The models of the Pauli basis test's stages 4 and 5 #
The soundness analysis of the Pauli basis test runs, from stage 2 on, in the register model
N.reg I of a bipartite model N: an EPR register I, whose first half A' the first player
holds and whose second half A'' the second player holds. Stages 4 and 5 adjoin more registers,
and read one state along several assignments of the registers to the players. This file builds
those models once, generically in the register type I (the analysis takes I := Anc F m), and
the maps between them. The names of the registers are the paper's: A is the first player's own
space (the algebra 𝒜), A' and A'' the two halves of the pair, Ea the padding register;
B, B', B'', Eb for the second player.
Stated in a bipartite model #
Stated in a bipartite model (Phase 5 of planning/mipco-track.md). The matrix analysis built each
of these states as an explicit vector on a product of index sets and moved between them by
Matrix.reindex along regroupings of those sets (PadReg, padState, extVec2, mirrorVec,
physLiftEquiv, outerPairEquiv, endEquiv). Here every state is an extension N.expand e of
the bipartite model by finite registers in a joint vector e, only the finite registers are ever
regrouped, and a regrouping is a local isometry or an isomorphism of models
(MIPRE/Foundations/ModelEmbedding.lean, MIPRE/Foundations/ModelIso.lean); several readings of
one extended state are readings (BipartiteModel.recut, MIPRE/Foundations/ModelReading.lean)
of one state model.
Stage 4: the Halmos-dilated padded strategy #
- The padding register
PadAnc F m d K: the ancillaDilationAncillaof the Halmos dilation of a POVM with the seeded test's answers as outcomes, with reference vectort₀at the zero answeransZero. Both players' dilations are taken on it with one commonK. - The padded model
padModel I F m d N K: the register modelN.reg Iextended by the padding register of each player, in the product basis vector|t₀⟩ ⊗ |t₀⟩. A projective strategy of it is the two one-sided Halmos dilations, and its Born probabilities are those of the reference entries (BipartiteModel.bornProb_expand_basisVec). - Where the seeded test's soundness is applied (
soundIn_padModel): the padded model is the extension ofNby the registers(Ea, A')and(Eb, A'')in the flattened vectorpadE, by associativity (BipartiteModel.assocIso), so the test's soundness in every extension ofNby a unit vector gives it in the padded model; and, the test being symmetric in the players, in the padded model ofN.swap(soundIn_padModel_swap).
Stage 5: the physical state and its readings #
- The physical registers
PhysReg I F m d K = I × (PadAnc F m d K × I)of each player,(A'', (Ea, A'))and(B'', (Eb, B')), in the physical statephysE: each player's own EPR pair(A'', A'),(B'', B'), and the two padding registers in|t₀⟩ ⊗ |t₀⟩. The physical model isphys I F m d N K = N.expand (physE …). The state is symmetric in the two players' blocks (physE_symm), a unit vector (norm_physE), and each player's pair satisfies the mirror identity (physE_mirror_A,physE_mirror_B, and their product forms): an operator on the near half acts on the state as its transpose on the far half. - The first cut
cut1 I F m d N K: the reading of the physical state model that hands the first player's far halfA''to the second player,A A' Ea | B A'' Eb (B' B''). It has the physical model's space, state and representation; only the players' operators differ, and the reading lemmascut1_πA,cut1_πBwrite its operators as physical ones. The second cut iscut1ofN.swap, a reading of the same state model with the second player's registers first. - The stage-4 models embed in the first cut:
j₁embeds the padded model, andι₁the register model (withι₁_ΦA,ι₁_ΦB), each carrying the state to the physical state. - Item 1's cut
tgt I F m d N K: the extension ofNby the padded registers, with an EPR register(A'', B'')outside;unassocregroups the physical registers into it, with no state condition. - The exchange of the blocks
blockSwapIso: the physical model ofN.swapis the physical model ofNwith the players exchanged, the players' algebras unchanged.
The padding register #
The zero answer to the seeded test at (q, 4m, d, 1): the reference answer of the Halmos
dilation of the padded strategy.
Equations
- MIPRE.QLD.ansZero = MIPRE.LIDT.CL.Answer.values fun (x : Fin 1) => 0
Instances For
The padding register Ea (and Eb): the ancilla of the Halmos dilation of a POVM with the
seeded test's answers as outcomes (DilationAncilla), with K + 1 terms. Both players' dilations
are taken on it with one common K, so that the two runs of stage 4 read one physical state.
Equations
- MIPRE.QLD.PadAnc F m d K = MIPRE.DilationAncilla (MIPRE.LIDT.CL.Answer F (4 * m) d 1) K
Instances For
The reference vector of the padding register: the zero answer, at the first term, where the
dilation compresses to the POVM it dilates (exists_pvm_dilation_ge).
Equations
Instances For
Equations
- MIPRE.QLD.instFintypePadAnc = { elems := Finset.univ.disjSum Finset.univ, complete := ⋯ }
Decidable equality on the padding register, assembled once.
Equations
The padded model, where the seeded test's soundness is applied #
The joint state of the padded model's registers, flattened: the first player's (Ea, A')
and the second player's (Eb, A''), the padding registers in |t₀⟩ ⊗ |t₀⟩ and the EPR pair
across.
Equations
- MIPRE.QLD.padE I F m d K q = MIPRE.basisVec MIPRE.QLD.t₀ MIPRE.QLD.t₀ (q.1.1, q.2.1) * MIPRE.Introspection.registerEPR I (q.1.2, q.2.2)
Instances For
The product condition of the associativity from the padded model (padModel).
The flattened padded state is the expanded vector of the two padding vectors and the pair.
The model of the Halmos-dilated padded strategy: the register model N.reg I extended by
the padding register of each player, in the product basis vector |t₀⟩ ⊗ |t₀⟩. Each player's
dilation acts on that player's padding register only, so the two dilations are one-sided; a Born
probability of the padded model is that of the two reference entries
(BipartiteModel.bornProb_expand_basisVec).
Equations
- MIPRE.QLD.padModel I F m d N K = (N.reg I).expand (MIPRE.basisVec MIPRE.QLD.t₀ MIPRE.QLD.t₀)
Instances For
The seeded test is sound in the padded model as soon as it is sound in every extension of
N by a unit vector: the padded model is the extension of N by the registers (Ea, A') and
(Eb, A'') in the flattened vector padE, by associativity (BipartiteModel.assocIso).
The seeded test is sound in the padded model of the model with the players exchanged, from
its soundness in the extensions of N: the test is symmetric in the players
(LIDT.Simul.soundIn_swap_expand_of_norm).
The physical registers and the physical state #
One player's physical registers (A'', (Ea, A')): the far half of the player's own EPR
pair, then the padding register and the near half.
Equations
- MIPRE.QLD.PhysReg I F m d K = (I × MIPRE.QLD.PadAnc F m d K × I)
Instances For
Equations
- MIPRE.QLD.instFintypePhysReg K = instFintypeProd I (MIPRE.QLD.PadAnc F m d K × I)
Decidable equality on the physical registers, assembled by hand once.
Equations
The physical state of the registers: each player's own EPR pair, (A'', A') and
(B'', B'), and the two padding registers in |t₀⟩ ⊗ |t₀⟩.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The physical state, entry by entry.
The physical state is symmetric in the two players' blocks: exchanging the blocks exchanges the two pairs and the two padding registers, which are in the same state.
The regrouping of the physical registers as the three pairs ((A'', B''), Ea) and
((A', B'), Eb), along which the physical state is a product.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The physical state, read along physSplit, is the product of the two pairs and the padding
vector.
The mirror identities of the two pairs #
The mirror identity of the first player's pair, on the registers: an operator P on the
near half A' acts on the physical state as its transpose on the far half A'', whatever acts on
the second player's registers.
The mirror identity of the second player's pair, on the registers: an operator P on the
near half B' acts on the physical state as its transpose on the far half B'', whatever acts on
the first player's registers.
The physical model and its first cut #
The physical model: N with each player's physical registers (A'', (Ea, A')) and
(B'', (Eb, B')), in the physical state. The first player's operators are the matrices over 𝒜
on the first player's physical registers, the second's over ℬ on the second's.
Equations
- MIPRE.QLD.phys I F m d N K = N.expand (MIPRE.QLD.physE I F m d K)
Instances For
The mirror identity of the first player's pair, in the physical model: an operator of the
first player that acts as P on the near half A' acts on the state as the one that acts as Pᵀ
on the far half A'', whatever the second player applies.
The mirror identity of the second player's pair, in the physical model: an operator of the
second player that acts as P on the near half B' acts on the state as the one that acts as
Pᵀ on the far half B'', whatever the first player applies.
The mirror identity of the first player's pair (reg_mirror for the physical model): a
matrix P on the near half A' acts on the state as its transpose on the far half A''.
The mirror identity of the second player's pair (reg_mirror for the physical model): a
matrix P on the near half B' acts on the state as its transpose on the far half B''.
The first cut, A A' Ea | B A'' Eb (B' B''): the reading of the physical state model that
hands the first player's far half A'' to the second player. It has the physical model's space,
state and representation; the first player's operators act on (Ea, A'), the second player's on
(A'', (B'', (Eb, B'))).
Equations
- MIPRE.QLD.cut1 I F m d N K = N.recut (MIPRE.QLD.physE I F m d K) (MIPRE.moveEquiv I (MIPRE.QLD.PadAnc F m d K × I) (MIPRE.QLD.PhysReg I F m d K))
Instances For
Reading lemma for the first player of the first cut: an operator X of (Ea, A') is the
physical operator 1_{A''} ⊗ X.
Reading lemma for the second player of the first cut: an operator that is a product
Q ⊗ Y of a matrix Q on the moved half A'' and an operator Y of the second player's
physical registers is the product of the physical first-player operator Q ⊗ 1 and the physical
second-player operator Y.
The stage-4 models in the first cut #
The second player's pair (B'', B'), adjoined to the padded model's registers with nothing for
the first player.
Equations
- MIPRE.QLD.pair2 I p = MIPRE.Introspection.registerEPR I p.2
Instances For
The registers of the first cut before their regrouping: the first player's ((), (Ea, A')),
the second player's ((B'', B'), (Eb, A'')), in the product of the adjoined pair and the flattened
padded state.
Equations
- MIPRE.QLD.pairPadE I F m d K q = MIPRE.QLD.pair2 I (q.1.1, q.2.1) * MIPRE.QLD.padE I F m d K (q.1.2, q.2.2)
Instances For
The regrouping of the second player's registers of the first cut: (A'', (B'', (Eb, B'))) as
((B'', B'), (Eb, A'')).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The relabelling condition of the first cut: the physical state, read along the first cut, is the adjoined pair times the flattened padded state, regrouped.
The padded model embeds in the first cut: flatten the padded registers
(BipartiteModel.assocEmb), adjoin the second player's pair (BipartiteModel.inertEmb), flatten
again, regroup the second player's registers (BipartiteModel.relabelEmb), and read the result
as the first cut (BipartiteModel.recutEmb).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The register model embeds in the first cut: the padding registers in |t₀⟩ ⊗ |t₀⟩
(BipartiteModel.inertEmb), then j₁. Both runs of stage 4 carry the register state to the one
physical state.
Equations
- MIPRE.QLD.ι₁ I F m d N K = (MIPRE.QLD.j₁ I F m d N K).comp ((N.reg I).inertEmb (MIPRE.basisVec MIPRE.QLD.t₀ MIPRE.QLD.t₀) ⋯)
Instances For
The first player's operators of the padded model, in the first cut: the flattening.
The first player's register operators, in the first cut: an operator Z of A' acts as
1_{Ea} ⊗ Z.
The second player's register operators, in the first cut: an operator Z of A'' acts
as Z ⊗ 1 on (A'', (B'', (Eb, B'))).
Item 1's cut, and the exchange of the blocks #
Item 1's cut (A A' Ea, B B' Eb) × (A'', B''): the extension of N by the padded
registers (Ea, A') and (Eb, B'), with an EPR register (A'', B'') outside.
Equations
- MIPRE.QLD.tgt I F m d N K = (N.expand (MIPRE.QLD.padE I F m d K)).reg I
Instances For
The isometry of unassoc: a vector of the physical registers, read with the outer registers
(A'', B'') of the two players outside and the inner ones ((Ea, A'), (Eb, B')) inside.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The components of unassocIsometry.
The associativity of item 1's cut undoes unassocIsometry.
The physical registers regrouped as item 1's cut: the inverse of the associativity
BipartiteModel.assoc of item 1's cut, with no state condition --- the two cuts hold different
pairs, so it does not carry the physical state to the state of item 1's cut, and it is read on a
transported state (LocalIsometry.bornProb_withState).
Equations
- One or more equations did not get rendered due to their size.
Instances For
The first player's physical operators, in item 1's cut: the block matrices over A''.
The second player's physical operators, in item 1's cut: the block matrices over B''.
The exchange of the blocks: the physical model of N.swap is the physical model of N
with the players exchanged, by the unitary exchanging the two blocks of registers
(BipartiteModel.swapExpand), which fixes the physical state (physE_symm). The players'
algebras are the same on both sides and the isomorphism is the identity on them, so the second
player's objects of one are the first player's of the other.
Equations
- One or more equations did not get rendered due to their size.
Instances For
The unitary of blockSwapIso exchanges the two blocks of registers.