Using pyEPR without Ansys HFSS#

pyEPR has two independent functional layers. The quantum analysis layer (DistributedAnalysis, QuantumAnalysis) requires only eigenfrequencies and energy-participation ratios as inputs — it does not need a live Ansys HFSS connection or even an Ansys installation.

This page explains the three main ways to use pyEPR’s quantum analysis without HFSS:

  1. Re-loading saved HDF5 data — re-run quantum analysis on a previously saved dataset.

  2. Bring your own participation ratios — supply participation ratios from any solver.

  3. Third-party and future solver integrations — upcoming and third-party solver integrations.

Re-loading saved HDF5 data#

After DistributedAnalysis.do_EPR_analysis() completes, pyEPR saves all field-integral results to an HDF5 file. This file is fully self-contained: you can close HFSS, move to another machine, and run the quantum analysis at any time.

import pyEPR as epr

# Path to the HDF5 file written by do_EPR_analysis()
data_file = r'C:\sim_results\my_device.hdf5'   # Windows path
# data_file = '/home/user/sim_results/my_device.hdf5'  # Linux/macOS

epra = epr.QuantumAnalysis(data_file)
epra.analyze_all_variations(cos_trunc=8, fock_trunc=7)
epra.report_results(numeric=True)
epra.plot_hamiltonian_results()

This is the recommended workflow for:

  • Sweeping cos_trunc / fock_trunc without re-running simulations.

  • Sharing results with collaborators who do not have HFSS.

  • Running the quantum analysis on macOS or Linux.

  • Archiving simulation results for reproducibility.

Bring your own participation ratios#

If you have eigenfrequencies and participation ratios from another source (a different FEM solver, an analytic calculation, or a circuit model), you can feed them directly into QuantumAnalysis by constructing the HamiltonianResultsContainer that it expects.

The key quantities pyEPR needs are:

Quantity

Description

freqs_hfss_GHz

Eigenfrequencies in GHz (one per mode)

Qs

Quality factors (use np.inf if not computing dissipation)

Ljs

Junction inductances in Henries (one per junction)

Cjs

Junction capacitances in Farads (one per junction; use 0 if unknown)

p_mj

DataFrame of inductive participation ratios: rows = modes, columns = junctions

sign_mj

DataFrame of participation signs ±1 (same shape as p_mj)

Below is a minimal worked example for a single transmon qubit coupled to a readout cavity:

import numpy as np
import pandas as pd
import pyEPR as epr
from pyEPR.core_quantum_analysis import QuantumAnalysis, HamiltonianResultsContainer

# ── Device parameters ────────────────────────────────────────────────────
modes     = ['qubit', 'cavity']
junctions = ['jj']

freqs_GHz = np.array([5.0,  7.2])   # GHz
Qs        = np.array([1e6,  1e4])   # quality factors

Lj_H = 10e-9   # 10 nH junction inductance
Cj_F = 2e-15   # 2 fF junction capacitance

# Participation ratios (rows=modes, cols=junctions)
# Qubit is nearly all junction; cavity has tiny participation
p_mj = pd.DataFrame(
    [[0.95],
     [0.02]],
    index=modes, columns=junctions,
)
sign_mj = pd.DataFrame(
    [[+1],
     [+1]],
    index=modes, columns=junctions,
)

# ── Build the results container ───────────────────────────────────────────
# HamiltonianResultsContainer is an OrderedDict keyed by "variation" label.
# A single variation (no parametric sweep) uses key '0'.
results = HamiltonianResultsContainer()
results['0'] = {
    'freqs_hfss_GHz': pd.Series(freqs_GHz, index=modes),
    'Qs':             pd.Series(Qs,         index=modes),
    'Ljs':            pd.Series([Lj_H],     index=junctions),
    'Cjs':            pd.Series([Cj_F],     index=junctions),
    'p_mj':           p_mj,
    'sign_mj':        sign_mj,
}

# ── Save and reload via QuantumAnalysis ───────────────────────────────────
import tempfile, os
tmp = tempfile.mktemp(suffix='.hdf5')
results.save(tmp)

epra = QuantumAnalysis(tmp)
epra.analyze_all_variations(cos_trunc=8, fock_trunc=7)
epra.report_results(numeric=True)

os.unlink(tmp)   # clean up temp file

Note

The sign convention for sign_mj follows the current orientation of the polyline defined in HFSS. For custom data, a consistent choice of +1 is fine unless you need to track relative phase between junctions.

Third-party and future solver integrations#

pyEPR’s quantum analysis layer is deliberately solver-agnostic. The only hard requirement is a set of eigenfrequencies and field-energy integrals. Below is the current status of integrations beyond Ansys HFSS.

PyAEDT (github.com/ansys/pyaedt)

Ansys’s official cross-platform Python scripting library for AEDT. PyAEDT can drive HFSS simulations on Windows, and in principle the field results it extracts can be passed into pyEPR’s quantum analysis layer. Direct integration is not yet packaged; contributions welcome.

Palace (github.com/awslabs/palace)

AWS open-source, GPU-accelerated FEM solver with an eigenmode solver. Palace outputs field data in a format that can be post-processed to yield participation ratios. A pyEPR–Palace bridge is a natural future contribution; the Bring your own participation ratios approach above is the recommended path in the meantime.

Qiskit Metal

Metal uses pyEPR internally. If you are using Metal’s EPR analysis renderer, the data flows through pyEPR automatically.

Custom / analytic

For simple geometries (e.g. lumped transmon estimates), you can compute participation ratios analytically and plug them directly into HamiltonianResultsContainer as shown in Bring your own participation ratios.

Tip

If you have added a solver integration or have working Palace / OpenEMS / COMSOL scripts that produce pyEPR-compatible output, please open a pull request or issue at github.com/zlatko-minev/pyEPR. The core maintainers are happy to help integrate and document new solver backends.