aa r X i v : . [ m a t h . N T ] D ec NUMBERS AS FUNCTIONS Yuri I. Manin
Max–Planck–Institut f¨ur Mathematik, Bonn, Germany
ABSTRACT. In this survey I discuss A. Buium’s theory of “differential equa-tions in the p –adic direction” ([Bu05]) and its interrelations with “geometry overfields with one element”, on the background of various approaches to p –adic modelsin theoretical physics (cf. [VlVoZe94], [ACG13]). Introduction
One of the most beautiful (arguably, the most beautiful) mathematical formulasis Euler’s identity e πi = − . (0 . π = 3 , . . . , e = 2 , . . . , i = √−
1, and − −
1” in the right hand side of (0.1) shows how two quantumstates with opposite phases may annihilate each other after superposition.On the other hand, of these four numbers π, e, i, − only π looks as somethingsimilar to a “physical constant” in the sense that it can be (and was) measured,with a certain approximation.Moreover, the traditional names of the respective classes of numbers, which wenowadays tend to perceive as mathematical terms introduced by precise definitionsin courses of calculus, – irrational, transcendent, imaginary, negative, – in the courseof history conveyed the primeval bafflement of the rational mind, discovering thesenumbers but reluctant to accept them.We may recall that at the time of their discovery these numbers had very dif-ferent sources of justification: π in Euclidean geometry (which describes essentiallykinematics of solids in gravitational vacuum), − e in theearly history of computer science (Napier’s implementation of the discovery thata specific precomputation can facilitate everyday tasks of multiplication), i in theearly history of polynomial equations. Based on talks at the
International Workshop on p –adic methods for modelling of complexsystems , Bielefeld, April 15–19, 2013, and at Journ´ees Arithm´etiques , Grenoble, June 2–5, 2013. When I was asked to deliver a talk at the Workshop on p –adic Methods forModelling of Complex Systems, I decided to present first a p –adic environment of π and e .Probably, the earliest “arithmetic” formula involving π is due to Euler (as wellas (0.1)): π Y p (1 − p − ) − . (0 . all primes p simultaneously, and in fact, can be best understoodas a fact from ad`elic geometry. As such, it looks as a generalisation of the simple–minded product formula Q v | a | v = 1 valid for all a ∈ Q ∗ , where v runs overall valuations of Q , p –adic ones and archimedean one. To be more precise, (0.2)expresses the fact that the natural adelic measure of SL (2 , A Q ) /SL (2 , Q ) equals1. For some more details, cf. [Ma89], where it was suggested that fundamentalquantum physics might be related to number theory via this ad`elic philosophy, “democracy of all valuations” , and the exclusive use of real and complex numbersin our standard formalisms is the matter of tradition, which we now try to overcomeby replacing “the first among equals” archimedean valuation by an arbitrary non–archimedean one.Now we turn to e . Here, as the discoverer of p –adic numbers Kurt Hensel himselfremarked, we have a candidate for e p in each p –adic field, since the (archimedean)series for e p converges also p –adically: e p = ∞ X n =0 p n n ! . (0 . p of the right hand side of (0.3) understood as p –adicnumber generates an extension of Q p of degree p , there can be no algebraic numberwith such local components.This argument looks tantalisingly close to a proof of transcendence of e , although,of course, it is not one. On the other hand, I do not know any ad`elic formulainvolving e in such a way as (0.2) involves π .In this survey, I proceed with discussion consisting of three main parts. A. I will describe a class of numbers (including transcendental ones) relevant forQuantum Field Theory in the sense that they define the coefficients of perturbative series for Feynmann’s path integrals. These numbers are called (numerical) periods,they were introduced and studied in [KoZa01].
Roughly speaking, numerical periods are values at algebraic points of certain multi–valued transcendental functions , naturally defined on various moduli spaces,and also traditionally called (functions–)periods.These functions–periods satisfy differential equations of Picard–Fuchs type, andsuch equations furnish main tools for studying them.In the second part of this survey, I focus on the following program:
B. For a prime p , numerical periods also can be considered as solutions of “dif-ferential equations in the p –adic direction” .The whole machinery of such differential equations was suggested and developedby Alexandru Buium, cf. his monograph [Bu05], and I briefly review it. I use thecatchword “numbers as functions” to name this analogy.Alexandru Buium has convincingly shown that the right analog of the p –adicderivation is (a natural generalization of) the Fermat quotient δ p ( a ) := ( a − a p ) /p initially defined for a ∈ Z . Unexpectedly, this formal idea had rich consequences:Buium was able to construct analogs of classical jet spaces “in the p –adic direction”,together with a theory of functions on these jet spaces, containing an incredibleamount of analogs of classical constructions traditionally requiring calculus.Those numerical periods that were already treated by Buium include periodsof abelian varieties defined over number (or even p –adic) fields. (But the readershould be aware that, in the absence of uniformization, this last statement onlyvery crudely describes a pretty complicated picture; see more details in the maintext.) C. For Buium’s differential equations, “constants in the p –adic direction” turnout to be roots of unity and zero: Teichm¨uller’s representatives of residue classesmodulo p . Until recently, algebraic geometry over such constants was motivated by verydifferent insights: for a more detailed survey cf. [Ma95], [Ma08]. It is known as“theory of the field F ”.Briefly, this last field of inquiry is focused on the following goal: to make theanalogy between, say, Spec Z (or spectra of rings of algebraic integers) on the onehand, and algebraic curves over finite fields, on the other hand, so elaborated and precise that one could use a version of the technique of Andr´e Weil, AlexanderGrothendieck and Pierre Deligne in order to approach Riemann’s conjecture forRiemann’s zeta and similar arithmetic functions.The solid bridge between F –geometry and arithmetic differential equations wasconstructed by James Borger: cf. [Bor11a,b], [Bor09], [BorBu09]. Roughly speak-ing, in order to define the p –adic derivative δ p of elements of a commutative ring A , one needs a lift of the Frobenius map, that is an endomorphism a F ( a ),such that F ( a ) ≡ a p mod p. Borger remarked that a very natural system of suchlifts for all p simultaneously is encoded in the so called psi–structure or its slightmodification, lambda–structure, and then suggested to consider such a structureas descent data on Spec A to F . A related notion of “cyclotomic coordinates” in F was independently suggested in [Ma08]. In particular, a ∈ A is a cyclotomicco–ordinate (wrt a prime p ) if F ( a ) = a p . I will return to these ideas in the lastpart of this survey.Finally, I should mention that there exists a very well developed deep theory of“ p –adic periods” for algebraic varieties defined over p –adic fields that replaced theclassic integration of differential forms over topological cycles with comparison ofalgebraic de Rham and ´etale cohomology theories: see [Fa88] and a recent contribu-tion and brief survey [Be11]. Periods in this setting belong to a very big Fontaine’sfield B dR . The approach to periods via Buium’s p –adic geometry that we describein this survey has a very different flavour. It would certainly be important to findconnections between the two theories.
1. Periods1.1. Numerical periods.
M. Kontsevich and D. Zagier introduced an im-portant subring
P ⊂ C containing all algebraic numbers and a lot of numbersimportant in physics (see [KoZa01]). α ∈ P if and only if the real and imaginary parts of α arevalues of absolutely convergent integrals of functions in Q ( x , . . . , x n ) over chainsin R n given by polynomial (in)equalities with coefficients in Q . a) All algebraic numbers are periods.b) π = R R x + y ≤ dxdy .c) Γ ( p/q ) q ∈ P . It is not difficult to prove that periods form a subring of C . Feynman integrals(of a certain class) are periods. But it is still not known whether π − , e , or Euler’sconstant γ are periods (probably, not). There is a close connection between periodsand Grothendieck motives (see [KoZa01]), and 2 πi corresponds to the Tate’s motive.Since in the motivic formalism one formally inverts the Tate motive, it is also usefulto extend the period ring by (2 πi ) − .d) The multiple ζ –values (Euler) ζ ( n , ..., n m ) = X
Non–linear DE for the periods of dX/Y over relative 1–cycles withboundaries at sections P := ( X ( t ) , Y ( t )) of finite order: µ ( P ) = 0 , (1 . µ ( P ) := Y ( t )2( X ( t ) − t ) − ddt (cid:20) t ( t − X ′ ( t ) Y ( t ) (cid:21) + 2 t ( t − X ′ ( t ) Y ′ ( t ) Y ( t ) . (1 . µ defined by (1 .
3) and extended to the function on the set of L –points of the generic fiber E t with values in any differential extension L of Q ( t ) is“a differential character”: µ ( P + Q ) = µ ( P ) + µ ( Q ) (1 . µ ( P ) = L t Z P ∞ dX/Y because L t ( dX/Y ) = d Y ( X − t ) . Here I will briefly describe the heuris-tic origin of a set of numerical periods (and periods–functions) indexed by labeledgraphs relevant for quantum field theory, following [Ma09], sec. 1. For a more fo-cussed study of (some) of the integrals appearing in this way see [M¨uWZa12] and[W13].
A Feynman path integral is an heuristic expression of the form R P e S ( ϕ ) D ( ϕ ) R P e S ( ϕ ) D ( ϕ ) (1 . correlation functions .Here the integration domain P stands for a functional space of classical fields ϕ on a space–time manifold M . Space–time may be endowed with a fixed Minkovskior Euclidean metric. In models of quantum gravity metric is one of the fields.Fields may be scalar functions, tensors of various ranks, sections of vector bundles,connections. S : P → C is a functional of classical action : generally S ( ϕ ) is expressedas an integral over M of a local density on M which is called Lagrangian.
Inour notation (1.6) S ( ϕ ) = − R M L ( ϕ ( x )) dx. Lagrangian density may depend onderivatives, include distributions etc.Usually S ( ϕ ) is represented as the sum of a quadratic part S ( ϕ ) (Lagrangian offree fields) and remaining terms which are interpreted as interaction and treatedperturbatively.Finally, the integration measure D ( ϕ ) and the integral itself R P should be consid-ered as simply a part of the total expression (1.6) expressing the idea of “summingthe quantum probability amplitudes over all classical trajectories”.To explain the appearance and combinatorics of Feynman graphs, we considera toy model, in which P is replaced by a finite–dimensional real space. We endowit with a basis indexed by a finite set of “colors” A , and an Euclidean metric g encoded by the symmetric tensor ( g ab ) , a, b ∈ A. We put ( g ab ) = ( g ab ) − . The action functional S ( ϕ ) will be a formal series in linear coordinates on P ,( ϕ a ), of the form S ( ϕ ) = S ( ϕ ) + S ( ϕ ) , S ( ϕ ) := − X a,b g ab ϕ a ϕ b ,S ( ϕ ) := ∞ X k =1 k ! X a ,...,a k ∈ A C a ,...,a k ϕ a . . . ϕ a k (1 . C a ,...,a n ) are certain symmetric tensors. If these tensors vanish for allsufficiently large ranks n , S ( ϕ ) becomes a polynomial and can be considered as agenuine function on P . Below we will treat ( g ab ) and ( C a ,...,a n ) as independentformal variables, “formal coordinates on the space of theories”.Now we can express the toy version of (1.6) as a series over (isomorphism classesof) graphs.Here a graph τ consists of two finite sets, edges E τ and vertices V τ , and theincidence map sending E τ to the set of unordered pairs of vertices. Each vertex issupposed to be incident to at least one edge. There is one empty graph. The formula for (1.6) including one more formal parameter λ (“Planck’s con-stant”) looks as follows: R P e λ − S ( ϕ ) D ( ϕ ) R P e λ − S ( ϕ ) D ( ϕ ) = X τ ∈ Γ λ − χ ( τ ) | Aut τ | w ( τ ) (1 . τ . The weight w ( τ ) of such a graph isdetermined by the action functional (1.2) as follows: w ( τ ) := X u : F τ → A Y e ∈ E τ g u ( ∂e ) Y v ∈ V τ C u ( F τ ( v )) . (1 . F τ is the set of flags, or “half–edges” of τ . Each edge e consists of a pair offlags denoted ∂e , and each vertex v determines the set of flags incident to it denoted F τ ( v ). Finally, χ ( τ ) is the Euler characteristic of τ . The passage of the left hand side of (1.8) to the right hand side is by definitionthe result of term–wise integration of the formal series which can be obtained fromthe Taylor series of the exponent in the integrand. Concretely Z P e λ − S ( ϕ ) D ( ϕ ) = Z P e λ − S ( ϕ ) ∞ X N =1 λ − N S ( ϕ ) N N ! ! Y a dϕ a := Z P e λ − S ( ϕ ) Y a dϕ a + ∞ X N =1 λ − N N ! ∞ X k ,...,k N =1 k ! . . . k N ! X a ( i ) j ∈ A, ≤ j ≤ k i N Y i =1 C a ( i )1 ,...,a ( i ) ki Z P e λ − S ( ϕ ) N Y i,j ϕ a ( i ) j Y a dϕ a . (1 . C a ,...,a k , weightof C a ,...,a k being k . In fact, the Gaussian integrals in the coefficients uniformlyconverge, and one can use the so called Wick’s lemma.The last remark is that periods appearing in concrete models of quantum fieldtheories are weights (1.9) , in which the summation over maps u : F τ → A is replacedby the integration over some continuous variables such as positions/momenta/coloursof particles moving along the edges of the respective Feymann graph: cf. [W13],[M¨uWZa12] and references therein.
2. Arithmetic differential equations2.1. Analogies between p –adic numbers and formal series. Combin-ing the lessons of previous examples we suggest now that in order to see “ p –adicproperties” of numerical periods, transcendental numbers important for physics,one could try to design a theory of “derivations in p –adic direction” and interpretnumerical periods as solutions of differential equations in the p –adic direction.Below we present basics of such a theory due to A. Buium. We start with thefollowing table of analogies. On the formal series side, we consider rings of the form k [[ t ]] where k is a field of characteristics zero. On the p –adic side, we consider themaximal unramified extension R of Z p . P OW ER SERIES p − ADICS P a i t i ∈ k [[ t ]] =: L P ε i p i ∈ R := Z unp Field of constants: a i ∈ k Monoid: ε i ∈ µ ∞ ∪ { } (Teichm¨uller representatives) Derivation: d/dt δ p ( ∗ ) := Φ( ∗ ) −∗ p p (Φ := lif t of F robenius )Polynomial Diff. Operators (PDO): p -adic PDO: D ∈ L [ T , T , . . . , T n ] D p ∈ R [ T , T , . . . , T n ] ( p –adic completion!) —————————————————————————-Action of PDO: f D ( f, f ′ , . . . f ( n ) ) or D p ( f, δ p f, . . . , δ np f )—————————————————————————-The Frobenius lift Φ : R → R involved in the definition of the p –adic derivative δ p is given explicitly as Φ( P ε i p i ) := P ε pi p i . Here we give a sample of interesting p –adicdifferential operators. p –adic logarithmic derivative. It is an analog of the map G m ( L ) → G a ( L ) : f f ′ /f (2 . x ∈ G m ( L ) is represented by the value f ∈ L ∗ at x of a fixed algebraiccharacter t of G m such that G m = Spec [ t, t − ]. Similarly, its p –adic version is thedifferential character G m ( R ) → G a ( R ) : a δ p a · a − p − p δ p a · a − p ) + p δ p a · a − p ) − . . . (2 . Example 2:
Quadratic reciprocity symbol: (cid:18) ap (cid:19) = a p − ∞ X k =1 ( − k − (2 k − k − ( k − k ! ( δ p a ) k a − pk ! . Example 3: a p –adic analog of the differential character µ of the group ofsections of a generic elliptic curve: µ ( P ) = (4 t (1 − t ) d dt + 4(1 − t ) ddt − Z P ∞ dXY as a non–linear p –adic DO acting upon coordinates of P .Such analogs were constructed in [Bu95] also for abelian varieties of arbitrarydimension and called δ p –differential characters ψ ( P ). More precisely, let E be anelliptic curve over R . Then there exist a differential additive map ψ : E ( R ) → R + of order 2 (as in the geometric case) or 1 (as for G m ).A character of order 2 exists if E has a good reduction and is not the canonicallift of its reduction in the sense of Serre–Tate: cf. additional discussion in 4.4 below.A character of order 1 exists if either E has good ordinary reduction and is thecanonical lift, or E has a bad multiplicative reduction.Using these multiplicative characters, A. Buium and the author constructed in[BuMa13] “Painlev´e VI equations with p –adic time.” p –derivations. In the commutative algebra,given a ring A and an A –module N , a derivation of A with values in N is anyadditive map ∂ : A → N such that ∂ ( ab ) = b∂a + a∂b . Equivalently, the map A → A × N : a ( a, ∂a ) is a ring homomorphism, where A × N is endowedwith the structure of commutative ring with componentwise addition, inheritingmultiplication from A on A × { } and having { } × N as an ideal of square zero.Similarly, in arithmetic geometry Buium defines a p –derivation of A with valuesin an A –algebra B , f : A → B , as a map δ p : A → B such that the map A → B × B : a ( f ( a ) , δ p ( a )) is a ring homomorphism A → W ( B ) where W ( B )is the ring of p –typical Witt vectors of length 2. Here Witt vectors of the form(0 , b ) form the ideal of square zero only if pB = { } . Making this definition explicit, we get δ p (1) = 0, and the following versions ofadditivety and Leibniz’s formula: δ p ( x + y ) = δ p ( x ) + δ p ( y ) + C p ( x, y ) , (2 . δ p ( xy ) = f ( x ) p · δ p ( y ) + f ( y ) p · δ p ( x ) + p · δ p ( x ) · δ p ( y ) , (2 . C p ( X, Y ) := X p + Y p − ( X + Y ) p p ∈ Z [ X, Y ] . (2 . p –derivation δ p : A → B the respective map φ p : A → B defined by φ p ( a ) := f ( a ) p + pδ p ( a ) is a ring homomorphism satisfying φ p ( x ) ≡ f ( x ) p mod p , that is “a lift of the Frobenius map applied to f ”.Conversely, having such a lift of Frobenius, we can uniquely reconstruct therespective derivation δ p under the condition that B has no p –torsion: δ p ( a ) := φ p ( a ) − f ( a ) p p generalising the definition given in 2.1 for A = B = R and identical morphism.Working with p –derivations A → A with respect to the identity map A → A and keeping p fixed, we may call ( A, δ ) a δ –ring. Morphisms of δ –rings are algebramorphisms compatible with their p –derivations. p –jet spaces. Let A be an R –algebra. A prolongation sequence for A consists of a family of p –adically complete R –algebras A i , i ≥
0, where A = A b isthe p –adic completion of A , and of maps ϕ i , δ i : A i → A i +1 satisfying the followingconditions: a) ϕ i are ring homomorphisms, each δ i is a p –derivation with respect to ϕ i ,compatible with δ on R .b) δ i ◦ ϕ i − = ϕ i ◦ δ i − for all i ≥ . Prolongation sequences form a category with evident morphisms, ring homomor-phisms f i : A i → B i commuting with ϕ i and δ i , and in its subcategory with fixed A there exists an initial element, defined up to unique isomorphism (cf. [Bu05],Chapter 3). It can be called the universal prolongation sequence. In the geometric language, if X = Spec A , the formal spectrum of the i –th ring A i in the universal prolongation sequence is denoted J i ( X ) and called the i –th p –jetspace of X . Conversely, A i = O ( J i ( X )), the ring of global functions.The geometric morphisms (of formal schemes over Z ) corresponding to φ i aredenoted φ i : J i ( X ) → J ( X ) =: X b (formal p –adic completion of X ).This construction is compatible with localisation so that it can be applied to thenon–necessarily affine schemes: cf. [Bu05], Chapter 3.
3. An arithmetically global version of Buium’s calculusand lambda–rings3.1. Introduction. p –adic numbers were considered in sec. 2 above as analogsof formal functions/local germs of functions of one variable.In this section, we discuss the following question: does there exist a (more) globalversion of “arithmetic functions”, elements of a ring A , admitting p –adic derivations δ p with respect to several, eventually all primes p ?An obvious example is Z : δ p ( m ) = m − m p p . Generally, we need “lifts of Frobenii”: such ring endomorphisms Φ p : A → A that Φ( a ) ≡ a p mod p . Then we may put δ p ( a ) = Φ p ( a ) − a p p . A general framework for a coherent system of such lifts is given by the followingdefinition:
A system of psi–operations on a commutative unitary ring A is a family of ring endomorphisms ψ k : A → A , k ≥ , such that: ψ = id A , ψ k ψ r = ψ kr ,ψ p x ≡ x p mod pA f or all primes p. Another important structure is introduced by the following definition:
A system of lambda–operations on a commutative unitary ring A is a family of additive group endomorphisms λ k : A → A , k ≥ , such that λ ( x ) = 1 , λ = id A ,λ n ( x + y ) = X i + j = n λ i ( x ) λ j ( y ) . These structures are related in the following way: (a) If A has no additive torsion, then any system of psi–operations defines a unique system of lambda–operations satisfying the compatibilityrelations: ( − k +1 kλ k ( x ) = X i + j = k, j ≥ ( − j +1 λ i ( x ) ψ j ( x ) . (b) Generally, any system of lambda–operations defines a unique system of psi–operations satisfying the same compatibility relations. Briefly, such a ring, together with psi’s and lambda’s, is called a lambda–ring.
Let R = a commutative unitary ring.Denote by A = A R the Grothendieck K –group of the additive category, con-sisting of pairs ( P, ϕ ), where P is a projective R –module of finite type, ϕ : P → P an endomorphism. Denote by [( P, ϕ )] ∈ A the class of ( P, ϕ ) . The ring structure on A is induced by the tensor product: [( P, ϕ )][(
Q, ψ )] :=[( P ⊗ Q, ϕ ⊗ ψ ].The lambda–operations on A are defined by λ k [( P, ϕ )] := [(Λ k P, Λ n ϕ )]. W ( R ) . Again, let R = a commutativeunitary ring.Define the additive group of W ( R ) as the multiplicative group T R [[ T ]].The multiplication ∗ in W ( R ) is defined on elements (1 − at ) as (1 − aT ) ∗ (1 − bT ) := 1 − abT , and then extended to the whole W ( R ) by distributivity, continuityin the ( T )–adic topology, and functoriality in R . Similarly, lambda–operations in W ( R ) are defined by λ k (1 − aT ) := 0 for k ≥
4. Roots of unity as constants:geometries over “fields of characteristic 1”4.1. Early history.
In the paper [T57], J. Tits noticed that some basic nu-merical invariants related to the geometry of classical groups over finite fields F q have well–defined values for q = 1, and these values admit suggestive combinatorialinterpretations.For example, if q = p k , p a prime, k ≥
1, thencard P n − ( F q ) = card ( A n ( F q ) \ { } )card G m ( F q ) = q n − q − n ] q , card Gr ( n, j )( F q ) = card { P j ( F q ) ⊂ P n ( F q ) } =: (cid:18) nj (cid:19) q , and the q = 1 values of the right hand sides are cardinalities of the sets P n − ( F ):= a finite set P of cardinality n , Gr ( n, j )( F ) := the set of subsets of P of cardinality j .Tits suggested a program: make sense of algebraic geometry over “a field ofcharacteristic one” so that the “projective geometry” above becomes a special caseof the geometry of Chevalley groups and their homogeneous spaces.The first implementation of Tits’ program was achieved only in 2008 by A. Connesand C. Consani, cf. [CC11], after the foundational work by C. Soul´e [So04]. How-ever, they required F as a definition field.Earlier, in an unpublished manuscript [KaS], M. Kapranov and A. Smirnov in-troduced fields F n on their own right.They defined F n as the monoid { } ∪ µ n , where µ n is the set of roots of unityof order n . Moreover, they defined a a vector space over F n as a pointed set ( V, µ n free on V \ { } . The group GL ( V ), by definition, consists ofpermutations of V compatible with action of µ n . Kapranov and Smirnov definedthe determinant map det : GL ( V ) → µ n and proved a beautiful formula for thepower residue symbol. Namely, if q = p k ≡ n and µ n is embedded in F ∗ q , F q becomes a vectorspace over F n , and the power residue symbol (cid:18) a F q (cid:19) n := a q − n ∈ µ n is the determinant of the multiplication by a in F n –geometry.Cf. also [Sm92], [Sm94].As we noticed in sec. 2, constants with respect to Buium’s derivation δ p in R := Z unp are roots of unity (of degree prime to p ) completed by 0 . Therefore, in the context of the differential geometry “in the p –adic direction”an independent project of Algebraic Geometry “over roots of unity”, or “in charac-teristic 1”, or else “over fields F , F n , F ∞ ” acquires a new motivation. Moreover,it becomes enriched with new insights: whereas at the first stage schemes in charac-teristic 1 were constructed by glueing “spectra of commutative monoids”, now theycould be conceived as Z –schemes endowed with lambda–structure considered asdescent data: see [Bor11a,b], [Bor09]. Here is a brief survey of Borger’s philosophy,showing that his schemes form a natural habitat for p –adic differential geometriesas well. The category of affine F –schemes Af f can bedefined as the opposite category of rings endowed with lambda–structures, ( A, Λ A ),and compatible morphisms. The forgetful functor to the usual category of affineschemes Af f → Af f : ( A, Λ) A is interpreted as the functor of base extension ∗ 7→ ∗ ⊗ F Z .Thus, a lambda–structure on a ring A is a descent data on Spec A to F . In particular, W ( Z ) must be considered as (a completion of?) Z ⊗ F Z .More generally, using general topos theory, Borger globalizes this construction,constructing a natural algebraic geometry of λ –schemes, which should be thoughtof as a lifted algebraic geometry over F . Just as all of usual algebraic geometry is contained in the big ´etale topos of Z , λ –algebraic geometry is contained in a big topos, which should be thought of asthe big ´etale topos over F . There is a map of topoi from the big etale topos over Z to the one over F . Schemes of finite type over F (in this sense, as in most other approaches) arevery rigid, combinatorial objects. They are essentially quotients of toric varietiesby toric equivalence relations.Non–finite–type schemes over F are more interesting. The big de Rham–Wittcohomology of X “is” the de Rham cohomology of X “viewed as an F –scheme”.It should contain the full information of the motive of X and is probably a concreteuniversal Weil cohomology theory.The Weil restriction of scalars from Z to F exists and is an arithmetically globalversion of Buium’s p –jet space.In conclusion, we briefly mention some remaining challenges. –geometry. In [Ma95], I suggested thatthere should exist a category of F –motives visible through the q = 1 point countof F –schemes. Predictions about such a point count were justified in Soul´e’s ge-ometry, cf. [So04]. In particular the zetas of non–negative powers of the “Lefschetz(dual Tate) motive” L must be: Z ( L × n , s ) = s + n π . This provides a conjectural bridge between F –geometry and geometry of Spec Z at the archimedean infinity, that is, Arakelov geometry : a Γ–factor of classical zetas,e.g., Γ C ( s ) := [(2 π ) − s Γ( s )] − = Y n ≥ s + n π (regularized product) looks like F –zeta of the dualized inf–dim projective spaceover F .However, this phenomenon remains an isolated observation, and the archimedeanprime still remains “first among equals” breaking the democracy of all valuations. “under Spec Z ”. In the traditional algebraic geometry,the special role of Spec Z is related to the fact that it is the final object of thecategory of schemes. Since it is very far from being “a point–like object”, it seemednatural to imagine that Spec F , being “really point–like”, will replace it. However,the belief that in an extended algebraic geometry there should necessarily exist afinal object, is unfounded. Already in the simplest category of Deligne–Mumford stacks over a field k , admitting quotients with respect to the trivial action of anyfinite group G , there is no final object, because we have non–trivial morphismsSpec k → Spec k/G .This led several authors to the contemplation of more general geometries lying“under Spec Z ” but not necessarily at the bottom of the unfathomable abyss: cf.the To¨en–Vaqui´e project [TV05].For example, in the Borger–Buium’s framework we may consider schemes forwhich Frobenius lifts are given only for some subsets of primes, eventually one prime p , such as the Serre–Tate canonical liftings of Abelian varieties in characteristic p :cf. [Katz81].More precisely, for the simplest case of elliptic curves, denote by M the p –adiccompletion of the moduli stack of elliptic curves without supersingular locus. Onecan define Frobenius lift on this stack: it sends an elliptic curve to its quotient byits canonical subgroup. The latter is defined as the unique closed sub–groupschemewhose Cartier dual is the ´etale lift to Z p of the Cartier dual of the kernel of Frobeniuson the fiber modulo p . This endomorphism also lifts to a natural endomorphism ofthe universal elliptic curve. So James Borger suggests to say that M “descends tothe p –typical F ”, and the same can be said about the universal elliptic curve overit. The p –adic elliptic curves with Frobenius lift are called canonical liftings.Notice that if we replace the p –adic direction by the functional one, we wouldsimply speak about families of elliptic curves with constant absolute invariants. But p –adic absolute invariants of canonical liftings are by no means “constants” in thenaive sense, discussed in sec. 2, that is they are not Teichm¨uller representatives: cf.a recent paper by Finotti, ”Coordinates of the j p –adic differential geometry. Acknowledgements.
Collaboration with A. Buium on [BuMa13] helped me muchin conceiving this survey. J. Borger generously explained me some of his construc-tions and motivations. Igor Volovich stimulated the final writing by inviting meto give a talk at the International Workshop on p –adic methods for modelling ofcomplex systems, Bielefeld, April 15–19, 2013. I am grateful to them all. References [ACG13] A. Abdessalam, A. Chandra, G. Guadagni.
Rigorous quantum field the-ory functional integrals over the p –adics I: anomalous dimensions. arXiv:1302.5971[Be11] A. Beilinson. p –adic periods and derived De Rham cohomology. Journ.AMS, vol. 25, no. 3 (2012), 319–327. arXiv:1102.1294[Bor09] J. Borger.
Lambda–rings and the field with one element. arXiv:0906.3146[BorBu09] J. Borger, A. Buium.
Differential forms on arithmetic jet spaces.
Selecta Math. (N.S.) 17 (2011), no. 2, 301–335. arXiv:0908.2512[Bor11a] J. Borger.
The basic geometry of Witt vectors, I: The affine case.
Algebra Number Theory 5 (2011), no. 2, 231–285.[Bor11b] J. Borger. basic geometry of Witt vectors. II: Spaces.
Math. Ann. 351(2011), no. 4, 877–933.[Bu95] A. Buium.
Differential characters of Abelian varieties over p–adic fields.
Inv. Math., vol. 122 (1995), 309–340.[Bu05] A. Buium.
Arithmetic Differential Equations.
AMS Math Surveys andMonographs, vol. 118, 2005.[BuMa13] A. Buium, Y. Manin.
Arithmetic Differential Equations of Painlev´eVI Type. arXiv:1307.3841[CC11] A. Connes, C. Consani.
On the notion of geometry over F . J. AlgebraicGeom. 20 (2011), no. 3, 525–557.[CCMa08] A. Connes, C. Consani, M. Marcolli. Fun with F . J. Number Theory129 (2009), no. 6, 1532–1561. math.AG/0806.2401[De04] A. Deitmar.
Schemes over F . In: Number Fields and Function Fields –Two Parallel Worlds. Ed. by G. van der Geer, B. Moonen, R. Schoof. Progr. inMath, vol. 239, 2005. math.NT/0404185[Er13] A. Erdogan.
A universal formula for the j –invariant of the canonicallifting. arXiv:1211.1152[GoMa04] A. Goncharov, Yu. Manin. Multiple zeta–motives and moduli spaces M ,n . Compos. Math. 140:1 (2004), 1–14. math.AG/0204102[Fa88] G. Faltings. p –adic Hodge theory. J. Amer. Math. Soc., 1(1988), 255–288.[KaS] M. Kapranov, A. Smirnov.
Cohomology determinants and reciprocity laws:number field case.
Unpublished manuscript, 15 pp. [Katz81] Katz, N. Serre–Tate local moduli.
Algebraic surfaces (Orsay, 1976–78),pp. 138–202, Lecture Notes in Math., 868, Springer, Berlin-New York, 1981.[KoZa01] M. Kontsevich, D. Zagier.
Periods.
In: Mathematics unlimited—2001and beyond, 771–808, Springer, Berlin, 2001.[LeBr13] L. Le Bruyn.
Absolute geometry and the Habiro topology. arXiv:1304.6532[Ma89] Yu. Manin.
Reflections on arithmetical physics.
In: Conformal In-variance and string theory (Poiana Brasov, 1987), Academic Press, Boston, MA,1989, 293–303. Reprinted in “Mathematics as Metaphor”, Selected Essays byYu. I. Manin, AMS 2007, pp. 149–155.[Ma95] Yu. Manin.
Lectures on zeta functions and motives (according to Deningerand Kurokawa).
Ast´erisque 228:4 (1995), 121–163.[Ma08] Yu. Manin.
Cyclotomy and analytic geometry over F . In: Quanta ofMaths. Conference in honour of Alain Connes. Clay Math. Proceedings, vol. 11(2010), 385–408. Preprint math.AG/0809.2716.[Ma09] Yu. Manin.
Renormalization and computation I: motivation and back-ground.
In: Proceedings OPERADS 2009, eds. J. Loday and B. Vallette, S´eminaireset Congr`es 26, Soc. Math. de France, 2012, pp. 181–223. Preprint math.QA/0904.4921[M¨uWZa12] S. M¨uller–Stach, S. Weinzierl, R. Zayadeh.
Picard–Fuchs equationsfor Feynman integrals. arXiv:1212.4389[Sm92] A. L. Smirnov.
Hurwitz inequalities for number fields. (Russian).
Al-gebra i Analiz 4 (1992), no. 2, 186–209; translation in St. Petersburg Math. J. 4(1993), no. 2, 357–375.[Sm94] A. L. Smirnov.
Absolute determinants and Hilbert symbols.
PreprintMPI 94/72, Bonn, 1994.[So04] C. Soul´e.
Les vari´et´es sur le corps `a un ´el´ement.
Mosc. Math. J. 4:1(2004), 217–244.[Ti57] J. Tits.
Sur les analogues alg´ebriques des groupes semi–simples com-plexes.
Colloque d’alg`ebre sup´erieure, Centre Belge de Recherches Math´ematiques,´Etablissement Ceuterick, Louvain, 1957, 261–289.[TV05] B. To¨en, M. Vaqui´e.
Au–dessous de
Spec Z . J. K-Theory 3 (2009), no.3, 437–500. math.AG/0509684[VlVoZe94] V. S. Vladimirov, I. V. Volovich, E. I. Zelenov. p–adic analysisand mathematical physics.