NMR FACILITY ::: Division of Chemistry and Chemical Engineering ::: CALTECH

NMR Facility Home | Instruments and Sign-up | Documents | FAQs

NMR Facility FAQ's

Where is my NMR data?

The NMR Lab may have up to 3 copies of spectra that you save:

On the host workstation of the spectrometer where it was originally acquired. Any data you save is immediately accessible on the network drive from that spectrometer (e.g. \\indy.caltech.edu\nmrdata)
On our main lab server, mangia. Newly created data is copied from each spectrometer to mangia overnight. Consequently, data acquired today will not be visible on mangia until tomorrow.
On the cloud backup made from mangia. It may take a few days for new spectra to get copied to the cloud backup.

See the page on retrieving data for instructions on accessing the network drives on the spectrometer hosts and on mangia. Individual users do not have access to the cloud backup.

The data you save on the spectrometer workstation will be there until one of two things happens:

You leave Caltech and your account on the workstation is deleted.
The computer breaks. The commonest cause of this is hard drive failure. Typical hard drives last 2-3 years in computers that are on all the time. We have a dozen or so computers in the lab, so hard drive failures occur multiple times per year. After a hard drive failure, we reload VnmrJ on a new drive, but not the old data. You can retrieve the old data from the mangia server.

The archives on mangia go back to 2007-2008 and we plan to maintain them indefinitely--we never purposely delete archived data from mangia. A small number of spectra may be missing from mangia. For instance, if a hard drive fails in the middle of the day, spectra acquired since the last overnight backup won't get copied to the server.

We strongly recommend that you make regular and frequent backups of all your research data and have multiple copies that you can access yourself. You might, for example, copy all your data weekly to your laptop and have automatic cloud backup software running on it. In the event of a major earthquake, the mangia server could be offline for an extended period; you might also lose access to your computer. But if you can access your research data, you might be able to work on writing papers or your thesis if you don't have a lab to work in.

What is the best type of apodization to use for my spectra?

For 1D spectra, it is almost always advantageous to use some exponential multiplication, aka line broadening. This means you are multiplying the raw data (the free induction decay, or fid) by a decaying exponential function. Since the fid is a decaying exponential (or sum of multiple decaying exponentials) to begin with, this does not change its basic shape. However, it can meaningfully improve the signal to noise of the processed spectrum. The front of the fid is dominated by signal, whereas the tail should be mostly noise (assuming the acquisition time is well matched to the expected linewidth of the peaks in your spectrum). The decaying exponential will have little effect on the front of the fid, but will truncate the noise present in the tail of the fid. We normally specify the decay rate of the decaying exponential by the amount it will add to the apparent linewidth of the peaks in the transformed spectrum, thus its common name "line broading."

The most effective signal to noise reduction will be obtained if the line broadening applied is equal to the natural linewidth of the peak. Of course, in a multiline spectrum, each peak can have a unique width. An easy way to find out the average linewidth of your spectrum is to open the Processing -> Apodization window in MestReNova (you may find the shortcut "w" helpful). The average value is given at the top of the apodization window. For 1H decoupled carbon spectra, or other spectra with little in the way of J-coupled fine structure, setting the line broadening to this value has little down side. However, for proton NMR, this value may make J-coupled multiplets less well resolved, and it may be better to take less noise reduction so as to preserve the multiplet structure. In MestReNova, it is easy to watch what happens as you change the amount of line broadening if you have the "Interactive" button checked in the apodization window. With a 3 button wheel mouse, roll the wheel up and down to change the line broadening, and watch what happens. A value of 0.2-0.3 Hz is typical for 1H spectra. Sometimes, broad carbon peaks (for instance quaternary aromatic carbons alpha to a nitrogen atom) will be nearly inobservable with the default line broadening of 0.5-1 Hz, but become easy to see with 5-10 Hz.

For 2D spectra, there will be an apodization function for each axis. It matters a lot whether the 2D data has been acquired in phase sensitive or magnitude mode. In phase sensitive mode, the spectrum can and should be phased so that peaks are purely up or purely down. 2D fid's are normally acquired in a very truncated form compared with 1D spectra, to keep 2D data files from being unreasonably large. The apodization should be chosen to make sure the truncated fid is smoothly tailed down to zero at the end to avoid severe truncation artifacts ("sinc wiggles") in the transformed spectra. A cosine squared function (or sine squared with a 90 degree phase shift) is a safe choice which should give an artifact free transform. Some people may prefer to give a moderate amount of resolution enhancement to a phase sensitive 2D spectrum by use of some gaussian function, linear prediction, or both. It may be a good idea to compare the effects of those apodization schemes with the cosine squared function to make sure you are not overdoing it with resolution enhancement. For magnitude mode 2D spectra, phasing is not possible because the lineshape will have a mixed absorption/dispersion character. To work around the problems this would cause, a magnitude calculation is performed by squaring the spectrum and then taking the positive square root. This solves the phasing problem, but in turn leads to loss of resolution. To work around that, a strong resolution enhancing apodization function is used. A typical choice would be a sine bell (half sine wave) function with a zero degree phase shift. This function zeroes out the beginning and the end of the fid, but enhances the middle section.

Most 2D spectra are either phase sensitive in both dimensions or magnitude mode in both dimensions. An exception is an HMBC, which is advantageously acquired as magnitude mode in the 1H direction (because phase sensitive is not possible there) and phase sensitive in the X nucleus direction. This gives much sharper, better resolved peaks along the X direction. However, your apodization and processing parameters should match the way the spectrum was acquired.

How much solvent should I use to make up the sample?

Varian 5 mm probes will most reliably shim to spec with about 700 microliters of solvent. Bruker 5 mm probes may do so with somewhat less, 550 microliters or so. If you have a limited amount of compound to work with, adding this much solvent may produce a dilute solution. Using less than the recommended amount of solvent will produce a more concentrated solution, but one that will require more effort to shim, and will not shim as well. If possible, the best option for using a reduced amount of solvent is to use a Shigemi tube. These can be filled with 275-300 microliters of solvent and still give good resolution. Shigemi tubes can be purchased for water/D2O, chloroform, methanol, and DMSO. If you decide to use a smaller volume in a regular NMR tube, carefully center the liquid in the tube around the center of the NMR probe. The center line in the probe is marked in each sample depth gauge. Even if the sample is well centered, a short sample will mean some compromise on the peak width or shape in the spectrum, and the shorter the sample is, the worse the spectrum will look even when it is optimized.