work-around with sparse representation of QuadOperator?

[michaelkaicher]

Hi all,

I am currently using openfermion.ops.QuadOperator and I wanted to check out some relations I derived which required me to represent the symbolic QuadOperator as an array in the N^M x N^M Hilbertspace, where N is the truncation of the bosonic levels and M is the number of bosonic modes.

I have the following very simple problem: I want to write the QuadOperator a_quad_operator_on_mode_0 = of.ops.QuadOperator(((0, 'q')), 1) as a NM x NM array, by first converting it to sparse object and then to an array through array_example_1 = of.boson_operator_sparse(a_quad_operator_on_mode_0, N, hbar=2).toarray() The issue is that in this particular instance, if I e.g. choose N=5 and M = 3, i will get array_example_1.shape = (5,5), instead of (125,125).

If instead of a_quad_operator_on_mode_0 I compute a_quad_operator_on_mode_1 = of.ops.QuadOperator(((1, 'q')), 1) then I get a_quad_operator_on_mode_1.shape = (25,25) and similarly a_quad_operator_on_mode_2 = of.ops.QuadOperator(((2, 'q')), 1) then I get a_quad_operator_on_mode_1.shape = (125,125).

One work-around would be to artificially insert the identity [q_j, p_j] = 2i on the modes j which are not k in of.ops.QuadOperator(((k, 'q')), 1)... but this seems like a super inefficient way. Any suggestions? Even just including that identity on the first and last mode (if either one is not k) would be enough already

Thanks! Michael.

PS: In FermionOperator class it is a bit nicer, since there you can specify the number of modes in your system when using get_sparse_operator.

[mhucka]

@michaelkaicher Thank you for opening this issue (and our apologies for the long silence). I'm not completely clear on what is being asked. Could you clarify what the question is?