addBasis gives wrong results when Cutoff is very small

[chuffa]

Hello,

I have been using your TDVP implementation and tested it quite rigorously also against my own TDVP implementation and it works very well.

The only issue I found was that the basis expansion gives me very wrong results when the cutoff parameter is chosen too small. To be precise, I use the following code to expand the subspace:

 if(n < 3) // n is the time step
 {
       // Global subspace expansion
       std::vector<Real> epsilonK = {1E-10, 1E-10};
       addBasis(psiT,H,epsilonK,{"Cutoff",1E-12,
                                                  "Method","DensityMatrix",
                                                  "KrylovOrd",3,
                                                  "DoNormalize",true,
                                                  "Quiet",true});
}

If I use the Cutoff as above (1E-12), the results are reasonable but if I decrease it to 1E-15 they are very wrong. The quantity I am looking at is the single particle Green function in time and I can use such a small Cutoff because I am testing it on a very small system with complete bond-dimension. If necessary I can also provide the full code of course but since everything depends only on the one parameter Cutoff I chose not to (and also because it is a mess :) ).

Thanks, Daniel

[mingruyang]

Hi Daniel,

Thanks for the report! Could you please provide a minimal example that produces the wrong results? That would make me easier to fix the possible bugs.

Sincerely,
Mingru