In AMS and ASCM, several ions are not directly measured but are calculated based on the CO? (HR) and m/z 44 (UMR) signal, respectively. The details about this calculation can be found in the fragmentation table.
Necessity of removing CO2-related variables
Because these variables are derived from the same measurement, they do not provide independent information. Including them in the PMF input would therefore introduce redundancy and violate the PMF assumption that each variable contributes independent information. Excluding CO?-related variables is thus important for obtaining a physically meaningful and numerically stable PMF solution.
This issue becomes particularly relevant when PMF is run in robust mode, which is standard practice. During the iterative optimization, the algorithm evaluates each variable based on its residual-to-uncertainty ratio and may downweight variables that exceed the robust threshold. If several variables represent the same CO? signal, one copy could be downweighted while another is not. This leads to inconsistent treatment of identical signals, even though they correspond to the same physical quantity.
SoFi handles this automatically in a robust and transparent way. Before running PMF, SoFi removes all CO?-related variables from the dataset. After the analysis, they are reconstructed using the appropriate fragmentation table. This approach ensures that the CO? signal is treated consistently while still contributing to the interpretation of the results in a controlled and physically meaningful manner.
Why downweighting is not enough
Simply downweighting CO?-related variables would not fully solve the issue. Even with reduced weights, these variables would remain in the PMF input, so the redundancy and potential inconsistencies could still affect the solution. In addition, including them can bias the model toward explaining variability in CO?, which may disproportionately influence oxygenated factors and reduce sensitivity to less intense but chemically important ions.
To ensure that this process works reliably, the fragmentation table must match the selected instrument type. SoFi enforces instrument-consistent fragmentation tables automatically, based on the instrument chosen by the user. This design helps ensure that CO?-related variables are handled correctly and that PMF analyses remain reliable, reproducible, and physically meaningful.
So this is not an error message, just SoFi letting you know that your chosen fragmentation table does not match your data and that SoFi will take care of it.
