I received many useful replies to my questions about getting DOSY spectra
to look the way one might think they should, e.g. with all peaks known to
belong to a single compound appearing in the same row of the DOSY.
Several people asked for summaries. Here are the suggestions I found most
helpful.
The possibly imperfect DOSY plot is the result of subjecting your possibly
imperfect diffusion data to an inverse Laplace transform, which is
inherently not as stable, say, as the Fourier transform is at displaying
signal frequencies. My main interest was in trying to make my input data
less imperfect. A data set where the spectrum phase does not hold
constant as the gradient amplitude is ramped up is not going to give a
textbook DOSY--the phase instability is taken as a sign of convection in
the tube.
The biggest problem in our experimental setup was temperature gradients
within the sample. I believe I was doing a good deal to create these
gradients by using the FTS cooling bath to chill the VT air on our system
and then reheat to achieve regulation at 25 C. The instrument is a 600
with a protein type triple resonance HCN triax gradient probe. Running
the cooling bath all the time is fairly common, I think, on protein
instruments. For low viscosity organic solvents, this proved to be highly
counterproductive for diffusion experiments, as a pretty small temperature
gradient is enough to sustain convection. This down side may be less, or
not, apparent with aqueous protein samples, as water is much more viscous
than e.g. THF or acetonitrile. But I always took it on faith that I was
doing a good thing by using precooled air in protein experiments--now I'm
not so sure. It would seem to me that any convection would lead to signal
losses in pulse sequences with many, many gradient pulses. What I can say
for sure is that the bigger the amount of precooling (we were normally
using 15 C), and the lower the airflow, the worse the temperature
gradients become. When it was especially bad, obvious problems appear
with gradient shimming--the profiles develop large gaps in them. But
phase instability in the diffusion spectra was evident when there were no
other obvious problems with other gradient-based experiments.
I received a number of suggestions about ways to reduce the temperature
gradients within the sample, the simplest of which was just to turn off
active temperature regulation. This amazingly improved the phaseability
of the 1D spectra; I have to think the data would better define the
diffusion coefficient. The processed DOSY data did not look that much
better, however, in terms of alignment of peaks from the same molecule.
Additionally using the convection compensated pulse sequence with no
active temperature regulaion cost about half the signal overall, but gave
a significantly better DOSY. It was not clear what about the convection
compensated 1D data going in to the DOSY was better--this data also phased
well, but no better, as the data taken without convection compensation
already had a completely stable spectrum phase and flat baseline.
Other suggestions for improved temperature homogeneity included using a 3
mm probe, putting a 3 mm tube into a 5 mm probe, and using a slotted
spinner turbine which allows VT air to run completely past the sample tube
and exit out the top.
One respondent said he had been told by a vendor that the linear region of
z gradient in an xyz-gradient probe is indeed shorter than a z-only
gradient probe--something on the order of only a centimeter. He got
better results by using a Shigemi tube to put sample only in the center of
the gradient coil.
Other replies focused on postacquisition processing, including very
meticulous phase and baseline corrections to the 1D data, using the
"fiddle" function to correct for any non-Lorentzian character of the
lineshape, and using absolute value instead of phase sensitive spectra.
In high S/N spectra, baseline problems are the largest source of error,
either from the baseline not being flat, or a small peak on the tail of a
large one.
David VanderVelde
Manager, High Resolution NMR Facility, Caltech
davidv_at_caltech.edu
Received on Fri Sep 11 2009 - 14:06:46 MST