The Cyclic and Epicyclic Sites 2015

Rend. Sem. Mat. Univ.
Padova 1xx (201x) Rendiconti del Seminario Matematico

della Università di Padova
© European Mathematical Society
The Cyclic and Epicyclic Sites
Alain Connes – Caterina Consani∗
Abstract – We determine the points of the epicyclic topos which plays a key role
in the geometric encoding of cyclic homology and the lambda operations. We show
that the category of points of the epicyclic topos is equivalent to projective geometry in
characteristic one over algebraic extensions of the infinite semifield of “max-plus integers”
Zmax . An object of this category is a pair (E, K) of a semimodule E over an algebraic
extension K of Zmax . The morphisms are projective classes of semilinear maps between
semimodules. The epicyclic topos sits over the arithmetic topos N ̂× of [6] and the fibers
of the associated geometric morphism correspond to the cyclic site. In two appendices we
review the role of the cyclic and epicyclic toposes as the geometric structures supporting
cyclic homology and the lambda operations.
Mathematics Subject Classification (2010). 18B25; 20L05, 19D55.
Keywords. Grothendieck topos, Cyclic category, Groupoids, Characteristic one, Pro-

jective geometry.
1. Introduction
The theory of topoi of Grothendieck provides the best geometric framework to

understand cyclic homology and the λ-operations using the topos associated to
the cyclic category [3] and its epicyclic refinement [5]. Given a topos T a basic
question is to determine the category of points of T , i.e. of geometric morphisms
from the topos of sets to T . In this paper we show how to describe the category
of points of the epicyclic topos in terms of projective geometry in characteristic
1. Given a small category C, we denote by Cˆ the topos of contravariant functors
∗ C. Consani would like to thank the Collège de France for some financial support.
Alain Connes, Collège de France, 3 rue d’Ulm, Paris F-75005 France

I.H.E.S. and Ohio State University
E-mail: [email protected]
Caterina Consani, Department of Mathematics, The Johns Hopkins University
Baltimore MD 21218 USA
E-mail: [email protected]
2 Alain Connes – Caterina Consani
from C to the category Sets of sets. The epicyclic topos (Λ̃op )∧ is obtained by
taking the opposite of the epicyclic category Λ̃. This choice is dictated by the
following natural construction. A commutative ring R determines a covariant
functor (−)♮ ∶ Fin Ð→ Ab from the category of finite sets to that of abelian groups.
This functor assigns to a finite set J the tensor power R⊗J = ⊗j∈J R. As explained
in geometric terms here below, there is also a natural covariant functor Λ̃ Ð→ Fin.
The composite of these two functors Λ̃ Ð→ Ab provides, for any commutative
ring R, a covariant functor R♮ from the epicyclic category to the category of
abelian groups. In geometric terms R♮ is a sheaf of abelian groups over the topos
(Λ̃op )∧ . Both the cyclic homology of R and its λ-operations are completely encoded
by the associated sheaf R♮ . In [5], we provided a conceptual understanding of
the epicyclic category as projective geometry over the semifield F ∶= Zmax of the
tropical integers. In these terms the functor (−)♮ considered above assigns to a
projective space the underlying finite set. This article pursues the relation between
the epicyclic topos and (projective) geometry in characteristic 1 in more details.
Our main result is the following (cf. Theorem 4.1)
Theorem. The category of points of the epicyclic topos (Λ̃op )∧ is equivalent to

the category P whose objects are pairs (K, E), where K is an algebraic extension
of F = Zmax and E is an archimedean semimodule over K. The morphisms in P
are projective classes of semilinear maps and injective semifield morphisms.
It is important to realize the relevance of the language of Grothendieck topoi to

interpret, for instance, the action of the edgewise subdivision on the points of the
ˆ The preliminary Section 2 is dedicated to this description. It
simplicial topos ∆.
is well known (cf. [10]) that the points of ∆ ˆ correspond to intervals, i.e. totally
ordered sets I with a smallest element b and a largest element t ≠ b. For each
integer k > 0, the edgewise subdivision Sdk defines an endofunctor of the simplicial
category ∆ and one obtains in this way an action of the monoı̈d N× by geometric
morphisms on the topos ∆. ˆ We show that the action of the edgewise subdivision
on the points of ∆ ˆ is given by the operation of concatenation of k copies of the
interval I: the intermediate top point tj of the copy Ij is identified with the bottom
point bj+1 of the subsequent copy Ij+1 . Then, we form the small category ∆op ⋉N×
crossed product of ∆op by the transposed action Sd∗ of N× (i.e. Sdk (f )∗ = Sd∗k (f ∗ ),
where f ↦ f ∗ is the anti-isomorphism ∆ Ð→ ∆op ).
Section 3 gives a description of the epicyclic category in terms of oriented groupoı̈ds.
The ambiguı̈ty in the choice of a representative of a projective class of semilinear
maps in the category P of Theorem 4.1 is inconvenient when working, for example,
with colimits. In Section 3.3 we provide a description of the cyclic and the epicyclic
categories in terms of a category g of oriented groupoı̈ds whose morphisms are
no longer given by equivalence classes. There are by now a number of equivalent
descriptions of the cyclic and epicyclic categories, ranging from the most concrete
i.e. given in terms of generators and relations, to the most conceptual as in [5].
The description of these categories in terms of oriented groupoı̈ds turns out to be
very useful to determine the points of the epicyclic topos by considering filtering
The Cyclic and Epicyclic Sites 3
colimits, in the category g, of the special points provided by the Yoneda embedding
of the categories. It is in fact well known that any point of a topos of the form
Cˆ is obtained as a filtering colimit, in the category of flat functors C Ð→ Sets, of
these special points. On the other hand, there is no guarantee “a priori” that this
colimit process yields the same result as the colimit taken in the category g. This
matter is solved in two steps and in concrete terms in Section 4. In Proposition
4.3 we show how to associate to a pair (K, E) as in the above Theorem a point of
(Λ̃op )∧ . Conversely, in §§4.2-4.4 we explain a geometric procedure that allows one
to reconstruct the structure of an oriented groupoı̈d from the flat functor naturally
associated to a point of (Λ̃op )∧ .
In §4.8 we explore the relations of (Λ̃op )∧ with the arithmetic site N̂× , as recently
defined in [6]. Let N× be the small category with a single object ● and whose endo-
morphisms End(●) = N× form the multiplicative semigroup N× of positive integers.
One has a canonical functor Mod ∶ Λ̃op Ð→ N× which is trivial on the objects and
associates to a semilinear map of semimodules over F = Zmax the corresponding
injective endomorphism Frn ∈ End(F) (cf. [5] for details). This functor induces a
geometric morphism of topoi Mod ∶ (Λ̃op )∧ Ð→ N ̂× . The subcategory of Λ̃op which
is the kernel of this morphism is the cyclic category Λ (Λ ≃ Λop ).
In Appendix Appendix A we view the λ-operations as elements Λkn of the convolu-
tion ring Z[∆op ⋉N× ] with integral coefficients. We review their geometric meaning
and the geometric proof of their commutation (cf. [9]) with the Hochschild bound-
ary operator.
Appendix Appendix B is dedicated to the description of the cyclic homology of
cyclic modules (cf. [3]) and its extension to epicyclic modules [8]. We stress the
nuance between Λ̃ and Λ̃op in a hopefully clear form. An epicyclic module E is
a covariant functor Λ̃ Ð→ Ab. These modules correspond to sheaves of abelian
groups on the topos (Λ̃op )∧ . At this point the nuance between the epicyclic cat-
egory and its dual plays an important role since unlike the cyclic category the
epicyclic category is not anti-isomorphic to itself. As explained earlier on in this
introduction, a commutative ring R gives rise naturally to an epicyclic module R♮
and it is well known (cf. [8]) that the λ-operations on cyclic homology of R are
obtained directly through the associated epicyclic module. We provide a simple
and conceptual proof of the commutation of the λ-operations with the B operator
of cyclic theory. Finally, we point out that the extended framework of epicyclic
modules involves many more modules than those arising by composition, as ex-
plained earlier, from a covariant functor Fin Ð→ Ab. In fact, these particular
(epicyclic) modules have integral weights and the λ-operations decompose their
cyclic homology as direct sums of modules on which Λkn acts by an integral power
of k. This integrality property no longer holds for general epicyclic modules as
can be easily checked by applying a twisting argument.
ˆ
2. The action of the edgewise subdivision on points of ∆
We recall that the simplicial category ∆ is the small category with objects the
totally ordered sets [n] ∶= {0, . . . , n}, for each integer n ≥ 0, and morphisms non-
decreasing maps.
In this section we study, using the formalism of topoi, the edgewise subdivision
functors Sdk ∶ ∆ Ð→ ∆, for k ∈ N× and their action on the points of the simplicial
ˆ
topos ∆.
2.1 – The edgewise subdivision functors Sdk

Let F be a finite, totally ordered set and k ∈ N× a positive integer. We define the
set
(1) Sdk (F ) ∶= {0, . . . , k − 1} × F
to be the cartesian product of the finite ordered set {0, . . . , k − 1} with F , endowed
with the lexicographic ordering. For f ∈ Hom∆ (F, F ′ ) a non-decreasing map (of
finite, totally ordered sets), we let
(2) Sdk (f ) ∶= Id × f ∶ Sdk (F ) → Sdk (F ′ )
Proposition 2.1. For each k ∈ N× , (1) and (2) define an endofunctor Sdk ∶ ∆ Ð→
∆. They fulfill the property
Sdkk′ = Sdk ○ Sdk′ , ∀k, k ′ ∈ N× .
Proof. The totally ordered sets Sdk ([n]) and [k(n + 1) − 1] have the same cardi-
nality and are canonically isomorphic. The unique increasing bijection Sdk ([n]) →
[k(n + 1) − 1] is given by
(a, i) ↦ i + a(n + 1) , ∀a ∈ {0, . . . , k − 1}, i ∈ {0, . . . , n}.
Let f ∈ Hom∆ ([n], [m]) then by definition Sdk (f ) ∈ Hom∆ (Sdk ([n]), Sdk ([m]))
is given by
(3) Sdk (f )(i + a(n + 1)) = f (i) + a(m + 1) , ∀i, a, 0 ≤ i ≤ n, 0 ≤ a ≤ k − 1.
One checks directly that Sdkk′ = Sdk ○ Sdk′ .
We transfer the functors Sdk to the opposite category ∆op of finite intervals. Recall
that by definition, an interval I is a totally ordered set with a smallest element
b and a largest elements t ≠ b. The morphisms between intervals are the non-
decreasing maps respecting b and t, i.e. f ∶ I → J, f (bI ) = bJ , f (tI ) = tJ .
For all n ≥ 0 we denote by n∗ ∶= {0, . . . , n + 1}. The interval n∗ parametrizes the
hereditary subsets of [n]: indeed, to j ∈ n∗ corresponds [j, n] ∶= {x ∈ [n] ∣ x ≥ j},
the latter set is empty for j = n+1. The duality between ∆ and ∆op is then provided
∼
by the contravariant functor ∆ Ð→ ∆op , [n] ↦ n∗ , which acts on morphisms as
follows
(4)
Hom∆ ([n], [m]) ∋ f → f ∗ ∈ Hom∆op (m∗ , n∗ ), f −1 ([j, m]) = [f ∗ (j), n] , ∀j ∈ m∗
Let I be an interval, and k ∈ N× , then one lets Sd∗k (I) to be the quotient of the
totally ordered set {0, . . . , k − 1} × I = Sdk (I) (with lexicographic ordering) by
the equivalence relation (j, tI ) ∼ (j + 1, bI ) for j ∈ {0, . . . , k − 2}. This defines an
endofunctor Sd∗k of the category of intervals whose action on morphisms sends
f ∶ I → J to Sd∗k (f ) = Id × f . By restriction to finite intervals one obtains an
endofunctor Sd∗k ∶ ∆op Ð→ ∆op .
In particular, the interval Sd∗k (n∗ ) has k(n + 2) − (k − 1) = k(n + 1) + 1 elements
and one obtains a canonical identification of Sd∗k (n∗ ) with the hereditary subsets
of Sdk ([n]) as follows
{0, . . . , k − 1} × n∗ ∋ (b, j) ↦ {(a, i) ∈ {0, . . . , k − 1} × [n] ∣ a > b or a = b & j ≥ i}.
Note that the right hand side of the above formula depends only upon the class
of (b, j) ∈ Sd∗k (n∗ ).
Lemma 2.2. For f ∈ Hom∆ ([n], [m]), one has (Sdk (f ))∗ = Sd∗k (f ∗ ).
Proof. The morphism (Sdk (f ))∗ is defined by the equivalence

Sdk (f )(x) ≥ y ⇐⇒ (Sdk (f ))∗ (y) ≤ x
Let x = i + a(n + 1), y = j + b(m + 1) with 0 ≤ i ≤ n, 0 ≤ j ≤ m, 0 ≤ a ≤ k − 1,
0 ≤ b ≤ k − 1. Then by (3) one has Sdk (f )(x) = f (i) + a(m + 1), thus the condition
Sdk (f )(x) ≥ y determines (Sdk (f ))∗ (y) as follows
f (i) + a(m + 1) ≥ j + b(m + 1) ⇐⇒ a > b or a = b and f (i) ≥ j
⇐⇒ i + a(n + 1) ≥ f ∗ (j) + b(n + 1) = Sd∗k (f ∗ )(y).
This provides the required equality (Sdk (f ))∗ = Sd∗k (f ∗ ).
Theorem 2.3. The action Sd ̂k of the geometric morphism Sdk (k ∈ N× ) on the

ˆ is described by the endofunctor Sd∗k on the category of inter-
points of the topos ∆
vals.
Proof. One can prove this theorem using the fact that any point of ∆ ˆ is obtained
as a filtering colimit of the points associated to the Yoneda embedding of ∆op in
ˆ k coincides
ˆ One shows that on such points the action Sd
the category of points of ∆.
with the functor Sd∗k ∶ ∆op Ð→ ∆op . We shall nevertheless find it more instructive
to give, in §2.3, a concrete direct proof of the equality between the following two
flat functors ∆ Ð→ Sets associated to an interval I
(5) F1 ([n]) = Hom∆op (n∗ , Sd∗k (I)), F2 ([n]) =
= ( ∐ (Hom(m∗ , I) × Hom∆op (n∗ , Sd∗k (m∗ )))) / ∼

m≥0
ˆ applied to the contravariant

here F2 is the inverse image functor of the point pI of ∆
functor Y ∶ ∆ Ð→ Sets, Y = X ○ Sdk , where X = h[n] is the Yoneda embedding, so
that
(6) Y ([m]) = X(Sdk ([m])) = Hom∆ (Sdk ([m]), [n]) = Hom∆op (n∗ , Sd∗k (m∗ )).
Thus F2 corresponds to the point Sd̂k (pI ) and the equality between F1 and F2
(cf. Lemma 2.7) yields the result.
Corollary 2.4. : The point of the simplicial topos ∆ ˆ associated to the interval
×
[0, 1] ⊂ R is a fixed point for the action of N on ∆.
ˆ
Proof. The statement follows using the affine isomorphism

a x
(7) Sd∗k ([0, 1]) = {0, . . . , k − 1} × [0, 1] → [0, 1], (a, x) ↦ + .
k k
2.2 – Canonical decomposition of ϕ ∈ Hom∆op (n∗ , Sd∗k (I))

Let I be an interval, I/ ∼ be the quotient of I by the identification b ∼ t. Consider
the map π ∶ Sd∗k (I) → I/ ∼, (j, x) ↦ x. For ϕ ∈ Hom∆op (n∗ , Sd∗k (I)), we define the
rank of ϕ as the cardinality of the set Z = I o ∩ Range(π ○ ϕ), where I o = I ∖ {b, t}.
Proposition 2.5. Let ϕ ∈ Hom∆op (n∗ , Sd∗k (I)) and r its rank. Then, one has a
unique decomposition
ϕ = Sd∗k (α) ○ β, β ∈ Hom∆op (n∗ , Sd∗k (r∗ )), α ∈ Hom∆op (r∗ , I).
Moreover, the morphism α ∈ Hom∆op (r∗ , I) is the unique increasing injection

{1, . . . , r} ↪ I o which admits Z as range. The composite πr ○ β is surjective,
where πr ∶ Sd∗k (r∗ ) → r∗ / ∼ is the canonical surjection.
Proof. Let α ∈ Hom∆op (r∗ , I) be the map whose restriction to {1, . . . , r} is the
unique increasing injection into I o which admits Z as range. Recall that an el-
ement x ∈ Sd∗k (I) is given by a pair x = (j, y) ∈ {0, . . . , k − 1} × I = Sdk (I) with
the identifications (j, t) ∼ (j + 1, b) for j ∈ {0, . . . , k − 2}. Similarly an element
z ∈ Sd∗k (r∗ ) is given by a pair z = (i, u) ∈ {0, . . . , k − 1} × r∗ = Sdk (r∗ ) with the iden-
tifications (j, r + 1) ∼ (j + 1, 0) for j ∈ {0, . . . , k − 2}. Let s ∈ n∗ = {0, . . . , n + 1}, then
ϕ(s) ∈ Sd∗k (I) is given by a pair ϕ(s) = (j, y) ∈ {0, . . . , k − 1} × I = Sdk (I) unique
up to the above identifications. If y ∈ {b, t}, one defines β(s) = (j, 0) ∈ Sd∗k (r∗ )
if y = b, and β(s) ∶= (j, r + 1) ∈ Sd∗k (r∗ ) if y = t. This definition is compatible
with the identifications. Let us now assume that y ∉ {b, t}. Then y ∈ Z and
there exists a unique element v ∈ {1, . . . , r} such that y = α(v). One then defines
β(s) ∶= (j, v) ∈ Sd∗k (r∗ ). The map β ∶ n∗ → Sd∗k (r∗ ) so defined is non-decreasing,
i.e. for s < s′ one has β(s′ ) ≥ β(s) since the inequality ϕ(s′ ) ≥ ϕ(s) shows that
either j ′ > j in which case (j ′ , v ′ ) ≥ (j, v) is automatic, or j = j ′ and in that
case y ′ > y which shows that v ′ ≥ v. Moreover since Sd∗k (α) = Id × α one has
ϕ = Sd∗k (α) ○ β.
We prove the uniqueness of this decomposition. Since ϕ ∈ Hom∆op (n∗ , Sd∗k (I))
preserves the base points, Range (π ○ ϕ) contains the base point and its cardi-
nality is r + 1. Thus the map α ∈ Hom∆op (r∗ , I) is the unique map whose re-
striction to {1, . . . , r} is the increasing injection to I o and which admits Z = I o ∩
Range (π ○ ϕ) as range. Moreover α is injective and so is Sd∗k (α). Thus the map
β ∈ Hom∆op (n∗ , Sd∗k (r∗ )) is uniquely determined by the equality ϕ = Sd∗k (α) ○ β.
Finally πr ○ β ∶ n∗ → r∗ / ∼ is surjective since otherwise the range of Sd∗k (α) ○ β
would be strictly smaller than the range of ϕ.
Corollary 2.6. For any interval I the map
Hom∆op (n∗ , I)×Hom∆op (n∗ , Sd∗k (n∗ )) → Hom∆op (n∗ , Sd∗k (I)), (α, β) ↦ Sd∗k (α)○β
is surjective.
2.3 – Explicit description of the isomorphism F1 ≃ F2

Let Fj ∶ ∆ Ð→ Sets be the flat functors defined in (5). By definition
(8) F2 ([n]) = ( ∐ (Hom(m∗ , I) × Hom∆op (n∗ , Sd∗k (m∗ )))) / ∼

m≥0
where the equivalence relation is generated by
(α ○ f, β) ∼ (α, Sd∗k (f ) ○ β)
for f ∈ Hom∆op (m∗ , r∗ ), β ∈ Hom∆op (n∗ , Sd∗k (m∗ )), α ∈ Hom(r∗ , I).
Lemma 2.7. The map
Φ ∶ F2 ([n]) → F1 ([n]) = Hom∆op (n∗ , Sd∗k (I)), (α, β) ↦ Sd∗k (α) ○ β
is a bijection of sets.
Proof. The map Φ is well defined since Φ(α ○ f, β) = Φ(α, Sd∗k (f ) ○ β). Corollary
2.6 shows that Φ is surjective. To show the injectivity it is enough to prove that for
any (α, β) ∈ Hom(m∗ , I) × Hom∆op (n∗ , Sd∗k (m∗ )) one has (α, β) ∼ (αc , βc ) where
ϕ = Sd∗k (αc ) ○ βc , βc ∈ Hom∆op (n∗ , Sd∗k (r∗ )), αc ∈ Hom∆op (r∗ , I)
is the canonical decomposition of ϕ = Sd∗k (α) ○ β.

One has the canonical decomposition β = Sd∗k (α0 )○β0 with β0 ∈ Hom∆op (n∗ , Sd∗k (`∗ )),
Id` ○ β0 surjective. Thus (α, β) ∼ (α ○ α0 , β0 ). Since Id` ○ β0 is surjective, Range
α ○ α0 ⊂ Range Id ○ ϕ = Range αc , and thus α ○ α0 = αc ○ ρ
(α, β) ∼ (α ○ α0 , β0 ) = (αc ○ ρ, β0 ) ∼ (αc , Sd∗k (ρ) ○ β0 ) = (αc , βc ).

2.4 – The small category ∆op ⋉ N×

We denote by ∆op ⋉N× the small category semi-direct product of ∆op by the action
of N× implemented by the endofunctors Sd∗k , for k ∈ N× . It has the same objects as
∆op while one adjoins to the collection of morphisms of ∆op the new morphisms
∗
πnk ∶ Sd∗k (n∗ ) = (k(n + 1) − 1) → n∗ such that
∗
(9) πnk ○ πk(n+1)−1
`
= πnk` ∈ Hom∆op ⋉N× ((k`(n + 1) − 1) , n∗ )
where πnk implements the endofunctor Sd∗k , i.e.

(10) α ○ πnk = πm
k
○ Sd∗k (α) , ∀α ∈ Hom∆op (n∗ , m∗ ).
Using this set-up one checks that any morphism φ in ∆op ⋉ N× is uniquely of the
form φ = πnk ○ α with α a morphism in ∆op . Any such φ composes as follows
(11) (πm
k
○ β) ○ (πn` ○ α) = πm
k`
○ (Sd∗` (β) ○ α)
∗
where α ∈ Hom∆op (r∗ , (`(n + 1) − 1) ) and
∗ ∗
Sd∗` (β) ∈ Hom∆op ((`(n + 1) − 1) , (k`(m + 1) − 1) )
∗
so that Sd∗` (β) ○ α makes sense and belongs to Hom∆op (r∗ , (k`(m + 1) − 1) ). Us-
ing Proposition 2.1 one checks that, if one takes (11) as a definition, the product
is associative.
Let Fin∗ be the category of finite pointed sets and let F be the functor which
associates to an interval I the pointed set I∗ = I/ ∼ with base point the class
of b ∼ t. To any morphism of intervals f ∶ I → J corresponds the quotient map
f∗ which preserves the base point. By restricting F to ∆op one gets a covariant
functor F ∶ ∆op Ð→ Fin∗ . The following Proposition shows that F can be extended
to ∆op ⋉ N× .
Proposition 2.8. : For any n ≥ 0, k ∈ N× , let (πnk )∗ ∶ F(Sd∗k (n∗ )) → F(n∗ )

be given by the residue modulo n + 1. Then the extension of the functor F on
morphisms given by
φ = πnk ○ α ↦ φ∗ ∶= (πnk )∗ ○ α∗
determines a functor F ∶ ∆op ⋉ N× Ð→ Fin∗ .
Proof. One checks directly that the definition of (πnk )∗ is compatible with the
rules (9) and (10) so that the required functoriality follows.
3. The epicyclic category and the oriented groupoı̈ds
3.1 – Generalities on oriented groupoı̈ds

A groupoı̈d G is a small category where the morphisms are invertible. Given a
subset X ⊂ G of a groupoı̈d, we set X −1 ∶= {γ −1 ∣ γ ∈ X}. Let G(0) be the set
of objects of G and denote by r, s ∶ G → G(0) the range and the source maps
respectively. We view G(0) as the subset of units of G. The following definition is
a direct generalization to groupoı̈ds of the notion of right ordered group (cf. [7])
Definition 3.1. An oriented groupoı̈d (G, G+ ) is a groupoı̈d G endowed with a

subcategory G+ ⊂ G, such that the following relations hold
(12) G+ ∩ G−1
+ =G
(0)
, G+ ∪ G−1
+ = G.
Let (G, G+ ) be an oriented groupoı̈d and let x ∈ G(0) . The set Gx ∶= {γ ∈ G ∣

s(γ) = x} is endowed with the total order defined by
(13) γ ≤ γ ′ ⇐⇒ γ ′ ○ γ −1 ∈ G+ .
This order is right invariant by construction: i.e. for any β ∈ G, with r(β) = x, one
has
γ ≤ γ ′ ⇐⇒ γ ○ β ≤ γ ′ ○ β.
In the following subsections we describe two constructions of oriented groupoı̈ds

associated to a group action.
3.1.1 – G = X ⋉ H.
Let H be a group acting on a set X. Then the semi-direct product G ∶= X ⋉ H is
a groupoı̈d with source, range and composition law defined respectively as follows
s(x, h) ∶= x, r(x, h) ∶= hx, (x, h) ○ (y, k) ∶= (y, hk).
(As in any groupoı̈d the composition γ ○ γ ′ is only defined when s(γ) = r(γ ′ ) which
holds here if and only if x = ky). One has a canonical homomorphism of groupoı̈ds
ρ ∶ G → H, ρ(x, h) = h.
Lemma 3.2. Let (H, H+ ) be a right ordered group. Assume that H acts on a set
X. Then the semi-direct product G = X ⋉ H with G+ ∶= ρ−1 (H+ ) is an oriented
groupoı̈d.
Proof. By definition, the subset H+ ⊂ H of the group H is stable under product

and fulfills the equalities: H+ ∩ H+−1 = {1}, H+ ∪ H+−1 = H. This implies (12) using
ρ−1 ({1}) = G(0) .
Let, in particular, (H, H+ ) = (Z, Z+ ) act by translation on the set X = Z/(m + 1)Z
of integers modulo m + 1. Then one obtains the oriented groupoı̈d
(14) g(m) ∶= (Z/(m + 1)Z) ⋉ Z.
The oriented groupoı̈ds g(m) will play a crucial role in this article.
3.1.2 – G = (X × X)/H.
Let H be a group acting freely on a set X. Let G(X, H) = (X × X)/H be the
quotient of X × X by the diagonal action of H
G(X, H) ∶= (X × X)/ ∼ (x, y) ∼ (h(x), h(y)), ∀h ∈ H.
Let r and s be the two projections of G(X, H) on G(0) ∶= X/H defined by r(x, y) =
x and s(x, y) = y. Let γ, γ ′ ∈ G(X, H) be such that s(γ) = r(γ ′ ). Then, for
γ ∼ (x, y) and γ ′ ∼ (x′ , y ′ ), there exists a unique h ∈ H satisfying x′ = h(y): this
because s(γ) = r(γ ′ ) and H acts freely on X. Then, the pair (h(x), y ′ ) defines
an element of G(X, H) independent of the choice of the pairs representing the
elements γ and γ ′ . We denote by γ ○ γ ′ the class of (h(x), y ′ ) in G(X, H). This
construction defines a groupoı̈d law on G(X, H) = (X × X)/H.
Lemma 3.3. Let H be a group acting freely on a set X. Assume that X is totally
ordered and that H acts by order automorphisms. Then G(X, H) = (X × X)/H is
an oriented groupoı̈d with
(15) G+ (X, H) = {(x, y) ∈ G(X, H) ∣ x ≥ y}.
Proof. Since H acts by order automorphisms the condition x ≥ y is independent

of the choice of a representative (x, y) of a given γ ∈ G(X, H) = (X × X)/H. This
condition defines a subcategory G+ (X, H) of G(X, H). The conditions (12) then
follow since X is totally ordered.
Lemma 3.4. Let X = Z with the usual total order. Let m ∈ N and let the group
Z act on X by h(x) ∶= x + (m + 1)h, ∀x ∈ X, h ∈ Z. Then the oriented groupoı̈d
G = (X × X)/Z is canonically isomorphic to the oriented groupoı̈d g(m) of (14).
Proof. The associated oriented groupoı̈d (G, G+ ) is by construction the quotient

of Z × Z by the equivalence relation: (x, y) ∼ (x + `(m + 1), y + `(m + 1)), ∀` ∈ Z.
Thus the following map defines a bijective homomorphism of groupoids
ψ ∶ G → g(m) = (Z/(m + 1)Z) ⋉ Z, ψ(x, y) = (π(y), x − y)
where π ∶ Z → Z/(m + 1)Z is the natural projection. One has by restriction

ψ
G+ → g+ (m), since x ≥ y ⇐⇒ x − y ≥ 0, so that ψ is in fact an isomorphism of
∼
oriented groupoı̈ds.
3.2 – The oriented groupoı̈d associated to an archimedean set

In this section we explain how to associate an oriented groupoı̈d to an archimedean
set and describe the special properties of the oriented groupoı̈ds thus obtained. We
first recall from [4] the definition of an archimedean set.
Definition 3.5. An archimedean set is a pair (X, θ) of a non-empty, totally

ordered set X and an order automorphism θ ∈ AutX, such that θ(x) > x, ∀x ∈ X.
The automorphism θ is also required to fulfill the following archimedean property
∀x, y ∈ X, ∃n ∈ N s.t. y ≤ θn (x).
Let (X, θ) be an archimedean set and let G(X, θ) be the oriented groupoı̈d asso-
ciated by Lemma 3.3 to the action of Z on X by integral powers of θ. Thus
G(X, θ) ∶= (X × X)/ ∼, (x, y) ∼ (θn (x), θn (y)), ∀n ∈ Z
and
(16) G+ (X, θ) ∶= {(x, y) ∈ G(X, θ) ∣ x ≥ y}.
Next proposition describes the properties of the pair (G(X, θ), G+ (X, θ)) so ob-
tained.
Proposition 3.6. The oriented groupoı̈d (G, G+ ) = (G(X, θ), G+ (X, θ)) fulfills
the following conditions
(1) ∀x, y ∈ G(0) , ∃γ ∈ G+ s.t. s(γ) = y, r(γ) = x.
(2) For x ∈ X, the ordered groups Gxx ∶= {γ ∣ s(γ) = r(γ) = x} are isomorphic to
(Z, ≤).
(3) Let γ ∈ G with s(γ) = y and r(γ) = x. Then the map: Gyy ∋ ρ ↦ γ ○ρ○γ −1 ∈ Gxx
is an isomorphism of ordered groups.
Proof. Since θ is an order automorphism of X, the group Z acts by order auto-

morphisms. We check the three conditions (i)-(iii).
(i) For x, y ∈ X, there exists n ∈ N such that θn (x) ≥ y. Then γ ∼ (θn (x), y)
belongs to G+ and s(γ) = y, r(γ) = x.
(ii) Let x ∈ X. The conditions s(γ) = r(γ) = x imply that the class of γ ∈ G(X, θ)
admits a unique representative of the form (θn (x), x). One easily checks that the
map Gxx → (Z, ≤), (θn (x), x) ↦ n is an isomorphism of ordered groups.
(iii) Let x, y ∈ X with γ ∼ (x, y). Then for ρ ∼ (θn (y), y) ∈ Gyy one gets γ ○ ρ ○ γ −1 ∼
(θn (x), x), thus the unique isomorphism with (Z, ≤) is preserved.
Let (G, G+ ) be an oriented groupoı̈d fulfilling the three conditions of Proposition
3.6. Let x ∈ G(0) , consider the set Gx = {γ ∈ G ∣ s(γ) = x} with the total order
defined by (13) and with the action of Z given, for γx ∈ G+ the positive generator
of Gxx , by
(17) θ(γ) ∶= γ ○ γx .
When one applies this construction to the case (G, G+ ) = (G(X, θ), G(X, θ)+ ), for
(X, θ) an archimedean set, with x ∈ G(0) = X/θ one obtains, after choosing a lift
x̃ ∈ X of x, an isomorphism
∼
(18) jx̃ ∶ X → Gx , jx̃ (z) = (z, x).
The following proposition shows that the two constructions (X, θ) ↦ G(X, θ) and
(G, G+ ) ↦ (Gx , θ) are reciprocal.
Proposition 3.7. Let (G, G+ ) be an oriented groupoı̈d fulfilling the conditions of

Proposition 3.6 and let x ∈ G(0) . Consider the set Gx ∶= {γ ∈ G ∣ s(γ) = x} = X
endowed with the total order (13) and the action of Z on it given by (17). Then
(X, θ) is an archimedean set and one has an isomorphism of oriented groupoı̈ds
(G(X, θ), G(X, θ)+ ) ≅ (G, G+ ).
Proof. The implication γ ≤ γ ′ Ô⇒ θ(γ) ≤ θ(γ ′ ) follows since right multiplica-

tion preserves the order. Moreover, the condition (iii) of Proposition 3.6 implies
that θ(γ) > γ, ∀γ ∈ X = Gx . Next we show that the archimedean property holds
on (X, θ). Let γ ≤ γ ′ ∈ Gx , with y = r(γ) and y ′ = r(γ ′ ). By applying the condition
(i) of Proposition 3.6, we choose δ ∈ G+ such that s(δ) = y ′ and r(δ) = y. Then
γ ′′ = δ ○ γ ′ fulfills s(γ ′′ ) = s(γ) and r(γ ′′ ) = r(γ) and thus there exists n ∈ Z such
that γ ′′ = γ ○ γxn . Moreover, one has γ ′′ = δ ○ γ ′ ≥ γ ′ . It follows that (X, θ) is
an archimedean set. If one replaces x ∈ G(0) by y ∈ G(0) , then the condition (i)
implies that there exists α ∈ G+ with s(α) = y, r(α) = x. Then the map Gx → Gy ,
γ ↦ γ ○ α is an order isomorphism which satisfies
(γ ○ γx ) ○ α = (γ ○ α) ○ (α−1 ○ γx ○ α).
Since condition (iii) implies α−1 ○ γx ○ α = γy , one obtains an isomorphism of the
corresponding archimedean sets.
Finally, we compare the pair (G, G+ ) with (G(X, θ), G+ (X, θ)). We define a map
f ∶ G(X, θ) → G as follows: given a pair (γ, γ ′ ) of elements of X = Gx , one sets
f (γ, γ ′ ) ∶= γ ○ γ ′−1 . One has
f (θ(γ), θ(γ ′ )) = f (γ ○ γx , γ ′ ○ γx ) = γ ○ γ ′−1 = f (γ, γ ′ ).
To show that f is a groupoı̈d homomorphism it is enough to check that f (γ, γ ′ ) ○
f (γ ′ , γ ′′ ) = f (γ, γ ′′ ) and this can be easily verified. Next we prove that f is
bijective. Let α ∈ G. By applying condition (i) of Proposition 3.6, there exists
γ ∈ G such that r(γ) = r(α) and s(γ) = x. Let then γ ′ = α−1 γ. Since s(γ ′ ) = x,
both γ, γ ′ belong to X = Gx and moreover f (γ, γ ′ ) = α showing that f is surjective.
Let γj , γj′ be elements of X = Gx such that f (γ1 , γ1′ ) = f (γ2 , γ2′ ). One then has
γ1 ○ γ1′−1 = γ2 ○ γ2′−1 and hence γ2−1 ○ γ1 = γ2′−1 ○ γ2 = γxn for some n ∈ Z. It follows
that (γ1 , γ1′ ) = (θn (γ2 ), θn (γ2′ )) which shows that f is also injective. Finally, for
any γ, γ ′ ∈ X = Gx one has γ ≥ γ ′ ⇐⇒ γ ○ γ ′−1 ∈ G+ showing that f , so defined, is
an order isomorphism.
3.3 – The category of archimedean sets in terms of oriented groupoı̈ds

In this section we extend the above construction (X, θ) → G(X, θ) of the oriented
groupoid associated to an archimedean set to a functor G connecting the category
Arc ⋉ N× of archimedean sets to that of oriented groupoı̈ds. We recall, from [4],
the definition of the category of archimedean sets.
Definition 3.8. The objects of the category Arc ⋉ N× are the archimedean sets
(X, θ) as in Definition 3.5, the morphisms f ∶ (X, θ) → (X ′ , θ′ ) in Arc ⋉ N× are
equivalence classes of maps
(19)
f ∶ X → X ′ , f (x) ≥ f (y) ∀x ≥ y; ∃k > 0, f (θ(x)) = θ′k (f (x)), ∀x ∈ X
where the equivalence relation identifies two such maps f and g if there exists an
integer m ∈ Z such that g(x) = θ′m (f (x)), ∀x ∈ X.
Given a morphism f ∶ (X, θ) → (X ′ , θ′ ) in Arc ⋉ N× as in (19), the map G(f )

sending (x, y) ↦ (f (x), f (y)) is well defined since
X × X ∋ (x, y) ∼ (x′ , y ′ ) Ô⇒ (f (x), f (y)) ∼ (f (x′ ), f (y ′ )).
Moreover G(f ) ∶ G(X, θ) → G(X ′ , θ′ ) does not change if we replace f by g(x) =

θ′m (f (x)) since this variation does not alter the element
(f (x), f (y)) ∼ (g(x), g(y)) ∈ G(X ′ , θ′ ).
Proposition 3.9. The association (X, θ) ↦ G(X, θ), f ↦ G(f ) defines a faithful
functor G from Arc ⋉ N× to the category of oriented groupoids. For any non-trivial
morphism of oriented groupoı̈ds ρ ∶ G(X, θ) → G(X ′ , θ′ ) there exists a (unique)
morphism
f ∈ HomArc⋉N× ((X, θ), (X ′ , θ′ )) s.t. ρ = G(f ).
Proof. By construction the map G(f ) is a morphism of groupoı̈ds since
G(f ) ((x, y) ○ (y, z)) = (f (x), f (z)) = G(f )((x, y)) ○ G(f )((y, z))
Moreover it is a morphism of oriented groupoı̈ds since by (19), one has f (x) ≥ f (y),
∀x ≥ y so that
(x, y) ∈ G+ (X, θ) Ô⇒ (f (x), f (y)) ∈ G+ (X, θ)
We show that the functor G is faithful. Assume f, g ∈ HomArc⋉N× ((X, θ), (X ′ , θ′ ))

are such that G(f ) = G(g). Then for any (x, y) ∈ X × X there exists an integer
n = n(x, y) ∈ Z such that
g(x) = θ′n(x,y) (f (x)), g(y) = θ′n(x,y) (f (y)).
Since θ′ acts freely on X ′ , the integer n(x, y) is unique. The first and the second
equations prove that n(x, y) is independent of y and x respectively. Thus one
derives that f and g are in the same equivalence class, i.e. they define the same
element of HomArc⋉N× ((X, θ), (X ′ , θ′ )).
Let now ρ ∶ G(X, θ) → G(X ′ , θ′ ) be a non-trivial morphism of oriented groupoı̈ds.
Let x ∈ G(0) , y = ρ(x): the ordered group morphism ρ ∶ Gxx → G′y y is non-constant
and is given by ρ(γx ) = γyk , for some k > 0. The map ρ ∶ Gx → G′y is non-decreasing
since
γ ≤ γ ′ Ô⇒ γ ′ ○ γ −1 ∈ G+ Ô⇒ ρ(γ ′ ) ○ ρ(γ)−1 ∈ G′+ .
Given an archimedean set (X, θ) and an element z ∈ X, we claim that the map
ψX,z ∶ X → G(X, θ) defined by ψX,z (y) ∶= (y, z) is an order preserving bijection
of X with G(X, θ)x , where x is the class of z in X/θ. Indeed, every element of
G(X, θ)x admits a unique representative of the form (y, z) and one has
ψX,z (y) ≤ ψX,z (y ′ ) ⇐⇒ (y ′ , z) ○ (z, y) ∈ G+ ⇐⇒ y ′ ≥ y.
Moreover ψX,z is also equivariant since one has
ψX,z (θ(y)) = (θ(y), z) = (θ(y), θ(z)) ○ (θ(z), z) = ψX,z (y) ○ γx .
Let then x̃ ∈ X and ỹ ∈ X ′ be two lifts of x and y. The map f ∶= ψX
−1
′ ,ỹ ○ ρ ○ ψX,x̃ is
′
a non-decreasing map from X to X and one has, using ρ(γ ○ γx ) = ρ(γ) ○ γyk , that
f (θ(x)) = θ′k (f (x)), ∀x ∈ X.

One derives by construction the equality (f (a), ỹ) = ρ((a, x̃), ∀a ∈ X. Taking
a = x̃ this gives f (x̃) = ỹ since ρ((x̃, x̃)) is a unit. This shows that ρ(γ) = G(f )(γ)
∀γ ∈ Gx and the same equality holds for all γ ∈ G since both ρ and G(f ) are
homomorphisms while any element of G is of the form γ○(γ ′ )−1 with γ, γ ′ ∈ Gx .
Remark 3.10. The oriented groupoı̈ds associated to archimedean sets are all
equivalent, in the sense of equivalence of (small) categories, to the ordered group
(Z, Z+ ). It follows that a morphism φ of oriented groupoı̈ds induces an associ-
ated morphism Mod(φ) of totally ordered groups, i.e. an ordered group morphism
(Z, Z+ ) → (Z, Z+ ) given by multiplication by an integer Mod(φ) = k ∈ N. Proposi-
tion 3.9 suggests to refine the category g of oriented groupoı̈ds by considering only
the morphisms φ such that Mod(φ) ≠ 0. In other words what one requires is that
the associated morphism of totally ordered groups, obtained by working modulo
equivalence of categories, is injective. One can then reformulate Proposition 3.9
stating that the functor G is full and faithful.
Corollary 3.11. The epicyclic category Λ̃ (cf. Appendix Appendix B) is canon-

ically isomorphic to the category with objects the oriented groupoı̈ds g(m), m ≥ 0,
of equation 14 and morphisms the non-trivial morphisms of oriented groupoı̈ds.
The functor which associates to a morphism of oriented groupoı̈ds its class up to
equivalence coincides with the functor Mod ∶ Λ̃ Ð→ N× which sends a semilinear
map of semimodules over F = Zmax to the corresponding injective endomorphism
Frn ∈ End(F) (cf. [5])
Proof. By Proposition 2.8 of [5], the epicyclic category Λ̃ is canonically isomor-

phic to the full subcategory of Arc ⋉ N× whose objects are the archimedean sets
m ∶= (Z, x ↦ x + m + 1) for m ≥ 0. The oriented groupoı̈d G(m) is by Lemma 3.4
canonically isomorphic to g(m). The first statement then follows from Proposition
3.9 while the last one is checked easily and directly.
4. Points of (Λ̃op )∧ and projective geometry in characteristic one
Let C be a small category and Cˆ the topos of contravariant functors from C to

the category of sets Sets. Yoneda’s Lemma defines an embedding of the opposite
category C op into the category of points of the topos C. ˆ More precisely, to an
object c of C op one associates the flat covariant functor hc (−) ∶ C Ð→ Sets, hc (−) =
HomC (c, −). Then, one sees that through Yoneda’s embedding h ∶ C op Ð→ C, ˆ
c ↦ hc , any point of Cˆ can be obtained as a filtering colimit of points of the
form hc . We shall apply these well known general facts to C = Λ̃op : the opposite
of the epicyclic category Λ̃. We refer to Appendix Appendix B for the basic
notations on the cyclic and epicyclic categories. It follows from Corollary 3.11 that
Λ̃ is canonically isomorphic to the full subcategory of the category g of oriented
groupoı̈ds whose objects are the oriented groupoı̈ds of the form g(m). This fact
suggests that one should obtain the points of the topos (Λ̃op )∧ by considering
filtering colimits of the objects g(m) in g. In this section we compare the colimit
procedures taken respectively in the category of flat functors Λ̃op Ð→ Sets and in
the category g. The comparison is made directly by reconstructing the structure
of an oriented groupoı̈d starting from a flat functor as above. The main result is
the following
Theorem 4.1. The category of points of the epicyclic topos (Λ̃op )∧ is equivalent
to the category P whose objects are pairs (K, E) where K is an algebraic extension
of F = Zmax and E is an archimedean semimodule over K. The morphisms are
projective classes of semilinear maps and injective semifield morphisms.
One knows from [6] that an algebraic extension K of the semifield F = Zmax of
tropical integers is equivalently described by a totally ordered group (H, H+ ) iso-
morphic to a subgroup Z ⊆ H ⊆ Q of the rationals. An archimedean semimodule E
over K is in turn described (cf. [5]) by a totally ordered set X on which H acts
by order automorphisms of type: (x, h) ↦ x + h which fulfill the property
(20) h + x > x, ∀h ∈ H+ , h ≠ 0, x ∈ X
and the archimedean condition
(21) ∀x, y ∈ X, ∃h ∈ H+ s.t. h + x > y.
It follows from [1] (cf. also [10], Theorem 2 Chapter VII §5) that a point of a topos
of the form C,ˆ where C is a small category, is described by a covariant flat functor
F ∶ C Ð→ Sets. Next, we overview the strategy adopted to prove Theorem 4.1.
In §4.1 we associate to a pair (K, E) a point of (Λ̃op )∧ . This construction is
accomplished in two steps. First, we extend the construction (X, θ) ↦ G(X, θ)
of the oriented groupoı̈d associated to an archimedean set (as in §3.2) to a pair
(K, E) as in Theorem 4.1. Then, for any given pair (K, E), we provide a natural
construction of a point of (Λ̃op )∧ by means of the following associated flat functor
(n = (Z, x ↦ x + n + 1), n ≥ 0)
(22) F ∶ Λ̃op Ð→ Sets, F (n) = Homg (g(n), G(K, E)) .
Here, one implements Corollary 3.11 to identify the category Λ̃ with a full sub-
category of the category g of oriented groupoı̈ds with injective morphisms (up to
equivalence).
To produce the converse of the above construction, i.e. in order to show that any
point of (Λ̃op )∧ is obtained as in (22) by means of a uniquely associated pair
(K, E), we start from a covariant flat functor F ∶ Λ̃op Ð→ Sets and describe in
§§4.2-4.4 a procedure that allows one to reconstruct the semifield K by using the
natural geometric morphism of topoi associated to the functor Mod ∶ Λ̃op Ð→ N× .
The archimedean semimodule E (totally ordered set) is then reconstructed by
using a suitable restriction of F to obtain intervals from points of the simplicial
ˆ
topos ∆.
4.1 – The flat functor Λ̃op Ð→ Sets associated to a pair (K, E)

Let (H, H+ ) be a totally ordered abelian group, denoted additively and X a totally
ordered set on which the ordered group H acts preserving the order and fulfilling
(20). Let (G(X, H), G+ (X, H)) be the oriented groupoı̈d associated to the pair
(X, H) by Lemma 3.3, thus one has
(23) G(X, H) ∶= (X × X)/H, G+ (X, H) ∶= {(x, y) ∣ x ≥ y}.
The next lemma is used to show that the functor F ∶ Λ̃op Ð→ Sets naturally
associated to a pair (K, E) is filtering.
Lemma 4.2. Let (H, H+ ) be a non-trivial subgroup of (Q, Q+ ) and assume that the
totally ordered set X on which H acts fulfills the archimedean condition (21). Let
F = {φj ∣ 1 ≤ j ≤ n} be a finite set of morphisms φj ∈ Homg (g(mj ), G(X, H)).
Then, there exists a cyclic subgroup H0 ⊂ H, a subset X0 ⊂ X stable under the
action of H0 , morphisms ψj ∈ Homg (g(mj ), G(X0 , H0 )) and an integer m ∈ N
such that, denoting by ι ∶ G(X0 , H0 ) → G(X, H) the natural morphism, one has
(24) φj = ι ○ ψj ∀j, G(X0 , H0 ) ≃ g(m).
Moreover, let ψ, ψ ′ ∈ Homg (g(n), G(X0 , H0 )) be two morphisms such that ι ○ ψ =

ι ○ ψ ′ , then there exists a singly generated subgroup H1 with H0 ⊂ H1 ⊂ H, a subset
X1 ⊂ X containing X0 and stable under the action of H1 , such that the equality
ι1 ○ ψ = ι1 ○ ψ ′ holds in Homg (g(n), G(X1 , H1 )), with ι1 ∶ G(X0 , H0 ) → G(X1 , H1 )
the natural morphism.
Proof. We denote by α(m, i) ∶= (i, 1) ∈ (Z/(m + 1)Z) ⋉ Z = g(m) the natural

positive generators of the oriented groupoı̈d g(m). Let (G, G+ ) be an oriented
groupoı̈d . A morphism φ ∈ Homg (g(m), G) is uniquely specified by the m + 1
elements γi = φ(α(m, i)) ∈ G+ (cf. Figure 1) fulfilling the conditions (with γm+1 ∶=
γ0 )
r(γi ) = s(γi+1 ) , ∀i, 0 ≤ i ≤ m, γm ○ ⋯ ○ γ0 ∉ G(0) .
Figure 1. One encodes a morphism φ of oriented groupoı̈ds from g(n) to g(m) by the
arrows γi = φ(α(n, i)) associated to the generators α(n, i). To each generator (in blue)
one assigns an arrow (in red) specified by its source and range and by an integer which
gives the number of additional windings.
If G = G(X, H), it follows that Homg (g(m), G) is the quotient of the subset of
X m+2
(25) {(x0 , . . . , xm+1 ) ∈ X m+2 ∣ xj ≤ xj+1 , ∀j ≤ n, xm+1 ∈ x0 + H+ }
by the diagonal action of H. The morphism φ ∈ Homg (g(m), G) associated to

(x0 , . . . , xm+1 ) is given by
(26) φ(α(m, i)) = (xi+1 , xi ) ∈ G+ , ∀i, 0 ≤ i ≤ m.
For each φj ∈ F we get an hj ∈ H+ , then we choose a finite subset Zj ⊂ X such

that φj is represented by an (m+2)-uple (x0 , . . . , xm+1 ) ∈ X m+2 with all xi ∈ Zj .
Let H0 be the subgroup of H generated by the hj ’s and let X0 be the H0 invariant
subset of X generated by the union of the Zj . Then H0 is singly generated and the
pair (X0 , H0 ) fulfills the archimedean condition (21). Moreover by construction
one can lift the maps φj ∈ F to elements ψj ∈ Homg (g(mj ), G(X0 , H0 )) such that
φj = ι ○ ψj , where ι ∶ G(X0 , H0 ) → G(X, H) is the natural morphism.
It remains to show that G(X0 , H0 ) ≃ g(m) for some m ∈ N. By construction
there exists a finite subset Z ⊂ X0 such that X0 = Z + H0 . Using the archimedean
property (21) and since H0 ≃ Z, it follows that the pair (X0 , H0 ) is an archimedean
set such that the quotient X0 /H0 is finite. Thus G(X0 , H0 ) ≃ g(m), where m + 1
is the cardinality of X0 /H0 .
To prove the last statement, let (x0 , . . . , xm+1 ) ∈ X0m+2 (resp. (x′0 , . . . , x′m+1 ) ∈
X0m+2 ) represent ψ (resp. ψ ′ ). The equality ι ○ ψ = ι ○ ψ ′ implies that there exists
h ∈ H such that x′j = xj + h for all j. One then lets H1 be the subgroup of H
generated by H0 and h and X1 the H1 invariant subset of X generated by X0 .
Proposition 4.3. Let (H, H+ ) be a non-trivial subgroup of (Q, Q+ ) and assume

that the totally ordered set X on which H acts fulfills the archimedean condition
(21). Then the following formula defines a flat functor
(27) F ∶ Λ̃op Ð→ Sets F (n) = Homg (g(n), G), G = G(X, H)

θ
where n = (Z, x ↦ x + n + 1).
Proof. The statement follows from Lemma 4.2 showing that the functor F is ob-
tained as a filtering colimit of representable, flat functors. We provide the detailed
proof for completeness and to review the basic properties of flat functors which will
be used later in this article. Corollary 3.11 provides a canonical identification of
the epicyclic category Λ̃ with the full subcategory of g of oriented groupoids of the
form g(m). In particular, (27) defines a covariant functor. It remains to show that
this functor is flat. One knows from classical facts in the theory of Grothendieck
topoi (cf. e.g. [10], Chapter VII §6, Theorem 3) that a functor F ∶ C Ð→ Sets (C
a small category) is flat if and only if it is filtering i.e. the category ∫C F is filter-
ing (cf. [10] Chapter VII §6, Definition 2). The objects of the category ∫C F are
pairs (j, x) where j is an object of C and x ∈ F (j). The morphisms between two
such objects (j, x) and (k, y) are elements γ ∈ HomC (j, k) such that F (γ)x = y.
We recall that the filtering condition on a small category I is equivalent to the
fulfillment of the following conditions
(1) I is non empty.
(2) For any two objects i, j of I there exist an object k and morphisms k → i,
k → j.
(3) For any two morphisms α, β ∶ i → j, there exists an object k and a morphism
γ ∶ k → i such that α ○ γ = β ○ γ.
For each object i of the small category C one obtains a flat functor provided by the
Yoneda embedding hi ∶ C Ð→ Sets, j ↦ HomC (i, j). Here we take C = Λ̃op and F
given by (27). The filtering property of F only involves finitely many elements of
Homg (g(n), G) and hence by Lemma 4.2, it follows using the filtering property of
the functors hi . The first part of Lemma 4.2 is used to prove the filtering property
(ii) while the last part is implemented to prove (iii).
4.2 – The image of a flat functor F ∶ Λ̃op Ð→ Sets by the module morphism
The functor Mod ∶ Arc ⋉ N× Ð→ N× associates to any morphism of archimedean
sets the integer k ∈ N× involved in Definition 3.8 (cf. [5]). Its restriction to the
full subcategory whose objects are archimedean sets of type (Z, θ), where θ(x) =
x + n + 1, defines a functor Mod ∶ Λ̃ Ð→ N× . The category N× is isomorphic to
its opposite in view of the commutativity of the multiplicative monoı̈d of positive
integers. The functor Mod ∶ Λ̃op Ð→ N× determines a geometric morphism of
topoi. We recall once again ( [6]) that the category of points of the topos N ̂×
is canonically equivalent to the category of totally ordered groups isomorphic to
non-trivial subgroups of (Q, Q+ ), and injective morphisms of ordered groups. This

latter category is in turn equivalent to the category of algebraic extensions of the
semifield F = Zmax i.e. of extensions F ⊂ K ⊂ Qmax . The morphisms are the
injective morphisms of semifields.
This section is devoted to the description of the action of the geometric morphism
Mod ∶ (Λ̃op )∧ Ð→ N ̂× on points, in terms of the associated flat functors. This
process allows one to recover the extension K of F = Zmax involved in Theorem
4.1 from the datum of a flat functor Λ̃op Ð→ Sets.
Given two small categories Cj (j = 1, 2), a functor φ ∶ C1 Ð→ C2 determines a
geometric morphism (also noted φ) of topoi Cˆ1 Ð→ Cˆ2 (cf. e.g. [10], Chapter VII
§2, Theorem 2). The inverse image φ∗ sends an object of Cˆ2 , i.e. a contravariant
functor C2 Ð→ Sets to its composition with φ which determines a contravariant
functor C1 Ð→ Sets. The geometric morphism φ sends points of Cˆ1 to points of
Cˆ2 . In terms of the flat functors associated to points, the image by φ of a flat
functor F1 ∶ C1 Ð→ Sets associated to a point p1 ∶ Sets Ð→ Cˆ1 is the flat functor
F2 ∶ C2 Ð→ Sets obtained by composing the Yoneda embedding C2 Ð→ Cˆ2 with
p∗1 ○ φ∗ , where p∗1 ∶ Cˆ1 Ð→ Sets is the inverse image functor with respect to F1 .
Thus, for any object Z of C2 one obtains
(28) F2 (Z) = p∗1 (X), X ∶ C1op Ð→ Sets, X(c1 ) ∶= HomC2 (φ(c1 ), Z).
We apply this procedure to the functor Mod ∶ Λ̃op Ð→ N× : i.e. we take C1 = Λ̃op ,
C2 = N× and φ = Mod. Let F be a flat functor F ∶ Λ̃op Ð→ Sets. The inverse image
functor with respect to F of a covariant functor X ∶ Λ̃ Ð→ Sets coincides with the
geometric realization ∣X∣F and it is of the form
∣X∣F = ( ∐ (F (n) ×Λ̃ X(n))) / ∼

n≥0
The image of the flat functor F ∶ Λ̃op → Sets by the morphism Mod is thus the
flat functor H ∶ N× Ð→ Sets obtained as the geometric realization ∣X∣F of the
covariant functor
X ∶ Λ̃ Ð→ Sets, X(n) = HomN× (●, Mod(n)) ≅ N× Obj(N× ) = {●}.
The functor X associates to any object of Λ̃ the set N× and to a morphism γ of Λ̃
its module Mod(γ) acting by multiplication on N× . Hence we obtain
(29) H(●) = ( ∐ (F (n) ×Λ̃ N× )) / ∼

n≥0
The equivalence relation is exploited as follows: for (z, k) ∈ F (n) × N× , one has
(z, k) ∼ (F (γ)z, k) ∈ F (0) × N× for any γ ∈ HomΛ (0, n), since Mod(γ) = 1. More-
over for (z, k) ∈ F (0) × N× , one has (z, k) ∼ (F (γ)z, 1) for γ ∈ HomΛ̃ (0, 0) with
Mod(γ) = k. This shows that any element of H(●) is equivalent to an element
of the form (z, 1) for some z ∈ F (0). In particular one deduces that H(●) is a
quotient of F (0).
Lemma 4.4. Let F ∶ Λ̃op Ð→ Sets be a flat functor and (H, H+ ) the corresponding
point of N̂× through the morphism Mod. Then there is a canonical, surjective
N× -equivariant map
π ∶ F (0) → H+ , π(F (γ)z) = Mod(γ)π(z), ∀γ ∈ HomΛ̃ (0, 0).
Moreover, the equivalence relation x ∼ x′ ⇐⇒ π(x) = π(x′ ) is given by
(30) x ∼ x′ ⇐⇒ ∃z ∈ F (1), F (δ0 )z = x, F (δ1 )z = x′
where δj ∈ Hom∆ (0, 1), (j = 1, 2) are the two face maps.
Proof. Let π ∶ F (0) → H(●), π(z) = (z, 1) as in (29). As remarked above, the
map π is surjective. Let γ ∈ HomΛ̃ (0, 0), then one derives easily: π(F (γ)z) =
(F (γ)z, 1) ∼ (z, Mod(γ)) = Mod(γ)π(z). Moreover, π(x) = π(x′ ) if and only if
there exist n ∈ N, u ∈ F (n) and γ, γ ′ ∈ HomΛ̃ (0, n) such that F (γ)u = x, F (γ ′ )u =
x′ , Mod(γ) = Mod(γ ′ ). Let k = Mod(γ), and Ψk (n) the archimedean set obtained
from n = (Z, θ) by replacing θ with θk (cf. Appendix Appendix B). Then one
derives canonical factorizations involving the identity map Z → Z viewed as the
element Idkn ∈ HomΛ̃ (Ψk (n), n) with Mod(Idkn ) = k
γ = Idkn ○ α0 , γ ′ = Idkn ○ α1 , α0 , α1 ∈ HomΛ (0, Ψk (n)).
Let δj ∈ Hom∆ (0, 1), j = 0, 1, be the two face maps. One can then find α ∈
HomΛ (1, Ψk (n)) such that αj = α ○ δj . Thus it follows that γ = Idkn ○ α ○ δ0 , γ ′ =
Idkn ○ α ○ δ1 . and one also gets that x = F (γ)z = F (δ0 )F (α)F (Idkn )u = F (δ0 )z, z =
F (α)F (Idkn )u, x′ = F (δ1 )z, which proves (30).
Let ι ∶ N× Ð→ Λ̃op be the functor which associates to the unique object ● of the
small category N× the object 0 of Λ̃op and to a positive integer k the unique element
ι(k) ∈ HomΛ̃ (0, 0) such that Mod(ι(k)) = k. There is another natural equivalence
relation on F (0) deduced from the action of N× and explicitly given by
(31) x ∼′ y ⇐⇒ ∃n, m ∈ N× , F (ι(n))x = F (ι(m))y.
We let π ′ ∶ F (0) → Q be the map to the quotient Q ∶= F (0)/ ∼′ .
Lemma 4.5. Let F ∶ Λ̃op Ð→ Sets be a flat functor. Let (H, H+ ) be the correspond-
̂× through the morphism Mod. Then, with π ′ ∶ F (0) → Q defined just
ing point of N
above, the map
(π, π ′ ) ∶ F (0) → H+ × Q
is a bijection of sets.
Proof. We prove first that (π, π ′ ) is injective. Let x, y ∈ F (0) be such that
x ∼′ y. Let n, m ∈ N× with F (ι(n))x = F (ι(m))y. Then the N× -equivariance
of π shows that in the ordered group H one has nπ(x) = mπ(y). Thus the
equality π(x) = π(y) implies that n = m since H has no torsion and both π(x)
and π(y) are non-zero. Hence we derive F (ι(n))x = F (ι(n))y. So far we have
used F ∶ Λ̃op Ð→ Sets as a contravariant functor from Λ̃ to Sets. In this proof
we shall also use F directly as a covariant functor and for clarity of exposition
we shall denote F cov (α) = F (α∗ ) for any morphism α ∈ Λ̃op using the canonical
anti-isomorphism Λ̃ → Λ̃op . The flatness of F implies that there exists an object
(m, t) of ∫Λ̃op F , with t ∈ F (m), and two elements u, v ∈ HomΛ̃op (m, 0) such that
F cov (u)t = x and F cov (v)t = y. One thus obtains F cov (ι(n)∗ ○u)t = F cov (ι(n)∗ ○v)t.
The flatness of F provides an object (`, s) of ∫Λ̃op F , with s ∈ F (`), and a morphism
in ∫Λ̃op F , given by a w ∈ HomΛ̃op (`, m) such that ι(n)∗ ○ u ○ w = ι(n)∗ ○ v ○ w. The
fact that w is a morphism in ∫Λ̃op F connecting the object (`, s) to (m, t) means
that F cov (w)s = t. The equality ι(n)∗ ○ u ○ w = ι(n)∗ ○ v ○ w in Λ̃op implies that
the two elements u ○ w, v ○ w of HomΛ̃op (`, 0) are equal. Indeed, let us show that
for γ, γ ′ ∈ HomΛ̃op (`, 0), and n ∈ N×
(32) ι(n)∗ ○ γ = ι(n)∗ ○ γ ′ Ô⇒ γ = γ ′ .
To prove this we pass to the opposite category using the identification HomΛ̃op (`, 0) ≅
HomΛ̃ (0, `), thus we get elements α, α′ ∈ Homg (g(0), g(`)) such that α ○ ι(n) =
α′ ○ ι(n). The oriented groupoı̈d g(0) has a single object o and an element
α ∈ Homg (g(0), g(`)) is uniquely specified by the image α(o) and by its module
Mod(α). Replacing α by α ○ ι(n) does not alter the image α(o) while it replaces
the module by n Mod(α). Thus (32) follows. This shows that u ○ w = v ○ w and
x = F cov (u ○ w)s = F cov (v ○ w)s = y which proves that (π, π ′ ) is injective.
Next we show that (π, π ′ ) is also surjective. It is enough to prove that given
x, y ∈ F (0) there exists z ∈ F (0) such that
(33) x ∼′ z, y ∼ z.
Again from the flatness of F one derives that there exists an object (m, t) of ∫Λ̃op F ,
with t ∈ F (m), and two elements u, v ∈ HomΛ̃op (m, 0) such that F cov (u)t = x and
F cov (v)t = y. Let then u∗ , v ∗ ∈ HomΛ̃ (0, m) be the corresponding morphisms in
the dual category Λ̃. We view them as elements of HomG (g(0), g(m)). Then
there exists a unique w∗ ∈ HomG (g(0), g(m)) which has the same range object of
g(m) as u∗ and the same module as v ∗ . Since w∗ and v ∗ have the same module
k one can find, as in the proof of Lemma 4.4, α ∈ HomΛ (1, Ψk (m)) such that
w∗ = Idkm ○ α ○ δ0 , v ∗ = Idkm ○ α ○ δ1 .
Let z = F cov (w)t, one has z = F (δ0 )F (Idkm ○ α)t, y = F (δ1 )F (Idkm ○ α)t which
shows that y ∼ z. Since w∗ ∈ HomG (g(0), g(m)) has the same range object of
g(m) as u∗ , there exist k, k ′ ∈ N× such that u∗ ○ ι(k) = w∗ ○ ι(k ′ ). This implies
that F (ι(k))x = F cov (ι(k)∗ )F cov (u)t = F cov (ι(k ′ )∗ )F cov (w)t = F (ι(k ′ ))z, so that
x ∼′ z as required.
Corollary 4.6. Let F ∶ Λ̃op Ð→ Sets be a flat functor. Then, for any x ∈ F (0)
the map π ∶ F (0) → H+ induces a bijection between the commensurability class of
x under the action of N× and the positive part H+ of the ordered group image of
the point F by the geometric morphism Mod ∶ (Λ̃op )∧ Ð→ N̂× .
Figure 2. The morphism g(0) → g(n) is the unique element jn ∈ Hom∆op (0∗ , n∗ ) ⊂
HomΛ̃ (0, n). It gives a map F (jn ) ∶ F (n) → F (0) whose value fixes an object and the
length of the interval.
4.3 – The interval associated to a pair F ∶ Λ̃op Ð→ Sets, x ∈ F (0)

In this section we describe a process which associates an interval Fx to a pair
made by a flat functor F ∶ Λ̃op Ð→ Sets and an element x ∈ F (0). For each
integer n ≥ 0 we denote by jn ∈ Hom∆op (0∗ , n∗ ) ⊂ HomΛ̃ (0, n) the unique element
of Hom∆op (0∗ , n∗ ). One has jn (x) = (n + 1)x, ∀x ∈ Z. We refer to Appendix 2 for
the notations on the category ∆op . We identify the opposite ∆op of the simplicial
category ∆ with the subcategory of the epicyclic category Λ̃ with the same objects
and morphisms specified by the identification
(34) Hom∆op (n∗ , m∗ ) ≃ {γ ∈ HomΛ̃ (n, m) ∣ γ ○ jn = jm }.
Passing to the dual categories, we get jn∗ ∈ HomΛ̃op (n, 0) and these elements define
objects of the slice category Λ̃op /0. More precisely one obtains
Lemma 4.7. The simplicial category ∆ is canonically isomorphic to the full sub-
category of the slice category Λ̃op /0 with objects the jn∗ .
Proof. We define a functor κ ∶ ∆ Ð→ Λ̃op /0 sending the object [n] = {0, . . . , n} to

jn∗ . To define the action of κ on morphisms we first describe explicitly the identifi-
cation (34). Given φ ∈ Hom∆op (n∗ , m∗ ) i.e. a non-decreasing map preserving the
end points from the interval n∗ ∶= {0, . . . , n + 1} to the interval m∗ , one extends φ
uniquely to the following non-decreasing map
φ̃ ∶ Z → Z, φ̃(x) = φ(x) , ∀x ∈ {0, . . . , n+1}, φ̃(x+a(n+1)) = φ̃(x)+a(m+1) , ∀a ∈ Z.
One finds φ̃ ∈ HomΛ̃ (n, m) with Mod(φ̃) = 1. The map φ ↦ φ̃ defines a faithful,
covariant functor ∆op Ð→ Λ̃ which is bijective on objects and whose range is
characterized by (34). To check this statement, note that the equation γ ○ jn = jm
implies that Mod(γ) = 1. Moreover any non-decreasing map ψ ∶ Z → Z such that

ψ(0) = 0 and ψ(x + a(n + 1)) = ψ(x) + a(m + 1), ∀a ∈ Z is of the form φ̃ for a unique
φ ∈ Hom∆op (n∗ , m∗ ). Then we define κ on morphisms as follows
∗
κ(γ) ∶= (γ̃ ∗ ) ∈ HomΛ̃op /0 (jn∗ , jm
∗
) , ∀γ ∈ Hom∆ ([n], [m]).
Lemma 4.8. Let F ∶ Λ̃op Ð→ Sets be a flat functor and x ∈ F (0). The following
equality defines a flat functor ∆ Ð→ Sets (i.e. interval) which can be equally
interpreted as a contravariant functor
(35) Fx ∶ ∆op Ð→ Sets, Fx (n∗ ) = {z ∈ F (n) ∣ F (jn )z = x}.
Proof. Let X be a totally ordered set on which an ordered group H acts fulfilling
the conditions of Lemma 4.2. Assume first that the functor F is defined by
F (n) = Homg (g(n), G), G = G(X, H).
Then, by applying (25), F (n) is the quotient by the diagonal action of H on the
subset of X n+2
{(x0 , . . . , xn+1 ) ∈ X n+2 ∣ xj ≤ xj+1 , ∀j ≤ n, xn+1 = x0 + H+ }.
Thus one derives that F (0) = X/H × H+ . Let x ∈ F (0), x = (x̃0 , h) ∈ X/H × H+ .
After choosing a lift x0 ∈ X of x̃0 ∈ X/H, (35) provides the equality
(36) Fx (n∗ ) = {(x0 , . . . , xn+1 ) ∈ X n+2 ∣ xj ≤ xj+1 , ∀j ≤ n, xn+1 = x0 + h}
which can be equivalently described as Hom(n∗ , I), where I is the interval [x0 , x0 +
h] ⊂ X. Let m ∈ N and G = g(m), F (−) = HomΛ̃ (−, g(m)). Then x ∈ F (0) is
determined by an object of g(m) and an integer k, and it follows by the above
discussion and from the arguments developed in Appendix 2 that the associated
interval is Fx = Sd∗k (m∗ ).
The general case is deduced from the above one by writing the flat functor F
as a filtering colimit of functors of the form HomΛ̃ (−, g(m)). More precisely, let
F ∶ Λ̃op Ð→ Sets be a flat functor and x ∈ F (0). We prove, for instance, the filtering
property of Fx . An object α of the category ∫∆ Fx is of the form α = (n, z), z ∈
F (n), F (jn )z = x. Given two such objects αj = (nj , zj ), j = 1, 2 one has F (jn1 )z1 =
F (jn2 )z2 and since F is a filtering colimit of functors of the form HomΛ̃ (−, g(m))
one can find an integer m ∈ N and elements yj ∈ HomΛ̃ (g(nj ), g(m)) such that yj
represents zj in the colimit and that y1 ○ jn1 = y2 ○ jn2 . This last equality can be
realized as a consequence of F (jn1 )z1 = F (jn2 )z2 using the filtering colimit i.e.
the definition of equality in the limit. The existence of an object α = (n, z) and
morphisms uj in ∫∆ Fx to αj = (nj , zj ) then follows from the filtering property of
the category ∫∆ Fy′ where F ′ is the flat functor HomΛ̃ (−, g(m)) and y = y1 ○ jn1 =
y2 ○ jn2 .
Figure 3. The morphism η ∶ g(1) → g(2) is in Hom∆op (1∗ , 2∗ ) ⊂ HomΛ̃ (1, 2). It deter-
mines a map F (η) ∶ F (2) → F (1) such that F (η)Fx (2∗ ) ⊂ Fx (1∗ ), since η ○ j1 = j2 .
4.4 – The oriented groupoı̈d associated to a pair F ∶ Λ̃op Ð→ Sets, x ∈ F (0)
Let F ∶ Λ̃op Ð→ Sets be a flat functor. Lemma 4.8 shows how to associate an
interval Fx to an element x ∈ F (0). Next, we state two lemmas which will be used
to reconstruct the totally ordered set X with an action of (H, H+ ), where (H, H+ )
is the ordered group corresponding to F by means of the geometric morphism
Mod ∶ (Λ̃op )∧ Ð→ N̂× , as a point of N
̂× : cf. Lemma 4.4. We use the notations of
Appendix 2.
Lemma 4.9. Let F ∶ Λ̃op Ð→ Sets be a flat functor and x ∈ F (0). The following
map defines an isomorphism of intervals for any k ∈ N×
(37) ωk ∶ Sd∗k (Fx ) → FF (ι(k))x , ωk (z, ρ) = F (Idkm ○ ρ)z ∈ FF (ι(k))x (n∗ )
∀(z, ρ) ∈ Fx (m∗ ) × Hom∆op (n∗ , Sd∗k (m∗ )).
Proof. Using the same method as in the proof of Lemma 4.8, it is enough to check
that the map is well defined and that one gets an isomorphism in the special case
where the functor F is of the form HomΛ̃ (●, g(m)). We consider the subcategory
∆op ⋉ N× ⊂ Λ̃ also reviewed in Appendix 2.4.
In Appendix 2.3, Lemma 2.7, it is shown that the action of the edgewise subdivision
functor Sdk on points of the topos of simplicial sets ∆ˆ is described in terms of the
associated flat functors by replacing the given flat functor G ∶ ∆ Ð→ Sets with the
flat functor Sd∗k (G) ∶ ∆ Ð→ Sets. This latter associates to the object [n] of ∆ the
quotient set
(38) Sd∗k (G)([n]) = ( ∐ (G(m) × Hom∆op (n∗ , Sd∗k (m∗ )))) / ∼

m≥0
where the equivalence relation ∼ is given, for f ∈ Hom∆op (m∗ , r∗ ) = Hom∆ ([r], [m]),
β ∈ Hom∆op (n∗ , Sd∗k (m∗ )) and y ∈ G(r) as follows
G(m) × Hom∆op (n∗ , Sd∗k (m∗ )) ∋ (G(f )y, β) ∼

∼ (y, Sd∗k (f ) ○ β) ∈ G(r) × Hom∆op (n∗ , Sd∗k (r∗ )).
We prove that the map (37) is well defined. First, one has Idkm ○ ρ ∈ HomΛ̃ (n, m)
and since Idkm ○ ρ ○ jn = jm ○ ι(k) one gets
F (Idkm ○ ρ)z ∈ F (n), F (jn )F (Idkm ○ ρ)z = F (Idkm ○ ρ ○ jn )z = F (ι(k))x.
We let G(m) = Fx (m∗ ). We need to check that two equivalent elements in (38)
have the same image under ωk . Let f ∈ Hom∆op (m∗ , r∗ ) = Hom∆ ([r], [m]), β ∈
Hom∆op (n∗ , Sd∗k (m∗ )) and y ∈ G(r) = Fx (r∗ ). One then has, since f ○ Idkm =
Idkr ○ Sd∗k (f )
ωk (G(f )y, β) = F (Idkm ○ β)G(f )y = F (Idkm ○ β)F (f )y = F (f ○ Idkm ○ β)y =

= F (Idkr ○ Sd∗k (f ) ○ β)y.
Thus one derives ωk (G(f )y, β) = F (Idkr ○ Sd∗k (f ) ○ β)y = ωk (y, Sd∗k (f ) ○ β). This
shows that the map ωk is well defined. One easily checks that it is an isomorphism
in the special case when the functor F is of the form HomΛ̃ (●, g(m)). Then, one
obtains the general case working as in the proof of Lemma 4.8.
Given a flat functor H ∶ ∆ Ð→ Sets, the set I ∶= H([1]) is endowed with the total
order relation defined by
y ≤ y ′ ⇐⇒ ∃z ∈ H([2]) ∣ y = H(σ)(z), y ′ = H(σ ′ )(z)
where σ, σ ′ ∈ Hom∆ ([2], [1]) are the two surjections. Endowed with this order rela-
tion I is an interval and one recovers the flat functor H as H([n]) = Hom∆op (n∗ , I).
We shall use this fact to organize the intervals Ix = Fx ([1]) associated to the pair
of a flat functor F ∶ Λ̃op Ð→ Sets and x ∈ F (0).
Our final goal is in fact to show how to reconstruct an oriented groupoı̈d from a flat
functor Λ̃op Ð→ Sets. It will be enough to check that we can reconstruct the ori-
ented groupoı̈d G(X, H) for flat functors of the form F (m) = Homg (g(m), G(X, H))
as in Proposition 4.3.
The k elements of Σk1 (cf. Appendix A) determine the following elements ψk,j ∈
Hom∆op ⋉N× (1∗ , 1∗ ) ⊂ Homg (g(1), g(1)), for 0 ≤ j < k,
ψk,j ∶ Idk1 ○ αk,j , αk,j ∈ Hom∆op (1∗ , Sd∗k (1∗ )), αk,j (1) = 2j + 1.
and we view them as the elements of Homg (g(1), g(1)) associated to the endo-
morphisms φk,j (as in Definition 3.8) of the archimedean set (Z, u ↦ u + 2) given
by
(39) φk,j ∶ Z → Z, φk,j (2`) ∶= 2k`, φk,j (2` + 1) ∶= 2j + 1 + 2k`
Note that Mod(ψk,j ) = k. One lets ψk ∶= ψk,1 .
Note that any ψ ∈ HomΛ̃ (g(1), g(1)) = Homg (g(1), g(1)) such that ψ ○j1 = j1 ○ι(k)
defines a map
(40) Fψ ∶ Ix → IF (ι(k))x , Fψ (z) = F (ψ)z, ∀z ∈ F (1), F (j1 )z = x.
Indeed one has F (ψ)z ∈ F (1) and F (j1 )F (ψ)z = F (ι(k)F (j1 )z = F (ι(k))x. Thus
F (ψ)z ∈ IF (ι(k))x . This fact is used in the statement of the next proposition.
Proposition 4.10. Let F be a flat functor of the form F (m) = Homg (g(m), G(X, H))
as in Proposition 4.3. Let x ∈ F (0), x = (x̃0 , h) ∈ X/H × H+ and x0 ∈ X be a lift
of x̃0 .
(i) The following equality determines the ordered set Y = {y ∈ X ∣ y ≥ x0 }:
(41) Jx ∶= lim (IF (ι(k))x , Fψ`/k )
Ð→
where the colimit is taken using the maps Fψk ∶ Ix → IF (ι(k))x and the set of indices
is ordered by divisibility.
(ii) The graph of the action of h on Y is given as the union in the inductive limit
of the subsets
(42) ⋃ {(Fψk,j (z), Fψk,j+1 (z)) ∣ z ∈ Ix }
0≤j<k−1
(iii) The oriented groupoı̈d G(X, H) is isomorphic to the oriented groupoı̈d of

ordered pairs of elements of Jx modulo the action of H.
Proof. By (25), (26), a morphism φ ∈ Homg (g(m), G(X, H)) is determined by a

non-decreasing map f ∶ Z → X and an element h ∈ H+ , h ≠ 0 such that f (x+m+1) =
f (x) + h ∀x ∈ Z. Moreover, by (25), the maps f and g define the same morphism
if and only if there exists an a ∈ H such that g = f + a. In particular the map
f ∶ Z → X associated to an element z of the interval Ix = [x0 , x0 + h] of (36) is
given by f (2n) = x0 + nh, f (2n + 1) = z + nh ∀n ∈ Z. With these notations let us
determine the map Fψk,j . It corresponds to the composition f ○ φk,j and is hence
given by
f ○ φk,j (0) = f (0) = x0 , f ○ φk,j (1) = f (2j + 1) = z + jh,
while for any ` ∈ Z one has f ○ φk,j (u + 2`) = f ○ φk,j (u) + k`h. Thus the map
Fψk,j sends z ∈ [x0 , x0 + h] to z + jh ∈ [x0 , x0 + kh]. In particular the Fψ`/k give the
canonical inclusion of intervals [x0 , x0 + kh] ⊂ [x0 , x0 + `h] for k∣` and the inductive
limit (41) is Y = {y ∈ X ∣ y ≥ x0 }. Moreover, in the limit, the graph of the action of
h is given by the pairs (42). Finally one reconstructs in this way the action of H+
on the colimit set Jx as well as the oriented groupoı̈d of ordered pairs of elements
of Jx modulo the equivalence relation generated by the action of the semigroup
H+ . .
4.5 – Proof of Theorem 4.1

In Proposition 4.3 we have shown that a pair (K, E) of an algebraic extension
K of F = Zmax and an archimedean semimodule E over K, defines a flat functor
Λ̃op Ð→ Sets as in (27). Conversely, in §4.2 and Proposition 4.10, we have proven
that a flat functor F ∶ Λ̃op Ð→ Sets determines the data given by a totally ordered
group (H, H+ ) of rank one, a totally ordered set J and an action of H on J. One
easily checks that the pair (J, H) fulfills the conditions of Proposition 4.3 and that
F (n) = Homg (g(n), G), G = G(J, H).
Figure 4. The morphism ψ ∶ g(1) → g(1) has module 8, it gives a map Fψ ∶ Fx →

FF (ι(8))x ≅ Sd∗8 (Fx ).
The proof of the Theorem 4.1 is then complete.
4.6 – Exotic points

In general, a colimit involves the process of taking a quotient and there is no guar-
antee “a priori” that the quotient is a non-singular space, even at the set-theoretic
level. Next, we exhibit certain points of the epicyclic topos whose associated flat
functors show a singular behavior.
Consider K = Qmax and the semimodule E = Rmax . Then the associated functor
F is such that F (n) is the quotient of the following subset of Rn+2 by the diagonal
action of Q
{(x0 , . . . , xn+1 ) ∈ Rn+2 ∣ xj ≤ xj+1 , ∀j ≤ n, xn+1 − x0 ∈ Q+ }.
One gets for instance

F (0) = R/Q × Q+ .
The units of the associated groupoı̈d G = G(E, K) form the quotient set R/Q
which is singular. This result supports the point of view that the description of
the points of the epicyclic topos in terms of the category P as in Theorem 4.1 is
more appropriate than the description given in terms of oriented groupoı̈ds since
the latter interpretation involves singular quotients.
̂
4.7 – Relation between Λ op and (Λ̃op )∧
̂
By a result of [11] the points of the topos Λop dual to the (opposite of) the cyclic
category form the category of abstract circles and in turns by [4] this latter is
equivalent to the category Arc of archimedean sets with morphisms of module 1.
The inclusion of categories Λop ⊂ Λ̃op (dual to Λ ⊂ Λ̃) induces a geometric mor-
phism of topoi: in this section we determine its behavior on the points. We start
by stating a technical lemma which will be applied in the proof of Proposition 4.12
Lemma 4.11. Let (X, θ) be an archimedean set and φ ∈ Homg (g(u), G(X, θ)),
u ≥ 0. Denote by r + 1 the cardinality of φ(g(u)(0) ) ⊂ G(X, θ)(0) . Then there exists
g ∈ Homg (g(u), g(r)) and f ∈ HomArc (r, (X, θ)) such that φ = G(f ) ○ g. This
factorization is unique up to the cyclic group AutArc (r), i.e. the replacement
f ↦ f ○ σ −1 , g ↦ G(σ) ○ g , ∀σ ∈ AutArc (r).
Proof. Let φ ∈ Homg (g(u), G(X, θ)) and k = Mod(φ). Then by (25), φ is deter-
mined by u + 2 elements xj , 0 ≤ j ≤ u + 1 of X such that xj ≤ xj+1 ∀j ≤ u and that
xu+1 = θk (x0 ). One has a canonical isomorphism [x0 , θk (x0 )] ≃ Sd∗k ([x0 , θ(x0 )].
Let χ ∈ Hom∆op (u∗ , Sd∗k ([x0 , θ(x0 )]) be the morphism of intervals defined by
χ(j) = xj , ∀j, 0 ≤ j ≤ u + 1. Let Y = φ(g(u)(0) ) ⊂ G(X, θ)(0) = X/θ. Then
the rank of the morphism χ is the cardinality of Y ∖ {x̃0 } where x̃0 ∈ X/θ is the
class of x0 and it is thus equal to r. By Proposition 2.5 there exists a unique
decomposition
χ = Sd∗k (α) ○ β, β ∈ Hom∆op (u∗ , Sd∗k (r∗ )), α ∈ Hom∆op (r∗ , [x0 , θ(x0 )]).
The map α ∈ Hom∆op (r∗ , [x0 , θ(x0 )]) is uniquely determined by its restriction to
{1, . . . , r} which is the unique increasing injection of {1, . . . , r} in (x0 , θ(x0 )) whose
range gives Y ∖ {x̃0 } ⊂ X/θ. Let f ∈ HomArc (r, (X, θ)) be defined by
f (i + `(r + 1)) = θ` (α(i)) , ∀i, 0 ≤ i ≤ r, ∀` ∈ Z
Moreover β ∈ Hom∆op (u∗ , Sd∗k (r∗ )) determines g = Idkr ○ β ∈ Homg (g(u), g(r)).
One has by construction Mod(g) = k. The equality χ = Sd∗k (α) ○ β shows that
φ = G(f ) ○ g. The uniqueness of the decomposition follows from the uniqueness of
the archimedean set (X ′ , θ) obtained as the inverse image of Y under the canonical
projection X → X/θ. This gives the uniqueness of f up to cyclic permutations
and since f is injective the uniqueness of g follows.
Proposition 4.12. The geometric morphism j ∶ Λ ̂op Ð→ (Λ̃op )∧ associated to
the inclusion of categories Λ ⊂ Λ̃ induces on the points of the topoi the func-
op op
tor producing the inclusion of F-semimodules among semimodules over algebraic

extensions of F.
̂
Proof. A point p of Λ op is given by a flat functor of the form
P ∶ Λop Ð→ Sets, n ↦ P (n) = HomArc (n, (X, θ))
where (X, θ) is an archimedean set and, as above, n = (Z, tn+1 ) is the archimedean
set given by the translation x ↦ x + n + 1 on Z. By (28), j(p) is given by the flat
functor F ∶ Λ̃op Ð→ Sets
(43)
F (Z) = p∗ (X), X ∶ Λ Ð→ Sets, X(n) ∶= HomΛ̃op (g(n), Z) = Homg (Z, g(n)).
The inverse image p∗ (X) is defined as
p∗ (X) = ∐ P (n) × X(n)/ ∼ = ∐ HomArc (n, (X, θ)) × Homg (Z, g(n))/ ∼
where the equivalence relation ∼ is the simplification with respect to Λ, i.e. one
has
(f ○ h, g) ∼ (f, G(h) ○ g) , ∀h ∈ HomArc (n, m), f ∈ P (m), g ∈ X(n).
It follows that the map γ ∶ p∗ (X) → Homg (Z, G(X, θ)) which sends (f, g) ∈
P (n) × X(n) to γ(f, g) = G(f ) ○ g ∈ Homg (Z, G(X, θ)) is well defined. Next,
we show that when Z = g(u) (u ≥ 0) this map is bijective.
First, we prove that γ is surjective. Let φ ∈ Homg (g(u), G(X, θ)). By Lemma
4.11, there exists r ≥ 0, g ∈ Homg (g(u), g(r)) and f ∈ HomArc (r, (X, θ)) such that
φ = G(f ) ○ g. Thus φ = γ(f, g) is in the range of γ.
We prove that γ is injective. It is enough to show that for any n ≥ 0 and any
pair (f ′ , g ′ ) ∈ HomArc (n, (X, θ)) × Homg (g(u), g(n)) one has (f ′ , g ′ ) ∼ (f, g),
where the pair (f, g) is obtained from Lemma 4.11 applied to φ = f ′ ○ g ′ ∈
Homg (g(u), G(X, θ)). The element g ′ ∈ Homg (g(u), g(n)) satisfies Mod(g) =
Mod(f ′ ○ g ′ ) = k. One has g(n) = G(X0 , θ0 ) where X0 = Z and θ0 (x) = x + n + 1.
Thus by Lemma 4.11 applied to g ′ ∈ Homg (g(u), G(X0 , θ0 )) one obtains a decom-
position
g ′ = G(f0 ) ○ g0 , f0 ∈ HomArc (r0 , (X0 , θ0 )), g0 ∈ Homg (g(u), g(r0 ))
One has f0 ∈ HomArc (r0 , n) and thus (f ′ , g ′ ) ∼ (f ′ ○f0 , g0 ). The map g0 induces by
construction a surjection g(u)(0) → g(r0 )(0) and thus the range of the map induced
by f ′ ○ f0 ∈ HomArc (r0 , (X, θ)) from g(r0 )(0) to G(X, θ)(0) = X/θ is contained in
Y = φ(g(u)(0) ) ⊂ G(X, θ)(0) . Moreover Y = f (g(r)(0) ) and the map f (0) gives
an increasing bijection of {0, . . . r} with Y by construction. Thus, there exists
ρ ∈ HomArc (r0 , r) such that f ′ ○ f0 = f ○ ρ and one gets
(f ′ , g ′ ) ∼ (f ′ ○ f0 , g0 ) = (f ○ ρ, g0 ) ∼ (f, G(ρ) ○ g0 ) = (f, g)
which gives the required equivalence.

We have shown that the map γ ∶ p∗ (X) → Homg (Z, G(X, θ)) is bijective, it follows
that j(p) is given by the flat functor F ∶ Λ̃op Ð→ Sets, F (u) = Homg (g(u), G(X, θ)).
Translating this fact in terms of semifields and semimodules one obtains the re-
quired result.
4.8 – The sections ιn of Mod and their action on points

Let n ≥ 0 and consider, as above, the archimedean set n ∶= (Z, tn+1 ) given by the
translation x ↦ x + n + 1 on Z. The map x ↦ kx on Z defines, for any k ∈ N× , an
element ιn (k) ∈ HomArc⋉N× (n, n) and one obtains a homomorphism of semigroups
ιn ∶ N× → HomArc⋉N× (n, n). Moreover by construction one has Mod(ιn (k)) = k
∀k ∈ N× . This defines, for any n ≥ 0, a section ιn ∶ N× Ð→ Λ̃op of the functor Mod.
More precisely, ιn is associated to the functor which maps the only object ● of N×
to n and sends the morphism k ∈ EndN× (●) to ιn (k) ∈ HomΛ̃op (n, n) = Endg (g(n)).
One has Mod ○ ιn = Id.
Proposition 4.13. Let p be the point of the topos N ̂× associated to the ordered
group (H, H+ ). Let K = Hmax be the semifield associated to (H, H+ ). Then the
image ιn (p) by the section ιn is the point of (Λ̃op )∧ associated by Theorem 4.1
to the pair (K, K (n+1) ), where for m ≥ 1, K (m) denotes the semimodule over K
obtained from K by extension of scalars using the endomorphism Frm ∈ End(K),
Frm (x) ∶= xm ∀x ∈ K.
Proof. By (28), ιn (p) is given by the flat functor F ∶ Λ̃op Ð→ Sets
(44) F (Z) = p∗ (X), X ∶ N× Ð→ Sets, X(●) ∶= HomΛ̃op (ιn (●), Z).
One has HomΛ̃op (ιn (●), Z) = Homg (Z, g(n)). For any k ∈ N× = EndN× (●) the map
X(k) ∶ X(●) → X(●) is obtained by composition with ιn (k) ∈ Endg (g(n)). The
inverse image part p∗ of the geometric morphism associated to the point p is the
tensor product which associates to any N× -space X the set
p∗ (X) ∶= X ×N× H+ .
For X = Homg (Z, g(n)) one gets p∗ (X) = Homg (Z, G(H (n+1) , H)) where, for
m ≥ 1, H (m) is the totally ordered set H on which the ordered group H acts by
(h, x) ↦ x + mh, ∀x ∈ H (m) , h ∈ H. The flat functor F is thus given by F (Z) =
Homg (Z, G(H (n+1) , H)). By formulating this construction in terms of semifields
and semimodules one obtains the required result.
Appendix A. The λ-operations as elements of Z[∆op ⋉ N× ]
In this appendix we review the construction of the λ-operations. We use the

notations of section 2.
A.1 – The subset Σkn ⊂ Hom∆op (n∗ , Sd∗k (n∗ ))

For each n ≥ 0 and k ∈ N× , we introduce the following set
(45) Σkn ∶= {α ∈ Hom∆op (n∗ , Sd∗k (n∗ )) ∣ (πnk )∗ ○ α∗ is surjective} ⊂

⊂ Hom∆op (n∗ , Sd∗k (n∗ )).
Here (πnk )∗ ○ α∗ ∶ {0, . . . , n} → {0, . . . , n} preserves the base point 0, moreover it is

surjective if and only if it is a permutation of {0, . . . , n} fixing 0. One thus obtains
a canonical map to the group Sn of permutations of {1, . . . , n} defined as follows
(46) Perm ∶ Σkn → Sn , Perm(α) ∶= (πnk )∗ ○ α∗ .
One gets a parametrization of the set Σkn as follows, let f ∶ {1, . . . , n} → {0, . . . , k−1}
be an arbitrary map of sets and consider the following subset with n elements
(47) Xf ∶= {j + (n + 1)f (j) ∣ j ∈ {1, . . . , n}} ⊂ {1, . . . , k(n + 1) − 1}

Then, as shown by the next Lemma A.1, there exists a unique element s(f ) ∈
Hom∆op (n∗ , Sd∗k (n∗ )) whose range is Xf . It is obtained by labeling the elements
of Xf in the lexicographic order, with s(f )(0) = 0 and s(f )(n + 1) = k(n + 1).
Lemma A.1. The map f ↦ s(f ) is a bijection of the set of all maps {1, . . . , n} →
{0, . . . , k − 1} with Σkn .
Proof. First the range of s(f ) when taken modulo n+1 contains all j ∈ {1, . . . , n}
and thus s(f ) ∈ Σkn . By construction the map f ↦ s(f ) is injective since f (j) is
the only element of Xf which is congruent to j modulo n + 1. Let us show that it
is surjective. Let α ∈ Σkn , then since (πnk )∗ ○ α∗ is surjective there exists for each
j ∈ {1, . . . , n} an f (j) ∈ {0, . . . , k − 1} such that j + f (j)(n + 1) ∈ Range(α). One
has Range(α) = Xf and it follows that α = s(f ).
Proposition A.2. The range of Perm ∶ Σkn → Sn is the set of permutations whose
descent number is d = k − 1 − v < k. Moreover
n+v
#{α ∈ Σkn ∣ Perm(α) = σ} = ( ).
n
Proof. Given a permutation σ ∈ Sn , a lift of σ in Σkn is of the form
(48) α(j) = σ(j) + k(j)(n + 1)
where k ∶ {1, . . . , n} → {0, . . . , k − 1} is non-decreasing and such that
(49) σ(j + 1) < σ(j) Ô⇒ k(j + 1) > k(j)
By definition (cf. [8]) the descent number number d of σ is the number of j ∈
{1, . . . , n − 1} such that σ(j + 1) < σ(j). Thus (49) implies that k(n) − k(1) ≥ d
and one gets k − 1 ≥ d. Conversely assume that the descent number number d of σ
is d = k − 1 − v with v ≥ 0. Then the lifts α of σ correspond to sequences (j) ≥ 0,
j ∈ {1, . . . , n}, such that
∑ (j) ≤ k − 1, σ(j + 1) < σ(j) Ô⇒ (j) > 0.
Given such a sequence one lets k(j) = ∑j1 (i) and one derives a lift of σ using
(48). Replacing (j) by (j) − 1 when σ(j + 1) < σ(j) one gets that the number of
lifts is the number of sequences ′ (j) ≥ 0, j ∈ {0, . . . , n} such that ∑ ′ (j) = v, i.e.
(n+v
n
).
A.2 – The elements Λkn ∈ Z[∆op ⋉ N× ]

Let Z[∆op ] be the ring of finite formal sums of elements of ∆op with integral
coefficients, where the product law is given by the composition of morphisms
in ∆op whenever they are composable, otherwise is defined to be 0. One first
introduces the elements
(50) σnk ∶= ∑ (Perm(α))α ∈ Z[∆op ].
α∈Σk
n
Then, by composing with the maps πnk one obtains the following elements of the
ring Z[∆op ⋉ N× ]
(51) Λkn ∶= ∑ (Perm(α))πnk ○ α.

α∈Σk
n
Next, we show that these elements Λkn ∈ Z[∆op ⋉ N× ] fulfill the law
(52) Λkn Λ`n = Λk`

n , ∀k, `, n ∈ N.
One has Λkn Λ`n = πnk σnk πn` σn` = πnk` Sd∗` (σnk )σn` where Sd∗` is extended to Z[∆op ] by
linearity. Thus (52) follows from the next Lemma A.3 which shows that the map
∗
Σkn × Σ`n → Σk`
n , (α, β) ↦ Sd` (α)β
is bijective and that the signatures are multiplicative.
Lemma A.3. (i) Let α ∈ Σkn and β ∈ Σ`n . Then Sd∗` (α)β ∈ Σk` ∗
n and Perm(Sd` (α)β) =
Perm(α) ○ Perm(β). (ii) Let f ∶ {1, . . . , n} → {0, . . . , k − 1}, g ∶ {1, . . . , n} →
{0, . . . , ` − 1} be arbitrary maps of sets. Then one has
Sd∗` (s(f )) ○ s(g) = s(h)
where h ∶ {1, . . . , n} → {0, . . . , k` − 1} is given by
h(j) = f (j) + kg(σ −1 (j)), σ = Perm(s(f )).
Proof. (i) One has α ∈ Hom∆op (n∗ , Sd∗k (n∗ )), Sd∗` (α) ∈ Hom∆op (Sd∗` (n∗ ), Sd∗`k (n∗ ))
and β ∈ Hom∆op (n∗ , Sd∗` (n∗ )). Thus Sd∗` (α)β ∈ Hom∆op (n∗ , Sd∗`k (n∗ )). Moreover,
by applying the functor F ∶ ∆op ⋉ N × → Fin∗ one derives
F(πnk` (Sd∗` (α)β) = F(πnk α ○ πn` β) = F(πnk α)F(πn` β) = Perm(α) ○ Perm(β).
This shows that Sd∗` (α)β ∈ Σk` ∗

n and Perm(Sd` (α)β) = Perm(α) ○ Perm(β).
(ii) We determine the range of γ = Sd` (s(f )) ○ s(g). It is the image by Sd∗` (s(f ))
∗
of the range of s(g) which is by construction Xg = {j + g(j)(n + 1) ∣ j ∈ {1, . . . , n}}

(ignoring the base points). Now one has
Sd∗` (s(f ))(i + a(n + 1)) = s(f )(i) + ka(n + 1),
and thus
Sd∗` (s(f ))(j + g(j)(n + 1)) = s(f )(j) + kg(j)(n + 1).
Let σ = Perm(s(f )), then one has s(f )(j) = σ(j) + f (σ(j))(n + 1) and thus taking
j = σ −1 (i) one gets s(f )(σ −1 (i)) = i + f (i)(n + 1) and
Sd∗` (s(f ))(j + g(j)(n + 1)) = i + f (i)(n + 1) + kg(σ −1 (i))(n + 1)
which determines uniquely the range of γ = Sd∗` (s(f )) ○ s(g) in the required form.
A.3 – The commutation Λkn−1 b = bΛkn

We define the Hochschild boundary operator b, with one component bn for each
n ≥ 1, as the element
n
(53) bn ∶= ∑(−1)i dni ∈ Z[∆op ],
0
where for each n ≥ 1 and i ∈ {0, . . . , n} one lets
j, if j ≤ i;
dni ∈ Hom∆op (n∗ , (n − 1)∗ ), dni (j) ∶= {
j − 1, if j > i.
One has di ○ dj = dj−1 ○ di , ∀i < j, and this shows that bn ○ bn+1 = 0. Next, we
provide the geometric proof ([9]) of the commutation of the λ-operations with the
Hochschild boundary.
Proposition A.4. The elements Λkn ∈ Z[∆op ⋉ N× ] commute with the Hochschild
boundary b i.e. one has Λkn−1 bn = bn Λkn , ∀n ≥ 1.
Proof. Note first that since the number of terms in the formula (50) defining Λkn
is the cardinality k n of Σkn , the expansion of the sum
Λkn−1 bn = ∑ (−1)j (Perm(δ)) πn−1

k
○ δ ○ dnj
δ∈Σk
n−1
0≤j≤n
contains (n + 1)k n−1 terms while the expansion of
bn Λkn = ∑ (−1)i (Perm(α)) dni ○ πnk ○ α

α∈Σkn
0≤j≤n
contains (n + 1)k n terms. One has
dni ○ πnk = πn−1

k
○ Sd∗k (dni )
and thus the equality to be proved is

∗ n
∑ (−1) (Perm(δ)) δ ○ dj = ∑ (−1) (Perm(α)) Sdk (di ) ○ α.
j n i
δ∈Σk α∈Σkn
n−1
0≤j≤n 0≤i≤n
We show geometrically that among the (n+1)k n terms on the right hand side of the
above formula there are (n + 1)k n−1 terms which correspond to the left hand side
while the others cancel in pairs. Indeed, we construct a natural correspondence
(cf. (56) below) between the terms Sd∗k (dni )○α and the faces of the simplices in the
triangulation of the standard simplex ∆n = {(xb ) ∣ 0 ≤ x1 ≤ ⋯ ≤ xn ≤ 1} given by
the edgewise subdivision into k n simplices ∆(α) parametrized by α ∈ Σkn . The faces
which belong to the interior of ∆n appear twice and with opposite orientations.
This corresponds to the cancelation by pairs of (n + 1)(k n − k n−1 ) terms. On

the other hand, the faces which belong to the boundary of ∆n correspond to
the remaining (n + 1)k n−1 terms which correspond to the terms on the left hand
side. Consider the affine isomorphism (7), and associate to the elements x ∈
Hom∆op (n∗ , [0, 1]) = ∆n the map of intervals
Sd∗k (x) ∈ Hom∆op (Sd∗k (n∗ ), Sd∗k ([0, 1])) ≅ Hom∆op (Sd∗k (n∗ ), [0, 1]).
In terms of the coordinates x(j) one has
a x(b)
(54) Sd∗k (x)(j) = + , ∀j = b + a(n + 1) ∈ Sd∗k (n∗ ).
k k
One obtains in this way a triangulation of ∆n by the k n simplices
(55) ∆(α) = {Sd∗k (x) ○ α ∣ x ∈ ∆n } , ∀α ∈ Σkn .
To each simplex ∆(α) corresponds a permutation Perm(α) = (πnk ○ α)∗ . Geomet-

rically the map x ↦ Sd∗k (x) ○ α is of the form (xb ) ↦ (yb ), yb = k1 xσ(b) + sb , σ =
Perm(α). For each element β ∈ Hom∆op (n∗ , Sd∗k ((n − 1)∗ )) we consider the subset
of ∆n given by
(56) F (β) = {Sd∗k (y) ○ β ∣ y ∈ ∆n−1 }, ∆n−1 = Hom∆op ((n − 1)∗ , [0, 1])
Lemma A.5. (i) Let β, β ′ ∈ Hom∆op (n∗ , Sd∗k ((n−1)∗ )) be such that the faces F (β),
F (β ′ ) are non-degenerate and equal then β = β ′ .
(ii) Let α ∈ Σkn and ∆(α) the corresponding simplex as in (55). Then the faces of
∆(α) are the F (Sd∗k (dni )α) for i ∈ {0, . . . , n}.
Proof. (i) This follows since the barycenter of a non-degenerate face F (β) de-
termines β. More precisely since one assumes that F (β) is non-degenerate the
map ∆n−1 → F (β), y ↦ Sd∗k (y) ○ β is affine and injective and thus the barycenter
of F (β) is the image Sd∗k (z) ○ β of the barycenter z ∈ Hom∆op ((n − 1)∗ , [0, 1])
of ∆n−1 . One has z(j) = j/n ∈ [0, 1], ∀j ∈ {0, . . . , n}. It follows from (54) that
Sd∗k (z)(i) = i/(nk), ∀i ∈ {0, . . . , nk}. Thus the coordinates Sd∗k (z) ○ β(i) of the
barycenter of F (β) determine β.
(ii) Using (56), one has
F (Sd∗k (dni )α) = {Sd∗k (y) ○ Sd∗k (dni )α ∣ y ∈ ∆n−1 } = {Sd∗k (y ○ dni )α ∣ y ∈ ∆n−1 }
and the result follows since the faces of the standard simplex ∆n = Hom∆op (n∗ , [0, 1])
are the
∂i ∆n = {y ○ dni ∣ y ∈ ∆n−1 } ⊂ Hom∆op (n∗ , [0, 1]).
Appendix B. Epicyclic modules and the λ-operations
In this appendix we first review the definition of cyclic homology for cyclic mod-
ules, then we give a detailed description of the action of the λ-operations on the
(b, B) bicomplex in the context of epicyclic modules. Since the cyclic category is
isomorphic to its dual, we use a covariant definition of cyclic modules which turns
out to be more convenient when one discusses the epicyclic construction.
We recall that the cyclic category Λ is the full subcategory of the category Arc
with one object n for each non-negative integer n ≥ 0, given by the archimedean
set n = (Z, θ), θ(x) = x + n + 1 ∀x ∈ Z. Thus the morphisms between objects are
given by HomΛ (n, m) ∶= C(n + 1, m + 1), where for each pairs of integers a, b > 0,
C(a, b) denotes the set of equivalence classes of maps f ∶ Z → Z, with f (x) ≥ f (y),
∀x ≥ y and f (x + a) = f (x) + b, ∀x ∈ Z (cf. [4] §2.2). The equivalence relation is
defined as follows
f ∼ g ⇐⇒ ∃k ∈ Z, g(x) = f (x) + kb ∀x ∈ Z.
The cyclic category has a decomposition Λ = ∆C which describes Λ as an extension

of the small simplicial category ∆ by means of a new generator τn ∈ Cn+1 ∶=
AutΛ (n) for each n ≥ 0, and fulfilling the relations (in terms of the faces δj and
degeneracies σj , 2 ≤ j ≤ n, which describe a presentation of ∆: cf. op.cit. for
details)
τnn+1 = id,
τn ○ σ0 = σn ○ τn+1
2
τn ○ σj = σj−1 ○ τn+1 , ∀j ∈ {1, . . . , n}
τn ○ δ0 = δn τn ○ δj = δj−1 ○ τn−1 , ∀j ∈ {1, . . . , n}.
The epicyclic category is the full subcategory of the category Arc⋉N× of archimedean
sets whose objects are the archimedean sets n. We recall that for (X, θ) an
archimedean set and k > 0 an integer, the pair (X, θk ) is also an archimedean
set which we denote as
(57) Ψk (X, θ) ∶= (X, θk ).
The epicyclic category Λ̃ is obtained (cf. [2], Definition 1.1) by adjoining to Λ new
morphisms Idkn ∶ Ψk (n) → n for n ≥ 0, k ≥ 1, which fulfill the following relations
(1) Id1n = idn , Id`n ○ Idk`(n+1)−1 = Idk`
n
(2) α Idkm = Idkn Sdk (α), for any α ∈ Hom∆ ([m], [n])
(3) τn Idkn = Idkn τk(n+1)−1
where Sdk ∶ ∆ Ð→ ∆ is the edgewise subdivision functor ((1), (2)).
The canonical inclusion ∆op ⊂ Λ extends to an inclusion ∆op ⋉ N× ⊂ Λ̃ which
associates to the morphisms πnk ∈ Hom∆op ⋉N× (Sd∗k (n∗ ), n∗ ) the morphism Idkn ∈
HomΛ̃ (Ψk (n), n).
B.1 – Cyclic homology and cyclic modules

Definition B.1. A cyclic module E is a covariant functor Λ Ð→ Ab from the
cyclic category to the category of abelian groups.
We briefly recall the construction of the normalized (b, B)-bicomplex of a cyclic

module E. We keep the notations of A.2 and A.3. For each integer n ≥ 0, one
defines the element B(n) = B0 A ∈ Z[Λ], for A ∶= ∑ (Perm(τ ))τ , where the sum-
mation is taken over τ ∈ AutΛ (n), while the map B0 ∈ HomΛ (n, n + 1) is defined
as follows
(58) B0 (i + a(n + 1)) = 1 + i + a(n + 2) , ∀i, 0 ≤ i ≤ n, a ∈ Z.
Let sj = snj ∈ Hom∆op (n∗ , (n + 1)∗ ) be the degeneracies, for 0 ≤ j ≤ n. They are
given by
snj (i) = i if i ≤ j, snj (i) = i + 1 if i > j
so that the image of snj does not contain j + 1. Note that B0 ≠ snj , ∀j, n. For n ≥ 0
let
(59) Cn (E) ∶= E(n)/V (n), V (n) ∶= ⊕ Im(sj )
where the sj = sn−1

j ∈ Hom∆op ((n − 1)∗ , n∗ ) are the degeneracies for 0 ≤ j ≤ n − 1.
Both the operators b and B pass to the quotient and one has (cf. [3], [8])
Lemma B.2. (i) B (⊕ Im(sj )) ⊂ (⊕ Im(s` )) ∀j, `.

(ii) The subspace W (n) ∶= Im(B0 ) ⊕ V (n) ⊂ E(n) is invariant under cyclic per-
mutations: E(t)W (n) ⊂ W (n).
(iii) Im(B ○ B0 ) ⊂ (⊕ Im(s` )) , ∀`.
(iv) b (⊕ Im(sj )) ⊂ (⊕ Im(s` )) , ∀j, `.
(v) b2 = B 2 = bB + Bb = 0 on Cn (E).
We set the (b, B)-bicomplex in negative degrees as follows
(60) C α,β ∶= Cα−β (E) if α ≥ β, α ≤ 0, C α,β ∶= {0}, otherwise,
and one defines the cyclic homology of the cyclic module E by
HCn (E) ∶= H−n ( ⊕ C α,β , b + B)

α≤0
B.2 – Epicyclic modules

We generalize the above set-up to the epicyclic category.
Definition B.3. An epicyclic module E is a covariant functor Λ̃ Ð→ Ab from the

epicyclic category to the category of abelian groups.
The inclusion of categories Λ Ð→ Λ̃ turns the operator B into an element of the

ring Z[Λ̃] for each n ≥ 0 (we denote it by B(n)). Moreover, the inclusion of
categories ∆op ⋉ N× Ð→ Λ̃ allows one to view the operators Λkn ∈ Z[∆op ⋉ N× ] as
elements of the ring Z[Λ̃]. In this section we give a proof of the commutation
relation Λkn+1 B = kBΛkn (cf. [8], [9]) at the categorical level. We use the following
two variants of the sets Σkn as in (45), obtained by replacing the category ∆op with
∆ and Λ respectively. We define, using the canonical inclusion ∆ ⊂ Λ
(61) ∆kn = {α ∈ Hom∆ ([n], Sdk ([n])) ∣ (Idkn ○ α)∗ is bijective}
and
(62) Γkn = {α ∈ HomΛ (n, Ψk (n)) ∣ (Idkn ○ α)∗ is bijective}
where it follows from (57) that Ψk (n) = (Z, tk(n+1) ) is the archimedean set (Z, θk ),
with θk ∶ Z → Z, θk (x) = x + k(n + 1).
Lemma B.4. (i) The inclusion of categories ∆op ⊂ Λ induces an inclusion of sets
Σkn ⊂ Γkn and one has
(63) Γkn = AutΛ (Ψk (n))Σkn .
(ii) The canonical decomposition Λ = ∆C determines an inclusion of sets ∆kn ⊂ Γkn

and one has
(64) Γkn = ∆kn AutΛ (n)
(iii) For any α ∈ Σkn+1 there exists a unique element α′ ∈ ∆kn such that αB0 =
Sdk (B0 )α′ . Moreover the map α ↦ α′ defines a bijection of sets Σkn+1 ≃ ∆kn .
Proof. (i) Let α ∈ Hom∆op (n∗ , Sd∗k (n∗ )), then it is easy to see that its image
into the cyclic category α̃ ∈ HomΛ (n, Ψk (n)) belongs to Γkn if and only if α ∈ Σkn .
In that case, the associated permutation of {0, . . . , n} is obtained by extending
Perm(α) to a permutation of {0, 1, . . . , n} by fixing 0. One has Λ = C∆op and
every morphism φ in Λ uniquely decomposes as φ = ta ○ δ, where δ is in ∆op . For
φ ∈ Γkn , one has φ = ta ○ δ where δ ∈ Hom∆op (n∗ , Sd∗k (n∗ )), thus one gets (63).
(ii) By construction one has Γkn ∩ ∆ = ∆kn . For α ∈ HomΛ (n, Ψk (n)), the right
multiplication α ↦ ατ by an element τ ∈ AutΛ (n) does not affect the condition:
“(Idkn ○ α)∗ is bijective” thus the result follows from the decomposition Λ = ∆C.
(iii) For α ∈ Hom∆op ((n + 1)∗ , (k(n + 2) − 1)∗ ) we let α̃ its canonical lift to a
non-decreasing map
α̃ ∶ Z → Z, α̃(x) = α(x) , ∀x ∈ {0, . . . , n+1}, α̃(x+n+2) = α̃(x)+k(n+2) , ∀x ∈ Z.
By hypothesis the map Perm(α) which to x ∈ {1, . . . , n+1} associates α(x) modulo
n + 2 is a permutation of {1, . . . , n + 1}. Thus the map φ ∶ Z → Z, φ = α̃ ○ B0 is
increasing and fulfills the following properties:
● φ(x + a(n + 1)) = φ(x) + ka(n + 2), ∀x, a ∈ Z

● φ(x) ≠ 0 modulo n + 2, ∀x ∈ Z.
We show that these conditions imply that there exists α′ ∈ Hom∆ ([n], [k(n+1)−1])
such that φ = Sdk (B0 )α′ , i.e. the following diagram commutes
{0, . . . , n, n + 1}
α / {0, . . . , k(n + 2) − 1}
O O
B0 Sdk (B0 )
{0, . . . , n}
α′ / {0, . . . , k(n + 1) − 1}
One has B0 ∈ Hom∆ ([n], [n + 1]) ⊂ HomΛ (n, n + 1) and the morphism Sdk (B0 )
which belongs to Hom∆ (Sdk ([n]), Sdk ([n + 1])) is also given by (58). Thus the
range of φ ∶ Z → Z is contained in the range of the injection Sdk (B0 ) and for each
x ∈ Z there exists a unique y such that φ(x) = Sdk (B0 )(y). Set α′ (x) = y, then
α′ ∶ Z → Z is a non-decreasing map which satisfies:
● α′ (x + a(n + 1)) = α′ (x) + ka(n + 1), ∀x, a ∈ Z
● α′ (x) ∈ {0, . . . , k(n + 1) − 1}, ∀x ∈ {0, . . . , n}.
Thus α′ ∈ Hom∆ ([n], [k(n + 1) − 1]) and the map (Idkn ○ α′ )∗ is the same as the
permutation Perm(α) of the set {1, . . . , n+1} reindexed as {0, . . . , n}, thus α′ ∈ ∆kn .
Moreover the map α ↦ α′ determines a bijection Σkn+1 ≃ ∆kn since it is injective
by construction and the cardinality of Σkn+1 , which is k n+1 by Lemma A.1, is the
same as the cardinality of ∆kn (which is by (ii) equal to 1/n × #(Γkn ) = k n+1 using
(i)).
Theorem B.5. For any n, k ≥ 1 one has in Z[Λ̃]
(65) Λkn+1 B(n) = kB(n)Λkn
Proof. The left hand side of (65) is a sum of (n + 1)k n+1 terms which, up to sign,
are of the form
Idkn+1 αB0 τ, α ∈ Σkn+1 , τ ∈ AutΛ (n).
By Lemma B.4 one has αB0 = Sdk (B0 )α′ , so that these terms can be rewritten as
follows
(66) Idkn+1 Sdk (B0 )α′ τ, α′ ∈ ∆kn , τ ∈ AutΛ (n).
The coefficient of k in the right hand side of (65) is a sum of (n + 1)k n terms
which, up to sign, are of the form B0 ξ Idkn β, β ∈ Σkn , ξ ∈ AutΛ (n).
One also has ξIdkn = Idkn ξ˜ where ξ˜ ∈ AutΛ (Ψk (n)) is a lift of the cyclic permutation
ξ ∈ AutΛ (n). Moreover B0 Idkn = Idkn+1 Sdk (B0 ). Thus we conclude that B0 ξ Idkn β =
Idkn+1 Sdk (B0 )ξ˜β. One has ξ˜β ∈ AutΛ (Ψk (n))Σkn and using the coefficient k in
front of the right hand side of (65) together with the equality Idkn η = Idkn η ′ for
any two of the k lifts of ξ one can rewrite the right hand side of (65) as a sum of
(n + 1)k n+1 terms which up to sign are of the form
(67) Idkn+1 Sdk (B0 )ξ˜β, ξ˜ ∈ AutΛ (Ψk (n)), β ∈ Σkn .
Thus the validity of (65) follows from Lemma B.4 which gives AutΛ (Ψk (n))Σkn =
Γkn = ∆kn AutΛ (n) and the fact that the signs in front of the terms in (66) and (67)
are always given by the signature of the associated permutations.
B.3 – λ-operations on the cyclic homology of epicyclic modules
Let E be an epicyclic module. Then its restriction to Λ ⊂ Λ̃ is a cyclic module and

one has the associated abelian groups (59)
Cn (E) ∶= E(n)/V (n), V (n) ∶= ⊕ Im(sj ).
The operations Λkn are meaningful on the quotient in view of the following
Lemma B.6. Λkn (⊕ Im(sj )) ⊂ (⊕ Im(s` )) , ∀j, `.
Proof. For any σ ∈ Σkn and any degeneracy sj , there exists α such that with
i = Perm(σ)(j + 1), one has σ ○ sj = Sd∗k (si−1 ) ○ α. The conclusion follows using
Idkn Sd∗k (si−1 ) = si−1 Idkn .
Lemma B.7. (i) The following equality defines, for any integer k ≥ 1, an endo-
morphism of the (b, B)-bicomplex of E:
(68) θ(k)ξ = k −α Λkα−β ξ , ∀ξ ∈ C α,β
(ii) The translation (α, β) ↦ (α + 1, β + 1) defines an endomorphism S of the

(b, B)-bicomplex and one has Sθ(k) = kθ(k)S.
Proof. (i) The commutation of θ(k) with b and B follows from Proposition A.4
and Theorem B.5 respectively. More precisely one has for ξ ∈ C α,β , Bξ ∈ C α+1,β
and
Bθ(k)ξ = Bk −α Λkα−β ξ = k −α−1 kBΛkα−β ξ = k −(α+1) Λkα+1−β Bξ = θ(k)Bξ.
(ii) Follows directly from the definition of θ(k) as in (68).

The endomorphisms θ(k) as in (68) define, at the categorical level, the λ-operations
in homology. Lemma B.7 immediately implies the following well-known result
(cf. [8], Ex. 6.4.5)
Theorem B.8. Let E be an epicyclic module. The λ-operations define an action

of N× on HCn (E) given by the induced action of the operators θ(k).
When the epicyclic module E factors through Fin one has, after tensorisation of
the abelian groups by Q, a decomposition (cf. [8] Thm. 6.4.5)
HCn (E) = ⊕ HCn(j) (E)
j≥0
×
which diagonalizes the action of N as a sum of the representations given by the
characters N× ∋ k ↦ k j . This decomposition of an epicyclic module does not hold
in general. Let E be an epicyclic module and ρ ∶ N× → V be a representation of
N× , then the tensor product E ⊗ V endowed with the maps
(E ⊗ V )(f ) = E(f ) ⊗ ρ(Mod(f )) ∶ E(n) ⊗ V → E(m) ⊗ V , ∀f ∈ HomΛ̃ (n, m)
is still an epicyclic module. When working with vector spaces over a field, this
construction tensors the cyclic homology by V and replaces the action of N×
on HCn (E) by its tensor product with ρ. Thus this twisting process generates
epicyclic modules with arbitrary weights.
References
[1] M. Artin, A. Grothendieck and J. L. Verdier, Théorie de topos et cohomologie étale
des schémas, (“SGA4”) Springer LNM 269 and 270, Springer-Verlag, Berlin, 1972.
[2] D. Burghelea, Z. Fiedorowicz, W. Gajda, Power maps and epicyclic spaces, Journal
of Pure and Applied Algebra 96 (1994), 1–14.
[3] A. Connes, Cohomologie cyclique et foncteurs Extn , C. R. Acad. Sci. Paris Sér. I
Math. 296 (1983), no. 23, 953–958.
[4] A. Connes, C. Consani, Cyclic structures and the topos of simplicial sets, J. Pure
Appl. Algebra 219 (2015), no. 4, 1211–1235.
[5] A. Connes, C. Consani, Projective geometry in characteristic one and the epicyclic
category, Nagoya Mathematical Journal 217 (2015), 95–132.
[6] A. Connes, C. Consani, The Arithmetic Site, Comptes Rendus Mathematique Ser.
I 352 (2014), 971–975.
[7] A. M. W. Glass, Partially ordered groups, volume 7 of Series in Algebra. World
Scientific Publishing Co. Inc., River Edge, NJ, 1999.
[8] J.L. Loday, Cyclic homology. Grundlehren der Mathematischen Wissenschaften, 301.
Springer-Verlag, Berlin, 1998.
[9] R. McCarthy, On operations for Hochschild homology. Comm. Algebra 21 (1993),
no. 8, 2947–2965.
[10] S. Mac Lane, I Moerdijk, Sheaves in geometry and logic. A first introduction to
topos theory. Corrected reprint of the 1992 edition. Universitext. Springer-Verlag,
New York, 1994.
[11] I Moerdijk, Cyclic sets as a classifying topos (preprint).
Received submission date; revised revision date

The Cyclic and Epicyclic Sites 2015

Uploaded by

Copyright:

Available Formats

The Cyclic and Epicyclic Sites 2015

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

The Cyclic and Epicyclic Sites 2015

Uploaded by

Copyright:

Available Formats

Rend. Sem. Mat. Univ.

Padova 1xx (201x) Rendiconti del Seminario Matematico

The Cyclic and Epicyclic Sites

Alain Connes – Caterina Consani∗

Mathematics Subject Classification (2010). 18B25; 20L05, 19D55.

Keywords. Grothendieck topos, Cyclic category, Groupoids, Characteristic one, Pro-

The theory of topoi of Grothendieck provides the best geometric framework to

Alain Connes, Collège de France, 3 rue d’Ulm, Paris F-75005 France

Theorem. The category of points of the epicyclic topos (Λ̃op )∧ is equivalent to

It is important to realize the relevance of the language of Grothendieck topoi to

2.1 – The edgewise subdivision functors Sdk

(1) Sdk (F ) ∶= {0, . . . , k − 1} × F

(2) Sdk (f ) ∶= Id × f ∶ Sdk (F ) → Sdk (F ′ )

Sdkk′ = Sdk ○ Sdk′ , ∀k, k ′ ∈ N× .

(a, i) ↦ i + a(n + 1) , ∀a ∈ {0, . . . , k − 1}, i ∈ {0, . . . , n}.

(3) Sdk (f )(i + a(n + 1)) = f (i) + a(m + 1) , ∀i, a, 0 ≤ i ≤ n, 0 ≤ a ≤ k − 1.

One checks directly that Sdkk′ = Sdk ○ Sdk′ .

Proof. The morphism (Sdk (f ))∗ is defined by the equivalence

Theorem 2.3. The action Sd ̂k of the geometric morphism Sdk (k ∈ N× ) on the

(5) F1 ([n]) = Hom∆op (n∗ , Sd∗k (I)), F2 ([n]) =

= ( ∐ (Hom(m∗ , I) × Hom∆op (n∗ , Sd∗k (m∗ )))) / ∼

ˆ applied to the contravariant

Proof. The statement follows using the affine isomorphism

2.2 – Canonical decomposition of ϕ ∈ Hom∆op (n∗ , Sd∗k (I))

Moreover, the morphism α ∈ Hom∆op (r∗ , I) is the unique increasing injection

Corollary 2.6. For any interval I the map

2.3 – Explicit description of the isomorphism F1 ≃ F2

(8) F2 ([n]) = ( ∐ (Hom(m∗ , I) × Hom∆op (n∗ , Sd∗k (m∗ )))) / ∼

where the equivalence relation is generated by

Lemma 2.7. The map

Φ ∶ F2 ([n]) → F1 ([n]) = Hom∆op (n∗ , Sd∗k (I)), (α, β) ↦ Sd∗k (α) ○ β

ϕ = Sd∗k (αc ) ○ βc , βc ∈ Hom∆op (n∗ , Sd∗k (r∗ )), αc ∈ Hom∆op (r∗ , I)

is the canonical decomposition of ϕ = Sd∗k (α) ○ β.

(α, β) ∼ (α ○ α0 , β0 ) = (αc ○ ρ, β0 ) ∼ (αc , Sd∗k (ρ) ○ β0 ) = (αc , βc ).

2.4 – The small category ∆op ⋉ N×

where πnk implements the endofunctor Sd∗k , i.e.

Proposition 2.8. : For any n ≥ 0, k ∈ N× , let (πnk )∗ ∶ F(Sd∗k (n∗ )) → F(n∗ )

3. The epicyclic category and the oriented groupoı̈ds

3.1 – Generalities on oriented groupoı̈ds

Definition 3.1. An oriented groupoı̈d (G, G+ ) is a groupoı̈d G endowed with a

Let (G, G+ ) be an oriented groupoı̈d and let x ∈ G(0) . The set Gx ∶= {γ ∈ G ∣

In the following subsections we describe two constructions of oriented groupoı̈ds

s(x, h) ∶= x, r(x, h) ∶= hx, (x, h) ○ (y, k) ∶= (y, hk).

Proof. By definition, the subset H+ ⊂ H of the group H is stable under product

(14) g(m) ∶= (Z/(m + 1)Z) ⋉ Z.

G(X, H) ∶= (X × X)/ ∼ (x, y) ∼ (h(x), h(y)), ∀h ∈ H.

(15) G+ (X, H) = {(x, y) ∈ G(X, H) ∣ x ≥ y}.

Proof. Since H acts by order automorphisms the condition x ≥ y is independent

Proof. The associated oriented groupoı̈d (G, G+ ) is by construction the quotient

ψ ∶ G → g(m) = (Z/(m + 1)Z) ⋉ Z, ψ(x, y) = (π(y), x − y)

where π ∶ Z → Z/(m + 1)Z is the natural projection. One has by restriction

3.2 – The oriented groupoı̈d associated to an archimedean set

Definition 3.5. An archimedean set is a pair (X, θ) of a non-empty, totally

∀x, y ∈ X, ∃n ∈ N s.t. y ≤ θn (x).

G(X, θ) ∶= (X × X)/ ∼, (x, y) ∼ (θn (x), θn (y)), ∀n ∈ Z

(16) G+ (X, θ) ∶= {(x, y) ∈ G(X, θ) ∣ x ≥ y}.

Proof. Since θ is an order automorphism of X, the group Z acts by order auto-

Proposition 3.7. Let (G, G+ ) be an oriented groupoı̈d fulfilling the conditions of

Proof. The implication γ ≤ γ ′ Ô⇒ θ(γ) ≤ θ(γ ′ ) follows since right multiplica-

3.3 – The category of archimedean sets in terms of oriented groupoı̈ds