Approach to optimal quantum transport via states over time

This paper proposes a novel framework for quantum optimal transport by defining transport costs as linear functions of "states over time" (the Jordan product of a density matrix and a transport map), revealing that this approach yields qualitatively different results from classical Monge transport theory, particularly in the analytically tractable case of unitary-invariant costs.

The Gist

Imagine you are a logistics manager in a bustling city. Your job is to move a pile of sand (representing mass or probability) from one location to another. In the classical world, you have a map, and you want to find the cheapest way to move every grain of sand to its destination. This is the famous "Optimal Transport" problem, pioneered by mathematician Gaspard Monge. You calculate the cost based on how far each grain travels.

Now, imagine you are in the quantum world. Here, "sand" isn't just a pile of grains; it's a fuzzy, shifting cloud of possibilities (a quantum state). And the "truck" moving the sand isn't just a vehicle; it's a complex rule that changes the very nature of the sand as it moves (a quantum channel).

This paper, by Hoogsteder-Riera, Calsamiglia, and Winter, asks a big question: How do we calculate the "transport cost" in this fuzzy quantum world?

Here is the breakdown of their approach, using simple analogies:

1. The New "Coupling": The "Stote"

In the classical world, to move sand, you create a "coupling." Think of this as a master spreadsheet that lists: "If a grain is at point A, what is the chance it ends up at point B?" It connects the starting pile to the ending pile.

In the quantum world, the authors realized you can't just use a spreadsheet. You need a new object that combines the starting cloud (the initial state) and the moving rule (the channel) into a single package. They call this package a "Stote" (a cute pun on "state over time," though they jokingly note it sounds like a stoat, a type of weasel).

• The Analogy: Imagine you have a recipe (the channel) and a bag of ingredients (the initial state). In classical transport, you just list the ingredients and the destination. In this quantum version, the "Stote" is like a magic smoothie where the ingredients and the recipe are blended together. You can't separate them easily; the cost of the transport depends on how they are mixed.

2. The "Jordan Product": The Mixing Method

How do you blend the ingredients and the recipe? The authors use a specific mathematical operation called the Jordan product.

• The Analogy: Think of mixing paint. If you mix Red and Blue, you get Purple. But in the quantum world, the order and the way you mix matter. The Jordan product is a specific, symmetric way of blending the "starting state" and the "transport rule" so that the result captures the history of the journey.

3. The Cost: How Expensive Was the Trip?

Once you have your "Stote" (the blended package), you assign a cost to it.

• The Goal: Find the transport rule (channel) that moves your quantum state from Point A to Point B with the lowest possible cost.
• The Twist: In classical transport, the cost is usually just distance. In this quantum version, the cost is a linear function of the "Stote."

4. What They Found (The Surprises)

The authors tested this new system, particularly looking at a "fair" cost where the rules don't change if you rotate your coordinate system (Unitary Invariance). They found some results that are very different from the classical world:

• The "Square Root" Problem: In classical transport, if you move things twice as far, the cost doubles. In their quantum model, the cost behaves more like the square of a distance.

  • Analogy: If you walk 1 mile, the cost is 1. If you walk 2 miles, the cost isn't 2; it's 4. This suggests that to get a "true" distance in the quantum world, you might need to take the square root of their calculated cost, something that isn't necessary in the classical world.

• The "One-Way Street" (Asymmetry): In classical transport, the cost to go from A to B is usually the same as B to A. In their quantum model, this is not always true.

  • Analogy: Imagine a river. It might be easy to float a boat downstream (A to B), but very hard to row upstream (B to A). The authors found that even with a "fair" cost rule, the quantum transport cost can be different depending on which way you are moving.

• The "Ghostly" Influence (Discontinuity): This is perhaps the strangest finding. In the classical world, if you change your starting pile of sand just a tiny bit, the cost changes just a tiny bit. In their quantum model, if you have a "pure" state (a very specific, sharp quantum cloud) and you change it even slightly to be "mixed" (fuzzy), the cost can jump suddenly.

  • Analogy: Imagine a bridge that is perfectly stable for a single person. But if you add a tiny, almost invisible pebble to their backpack, the bridge suddenly collapses. The cost function is "jumpy" and discontinuous in the quantum realm.

• The "Far-Field" Effect: In classical transport, if you move a pile of sand, the cost only depends on where the sand is. If there is empty space nearby, it doesn't matter. In their quantum model, the cost does depend on the empty space around the sand.

  • Analogy: It's like the Aharonov-Bohm effect in physics. A charged particle can be affected by a magnetic field even if the particle never touches the field. Similarly, the "cost" of moving a quantum state depends on the "shape" of the empty universe around it, not just the state itself.

5. The Big Picture

The authors conclude that while they have built a beautiful mathematical machine (the "Stote" formalism) to calculate these costs, the results are qualitatively different from classical transport.

• The Open Question: They admit they don't yet have a simple, complete rulebook (a "dual cone") that tells them exactly which cost functions will behave nicely (like obeying the triangle inequality).
• The Takeaway: Quantum transport isn't just "classical transport with quantum math." It has its own unique, sometimes weird, rules where direction matters, small changes can cause big jumps, and the empty space around you matters.

In short, they have built a new way to measure the "effort" of moving quantum information, and it turns out the quantum universe is much more sensitive and asymmetric than the classical one we are used to.

Technical Summary

Problem Statement

The paper addresses the challenge of constructing a quantum analogue of the classical Monge optimal transport theory. In the classical setting, the optimal transport cost between two probability distributions $\mu$ and $\nu$ is defined as the infimum of a cost functional over all couplings (joint distributions) with the given marginals. The cost functional is bilinear in the mass distribution and the transport plan (stochastic kernel).

The authors aim to generalize this framework to quantum systems. The primary difficulty lies in defining a "quantum coupling" that preserves the bilinear structure of the classical cost (linear in the initial state and linear in the transport channel) while respecting quantum mechanical constraints. Previous approaches have utilized primal formulations (couplings), dual formulations, or continuous dynamical semigroups, but often lack a direct physical interpretation of the coupling as a bilinear combination of a state and a channel.

Methodology

The authors propose a formulation based on the concept of "states over time" (stotes). Their methodology proceeds as follows:

1. Definition of the Coupling (Stote):
   Instead of treating the coupling as a joint state in a spatial product space, they interpret it as an operator encoding the correlation between an initial state and its evolution through a quantum channel. They define the state over time $Q$ for an initial state $\rho$ and a quantum channel (represented by its Jamiołkowski matrix $J$) as the Jordan product:
   $$Q = \rho \star J = \frac{1}{2}(\rho \otimes \mathbb{1} \cdot J + J \cdot \rho \otimes \mathbb{1})$$
   This choice is motivated by the work of Fullwood and Parzygnat, who showed that the Jordan product is the unique function satisfying desirable axioms for states over time, including bilinearity, Hermiticity, marginal preservation, and composability.

2. Quantum Transport Cost:
   The cost of transporting a state $\rho$ through a channel $E$ (with Jamiołkowski matrix $J_E$) is defined as the expectation value of a Hermitian cost matrix $K$ with respect to the stote:
   $$\kappa(\rho, E) = \text{Tr}[K Q_{\rho, E}] = \text{Tr}[K (\rho \star J_E)]$$
   The quantum optimal transport cost $K(\rho, \sigma)$ between two states is the minimum of this cost over all admissible channels $E$ such that $E(\rho) = \sigma$.

3. Optimization Framework:
   The problem of finding the optimal cost is formulated as a Semidefinite Program (SDP). The authors provide both primal and dual SDP formulations, utilizing the Choi and Jamiołkowski isomorphisms to handle the positivity constraints of the channels.

4. Unitary Invariance:
   A significant portion of the analysis focuses on unitary invariant (UI) costs, where the cost matrix $K$ commutes with $U \otimes U$ for all unitaries $U$. They prove that, up to scaling, there is a unique cost matrix satisfying the conditions of assigning zero cost to the identity channel and being in the dual cone of stotes:
   $$K_0 = d\mathbb{1} - S$$
   where $S$ is the swap operator and $d$ is the dimension.

Key Contributions and Results

1. Mathematical Characterization of Stotes:
The authors rigorously define the set of states over time $\mathcal{Q}(\rho, \sigma)$ and its associated cone. They establish that the Jordan product allows for the reconstruction of the channel from the stote (via Theorem 2.3), generalizing Bayes' theorem to the quantum domain. They characterize the dual cone of stotes, which is necessary for determining valid cost matrices that yield non-negative costs.

2. Properties of the Transport Cost:
The paper investigates whether the proposed cost satisfies metric properties:

• Identity Cost: The cost is zero for the identity channel if and only if $\text{Tr}_B[S \star K] = 0$.
• Positivity: The cost is non-negative if $K$ lies in the dual cone of the stote cone.
• Symmetry: The authors demonstrate that even if the cost matrix $K$ is symmetric (e.g., $SKS=K$), the resulting optimal transport cost $K(\rho, \sigma)$ is not necessarily symmetric. They provide examples (e.g., depolarizing and dephasing channels) where the set of stotes is asymmetric, leading to $K(\rho, \sigma) \neq K(\sigma, \rho)$.
• Triangle Inequality: A condition is derived for the cost to satisfy the triangle inequality, involving the dual cone of stotes over three time steps.
• Convexity: The cost is shown to be convex in the second argument (target state) but not necessarily in the first.

3. Analytical Results for Unitary Invariant Cost:

• Pure States: For pure states $\rho = |0\rangle\langle 0|$ and $\sigma = |\psi\rangle\langle\psi|$ with overlap $\alpha = |\langle 0|\psi\rangle|$, the optimal cost is derived analytically as $K(\rho, \sigma) = (1-\alpha)(d + 2\alpha)$. The optimal channel is a unitary conjugation.
• Commuting States: For commuting states (diagonal in the same basis), the optimal cost can be calculated analytically. The authors show that the optimal quantum transport cost is strictly lower than the cost obtained by restricting the transport to classical stochastic maps, highlighting a quantum advantage.
• High-Dimensional Limit ($d \to \infty$): By embedding finite-dimensional states into a larger Hilbert space and taking the limit, the authors derive a renormalized cost formula. In this limit, the cost depends only on the supports of the states and relates to the entanglement fidelity. Specifically, $K_\infty(\rho, \sigma) = 1 - \max_E F(\rho, E)$, where $F$ is the channel fidelity.

4. Asymmetry and Discontinuity:
The paper highlights a striking feature: the optimal cost function can be discontinuous. When a pure state is approached by a sequence of mixed states, the set of feasible channels changes abruptly (from a single replacement channel to a set including unitaries), causing a jump in the optimal cost. This leads to a non-zero "symmetry gap" ($K(\rho, \sigma) \neq K(\sigma, \rho)$) even for symmetric cost matrices.

Significance and Claims

The authors claim that their approach offers a qualitatively different perspective on quantum optimal transport compared to classical Monge transport and other quantum generalizations.

• Physical Interpretation: The use of the Jordan product provides a coupling that is bilinear in the state and the channel, mirroring the classical structure and offering a clear physical interpretation of the transport plan.
• Quantum Specificity: The results suggest that quantum optimal transport costs possess unique features not found in the classical case. Notably:
  • The cost behaves like the square of a distance for pure states (violating the triangle inequality directly), suggesting that a square root might be necessary to recover a metric, a feature not seen in classical Wasserstein distances.
  • The cost is sensitive to the behavior of the channel on the orthogonal complement of the state's support (an effect reminiscent of the Aharonov-Bohm effect), whereas classical transport is insensitive to regions with zero probability mass.
  • The cost function is inherently asymmetric and discontinuous in certain regimes, challenging the notion that this formalism naturally yields a metric space on quantum states.

The paper concludes that while the formalism is mathematically robust and allows for analytical solutions in specific cases (unitary invariant, commuting states), the physical meaning of the optimal cost and the existence of a concise characterization of the dual cone of stotes remain open problems. The authors do not claim to have solved the problem of defining a perfect quantum metric but rather present a new, physically motivated framework that reveals distinct quantum behaviors in transport phenomena.