§III.D: distance and angle objectives (113)–(116)
ProvedQAlgorithms.Papers.Independent.Advance50.P38.distanceAngleObjectivesLet , let be arbitrary complex by matrices, and put for . Then
A unitary minimizes this distance exactly when it maximizes . If , define the squared cosine between their complex lines by . Maximizers of are exactly maximizers of the absolute trace. If the trace range is a disk centered at zero, some unitary achieves , the maximal absolute trace. This last assertion concerns optimal values and does not identify the two optimizer sets.
import Definitions.Def_QAlgorithms_Papers_Independent_Advance50_P38
namespace QAlgorithms.Papers.Independent.Advance50.P38
open scoped BigOperators Topology Manifold ContDiff
noncomputable section
/-- Distance and angle objectives (113)--(116), PDF p.21.
Printed: the two tasks are equivalent for a circular numerical-range disk.
Changed: the real and absolute overlaps have equal optimal values and a common
optimizer; equality of their optimizer sets is not asserted.
Why: for A=C=E12 in dimension 2 the range is the unit disk, but diag(i,-i)
maximizes the absolute overlap and has real overlap -1 instead of the maximum 1.
The distance identity and nonzero-line angle equivalence are unchanged. -/
theorem distanceAngleObjectives {n : ℕ} (hn : 0 < n) (A C : (Matrix (Fin n) (Fin n) ℂ)) :
(∀ U : Matrix.specialUnitaryGroup (Fin n) ℂ,
frobSq ((U : (Matrix (Fin n) (Fin n) ℂ)) * A * (U : (Matrix (Fin n) (Fin n) ℂ)).conjTranspose - C) =
frobSq A + frobSq C - 2 * realTraceOverlap A C U) ∧
(∀ U : Matrix.specialUnitaryGroup (Fin n) ℂ,
(∀ V : Matrix.specialUnitaryGroup (Fin n) ℂ,
frobSq ((U : (Matrix (Fin n) (Fin n) ℂ)) * A * (U : (Matrix (Fin n) (Fin n) ℂ)).conjTranspose - C) ≤
frobSq ((V : (Matrix (Fin n) (Fin n) ℂ)) * A * (V : (Matrix (Fin n) (Fin n) ℂ)).conjTranspose - C)) ↔
∀ V : Matrix.specialUnitaryGroup (Fin n) ℂ, realTraceOverlap A C V ≤ realTraceOverlap A C U) ∧
(A ≠ 0 → C ≠ 0 → ∀ U : Matrix.specialUnitaryGroup (Fin n) ℂ,
(∀ V : Matrix.specialUnitaryGroup (Fin n) ℂ, lineCosSq A C V ≤ lineCosSq A C U) ↔
∀ V : Matrix.specialUnitaryGroup (Fin n) ℂ,
‖Matrix.trace (C.conjTranspose * ((V : (Matrix (Fin n) (Fin n) ℂ)) * A * (V : (Matrix (Fin n) (Fin n) ℂ)).conjTranspose))‖ ≤
‖Matrix.trace (C.conjTranspose * ((U : (Matrix (Fin n) (Fin n) ℂ)) * A * (U : (Matrix (Fin n) (Fin n) ℂ)).conjTranspose))‖) ∧
(∀ R : ℝ, 0 ≤ R →
{z : ℂ | ∃ U : Matrix.specialUnitaryGroup (Fin n) ℂ,
z = Matrix.trace (C.conjTranspose * ((U : (Matrix (Fin n) (Fin n) ℂ)) * A * (U : (Matrix (Fin n) (Fin n) ℂ)).conjTranspose))} =
{z : ℂ | ‖z‖ ≤ R} →
∃ U : Matrix.specialUnitaryGroup (Fin n) ℂ,
realTraceOverlap A C U = numericalRadius A C) := by sorry
end
end QAlgorithms.Papers.Independent.Advance50.P38
Read-back
What the Lean code literally says, in plain math · gpt-6.1-sol
For every natural number and arbitrary complex matrices , write , , and . The following four assertions hold together:
If and , then for every ,
Finally, for every real , if , then there exists such that
The first two assertions and the final conditional assertion also apply when or .
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.