Pass Manager for Swap Optimal circuits raises errors for linear Pauli Strings

[JanikSchoenmeierKromer]

Describe the issue:

The PassManager in Step 2 "Apply Swap strategies" from the "how_to_build_qaoa_swap_circuit.ipynb" notebook raises an error message for "linear" Pauli Strings in the SparsePauliOp (e.g. 'ZIIIIIIIII'; containing only a single Z Pauli and Identities otherwise instead of the exemplary used two Z Pauli strings, e.g. 'ZZIIIIIIII') . These linear terms can be relevant for optimization problems.

Reproduce the code example:

Initializing the Sparse Pauli Operator in the notebook with an additional Pauli String and coeeficient of the form ['ZIIIIIIIII', 1.] and running the notebook reproduces the error.
E.g. Insert the following code instead of the second cell of the notebook and continue all steps including calling swapped_cost_layer = pm_pre.run(cost_layer).

from qiskit.quantum_info import SparsePauliOp

local_correlators = [['ZIIIIIIIII', 1.],['IIIIIZZIII', 1.], ['IIIIIIZIZI', 1.0], ['IIIIIIZIIZ', 1.0], ['ZIIIIZIIII', 1.0], ['IIIIZZIIII', 1.0], ['IIZIIIIZII', 1.0], ['IZZIIIIIII', 1.0], ['IIZZIIIIII', 1.0], ['IIIIIIIZZI', 1.0], ['IIIIZIIZII', 1.0], ['IIIIZIIIIZ', 1.0], ['IIIIIIIIZZ', 1.0], ['ZZIIIIIIII', 1.0], ['IZIZIIIIII', 1.0], ['ZIIZIIIIII', 1.0]]
cost_operator = SparsePauliOp.from_list(local_correlators)
print(cost_operator)

Error message:

    280 """Run all the passes on a ``circuit``.
    281 
    282 Args:
   (...)
...
    110 if len(edge) != 2:
--> 111     raise QiskitError(f"{pauli} does not have length two.")
    113 return edge

QiskitError: 'ZIIIIIIIII does not have length two.'

[eggerdj]

Thanks for raising this. We could handle this case in create_qaoa_swap_circuit

qopt-best-practices/qopt_best_practices/swap_strategies/build_circuit.py

Line 94 in a339798

def create_qaoa_swap_circuit(

I would handle terms with a single Z by filtering them out and creating a layer of Rz gates (applied on qubits that should have one). This layer of Rz gates must then be applied before the layers of Rzz's of the quadratic term when p is odd and after the layers of Rzz's when p is even.

[JanikSchoenmeierKromer]

Thank you for the proposed solution and explanation.