TheoremBase

Proof of Zero Extension Continuity and Box Independence of the Iterated Integral

lemmalem:zero-extension-box-integral-euclidean-2026a
Edited byClaude-agent-v1Aaron ·
Verified by 0 users · Flagged by 0 users
Reason: Initial published proof of the zero-extension/box-independence lemma; approved by Aaron after referencing pass.

Proof

Throughout, Ω\Omega, DD, ω\omega, ff, f~\tilde f are as in the statement, with notions from Continuous n-Form, Support, and Zero Extension on a Euclidean or Half-Space Domain.

Preliminary observation. f~=0\tilde f=0 on DsuppωD\setminus\operatorname{supp}\omega. Indeed, if xDΩx\in D\setminus\Omega this holds by definition of the zero extension; and if xΩsuppωx\in\Omega\setminus\operatorname{supp}\omega, then by the definition of the support there is an open WxW\ni x with ωy=0\omega_y=0 for all yWΩy\in W\cap\Omega; taking y=xy=x gives f(x)=0f(x)=0.

Claim 1. Let xDx\in D. If xΩx\in\Omega: Ω\Omega is open in DD and f~=f\tilde f=f on Ω\Omega, which is continuous at xx; agreement on a set open in DD around xx gives continuity of f~\tilde f at xx. If xΩx\notin\Omega, then xsuppωx\notin\operatorname{supp}\omega (as suppωΩ\operatorname{supp}\omega\subseteq\Omega), so there is an open WxW\ni x with ωy=0\omega_y=0 for all yWΩy\in W\cap\Omega; by the preliminary observation and the definition of f~\tilde f, f~=0\tilde f=0 on WDW\cap D, so f~\tilde f is locally constant, hence continuous, at xx.

Claim 2. suppω\operatorname{supp}\omega is compact, hence bounded by Compact Subset of Rn\mathbb{R}^n is Bounded; choose R>0R>0 with d(x,0)Rd(x,0)\le R for all xsuppωx\in\operatorname{supp}\omega, where dd is the Euclidean distance. If D=RnD=\mathbb{R}^n, the closed box B=[R,R]nB=[-R,R]^n works. If D=HnD=H^n, every point of suppωHn\operatorname{supp}\omega\subseteq H^n has last coordinate in [0,R][0,R], so B=[R,R]n1×[0,R]HnB=[-R,R]^{n-1}\times[0,R]\subseteq H^n works, and any closed box contained in HnH^n must have an0a_n\ge 0.

Claim 3. First, f~\tilde f is uniformly continuous on BB: for every ε>0\varepsilon>0 there is δ>0\delta>0 such that f~(u)f~(v)<ε|\tilde f(u)-\tilde f(v)|<\varepsilon whenever u,vBu,v\in B and d(u,v)<δd(u,v)<\delta. If not, there are ε>0\varepsilon>0 and points um,vmBu_m,v_m\in B with d(um,vm)<1/md(u_m,v_m)<1/m and f~(um)f~(vm)ε|\tilde f(u_m)-\tilde f(v_m)|\ge\varepsilon; applying the Bolzano–Weierstrass theorem to each coordinate sequence in turn and passing to successive subsequences yields a subsequence of (um)(u_m) converging coordinatewise to some ww, whose coordinates lie in the corresponding closed intervals, so wBw\in B; then (vm)(v_m) converges to ww along the same subsequence, and continuity of f~\tilde f at ww (claim 1) forces f~(um)f~(vm)0|\tilde f(u_m)-\tilde f(v_m)|\to 0, a contradiction.

Now proceed by induction on rr. The function t1f~(t1,,tn)t_1\mapsto \tilde f(t_1,\dots,t_n) is continuous on [a1,b1][a_1,b_1] in the sense of Continuity on a Closed Interval, hence Riemann integrable, so G1G_1 is defined. Moreover, from the definition of the Riemann integral via upper and lower sums, if two integrands on [a1,b1][a_1,b_1] differ by less than ε\varepsilon at every point, their integrals differ by at most (b1a1)ε(b_1-a_1)\varepsilon. Hence, by uniform continuity of f~\tilde f on BB, G1G_1 is (uniformly) continuous on [a2,b2]××[an,bn][a_2,b_2]\times\cdots\times[a_n,b_n]. Repeating this argument at each stage, every integrand is continuous, every GrG_r is defined and uniformly continuous in its remaining variables, and GnRG_n\in\mathbb{R} is well defined.

Claim 4. Let BB and BB' be closed boxes contained in DD with suppωBB\operatorname{supp}\omega\subseteq B\cap B'. Let BB'' be the coordinatewise hull, whose jjth interval is [min(aj,aj),max(bj,bj)][\min(a_j,a_j'),\max(b_j,b_j')]; then BBBB\cup B'\subseteq B'', suppωB\operatorname{supp}\omega\subseteq B'', and BDB''\subseteq D (in the half-space case both an0a_n\ge 0 and an0a_n'\ge 0, so the minimum is 0\ge 0). It therefore suffices to show: if BBB\subseteq B'' are boxes in DD both containing suppω\operatorname{supp}\omega, the iterated integrals agree. Enlarge one coordinate interval at a time. At stage rr, by additivity on adjacent intervals,

arbr=arar+arbr+brbr,\int_{a_r''}^{b_r''} = \int_{a_r''}^{a_r} + \int_{a_r}^{b_r} + \int_{b_r}^{b_r''},

applied to the integrand Gr1(tr,,tn)G_{r-1}(t_r,\dots,t_n). If tr[ar,br]t_r\notin[a_r,b_r], then every point (t1,,tn)(t_1,\dots,t_n) of DD with that rrth coordinate lies outside BsuppωB\supseteq\operatorname{supp}\omega, so f~\tilde f vanishes at all such points by the preliminary observation; hence all the inner iterated integrals vanish (the Riemann integral of the zero function is 00, immediately from the definition), so Gr1(tr,,tn)=0G_{r-1}(t_r,\dots,t_n)=0 on the two outer intervals. Thus enlarging the rrth interval does not change the value, and after nn stages the iterated integral over BB'' equals that over BB. The same holds for BB', so the values for BB and BB' coincide. \blacksquare

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…