Chap. XIII-12) that, if
w 2 = 1, w =1= 1, there is a quadratic extension k' of k such that
p(k, P, W, S)=p(k', pi, s) p(k, P, S)-i
where pi is the set of places of k' above P; in substance, this is equivalent
to the "law of quadratic reciprocity" for k. As both factors in the right-
hand side have a simple pole at s= 1, this proves the assertion. That
proof, however, can be replaced by a simple function-theoretic argument,
as follows. Note first that, for w 2 = 1, the product
Pi (s) = P(k, P, w, s) p(k, P, s)
is a product of factors respectively equal to
00
(1_q,;-s)-2= I(n+l)q;nS
n=O
or to
L q,;-2ns
00
(1_q,;-2s)-1 =
n=O
according as A.(v) is 1 or -1. Expanding this into a Dirichlet series, we

get for Pi (s) a series with coefficients in R+ which diverges for s = O. By
an elementary lemma, originally due to Landau (cf. e. g. E. C. Titchmarsh,
The Theory of. Functions (2nd ed.), Oxford 1939, § 9.2) the function
defined by such a series must have a singular point on R+. On the other
hand, in view of our results in §§ 6-7, Pi (s) would be holomorphic in the
whole plane if p(k, P, w, s) was 0 at s= 1. Cf. also the remark at the end
of the proof of tho 11, Chap. XIII-12, and the Notes to p. 288.
P.152: The theorem expressed by formula (11) is due to J. Herbrand

(J.de Math. (IX) 10 (1931), 481-498); hence the name we have given to
"the Herbrand distribution".
P.165: This argument is incomplete. Before applying prop. 2 to C/C,

Z, M, one should first observe that M, regarded as a (C/C)-module, is
both faithful and simple. For any ZEZ, the mapping m-tzm is an endo-
morphism of M as a (C/C)-module, hence also of M as a C-module,
hence of the form m-tem with eEK; therefore Z is isomorphic to K,
and C/ C is an algebra over K in the sense of § 1 (this was tacitly assumed
in the text). The proof proceeds then as before.
P.17S: Cf. R. Brauer, Math. Zeit. 28 (1928), 677-696.
P. 202: An alternative proof (communicated by A. Dress) is as follows.

Call N = n2 the dimension of A over k; take IX as in the text; identify
EnddA) with MN(k) by means of the basis IX. Then prop. 3 of Chap. IX-1
defines an isomorphism F of A®Ao onto MN(k). As IX®IX is a basis of
294 Notes to the text
A®AO over k, F determines, for almost all v, an isomorphism F" of the

rv-Iattice Av generated in Av®A~ by (X®(X onto MN(rv); when that is so, by
th.10f Chap. X-I, Av is a maximal compact subring of Av®A~; as easily
seen, this implies that (Xv is a maximal compact subring of Av. By tho 1 of
Chap. X-t, there is then a division algebra Dover kv, an integer v and an
isomorphism cp of Mv(D) onto Av such that cp maps Mv(R) onto (Xv,
R being the maximal compact subring in D. Let rr be a prime element
of kv; using prop. 5 of Chap. 1-4, one sees easily that R/rrR and M.(R)/
rrM.(R) are simple rings (i.e. that they have no non-trivial two-sided
ideals) if and only if D = kv. Consequently, A is unramified at v if and
only if the ring (Xv/rr(Xv is simple; but it must be so if v is as above, as one
sees at once by using the isomorphism F" and the fact that the ring
MN(rv)/rrMN(rv) is simple.
P. 206: Cf. M. Eichler, Math. Zeit. 43 (1938), 481-494.
P. 208: This statement is obviously false if K is of characteristic p> 1 ;
for instance, it contradicts the results of Chap. XII-3 if those of Chap. 11-3
are taken into account. If K is of characteristic 0, the statement is correct.
P. 241: The proof of the transfer theorem given here is the one due to
C. Chevalley (J. Math. Soc. Japan 3 (1951), 36-44). For another proof,
cf. Appendix I in this volume.
P. 256: Cf. H. Hasse, Math. Ann. 107 (1933), 731-760.
P. 262: The content of proposition 8 may be expressed by saying that,
in the duality between kA /(k A)" and itself defined by the Hilbert symbol
(cf. prop. 7), the image of k X in that group (which is a discrete subgroup
with compact factor-group) is self-dual, i. e. that it is the group" associated
by duality" with itself in the sense of Chap. 11-5.
P. 273: Cf. C. Chevalley, loco cit. (in the Note to p. 241).
P. 288: Cf. above, Note to p.126.
P. 288: Of course the same argument applies to w = 1; in other words,
't(l +it)+O for tER, t+O. As first shown by Hadamard for k=Q (loc.
cit., Note to p.125), this fact is essentially equivalent to the "prime
number theorem" (more precisely, the "prime ideal theorem") for k.
P. 291: This proof (originally arising from a suggestion by J.-P. Serre) is
taken from J. V. Armitage, Invent. Math. 2 (1967),238-246.
Appendix I
The transfer theorem

1. As in Chap. IX-3, take an arbitrary field K and an extension K'
of K of finite degree n, contained in Ksep; write (fj, (fj' for the Galois
groups of Ksep over K and over K', respectively. Call t the transfer
homomorphism of (fj/(fj(1) into (fj'/(fj'(1); as explained in Chap. XII-5,
this may be defined by means of any full set {a 1, ... , an} of representatives
of the co sets a(fj' of (fj' in (fj.
Let f' be any factor-set of K' (cf. Chap. IX-3, def.4). For any p, a, L
in (fj, and for 1 ~i~n, we can write pai, aai' Lai uniquely in the form
(1)
with 1 ~j, k, I ~ n and with (x;, Pi' Yi in (fj'. Then the formula
n
(p, a, L)-+f(p, a, L)= nf'((X;, Pi' yl·.- I
defines a factor-set f of K; we will write f =v(f'). If z' is a covariant

mapping of (fj' x (fj' into Ks~p, we can define quite similarly a covariant
mapping Z= v(z') of (fj x (fj into Ks~p; then, if f' is the coboundary of z',
v(f') is the coboundary of v(z'). Therefore v maps coboundaries into
co boundaries and determines a morphism, for which we also write v,
of factor-classes of K' into factor-classes of K. If, for each i, we replace
a i by a)'i with AiE(fj', then, for a given 1', v(f') is modified by the co-
boundary of the covariant mapping
(p, a)-+ TI [I' ((Xi' AT 1 (Xi' Ak 1 Pi) f' ((Xi' Pi' Ak 113;)-1 ]",-1,
i
where j, k, (Xi. Pi are as in (1). This shows that the morphism v for factor-
classes does not depend upon the choice of the ai •
2. Now let notations be as in Chap. IX-4; instead of {X, O}, however,
we will write {X,O}K; and we write {x',O'}K' for the similarly defined
symbol over K'.
LEMMA x' be a character of (fj'; then, for all () E K x :
A. Let
{x' t, O}K=V({x', Old.
0
As in Chap. IX-4, write x' =e fIJ', where fIJ' is a mapping of (fj'

0 into
the interval [0, 1[ on R; fIJ' is constant on co sets modulo (fj,(l). Then
296 Appendix I: The transfer theorem
x'ot=eotJ> with tJ>=tJ>'ot; if p and the IX; are as in (1), this gives (by
definition of the transfer) tJ> (p) = tJ>' (TI IX;). In the formula to be proved,
;
both sides are defined as the classes of certain factor-sets; one has to
show that those factor-sets differ only by the coboundary of some co-
variant mapping z. For any p, a in (fj, define the lXi' Pi as in (1), and put
z(p, a)= ()N where N is the integer
N= L tJ>' (PJii- l ) - tJ>' (TI Pi lXi-I).

i i
It is trivial to verify that z is then a covariant mapping with the required

property.
LEMMA B. Let X be a character of (fj, and x' its restriction to (fj'. Then,
for all ()' E K' x , we have
The proof is similar to that of lemma A. Write x=e o tJ>; both sides
of the formula to be proved are defined as the classes of certain factor-
sets; one verifies that the latter differ by the coboundary of the covariant
mapping z given, for all p, a, by the formulas
z(p, a) = TI WUC ')Ni,

i
where j, k, lXi' Pi are given by (1), so that the tv; are integers.
3. Now we take for K a commutative p-field. In view of the definition
of the canonical morphism in Chap. XII-2, the local "transfer theorem ",
i.e. theorem 6 of Chap. XII-5, is equivalent to the following statement:
THEOREM. Let K, K' be as in theorem 8 of Chapter XII-5; then, for all
X'EXK , and all ()EKX, we have
Consider the symbol 1} defined in Chap. XII-2; let 1}' be the corre-
sponding symbol for K'. In view of lemma A, the theorem will be proved
if we show that, for any factor-class c' of K', we have 1} [v (c')] =1}'(c'). By
th.l of Chap. XII-2, we may write c' in the form {x', ()'}K' with an un-
ramified character X' of (fj' and some ()' E K'. Then X' is attached to a
cyclic extension K' (Ji) of K' generated by a root Ji of 1 of order prime
to p, and it is the restriction to (fj' of a suitably chosen character X of (fj
attached to the cyclic unramified extension K(Ji) of K. Our conclusion
follows now at once from lemma B, combined with tho 2 of Chap. XII-2.
Appendix I: The transfer theorem 297
4. In order to deduce the global transfer theorem (theorem 8 of

Chap. XIII-9) from the local one, we first observe the following. Let
notations be as in Chap. XIII-l0; let k' be an extension of k of finite
degree, contained in ksep • For any place v of k, and any place w of k'
lying above v, let m~ be the Galois group of k~.ab over k~, and p~ the
restriction morphism of m~ into the Galois group m' of k~b over k'.
Call t, tw the transfer homomorphisms of minto m' and of mv
into m~,
respectively. Then we have
to Pv= n(p~o t w),
w/v
the product being taken over all the places w of k' lying above v; the
proof of this is easy (and purely group-theoretical) and will be left as an
exercise to the reader. This being granted, the global transfer theorem
is an immediate consequence of the local theorem and of the definitions.
Appendix II
W-groups for local fields

1. For the formulation of Shafarevitch's theorem and related results,
it is convenient to introduce modified Galois groups, to be called
W-groups, as follows. Let K be a commutative p-field; as in Chap. XII-2,
let Ko = K (IDl) be the subfield of Ksep generated over K by the set IDl of
all roots of 10f order prime to p in Ksep. Let 5\ be a Galois extension of
K between Ko and Ksep; let 6), 6)0 be the Galois groups of 5\ over K and
over K o , respectively. Let cp be the restriction to 5\ of a Frobenius
automorphism of Ksep over K. We put
and give to m the topology determined by a fundamental system of

neighborhoods of the identity in 6)0 (e.g., by all open subgroups of 6)0)'
This makes m into a locally compact group with the maximal compact
subgroup 6)0; m/6)o is discrete and isomorphic to Z. With this topology,
m will be called the W-group of 5\ over K; it has an obvious injective
morphism tJ into 6), which maps it onto a dense subgroup of 6).
Call q the module of K; the Frobenius automorphism cp determines
on IDl the bijective mapping J.l-+J.l"'=J.lq, and cp" determines on IDl, for
every neZ, a bijection which we write as J.l-+J.lQ with Q=q". Then m
may be described as consisting of those automorphisms ro of 5\ over K
which determine on IDl a bijection of the form J.l-+J.l())=J.lQ with Q=q",
neZ; when ro and Q are such, we will write Irol!ID=Q-1 and call Irol!ID
the module of ro in m. Clearly ro-+lrol!ID is a morphism of m into R~
with the compact kernel 6)0' and it maps m onto the subgroup of R~
generated by q.
2. If 5\' is any Galois extension of K between Ko and 5\, and r is the
Galois group of 5\ over 5\', we may clearly identify the W-group of 5\'
over K with mfr. On the other hand, let K' be any finite extension of K
between K and 5\; let 6)' be the Galois group of 5\ over K', and m' its
W-group over K'; clearly we have m' = tJ -1 (6)'). As 6)' and its cosets in 6)
are open in 6), m' is open in m and has a finite index, equal to that of 6)'
in 6) and to the degree of K' over K. If K' is a Galois extension of K, we
can identify its Galois group over K with m/m' as well as with 6)/6)'.
Conversely, let m' be any open subgroup of m of finite index in m.
Appendix II: W-groups for local fields 299
Then (fjo n W' is open in (fjo and therefore belongs, in the sense of Galois
theory, to some finite extension Ko(e) of K o, contained in R Let L be a
finite Galois extension of K between K(e) and R Let q/ be in W' and
not in (fjo; replacing q/ by qJ'-l if necessary, we may assume that
IqJ'lm=qn with n>O. Take an integer v>O such that qJ'v induces the
identity on L; call K" the compositum of L and of the unramified ex-
tension Knv of degree nv of Kin K o, and let W" be the W-group of 5\
over K". Take any WEW"; as W induces the identity on Knv' we have
Iwlm=qnvi with some iEZ. Then WqJ,-vi induces the identity on Ko and
on L, hence on Ko(e), so that it is in W'. Thus W" is contained in W'. As
we have seen that the Galois group of K" over K may be identified with
WjW", this shows that W' belongs to some field K' between K and K",
and, more precisely, that it is the W-group of 5\ over K'. Thus we see
that W-groups have the same formal properties as Galois groups.
In particular, a cyclic extension L of K of degree n corresponds to
an open subgroup W' of W of index n whose factor-group is cyclic and
may be identified with the Galois group of Lover K, and conversely. If
X is a character of (fj attached to L, it determines a character X0 () of W,
also of order n; conversely, a character of W is of the form X0 () if and
only if it is of finite order. We will frequently (by abuse of notation) make
no distinction between a character X of (fj and the corresponding
character of W.
3. In applying the above concepts, the field 5\ will mostly be taken
of the form Lab' where L is a finite Galois extension of K. In particular,
we will always denote by WK the W-group of Kab over K. It follows at
once from prop. 7 and corollary 2 of tho 3, Chap. XII-3, that the image
{)(WK ) of WK in the Galois group m: of Kab over K is the same as the
image Q (K X) of K x in m: under the canonical morphism Q. Consequently,
there is a canonical isomorphism wK of K x onto WK such that a = 0 0 WK'
Moreover, it follows from the same results that IwK(O)IJV,c= IOIK for all
OEK x •
Let for instance L be cyclic of degree n over K; as L is contained
in K ab , it corresponds to an open subgroup r of WK , of index n, and we
may identify WKjr with the Galois group 9 of Lover K; every character
of 9 may be regarded as a character of WK , trivial on r. If X is such a
character of order n, i. e. if it is attached to L (in the sense of Chap. IX-4),
then, by the definition of the canonical morphisms Q and W K, X [wK(O)],
for any OE K X. is the Hasse invariant h(A) = (X, O)K of the cyclic algebra
A=[LjK; X, 0] over K.
4. Let K' be any extension of K of finite degree; we assume that
Ksep is contained in K~ep' Let 5\, 5\' be Galois extensions of K and of K',
respectively, such that Ko c 5\ c Sl' c K~ep' Let W, W' be the W-groups
of 5\ over K and of 5\' over K', respectively. Then, just as for ordinary
300 Appendix II: W-groups for local fields
Galois groups, there is a restriction morphism of W' into W, which we

again denote by p; obviously Ip(w')lm=lw'l!lll' for all W'EW'. Such is the
case, for instance, if ~ = K ab , ~' = K~b; it is then an immediate conse-
quence of tho 2, Chap. XII-2 Gust as in corollary lof that theorem) that
poWK,=WKoNK'/K'
5. On the other hand, let ~, ~' be two Galois extensions of K such
that Ko c ~ c~' c Ksep; let W, W' be their W-groups over K, and let r
be the Galois group of ~' over R Then we can identify W with W'/ r,
and the canonical morphism of W' onto W preserves the module. Thus
~ is abelian over K if and only if r contains the closure of the commutator-
group ofW'.
Now take any finite extension K' of K, contained in Ksep; let ~' be
any Galois extension of K between K~b and K sep , e.g. Ksep itself. Call D,
fl the W-groups of~' over K and over K', respectively; write a, D'C for
the closures of their commutator-groups; as D' is an open subgroup of
finite index of D, we may introduce, just as in Chap. XII-5, the transfer
homomorphism t of D/a into D'/D'c. As K ab , K~b are respectively the
maximal abelian extensions of K and of K', contained in ~', the Galois
groups of ~' over Kab and over K~b are a and D'c, respectively, and we
may identify WK with D/a and WK, with D'/D'c, so that t maps WK into
WK ,. Combining now the transfer theorem (cf. Chap. XII-5 and Appen-
dix I) with our definitions for the W-groups, one sees at once that the
theorem in question may be expressed by the formula
toWK=WK,oj,
where j is the natural injection of K x into K' x. Clearly this implies

that t is injective and maps WK onto wK ' (K X).
Appendix III
Shafarevitch's theorem
This theorem gives the structure of the W-group of Lab over K when-
ever K is a commutative p-field and L a finite Galois extension of K.
We begin by supplementing the results of Chapter IX with some addi-
tional observations.
1. Let assumptions and notations be as in Chap. IX, so that K is an
arbitrary field, (fj the Galois group of Ksep over K, and all algebras
over K are understood to be as stated in Chap. IX-1. Let A be a central
simple algebra of dimension n2 over K. Let L be an extension of K of
degree n, and f a K-linear isomorphism of L into A. Call V the vector-
space of dimension n over L, with the same underlying space as A,
defined by (e, x)--+xf(e) for eEL, xEA. For every aEA, the mapping
x--+ax is an endomorphism F(a) of V; F is then a representation of A
into EndL(V), and, by corollary 5 of prop. 3, Chap. IX-I, its L-linear
extension FL to AL is an isomorphism of AL onto EndL(V). Let zEA be
such that zf(e)= f(e)z for all eEL; then x--+xz is in EndL(V) and com-
mutes with F(a) for all aEA; therefore it is in the center of EndL(V), i.e.
of the form x--+xf(O with some (EL, so that z=f(O. In other words,
f (L) is its own" com mutant " in A, and f (LX) its own centralizer in A x.
Let now f' be another embedding of L into A; let V', F' be to f' what V,
F are to f As noted in Chap. IX-2, it follows from prop. 4, Chap. IX-l
that there is an isomorphism Y of V onto V' such that F' = Y -1 F Y.
This means that Y is a bijection of A onto A such that Y(xf(e))=
Y(x)f'(e) and Y(ax)=aY(x) for all eELand all x, a in A. Take X= IA and
put b=Y(lA); then we see that bEA x and that f'=b- 1jb. In other
words, two embeddings f, f' can differ only by an inner automorphism
of A. In particular, let 9 be the group of all automorphisms of Lover K;
then, for every exEg, there is b(lEAx such that fW)=b;;lf(e)b(l for all
eEL; consequently, the normalizer N of f(U) in A x is given by N =
Ub(lf(U), and N/f(U) can be identified with g. For any ex, 13 in g,
(l
b;;/b(lbfJ commutes with U, so that we can write b(lbfJ = b(lfJ A(ex, 13) with
A(ex, f3)EU. Moreover, the b(l are linearly independent over L in V; for
otherwise, taking a maximal subset {b;J of linearly independent ones
L
among them, we could write, for any b(l not in that set, b(l= b)'f(e)');
then, writing that f(lJ) b(l = bJ(rt) for allIJEL, we get a contradiction.
302 Appendix III: Shafarevitch's theorem
2. In particular, assume that L is a Galois extension of K, so that 9

is its Galois group; at the same time, simplify notations by identifying L
with f(L) by means of f Then the b~ make up a basis of V over L, so that
I
A consists of the elements b~ ~~ with ~~EL for all IX. Clearly A is com-
pletely defined as an algebra by the multiplication laws:
(1)
for all IX, Pin 9 and all ~ EL. Moreover, writing that (b~ bp) by is the same
as b~(bpby), one gets
(2) A(IX P, y)A(IX, P)Y = A(IX, P y)A(P, y).
Conversely, let L be a Galois extension of K of degree n, contained in
Ksep; (fj being as before, call f) the Galois group of Ksep over L, so that
the Galois group of Lover K is g=(fj/f). For any PE(fj, write P* for the
image of P in 9 = (fj/fj. For any mapping A of 9 x 9 into U, we define an
f)-regular covariant mapping f of (fj x (fj x (fj into Ksep by
(3)
this is a factor-set if (and only it) A satisfies (2). It is now easily verified
that the algebra A defined by means of f by Brauer's construction
(as described in the proof of lemma 4, Chap. IX-3) is precisely as above
if we call b~ the element of A given (in terms of that construction) by the
covariant mapping (p, 0') -+ () p., ~a.'
As in § 1, consider the normalizer N = UbIZ U of U in A x ; write N C
for its commutator-group, and 't for the transfer homomorphism of
N INc into LX. As the definition of't is invariant with respect to all inner
automorphisms of N, and as such automorphisms determine the identity
on NINc, r must map NINc into the subgroup of the elements ofUwhich
are invariant under such automorphisms, i: e. into K x • On the other hand,
regarding r as a morphism of N into U, and calculating it (according to
definition) by means of the representatives b~ of the co sets of U in N,
one sees at once that, on U, r coincides with NL /K •
3. Assumptions being as in § 2, let K' be a field between K and L,
corresponding to a subgroup g' of g. One verifies at once that an element
of A commutes with all elements of 1A • K' if and only if it is of the form
I b~ ~~, with ~~ ELand ~~ = 0 unless IX Eg'. Clearly these elements make
up a subring A' of A (the" commutant" of K' in A) which is the algebra
over K' defined by means of K', L and the restriction of A to g' x g' just
as A was defined above by means of K, L, A; in particular, it is a central
simple algebra over K'.
4. Let K, L, A be again as in § 2; consider the case where 9 is cyclic;
if IX is a generator of g, we have 9 = {1, IX, ... , IXn - 1}. For P= lXi, we have
Appendix III: Shafarevitch's theorem 303
b;;i~b~=~P, so that we may take b{J=b~ for O:::;i:::;n-1. If Nand rare

as in §2, we also see at once that r(b~)=b:; similarly, for any aeb~L",
we may take 1, a, ... , an - 1 as the representatives of the cosets of L" in N,
and see thus that r(a) = an, hence an e K x • In particular, if we put () = b:, () is
in K x ; it is clear that A is then no other than the cyclic algebra defined in
prop. 11 of Chap. IX-4, i.e. the algebra [LIK; x, 0] if X is the character of
g given by x(a)=e(l/n).
Under those same assumptions, we have, for every ~ e L",
Clb;;l~b~=~~-l; therefore the image U of L" under ~-t~~-l is con-
tained in NC. Conversely, the image of b~ in N IU commutes with the
image of L" in N IU; as these images generate N IU, N IU is commutative,
so that U => N C. Therefore, in this case, NC is the same as U, i. e. (by
Hilbert's theorem) the same as the kernel of the morphism NL/K of L"
into KX.
5. As in § 2, let K be any field, and L a Galois extension of K of
degree n, with the Galois group g.
LEMMA A. Let qJ be a morphism of a group G onto g; let H be its kernel.
Let w be a morphism of H into L"; assume that we have, for all geG and
all heH:
(4) W(g-l h g)= w(h)lP(g).
Then there is a central simple algebra A of dimension n2 over K, containing
L, such that w can be extended to a morphism w* of G into A x satisfying
w* (g-l) ~ w* (g)= ~1P(g)
for all geG and all ~eL. Moreover, these conditions determine A and w*
uniquely, up to isomorphism; and w*(G) L" is then the normalizer of LX
in AX.
F or each a e g, choose g~ e G such that qJ (g~) = a. For any a, P in g,
we can write g~gp=g~ph(a, P) with h(a, p)eH. Writing that (g~gp) gy is
the same as g~(gp gy), we get
h(a p, y) . g; 1 h(a, P) gy = h(a, p y) h(P, y).
Putting A.(a,p)=w[h(a,p)], we see now, in view of (4), that A. satisfies

L
(2), so that we can construct an algebra A = b~ L with the multiplication
laws (1). It is then obvious that the formulas w*(g~h)=b~w(h), for all
aeg, heH, define a morphism w* with the required properties. If A', w'*
have the same properties, then, putting b~=w'*(g~), we see that the b~
satisfy relations similar to (1); from the results of § 2, it follows then
that they are a basis for A' over L and that A', w' * differ from A, w*
only by an isomorphism of A onto A'.
6. Now consider the following situation. Let K, L, 9 be as before;

let E be a Galois extension of K, containing L and contained in K sep ,
of finite degree dover L. Call r,,1 the Galois groups of E over K and
over L, respectively, so that 9 = rj,1. Let G be a group, q/ a morphism of
G onto r, H' the kernel of ql, and OJ' a morphism of H' into EX; we assume
that these data satisfy (4) when they are substituted there for G, qJ, H, OJ
respectively, so that we can apply lemma A to them. This determines an
algebra A' of dimension n2 d2 over K. Call t/I the canonical morphism of
r onto g=rj,1; put qJ=t/l0qJ' and H=qJ,-l(,1); qJ is a morphism of G
onto 9 with the kernel H. To simplify notations, assume that H' is com-
mutative, and let He be the commutator-group of H; then we can define
(as in Chap. XII-5) the transfer homomorphism t of HIB" into H', and
regard it as a morphism of H into H'; we have t(g-lhg)=g-lt(h)g for
all gEG and hEH. Now put OJ = OJ' ° t. We have, for all hEH, hEG:
OJ(g-l hg)=OJ(h)"'(g);
for gEH, this implies that OJ(h) is invariant under qJ'(H)=,1, so that it is
in LX and that we may replace qJ' by qJ in the above formula. Therefore
we can apply lemma A to G, H, qJ, OJ; this defines an algebra A of dimen-
sion n2 over K. The following lemma and its proof are due to Artin
and Tate (E. Artin and J. Tate, Classfield theory, Harvard 1961, Chap.
XIII-3, tho 6, p.188):
LEMMA B. Let A, A' be as above; then, in the Brauer group B(K), we
have CI(A)=CI(A')" with d= [E:LJ.
For each ~Er, choose g~EG such that qJ'(g~)=~; for all~, 11 in r, put
h'(~, 11)=g~,/g~g,,; A.'(~, 17)=OJ'[h'(~, 11)].
As in the proof of lemma A, CI(A') is determined by A.', or, in the language

of Chap. IX-3, by the factor-set f' of K determined in terms of A.' by the
formula similar to (3). On the other hand, the definition of the transfer
gives, for any hEH:
t(h)= n(g"'(h)~' h· g/J)'
8eLl
Choose a full set M of representatives of the cosets ~,1 of,1 inr; for any
~Er, call Jl(~) the representative in M of the coset ~,1. The elements gil'
for JlEM, make up a full set of representatives of the cosets of H in G,
so that we may use them, as in the proof of lemma A, to construct a
factor-set defining CI(A); this is done as follows. Take any two elements
~,11 of r; put cx=Jl(~), {3=Jl(11), Y=Jl(~11), {,=y- 1 cx{3; put
h(~, 11)= g:; 19<1gp, A.(~, 11)=OJ[h(~, 11)J~

as these are constant on cosets of .1 in r, A. may be regarded as a mapping

of 9 x 9 into U, and a factor-set f defining CI(A) is given in terms of A.
by (3). By the definition of w, we have
For any OEL1, put

0' = PO",-I, 0" =IXO' e- 1=lXpO(e",)-I.
When 0 runs through .1, so do 0', 0" and bO. In G, we have the following
(easily verified) group-theoretical identity:
giig:;lgagpgs=h'(y, bO)-1 h'(O", e,,)h'(e, "')

.g;; 1 [h' (0", e)-1 h' (IX, O')]g,,· h' (0', ,,)-1 h' (P, 0).
For every eE r, put
c(e)=w' [n h'(Jl(e), 0) h'(O, e)-I].
Sell
Then, taking into account the fact that w' satisfies (4), we get
This proves the lemma; in fact, if!,!, are as above,f!'-d is the coboundary
of(p, o')~C(O'p-l)P.
7. From now on, we will take for K a commutative p-field. Also, if G
is any topological group (e.g. a W-group), we will denote by GC its topo-
logical commutator-group, i. e. the closure of its commutator-group in
the algebraic sense.
Let A be any central simple algebra over K; if its dimension over K
is n 2 , we can write it as Md(D), where D is a division algebra of dimension
(n/d)2 over K; then CI(A) is the same as Cl(D), and this, as shown in
th.1 of Chap. XII-2 and its corollaries, is of order n/d in the Brauer
group B(K); in other words, the Hasse invariant h(A) is a root of 1 of
order n/d, and it is of order n if and only if A is a division algebra. By
corollary 2 of th.2, Chap. XII-2, combined with corollary 3 of th.3,
Chap. IX-3, this implies that every separable extension L of K of degree n
can be embedded in A; in view of § labove, this embedding is unique,
up to an inner automorphism of A, so that we can apply to K, L and A
all the results of that §. In particular, if L is a Galois extension of K with
the Galois group 9, and if N is the normalizer of 1: in A x, the inner
automorphisms x~a-lxa, for aEN, induce 9 on L, and we can thus
identify N/1: with 9.
8. A straightforward application of lemma A gives now:

THEOREM I. Let K be a commutative p-field, L a Galois extension of K
of degree n; let W, WLbe the W-groups of Lab over K and over L, respectively.
Then there is a central simple algebra A of dimension n2 over K, an em-
bedding of L into A, and an isomorphism W of the normalizer N of LX in
A x onto W, such that the restriction of w to r is the canonical isomorphism
W L of r onto WL, and that, for every aE N, the automorphism induced on L
by x--+a-lxa is the restriction ofw(a) to L.
In fact, in lemma A, substitute W, WL for G, H; for cp, substitute the
canonical morphism of W onto U1WL when U1WL is identified with the
Galois group of Lover K (cf. Appendix II, § 2); for OJ, substitute wLl.
Then (4) follows at once (by "transport of structure") from the fact that
WL is "canonically" attached to the pair (L, Lab)' Under these circum-
stances, it is obvious that the morphism OJ* oflemma A is an isomorphism
of Wonto N; its inverse w has then the required properties.
COROLLARY 1. In theorem I, A, Land ware uniquely characterized
(up to an isomorphism) by the properties stated there.
Also this is part of lemma A. The algebra A, with a given embedding
of L into A, will be called the canonical algebra for the pair (K, L); w will
be called the canonical isomorphism of N onto W.
COROLLARY 2. With A and w as above, write vAjK for the reduced norm
in A over K. Then, for all a EN:
IvA/K (a)IK = Iw(a)lw·
In fact, both sides define morphisms of N into R~; as such, they
must be equal if they coincide on a subgroup of N of finite index, e. g.
on LX. On r, vA/K coincides with NL/K (cf. the proof oflemma 4, Chap.
IX-3), and w with wL ; therefore, for a=~Er, the left-hand side is I~IL'
and the right-hand side is IwLWlwL; in view of Appendix II, § 3, this
proves our assertion.
COROLLARY 3. With A, L, N as above, the transfer homomorphism r of
N / N Cinto r is injective and maps N / NC onto K x ; if it is regarded as a
morphism of N onto K X, then,for every aEN, wK[r(a)] is the restriction
to Kab of the automorphism w(a) of Lab over K.
This follows at once from the transfer theorem, as reformulated for
W-groups at the end of Appendix II, § 5, when this is combined with the
above results.
9. With the same notations as in § 8, the structure of W will now be
completely determined by Shafarevitch's theorem (I. R. Shafarevitch,
C. R. Ac. Sc. URSS 53 (1946), 15-16):
THEOREM II. Let K, L, n be as in theorem I. Then the canonical algebra

for (K, L) is the division algebra with the Hasse invariant e(l/n) over K.
We will write (K; L) for the Hasse invariant h(A) of A; we have to
prove (K; L)=e(1/n), and this will imply that A is a division algebra.
The proof will consist of three steps:
(a) Let K be as above; let L, E be as in § 6; call W', W", Wv the
W-groups of Eab over K, L, E, respectively; take for OJ' the inverse of the
canonical isomorphism lDv of E x onto Wv ; let t be the transfer homo-
morphism of W" /W"c into Wv . As in Appendix II, § 5, we can identify
W"/W"c with WL . If then we call OJ the inverse of the canonical iso-
morphism lDL of V onto WL = W"/W"c, the transfer theorem, as restated
in Appendix II, § 5, gives OJ = OJ' 0 t. We are therefore exactly in the situ-
ation described in § 6, and lemma B gives (K; L)=(K; Ef with d= [E:LJ.
(b) Let K, L, n be as in theorem I; let K' be any cyclic extension of K
of degree n, e. g. the unramified one. Call E the compositum of Land K'
in Ksep; put Kl =K' nL and d= [L:KIJ. Then E is of degree d both
over L and over K'; consequently, by (a), (K; L) and (K; K') are both
equal to (K; Et In particular, we see that (K; L) depends only upon n,
and that it is enough to prove our theorem in the cyclic case.
(c) Take L cyclic over K; take notations as in §4 above; with those
notations, A is the cyclic algebra [L/K; x, OJ with O=t(br. }. In view of
Appendix II, § 3, we have h(A)=X[lDK(O)]. By corollary 3 of theorem I,
§ 8, lDK(O) is the restriction to Kab of the automorphism lD(b~) of Lab
over K; by theorem I, the restriction to L of the latter automorphism,
and therefore also of the former one, is the one induced on L by the
automorphism x~b;lxb~ of A, which is 0(. This gives h(A)=X(O()=
e(l/n), which completes the proof of Shafarevitch's theorem.
Appendix IV
The Herbrand distribution

1. We begin by stating some general facts about the Herbrand
distributions, as defined in Chap. VIII-3. Let again K be a commutative
p-field.
LEMMA A. Let 5\, 5\' be two Galois extensions of K, finite or not, such
that K c 5\ c 5\'. Let (fj, (fj', ~ be the Galois groups of 5\ over K, of 5\'
over K and of 5\' over 5\, respectively; let tP be the canonical morphism of (fj'
onto (fj = (fj'/~. Let H, H' be the H erbrand distributions on (fj and on (fj',
respectively. Then, for every locally constant function f on (fj, we have
H(f) = H'(fo tP).
This is obvious. We may express the conclusion by saying that H is

the image (more precisely, the "direct image") of H' under tP.
LEMMA B. Let 5\ be a Galois extension of K, finite or not. Let K' be
an extension of K of finite degree, contained in 5\, with the order of rami-
fication e and the differ ental exponent dover K. Let (fj, (fj' be the Galois
groups of 5\ over K and over K', respectively; let H, H' be the Herbrand
distributions on (fj and on (fj'. Then, for every locally constant function f
on (fj, equal to 0 outside (fj', we have H'(f)=eH(f)-df(e), where e is the
identity in (fj.
This is also obvious. It may be expressed by saying that H' coincides
with eH on open and compact subsets of (fj', disjoint from e, or more
briefly that it coincides with eH on (fj' outside e; this fact, together with
the trivial condition H'(l)=O, determines H' completely in terms of H.
2. Now let assumptions and notations be as in § 1 of Appendix II.
Let H be the Herbrand distribution on (fj; as has been shown in the
proof of lemma 3, Chap. XII -4, it is 0 outside (fjo. More precisely:
LEMMA C. Let K, K o , 5\,
(fj, (fjo be as in § 1 of Appendix II. Then the
support of the Herbrand distribution H on (fj is (fjo.
Take any AE(fjo, other than the identity; take any open subgroup (fj'
of (fj, not containing A; then, by Chap. VIII-3, we have H(fj" A)<O for
all open subgroups (fj" of (fj'. Now take AE(fj - (fjo; then there is a root f..l
of 1 of order prime to p such that f..lA +- J1; call (fj' the open subgroup of (fj
corresponding to K(f..l). Then A is not in (fj', and, by Chap. VIII-3, we
Appendix IV: The Herbrand distribution 309
have H(fj" X)=O whenever XE(fj' A. and (fj" is an open subgroup of (fj',
so that H is 0 on (fj' A..
As noted above, it also follows from the definition of H that H(fj)=O,
so that, in view of lemma C, H(fjo) = O.
Let now W be the W-group of ~ over K. Clearly there is a unique
distribution H on W which coincides with H on (fjo and is 0 outside (fjo.
This will be called the H erbrand distribution on W. As explained in
Chap. VIII-3, we extend it to a linear form, also denoted by H, on the
space of locally constant functions on W.
3. Now we will apply theorems I and II of Appendix III, §§ 8-9. As
in those theorems, we take a Galois extension L of K, of finite degree n;
we call W, WL the W-groups of Lab over K and over L, respectively. We
call A the canonical algebra for (K, L), N the normalizer of r in A x,
and w the canonical isomorphism of N onto W. We use the isomorphism
w -1 of Wonto N to transport to N the Herbrand distribution on W, and
denote again by H this distribution on N. Our purpose is to give an
explicit formula for H on N.
As before, we write vA/K for the reduced norm in A over K; moreover,
we put Ilxll = IvA/K(X)IK for every XEA. In view of corollary 2 of theorem I,
Appendix III, § 8, and of lemma C above, the support of H on N is the
compact subgroup No of N determined by Ilall= 1, i.e. the kernel of the
morphism a-..llall of N into R~. As noted above, we have H(1)=0.
Let da be the Haar measure on N, normalized so that the measure of
No is 1. The following theorem, in substance, is due to 1. Tate and Shankar
Sen (J. Ind. Math. Soc. 27 (1964),197-202):
THEOREM. For any locally constant function f on N, we have:
(1) H(f)= - Hf(a)- f(1 A )] ·111 A _all- 1 da.

No
As both sides of (1) are 0 for f = 1, it is enough to prove it for the case
f(1A)=O; this will be assumed from now on. The proof will consist of
several steps:
(a) Take the "abelian case" where L=K, n= 1, A=K, N =K x ,
No = R x ; as usual, we write R for the maximal compact subring of K,
and P for its maximal ideal. Clearly it is enough to verify (1) when f is
the characteristic function of any set X of the form X = (1 + PV) ~ with
O::;ord(1-~)<v; then H(X) is given by theorem 5 ofChap.XII-4. At
the same time, the integrand in (1) is 0 outside X and has on X the
constant value qP with p =ord(1- ~). As 1 + Pv has the index qv-1(q -1)
in R x , this proves (1) in this case.
(b) Take now the general case, and take a field K' between K and L,
corresponding to a subgroup g' of the Galois group 9 of Lover K; put
310 Appendix IV: The Herbrand distribution
n' = [L: K']. Then, so far as A, L, N, K and K' are concerned, we are in
the situation considered in § 3 of Appendix III; if we write N = Ub"Lx
and A = I b" L as there, we have seen in that § that the "commutant"
algebra of K' in A is A'=Ib;.L, where the sum is taken over all AEg';
this is a central simple algebra of dimension n,2 over K', and the nor-
malizer of V in A' x is the subgroup N' = Ub;. V of N, the union being
taken again over all AEg'. In view of our observations in § 2 of Appen-
dix II, it is clear that the canonical isomorphism w of N onto W maps N'
onto the W-group W' of Lab over K'; corollary 1 of theorem I, Appen-
dix III, § 8, shows now that A' is the canonical algebra for (K', L), and
that the canonical isomorphism of N' onto W' is the restriction w' of w
to N'. Consequently, if H' is the Herbrand distribution on N', lemma B
shows that, on N' and outside lA, H' coincides with eH, where e is the
order of ramification of K' over K. Now call (1') the formula, similar
to (1), with H', A', N' substituted for H, A, N. For any f, equal to 0 out-
side N' (and at lA' as assumed above), call H 1 (f), H 1(f) the right-hand
sides of (1) and of (1'), respectively; it will be shown now that H 1(f)=
eH1(f)·
Take any x'EA' x; by corollary 1 of prop. 6, Chap. IX-2, and corol-
lary 3 of th.3, Chap. 1-2, the automorphism y' ~x' y' of the additive
group of A' has the module
modA,(x')= IvA'/K' (x')I~'·
Similarly, the module of y~ x' y in A is
modA(x') = IvA/K (x') I~.
But we may also regard A as a left vector-space over the division alge-
bra A'; as the dimensions of A and A' over K are n2 and n,2 d with
d = [K': K] = n/n', A has the dimension d over A'. By corollary 2 of
tho 3, Chap. 1-2, we have then modA(x') = modA,(x't This gives
IVA/K(X')IK = IvA'/K' (X')IK"
Therefore the integrands in H1 (f) and HI (f) are the same. Put now
N~ = N' n No. If Ko is as before (cf. § 2), w maps No onto the Galois
group (f;o of Lab over K o , and similarly it maps N~ onto the Galois
group of Lab over the compositum K'o = K' Ko; therefore the index of
N~ in No is equal to the degree of K'o over K o , which is the same as that
of K' over K' n Ko; this is e, by corollary 4 of tho 7, Chap. 1-4. Conse-
quently, if d'a is the Haar measure on N', normalized so that the measure
of N~ is 1, we have d'a=e,da on N'. This gives H 1(f)=eH 1(f), as we
had asserted.
(c) In particular, apply (b) to the case K' = L. In view of (a), it shows
that (1) holds whenever f is 0 outside V (and at lA)'
Appendix IV: The Herbrand distribution 311
(d) To prove (1), it is enough to show that both sides coincide on

each coset of V in N; we have found in (c) that they do so on V itself;
we still have to verify that they coincide on all other co sets. In other
words, let ba V be any coset of LX in N, other than V ; we have to show
that (1) holds whenever f is 0 outside ba V . Let g' be the cyclic subgroup
of 9 generated by tX; let K' be the field between K and L, corresponding
to g'; apply to K' what has been proved above in (b). We see th,us that it
is enough to verify (1') for f equal to 0 outside bIZ V . Writing now K, N,
9 instead of K', N', g', we see that our theorem will be proved if we
verify (1) under the additional assumptions that 9 is cyclic of order
n> 1, generated by tX. and that f is 0 outside the coset ba V .
(e) That being now assumed, we are once more in the situation
described in § 4 of Appendix III. Let notations be the same as there;
N C is then the same as the kernel U of the morphism NL/K of V into K x ;
as we have I~IL=INL/K(~)IK for all (,EV, U is compact. By corollary 3
of theorem I, Appendix III, § 8, the transfer r of N into LX has the kernel
N C = U and maps N onto K x ; the same corollary shows also that, if we
identify N with W by means of w, and K x with WK = U1Wc by means of
W K , r becomes the canonical morphism of Wonto a;'wr. therefore we
can apply lemma A of § 1 above, and conclude that the direct image
under r of the Herbrand measure H on W is the Herbrand measure HK
on K x, as given by (a) above. In other words, for any locally constant
function F on K x , equal to 0 at 1, we have
(2) H(For)=HK(F)= - J F(x)-ll-xli1dxx,

RX
with the Haar measure dXx on K normalized so that the measure of

R x is 1. Clearly a locally constant function on N can be written as
For if and only if it is constant on the cosets of U.
We have to prove (1) for those functions f on N which are 0 out-
side bIZ V . Observe now that both sides of (1) are clearly invariant under
all inner automorphisms of N, and in particular under any automorphism
a--+~ a ~-l with ~EV. ForaEbaV, wehavea- 1 ~ a= ~a,hence~ a ~-l =a U
with u=~a-l. By Hilbert's theorem, the kernel U of NL/K is the group
consisting of the elements u = ~a-l for ~ E V; consequently, on the coset
baV, the inner automorphisms a--+~a~-l, for ~ELX, induce the same
mappings as the translations a--+a u for UE U. Therefore, if f is 0 outside
blZLX, both sides of (1) remain unchanged, for any UEU, when one
replaces f by the function a--+ f(a u); hence they are still unchanged if
we replace f by a--+ j(a), where j(a) is the mean value of u--+ f(a u) on U
for the Haar measure on U. Thus our theorem will be proved if we
verify (1) for such a function j, i.e. for one which is 0 outside blZLx and
312 Appendix IV: The Herbrand distribution
constant under the translations a-+a u. From now on, let f be such a
function; as we have seen, it can be written asf=For.
As before, put () = r (b"J As r coincides with NL/K on U, it maps the
co sets b~U of U in N, for O~i<n, onto the co sets (}i NL/K(U) of NL/K(U)
in K x, respectively; as it maps N onto K x , and as its kernel U is con-
tained in U, K x is the disjoint union of these n co sets (a result which is
substantially contained in corollary 2 ofth. 3, Chap. XII-3). In particular,
a function f = For is 0 outside baU if and only if F is 0 outside () NL/K(U ).
To complete our proof, we have to compare the right-hand sides of (l)
and of (2) for such a pair of functions j; F. By corollaries 2 and 3 of
theorem I, Appendix III, § 8, we have Ir(a)IK=llall for allaEN; therefore
r -1 (R X) = No, and the direct image of the measure da in (1) is the measure
dXx in (2). Now take any aEbaU; in §4 of Appendix III, we have seen
that r(a)=anEK\ and that the d, for O~i<n, may be taken as re-
presentatives of the co sets of U in N and therefore also as a basis of A
over L; consequently, if {lJo, ... , IJn- d is a basis of Lover K, the elements
dlJj, for O~i,j<n, make up a basis of A over K. In order to evaluate
the integrand of (1) for o'EbaU, put Jl= lilA - all. Then the automorphism
z-+(lA-a)z of the additive group of A has the module Jl n. On the other
hand, this module may also be expressed by means of corollary 3 of
th.3, Chap. 1-2, in terms of the determinant of the linear substitution
determined by that automorphism on the basis {d IJj}; this determinant
is easily seen to have the value (l_x)n, with x=an=r(a). This gives
Jl=ll-xI K. Therefore, for x=r(a), the integrands in the right-hand sides
of(1) and of (2) are the same. This concludes the proof.
Index of definitions
(This index contains all concepts and terms whose definition is given or recalled in the
text, even if this is not done by way of a formal definition. A reference such as IV-l (62)
means Chapter IV, § 1, page 62; IV-I, d. 1 (59) means Chapter IV, § 1, definition 1, page S9;
P & N (XII) means Prerequisites and Notations, page XII).
above, III-I, d. 4 (45). congruence group, XIII-ll, d. 1 (282).
adele, adele-ring, IV-I, d. 1 (60). constant: - -field extension, XIII-I (247);
admissible (for G, for (G, r)), VII-2 (lOS). field of -s, IV-4 (77); locally -, VII-2
A-field, III-I, d. I (43). (106).
algebraic: - dual, II-S (39); - number- covariant, IX-3, d. 3 (172).
field, III -I (43). cyclic: - algebra, IX-4, d. 6 (1841; - ex-
almost all, III-I (47). tension, IX-4 (180); - factor-class,
annihilator, IX-I (162). - factor-set, IX-4 (181).
associated by duality, IJ-S n~).
attached to: factor-class -, IX-3 (176); char- Dedekind zeta-function, VII-6, d. 8 (129).
acter -, IX-4 (180); cyclic extension defining group, XIII-ll, d. I (282).
-, IX-4 (lBI). deg, VI (96).
Aut, IV-3 (71), X-I (188-189). degree: (of a place, of a divisor), VI (96);
automorphism, P & N (XV); Frobenius, modular -, 1-4, d. 4 (15).
cf. Frobenius. denominator, V-3 (87).
different, VIII-I, d. I (140), VIII-4, d. 3
basic character, II-S (41), IV-2 (70).
(IS3).
belonging to (factor-class -), IX-3 (176).
differental: - exponent, VIII-I, d. I (140);
below, III-I, d. 4 (4S).
- idele, VII-2, d. 4 (113).
Brauer group, IX-3 (171).
disconnected: totally -, VII-3 (114).
canonical: - class, - divisor, VI (100); discriminant, V-4, d. 6 (91), VIII-2, d. 2
-injection, IV-I (60), V-2(83);-mor- (144), VIII-4, d. 4 (156), VIII-4 (157).
phism, XII-I (2IS), XII-2 (224), XIII-I div, VI (97, 100).
(24S); - pairing, XIJ-I (215), XII-2 divisor, VI (96), (of a character) VI (100);
(224), XIII-I (245). - -class, VI (97); canonical --, VI
central, IX -I (162). (100); principal-, VI (97).
character, P & N (XV), II-S (38); basic-, dual: -lattice, II-S (39); -measure, VII-2
cf. basic; multiplicative -, XIII-2 (10S); - system, VI (99); algebraic-,
(2S0); order of -, II-S, d. 4 (41). II-S (39); topological-, II-S (38).
CI, IX-3 (170). duality (associated by -), II-S (38).
class: - of algebras, IX -3 (170) ; - ofideals,
V-3 (87); - of divisors, VI (97); canon- Eisenstein polynomial, VIII-2 (147).
ical-, VI (100). embedding, P & N (XVI); natural-,
classfield (for a congruence group), XIII-II VIII-4 (IS4); proper -, III-2 (50).
(28S). End, 111-3 (S3), X-I (188-189).
coboundary, IX-3, d. S (17S). endomorphism, P & N (XV).
coherent, VII-2 (110); - system, VI (97). Euler product, VII-l (102).
completion, III-I, d. 2 (43-44); -at (a exponent (differental-), VIII-I, d. 1(140).
place), III-I (44).
conductor, VII-3, d. 7 (117), VII-7 factor-class, IX-3 (17S); cyclic -, IX-4
(133-134). (181).
factor-set, IX-3, d. 4 (175); cyclic -, IX-4 maximal order, V-2 (82), X-I (191), XI-4
(181). (211).
faithful, IX-I (162). mod, 1-2 (3-4).
field of constants, IV-4 (77). modular degree, 1-4, d. 4 (15).
finite: - field, I-I (I); - place, III-I, d. 3 module: (of an automorphism) 1-2 (3); (of
(44). a p-field) 1-4, d.3 (13); (of an idele)
Fourier transform, VII-2 (105). IV-3 (73).
fractional ideal, V-3, d. 4 (85). monic, P & N (XIV).
Frobenius automorphism, 1-4, d. 5 (20), morphism, P & N (XV); canonical -,
XII-2 (221), XIII-I (247), XIII-ll XII -I (215), XII -2 (224), XIII-I (245);
(284-285). natural -, IX-3 (171); restriction -,
fully: - ramified, 1-4, d. 4 (15); to split-, IX-3 (173).
VIII-5 (158). multiplicative character, XIII-2 (250).
fundamental set, V-4 (89). N, III-3 (53).
genus, VI, d. I (100). 91, V-3, d. 5 (88), VIII-4 (155).
natural: - embedding, VIII-4 (154);
Hasse invariant, XII-2 (221, 224), XIII-3 - morphism, IX-3 (171).
(252). norm, III-3 (53), V-3, d. 5 (88), VIII-4(155);
Herbrand distribution, VIII-3 (152), XH-4 K- -, II-I, d. I (24); reduced -, IX-2
(239). (169); regular -, III-3 (53).
hermitian form, X-3 (198). normal lattice, X-I (193), XI-4 (212).
Hilbert symbol, XIII-5 (262). N-orthogonal, II-I (25-26).
Hom, X-I (188-189). numerator, V-3 (87).
homomorphism, P & N (XIV); transfer-, ord, 1-4, d. 3 (13), II-5, d. 4 (41).
XII-5 (241). order, P & N (XV), V-I, d. 2 (81), X-I (191
t;-regu!:u, IX-3, d. 2 (171). -192), XI-4 (211-212); - (of a char-
acter), II-5. d. 4(41); -of ramification,
id, V-3 (87).
1-4, d. 4 (15); left, right -, X-I (192),
ideal: fractional -, V-3, d. 4 (85).
XI-4 (212); maximal -, V-2 (82), X-I
ideal-class, V-3 (87).
(191), XI-4 (211).
idele, idele group, IV-3 (71); differental-,
orthogonal: N- -, II-I (25-26).
VII-2, d. 4 (113).
orthonormal, X-3 (198).
image (of a Haar measure), II-4 (36).
imaginary (place), III-I, d. 3 (44). p-adic, 1-3 (II).
infinite (place), III-I, d. 3 (44). pairing (canonical-), XII-l (215), XII-2
integral (over a ring), 1-4 (22). (224), XIII-I (245).
invariant (Hasse -), cf. Hasse. p-field, 1-3, d. 2 (12).
inverse (algebra), IX-I (164). place, III-I, d. 2 (43-44); finite, imaginary,
isomorphism, P & N (XV). infinite, real -, III-I, d. 3 (44).
Poisson summation formula, VII-2 (106).
k-Iattice, V-2, d. 3 (83).
polynomial: - mapping, 1II-3 (52); Eisen-
K-lattice, II-2, d. 2 (28).
stein -, VIII-2 (147); prime -, III-I
K-norm, II-I, d. I (24).
(46).
lattice: dual -,11-5 (39); k- -, V-2, d. 3 power-series (formal-), 1-4 (20).
(83); K- - , II-2, d. 2 (28); normal-, prime: - element, 1-4, d. 3 (13); - poly-
X-I (193), XI-4 (212); Q- -, V-I, d. 1 nomial, III-I (46); - to P, VII -8 (136);
(80); R- -, II-4, d. 3 (35). mutually -, V-3 (86).
left order, X-I (192), XI-4 (212). primitive root of I, P & N (XV).
L-function, VII -7 (133). principal: - divisor, VI (97); - ideal, V-3
lie (to - above, below), III-I, d. 4 (45). (87); - quasicharacter, VII-3, d. 6
local field, 1-4 (20). (115).
locally constant, VII-2 (106). proper embedding, III-2 (50).
Q-Iattice, V-I, d. I (80). separably algebraically closed, I1I-2 (52).

quasicharacter, VII-3, d. 5 (115); principal similar, IX-3 (170).
-, VII-3, d. 6 (115). simple, IX-I, d. I (162).
quasicompact, VII-3, d. 5 (115). split (to - fully), VIII-5 (158).
quasifactor, IV -I (60). standard (function), VII-2, d. 1 (106), d. 2
(108), d. 3 (110).
ramification: groups of -, VIII-3 (149);
symbol: Hilbert - , XIII-5 (262).
order of -,1-4, d. 4 (15).
ramified, XI-I (202) ; cf. unramified; fully -,
Tamagawa measure, VII-2 (113).
1-4, d. 4 (15); tamely - , VIII-I (142).
tamely ramified, VIII -1 (142).
rank, II-2 (31).
topological dual, II-5 (38).
real (place), III-I, d. 3 (44).
totally disconnected, VII-3 (114).
reduced (norm, trace), IX-2 (169).
Tr, III-3 (53).
regular: (norm, trace), 1II-3 (53); - repre-
trace, III-3 (53); reduced -, IX-2 (169); re-
sentation, III-3 (53); ~- -, IX-3, d. 2
gular -, III-3 (53).
(171).
transfer, XII-5 (241).
regulator, V-4, d. 7 (94).
trivial: (- algebra), IX-3 (170); (- factor-
representation, P & N (XV); L- -, IX-2
set), IX-3, d. 5 (175).
(168), regular - , III-3 (53).
restriction morphism, IX-3 (173).
R-field, 1-3, d. 2 (12). ultrametric, 1-2, d. 1 (8).
right order, X-I (192), XI-4 (212). unramified, 1-4, d. 4 (15), VII-3 (117), XI-I
R-Iattice, II-4, d. 3 (35). (202), XII-2, d. 1 (222).
root of I, P & N (XV).
valuation, 1-3 (11).
self-dual (Haar measure), VII-2 (105).
separable algebraic closure, IX-2 (168). zeta-function, VII-6, d. 8 (129).
Spri nger-Verlag
and the Environment
We at Springer-Verlag firmly believe that an

international science publisher has a special
obligation to the environment, and our corpo-
rate policies consistently reflect this conviction.
We also expect our busi-

ness partners - paper mills, printers, packag-
ing manufacturers, etc. - to commit themselves
to using environmentally friendly materials and
production processes.
The paper in this book is made from

low- or no-chlorine pulp and is acid free, in
conformance with international standards for
paper permanency.

Basic Number Theory - Andre Weil PDF

Uploaded by

Copyright:

Available Formats

Basic Number Theory - Andre Weil PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Basic Number Theory - Andre Weil PDF

Uploaded by

Copyright:

Available Formats

Andre Weil was born May 6, 1906 in Paris.

Andre Weil's work laid the foundation for abstract

Originally published as Vol. 144 of the

Mathematics Subject Classification (1991): llR

ISBN-13: 978-3-540-58655-5 e-ISBN-13: 978-3-642-61945-8

Photograph by kind permission of Lucien Gillet

elP data applied for

SPIN 10701941 4113111 - 5 4 3 21 - Printed on acid-free paper

Basic Number Theory

Third Edition 1974

Springer-Verlag Berlin • Heidelberg • New York

Aiux., llpop.. .dsup..

The first part of this volume is based on a course taught at Princeton

Princeton, May 1967. ANDRE WElL

Foreword to the second edition

Princeton, December 1971. ANDRE WElL

PART I. ELEMENTARY THEORY

Chapter I. Locally compact fields. . . . . 1

Chapter II. Lattices and duality over local fields 24

Chapter III. Places of A·fields . . . . . . . 43

Chapter IV. Adeles . . . 59

Chapter V. Algebraic number-fields. . 80

Chapter VI. The theorem of Riemann-Roch 96

Chapter VII. Zeta-functions of A-fields . . 102

Chapter VIII. Traces and norms . . 139

PART II. CLASSFIELD THEORY

Chapter IX. Simple algebras. . . . . . . . 162

Chapter X. Simple algebras over local fields 188

Chapter XI. Simple algebras over A-fields 202

Chapter XII. Local classfield theory. . . 213

Chapter XIII. Global classfield theory. 244

Notes to the text. . . . . . . . . . 292

Fermat (1601-1665) Riemann (1826-1866)

As morphisms of fields into one another are assumed to be "unitary"

of quotation, and not as an indication of their importance. When a

Locally compact fields

(1) q"-I=q-l +L q" - 1 ,

LEMMA 1. If K is a commutative field, every finite subgroup of K x

J f(x)d rx(x) = d·r (JG' f(x + y)d rx'(y»)d rx"(x);

In particular, for any locally compact field K, we define modK(a) to be

PROPOSITION 2. For all m> 0, the set Bm of the elements x of K such

Let V be a compact neighborhood of 0 in K; let W be a neighborhood

show that it is an isomorphism, it is enough to prove that f- 1 is conti-

As (x~, ... ,x~) is in S, we have v'ef(S); as veaW, we have v'eyW with

with aeKx; (c) mappings ofthe type

(Xl'X 2 , ... , X,,)-+{Xl

hence the conclusion.

LEMMA 3. Let F be a function on the set N of natural integers, with

b=sup(f(O),J(l), ... ,J(1l-1»).

and therefore, in view of (4):

As nr~m, i.e. r logn~logm, this gives

Now replace n by nk; for k-+ + 00, we get

LEMMA 4. Assume that F is bounded, i.e. that mod K is bounded on

modK(x+ y)2"~A"+1 SUPi (F (e")) mOdK(X)imodK(y)2"-i).

defined in this way on Q is known as the p-adic valuation on Q; clearly it

§ 4. Structure of p-fields. In this section, p will be a prime and K will

THEOREM 6. Let K be a p-field; call R, R x and P the subsets of K

Then K is ultrametric; R is the unique maximal compact subring of K;

isomorphic to the dual of GIH. As H is open in G if and only if GIH

(v, w)v= (IX(X)V,V)v= (V,V*IX(X)v