Supported Observables
What is supported
Pauli sums: weighted sums of tensor products of I, X, Y and Z, with complex coefficients.
| Form | Supported | Notes |
|---|---|---|
| Single Pauli string | Yes | SparsePauliOp("ZZI") |
| Weighted sum of Pauli strings | Yes | All terms evaluated in one run |
| Complex coefficients | Yes | The Qiskit estimator returns the real part |
| Array of observables | Yes | Evaluated element by element |
| Observable on more than 64 qubits | Yes | Masks span several 64-bit words |
| Identity term alongside other terms | Yes | Exact |
| Only identity terms | No | Raises InternalError; add the constant yourself (details) |
| Non-Pauli operators (general matrices, projectors) | No | Not representable |
| Hermiticity | Not checked | A non-Hermitian sum gives a complex value natively, its real part through Qiskit |
Which engines can evaluate one
| Engine | Condition |
|---|---|
| Dense | Always |
| Stabilizer | The circuit is entirely Clifford |
| Pauli propagation | Clifford gates and rz rx ry p u only, and the circuit is not entirely Clifford |
| MPS | Always (exact mode) within the memory budget |
| Tensor network | Always, within the memory budget |
| Cut and recombine | A cut exists (at most two cuts) |
Which one runs is the planner's choice; see Plan.
Estimator-specific limits
MENOSEstimator additionally requires that the circuit carry no unbound parameters, and that the observable be a SparsePauliOp (or a mapping of Pauli labels to coefficients, as the Qiskit pub format produces).