Removing Trace Preservation from a Universal Spectral Bound for 2-Positive Maps

Technical Summary: Removing Trace Preservation from a Universal Spectral Bound for 2-Positive Maps

Problem Statement
The paper addresses a conjecture posed by Vom Ende, Chruściński, Kimura, and Muratore-Ginanneschi regarding the spectral bounds of quantum dynamical maps. Previous work established that for any 2-positive, trace-preserving map $\Phi$ on a $d$-dimensional matrix algebra $M_d$, the trace satisfies the inequality:
$$\text{Tr } \Phi \leq d \cdot m(\Phi) + d^2 - d$$
where $m(\Phi)$ is the smallest real part of an eigenvalue of $\Phi$. This result was used to derive a universal bound on relaxation rates for generators of 2-positive, trace-preserving semigroups.

However, the authors of the prior work noted that the trace-preserving hypothesis is crucial for fixing the dominant eigenvalue (which equals 1 for maps and 0 for generators). They asked whether this hypothesis could be removed if the dominant eigenvalue is explicitly accounted for. Specifically, they conjectured that for any conditionally 2-positive map $L$ (where $e^{tL}$ is 2-positive for all $t \geq 0$), the following inequality holds:
$$\text{Tr } L \leq d \cdot m(L) + (d^2 - d) s(L)$$
where $s(L)$ is the largest real part of an eigenvalue of $L$. The difficulty in proving this lies in the fact that without trace preservation, a general 2-positive map lacks a distinguished fixed point, may be reducible, or may have a nilpotent adjoint, rendering the fixed-point arguments used in the trace-preserving case inapplicable.

Methodology
The paper proves the conjecture by reducing the general case to the known trace-preserving case (Theorem A) through a two-step normalization and perturbation strategy:

1. Quantum Doob Transform (Normalization):
   The author observes that if a 2-positive map $\Phi$ has a Hilbert–Schmidt adjoint $\Phi^*$ with a positive definite eigenmatrix $H$ (i.e., $\Phi^*(H) = rH$ for $r > 0$), one can construct a new map $T$ via conjugation and scaling:
   $$T = \frac{1}{r} C_H \circ \Phi \circ C_H^{-1}$$
   where $C_H(X) = H^{1/2} X H^{1/2}$. This transformation, analogous to a quantum Doob transform, converts $\Phi$ into a 2-positive, trace-preserving map $T$ that shares the same spectrum up to the factor $r$. Since $T$ is trace-preserving, Theorem A applies to it. By reversing the transformation, the inequality is recovered for $\Phi$ with the term $s(\Phi)$ replacing the fixed dominant eigenvalue.

2. Perturbation and Continuity:
   A general 2-positive map does not necessarily possess a positive definite eigenmatrix for its adjoint. To overcome this, the author introduces a perturbation $\Phi_\varepsilon = \Phi + \varepsilon \Delta$, where $\Delta(X) = \text{tr}(X)I$ is the unnormalized depolarizing map.

   • By Brouwer's fixed-point theorem, the adjoint of this perturbed map, $\Phi_\varepsilon^*$, acting on the compact convex set of density matrices, admits a fixed point $H_\varepsilon$.
   • Crucially, the perturbation ensures $H_\varepsilon$ is strictly positive definite ($H_\varepsilon \geq (\varepsilon/r_\varepsilon)I > 0$).
   • The inequality is established for $\Phi_\varepsilon$ using the normalization method above.
   • Finally, the result is extended to the original map $\Phi$ by taking the limit $\varepsilon \to 0$. The proof relies on the continuity of the spectrum and the trace, noting that the inequality depends only on these continuous quantities, not on the convergence of the eigenmatrices $H_\varepsilon$ themselves.

Key Contributions and Results

• Proof of the Conjecture: The paper proves that every conditionally 2-positive map $L: M_d \to M_d$ satisfies $\text{Tr } L \leq d \cdot m(L) + (d^2 - d) s(L)$. This result holds without assumptions of irreducibility, diagonalizability, or the existence of a faithful invariant state.
• Optimality of the Constant: The author demonstrates that the coefficient $d$ in the inequality is optimal. They construct a family of generators for completely positive, non-trace-preserving semigroups where equality holds. They show that replacing $d$ with any larger constant $a$ would violate the inequality for specific generators.
• Extension to Generators: The result is applied to generators of quantum dynamical semigroups. By analyzing the Taylor expansion of $e^{tL}$ near $t=0$, the map-level inequality is differentiated to yield the bound for generators.
• Universal Constraint on Decay Rates: The paper recasts the result in terms of decay rates relative to the dominant spectral growth rate. Defining $\Gamma_j = s(L) - \text{Re}(\lambda_j)$, the inequality is shown to be equivalent to:
  $$\max_k \Gamma_k \leq \frac{1}{d} \sum_j \Gamma_j$$
  This extends the known universal bound for trace-preserving dynamics to non-trace-preserving dynamics (such as tilted generators in quantum trajectory theory).

Significance and Claims
The paper claims to resolve an open question regarding the necessity of trace preservation in universal spectral bounds for 2-positive maps. The significance lies in:

1. Generalization: It removes a restrictive hypothesis (trace preservation) from a fundamental inequality, showing that the bound depends only on the spectral extremities ($m$ and $s$) and the dimension.
2. Methodological Insight: The proof utilizes a "soft" reduction technique, treating the trace-preserving theorem as a black box and using perturbation theory to bridge the gap to the general case. This avoids the need for a direct, complex algebraic proof for the non-trace-preserving case.
3. Physical Interpretation: The result provides a rigorous constraint on relaxation rates for open quantum systems that are not strictly trace-preserving, which is relevant for studying quantum trajectory fluctuations and tilted generators.

The author explicitly states that they do not address whether complete positivity can be spectrally distinguished from 2-positivity or whether positive trace-preserving maps in $d \geq 3$ can violate the original bound, leaving those questions open. The work is presented as a preprint, with the conjecture noted as unresolved in the literature at the time of writing.