Matrix analysis · PhysLib theorem
Araki–Lieb–Thirring Trace Inequality
For finite positive-semidefinite Hermitian matrices A and B and 0 < r ≤ 1, PhysLib proves Tr(BʳAʳBʳ) ≤ Tr((BAB)ʳ), using its Hermitian-matrix conjugation and real-power operations.
What it says
What the theorem says
For finite positive-semidefinite Hermitian matrices A and B, the trace of BʳAʳBʳ is at most the trace of (BAB)ʳ whenever 0 < r ≤ 1. The statement is a finite-matrix trace inequality and does not assert an operator inequality or cover exponents above one.
Context
Why it matters
The Araki–Lieb–Thirring inequality is a major bridge between spectral matrix powers and trace comparison, with applications across operator theory and quantum information.
Inside the proof
Proof architecture
The retained declaration extends a positive-definite Lieb–Thirring estimate to the positive-semidefinite boundary by regularizing B, proving continuity of both trace expressions, and taking a one-sided limit.
Step 1
Regularize the semidefinite matrix
The proof replaces B by Bε = B + εI. For every ε > 0, the shifted matrix is positive definite, so the previously established positive-definite inequality applies.
Bε = B + εI, ε > 0
Show the Lean ingredients
set Bε : ℝ → HermitianMat d ℂ := fun ε => B + ε • 1have hPos : ∀ ε > 0, (Bε ε).mat.PosDefStep 2
Track both trace expressions continuously
Continuity of real matrix powers, conjugation, and trace shows that the regularized left and right sides converge to the desired semidefinite expressions as ε approaches zero from above.
Bε → B ⟹ L(Bε) → L(B), R(Bε) → R(B)
Show the Lean ingredients
have hleft_tendstohave hright_tendstotendsto_helperStep 3
Pass the inequality to the boundary
The positive-definite estimate holds for every ε > 0. The two convergence results transfer that eventual inequality to the limiting positive-semidefinite matrices.
L(Bε) ≤ R(Bε) for ε > 0 ⟹ L(B) ≤ R(B)
Show the Lean ingredients
exact le_of_tendsto_of_tendstolieb_thirring_le_one_posDefAuthoritative formal type
Exact PhysLib statement
HermitianMat.lieb_thirring_le_one.{u_1} {d : Type u_1} [Fintype d] [DecidableEq d] {A B : HermitianMat d ℂ} (hA : 0 ≤ A) (hB : 0 ≤ B) {r : ℝ} (hr0 : 0 < r) (hr1 : r ≤ 1) : ((HermitianMat.conj ↑(B ^ r)) (A ^ r)).trace ≤ ((HermitianMat.conj ↑B) A ^ r).traceThis is the theorem's exact Lean type. The diagrams and commentary are explanatory; the pinned statement and source are authoritative.
Recorded connections
How this result fits
A one-step view of the published results directly connected to this page.
3 direct connections
- Related resultAraki–Lieb–Thirring Trace Inequalityis related toHorn's Weak Log-Majorization for Singular Values
- Related resultAraki–Lieb–Thirring Trace Inequalityis related toSchatten–Hölder Inequality for Matrix Products
- Related resultAraki–Lieb–Thirring Trace Inequalityis related toLieb's Trace Functional Is Jointly Concave
Only recorded public relationships are shown. A shared topic does not imply a dependency, and this is not a complete Lean proof-term dependency graph.
Open this result in the full theorem map3 direct connectionsProvenance and verification
Sources and technical checks
Show declaration provenance and Lean checks
HermitianMat.lieb_thirring_le_oneQuantumInfo.ForMathlib.HermitianMat.Rpow6bd0b4b99d3dsorryAx reportedThe declaration footprint describes source layout, not proof difficulty or the size of its dependencies. The pinned checks verify the cited Lean theorem; semantic review remains a separate PhysAtlas judgment.



