Data Processing

Calibration

The calibration file (.xlsm) maps OMS partial pressures (Torr) to species molar fractions using a linear relationship:

pct = (P_torr - intercept) / slope

This fraction is then converted to a molar flux (nmol/s) via the ideal gas law using the user-supplied flow rate, temperature, and pressure.

Baseline correction

After conversion to nmol/s, a baseline correction is applied to each species to remove the instrument background signal and slow drift.

Method: percentile envelope

A rolling window percentile (default: 5th percentile, window size 101 points) is computed across the signal. This envelope tracks the lower bound of the signal — i.e., the background — while being largely insensitive to peaks of interest. The envelope is then smoothed with a Savitzky-Golay filter (default: window 201 points, polynomial order 2) to remove artefacts at the window edges.

The corrected signal is:

corrected = signal - baseline

Why this method?

OMS signals typically consist of sharp peaks on top of a slowly varying background. A simple global offset would not account for drift over long experiments. A percentile-based rolling window was chosen over a mean or median because:

• The 5th percentile selects the background floor rather than the signal midpoint, making it robust to large peaks that would bias a mean or median upward
• Rolling (local) computation handles slow drift that a single global estimate would miss
• The subsequent Savitzky-Golay smoothing removes the stepped artefacts that rolling percentile windows introduce at boundaries, giving a smooth baseline that can be visually validated in the plot

The raw signal and fitted baseline are shown together in the first plot panel, and the baseline-corrected signal in the second, so the user can visually verify the correction is sensible before interpreting the extracted values.