Technical Notes

Electronic Referencing Techniques forQuantitative NMR: Pitfalls and How To Avoid ThemUsing Amplitude-Corrected Referencing throughSignal Injection

Knut Mehr, Boban John, David Russell, and Daina Avizonis*

Varian MR Systems, 3120 Hansen Way M/S D-298 Palo Alto California, California 94304

NMR spectroscopy can be a superior analytical tech-nique for quantification of compounds dissolved insolution. Traditionally a chemical reference standardof known concentration is added to the sample. Theconcentration of the solute can then be determined bycomparing the signal integrals. However, it can beinconvenient or impossible to use internal references.Electronic referencing was developed to circumventproblems with internal standards and has been usedsuccessfully in well-controlled situations. However, itis not always possible or convenient to have sampleswhere the dielectric sample properties do not changefrom one to the next. We propose a modification of theold electronic referencing technique that takes intoaccount the electronic changes between dissimilarsamples. We have called this new technique Amplitude-corrected Referencing Through Signal Injection orARTSI.

NMR is a nondiscriminate linear detector. This means that ifa solution of a compound is concentrated enough and has an NMR“active” nucleus such as 1H or 19F the spectrometer will detect it.The intensity integral of the resonance detected is directlyproportional to its concentration in solution. Very few analyticaltechniques have such a clean and direct path to quantification.Today’s modern NMR spectrometers have improved in sensitivity,making quantitation practical, but they also have very stabletransmitters and receivers that remove the necessity of requiringan internal standard of known concentration. This is especiallyimportant where it is impossible or inconvenient to add an internalconcentration standard to the sample. One may measure astandard of known concentration and use that NMR spectrum toset integral values. These values may be transferred, with a fewmodifications, to another sample of unknown solute concentrationand used to quantify the concentration of the unknown solute.One can imagine a few clever macros or software programs thata chemist could execute. However, even for an experienced

spectroscopist this is an inconvenient and cumbersome solution.The fear is 2-fold. First, if the quantitation sample integral valuesare overwritten or lost then the calibration is lost and the chemistwould need to return to the calibrant spectrum to retrieve andcorrect the integral values. The standard calibration integrals arein a spectrum or data set separate from the unknown’s data set.Second, the calibrant integrals may potentially be mis-set uponretrieval of the reference integral intensities. In the end having apeak, albeit an electronically synthetized reference peak, storedalong with the unknown’s NMR data set leaves much less roomfor error and ambiguity. The calibrated internal synthetic referencesignal represents a calibrated concentration that can be easilyintegrated and used. It is something that is easily used by allchemists, and because it is part of the free induction decay of theunknown sample’s data set, it cannot be easily lost or confusedwith other standard data sets.

Approximately thirteen years ago, Serge Akoka patented anelectronic referencing technique Electronic Referencing To accessIn vivo Concentrations” or ERETIC1 for imaging. This techniquewas later modified to work for high resolution spectrometers.2

The ERETIC technique has been used for a number of applicationsranging from high-resolution NMR, solids NMR, to MRSimaging.3-6 For high-resolution NMR, a radio frequency signalat the observed frequency, typically proton, is electronicallygenerated via a very low power shaped pulse on a decoupler orsecond channel. The signal is introduced directly into the probeon a spare channel, typically the X-channel, during acquisition.The X-channel is not necessarily tuned to the operating frequency.During acquisition, the proton coil may then detect the syntheticsignal emanating, in theory, from the X-coil or channel. Since theERETIC signal is detected on the same coil as the NMR signal,it has been assumed that it too experiences the changes in the

* To whom correspondence should be addressed. E-mail:daina.avizonis@varianinc.com.

(1) Barantin, L.; Akoka, S.; LePape, A., CNRS, F. P., Ed. France, 1995.(2) Akoka, S.; Barantin, L.; Trierweiler, M. Anal. Chem. 1999, 71, 2554–2557.(3) Burton, I. W.; Quilliam, M. A.; Walter, J. A. Anal. Chem. 2005, 77, 3123–

3131.(4) Dalvit, C. Prog. Nucl. Magn. Reson. Spectrosc. 2007, 51, 243–271.(5) Franconi, F.; Chapon, C.; Lemaire, L.; Lehmann, V.; Barantin, L.; Akoka,

S. Magnetic Resonance Imaging 2002, 20, 587–592.(6) Ziarelli, F.; Caldarelli, S. Solid State Nucl. Magn. Reson. 2006, 29, 214–

218.

Anal. Chem. 2008, 80, 8320–8323

10.1021/ac800865c CCC: $40.75 2008 American Chemical Society8320 Analytical Chemistry, Vol. 80, No. 21, November 1, 2008Published on Web 10/10/2008

quality factor of the probe caused by the insertion of a sampleinto the detection coil.2 The ERETIC signal integral is calibratedagainst a reference standard sample. The same signal is thengenerated in the sample of unknown concentration and used as aconcentration reference integral there by allowing the chemist tocalculate the concentration of the unknown sample.

Since this technique was adapted to high-resolution NMR,spectroscopists have encountered inconsistencies when applyingit to a wider range of applications. Variations in quantitativeperformance appear to be associated with two different experi-mental aspects: probes and samples. From anecdotal evidence andour own experience, some probes behave more consistently thanothers. Likewise, samples that have dielectric properties verysimilar to or the same as the concentration reference standarddo well with the ERETIC technique while those in differentsolvents or salt compositions can give large errors in quantitation.In short, it is important to recognize the limitations of a very usefultechnique. In this technical note, we explain and solve theseshortcomings by taking advantage of stable and flexible instru-mentation and using the basic reciprocity principle 7,8 to overcomethe most serious limitations of earlier electronic referencingschemes. We call this new technique “Amplitude-correctedReferencing Through Signal Injection” or ARTSI.

EXPERIMENTAL SECTIONTo challenge the quality factor of the NMR probe, a series of

2 mM sucrose, 0.24 mM 2,2-dimethyl-2-silapentane-5-sulfonic acid(DSS), and 0-250 mM sodium chloride solutions were preparedfrom a 200 mM sucrose and 24 mM DSS stock solution, and 2.5M sodium chloride stock D2O solutions. The final volume of thesamples was 700 µL in D2O. The NMR data were collected on a600-MHz Varian NMR System (installed 2006) equipped a tripleresonance probe (HCN) and VnmrJ Software (rev 2.2C). For eachsucrose/DSS sample, the probe was tuned and matched followedby gradient shimming and a 90° pulse width calibration. The 90°pulse was manually calculated from 360° and 720° tip anglecalibrations on the residual HOD and DSS signals. Each samplewas inserted into the magnet, tuned, and calibrated at least sixseparate times. In general, the 90° pulses calculated from the 360°and the 720° pulse calibrations were within 0.1 µs of each other.This indicates a calibration accuracy of ∼1-3% for the given pulsepower. Spectra with electronic referencing were collected usingfour scans, a calibrated 90° pulse for each sample, an acquisitiontime of 5 s, and interpulse delay of 25 s with a low-powerpresaturation of 5 s (35 s for each scan). Data were processedusing a 0.5-Hz exponential line broadening and zero-filled to 524288 data points. Integral resets were set according to Weiss.9 Datawere acquired 10 times for each sample using both ERETIC andARTSI techniques.

The electronic reference signal injection required some minorsystem recabling as described by Akoka et al.2 or using animproved cabling scheme as shown in Figure 1. It was necessaryto split the synthesizer input from the proton channel to thereference channel to maintain proper phase of the reference signal.The standard presaturation experiment was modified to generate

the electronic reference signal during acquisition. The electronicreference signal was programmed as an exponentially decayingshaped pulse at the proton frequency using pbox.10 The shapedpulse was executed on the second or third channel (dependingon how the system was cabled) during the acquisition time. Theshaped pulse is automatically generated by the pulse sequencesuch that the operator can set acquisition time, offset, line width,phase, and intensity of the reference signal. The same pulsesequence was used for both ERETIC and ARTSI style dataacquisition.

In the “proof-of-concept” study presented here, 2 mM sucrosewith 0.24 mM DSS in D2O was used as the concentration referencestandard. The fine power used for the synthetic signal wasadjusted until a reasonable integration value was achievedcompared to the anomeric proton of sucrose. For the ARTSItechnique, the reference signal power and reference sample’s 90°pulse width were also recorded (and maintained in the probe file).For the ARTSI technique, the intensity (fine power) of theelectronic signal was adjusted based on the 90° pulse width ofthe sample in the magnet compared to that of the referencesample’s 90° pulse width as described Results and Discussion.

RESULTS AND DISCUSSIONIn order to test how well the ERETIC technique works when

the quality factor of the probe is changed by samples withincreasing salt concentration, we acquired data using the ERETICtechnique and analyzed it for six samples of 2 mM sucrosesamples with sodium chloride concentrations ranging from no saltto levels above 250 mM. Each sample was carefully tuned,shimmed, and its 90° pulse calibrated. The data were analyzedaccording to Akoka et al.2 The results are summarized in Figure2 and Table 1. As the salt concentration increases, the qualityfactor of the probe is decreased. From the data analyzed, it isclear that the ERETIC signal intensity did not accurately track(7) Hoult, D. I. Concepts in Magnetic Resonance 2000, 12, 173–187.

(8) Hoult, D. I.; Richards, R. E. J. Magn. Reson. 1976, 24 (1969), 71–85.(9) Weiss, G. H.; Ferretti, J. A. J. Magn. Reson. 1983, 55 (1969), 397–407. (10) Varian, 2.2C ed.; Varian Inc.: Palo Alto, 2008 Varian NMR Software.

Figure 1. Schematic drawing of how one may recable a spectrom-eter so that the reference signal is coupled into the receive pathbetween the probe and rf filter/preamplifier using a low-loss directionalcouplers. A splitter may be necessary for the synthesizer output fromthe observe channel to the reference generating channel to avoidreference signal phase changes. Abbreviations: TX, transmitter, RX,receiver, T/R, transmit/receive switch; CH1, spectrometer channelone, CH2, or CH3; spectrometer channel two or three, triangularsymbols represent amplifiers.

8321Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

the changes in probe efficiency due to the changing dielectricproperties of the samples in the probe. This is due to the factthat the intensity of the ERETIC signal, rather than being afunction of only the inductive coupling between the coils and thusmodulated only by the Q of the receive coil, is typically modulatedby a variety of coupling factors. Those commonly include bothinductive and capacitive coupling mechanisms between the tune/match networks of both channels as well as between the NMRcoils, but also by the return loss of the probe channel used tocouple in the ERETIC signal. All these coupling factors combinedare accurately described by the transmission factor, often referredto as isolation, between the observe port and the port used tocouple in the ERETIC signal. While this transmission factor isalso modulated as a function of changes in the observe coil Qand the associated tuning and matching changes, it is not normallymodulated proportional to the intensity changes of the observesignal. As a general guideline, as soon as changes betweensamples require an adjustment of the tune and matching controls,the quantitative accuracy of the ERETIC method is beingcompromised.

For these reasons, it is best not to inject a reference signaldirectly into the probe on any channel. The ARTSI technique,presented here, routes a reference signal through the full receivepath of the spectrometer (Figure 1) using a directional couplerwith very low insertion loss. This has the advantage that the pathof the reference signal is well defined and controlled as well as itserves to test the electronics of the receive path for any degrada-tion over time. The electronic reference signal intensity is adjustedthe same as one would adjust it for the ERETIC technique excepta reference pulse width and power are also recorded (seeSupporting Information Figure 1 for example). Since each samplemay change the receptivity of the probe coil and change thesensitivity of NMR signal received, the electronic signal intensity

must be scaled accordingly (see Supporting Information Figures2 and 3). The reciprocity principle, introduced by Hoult andRichards in 1976 and subsequently reviewed in 2000,7,8 states thatthe signal strength is directly proportional to the square root ofthe quality factor of the probe. The quality factor is inverselyproportional to the square of the 90° pulse width. Thus, in a givenprobe and sample, if the 90° pulse width is shorter then the qualityfactor of the probe is larger and vice versa. Since these relation-ships are straightforward, we can easily calculate the necessarypower for the reference signal intensity based on the calibrationsample’s pulse width and electronic reference signal power asfollows:

ERPWRsamp ) ERPWRcal(pw90cal/pw90samp)

where ERPWRsamp is the resulting power for the reference signal,ERPWRcal is the electronic reference signal power based on thecalibration sample, pw90cal is the 90° pulse width of the tuned andmatched calibration sample, and pw90samp is the 90° pulse widthof the tuned and matched sample. The pulse power is assumedto be the same for both calibrant sample and unknown sample.The results for the ARTSI technique are shown in Figure 2 andTable 1. Clearly this technique successfully predicts the 2 mMconcentration of sucrose in each of the samples measured to amuch better degree than the data acquired using the ERETICmethod. There is still some scatter in the data, which may beaccounted for by various experimental errors, in particular howaccurately the 90° pulse width of each sample was measured,variations in the NMR tubes, variations in volume measurements,and so forth. This scatter is no greater than that if one were touse the 0.24 mM DSS internal concentration standard (instead ofthe electronic reference) as shown in Table 1. A 5% variation orerror in quantitative measurement is considered acceptable formost applications. It should be noted that the ARTSI techniquecan only be applied with accuracy when the probe can be tunedand matched to 50 Ω. If this requirement is not met, the reciprocityprinciple no longer applies and one cannot use the simplerelationships described here.

CONCLUSIONSWe have shown that the ERETIC signal does not always

accurately factor in changes in the receptivity of the proton coilwith changing sample properties (see Supporting InformationFigure 2). Instead, it reflects how the isolation between thereference port and observe port of the probe changes as a functionof tuning and sample properties. The ERETIC technique can beapplied where the tuning and matching between the referencesample and unknown samples does not change or in probes werethe sample does not change the receptivity of the probe’s coil.This might be the case for probes that have very small samplevolumes such as a room-temperature 1-mm probes or microcoilprobes. The ARTSI technique does not send the reference signalthrough the NMR probe. It sends the electronic signal directlyto the well-controlled receive path of the spectrometer and utilizesthe well-established reciprocity principle (see Supporting Infor-mation Figure 3). The 90° pulse width of the sample is used asan inverse measure of the effective probe sensitivity change withdifferences in sample properties. The amplitude of the reference

Figure 2. Calculated concentration of 2 mM sucrose samples withincreasing salt concentration results obtained using the ERETIC (filledin boxes) and ARTSI (filled in circles) techniques. Calculations werebased on the anomeric signal compared to the reference signal placedat -1 ppm. Data for each sample were collected 10 times for eachreferencing technique. The error bars show the standard deviationof 10 measurements.

8322 Analytical Chemistry, Vol. 80, No. 21, November 1, 2008

signal is adjusted proportionally. We propose that the ARTSItechnique can be used for a wide variety of samples and need notbe limited to samples of very similar probe receptivity. The ARTSItechnique takes into account changes in probe receptivity andcorrects the reference signal intensity, which in turn gives a moreaccurate quantitation of the unknown sample’s concentration.

We have presented only one way of solving the electronicreferencing problem. Recently, Upton presented another solutionwhere the reference signal is fully simulated and simply added tothe digital NMR data.11,12 The intensity of this artificial signal isadjusted based on solvent, receiver gain, number of transientsacquired, and other experimental factors are taken into account.While Upton presents a good general solution that can be usedfor a wide range of applications, it does require additional software,does not test the receiver path, and requires the accurateconsideration of a wider range of parameters than required forARTSI.

The electronic referencing procedure introduced here asARTSI is a simple improvement that can be easily implemented

on any modern spectrometer without any additional software andminimal additional hardware. It can also be applied to a wide rangeof samples without recalibration.

ACKNOWLEDGMENTWe thank Phil Hornung, Christine Hofstetter, and Mark Dixon

for their constructive input and technical advice, as well as DimitrisArgyropoulos and Peter Sandor for their pulse sequence writingexpertise for this project. We also acknowledge the VarianApplications laboratory in Palo Alto for allowing the use of theirspectrometers for this work.

SUPPORTING INFORMATION AVAILABLEAdditional information as noted in text. This material is

available free of charge via the Internet at http://pubs.acs.org.

Received for review April 29, 2008. Accepted September 4,2008.

AC800865C(11) Upton, R., SMASH, Chamonix, France, 2007.(12) Upton, R., ENC, Asilomar, California, March 10, 2008.

Table 1. Comparison of Results for the ERETIC and ARTSI Techniques to the Internal 0.24 mM DSS InternalStandarda

ARTSI technique ERETIC technique

[NaCl](mM)

[sucrose](mM) based

on DSS

[sucrose](mM) based

on ERb

anomericsignal

to noise

[sucrose](mM) based

on DSS

[sucrose](mM) based

on ERb

anomericsignal

to noise90° pulse

(µs)

0 2.00±0.04 2.00±0.04 306±10 2.00±0.04 2.00±0.04 319±09 6.43±0.0750 2.04±0.03 2.06±0.03 293±12 1.98±0.04 1.89±0.04 291±10 7.03±0.08100 2.04±0.04 2.04±0.04 258±09 2.02±0.03 1.81±0.03 258±09 7.50±0.14150 2.04±0.06 2.05±0.06 253±12 2.00±0.04 1.63±0.04 255±10 7.96±0.06200 2.03±0.05 2.07±0.04 247±09 2.02±0.02 1.84±0.04 245±06 8.16±0.07250 2.03±0.03 2.09±0.03 237±06 2.03±0.04 1.73±0.04 245±09 8.59±0.09

a Standard deviations are shown for 10 measurements. The average calibrated 90° pulse witdths (for 6 independent measurements) and averagesignal to noise (10 measurements) are also shown. The anomeric proton of sucrose was used for signal to noise measurements with a 200 Hz noiseregion. Note, at 250mM the probe had reached the edge of its match window. This sample could not be matched as well as the others thereforethe Principle of Reciprocity may no longer apply as can be seen as a loss of accuracy in the concentration measurement. This is a limitation of theprobe used to gather the data and should not be interpreted to be a limit of the ARTSI technique. b ER, electronic reference.

8323Analytical Chemistry, Vol. 80, No. 21, November 1, 2008