Automatic Baseline (Baseplane) Correction
It is called “automatic” because iNMR automatically samples the baseline and calculate the best fitting polynomial, yet the user has a number of choices:
- if the baseline must be fitted to a polynomial model or if it will, instead, simply smoothed and subtracted. Practice demonstrates that, in 1D spectroscopy, is often better NOT to fit the baseline! The exception is represented by broad signals that you want to retain. Broad signals are not recognized as peaks and therefore removed. If you want to retain them, opt for the polynomial fitting.
- the sequence of corrections, if first along rows, or first along columns;
- the degree of the polynomial, form 0 to 7 (“?” means no correction);
- the region to fit; if not specified, the whole spectrum is fitted. In 1D spectroscopy, if you specify to fit only a limited region, iNMR will indeed do that, but the calculated polynomial will be subtracted from the whole spectrum, in other words the range of application is always the whole spectrum. In 2D spectroscopy, instead, the application range coincides with the fitted range.
- a filter parameter to guide the autosampling algorithm: in theory the higher is the parameter, the more refined is the search. If, erroneously, a peak is sampled, the correction will fit the peak instead of the baseline... The filter of value 'n' broadens (for an instant only) all the peaks to an extent of n extra points on each side, thus preventing the sampling nearby a peak.
You launch this task with the command ‘Process/Automatic’. If you want to narrow the fitted range, you have to select that range with the mouse before opening the dialog. The ppm limits will appear in the dialog. They are stored into the document. The next time you try to correct the baseplane, the limits will be shown again, even if nothing is selected, and EVEN IF ANOTHER REGION IS SELECTED. If you need to correct a second region, click the button ‘Reset Region’.
There is also another reason why you should avoid to restrict the fitted range. Good fitting requires many sampled points. The higher the degree of the polynomial, the MUCH higher the need for sampled points. If the range to fit contains few and small regions of pure baseline, fitting will be quite poor. That's why in the 1d case you have the possibility of fitting the baseline by eye.
With 1D spectra, iNMR shows the correction while the dialog is open. You can change the parameters and monitor how they affect the correction. In practice, you should, in this order:
- Uncheck the option “Fit with a Polynomial”
- Find the optimal value for the “filter” parameter. This is the most important parameter in this kind of correction.
- Refine the degree. A lower degree smoothes the correction, an higher degree makes it more similar to the spectrum.
- Use the keyboard keys, as always, to move to different parts of the spectrum and check the result.
- Press “OK” when satisfied.
After the first minute of practice, you'll have understood the whole theory behind baseline correction. To learn more, read the article: Cobas, J.C. et al. J. Mag. Res. 183, 2006, 145. It describes the smoothing algorithm applied.
How you Decide the Degree (order) of the Polynomial
In theory you can always specify the highest available degree, because the
fitting algorithm can decrease it, but can never increase it. This approach, unfortunately,
is not always good. Higher degree polynomials are quite troublesome for a computer. You should only use
them if necessary.
Another approach is to figure out, by visual inspection, the order of the baseline shape, and specify the same order for the fitting polynomial. This approach is too individualistic to be publicly suggested.
A generalized approach is to start with a middle-of-the-range value, like order 4. If the baseline, after correction, is already flat, there's no need for a different solution. If, instead, the corrected baseline is worse than the uncorrected one, decrease the order (after reloading and reprocessing the fid, in case of a 2D spectrum). If the corrected baseline in better than the original, but still not satisfactory, you can try with an higher degree. In this case you don't need to reload the FID, you can just add correction on correction.