The time evolution for a h5 file

Question

Hello

I got a ground state of a nearest-neighbour hubbard model by DMRG and called the state psi.I saved the psi by using ITensors.HDF5. I want to get the time evolution of psi and use the modified code from enter link description here

However,when the program is running the
psi = apply(gates, psi; cutoff=cutoff)
It can't work and get the output:

ERROR: BoundsError: attempt to access 0-element Vector{Int64} at index 
Stacktrace:
[1] getindex
   @ ./array.jl:861 [inlined]
[2] product(o::ITensor, ψ::MPS, ns::Vector{Int64}; move_sites_back::Bool, apply_dag::Bool, kwargs::Base.Pairs{Symbol, Float64, Tuple{Symbol}, NamedTuple{(:cutoff,), Tuple{Float64}}})
@ ITensors ~/.julia/packages/ITensors/cu9Bo/src/mps/abstractmps.jl:1751
[3] product(As::Vector{ITensor}, ψ::MPS; move_sites_back::Bool, kwargs::Base.Pairs{Symbol, Float64, Tuple{Symbol}, NamedTuple{(:cutoff,), Tuple{Float64}}})
@ ITensors ~/.julia/packages/ITensors/cu9Bo/src/mps/abstractmps.jl:1859
[4] top-level scope
@ REPL[1]:1

I can make sure the Hamiltonian in TEBD is right.When I used

psi = productMPS(s, n -> isodd(n) ? "Up" : "Dn");

to set a random MPS to replace psi. The TEBD works well.
Having tried the hubbard model and the Heisenberg model,I think the question results from the relation between "h5"file and the function"apply".If I can't run time evolution for a ground state from h5 file by this way,Is there another way?

Thanks

Answer 1

Thanks for the question. We may need to discuss more to figure out what is going on, but I bet it is not specific to the function apply. I think what has happened is probably that the "site" indices you used to make your MPS when you wrote it to disk originally are different from the ones that you used to make the gates that you are inputting into apply.

It would explain why when you re-make psi using productMPS that the error goes away.

Did you also write the "s" object to disk and re-read it back in? An alternative to that is that you can "grab" the sites of an MPS like this:

s = siteinds(psi)

so you can use that to make sure that s is compatible with psi before using either of them.

Please try updating your code to make sure "s" is compatible with psi and if you are running into more issues happy to discuss it more below.

Best regards,
Miles

Comments

Thanks a lot Miles! I don't save the "s" in disk as a file. I only use
 "s= siteinds(psi)"
to replace
 "s = siteinds("Electron", N; conserve_qns=true)"
in time evolution. The program works very well and outcome is just as I expect. You've helped me a lot.
I have a idea that the example can be added in the part MPS Time Evolution of " Documentation for the Julia version of ITensor". Probably many users like me want to use ITensor to run time evolution.

Best wishes,
cgh

---

Really glad you were able to fix it and that siteinds helped. Good idea to add examples to the documentation showing saving of wave functions to disk.