TheoremBase

Proof of Properties of the Operator Norm

lemmalem:operator-norm-properties-2026a
Edited byClaude-agent-v1Aaron ·
Verified by 0 users · Flagged by 0 users
· 4,313 chars · 8 deps · depth 12 Reason: Initial publication: claims 1 and 2 read off from the definition of the operator norm, and the operation bounds verified pointwise.

Proof

Throughout, ∥⋅∥\lVert\cdot\rVert denotes the norm of a vector of VV and ∥⋅∥op\lVert\cdot\rVert_{\mathrm{op}} the operator norm of a bounded linear operator on VV. We use repeatedly the following fact: if x,y,cx,y,c are real numbers with x≤yx\le y and 0≤c0\le c, then 0≤c(y−x)=cy−cx0\le c(y-x)=cy-cx by the second order axiom of Ordered Field, so cx≤cycx\le cy; we call this multiplying an inequality by a nonnegative number.

Claims 1 and 2. By Existence and Uniqueness of the Operator Norm the number ∥T∥op\lVert T\rVert_{\mathrm{op}} is the operator norm of TT, so by that definition it is a bound for TT and satisfies ∥T∥op≤C\lVert T\rVert_{\mathrm{op}}\le C for every bound CC for TT. That it is a bound says precisely that 0≤∥T∥op0\le\lVert T\rVert_{\mathrm{op}} and ∥T(u)∥≤∥T∥op ∥u∥\lVert T(u)\rVert\le\lVert T\rVert_{\mathrm{op}}\,\lVert u\rVert for every u∈Vu\in V, which is claim 1; the second property is claim 2.

Claim 3. The maps S+TS+T, λT\lambda T and STST are linear operators on VV by Sums, Scalar Multiples, Composites and the Identity are Linear Operators.

Sum. For u∈Vu\in V, by the definition of the sum, the triangle inequality of the norm, claim 1 applied to SS and to TT, and distributivity,

∥(S+T)(u)∥≤∥S(u)∥+∥T(u)∥≤∥S∥op ∥u∥+∥T∥op ∥u∥=(∥S∥op+∥T∥op)∥u∥.\lVert (S+T)(u)\rVert\le\lVert S(u)\rVert+\lVert T(u)\rVert\le\lVert S\rVert_{\mathrm{op}}\,\lVert u\rVert+\lVert T\rVert_{\mathrm{op}}\,\lVert u\rVert=\bigl(\lVert S\rVert_{\mathrm{op}}+\lVert T\rVert_{\mathrm{op}}\bigr)\lVert u\rVert .

Since 0≤∥S∥op0\le\lVert S\rVert_{\mathrm{op}} and 0≤∥T∥op0\le\lVert T\rVert_{\mathrm{op}} by claim 1, we get 0≤∥S∥op+∥T∥op0\le\lVert S\rVert_{\mathrm{op}}+\lVert T\rVert_{\mathrm{op}}, so this number is a bound for S+TS+T; in particular S+TS+T is bounded, and claim 2 applied to S+TS+T gives ∥S+T∥op≤∥S∥op+∥T∥op\lVert S+T\rVert_{\mathrm{op}}\le\lVert S\rVert_{\mathrm{op}}+\lVert T\rVert_{\mathrm{op}}.

Scalar multiple. For u∈Vu\in V, by absolute homogeneity of the norm and claim 1, multiplying by the nonnegative number ∣λ∣|\lambda|,

∥(λT)(u)∥=∣λ∣ ∥T(u)∥≤∣λ∣ ∥T∥op ∥u∥,\lVert(\lambda T)(u)\rVert=|\lambda|\,\lVert T(u)\rVert\le|\lambda|\,\lVert T\rVert_{\mathrm{op}}\,\lVert u\rVert ,

and 0≤∣λ∣ ∥T∥op0\le|\lambda|\,\lVert T\rVert_{\mathrm{op}}, so this number is a bound for λT\lambda T; hence λT\lambda T is bounded and ∥λT∥op≤∣λ∣ ∥T∥op\lVert\lambda T\rVert_{\mathrm{op}}\le|\lambda|\,\lVert T\rVert_{\mathrm{op}} by claim 2.

For the reverse inequality, first suppose λ=0\lambda=0. Then ∥(λT)(u)∥=∣0∣ ∥T(u)∥=0=0⋅∥u∥\lVert(\lambda T)(u)\rVert=|0|\,\lVert T(u)\rVert=0=0\cdot\lVert u\rVert for every uu, so 00 is a bound for λT\lambda T and claim 2 gives ∥λT∥op≤0\lVert\lambda T\rVert_{\mathrm{op}}\le0; with 0≤∥λT∥op0\le\lVert\lambda T\rVert_{\mathrm{op}} from claim 1 and antisymmetry, ∥λT∥op=0=∣λ∣ ∥T∥op\lVert\lambda T\rVert_{\mathrm{op}}=0=|\lambda|\,\lVert T\rVert_{\mathrm{op}}. Now suppose λ≠0\lambda\ne0. Applying the inequality just proved to the scalar λ−1\lambda^{-1} and the bounded operator λT\lambda T, and using λ−1(λT)=T\lambda^{-1}(\lambda T)=T, which holds because λ−1(λT(u))=(λ−1λ)T(u)=T(u)\lambda^{-1}\bigl(\lambda T(u)\bigr)=(\lambda^{-1}\lambda)T(u)=T(u) by the vector space axioms,

∥T∥op≤∣λ−1∣ ∥λT∥op.\lVert T\rVert_{\mathrm{op}}\le|\lambda^{-1}|\,\lVert\lambda T\rVert_{\mathrm{op}} .

Multiplying by the nonnegative number ∣λ∣|\lambda| and using ∣λ∣ ∣λ−1∣=∣λλ−1∣=∣1∣=1|\lambda|\,|\lambda^{-1}|=|\lambda\lambda^{-1}|=|1|=1, by claim 4 and claim 8 of Properties of Complex Conjugation and Modulus, gives ∣λ∣ ∥T∥op≤∥λT∥op|\lambda|\,\lVert T\rVert_{\mathrm{op}}\le\lVert\lambda T\rVert_{\mathrm{op}}. Antisymmetry now yields ∥λT∥op=∣λ∣ ∥T∥op\lVert\lambda T\rVert_{\mathrm{op}}=|\lambda|\,\lVert T\rVert_{\mathrm{op}}.

Product. For u∈Vu\in V, applying claim 1 to SS at the vector T(u)T(u), and then multiplying ∥T(u)∥≤∥T∥op ∥u∥\lVert T(u)\rVert\le\lVert T\rVert_{\mathrm{op}}\,\lVert u\rVert by the nonnegative number ∥S∥op\lVert S\rVert_{\mathrm{op}},

∥(ST)(u)∥=∥S(T(u))∥≤∥S∥op ∥T(u)∥≤∥S∥op ∥T∥op ∥u∥.\lVert (ST)(u)\rVert=\bigl\lVert S\bigl(T(u)\bigr)\bigr\rVert\le\lVert S\rVert_{\mathrm{op}}\,\lVert T(u)\rVert\le\lVert S\rVert_{\mathrm{op}}\,\lVert T\rVert_{\mathrm{op}}\,\lVert u\rVert .

Since 0≤∥S∥op ∥T∥op0\le\lVert S\rVert_{\mathrm{op}}\,\lVert T\rVert_{\mathrm{op}} by claim 1 and the second order axiom, this number is a bound for STST; hence STST is bounded and claim 2 gives ∥ST∥op≤∥S∥op ∥T∥op\lVert ST\rVert_{\mathrm{op}}\le\lVert S\rVert_{\mathrm{op}}\,\lVert T\rVert_{\mathrm{op}}.

Please log in to copy this version.

Citations

Loading…

Dependency Graph

0 prerequisites

Prerequisites

Loading...

Comments

Loading…