AI 中文总结
研究准周期磁场作用下多体局域 XXZ 自旋链的热化，通过雪崩机制分析，用两点关联函数和粒子数熵两个可观测量研究热雪崩性质，发现两点关联识别热雪崩不可靠，结果对标准雪崩框架修正有启示。
AI 中文摘要
我们研究了受准周期磁场作用的多体局域 XXZ 自旋链中由浴引起的热化。通过将链的一部分中的场强设置在局域化阈值以下来构建热包含物，并通过雪崩机制分析热化。为研究此类雪崩的性质，我们使用两个互补的可观测量，即两点关联函数和粒子数熵。令人惊讶的是，仅相关性未显示雪崩特征，而是在整个动力学过程中显示出关联长度的对数增长，并预测局域运动积分的局域长度远低于雪崩阈值。相反，粒子数熵显示出热雪崩的明显特征，对于足够大的浴，雪崩前沿逐渐深入到局域子系统中。两个可观测量之间的差异表明，与随机情况相反，两点关联在识别准周期系统中的热雪崩时不可靠。一方面，定性结果与标准雪崩理论的分析预测一致。另一方面，仍存在显著的定量偏差，这可能是由于准周期势产生的短程共振所致。我们的结果表明，标准雪崩框架需要修正以考虑准周期势的短程关联。
英文摘要
We study bath-induced thermalization in the many-body localized XXZ spin chain subjected to a quasiperiodic magnetic field. We engineer a thermal inclusion by setting the field strength in one part of the chain below the localization threshold and analyze thermalization via the avalanche mechanism. To study the nature of such avalanches, we use two complementary observables, namely the two-point connected correlation function and the particle-number entropy. Surprisingly, the correlations alone show no signatures of avalanches. Instead, they display a logarithmic growth of the correlation length throughout the dynamics and predict localization lengths of the local integrals of motion that remain well below the avalanche threshold. In contrast, the particle number entropy shows clear signatures of the thermal avalanche, with the avalanche front progressively propagating deeper into the localized subsystem for sufficiently large baths. The discrepancy between the two observables shows that two-point correlations are unreliable in identifying thermal avalanches in quasiperiodic systems, as opposed to the random case. On one hand, qualitative results are consistent with the analytical predictions of the standard avalanche theory. On the other hand, significant quantitative deviations persist, which could be due to short-range resonances generated by the quasiperiodic potential. Our results suggest that the standard avalanche framework requires revision to account for the short-range correlations of quasiperiodic potentials.
Comments15 pages, 10 figures, comments are welcome
