F1 for everyone

Oliver Lorscheid

Abstract. This text serves as an introduction to F1 -geometry for the general math-
ematician. We explain the initial motivations for F1 -geometry in detail, provide an
overview of the different approaches to F1 and describe the main achievements of the
field.
arXiv:1801.05337v1 [math.AG] 16 Jan 2018

Prologue
F1 -geometry is a recent area of mathematics that emerged from certain heuristics in
combinatorics, number theory and homotopy theory that could not be explained in the
frame work of Grothendieck’s scheme theory. These ideas involve a theory of algebraic
geometry over a hypothetical field F1 with one element, which includes a theory of
algebraic groups and their homogeneous spaces, an arithmetic curve Spec Z over F1
that compactifies the arithmetic line Spec Z and K-theory that stays in close relation to
homotopy theory of topological spaces such as spheres.
Starting in the mid 2000’s, there was an outburst of different approaches towards
such theories, which bent and extended scheme theory. Meanwhile some of the initial
problems for F1 -geometry have been settled and new applications have been found, for
instance in cyclic homology and tropical geometry.
The first thought that crosses one’s mind in this context is probably the question:

What is the “field with one element”?

Obviously, this oxymoron cannot be taken literally as it would imply a mathematical

contradiction. It is resolved as follows. First of all we remark that we do not need to
define F1 itself—what is needed for the aims of F1 -geometry is a suitable category of
schemes over F1 .
However, many approaches contain an explicit definition of F1 , and in most cases, the
field with one element is not a field and has two elements. Namely, the common answer
of many theories is that F1 is the multiplicative monoid {0, 1}, lacking any additive
structure.
Continuing the word plays, the protagonist F1 is also written as Fun , coming from the
French word “un” for 1, and it enters the title of the paper “Fun with F1 ” of Connes,
Consani and Marcolli ([CCM09]). In other sources, F1 -geometry disguises itself as
“non-additive geometry” or “absolute geometry”.

Content overview. The first chapter 1 of this text is dedicated to a description of the
main ideas that led to F1 -geometry. We explain the two leading problems in detail: Tits’
dream of explaining certain combinatorial geometries in terms of an algebraic geometry
over F1 in sections 1.1 and the ambitious programme to prove the Riemann hypothesis
in section 1.2.
1
2 Oliver Lorscheid

The second chapter 2 serves as an overview of the manifold approaches towards F1 -

geometry. We will describe two prominent theories of F1 -schemes in detail: Deitmar’s
theory of monoid schemes in section 2.1 and the author’s theory of blue schemes in
section 2.2.
The third chapter 3 summarizes the impact of F1 -geometry on mathematics today.
We spend a few words on developments around the Riemann hypothesis in section 3.1,
describe in detail the realization of Tits’ dream via blue schemes in section 3.2 and
outline promising recent developments in tropical geometry that involve F1 -schemes in
section 3.3.

1. Conception: the heuristics leading to F1

The first mentioning of a “field of characteristic one” in the literature can be found in
the 1957 paper [Tit57] by Jacques Tits. The postulation of such a field origins in his
observation that certain projective geometries over finite fields Fq with q elements have
a meaningful analogue for q = 1. Tits remarks that these latter geometries should have
an explanation in terms of projective geometry over a field with one element.
In the modern literature on the topic, the analogue for q = 1 is often called the “limit
q → 1”, a notion that finds its origin in the connections to quantum groups where q
occurs indeed as a complex parameter.
For instance, the group of invertible matrices with coefficient in Fq converges towards
the symmetric group Sn on n elements,

GL(n, Fq ) −−→ GL(n, F1 ) = Sn ,

q→1

compatible with the respective actions on Grassmannians Gr(k, n)(Fq ) and the family of
k-subsets of {1, . . . , n}. This is explained in detail in section 1.1.
According to a private conversation with Cartier, Tits’ idea did not find much res-
onance at the time—one has to bear in mind that this was at a moment in which the
community struggled with a generalization of algebraic geometry from fields to other
types of rings; Grothendieck’s clarifying invention of schemes was still several years
ahead. In so far a geometry over the hypothetical object F1 was too far away from
conceivable mathematics at the time.
As a consequence, it took more than three decades until the field with one element
gained popularity, this time thanks to one of the most famous riddles in number theory,
the Riemann hypothesis. Alexander Smirnov gave talks about how F1 -geometry could
be involved in a proof of the Riemann hypothesis in the late 1980s. This idea finds its
first mentioning in the literature in Manin’s influential lecture notes [Man95], which are
based on his talks at Harvard, Yale, Columbia and MSRI in 1991/92.
In a nutshell, this ansatz postulates a completed arithmetical curve Spec Z over F1 ,
which would allow one to mimic Weil’s proof for function fields. A particular ingredient
of this line of thought is that Z is an algebra over F1 and that there is a base extension
functor

− ⊗F1 Z : F1 -algebras / schemes Z-algebras / schemes .

 
−→

This is explained in more detail in section 1.2.

 F1 for everyone 3

Around the same time, Smirnov explained another possible application of F1 -geometry
in [Smi92]: conjectural Hurwitz inequalities for a hypothetical map
Spec Z −→ P1F1
from the completed arithmetical curve to the projective line over F1 would imply the
abc-conjecture.
Soon after, Kapranov and Smirnov aim in the unpublished note [KS95] to calculate
cohomological invariants of arithmetic curves in terms of cohomology over F1n . The
unfinished text contains an outburst of different ideas: linear and homological algebra
over F1n , distinguished morphisms as cofibrations, fibrations and equivalences (which
might be seen as a first hint of the connections of F1 -geometry to homotopy theory),
Arakelov theory modulo n and connections to class field theory and reciprocity laws,
which can be seen in analogy to knots and links in 3-space.
In [Sou04], Soulé explains a connection to the stable homotopy groups of spheres, an
idea that he attributes to Manin. Elaborating the formula GL(n, F1 ) = Sn , there should
be isomorphisms
K∗ (F1 ) = π∗ (B GL(∞, F1 )+ ) = π∗ (BS∞ + ) ' π∗s (S)
where the first equality is the definition of K-theory via Quillen’s plus construction,
naively applied to the elusive field F1 . The second equality is derived from the hypothet-
ical formula
GL(∞, F1 ) GL(n, F1 ) Sn S∞ .
[ [
= = =
n>1 n>1

The last isomorphism is the Barratt-Priddy-Quillen theorem.

1.1. Incidence geometry and F1 . In his seminal paper [Tit57] from 1957, Tits investi-
gates analogues of homogeneous spaces for Lie groups over finite fields. In the following,
we explain his ideas in the example of the general linear group GL(n) of invertible n × n-
matrices acting by base change on a Grassmannian Gr(k, n) of k-dimensional subspaces
of an n-dimensional vector space.1
We remark that all these thoughts should find a much more conceptual explanation
within the theory of buildings that was introduced a few years later Tits; the interested
reader will find more information on the developments of buildings in [Rou09]. However,
we will refrain from such a reformulation in order to stay historically accurate and to
avoid burdening the reader with an introduction to buildings.

1.1.1. Tits’ notion of a geometry. The example of GL(n) acting on Gr(k, n) makes sense
over the real or complex numbers as well as over a finite field Fq with q elements. In the
latter case, we are concerned with the group G = GL(n, Fq ) and the Fq -rational points
Gr(k, Fnq ) = Gr(k, n)(Fq ) of the Grassmannian.
Since the action of G on Gr(k, Fnq ) is transitive, Gr(k, Fnq ) stays in bijection to the left
cosets of the stabilizer of a k-subspace V of Fnq in G. If we choose V to be spanned

1To be precise, Tits considers in [Tit57] only semi-simple algebraic groups and he considers PGL(n)
in place of GL(n). However, we can illustrate Tits’ idea in the case of either group and we will allow
ourselves this inaccuracy for the sake of a simplified account.
4 Oliver Lorscheid

by the first k standard basis vectors, then the stabilizer Pk = StabG (V ) consists of all
matrices in G of the form
∗. ··· ∗. ∗. ··· ∗.
 
.. .. .. ..
∗ ··· ∗ ∗ ··· ∗ 
 0. ··· 0. ∗. ··· ∗. 
.. .. .. ..
0 ··· 0 ∗ ··· ∗
where the upper left block contains an invertible k × k-matrix and the lower right block
contains an invertible (n − k) × (n − k)-matrix. Thus we obtain an identification of
Gr(k, Fnq ) with G/Pk .
The containment relation V 0 ⊂ V of different subspaces of Fnq defines an incidence
relation ι between the elements V 0 ∈ Gr(k0 , Fnq ) and V ∈ Gr(k, Fnq ) for different k0 and k.
Tits dubs a collection of various homogeneous spaces for a fixed group G together
with an incidence relation a geometry. He investigates various properties that are satisfied
by geometries coming from matrix groups over Fq , like the one described above or its
analogues for symplectic groups or orthogonal groups.
The name “geometry” can be motivated in the above example. The points of the
different Grassmannians Gr(1, Fnq ), Gr(2, Fnq ), . . . , Gr(n−1, Fnq ) correspond to the points,
lines, . . . , (n − 2)-dimensional subspaces of the projective space Pn−1 (Fq ) = Gr(1, Fnq )
and the action of G on the different Grassmannians corresponds to the permutation of
linear subspaces by the action of G on Pn−1 (Fq ).

1.1.2. The limit geometry. Note that every finite field Fq with q elements produces such
a geometry. In other words, the geometry depends on the “parameter” q. The crucial
observation that led Tits to postulate the existence of a field F1 with 1 element is that
there is a meaningful limit of a geometry when q goes to 1. More precisely, for every
geometry coming from a matrix group over Fq , there is a geometry satisfying the same
aforementioned properties and which looks like the limit q → 1.
We explain this limit in our example of G = GL(n, Fq ) and the Grassmannians
Gr(k, Fnq ) for various n. The group for the limit geometry is the symmetric group
Sn on n elements. The homogeneous spaces are the families Σ(k, n) of all k-subsets of
{1, . . . , n}. The incidence relation is defined in terms of the containment X 0 ⊂ X for
different subsets X 0 and X of {1, . . . , n}. Note that the stabilizer of the subset {1, . . . , k}
is Sk × Sn−k , thus we have Σ(k, n) = Sn /(Sk × Sn−k ).
As mentioned before, the geometry that consists of the homogeneous spaces Σ(k, n)
of Sn satisfies analogous properties to the geometry of Grassmannians Gr(k, Fnq ).
A first link between these two geometries is laid in terms of the Weyl groups of GL(n)
and the stabilizers Pk . To explain, the Weyl group of a matrix group G is defined as the
quotient W = NormG (T )/T of the normalizer of the diagonal torus T in G by T itself.2
In our case, the normalizer of T consists of all monomial matrices, i.e. matrices that
have precisely one non-zero entry in each row and each column. The elements of the
quotient W = NormG (T )/T can thus be represented by permutation matrices and we
conclude that the Weyl group W of G is isomorphic to Sn . Similarly, the Weyl group of
Pk is Sk × Sn−k .

2This is, again, slightly inaccurate. In general, one can consider the Weyl group for any torus of a
matrix group. However, if the torus is not specified, it is assumed that the torus is of maximal rank. For
G = GL(n), the diagonal torus is of maximal rank, but this is not true for all matrix groups.
 F1 for everyone 5

The idea that the geometry of the Σ(k, n) should be thought as the limit q → 1 of
the geometry of the Gr(k, Fnq ) is suggested by the behaviour of the invariants counting
points, lines, et cetera.
To start with, these counts are immediate for the limit geometry:
n! n
 
# Sn = n!, # Sk × Sn−k = k!(n − k)!, and # Σ(k, n) = = .
k!(n − k)! k
The corresponding counts for the geometry of Grassmannians will employ the quantities
n−1 n [n]q !
n
[n]q = ∑ q , [n]q ! = ∏[i]q ,
h i
i
and = ,
i=0 i=1
k q [k]q ![n − k]q !
which are called the Gauss number, the Gauss factorial and the Gauss binomial, respec-
tively, or sometimes quantum number, quantum factorial and quantum binomial because
of their relevance in theoretical physics. Note that we recover the classical quantities in
the limit q → 1:
n n
h i  
lim [n]q = n, lim [n]q ! = n!, and lim = .
q→1 q→1 q→1 k q k
The elements of G = GL(n, Fq ) correspond to ordered bases of Fnq . For the first basis
vector, we have qn − 1 choices, for the second qn − q choices and so forth. Therefore we
have
n n 1
∏(q ∏(q − 1)qi−1[i]q
n i−1 2 −n)
# GL(n, Fq ) = −q ) = = (q − 1)n q 2 (n [n]q !
i=1 i=1
The elements of Pk decompose into four blocks, consisting of an invertible k × k-matrix,
an arbitrary k × (n − k)-matrix, a zero matrix and an invertible (n − k) × (n − k)-matrix.
Thus we obtain
1 2 −n)
#Pk = # GL(k, Fq ) # Mat(k × n, Fq ) # GL(n − k, Fq ) = (q − 1)n q 2 (n [k]q ![n − k]q !
Dividing the former two quantities yields
# GL(n, Fq ) [n]q ! n
h i
# Gr(k, Fnq ) = = = .
#Pk [k]q ![n − k]q ! k q
Note that the limit q → 1 of the cardinalities of G and Pk is 0 due to the term (q − 1)n .
But if we resolve this zero, i.e. if we divide by #T = (q − 1)n , then we obtain
lim #G/#T = #Sn , lim #Pk /#T = #Sk × Sn−k , lim # Gr(k, Fnq ) = #Σ(k, n).
q→1 q→1 q→1

Based on these observations, Tits dreamt about the existence of a geometry over a
field F1 with one element that is capable to explain these effects. Later, this idea has
been summarized in hypothetical formulas such as
GL(n, F1 ) = Sn and Gr(n, F1 ) = Σ(k, n).
However, we caution the reader not to take these formulas too literal, for the reasons
explained in [Lor16, Prologue].
Example 1.1. We illustrate the ideas of Tits in the example of GL(3, Fq ). In this case,
we consider the two Grassmannians Gr(1, F3q ) and Gr(2, F3q ), whose points corresponds
6 Oliver Lorscheid

to the points and lines in P2 (Fq ), respectively. The incidence relation consists of pairs
of a point P and a line L such that P ∈ L. In the case q = 2, we calculate
3 3
h i h i
# Gr(1, F32 ) = = 7 and # Gr(2, F32 ) = = 7.
1 q 2 q
Moreover, note that every line contains q + 1 = 3 points and that every point is contained
in q + 1 = 3 lines. The corresponding geometry is illustrated on the left hand side of
Figure 1 where the dots correspond to the points in P2 (Fq ), the circles correspond to the
lines in P2 (Fq ) and an edge between a dot and a circle indicates that the corresponding
point is contained in the corresponding line.
Considering the limit q → 1 yields the  geometry for S3 that consists of the sets
Σ(1, 3) = {1}, {2}, {3} and Σ(2, 3) = {1, 2}, {1, 3}, {2, 3} . Note that every point

of this geometry, i.e. an one element subset of {1, 2, 3}, is contained in q+1 = 2 lines, i.e.
a 2-subset of {1, 2, 3}. Similarly, every line contains q +1 = 2 points. The corresponding
geometry is depicted on the right hand side of Figure 1 .

q→1

Figure 1. The geometry of GL(3, Fq ) for q = 2 and its limit q → 1

1.2. The Riemann hypothesis. One of the most profound problems in mathematics is
the Riemann hypothesis. We refrain from alluding to its importance, but refer the reader
to one of the numerous overview texts on the topic. A good start is the book [BCRW08].
Arguably, the most influential publication for shaping the area of F1 -geometry was
Manin’s lecture notes [Man95], in which Deninger’s programme for a proof of the
Riemann hypothesis ([Den91], [Den92], [Den92]) and Kurokawa’s work on absolute
tensor products of zeta functions ([Kur92]) are combined with the idea to realize the
integers as a curve over the elusive field F1 with one element.
In the following, we briefly review the Riemann hypothesis and explain how the proof
of its analogue for function fields leads to the desire for algebraic geometry over F1 .
1.2.1. The Riemann zeta function. The Riemann zeta function is defined by the formulas
1 1
ζ(s) = ∑ s = ∏ ,
n>1 n p prime 1 − p
−s

both of which are expressions that converge absolutely in the half plane {s ∈ C|Re s > 1}.
It extends to a meromorphic function on the whole complex plane with simple poles
at 0 and 1 and it satisfies a certain functional equation. The zeros of the Riemann zeta
function are more delicate and the protagonists of the Riemann hypothesis.
 F1 for everyone 7

Let us begin with explaining the functional equation. The Γ-function is defined in
terms of the formula Z ∞
Γ(t) = xt−1 e−x dx,
0
which converges on the half plane {t ∈ C|Re t > 0} and can be extended to a meromor-
phic function on the whole complex plane. It has simple poles at all negative integers
−1, −2, . . . and no zeros.
The completed zeta function is the meromorphic function
ζ ∗ (s) = Γ(s/2) π −s/2 ζ(s)
and it satisfies the functional equation
ζ ∗ (1 − s) = ζ ∗ (s).
It is immediate from the definition that ζ(s) does not have any zero for Re s > 1. The
functional equation implies that ζ(s) compensates the poles of Γ(s/2) with simple zeros
at the negative even integers −2, −4, . . . , which are called the trivial zeros of the zeta
function. We see that all other zeros of ζ(s) lie on the critical strip {s ∈ C|0 6 Re s 6 1}.
The Riemann hypothesis claims the following.
Riemann hypothesis. Every non-trivial zero of ζ(s) has real part 1/2.
In other words, the Riemann hypothesis states that all zeros of completed zeta function
ζ ∗ (s)
lie on the critical line {s ∈ C|Re s = 1/2}. See Figure 2 for an illustration of the
poles and zeros of ζ(s).

critical line

1/2 + i · 25.01 . . .
1/2 + i · 21.02 . . .

1/2 + i · 14.13 . . .
trivial zeros

pole pole nontrivial

−4 −2 0 1 zeros

1/2 − i · 14.13 . . .

1/2 − i · 21.02 . . .
1/2 − i · 25.01 . . .

critical strip

Figure 2. The poles and zeros of ζ(s)

1.2.2. Absolute values. While the Riemann hypothesis is still an open problem, its
function field analoguea has been proven by André Weil. In order to explain the analogy
between the Riemann zeta function ζQ (s) = ζ(s) and the zeta function ζF (s) of a function
field, we reinterpret the factors of the Euler product ζQ (s) = ∏(1 − p−s )−1 in terms of
absolute values of Q.
8 Oliver Lorscheid

An absolute value of a field F is a map v : F → R>0 to the non-negative real numbers

that satisfies v(0) = 0, v(1) = 1, v(ab) = v(a)v(b) and v(a + b) 6 v(a) + v(b) for all
a, b ∈ F.
We illustrate some examples. Every field F has a trivial absolute value, which is the
absolute value v0 : F → R>0 that maps every nonzero element of F to 1. For F = Q, the
usual absolute value or archimedean absolute value v∞ = | | : Q → R>0 is an absolute
value. But there are other, substantially different absolute values for Q. Namely, for
every prime number p, the p-adic absolute value is defined as the map v p : Q → R>0
with v p (pi · ba ) = p−i if neither a nor b are divisible by p.
An absolute value v : F → R>0 is nonarchimedean if it satisfies the strong triangle
inequality v(a + b) 6 max{v(a), v(b)} for all a, b ∈ F. In fact, most absolute values
are nonarchimedean. By a theorem of Ostrowski, the only exceptions come from
restricting the usual absolute value of C to a subfield, and possibly taking powers. If v is
nonarchimedean, then the subsets
Ov = { a ∈ F | v(a) 6 1 } and mv = { a ∈ F | v(a) < 1 }
are a local ring, called the valuation ring of v, and its maximal ideal, respectively. We
define the residue field of v as k(v) = Ov /mv . Examples of nonarchimedean places are
the trivial absolute value v0 and the p-adic absolute value v p : Q → R>0 . The valuation
ring of v p and its maximal ideal are
a a
Ov p = { ∈ Q | p does not divide b } and mv p = { ∈ Ov p | p divides a },
b b
respectively. The residue field k(v p ) is isomorphic to the finite field F p with p elements.
Two absolute values v1 and v2 of F are equivalent if there is a t ∈ R>0 such that
v2 (a) = v1 (a)t for all a ∈ F. A place of F is an equivalence class of nontrivial absolute
values. Note that two equivalent nonarchimedean absolute values define the same subsets
Ov and mv , and thus have the same residue field k(v).
By another theorem of Ostrowski, the places of Q are the archimedean place, rep-
resented by v∞ , and the p-adic places, represented by v p , where p ranges through all
prime numbers.
By a slight abuse of notation, we will denote a place represented by an absolute value
v by the same symbol v. If v is nonarchimedean and k(v) is finite, then we define the
local zeta factor at v as
1
ζv (s) = .
1 − #k(v)s
If we define the zeta factor of the archimedean absolute value v∞ of Q as ζv∞ (s) =
π −s/2 Γ(s/2), then the completed zeta function can be expressed as
∗
ζQ (s) = ∏ ζv (s)
places v of Q

in the region of convergence, i.e. for Re s > 1. While the different shape of the factor
ζ∞ at infinity has been the cause for much musing, the nonarchimedean factors have a
direct analogue in the function field setting.
1.2.3. Zeta functions for function fields. Let F be a global function field, i.e. a finite
field extension of a rational function field Fq (T ) over a finite field Fq with q elements.
For later reference, we assume that q is maximal with the property that F contains Fq (T )
 F1 for everyone 9

as a subfield. Every nontrivial absolute value of F is nonarchimedean and the residue

field is a finite field extension of Fq and therefore finite. Thus the local zeta factors
ζv (s) = (1 − #k(v)−s )−1 make sense, and we define the zeta function of F as
ζF (s) = ∏ ζv (s),
places v of F

which is an expression that has analogous properties to the completed Riemann zeta
function: it converges for Re s > 1 and has a meromorphic continuation to all s in C. It
satisfies a functional equation of the form
ζF (1 − s) = ±q(2g−2)(1−s) ζF (s)
where g is the genus of F, a number which plays an analogue role as the genus of a
Riemann surface. As explained below, the field F occurs indeed as the function field of
a certain curve, but explaining the definition of the genus would lead us too far astray.
A particular property for the function field setting is that the zeta function can be
expressed in terms of a rational function. Namely, ζF (s) = ZF (q−s ) for a function ZF (T )
in T of the form
P(X)
ZF (T ) = .
(1 − T )(1 − qT )
2πi
This implies that ζF (s) is periodic modulo ln q , and it has simple poles in all complex
2πi 2πi
numbers of the form k ln q and 1 + k ln q with k ∈ Z; cf. the illustration in Figure 3 below.
Example 1.2. To explain the analogy to Q in more detail below, we consider the example
of a rational function field F = Fq (T ). Then F is the field of fraction of the polynomial
ring Fq [T ].
The analogue of the p-adic absolute value of Q is the f -adic absolute value v f :
Fq (T ) → R>0 where f is an irreducible polynomial in Fq [T ], i.e. f is a polynomial of
positive degree d > 1 and does not equal the product of two polynomials of positive
degree. Note that we can write every nonzero element of Fq (T ) in the form f i gh where
i ∈ Z and g and h are polynomials that are not multiples of f by another polynomial.
Then the f -adic valuation v f : Fq (T ) → R>0 is defined by the formula
g
v f ( f i ) = q−di .
h
Note that, similar to the p-adic valuation, v f is nonarchimedean and we have
g g
Ov f = { ∈ Fq (T ) | f does not divide h } and mv f = { ∈ Ov f | f divides g }.
h h
The residue field k(v f ) = Ov f /mv f is isomorphic to Fqd , the unique degree d-extension
field of Fq .
Up to equivalence, the f -adic absolute values v f represent all places of Fq (T ), with
the exception of the place at infinity. This latter place is represented by the absolute
value v∞ , which is defined by
g
v∞ ( ) = qdeg h−deg g
h
for nonzero polynomials g and h of respective degrees deg g and deg h. The absolute
value v∞ is also nonarchimedean, in stark contrast to the situation for Q where the place
10 Oliver Lorscheid

at infinity is archimedean. We have

g g
Ov∞ = { ∈ Fz (T ) | deg g 6 deg h } and mv∞ = { ∈ Ov∞ | deg g < deg h }.
h h
The residue field of v∞ is k(v∞ ) = Fq .
To conclude this example, we note that it is not too hard to show that the zeta function
of F = Fq (T ) has the following form:
1 1
ζF (s) = ∏ =
(1 − q )(1 − q1−s )
.
places v of F 1 − #k(v)
−s −s

In particular, ζF (s) does not have any zero at all in this case.
1.2.4. From the Hasse-Weil theorem to F1 . The analogue of the Riemann hypothesis for
ζF (s) has been proven by Hasse ([Has36]) in the case of elliptic function fields and by
Weil ([Wei48]) for all function fields. It is also known as the Hasse-Weil theorem.
Theorem 1.3. Every zero of ζF (s) has real part 1/2.

zeros

pole pole
2π i 2π i
ln q 1 + ln q

pole pole
0 1

pole pole
2π i 2π i
− ln q 1 − ln q

Figure 3. The poles and zeros of ζF (s)

Example 1.4. To give a few examples, the case of genus g = 0 is the case of a rational
function field F = Fq (T ) where the curve C is a projective line over Fq . As we have
seen before, the Hasse-Weil theorem is trivial in this case since ζF (s) does not have any
zeroes. The case of genus g = 1 is the case of an elliptic curve, which has been treated
2πi
by Hasse. In this case, the Riemann zeta function has two zeros modulo ln q and the
√ √
Riemann hypothesis for F is equivalent to the estimate q − 2 q 6 h 6 q + 2 q where
h is the number of places v of F with residue field k(v) equal to Fq .
In the following, we will explain a few of the key ingredients in the proof of the
Hasse-Weil theorem and make clear how this leads to the postulation of F1 .
As mentioned before, F is the function field of a curve C over Fq . The points of C are,
by definition, the places of F where we neglect the “generic point” in our description.
The curve fibres over its “base point” Spec Fq .
 F1 for everyone 11

We can embed the curve C diagonally into the fibre product C ×Spec Fq C. The Riemann
hypothesis for ζF (s) can be tied to an estimate for the number of intersection points of the
diagonally embedded curve C with its “Frobenius twist” inside the surface C ×Spec Fq C.
This estimate can be established by an explicit calculation.
The analogies between number fields and function fields lead to the hope that one can
mimic these methods for Q and approach the Riemann hypothesis. Grothendieck’s theory
of schemes provides a satisfying framework to view the collection of all nonarchimedean
places v p of Q as a curve, namely, as the arithmetic curve Spec Z, the spectrum of Z. In
analogy to the function field setting, we would like to include the archimedean place v∞
in a hypothetical completion Spec Z of Spec Z and we would like to have a base field
for Spec Z, namely F1 , the field with one element.
In such a theory, we should have a base extension functor − ⊗F1 Z from F1 -schemes to
usual schemes and we should be able to define the arithmetic surface Spec Z ×F1 Spec Z.

2. Birth hour: F1 -varieties and their siblings

The first person courageous enough to propose a notion of an F1 -variety was Cristophe
Soulé who presented a first attempt [Sou99] at the Mathematische Arbeitstagung of the
MPI in 1999. He published a refinement [Sou04] of his approach in 2004.
Soon after, ideas were blooming and the same decade saw more than a dozen different
definitions for F1 -geometry. In the following, we give a brief and incomplete overview
of such approaches. For more details and references, confer [LPL11a, Lor16].
Soulé’s approach underwent a number of further variations by himself in [Sou11] and
by Connes and Consani in [CC11b, CC10b]. These approaches are closely related to the
notion of a torified scheme as introduced by López Peña and the author in [LPL11b].
One of the most important approaches is Deitmar’s variation [Dei05] of Kato fans
([Kat94]), which he calls F1 -schemes. It is the most minimalistic approach to F1 -
geometry since it is included in every other theory about F1 . In this sense, Deitmar’s
theory constitutes the very core of F1 -geometry. Subsequently, this theory has found
various applications under the names of Deitmar schemes, monoid schemes or monoidal
schemes.
Toën and Vaquié generalize in [TV09] the functorial viewpoint on scheme theory to
any category C that looks sufficiently like a categories of modules over rings. Durov
gains a notion of F1 -varieties in his unpublished approach [Dur07] to Arakelov theory.
Borger defines in [Bor09] an F1 -variety as a scheme together with a lift of the Frobenius
automorphisms for every prime number p. Haran has developed various approaches in
[Har07, Har10, Har17].
Lescot develops in [Les11, Les12] an algebraic geometry over idempotent semirings,
which he dubs F1 -geometry. Connes and Consani extend Lescot’s viewpoint to the
context of Krasner’s hyperrings, which they promote as a geometry over F1 in [CC10a,
CC11a]. This point of view has been extended by Jun in [Jun15].
Berkovich introduces a theory of congruence schemes for monoids ([Ber]), which can
be seen as an enrichment of Deitmar’s F1 -geometry. Deitmar modifies this approach in
[Dei11].
The author develops in [Lor12b, Lor12c, LPL12, Lor14, Lor17] (partly in collabo-
ration with López Peña) a notion of F1 -geometry, based on the notion of a so-called
blueprint.
12 Oliver Lorscheid

We will examine Deitmar’s and the author’s approach to F1 -geometry in more detail
in the following sections.
2.1. Monoid schemes. In this section, we review a slight modification of Deitmar’s
approach to F1 -schemes in [Dei05], which can be seen as the very core of F1 -geometry.
It realizes the motto “non-additive geometry” literally and it forms a subclass of every
other approach to varieties over F1 , up to some finiteness conditions in certain cases. We
make the definitions approachable to the non-expert, but warn the inexperienced reader
that a motivation for scheme theory lies outside the scope of this overview paper.
2.1.1. Monoids. The underlying algebraic objects in [Dei05] are monoids, i.e. semi-
groups with an identity element. For the purpose of F1 -geometry, it has been proven
useful to consider monoids with zero, which are monoids A together with an absorbing
element 0, i.e. 0 · a = 0 for all a ∈ A where we write the monoid multiplicatively. In this
exposition, we will consider the variation of monoid schemes for monoids with zero.
In the following, we will agree that all of our monoids are commutative and with zero.
A monoid morphism is a map f : A1 → A2 between monoids A1 and A2 with f (0) = 0,
f (1) = 1 and f (ab) = f (a) f (b) for all a, b ∈ A1 .
There is a base extension functor − ⊗F1 Z that sends a monoid to a ring. Namely,
given a monoid A, we define
A ⊗F1 Z = Z[A] / h0A i
where Z[A] is the semigroup ring Z[A] = {∑ na a|na ∈ Z, almost all 0} of finite Z-linear
combinations of elements of A and where h0A i is the ideal generated by the zero 0A of A.
In other words, we identify the zero of the monoid A with the zero of the ring Z[A]. A
monoid morphism f : A1 → A2 can be extended by linearity to a ring homomorphism
A1 ⊗F1 Z → A2 ⊗F1 Z between the associated rings, which defines the base extension
functor − ⊗F1 Z from monoids to rings.
Example 2.1. We provide some first examples of monoids. The trivial monoid is the
monoid with a single element 0 = 1. Its base extension {0} ⊗F1 Z is the trivial ring.
The smallest nontrivial monoid consists solemnly of two elements 0 and 1, and it is
this monoid that we call F1 . Its base extension F1 ⊗F1 Z is Z.
The free monoid generated by a number of indeterminants T1 , . . . , Tn consists of all
monomials T1e1 . . . Tnen in T1 , . . . , Tn together with a distinct element 0. We denote the
free monoid in T1 , . . . , Tn by F1 [T1 , . . . , Tn ]. Its base extension F1 [T1 , . . . , Tn ] ⊗F1 Z is the
polynomial ring Z[T1 , . . . , Tn ].
Note that every ring is a monoid if we omit its addition.
2.1.2. The spectrum. Let A be a monoid. An ideal of A is a multiplicative subset I of A
such that 0 ∈ I and IA = I. A prime ideal of A is an ideal p of A such that its complement
S = A − p is a multiplicative subset, i.e. it contains 1 and is closed under multiplication.
The spectrum Spec A of A is the set of all prime ideals of A together with a topology and
a structure sheaf, which we will describe below.
The topology of Spec A is generated by the principal open subsets
Uh = { p ∈ Spec A | h ∈
/ p}
where h ranges through all elements of A. Note that U1 = Spec A, that U0 = 0/ and that
Ugh = Ug ∩Uh . Thus every open subset of Spec A is a union of principal open subsets.
 F1 for everyone 13

Example 2.2. We describe the topological spaces of some spectra of monoids. We begin
with some general observations that are helpful in the calculation of the prime ideals.
The unit group A× of a monoid A is the set of invertible elements and forms an abelian
group. If a ∈ A× and p is a prime ideal, then a ∈ / p. Note that A − A× is always a prime
ideal and that it contains every other prime ideal.
As a consequence, a monoid A with A× = A − {0} has a single prime ideal, which is
{0}. In particular, Spec F1 consists of a single point.
The ideal I = (J) generated by a subset J of A is, by definition, the smallest ideal
containing J. If A is generated by elements a1 , . . . , an as a monoid, i.e. every element of
A is a finite product of these elements or 0, then every prime ideal of A is generated by a
subset of the generators.
The free monoid F1 [T1 , . . . , Tn ] is generated by T1 , . . . , Tn . In this case, every subset of
{T1 , . . . , Tn } generates a prime ideal. We illustrate the spectra of F1 [T ] and F1 [T1 , T2 ] in
Figure 4. The labelled dots stay for the corresponding prime ideals, and a line between
two different prime ideals indicates that the prime ideal on the bottom end is contained
in the prime ideal on the top end.
Note that all our examples are topological spaces with finitely many points. In this
case, the open subsets are precisely those that are closed from below with respect to the
inclusion relation, and the closed subsets are those that are closed from above.

(T1 , T2 )
(T )

(T1 ) (T2 )
{0}
{0}

Figure 4. The spectra of F1 [T ] and F1 [T1 , T2 ]

2.1.3. The structure sheaf. It is somewhat more difficult to describe the structure sheaf
OX of X = Spec A. It is an association that sends an open subset U of X to a monoid
OX (U) and an inclusion V ⊂ U to a monoid morphism resU,V : OX (U) → OX (V ). We
will restrict ourselves to an explicit description of its values on principal open subsets,
which are given in terms of localizations.
Namely, let A be a monoid and S a multiplicative subset. The localization of A at S is
the monoid
S−1 A = S × A/ ∼
where ∼ is the equivalence relation on the Cartesian product S × A given by (s, a) ∼
(s0 , a0 ) if and only if there exists a t ∈ S such that tsa0 = ts0 a. If we denote the equivalence
class of (s, a) by as , then the multiplication of S−1 A is given by the rule as · bt = ab st . The
−1 1 0
identity element of S A is 1 and its zero is 1 . Note that the localization comes with a
monoid morphism A → S−1 A, defined by a 7→ 1a , which sends every s ∈ S to an invertible
element in S−1 A. To wit, the inverse of 1s is 1s .
14 Oliver Lorscheid

Cases of particular importance are the localizations A[h−1 ] = S−1 A of A in an element

h where S = {hi }i>0 and the localization Ap = S−1 A of A at a prime ideal p where
S = A − p.
Example 2.3. Let A = F1 [T ] be the free monoid in one indeterminate T . We denote the
localization A[T −1 ] of F1 [T ] in T by F1 [T ±1 ], which consists in 0 and all integer powers
T i of T . The base extension F1 [T ±1 ] ⊗F1 Z is the ring Z[T ±1 ] of Laurent polynomials
over Z.
We are prepared to describe the structure sheaf for principal open subsets. Namely,
we define OX (Uh ) = A[h−1 ]. It follows easily from the definitions that Ug ⊂ Uh if and
only if g is divisible by h, i.e. we have g = f h for some f ∈ A. This observation lets us
define the monoid morphism
resUh ,Ug : OX (Uh ) = A[h−1 ] −→ A[g−1 ] = OX (Ug ).
a afi
hi
7−→ gi
For example, if h = 1, then Uh = Spec A and OX (Spec A) = A. In this case, the restriction
map resUh ,Ug : A → A[g−1 ] is nothing else than the aforementioned map that sends a → a1 .
The structure sheaf OX of X = Spec A derives from our description on its values on
principal open subsets due to the fact that this definition extends uniquely to a sheaf of
monoids3 on Spec A. Thus Spec A is a monoidal space, i.e. a topological space together
with a sheaf in monoids.
2.1.4. Monoid schemes. A monoid scheme, or F1 -scheme in the terminology of [Dei05],
is a monoidal space that has a covering by open subsets that are isomorphic to the spectra
of monoids. We say that a monoid scheme is affine if it is isomorphic to the spectrum of
a monoid.
The base extension functor − ⊗F1 Z extends to a functor from monoid schemes to
usual schemes in terms of open coverings and the definition
(Spec A) ⊗F1 Z = Spec A ⊗F1 Z .

Example 2.4. In analogy to the affine space AnZ = Spec Z[T1 , . . . , Tn ] over Z, we define
the affine space AnF1 over F1 as Spec F1 [T1 , . . . , Tn ]. We obtain AnF1 ⊗F1 Z = AnZ .
Similarly, we define the multiplicative group scheme Gm,F1 over F1 as Spec F1 [T ±1 ]
in analogy to the multiplicative group scheme Gm,Z = Spec Z[T ±1 ] over Z. We obtain
Gm,F1 ⊗F1 Z = Gm,Z . For details about the group law of Gm,F1 , see section 3.2.
This view on F1 -geometry is very appealing thanks to its simple and clean approach.
However, it is often too limited for applications. Namely, the only varieties that are base
extensions of monoid schemes are toric varieties. In more detail, Deitmar proves the
following in [Dei08].
Theorem 2.5. Let X be a monoid scheme such that X ⊗F1 Z is a connected, separated
and flat scheme of finite type. Then X ⊗F1 Z is a toric variety over Z.
Example 2.6. As a first example of a monoid scheme that is not affine, we consider the
projective line PF11 over F1 . It consists of two closed points x0 and x1 and a so-called
3In order to avoid a digression into technicalities, we do not introduce sheaves. The reader can safely
omit all details concerning sheaves.
 F1 for everyone 15

generic point η, which is contained in every non-empty open subset. It can be covered
by the two open subsets U0 = {x0 , η} and U1 = {x1 , η}, which can be identified with
affine lines Spec F1 [Ti ] over F1 where i ∈ {0, 1}. Their

intersection is U01 = {η}, which
coincides with Gm,F1 = Spec F1 [T0 , T1 ]/(T0 T1 = 1) . The restriction maps from Ui to
U01 are the obvious inclusions
resUi ,U01 : F1 [Ti ] −→ F1 [T0 , T1 ]/(T0 T1 = 1).
The projective line P1F1 is illustrated on the left hand side of Figure 5 where we label
the different points with homogeneous coordinates with coefficients in F1 . However, a
coefficient 1 should be read as a generic value different from zero. This explains why
[1 : 1] is the generic point.
In a similar vain, it is possible to define projective spaces or any toric variety as a
monoid scheme. We illustrate the projective surface P2F1 over F1 on the right hand side
of Figure 5.

[1 : 0 : 0] [0 : 1 : 0] [0 : 0 : 1]
[1 : 0] [0 : 1]

[1 : 1 : 0] [1 : 0 : 1] [0 : 1 : 1]
[1 : 1]
[1 : 1 : 1]

Figure 5. The projective line P1F1 and the projective plane P2F1 over F1

Remark 2.7. We see in these examples already a first relation to the incidence geome-
tries considered by Tits. The homogeneous coordinates [x1 : · · · : xn ] of a point in Pn−1 F1
defines a the subset I = {i|xi 6= 0} of {1, . . . , n} and thus a point in Σ(k, n) if k = #I. If
we consider the action of Sn on PFn−1 1
that permutes the coordinates of points, then we
obtain the limit geometry of Sn as considered in section 1.1.
We illustrate this relation in Figure 6. If we remove the generic point [1 : 1 : 1] from
2
PF1 , which corresponds to the full subset of {1, 2, 3}, then we are left with the space
illustrated on the left hand side of Figure 6 where we indicate the points corresponding
to 2-subsets by circles. This is the same as the limit geometry from Figure 1, as repeated
on the right hand side of Figure 6.

Figure 6. The relation of P2F1 to the Tits geometry of S3

16 Oliver Lorscheid

2.2. Blueprints and blue schemes. Since semi-simple algebraic groups are not toric
varieties, Theorem 2.5 testifies that monoid schemes are not sufficient to realize Tits’
dream of algebraic groups over F1 . This led to the refinement of monoid schemes in
terms of blueprints.
As a leading example for the exposition in this section, we consider the special linear
group SL(2). As a scheme over the integers, we have
SL(2)Z = Spec Z[T1 , T2 , T3 , T4 ] (T1 T4 − T2 T3 − 1) ,


which encodes all 2 × 2-matrices with coefficients T1 , . . . , T4 with determinant T1 T4 −

T2 T3 equal to 1. We would like to make sense of the formula
SL(2)F1 = Spec F1 [T1 , T2 , T3 , T4 ] (T1 T4 = T2 T3 + 1) .


2.2.1. Semirings. In order to understand the concept of a blueprint, we have to review

some facts about semirings. Although the axioms of a semiring are very similar to those
of a ring—we just omit the axiom about additive inverses—, we will see that the theory
of semirings shows certain effects that do not occur in ring theory.
A commutative semiring with 0 and 1, or for short a semiring, is a set R together with
an addition +, a multiplication · and constants 0 and 1 such that (R, +) is a commutative
semigroup with neutral element 0, such that (R, ·) is a commutative semigroup with
neutral element 1 and with zero 0 and such that a(b + c) = ab + ac for all a, b, c ∈ R. A
semiring homomorphism is an additive and multiplicative map between semirings that
maps 0 to 0 and 1 to 1.
Example 2.8. Commutative rings with 1 are semirings. Other examples are the natural
numbers N and the non-negative real numbers R>0 together with the usual addition and
multiplication.
Given a semiring R, we can form the polynomial semiring R[T1 , . . . , Tn ] in n inde-
terminants T1 , . . . , Tn , which consists in all polynomials ∑ ae T e where e = (e1 , . . . , en )
is a multi-index and T e = T1e1 · · · Tnen . Addition and multiplication of R[T1 , . . . , Tn ] are
defined in the same way as for polynomial rings.
Given a multiplicative monoid
A with 0, then we define the monoid semiring as the
set N[A] = ∑ ai ai ∈ A − {0} of finite formal sums of nonzero elements in A whose
 

product is inherited by the product of A. Note that if ab = 0 in A, then 1a · 1b equals the

neutral element for addition of N[A], which is the empty sum. In other words, the zero
of A is identified with the zero of N[A].
An example of a more exotic, or tropical, nature is the Boolean semifield B = {0, 1}
where 1 + 1 = 1. Note that all other sums and products are determined by the semiring
axioms. This definition extends to the semiring T whose underlying set are the non-
negative numbers R>0 , whose multiplication is defined as usual and addition is defined
as the maximum, i.e.
a + b = max{a, b}.
This semiring is called the tropical semifield.
Note that the tropical semifield is typically described differently in the literature about
tropical geometry. Namely, it takes the shape of the “max-plus-algebra” R ∪ {−∞} or
the “min-plus-algebra” R ∪ {∞}. However, all the three semirings, T, R ∪ {−∞} and
R ∪ {∞}, are isomorphic. For our purposes, it is least confusing to use the non-negative
real numbers with usual multiplication as a model for the tropical numbers.
 F1 for everyone 17

Given a semiring R, we can construct the ring RZ = R ⊗N Z of the formal differences

of elements of R. In more detail, RZ is defined as follows. As a set, RZ = R × R/ ∼,
where (x, y) ∼ (x0 , y0 ) if and only if x + y0 + z = x0 + y + z for some z ∈ R. We write x − y
for the equivalence class of (x, y) in RZ . We can extend the addition and multiplication
from R to RZ by the formulas
(x − y) + (z − t) = (x + z) − (y + t) and (x − y) · (z − t) = (xz + yt) − (xt + yz).
Example 2.9. We have N ⊗N Z = Z, R>0 ⊗N Z = R, R[T1 , . . . , Tn ]Z = RZ [T1 , . . . , Tn ]
and B ⊗N Z = T ⊗N Z = {0}.
We can extend the notion of an ideal from rings to semirings: an ideal of a semiring
R is a subset I that contains 0, x + y and tx for all x, y ∈ I and t ∈ R.
In contrast to ring theory, there is, however, no perfect unison between ideals and
quotient semirings. This leads to the notion of a congruence, which is an equivalence
relation R on a semiring R such that the quotient set R/R is a semiring, i.e. we can
define addition and multiplication on equivalence classes by evaluating unambiguously
on representatives. More explicitly, a congruence is an equivalence relation R on R such
that x ∼ y and z ∼ t imply
x+z ∼ y+t and xz ∼ yt
for all x, y, z,t ∈ R. For any set of relations xi ∼ yi between elements xi and yi of a
semiring R, there is a smallest congruence R containing these relations. Thus we can
define
R / hxi ∼ yi i = R / ∼ .
Conversely, every morphism f : R1 → R2 of semirings has a congruence kernel, which is
the congruence R on R1 that is defined by x ∼ y if and only if f (x) = f (y). In contrast,
the kernel of f is the ideal ker f = {x ∈ R| f (x) = 0} of R1 .
Thus we gain two opposing constructions: we can associate with a congruence R on
R the kernel of the quotient map R → R/R, and we can associate with an ideal I of R
the congruence R = hx ∼ 0|x ∈ Ii. We write R/I for R/R in this case.
However, returning to our earlier remark, in general it is neither true that every
congruence comes from an ideal nor that every ideal comes from a congruence. We call
an ideal that occurs as the kernel of a morphism a k-ideal. Note that if R is a ring, then
every ideal is a k-ideal.
Example 2.10. Every semiring can be written as a quotient of a polynomial semiring
over N, possibly in infinitely many indeterminants. For instance, we can define
R = N[T1 , . . . , T4 ] / hT1 T4 ∼ T2 T3 + 1i.
Then RZ = Z[T1 , . . . , T4 ]/(T1 T4 − T2 T3 − 1) is the coordinate ring of SL(2)Z .
2.2.2. Blueprints. A blueprint is a pair B = (R, A) of a semiring R and a multiplicative
subset A of R that contains 0 and 1 and that generates R as a semiring.
The last condition in this definition is equivalent with the fact that R is a quotient
of the monoid semiring N[A] by a congruence R on N[A]. We also write AR for the
blueprint B = (R, A). Note that R = N[A]/R is determined by A and R. Further, we
write B• = A and B+ = R. We write x ≡ y if (x, y) ∈ R.
A blueprint morphism f : B1 → B2 is a semiring morphism f + : B+ +
1 → B2 such that
+ • • • • • +
f (B1 ) ⊂ B2 . Note that the restriction f : B1 → B2 of f is necessarily a monoid
morphism and that f + is uniquely determined by f • .
18 Oliver Lorscheid

Example 2.11. A monoid A can be identified with the blueprint (A, N[A]) and a semiring
R can be identified with the blueprint (R, R). In this sense, blueprints are a simultaneous
generalization of monoids and semirings. Note that if B = (A, N[A]), then the base
extension functor − ⊗F1 Z from section 2.1 can be recovered in terms of the formula
A ⊗F1 Z = B+ ⊗N Z.
There are also a series of novel constructions for blueprints. One of these are the
cyclotomic field extensions F1n of F1 , which are defined as follows for n > 2. Let Z[ζn ]
be the ring of integers in the cyclotomic number field Q[ζn ] generated by a primitive n-th
root ζn of unity and let µn,0 = {0} ∪ {ζni |i ∈ Z} be the submonoid generated by ζn . Then
F1n = (µn,0 , Z[ζn ]) is a blueprint. It incorporates certain properties of the cyclotomic
field Q(ζn ): we have F+ 1n = Z[ζn ] and the Galois group of Q(ζn )/Q equals the group of
automorphisms of F1n that fix F1 .
Of particular importance is the case n = 2: the “quadratic” extension F12 = {0, 1, −1}
h1 + (−1) ≡ 0i of F1 contains an additive inverse of 1, which is important for several
purposes. For instance, it is possible to define sheaf cohomology over F12 , cf. [FLS17].
The blueprint F12 also plays a role in our approach to algebraic groups over F1 , see
section 3.2.2.
Most interestingly for our purposes is that it is possible to define the blueprint
B = F1 [T1 , . . . , T4 ]hT1 T4 ≡ T2 T3 + 1i.
The coordinate ring of SL(2)Z can be recovered from this blueprints as B+
Z.
2.2.3. The spectrum. Let B be a blueprint. We will not enter all details in the definition
of the spectrum Spec B of B, but concentrate on the description of the topological space
associated with Spec B. In fact, one can associate several topological spaces with Spec B,
cf. [Lor15] for more details. The relevant one for a theory of algebraic groups over F1 is
based on the notion of a k-ideal, which is an ideal of the monoid B• that spans a k-ideal
in the semiring B+ .
More explicitly, a k-ideal of a blueprint B is a subset I of B• such that 0 ∈ I, ab ∈ I
for all a ∈ I and b ∈ B• and c ∈ I whenever there are a1 , . . . , an , b, . . . , bm ∈ B• such that
∑ ai + c = ∑ b j in R. A k-ideal p of B is prime if S = B• − p is a multiplicative subset of
B• .
We define Spec B as the set of all prime ideals p of B together with the topology
generated by the open subsets
Uh = p ∈ Spec B  h ∈
 
/p
where h ranges through all elements of B• . It carries a structure sheaf in blueprints, but
its definition is somewhat more involved and we omit these details from our account.
Example 2.12. If B = (A, N[A]) is a monoid, then the last condition in the definition of a
k-ideal is automatically satisfied. Thus Spec B equals the monoid spectrum Spec A. For
an arbitrary blueprint B = (A, R), the spectrum Spec B is a subspace of Spec A, together
with the subspace topology.
If B = (R, R) is a semiring, then a k-ideal of B is the same as a k-ideal of R. In
particular if R is a ring, then the spectrum Spec B coincides with the usual spectrum
Spec R of the ring R. For an arbitrary blueprint B = (A, R), we have a continuous map
Spec RZ → Spec B, which is surjective if R = RZ is a ring. As a by-product, we obtain
the notion of the spectrum of a semiring.
 F1 for everyone 19

As the next case, we inspect the spectrum of our leading example, B = F1 [T1 , . . . , T4 ]
hT1 T4 ≡ T2 T3 + 1i. As explained above, it is a subspace of Spec F1 [T1 , . . . , T4 ], whose
prime ideals are of the from pI = (Ti )i∈I for any subset I of {1, . . . , 4}, cf. Example 2.2.
In order to determine Spec B, we have to verify, for which I, the set pI satisfies the
additive axiom of a prime ideal of B. This can be tested on the generators of the defining
congruence of B, which is T1 T4 ≡ T2 T3 + 1.
This relation implies that if both T1 T4 and T2 T3 are in a prime ideal p, then 1 ∈ p,
which is not the case for any prime ideal. Thus we conclude that either T1 T4 or T2 T3
is not in p, which means that either T1 and T4 are not in p or T2 and T3 are not in p.
Thus Spec B consists of the prime ideals {0}, (T1 ), . . . , (T4 ), (T1 , T4 ) and (T2 , T3 ). We
illustrate Spec B in Figure 7.

(T2 , T3 ) (T1 , T4 )

(T2 ) (T3 ) (T1 ) (T4 )

{0}

Figure 7. The spectrum of F1 [T1 , . . . , T4 ]hT1 T4 ∼ T2 T3 + 1i

We draw the reader’s attention to the fact that the two closed points (T2 , T3 ) and
(T1 , T4 ) stay in bijection to the elements of the Weyl group W of SL(2), whose two
elements are the subgroup T of diagonal matrices in SL(2) and the set of antidiagonal
matrices. We shall investigate this fact in more depth in section 3.2.
2.2.4. Blue schemes. We omit a rigorous definition of blue schemes, which would
deviate into too heavy technicalities for the flavour of this overview paper. To give a
taste, a blue scheme can be seen as a topological space together with a structure sheaf
in blueprints that is locally isomorphic to the spectrum of a blueprint. This approach
recovers monoid schemes and usual schemes as special cases; moreover, it provides a
notion of semiring schemes.
We remark that there are other types of blue schemes, which are based on other types
of prime ideals. These alternate approaches are relevant for different applications.
Meaningful variants are the following. The notion of a k-ideal, as considered above,
yields blue schemes as introduced in [Lor12b]. The notion of an ideal of the underlying
monoid B• of a blueprint B yields blue schemes that were dubbed subcanonical in
[Lor17]. Another variant is based on congruences. This has not worked out in the full
generality of blueprints, but for certain subcategories by Berkovich ([Ber]) and Deitmar
([Dei11]).

3. Growing up: achievements of F1 -geometry

Most of the initial goals of F1 -geometry have been solved, with the exception of the
most influential one, the Riemann hypothesis.
The abc-conjecture has been claimed to be proven by Mochizuki in his monumental
work [Moc12]. There are claims that some mathematicians have verified all details of
20 Oliver Lorscheid

the proof, but the acceptance by the community at large is still not clear at the time of
writing. Although Mochizuki’s proof follows a different line of thought, it contains ideas
from F1 -geometry, cf. [Moc12, Remark III.3.12.4 (iii)] and [Moc15, Remark 5.10.2
(iii)].
K-theory has been developed for monoids and monoid schemes by Deitmar in [Dei06]
and by the author’s collaboration [CLS12] with Chu and Santhanam, respectively, and it
has been shown that the K-theory of F1 coincides with the stable homotopy groups of
the sphere.
Tits’ dream of algebraic groups over F1 has been realized by the author in [Lor12c],
based on the blueprint approach to F1 .
Besides settling these old scores, F1 -geometry has found some further applications.
Monoid schemes have been utilized by Cortiñas, Haesemeyer, Walker and Weibel
([CHWW15]) to connect the algebraic K-theory of toric varieties to cyclic homology in
characteristic p.
The Giansiracusa brothers have used in [GG16] F1 -schemes and semiring schemes to
describe the tropicalization of a classical variety as a tropical scheme. This approach
towards tropical geometry promises to be a major breakthrough in the field.
In the following, we explain some of these applications in more detail. In section
3.1, we devote a few words to the progress that has been made towards the Riemann
hypothesis. In section 3.2, we explain in a certain depth how blue schemes can be used
to realize Tits’ dream of algebraic groups over F1 . Though we try to make this accessible
to the general reader, we have to assume a certain familiarity with group schemes when
stating our main results. In section 3.3, we describe the relevance of tropical scheme
theory for tropical geometry.

3.1. Steps towards the Riemann hypothesis. Several authors have considered com-
pactifications Spec Z of Spec Z and the arithmetic surface Spec Z ×F1 Spec Z; for in-
stance, see [Dur07], [Har07], [Lor14], [Tak12]. However, none of these ideas have been
pursued further to the authors knowledge.
A somewhat different route, employing idempotent semirings, is taken by Connes and
Consani who follow an ambitious programme around the Riemann hypothesis. Their
research has already lead to a large number of publications; to name a few, cf. [CC15],
[CC16], [CC17]. We are not attempting an outline of this programme, but refer the
interested reader to Connes’ chapter in [NR16] for such a summary.

3.2. Algebraic groups over F1 . In this section, we explain how we can make sense of
Tits’ dream of algebraic groups over F1 in the language of blueprints and blue schemes.
Let k be a ring and G be a group scheme over k, by which we mean a k-scheme G
together with a group law µ : G × G → G, a unit  : Spec Z → G and an inversion
ι : G → G, which are k-linear morphisms that satisfy the usual axioms of a group, i.e.
the diagrams
id×µ ∆ id×ι
G×G×G G×G G×G G G×G G×G
µ id× µ µ
µ×id
µ id 
G×G G G G Spec k G
 F1 for everyone 21

commute where ∆ : G → G × G is the diagonal and the G → Spec k is the unique

morphism to Spec k. As it is the case for usual groups, the unit  and the inversion ι are
uniquely determined by G and µ.
Example 3.1. As our main example that guides us through the concepts of this section,
we will inspect the special linear group SL(2). As a scheme, SL(2)Z is defined as the
spectrum of Z[T1 , . . . , T4 ]/(T1 T4 − T2 T3 − 1). We have already seen in Example 2.12 that
SL(2)Z descends to F1 as a blue scheme.
The group law µZ : SL(2)Z × SL(2)Z → SL(2)Z is determined by the matrix multipli-
cation
 0
T1 T20
  00
T1 T200
 0 00
T1 T2 T1 T1 + T20 T300 T10 T200 + T20 T400
   
= · =
T3 T4 T30 T40 T300 T400 T30 T100 + T40 T300 T30 T200 + T40 T400
of 2 × 2-matrices with determinant 1.
To put Tits’ idea of algebraic groups over F1 into modern language, we ask for the
following: given a group scheme G (over Z) together with a group law µ : G × G → G,
is there an F1 -scheme GF1 together with a group law µF1 : GF1 × GF1 → GF1 whose base
extensions to usual schemes are G and µ, respectively?
While some approaches to F1 -geometry contain F1 -models GF1 for large classes of
group schemes G it is a problem to descend the group law to a morphism of F1 -schemes.
Heuristically speaking, the group law of most group schemes involves the addition of
coordinates, as it is the case for SL(2)Z , cf. Example 3.1. Therefore it is not possible to
consider such a morphism in a non-additive geometry.
There is a second, more subtle, problem concerning the formula G(F1 ) = W where
G(F1 ) = Hom(Spec F1 , G) and where W is the Weyl group of G. Namely, if this
equality was an identity of groups, then it would imply that the Weyl group W of G
could be embedded as a subgroup of G(Z) such that each element of W lies in the
coset representing it. But this leads already in the simplest case of G = SL(2) to a
contradiction: its Weyl group is {±1}, but the coset of −1 is the anti-diagonal in
SL(2, Z), which does not contain an element of order 2. For more details on this, cf.
[Lor16].
Our conclusion is the following: we are required to alter the notion of a morphism
to make it possible to descend group laws and we cannot take the formula G(F1 ) = W
literally, but have to find an appropriate meaning for it. This has been done in [Lor12a],
using torified schemes and F1 -schemes after Connes and Consani, and in [Lor12c],
using blue schemes. We describe the latter approach in the upcoming paragraphs.

3.2.1. Algebraic tori. Because of their central role for what is to come we begin with
a description of (split) algebraic tori and their F1 -models. Algebraic tori are the only
connected group schemes that fit easily into any concept of F1 -geometry. We examine
the case of a torus of rank 1 from the perspective of blue schemes; the case of higher
rank can be deduced easily from the following description.
Let B be a blueprint. We define the multiplicative group scheme Gm,B over B as
Spec B[T ±1 ]. The multiplication µ : Gm,B × Gm,B → Gm,B is given by the morphism
µ# : B[T ±1 ] −→ B[T ±1 ] ⊗B B[T ±1 ],
aT 7−→ aT ⊗ T
22 Oliver Lorscheid

the unit  : Spec B → Gm,B is given by the morphism

# : B[T ±1 ] −→ B
aT 7−→ a

and the inversion ι : Gm,B → Gm,B is given by the morphism

ι# : B[T ±1 ] −→ B[T ±1 ].
aT 7−→ aT −1

If B is F1 or F12 = {0, ±1}h1 + (−1) ≡ 0i, then the base extension of Gm,B to Z is the
multiplicative group scheme Gm,Z over Z as a group scheme.4

3.2.2. The rank space. A blueprint B is cancellative if ∑ ai + c ≡ ∑ b j + c implies

∑ ai ≡ ∑ b j . Note that a blueprint B is cancellative if and only if the canonical morphism
B → B+ Z is injective. A blue scheme is cancellative if its structure sheaf takes values in
the class of cancellative blueprints.
Let X be a blue scheme and x ∈ X a point. As in usual scheme theory, every closed
subset of X comes with a natural structure of a (reduced) closed subscheme of X, cf.
[Lor12c, Section 1.4]. The rank rk x of x is the dimension of the Q-scheme x+ Q where x
denotes the closure of x in X together with its natural structure as a closed subscheme.
Define
r = min { rk x | x ∈ X }.
For the sake of simplicity, we will make the following general hypothesis on X.
(H) The blue scheme X is connected and cancellative. For all x ∈ X with rk x = r,
the closed subscheme x of X is isomorphic to either Grm,F1 or Grm,F 2 .
1

This hypothesis allows us to surpass certain technical aspects in the definition of the
rank space.
Assume that X satisfies (H). Then the number r is denoted by rk X and is called the
rank of X. The rank space of X is the blue scheme

X rk =
a
x,
rk x=r

and it comes together with a closed immersion ρX : X rk → X.

Note that Hypothesis (H) implies that X rk is the disjoint union of copies of tori Grm,F1
and Grm,F 2 . Since the underlying set of both Grm,F1 and Grm,F 2 is the one-point set, the
1 1
underlying set of X rk is W (X) = {x ∈ X|rk x = r}.

4Please note that we face a clash of notation at this point: while we denote by G
m,Z the spectrum of
the polynomial ring Z[T ]+ , the very same notation is also used for the spectrum of the free blueprint
({aT i }, Z[T ]+ ) in the definition of Gm,B in the case B = Z. However, for the sake of a more intuitive
notation, we do not dissolve this contradiction, but refer the reader to [Lor12c] and [Lor16] for a more
sophisticated treatment.
 F1 for everyone 23

rank
(T2 , T3 ) (T1 , T4 ) 1

(T2 ) (T3 ) (T1 ) (T4 ) 2

{0} 3

Figure 8. The ranks of the points of SL(2)F1

For a blue scheme Y , we denote by βY : YZ+ → Y the base extension morphism. We

have XZrk,+ ' rk x=r Grm,Z and we obtain a commutative diagram
`

ρ+
X,Z
XZrk,+ / XZ+
βX rk βX
 
ρX
X rk / X.
Example 3.2. We determine the rank space of our leading example SL(2) and its F1 -
model SL(2)F1 . It is not hard to verify that SL(2)F1 is cancellative. As already explained
in Example 2.12, the points of SL(2)F1 are of the form (Ti )i∈I where I is a subset of
{1, 2, 3, 4} that does not contain elements of both {2, 3} and {1, 4}. The subscheme

+
(Ti )i∈I Q of SL(2)Q represents the set of 2 × 2-matrices TT13 TT24 with determinant 1 for
+
which Ti = 0 for all i ∈ I. For instance, (T2 , T3 )Q is the diagonal torus in SL(2)Q , which
+ +
is one dimensional, (T1 , T4 )Q is the antidiagonal torus, also one dimensional, (T2 )Q is
+
the Borel subgroup of upper triangular matrices, which is of dimension 2, and {0}Q
equals SL(2)Q . In Figure 8, we illustrate all points of SL(2)F1 , together with their ranks.
We conclude that (T2 , T3 ) and (T1 , T4 ) are the points of minimal rank. Since

(T2 , T3 ) = Spec F1 [T1 , T2 , T3 , T4 ]hT1 T4 ≡ T2 T3 + 1, T2 = T3 = 0i

= Spec F1 [T1 , T4 ]hT1 T4 ≡ 1i

' Gm,F1

and

(T1 , T4 ) = Spec F1 [T1 , T2 , T3 , T4 ]hT1 T4 ≡ T2 T3 + 1, T1 = T4 = 0i

= Spec F1 [T2 , T3 ]hT2 T3 + 1 ≡ 0i

' Gm,F12 ,

we conclude that hypothesis (H) is satisfied by SL(2)F1 and its rank space is

SL(2)rk
F1 = (T2 , T3 ) q (T1 , T4 ) ' Gm,F1 q Gm,F12 .

3.2.3. The Tits category and the Weyl extension. A Tits morphism ϕ : X → Y between
two blue schemes X and Y is a pair ϕ = (ϕrk , ϕ+ ) of a morphism ϕrk : X rk → Y rk and a
24 Oliver Lorscheid

morphism ϕ+ : X + → Y + such that the diagram

ϕrk,+
XZrk,+ Z
YZrk,+
ρ+
X,Z
+
ρY,Z
ϕ+
XZ+ Z
YZ+
rk,+
commutes where ϕ+ Z and ϕZ are the base extension of ϕ+ and ϕrk , respectively, to
usual schemes. We denote the category of blue schemes together with Tits morphisms
by SchT and call it the Tits category.
The Tits category comes together with two important functors. The Weyl extension
W : SchT → Sets sends a blue scheme X to the underlying set W (X) = {x ∈ X|rk x = r}
of X rk and a Tits morphism ϕ : X → Y to the underlying map W (ϕ) : W (X) → W (Y )
of the morphism ϕrk : X rk → Y rk .
The base extension (−)+ : SchT → Sch+ sends a blue scheme X to its universal
semiring scheme X + and a Tits morphism ϕ : X → Y to ϕ+ : X + → Y + . We obtain the
following diagram of “base extension functors”

Sets SchZ SchR

(−)+
Z (−)+
R

W
SchN

(−)+
SchT

from the Tits category SchT to the category Sets of sets, to the category Sch+
Z of usual
+
schemes and to the category SchR of semiring schemes over any semiring R.
Theorem 3.3 ([Lor12c, Thm. 3.8]). All functors appearing in the above diagram
commute with finite products. Consequently, all functors send (semi)group objects
to (semi)group objects.
Example 3.4. The group law µZ of SL(2)Z descends uniquely to a Tits morphism
µF1 = (µ+ , µrk ) : SL(2)F1 × SL(2)F1 → SL(2)F1
in SchT . To wit, µZ descends uniquely to a morphism µ+ : SL(2)N × SL(2)N → SL(2)N
where
SL(2)N = SL(2)+ Spec T T T T T T 1)


F1 = N[T 1 , 2 , 3 , 4 ]/(T1 4 ∼ 2 3 +
since µZ is defined by means of sums and products, without employing subtraction. The
morphism µrk is determined by the commutativity of the diagram

+ µrk,+
+
SL(2)rk rk
F1 × SL(2)F1 Z
Z
SL(2)rk
F1 Z

µ+
SL(2)N × SL(2)N SL(2)N

Z


Z Z
.
 F1 for everyone 25

The Weyl extension of SL(2)F1 consists of two points x23 and x14 , which are the respec-
tive unique points of the components (T2 , T3 ) and (T1 , T4 ) of the rank space SL(2)rk
F1 . The
Weyl extension of µF1 endows W SL(2)F1 with the structure of a group with neutral


element x23 .
3.2.4. Tits-Weyl models. A Tits monoid is a (not necessarily commutative) monoid in
SchT , i.e. a blue scheme G together with an associative multiplication µ : G × G → G
in SchT that has an identity  : Spec F1 → G. We often understand the multiplication µ
implicitly and refer to a Tits monoid simply by G. The Weyl extension W (G) of a Tits
monoid G is a unital associative semigroup. The base extension G+ is a (not necessarily
commutative) monoid in SchN .
Given a Tits monoid G satisfying (H) with multiplication µ and identity , the image
of  : Spec F1 → G consists of a closed point e of X. The closed reduced subscheme
e = {e} of G is called the Weyl kernel of G.
Lemma 3.5 ([Lor12c, Lemma 3.11]). The multiplication µ restricts to e, and with this,
e is isomorphic to the torus Grm,F1 as a group scheme where r = rkX.
This means that e+ r +
Z is a split torus T ' Gm,Z of G = GZ , which we call the canonical
torus of G (w.r.t. G).
If G is an affine smooth group scheme of finite type, then we obtain a canonical
morphism
Ψe : Grk,+
Z /eZ
+
−→ W (T )
of schemes where W (T ) = NormG (T )/ CentG (T ) is the Weyl group of G w.r.t. T (cf.
[ABD+ 64b, XIX.6]). We say that G is a Tits-Weyl model of G if T is a maximal torus of
G (cf. [ABD+ 64a, XII.1.3]) and Ψe is an isomorphism.
Example 3.6. The blue scheme SL(2)F1 together with the Tits morphism µF1 is a Tits-
Weyl model of SL(2)Z , which can be reasoned as follows. It can be easily seen that
µF1 is an associative multiplication for SL(2)F1 in the Tits category, which has a unit
 : Spec F1 → SL(2)F1 , which is the pair of the morphism + : Spec N → SL(2)N , given
by TT13 TT24 7→ 10 01 , and the morphism rk : Spec F1 → (T2 , T3 ) ' Spec F1 [T1±1 ] , given
   

by T1 7→ 1.
The Weyl kernel e of the Tits monoid SL(2)F1 is therefore (T2 , T3 ) ' Spec F1 [T1±1 ] ,


and we have
+
e+ T Spec T T 1) = T = CentSL(2)Z (T )


Z = (T 2 , 3 )Z = Z[T1 , 4 ]/(T1 4 −
where the canonical torus T is the diagonal torus, which is a maximal torus of SL(2)Z .
Since the normalizer of T in SL(2)Z is the union of the diagonal torus T with the
antidiagonal torus, we obtain an isomorphism

rk,+ ∼
Ψe : SL(2)F1 Z /e+ Z −→ NormSL(2)Z (T )/ CentSL(2)Z (T ) = W (T ).
This shows that SL(2)F1 together with µF1 is a Tits-Weyl model of SL(2)Z .
We review some definitions, before we formulate the properties of Tits-Weyl models
in Theorem 3.7 below. The ordinary Weyl group of G is the underlying group W of
W (T ). The reductive rank of G is the rank of a maximal torus of G . For a split reductive
group scheme, we denote the extended Weyl group or Tits group NormG (T )(Z) by W e
(cf. [Tit66] or [Lor12c, Section 3.3]).
26 Oliver Lorscheid

For a blueprint B, the set GT(B) of Tits morphisms from Spec B to G inherits the
structure of an associative unital semigroup. In case G has several connected components,
we define the rank of G as the rank of the connected component of G that contains the
image of the unit  : Spec F1 → G.
Theorem 3.7 ([Lor12c, Thm. 3.14]). Let G be an affine smooth group scheme of finite
type. If G has a Tits-Weyl model G, then the following properties hold true.
(i) The Weyl group W (G) is canonically isomorphic to the ordinary Weyl group W
of G .
(ii) The rank of G is equal to the reductive rank of G .
(iii) The semigroup HomT (Spec F1 , G) is canonically a subgroup of W (G).
(iv) If G is a split reductive group scheme, then HomT (Spec F12 , G) is canonically
isomorphic to the extended Weyl group W e of G .

The following theorem is proven in [Lor12c] for a large class of split reductive group
schemes G and their Levi- and parabolic subgroups. An additional idea of Reineke
extended this result to all split reductive group schemes; cf. [Lor16].
Theorem 3.8. (i) Every split reductive group scheme G has a Tits-Weyl model G.
(ii) Let T be the canonical torus of G and M a Levi subgroup of G containing T .
Then M has a Tits-Weyl model M that comes together with a locally closed
embedding M → G of Tits-monoids that is a Tits morphism.
(iii) Let P a parabolic subgroup of G containing T . Then P has a Tits-Weyl model
P that comes together with a locally closed embedding P → G of Tits-monoids
that is a Tits morphism.
(iv) Let U be the unipotent radical of a parabolic subgroup P of GLn,Z that
contains the diagonal torus T . Then U , P and GLn,Z have respective Tits-Weyl
models U, P and GLn,F1 , together with locally closed embeddings U → P →
GLn,F1 of Tits-monoids that are Tits morphisms and such that T is the canonical
torus of P and GLn,Z .
3.2.5. Tits’ dream. With the formalism developed in the previous sections, it is possible
to make Tits’ idea precise: the combinatorial counterparts of geometries over Fq can be
seen as geometries over F1 . This is explained in detail in [Lor16, section 6], using the
language of buildings.
We explain how this works in the previously considered example GL(3), cf. Example
1.1. Similar to the example of SL(2), the group scheme GL(3)Z has a Tits-Weyl model
GL(3)F1 , which is a monoid in the Tits category. The Weyl extension of GL(3)F1 is the
symmetric group S3 on three elements, which equals the Weyl group of GL(3).
The standard action of GL(3)Z on the Grassmannians Gr(k, 3)Z for k = 1, 2 descends
to monoid actions of GL(3)F1 on the standard F1 -models Gr(k, 3)F1 of the Grassman-
nians, considered as objects of the Tits category SchT . The Weyl extension of these
actions corresponds to the action of S3 on the set Σ(1, 3)
= {1, 2, 3} for k = 1 and to the
action of S3 on the set Σ(2, 3) = {1, 2}, {1, 3}, {2, 3} for k = 2.
Replacing Hom(Spec F1 , X) by W (X) makes precise the heuristics
W (GL(3)F1 ) = S3 and W (Gr(k, 3)) = Σ(k, 3).
This example is generalized to all other classical groups in [Lor16, section 6].
 F1 for everyone 27

3.3. Applications to tropical geometry. An exciting new application of F1 -geometry

lies in the field of tropical geometry. Such an application was already foreseen in various
works on F1 -geometry, cf. [Dur07], [TV09] and [Lor12b], but it took until 2013 till
Jeffrey and Noah Giansiracusa ([GG16]) used F1 -schemes and semiring schemes to
underlay tropical varieties with the structure of a semiring scheme—at least in the case
of the tropicalization of a classical scheme.
We will not dwell into this theory, but restrict ourselves to an explanation of the
relevance of a scheme theoretic approach for tropical geometry and a description of the
first results in this direction. For more details, we refer to Giansiracusas’ papers [GG14]
and [GG16], to Maclagan and Rincón’s papers [MR14] and [MR16] as well as to the
author’s paper [Lor15].

3.3.1. Some history. Tropical geometry was born with Mikhalkin’s calculation ([Mik00])
of Gromov-Witten invariants around twenty years ago. He was able to convert the classi-
cal problem of counting the number of nodal algebraic curves passing through a given
number of points to a counting problem of tropical curves, and to solve the latter problem
by means of elementary combinatorics.
Since then tropical geometry has developed rapidly. While during the first decade
of tropical geometry, researchers concentrated mainly on applications in the vein of
Mikhalkin’s results, tropical geometry entered a second era with the fundamental works
of Kajiwara ([Kaj08]) and Payne ([Pay09]) around 10 years ago. They provided an
elegant framework for tropical geometry and tied it to nonarchimedean analytic geometry
in the sense of Berkovich. Soon after these insights, tropical geometry found applications
in Brill-Noether theory, moduli spaces, skeleta of Berkovich spaces and rational points.

3.3.2. Tropical varieties. In order to explain the relevance of tropical scheme theory,
we have to explain what a tropical variety is. To this end, it is easiest to work with the
max-plus algebras R ∪ {−∞} as the tropical numbers T where addition is the maximum,
a +trop b = max{a, b}, and its multiplication the usual addition, a ∗trop b = a + b, with
the obvious extensions of these operations to −∞; cf. Example 2.8 for alternative
descriptions of the tropical numbers. In this form, the multiplicative units of T are
T× = R.
In its simplest incarnation, a tropical variety is a closed subset of (T× )n = Rn together
with a subdivision into finitely many rational polyhedra and together with a weight
function on the top dimensional polyhedra that satisfies a certain balancing condition.
These definitions somewhat difficult to explain for higher dimensions—and we choose
to omit them—, but in the case of dimension 1, they boil down to the following.
A tropical curve in Rn is a finite graph with possibly unbounded edges, embedded in
n
R , together with a weight function in Z>0 on the edges such that all edges have rational
slope and such that for every vertex p the following balancing condition is satisfied:
given an edge e adjacent to p, let ve be primitive vector at p pointing in the direction of
e, i.e. ve is the shortest vector in Z2 − {0} such that ve is contained in the ray generated
by {q − p|q ∈ e} ⊂ R2 ; then we have
∑ ve = 0.
e adjacent to p

The ties to classical algebraic geometry are given in terms of tropicalizations. Let k
be an algebraically closed field together with a valuation v : k× → R with dense image.
28 Oliver Lorscheid

1
v1
1
v2 v3 1
balancing condition
1
u1 1v1 + 1v2 + 1v3 = 0
w3
u3 2 w1
1u1 + 1u2 + 2u3 = 0
w2
u2
2w1 + 3w2 + 1w3 = 0
1 3

Figure 9. A tropical curve in R2 and the balancing condition

Let X be a closed subvariety of the algebraic torus (k× )n . Then the coordinatewise
application of v defines a subset of Rn whose topological closure is defined as the
tropicalization X trop of X. The structure theorem of tropical geometry asserts that X trop
can be subdivided into finitely many polyhedra and that the classical variety X determines
weights for the top dimensional polyhedra that satisfy the balancing condition. This can
be extended to subvarieties of any toric variety, e.g. subvarieties of a projective space.

tropicalization
1
0 4
−1 2
0
−2 0
2 −2

Figure 10. Tropicalization of an elliptic curve, including its points at infinity

The central problem. The definition of a tropical variety is problematic in several

senses. First of all, the structure of a polyhedral complex of X trop is not determined by
the classical variety, but involves choices. Thus strictly speaking, the tropicalization of a
classical variety is not a tropical variety. Secondly, there is a discrepancy between the
polynomial algebra over the tropical numbers and the functions determined by these
polynomials: different polynomials can define the same function. These defects call for
more sophisticated foundations of tropical geometry.
3.3.3. Tropical scheme theory. The above mentioned problem has been overcome by
the paradigm changing approach of Jeff and Noah Giansiracusa ([GG16]), in which they
realize the tropicalization X trop of a classical variety X as the set of T-rational points
of a T-scheme X , which is a semiring scheme with a structure morphism to Spec T.
This development can be seen as the tropical analogue of the passage from classical
algebraic geometry, where varieties were considered as point sets in an ambient affine or
projective space, to modern algebraic geometry in the sense of Grothendieck.
Soon after the initial paper [GG16] appeared, Maclagan and Rincón ([MR14]) have
shown that the balancing condition can be recovered from the structure as a T-scheme
together with its embedding into an algebraic torus.
 F1 for everyone 29

To complete the step from embedded varieties to abstract schemes, the author has
developed in [Lor15] a coordinate-free theory of tropicalizations, which is based on
blueprints and blue schemes. In more detail, the Giansiracusa tropicalization of a
classical variety is not only a semiring scheme, but comes with the richer structure as
a blue scheme. The structure of the tropicalization as a blue scheme is sufficient to
determine the weights of the underlying tropical variety. This makes it possible to pass
from embedded tropical schemes to abstract tropical schemes. The gain of this change
of perspective is that it applies, under suitable conditions, to more general situations,
such as skeleta of Berkovich spaces and tropicalizations of moduli spaces of curves.
3.3.4. Future applications. At the time of writing, there are high hopes that this new
approach to tropical geometry will lead to a realization of a conjectured tropical sheaf
cohomology and subsequently allows for progress in Brill-Noether and Baker-Norine
theory. It also might put tropical intersection theory on a new footing. Finally, we expect
that tropical scheme theory will interplay with Connes and Consani’s program around
the Riemann hypothesis, which is based on idempotent semirings as tropical scheme
theory is.

References
[ABD+ 64a] M. Artin, J. E. Bertin, M. Demazure, P. Gabriel, A. Grothendieck, M. Raynaud, and J.-P.
Serre. Schémas en groupes II: Groupes de Type Multiplicatif, et Structure des Schémas
en Groupes Généraux. Lecture Notes in Mathematics, Vol. 152. Springer-Verlag, Berlin,
1962/64.
[ABD+ 64b] M. Artin, J. E. Bertin, M. Demazure, P. Gabriel, A. Grothendieck, M. Raynaud, and J.-P.
Serre. Schémas en groupes. III: Structure des schémas en groupes réductifs. Lecture Notes
in Mathematics, Vol. 153. Springer-Verlag, Berlin, 1962/64.
[BCRW08] Peter Borwein, Stephen Choi, Brendan Rooney, and Andrea Weirathmueller, editors. The
Riemann hypothesis. CMS Books in Mathematics/Ouvrages de Mathématiques de la SMC.
Springer, New York, 2008. A resource for the afficionado and virtuoso alike.
[Ber] Vladimir G. Berkovich. Analytic geometry over F1 . Slides, 2011. Online available at
www.wisdom.weizmann.ac.il/~vova/Padova-slides_2011.pdf.
[Bor09] James Borger. Λ-rings and the field with one element. Preprint, arXiv:0906.3146, 2009.
[CC10a] Alain Connes and Caterina Consani. From monoids to hyperstructures: in search of an
absolute arithmetic. In Casimir force, Casimir operators and the Riemann hypothesis, pages
147–198. Walter de Gruyter, Berlin, 2010.
[CC10b] Alain Connes and Caterina Consani. Schemes over F1 and zeta functions. Compos. Math.,
146(6):1383–1415, 2010.
[CC11a] Alain Connes and Caterina Consani. The hyperring of adèle classes. J. Number Theory,
131(2):159–194, 2011.
[CC11b] Alain Connes and Caterina Consani. On the notion of geometry over F1 . J. Algebraic Geom.,
20(3):525–557, 2011.
[CC15] Alain Connes and Caterina Consani. Universal thickening of the field of real numbers. In
Advances in the theory of numbers, volume 77 of Fields Inst. Commun., pages 11–74. Fields
Inst. Res. Math. Sci., Toronto, ON, 2015.
[CC16] Alain Connes and Caterina Consani. Geometry of the arithmetic site. Adv. Math., 291:274–
329, 2016.
[CC17] Alain Connes and Caterina Consani. Geometry of the scaling site. Selecta Math. (N.S.),
23(3):1803–1850, 2017.
[CCM09] Alain Connes, Caterina Consani, and Matilde Marcolli. Fun with F1 . J. Number Theory,
129(6):1532–1561, 2009.
[CHWW15] Guillermo Cortiñas, Christian Haesemeyer, Mark E. Walker, and Charles Weibel. Toric
varieties, monoid schemes and cdh descent. J. Reine Angew. Math., 698:1–54, 2015.
30 Oliver Lorscheid

[CLS12] Chenghao Chu, Oliver Lorscheid, and Rekha Santhanam. Sheaves and K-theory for F1 -
schemes. Adv. Math., 229(4):2239–2286, 2012.
[Dei05] Anton Deitmar. Schemes over F1 . In Number fields and function fields—two parallel worlds,
volume 239 of Progr. Math., pages 87–100. Birkhäuser Boston, Boston, MA, 2005.
[Dei06] Anton Deitmar. Remarks on zeta functions and K-theory over F1 . Proc. Japan Acad. Ser. A
Math. Sci., 82(8):141–146, 2006.
[Dei08] Anton Deitmar. F1 -schemes and toric varieties. Beiträge Algebra Geom., 49(2):517–525,
2008.
[Dei11] Anton Deitmar. Congruence schemes. Preprint, arXiv:1102.4046, 2011.
[Den91] Christopher Deninger. On the Γ-factors attached to motives. Invent. Math., 104(2):245–261,
1991.
[Den92] Christopher Deninger. Local L-factors of motives and regularized determinants. Invent.
Math., 107(1):135–150, 1992.
[Dur07] Nikolai Durov. New approach to arakelov geometry. Thesis, arXiv:0704.2030, 2007.
[FLS17] Jaret Flores, Oliver Lorscheid, and Matt Szczesny. Čech cohomology over F12 . J. Algebra,
485:269–287, 2017.
[GG14] Jeffrey Giansiracusa and Noah Giansiracusa. The universal tropicalization and the Berkovich
analytification. Preprint, arXiv:1410.4348, 2014.
[GG16] Jeffrey Giansiracusa and Noah Giansiracusa. Equations of tropical varieties. Duke Math. J.,
165(18):3379–3433, 2016.
[Har07] M. J. Shai Haran. Non-additive geometry. Compos. Math., 143(3):618–688, 2007.
[Har10] Shai M. J. Haran. Invitation to nonadditive arithmetical geometry. In Casimir force, Casimir
operators and the Riemann hypothesis, pages 249–265. Walter de Gruyter, Berlin, 2010.
[Har17] M. J. Shai Haran. New foundations for geometry-two non-additive languages for arithmetic
geometry. Mem. Amer. Math. Soc., 246(1166):x+200, 2017.
[Has36] Helmut Hasse. Zur Theorie der abstrakten elliptischen Funktionenkörper III. Die Struktur des
Meromorphismenrings. Die Riemannsche Vermutung. J. Reine Angew. Math., 175:193–208,
1936.
[Jun15] Jaiung Jun. Algebraic geometry over hyperrings. Preprint, arXiv:1512.04837, 2015.
[Kaj08] Takeshi Kajiwara. Tropical toric geometry. In Toric topology, volume 460 of Contemp.
Math., pages 197–207. Amer. Math. Soc., Providence, RI, 2008.
[Kat94] Kazuya Kato. Toric singularities. Amer. J. Math., 116(5):1073–1099, 1994.
[KS95] Mikhail Kapranov and Alexander Smirnov. Cohomology determinants and reciprocity laws:
number field case. Unpublished preprint, 1995.
[Kur92] Nobushige Kurokawa. Multiple zeta functions: an example. In Zeta functions in geometry
(Tokyo, 1990), volume 21 of Adv. Stud. Pure Math., pages 219–226. Kinokuniya, Tokyo,
1992.
[Les11] Paul Lescot. Absolute algebra II—ideals and spectra. J. Pure Appl. Algebra, 215(7):1782–
1790, 2011.
[Les12] Paul Lescot. Absolute algebra III—the saturated spectrum. J. Pure Appl. Algebra,
216(5):1004–1015, 2012.
[Lor12a] Oliver Lorscheid. Algebraic groups over the field with one element. Math. Z., 271(1-2):117–
138, 2012.
[Lor12b] Oliver Lorscheid. The geometry of blueprints. Part I: Algebraic background and scheme
theory. Adv. Math., 229(3):1804–1846, 2012.
[Lor12c] Oliver Lorscheid. The geometry of blueprints. Part II: Tits-Weyl models of algebraic groups.
Preprint, arXiv:1201.1324, 2012.
[Lor14] Oliver Lorscheid. Blueprints—towards absolute arithmetic? J. Number Theory, 144:408–
421, 2014.
[Lor15] Oliver Lorscheid. Scheme theoretic tropicalization. Preprint, arXiv:1508.07949, 2015.
[Lor16] Oliver Lorscheid. A blueprinted view on F1 -geometry. In Absolute arithmetic and F1 -
geometry (edited by Koen Thas). European Mathematical Society Publishing House, 2016.
[Lor17] Oliver Lorscheid. Blue schemes, semiring schemes, and relative schemes after Toën and
Vaquié. J. Algebra, 482:264–302, 2017.
 F1 for everyone 31

[LPL11a] Javier López Peña and Oliver Lorscheid. Mapping F1 -land: an overview of geometries over
the field with one element. In Noncommutative geometry, arithmetic, and related topics,
pages 241–265. Johns Hopkins Univ. Press, Baltimore, MD, 2011.
[LPL11b] Javier López Peña and Oliver Lorscheid. Torified varieties and their geometries over F1 .
Math. Z., 267(3-4):605–643, 2011.
[LPL12] Javier López Peña and Oliver Lorscheid. Projective geometry for blueprints. C. R. Math.
Acad. Sci. Paris, 350(9-10):455–458, 2012.
[Man95] Yuri I. Manin. Lectures on zeta functions and motives (according to Deninger and Kurokawa).
Astérisque, (228):4, 121–163, 1995. Columbia University Number Theory Seminar.
[Mik00] Gregory Mikhalkin. Real algebraic curves, the moment map and amoebas. Ann. of Math.
(2), 151(1):309–326, 2000.
[Moc12] Shinichi Mochizuki. Inter-universal Teichmüller theory I–IV. Preprint, 2012.
[Moc15] Shinichi Mochizuki. Topics in absolute anabelian geometry III: global reconstruction algo-
rithms. J. Math. Sci. Univ. Tokyo, 22(4):939–1156, 2015.
[MR14] Diane Maclagan and Felipe Rincón. Tropical schemes, tropical cycles, and valuated matroids.
Preprint, arXiv:1401.4654, 2014.
[MR16] Diane Maclagan and Felipe Rincón. Tropical ideals. Preprint, arXiv:1609.03838, 2016.
[NR16] John Forbes Nash, Jr. and Michael Th. Rassias, editors. Open problems in mathematics.
Springer, [Cham], 2016.
[Pay09] Sam Payne. Analytification is the limit of all tropicalizations. Math. Res. Lett., 16(3):543–
556, 2009.
[Rou09] Guy Rousseau. Les immeubles, une théorie de Jacques Tits, prix Abel 2008. Gaz. Math.,
(121):47–64, 2009.
[Smi92] Alexander Smirnov. Hurwitz inequalities for number fields. Algebra i Analiz, 4(2):186–209,
1992.
[Sou99] Christophe Soulé. On the field with one element. Lecture notes from the Arbeitstagung 1999
of the Max Planck Institute for Mathematics. Online available at www.mpim-bonn.mpg.
de/preblob/175, 1999.
[Sou04] Christophe Soulé. Les variétés sur le corps à un élément. Mosc. Math. J., 4(1):217–244, 312,
2004.
[Sou11] Christophe Soulé. Lectures on algebraic varieties over F1 . In Noncommutative geometry,
arithmetic, and related topics, pages 267–277. Johns Hopkins Univ. Press, Baltimore, MD,
2011.
[Tak12] Satoshi Takagi. Compactifying Spec Z. Preprint, arXiv:1203.4914, 2012.
[Tit57] Jacques Tits. Sur les analogues algébriques des groupes semi-simples complexes. In Col-
loque d’algèbre supérieure, tenu à Bruxelles du 19 au 22 décembre 1956, Centre Belge de
Recherches Mathématiques, pages 261–289. Établissements Ceuterick, Louvain, 1957.
[Tit66] Jacques Tits. Normalisateurs de tores I. Groupes de Coxeter étendus. J. Algebra, 4:96–116,
1966.
[TV09] Bertrand Toën and Michel Vaquié. Au-dessous de Spec Z. J. K-Theory, 3(3):437–500, 2009.
[Wei48] André Weil. Sur les courbes algébriques et les variétés qui s’en déduisent. Actualités Sci.
Ind., no. 1041 = Publ. Inst. Math. Univ. Strasbourg 7 (1945). Hermann et Cie., Paris, 1948.
E-mail address: [email protected]

I NSTITUTO NACIONAL DE M ATEMÁTICA P URA E A PLICADA , R IO DE JANEIRO , B RAZIL