SchottkySA — Schottky spectra Shape Analyzer

SchottkySA fits spectral components with empirical line shapes measured in the same experiment. It returns centroids, integrated areas, widths and joint resampling distributions through Python, a command line and an interactive Qt application. Version 1.1 adds component-specific templates, profile comparison and averaging, independent edge controls and expanded saved-session replay.

Installation

python -m pip install SchottkySA         # scientific core
python -m pip install 'SchottkySA[gui]'   # desktop application
python -m pip install 'SchottkySA[all]'   # GUI and spreadsheet exports

Python 3.10 or later is required. From a checkout, use python -m pip install -e '.[dev]' for development.

ssa
ssa --version

A fit in Python

from ssa import build_peak_template, fit_template_region

template = build_peak_template(f_reference, power_reference, name="reference")
fit = fit_template_region(f_target, power_target, template=template, n_peaks=2, init_mus=centres, mu_bounds=bounds)
print(fit["info"]["mus"], fit["info"]["areas"], fit["red_chi2"])

Frequencies and centroid bounds are in Hz; power is linear. The template is normalized to unit area and centered at zero center of gravity. A component is a translated, optionally dilated copy of that shape. Its area has units of the input ordinate times Hz.

component_templates=[template_1, template_2] assigns a different profile to each component. Ordered centroid bounds retain identity even when the supports overlap. A common dilation means a common fractional stretch; different templates can still give different component widths.

python examples/quickstart.py demonstrates the numerical interfaces using constructed test spectra. These spectra are software fixtures, not experiments.

Interactive analysis

Load an NPZ containing frequency and linear-power arrays. Select the reference and target regions, build a template and inspect its edges. The profile workbench provides two linked spectrum views so distant regions can be compared without losing the current analysis window.

Double-click to add a centroid, move its marker and edit or drag its allowed range. The multiple-template selector assigns a profile to each component. Both template edges have separate measured-point counts and zero-anchor distances. The preview shows the selected samples and closure before fitting.

Harmonic fields accept integers, decimals and positive fractions such as 214/214 and 215/214. Profile transfer scales the frequency coordinate by the harmonic ratio and preserves unit area. Bidirectional cross-fits compare each selected reference with the other's measured region.

Several centered shapes can form a weighted empirical mean without extra smoothing. A source-sensitivity bank retains differences among the source profiles. Independent-reference mean draws are available when their independence is justified. Different harmonics of one acquisition are not automatically independent reference measurements.

Uncertainties and fit diagnostics

Target residual resampling refits a fixed template after circular-block resampling. Reference propagation refits the unchanged target with a bank of perturbed reference shapes. Seeds and replicate counts are explicit. Each parameter retains the 16th, 50th and 84th percentiles.

Reported combined errors add the two symmetric percentile halfwidths in quadrature under an independence assumption. Separations and area ratios are formed from joint component draws, preserving within-fit covariance. Repeated uses of one reference bank share the selected draw. Distinct banks use an independent-draw model unless the caller supplies a different treatment.

Reduced chi-square, residuals, optimization status, search-bound proximity and returned resampling counts accompany each fit. Formal Jacobian covariance is a separate diagnostic; it is not added again to the resampling result. Fixed edge choices and source-to-target shape mismatch require separate sensitivity checks. See Scientific method.

Headless run and replay

ssa run reads named-column CSV, TSV or whitespace tables and writes a JSON summary plus per-bin curves. ssa replay reconstructs a saved TOML session. Use ssa run --help and ssa replay --help for their options.

The GUI's Save config + tables and Replay config… controls create or restore DATE_SchottkySA_CONFIG_suffix bundles containing config.toml, fit history and a current-fit overlay. The input spectrum is referenced by path and SHA-256. Completed fits preserve template recipes, component assignments, centroid bounds, harmonics, resampling settings and numerical checkpoints.

Replay reports changed-input warnings and numerical mismatches. Absolute frequency positions are compared at 1e-5 Hz tolerance; other scalar fields retain their documented relative/absolute tolerances. See Reproducibility and the Profile workbench guide.

Development and citation

pytest
ruff check .
ruff format --check .
pyright
python -m build
twine check dist/*
python scripts/check_release.py dist/*

Use CITATION.cff to cite the software. The badge identifies the software family; a published version has its own archive DOI. See release instructions, CONTRIBUTING.md and SECURITY.md. SchottkySA is licensed under GPL-3.0-only.