Each result cited is universally quantified over the data in its own statement. Points of R n \mathbb{R}^{n} R n , and of Z n ⊆ R n \mathbb{Z}^{n}\subseteq\mathbb{R}^{n} Z n ⊆ R n (Lattice-Periodic Functions and the Periodic Function Classes §lattice ), are added, subtracted and multiplied by real numbers componentwise (Scalar Multiple of a Point of R n \mathbb{R}^n R n ). The side L L L is a natural number, read in R \mathbb{R} R as in The Real Numbers: Standing Notation and Background §numbers , so L ∈ Z L\in\mathbb{Z} L ∈ Z by Arithmetic, Order, Discreteness and Intervals of the Integers §sign ; and 0 , 1 ∈ Z 0,1\in\mathbb{Z} 0 , 1 ∈ Z and Z \mathbb{Z} Z is closed under + + + , − - − and multiplication by Arithmetic, Order, Discreteness and Intervals of the Integers §ring , so that sums, differences and integer multiples of points of Z n \mathbb{Z}^{n} Z n lie in Z n \mathbb{Z}^{n} Z n . For m ∈ Z m\in\mathbb{Z} m ∈ Z , C m C_{m} C m and S m S_{m} S m are the trigonometric monomials C m ( t ) = cos ( 2 π m t ) C_{m}(t)=\cos(2\pi mt) C m ( t ) = cos ( 2 πm t ) and S m ( t ) = sin ( 2 π m t ) S_{m}(t)=\sin(2\pi mt) S m ( t ) = sin ( 2 πm t ) of Cell Integrals of the Trigonometric Monomials , and ϕ m \phi_{m} ϕ m is the one-dimensional map of The Trigonometric System on the Torus §one-dimensional ; since ψ k \psi_{k} ψ k is the map written e k e_{k} e k in The Trigonometric System on the Torus §system (The Lattice Torus of Odd Side, Lattice Fields, and the Discrete Fourier Coefficients of a Lattice Field ),
ψ k ( x ) = ∏ i = 1 n ϕ k i ( x i ) ( k ∈ Z n , x ∈ R n ) . \psi_{k}(x)=\prod_{i=1}^{n}\phi_{k_{i}}(x_{i})\qquad(k\in\mathbb{Z}^{n},\ x\in\mathbb{R}^{n}). ψ k ( x ) = i = 1 ∏ n ϕ k i ( x i ) ( k ∈ Z n , x ∈ R n ) .
The map w ↦ L − 1 w w\mapsto L^{-1}w w ↦ L − 1 w from W L W_{L} W L to L M \mathbb{L}_{M} L M is a bijection : it is onto by The Lattice Torus of Odd Side, Lattice Fields, and the Discrete Fourier Coefficients of a Lattice Field §lattice , and L − 1 w = L − 1 w ′ L^{-1}w=L^{-1}w' L − 1 w = L − 1 w ′ implies w = w ′ w=w' w = w ′ after multiplying each component by L L L . Every site y y y is therefore L − 1 w L^{-1}w L − 1 w for exactly one w ∈ W L w\in W_{L} w ∈ W L , and by Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols §matrix , J ( L − 1 w , L − 1 w ′ ) = j ( w − w ′ ) J(L^{-1}w,L^{-1}w')=\mathfrak{j}(w-w') J ( L − 1 w , L − 1 w ′ ) = j ( w − w ′ ) for w , w ′ ∈ W L w,w'\in W_{L} w , w ′ ∈ W L .
Step 1: symmetry (claim 1). For p ∈ [ n ] p\in[n] p ∈ [ n ] the reflection R p R_{p} R p of Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols maps Z n \mathbb{Z}^{n} Z n to Z n \mathbb{Z}^{n} Z n . Let b ∈ Z n b\in\mathbb{Z}^{n} b ∈ Z n . Put b ( 0 ) = b b^{(0)}=b b ( 0 ) = b and b ( p ) = R p b ( p − 1 ) ∈ Z n b^{(p)}=R_{p}b^{(p-1)}\in\mathbb{Z}^{n} b ( p ) = R p b ( p − 1 ) ∈ Z n for p ∈ [ n ] p\in[n] p ∈ [ n ] . By induction on p p p , the i i i -th component of b ( p ) b^{(p)} b ( p ) is − b i -b_{i} − b i for i ≤ p i\le p i ≤ p and b i b_{i} b i for p < i p<i p < i , so b ( n ) = − b b^{(n)}=-b b ( n ) = − b ; and j ( b ( p ) ) = j ( R p b ( p − 1 ) ) = j ( b ( p − 1 ) ) \mathfrak{j}(b^{(p)})=\mathfrak{j}(R_{p}b^{(p-1)})=\mathfrak{j}(b^{(p-1)}) j ( b ( p ) ) = j ( R p b ( p − 1 ) ) = j ( b ( p − 1 ) ) by Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols §interaction . Hence j ( − b ) = j ( b ) \mathfrak{j}(-b)=\mathfrak{j}(b) j ( − b ) = j ( b ) . Now let y = L − 1 w y=L^{-1}w y = L − 1 w and z = L − 1 w ′ z=L^{-1}w' z = L − 1 w ′ with w , w ′ ∈ W L w,w'\in W_{L} w , w ′ ∈ W L . Then J ( z , y ) = j ( w ′ − w ) = j ( − ( w − w ′ ) ) = j ( w − w ′ ) = J ( y , z ) J(z,y)=\mathfrak{j}(w'-w)=\mathfrak{j}(-(w-w'))=\mathfrak{j}(w-w')=J(y,z) J ( z , y ) = j ( w ′ − w ) = j ( − ( w − w ′ )) = j ( w − w ′ ) = J ( y , z ) .
Step 2: two involutions of W L W_{L} W L . A map τ : W L → W L \tau:W_{L}\to W_{L} τ : W L → W L with τ ( τ ( u ) ) = u \tau(\tau(u))=u τ ( τ ( u )) = u for every u ∈ W L u\in W_{L} u ∈ W L is a bijection: for v ∈ W L v\in W_{L} v ∈ W L one has τ ( τ ( v ) ) = v \tau(\tau(v))=v τ ( τ ( v )) = v , and τ ( u ) = v \tau(u)=v τ ( u ) = v forces u = τ ( τ ( u ) ) = τ ( v ) u=\tau(\tau(u))=\tau(v) u = τ ( τ ( u )) = τ ( v ) .
(a) Fix w ∈ W L w\in W_{L} w ∈ W L and put y = L − 1 w y=L^{-1}w y = L − 1 w . For u ∈ W L u\in W_{L} u ∈ W L and i ∈ [ n ] i\in[n] i ∈ [ n ] let λ i ( u ) = 1 \lambda_{i}(u)=1 λ i ( u ) = 1 if w i < u i w_{i}<u_{i} w i < u i and λ i ( u ) = 0 \lambda_{i}(u)=0 λ i ( u ) = 0 otherwise, let λ ( u ) ∈ Z n \lambda(u)\in\mathbb{Z}^{n} λ ( u ) ∈ Z n have components λ i ( u ) ∈ { 0 , 1 } ⊆ Z \lambda_{i}(u)\in\{0,1\}\subseteq\mathbb{Z} λ i ( u ) ∈ { 0 , 1 } ⊆ Z , and let τ ( u ) ∈ Z n \tau(u)\in\mathbb{Z}^{n} τ ( u ) ∈ Z n have components τ ( u ) i = w i − u i + L λ i ( u ) \tau(u)_{i}=w_{i}-u_{i}+L\lambda_{i}(u) τ ( u ) i = w i − u i + L λ i ( u ) , integers by Arithmetic, Order, Discreteness and Intervals of the Integers §ring as w i , u i , L , λ i ( u ) ∈ Z w_{i},u_{i},L,\lambda_{i}(u)\in\mathbb{Z} w i , u i , L , λ i ( u ) ∈ Z . If u i ≤ w i u_{i}\le w_{i} u i ≤ w i , then 0 ≤ τ ( u ) i = w i − u i ≤ w i ≤ L − 1 0\le\tau(u)_{i}=w_{i}-u_{i}\le w_{i}\le L-1 0 ≤ τ ( u ) i = w i − u i ≤ w i ≤ L − 1 . If w i < u i w_{i}<u_{i} w i < u i , then w i + 1 ≤ u i w_{i}+1\le u_{i} w i + 1 ≤ u i by Arithmetic, Order, Discreteness and Intervals of the Integers §discrete , that is, w i − u i ≤ − 1 w_{i}-u_{i}\le-1 w i − u i ≤ − 1 , and w i − u i ≥ − u i ≥ − ( L − 1 ) w_{i}-u_{i}\ge-u_{i}\ge-(L-1) w i − u i ≥ − u i ≥ − ( L − 1 ) , so 1 ≤ τ ( u ) i ≤ L − 1 1\le\tau(u)_{i}\le L-1 1 ≤ τ ( u ) i ≤ L − 1 . Thus τ ( u ) ∈ W L \tau(u)\in W_{L} τ ( u ) ∈ W L . Moreover τ ( τ ( u ) ) = u \tau(\tau(u))=u τ ( τ ( u )) = u : if u i ≤ w i u_{i}\le w_{i} u i ≤ w i , then τ ( u ) i ≤ w i \tau(u)_{i}\le w_{i} τ ( u ) i ≤ w i , so λ i ( τ ( u ) ) = 0 \lambda_{i}(\tau(u))=0 λ i ( τ ( u )) = 0 and τ ( τ ( u ) ) i = w i − ( w i − u i ) = u i \tau(\tau(u))_{i}=w_{i}-(w_{i}-u_{i})=u_{i} τ ( τ ( u ) ) i = w i − ( w i − u i ) = u i ; if w i < u i w_{i}<u_{i} w i < u i , then τ ( u ) i = w i + ( L − u i ) > w i \tau(u)_{i}=w_{i}+(L-u_{i})>w_{i} τ ( u ) i = w i + ( L − u i ) > w i as u i ≤ L − 1 u_{i}\le L-1 u i ≤ L − 1 , so λ i ( τ ( u ) ) = 1 \lambda_{i}(\tau(u))=1 λ i ( τ ( u )) = 1 and τ ( τ ( u ) ) i = w i − τ ( u ) i + L = u i \tau(\tau(u))_{i}=w_{i}-\tau(u)_{i}+L=u_{i} τ ( τ ( u ) ) i = w i − τ ( u ) i + L = u i . So τ \tau τ is a bijection of W L W_{L} W L , and
w − τ ( u ) = u + L ( − λ ( u ) ) , L − 1 τ ( u ) = ( y − L − 1 u ) + λ ( u ) ( u ∈ W L ) . w-\tau(u)=u+L\bigl(-\lambda(u)\bigr),\qquad L^{-1}\tau(u)=\bigl(y-L^{-1}u\bigr)+\lambda(u)\qquad(u\in W_{L}). w − τ ( u ) = u + L ( − λ ( u ) ) , L − 1 τ ( u ) = ( y − L − 1 u ) + λ ( u ) ( u ∈ W L ) .
Here w − τ ( u ) w-\tau(u) w − τ ( u ) and − λ ( u ) -\lambda(u) − λ ( u ) lie in Z n \mathbb{Z}^{n} Z n by Arithmetic, Order, Discreteness and Intervals of the Integers §ring . Hence j ( w − τ ( u ) ) = j ( u ) \mathfrak{j}(w-\tau(u))=\mathfrak{j}(u) j ( w − τ ( u )) = j ( u ) by Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols §interaction , applied with l = − λ ( u ) l=-\lambda(u) l = − λ ( u ) . Also ψ k ∈ C p e r \psi_{k}\in C_{\mathrm{per}} ψ k ∈ C per by The Trigonometric System on the Torus is Orthonormal §classes , so ψ k \psi_{k} ψ k is Z n \mathbb{Z}^{n} Z n -periodic by Lattice-Periodic Functions and the Periodic Function Classes §classes and Lattice-Periodic Functions and the Periodic Function Classes §periodic , and ψ k ( L − 1 τ ( u ) ) = ψ k ( y − L − 1 u ) \psi_{k}(L^{-1}\tau(u))=\psi_{k}(y-L^{-1}u) ψ k ( L − 1 τ ( u )) = ψ k ( y − L − 1 u ) for every k ∈ Z n k\in\mathbb{Z}^{n} k ∈ Z n .
(b) Let p ∈ [ n ] p\in[n] p ∈ [ n ] . For u ∈ W L u\in W_{L} u ∈ W L let ϱ p u ∈ Z n \varrho_{p}u\in\mathbb{Z}^{n} ϱ p u ∈ Z n have components ( ϱ p u ) i = u i (\varrho_{p}u)_{i}=u_{i} ( ϱ p u ) i = u i for i ≠ p i\ne p i = p , and ( ϱ p u ) p = 0 (\varrho_{p}u)_{p}=0 ( ϱ p u ) p = 0 if u p = 0 u_{p}=0 u p = 0 , ( ϱ p u ) p = L − u p (\varrho_{p}u)_{p}=L-u_{p} ( ϱ p u ) p = L − u p if u p ≠ 0 u_{p}\ne0 u p = 0 ; these components are integers by Arithmetic, Order, Discreteness and Intervals of the Integers §ring . If u p ≠ 0 u_{p}\ne0 u p = 0 then 0 < u p 0<u_{p} 0 < u p , as 0 ≤ u p 0\le u_{p} 0 ≤ u p , so 1 ≤ u p 1\le u_{p} 1 ≤ u p by Arithmetic, Order, Discreteness and Intervals of the Integers §discrete ; thus 1 ≤ u p ≤ L − 1 1\le u_{p}\le L-1 1 ≤ u p ≤ L − 1 , so 1 ≤ L − u p ≤ L − 1 1\le L-u_{p}\le L-1 1 ≤ L − u p ≤ L − 1 ; thus ϱ p u ∈ W L \varrho_{p}u\in W_{L} ϱ p u ∈ W L , and ϱ p ( ϱ p u ) = u \varrho_{p}(\varrho_{p}u)=u ϱ p ( ϱ p u ) = u since L − ( L − u p ) = u p L-(L-u_{p})=u_{p} L − ( L − u p ) = u p . So ϱ p \varrho_{p} ϱ p is a bijection of W L W_{L} W L . Further ϱ p u = R p u + L η ( u ) \varrho_{p}u=R_{p}u+L\,\eta(u) ϱ p u = R p u + L η ( u ) , where η ( u ) ∈ Z n \eta(u)\in\mathbb{Z}^{n} η ( u ) ∈ Z n has p p p -th component 1 1 1 if u p ≠ 0 u_{p}\ne0 u p = 0 and 0 0 0 if u p = 0 u_{p}=0 u p = 0 , and all other components 0 0 0 , so that η ( u ) ∈ Z n \eta(u)\in\mathbb{Z}^{n} η ( u ) ∈ Z n as 0 , 1 ∈ Z 0,1\in\mathbb{Z} 0 , 1 ∈ Z (Arithmetic, Order, Discreteness and Intervals of the Integers §ring ); by Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols §interaction , applied with l = η ( u ) l=\eta(u) l = η ( u ) and then with i = p i=p i = p , j ( ϱ p u ) = j ( R p u ) = j ( u ) \mathfrak{j}(\varrho_{p}u)=\mathfrak{j}(R_{p}u)=\mathfrak{j}(u) j ( ϱ p u ) = j ( R p u ) = j ( u ) . Finally, for q ∈ Z q\in\mathbb{Z} q ∈ Z ,
S q ( L − 1 ( ϱ p u ) p ) = − S q ( L − 1 u p ) . S_{q}\bigl(L^{-1}(\varrho_{p}u)_{p}\bigr)=-S_{q}\bigl(L^{-1}u_{p}\bigr). S q ( L − 1 ( ϱ p u ) p ) = − S q ( L − 1 u p ) .
If u p = 0 u_{p}=0 u p = 0 , both sides are ± sin 0 = 0 \pm\sin0=0 ± sin 0 = 0 by Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §zero . If u p ≠ 0 u_{p}\ne0 u p = 0 , put s = L − 1 u p s=L^{-1}u_{p} s = L − 1 u p ; then L − 1 ( L − u p ) = 1 − s L^{-1}(L-u_{p})=1-s L − 1 ( L − u p ) = 1 − s and S q ( 1 − s ) = sin ( − 2 π q s + 2 π q ) = sin ( − 2 π q s ) = − sin ( 2 π q s ) = − S q ( s ) S_{q}(1-s)=\sin(-2\pi qs+2\pi q)=\sin(-2\pi qs)=-\sin(2\pi qs)=-S_{q}(s) S q ( 1 − s ) = sin ( − 2 π q s + 2 π q ) = sin ( − 2 π q s ) = − sin ( 2 π q s ) = − S q ( s ) , by Quarter-Turn Identities and Periodicity of Sine and Cosine §integer and Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §parity .
Step 3: a one-variable expansion. Let m ∈ Z m\in\mathbb{Z} m ∈ Z . By Arithmetic, Order, Discreteness and Intervals of the Integers §sign , exactly one of m = 0 m=0 m = 0 , 0 < m 0<m 0 < m and m < 0 m<0 m < 0 holds. By Absolute Value in an Ordered Field , ∣ m ∣ |m| ∣ m ∣ is m m m or − m -m − m , so ∣ m ∣ ∈ Z |m|\in\mathbb{Z} ∣ m ∣ ∈ Z by Arithmetic, Order, Discreteness and Intervals of the Integers §ring , and S ∣ m ∣ S_{|m|} S ∣ m ∣ is defined. Let ϵ m = 1 \epsilon_{m}=1 ϵ m = 1 if 0 ≤ m 0\le m 0 ≤ m and ϵ m = − 1 \epsilon_{m}=-1 ϵ m = − 1 if m < 0 m<0 m < 0 . We show that for all m ∈ Z m\in\mathbb{Z} m ∈ Z and t , s ∈ R t,s\in\mathbb{R} t , s ∈ R
ϕ m ( t − s ) = ϕ m ( t ) C m ( s ) + ϵ m ϕ − m ( t ) S ∣ m ∣ ( s ) . (A) \phi_{m}(t-s)=\phi_{m}(t)\,C_{m}(s)+\epsilon_{m}\,\phi_{-m}(t)\,S_{|m|}(s).\qquad\text{(A)} ϕ m ( t − s ) = ϕ m ( t ) C m ( s ) + ϵ m ϕ − m ( t ) S ∣ m ∣ ( s ) . (A)
If m = 0 m=0 m = 0 : ϕ 0 ( t − s ) = 1 = ϕ 0 ( t ) C 0 ( s ) \phi_{0}(t-s)=1=\phi_{0}(t)\,C_{0}(s) ϕ 0 ( t − s ) = 1 = ϕ 0 ( t ) C 0 ( s ) and S 0 ( s ) = 0 S_{0}(s)=0 S 0 ( s ) = 0 , by The Trigonometric System on the Torus §one-dimensional and Cell Integrals of the Trigonometric Monomials §calculus . If 0 < m 0<m 0 < m : then ∣ m ∣ = m |m|=m ∣ m ∣ = m and − m < 0 -m<0 − m < 0 , so ϕ − m ( t ) = 2 S m ( t ) \phi_{-m}(t)=\sqrt{2}\,S_{m}(t) ϕ − m ( t ) = 2 S m ( t ) ; by Addition Formulas for Sine and Cosine §cosine with a = 2 π m t a=2\pi mt a = 2 πm t and x = − 2 π m s x=-2\pi ms x = − 2 πm s , and by Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §parity ,
ϕ m ( t − s ) = 2 cos ( 2 π m t − 2 π m s ) = 2 C m ( t ) C m ( s ) + 2 S m ( t ) S m ( s ) = ϕ m ( t ) C m ( s ) + ϕ − m ( t ) S ∣ m ∣ ( s ) . \phi_{m}(t-s)=\sqrt{2}\cos(2\pi mt-2\pi ms)=\sqrt{2}\,C_{m}(t)\,C_{m}(s)+\sqrt{2}\,S_{m}(t)\,S_{m}(s)=\phi_{m}(t)\,C_{m}(s)+\phi_{-m}(t)\,S_{|m|}(s). ϕ m ( t − s ) = 2 cos ( 2 πm t − 2 πm s ) = 2 C m ( t ) C m ( s ) + 2 S m ( t ) S m ( s ) = ϕ m ( t ) C m ( s ) + ϕ − m ( t ) S ∣ m ∣ ( s ) .
If m < 0 m<0 m < 0 : put q = − m = ∣ m ∣ q=-m=|m| q = − m = ∣ m ∣ , so 0 < q 0<q 0 < q , ϕ m ( t ) = 2 S q ( t ) \phi_{m}(t)=\sqrt{2}\,S_{q}(t) ϕ m ( t ) = 2 S q ( t ) , ϕ − m ( t ) = 2 C q ( t ) \phi_{-m}(t)=\sqrt{2}\,C_{q}(t) ϕ − m ( t ) = 2 C q ( t ) , and C m ( s ) = cos ( − 2 π q s ) = C q ( s ) C_{m}(s)=\cos(-2\pi qs)=C_{q}(s) C m ( s ) = cos ( − 2 π q s ) = C q ( s ) by Values at Zero, Parity, Addition and Double-Angle Identities for Sine and Cosine §parity ; by Addition Formulas for Sine and Cosine §sine with a = 2 π q t a=2\pi qt a = 2 π qt and x = − 2 π q s x=-2\pi qs x = − 2 π q s , and by the same parity clause,
ϕ m ( t − s ) = 2 sin ( 2 π q t − 2 π q s ) = 2 S q ( t ) C q ( s ) − 2 C q ( t ) S q ( s ) = ϕ m ( t ) C m ( s ) − ϕ − m ( t ) S ∣ m ∣ ( s ) . \phi_{m}(t-s)=\sqrt{2}\sin(2\pi qt-2\pi qs)=\sqrt{2}\,S_{q}(t)\,C_{q}(s)-\sqrt{2}\,C_{q}(t)\,S_{q}(s)=\phi_{m}(t)\,C_{m}(s)-\phi_{-m}(t)\,S_{|m|}(s). ϕ m ( t − s ) = 2 sin ( 2 π qt − 2 π q s ) = 2 S q ( t ) C q ( s ) − 2 C q ( t ) S q ( s ) = ϕ m ( t ) C m ( s ) − ϕ − m ( t ) S ∣ m ∣ ( s ) .
Step 4: eigenvectors (claim 2). Let k ∈ Γ M k\in\Gamma_{M} k ∈ Γ M and y ∈ L M y\in\mathbb{L}_{M} y ∈ L M , let w ∈ W L w\in W_{L} w ∈ W L with y = L − 1 w y=L^{-1}w y = L − 1 w , and let τ \tau τ be the bijection of Step 2(a) for this w w w . Reindexing along w ′ ↦ L − 1 w ′ w'\mapsto L^{-1}w' w ′ ↦ L − 1 w ′ and along τ \tau τ (claim 2 of Properties of a Sum over a Finite Index Set ), and using Step 2(a),
∑ z ∈ L M J ( y , z ) ψ k ( z ) = ∑ w ′ ∈ W L j ( w − w ′ ) ψ k ( L − 1 w ′ ) = ∑ u ∈ W L j ( w − τ ( u ) ) ψ k ( L − 1 τ ( u ) ) = ∑ u ∈ W L j ( u ) ψ k ( y − L − 1 u ) . \sum_{z\in\mathbb{L}_{M}}J(y,z)\,\psi_{k}(z)=\sum_{w'\in W_{L}}\mathfrak{j}(w-w')\,\psi_{k}(L^{-1}w')=\sum_{u\in W_{L}}\mathfrak{j}(w-\tau(u))\,\psi_{k}(L^{-1}\tau(u))=\sum_{u\in W_{L}}\mathfrak{j}(u)\,\psi_{k}(y-L^{-1}u). z ∈ L M ∑ J ( y , z ) ψ k ( z ) = w ′ ∈ W L ∑ j ( w − w ′ ) ψ k ( L − 1 w ′ ) = u ∈ W L ∑ j ( w − τ ( u )) ψ k ( L − 1 τ ( u )) = u ∈ W L ∑ j ( u ) ψ k ( y − L − 1 u ) .
For an integer p p p with 0 ≤ p ≤ n 0\le p\le n 0 ≤ p ≤ n , i ∈ [ n ] i\in[n] i ∈ [ n ] and u ∈ W L u\in W_{L} u ∈ W L let g p , i ( u ) = ϕ k i ( y i ) C k i ( L − 1 u i ) g_{p,i}(u)=\phi_{k_{i}}(y_{i})\,C_{k_{i}}(L^{-1}u_{i}) g p , i ( u ) = ϕ k i ( y i ) C k i ( L − 1 u i ) if i ≤ p i\le p i ≤ p and g p , i ( u ) = ϕ k i ( y i − L − 1 u i ) g_{p,i}(u)=\phi_{k_{i}}(y_{i}-L^{-1}u_{i}) g p , i ( u ) = ϕ k i ( y i − L − 1 u i ) if p < i p<i p < i , and let F p ( u ) = ∏ i = 1 n g p , i ( u ) F_{p}(u)=\prod_{i=1}^{n}g_{p,i}(u) F p ( u ) = ∏ i = 1 n g p , i ( u ) . As the i i i -th component of y − L − 1 u y-L^{-1}u y − L − 1 u is y i − L − 1 u i y_{i}-L^{-1}u_{i} y i − L − 1 u i , F 0 ( u ) = ψ k ( y − L − 1 u ) F_{0}(u)=\psi_{k}(y-L^{-1}u) F 0 ( u ) = ψ k ( y − L − 1 u ) ; and by claim 2 of Properties of Finite Products , F n ( u ) = ψ k ( y ) ∏ i = 1 n C k i ( L − 1 u i ) F_{n}(u)=\psi_{k}(y)\prod_{i=1}^{n}C_{k_{i}}(L^{-1}u_{i}) F n ( u ) = ψ k ( y ) ∏ i = 1 n C k i ( L − 1 u i ) .
Let p ∈ [ n ] p\in[n] p ∈ [ n ] . For i ≠ p i\ne p i = p one has g p − 1 , i = g p , i g_{p-1,i}=g_{p,i} g p − 1 , i = g p , i , since i ≤ p − 1 i\le p-1 i ≤ p − 1 if and only if i ≤ p i\le p i ≤ p . Let r i ( u ) = g p , i ( u ) r_{i}(u)=g_{p,i}(u) r i ( u ) = g p , i ( u ) for i ∈ [ n ] i\in[n] i ∈ [ n ] with i ≠ p i\ne p i = p , r p ( u ) = 1 r_{p}(u)=1 r p ( u ) = 1 , and Π p ( u ) = ∏ i = 1 n r i ( u ) \Pi_{p}(u)=\prod_{i=1}^{n}r_{i}(u) Π p ( u ) = ∏ i = 1 n r i ( u ) . Writing each factor g p , i ( u ) g_{p,i}(u) g p , i ( u ) , respectively g p − 1 , i ( u ) g_{p-1,i}(u) g p − 1 , i ( u ) , as r i ( u ) r_{i}(u) r i ( u ) times the number equal to 1 1 1 for i ≠ p i\ne p i = p and to g p , p ( u ) g_{p,p}(u) g p , p ( u ) , respectively g p − 1 , p ( u ) g_{p-1,p}(u) g p − 1 , p ( u ) , for i = p i=p i = p , claims 2 and 3 of Properties of Finite Products give F p ( u ) = Π p ( u ) g p , p ( u ) F_{p}(u)=\Pi_{p}(u)\,g_{p,p}(u) F p ( u ) = Π p ( u ) g p , p ( u ) and F p − 1 ( u ) = Π p ( u ) g p − 1 , p ( u ) F_{p-1}(u)=\Pi_{p}(u)\,g_{p-1,p}(u) F p − 1 ( u ) = Π p ( u ) g p − 1 , p ( u ) . By (A) with m = k p m=k_{p} m = k p , t = y p t=y_{p} t = y p and s = L − 1 u p s=L^{-1}u_{p} s = L − 1 u p ,
F p − 1 ( u ) = F p ( u ) + ϵ k p ϕ − k p ( y p ) D ( u ) , D ( u ) = Π p ( u ) S ∣ k p ∣ ( L − 1 u p ) . F_{p-1}(u)=F_{p}(u)+\epsilon_{k_{p}}\,\phi_{-k_{p}}(y_{p})\,D(u),\qquad D(u)=\Pi_{p}(u)\,S_{|k_{p}|}(L^{-1}u_{p}). F p − 1 ( u ) = F p ( u ) + ϵ k p ϕ − k p ( y p ) D ( u ) , D ( u ) = Π p ( u ) S ∣ k p ∣ ( L − 1 u p ) .
Each r i ( u ) r_{i}(u) r i ( u ) depends on u u u only through u i u_{i} u i for i ≠ p i\ne p i = p , and ( ϱ p u ) i = u i (\varrho_{p}u)_{i}=u_{i} ( ϱ p u ) i = u i for i ≠ p i\ne p i = p , so Π p ( ϱ p u ) = Π p ( u ) \Pi_{p}(\varrho_{p}u)=\Pi_{p}(u) Π p ( ϱ p u ) = Π p ( u ) ; with Step 2(b) this gives j ( ϱ p u ) D ( ϱ p u ) = − j ( u ) D ( u ) \mathfrak{j}(\varrho_{p}u)\,D(\varrho_{p}u)=-\mathfrak{j}(u)\,D(u) j ( ϱ p u ) D ( ϱ p u ) = − j ( u ) D ( u ) . Put Σ = ∑ u ∈ W L j ( u ) D ( u ) \Sigma=\sum_{u\in W_{L}}\mathfrak{j}(u)\,D(u) Σ = ∑ u ∈ W L j ( u ) D ( u ) . Reindexing along the bijection ϱ p \varrho_{p} ϱ p (claim 2 of Properties of a Sum over a Finite Index Set ) and then claim 4 of that lemma give Σ = ∑ u ∈ W L j ( ϱ p u ) D ( ϱ p u ) = ∑ u ∈ W L ( − 1 ) j ( u ) D ( u ) = − Σ \Sigma=\sum_{u\in W_{L}}\mathfrak{j}(\varrho_{p}u)\,D(\varrho_{p}u)=\sum_{u\in W_{L}}(-1)\,\mathfrak{j}(u)\,D(u)=-\Sigma Σ = ∑ u ∈ W L j ( ϱ p u ) D ( ϱ p u ) = ∑ u ∈ W L ( − 1 ) j ( u ) D ( u ) = − Σ , so Σ = 0 \Sigma=0 Σ = 0 . By claims 3 and 4 of Properties of a Sum over a Finite Index Set ,
∑ u ∈ W L j ( u ) F p − 1 ( u ) = ∑ u ∈ W L j ( u ) F p ( u ) + ϵ k p ϕ − k p ( y p ) Σ = ∑ u ∈ W L j ( u ) F p ( u ) . \sum_{u\in W_{L}}\mathfrak{j}(u)\,F_{p-1}(u)=\sum_{u\in W_{L}}\mathfrak{j}(u)\,F_{p}(u)+\epsilon_{k_{p}}\,\phi_{-k_{p}}(y_{p})\,\Sigma=\sum_{u\in W_{L}}\mathfrak{j}(u)\,F_{p}(u). u ∈ W L ∑ j ( u ) F p − 1 ( u ) = u ∈ W L ∑ j ( u ) F p ( u ) + ϵ k p ϕ − k p ( y p ) Σ = u ∈ W L ∑ j ( u ) F p ( u ) .
By induction on p p p the sum ∑ u ∈ W L j ( u ) F p ( u ) \sum_{u\in W_{L}}\mathfrak{j}(u)\,F_{p}(u) ∑ u ∈ W L j ( u ) F p ( u ) is the same for all p p p with 0 ≤ p ≤ n 0\le p\le n 0 ≤ p ≤ n . Hence, by claim 4 of Properties of a Sum over a Finite Index Set and Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols §symbol ,
∑ z ∈ L M J ( y , z ) ψ k ( z ) = ∑ u ∈ W L j ( u ) F 0 ( u ) = ∑ u ∈ W L j ( u ) F n ( u ) = ψ k ( y ) ∑ u ∈ W L j ( u ) ∏ i = 1 n C k i ( L − 1 u i ) = J ^ ( k ) ψ k ( y ) . \sum_{z\in\mathbb{L}_{M}}J(y,z)\,\psi_{k}(z)=\sum_{u\in W_{L}}\mathfrak{j}(u)\,F_{0}(u)=\sum_{u\in W_{L}}\mathfrak{j}(u)\,F_{n}(u)=\psi_{k}(y)\sum_{u\in W_{L}}\mathfrak{j}(u)\prod_{i=1}^{n}C_{k_{i}}(L^{-1}u_{i})=\hat{J}(k)\,\psi_{k}(y). z ∈ L M ∑ J ( y , z ) ψ k ( z ) = u ∈ W L ∑ j ( u ) F 0 ( u ) = u ∈ W L ∑ j ( u ) F n ( u ) = ψ k ( y ) u ∈ W L ∑ j ( u ) i = 1 ∏ n C k i ( L − 1 u i ) = J ^ ( k ) ψ k ( y ) .
Step 5: the interaction energy (claim 3). Let f f f and g g g be lattice fields and y ∈ L M y\in\mathbb{L}_{M} y ∈ L M . By Discrete Fourier Analysis on the Lattice Torus: Counting, Orthogonality, Inversion, Parseval, and the Spin Sum §inversion-formula , g ( z ) = ∑ k ∈ Γ M g ^ ( k ) ψ k ( z ) g(z)=\sum_{k\in\Gamma_{M}}\hat{g}(k)\,\psi_{k}(z) g ( z ) = ∑ k ∈ Γ M g ^ ( k ) ψ k ( z ) for z ∈ L M z\in\mathbb{L}_{M} z ∈ L M . Using claim 4 of Properties of a Sum over a Finite Index Set , the interchange of claim 5 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set , and Step 4,
∑ z ∈ L M J ( y , z ) f ( y ) g ( z ) = f ( y ) ∑ z ∈ L M ∑ k ∈ Γ M g ^ ( k ) J ( y , z ) ψ k ( z ) = f ( y ) ∑ k ∈ Γ M g ^ ( k ) ∑ z ∈ L M J ( y , z ) ψ k ( z ) = ∑ k ∈ Γ M J ^ ( k ) g ^ ( k ) f ( y ) ψ k ( y ) . \sum_{z\in\mathbb{L}_{M}}J(y,z)\,f(y)\,g(z)=f(y)\sum_{z\in\mathbb{L}_{M}}\sum_{k\in\Gamma_{M}}\hat{g}(k)\,J(y,z)\,\psi_{k}(z)=f(y)\sum_{k\in\Gamma_{M}}\hat{g}(k)\sum_{z\in\mathbb{L}_{M}}J(y,z)\,\psi_{k}(z)=\sum_{k\in\Gamma_{M}}\hat{J}(k)\,\hat{g}(k)\,f(y)\,\psi_{k}(y). z ∈ L M ∑ J ( y , z ) f ( y ) g ( z ) = f ( y ) z ∈ L M ∑ k ∈ Γ M ∑ g ^ ( k ) J ( y , z ) ψ k ( z ) = f ( y ) k ∈ Γ M ∑ g ^ ( k ) z ∈ L M ∑ J ( y , z ) ψ k ( z ) = k ∈ Γ M ∑ J ^ ( k ) g ^ ( k ) f ( y ) ψ k ( y ) .
By The Lattice Torus of Odd Side, Lattice Fields, and the Discrete Fourier Coefficients of a Lattice Field §coefficients , ∑ y ∈ L M f ( y ) ψ k ( y ) = L n f ^ ( k ) \sum_{y\in\mathbb{L}_{M}}f(y)\,\psi_{k}(y)=L^{n}\hat{f}(k) ∑ y ∈ L M f ( y ) ψ k ( y ) = L n f ^ ( k ) , as L n L − n = 1 L^{n}L^{-n}=1 L n L − n = 1 . Summing over y y y and using the same interchange and claim 4 of Properties of a Sum over a Finite Index Set again,
∑ y ∈ L M ∑ z ∈ L M J ( y , z ) f ( y ) g ( z ) = ∑ k ∈ Γ M J ^ ( k ) g ^ ( k ) ∑ y ∈ L M f ( y ) ψ k ( y ) = ∑ k ∈ Γ M J ^ ( k ) g ^ ( k ) L n f ^ ( k ) = L n ∑ k ∈ Γ M J ^ ( k ) f ^ ( k ) g ^ ( k ) . \sum_{y\in\mathbb{L}_{M}}\sum_{z\in\mathbb{L}_{M}}J(y,z)\,f(y)\,g(z)=\sum_{k\in\Gamma_{M}}\hat{J}(k)\,\hat{g}(k)\sum_{y\in\mathbb{L}_{M}}f(y)\,\psi_{k}(y)=\sum_{k\in\Gamma_{M}}\hat{J}(k)\,\hat{g}(k)\,L^{n}\hat{f}(k)=L^{n}\sum_{k\in\Gamma_{M}}\hat{J}(k)\,\hat{f}(k)\,\hat{g}(k). y ∈ L M ∑ z ∈ L M ∑ J ( y , z ) f ( y ) g ( z ) = k ∈ Γ M ∑ J ^ ( k ) g ^ ( k ) y ∈ L M ∑ f ( y ) ψ k ( y ) = k ∈ Γ M ∑ J ^ ( k ) g ^ ( k ) L n f ^ ( k ) = L n k ∈ Γ M ∑ J ^ ( k ) f ^ ( k ) g ^ ( k ) .
Step 6: the bound (claim 4). Let k ∈ Γ M k\in\Gamma_{M} k ∈ Γ M and u ∈ W L u\in W_{L} u ∈ W L . By Cell Integrals of the Trigonometric Monomials §calculus , ∣ C k i ( L − 1 u i ) ∣ ≤ 1 |C_{k_{i}}(L^{-1}u_{i})|\le1 ∣ C k i ( L − 1 u i ) ∣ ≤ 1 for every i ∈ [ n ] i\in[n] i ∈ [ n ] . By induction on j ∈ [ n ] j\in[n] j ∈ [ n ] , using the recursion of claim 1 of Properties of Finite Products and ∣ α β ∣ = ∣ α ∣ ∣ β ∣ |\alpha\beta|=|\alpha|\,|\beta| ∣ α β ∣ = ∣ α ∣ ∣ β ∣ for α , β ∈ R \alpha,\beta\in\mathbb{R} α , β ∈ R , one has ∣ ∏ i = 1 j C k i ( L − 1 u i ) ∣ ≤ 1 \bigl|\prod_{i=1}^{j}C_{k_{i}}(L^{-1}u_{i})\bigr|\le1 ∏ i = 1 j C k i ( L − 1 u i ) ≤ 1 : for j = 1 j=1 j = 1 the product is C k 1 ( L − 1 u 1 ) C_{k_{1}}(L^{-1}u_{1}) C k 1 ( L − 1 u 1 ) , and if the bound holds for j j j with j + 1 ≤ n j+1\le n j + 1 ≤ n , then the product up to j + 1 j+1 j + 1 has absolute value ∣ ∏ i = 1 j C k i ( L − 1 u i ) ∣ ∣ C k j + 1 ( L − 1 u j + 1 ) ∣ ≤ 1 ⋅ 1 \bigl|\prod_{i=1}^{j}C_{k_{i}}(L^{-1}u_{i})\bigr|\,|C_{k_{j+1}}(L^{-1}u_{j+1})|\le1\cdot1 ∏ i = 1 j C k i ( L − 1 u i ) ∣ C k j + 1 ( L − 1 u j + 1 ) ∣ ≤ 1 ⋅ 1 . Hence ∣ j ( u ) ∏ i = 1 n C k i ( L − 1 u i ) ∣ ≤ ∣ j ( u ) ∣ \bigl|\mathfrak{j}(u)\prod_{i=1}^{n}C_{k_{i}}(L^{-1}u_{i})\bigr|\le|\mathfrak{j}(u)| j ( u ) ∏ i = 1 n C k i ( L − 1 u i ) ≤ ∣ j ( u ) ∣ , and by Reflection-Symmetric Periodic Pair Interactions on the Lattice Torus, Their Interaction Matrices and Their Symbols §symbol , Real Sums over a Finite Index Set: Comparison, Nonnegativity, Monotonicity, Term Bounds, Absolute Values, Counting and Limits §absolute and Real Sums over a Finite Index Set: Comparison, Nonnegativity, Monotonicity, Term Bounds, Absolute Values, Counting and Limits §comparison ,
∣ J ^ ( k ) ∣ ≤ ∑ u ∈ W L ∣ j ( u ) ∏ i = 1 n C k i ( L − 1 u i ) ∣ ≤ ∑ u ∈ W L ∣ j ( u ) ∣ . |\hat{J}(k)|\le\sum_{u\in W_{L}}\Bigl|\mathfrak{j}(u)\prod_{i=1}^{n}C_{k_{i}}(L^{-1}u_{i})\Bigr|\le\sum_{u\in W_{L}}|\mathfrak{j}(u)|. ∣ J ^ ( k ) ∣ ≤ u ∈ W L ∑ j ( u ) i = 1 ∏ n C k i ( L − 1 u i ) ≤ u ∈ W L ∑ ∣ j ( u ) ∣.