Eq. (9), operator form — (corrected: )
ProvedQChernoff.Trace.eq_9functional-calculusintegral-representationp2o-batch-pfp2ap2o-correctedp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1quantum-chernoff
Let be a complex Hilbert space, a positive bounded operator on , and . Then the integrand below is integrable on and
where is the inverse of (which exists for ) and the integral is the Bochner integral of an operator-valued function, which converges absolutely in operator norm. For this is the scalar identity for .
In the proof of Lemma 1 this representation turns the difference into an integral of resolvent differences, which can then be handled one value of at a time.
Formalization Note A ^ t is Mathlib's CFC.rpow; real scalars act on operators through the real vector-space structure; Ring.inverse is the inverse in the ring of bounded operators.
Differences from the source:
- correction: valid for a ≥ 0 and 0 ≤ t ≤ 1 → 0 < t < 1; reason: at t = 0 and t = 1 the factor sin(tπ) vanishes while the integral diverges: for a = 1, t = 0 the left side is 1 and the right side is 0·∞ (in Lean a non-integrable integral is 0, so the printed range would be false)
- generalization: scalar a ≥ 0 (then "extending this representation to positive operators in the usual way") → positive bounded operator A on an arbitrary complex Hilbert space; reason: the paper uses the operator form; H = ℂ gives the scalar identity
- encoding: ∫_0^{+∞} dx a x^{t−1}/(a + x) → Bochner integral over Set.Ioi 0 of x^{t−1} • (A * Ring.inverse (A + x•1)) in H →L[ℂ] H; reason: for x > 0 the operator A + x is invertible, so Ring.inverse is the true inverse; the integrand is absolutely integrable, so the Lebesgue integral is the improper integral
- encoding: convergence of the integral (implicit in the display) → an explicit IntegrableOn conjunct in the conclusion; reason: it asserts that the Bochner integral is not Lean's junk value 0 of a non-integrable function; it makes the statement stronger, not weaker
Preamble
import Mathlib
Formal statement
namespace QChernoff.Trace
/-- Eq. (9), p. 4, in the operator form used in Appendix A (corrected: `0 < t < 1`): the
integrand is Bochner integrable on `(0, ∞)` and `A ^ t = (sin(tπ)/π) ∫₀^∞ x^{t−1} A (A + x)⁻¹ dx`. -/
theorem eq_9 {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℂ H] [CompleteSpace H]
(A : H →L[ℂ] H) (hA : 0 ≤ A) (t : ℝ) (ht0 : 0 < t) (ht1 : t < 1) :
MeasureTheory.IntegrableOn
(fun x : ℝ => x ^ (t - 1) • (A * Ring.inverse (A + (x : ℂ) • 1))) (Set.Ioi 0) ∧
A ^ t = (Real.sin (t * Real.pi) / Real.pi) •
∫ x in Set.Ioi (0 : ℝ), x ^ (t - 1) • (A * Ring.inverse (A + (x : ℂ) • 1)) := by sorry
end QChernoff.Trace
Source
Audenaert et al., The Quantum Chernoff Bound, arXiv:quant-ph/0610027v1, p. 4, eq. (9), proof of Lemma 1, Appendix A
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.