The universal thickening of real & complex numbers: towards an archimedean analogue of Fontaine’s theory Caterina Consani – Johns Hopkins U. (project in collaboration with Alain Connes) Texas A&M – May 2, 2014 MAIN GOAL: Define an ∞-adic theory of periods in direct analogy with p-adic Hodge theory Aut(R) = {1}, Theorem 1 AutR(C) = Z/2Z There exist universal field covers: θ W (R♭ ) R, θ W (C♭ ) C with larger groups of automorphisms ({Fλ}λ∈R× ∈ Aut(W (R♭))) + Theorem 2 The infinite dimensional real, resp. complex vector spaces Ω := Ker(θ)/Ker(θ2) contain the independent set of periods: {πp := [p] − p | p ∈ Z prime} resp. the independent complex periods: {πp := [e(log p)] − p | p ∈ Z prime}, ω := ([e(2πi)] − 1)/([e(πi)] − 1) I. Review of p-adic Hodge theory L = number field, X/L = proper, smooth alg. variety i (X ) ≃ C ⊗ H i (X, Z) C ⊗L HdR Z B /L Goal of the theory of p-adic periods: K = finite extension of Qp, GK := Gal(K̄/K) X/K = proper, smooth alg. variety construct analogous isomorphisms: (HT/dR conjectures) i i (X (??) ⊗K HdR /K ) ≃ (??) ⊗Qp Hét (XK̄ , Qp ) Fontaine’s strategy: Introduce rings of periods B (Qp-algebras) with an action of GK and additional structures so that: DB(Vp) := (B ⊗Qp Vp)GK BGK − module becomes an invariant of any “good” (B-admissible) GK -module Vp: B ⊗BGK DB(Vp) ≃ B ⊗Qp Vp B = BdR (filtered) field of p-adic periods, BGK = K i (X , Q ) X/K = smooth, proper variety: Vp := Hét p K̄ i (X , Q )) ≃ H i (X DdR(Hét p /K ) K̄ dR (Faltings) (filtered) K-isomorphism i i (X BdR ⊗K HdR /K ) ≃ BdR ⊗Qp Hét(XK̄ , Qp) B+ dR Cp : the universal Galois-equivariant cover 3 fundamental steps (K = Qp): (1) Universal perfection C♭p of the p-perfect field Cp := (Q̄p)∧ (2) p-isotypical Witt construction: C♭p ⊃ OC♭ p W (OC♭ ) p 1] )[ (3) B+ is a suitable completion of W (O ♭ dR p C p C♭p := {x := (x(n))n≥0| x(n) ∈ Cp; (x(n+1))p = x(n)} m (n+m) (n+m) p (n) (x + y) := lim (x +y ) , m→∞ (xy)(n) := x(n)y (n) perfect, complete field of characteristic p ∼ OC♭ := lim ←−pOCp −→ lim ←−pOCp /pOCp , p x7→x v7→v x 7→ x̄ := (x(n)mod. p)n≥0 perfect, associative, commutative Fp-algebra ∑ 1 b,+ B := W (OC♭ )[ p ] = { [xk ]pk ∈ W (C♭p) ⊗ Qp| xk ∈ OC♭ } p p k≫−∞ Zp-algebra (not complete dvr) θ : W (OC♭ ) OCp p lift of Fix: ∑ k≥0 θ0 : OC♭p OCp /(p), ε = (ε(n) ∈ µpn (K̄))n≥0 ∈ Zp(1) ω := ([ε] − 1)/([ε1] − 1), !! [xk ]pk → 7 ker(θ) = (ω) = ([p̃] − p) ∑ (0) xk pk k≥0 x 7→ x(0) [ε(0) = 1, ε(1) ̸= 1, (ε(n+1))p = ε(n)] [ε1] := [(ε(1), . . .)], p̃(0) := p, p ε1 = ε p̃ := (p̃(n))n≥0 ∞ ∑ + b,+ /(ker(θ))n) = {f = n|x ∈ BdR := lim ( B x ω n n ←− n=0 n∈N Qp-algebra; complete dvr; with residue field Cp Bb,+} θ(1 − [ε]) = 0 =⇒ 1 − [ε] is “small” for the topology of B+ dR t := log([ε]) = ∞ ∑ ([ε] − 1) n−1 (−1) n n=1 Zp(1) ↩→ B+ dR , g ∈ Gal(Q̄p/Qp) : n ε 7→ t ∈ B+ dR GQp -equivariant g(t) = g(log([ε])) = log([εχ(g) ]) = χ(g)t t is a period of the cyclotomic character GQ BdR := B+ dR [1/t]; F iliBdR := tiB+ dR ; Caveat: φ( ∑ grBdR ≃ ⊕ BdR p = Qp Cp(i) =: BHT i∈Z (GQp -equivariant) NO natural Frobenius structure on B+ dR (or on BdR) k k [xk ]p ) = ∑ p k k [xk ]p on Bb,+ BUT: φ([p̃] − p) = [p̃p] − p ∈ / ker(θ) II. Fix: R κ∈Q R♭ Perfection to characteristic one |κ|∞ < 1, |κ|2 = 1 R♭ := {x = (xn)n≥0 | xn ∈ R; (xn+1)κ = xn} (∗) m κ (x + y)n := lim (xn+m + yn+m) , (xy)n := xnyn m→∞ THEOREM (1) ∼ R♭ −→ R, x 7→ x0 bijection of multiplicative monoids. (2) −(n+m) −(n+m) κm κ κ ′ (x + y)n = lim (x0 + y0 ) m→∞ on R ∼ R♭ is not associative. (3) The addition ⌣ defined by taking the limit of the graphs Gκ,m in (∗), is multivalued and associative Gκ= 1 ,m=1,n=0 1 (x, y) := (x, y, (x3 + y 3) 3 ) 3 At m = 1 (cf. above): Gκ= 1 ,m(x, y) is a univalent map 3 lim G m→∞ κ= 1 3 ,m (x, y) fails to be univalent on the line y = −x: At m ≫ 0 (cf.above): Gκ= 1 ,m 3 1 m m (x, y) := (x, y, (x3 +y 3 ) 3m ) x +′ y can be expressed in terms of x0 ∈ R ∼ R♭: −(n+m) x +′ y := ((x +′ y)n)n≥0 = ( lim (xκ 0 m→∞ x +′ y = x, if |x| > |y| or x = y; y, if |x| < |y| or x = y; 0, if y = −x In particular: However: + y0κ x +′ x = x −(n+m) m )κ )n≥0 |x| := |x0| ∀x ∈ R♭ the associativity wrt +′ fails (y +′ x) +′ (−x) = x +′ (−x) = 0 ̸= y +′ (x +′ (−x)) = y +′ 0 = y ∀y ∈ R♭, |y| < |x| or y = x Hyperfield structure on R♭ ∼ R x, if |x| > |y| or x = y; x ⌣ y := y, if |x| < |y| or x = y; [-x,x], if y = −x .x . . y .0 . .x .y .0 .x . .y .0 R♭ = (R, ⌣, ·) x, y ∈ R .x ⌣ y . . .0 .x ⌣ y . .0 y ..x ⌣ . . .0 (Viro) real tropical hyperfield What is a hyperfield (F, +, ·) ? (M. Krasner 1957-58) Sum: + : F × F → P(F )∗, (F, +) = canonical hypergroup − x + y = y + x, ∀ x, y ∈ F ; ∃! 0, 0+x=x=x+0 − (x + y) + z = {∪(α + z) | α ∈ x + y} = x + (y + z) − For x ∈ F, ∃! y ∈ F : − (reversibility): Product: 0∈x+y (y := −x) x ∈ y + z =⇒ z ∈ x − y · : F × F → F, (F ∗ = F \ {0}, ·) = group, 1F ̸= 0F • the distributivity laws hold Examples of hyperfields 1) K = {0, 1}, 1 + 1 = {0, 1} 2) S = {−1, 0, 1}, 1 + 1 = 1, − 1 − 1 = −1 1 − 1 = {−1, 0, 1}, sign : Z −→ S; hyperfield of signs | · | : S −→ K; Spec (S) Krasner hyperfield F1 := (K, ·), q qqq q q q ∃!qqqqq q qq qqq q q x q q Spec (R♭ ) o o B := S+ = {0, 1} real tropical hypergeometry Spec (Z) Spec (B) qq qqq q q qq qqq q q q qqq q q x q q Spec (F1 ) o Spec (Rmax ) o tropical geometry Why NCG connects with hyperring theory (R, +, ·)? - (R, +) canonical hypergroup - (R, ·) multiplicative monoid with 1 ̸= 0 - distributivity laws hold. Fact: R = commutative ring, G ⊂ R× subgroup. The set R/G with: x̄ + ȳ := (xG + yG)/G x̄ · ȳ := xyG ∀x̄, ȳ ∈ R/G (coset) is a hyperring Theorem R = commutative ring, {1} ̸= G ⊂ R× subgroup (Krasner) {0, 1} = K ⊂ R/G ⇐⇒ (G ∪ {0}) ⊂ R Main example: subfield K ⊂ AQ/Q× adèle class space as K-hyperalgebra The hyperfield R♭: main properties (1) R♭ has characteristic one (i.e. x + x = x, ∀x ∈ R♭) ∼ ♭ ), [φ(λ)](x) := sign(x)|x|λ group isom (2) φ : R× −→ Aut( R + if λ ∈ Q× + , |λ|2 = 1: (3) R♭ is perfect [φ(λ)](x) = xλ ∼ (i.e. R♭ −→ R♭, x 7→ xn, ∀n = 2Z + 1) (4) O := [−1, 1] ⊂ R♭ maximal compact sub-hyperring (5) R♭ is complete for d(x, y) := |x0 − y0| on R♭ ∼ R♭ → R, x 7→ x0 isometry for d III. Witt construction for hyperfields H = hyperfield Definition (1) {W (H), ρ, τ } = Witt-model (W-model) of H: W (H) = field, ρ : W (H) −→ H hyperfield homomorphism τ : H −→ W (H) hom of multiplicative monoids: ρ ◦ τ = idH α (2) {W (H)1, ρ1, τ1} −→ {W (H)2, ρ2, τ2} morphism of W -models: α : W (H)1 → W (H)2 field homom: τ2 = α ◦ τ1, ρ1 = ρ2 ◦ α (3) {W (H)u, ρu, τu} universal W -model of H if ∀{W (H), ρ, τ } ∃! α u {W (H)u, ρu, τu} −→ {W (H), ρ, τ } (morphism of W -models) Universal W -models are unique up-to unique isomorphism The universal W -model of R♭ THEOREM (1) W (R♭) := {Frac(Q[R× +]), ρu , τu = [·]} universal W -model × −→ Q [ R R× + ], + x 7→ [x] ∈ Q[R× +] [·] τu : R♭ ∼ R −→ W (R♭), ρu ♭ W (R ) −→ R♭ ∼ R, x0 := sup{xi}, (canonical generator) [0] = 0, τu(−x) = −[x] ( ) ∑ a0 x0 ai [xi ] i ∑ ρu( b [y ] ) := sign j j j b0 y0 y0 := sup{yj }, ai, bj ∈ Q (2) ρu induces an isomorphism of hyperfields: ∼ W (R♭)/G −→ R♭, × × → R♭ ) G := Ker(ρu : Frac(Q[R× ]) + From: ♭ ), R× −→ Aut( R + ∼ φ λ 7→ [φ(λ)](x) = sign(x)|x|λ It follows (for each fixed λ): −1 ◦ ρ , τ ◦ φ(λ)} ]), φ(λ) {Frac(Q[R× u u + ρu ]) −→ Frac(Q[R× + φ(λ)−1 ♭ R −→ (universality ⇒) for each fixed λ: τu ◦ φ(λ) = Fλ ◦ τu; Corollary R♭ ; ∃! is also a W -model of R♭ φ(λ) u R♭ −→ R♭ −→ Frac(Q[R× + ]) τ Fλ × Frac(Q[R+]) −→ Frac(Q[R× +]) ∼ ρu = φ(λ)−1 ◦ ρu ◦ Fλ Aut(R♭) −→ Aut(W (R♭)), φ(λ) 7→ Fλ {Fλ}λ∈R× ∈ Aut(W (R♭)) 1-parameter group of Frobenius automs + The universal W -model W (H)u of a hyperfield H (when exists) is the field of quotients of a natural ring: H = hyperfield, {W (H)u, ρu, τu} = universal W -model WZ(H) ⊂ W (H)u integral subring generated by the τu(x)’s (resp. WQ(H) ⊂ W (H)u Then: sub Q-algebra) W (H)u ≃ Frac(WZ(H)) = Frac(WQ(H)) Example 1 S = {0, ±1} hyperfield of signs: WZ(S) ≃ Z, W (S) = Q Example 2 ♭) WQ(R♭) = Q[R× ] ⊂ W ( R + sub Q-algebra !! NO universal W -model for K = {0, 1} Krasner hyperfield IV. Universal pro-infinitesimal thickening of R There is an archimedean counterpart of the ring homomorphism of p-adic Hodge theory: θ 1 b,+ B := W (OC♭ )[ p ] p Cp, ∑ (0) k [xk ]p ) = xk pk θ( k≫−∞ k≫−∞ ∃! Q-algebra homomorphism θ( ∑ i ai[xi]) = ∑ ai(x0)i, ∑ ♭ ) −→ R Q[R× ] =: W ( R Q + θ([x]) = θ(τ (x)) = x0, θ ∀x = (xn)n≥0 ∈ R♭ i DEFINITION The universal pro-infinitesimal cover of R: ♭)/Ker(θ)n) R∞ := lim (W ( R Q ←−n b,+ /(ker(θ))n ) ( B is the archimedean counterpart of B+ dR := lim ←−n THEOREM (1) ℓ : R× +→R group homomorphism Tℓ : WQ(R♭) −→ C ∞(R) Tℓ(X)(z) := ∑ aielog(xi)+(z−1)ℓ(xi) i Tℓ(X)(1)≡θ(X) = (2) ∑ X := ring homomorphism ∑ ai[xi] ∈ WQ(R♭) = Q[R× +] i i ai xi θ WQ(R♭) −→ R The Taylor expansion at z = 1 induces the ring hom T ℓ ♭ )/Ker(θ)n −→ R∞ := lim W ( R R[[z − 1]] Q ←−n ( ) d (T (1 − [x] − [1 − x])(z) = −xℓ(x) − (1 − x)ℓ(1 − x) dz ℓ z=1 (3) {πp := [p] − p | p ∈ Z prime} ⊂ Ker(θ)/Ker(θ)2 linearly independent “periods” over R THEOREM (Ker(θ)/Ker(θ)2) is the infinite dimensional R-linear space generated by the symbols ε(x), x ∈ R, with relations: (A) ε(1 − x) = ε(x) (B) x ) + y ε(− x ) , ε(x + y) = ε(y) + (1 − y) ε( 1−y y (C) x ε(1/x) = −ε(x) , ∀y ∈ / {0, 1} ∀x ̸= 0. The only Lebesgue measurable group homomorphism ℓ : R× +→R yields the linear form on Ker(θ)/Ker(θ)2: s(x) := 1 − [x] − [1 − x] 7→ −x log(x) − (1 − x) log(1 − x) =: S(x) given by the entropy function S(x) V. Perfection of C C♭ := {x = (xn)n≥0, convergent, xn ∈ C | (xn+1)3 = xn, ∀n} (C ⨿ ∼ n C) ∪ {0} −→ C♭ ϵ e(z) 7→ (xn)n≥0 xn := ϵ ez/3 ; ϵ ∈ {±1} (ϵ = limn→∞ xn, z = limn→∞ 3n(ϵxn − 1)) Product : Sum : ϵe(z) · ϵ′e(z ′) := ϵϵ′e(z + z ′) ϵ e(z), ϵ′ e(z ′), ϵ e(z) + ϵ′ e(z ′) := ϵ e(z), {0} ∪ {ϵ′′e(z ′′) | z ′′ ≤ C+ z}, C+ := {z ∈ C|ℜe(z) ≥ 0; ℑ(z) ≥ 0 if ℜ(z) = 0} if if if if z ′ < C+ z z < C+ z ′ z = z ′, ϵ = ϵ′ z = z ′, ϵ = −ϵ′ (total) cone z ≤C+ z ′ ⇐⇒ z ′ − z ∈ C+ total order on C C♭ is a perfect hyperfield of characteristic one Embedding of hyperfields: R♭ ↩→ C♭ ∀α ∈ (− π2 , π2 ], ∃ injective homom of hyperfields Proposition ι α R♭ −→ C♭ , ( B := {g = ) φ B −→ Aut(C♭), φ −→ Aut(C♭) ⊂ ⊂ z∈C+ a 0 ∈ GL2(R)|a, d > 0} c d FACT: B x 7→ sign(x)e(log |x| |{z} eiα ) ∼ ♭) R× −→ Aut( R + φ ( isotropy group of C+ [φ(g)](ϵe(z)) = ϵe(g · z) λ 0 gλ = 0 1 ↑ λ ) → φgλ (ι0(x)) ↑= → [φ(λ)](x) := sign(x)|x|λ VI. The universal W -model of C♭ W (C♭) := {Frac(Q[C]), ρC+ , [·]} THEOREM ρC + Q[C] −→ C♭, [·] ♭ C −→ ρC + ( Q[C], universal W -model n ∑ aj ϵj u(zj )) := ϵ0e(z0) |{z} j=1 ∈Q× + [ϵ e(z)] := ϵ u(z) z0 := ∨ C+ (u(z) ∈ Q[C] canonical generator) ρC+ ◦ [·] = id ρC+ induces an isomorphism of hyperfields: ∼ W (C♭)/G −→ C♭, G := Ker(ρC+ : W (C♭)× −→ (C♭)×) ∃! Q-algebra homomorphism θ( ∑ i ai[zi]) := ∑ i {zj } θ WQ(C♭) := Q[C] −→ C aiev(zi) := ∑ i aiezi The universal formal pro-infinitesimal thickening of C is the Ker(θ)-adic completion of WQ(C♭): ♭ )/Ker(θ)n ) C∞ := lim (W ( C Q ←−n ℓ ∈ V := HomZ(C, C), X= ∑ aiu(zi) ∈ WQ(C♭) i T : WQ(C♭) → F(V, C), T (X)(ℓ) := ∑ ℓ(zi ) , i ai e θ(X) = T (X)(id) THEOREM The Taylor expansion at (id) ∈ V1 := HomR(C, C) ⊂ V (−1,0) of T (X)(L), for L = id + a(id) + b(id) ∈ V1 = V1 induces a surjective ring homomorphism: ρ C∞ −→ C[[a, b]], a, b = variables (0,−1) ⊕ V1 The “periods”: {πp := [e(log p)] − p ∈ Ker(θ)/Ker(θ)2, (p prime)} are linearly independent over C. There is another “period” which is purely complex: it arises as the analogue of ε ∈ C♭p: (ε(n) ∈ µpn (K̄))n≥0 generator of Zp(1) ε := e(2πi) ∈ C♭ θ([ε]) = e2πi = 1, ω := ([e(2πi)] − 1)/([e(πi)] − 1) ∈ Ker(θ) ε1/2 = e(πi) t := log([ε]) = − ∑ (1 − [ε])n n≥1 n ρ(t) = 2πi(a − b), ∈ C∞, u(p) := log( ρ(u(p)) = log(p)(a + b) [p] ) ∈ C∞ p R∞ θ WZ(S)O = Z / Q = W (S ) / C∞ θ / l i i O l i l i i l l i i l i l i i i l l /F tti i i ul u l / 1 / R ? F12 = {0, ±1} ϵ C ρ τ/F C♭ ev l R♭ , 1 b b b b b1 b b b b b b b b b b b b b b b b b b b b b b b b b b b b b b S = {0, ±1} ∼ B+ dR θ W (Fp) = Zp l / Qp C♭p m mm m m mm vvm m O ˆ= C ( Q̄ ) p p ρ / τ/F 1 , ? Fp