Chemical shift referencing your NMR spectra

There are various ways of referencing NMR spectra; however, they all have potential problems with either systematic errors or practicality.   In general, there are three methods in wide use:
The main drawbacks to using internal standards in general are that their shifts change as they are diluted into different solvents, and removing the standard may be a problem if you want to recover the sample in pure form.  Although we normally think of TMS as an inert molecule, its 1H chemical shift moves more than 0.6 ppm as it is added to common organic solvents, relative to neat TMS (e.g. benzene +0.30 ppm, chloroform -0.14 ppm).  Some specific primary chemical shift standards are chemically reactive (such as phosphoric acid) and either inconvenient (such as liquid ammonia) or very dangerous (such as methylmercury) to handle.  Phosphoric acid and ammonia are also good examples of compounds whose chemical shifts are so dependent on pH and concentration that they are useless as dilute internal standards.  In proton NMR of aqueous solutions, DSS and TSP are both susceptible to changes in their shift at low sample pH, when they become protonated.  They are also capable of intermolecular interactions with other solute molecules.  For example, they form host-guest complexes with cyclodextrins, which changes the shift of the reference peak, and may disrupt other host-guest equilibria in the sample.

Shift changes with dilution can be avoided by keeping the standard as a neat liquid, unmixed with the sample to be referenced.  However, switching from an analyte solution to a separate NMR tube containing the reference is frequently quite inaccurate because different samples have different bulk magnetic susceptibilities; if this variation is not corrected, it may introduce an error of up to several ppm.  External referencing of 19F and 31P to neat CFCl3 and 85% phosphoric acid (this is the strength of the acid straight out of the reagent bottle--the balance is water, and it must be used without further dilution), respectively,  suffer from this problem.   These solutions necessarily contain nothing deuterated to lock on, so there is no way to compensate for the bulk susceptibility effects when you switch samples and run the standard unlocked.  The only accurate way to use a neat reference solution is to put it in a concentric insert in your NMR tube, the type used for determination of paramagnetic susceptibilities by the Evans method.  The concentricity is important for good lineshape and best accuracy--inserting e.g. a sealed melting point capillary into the sample, with nothing to hold it in the center of the outer tube, frequently will not give very good resolution.  (As an aside, the Evans method would not work correctly either if the two solutions were run in separate NMR tubes at different times.)

Some time ago, the concept of expressing NMR frequencies in Xi (as in the Greek capital letter) units was introduced in multinuclear NMR.  This is the calculated frequency of a signal where the 1H frequency of TMS is exactly 100 MHz.  These numbers are calculated by measuring the exact frequency of the peak of interest, and the exact frequency of the 1H TMS peak, on the same sample at the same time with the spectrometer locked; then determining the ratio of the frequencies, and multiplying by 100 million.  To translate that number into the expected frequency of that peak on your spectrometer, you find your 1H TMS frequency, divide by 100 million, and multiply the reported Xi by that number.  NMR frequencies can be measured with very high accuracy, better than 0.1 Hz; the ratios of these numbers are equally precise, and they are instrument independent.   In this way,  the chemical shift scales of all nuclei can be tied together with ratios. If we have one careful measurement of the ratio of the 1H TMS frequency to the 31P frequency of phosphoric acid in a coaxial insert, or the 15N frequency of liquid ammonia, and we know our 1H TMS frequency, then we don't need the actual multinuclear external standard to produce correctly referenced multinuclear spectra.  The IUPAC has adopted this method of chemical shift referencing and published tables of Xi values for the NMR active isotopes.  Varian's implementation of the IUPAC values is in the file /vnmr/nuctabref.  For some nuclei, more than one chemical shift reference has been used over time.   In those cases, the one used by VnmrJ will be the first uncommented line for a given nucleus.  For example, nitrogen chemical shifts have been reported relative to neat liquid ammonia, neat nitromethane, and tetramethylammonium iodide in DMSO. Biological NMR spectroscopists report nitrogen shifts on the ammonia scale, while organic chemists tend to use the nitromethane scale.  These scales differ by 380.2 ppm. Although they are less different, bio-NMR spectroscopists report 13C shifts relative to DSS in D2O, which is offset from the 1% TMS in CDCl3 scale used by organic chemists by a few hundredths of a ppm.

A ratio scale can be converted to an absolute scale if we know the frequency of any one point on the scale in absolute numbers.  In practice, what both Varian and Bruker do is to use the known absolute deuterium lock frequency, rather than the 1H TMS frequency, as the basis for all multinuclear chemical shift calibrations.  This has the advantage that you do not have to have TMS present in every sample in order to establish chemical shifts.  Both vendors' software can be optionally configured to check for TMS in the spectrum and set the shift to exactly zero if the peak is found, or make no changes if it is not.  This system can work extremely well, although there are various potential problems to be aware of.
My personal preference is to use the solvent based referencing scheme for all nuclei, unless I know the prerequisites for having it work are not met, such as running unlocked, or with a nonstandard solvent.

Dave 7/23/09