TheoremBase

Both orthogonality relations factorise over coordinates into one-dimensional sums, which product-to-sum formulas, a periodic shift of the samples and the Dirichlet kernel identity evaluate to L or 0; inversion, the linear bijection and Parseval follow by interchanging finite sums, and the spin sum is proved by induction on the number of sites.

Proof

Each result cited is universally quantified over the data in its own statement.

Notation. ϕa\phi_{a} (a∈Za\in\mathbb{Z}) is the one-dimensional system of The Trigonometric System on the Torus §one-dimensional, so that ψk(x)=∏i=1nϕki(xi)\psi_{k}(x)=\prod_{i=1}^{n}\phi_{k_{i}}(x_{i}) by The Trigonometric System on the Torus §system; 22=2\sqrt{2}\sqrt{2}=2 by the definition of 2\sqrt{2} recorded there. Natural numbers are read in R\mathbb{R} through the canonical map ι\iota of The Real Numbers: Standing Notation and Background §numbers; by Properties of the Canonical Map from the Natural Numbers to an Ordered Field §additive, Properties of the Canonical Map from the Natural Numbers to an Ordered Field §multiplicative and Properties of the Canonical Map from the Natural Numbers to an Ordered Field §base-step, the side L=2M+1L=2M+1 read in R\mathbb{R} is 2M+12M+1 formed in R\mathbb{R} with MM read in R\mathbb{R}, so L−1=2ML-1=2M. Natural numbers lie in Z\mathbb{Z} by Arithmetic, Order, Discreteness and Intervals of the Integers §sign, and Z\mathbb{Z} is closed under ++, −- and negation by Arithmetic, Order, Discreteness and Intervals of the Integers §ring. Put

E={a∈Z:−M≤a≤M},D={ν∈Z:0≤ν≤2M},P={b∈Z:1≤b≤M},E=\{a\in\mathbb{Z}:-M\le a\le M\},\qquad D=\{\nu\in\mathbb{Z}:0\le\nu\le 2M\},\qquad P=\{b\in\mathbb{Z}:1\le b\le M\},

nonempty finite sets by Arithmetic, Order, Discreteness and Intervals of the Integers §intervals. Points of Zn⊆Rn\mathbb{Z}^{n}\subseteq\mathbb{R}^{n} are maps [n]→R[n]\to\mathbb{R} by Euclidean Points as Tuples of Real Numbers; hence ΓM\Gamma_{M} is the set EnE^{n} of nn-tuples in EE and WLW_{L} is the set DnD^{n} of nn-tuples in DD, in the sense of Tuples in a Set. For y∈Ry\in\mathbb{R} we write y/L=yL−1y/L=yL^{-1}.

Step 1 (Dirichlet values). For r∈Zr\in\mathbb{Z} with −2M≤r≤2M-2M\le r\le 2M put K(r)=1+2∑a=1Mcos⁡(2πar/L)K(r)=1+2\sum_{a=1}^{M}\cos(2\pi a r/L). Then K(0)=LK(0)=L and K(r)=0K(r)=0 for r≠0r\neq0. If r=0r=0, each summand is cos⁡0=1\cos0=1 by Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §zero; since [M][M] has MM elements (claim 1 of Basic Properties of Finite Sets), Real Sums over a Finite Index Set: Comparison, Nonnegativity, Monotonicity, Term Bounds, Absolute Values, Counting and Limits §count and claim 1 of Properties of a Sum over a Finite Index Set give ∑a=1M1=M\sum_{a=1}^{M}1=M, so K(0)=1+2M=LK(0)=1+2M=L. Let r≠0r\neq0. By The Dirichlet and Fejer Kernel Identities §dirichlet with N=MN=M and t=r/Lt=r/L,

K(r) sin⁡(πr/L)=sin⁡(Lπr/L)=sin⁡(πr).K(r)\,\sin(\pi r/L)=\sin(L\pi r/L)=\sin(\pi r).

By Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §double-angle and Quarter-Turn Identities and Periodicity of Sine and Cosine §integer (with m=rm=r), 1−2(sin⁡(πr))2=cos⁡(2πr)=11-2(\sin(\pi r))^{2}=\cos(2\pi r)=1, so (sin⁡(πr))2=0(\sin(\pi r))^{2}=0 and sin⁡(πr)=0\sin(\pi r)=0. On the other hand π=2x0>0\pi=2x_{0}>0 by The Number Pi §pi and The Least Positive Zero of the Cosine §least-zero. If 0<r0<r, then 0<πr/L<π0<\pi r/L<\pi because r≤2M<Lr\le 2M<L, so 0<sin⁡(πr/L)0<\sin(\pi r/L) by Quarter-Turn Identities and Periodicity of Sine and Cosine §positive. If r<0r<0, the same applies to −r-r, and sin⁡(πr/L)=−sin⁡(π(−r)/L)<0\sin(\pi r/L)=-\sin(\pi(-r)/L)<0 by Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §parity. In both cases sin⁡(πr/L)≠0\sin(\pi r/L)\neq0, hence K(r)=0K(r)=0.

Step 2 (sums over EE). For every map g:E→Rg:E\to\mathbb{R},

∑a∈Eg(a)=g(0)+∑b=1M(g(b)+g(−b)).\sum_{a\in E}g(a)=g(0)+\sum_{b=1}^{M}\bigl(g(b)+g(-b)\bigr).

By Arithmetic, Order, Discreteness and Intervals of the Integers §sign and Arithmetic, Order, Discreteness and Intervals of the Integers §discrete, EE is the union of the pairwise disjoint nonempty sets {0}\{0\}, PP and −P={−b:b∈P}-P=\{-b:b\in P\}. By Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §disjoint-union (twice) and claim 1 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set, ∑a∈Eg(a)=g(0)+∑b∈Pg(b)+∑a∈−Pg(a)\sum_{a\in E}g(a)=g(0)+\sum_{b\in P}g(b)+\sum_{a\in -P}g(a). The map b↦−bb\mapsto-b is a bijection P→−PP\to-P, so claim 2 of Properties of a Sum over a Finite Index Set turns the last sum into ∑b∈Pg(−b)\sum_{b\in P}g(-b), and claim 3 of that lemma combines the two sums over PP into ∑b∈P(g(b)+g(−b))\sum_{b\in P}(g(b)+g(-b)). Finally the restriction of ι\iota to [M][M] is a bijection onto PP: it is injective by Properties of the Canonical Map from the Natural Numbers to an Ordered Field §injective; it maps into PP by Properties of the Canonical Map from the Natural Numbers to an Ordered Field §lower-bound and Properties of the Canonical Map from the Natural Numbers to an Ordered Field §monotone; and every b∈Pb\in P equals ι(k)\iota(k) for some k∈Nk\in\mathbb{N} by Arithmetic, Order, Discreteness and Intervals of the Integers §sign, where k≤Mk\le M since M<kM<k would give M<bM<b by Properties of the Canonical Map from the Natural Numbers to an Ordered Field §monotone (and Properties of the Order on the Natural Numbers §trichotomy). So claims 2 and 1 of Properties of a Sum over a Finite Index Set give ∑b∈P(g(b)+g(−b))=∑b=1M(g(b)+g(−b))\sum_{b\in P}(g(b)+g(-b))=\sum_{b=1}^{M}(g(b)+g(-b)).

Step 3 (sample sums). For m∈Zm\in\mathbb{Z} with −2M≤m≤2M-2M\le m\le 2M, put cm=∑ν∈Ecos⁡(2πmν/L)c_{m}=\sum_{\nu\in E}\cos(2\pi m\nu/L) and sm=∑ν∈Esin⁡(2πmν/L)s_{m}=\sum_{\nu\in E}\sin(2\pi m\nu/L). Then c0=Lc_{0}=L, cm=0c_{m}=0 for m≠0m\neq0, and sm=0s_{m}=0. By Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §parity the summand of cmc_{m} takes equal values at ν\nu and −ν-\nu, and the summand of sms_{m} takes opposite values, vanishing at ν=0\nu=0 by Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §zero. By Step 2 and claim 3 of Properties of Finite Sums, cm=cos⁡0+2∑ν=1Mcos⁡(2πνm/L)=K(m)c_{m}=\cos0+2\sum_{\nu=1}^{M}\cos(2\pi\nu m/L)=K(m), which is given by Step 1; and sm=0+∑ν=1M0=0s_{m}=0+\sum_{\nu=1}^{M}0=0 by Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §vanishing (with claim 1 of Properties of a Sum over a Finite Index Set).

Step 4 (shifting the samples). Let h:R→Rh:\mathbb{R}\to\mathbb{R} satisfy h(x+1)=h(x)h(x+1)=h(x) for every x∈Rx\in\mathbb{R}. Then ∑ν∈Dh(ν/L)=∑ν∈Eh(ν/L)\sum_{\nu\in D}h(\nu/L)=\sum_{\nu\in E}h(\nu/L). Define τ:E→D\tau:E\to D by τ(ν)=ν\tau(\nu)=\nu if 0≤ν0\le\nu and τ(ν)=ν+L\tau(\nu)=\nu+L if ν<0\nu<0, and ρ:D→E\rho:D\to E by ρ(s)=s\rho(s)=s if s≤Ms\le M and ρ(s)=s−L\rho(s)=s-L if M<sM<s; all these values are integers by Arithmetic, Order, Discreteness and Intervals of the Integers §ring. Below we add a number to both sides of an inequality and chain inequalities by the order rules of Elementary Order Arithmetic in an Ordered Field (in force by The Real Numbers: Standing Notation and Background §background), and we use L=2M+1L=2M+1 (Notation paragraph) and 0<M0<M, which holds by Arithmetic, Order, Discreteness and Intervals of the Integers §sign since MM lies in the image of N\mathbb{N}; in particular −M≤0≤M≤2M-M\le0\le M\le 2M. Since the order of R\mathbb{R} is total, every ν∈E\nu\in E falls under exactly one of the cases (a), (b), and every s∈Ds\in D under exactly one of (c), (d). (a) ν∈E\nu\in E, 0≤ν0\le\nu: then τ(ν)=ν\tau(\nu)=\nu satisfies 0≤ν≤M≤2M0\le\nu\le M\le 2M, so it lies in DD, and since τ(ν)≤M\tau(\nu)\le M we get ρ(τ(ν))=ν\rho(\tau(\nu))=\nu. (b) ν∈E\nu\in E, ν<0\nu<0: then ν+1≤0\nu+1\le0 by Arithmetic, Order, Discreteness and Intervals of the Integers §discrete, so −M≤ν≤−1-M\le\nu\le-1, and adding L=2M+1L=2M+1 gives M+1≤ν+L≤2MM+1\le\nu+L\le 2M. So τ(ν)=ν+L\tau(\nu)=\nu+L lies in DD (as 0<M+10<M+1), and since M<M+1≤τ(ν)M<M+1\le\tau(\nu) we get ρ(τ(ν))=(ν+L)−L=ν\rho(\tau(\nu))=(\nu+L)-L=\nu. (c) s∈Ds\in D, s≤Ms\le M: then −M≤0≤s≤M-M\le0\le s\le M, so ρ(s)=s\rho(s)=s lies in EE, and since 0≤ρ(s)0\le\rho(s) we get τ(ρ(s))=s\tau(\rho(s))=s. (d) s∈Ds\in D, M<sM<s: then M+1≤sM+1\le s by Arithmetic, Order, Discreteness and Intervals of the Integers §discrete, and s≤2Ms\le 2M; subtracting L=2M+1L=2M+1 gives −M≤s−L≤−1-M\le s-L\le-1. So ρ(s)=s−L\rho(s)=s-L lies in EE, and since s−L≤−1<0s-L\le-1<0 we get τ(ρ(s))=(s−L)+L=s\tau(\rho(s))=(s-L)+L=s. By (a) and (b), τ\tau maps EE into DD and ρ∘τ\rho\circ\tau is the identity of EE; by (c) and (d), ρ\rho maps DD into EE and τ∘ρ\tau\circ\rho is the identity of DD. So τ\tau is a bijection by claim 3 of Inverse of a Bijection. Since h(τ(ν)/L)h(\tau(\nu)/L) is h(ν/L)h(\nu/L) or h((ν+L)/L)=h(ν/L+1)=h(ν/L)h((\nu+L)/L)=h(\nu/L+1)=h(\nu/L), claim 2 of Properties of a Sum over a Finite Index Set gives ∑ν∈Dh(ν/L)=∑ν∈Eh(τ(ν)/L)=∑ν∈Eh(ν/L)\sum_{\nu\in D}h(\nu/L)=\sum_{\nu\in E}h(\tau(\nu)/L)=\sum_{\nu\in E}h(\nu/L).

Every ϕa\phi_{a} satisfies ϕa(x+1)=ϕa(x)\phi_{a}(x+1)=\phi_{a}(x): this is clear for a=0a=0, and for a≠0a\neq0 it is Quarter-Turn Identities and Periodicity of Sine and Cosine §integer applied with m=am=a or m=−am=-a, because 2πa(x+1)=2πax+2πa2\pi a(x+1)=2\pi ax+2\pi a. Hence so does every product ϕaϕb\phi_{a}\phi_{b}.

Step 5 (one-dimensional orthogonality). For a,b∈Ea,b\in E, ∑ν∈Dϕa(ν/L)ϕb(ν/L)\sum_{\nu\in D}\phi_{a}(\nu/L)\phi_{b}(\nu/L) equals LL if a=ba=b and 00 if a≠ba\neq b. By Step 4 the sum equals ∑ν∈Eϕa(xν)ϕb(xν)\sum_{\nu\in E}\phi_{a}(x_{\nu})\phi_{b}(x_{\nu}) with xν=ν/Lx_{\nu}=\nu/L. The expression is symmetric in aa and bb, so it suffices to treat the following cases, in which we use Step 3 together with claims 3 and 4 of Properties of a Sum over a Finite Index Set; all frequencies below lie between −2M-2M and 2M2M. (i) a=b=0a=b=0: the summand is 1=cos⁡(2π⋅0⋅ν/L)1=\cos(2\pi\cdot0\cdot \nu/L), so the sum is c0=Lc_{0}=L. (ii) a=0<ba=0<b: the sum is 2 cb=0\sqrt{2}\,c_{b}=0. (iii) a=0>ba=0>b: the sum is 2 s−b=0\sqrt{2}\,s_{-b}=0. (iv) a,b>0a,b>0: with u=2πaxνu=2\pi ax_{\nu}, v=2πbxνv=2\pi bx_{\nu}, Product-to-Sum Formulas for Sine and Cosine §cosine-cosine gives ϕaϕb=2cos⁡ucos⁡v=cos⁡(2π(a−b)xν)+cos⁡(2π(a+b)xν)\phi_{a}\phi_{b}=2\cos u\cos v=\cos(2\pi(a-b)x_{\nu})+\cos(2\pi(a+b)x_{\nu}), so the sum is ca−b+ca+bc_{a-b}+c_{a+b}; here 2≤a+b≤2M2\le a+b\le 2M, so ca+b=0c_{a+b}=0, and ca−bc_{a-b} is LL if a=ba=b and 00 otherwise. (v) a,b<0a,b<0: with a′=−aa'=-a, b′=−bb'=-b, Product-to-Sum Formulas for Sine and Cosine §sine-sine gives ϕaϕb=2sin⁡(2πa′xν)sin⁡(2πb′xν)=cos⁡(2π(a′−b′)xν)−cos⁡(2π(a′+b′)xν)\phi_{a}\phi_{b}=2\sin(2\pi a'x_{\nu})\sin(2\pi b'x_{\nu})=\cos(2\pi(a'-b')x_{\nu})-\cos(2\pi(a'+b')x_{\nu}), so the sum is ca′−b′−ca′+b′c_{a'-b'}-c_{a'+b'}, which is LL if a=ba=b and 00 otherwise, as in (iv). (vi) a>0>ba>0>b: with b′=−bb'=-b, Product-to-Sum Formulas for Sine and Cosine §sine-cosine (with u=2πb′xνu=2\pi b'x_{\nu}, v=2πaxνv=2\pi ax_{\nu}) gives ϕaϕb=2sin⁡(2πb′xν)cos⁡(2πaxν)=sin⁡(2π(b′+a)xν)+sin⁡(2π(b′−a)xν)\phi_{a}\phi_{b}=2\sin(2\pi b'x_{\nu})\cos(2\pi ax_{\nu})=\sin(2\pi(b'+a)x_{\nu})+\sin(2\pi(b'-a)x_{\nu}), so the sum is sb′+a+sb′−a=0s_{b'+a}+s_{b'-a}=0, and a≠ba\neq b.

Step 6 (one-dimensional dual orthogonality). For ν,ν′∈D\nu,\nu'\in D, ∑a∈Eϕa(ν/L)ϕa(ν′/L)\sum_{a\in E}\phi_{a}(\nu/L)\phi_{a}(\nu'/L) equals LL if ν=ν′\nu=\nu' and 00 if ν≠ν′\nu\neq\nu'. Put x=ν/Lx=\nu/L, y=ν′/Ly=\nu'/L. For b∈Pb\in P, The Trigonometric System on the Torus §one-dimensional gives ϕb(x)ϕb(y)=2cos⁡(2πbx)cos⁡(2πby)\phi_{b}(x)\phi_{b}(y)=2\cos(2\pi bx)\cos(2\pi by) and ϕ−b(x)ϕ−b(y)=2sin⁡(2πbx)sin⁡(2πby)\phi_{-b}(x)\phi_{-b}(y)=2\sin(2\pi bx)\sin(2\pi by); adding Product-to-Sum Formulas for Sine and Cosine §cosine-cosine and Product-to-Sum Formulas for Sine and Cosine §sine-sine with u=2πbxu=2\pi bx, v=2πbyv=2\pi by, their sum is 2cos⁡(2πb(x−y))=2cos⁡(2πbr/L)2\cos(2\pi b(x-y))=2\cos(2\pi br/L) with r=ν−ν′∈Zr=\nu-\nu'\in\mathbb{Z}, −2M≤r≤2M-2M\le r\le 2M. Since ϕ0(x)ϕ0(y)=1\phi_{0}(x)\phi_{0}(y)=1, Step 2 and claim 3 of Properties of Finite Sums give

∑a∈Eϕa(x)ϕa(y)=1+2∑b=1Mcos⁡(2πbr/L)=K(r),\sum_{a\in E}\phi_{a}(x)\phi_{a}(y)=1+2\sum_{b=1}^{M}\cos(2\pi br/L)=K(r),

which by Step 1 is LL if r=0r=0, that is ν=ν′\nu=\nu', and 00 otherwise.

Step 7 (factorisation). Let GG be a nonempty finite set and j∈Nj\in\mathbb{N}. For all maps g1,…,gj:G→Rg_{1},\dots,g_{j}:G\to\mathbb{R},

∑w∈Gj∏i=1jgi(wi)=∏i=1j(∑e∈Ggi(e)).\sum_{w\in G^{j}}\prod_{i=1}^{j}g_{i}(w_{i})=\prod_{i=1}^{j}\Bigl(\sum_{e\in G}g_{i}(e)\Bigr).

Here GjG^{j} is nonempty and finite by claim 3 of Finiteness of Cartesian Products, Tuple Sets, and Permutation Sets. Let JJ be the set of j∈Nj\in\mathbb{N} for which this holds for every family g1,…,gjg_{1},\dots,g_{j}; we apply Principle of Induction for the Natural Numbers. For j=1j=1: [1]={1}[1]=\{1\} by claim 2 of Basic Properties of Initial Segments of the Natural Numbers. By Tuples in a Set an element w∈G1w\in G^{1} is a map [1]={1}→G[1]=\{1\}\to G, so it is determined by its value w1w_{1} at 11, and every e∈Ge\in G is such a value (of the map sending 11 to ee); hence w↦w1w\mapsto w_{1} is a bijection G1→GG^{1}\to G; both products reduce to their only factor by claim 1 of Properties of Finite Products, and claim 2 of Properties of a Sum over a Finite Index Set gives the identity. Let j∈Jj\in J and let g1,…,gS(j)g_{1},\dots,g_{S(j)} be given. Let q:Gj×G→GS(j)q:G^{j}\times G\to G^{S(j)} be the bijection of claim 2 of Finiteness of Cartesian Products, Tuple Sets, and Permutation Sets. For (v,e)∈Gj×G(v,e)\in G^{j}\times G, the tuple q((v,e))q((v,e)) has iith component viv_{i} for every i∈[j]i\in[j] and S(j)S(j)th component ee, by the component property of qq in claim 2 of Finiteness of Cartesian Products, Tuple Sets, and Permutation Sets; so claim 1 of Properties of Finite Products gives ∏i=1S(j)gi(q((v,e))i)=(∏i=1jgi(vi))gS(j)(e)\prod_{i=1}^{S(j)}g_{i}(q((v,e))_{i})=\bigl(\prod_{i=1}^{j}g_{i}(v_{i})\bigr)g_{S(j)}(e). Hence, by claim 2 of Properties of a Sum over a Finite Index Set, then The Product of Two Sums over Finite Index Sets is a Sum over the Cartesian Product, then the hypothesis j∈Jj\in J for the family g1,…,gjg_{1},\dots,g_{j}, then claim 1 of Properties of Finite Products again,

∑u∈GS(j)∏i=1S(j)gi(ui)=∑(v,e)∈Gj×G(∏i=1jgi(vi))gS(j)(e)=(∑v∈Gj∏i=1jgi(vi))(∑e∈GgS(j)(e))=∏i=1S(j)(∑e∈Ggi(e)).\sum_{u\in G^{S(j)}}\prod_{i=1}^{S(j)}g_{i}(u_{i})=\sum_{(v,e)\in G^{j}\times G}\Bigl(\prod_{i=1}^{j}g_{i}(v_{i})\Bigr)g_{S(j)}(e)=\Bigl(\sum_{v\in G^{j}}\prod_{i=1}^{j}g_{i}(v_{i})\Bigr)\Bigl(\sum_{e\in G}g_{S(j)}(e)\Bigr)=\prod_{i=1}^{S(j)}\Bigl(\sum_{e\in G}g_{i}(e)\Bigr).

So S(j)∈JS(j)\in J, and J=NJ=\mathbb{N}.

Step 8 (reduction to one dimension). The map β:WL→LM\beta:W_{L}\to\mathbb{L}_{M}, β(w)=L−1w\beta(w)=L^{-1}w, is surjective by The Lattice Torus of Odd Side, Lattice Fields, and the Discrete Fourier Coefficients of a Lattice Field §lattice, and injective because (L−1w)i=L−1wi(L^{-1}w)_{i}=L^{-1}w_{i} by Scalar Multiple of a Point of Rn\mathbb{R}^n and L−1≠0L^{-1}\neq0, together with claim 1 of Euclidean Points as Tuples of Real Numbers; so it is a bijection. By The Trigonometric System on the Torus §system and claim 2 of Properties of Finite Products, for k,l∈ΓMk,l\in\Gamma_{M} and w,w′∈WLw,w'\in W_{L},

ψk(β(w)) ψl(β(w′))=∏i=1nϕki(wi/L) ϕli(wi′/L).\psi_{k}(\beta(w))\,\psi_{l}(\beta(w'))=\prod_{i=1}^{n}\phi_{k_{i}}(w_{i}/L)\,\phi_{l_{i}}(w'_{i}/L).

Hence claim 2 of Properties of a Sum over a Finite Index Set and Step 7 (with G=DG=D, WL=DnW_{L}=D^{n}, respectively G=EG=E, ΓM=En\Gamma_{M}=E^{n}) give

∑z∈LMψk(z)ψl(z)=∏i=1n(∑ν∈Dϕki(ν/L)ϕli(ν/L)),∑k∈ΓMψk(β(w))ψk(β(w′))=∏i=1n(∑a∈Eϕa(wi/L)ϕa(wi′/L)).\sum_{z\in\mathbb{L}_{M}}\psi_{k}(z)\psi_{l}(z)=\prod_{i=1}^{n}\Bigl(\sum_{\nu\in D}\phi_{k_{i}}(\nu/L)\phi_{l_{i}}(\nu/L)\Bigr),\qquad \sum_{k\in\Gamma_{M}}\psi_{k}(\beta(w))\psi_{k}(\beta(w'))=\prod_{i=1}^{n}\Bigl(\sum_{a\in E}\phi_{a}(w_{i}/L)\phi_{a}(w'_{i}/L)\Bigr).

Step 9 (products of Kronecker factors). Let d1,…,dn∈{0,1}d_{1},\dots,d_{n}\in\{0,1\}. If di=1d_{i}=1 for every ii, then ∏i=1n(L di)=Ln\prod_{i=1}^{n}(L\,d_{i})=L^{n} by Natural Number Power of an Element of a Field; otherwise some factor vanishes and ∏i=1n(L di)=0\prod_{i=1}^{n}(L\,d_{i})=0 by claim 4 of Properties of Finite Products. By claim 1 of Euclidean Points as Tuples of Real Numbers, k=lk=l in ΓM\Gamma_{M} exactly when ki=lik_{i}=l_{i} for every ii, and β(w)=β(w′)\beta(w)=\beta(w') exactly when wi=wi′w_{i}=w'_{i} for every ii (β\beta being injective). Combining this with Steps 5, 6 and 8:

∑z∈LMψk(z)ψl(z)=Lnδkl(k,l∈ΓM),∑k∈ΓMψk(y)ψk(z)={Lny=z,0y≠z(y,z∈LM).(∗)\sum_{z\in\mathbb{L}_{M}}\psi_{k}(z)\psi_{l}(z)=L^{n}\delta_{kl}\quad(k,l\in\Gamma_{M}),\qquad \sum_{k\in\Gamma_{M}}\psi_{k}(y)\psi_{k}(z)=\begin{cases}L^{n}&y=z,\\0&y\neq z\end{cases}\quad(y,z\in\mathbb{L}_{M}).\tag{$\ast$}

Claim 2 (Counting the modes). By The Cubes of the Integer Lattice are Finite, Nested, Exhaust the Lattice and Have (2N+1)^n Points §count with N=MN=M, ∑k∈ΓM1=(2M+1)n=Ln\sum_{k\in\Gamma_{M}}1=(2M+1)^{n}=L^{n}, using the Notation paragraph.

Claim 1 (Counting the sites). By Euclidean Points as Tuples of Real Numbers (claim 2) define γ:ΓM→WL\gamma:\Gamma_{M}\to W_{L} by γ(k)i=ki+M\gamma(k)_{i}=k_{i}+M and γ′:WL→ΓM\gamma':W_{L}\to\Gamma_{M} by γ′(w)i=wi−M\gamma'(w)_{i}=w_{i}-M (i∈[n]i\in[n]); they land in the stated sets because −M≤ki≤M-M\le k_{i}\le M if and only if 0≤ki+M≤2M=L−10\le k_{i}+M\le 2M=L-1, and they are mutually inverse by claim 1 of that lemma, so γ\gamma is a bijection by claim 3 of Inverse of a Bijection. Reindexing along β\beta and γ\gamma by claim 2 of Properties of a Sum over a Finite Index Set, and using claim 2 of this lemma (proved above),

∑z∈LM1=∑w∈WL1=∑k∈ΓM1=Ln.\sum_{z\in\mathbb{L}_{M}}1=\sum_{w\in W_{L}}1=\sum_{k\in\Gamma_{M}}1=L^{n}.

Claim 3 (Orthogonality). Multiply the first identity of (∗)(\ast) by L−nL^{-n}.

Claim 4 (Inversion formula). Let ff be a lattice field and y∈LMy\in\mathbb{L}_{M}. By The Lattice Torus of Odd Side, Lattice Fields, and the Discrete Fourier Coefficients of a Lattice Field §coefficients, claim 4 of Properties of a Sum over a Finite Index Set, and claim 5 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set (interchange over ΓM×LM\Gamma_{M}\times\mathbb{L}_{M}),

∑k∈ΓMf^(k)ψk(y)=∑k∈ΓM(∑z∈LML−nf(z)ψk(z)ψk(y))=∑z∈LML−nf(z)(∑k∈ΓMψk(z)ψk(y)).\sum_{k\in\Gamma_{M}}\hat{f}(k)\psi_{k}(y)=\sum_{k\in\Gamma_{M}}\Bigl(\sum_{z\in\mathbb{L}_{M}}L^{-n}f(z)\psi_{k}(z)\psi_{k}(y)\Bigr)=\sum_{z\in\mathbb{L}_{M}}L^{-n}f(z)\Bigl(\sum_{k\in\Gamma_{M}}\psi_{k}(z)\psi_{k}(y)\Bigr).

By the second identity of (∗)(\ast) the summand vanishes for z≠yz\neq y and equals L−nf(y)Ln=f(y)L^{-n}f(y)L^{n}=f(y) for z=yz=y, so Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §vanishing (with the nonempty subset {y}\{y\}) and claim 1 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set show that the sum is f(y)f(y).

Claim 5 (Bijection). Write A(f)=f^\mathcal{A}(f)=\hat{f}. The vector-space operations of Map(LM,R)\mathrm{Map}(\mathbb{L}_{M},\mathbb{R}) and Map(ΓM,R)\mathrm{Map}(\Gamma_{M},\mathbb{R}) are pointwise by The Real Vector Space of Real-Valued Functions on a Set §vector-space. For lattice fields f,gf,g, λ∈R\lambda\in\mathbb{R} and k∈ΓMk\in\Gamma_{M}, claims 3 and 4 of Properties of a Sum over a Finite Index Set give f+g^(k)=f^(k)+g^(k)\widehat{f+g}(k)=\hat{f}(k)+\hat{g}(k) and λf^(k)=λf^(k)\widehat{\lambda f}(k)=\lambda\hat{f}(k), so A\mathcal{A} is linear in the sense of Linear Map. For b∈Map(ΓM,R)b\in\mathrm{Map}(\Gamma_{M},\mathbb{R}) let S(b)\mathcal{S}(b) be the lattice field z↦∑k∈ΓMb(k)ψk(z)z\mapsto\sum_{k\in\Gamma_{M}}b(k)\psi_{k}(z). By claim 4, S(A(f))(z)=f(z)\mathcal{S}(\mathcal{A}(f))(z)=f(z) for every zz, so S∘A\mathcal{S}\circ\mathcal{A} is the identity. For bb and l∈ΓMl\in\Gamma_{M}, the same three sum rules as in claim 4 and claim 3 give

A(S(b))(l)=∑z∈LML−n(∑k∈ΓMb(k)ψk(z))ψl(z)=∑k∈ΓMb(k)(L−n∑z∈LMψk(z)ψl(z))=∑k∈ΓMb(k)δkl=b(l),\mathcal{A}(\mathcal{S}(b))(l)=\sum_{z\in\mathbb{L}_{M}}L^{-n}\Bigl(\sum_{k\in\Gamma_{M}}b(k)\psi_{k}(z)\Bigr)\psi_{l}(z)=\sum_{k\in\Gamma_{M}}b(k)\Bigl(L^{-n}\sum_{z\in\mathbb{L}_{M}}\psi_{k}(z)\psi_{l}(z)\Bigr)=\sum_{k\in\Gamma_{M}}b(k)\delta_{kl}=b(l),

the last step by Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §vanishing and claim 1 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set. So A∘S\mathcal{A}\circ\mathcal{S} is the identity, and claim 3 of Inverse of a Bijection shows that A\mathcal{A} is a bijection with inverse S\mathcal{S}.

Claim 6 (Parseval). Let f,gf,g be lattice fields. Expanding g(z)g(z) by claim 4 and using claim 4 of Properties of a Sum over a Finite Index Set and claim 5 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set,

L−n∑z∈LMf(z)g(z)=∑z∈LM(∑k∈ΓML−nf(z)g^(k)ψk(z))=∑k∈ΓMg^(k)(L−n∑z∈LMf(z)ψk(z))=∑k∈ΓMf^(k)g^(k).L^{-n}\sum_{z\in\mathbb{L}_{M}}f(z)g(z)=\sum_{z\in\mathbb{L}_{M}}\Bigl(\sum_{k\in\Gamma_{M}}L^{-n}f(z)\hat{g}(k)\psi_{k}(z)\Bigr)=\sum_{k\in\Gamma_{M}}\hat{g}(k)\Bigl(L^{-n}\sum_{z\in\mathbb{L}_{M}}f(z)\psi_{k}(z)\Bigr)=\sum_{k\in\Gamma_{M}}\hat{f}(k)\hat{g}(k).

Claim 7 (Spin sum). For s∈Rs\in\mathbb{R} put ℓ(s)=log⁡(exp⁡(s)+exp⁡(−s))\ell(s)=\log(\exp(s)+\exp(-s)), defined as recorded in the statement; then exp⁡(ℓ(s))=exp⁡(s)+exp⁡(−s)\exp(\ell(s))=\exp(s)+\exp(-s) by The Natural Logarithm. The set {−1,1}={−1}∪{1}\{-1,1\}=\{-1\}\cup\{1\} is nonempty and finite by claim 2 of Basic Properties of Finite Sets and claim 1 of Peeling an Element off a Finite Set, and Unions of Finite Sets, so Map(F,{−1,1})\mathrm{Map}(F,\{-1,1\}) is nonempty and finite for every nonempty finite set FF by The Maps from a Finite Set to a Nonempty Finite Set Form a Nonempty Finite Set §finite. Let ZZ be the set of those p∈Np\in\mathbb{N} such that for every set FF with pp elements and every map t:F→Rt:F\to\mathbb{R}

∑σ∈Map(F,{−1,1})exp⁡(∑z∈Fσ(z)t(z))=exp⁡(∑z∈Fℓ(t(z))).(∗∗)\sum_{\sigma\in\mathrm{Map}(F,\{-1,1\})}\exp\Bigl(\sum_{z\in F}\sigma(z)t(z)\Bigr)=\exp\Bigl(\sum_{z\in F}\ell(t(z))\Bigr).\tag{$\ast\ast$}

We show Z=NZ=\mathbb{N} by Principle of Induction for the Natural Numbers; the claim is then (∗∗)(\ast\ast) for F=LMF=\mathbb{L}_{M}, which is nonempty and finite by The Lattice Torus of Odd Side, Lattice Fields, and the Discrete Fourier Coefficients of a Lattice Field §lattice, and so has pp elements for some pp by Finite Set.

1∈Z1\in Z. A set with 11 element is {x}\{x\} for some xx, as [1]={1}[1]=\{1\} by claim 2 of Basic Properties of Initial Segments of the Natural Numbers. Map({x},{−1,1})\mathrm{Map}(\{x\},\{-1,1\}) is the union of the singletons {σ+}\{\sigma_{+}\} and {σ−}\{\sigma_{-}\}, where σ±(x)=±1\sigma_{\pm}(x)=\pm1, which are disjoint since −1≠1-1\neq1. By Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §disjoint-union and claim 1 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set, the left side of (∗∗)(\ast\ast) is exp⁡(t(x))+exp⁡(−t(x))=exp⁡(ℓ(t(x)))\exp(t(x))+\exp(-t(x))=\exp(\ell(t(x))), which is the right side by claim 1 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set.

If p∈Zp\in Z then S(p)∈ZS(p)\in Z. Let FF have S(p)S(p) elements and t:F→Rt:F\to\mathbb{R}. By claim 2 of Peeling an Element off a Finite Set, and Unions of Finite Sets, F=F′∪{x}F=F'\cup\{x\} with F′F' having pp elements (so F′≠∅F'\neq\emptyset) and x∉F′x\notin F'. Write Σ′=Map(F′,{−1,1})\Sigma'=\mathrm{Map}(F',\{-1,1\}) and, for ε∈{−1,1}\varepsilon\in\{-1,1\}, Aε={σ∈Map(F,{−1,1}):σ(x)=ε}A_{\varepsilon}=\{\sigma\in\mathrm{Map}(F,\{-1,1\}):\sigma(x)=\varepsilon\}; then A1∩A−1=∅A_{1}\cap A_{-1}=\emptyset and A1∪A−1=Map(F,{−1,1})A_{1}\cup A_{-1}=\mathrm{Map}(F,\{-1,1\}). Let χε:Σ′→Aε\chi_{\varepsilon}:\Sigma'\to A_{\varepsilon} send τ\tau to its extension by x↦εx\mapsto\varepsilon; restriction to F′F' is a two-sided inverse, so χε\chi_{\varepsilon} is a bijection by claim 3 of Inverse of a Bijection, and AεA_{\varepsilon} is nonempty and finite by claim 4 of Basic Properties of Finite Sets. For τ∈Σ′\tau\in\Sigma', claim 2 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set and claim 1 of Basic Properties of the Exponential Function give

exp⁡(∑z∈Fχε(τ)(z)t(z))=exp⁡(∑z∈F′τ(z)t(z)+εt(x))=exp⁡(εt(x))exp⁡(∑z∈F′τ(z)t(z)).\exp\Bigl(\sum_{z\in F}\chi_{\varepsilon}(\tau)(z)t(z)\Bigr)=\exp\Bigl(\sum_{z\in F'}\tau(z)t(z)+\varepsilon t(x)\Bigr)=\exp(\varepsilon t(x))\exp\Bigl(\sum_{z\in F'}\tau(z)t(z)\Bigr).

By claims 2 and 4 of Properties of a Sum over a Finite Index Set and the hypothesis p∈Zp\in Z applied to F′F' and the restriction of tt, the sum of the left side over σ∈Aε\sigma\in A_{\varepsilon} is exp⁡(εt(x))exp⁡(∑z∈F′ℓ(t(z)))\exp(\varepsilon t(x))\exp\bigl(\sum_{z\in F'}\ell(t(z))\bigr). Adding the cases ε=1\varepsilon=1 and ε=−1\varepsilon=-1 by Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §disjoint-union, and using claim 1 of Basic Properties of the Exponential Function and claim 2 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set,

∑σ∈Map(F,{−1,1})exp⁡(∑z∈Fσ(z)t(z))=exp⁡(ℓ(t(x)))exp⁡(∑z∈F′ℓ(t(z)))=exp⁡(∑z∈Fℓ(t(z))).\sum_{\sigma\in\mathrm{Map}(F,\{-1,1\})}\exp\Bigl(\sum_{z\in F}\sigma(z)t(z)\Bigr)=\exp(\ell(t(x)))\exp\Bigl(\sum_{z\in F'}\ell(t(z))\Bigr)=\exp\Bigl(\sum_{z\in F}\ell(t(z))\Bigr).

So S(p)∈ZS(p)\in Z, hence Z=NZ=\mathbb{N}, which proves claim 7.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…