I need to create equal amplitude superposition (Rokhsar-Kivelson type) superpositions to use as initial states in dynamics.
I checked and the required bond-dimension is not onerous, so it should be possible to represent these using MPS
Claude suggested two methods that I knew of – brute force construction and imaginary time evolution approximation to the projector – and the third which seemed more elegant using
MPO(sites, “MaxEntangler”)
seems like a neat trick but I never heard of MaxEntangler .. any quick thoughts?
I want to project your “X” (product) state to just one total Z sector – this will give an equal amplitude superposition of all states in that sector, which has entanglement but not much – only \chi\sim L/2 nonzero Schimdt values, which I assumed to mean it’s “easy”.
Amended:
I asked Claude for documentation for MaxEntagler just now and it backed off and elaborated on this approach which now makes sense – basically run DMRG to find a GS in that sector of a properly designed H, but I still need to check if it has any other hiccups. Let me know if you have a “standard” method for this.
Thanks again, Vadim
PS
I did find references to MaxEntangler on some forum last night but can’t find anymore), since I too suspected it hallucinated.
PPS
Hi Vadim,
If you happen to know of a good parent Hamiltonian for the states you want to make, then making the Hamiltonian and using DMRG to get the state can be a good option.
Also if you have a specific Z sector you are thinking of, then there may be some explicit constructions that are not too hard to pull off. But to have a general code for a arbitrary Z sector of a product state is a bit tough, I think. (I have thought of a more fancy idea to do this before but never wrote it up – it is like a kind of site-by-site truncation algorithm to gradually project an arbitrary state into a certain sector.)