1: \begin{abstract}

2: We propose a flexible power method for computing the leftmost, i.e.,

3: algebraically smallest, eigenvalue of an infinite dimensional tensor

4: eigenvalue problem, $H x = \lambda x$, where the infinite dimensional

5: symmetric matrix $H$ exhibits a translational invariant structure.

6: We assume the smallest eigenvalue of $H$ is simple and apply a power

7: iteration of $e^{-H}$ with the eigenvector represented in a compact way

8: as a translational invariant infinite Tensor Ring (iTR).

9: Hence, the infinite dimensional eigenvector can be represented by a

10: finite number of iTR cores of finite rank.

11: In order to implement this power iteration, we use a small parameter $t$

12: so that the infinite matrix-vector operation $e^{-Ht}x$ can efficiently

13: be approximated by the Lie product formula, also known as Suzuki--Trotter

14: splitting, and we employ a low rank approximation through a truncated

15: singular value decomposition on the iTR cores in order to keep the cost

16: of subsequent power iterations bounded.

17: We also use an efficient way for computing the iTR Rayleigh quotient and

18: introduce a finite size iTR residual which is used to monitor the convergence

19: of the Rayleigh quotient and to modify the timestep $t$.

20: In this paper, we discuss 2 different implementations of the flexible power

21: algorithm and illustrate the automatic timestep adaption approach

22: for several numerical examples.

23: \end{abstract}

24: