arXiv:2610.26483
Majorization for superpositions of nonorthogonal states and the existence of golden states
Gökhan Torun
Abstract
Golden states, from which every state can be prepared by free operations, are the units of a resource theory. In the resource theory of superposition relative to linearly independent, nonorthogonal reference states, they were known only for two reference states and for special overlap matrices, all of them maximally coherent in the Löwdin basis. We determine what decides their existence and when this Löwdin pattern holds. For every overlap matrix, the robustness and the geometric measure of superposition are maximized by the same states, which overlap equally with all reference states; the maximal values are set by the ground-state energy of a classical $XY$ model on the dual basis. These states, and only these, are golden under the maximal superposition-free operations, but they need not be maximally coherent in the Löwdin basis. For the Kraus-type free operations, we prove Gram majorization, a necessary and sufficient criterion for converting pure states into targets with nonzero coefficients, which reduces to majorization for orthonormal reference states. It shows that every golden state of these operations is maximally coherent in the Löwdin basis, yet golden states exist only for special overlap matrices: three reference states that are not orthonormal admit one if and only if their overlaps have a common modulus $s$ and their Bargmann phase $Φ$ obeys $s\geq2\cos[(π+|Φ|)/3]$, so that a geometric phase decides whether the theory has a unit. Finally, equal-overlap states yield a complete family of superposition witnesses in the Löwdin basis, which show that white noise is a resource unless the reference states are orthonormal.
AI Summary
The paper studies deterministic state conversion and the existence of universal resource states (“golden states”) for superposition relative to $d$ normalized, linearly independent, nonorthogonal reference states with Gram matrix $S$. It distinguishes Kraus-wise superposition-free operations (FO) from all free-state-preserving channels (MFO). For every $S$, maximal robustness is $\Gamma(S)-1$ and maximal geometric superposition is $1-1/\Gamma(S)$, where $\Gamma(S)=\max_{|u_j|=1}u^\dagger S^{-1}u$ and $d\leq\Gamma\leq d/\lambda_{\min}$. Their pure-state maximizers are equal-overlap superpositions of dual-basis vectors; these are exactly the golden states under MFO. The phase optimization corresponds to a classical $XY$ ground-state problem. These maximizers are maximally coherent in the Löwdin orthonormal basis precisely when $\Gamma=d/\lambda_{\min}$. For FO, a necessary-and-sufficient “Gram majorization” criterion characterizes pure-state conversion into targets with nonzero reference-basis coefficients: the source component Gram matrix must dominate a convex mixture of permutations of the target Gram matrix. The proof constructs Kraus operators from the positive-semidefinite slack. This establishes necessary golden-state conditions and a complete classification for $d=3$: apart from orthonormal references, golden states exist exactly when all overlap moduli equal $s$ and the Bargmann phase satisfies $s\geq2\cos[(\pi+|\Phi|)/3]$. Equal-overlap vectors also provide complete superposition witnesses; white noise has robustness at least $\Gamma/d-1>0$ for nonorthogonal references. Limitations include the nonzero-target-coefficient restriction, nonconvex phase optimization in higher dimensions, and unresolved FO existence for some higher-dimensional Gram matrices, including two-eigenvalue cases with both eigenvalues degenerate.
the author used Claude Opus 5.5 (Anthropic) to assist in refining the mathematical proofs, writing the code for the numerical checks and the figures, and preparing the manuscript.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 30 pages, 4 figures