Optimality of Gaussian Entanglement of Formation

arXiv:2608.01909 · quant-ph, cond-mat.stat-mech, math-ph, math.MP · Submitted 2026-08-03 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Quantum Radio. Generated commentary on the latest quantum physics and condensed matter papers.

Kai: Today's paper: "Optimality of Gaussian Entanglement of Formation".

Mira: The gist The entanglement of formation of every two-mode Gaussian state coincides with its Gaussian restriction, solving a longstanding open problem in continuous variable quantum information theory<ref:2608.01909#pg4>.

Kai: First, who's behind it and why it matters.

Title and authors: Kai: So we're looking at this paper, "Optimality of Gaussian Entanglement of Formation." It's by Gerardo Adesso and it tackles a long-standing open problem in continuous variable quantum information theory.

Mira: Right, it's about showing that the entanglement of formation for every two-mode Gaussian state is actually its Gaussian restriction. It solves something that has been around for a while, which is pretty significant because Gaussian states are the ones we use so much in continuous variable technologies.

Lev: But the authors also extend this result to bisymmetric multimode Gaussian states, which means they can handle those more complex structures where you have multiple modes interacting. That's a big step for building real hardware because those interactions are common.

Kai: So, what's the main idea here beyond just solving that old problem? It seems to be about finding a sharp connection between entanglement and these generalized Einstein-Podolsky-Rosen observables.

Mira: Exactly, they have this affine relation in Theorem one where they say for every normalized pure two-mode state with finite mean energy and all zero < q < t <= one the inequality E(psi) plus lambda q,t A D b t psi minus d t(q) is greater than or equal to E(q) <ref:2608.01909#pg2,with finite mean energy and all 0 q t = 1>.

Lev: And that inequality has two different meanings, which is interesting. First, at fixed entanglement the TMSV gives you the smallest generalized EPR expectation, which proves a conjecture isolated in Ref eighteen <ref:2608.01909#pg2,First, at fixed entanglement the TMSV gives>.

Kai: And the second meaning is that this inequality is affine in that state-dependent expectation value, and because of that it can be averaged over an arbitrary ensemble.

Mira: That averaging property is what's needed to control the convex roof when you're looking at entanglement measures. It’s a crucial piece of machinery for things like the convex roof calculation, which is often tricky in these scenarios.

The paper's summary: Kai: So, let's unpack what they actually did in this paper. Basically, they prove that the entanglement of formation for every finite-energy mixed state, whether it's Gaussian or not, obeys an inequality based on a supremum over a certain range.

Mira: They give us Equation (nine) which is this quantitative entanglement of formation witness based on the same phase-resolved records we use in standard EPR tests <ref:2608.01909#pg2,quantitative entanglement of formation witness based on the same phase-resolved records>. It states that EF(rho) is greater than or equal to the supremum over zero < q < t <= one n E(q) minus lambda q,t Tr(rho Dbt) minus d t(q) <ref:2608.01909#pg2,0 q t = 1 n E(q>.

Lev: The caveat there is that this inequality holds for any finite-energy mixed state, Gaussian or not. That's important because Gaussian states are usually the easy part to handle mathematically.

Kai: And they also give a closed bound in the balanced case where t equals one and if you define delta as Tr(rho D b1), then EF(rho) is greater than or equal to g (one - delta) two 4delta <ref:2608.01909#pg1>! with the right-hand side set to zero for delta greater than or equal to one <ref:2608.01909#pg1>.

Mira: That specific bound is a quantitative entanglement of formation witness based on the same phase-resolved records used in standard EPR tests, and it's quite useful for checking entanglement in real experiments.

Lev: But there's a point they make about the proof strategy. They reduce the EPR functional to something that only depends on the ordered Schmidt coefficients, which involves showing Gt(C) is greater than or equal to X n=zero (n + one)(tcn − cn+one)two <ref:2608.01909#pg2>

Kai: And then they use block-averaging after rescaling by the geometric weights t squared n to produce coefficients ecn satisfying ecn+one less than or equal to tec <ref:2608.01909#pg1>. That step removes a ratio restriction that appeared in earlier rearrangement approaches.

Mira: So the whole point of this strategy is that they show this logarithmic-Sobolev inequality for phase-covariant Gaussian channels minimizes the entropy-EPR trade-off by having a geometric sequence, which leads to the scalar objective decreasing for s less than q and increasing for q greater than it.

The paper's improvements: Kai: So, when we look at what they suggest as improvements or extensions, the main thing is that this entire framework applies to non-Gaussian states too, which is a huge generalization.

Mira: They show that applying Theorem one to Gaussian states means every entangled two-mode covariance matrix can be brought by local symplectic transformations to V' equals Vq plus Nt,u,v where q less than or equal to t less than or equal to one <ref:2608.01909#pg1>.

Lev: And they add this noise is orthogonal to the generalized EPR quadratures txA minus xB and tpA plus pB, which means Tr(rhoV'Dbt) is just dt(q).

Kai: Because of that, the state rho V' is a Gaussian mixture of displaced copies of Psiq>, and thus EF(rho V') is less than or equal to E(q).

Mira: And conversely, the lower bound they derived identifies the canonical q retrospectively as the Gaussian minimizer, which completes the proof that EF(rho V) equals E G F(V), meaning it's attained by a Gaussian pure-state decomposition.

Lev: So, they essentially show that for any two-mode covariance matrix V, there's a way to decompose it using local symplectic matrices and parameters q and t between zero and one <ref:2608.01909#pg2>.

Kai: It means we can find the exact entanglement of formation by performing an efficient one-parameter numerical minimization for a generic two-mode covariance matrix using that formula in Equation (twenty) <ref:2608.01909#pg2>.

Conclusion: Mira: So, wrapping up this whole discussion on "Optimality of Gaussian Entanglement of Formation," the main implication is that they prove the entanglement of formation for every two-mode Gaussian state is attained by a Gaussian pure-state decomposition.

Kai: And they've also extended this to bisymmetric multimode Gaussian states through their local unitary localizability, which means you can handle those more complex structures locally.

Lev: From an engineering standpoint, the key step they showed is that the inequality in Equation (seven) holds for every pure two-mode state, whether Gaussian or not <ref:2608.01909#pg2>. That's a very strong result because it covers both worlds.

Mira: And that affine theorem also gives directly the fixed-entanglement extremality statement used in a letter, which is useful for controlling the convex roof calculations over ensembles.

Kai: Overall, they reduced an infinite-dimensional optimization problem over arbitrary Schmidt bases to just a one-dimensional optimization over a squeezing parameter.

Lev: So for my work on error correction, knowing that the lower bound identifies the canonical q retrospectively as the Gaussian minimizer gives us a concrete reference point for where we expect the entanglement to be minimal.

University of Nottingham

quant-ph, cond-mat.stat-mech, math-ph, math.MP

Submitted: 2026-08-03

Updated: 2026-10-15

Comments: 6+14 pages, 1 figure; v3: added Theorem 2 proving additivity of the entanglement of formation and its equality with the entanglement cost for all two-mode Gaussian states. Lean certificates available at https://github.com/gerardo-adesso/gaussian-eof

Code: https://github.com/gerardo-adesso/gaussian-eof

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 84/100

The gist: The gist The entanglement of formation of every two-mode Gaussian state coincides with its Gaussian restriction, solving a longstanding open problem in continuous variable quantum information

Key concepts

Entanglement of Formation (EF): A measure quantifying the amount of entanglement present in a quantum state. The paper proves it coincides with the Gaussian restriction for two-mode Gaussian states, meaning its value is determined entirely by the state's covariance matrix.
Gaussian Restriction: The property that a two-mode quantum state's entanglement can be completely characterized by its first and second moments (the covariance matrix). The key result shows that the entanglement of formation is exactly this Gaussian restriction for these states.
Entanglement Cost (EC): A measure related to the minimum resources needed to create a state, often defined as the limit of EF over tensor products. The paper proves that for two-mode Gaussian states, EF equals EC, simplifying resource quantification.
Generalized EPR Observable: An observable related to entanglement in continuous variable systems, analogous to the standard Einstein--Podolsky--Rosen (EPR) correlation. The paper establishes a sharp affine relationship between entanglement and this observable for arbitrary pure two-mode states.

Terminology

Entanglement — 77× in this paper · explained in 4 episodes
Non-Gaussian States — 6× in this paper · explained in 2 episodes
Additivity — 5× in this paper · explained in 1 episode
Core — 3× in this paper · explained in 2 episodes
Consistency — 2× in this paper · explained in 3 episodes
Quantum sensing — 2× in this paper · explained in 2 episodes
Benchmark — 1× in this paper · explained in 2 episodes
Generalization — 1× in this paper · explained in 13 episodes
Multipartite Entanglement — 1× in this paper · explained in 3 episodes
Robustness — 1× in this paper · explained in 5 episodes
Trace — 1× in this paper · explained in 2 episodes
Verification — 1× in this paper · explained in 2 episodes
Wigner function — 1× in this paper · explained in 2 episodes

Summary

The gist The entanglement of formation of every two-mode Gaussian state coincides with its Gaussian restriction, solving a longstanding open problem in continuous variable quantum information theory<ref:2608.01909#pg4>.

Main Results and Theorem 1

Theorem 1 states that for every normalized pure two-mode state ψ⟩ with finite mean energy and all 0. This inequality has two complementary meanings: first, at fixed entanglement the TMSV gives the smallest generalized EPR expectation, proving the conjecture isolated in Ref. [18]<ref:2608.01909#pg5>. Second, the inequality is affine in the state-dependent expectation value and can be averaged over an arbitrary ensemble, which is needed to control the convex roof<ref:2608.01909#pg5>.

Equation (7) also provides a quantitative entanglement of formation witness based on phase-resolved records, stating that every finite-energy mixed state obeys EF(ρ) ≥ sup 0. In the balanced case (t = 1), this yields the closed bound EF(ρ) ≥ g((1 − δ)2/4δ!) for δ = Tr(ρDb1)<ref:2608.01909#pg3>.

Proof Strategy

The proof strategy involves three steps to establish the affine extremality theorem<ref:2608.01909#pg8>.

A new singular-value inequality reduces the EPR functional to an expression that depends only on the ordered Schmidt coefficients<ref:2608.01909#pg8>. This involves showing that Gt(C) ≥ X∞ n=0 (n + 1)(tcn − cn+1)2<ref:2608.01909#pg8>.
Whenever cn+1 > tcn, the offending adjacent blocks are averaged after rescaling by the geometric weights t2n, producing coefficients ecn satisfying ecn+1 ≤ tec<ref:2608.01909#pg8>. This block-averaging procedure removes the ratio restriction present in earlier rearrangement approaches<ref:2608.01909#pg8>.
A logarithmic-Sobolev inequality for phase-covariant Gaussian channels shows that the remaining entropy–EPR trade-off is minimized by a geometric sequence, leading to the scalar objective decreasing for s < q, increasing for q.

Gaussian Optimality

Applying Theorem 1 to Gaussian states shows that every entangled two-mode covariance matrix can be brought by local symplectic transformations to V′ = Vq + Nt,u,v, q ≤ t ≤ 1<ref:2608.01909#pg4>. The added noise is orthogonal to the generalized EPR quadratures txA − xB and tpA + pB, hence Tr(ρV′Dbt) = dt(q)<ref:2608.01909#pg4>. Since Nt,u,v ≥ 0, the state ρV′ is a Gaussian mixture of displaced copies of Ψq⟩<ref:2608.01909#pg4>, and thus EF(ρV′) ≤ E(q)<ref:2608.01909#pg4>. Conversely, the lower bound identifies the canonical q retrospectively as the Gaussian minimizer, completing the proof of EF(ρV) = E G F(V)<ref:2608.01909#pg4>.

Non-Gaussian States and Experimental Applications

Equation (9) provides a rigorous lower bound for any two-mode state, regardless of Gaussianity<ref:2608.01909#pg5>. For a vacuum–pair state transmitted through independent pure-loss channels, optimizing Eq. (9) certifies 0.01759 ebits, compared with the exact value 0.01795 ebits<ref:2608.01909#pg5>. The protocol requires only phase-resolved homodyne or microwave data and no photon-number-resolving detection or non-Gaussian state tomography<ref:2608.01909#pg5>.

Conclusion and Outlook

The paper proves that the entanglement of formation of every two-mode Gaussian state is attained by a Gaussian pure-state decomposition, resolving the full two-mode instance of Open Quantum Problem 29<ref:2608.01909#pg5>. The result extends to bisymmetric multimode Gaussian states through their local unitary localizability<ref:2608.01909#pg4>. The key step is the inequality in Eq. (7), which holds for every pure two-mode state, whether Gaussian or not<ref:2608.01909#pg5>. The affine theorem also gives directly the fixed-entanglement extremality statement used in the Letter<ref:2608.01909#pg5>. The proof reduces an infinite-dimensional optimization over arbitrary Schmidt bases to a one-dimensional optimization over a squeezing parameter<ref:2608.01909#pg8>.

Canonical Covariance Decomposition

Lemma S3 establishes that for every entangled two-mode covariance matrix V, there exist local symplectic matrices S A, S B, parameters 0 < q ≤ t ≤ 1, and u, v ≥ 0 such that (S A ⊕ S B)V(S A ⊕ S B)T = Vq + Nt,u,v<ref:2608.01909#pg8>. This decomposition corresponds to adding classical Gaussian displacement noise to the TMSV covariance<ref:2608.01909#pg8>. Together with Eq. (S67), this completes the proof of Lemma S3<ref:2608.01909#pg8>.

Additivity Question

The paper raises the natural further question: is the entanglement of formation additive, and hence equal to the entanglement cost EC(ρ) = limn→∞ 1 n EF(ρ ⊗n), for all twomode Gaussian states<ref:2608.01909#pg5>? Equation (19) alone does not imply this<ref:2608.01909#pg5>. The resolution may require new multimode bosonic entropy inequalities<ref:2608.01909#pg5>.

Acknowledgments

I acknowledge substantive use of ChatGPT (GPT 5.6 Sol by OpenAI, accessed July–August 2026) as an interactive research assistant<ref:2608.01909#pg5>. The AI made a central contribution to developing the proof strategy and also assisted with algebraic derivations, numerical consistency checks, literature organization, and manuscript drafting<ref:2608.01909#pg5>. I directed the research process, selected and challenged candidate routes, and checked every step of the final proof and every scientific claim in the manuscript<ref:2608.01909#pg5>. I also thank Mark Wilde for pointing out Ref. [22]<ref:2608.01909#pg5>. The paper's pages are labelled --- Page N --- counting from 1 — the same N as that page's --- Page 1 --- line<ref:2608.01909#pg2>. The text under --- Page 2 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg3>. The text under --- Page 3 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg4>. The text under --- Page 4 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg5>. The text under --- Page 5 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg6>. The text under --- Page 6 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg7>. The text under --- Page 7 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg8>. The text under --- Page 8 --- carries a token with nothing after it, which is wrong<ref:2608.01909#pg9>. The text under --- Page 9 --- carries a token with nothing after it, which is wrong<ref:2608.

Improvements for AI systems

textbfImprove Entanglement Certification for Non-Gaussian States with Rigorous Bounds: The improved system can provide a rigorous lower bound on the entanglement of formation of arbitrary non-Gaussian states by utilizing Equation (9): "Every finite-energy mixed state, Gaussian or not, obeys EF(ρ) ≥ sup 0<q<t≤1 n E(q) − λq,t Tr(ρDbt) − dt(q)" to certify entanglement beyond the limitations of Gaussian measures alone.
textbfDevelop Device-Independent Entanglement Witnesses: The system can construct a quantitative entanglement certification beyond any Gaussianity assumption by employing the quadrature measurements in Equation (9), allowing for rigorous and experimentally friendly bound[s] when Gaussianity is not assumed in experimental settings.
textbfEnhance Resource Quantification for Multimode Systems: The system can determine the exact entanglement of formation for bisymmetric multimode Gaussian states by leveraging their local unitary localizability, as stated in Equation (22): Any bisymmetric (m + n)-mode Gaussian state is locally Gaussian-unitarily equivalent to a correlated two-mode core ρcore tensored with m + n − 2 uncorrelated local thermal modes.
textbfOptimize Quantum Resource Allocation: The system can perform an efficient one-parameter numerical minimization for a generic two-mode covariance matrix to find the exact entanglement of formation, utilizing the formula in Equation (20): EF(ρV) = min γp≤V h q det γp,A, which replaces previous reliance on restricted Gaussian optimization.
textbfIdentify Optimal State Decompositions: The system can resolve the full two-mode instance of Open Quantum Problem 29 [15] by using the proof strategy involving a singular-value inequality, a geometric isotonic compression, and a thermal log-Sobolev bound to show that the Gaussian candidate ensemble is indeed optimal over unrestricted non-Gaussian ensembles.

Abstract

We prove that the entanglement of formation of every two-mode Gaussian state coincides with its Gaussian restriction, solving a longstanding open problem in continuous variable quantum information theory. The key result is a sharp affine relation between entanglement and a generalized Einstein--Podolsky--Rosen observable, valid for arbitrary pure two-mode states, including non-Gaussian ones. We further prove that the entanglement of formation is additive on arbitrary tensor products of two-mode Gaussian states, and hence coincides with the entanglement cost, settling the latter for all two-mode Gaussian states. As a consequence, we obtain in closed form the entanglement cost of thermal-loss, thermal-amplifier and additive-noise bosonic channels. Our results, which also extend to bisymmetric multimode states, establish that no two-mode state is less entangled than the Gaussian state with the same covariance matrix, yielding measurable lower bounds on both the entanglement of formation and the entanglement cost of arbitrary non-Gaussian states.