NMR Facility - Chemistry Department

The Point:

A new parameter set with a custom pulse sequence has been implemented on 400-1 that yields 19F spectra without the usual broad signal from the Teflon in the probe. This improves automatic phasing, provides flatter baselines, and sets the stage for quantitative 19F spectroscopy.

Fig 1: New vs Old 19F Parameters

Fig 1: New 19F parameters eliminate background, flatten baseline


The next time you need 19F NMR on 400-1, choose the simply-named “FLUORINE” experiment. A new pulse sequence is implemented that includes a simple spin echo between 90° excitation and acquisition. The duration of the echo has been tuned so the rapidly-decaying signals from teflon have completely relaxed, but the signals of interest remain virtually unaffected. This pulse sequence is part of the FLUORINE parameter set, and you do not need to make any adjustments to make it work. Furthermore, this experiment includes 1H decoupling to ensure sharp lines. D1 and AQ values have been set to enable full relaxation, based on T1 relaxation times measured for a small set of representative molecules.


The Problem: Strong Background Signal

If you’ve taken a 19F spectrum before, especially if your molecule of interest was dilute, you’ll have noticed that your background is not flat and/or difficult to phase well. There is a very broad two-humped signal spread out between -150 and -225 PPM, and it comes from 19F in the teflon components of the NMR probe. You can tell it comes from outside the sample by acquiring a 19F spectrum with no sample present:

Fig 2: 19F Background (No sample signal)

Fig 2: 19F Background (Signal comes from teflon in probe)

One could simply ignore or live with this background, but that leads to some other problems. Automatic phasing, for instance, doesn’t work quite right when this signal is present (especially using Topspin’s “apk” autophasing):

Fig 3: Teflon background ruins "apk" autophasing

Fig 3: Teflon background ruins “apk” autophasing

Further, if one’s 19F signals are near the teflon bumps, it is difficult or impossible to obtain quantitative integrals. Because 1H-decoupled 19F integrals are usually sharp singlets with high intensity, they’re great for quantitation, which is really useful when assessing impurity levels or isomer ratios. A properly-administered NMR spectrometer can even be calibrated to provide molar concentrations using 19F NMR signals … but only if the baseline is flat.

Broad Signal

There are several NMR techniques for eliminating or reducing signals. Most, like solvent suppression, focus on obliterating signals from a very narrow (~50 Hz) frequency range. That doesn’t help if your sharp signal falls on top of a very broad unwanted signal. But we CAN exploit that fact that major difference in peak sharpness, effectively eliminating all signals that are broad.

The reason teflon 19F signals are so broad is because of their very short T2 relaxation time. One can see just how quickly they go away by inspecting the beginning of a 19F FID:

Fig 4: 19F FID shows fast decay of teflon signal

Fig 4: 19F FID shows fast decay of teflon signal

You can see the strong first few points (after the “ramp-up” period for digital filtering – another topic altogether), which correspond to the very large teflon signal, are much more intense than the rest of the FID. But they drop off very rapidly, due to T2 spin-spin relaxation and chemical shielding anisotropies intrinsic to their solid matrix (the overall apparent T2 relaxation, which includes these other effects, is referred to as T2*). In contrast, nice sharp signals are quite long-lived, with T2* values approaching their T1 values. (Which, you recall from the posts about optimizing 1H and 13C parameters, typically range from 0.2 to 45 seconds for 1H and 13C signals in typical organic molecules with MW < 1000.)

Note that you can estimate a signal’s T2 value from its linewidth using the magic of the Fourier Transform:

Fig 5: Linewidth and T2 rate are inversely related

Fig 5: Linewidth and T2 rate are inversely related

Wouldn’t it be better to simply wait to start acquiring the FID until after all that nasty teflon signal had died off? It turns out there’s a problem with that: the resulting peaks would all be very out of phase, requiring huge phase corrections, which probably could not be achieved manually. Here’s an example of a pair of 13C spectra, one acquired normally and one acquired with a 2.0 millisecond delay between excitation and acquisition. (13C is used here because it has a large spectral width, like 19F, but has more peaks, which is more useful for showing phasing problems acros the whole spectrum).

Fig 6: Adding delay between pulse and acquisition ruins phasing

Fig 6: Adding delay between pulse and acquisition ruins phasing

You can see that this approach may remove broad signals, but introduces another difficult problem.

Solution: The Spin Echo

Instead of just beginning acquisition of the FID after that nice “filtering” delay period, how about refocusing the spins so they don’t need phase correction? That’s achieved handily using one of the simplest pulse sequence elements, the spin echo.

Fig 7: The Spin Echo

Fig 7: The Spin Echo

In a spin echo, signals that have been given 90° excitation are in the XY plane, aligned along, say the X axis. As time proceeds, they rotate in the XY plane because of their different chemical shifts, some proceeding slower and some proceeding faster. The “fast” and “slow” spins gradually move apart. After a period of evolution, a 180° pulse is applied that flips them to the opposite axis. Once there, both “fast” and “slow” spins continue moving in the same rotational direction as they had been. But because they’ve been flipped, this time the fast and slow spins move closer together. After a certain time (equal to the amount of time between their original 90° excitation and the 180° pulse), they converge and are once again all aligned, but on a different axis. (See the Wikipedia entry for an illustrative animation: https://en.wikipedia.org/wiki/Spin_echo. The name “echo” refers to sound made at one instant that bounces off a hard surface and returns shortly afterwards; listen to an example here: https://youtu.be/5OoaHjiZ91c )

If acquisition begins at this instant, it is almost as if the delay-180°-delay period had never happened – the spins are simply aligned in the XY plane, acquisition proceeds, and a normal FID results.

This treatment, however, neglects the effects of relaxation. This is no big deal if relaxation is occurring on a timescale much longer than that of the spin echo. But if relaxation is happening more quickly than the echo, the spins cannot fully refocus, and signal intensity is reduced. This is a problem if we’re interested in keeping spins intense through an echo, but it’s great if we want to eliminate them. To obtain an FID in which the fast-relaxing signals (like 19F in Teflon) are eliminated, but long-relaxing sample signals are retained, we just need to include a spin echo and set its delays so they are longer than the unwanted signals’ T2 values, and much shorter than the sample’s T2 values.

Application to 19F

If we quantitatively examine the 19F FID above, we see that the large signal from background 19F decays within the the first millisecond. If we include a spin echo with delays of 1.0 milliseconds, the resulting FID should therefore not include a significant amount of background 19F signal.

Examination of typical 19F signals from organic molecules (processed without line line broadening) shows linewidths of ~1.0 Hz, much like 1H signals. We can use that number to estimate the typical T2 value to be about 1/(pi*1.0 Hz) = ~318 milliseconds. This is very long compared to a 2.0 millisecond spin echo.

If we apply the pulse sequence 90°-(1.0 msec delay)-180°-(1.0 msec delay)-acquire to the following simple dinitro derivative (courtesy of Vladislav Lisnyak), we effectively eliminate the background and retain full intensity:

Fig. 8: New pulse sequence yields background-free, easily-phased spectrum with flat baseline

Fig. 8: New pulse sequence yields background-free, easily-phased spectrum with flat baseline


Optimizing D1 and AQ

To set good default values for D1 and AQ that provide the most signal in the least amount of time and provide quantitative integrals, we need to run some more tests.

First, we need to assume we’ll be performing 90° excitation, which practically eliminates use of the Ernst Angle equation, which we applied in the 1H and 13C parameter optimizations. We cannot use 30° excitation like in those cases because application of the 180° pulse in the echo would place the spins a position not suitable for good acquisition. As with other experiments requiring practically full relaxation, we’ll thus aim for a D1+AQ experiment time of 4-5*T1.

Second, we need to determine typical T1 values for 19F signals of interest. Sampling a small number of fluorinated molecules, it was determined they ranged from approximately 1.0 to 4.5 seconds. Therefore, setting D1+AQ to 10 seconds should result in reasonable quantitative spectra without employing painfully long experiments.

Third, we need to set AQ. When setting AQ for 1H, our primary consideration was achieving sharp linewidth, which requires longer AQ time. Unweighted 19F signals in our sample molecules were approximately 1.0 Hz, so to ensure we get good sharpness naturally, we should use AQ of at least 1.0 seconds. To be sure, we can safely use AQ=3.0 sec without much penalty, like we do with 1H. Longer acquisitions may heat the sample a little because of the 1H decoupling being applied during acquisition. We can thus set D1=7.0 sec.

Note that, unlike 13C NMR, in which we apply low-power 1H pulses to increase the 13C signal via the 1H-13C NOE, we leave it off for 19F. The 1H-19F NOE is relatively small – on the order of the 1H-1H NOE, so it does not offer much benefit. Further, the 19F signal, like that of 1H, is intrinsically strong, and 19F is 100% naturally abundant, so does not need enhancement.

For default parameters, NS=16 in the FLUORINE parameter set to provide good S/N and artifact averaging. Because D1+AQ=10.0 sec, the default FLUORINE experiment takes a little less than 3 minutes for acquisition.

Enjoy the new FLUORINE parameters! They’ll soon be copied to 500-2 and 500-3.