REVIEW 2 major objections 6 minor 20 references
On the two-copy distillability of Werner states and a new partial trace inequality
T0 review · 2 major / 6 minor · reviewed 2026-07-31 · grok-4.5
Pith's one-line read Werner states are two-copy undistillable exactly when they are one-copy undistillable: the threshold is α ≥ −1/2 for every dimension.
desk verdict Clean linear-algebra resolution of Problem 5: Werner two-copy undistillability coincides with the one-copy threshold α≥−1/2, via a new rank-constrained partial-trace inequality that checks out line-by-line. read the letter →
The pith
A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.
The reading
What carries the argument
The rank-constrained partial-trace inequality (Theorem A): every matrix C of rank at most r satisfies ‖tr₁(C)‖_F² + ‖tr₂(C)‖_F² ≤ r ‖C‖_F² + (1/r)|tr(C)|². At r = 2 this inequality is equivalent, by a prior criterion, to two-copy undistillability of Werner states at α = −1/2.
What would settle it
Exhibit a single rank-at-most-two matrix C on two qudits for which ‖tr₁(C)‖_F² + ‖tr₂(C)‖_F² exceeds 2‖C‖_F² + (1/2)|tr(C)|², or produce an explicit Schmidt-rank-two vector that makes the two-copy partially transposed Werner state at α = −1/2 negative.
Extended reading notes
Core claim
For every local dimension d ≥ 2 the Werner state ρ(d, α) is two-copy undistillable if and only if α ≥ −1/2. In particular the two-ququart state ρ(4, −1/2) is not two-copy distillable. The one-copy and two-copy distillability regions therefore coincide.
Load-bearing premise
The argument treats as given an external equivalence that says two-copy undistillability of a Werner state is exactly the same as the rank-two case of the new inequality; if that equivalence fails for some matrices, the distillability conclusions fall even if the inequality itself holds.
Editorial extensions
If this is right
- ρ(4, −1/2) is settled as two-copy undistillable, answering the listed open problem in the negative.
- For every d the one-copy and two-copy distillability thresholds of Werner states are identical: both sit at α = −1/2.
- Any Werner state with α ≥ −1/2 remains undistillable at two copies; distillation, if it exists at all, requires more than two copies or a different family.
- The new partial-trace inequality holds for arbitrary (not necessarily Hermitian or positive) matrices of bounded rank and any local dimensions.
Reading between the lines
- The same inequality may constrain multi-copy or multipartite distillability criteria that likewise reduce to bounds on partial traces of low-rank operators.
- Because the inequality is rank-constrained rather than positivity-constrained, it can be fed directly into numerical searches over non-normal witnesses that older positive-semidefinite methods miss.
- Closing the two-copy gap for Werner states sharpens the remaining open question to whether some NPT Werner state with α ≥ −1/2 is distillable at three or more copies.
Editorial analysis
A structured set of objections, weighed in public.
Referee Report
Summary. The manuscript resolves Problem 5 of Horodecki–Rudnicki–Życzkowski's "Five Open Problems in Quantum Information Theory" (PRX Quantum 3, 010101): the two-ququart Werner state ρ(4,−1/2) is not two-copy distillable (Corollary B). The route is a new rank-constrained partial-trace inequality (Theorem A, Eq. (4)): for every C ∈ M_{d1·d2}(ℂ) of rank ≤ r — with no hermiticity, normality, or positivity assumption — ‖tr₁C‖²_F + ‖tr₂C‖²_F ≤ r‖C‖²_F + (1/r)|tr C|². Combined with Costa Rico's equivalence ([10, Thm. 1], quoted as Theorem 2.1) at r=2, this yields the full classification (Corollary C): ρ(d,α) is two-copy undistillable iff α ≥ −1/2, so the one- and two-copy distillability regions coincide for all d. Theorem A is proved in §8 via a balanced rank-r decomposition (SVD + Parker–Fillmore constant-diagonal form, Lemma 7.1), the known rank-one inequality (Lemma 6.4), a four-state identity of Eltschka–Siewert recast as crossed polarization (Cor. 6.7), and a Lagrange-identity remainder.
Significance. If it holds, this settles a named open problem from a high-visibility list and closes the two-copy distillability question for Werner states in all dimensions — the first all-d result at the boundary α = −1/2, going beyond the α ≥ −1/3 range of Costa Rico–Wolf. Theorem A itself is of independent matrix-analysis interest: it extends the positive-matrix case (Audenaert) and the rank-one case to arbitrary rank and arbitrary, possibly non-normal, C, with a clean r‖C‖²_F + (1/r)|tr C|² form. The proof is elementary and, modulo one published equivalence theorem, self-contained; the manuscript is unusually transparent about provenance (the dependency table and graph in §4 correctly flag Lemma 6.5 as unused, and every external lemma is pinned to a numbered source). I verified the §8 derivation line by line and found it correct; the six steps are short and each is checkable with standard tools. No numerical or formal verification is shipped, but none is needed given the proof's length.
major comments (2)
- [§2, Theorem 2.1; proofs of Cors. B and C] §2, Theorem 2.1 (the bridge used in Corollaries B and C): the entire physical conclusion rests on Costa Rico's equivalence, invoked as a black box. Two things should be stated explicitly. (i) Only the 'if' direction (inequality (3) ⟹ undistillability) is ever used — the converse direction is not, since distillability for α<−1/2 comes from the cited one-copy result. (ii) The criterion quantifies over ALL C of rank ≤ 2 in M_{d²}(ℂ), including non-normal C; this is essential because Theorem A's contribution is precisely the non-normal generality (the normal case being covered by prior work). The quoted statement in §2 does quantify over all C, so the bridge is intact, but the manuscript should say (i)–(ii) in one sentence so the reader can confirm the quantifiers match without consulting [10].
- [§2, paragraph after Theorem 2.1; §9] §2 remarks that Costa Rico states the criterion 'for n copies' and that 'only n=2 is used here.' A careful reader will ask the obvious follow-up: Theorem A is proved for general rank r, and at r=n it has exactly the form of (3) with α=−1/n; moreover, simple algebra analogous to Eq. (76) gives n a + b/n ≤ 2a + b/2 using b ≤ n a, so if the n-copy criterion were the norm form (3) at rank ≤ n, Corollary C would apparently extend to all copy numbers. Since the manuscript deliberately does not claim this, the n-copy criterion presumably differs from rank-n (3) in some essential way — but the paper gives the reader no way to see why. Please add a remark explaining precisely why the general-r strength of Theorem A does (or does not) propagate beyond two copies; otherwise the scope of Corollary C is easy to misread.
minor comments (6)
- [§8.1] §8.1, sentence after Eq. (55): 'Our aim in steps 2 and 3 in then to bound' — 'in' should be 'is'.
- [§9, proof of Cor. C, Eq. (76)] Proof of Corollary C, Eq. (76): the parenthetical justification 'bt ≤ 2at ≤ a ≤ 2a' is hard to parse. Simpler: tb ≤ b (since t ≤ 1/2 < 1) and b ≤ 2a by (23) at rank 2, hence 2a ≥ tb, i.e. a/t ≥ b/2.
- [§8 / Theorem A] Theorem A: a brief remark on equality conditions would be valuable — e.g., when does equality hold in (4) (the Lagrange term vanishes iff all δ_i coincide, but the C–S and rank-one steps in (58)/(61)–(62) may be strict)? Even a conjectural statement or small-d examples would help readers gauge sharpness.
- [§1.3] The note on concurrent work (Fu et al., arXiv:2607.21367; Bharti–Gajjala–Haug in preparation) should be upgraded to formal citations with a sentence comparing proof strategies once those manuscripts are public, both for priority clarity and because the manuscript already notes a shared ingredient (a variant of Lemma 6.6).
- [§5.6] Notation: tr₁ traces out the FIRST factor (yielding an operator on the second), which is the reverse of the most common convention; it is defined unambiguously in (13)–(14), but a one-line warning would prevent misreadings.
- [§6.3, Lemma 6.3] Reference [16] (Hardy–Littlewood–Pólya, Theorem 7) for the Lagrange identity: please confirm the theorem number matches the cited edition (1934 vs 1952 2nd ed.), since numbering differs between editions.
Circularity Check
No circularity: Theorem A is derived from independent linear-algebra lemmas; Costa Rico's external equivalence is used only as a one-way bridge.
full rationale
The central claim (Theorem A) is proved in §8 from a balanced rank-r decomposition (Lemma 7.1, built from SVD and Parker–Fillmore), the rank-one partial-trace bound (Lemma 6.4), the Eltschka–Siewert four-state identity (Lemma 6.6 / Cor. 6.7), and the Lagrange variance identity (Lemma 6.3). None of these ingredients encode two-copy undistillability or the target constant −1/2. Corollaries B and C then invoke only the “if” direction of Costa Rico’s external equivalence (Theorem 2.1): the newly proved inequality implies undistillability. Distillability for α<−1/2 is taken from the classical one-copy result, not from the converse of Theorem 2.1. There are no fitted parameters, no self-defining normalizations, and no load-bearing self-citations. The derivation is self-contained against its stated external inputs.
Assumptions & free parameters
assumptions (7)
- domain assumption Costa Rico equivalence: ρ(d,α) is two-copy undistillable iff every rank-≤2 C satisfies ‖tr1(C)‖_F²+‖tr2(C)‖_F² ≤ (1/|α|)‖C‖_F² + |α||tr(C)|² (Theorem 2.1 / [10]).
- standard math Parker–Fillmore / Fillmore constant-diagonal unitary form: every square complex matrix is unitarily similar to one with constant diagonal equal to the average eigenvalue (trace/r).
- standard math Singular-value decomposition of any complex matrix.
- domain assumption Rank-one partial-trace inequality ‖Φ(xy*)‖² ≤ ‖x‖²‖y‖² + |y*x|² (Costa Rico / Costa Rico–Wolf).
- standard math Eltschka–Siewert four-state marginal identity relating Frobenius products of partial traces.
- domain assumption Two-copy distillability definition: existence of a Schmidt-rank-≤2 vector ψ with ⟨ψ|(ρ^{T_B})⊗2|ψ⟩<0.
- domain assumption Werner-state NPT/separability thresholds: NPT iff α<−1/d, separable iff α≥−1/d; one-copy undistillable iff α≥−1/2.
Cite this review
Pith. "Pith review of On the two-copy distillability of Werner states and a new partial trace inequality." pith.science (2026). https://pith.science/paper/53K57DJS
@misc{pith2026260724309,
author = {Pith},
title = {Pith review of: On the two-copy distillability of Werner states and a new partial trace inequality},
year = {2026},
howpublished = {\url{https://pith.science/paper/53K57DJS}},
note = {Machine review of arXiv:2607.24309}
}
abstract
Problem 5 in {\it Five Open Problems in Quantum Information Theory} [PRX Quantum 3, 010101 (2022)], asks whether the two-ququart Werner state $\varrho(4,-\tfrac12)$ is two-copy distillable, where $\varrho(d,\alpha)=(I+\alpha F)/(d^2+\alpha d)$. We answer it in the negative. To this end, we show the following stronger statement: for all $C\in M_{d_1d_2}(\mathbb{C})$ of rank at most $r \le d_1 d_2$, $\mathrm{tr}_1(C)\|_F^2+\|\mathrm{tr}_2(C)\|_F^2 \le r\|C\|_F^2+\frac{1}{r}|\mathrm{tr}(C)|^2$. A result by Costa Rico on the equivalence of this inequality with two-copy undistillability at $r = 2$ then settles Problem 5: $\varrho(4,-\tfrac{1}{2})$ is not two-copy distillable. Furthermore, we show that $\varrho(d,\alpha)$ is two-copy undistillable for every $d\ge2$, if and only if $\alpha\ge-\tfrac{1}{2}$. Thus, the one and two-copy distillability regions of $\varrho(d,\alpha)$ coincide. These results have been found and written up with AI tools, pointing towards a structural change affecting the field of quantum information and computation.
Reference graph
Works this paper leans on
-
[10]
New Partial Trace Inequalities and Distillability of Werner States,
P. Costa Rico, “New Partial Trace Inequalities and Distillability of Werner States, ”Letters in Mathematical Physics115, 47 (2025). doi:10.1007/s11005-025-01935-y; arXiv:2310.05726
arXiv 2025
-
[1]
Quantum states with Einstein–Podolsky–Rosen correlations admitting a hidden-variable model,
R. F. Werner, “Quantum states with Einstein–Podolsky–Rosen correlations admitting a hidden-variable model, ”Physical Review A40, 4277–4281 (1989). doi:10.1103/PhysRevA.40.4277
-
[2]
Mixed-state entanglement and distillation: Is there a ‘bound’ entanglement in nature?,
M. Horodecki, P. Horodecki, and R. Horodecki, “Mixed-state entanglement and distillation: Is there a ‘bound’ entanglement in nature?, ”Physical Review Letters80, 5239–5242 (1998). doi:10.1103/PhysRevLett.80.5239
-
[3]
Separability criterion for density matrices,
A. Peres, “Separability criterion for density matrices, ”Physical Review Letters77, 1413–1415 (1996). doi:10.1103/PhysRevLett.77.1413
-
[4]
Reduction criterion of separability and limits for a class of distillation protocols,
M. Horodecki and P. Horodecki, “Reduction criterion of separability and limits for a class of distillation protocols, ”Physical Review A59, 4206–4216 (1999). doi:10.1103/PhysRevA.59.4206
-
[5]
Distillability and partial transposition in bipartite systems,
W. Dür, J. I. Cirac, M. Lewenstein, and D. Bruß, “Distillability and partial transposition in bipartite systems, ” Physical Review A61, 062313 (2000). doi:10.1103/PhysRevA.61.062313; arXiv:quant-ph/9910022
arXiv 2000
-
[6]
Distillability of Werner states using entanglement witnesses and robust semidefinite programs,
R. O. Vianna and A. C. Doherty, “Distillability of Werner states using entanglement witnesses and robust semidefinite programs, ”Physical Review A74, 052306 (2006). doi:10.1103/PhysRevA.74.052306
-
[7]
A Few Steps More Towards NPT Bound Entan- glement,
Ł. Pankowski, M. Piani, M. Horodecki, and P. Horodecki, “A Few Steps More Towards NPT Bound Entan- glement, ”IEEE Transactions on Information Theory56(8), 4085–4100 (2010). doi:10.1109/TIT.2010.2050810
arXiv 2010
Show all 20 references
-
[8]
Joint Schmidt-type decomposition for two bipartite pure quantum states,
C. Eltschka and J. Siewert, “Joint Schmidt-type decomposition for two bipartite pure quantum states, ” Physical Review A101, 022302 (2020). doi:10.1103/PhysRevA.101.022302; arXiv:1907.07976
2020 arXiv
-
[9]
Five Open Problems in Quantum Information Theory,
P. Horodecki, Ł. Rudnicki, and K. Życzkowski, “Five Open Problems in Quantum Information Theory, ” PRX Quantum3, 010101 (2022). doi:10.1103/PRXQuantum.3.010101; arXiv:2002.03233
2022 arXiv
- [11]
-
[12]
On Similarity and the Diagonal of a Matrix,
P. A. Fillmore, “On Similarity and the Diagonal of a Matrix, ”The American Mathematical Monthly76(2), 167–169 (1969). doi:10.1080/00029890.1969.12000162
1969
-
[13]
Sets of Complex Numbers Associated with a Matrix,
W. V. Parker, “Sets of Complex Numbers Associated with a Matrix, ”Duke Mathematical Journal15(3), 711–715 (1948). doi:10.1215/S0012-7094-48-01560-9
1948 doi
-
[14]
The vec-permutation matrix, the vec operator and Kronecker products: a review,
H. V. Henderson and S. R. Searle, “The vec-permutation matrix, the vec operator and Kronecker products: a review, ”Linear and Multilinear Algebra9(4), 271–288 (1981). doi:10.1080/03081088108817379
1981 doi
-
[15]
Subadditivity of q-entropies for q >1,
K. M. R. Audenaert, “Subadditivity of q-entropies for q >1, ”Journal of Mathematical Physics48, 083507 (2007). doi:10.1063/1.2771542
2007 doi
-
[16]
G. H. Hardy, J. E. Littlewood, and G. Pólya,Inequalities, Cambridge University Press, Cambridge (1934); 2nd ed. (1952)
1934
-
[17]
M. A. Nielsen and I. L. Chuang,Quantum Computation and Quantum Information, 10th anniversary ed., Cambridge University Press, Cambridge (2010). doi:10.1017/CBO9780511976667
2010 doi
-
[18]
Watrous,The Theory of Quantum Information, Cambridge University Press, Cambridge (2018)
J. Watrous,The Theory of Quantum Information, Cambridge University Press, Cambridge (2018). doi:10.1017/9781316848142
2018 doi
-
[19]
R. A. Horn and C. R. Johnson,Topics in Matrix Analysis, Cambridge University Press, Cambridge (1991)
1991
-
[20]
Momentum
R. Bhatia,Matrix Analysis, Springer, New York (1997). ON THE TWO-COPY DISTILLABILITY OF WERNER STATES AND A NEW PARTIAL TRACE INEQUALITY 13 1 Department of Mathematical Sciences,University of Copenhagen, Copenhagen, 2100, Denmark Email address:tcf@math.ku.dk 2 Division of Quan...
1997
Reviewed July 31, 2026 · model on record in the stance chip above.
Discussion (0). Sign in to comment.