Metricity of separable quantum optimal transport with the Hilbert-Schmidt cost

arXiv:2609.03769 · quant-ph, math-ph, math.MP · Submitted 2026-09-03 · Read on arXiv

Listen

Radio episode about this paper

Transcript

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

Kai: I'm Kai, and with me are Mira and Lev, guest researcher.

Mira: Today's paper: "Metricity of separable quantum optimal transport with the Hilbert-Schmidt cost".

Kai: The paper establishes that a specific distance derived from separable quantum optimal transport costs defines a genuine metric between density matrices,

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

Title and authors: Kai: Following up on that idea of a genuine metric, the paper lays out exactly what they did by summarizing the main contribution of "Metricity of separable quantum optimal transport with the Hilbert-Schmidt cost" <ref:2609.03769#pg0>.

Mira: They summarize that they proved that d sep, defined via the separable quantum two-Wasserstein distance, satisfies the triangle inequality and is indeed a genuine distance on Dn <ref:2609.03769#pg1>.

Lev: That means they managed to overcome the known "quantum marginal obstruction" that plagued previous constructions when trying to use coupling-based versions of quantum optimal transport <ref:2609.03769#pg1>.

Kai: Exactly, and they achieved this by moving away from the standard gluing argument and instead employing convex-roof duality along with a dimension-independent interpolation result for Hermitian operators <ref:2609.03769#pg2>.

Mira: The summary also highlights that the proof involves expressing the squared transport cost as a convex roof, T(rho, sigma) = (Tr(K rho) + Tr(M sigma)) over Hermitian operators K and M satisfying a specific feasibility constraint <ref:2609.03769#pg2>.

Lev: That dual formulation is interesting because it gives us an alternative computational path; if you can solve the dual problem involving K and M, you bypass the direct minimization over all separable couplings <ref:2609.03769#pg2>.

Kai: And they show that this dual representation leads to T(rho, sigma) at most (d sep(rho, tau) + d sep(tau, sigma)) squared, which is the key step leading to the triangle inequality for d sep <ref:2609.03769#pg2>.

Mira: They also point out an equivalent formulation where this metricity result is shown by demonstrating that the corresponding separable SWAP fidelity, F sep(rho, sigma) = omega in sep(rho, sigma) Tr(S omega), satisfies p/one - F sep as a metric on Dn <ref:2609.03769#pg0>.

Lev: If the fidelity metric works, it means we can use measures of how well states are preserved under SWAP operations to quantify the distance between them <ref:2609.03769#pg1>.

The paper's summary: Kai: Now let's talk about the improvements they suggest, which really focus on how this new metric structure helps us move forward with quantum state comparison <ref:2609.03769#pg0>.

Mira: One major improvement is the introduction of a new distance function, d sep(rho, sigma) = sqrt one over two X j p j d HS(P X j, P Z j), which relies on sequences of triples (p j, X j, Z j) <ref:2609.03769#pg1>.

Lev: I'm interested in the practical implication of that definition; how does that specific minimization over sequences of pure states translate into something we could actually implement on hardware? <ref:2609.03769#pg1>.

Kai: The improvement is that this distance, d sep, is equivalent to the folded optimal transport distance D two associated with sqrt one over two d HS, which they show is precisely the Beatty–França quantity W d HS/sqrt two squared <ref:2609.03769#pg0>.

Mira: This equivalence means that passing to the chain envelope doesn't actually decrease the one-step separable transport cost, because D two equals d sep <ref:2609.03769#pg1>.

Lev: For error correction, this tells us that we don't lose distance when we simplify the coupling structure to a chain envelope; it maintains the actual transport cost measure <ref:2609.03769#pg1>.

Kai: Another key improvement is providing a framework for "gluing" quantum transport problems without needing classical marginal gluing, using an interpolating operator derived from convex-roof duality and the non-strict Finsler lemma <ref:2609.03769#pg2>.

Mira: That mechanism allows them to establish metric properties like the triangle inequality even when the intermediate state's pure-state decomposition isn't uniquely defined, which was the core issue before <ref:2609.03769#pg1>.

The paper's improvements: Kai: So to wrap up "Metricity of separable quantum optimal transport with the Hilbert-Schmidt cost," the main conclusion is that d sep satisfies the triangle inequality and is a genuine distance on Dn <ref:2609.03769#pg0>.

Mira: They also established that this result confirms that the folded optimal transport distance D two associated with sqrt one over two d HS is exactly equal to the Beatty–França quantity W d HS/sqrt two squared, which equals d sep <ref:2609.03769#pg0>.

Lev: For running this on real hardware, this metricity means we have a mathematically sound way to measure state separation, which is a necessary foundation before we can even start designing protocols that require measuring distances between quantum states <ref:2609.03769#pg2>.

Kai: It’s a solid piece of math because it gives us this rigorous metric structure using separable couplings and pure-state geometry, resolving the prior obstruction <ref:2609.03769#pg1>.

Mira: The implication for our field is that we now have a reliable way to compare quantum states based on the minimal effort required to generate them from common pure-state resources <ref:2609.03769#pg1>.

Lev: I just want to reiterate that if this metricity holds, it opens up a path for more rigorous analysis of how quantum transport costs behave in complex systems, which could be important for understanding error propagation in larger quantum circuits <ref:2609.03769#pg2>.

Kai: Alright folks, we've covered a lot about this paper and its results regarding the "Metricity of separable quantum optimal transport with the Hilbert-Schmidt cost." We’re going to take a quick breather before we move on to another interesting piece of research.

Conclusion: Kai: So, to recap, they proved that the square root of the separable quantum optimal transport cost defines a genuine metric between density matrices <ref:2609.03769#pg1>. That’s a significant result because it solves an open problem regarding the triangle inequality for certain quantum transport constructions <ref:2609.03769#pg1>.

Mira: Exactly, and they did this by using convex-roof duality to reformulate the squared transport cost as a supremum over Hermitian operators K and M <ref:2609.03769#pg2>. That approach bypassed the previous issues with quantum marginals that had stopped other researchers from proving these metric properties <ref:2609.03769#pg1>.

Lev: From a real hardware standpoint, this means we have a concrete distance measure d sep that we can actually use to compare states if we can implement the necessary pure-state projections and couplings <ref:2609.03769#pg1>. It’s a big step toward validating quantum transport models against empirical measurements, even if the actual implementation is still tough.

Kai: I agree with Lev; having a rigorous distance function is what experimentalists need to connect theory to the lab results <ref:2609.03769#pg1>. It’s not just abstract math; it’s a tool for characterizing state preparation quality, which is something we see constantly in quantum computing <ref:2609.03769#pg1>.

Mira: And the implication for condensed matter theory is that we can now use this distance to study how physical systems evolve under transport constraints, moving beyond purely static descriptions <ref:2609.03769#pg1>. It gives us a new lens to look at state evolution using separable couplings <ref:2609.03769#pg1>.

Lev: I think the dual formulation they developed is really promising for error correction research, because if we can solve that supremum problem efficiently, it could potentially lead to faster methods for characterizing the stability of quantum states under noise <ref:2609.03769#pg2>.

Kai: So this paper "Metricity of separable quantum optimal transport with the Hilbert-Schmidt cost" provides a very solid foundation for defining distance in this area <ref:2609.03769#pg1>. It’s a lot to take in, but it’s a necessary step forward.

Mira: Absolutely, it moves us past constructions that were mathematically obstructed by quantum marginal issues <ref:2609.03769#pg1>. We now have a rigorous metric structure defined by separable couplings <ref:2609.03769#pg1>.

Lev: I'm just excited to see how this metric can be applied in tandem with our work on fault-tolerant codes, because having a well-defined distance is crucial for measuring the fidelity of those error correction processes <ref:2609.03769#pg2>.

Copernicus Center for Interdisciplinary Studies, Jagiellonian University

quant-ph, math-ph, math.MP

Submitted: 2026-09-03

Updated: 2026-10-04

Comments: 12 pages, comments are welcome!

License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/

Importance score: 86/100

The gist: The paper establishes that a specific distance derived from separable quantum optimal transport costs defines a genuine metric between density matrices, thereby resolving an open problem regarding

Key concepts

Separable Quantum Optimal Transport Cost
This cost measures the minimum effort to move one quantum state ($ ho$) to another ($ au$) using a coupling that can be separated into independent parts. It is defined based on the Hilbert-Schmidt distance between pure states and involves an orthogonal projection onto an antisymmetric subspace.
Density Matrix Metric (dsep)
This is the specific distance derived from the separable quantum optimal transport cost. The authors show that this distance satisfies the triangle inequality, meaning it behaves like a true metric on density matrices, unlike previous constructions that failed this test.
Convex-Roof Duality
This mathematical technique is used to transform a difficult optimization problem (finding the minimum transport cost) into an equivalent problem involving finding an upper bound (a supremum). This allowed the authors to prove the necessary inequalities for the metric structure.
Antisymmetric Subspace Projection
This is a specific mathematical operation applied to vectors in complex space ($ ext{C}_n$). It projects vectors onto a subspace where the resulting vectors are antisymmetric. The cost function for transport is directly related to this projection, linking the geometry of the states to the transport distance.

Terminology

Summary

The paper establishes that a specific distance derived from separable quantum optimal transport costs defines a genuine metric between density matrices, thereby resolving an open problem regarding the triangle inequality for certain quantum transport constructions. This result is significant because it provides a rigorous metric structure for density matrices using separable couplings and pure-state geometry, which was previously obstructed by quantum marginal issues.

The gist

The square root of the separable quantum optimal transport cost associated with the orthogonal projection onto the antisymmetric subspace defines a genuine distance between density matrices.

Problem Addressed

Classical Monge–Kantorovich optimal transport faces challenges when applied to quantum states because a density matrix admits many inequivalent decompositions, making the standard gluing argument for proving the triangle inequality unavailable for coupling-based versions of quantum optimal transport. Specifically, constructions using the orthogonal projection onto the antisymmetric subspace as a cost operator were shown to violate the triangle inequality on diagonal states. A different approach based on minimizing an actual metric on pure states over separable bipartite couplings also faced this quantum marginal obstruction.

The Proposed Metric and Construction

The paper focuses on the order-two Beatty–França quantum optimal transport construction induced by the Hilbert–Schmidt distance between pure states. The distance is defined as:

  1. For unit vectors x, z in Cn, the cost function is defined as: c(x, z):= Tr (PA(Px ⊗ Pz)) = 1/2 (1 − ⟨x, z⟩2), where PA is the orthogonal projection onto the antisymmetric subspace.

  2. The separable quantum 2-Wasserstein distance is defined as: dsep(ρ, σ):= q T(ρ, σ), where T(ρ, σ) is the minimum transport cost over separable couplings Γsep(ρ, σ).

  3. This definition is equivalent to the construction based on pure-state distances: dsep(ρ, σ) = minXj pjd squared HS(Pxj, Pzj), where the minimum runs over all sequences of triples (pj, xj, zj)j with pj > 0 and xj, zj ∈ Cn are unit vectors such that P j pjPx j = ρ and P j pjPzj = σ.

Proof Strategy: Dual Formulation

The proof replaces the unavailable gluing argument by utilizing convex-roof duality and a dimension-independent interpolation result for Hermitian operators. The core of the proof involves showing that the squared transport cost, T(ρ, σ), can be expressed as a supremum over Hermitian operators K and M satisfying a feasibility constraint (5): ⟨x, Kx⟩ + ⟨z, Mz⟩ ≤ c(x, z) for all unit x, z ∈ Cn.

Key Steps in the Proof

  1. The squared transport cost is expressed as a convex roof: T(ρ, σ) = sup (Tr(Kρ) + Tr(Mσ)) over K, M satisfying constraint (5).

  2. The function g on the set of extreme points of the affine hull of C is extended to an extended-real-valued function ge.

  3. Carathéodory’s theorem is used to obtain the convex roof g∪, which equals d2sep = T by formula (1).

  4. The dual problem for the convex roof g∪ is solved using the Fenchel–Moreau theorem, leading to an expression for g∪ as a pointwise supremum of affine minorants: g∪(a) = sup l(a) l is affine on C, l ≤ g∪ on C.

  5. The minorant condition is shown to be equivalent to the constraint (5): ⟨x, Kx⟩ + ⟨z, Mz⟩ ≤ c(x, z).

  6. By substituting the dual representation into the definition of T(ρ, σ), one obtains T(ρ, σ) ≤ (dsep(ρ, τ) + dsep(τ, σ))2, which implies the triangle inequality for dsep by taking square roots.

Equivalent Formulations and Implications

The metricity result is equivalent to showing that the separable SWAP fidelity FS,sep(ρ, σ) = max ω∈Γsep(ρ,σ) Tr(Sω) satisfies p/1 - FS,sep is a metric on Dn. Furthermore, this result confirms that the folded optimal transport distance D2 associated with √1/2 dHS is precisely the Beatty–França quantity WdHS/√22 = dsep. This implies that passing to the chain envelope does not decrease the one-step separable transport cost, as D2 = WdHS/√22 = dsep.

Conclusion

The paper proves that for every n, the function dsep defined in (2) satisfies the triangle inequality and is indeed a genuine distance on Dn.

Improvements for AI systems

Here are the specific improvements to AI systems that can be derived from this scientific paper, focusing on areas where quantum optimal transport (QOT) is relevant:

  1. A new metric for comparing quantum states based on separable couplings: The paper introduces the distance function, denoted as the square root of the separable quantum optimal transport cost, defined by:

dsep(ρ, σ) = r1/2 minX j pjd 2 HS(Px j, Pz j), (1)

where the minimum runs over all sequences of triples (pj, xj, zj)j with pj > 0 and xj, zj ∈ C n are unit vectors such that Pj pjPxj = ρ and Pj pjPzj = σ.

  1. The ability to verify the triangle inequality for this distance: The paper rigorously proves that this distance, dsep, is a genuine metric on the space of density matrices (Dn), establishing the triangle inequality:

dsep(ρ, σ) ≤ dsep(ρ, τ) + dsep(τ, σ).

  1. A method for comparing quantum states based on separable SWAP fidelity: The paper proves that the square root of the cost operator associated with the orthogonal projection onto the antisymmetric subspace defines a genuine distance between density matrices. This is equivalent to showing that its corresponding separable SWAP fidelity metric satisfies a triangle inequality:

p 1 − FS,sep is a metric on Dn.

  1. A framework for gluing quantum transport problems without relying on classical marginal gluing: The proof introduces an interpolating operator (Lemma 2.3) derived from convex-roof duality and the non-strict Finsler lemma. This provides a mathematical mechanism to establish metric properties (like the triangle inequality) for quantum transport costs even when the intermediate state's pure-state decomposition is not uniquely defined, overcoming a known quantum marginal obstruction.

  2. Derivation of dual formulations for optimal transport problems: The paper establishes that the primal optimization problem (minimizing Tr(PAomega)) can be reformulated as a supremum over Hermitian operators K and M (Proposition 3.1). This duality provides alternative computational paths for solving quantum optimal transport problems, allowing researchers to use the simpler dual formulation if it is more tractable.

  3. Application in Quantum Machine Learning (QML) and Quantum Information Processing:

The improved AI system can perform tasks requiring the comparison or ordering of quantum states where the underlying structure involves separable decompositions (e.g., when training variational quantum circuits, comparing entanglement measures, or analyzing state preparation fidelity).

  1. Verification of State Preparation Quality: The metric dsep provides a rigorous way to measure how close two target density matrices are in terms of the minimal effort required to generate them from common pure-state resources (separable couplings), which is valuable for benchmarking quantum state preparation algorithms.

Sources

Related papers