DMRG with fixed bond dimension

How do I use this algorithm? What does that parameter do?
Post Reply
rike
Posts: 3
Joined: 06 Jan 2021, 02:38

DMRG with fixed bond dimension

Post by rike »

Hello,

I was wondering if there is an easy way to force the bond dimensions of a finite MPS (obtained via DMRG) to always be maximal up to some chi_max, e.g.:
chi = [2, 4, 8, 16, 32, 40, 40, …, 40, 40, 32, 16, 8, 4, 2] for spin-1/2 and chi_max = 40.
I played around with chi_min, but it doesn’t seem to do what I want.

Thank you!
Rike
User avatar
Johannes
Site Admin
Posts: 413
Joined: 21 Jul 2018, 12:52
Location: TU Munich

Re: DMRG with fixed bond dimension

Post by Johannes »

Welcome, Rike!

chi_min can only avoid reducing the bond dimension, not increase it further.
In addition to setting chi_min to the same value as chi_max, you should set svd_min to 0.
rike
Posts: 3
Joined: 06 Jan 2021, 02:38

Re: DMRG with fixed bond dimension

Post by rike »

Thanks Johannes!

In the end these truncation parameters did the job for me: 'trunc_params': {'chi_max': 40, 'svd_min': None, 'trunc_cut': 0}

(setting svd_min to zero returned a warning in my case since its logarithm is calculated somewhere.)

Best,
Rike

Post Reply