The 1D 13C spectra collected in the SSPPS NMR Facility often show uneven baselines. The usual baseline correction routines do not improve the spectra much, but recently I discovered that Bruker provides a processing macro that can dramatically improve the appearance of these spectra.
The figure below shows two 1D 13C spectra. The top spectrum in red has been processed in the default fashion, i.e. apodisation, zero filling, fourier transformation, and phase correction. The lower spectrum in blue used the Bruker macro "cryoproc1d" prior to applying these steps. Both spectra were generated from the same raw data, only the processing steps were different.
The macro works by using linear prediction to modify the points at the start of the FID. Typically, linear prediction is used to calculate additional points beyond what is collected, but it can also be used to replace points. The cryproc1d macro replaces points at the beginning of the FID. Distortions in the baseline are likely due to data points near the start of the FID that are inconsistent with the rest of it. Since these distorted points are only near the start of the FID they appear to come from rapidly relaxing nuclei that give rise to broad peaks. The top panel of the figure below shows the start of the FID as it was actually recorded, and the lower panel shows the FID after correction.
The points at the start of the FID are distorted when recording data with a cryoprobe, as can be seen by the large spike in the values. Additionally, Bruker's digital oversampling adds small intensity points at the start of the FID. The cryoproc1d macro removes the points added by the oversampling and uses linear prediction to replace the spike with more reasonable values.
Baseline distortions are mentioned in Bruker's cryoprobe manual:
Cryogenically cooled probes are more prone to background signals for two reasons: (1) the high signal-to-noise ratio of the receiving system enhances the background by the same factor as the desired signals; (2) the nuclear magnetization of any material increases in proportion to 1/T (Curie law) which results in a considerably stronger NMR response from substances located in the cold parts of the resonant circuit. Consequently, a background virtually cannot be avoided and is usually much larger than for conventional probes.
It seems Bruker is suggesting the baseline distortions are due to materials in the probe producing broad background signals. Since the baseline distortions are not always the same and the data can be corrected so easily by replacing points at the start of the FID I think the baseline distortions are more likely due to problems with the FID.
The cryproc1d macro has been integrated into the standard processing parameters on the SSPPS NMR for all the 1D 13C experiments. Processing off-spectrometer with other software will use the corrected data.
Acknowledgments
Thanks to Karol Francisco in the Ballatore lab for allowing me to use her data.
Brilliant fix for a recurrent irritating problem. I take it’s a reverse linear prediction - later points used to back-predict the deleted leading points. Thanks for implementing this in the 1D 13C processing, Brendan!
ReplyDeleteYes this is often called backwards linear prediction, and it seems to work quite well. Our users have already noticed how much flatter their 1D 13C spectra are.
ReplyDeleteIm not sure what causes the FID artifact, pulse ringdown or something else, but backwards LP is indeed a good way to improve the spectra. Increasing the pre-scan delay can also help.
ReplyDeleteEven on room temperature probes I see the same phenomenon in specific cases, for instance in 19F spectra. There too backwards LP is a good fix, though it might require manually setting the LP parameters instead of running cryoproc1d.