Partial majorization and Schur concave functions on the sets of quantum and classical states

This paper establishes tight upper bounds on the difference of Schur concave functions between quantum states related by $m$-partial majorization, with applications to von Neumann entropy, the introduction of $\varepsilon$-sufficient majorization rank, and extensions to classical probability distributions.

The Gist

The Big Picture: Measuring the "Messiness" of Quantum States

Imagine you are a chef trying to bake the perfect cake. In the world of quantum physics, a "state" (like a particle or a system) is like a cake recipe. Some recipes are very specific and ordered (low entropy), while others are chaotic and mixed up (high entropy).

The paper focuses on a specific mathematical tool called Schur concavity. Think of this as a "disorder meter."

• If you have a very ordered cake (State A) and you mix it up a bit to make a messier cake (State B), the disorder meter goes up.
• The paper asks: If we know State A is "more ordered" than State B in a specific way, how much more disorderly can State B actually be?

The Problem: Partial Knowledge

Usually, to say State A is "more ordered" than State B, you have to check the entire recipe from start to finish. This is called majorization. It's like checking every single ingredient and step in a cookbook.

However, in the real world (and in quantum physics), we often can't check the whole infinite recipe. We might only know the first few ingredients.

• The Paper's Concept: Partial Majorization.
• The Analogy: Imagine you are comparing two novels. You only read the first 10 chapters of both. If the first novel seems more structured than the second in those first 10 chapters, we say it "partially majorizes" the second.
• The Question: If Novel A looks more structured than Novel B in the first 10 chapters, how different can the ending of the stories be? Could Novel B turn out to be a total disaster compared to Novel A?

The Solution: The "Safety Net"

The author, M.E. Shirokov, builds a mathematical "safety net" (an upper bound).

1. The Setup: You have a state $\rho$ (the "good" ordered cake) and a state $\sigma$ (the "messy" cake).
2. The Constraints:
   • You know $\rho$ is "partially better" than $\sigma$ for the first $m$ ingredients (partial majorization).
   • You also know $\rho$ and $\sigma$ are very similar overall (they are close in "trace norm," meaning the total difference in ingredients is small, say $\epsilon$).
3. The Result: The paper calculates the maximum possible difference in "disorder" (entropy) between the two cakes.

The Metaphor:
Imagine you are grading two students.

• Student A has a perfect record for the first 10 tests ($m=10$).
• Student B has a slightly worse record for those 10 tests.
• You also know that Student B's overall average is very close to Student A's ($\epsilon$ is small).
• The Paper's Answer: "Based on these facts, Student B's final grade cannot be lower than X. Here is the exact formula for X."

The "Magic" State ($\rho_{m,\epsilon}$)

The most clever part of the paper is the construction of a specific "worst-case scenario" state, which the author calls $\rho_{m,\epsilon}$.

• How it works: The author creates a hypothetical "monster" state that is just barely different from the original state $\rho$ but is arranged in the most chaotic way possible, while still obeying the rules of partial majorization and closeness.
• Why it matters: By calculating the disorder of this "monster" state, the author finds the absolute limit. No matter what the actual messy state $\sigma$ is, its disorder cannot exceed the disorder of this monster.
• The Result: The difference in disorder is bounded by the difference between the original state and this monster state.

Why Does This Matter? (The "So What?")

The paper applies this to Von Neumann Entropy, which is the quantum version of "information content" or "uncertainty."

1. Quantum Oscillators: The author tests this on a "quantum oscillator" (like a vibrating atom). They calculate how many "steps" ($m$) you need to check to be sure that the system's entropy is stable.
2. The "Sufficient Rank": They introduce a new concept called the $\epsilon$-sufficient majorization rank.
   • Analogy: Imagine you are trying to describe a complex painting. You don't need to look at every single pixel. You only need to look at the top 50% of the most important brushstrokes to get 99% of the picture right.
   • The paper tells you exactly how many brushstrokes (how large $m$ needs to be) you need to look at to guarantee that your description of the painting's "messiness" is accurate within a certain error margin ($\epsilon$).

The Takeaway for Everyone

This paper is a guide on how much you can trust a partial description.

• In everyday life: If you know the first few chapters of a mystery novel are very structured, and the book is generally similar to another one, you can mathematically prove that the ending of the second book can't be too chaotic compared to the first.
• In quantum physics: It gives scientists a precise tool to say, "We only measured the first $m$ energy levels of this particle, but because of this math, we know the total uncertainty of the system is within this specific range."

It turns a vague feeling of "they look similar" into a hard, tight mathematical limit on how different they can really be.

Technical Summary

1. Problem Statement

The paper addresses the problem of quantifying the violation of the inequality $f(\rho) \leq f(\sigma)$ for Schur concave functions $f$ defined on the set of quantum states $\mathcal{S}(\mathcal{H})$.

• Context: In standard majorization theory, if a state $\rho$ majorizes a state $\sigma$ (denoted $\rho \succ \sigma$), then for any Schur concave function $f$, $f(\rho) \leq f(\sigma)$. This is fundamental in quantum information theory (e.g., for von Neumann entropy).
• The Gap: For infinite-dimensional systems or when only partial information is available, one often encounters $m$-partial majorization ($\rho \overset{m}{\succ} \sigma$), where the majorization inequalities hold only for the first $m$ eigenvalues. In this case, the standard inequality $f(\rho) \leq f(\sigma)$ does not necessarily hold.
• Objective: The author seeks to derive tight upper bounds on the difference $f(\rho) - f(\sigma)$ under the conditions:
  1. $\rho \overset{m}{\succ} \sigma$ (partial majorization).
  2. $\frac{1}{2}\|\rho - \sigma\|_1 \leq \varepsilon$ (states are close in trace norm).
  3. $f(\rho)$ is finite.

The ultimate goal is to determine conditions under which this difference vanishes as $m \to \infty$ and/or $\varepsilon \to 0$.

2. Methodology

The author develops a universal technique based on spectral analysis and the construction of specific "worst-case" states.

A. Definitions and Tools

• Partial Majorization: $\rho \overset{m}{\succ} \sigma$ if $\sum_{i=1}^k p_i \geq \sum_{i=1}^k q_i$ for all $k=1, \dots, m$, where $p_i, q_i$ are eigenvalues in non-increasing order.
• Key Inequality: The Mirsky inequality is utilized: $\sum |p_i - q_i| \leq \|\rho - \sigma\|_1$.
• Auxiliary Sets: The author defines sets $T_m(\rho)$ containing states $\sigma$ that share the first $m$ eigenvalues with $\rho$ and satisfy specific ordering constraints.
• Proposition 1 (Reduction): The problem of finding the supremum of $f(\rho) - f(\sigma)$ over the intersection of the $\varepsilon$-vicinity $U_\varepsilon(\rho)$ and the partial majorization constraint is reduced to finding the supremum over a specific subset $T_m(\rho) \cap U_\varepsilon(\rho)$. This is achieved by constructing a state $\sigma^*$ that majorizes any arbitrary $\sigma$ while preserving the distance constraints.

B. Construction of the Extremal State $\rho_{m,\varepsilon}$

The core of the methodology is the explicit construction of a state $\rho_{m,\varepsilon}$ that acts as a "majorant" for all states $\sigma$ satisfying the constraints.

• Definition: $\rho_{m,\varepsilon}$ is constructed by modifying the spectrum of $\rho$. It retains the first $m$ eigenvalues of $\rho$. The remaining probability mass is redistributed to minimize the value of the Schur concave function $f$ (since $f$ is Schur concave, minimizing $f(\sigma)$ maximizes the difference $f(\rho) - f(\sigma)$).
• Cases: The construction depends on the relationship between $\varepsilon$ and the tail sum of eigenvalues $d_{m+1} = 1 - \sum_{i=1}^{m+1} p_i$.
  • If $\varepsilon$ is large, the mass is concentrated on the $(m+1)$-th eigenvalue.
  • If $\varepsilon$ is small, the mass is distributed to satisfy the trace norm constraint while keeping the majorization order.

C. Reformulation for Classical Systems

The paper demonstrates that these results apply to probability distributions (classical states) by mapping distributions to diagonal quantum states, utilizing the total variation distance.

3. Key Contributions and Results

A. Main Theorem (Theorem 1)

The paper establishes a tight upper bound for any Schur concave function $f$:
$$\sup \{ f(\rho) - f(\sigma) \mid \sigma \in U_\varepsilon(\rho), \sigma \overset{m}{\prec} \rho \} \leq f(\rho) - f(\rho_{m,\varepsilon})$$

• Optimality: The bound is tight; there exist states where equality holds.
• Convergence: The bound tends to zero as $\min\{\varepsilon, 1/m\} \to 0$, provided $f$ is lower semicontinuous (or satisfies a specific limit condition). This applies to von Neumann, Rényi, and Tsallis entropies.

B. Application to von Neumann Entropy (Section 5)

The results are specialized for the von Neumann entropy $S(\rho) = -\text{Tr}(\rho \ln \rho)$.

• Explicit Bound: The bound $B(\rho, m, \varepsilon)$ is derived explicitly in terms of the spectrum of $\rho$ and the homogeneous extension of entropy $\hat{S}$.
• Gibbs States: The bound is applied to the Gibbs state of a quantum oscillator, providing a concrete formula for the error bound based on the mean number of quanta $N$.

C. $\varepsilon$-Sufficient Majorization Rank

A new concept is introduced: the $\varepsilon$-sufficient majorization rank, denoted $mr_\varepsilon(\rho)$.

• Definition: The minimal $m$ such that the relative error $\frac{S(\rho) - S(\sigma)}{S(\rho)} \leq \varepsilon$ for all $\sigma \overset{m}{\prec} \rho$.
• Significance: This characterizes how "fast" the spectrum of a state decays. For finite rank states, $mr_0(\rho) = \text{rank}(\rho)$. For infinite rank states with finite entropy, $mr_\varepsilon(\rho)$ is finite for any $\varepsilon > 0$.
• Result: An upper bound for $mr_\varepsilon(\rho)$ is derived, showing it depends only on $\varepsilon$ and the spectrum of $\rho$.

D. Generalization to Probability Distributions (Section 6)

The paper proves that all derived bounds and the concept of partial majorization rank hold for Schur concave functions on probability distributions with finite or countable outcomes, replacing the trace norm with the total variation distance.

4. Significance and Impact

1. Universal Continuity Bounds: The paper provides a unified framework for deriving continuity bounds for a wide class of Schur concave functions, not just entropy. This is crucial for analyzing the stability of quantum information measures under perturbations.
2. Partial Information Handling: It rigorously quantifies the error introduced when only partial spectral information (the first $m$ eigenvalues) is known, a common scenario in experimental physics and numerical simulations.
3. New Characteristic: The introduction of the $\varepsilon$-sufficient majorization rank offers a new metric for the "effective dimension" or complexity of a quantum state, bridging the gap between finite-rank approximations and infinite-dimensional reality.
4. Tightness: Unlike many loose continuity bounds in literature, these bounds are proven to be tight (optimal), meaning they cannot be improved without additional assumptions.
5. Practical Application: The explicit formulas for Gibbs states allow for precise estimation of entropy errors in quantum thermodynamics and oscillator models.

In summary, Shirokov's work bridges the gap between partial spectral knowledge and global functional properties of quantum states, providing rigorous, tight, and computable error bounds essential for the theoretical and practical analysis of quantum systems.