TheoremBase

Along a mode the penalty is the quadratic polynomial P(x) + (beta x(k)/mu_k) t + (beta/(2 muk))mu_k)) t2t^2, since the unit family changes the lattice sum in one term only, and its derivatives give the mode derivatives. The generator summands then equal -beta(x(k)2x(k)^2 - ck)c_k), and linearity of the mode derivatives and of finite sums turns the cutoff operators into the penalised form.

Proof

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

Order and arithmetic in R\mathbb{R} are handled with Elementary Order Arithmetic in an Ordered Field and Zero Products and Elementary Identities in a Field, whose claims are cited by number. For every mode kk the weight μk\mu_{k} is positive, since 1≤μk1\le\mu_{k} by The Wick-Square Problem on the Torus: Standing Notation §modes and 0<10<1 by claim 6 of Elementary Order Arithmetic in an Ordered Field; so 1μk\frac{1}{\mu_{k}} is defined by claim 7 of that lemma. Differentiability of a function R→R\mathbb{R}\to\mathbb{R} is that of Single-Variable Calculus on an Interval §derivative with I=RI=\mathbb{R}, as fixed in Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes; every point of R\mathbb{R} is an interior point of R\mathbb{R}, and the restriction of a map R→R\mathbb{R}\to\mathbb{R} to I=RI=\mathbb{R} is the map itself, so the rules of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives and Derivative of a Polynomial Function on the Real Line (the ones named in Single-Variable Calculus on an Interval §derivative) apply at every point.

Clause 1. Fix x∈H−1x\in H^{-1} and a mode kk, and let t∈Rt\in\mathbb{R}. Put y=x+teky=x+te_{k}, which lies in H−1H^{-1} by The Wick-Square Problem on the Torus: Standing Notation §units; the operations are the pointwise ones of Map(Zn,R)\mathrm{Map}(\mathbb{Z}^{n},\mathbb{R}) (The Wick-Square Problem on the Torus: Standing Notation §state-space), so y(k′)=x(k′)+tek(k′)y(k')=x(k')+te_{k}(k') for every mode k′k'. Hence y(k′)=x(k′)y(k')=x(k') for k′≠kk'\ne k, and y(k)=x(k)+ty(k)=x(k)+t, so that y(k)2=x(k)2+2t x(k)+t2y(k)^{2}=x(k)^{2}+2t\,x(k)+t^{2} by claim 5 of Zero Products and Elementary Identities in a Field. Define the families

u(k′)=x(k′)2μk′,dt(k′)=y(k′)2μk′−x(k′)2μk′(k′∈Zn).u(k')=\frac{x(k')^{2}}{\mu_{k'}},\qquad d_{t}(k')=\frac{y(k')^{2}}{\mu_{k'}}-\frac{x(k')^{2}}{\mu_{k'}}\qquad(k'\in\mathbb{Z}^{n}).

Then dt(k′)=0d_{t}(k')=0 for k′≠kk'\ne k and dt(k)=2t x(k)+t2μkd_{t}(k)=\frac{2t\,x(k)+t^{2}}{\mu_{k}}. The set {k}\{k\} is nonempty and finite, so by Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §finite-support the family dtd_{t} is cube-summable with lattice sum ∑k′∈{k}dt(k′)=dt(k)\sum_{k'\in\{k\}}d_{t}(k')=d_{t}(k), the last equality by claim 1 of Peeling, Splitting, and Interchange for Sums over a Finite Index Set. By The Wick-Square Problem on the Torus: Standing Notation §state-space, the family uu is cube-summable with lattice sum ∣x∣H−12|x|_{H^{-1}}^{2}, and the family k′↦y(k′)2μk′=u(k′)+dt(k′)k'\mapsto\frac{y(k')^{2}}{\mu_{k'}}=u(k')+d_{t}(k') is cube-summable with lattice sum ∣y∣H−12|y|_{H^{-1}}^{2}. By Lattice Sums along Cubes: Linearity, Nonnegative Families, Absolute Summability, Comparison and Finitely Supported Families §linear (and uniqueness of limits, claim 1 of Uniqueness of Limits and Boundedness of Convergent Real Sequences) we get ∣y∣H−12=∣x∣H−12+dt(k)|y|_{H^{-1}}^{2}=|x|_{H^{-1}}^{2}+d_{t}(k). By The Gaussian Penalty of the Wick-Square Problem this means that the section σ=σx,k\sigma=\sigma_{x,k} of PP at xx along kk (Derivatives of a Function on the Sobolev Space of Order -1 along the Fourier Modes) is

σ(t)=P(x+tek)=β2(∣x∣H−12+2t x(k)+t2μk)=A+B t+C t2(t∈R),\sigma(t)=P(x+te_{k})=\frac{\beta}{2}\Bigl(|x|_{H^{-1}}^{2}+\frac{2t\,x(k)+t^{2}}{\mu_{k}}\Bigr)=A+B\,t+C\,t^{2}\qquad(t\in\mathbb{R}),

with the constants A=P(x)A=P(x), B=β x(k)μkB=\frac{\beta\,x(k)}{\mu_{k}} and C=β2μkC=\frac{\beta}{2\mu_{k}} (field arithmetic).

Now t2=tS(1)t^{2}=t^{S(1)} and t=t1t=t^{1} by claim 1 of Properties of Natural Number Powers in a Field. By claim 1 of Derivative of a Polynomial Function on the Real Line, at every t0∈Rt_{0}\in\mathbb{R} the map t↦tt\mapsto t is differentiable with derivative 11 and the map t↦t2t\mapsto t^{2} is differentiable with derivative 2t02t_{0}, where 22 is the image of S(1)S(1) under the canonical map, equal to 1+11+1 by claim 1 of Properties of the Canonical Map from the Natural Numbers to an Ordered Field. By claims 1 and 2 of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives (constants, sums and constant multiples), σ\sigma is differentiable at every t0∈Rt_{0}\in\mathbb{R} with

σ′(t0)=B+2C t0.\sigma'(t_{0})=B+2C\,t_{0} .

Applying the same two claims to the function σ′:t0↦B+2C t0\sigma':t_{0}\mapsto B+2C\,t_{0}, it is differentiable at 00 (indeed everywhere) with derivative 2C=βμk2C=\frac{\beta}{\mu_{k}}. Since xx and kk were arbitrary, PP is twice differentiable along the modes in the sense of 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

∂kP(x)=σ′(0)=B=β x(k)μk,∂k2P(x)=(σ′)′(0)=βμk.\partial_{k}P(x)=\sigma'(0)=B=\frac{\beta\,x(k)}{\mu_{k}},\qquad\partial_{k}^{2}P(x)=(\sigma')'(0)=\frac{\beta}{\mu_{k}} .

Clause 2. Let N∈NN\in\mathbb{N} and x∈H−1x\in H^{-1}. By clause 1, LNP(x)L_{N}P(x) is defined, and by The Free-Field Generator with a Mode Cutoff §generator and clause 1,

LNP(x)=∑k∈ΓN(ν2⋅βμk−μk x(k) β x(k)μk).L_{N}P(x)=\sum_{k\in\Gamma_{N}}\Bigl(\frac{\nu}{2}\cdot\frac{\beta}{\mu_{k}}-\mu_{k}\,x(k)\,\frac{\beta\,x(k)}{\mu_{k}}\Bigr).

For each mode kk, field arithmetic gives ν2⋅βμk=β ν2μk=βck\frac{\nu}{2}\cdot\frac{\beta}{\mu_{k}}=\beta\,\frac{\nu}{2\mu_{k}}=\beta c_{k} by The Free-Field Variances of the Fourier Modes §variances, and μk x(k) βx(k)μk=β x(k)2\mu_{k}\,x(k)\,\frac{\beta x(k)}{\mu_{k}}=\beta\,x(k)^{2} because μk1μk=1\mu_{k}\frac{1}{\mu_{k}}=1; so, by claim 2 of Zero Products and Elementary Identities in a Field and distributivity, the summand equals −β(x(k)2−ck)-\beta\bigl(x(k)^{2}-c_{k}\bigr). By claim 4 of Properties of a Sum over a Finite Index Set (homogeneity, with λ=−β\lambda=-\beta) and The Wick Square with a Mode Cutoff and the Wick Domain §cutoff,

LNP(x)=−β∑k∈ΓN(x(k)2−ck)=−β :x2:N.L_{N}P(x)=-\beta\sum_{k\in\Gamma_{N}}\bigl(x(k)^{2}-c_{k}\bigr)=-\beta\,{:}x^{2}{:}_{N}.

Clause 3 (differentiability of φ−P\varphi-P). Let φ\varphi be twice differentiable along the modes, and fix x∈H−1x\in H^{-1} and a mode kk. The section of φ−P\varphi-P at xx along kk is t↦φ(x+tek)−P(x+tek)t\mapsto\varphi(x+te_{k})-P(x+te_{k}), that is σφ+(−1)σP\sigma^{\varphi}+(-1)\sigma^{P}, where σφ\sigma^{\varphi} and σP\sigma^{P} are the sections of φ\varphi and PP. Both are differentiable at every point of R\mathbb{R} (by hypothesis, and by clause 1), so by claim 2 of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives the section of φ−P\varphi-P is differentiable at every point with derivative (σφ)′+(−1)(σP)′(\sigma^{\varphi})'+(-1)(\sigma^{P})'. Both derivatives are differentiable at 00 (by hypothesis, and by clause 1), so by the same claim this derivative is differentiable at 00. Hence φ−P\varphi-P 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

∂k(φ−P)(x)=∂kφ(x)−∂kP(x),∂k2(φ−P)(x)=∂k2φ(x)−∂k2P(x).\partial_{k}(\varphi-P)(x)=\partial_{k}\varphi(x)-\partial_{k}P(x),\qquad\partial_{k}^{2}(\varphi-P)(x)=\partial_{k}^{2}\varphi(x)-\partial_{k}^{2}P(x).

Clause 3 (the penalised form). Let N∈NN\in\mathbb{N}, x∈H−1x\in H^{-1} and b:Zn→Rb:\mathbb{Z}^{n}\to\mathbb{R}. By The Free-Field Generator with a Mode Cutoff §generator, the last display, distributivity, and claims 3 and 4 of Properties of a Sum over a Finite Index Set (additivity, and homogeneity with λ=−1\lambda=-1),

LN(φ−P)(x)=∑k∈ΓN(ν2∂k2φ(x)−μkx(k)∂kφ(x))−∑k∈ΓN(ν2∂k2P(x)−μkx(k)∂kP(x))=LNφ(x)−LNP(x).L_{N}(\varphi-P)(x)=\sum_{k\in\Gamma_{N}}\Bigl(\frac{\nu}{2}\partial_{k}^{2}\varphi(x)-\mu_{k}x(k)\partial_{k}\varphi(x)\Bigr)-\sum_{k\in\Gamma_{N}}\Bigl(\frac{\nu}{2}\partial_{k}^{2}P(x)-\mu_{k}x(k)\partial_{k}P(x)\Bigr)=L_{N}\varphi(x)-L_{N}P(x).

With clause 2 this gives −LNφ(x)=−LN(φ−P)(x)+β∑k∈ΓN(x(k)2−ck)-L_{N}\varphi(x)=-L_{N}(\varphi-P)(x)+\beta\sum_{k\in\Gamma_{N}}\bigl(x(k)^{2}-c_{k}\bigr), using The Wick Square with a Mode Cutoff and the Wick Domain §cutoff. Substituting into The Cutoff Hamilton-Jacobi-Bellman Operators of the Wick-Square Problem, with a General Counterterm §operator,

FNb[φ](x)=γ φ(x)−LN(φ−P)(x)+12 ∣DNφ(x)∣2+β∑k∈ΓN(x(k)2−ck)−β∑k∈ΓN(x(k)2−b(k))−g(x).F_{N}^{b}[\varphi](x)=\gamma\,\varphi(x)-L_{N}(\varphi-P)(x)+\tfrac12\,|D_{N}\varphi(x)|^{2}+\beta\sum_{k\in\Gamma_{N}}\bigl(x(k)^{2}-c_{k}\bigr)-\beta\sum_{k\in\Gamma_{N}}\bigl(x(k)^{2}-b(k)\bigr)-g(x).

By claims 3 and 4 of Properties of a Sum over a Finite Index Set, the two sums combine to

β∑k∈ΓN((x(k)2−ck)−(x(k)2−b(k)))=β∑k∈ΓN(b(k)−ck)=−β∑k∈ΓN(ck−b(k)),\beta\sum_{k\in\Gamma_{N}}\Bigl(\bigl(x(k)^{2}-c_{k}\bigr)-\bigl(x(k)^{2}-b(k)\bigr)\Bigr)=\beta\sum_{k\in\Gamma_{N}}\bigl(b(k)-c_{k}\bigr)=-\beta\sum_{k\in\Gamma_{N}}\bigl(c_{k}-b(k)\bigr),

where the summands were simplified by field arithmetic and claim 2 of Zero Products and Elementary Identities in a Field. This is the asserted formula for FNb[φ](x)F_{N}^{b}[\varphi](x). Finally, by The Cutoff Hamilton-Jacobi-Bellman Operators of the Wick-Square Problem, with a General Counterterm §wick, FN[φ]=FNb[φ]F_{N}[\varphi]=F_{N}^{b}[\varphi] with b(k)=ckb(k)=c_{k}; then every summand ck−b(k)c_{k}-b(k) is 00, so the sum vanishes by Sums over Finite Index Sets: Finite Unions, Disjoint Unions, Vanishing Terms, Dependent Pairs, Conjugation and the Modulus §vanishing, and the formula reduces to

FN[φ](x)=γ φ(x)−LN(φ−P)(x)+12 ∣DNφ(x)∣2−g(x).F_{N}[\varphi](x)=\gamma\,\varphi(x)-L_{N}(\varphi-P)(x)+\tfrac12\,|D_{N}\varphi(x)|^{2}-g(x).

This completes the proof.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…