arXiv:2609.23633
Spectral width and polynomial degree in perfect state transfer
Xingkun Song
Abstract
We study the minimum time for perfect state transfer under polynomial Hamiltonians with bounded degree and spectral width. For a strongly cospectral pair and width bound $W$, the optimum, when finite, is an integer multiple of $π/W$, determined by integer interpolation with prescribed parities. For equally spaced supported eigenvalues with alternating signs, we give degree bounds under which every minimizer is affine, and sharp asymptotics for each fixed exact degree. Near-minimizing phase polynomials satisfy a quantitative Chebyshev stability estimate. We determine the optimal transfer time for every degree bound on hypercubes of odd prime dimension. For complementary vertices of $J(2m,m)$, the optimal time at fixed spectral width grows exponentially in $m$ throughout an interval of feasible degrees. We also construct polynomial Hamiltonians showing that every feasible degree $m-t$ with $t=o(m)$ admits subexponential transfer time.
AI Summary
The paper asks how quickly polynomial Hamiltonians $p(H)$ can achieve perfect state transfer between strongly cospectral vertices when spectral width is bounded by $W$ and polynomial degree is restricted. Degree is measured modulo the supported minimal polynomial; most general results constrain local spectral width. Without a degree restriction, the optimal time is $\pi/W$, attained by assigning two energies to the two cospectral sign classes. With degree bound $d$, the optimum is $T_d(W)=\pi R_d/W$, where $R_d$ is the minimum range of integer-valued interpolation data with prescribed parities. This converts time optimization into an arithmetic polynomial problem. For equally spaced supported eigenvalues with alternating signs, every minimizer is affine through degree $d\leq(0.6298694-o(1))n$, giving $R_d=n$. For antipodes of an odd-prime-dimensional hypercube $Q_p$, the exact optimum is $p\pi/W$ for every $1\leq d<p$, dropping to $\pi/W$ at $d=p$. Fixed exact degree $k\geq2$ instead gives time asymptotic to $\pi n^k/(W2^{2k-3}k!)$, with quantitative Chebyshev stability near optimality. For complementary vertices of $J(2m,m)$, feasible degrees begin at $q=2^{\lfloor\log_2m\rfloor}$. Transfer time is exponential in diameter $m$ throughout any nonempty interval $q\leq d\leq\beta m$, with $1/2<\beta<0.6298694$, whereas $R_{m-t}\leq\exp[O(t\log(em/t))]$ makes feasible sublinear degree losses subexponential. Proofs combine phase interpolation, spectral-moment identities, discrete orthogonal-polynomial bounds, and coefficient rounding. The precise transition thresholds and exponential rates remain open. Arbitrary real polynomial coefficients are allowed; coupling-sign and individual-coupling constraints are not imposed.
Central claim may not hold: The abstract's unqualified integer-multiple claim fails for ordinary polynomial degree combined with global spectral width: affine Hamiltonians on A(P_3)⊕[2] have optimal endpoint transfer time (1+√2)π/W.
Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.
- Comments
- 20 pages, 0 figures