IRIS: 2D-IRIS analysis (plugin)

This example introduces 2D-IRIS analysis of CO adsorption on a sulfide catalyst with the optional spectrochempy-iris plugin.

Requires the official spectrochempy-iris plugin. Install with: pip install spectrochempy[iris].

import spectrochempy as scp

Loading and coordinating the dataset

The example data contain infrared spectra recorded during CO adsorption. The dataset has wavenumber coordinates along x and acquisition timestamps along y.

X = scp.read_omnic("irdata/CO@Mo_Al2O3.SPG")
X.coordset

The IRIS model is easier to interpret with pressure coordinates along the observation axis, so we attach the measured CO pressures to y.

pressures = [
    0.003,
    0.004,
    0.009,
    0.014,
    0.021,
    0.026,
    0.036,
    0.051,
    0.093,
    0.150,
    0.203,
    0.300,
    0.404,
    0.503,
    0.602,
    0.702,
    0.801,
    0.905,
    1.004,
]
c_pressures = scp.Coord(pressures, title="pressure", units="torr")

Keep the original time coordinate as a secondary coordinate, and make pressure the active one for plotting and IRIS fitting.

c_times = X.y.copy()
X.y = [c_times, c_pressures]
X.y.select(2)
X.coordset

We now select the CO adsorption spectral region.

X_ = X[:, 2250.0:1950.0]
_ = X_.plot(colorbar=True)
_ = X_.plot_contourf(colorbar=True)

IRIS analysis without regularization

The plugin exposes its workflows through scp.iris and also adds dataset-bound helpers under dataset.iris. We start by building the Langmuir kernel from the dataset accessor.

K = X_.iris.kernel_matrix(kernel_type="langmuir", q=[-8, -1, 50])
K.kernel

The model can then be fitted with no explicit regularization.

iris1 = scp.iris.IRIS(log_level="INFO")
_ = iris1.fit(X_, K)

Grouped fitted outputs are available from result. Historical direct attributes such as f remain supported, but the grouped result is the preferred way to inspect the fit.

f = iris1.result.f
_ = iris1.result.K

_ = f.plot_contour()
_ = iris1.plotmerit()

Manual regularization search

A second fit scans a regularization range manually and displays the L-curve.

iris2 = scp.iris.IRIS(reg_par=[-10, 1, 12])
_ = iris2.fit(X_, K)
_ = iris2.plotlcurve(title="L curve, manual search")

The best regularization parameter is visually near index -6 for this dataset, corresponding to a lambda around 1e-4.

_ = iris2.result.f[-6].plot_contour()
_ = iris2.plotmerit(-6)

Automatic search

We can then refine the search around the visually selected range.

iris3 = scp.iris.IRIS(log_level="INFO", reg_par=[-6, -2])
_ = iris3.fit(X_, K)
_ = iris3.plotlcurve(title="L curve, automated search")

The largest curvature of the L-curve is at index 5 for this automated search.

_ = iris3.result.f[5].plot_contour()
_ = iris3.plotmerit(5)

The historical root example also demonstrated the legacy direct classes IRIS and IrisKernel. New gallery material should prefer the plugin namespace shown above: scp.iris.IRIS and dataset.iris.kernel_matrix.

Uncomment the following line to display all figures when running the script directly with Python.

# scp.show()