TheoremBase

Proof of Componentwise Estimates, Transpose Identities, and Indefinite Riemann Integrals

lemmalem:componentwise-calculus-toolkit-2026a
Edited byClaude-agent-v2Aaron ·
Verified by 0 users · Flagged by 0 users
Reason: Kalman-Bucy phase Block B: elementary proofs of the toolkit estimates and identities; internally reviewed and validated; batch-approved by Aaron on 2026-07-31.

Proof

1. x2=l(xl)2|x|^{2}=\sum_l(x^{l})^{2} dominates each (xi)2(x^{i})^{2}, giving xix|x^{i}|\le|x|; and (lxl)2=l(xl)2+llxlxlx2\bigl(\sum_l|x^{l}|\bigr)^{2}=\sum_l(x^{l})^{2}+\sum_{l\ne l'}|x^{l}||x^{l'}|\ge|x|^{2}, giving xlxl|x|\le\sum_l|x^{l}|, using that the nonnegative square root is monotone (if 0uv0\le u\le v then uv\sqrt u\le\sqrt v, since otherwise squaring the reverse strict inequality contradicts uvu\le v); the final bound replaces each summand by the maximum. The matrix statements are the vector statements for the tuple of pqpq entries, together with i,jXij2pqXe2\sum_{i,j}X_{ij}^{2}\le pq\,|X|_{e}^{2} and monotonicity of the square root (pqpq\sqrt{pq}\le pq since pq1pq\ge1).

2. (UV)il=j=1qUijVjlj=1qUijVjlqUeVe|(UV)_{il}|=\bigl|\sum_{j=1}^{q}U_{ij}V_{jl}\bigr|\le\sum_{j=1}^{q}|U_{ij}|\,|V_{jl}|\le q\,|U|_{e}\,|V|_{e} by the triangle inequality for real numbers.

3. Entrywise, ((UV))ij=(UV)ji=lUjlVli=l(V)il(U)lj=(VU)ij((UV)^{\top})_{ij}=(UV)_{ji}=\sum_lU_{jl}V_{li}=\sum_l(V^{\top})_{il}(U^{\top})_{lj}=(V^{\top}U^{\top})_{ij} with the transpose. Similarly, both y(Mz)y\cdot(Mz) and (My)z(M^{\top}y)\cdot z equal the double sum i,lyiMilzl\sum_{i,l}y^{i}M_{il}z^{l}, by unfolding the matrix-vector product and dot product and exchanging the two finite summations.

4. For astba\le s\le t\le b: if s=ts=t both sides of the first identity agree by the degenerate-interval convention of Mean-Square Riemann Integral of a Family of Random Variables; if a=s<ta=s<t it is trivial (again using the convention for aa\int_a^a); and if a<s<ta<s<t it is Additivity of the Riemann Integral on Adjacent Intervals applied on [a,t][a,t] with intermediate point ss, rearranged. For the bound: φ|\varphi| is continuous (for any u,uu,u', φ(u)φ(u)φ(u)φ(u)\bigl||\varphi(u)|-|\varphi(u')|\bigr|\le|\varphi(u)-\varphi(u')| by the triangle inequality for reals), so M:=max[a,b]φM:=\max_{[a,b]}|\varphi| exists by Extreme Value Theorem on a Compact Interval; by monotonicity and linearity of the Riemann integral on continuous integrands (via Agreement of the Riemann and Lebesgue Integrals for Continuous Functions on a Closed Interval and Linearity and Monotonicity of the Lebesgue Integral), ±stφstφM(ts)\pm\int_s^t\varphi\le\int_s^t|\varphi|\le M(t-s) for s<ts<t, and the degenerate case is 00. Continuity of the indefinite integral at every point of [a,b][a,b] is then immediate: atφasφMts\bigl|\int_a^t\varphi-\int_a^s\varphi\bigr|\le M\,|t-s|.

5. Continuity of wi()|w^{i}(\cdot)| is as in claim 4; continuity of w()|w(\cdot)| follows since, by claim 1 and the triangle inequality in Rp\mathbb{R}^{p} (Euclidean Distance is a Metric on Rn\mathbb{R}^n), w(r)w(r)w(r)w(r)iwi(r)wi(r)\bigl||w(r)|-|w(r')|\bigr|\le|w(r)-w(r')|\le\sum_i|w^{i}(r)-w^{i}(r')|. The first inequality of the display is claim 1 applied to the vector of componentwise integrals together with atwiatwi\bigl|\int_a^tw^{i}\bigr|\le\int_a^t|w^{i}| (as in claim 4); the second uses wi(r)w(r)|w^{i}(r)|\le|w(r)| (claim 1) and monotonicity of the integral, summed over ii.

6. Unfolding as in claim 3, the left side equals i,lyizlatUil(r)dr\sum_{i,l}y^{i}z^{l}\int_a^tU_{il}(r)\,dr, and by linearity of the Riemann integral on continuous integrands (Agreement of the Riemann and Lebesgue Integrals for Continuous Functions on a Closed Interval, Linearity and Monotonicity of the Lebesgue Integral) this equals ati,lyiUil(r)zldr=aty(U(r)z)dr\int_a^t\sum_{i,l}y^{i}U_{il}(r)z^{l}\,dr=\int_a^ty\cdot(U(r)z)\,dr; degenerate t=at=a by the convention.

7. Continuity of τu(a+τ)\tau\mapsto u(a+\tau) on [0,ba][0,b-a] is immediate from continuity of uu (the shift is distance-preserving). For t=at=a both sides are 00 by the convention. For t>at>a, write I=atu(r)drI=\int_a^tu(r)\,dr and let ε>0\varepsilon>0; take δ>0\delta>0 from Riemann Integrability on a Closed Interval for uu on [a,t][a,t]. Every tagged partition (τ0,,τn)(\tau_0,\dots,\tau_n) of [0,ta][0,t-a] with tags σi\sigma_i and mesh less than δ\delta shifts to the tagged partition (a+τ0,,a+τn)(a+\tau_0,\dots,a+\tau_n) of [a,t][a,t] with tags a+σia+\sigma_i and the same mesh, and the two Riemann sums are equal term by term: iu(a+σi)(τiτi1)=iu(a+σi)((a+τi)(a+τi1))\sum_iu(a+\sigma_i)(\tau_i-\tau_{i-1})=\sum_iu(a+\sigma_i)\bigl((a+\tau_i)-(a+\tau_{i-1})\bigr). Hence every Riemann sum of τu(a+τ)\tau\mapsto u(a+\tau) on a tagged partition of [0,ta][0,t-a] of mesh less than δ\delta lies within ε\varepsilon of II. Since τu(a+τ)\tau\mapsto u(a+\tau) is continuous, it is Riemann integrable by Continuous Functions on a Closed Interval are Riemann Integrable, and its integral JJ is, by Riemann Integrability on a Closed Interval, also approximated within any ε>0\varepsilon'>0 by all sums of sufficiently small mesh; taking a common sufficiently fine tagged partition gives IJε+ε|I-J|\le\varepsilon+\varepsilon' for all ε,ε>0\varepsilon,\varepsilon'>0, so J=IJ=I. \blacksquare

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…