Central limit theorems for non-symmetric random walks on nilpotent covering graphs: Part I

Central limit theorems for non-symmetric random walks on nilpotent covering graphs: Part I
复制标题

DOI:
10.1214/20-ejp486
复制
发表时间:
2018-06
期刊:
arXiv: Probability
影响因子:
--
通讯作者:
Satoshi Ishiwata;Hiroshi Kawabi;Ryuya Namba
Satoshi Ishiwata;Hiroshi Kawabi;Ryuya Namba
中科院分区:
其他
文献类型:
--
作者:
Satoshi Ishiwata;Hiroshi Kawabi;Ryuya Namba

文献摘要

相似文献

本文从离散几何分析的角度研究了幂零覆盖图上非对称随机游动的中心极限定理。建立了幂零覆盖图上非对称随机游动的半群CLT。通过离散调和映射将幂零覆盖图实现为幂零李群,给出了幂零李群上极限半群的一个几何刻画。更精确地说,我们证明了极限半群是由具有非平凡漂移的次拉普拉斯算子在具有Albanese度量的幂零李群上生成的。漂移项由随机游动的非对称性引起,当随机游动对称时漂移项消失。此外,通过施加“中心条件”,我们建立了一个功能性的交际法(即,Donsker型不变性原理)在幂零李群上的Hoelder空间中。功能CLT的情况下,实现不一定是谐波。我们还得到了幂零李群上极限扩散过程的一个显式表示,并讨论了它与粗糙路理论的关系。最后,我们给出了几个幂零覆盖图上的随机游动的例子,并给出了显式计算。
In the present paper, we study central limit theorems (CLTs) for non-symmetric random walks on nilpotent covering graphs from a point of view of discrete geometric analysis developed by Kotani and Sunada. We establish a semigroup CLT for a non-symmetric random walk on a nilpotent covering graph. Realizing the nilpotent covering graph into a nilpotent Lie group through a discrete harmonic map, we give a geometric characterization of the limit semigroup on the nilpotent Lie group. More precisely, we show that the limit semigroup is generated by the sub-Laplacian with a non-trivial drift on the nilpotent Lie group equipped with the Albanese metric. The drift term arises from the non-symmetry of the random walk and it vanishes when the random walk is symmetric. Furthermore, by imposing the "centered condition", we establish a functional CLT (i.e., Donsker-type invariance principle) in a Hoelder space over the nilpotent Lie group. The functional CLT is extended to the case where the realization is not necessarily harmonic. We also obtain an explicit representation of the limiting diffusion process on the nilpotent Lie group and discuss a relation with rough path theory. Finally, we give several examples of random walks on nilpotent covering graphs with explicit computations.