Hi folks,
After many years I have spent studiously ignoring the problems detailed in
this email, some users in my new home have asked general
questions about chemical shift referencing which prompted me to revisit
some old ground and write detailed notes for our users. As part of this I
am revisiting the question of basic referencing of proton spectra in all the
usual solvents. This should be simple, but I can't reconcile various data,
so I wanted to explore the ammrl wisdom to find out
a) if I'm being stupid,
and b) what everybody does.
Be warned that this is a very long email...
As far as this discussion is concerned, I'm only interested in spectra acquired
locked, in normal tubes (no capillaries etc). I'm personally interested
in Bruker systems at this point but would welcome discussion of how things
are handled on other manufacturers systems (and in software other than TopSpin(tm)).
I will of course collate responses and report back!
I have not yet done measurements on solvents other than CDCl3, partly
because others may already have done so, so will collate also any comments
about other solvents.
The fact that the solvent shift is much less reproducible than that of TMS is
also noted, but that's a separate question. The J Org Chem paper indicated
below addresses this and makes a case for using TMS, but I think that's
unlikely to happen routinely in our case. A lot of our samples are likely to be
sufficiently dilute that this source of error is not likely to be disastrous, I feel.
This may however mean that I am chasing meaningless precision, but I think
it is correct to say that the current situation on Bruker systems at least is
"reproducibly wrong", and this ought to be fixable.
Firstly, how should things work for proton spectra in CDCl3?
According to the IUPAC guidelines
https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/mrc.2225),
we should reference proton spectra to the resonance frequency of dilute (1% or less)
TMS in CDCl3.
For proton spectra of samples in CDCl3 this is easily achieved by adding TMS to the sample and
somehow setting that peak position to 0ppm. On Bruker systems this is done
by using either the SREF command, which searches for a peak around
(in this case) 0ppm, and adjusts the axis until so that the peak is at exactly 0ppm;
or doing this manually with the "calib axis" button / .cal command.
In either case this sets the parameter SR to the difference in Hz between the
measured shift and the corrected shift. A positive SR value means that in
the unreferenced spectrum the reference signal had a positive shift relative
to the expected position.
As far as the default axis calculation is concerned, on a Bruker system the
actual "reference frequency", for proton is simply the "base frequency"
of the spectrometer (400.13MHz by default for a 400 for example, set during CF).
So for this to make sense, the magnetic field has to be set such that the protons
of dilute TMS would indeed resonate at e.g. 400.13MHz exactly.
This is usually achieved by the lock: the lock measures the 2H frequency
of the solvent , and adjusts the field so that it is at the "right value", defined by:
<base 2H frequency> + <lockshift> ppm
For this to work exactly, the 2H base frequency and the lock shift in ppm must
be set correctly. If it works exactly, then modulo the temperature and other
things dependence of the solvent chemical shift , the TMS signal will appear
at exactly 0ppm even without setting SR.
For nuclei other than protons, this referencing to the protons of TMS means using
a strictly defined reference frequency, derived from the resonance of protons in
dilute TMS in CDCL3 by
Vref(Nucleus) = Vref(1H,TMS) * 100 * Xi(Nucleus)
Where Xi(Nucleus) is defined in the IUPAC guidelines, and is essentially the
experimentally determined ratio of the resonance frequency of some
reference compound for that nucleus to the TMS resonance frequency ,
expressed as a percentage. For example for 13C, Xi(13C) is given by
100* V(13C, TMS) / V(1H, TMS).
Again on a Bruker system the actual reference frequency values for the various nuclei
are derived from the base frequency and a set of ratios, defaults defined in the file
<topspsin>/exp/stan/nmr/lists/nuclei.all
These should in principle correspond to the IUPAC Xi ratios for all nuclei.
The actual used ratios can be changed during CF.
The first problem:
In general, on a Bruker system with standard lock shift values, if you acquire
a spectrum of TMS in any solvent, locked, at 25C, and process the spectrum
with SR=0 , the TMS signal appears at some more or less reproducible shift
that is not exactly 0ppm, where for dilute samples the difference from zero is
noticeably more than the jitter in the measured value.
Over a few different instruments in our lab the position of the TMS peak in a
0.1% ethylbenzene sample in CDCl3, locked but with SR=0, is essentially
0.0255ppm +/- 0.001ppm. This corresponds to just under 10Hz at 400MHz,
and is comparable with the 0.02ppm that crops up in comments sometimes
(eg Hoffman paper) as the expected reproducibility of chemical shifts, so would
seem to be worth bothering about.
For comparison in another lab, in a recent study, J. Org. Chem. 2022, 87, 2, pp905-909
https://pubs.acs.org/doi/10.1021/acs.joc.1c02590 ,
samples with least interacting solutes gave approx. 0.023ppm for TMS with SR=0 -
slightly different to my values, but probably within the bounds of error due to lock phase,
temperature and so on. Spectra in that study were acquired at nominal 298K but
no details of temperature calibration given, it was not critical in their context.
In all above cases, after referencing to TMS=0, the residual CHCl3 signal is 7.260+/- 0.002 ppm.
This is in agreement with Roy Hoffman's paper, Magn. Reson. Chem. 2006; 44 pp606-616
https://analyticalsciencejournals.onlinelibrary.wiley.com/doi/10.1002/mrc.1801,
on which part of the IUPAC numbers are based.
This just means that the lock shift values are "wrong". The default Bruker
value for CDCl3 is 7.24ppm, but according to the above data on my systems
this is about 0.0255 ppm too high, so should be 7.2145ppm (this is indeed the peak
position in an unlocked 2H spectrum, referenced using the Xiref command to an
unlocked proton spectrum that has been referenced to TMS=0).
However, this 2H shift is *not* in agreement with Hoffman's paper: he has the 2H
shift of CDCl3 =7.290 at 25C, a difference of 0.0745ppm .
This difference is not explained by the (small) difference in the used Xi factors for 2H,
noted in Hoffman's paper.
The Bruker Xi value is 15.35060886 , whereas the IUPAC value is 15.350609, which is a
difference of 0.009ppm, almost an order of magnitude smaller than the difference between
my value for the shift and Hoffman's.
It's also not a typo, because fig 4 shows that the 2H shift of CDCl3 is always *higher*
than the 1H shift of CHCl3, in contrast to my measurement.
The Bruker Xi value looks like a value that has not been rounded to the 6 decimal
places recommended by IUPAC in the 2001 paper
https://stats.iupac.org/publications/pac/2001/pdf/7311x1795.pdf),
but unhelpfully the literature reference for the 2H Xi value indicated in the
IUPAC paper is "previously unpublished work" by Bernard Ancian so this is not
so easy to check.
Considering both the correct Xi value for 2H (you could modify this during CF)
and the correct 1H shift of TMS, one should have lockshift = 7.2055 for CDCl3.
As an aside, the 2H reference used to derive the Xi value is neat deuterated
TMS, rather than TMS in CDCl3, which I guess explains why the 1H and 2H shifts of
solvents are not the same despite the expected absence of a primary 1H-2H
isotope shift.
So, it seems to me that the lock shift for CDCl3 should really be 7.2145, (if using
the default Bruker 2H Xi value), or 7.2055 (if using the exact IUPAC Xi value, which
you could set that during CF).
But that's not consistent with the measured CDCl3 chemical shift in Hoffman's paper.
I suspect an error or misunderstanding on my part, but I've gone through this a number
of times , and verified that if I set the lock shift to 7.2145 the unreferenced shift
of residual
CHCl3 is really closer to 7.261ppm, something like an order of magnitude closer
than with lock shift of 7.24ppm.
I would be happy with this were it not for the rather different published 2H shifts,
and the fact that the Bruker Xi value is not exactly consistent with IUPAC
numbers (although possibly within experimental error).
Secondly, how do we deal with solvents other than CDCl3?
IUPAC recommendation is that TMS is CDCl3 is the one true reference.
As shown in Hoffman's paper, if you can account for magnetic susceptibility
effects (e.g. by MAS), you can really use this as a reference for samples in
other solvents. If you do that, you find the proton shift of TMS in other solvents
is in general somewhat nonzero (e.g. -0.16ppm for acetone, which is a very large
difference compared with other sources of error).
The Bruker lockshift values give, for locked spectra, a reasonable
(but like CDCl3, not exact) approximation of the shifts given in Hoffman's
paper adjusted to be relative to TMS=0 (as expected, because until relatively
recently the deviations of the TMS shifts from zero had not been accurately
determined so referencing was usually to "TMS in current solvent =0").
So, does everyone just ignore the fact that referencing should in theory be to
*TMS in CDCl3=0ppm*, and accept referencing to *TMS in the current solvent=0ppm*?
Thirdly, how do we deal with other nuclei?
IUPAC recommendation is to use Xi factors only, here. In general in CDCl3 this is easy:
having set the correct SR value in a proton spectrum to give TMS=0, the
SR value for any other nucleus X is just SR(proton) * Xi(X)/Xi(proton).
Setting this can be done using the Bruker a program xiref for example.
If the lockshift is set correctly, the referencing step is not needed assuming
reproducibility of the CDCl3 shift.
However, there are some anomalies here. The ratio of the proton and
carbon frequencies of TMS, in DMSO, is different to the ratio in CDCl3.
So even with the lock shift set so that the proton shift of TMS =0 in a
DMSO sample, the carbon shift of TMS will be 0.5ppm in an unreferenced
spectrum. What to do about that?
In old topspin versions, the "reference" parameter in edlock was set to 0ppm, so on
finding TMS at +0.5 the SREF command would shift the spectrum by -0.5ppm; the
"reference shift" parameter in edlock was set to -0.5, so the SREF command
would shift the spectrum by -0.5ppm also If it did not find the reference peak,
essentially the same value as if it did find a TMS peak.
Modern topspin however handles this differently - if no reference peak is found
SR is left at 0, which is rather different...
This leaves a number of questions:
Do you just accept the current Bruker default lock shift values?
If so, do you use some default SR value that is not zero? (we seem in my current
home to have a situation where SR is set in proton parameter sets to be the right
value for CDCl3, for example, and in general don't use SREF). Is the observed
difference so small you feel it's reasonable to just ignore it?
Do you routinely use TMS, or not?
For solvents other than CDCl3 do you reference effectively to TMS(current solvent)=0?
Has anyone done similar solvent shift measurements relative to TMS in the sample for other solvents?
Where have I gone wrong in understanding the 2H shifts of CDCL3?
How do you handle 13C spectra (particularly in DMSO)?
Do you use SREF at all ?
How is this handled in other software ? Does e.g. mNova have any form of automatic
referencing?
And possibly most importantly What do you tell your users to write in papers?
Any thoughts about any of this gratefully received!
Kind regards
Pete
--
Peter Gierth
Senior NMR Specialist
Yusuf Hamied Department of Chemistry
University of Cambridge
https://www.ch.cam.ac.uk/analytical/nmr
Received on Tue Apr 05 2022 - 10:24:16 MST