NumPyEigensolver does not support all BaseOperator instances.

[declanmillar]

Environment

• Qiskit Terra version: 3ce1737
• Python version: 3.10.6
• Operating system: macOS 13.0

What is happening?

NumPyEigensolver and by extension NumPyMinimumEigensolver do not support all BaseOperator instances, in particular SparsePauliOp. This is because it requires the data attribute, which Operator exposes but e.g. SparsePauliOp does not.

How can we reproduce the issue?

from qiskit.algorithms.eigensolvers import NumPyEigensolver
from qiskit.quantum_info import SparsePauliOp

op = SparsePauliOp("Z")
npe = NumPyEigensolver()
result = npe.compute_eigenvalues(op)

Output:

Traceback (most recent call last):
  File "/Users/declanmillar/Projects/qiskit/qiskit-terra/test.py", line 10, in <module>
    result = npe.compute_eigenvalues(op)
  File "/Users/declanmillar/Projects/qiskit/qiskit-terra/qiskit/algorithms/eigensolvers/numpy_eigensolver.py", line 248, in compute_eigenvalues
    self._solve(operator)
  File "/Users/declanmillar/Projects/qiskit/qiskit-terra/qiskit/algorithms/eigensolvers/numpy_eigensolver.py", line 144, in _solve
    if operator.data.all() == operator.data.conj().T.all():
AttributeError: 'SparsePauliOp' object has no attribute 'data'

What should happen?

The code should compute the eigenvalues without error, similarly to:

from qiskit.algorithms.eigensolvers import NumPyEigensolver
from qiskit.quantum_info import SparsePauliOp
from qiskit.opflow import PauliSumOp

op = SparsePauliOp("Z")
op = PauliSumOp(op)
npe = NumPyEigensolver()
result = npe.compute_eigenvalues(op)

or:

from qiskit.algorithms.eigensolvers import NumPyEigensolver
from qiskit.quantum_info import SparsePauliOp, Operator

op = SparsePauliOp("Z")
op = Operator(op.to_matrix())
npe = NumPyEigensolver()
result = npe.compute_eigenvalues(op)

Any suggestions?

Generalize the computation inside NumPyEigensolver to avoid the data attribute, perhaps via to_matrix() or, if possible, a sparse implementation that works for BaseOperator instances.

[declanmillar]

I'm happy to work on this issue myself.