wrong excited state population when building noise model from backend - frequency unit

[ctlila]

Informations

• Qiskit Aer version: 0.12.2
• Python version: 3.10
• Operating system: WSL - Ubuntu

What is the current behavior?

When building a noise model from a backend with a non zero temperature, the excited state population is not correctly computed.
The Maxwell Boltzmann distribution is computed in qiskit_aer.noise.device.models._excited_population(freq, temperature) where the frequency and temperature are normally in GHz and mK, but the frequency is passed in Hertz, yielding a 1e9 error in exp_param and an incorrect population.

The frequency is taken from relax_params in qiskit_aer.noise.device.models._device_thermal_relaxation_error, called in qiskit_aer.noise.device.models_basic_device_target_gate_errors, where the params are taken from Target.qubit_properties, that is an IBMQubitProperties object, where the frequency is stored in Hz.

Steps to reproduce the problem

Build a noise model from backend with non zero temperature :

from qiskit_aer.noise import NoiseModel
provider = IBMProvider()
machine='ibmq_jakarta'
model = NoiseModel.from_backend(provider .get_backend(machine), temperature=10)

On my end this yields a warning :
qiskit_aer/noise/device/models.py:394: RuntimeWarning: overflow encountered in exp exp_param = exp((47.99243 * freq) / abs(temperature))

And then look at freq in qiskit_aer.noise.device.models._excited_population, it's in Hz.

What is the expected behavior?

The frequency should be expected in Hz, and a factor of 1e-9 should be added to the exp_param computation.

Suggested solutions

Change the function to accept freq in Hz and then add a 1e-9 factor in exp_param.