Enlarging Lemma, §D.2 (corrected: commutators of the generated algebras, )
ProvedBaconThesis.ExchangeUniv.enlarging_lemmalie-algebrasp2o-correctedp2o-gran-per-chapterp2o-plan-bookp2o-strengthenedp2o-v1quantum-information
Let , let be linearly independent vectors, and let be sets of Hamiltonians (represented by the anti-Hermitian matrices ). Suppose that
- the real Lie algebra generated by contains acting on the orthogonal complement (and as zero on ), and
- the real Lie algebra generated by contains acting on (and as zero on its orthogonal complement).
Then:
- the Lie algebra generated by contains ;
- if , the commutators with in the Lie algebra generated by and in the one generated by generate exactly as a Lie algebra.
The lemma is the last step of the special case of the proof of Theorem C.2.1.
Read literally, with commutators of the generating sets only, the book's conclusion is false. Take with orthonormal , and : these generate on and on , but their four commutators generate only on . At , and all commutators vanish. The book's proof itself takes in the algebra generated by , which is the reading stated here.
Formalization Note extended by zero is suOn. Lie algebras are real Lie spans of matrices.
Differences from the source:
- correction: "[H1, H2] (all commutators) generates su(d) under closure as a Lie-algebra" → for d ≥ 3, the commutators [A, B] with A ∈ Lie(H1), B ∈ Lie(H2) generate exactly su(d); reason: with commutators of the generating sets it is false (d = 3, H1 = {iX_jk, iY_jk}, H2 = {iX_ij, iY_ij}: the four commutators generate a 3-dimensional algebra, su(2) on span{i, k}), and at d = 2 every commutator is 0 since su(1) = 0
- generalization: "H1 generates su(d − 1)", "H2 generates su(2)" (equalities) → the generated Lie algebras contain these; reason: weaker hypotheses with the same conclusion; specializes to the equality case
- generalization: the conclusion about [H1, H2] alone → also clause 1, Lie(H1 ∪ H2) ⊇ su(d), for every d; reason: it is the form in which Appendix C uses the lemma ("by the Enlarging Lemma it follows that we can obtain su(n − 1)"), and it covers d = 2
- reading: "the subspace of H that does not contain |i⟩" → the orthogonal complement of |i⟩; reason: the only (d − 1)-dimensional subspace singled out by |i⟩ on which su(d − 1) is meant
- reading: "|j⟩ is another state in H" → |j⟩ linearly independent of |i⟩; reason: otherwise span{|i⟩, |j⟩} is one-dimensional and su(2) cannot act on it; span{|i⟩, |j⟩} is the same as for the component of |j⟩ orthogonal to |i⟩, which the proof uses
Preamble
import Mathlib import Definitions.Def_BaconThesis_ExchangeUniv_suAlg import Definitions.Def_BaconThesis_ExchangeUniv_suOn open scoped Matrix
Formal statement
attribute [local instance 100] LieRing.ofAssociativeRing
namespace BaconThesis.ExchangeUniv
/-- **Enlarging Lemma (§D.2), with the commutator clause corrected.**
Bacon, *Decoherence, Control, and Symmetry in Quantum Computers*, arXiv:quant-ph/0305025v1, Appendix D.2, "Enlarging Lemma", p.256 (PDF p.273), proof to p.257 (PDF p.274).
Let `H = ℂ^ι` have dimension `d`, let `|i⟩ ≠ 0` and let `|j⟩` be a vector with `|i⟩, |j⟩`
linearly independent. Let `H₁, H₂` be sets of Hamiltonians (represented as anti-Hermitian `iH`)
such that the Lie algebra generated by `H₁` contains `su(d − 1)` on the orthogonal complement of
`|i⟩` (acting as zero on `|i⟩`), and the Lie algebra generated by `H₂` contains `su(2)` on
`span{|i⟩, |j⟩}` (zero outside). Then
1. the Lie algebra generated by `H₁ ∪ H₂` contains `su(d)`;
2. if `d ≥ 3`, the commutators `[A, B]` with `A` in the Lie algebra generated by `H₁` and `B` in
the Lie algebra generated by `H₂` generate exactly `su(d)` as a Lie algebra.
**Formalization Note.** `su(W)` extended by zero is `suOn`; the orthogonal complement of `|i⟩` is
`{w | ⟨i, w⟩ = 0}`. Lie algebras are real Lie spans in `Matrix ι ι ℂ`. The book's "[H₁, H₂] (all
commutators)" is read as commutators of elements of the generated algebras, which is how its proof
uses it (`X_jk ∈ H₁`, `X_ij ∈ H₂`); with literal generating sets it fails (see the mission notes),
as it does for `d = 2`. -/
theorem enlarging_lemma {ι : Type*} [Fintype ι] [DecidableEq ι]
(vi vj : ι → ℂ) (hij : LinearIndependent ℂ ![vi, vj])
(H₁ H₂ : Set (Matrix ι ι ℂ)) (hH₁ : ∀ A ∈ H₁, Aᴴ = -A) (hH₂ : ∀ A ∈ H₂, Aᴴ = -A)
(h₁ : suOn {w | star vi ⬝ᵥ w = 0} ⊆
(LieSubalgebra.lieSpan ℝ (Matrix ι ι ℂ) H₁ : Set (Matrix ι ι ℂ)))
(h₂ : suOn {vi, vj} ⊆ (LieSubalgebra.lieSpan ℝ (Matrix ι ι ℂ) H₂ : Set (Matrix ι ι ℂ))) :
suAlg ι ⊆ (LieSubalgebra.lieSpan ℝ (Matrix ι ι ℂ) (H₁ ∪ H₂) : Set (Matrix ι ι ℂ)) ∧
(2 < Fintype.card ι →
(LieSubalgebra.lieSpan ℝ (Matrix ι ι ℂ)
{C | ∃ A ∈ LieSubalgebra.lieSpan ℝ (Matrix ι ι ℂ) H₁,
∃ B ∈ LieSubalgebra.lieSpan ℝ (Matrix ι ι ℂ) H₂, C = ⁅A, B⁆} :
Set (Matrix ι ι ℂ)) = suAlg ι) := by sorry
end BaconThesis.ExchangeUniv
Source
Bacon, Decoherence, Control, and Symmetry in Quantum Computers, arXiv:quant-ph/0305025v1, Appendix D.2, Enlarging Lemma, p.256 (PDF p.273), proof to p.257 (PDF p.274)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.