TheoremBase

Along a mode the diagonal lattice sum changes by a one-term finitely supported family and the cylindrical part moves along a coordinate line, which gives the mode derivatives; the generator and gradient-energy families then split into a part dominated by summable weight families plus a finitely supported part.

Proof

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

The data η\eta, mm, k1,…,kmk_{1},\dots,k_{m} and ff are those of one witness of the regularity of ψ\psi in Regular Functions for the Wick-Square Problem: Diagonal Quadratics of Curvature of Order the Inverse Squared Weight, plus Twice Differentiable Functions on the Sobolev Space of Order -3; we fix a curvature bound C≥0C\ge0 of η\eta as in Regular Functions for the Wick-Square Problem: Diagonal Quadratics of Curvature of Order the Inverse Squared Weight, plus Twice Differentiable Functions on the Sobolev Space of Order -3 §curvature, so that μk2∣η(k)∣≤C\mu_{k}^{2}|\eta(k)|\le C for every mode kk, and for every y∈H−1y\in H^{-1}

ψ(y)=Q(y)+f(ξ(y)),Q(y)=∑k∈Znη(k) y(k)2,\psi(y)=Q(y)+f(\xi(y)),\qquad Q(y)=\sum_{k\in\mathbb{Z}^{n}}\eta(k)\,y(k)^{2},

the lattice sum Q(y)Q(y) existing as recorded in that definition. Let E={k1,…,km}E=\{k_{1},\dots,k_{m}\}. Natural numbers start at 11, so [m][m] is nonempty and finite, and since the modes kik_{i} are pairwise distinct the map i↦kii\mapsto k_{i} is a bijection from [m][m] onto EE; in particular EE is a nonempty finite subset of Zn\mathbb{Z}^{n}.

Preliminary facts.

(P1) Real arithmetic. For a real yy, ∣y∣|y| equals yy or −y-y and 0≤∣y∣0\le|y| (claim 1 of Properties of the Absolute Value in an Ordered Field); since (−y)(−y)=yy(-y)(-y)=yy (claim 2 of Zero Products and Elementary Identities in a Field), y2=∣y∣ ∣y∣y^{2}=|y|\,|y|, and 0≤∣y∣ ∣y∣0\le|y|\,|y| by claim 5 of Elementary Arithmetic in an Ordered Field (multiply 0≤∣y∣0\le|y| by ∣y∣|y|). Thus 0≤y20\le y^{2} and ∣y2∣=y2|y^{2}|=y^{2}. If 0≤y0\le y, then ∣y∣=y|y|=y: otherwise ∣y∣=−y|y|=-y with 0≤−y0\le -y and 0≤y0\le y, forcing y=0y=0 by claim 4 of Elementary Order Arithmetic in an Ordered Field and antisymmetry.

(P2) Weights. By The Wick-Square Problem on the Torus: Standing Notation §modes, 1≤μk1\le\mu_{k}, so 0<μk0<\mu_{k} (claims 6 and 2 of Elementary Order Arithmetic in an Ordered Field); 0<μk2=μkμk0<\mu_{k}^{2}=\mu_{k}\mu_{k} by claim 5 of that lemma, and the inverses μk−1\mu_{k}^{-1}, μk−2\mu_{k}^{-2} exist and are positive by its claim 7. Multiplying 1≤μk1\le\mu_{k} by μk−2≥0\mu_{k}^{-2}\ge0 (claim 5 of Elementary Arithmetic in an Ordered Field) gives 1μk2≤1μk\frac{1}{\mu_{k}^{2}}\le\frac{1}{\mu_{k}}.

(P3) Curvature. Multiplying μk2∣η(k)∣≤C\mu_{k}^{2}|\eta(k)|\le C by μk−2≥0\mu_{k}^{-2}\ge0, respectively by μk−1≥0\mu_{k}^{-1}\ge0 (claim 5 of Elementary Arithmetic in an Ordered Field), gives for every mode kk

∣η(k)∣≤Cμk2,μk ∣η(k)∣≤Cμk.|\eta(k)|\le\frac{C}{\mu_{k}^{2}},\qquad\mu_{k}\,|\eta(k)|\le\frac{C}{\mu_{k}} .

(P4) The diagonal part along a mode. Let y∈H−1y\in H^{-1}, k∈Znk\in\mathbb{Z}^{n} and t∈Rt\in\mathbb{R}; then y+tek∈H−1y+te_{k}\in H^{-1} by The Wick-Square Problem on the Torus: Standing Notation §units, and pointwise (y+tek)(k′)=y(k′)(y+te_{k})(k')=y(k') for k′≠kk'\ne k and (y+tek)(k)=y(k)+t(y+te_{k})(k)=y(k)+t. Let a(k′)=η(k′)y(k′)2a(k')=\eta(k')y(k')^{2}, b(k′)=η(k′)(y+tek)(k′)2b(k')=\eta(k')(y+te_{k})(k')^{2} and δ=b−a\delta=b-a. Both aa and bb are cube-summable (the families whose lattice sums are Q(y)Q(y) and Q(y+tek)Q(y+te_{k})), δ(k′)=0\delta(k')=0 for k′≠kk'\ne k, and by claim 5 of Zero Products and Elementary Identities in a Field

δ(k)=η(k)((y(k)+t)2−y(k)2)=2η(k)y(k) t+η(k) t2.\delta(k)=\eta(k)\bigl((y(k)+t)^{2}-y(k)^{2}\bigr)=2\eta(k)y(k)\,t+\eta(k)\,t^{2}.

By Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §finite-support with the set {k}\{k\}, δ\delta is cube-summable with lattice sum ∑k′∈{k}δ(k′)=δ(k)\sum_{k'\in\{k\}}\delta(k')=\delta(k) (Sum over a Finite Index Set, with the bijection from [1][1] sending 11 to kk). Since b=a+δb=a+\delta, Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §linear gives

Q(y+tek)=Q(y)+2η(k)y(k) t+η(k) t2.Q(y+te_{k})=Q(y)+2\eta(k)y(k)\,t+\eta(k)\,t^{2}.

(P5) Coordinate lines. Let x∈H−1x\in H^{-1}, i∈[m]i\in[m], and let G:Rm→RG:\mathbb{R}^{m}\to\mathbb{R} be a function whose partial derivative with respect to the iith variable (Partial Derivative on a Euclidean Open Set) exists at every point of Rm\mathbb{R}^{m}. Put g(t)=G(ξ(x+teki))g(t)=G(\xi(x+te_{k_{i}})) for t∈Rt\in\mathbb{R}. Then gg is differentiable at every t0∈Rt_{0}\in\mathbb{R}, in the sense of Single-Variable Calculus on an Interval §derivative with I=RI=\mathbb{R}, and g′(t0)=∂iG(ξ(x+t0eki))g'(t_{0})=\partial_{i}G(\xi(x+t_{0}e_{k_{i}})). Indeed, by (P4) and the distinctness of the modes, the jjth component of ξ(x+teki)\xi(x+te_{k_{i}}) is x(kj)x(k_{j}) for j≠ij\ne i and x(ki)+tx(k_{i})+t for j=ij=i. Hence, for a=ξ(x+t0eki)a=\xi(x+t_{0}e_{k_{i}}) and h∈Rh\in\mathbb{R}, the point (a1,…,ai−1,ai+h,ai+1,…,am)(a_{1},\dots,a_{i-1},a_{i}+h,a_{i+1},\dots,a_{m}) equals ξ(x+(t0+h)eki)\xi(x+(t_{0}+h)e_{k_{i}}) by claim 1 of Euclidean Points as Tuples of Real Numbers, it lies in Rm\mathbb{R}^{m}, and the difference quotient in Partial Derivative on a Euclidean Open Set at aa with increment hh is (g(t0+h)−g(t0))/h(g(t_{0}+h)-g(t_{0}))/h. So the defining condition of the partial derivative at aa with value ∂iG(a)\partial_{i}G(a) is exactly the defining condition of differentiability of gg at t0t_{0} with derivative ∂iG(a)\partial_{i}G(a) in Single-Variable Calculus on an Interval §derivative (every point of R\mathbb{R} being an interior point of R\mathbb{R}, as recorded in Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes), and the derivative is unique by that clause.

Clause 1. Fix x∈H−1x\in H^{-1} and k∈Znk\in\mathbb{Z}^{n}, and let σ(t)=ψ(x+tek)\sigma(t)=\psi(x+te_{k}) be the section of Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes. By (P4), σ=p+g\sigma=p+g with

p(t)=Q(x)+2η(k)x(k) t+η(k) t2,g(t)=f(ξ(x+tek)).p(t)=Q(x)+2\eta(k)x(k)\,t+\eta(k)\,t^{2},\qquad g(t)=f(\xi(x+te_{k})).

By the constant, power, sum and constant-multiple rules of Single-Variable Calculus on an Interval §derivative (the map t↦tt\mapsto t has derivative 11 and t↦t2t\mapsto t^{2} has derivative 2t2t), pp is differentiable at every tt with p′(t)=2η(k)x(k)+2η(k) tp'(t)=2\eta(k)x(k)+2\eta(k)\,t, and p′p' is differentiable at 00 with (p′)′(0)=2η(k)(p')'(0)=2\eta(k).

Suppose first that k=kik=k_{i} for some i∈[m]i\in[m], unique by distinctness. Since ff is of class C2=C1+1C^{2}=C^{1+1} on Rm\mathbb{R}^{m}, clause 2 of C^k Maps on a Euclidean Open Set (read with the scalar convention of its clause 3) shows that ff is of class C1C^{1}, so its partial derivative with respect to the iith variable exists at every point, and that ∂if\partial_{i}f is of class C1C^{1}, so the partial derivative of ∂if\partial_{i}f with respect to the iith variable exists at every point, its value being ∂i∂if\partial_{i}\partial_{i}f by clause 4 of that definition. By (P5) with G=fG=f, gg is differentiable at every tt with g′(t)=∂if(ξ(x+teki))g'(t)=\partial_{i}f(\xi(x+te_{k_{i}})); by (P5) with G=∂ifG=\partial_{i}f, g′g' is differentiable at 00 with (g′)′(0)=∂i∂if(ξ(x))(g')'(0)=\partial_{i}\partial_{i}f(\xi(x)). Also g′(0)=∂if(ξ(x))g'(0)=\partial_{i}f(\xi(x)). These are dk(x)d_{k}(x) and dk2(x)d_{k}^{2}(x).

Suppose next that k∉Ek\notin E. Then kj≠kk_{j}\ne k for every jj, so by (P4) ξ(x+tek)=ξ(x)\xi(x+te_{k})=\xi(x) for every tt (claim 1 of Euclidean Points as Tuples of Real Numbers); gg is constant, hence differentiable everywhere with g′g' the zero function, whose derivative at 00 is 00 (constant rule of Single-Variable Calculus on an Interval §derivative). Thus again g′(0)=0=dk(x)g'(0)=0=d_{k}(x) and (g′)′(0)=0=dk2(x)(g')'(0)=0=d_{k}^{2}(x).

In both cases the sum rule of Single-Variable Calculus on an Interval §derivative shows that σ\sigma is differentiable at every point with σ′=p′+g′\sigma'=p'+g', and that σ′\sigma' is differentiable at 00 with (σ′)′(0)=2η(k)+dk2(x)(\sigma')'(0)=2\eta(k)+d_{k}^{2}(x). So ψ\psi is twice differentiable along the modes (Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §twice), and by Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes §derivatives

∂kψ(x)=σ′(0)=2η(k)x(k)+dk(x),∂k2ψ(x)=2η(k)+dk2(x).\partial_{k}\psi(x)=\sigma'(0)=2\eta(k)x(k)+d_{k}(x),\qquad\partial_{k}^{2}\psi(x)=2\eta(k)+d_{k}^{2}(x).

Clause 2. Fix x∈H−1x\in H^{-1} and define families on Zn\mathbb{Z}^{n} by

h(k)=ν η(k)−2μkη(k) x(k)2,e(k)=ν2 dk2(x)−μk x(k) dk(x).h(k)=\nu\,\eta(k)-2\mu_{k}\eta(k)\,x(k)^{2},\qquad e(k)=\frac{\nu}{2}\,d_{k}^{2}(x)-\mu_{k}\,x(k)\,d_{k}(x).

Cube-summability of hh: by claims 5 and 2 of Properties of the Absolute Value in an Ordered Field, claim 4 of that lemma, (P1) (for the nonnegative numbers ν\nu, 2μk2\mu_{k}, x(k)2x(k)^{2}) and (P3), with claim 5 of Elementary Arithmetic in an Ordered Field for scaling and claims 3 and 2 of that lemma for adding two inequalities,

∣h(k)∣≤ν ∣η(k)∣+2x(k)2 μk∣η(k)∣≤νC 1μk2+2C x(k)2μk=:B(k).|h(k)|\le\nu\,|\eta(k)|+2x(k)^{2}\,\mu_{k}|\eta(k)|\le\nu C\,\frac{1}{\mu_{k}^{2}}+2C\,\frac{x(k)^{2}}{\mu_{k}}=:B(k).

Since n≤3<4=2⋅2n\le3<4=2\cdot2 (The Wick-Square Problem on the Torus: Standing Notation §dimension), k↦1/μk2k\mapsto1/\mu_{k}^{2} is cube-summable by Lattice Sums of the Fourier Weights of the Torus, and the Negative Sobolev Norms as Lattice Sums §convergent with s=2s=2; k↦x(k)2/μkk\mapsto x(k)^{2}/\mu_{k} is cube-summable by The Wick-Square Problem on the Torus: Standing Notation §state-space; so BB is cube-summable by Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §linear, and hh is cube-summable by Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §comparison. This is the first assertion of clause 2.

The family ee vanishes off EE: for k∉Ek\notin E, dk(x)=dk2(x)=0d_{k}(x)=d_{k}^{2}(x)=0, and products with 00 vanish (claim 1 of Zero Products and Elementary Identities in a Field). By Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §finite-support, ee is cube-summable with lattice sum ∑k∈Ee(k)\sum_{k\in E}e(k), and by claims 2 and 1 of Properties of a Sum over a Finite Index Set along the bijection i↦kii\mapsto k_{i} from [m][m] onto EE,

∑k∈Zne(k)=∑i=1me(ki)=∑i=1m(ν2 ∂i∂if(ξ(x))−μki x(ki) ∂if(ξ(x))),\sum_{k\in\mathbb{Z}^{n}}e(k)=\sum_{i=1}^{m}e(k_{i})=\sum_{i=1}^{m}\Bigl(\frac{\nu}{2}\,\partial_{i}\partial_{i}f(\xi(x))-\mu_{k_{i}}\,x(k_{i})\,\partial_{i}f(\xi(x))\Bigr),

because dki(x)=∂if(ξ(x))d_{k_{i}}(x)=\partial_{i}f(\xi(x)) and dki2(x)=∂i∂if(ξ(x))d_{k_{i}}^{2}(x)=\partial_{i}\partial_{i}f(\xi(x)).

By clause 1 and field arithmetic, for every mode kk

ν2 ∂k2ψ(x)−μk x(k) ∂kψ(x)=ν2(2η(k)+dk2(x))−μkx(k)(2η(k)x(k)+dk(x))=h(k)+e(k),\frac{\nu}{2}\,\partial_{k}^{2}\psi(x)-\mu_{k}\,x(k)\,\partial_{k}\psi(x)=\frac{\nu}{2}\bigl(2\eta(k)+d_{k}^{2}(x)\bigr)-\mu_{k}x(k)\bigl(2\eta(k)x(k)+d_{k}(x)\bigr)=h(k)+e(k),

so by The Free-Field Generator with a Mode Cutoff §generator and Cube Sums of Families on the Integer Lattice §cube-sums, LNψ(x)L_{N}\psi(x) is the NNth cube sum of h+eh+e for every N∈NN\in\mathbb{N}. By Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §linear, h+eh+e is cube-summable, that is (Cube Sums of Families on the Integer Lattice §lattice-sum) the sequence (LNψ(x))N∈N(L_{N}\psi(x))_{N\in\mathbb{N}} converges, to ∑k∈Znh(k)+∑k∈Zne(k)\sum_{k\in\mathbb{Z}^{n}}h(k)+\sum_{k\in\mathbb{Z}^{n}}e(k). Hence LψL\psi is defined at xx (The Free-Field Generator with a Mode Cutoff §limit) and Lψ(x)L\psi(x) is the displayed expression of clause 2.

Clause 3. By The Cutoff Wick Square is Minus the Free-Field Generator of the Gaussian Penalty, and the Penalised Form of the Cutoff Operators §derivatives, PP is twice differentiable along the modes with ∂kP(x)=βx(k)/μk\partial_{k}P(x)=\beta x(k)/\mu_{k}. Fix x∈H−1x\in H^{-1} and k∈Znk\in\mathbb{Z}^{n}. The section of φ\varphi at xx along kk is t↦θP(x+tek)+ψ(x+tek)t\mapsto\theta P(x+te_{k})+\psi(x+te_{k}), that is, θ\theta times the section of PP plus the section of ψ\psi. Both sections are differentiable at every point and their derivatives are differentiable at 00 (by The Cutoff Wick Square is Minus the Free-Field Generator of the Gaussian Penalty, and the Penalised Form of the Cutoff Operators §derivatives and clause 1), so by the sum and constant-multiple rules of Single-Variable Calculus on an Interval §derivative the section of φ\varphi is differentiable everywhere, its derivative is θ\theta times that of the section of PP plus that of the section of ψ\psi, and this derivative is differentiable at 00. So φ\varphi is twice differentiable along the modes, and by clause 1

∂kφ(x)=λk x(k)+dk(x),λk=θβμk+2η(k).\partial_{k}\varphi(x)=\lambda_{k}\,x(k)+d_{k}(x),\qquad\lambda_{k}=\frac{\theta\beta}{\mu_{k}}+2\eta(k).

Let w(k)=(∂kφ(x))2w(k)=(\partial_{k}\varphi(x))^{2}, w1(k)=(λkx(k))2w_{1}(k)=(\lambda_{k}x(k))^{2} and w2=w−w1w_{2}=w-w_{1}. For k∉Ek\notin E, dk(x)=0d_{k}(x)=0, so w(k)=w1(k)w(k)=w_{1}(k) and w2(k)=0w_{2}(k)=0; by Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §finite-support, w2w_{2} is cube-summable.

Put K=∣θ∣β+2CK=|\theta|\beta+2C; 0≤K0\le K by (P1), claim 5 of Elementary Arithmetic in an Ordered Field and claim 2 of that lemma. By claims 5 and 4 of Properties of the Absolute Value in an Ordered Field, (P1), (P3) and (P2) (with claim 5 of Elementary Arithmetic in an Ordered Field),

∣λk∣≤∣θ∣βμk+2Cμk2≤∣θ∣βμk+2Cμk=Kμk.|\lambda_{k}|\le\frac{|\theta|\beta}{\mu_{k}}+\frac{2C}{\mu_{k}^{2}}\le\frac{|\theta|\beta}{\mu_{k}}+\frac{2C}{\mu_{k}}=\frac{K}{\mu_{k}} .

Both ∣λk∣|\lambda_{k}| and K/μkK/\mu_{k} are nonnegative, so claim 2 of Monotonicity of Squaring on the Nonnegative Elements of an Ordered Field gives ∣λk∣2≤K2/μk2|\lambda_{k}|^{2}\le K^{2}/\mu_{k}^{2}; multiplying by x(k)2≥0x(k)^{2}\ge0 (claim 5 of Elementary Arithmetic in an Ordered Field) and using ∣λk∣2=λk2|\lambda_{k}|^{2}=\lambda_{k}^{2} from (P1) and then (P2),

0≤w1(k)=λk2 x(k)2≤K2 x(k)2μk2≤K2 x(k)2μk.0\le w_{1}(k)=\lambda_{k}^{2}\,x(k)^{2}\le K^{2}\,\frac{x(k)^{2}}{\mu_{k}^{2}}\le K^{2}\,\frac{x(k)^{2}}{\mu_{k}} .

Here w1(k)≥0w_{1}(k)\ge0 by (P1), so ∣w1(k)∣=w1(k)|w_{1}(k)|=w_{1}(k). The right-hand family is cube-summable by The Wick-Square Problem on the Torus: Standing Notation §state-space and Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §linear, so w1w_{1} is cube-summable by Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §comparison. Then w=w1+w2w=w_{1}+w_{2} is cube-summable by Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §linear. Since ∣DNφ(x)∣2|D_{N}\varphi(x)|^{2} is the NNth cube sum of ww (The Gradient Energy with a Mode Cutoff §gradient, Cube Sums of Families on the Integer Lattice §cube-sums), the sequence (∣DNφ(x)∣2)N∈N(|D_{N}\varphi(x)|^{2})_{N\in\mathbb{N}} converges (Cube Sums of Families on the Integer Lattice §lattice-sum), and ∣Dφ(x)∣2|D\varphi(x)|^{2} is defined by The Gradient Energy with a Mode Cutoff §limit.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…