Cannot initialize binhamiltonian from qubithamiltonian

[andrew-koren]

Describe the bug An error occurs when trying to initialize a binary hamiltonian from qubit hamiltonian when the qubit hamiltoian has identity in it.

To Reproduce Steps to reproduce the behavior: Ideally an executable code snipped like

import tequila as tq
from tequila.grouping.binary_rep import BinaryHamiltonian

pauli_string = tq.PauliString.from_string('X(0)X(2)')
hamiltonian = tq.QubitHamiltonian.from_paulistrings(pauli_string)
assert hamiltonian.n_qubits == 2 #Should be 3 if you count I(1)?

BinaryHamiltonian.init_from_qubit_hamiltonian(hamiltonian) 
#-> TequilaException: PauliString acts on qubit number larger than n_qubits given PauliString=+1.0000X(0)X(2), n_qubits=2

Expected behavior Should create Binary Hamiltonian without throwing error

Computer (please complete the following information):

• OS: Windows
• Version: Python 3.11.9 Tequila 1.9.6dev

[andrew-koren]

I have a poor-mans solution, I changed the QubitHamiltonian class

@property
def n_qubits(self):
    return len(self.qubits)

to

@property
def n_qubits(self):
    self.qubits[-1]+1

Which selects the value of the largest qubit number and adds 1 for 0-indexing. This allowed the binary hamiltonian to be made without error and function correctly for my purposes.

[kottmanj]

Hi,

Thanks for reporting the bug. The issue should be resolved with the latest pull request to the devel branch. This fix will soon be available in the master branch/PyPI release, providing a reliable solution to the problem you encountered.

In the meantime you can upgrade with

pip install --upgrade git+https://github.com/tequilahub/tequila.git@devel

If the problem persists please let us know again.

I couldn't use the proposed fix as this would break the package in other instances. Overall, the contention is that .n_qubits always gives back the number of qubits where non-trivial actions occur. Tequila is then able to map to smaller qubit systems before simulating (an application where this often pays off is adapt-vqe calculations, where in the beginning, the circuit is not acting on all the qubits).

Example:

if you are initializing

U = tq.gates.H(target=10) + tq.gates.X(target=15, control=10) 
H = tq.paulis.X(10) + tq.paulis.Y(15) + tq.paulis.Z(200) 
E = tq.ExpextationValue(H=H, U=U)
result= tq.simulate(E)

before compiling to the simulation backend tequila will map the circuit to a 2-qubit circuit, send this circuit to the simulator, and will afterwards map the result back. The Pauli-Z in the Hamiltonian will be traced out in the preprocessing.