Error when adding a process matrix to the noise model about pygsti HOT 4 CLOSED

Quick answer: you need to insert the line
XX = pygsti.tools.change_basis(XX, 'std', 'pp')
before the final line of your example.

Explanation:
You're getting this error because you're assigning a complex numpy array (see the elements of XX in your code above) to a model operation. Since pyGSTi uses Hermitian bases for density matrices, it assumes that if you're specifying a super-operator (a map from density matrices to density matrices) as a dense array then this array will be real because a density matrix expressed in a Hermitian basis is a real vector. Since your array is complex, it interprets it as a unitary operator on (complex) state-vectors as opposed to density matrices - i.e. it has evolution type "statevec". But model1 in your example has already been setup to represent quantum states as density matrices (evolution type "densitymx"), so it gives you the error you see.

To fix this, you need to convert XX into the Pauli basis - that is, write the action of XX as you have it as the action on a state represented as a length-16 real vector specifying the 2-qubit density matrix as a linear combination of the 16 2-qubit product-of-two-Pauli matrices. To do this, you can use the change_basis function to move from the "standard" ("std") matrix-unit, basis you've expressed XX in to the "Pauli-product" ("pp") basis the your model expects - that's what the line:
XX = pygsti.tools.change_basis(XX, 'std', 'pp') does.

from pygsti.

Thank you for the explanation! Makes sense to me and works now.

As for the second question: when I'm defining the processor specification that I'm trying to benchmark or run any simulation on, XX gate is not one of the standard set of gates. Just like I did for the model building, is it possible to input XX as a basis get for the processor?

As an example, my goal is to do something like this:

#Specify the device to be benchmarked - in this case 2 qubits
nQubits = 2
qubit_labels = [0,1] 

Gxx = ...........

gate_names = ['Gx', 'Gy','Gxx'] 


availability = {'Gxx':[(0,1)]}
pspec = pygsti.obj.ProcessorSpec(nQubits, gate_names, availability=availability, 
                                 qubit_labels=qubit_labels)

from pygsti.

You should be able to just supply the nonstd_gate_unitaries argument to the ProcessorSpec constructor, which as its name implies, should contain the unitary for your MS gate, not the super operator. So, using the U1_xx variable from your original post, you can just write:

gate_names = ['Gx', 'Gy','Gxx'] 
availability = {'Gxx':[(0,1)]}
pspec = pygsti.obj.ProcessorSpec(nQubits, gate_names, availability=availability, 
                                 qubit_labels=qubit_labels,
                                 nonstd_gate_unitaries={'Gxx': U1_xx})

from pygsti.

That's perfect! Thanks for all the help, Erik!

from pygsti.