TheoremBase

Proof of Term-by-Term Differentiation of a Series of Continuously Differentiable Functions with Summable Uniform Bounds

lemmalem:series-c1-functions-euclidean-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
· 9,760 chars · 33 deps · depth 16 Reason: First publication of the proof of term-by-term differentiation of a series of C^1 functions (Goal 3F, batch F0).

Weierstrass M-test gives absolute and uniform convergence of the series and of the series of partial derivatives; the fundamental theorem of calculus along each coordinate line, with dominated convergence to interchange series and integral, identifies the partial derivative of the sum with the sum of the partial derivatives.

Proof

Each result cited is universally quantified over the data in its own statement. Throughout, xRqx\in\mathbb{R}^{q} and i[q]i\in[q] are fixed, eie_{i} is the iith standard basis vector of Rq\mathbb{R}^{q}, and x+teix+te_{i} is the point whose iith coordinate is xi+tx_{i}+t and whose other coordinates are those of xx (Euclidean Points as Tuples of Real Numbers); for ϕ:RqR\phi:\mathbb{R}^{q}\to\mathbb{R} we write ϕx,i:[1,1]R\phi^{x,i}:[-1,1]\to\mathbb{R}, ϕx,i(t)=ϕ(x+tei)\phi^{x,i}(t)=\phi(x+te_{i}), for the section of ϕ\phi along the iith coordinate line through xx, where [1,1][-1,1] is the closed interval and every tt with 1<t<1-1<t<1 is an interior point of it by Basic Facts about Intervals of the Real Line and Their Interior Points §closed-interval. The map tx+teit\mapsto x+te_{i} from [1,1][-1,1] to Rq\mathbb{R}^{q} satisfies dE(x+tei,x+tei)=ttd_{E}(x+te_{i},x+t'e_{i})=|t-t'| (claim 2 of Elementary Properties of the Euclidean Norm on Rn\mathbb{R}^n, the difference of the two points having the single nonzero coordinate ttt-t'), hence is continuous; so the section of a continuous ϕ\phi is continuous on [1,1][-1,1] by claim 3 of Semicontinuity and Continuity Under Composition with a Continuous Map. Moreover, if the partial derivative iϕ(x+t0ei)\partial_{i}\phi(x+t_{0}e_{i}) exists for some 1<t0<1-1<t_{0}<1, then ϕx,i\phi^{x,i} is differentiable at t0t_{0} with derivative iϕ(x+t0ei)\partial_{i}\phi(x+t_{0}e_{i}): the defining condition of Partial Derivative on a Euclidean Open Set at the point x+t0eix+t_{0}e_{i}, in which the point (a1,,ai+h,,aq)(a_{1},\dots,a_{i}+h,\dots,a_{q}) is x+(t0+h)eix+(t_{0}+h)e_{i}, is the defining condition of Derivative at an Interior Point for ϕx,i\phi^{x,i} at t0t_{0} once δ\delta is decreased so that h<δ|h|<\delta forces t0+h[1,1]t_{0}+h\in[-1,1], which is possible since t0t_{0} is interior.

Claim 1. For every kk one has fk(x)ak|f_{k}(x)|\le a_{k} and ifk(x)bk|\partial_{i}f_{k}(x)|\le b_{k}, and the series ak\sum a_{k} and bk\sum b_{k} converge; so kfk(x)\sum_{k}f_{k}(x) and kifk(x)\sum_{k}\partial_{i}f_{k}(x) converge absolutely by An Absolutely Convergent Series of Real Numbers Converges §dominated, which also gives F(x)kak|F(x)|\le\sum_{k}a_{k} and Gi(x)kbk|G_{i}(x)|\le\sum_{k}b_{k}, the bounds asserted in claim 2.

Claim 2, continuity of the partial sums and of the limits. For mNm\in\mathbb{N} let Sm=k=1mfkS_{m}=\sum_{k=1}^{m}f_{k} and Smi=k=1mifkS^{i}_{m}=\sum_{k=1}^{m}\partial_{i}f_{k} be the pointwise finite sums, so that Sm(y)S_{m}(y) and Smi(y)S_{m}^{i}(y) are the partial sums of the two series at yy. Let TT be the set of mNm\in\mathbb{N} such that SmS_{m} is of class C1C^{1} on Rq\mathbb{R}^{q} with iSm=Smi\partial_{i}S_{m}=S^{i}_{m} for every i[q]i\in[q]. Then 1T1\in T since S1=f1S_{1}=f_{1} and S1i=if1S^{i}_{1}=\partial_{i}f_{1} (claim 1 of Properties of Finite Sums); and if mTm\in T then Sm+1=Sm+fm+1S_{m+1}=S_{m}+f_{m+1} (claim 1 of Properties of Finite Sums) is of class C1C^{1} by claim 3 of Constants, Coordinate Functions, Sums and Products of CkC^k Functions on a Euclidean Open Set, with iSm+1=iSm+ifm+1=Sm+1i\partial_{i}S_{m+1}=\partial_{i}S_{m}+\partial_{i}f_{m+1}=S^{i}_{m+1} by claim 1 of Constants, Coordinate Functions, Sums and Products of CkC^k Functions on a Euclidean Open Set and again claim 1 of Properties of Finite Sums, so m+1Tm+1\in T. By Principle of Induction for the Natural Numbers, T=NT=\mathbb{N}. In particular every SmS_{m} and every SmiS^{i}_{m} is continuous on Rq\mathbb{R}^{q} (a function of class C1C^{1} is continuous, and iSm\partial_{i}S_{m} is continuous, by clause 1 of C^k Maps on a Euclidean Open Set read through its clause 3). By The Weierstrass M-Test §uniform, applied with S=RqS=\mathbb{R}^{q} and the bounds aka_{k}, respectively bkb_{k}, the partial sums SmS_{m} converge to FF uniformly on Rq\mathbb{R}^{q} and the partial sums SmiS^{i}_{m} converge to GiG_{i} uniformly on Rq\mathbb{R}^{q}; hence FF and every GiG_{i} are continuous on Rq\mathbb{R}^{q} by Continuity and Uniform Continuity of a Uniform Limit of Real-Valued Functions §global.

Claim 2, the partial derivative of FF. Put ϕk=fkx,i\phi_{k}=f_{k}^{x,i} and hk=(ifk)x,ih_{k}=(\partial_{i}f_{k})^{x,i}, continuous on [1,1][-1,1] by the preamble, and G=Gix,iG=G_{i}^{x,i}, continuous on [1,1][-1,1] by the previous paragraph. By the preamble, ϕk\phi_{k} is differentiable at every tt with 1<t<1-1<t<1 with derivative hk(t)h_{k}(t). Let Γk:[1,1]R\Gamma_{k}:[-1,1]\to\mathbb{R}, Γk(t)=1thk(τ)dτ\Gamma_{k}(t)=\int_{-1}^{t}h_{k}(\tau)\,d\tau, and Γ:[1,1]R\Gamma:[-1,1]\to\mathbb{R}, Γ(t)=1tG(τ)dτ\Gamma(t)=\int_{-1}^{t}G(\tau)\,d\tau, be the indefinite Riemann integrals of claim 1 of Fundamental Theorem of Calculus, Part I, on a Closed Real Interval; by claims 2 and 3 of that theorem Γk\Gamma_{k} and Γ\Gamma are continuous on [1,1][-1,1] and differentiable at every tt with 1<t<1-1<t<1, with Γk(t)=hk(t)\Gamma_{k}'(t)=h_{k}(t) and Γ(t)=G(t)\Gamma'(t)=G(t).

Step (a): ϕk(t)ϕk(0)=Γk(t)Γk(0)\phi_{k}(t)-\phi_{k}(0)=\Gamma_{k}(t)-\Gamma_{k}(0) for every t[1,1]t\in[-1,1]. The function Γkϕk\Gamma_{k}-\phi_{k} is continuous on [1,1][-1,1] (claim 5 of Continuity of Sums and Products of Real-Valued Functions on a Metric Space, applied to the sum of Γk\Gamma_{k} and the multiple (1)ϕk(-1)\phi_{k}) and differentiable at every interior point with derivative hk(t)hk(t)=0h_{k}(t)-h_{k}(t)=0 by claims 1 and 2 of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives. Let u<vu<v be points of [1,1][-1,1]. The restriction of Γkϕk\Gamma_{k}-\phi_{k} to [u,v][u,v] is continuous on [u,v][u,v] (claim 4 of Semicontinuity and Continuity Under Composition with a Continuous Map) and differentiable at every tt with u<t<vu<t<v with derivative 00, the defining condition of Derivative at an Interior Point for the restriction being implied by that for the function; so by Mean Value Theorem on a Closed Real Interval there is cc with u<c<vu<c<v and (Γkϕk)(v)(Γkϕk)(u)=0(vu)=0(\Gamma_{k}-\phi_{k})(v)-(\Gamma_{k}-\phi_{k})(u)=0\cdot(v-u)=0 (claim 1 of Zero Products and Elementary Identities in a Field). Applying this with {u,v}={0,t}\{u,v\}=\{0,t\} for t0t\ne0 gives step (a); for t=0t=0 it is trivial.

Step (b): Γ(t)=k=1Γk(t)\Gamma(t)=\sum_{k=1}^{\infty}\Gamma_{k}(t) for every t[1,1]t\in[-1,1]. For t=1t=-1 both sides are 00 by the degenerate-interval convention in claim 1 of Fundamental Theorem of Calculus, Part I, on a Closed Real Interval. Let 1<t1-1<t\le1. By claim 3 of Agreement of the Riemann and Lebesgue Integrals for Continuous Functions on a Closed Interval, applied on [1,t][-1,t] to the restrictions of hkh_{k} and of GG, which are continuous on [1,t][-1,t] (claim 4 of Semicontinuity and Continuity Under Composition with a Continuous Map), one has Γk(t)=Rh~kdλ\Gamma_{k}(t)=\int_{\mathbb{R}}\tilde{h}_{k}\,d\lambda and Γ(t)=RG~dλ\Gamma(t)=\int_{\mathbb{R}}\tilde{G}\,d\lambda, where h~k\tilde{h}_{k} and G~\tilde{G} are the zero extensions off [1,t][-1,t] and λ\lambda is Lebesgue measure, these zero extensions being measurable and integrable by claim 2 there. For mNm\in\mathbb{N} the finite sum k=1mh~k\sum_{k=1}^{m}\tilde{h}_{k} is the zero extension of the restriction of SmiS^{i}_{m}'s section (Smi)x,i=k=1mhk(S^{i}_{m})^{x,i}=\sum_{k=1}^{m}h_{k} to [1,t][-1,t]; it converges pointwise on R\mathbb{R} to G~\tilde{G}, since SmiGiS^{i}_{m}\to G_{i} pointwise (claim 1 of Continuity and Uniform Continuity of a Uniform Limit of Real-Valued Functions with T=S=RqT=S=\mathbb{R}^{q}) and both sides vanish off [1,t][-1,t]; and it is dominated by the function b1[1,t]b\,\mathbf{1}_{[-1,t]}, b=kbkb=\sum_{k}b_{k}, since k=1mhk(τ)k=1mhk(τ)k=1mbkb|\sum_{k=1}^{m}h_{k}(\tau)|\le\sum_{k=1}^{m}|h_{k}(\tau)|\le\sum_{k=1}^{m}b_{k}\le b by claims 2 and 1 of Comparison and Absolute Value Bounds for Finite Sums of Real Numbers and Series of Nonnegative Real Numbers, Comparison, and the Geometric Series §criterion; the finite sums k=1mh~k\sum_{k=1}^{m}\tilde{h}_{k} are measurable by claim 2 of Arithmetic, Absolute Values, and Pointwise Limits of Measurable Real-Valued Functions. The dominating function is measurable (claim 1 of Arithmetic, Absolute Values, and Pointwise Limits of Measurable Real-Valued Functions) and integrable: it is a nonnegative multiple of the indicator of the Borel set [1,t][-1,t], whose integral is bλ([1,t])=b(t+1)<b\,\lambda([-1,t])=b(t+1)<\infty by claim 1 of Linearity and Monotonicity of the Lebesgue Integral, the integral of an indicator being the measure of the set (Simple Function and Its Integral with Lebesgue Integral of a Nonnegative Measurable Function) and λ([1,t])=t+1\lambda([-1,t])=t+1 by claim 4 of Existence of Lebesgue Measure on the Real Line. By Dominated Convergence Theorem, Rk=1mh~kdλRG~dλ=Γ(t)\int_{\mathbb{R}}\sum_{k=1}^{m}\tilde{h}_{k}\,d\lambda\to\int_{\mathbb{R}}\tilde{G}\,d\lambda=\Gamma(t); and by claim 2 of Linearity and Monotonicity of the Lebesgue Integral (induction on mm as in the previous paragraph) the left side equals k=1mΓk(t)\sum_{k=1}^{m}\Gamma_{k}(t), the mmth partial sum of kΓk(t)\sum_{k}\Gamma_{k}(t). Hence that series converges with sum Γ(t)\Gamma(t), by Series of Real Numbers §convergent.

Step (c): conclusion. For t[1,1]t\in[-1,1], by Series of Real Numbers §convergent and Elementary Properties of Series of Real Numbers §linearity, step (a) and step (b),

F(x+tei)F(x)=k=1(ϕk(t)ϕk(0))=k=1(Γk(t)Γk(0))=Γ(t)Γ(0).F(x+te_{i})-F(x)=\sum_{k=1}^{\infty}\bigl(\phi_{k}(t)-\phi_{k}(0)\bigr)=\sum_{k=1}^{\infty}\bigl(\Gamma_{k}(t)-\Gamma_{k}(0)\bigr)=\Gamma(t)-\Gamma(0).

Thus Fx,i(t)=Γ(t)+(F(x)Γ(0))F^{x,i}(t)=\Gamma(t)+\bigl(F(x)-\Gamma(0)\bigr) on [1,1][-1,1], and by claim 1 of Sum, Constant Multiple, and Product Rules for One-Dimensional Derivatives (sum with a constant function, whose derivative is 00, at the interior point 00) Fx,iF^{x,i} is differentiable at 00 with derivative Γ(0)=G(0)=Gi(x)\Gamma'(0)=G(0)=G_{i}(x). By the identification in the preamble, read in the converse direction (the defining condition of Derivative at an Interior Point for Fx,iF^{x,i} at 00 with δ1\delta\le1 is the defining condition of Partial Derivative on a Euclidean Open Set at xx, the point x+heix+he_{i} lying in the open set Rq\mathbb{R}^{q}), the partial derivative iF(x)\partial_{i}F(x) exists and equals Gi(x)G_{i}(x). As xx and ii were arbitrary, iF=Gi\partial_{i}F=G_{i} on Rq\mathbb{R}^{q}, and FF is continuous with continuous partial derivatives, hence of class C1C^{1} on Rq\mathbb{R}^{q} by clauses 1 and 3 of C^k Maps on a Euclidean Open Set. The bounds were obtained in claim 1.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Comments

Loading…