24 DR. H. F. BAKER [Jan.

12,

ALTERNANTS AND CONTINUOUS GROUPS

By H. F. BAKER.

[Received January 8th, 1905.—Read January 12th, 1905.]

THE present paper is concerned with the proof of a fundamental pro-

position of non-cominutative algebra, and with shewing that this leads at
once to various results fundamental for the theory of continuous groups
which are usually developed only with much more detailed considerations
than are here employed.
One of the most noticeable peculiarities of the algebra of two non-
commutative quantities A, B is the occurrence of the alternant
{A, B) = AB-BA ;
from this we may form the alternants [A, {A, B)), (B, (A, B)),
and so on, each new quantity being the alternant of a previous quantity
with either A or B ; and, if these be called simple alternants, we may
proceed to form alternants of simple alternants; it is, however, easy to
see that these are expressible by simple alternants with numerical co-
efficients. Now it is an obvious suggestion of Lie's theory of groups
that the product

is of the form ec, where C is a series of alternants of A and B. For, if,

in -the ordinary notation,
X = exXx + ...+erXr) X' = e'1Xl+...+e'rXn X" = elXl-\-...+elXn
we have
exex'xt = exft(x, e') = /«[/(«, e), «'] = /,(«, e") = ex"xif
and the ground for this simplification can only be the alternant relations
(Xp, -A.tr) = 2jCpcrrX.T.

The fact that the product eAeb is capable of the form eu has been insisted
011 by Mr. Campbell in his recent interesting book on groups, and he has
k'iven an elementary proof of it (Proc. London Math. Soc, Vol. xxix., 1898,
p. 14), which cannot, however, I think, be regarded as final. It follows
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 25

from this that the terms of any the same dimension in A and B
in the expansion of
(eAeB-l)-%{eAeli-l?+$(eAeB-l)ii-...
must be simple alternants or sums of such with numerical coefficients;
and to the fourth dimension we, in fact, find for this expansion
, B)+&(A-B, (A, B))-&(A, (B, (A, B))).
On the other hand, starting with Schur's formulae for the first
parameter group, and using the notation oi the theory of matrices, I have,
in several papers in these Proceedings, deduced from the theory of groups
various propositions primarily for the parameter group; and, in
particular, if e, = ^<r(e, c'), which we shall write e" = \fr(e, e'), be the
equations of that group, with canonical variables e, e" and canonical
parameters e', if E be the matrix whose general element is
EP<T = 2 cC7per (p, or, T = 1 , 2 , ..., r),
and A = eE, while E', A' and E", A" depend respectively on e' and e", as
E, A depend on e, it has been shown {Proc. London Math. Soc, Vol. xxxiw,
1901, p. 98) that A" = A'A,

leading {ib., Vol. xxxv., p. 338) to

E" = ( A ' A - l ) - i ( A ' A - l ) 2 + . . . .
The present paper professes to show that these investigations for the
matrix E, so long as they refer to the general group of unassigned order,
have a much wider bearing, and that, in fact, they furnish a proof
of the general theorem eAen = cc, of a simple character, and so also of the
fundamental group property which arises when, as in Mr. Campbell's work,
A and B are linear differential operators. But there is more than this :
we have proved (ib., Vol. xxxrv., 1902, p. 352) that the E matrix of the
row Ee' is the alternant (E, E'); hence, if Ev ..., Em be each E matrices,
either of e, or e\ or of some other row, the E matrix of the row
E1E2...EmEe' is the alternant (Ev (E2, ..., (Em, (E, E'))...)); thus every
alternant of E matrices is an E matrix. From this it follows that the
equations of the first parameter group are of the form

where Hn is a homogeneous polynomial of dimension n in the matrices

E, E'—in itself a remarkable result, since this is then the parameter
group of every possible continuous group when the appropriate matrix E
26 DR. H. F. BAKER [Jan. 12,

is employed—but it also follows that every term in this expression gives

rise to one of the terms in the expansion of

and that there are no terms in this latter but such as arise in this way.
The series for e" is manifestly simpler to deal with than the series of
alternants : in this paper it is obtained without the assumption of any of
the propositions of the theory of groups, or the use of the theory
of matrices; and a law of recurrence for its terms, whereby any term is
deducible from the preceding, is obtained. The result can then conversely
be used to establish the existence of general functions possessing the
group property, and in particular it gives incidentally the origin of
Schur's formulae. The method employed is of the kind usually called
symbolical, the quantities being defined by their laws of operation and not
belonging to the ordered aggregate of natural number; if it be not con-
ceded that sufficient explanation is given to shew that these laws are in
this case self-consistent, the work retains its validity when interpreted
in terms of the theory of matrices; but it is claimed that the paper is a
contribution to the calculus of alternants of any non-commutative
quantities.

1. Alternants and their Bases.

The capitals A, B, G, ... denote any quantities which can be added
and subtracted associatively and commutatively, and can be multiplied
associatively and distributively, but not commutatively. We do not
recognise the existence of solutions of the equations xA = 1 or Ax = 1.
With each capital, A, we associate a quantity, a, which we call its base,
and we call A the derivative of a, either determining the other uniquely.
By the base of a sum or difference of derivatives we understand the sum
or difference of their bases; thus the base of —A is —a, and the base of
zero is zero, and conversely; and bases can be added associatively and
commutatively. Further, with every alternant of capitals we associate a
definite base of which it is the derivative; and a derivative is either a
capital or an alternant of capitals. As we have not as yet introduced
products of capitals and bases, it is legitimate to denote the base of (A, B)
by Ab ; the identity (A, B)-\-{B, A) = 0 will then require Ab-\-Ba = 0;
in particular Aa = 0. The base of an alternant {A, (B, G)) will then
naturally be denoted by Ad, where d is Be, the base of (B, C), and so may
be denoted by A (Be); we shall, however, denote this by ABc; it being
observed that the notation Ab has as yet been introduced only for the
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 27

case when A is a single capital, and b a base, there is no contusion

possible of ABc with the symbol AB.c regarded .as the base of {AB, C) or
ABC—CAB. In general we assign similarly, to an alternant
U i , U 2 , {..., A , , , B) ...))),
a base denoted by AXAZ... Anb; if c = AhAh+\... Anb, the alternant
takes the form (Av (A2, ... (Ah_i, C)...)), and its base has the notation
Ax... Ah-iC; so that in the product Al...Allh the symbols are
associative.
We shew next that this notation for the base of a simple alternant
may conveniently be extended to the base of a compound alternant; as a
compound alternant is expressible by a sum of simple alternants, its base
is a sum of bases of the form considered; we shew, however, that its
expression in a succinct form is equivalent to allowing the symbols in the
base of a compound alternant to obey a distributive law. First, the
alternant (04, B), C) is easily verified to be the sum of {A, {B, G)) and
—CB, (A, O), of which the bases are respectively A Be and — BAc ; if then
the base (A, B)c be assigned to ((A, B), C), we have
U, B)c = ABc—BAc or (AB-BA)c = ABc—BAc,
as is most natural. In passing we notice that the base of
U,B)t'!) =-(C,(A,B))
is also denoted by —CAb; so th ,t we have
ABc+BCa+CAb = 0.
In general, if A = (Av (A2, ... (An-\, An) ...)) be any alternant, and B be
a single capital, we denote the base of (A, B) by Ab, and we examine now
the implications of this notation. Suppose, in expanded form,
A= 2 ±K1K2...Kn,
where K^^... Kn is a permutation of AXA2... An, so that
Ab = (2±K1K2...Kn)b.
Any term KiK2... Knb is the base of an alternant (jBTlf (K2i... (Kn, B) ...));
we show that
(A,B) = 2±(iST l t (tf2, ...(Kn,B) ...));
when this is proved we can write
(X±K1K2...Kn)b = 2±(K1K2...KHb).
Let Pn_x denote (A»Ua, ... U*-x,Atd •••)); then (A,B), = (U1,PW_,),.B)>
is, as just remarked, equal to (Av (Pn-u B))—(Pn-\, (Alf B)); denote
28 DR. H. F. BAKER [Jan. 12,

(Av B) by C, and assume the law of expansion just stated, which we wish
to prove true for an alternant (A, B) of (n-f-1) dimensions, to be true for
alternants (Pn-u B), (P n -i, O, of n dimensions; we have remarked that it
holds for n = 8. Supposing
Pu_i = 2 + Hl... Hn-i,
we have thence (Pn.u B) = 2± (Hv (... (#„-!, B)...)),
(P..,, C) = 2 ± (fflf (... (fl,.,, O ...)),
and thus
(A, B) = (Ax, (?„_,, £))-(?»_!, (7)
= 2 ± (Alt (ff lf (... ( H U , JB) ...)))-2 ± (fflf (... (Hn-i, UVB)) ...)),
while .4 = (Alt P.n-{) = 2 ± ^ H x . . . F n - i - 2 ± JBTj... ffll_1^1;
so that the result is established.
More generally, let B = (Bv (J52, ... (J5,,t_i, B»i)...)) be any alternant,
of which the base is B1B2... Bm-\bm. When A was a single capital, we
agreed to denote the base of (A, B) by ABX... BM-ibm; it is natural to
seek to extend this to the case when, as above,
A = (Alt (A2, ... (An-u An)...)),
and to denote the base of {A, B) by (2 ± K1K2 ... Kn)B1B2... Bm-Xbin ; in
fact, the proof given above for the expansion of (A, B) when B is a single
capital holds equally here, and we have
2 ± {K^Kt... KHBX ... £»_ibm) = (Z±K1K2... Kn)BxB2... Bm^bm.
Again, if C = (C^ (C 2 ... {Ch-i, Ch) ...)), the equation
2 + CiC2--. GhK-^K^... KnB1... Bm-\bm
= ClCi...Ch(2±K1...Kn)Bl...Bm_1bm
= CxCa ... Cfc2 ± {Kx ... Z , ^ ... JB,.! bm)
requires the obvious identity
2 ± (Clt ( C 2 . . . (Ch, (K,... (Kn, B)...))...))
= (Cv (C2... [Ch, ( 2 ± (Kv (... (Kn, B) ...)))] ...)).
Thus in a product or sum of products of the form Ax... An-i(in, where
Av ..., An are single capitals, when the sum is of the form
Ax ... AhPAk ... An-ldn,
where P is an alternant, the symbols obey the associative and distributive
laws. The same is true when Ax, ..., An are themselves alternants, as
may be similarly shewn. For instance, returning to the theorem of
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 29

expansion proved above, the base of (A, B) or {{Av (A2... (An-i, An) •••))> B)
may equally be denoted by —BA1 ... An-\an; let pn_i = A2A3 ... /4 n _ia n ,
so that the derivative Pn-i is as above ; then

—BAX... An.xan =

the alternants shew, as in the work above, that this is the same as

A1Pn-ib+Pn-iBal = A^n^b-Pn-iA^;
but the base of (A, B), = ({Alt P»_i), B), is (AlfPu^i)b; thus we have
(AxPa-i—Pn-\AJ b equal to A^P^-ib —Ptl-\Axh. And so on. And by
the way we notice the identities such as

BAl... An-iOn+2 ±Kl... Kvb = 0,

the general form of which is Ab-\-Ba = 0, including Aa = 0, where

A, B are any two derivatives, that is, either single capitals or alternants
of capitals.
It is manifest that the notation of bases, once established, as here,
furnishes a compendious way of expressing the relations among alternants,
and may be of great use in expressing a compound alternant by simple
alternants.
Many identities arise from the relation Ab-\-Ba = 0, by supposing a
to be a base of the form AXA2... Amam+i, and b a base of the form
BXB2... Bnbll+i, where Alt ..., Am+i, Bx, ..., Bn+\ are any derivatives.
^Restricting ourselves, as sufficient for purposes of illustration and as
giving results to be utilised below, to the cases when each of Av ..., Am+i,
Bv ..., Bn+\ is either A or B, we find on examination that the following
are all the identities so obtainable up to those of the sixth dimension.

(1) From Aa = 0, putting Ah for a, we have (AB—BA) Ah = 0.

This corresponds to the alternant relation ((A, B), (A, B)) = 0, which is
the same as (A, (B, (A, B))) = (B, (A, (A, B))).

(2) Also from Aa= 0, by putting A2b for a, we have

U,lA,B))A*b = 0,
that is (A B-2ABA+BA2)
2
A2b = 0,
or (A3B-2ABA2+BA3) Ab = 0.
This corresponds to th€ identity
UAA,B)), {A, (A,B))) = 0,
80 DR. H. F. BAKER [Jan. 12,

which is equivalent to
(A, (A, (B, U, (A, B)))))-2W, (B, (A, (A, {A, B)))))
+(B,(AAA,(A,(A,B)))))=0,
an identity which we may agree to denote by
[l 2 21 9 2]-2 [12182] + [21*2] = 0.
By interchange of A and B we derive
(AB*-2BAB*+B*A)Ab = 0.

(8) From Ab+Ba — 0, putting A*b for a, BAb for b, we derive

(A, (A, B))BAb+(Bf (A, B))A*b = 0,
which is the same as
(A2B*-B2A*+2BABA + BA2B-2ABAB-AB*A)Ab = 0,
or, in virtue of (AB—BA) Ab = 0, the same as
(AW-BW+dBA^B-SAB^Ab = 0.
This corresponds to the identity
((A, (A,B)), (B, U , J5)))+((B, U , B)), W, U , B))) = 0,
which expanded is the same as
[I a 2 a 12]-[2 2 l 3 2]+8[21 a 212]-8[12 2 l a 2] = 0.
The identities
S(A,B)Ab= (ABB+BA*-2ABA*)Ab = 0,
T(A,B) Ab = W a JB a -BM a -h3JB^l a B-8^B a ^) 46 = 0
are not really independent; for we find
S(A +B, B)Ab = \_S(A, B)-S(B, A) + T(A, Bf]Ab,
T(A+B,B)Ab = [T(A,B)-2S(B,A)]Ab,
while {A+B)b = Ab.
In what has preceded we have considered only bases of the form Ab,
.vherein A and B are alternants (or simple capitals); and, if
A =2±KlKi...Kn,
it has been shown that Ab = 2 ± {KlK^... Knb).
If now P l f P2» • • • b e simple capitals or alternants, it is natural to regard
(PlP2... Ph-\-Ph+\Ph+z •" Pa+...)& as a n appropriate notation for the
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 31

which is the sum of the bases PlP2... Phb, Ph+iPh+2... Pk.b, ... ; so
that its derivative will be
(Pv (P2, ... (Ph,B) ...)) + (P/l+i,(Pfc+2, ... (P,, £ ) . . .
When PxPg... Pfc+Pzi+i-Pfc+2 •••-Pjt+--- is an alternant we have shewn
this to be legitimate ; we extend it to the general case, for which there is
no corresponding condition for alternants satisfied, the derivative not
being equal to ( ( P j ... pfc + p m ... P H _ . . . ) , B).

this expression, in fact, not being an alternant of alternants, is not under

consideration.
A particular case of this definition is

where A is a derivative ; this we regard as the base of

C = \B + \iAt B) + \2G4, (A,
If A*o> Mi> • • • be such functions of Xo, Xv ... that

we have b = (yuo+Mi-44-M2^2H"- ••) c>

and the substitution in the series above given for C of the series

reduces that series to C.

2. Of the Substitutional Operation.

We introduce a symbol to express the operation of replacing a base a
by a base b, and write

the operation of replacing the capital A by the capital J5 is also con-

sidered, and may be denoted by the same symbol, so that

and, if P be a product of capitals, some of which are A, by the effect of

the operator upon P is meant a sum of terms each differing from P by the
substitution of B for A in one factor without change of order ; for instance.

ABA = B2A+AB2, (b£\ A2Ca = BACa+ABCa + A*Cb.

32 DR. H. F. BAKER [Jan. 12,

Denoting (b ^-j by 8, and the operation of forming the derivative A

of a by e, so that ea = A, we have
Sea = SA = B = eb = eSa.
Let the alternant {A, {A, ... {A, O)...)), where A occurs m times, be, for
brevity, denoted by {AmC); then
8e(Ac) = 8(AC-CA) = BC-CB - e(Bc) = eS(Ac),
m
8e(A c) = 8(AniQ = (B,(An.,C))+(A, (B, (Am^O)) + ... + (Ain.l(B, C))
l } w 2 m i
= e(BA "- c+ABA - c+...+A - Bc) = e8(Amc),
6e{A*DA"c) = 8(An(D(AmQ)) = J 1 (iUD(il m 0))+a a (i4,(DM J .Q)) >
where 8X operates on A occurring under An, and 82 on A occurring
under Am,
) = eS{A"DAwc)\
and so on. Thus, if F{A) be a symbol for the sum of a finite or an infinite
number of products of capitals, of which each product involves A, and the
other capitals in the product are unaffected by o, we infer that
SeF(A)c - e8F{A)c.
We shall presently be concerned with a base b which is a particular
case of the form ^(^4)^, namely, is
b= {\Q-\-XlA-\-X2A2-\-...)c,
theu
and 82a = (b | - ) 2 a = 8b = [\l

is also, clearly, on substitution for B, of the form F(A)c, and the same is
similarly true of every expression 8ma ; thus
e&a = e8(8a) = eSb = 8e(8a) = S(e8a) = 8 (Sea) = 82ea,
e8*a = e8(S2a) = 8e(S?a) = SieS^a) - S»ea,
and so on ; and in general e8ma = Smea.
Now, with the same b, let t be an ordinary number, and put

so that

Put, further. V =
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 88

so that, after substitution for F and arrangement according to powers of t,

We have ^ = [<S+*<Sa+...],4 = i[l + tf+...]A = SF,

and thus

= »18F+H2(8F.F+F.8F)+VL3(8F.F2+FSF.F+F28F) +

from which (£g) = (<TV),=o = a"V,.o.

Therefore, the substitution

in

where Vo = /xo~ '

Here b is any series of the ft in

We shall presently utilise this result. It is convenient to give here

another particular result. If b reduce to Ac, and F(A,C), or simply F,
denote any sum of terms each of which is a product of powers of A and
C, with a numerical coefficient, we have, as we shall now prove,

) F(A, C) = FC-CF = (F, C).

oal
In particular, if F (A, G) reduce to A,

To prove the result, notice

S) A2 = (A, QA+A U, C) = A2C-CA2 = (A2, O,

and, if {b^)H= {H C)
> > {b ^ ) K = {K C)
' '
8BB. 2 . VOL. 3 . NO. 8 8 4 .
84 DR. H. F. BAKER [Jan. 12,

then (b—^HC1 = Cm(H, QCn = (THC^-CT^HC* = (C"HC*, ®

and (*>^)HK = (#> OK+H(K, O = HKC-CHK = (HK;Q.

Another result to be utilised follows hence. We have

(AzO = (A,U,O) = A2C-2ACA + CA*,
and in general, easily, by induction,

and so

(AXO

and thus, if <j>(A) = v o +j/ 1 i4+»'ih4 2 +..., b = Ac,

where 0 ' U ) = vx+%<iA+%v<iA*+...> 4>"{A) = 2»/2+S.2.>v4 + ....

This is the same as

and this result, though proved for B = (A^C), can easily be shown to hold
for arbitrary B. A particular case, given by Mr. Campbell, Proc. London
Math. Soc, Vol. xxix., 1897, p. 16, is
BAm~1-\-ABAm-2+...-\-Am-1B =
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 85

8. The Exponential Theorem.

Applying the theorem just obtained for arbitrary A and B,

to the case where for <f>{A) we have

A =

so that each of <f>'(A), <f>"(A), ... is also A, we obtain

Now put, with arbitrary a',

where rsx = ^, BT2 = ~~ ^j» • • • are

Bernoulli's numbers with associated
signs; then,as

2! ' 8!
/ . A2
is the derivative of I * ~^~ i7[ ~^~ q7 ~^~""

or d, we shall obtain

D , (A,B) , (A,(A,B)) r, _ A,
*-r—2P"1 3I " ~ A*
Thus, with the particular b so defined from A and ^4', we have

and hence (ft # Y " A = 4""A

Next, put "

and A"
p 2
36 DR. H. F. BAKER [Jan. 12,

then, as follows from a result proved in the previous section (p. 33),

which we appropriately write A" = A'A;

this is then the consequence of

where, a and a' being arbitrary,

, /- . A . A2 . N"1
ft= 1+
( 2l+8!+)
This, however, gives

where [BA™-1] = BAm-1+ABAm-2+...+Am~1B,

so that (& ~-) b is of the form F(A, A')a', where F(A,A') is a series of
polynomials in A and A' with numerical coefficients. So therefore is
every term (b —) a. Hence, as A'a' = 0, we have an equation

a"= a+a' + (H0+H1+H2+H3+...)Aa',

where Hr is a homogeneous polynomial in A and A', of dimensions r.
As A", the derivative of this, is a series of alternants, we infer therefore,
with arbitrary non-commutative symbols A, A', if

A = l+A+ +
and the expansion of

be arranged according to homogeneous polynomials in A and A' of

ascending dimension, that each polynomial is an aggregate of alternants
of A and A', any one such aggregate being the derivative of the terms of
the same dimension in
a" = a+a' + (H0+Hl+H2+...)Aa'.
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 87

For instance, as will be presently seen,

and we find on trial, as was remarked in the introduction, that to the

fourth dimension the expansion in question is
A+A'-l(A, A')+T\(A-A', (A, A'))+^{A, {A', (A, A'))).
The theorem has a converse. When A, A' are perfectly general there
exists no other derivative An such that

is equal to A'A.
For otherwise A n = A", An1 A" = 1, and so, if we put
i,(a, a') = a+a'+(H0+Hl
and denote by Ao the derivative of \^(a", — a u ), we shall have Ao = 1,
namely,

we s
or, on multiplying by 1—£^0+Hy^o-H•••» ^ a ^ ^ a v e ^0 = 0»
\fs(a", —an) = 0. The only solution of this which holds for all values of
a" must be one for which the terms of any order are separately zero; in
particular, from the terms of the first order a"—an = 0 ; while, con-
versely, if this be so, A" = A n .

4. On certain Properties of \Js(a, a'), and the Determination of its

Coefficients.

The actual form of \Js(a, a') can be determined from its definition

it is at once seen, however, that beyond the first few terms this is an
intricate process, and, moreover, the terms then appear arranged according
to ascending dimension in a' only, and not, as will appear more proper,
according to ascending dimension in a and a' jointly.
Some information appears on the face of the series. First we have
i/r(—a', —a) = — \^(a, a').
88 DR. H. F. BAKER [Jan. 12,

For A-'A'-'A'A = 1, or A""1 = A~ l A'-\ namely, as

Alii
A""1 = l—A"-\--
2!
l l
we have -A" = (A- A'- -l)-% (A^A'" 1 -!) 3 *....
Thus in the series
\lr(a,a') = a+a'+(H0+H1+H2+...) Aa'
Hin is symmetrical in A and A', while iZ"2ft_i is changed in sign by inter-
changing A and A'. Next, the terms of the first dimension in a' in

thus # 0 = — £, while if2n has no terms in A2n or A'2n, and #2n-i contains
the terms ^ U a n - 1 - ^ ' 2 n - 1 ) . This gives H1 = ^2(A-A')i and tf2
= numerical multiple of AA'-\-A'A ; we have seen, however, that
{AA'—A 'A) A a' = 0, so that H2 is a numerical multiple of AA'.
To obtain further information we prove first two lemmas: (a) if a, a'
be arbitrary and c = Aa', then C = A-4'A"1, and hence

^ + . . . = AA'A"1;

(/3) we have >ff(a', a) = A^(a, a') = \^(a, Aa'), which are both included in
^(a, AV) = A*-lf(a', a).
The result (a) is obvious on forming the expression

in which the terms of dimension (n-\-l) are

An An~l A An~z A2 A11
A At_ •"• A' \ A' 4-(— 1VM'

which is —, (AnA'), the derivative of ——. Or it follows thus : From

n! n\
c = Aa', we have
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 89

where b = Aa'; by previous theorems (§ 2), this is

A = i

so that C = Ail'A"1. To prove the result \fs(a',a) = Ax/r(a, a') of

lemma (/3) we have only to notice that, if a n = Aa", then A n = A A" A"1,
and so, if a" = \j/(a, a'), and therefore an = A\/r(a, a'), we have
A u = AA'AA"1 = AA',
which establishes au — yjs{a', a). The same proof stated differently
gives, from a u = A\js (a, a'), the series

which shows that a n = \Jr{a\ a). The result ^(a, Aa') = Ai/r(a., a') of
lemma (/S) is obvious from the form
i,(a, a') = a+a'+(J?0+£Ti-f Ht+...)Aa'.
For consider any term P = AkA'K>AW'... Aa';
if herein we replace a' by Aa', and therefore -4' by AA'A'1, it becomes
i*AAi4'x#A-1il'1Ai!"1'A-1... A Aa',

or, since A and A. are commutable,

or A P : while a-\-a' becomes Aa+Aa', it being noticed that in virtue of

Aa = 0 we have Aa = a ; thus ^(a, Aa') = Ai/^(a, a').
Consider now the equation
—\fr(—a, —a') = Ax/r(a, a');
it gives
a+a'-(H0-Hl+Hz-...)Aat

comparing here terms of second dimension, we have

-H0Aa' = 2J 0 .4a'+,4(a+a') = (H0+l)Aa',
so that iZo = — £; comparing terms of third dimension,
HxAa' = H.Aa' ^
40 DR. H. F. BAKER [Jan. 12,

giving also Ho — — \ ; comparing terms of (n-f 2)-th dimension, we have

thus when n is even, = 2m, say, we obtain

( ^ ) ] ' = 0,
from which HimAa' can at once be calculated when all the preceding
terms are known ; for n = 2 m + l we obtain a less convenient equation in
which the term of highest dimension is again the term involving if2.»; if,
however, in the equation just obtained we interchange a and a', and
subtract the result from this equation, we obtain

we proceed to show that this equation determines H2m-.i Aa' when all
preceding terms are known. To explain this, let Ka denote what Hn
becomes when A—A' is put for A ; if Ktl were determined, Hn could then
be found by putting A-\-A' for A ; we have thus, since (A—A')a' = Aa', •

• U-AT-A12,, u-A')**-A'*~

XAa' = 0;
this gives AKim-iAa,' when preceding terms are known; we say two
solutions for K-2M-\ are not then possible; for, if so, and P were their
difference, there would exist an equation APAa' = 0, in which P is an
integral polynomial in ^-1 and A'; we have remarked the existence of
identities QAa' = 0, in which Q is such an integral polynomial; we say
that Q is incapable of the form AP; for let PAa' = f, so that APAar= 0
is equivalent to Af= 0 or AF = FA ; let F = AhG, where G is incapable
of a form AGX, in which Gx is an integral polynomial in A and A'; then
Ah+1G = AhGA, clearly an impossible result, since the highest powers
of A occurring as left-side factors are different for these two quanti-
ties.
As illustrations of these formulae we proceed to calculate Hu H2,
H3, Hv For Hl we have

^ ^ ~\a' = 0,
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 41

or, in the notation above,

in virtue of (AA'-A'A)Aaf = 0, this gives HxAa' = ^(A-A')Aa'.

For JET, we then have

( ^ 4 ) ] ^ a ' = 0,
and hence at once H2Aa' = -£$AA'.Aa'. Hence we have for Kz
U-A'f-A'2 (A-A')A' 3 13
-j +. (A-A') +A A-2A'
—

Taking account of (AAf—A'A)Aa' = 0, this is found to be

[720AK3+Ai+3A3A'-4:A2A''i+4AA's-3AAr2A-AA'A2+Q']Aa' = 0,
where QAa' = [4A'AA'2-8A'A2A'-A'A3+A'2A2-2A'3A]Aa!,
which must then be capable of the form APAa'; in fact the previously
given identities (§ 1) lead to
W3A-2A'AA'2)Aa' = -AA'aAa'}
Ut2A2-3A'A2A')Aa' = (A2A'2-3AA'2A)Aa',
A'A*Aa' = (-A
and QAa' then takes the form in question; putting then A-\-A' for A, we
find
H3Aa' = -Y
in which Hs is changed in sign when A, A' are interchanged, and has
CT2/4!for coefficient of A3. Lastly, for H4 we obtain, on substituting for

[i '2-§AA'2A-(lAA'A2-AA'3']Aa' =0;
putting here, from the identities in question,
(A3A'-2AA'A2)Aa' = - A'A3Aa',
and (2A2A'2-6AA'2A)Aa' = (A2A'2-3AA'2A+A'2A2-3A'A2A')Aa',
42 DR. H. F. BAKER [Jan. 12,

we find
i 2 2 8 8
HtAa' = Tjtv[A*A'*+A' A*-SAA' A-SA'A A'-AA' -A'A ]Aa',

wherein if4 is symmetrical in form. We have therefore, to the terms

found,
a" =

By taking the derivative of this we have therefore the expansion by

simple alternants of H(a, a') = (A'A—1)— £(A'A—l)2+---> up to terms
of the sixth dimension in A and A'. And the law of succession for the
calculation of the terms of \[r(a, a') corresponds of course to a law for the
terms of H(a, a'). In fact, H(a', a) = —H(—a, —a') = AH(a,a')A~l, say

giving, if we notice (AmC) = 2 (—1)* ( ) Am~*CAs, on equating terms

3=0 \S /
of dimension r,
{-ir*Cr = Cr+U, Cr-l)+ ^ (A,(A, Cr-j)+ •.. + J^^ Wr-1 C,).

It is manifest, however, that the application of this law is more laborious

than the rule we have followed for the bases of the alternants.

In what has preceded the derivatives A, A' have been perfectly general.
But the fact that the series H{a, a') is obtainable by forming the derivative
of the series \fs{a,a') remains true when A, A' are less general. One of
the most natural ways in which we may allow restrictions is to suppose
that the powers and products of A and A' are also derivatives. We carry
out a particular example of this suggestion, of a very simple kind, which
is yet general enough to give a great deal of information in regard to the
expansion of y^{a,a'). -We suppose A2 to be a derivative, in fact equal to
/J.A, where fi is a number; and similarly A'2 = ft'A'. Then A2a' — pAa',
and therefore
{A, {A, A')) = fi(A, A'), or A*A'-2AA'A+A'A* = n(AA'-A'A),
which gives fxA'A =AA'A; and similariy /JL'AA' = A'AA'. It is con-
sistent with these to suppose A A' = fxA' and A'A = n'A. These four
 1905.] ALTERNANTS AND CONTINUOUS GROUPS. 43

equations give, if B = pA+qA', whatever numbers p and q may be,

JB3 = {pix-\-qix')B, and conversely. It is then easily seen that

and we have

say, where X is a number, giving A" = (1—\H')A-\-(1-\-\M.)A'. On the

m m x
other hand, from A = fx ~ A, we have A = l-\ -A, where
()

also, from the form A" = (1—X/x'M-Hl+A/xM', we have

AM = (n+»')A",
A
and A» l+ "

putting this equal to A'A, 'v>, have

which are consistent in virtue of

— = - £ _ + -iiL- (l+JL)
u.') cr(u) -O*') \
<T(U') criu))'
t+ x
or ' '' 1 ^ l
and give X; namely, in this case Hn (ju., n') consists of the terms of order m
in the expansion of [o-Ou')]"1 < r ° A + ^ - ( r f c ) . We find

<T\JX )

thus we have at once the values of if0, -ff^ H2 and infer, putting down
for H3(A, A'), H^Ay A') the most general possible respectively un-
symmetrical and symmetrical polynomials of dimensions three and four
certain relations among the numerical coefficients in these.
44 DR. H. F. BAKER [Jan. 12,

An important Property of the Series \fs{a, a').

Denoting l - M + - 4 a / 2 ! + . . . by Aa, we manifestly have

Aa" Al^(f(ia') = Aa» A a 'A 0 = A ^ 0")An,
leading to \f^[\^(a, a'), a"] = \fs[a, \fs(a', a")\
which can be verified directly from the series \}/(a, a').

5. Some Applications to the Theory of Groups.

The applications we shall notice are of two kinds : the first, that which
was in view in Mr. Campbell's paper referred to in the introduction to this
paper, when the capitals A, B, ... are linear operators of the form

X = ii(xx ... <Cn) 3— + . •. +&(&! • • • *n) 3— J

OX i OXn

the second when the capitals are certain matrices. As regards the first,
it is easy to show that, if

where t is a number, then, at least in a neighbourhood where the analytic

function F(xt... xn) is regular, a similar condition being assumed for

F(x\... x'n) = (l + tX+ -^ X*+...) F(xx ... *„)•

if then we put Fx(x) = F(etXlx) = etx'F(x),

where Zx is one linear operator, and define similarly F 2 from another
operator Z2, the exponential theorem proved above gives
etxletx, _ etxs>

where tXB = (e^etx>-1) -£ (e1**etx*-l)a+...

is a series of alternants and therefore also a linear operator, and therefore
FB(x) = etx>etx>F(x) = etx>F2(x) = F^e'^x).
In particular, putting F(x) = xu and, with linear operators Pv ..., Pr,
X, = '
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 45

where eXt ..., er, e'u ..., e'r are numerical constants, we have, if
e*1 Xi = fi («. e), eXl Xi = /»(a?, e'),

the equation e^'Xi =fi[f(x, e), e'].

If, further, every alternant of two of the operators Pv ...,P r is expressible
linearly by P x . . . P r , say in the form
(Pp, P r ) = c p(rl P 1 +...+c p<rr P r 0>, o- = 1, 2, ..., r),
then X3 will be also so expressible, say in the form e"Pl-\-... + e','Pr, and
we shall have , r *, ,-• _ , „,

the fundamental equation of Lie's group theory.

Now we have seen (§ 8) that
X3 = ITi
and we have
(Xlt Za) = (e^+.-.+^Pr, «'iP!+...+«;Pr) = 2 p <F
T

(p, a; T = 1 , 2 , ..., ? ) .

We introduce now a matrix of r rows and columns whose general element

is
Epa = 2 CrTpeT (T = 1, ..., r),
so that each element is a linear function of ev ..., er with coetticients chosen
from the constants cp<rT; then we have, since from (P p , Pa) = — (P a , Pp)
follows cp<rT = — c,pT,

(XIf Xa) = - 2 2 ETXPr = -

T <r T

where Ee' is a symbol for the set of r quantities

(Ee')T = ETle[+ET2e'2+...+ETrer,
and clearly Ee' = — E'e, if E' be the same function of eu ..., e). as is £ of
g t ... er. Hence it follows, if {Ee')T be momentarily denoted by/ T ,
(Xv (Xv X2)) = -{e^+^.+erPr, yiPi+-..+/rPr) = 2 (^/)T PT)
(X2, (Xv X2)) = -(e[pl+...+*rpn /iPi+.-.+ZrPr) = 2(^'/) T P T ,
and therefore again, if (£'/) be momentarily denoted by gT,
(Xv(X2,(XvX2))) = (exP1 + ...+erPr, g1Pl + ...+grPr) = -2(Eg)TPT;
46 DR. H. F. BAKER [Jan. 12,

but Ef = EV, E'f= E'Ee', Eg = EE'f = EE'Ee'; thus we have

X3 = 2 (eT+e:-
r
or, since X3 = 2 eT'PT, we have r equations
T=l

eT = eT-\-eT—:
all of which are represented by
e" = g-f-e'4- f i-|-
and this is the series we have previously denoted by \fs(e, e'), there being
r functions xfs^e, e')t ...,\fsr(e, e'). Thus it appears that the fundamental
equations of Lie's group theory are
f.[ f(r A p'~\ — f['r \\AP P'\\

and we have proved, to employ the ordinary nomenclature, not only that
the equations of the parameter group have the form given by e" = ^-(e, e')
expressible by the matrices E, E', but that there is a one to one corre-
spondence between the terms of this, and the terms, which are alternants,
in the expansion of A n_-t _i( A i»'_i\2_j_

And this brings us to the second application of our preceding general

theory to the theory of groups. It is, in fact, easily recognised that the
laws fundamental for derivatives and bases which were set out at the
beginning of this paper hold for matrices E, such as have been here intro-
duced, and for the sets elt..., er. To see this it is only necessary to notice
(1) the fact proved above that Ee'+E'e = 0, (2) the fact that the E matrix
of the set Ee' is EE'—E'E. This last follows from the equation
EE'e"+E'E"e+E"Ee' = 0,
which expresses that relation among the constants of structure which
follow from the identity
(Xv (X2, X3))+(A-2, (X3,
for this equation is (EE'-E'E)e"+E"Ee' = 0.
It must be remarked, however, that it is only for groups of general form
that the equation E = 0 involves ex = 0, ..., er = 0, or e = 0. The group
must have no special transformations.
Conversely that intricate part of the algebra of the theory of groups
which is concerned with the theory of structure may conveniently be
stated in the general terms of the earlier part of this paper. As the
1905.] ALTERNANTS AND CONTINUOUS GROUPS. 47

theory is there set out, the primary set of capitals may be of infinite
number, and in any case the alternants formed from them of successive
dimensions form an indefinitely extended system from which there is no
return to the original capitals, save perhaps by reverting an infinite series
\B+\(A, B)+\,(4, U, B))+....
If we wish to have a finite system of elements, we may then naturally
suppose, to speak in terms of the bases of the alternants, that there is but
a finite number of bases, say av a2, ..., ar, in terms of which every other
base is expressible linearly with numerical coefficients. We shall thus
have, for instance, equations such as
ApCL* = — (cp<rl « ! + . . . + Cp<rr ft,).
It is at once evident that this leads to equations F(Ap)aa = 0, where
F(AP) is an integral polynomial in Ap with numerical coefficients ; and so
on. It is interesting to consider from this point of view such propositions,
for instance, as one, due to Gartan, relating to the necessary and sufficient
condition that a group be integrable; with our proposition that the E
matrix of Ee' is (E, E'), this condition is that the constants of structure
should be such that for some positive integral A we should have, for every
e and e', an equation {EE'—E'EY = 0.