Distances between pure quantum states induced by a distance matrix

arXiv:2509.14727 · math-ph, math.MP, quant-ph · Submitted 2025-09-18 · 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: Today's paper: "Distances between pure quantum states induced by a distance matrix".

Mira: With the help of a given distance matrix, an infinite family of distances can be constructed on the complex projective space modeling pure quantum states,

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

Title and authors: Kai: So we’re looking at the paper titled "Distances between pure quantum states induced by a distance matrix." It seems like they're taking an arbitrary distance matrix and using it to create a new way to measure how far apart pure quantum states are in complex projective space.

Mira: That sounds intriguing, Kai, because it’s moving away from standard measures like the Hilbert–Schmidt distance that we see everywhere. It suggests a much more flexible geometry for the space of pure states, which is what we need when modeling things that aren't just simple orthogonal projections.

Lev: From an error correction standpoint, I wonder how robust this construction would be when applied to actual physical systems where we have noise and errors affecting the state representation.

Kai: Exactly Lev, that’s a big question about practicality. The paper introduces a map d p(x, y) using a distance matrix E ij of size n, and it claims this method can isometrically embed any finite metric space into this quantum state space.

Mira: That embedding capability is what I find most compelling; it means we can map any real-world distance structure onto the pure states, which opens up new avenues for relating classical metrics to quantum geometry.

Lev: If we take those finite metric spaces and try to run error correction algorithms on them, the complexity of verifying the triangle inequality for d p will be a major hurdle for hardware implementation.

Kai: Well, the paper tackles that head-on by showing they can prove that these maps d p are bona fide distance functions by using methods from analysis and convex geometry to get some auxiliary convexity result.

Mira: That reliance on multilinear algebra to establish the triangle inequality is interesting; it suggests the geometric structure is quite deep, not just a superficial tweak on existing metrics.

The paper's summary: Kai: Moving onto what they actually summarize in this work, they are proposing the construction of d p(x, y) defined by formula (four), which involves an operator E acting on wedge products of basis vectors and a parameter p where p two <ref:2509.14727#pg2>.

Mira: The core idea is that this map generalizes the Hilbert–Schmidt distance, which they note is a special case when you use the discrete metric for your matrix E ij, showing that it covers known concepts under different conditions.

Lev: But generalizing means we have to be careful; if we're talking about actual quantum hardware, ensuring that this new distance behaves predictably under the noisy operations we apply is where the real engineering challenge lies.

Kai: The paper also focuses on how this construction extends from pure states to mixed states by proposing a formula p(rho, sigma) for density matrices in n.

Mira: That extension requires a specific form for the distance matrix E ij, specifically that it must be of the form E ij = v i - v j two/p for some set of affinely independent vectors, which is a significant restriction <ref:2509.14727#pg0>.

Lev: That restriction on the distance matrix seems like a major limitation if we want to apply this to general noisy channels, because most physical noise structures won't fit that specific quadratic form.

Kai: The paper then shows that this extension establishes an isometric embedding of the pure state space (P(C n), d p) into the space of density matrices equipped with a specific Hilbertian norm, H 0n <ref:2509.14727#pg2>.

Mira: It’s interesting how they tie the metric structure to a specific norm on the space of density matrices; that norm, defined by N = sum i<j E ij (ij squared - ii jj)/p, is what makes the mixed state distance work <ref:2509.14727#pg0>.

Lev: So, if we look at the paper "Distances between pure quantum states induced by a distance matrix," this construction seems very powerful theoretically, but its applicability to realistic noise models might be highly constrained by those necessary forms for E ij.

The paper's improvements: Kai: Regarding the improvements they suggest, the authors emphasize that they need to verify that the map d p satisfies the triangle inequality for any p two and any distance matrix, which is what their main result establishes <ref:2509.14727#pg0>.

Mira: They do this by relying on a nontrivial auxiliary convexity result derived from methods of analysis and multilinear algebra, which is a bit abstract but mathematically sound.

Lev: That reliance on these high-level analytical tools means that trying to verify this in a real quantum computation environment will be incredibly difficult because we can't easily check those abstract convexity conditions.

Kai: They do address the case for n = three specifically, showing the triangle inequality holds if and only if the spectral radius rho(E) satisfies two rho(E) Tr E <ref:2509.14727#pg0>.

Mira: That specific condition for n=three gives us a concrete constraint on how the underlying distance matrix must behave for this pure state metric to function correctly <ref:2509.14727#pg0>.

Lev: For hardware, having such a precise spectral radius condition means we’d need to know the exact eigenvalues of our error structure beforehand, which is rarely possible in practice.

Kai: For any general n three they reduce the triangle inequality verification to checking a condition for orthonormal systems where the norm squared term must be less than or equal to the sum of two other terms <ref:2509.14727#pg0>.

Mira: So, it boils down to verifying that specific inequality for any orthonormal set, which is a more tangible check than solving some general matrix problem.

Conclusion: Kai: To wrap up on this paper, the construction presented in "Distances between pure quantum states induced by a distance matrix" gives us a systematic way to quantify any finite metric space using the geometry of pure quantum states.

Mira: They’ve shown how to extend this idea from pure states to mixed states, but they highlighted that for that extension to hold, the input distance matrix E ij has to follow a very specific quadratic form.

Lev: So, while it provides a beautiful theoretical framework for embedding metrics into quantum geometry, I think the real challenge remains in bridging the gap between this theory and running this on actual hardware where we deal with noisy systems.

Kai: It certainly points toward future work where researchers can use these distances as a guide for designing more physically motivated quantum error correction codes or state preparation routines.

Mira: I agree, because constraining the distance matrix form helps narrow the search space for applicable physical models of quantum channels.

Lev: And my final thought is that until we see a way to verify those abstract convexity proofs on real hardware, this remains mostly a theoretical tool for structuring our thinking about quantum data.

Tomasz Miller, Rafał Bistroń

Copernicus Center for Interdisciplinary Studies, Jagiellonian University

math-ph, math.MP, quant-ph

Submitted: 2025-09-18

Updated: 2026-10-04

Comments: 27 pages, 1 figure

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

Importance score: 79/100

The gist: With the help of a given distance matrix, an infinite family of distances can be constructed on the complex projective space modeling pure quantum states, providing a natural way to isometrically

Key concepts

Distance Matrix (Eij)
This is an arbitrary $n imes n$ matrix that defines the metric structure of a finite set. It dictates how distances between elements in that set should be measured. The construction uses this matrix to define a new, generalized distance function on quantum states.
dp(x, y)
This is the constructed distance function between two pure quantum states $x$ and $y$. It is defined using the matrix Eij and the wedge product of the states, essentially mapping classical distances from Eij into a geometric structure on complex projective space.
Pure Quantum States P(Cn)
These are vectors in complex projective space representing pure quantum states. The paper shows that by defining dp(x, y) this way, any finite metric space can be perfectly embedded (isometrically mapped) into this continuous quantum state space.

Terminology

Summary

With the help of a given distance matrix, an infinite family of distances can be constructed on the complex projective space modeling pure quantum states, providing a natural way to isometrically embed any finite metric space into this quantum state space.

Construction of the Distance Function

The core idea involves defining a map on the pure state space, denoted as dp(x, y), based on an arbitrary distance matrix (Eij) of size n. This map is explicitly given by:

dp(x, y):= X i<j E p ij xiyj − xjyi 2/p = E p/2 (x ∧ y) 2/p, where x and y are pure states in P(C n). This construction generalizes the Hilbert–Schmidt distance (dHS), which is a special case when Eij is the discrete metric. The operator E on the space of wedge products, Λ2C n, is defined such that E(ei ∧ ej) = Eij ei ∧ ej.

Properties and Proof of Metric Axioms

The primary result establishes that this map dp is a bona fide distance function on P(C n) for any p ≥ 2 and any distance matrix (Eij). This involves demonstrating that it satisfies the triangle inequality: dp(x, y) ≤ dp(x, z) + dp(y, z). The proof relies on several key analytical tools:

  1. The use of methods of analysis, multilinear algebra, and convex geometry to obtain a nontrivial auxiliary convexity result.

  2. For the case n = 3, the triangle inequality is proven by showing that it holds if and only if the spectral radius ρ(E) satisfies 2ρ(E) ≤ Tr E.

  3. For general n ≥ 3, the proof reduces to verifying a condition for orthonormal systems: Triangle inequality (5) holds if and only if E p/2 (u ∧ v) 2/p ≤ E p/2 (u ∧ w) 2/p + E p/2 (v ∧ w) 2/p for any orthonormal system.

Symmetries and Invariance

The paper investigates the projective unitary transformations that leave the distance function dp invariant. The equivalence between invariance under a unitary U and the commutation relation [E, U ∧ U] = 0 is established via Lemma 8. This means that dp(Ux, Uy) = dp(x, y) if and only if E commutes with U ∧ U. The paper notes that the subgroup of unitaries leaving dp invariant depends on the specific distance matrix (Eij).

Extension to Mixed States

The construction is extended from pure states to mixed states (density matrices in Ωn) by proposing a formula for a bona fide metric D(ρ, σ) on Ωn. This is achieved via Theorem 16, which shows that the map x → xx† is an isometric embedding of (P(C n), dp) into the space of density matrices equipped with a specific Hilbertian norm.H0n. The resulting distance function for mixed states is given by:

ˆdp(ρ, σ):= Φ(ρ − σ)2/p F⊕E = X i<j E p ij ρij − σij 2 - (ρii − σii)(ρjj − σjj) 1/p. This extension requires the distance matrix to be of the form Eij = vi − vj 2/p for some affinely independent vectors, and the norm.H0n is defined by N(∆) = sX i<j E p ij (∆ij 2 - ∆ii∆jj).

Generalization and Applications

The constructed distance dp provides a canonical way to quantize any finite metric space (Remark 6), allowing for the isometric embedding of an n-element metric space X into P(C n). Furthermore, it is shown that the distances are uniformly equivalent to the Frobenius distance raised to the power 2/p, specifically: √2n 2/p cT2/p(ρ, σ) ≤ ˆdp(ρ, σ) ≤ C2/p T2/p(ρ, σ). This framework is useful for defining a quantum p-Hamming distance between arbitrary pure states of an m-qubit system using Hamming distances as the distance matrix. The extension to mixed states allows for the study of quantum channels and their monotonicity properties.

The Gist

dp(x, y):= X i<j E p ij xiyj − xjyi 2/p = E p/2 (x ∧ y) 2/p, where x and y are pure states in P(C n).

Improvements for AI systems

Based on the provided scientific paper, here are specific improvements that could be implemented in AI systems, categorized by the area of application:

)1. Enhanced State Representation and Geometric Distance Learning (From Theorem 1 & Remark 6)

The paper establishes a method to isometrically embed any finite metric space (like a set of training data points or configuration states) into the space of pure quantum states on an n-level system, using the distance function:

dp(x, y):=∥E p/2(x ∧ y)∥ 2/p

The resulting map is a genuine metric for pure states derived from an arbitrary distance matrix (Eij).

Improving AI systems could involve:

  • Instead of relying solely on standard Hilbert-Schmidt or Fubini-Study distances, AI models (especially those dealing with quantum state estimation or classification) could utilize the generalized distance function, which is tailored to the specific geometric structure defined by the problem's inherent metric (Eij).

  • AI systems designed for metric learning could be trained to learn a distance matrix Eij that accurately reflects the underlying physical or semantic relationships between states, allowing them to perform more robust classification and clustering on quantum data.

)2. Quantum Optimal Transport (QOT) Modeling for State Evolution and Generation (From Section 2 & Theorem 16)

The paper connects the constructed distance to the quantum optimal transport problem, suggesting its use in modeling state transitions:

  • AI systems could be developed to solve complex quantum state evolution problems by formulating them as a minimization of the quantum p-Wasserstein semidistance (Equation 2). This would allow for finding the most physically plausible or shortest path evolution between two quantum states, which is highly relevant for quantum machine learning and simulation.

  • The extension to mixed states (Theorem 16) allows these transport models to be applied directly to density matrices representing noisy or partially mixed systems, enabling the prediction of state distributions under non-ideal conditions.

)3. Quantum Feature Engineering via Quantum p-Hamming Distance (From Remark 7)

The paper introduces a specific distance for m-qubit systems:

dp(x, y) = X i<j E p/2 ij x iyj − xjy ij2/p

This is interpreted as a quantum p-Hamming distance between pure states.

Improving AI systems could involve:

  • In quantum machine learning tasks (e.g., distinguishing entangled states), feature engineering could be augmented by calculating this specific distance metric, which is more sensitive to the relative phase and overlap structure than simple overlap measures. This would improve the system's ability to discriminate between highly entangled quantum states.

)4. Metric Space Quantization for Data Compression and Indexing (From Remark 6)

Remark 6 suggests that the construction provides a way to isometrically embed any n-element metric space into the space of pure quantum states:

dp(ι(ai), ι(aj)) = E ij = dist(ai, aj)

This means arbitrary discrete metrics can be mapped to quantum geometry.

Improving AI systems could involve:

  • AI systems dealing with high-dimensional discrete data (e.g., graph structures or complex network states) could use this mapping to represent them as pure quantum states in a high-dimensional Hilbert space, potentially enabling more compact representations or better topological analysis of the data structure using quantum geometric tools.

)5. Robustness and Parameter Sensitivity Analysis (From Remark 4 & Proposition 9)

The paper explicitly proves that the distance function is only a true metric for orders where p ≥ 2:

Remark 4: The assumption p ≥ 2 in (1) is necessary. More concretely, for p < 2 the map dp no longer satisfies the triangle inequality.

Improving AI systems could involve:

  • AI researchers developing algorithms that are specifically designed to operate within the regime where this metric is guaranteed to be a true distance function (p ≥ 2). This ensures mathematical rigor and reliable convergence guarantees when using this distance for optimization or learning tasks.

  • For applications requiring sensitivity analysis, understanding the critical threshold at p=2 (where it transitions from non-metric to metric) provides a clear boundary for system stability.

)6. Metric Space Learning via Auxiliary Convexity Results (From Theorem 13 & Lemma 14)

The paper provides powerful auxiliary convexity results (Theorem 13 and Lemma 14) that establish the triangle inequality for the distance function based on properties of symmetric matrices:

Theorem 13: The triangle inequality holds if and only if a specific matrix inequality involving a jointly convex, totally symmetric map f holds.

Improving AI systems could involve:

  • AI systems tasked with learning complex, non-linear decision boundaries or energy landscapes could be guided by these geometric constraints. Instead of searching through the entire space of functions, the system can constrain its search to those that satisfy the structural properties implied by Theorem 13, leading to more efficient and geometrically sound learning algorithms.

In summary, this paper provides a mathematically rigorous framework for creating quantum-geometrically motivated distance metrics based on arbitrary input metrics (Eij). The improvements focus on using these metrics not just as descriptive tools, but as fundamental structural constraints for:

  1. Learning robust state representations (Metric Learning).

  2. Modeling physically constrained quantum dynamics (State Evolution/QOT).

  3. Developing specialized feature extraction methods for quantum data (Feature Engineering).

Sources

Related papers