OctoInform Beta

OpenAI preprint, September 2026 · §7. Proof of the main theorem · §7

§7 (proof of Theorem 1.1): assembly: For k≥3k \ge 3 continuous functions on a compact metrizable mixing system, one a centered function of a zero-entropy observable, along a family of time tuples

Zero-entropy correlations have no positive lower bound.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Rokhlin asked whether every mixing transformation is mixing of all orders. Let XX be a standard Borel space, a nonempty measurable space isomorphic to a Polish space with its Borel sets.✦ A measurable automorphism TT of XX preserving a probability measure μ\mu is mixing when, for all measurable sets AA and BB, μ(A∩T−nB)→μ(A)μ(B)(n→∞).\mu\bigl(A\cap T^{-n}B\bigr)\longrightarrow\mu(A)\,\mu(B)\qquad(n\to\infty).(1) For a finite index set II of k=|I|k=|I| slots, functions fjf_j on XX bounded by 11, one for each j∈Ij\in I, and a tuple d=(d(j))j∈Id=(d(j))_{j\in I} of times, the correlation along dd is ∫X∏j∈Ifj(Td(j)x)dμ(x)\int_X\prod_{j\in I}f_j(T^{d(j)}x)\,d\mu(x). The main theorem concerns these correlations for k≥3k\ge3 when, for some index i∈Ii\in I, the factor fif_i is a centered function of a finite-valued observable of word-entropy rate zero, notions defined next.

A finite-valued observable is a measurable map α\alpha from XX to a finite set 𝒜\mathcal{A}. For a sequence F=(Fm)m≥0F=(F_m)_{m\ge0} of maps from XX to 𝒜\mathcal{A}, the word map of length nn sends xx to the word F(n)(x)=(F0(x),…,Fn−1(x))F^{(n)}(x)=(F_0(x),\dots,F_{n-1}(x)) in 𝒜n\mathcal{A}^n✦, and the word entropy Hn(F)H_n(F) of FF at length nn is the Shannon entropy H(p)=−∑wp(w)log⁡p(w)H(p)=-\sum_w p(w)\log p(w), with 0log⁡0=00\log 0=0, of the law p(w)=μ{x:F(n)(x)=w}p(w)=\mu\{x:F^{(n)}(x)=w\} of that word.✦ The observable α\alpha has word-entropy rate zero when 1nHn((α∘Tm)m≥0)→0(n→∞),\frac{1}{n}\,H_n\bigl((\alpha\circ T^m)_{m\ge0}\bigr)\longrightarrow 0\qquad(n\to\infty),(2) and for a real function vv on 𝒜\mathcal{A} the function v∘αv\circ\alpha is centered when ∫Xv∘αdμ=0\int_X v\circ\alpha\,d\mu=0. The statements below assemble the main theorem; the sections before them set out the arrays, array objects and Hilbert codes they are stated over.

Arrays

Fix the finite index set II. An array of order nn over a space YY is a function yy from the index tuples v∈Inv\in I^n to YY, so that the arrays of order nn form the product YInY^{I^n}.✦ The only index tuple of length zero is the empty one, and the arrays of order zero form a space measurably isomorphic to YY, an array corresponding to its one entry; through this identification an array yy of order one is the II-tuple (yj)j∈I(y_j)_{j\in I} of its entries.✦ For an axis a∈{0,…,n}a\in\{0,\dots,n\} and an index i∈Ii\in I, the plane restriction πa,i:YIn+1→YIn\pi_{a,i}\colon Y^{I^{n+1}}\to Y^{I^n} fixes the axis aa at ii: (πa,iy)(v0,…,vn−1)=y(v0,…,va−1,i,va,…,vn−1).(\pi_{a,i}y)(v_0,\dots,v_{n-1})=y(v_0,\dots,v_{a-1},i,v_a,\dots,v_{n-1}).(3)✦ The tuple of planes of yy along a subset JJ of II is the family (πn,jy)j∈J(\pi_{n,j}y)_{j\in J} of its planes that fix the last axis.✦ A map f:Y→Zf\colon Y\to Z acts on arrays coordinatewise, the array fIn(y)f^{I^n}(y) being v↦f(y(v))v\mapsto f(y(v)).✦

Array Objects and Saturation

Let TT be a measurable automorphism of a standard Borel space XX. Initial array laws of TT over II are probability measures mnXm^X_n on XInX^{I^n}, one for each order n≥0n\ge0, each invariant under TInT^{I^n}, that are consistent and proper: for every order nn, every axis aa, every index ii and every subset JJ of II other than II itself, (πa,i)*mn+1X=mnX,((πn,j)j∈J)*mn+1X=⨂j∈JmnX.(\pi_{a,i})_*\,m^X_{n+1}=m^X_n,\qquad \bigl((\pi_{n,j})_{j\in J}\bigr)_*\,m^X_{n+1}=\bigotimes_{j\in J}m^X_n.(4)✦

An array object MM over initial array laws mXm^X of TT consists of a standard Borel space YY of states and a standard Borel space UU of parameters with measurable automorphisms TYT_Y and TUT_U, measurable maps x:Y→Xx\colon Y\to X and par:Y→U\mathrm{par}\colon Y\to U with x∘TY=T∘xx\circ T_Y=T\circ x and par∘TY=TU∘par\mathrm{par}\circ T_Y=T_U\circ\mathrm{par}, and probability measures mnm_n on YInY^{I^n}, its laws, with the properties below.✦ The parameter law of MM at order nn is the image (parIn)*mn(\mathrm{par}^{I^n})_*m_n, the law of the parameter array parIn(y)\mathrm{par}^{I^n}(y), and the own kernel κn\kappa_n of MM is the Markov kernel from UInU^{I^n} to YInY^{I^n} giving the conditional law under mnm_n of an array given its parameter array. Each mnm_n is invariant under TYInT_Y^{I^n} and lies over XX, in that (xIn)*mn=mnX(x^{I^n})_*m_n=m^X_n; the laws are consistent and proper as in (4); they are conditionally local, in that for every order nn, every subset JJ of II other than II and almost every parameter array pp of order n+1n+1, Lawmn+1((πn,jy)j∈J|parIn+1(y)=p)=⨂j∈Jκn(πn,jp);\mathrm{Law}_{m_{n+1}}\Bigl((\pi_{n,j}y)_{j\in J}\mathrel{\Big|}\mathrm{par}^{I^{n+1}}(y)=p\Bigr) =\bigotimes_{j\in J}\kappa_n(\pi_{n,j}p);(5) and xx and par\mathrm{par} are independent under the local law of MM, the measure m0m_0 read on YY.

An extension ee of an array object MM by an array object NN over the same initial laws, with states YMY_M, YNY_N, parameters UMU_M, UNU_N, base maps xMx_M, xNx_N, parameter maps parM\mathrm{par}_M, parN\mathrm{par}_N and laws mnMm^M_n, mnNm^N_n, is a pair of measurable maps, σ:YN→YM\sigma\colon Y_N\to Y_M on states and ρ:UN→UM\rho\colon U_N\to U_M on parameters, that intertwine the automorphisms, satisfy xM∘σ=xNx_M\circ\sigma=x_N and parM∘σ=ρ∘parN\mathrm{par}_M\circ\sigma=\rho\circ\mathrm{par}_N, and carry laws to laws, (σIn)*mnN=mnM(\sigma^{I^n})_*m^N_n=m^M_n for every nn.✦ The pullback along ee at order nn is the linear isometry of complex L2L^2 spaces e*:L2(mnM)→L2(mnN),e*F=F∘σIn.e^*\colon L^2(m^M_n)\to L^2(m^N_n),\qquad e^*F=F\circ\sigma^{I^n}.(6)✦

A data kind over II is a label naming one of 2k+12k+1 data fields: the parameter field, and for each j∈Ij\in I an outside field and an enlarged field.✦ The parameter field 𝒫n+1\mathcal{P}_{n+1}, the outside field 𝒪n+1,j\mathcal{O}_{n+1,j} and the enlarged field 𝒢n+1,j\mathcal{G}_{n+1,j} are the σ\sigma-algebras on YIn+1Y^{I^{n+1}} that an array object MM comes with, for every n≥0n\ge0 and j∈Ij\in I. The data σ\sigma-algebra of MM of kind dd at order nn is the σ\sigma-algebra on YIn+1Y^{I^{n+1}}, among the parameter, outside and enlarged fields of MM, that dd names.✦ The projection onto data of MM of kind dd at order nn, written Πn,dM\Pi^M_{n,d}, is the orthogonal projection of L2(mn+1)L^2(m_{n+1}) onto the functions measurable for the data σ\sigma-algebra of MM of kind dd at order nn.✦ An array object MM is saturated when its projections onto data commute with pullback along every extension: for every array object NN over the same initial laws, every extension ee of MM by NN, every order nn and every data kind dd, Πn,dN(e*F)=e*(Πn,dMF)(F∈L2(mn+1M)).\Pi^N_{n,d}(e^*F)=e^*\bigl(\Pi^M_{n,d}F\bigr)\qquad\bigl(F\in L^2(m^M_{n+1})\bigr).(7)✦

Hilbert Codes

Let η\eta be a finite measure on a standard Borel space YY. The test set algebra of YY is the algebra of sets generated by the standard generating sequence of its σ\sigma-algebra, a fixed sequence of measurable sets generating it.✦ A Hilbert test key on YY is a finite list of pairs ((a,b),A)((a,b),A) of two rational numbers and a set of the test set algebra✦; its Hilbert test function is the simple function ∑(a+bi)𝟏A\sum(a+bi)\mathbf{1}_A✦, and its Hilbert test vector is the class of that function in L2(η)L^2(\eta).✦ A fixed enumeration (tn)n≥0(t_n)_{n\ge0} runs through all test keys✦, and the Hilbert test sequence (hn(η))n≥0(h_n(\eta))_{n\ge0} of η\eta lists their test vectors in that order.✦

A real sequence zz is a distance code for a matrix (d(n,m))(d(n,m)) of distances when, for all indices nn and mm, zn≥0,zn≤d(n,m)+zm,d(n,m)≤zn+zm,infnzn=0.z_n\ge0,\qquad z_n\le d(n,m)+z_m,\qquad d(n,m)\le z_n+z_m,\qquad \inf_n z_n=0.(8)✦

Let κ\kappa be a kernel of finite measures from a space UU to YY. A Hilbert code for κ\kappa is a pair q=(u,z)q=(u,z) of a point uu of UU, its base point, and a distance code zz for the distances ∥hn(κ(u))−hm(κ(u))∥\|h_n(\kappa(u))-h_m(\kappa(u))\| between test vectors.✦ The decoded vector of a Hilbert code q=(u,z)q=(u,z) for κ\kappa is a vector ww of L2(κ(u))L^2(\kappa(u)) chosen with ∥w−hn(κ(u))∥=zn\|w-h_n(\kappa(u))\|=z_n for every nn, and the encoding of a vector ww of L2(κ(u))L^2(\kappa(u)) is the code (u,(∥w−hn(κ(u))∥)n≥0)\bigl(u,(\|w-h_n(\kappa(u))\|)_{n\ge0}\bigr).✦ The approximation index of qq at precision mm is the least nn with zn<2−mz_n<2^{-m}✦; the approximant of qq of level mm is the test function of the key tnt_n at that index✦, and the representative fqf_q of qq is the pointwise limit of its approximants as m→∞m\to\infty.✦ The circle representative ϕq:Y→𝕋\phi_q\colon Y\to\mathbb{T} sends yy to the point of the unit circle attached to fq(y)f_q(y), which is fq(y)f_q(y) itself when |fq(y)|=1|f_q(y)|=1✦; the circle carries the Borel σ\sigma-algebra it inherits from ℂ\mathbb{C}.✦

A vector ww of a complex inner product space is canonical in its ray with respect to a sequence (hn)(h_n) when the first hnh_n not orthogonal to ww has ⟨hn,w⟩>0\langle h_n,w\rangle>0.✦ A Hilbert code q=(u,z)q=(u,z) for κ\kappa is canonical when its decoded vector is canonical in its ray with respect to the test sequence of κ(u)\kappa(u)✦, and unimodular when |fq|=1|f_q|=1 almost everywhere for κ(u)\kappa(u).✦

Projective Circle Groups

The measurable circle subgroup ℳ(Y)\mathcal{M}(Y) is the group of measurable maps from YY to the unit circle 𝕋\mathbb{T} under pointwise multiplication.✦ The constant phase subgroup Φ(η)\Phi(\eta) of a measure η\eta on YY is the subgroup of the maps in ℳ(Y)\mathcal{M}(Y) that are η\eta-almost everywhere equal to a constant✦, and the projective circle group of η\eta is the quotient 𝒫𝒞(η)=ℳ(Y)/Φ(η)\mathcal{PC}(\eta)=\mathcal{M}(Y)/\Phi(\eta)(9) The projective circle class [g][g] of a measurable map g:Y→𝕋g\colon Y\to\mathbb{T} is its image in 𝒫𝒞(η)\mathcal{PC}(\eta)✦, and the projective class at uu of a Hilbert code qq for κ\kappa is the class [ϕq][\phi_q] in 𝒫𝒞(κ(u))\mathcal{PC}(\kappa(u)).✦ The projective representative rgr_g of an element gg of 𝒫𝒞(η)\mathcal{PC}(\eta) is a map chosen in the class gg, the constant 11 when g=1g=1✦, and the projective character of gg is rgr_g read as a complex function on YY.✦

For an abelian group LL, a family (χl)l∈L(\chi_l)_{l\in L} of complex functions on (Y,η)(Y,\eta) is a family of projective characters of LL when each χl\chi_l is η\eta-measurable with |χl|=1|\chi_l|=1 almost everywhere, χ0=1\chi_0=1 almost everywhere, ∫Yχldη=0\int_Y\chi_l\,d\eta=0 for every l≠0l\ne0, and for all ll and l′l' in LL there is a constant cc of modulus one with χlχl′=cχl+l′η-almost everywhere.\chi_l\,\chi_{l'}=c\,\chi_{l+l'}\qquad\eta\text{-almost everywhere}.(10)✦

Let SS be a set of pairs (a,q)(a,q) of a point of UU and a Hilbert code for κ\kappa. The local projective group GaG_a of SS at a point aa of UU is the subgroup of 𝒫𝒞(κ(a))\mathcal{PC}(\kappa(a)) generated by the projective classes at aa of the codes qq with (a,q)∈S(a,q)\in S.✦ The local group code of an element gg of GaG_a is a code qq chosen with (a,q)∈S(a,q)\in S and projective class gg when there is one, and the encoding of the zero vector otherwise.✦ For a map ψ\psi from the pairs (a,q)(a,q) of a point of UU and a Hilbert code for κ\kappa to a set 𝒞\mathcal{C}, the local group block of an element gg of GaG_a is the value of ψ\psi at the pair of aa and the local group code of gg.✦

Exact Codes and Fibre Sums

The absolute coefficient βf(u,q)\beta_f(u,q) of a complex function ff on YY at a pair (u,q)(u,q) of a point of UU and a Hilbert code for κ\kappa is the modulus of the correlation of ff with the circle representative ϕq\phi_q: βf(u,q)=|∫Yfϕq¯dκ(u)|.\beta_f(u,q)=\Bigl|\int_Y f\,\overline{\phi_q}\,d\kappa(u)\Bigr|.(11)✦

Let Λ\Lambda be a kernel from a space PP to YIY^I and let uu assign to each point pp of PP an II-tuple u(p)u(p) of points of UU. The exact code mean of a pair (p,𝒒)(p,\mathbf{q}) of a point pp of PP and an II-tuple 𝒒=(qi)i∈I\mathbf{q}=(q_i)_{i\in I} of Hilbert codes for κ\kappa is ∫∏i∈Ifqi(yi)dΛ(p)(y)\int\prod_{i\in I}f_{q_i}(y_i)\,d\Lambda(p)(y).✦ The exact code set for κ\kappa, Λ\Lambda and uu is the set of the pairs (p,𝒒)(p,\mathbf{q}) in which every qiq_i is canonical, unimodular and based at u(p)iu(p)_i, and whose exact code mean has modulus 11.✦ For a space QQ and maps row\mathrm{row} and col\mathrm{col} from Q×IQ\times I to PP, the exact square set for κ\kappa, Λ\Lambda, uu, row\mathrm{row} and col\mathrm{col} is the set of pairs (e,(qrs)r,s∈I)(e,(q_{rs})_{r,s\in I}) of a point ee of QQ and an I×II\times I matrix of Hilbert codes such that, for every rr, the pair (row(e,r),(qrs)s∈I)(\mathrm{row}(e,r),(q_{rs})_{s\in I}) lies in the exact code set and, for every ss, the pair (col(e,s),(qrs)r∈I)(\mathrm{col}(e,s),(q_{rs})_{r\in I}) lies in the exact code set.✦

Let ZZ be a space. A Borel fibre enumeration EE of a set S⊂U×ZS\subset U\times Z is a sequence of measurable sets Dn⊂UD_n\subset U and measurable maps ζn:U→Z\zeta_n\colon U\to Z such that (u,ζn(u))∈S(u,\zeta_n(u))\in S for u∈Dnu\in D_n, every point of SS is (u,ζn(u))(u,\zeta_n(u)) for some nn with u∈Dnu\in D_n, and u∈Dn∩Dmu\in D_n\cap D_m with ζn(u)=ζm(u)\zeta_n(u)=\zeta_m(u) forces n=mn=m.✦ The fibre sum ΣEF\Sigma_E F of a real function FF on U×ZU\times Z along EE is the function on UU whose value at a point aa of UU is the sum of FF over the points of SS above aa: (ΣEF)(a)=∑n:a∈DnF(a,ζn(a)).(\Sigma_E F)(a)=\sum_{n:\,a\in D_n}F\bigl(a,\zeta_n(a)\bigr).(12) For a measurable map ψ\psi from U×ZU\times Z to a set 𝒞\mathcal{C} and a value bb in 𝒞\mathcal{C}, the level restriction Eψ=bE^{\psi=b} is the Borel fibre enumeration of {s∈S:ψ(s)=b}\{s\in S:\psi(s)=b\} with domains {a∈Dn:ψ(a,ζn(a))=b}\{a\in D_n:\psi(a,\zeta_n(a))=b\} and the same maps ζn\zeta_n.✦

Raw Labels

For an array object MM, identify parameter arrays of order zero with points of UU, so that κ0\kappa_0 is a kernel from UU to YY. The raw label space of MM is the product of UU with the Hilbert codes for κ0\kappa_0.✦ The raw tuple of a state array yy of order one is the II-tuple (π0,jy)j∈I(\pi_{0,j}y)_{j\in I} of its entries✦, and the tuple parameter of a parameter array pp of order one is likewise (π0,jp)j∈I(\pi_{0,j}p)_{j\in I}.✦ The raw tuple kernel of MM, from UIU^I to YIY^I, is the image of κ1\kappa_1 under the map sending a state array of order one to its raw tuple.✦ For a kernel Λ\Lambda from UIU^I to YIY^I and an index ii, the raw line graph at ii is the set of pairs (p,z)(p,z) of a parameter array of order one and a code zz for κ0\kappa_0 with z=qiz=q_i for some tuple 𝒒\mathbf{q} such that (p,𝒒)(p,\mathbf{q}) lies in the exact code set for κ0\kappa_0, Λ\Lambda and the tuple parameter.✦ The raw exact square of MM for a kernel Λ\Lambda from UIU^I to YIY^I is the exact square set for κ0\kappa_0, Λ\Lambda and the tuple parameter, where QQ is the set of parameter arrays ee of order two and the rows and columns of ee are read on its planes π0,re\pi_{0,r}e and π1,se\pi_{1,s}e.✦

The Claim

Theorem 1. (a) Orbit cancellation. Let κ\kappa be a Markov kernel from a space UU to a measurable space YY with a countably generated σ\sigma-algebra; let EE be a Borel fibre enumeration of a set SS of pairs of a point of UU and a Hilbert code for κ\kappa, ψ\psi a measurable map from these pairs to a measurable space 𝒞\mathcal{C} whose points are measurable, and DD a bijection of 𝒞\mathcal{C}. Let a0a_0 and aa be points of UU and bb a value in 𝒞\mathcal{C}, and write βa′(g)\beta_{a'}(g) for the local group block of g∈Ga′g\in G_{a'}. Suppose that the block {g∈Ga0:βa0(g)=b}\{g\in G_{a_0}:\beta_{a_0}(g)=b\} is finite; that some group automorphism Γ\Gamma of Ga0G_{a_0} has βa0(Γg)=D−1βa0(g)\beta_{a_0}(\Gamma g)=D^{-1}\beta_{a_0}(g) and some group isomorphism J:Ga0→GaJ\colon G_{a_0}\to G_a has βa(Jg)=βa0(g)\beta_a(Jg)=\beta_{a_0}(g) for every g∈Ga0g\in G_{a_0}; that some bijection from {q:(a,q)∈S}\{q:(a,q)\in S\} onto GaG_a sends every code to its projective class at aa; and that the projective characters of the elements of GaG_a, indexed by GaG_a written additively, form a family of projective characters of GaG_a on (Y,κ(a))(Y,\kappa(a)). Let μ\mu be a probability measure on a measurable space XX, x:Y→Xx\colon Y\to X a measure-preserving map from κ(a)\kappa(a) to μ\mu, and (tm)m≥0(t_m)_{m\ge0} maps of XX preserving μ\mu with tm+m′=tm′∘tmt_{m+m'}=t_{m'}\circ t_m and μ(A∩tm−1B)→μ(A)μ(B)\mu(A\cap t_m^{-1}B)\to\mu(A)\,\mu(B) as m→∞m\to\infty for all measurable sets AA and BB. Let α\alpha be a measurable map from XX to a finite set 𝒜\mathcal{A} with 1nHn((α∘tm)m≥0)→0\frac{1}{n}H_n\bigl((\alpha\circ t_m)_{m\ge0}\bigr)\to0 under μ\mu, vv a complex function on 𝒜\mathcal{A} with ∫Xv∘αdμ=0\int_Xv\circ\alpha\,d\mu=0, RR a map of YY and ff a complex function on YY with f(Rny)=v(α(tn(x(y))))f(R^ny)=v\bigl(\alpha(t_n(x(y)))\bigr) for every n≥0n\ge0 and every y∈Yy\in Y. Then 1N∑n=0N−1(ΣEψ=D−nbβf∘Rn)(a)→0(N→∞).\frac{1}{N}\sum_{n=0}^{N-1}\bigl(\Sigma_{E^{\psi=D^{-n}b}}\,\beta_{f\circ R^n}\bigr)(a) \longrightarrow0\qquad(N\to\infty).(13)✦

(b) Bounded Cesàro averages. Let ν\nu be a probability measure on a space UU, C≥0C\ge0 and cc real numbers, and (sn)n≥0(s_n)_{n\ge0} a sequence of ν\nu-measurable real functions on UU with 0≤sn≤C0\le s_n\le C almost everywhere and ∫Usndν=c\int_U s_n\,d\nu=c for every nn, whose Cesàro averages 1N∑m<Nsm(u)\frac{1}{N}\sum_{m<N}s_m(u) tend to 00 for ν\nu-almost every uu. Then c=0.c=0.(14)✦

(c) Saturated objects. Suppose k=|I|≥3k=|I|\ge3. Let TT be a measurable automorphism of a standard Borel space XX that preserves a probability measure μ\mu and is mixing, α\alpha a finite-valued observable on XX of word-entropy rate zero, and vv a complex function on its values with ∫Xv∘αdμ=0\int_Xv\circ\alpha\,d\mu=0. Let MM be a saturated array object over initial array laws mXm^X of TT, with state space YY, base map xx and laws mnm_n, such that μ\mu is the image under xx of the local law of MM; fix i∈Ii\in I, and let (fj)j∈I(f_j)_{j\in I} be measurable complex functions on YY with |fj|≤1|f_j|\le1 for every jj and fi=v∘α∘xf_i=v\circ\alpha\circ x. Then the correlation of order one vanishes, read on the planes of the array or on its entries. The library proves it twice. Along the raw route it assumes in addition a Markov kernel Λ\Lambda from UIU^I to YIY^I equal to the raw tuple kernel of MM at almost every parameter array of order one for the parameter law of MM, under which, for every pp and every j∈Ij\in I, the entry yjy_j has law κ0(π0,jp)\kappa_0(\pi_{0,j}p) and the entries other than yjy_j are independent; a set SS of raw labels such that, for almost every parameter array pp of order one for the same law and every j∈Ij\in I, the codes zz with (p,z)(p,z) in the raw line graph at jj for Λ\Lambda are exactly those with (π0,jp,z)∈S(\pi_{0,j}p,z)\in S; Borel fibre enumerations of SS, of the exact code set for κ0\kappa_0, Λ\Lambda and the tuple parameter, and of the raw exact square of MM for Λ\Lambda; and two indices s≠ts\ne t of II. It then gives ∫YI∏j∈Ifj(π0,jy)dm1(y)=0,\int_{Y^I}\prod_{j\in I}f_j(\pi_{0,j}y)\,dm_1(y)=0,(15)✦ and without these additional data it gives ∫YI∏j∈Ifj(yj)dm1(y)=0.\int_{Y^I}\prod_{j\in I}f_j(y_j)\,dm_1(y)=0.(16)✦

(d) Initial laws. With k≥3k\ge3 and TT, μ\mu, α\alpha, vv, mXm^X and ii as in (c), suppose that μ\mu is the law m0Xm^X_0 read on XX, and let (fj)j∈I(f_j)_{j\in I} be measurable complex functions on XX with |fj|≤1|f_j|\le1 for every jj and fi=v∘αf_i=v\circ\alpha. Then ∫XI∏j∈Ifj(xj)dm1X(x)=0.\int_{X^I}\prod_{j\in I}f_j(x_j)\,dm^X_1(x)=0.(17)✦

(e) Compact systems. Let XX be a compact metrizable space, with its Borel sets a standard Borel space, and TT a homeomorphism of XX that preserves a Borel probability measure μ\mu and is mixing, suppose k=|I|≥3k=|I|\ge3, let α\alpha be as in (c) and vv a real function on its values with v∘αv\circ\alpha centered, fix i∈Ii\in I, let (fj)j∈I(f_j)_{j\in I} be continuous real functions on XX with |fj|≤1|f_j|\le1 for every jj and fi=v∘αf_i=v\circ\alpha, and let (db)b≥0(d_b)_{b\ge0} be a sequence of tuples db=(db(j))j∈Id_b=(d_b(j))_{j\in I} of natural numbers whose proper sublayouts mix: for every subset J≠IJ\ne I of II and all measurable sets AjA_j, j∈Jj\in J, μ(⋂j∈JT−db(j)Aj)→∏j∈Jμ(Aj)\mu\bigl(\bigcap_{j\in J}T^{-d_b(j)}A_j\bigr)\to\prod_{j\in J}\mu(A_j) as b→∞b\to\infty. Then for every δ>0\delta>0 there is an index bb with ∫X∏j∈Ifj(Tdb(j)x)dμ(x)<δ.\int_X\prod_{j\in I}f_j\bigl(T^{d_b(j)}x\bigr)\,d\mu(x)<\delta.(18)✦

In the paper, the result reads: in each typical fibre, orbit cancellation applied to every label ll of the positive block BB gives 1N∑r<Nsr(u)→0\frac{1}{N}\sum_{r<N}s_r(u)\to0, and bounded convergence makes the integrated averages tend to 00, contradicting ∫Usrdν=c>0\int_U s_r\,d\nu=c>0.

What Is Checked

The claims are stated over arrays and their plane restrictions, initial array laws of a mixing automorphism, saturated array objects with their laws, data fields, projections onto data and extensions, the word entropy of a finite-valued observable, Hilbert codes with their circle representatives and absolute coefficients, and Borel fibre enumerations with their fibre sums and level restrictions. About these constructions the library proves that at every order nn the coordinatewise state map of an extension is measure preserving from mnNm^N_n to mnMm^M_n✦, and that the circle representative of every Hilbert code is a measurable map into the unit circle.✦