NMR Sensitivity Enhancement

NMR spectroscopy, although incredibly powerful, is an inherently insensitive analytical technique. That this is so can be seen from the difference in the populations of the |a> and |b> spin states at room temperature ... works out to about 1 part in 105, depending on field strength. The bulk magnetisation is proportional to this population difference. This is a very small number to be dealing with ... the raw nmr signal is very weak indeed. Anything that we can do to increase the signal strength will be of value to us. Thus, almost all technical advances in nmr spectroscopy have been (and still are) linked with increases in sensitiviy as measured by the signal-to-noise (SNR) ratio.

The first major advance was the move from cw nmr spectrometers to ft nmr spectrometers. In a cw spectrometer, one sweeps the field or the transmitter frequency through the resonances. In order to get decent resolution the sweep rate had to be rather slow .. on the order of 1 Hz/s. So, for a 1000 Hz wide spectrum a single scan would take ~1000 sec. or approximately 15 minutes! The ft spectrometer could accomplish this in 1 second! Along with this came signal averaging which basically means that one adds the results of a second and subsequent scans to the first. Actually, it's misleading to say scans with respect to ft nmr ... much better to say acquisitions. When this addition process is performed the magitude of the nmr signal grows linearly but the magnitude of the noise grows as &radic(2). Thus the SNR increases with more scans. In order to double the SNR the number of acquisitions must be quadrupled. So, with ft nmr, we can get a comparable spectrum to that of a cw instrument in 1/1000 of the time or, assuming a acquisition+delay time of 2 sec., get a spectrum that has a SNR 32 times better than that obtainable by cw nmr methods!

So ... sensitivity problem solved, right? We can just acquire data until we have the desired SNR and stop our experiment. Well, up to a point this is true but, as I said above, to double the signal to noise ratio you must quadruple the number of acquisitions. This is ok if you want to double, quadruple or perhaps get eight times the signal to noise. How many more scans would you need to get 16 times the SNR? It will take 162 times the number of scans or 256. You can see that it begins to get rapidly out of hand. SNR is not usually a problem for 1H nmr since it is a sensitive nucleus and of high abundance. It is a problem for 13C nmr spectroscopy however. 13C is not a sensitive nucleus and has a % abundance of 1.1%. Getting a decent 13C spectrum is sometimes a big challenge ... especially when relaxation issues play a role. Let me put this another way to try to get it into perspective. Let's suppose that we acquire a usable 13C nmr spectrum of 100 mg of a sample dissolved in our solvent of choice in 1 scan. To get the same signal-to-noise with 50 mg would require 4x1=4 scans. If we were to use only 10 mg it would take 100 scans. 1 mg would need 10,000 scans! You can see that, for 13C nmr at least, it is to your advantage to have as much as possible solute in solution to maximize the SNR in the time available to you. Many new users make the mistake of thinking that doubling the number of scans will result twice the SNR. Not so.

Another increase in sensitivity was the use of quadrature detection. Prior to this the transmitter frequency was set at one side of the spectrum and (hopefully) all of the peaks in the spectrum appeared on one side of the transmitter frequency. If any peaks appeared on the wrong side of the transmitter frequency they would still be digitized but would be 'folded' or aliased into the resulting spectrum. Even if all of the peaks were on the correct side there would always be noise on the wrong side and, after all, noise is just random signals of varying frequencies and they would be aliased into the spectrum. Thus, at least a full spectral width's worth of noise was being added to the spectrum. Using quadrature detection and the appropriate analog filters it was possible to cut the amount of aliased noise in the spectrum.

Other SNR enhancements include improvements in magnet and probe design to provide for better, more homogeneous fields and improved signals from the coil material. Higher field strength means a larger population difference between |a> and |b> energy levels which means larger signals.

Pulse sequences have also advanced with sensitivity enhancement in mind. INEPT and related pulse sequences allow for enhanced sensitivity of a relatively insensitive nucleus by the use of polarisation transfer.

For two dimensional heteronuclear correlation spectroscopy in which one nucleus is proton, sensitivity is enhanced through the use of an 'inverse' probe.

The most current advance in SNR enhancement is the cryoprobe. This is a special probe in which the receiver coils are cooled to 77K thereby reducing the thermal noise. This is a major source of noise introduced into the system prior to preamplification. SNR is enhanced 10 fold or more.

Thus, the lesson is that almost all major advances in nmr spectroscopy are advances in signal-to-noise ratio.