• Investigate the potential of NMR spectrometry in process analysis
• Develop a low-field high-resolution NMR spectrometer for use in process analysis
• Evaluate potential of at-line and on-line NMR spectrometry using example applications
• Develop signal and data processing methods to analyse overlapping signals
• Compare NMR spectrometry with other process analytical techniques
• Conduct a plant-based evaluation of technique

Achievements

1 Instrumentation
A low-field NMR spectrometer (operating frequency of 29 MHz for ¹H) has been developed by Resonance Instruments (Witney, UK). The homogeneity of the magnet was increased through improvements in magnet design and the addition of shim coils. A lock channel was also employed to improve the stability of the magnetic field. Using the instrument, it is possible to acquire ¹H, ¹⁹F and ³¹P NMR spectra with sufficient resolution to enable chemical shift information to be obtained. There are holes in the top and bottom of the magnet box, and the probes employ a removable stop at the base to allow on-line measurements to be performed.

2 Applications
At-line ¹H NMR spectrometry

• determination of strong acid concentration
• determination of alkyl chain length in nonyl phenol ethoxylates
• compositional analysis of samples from a methacrylamide process
• qualitative analysis of samples from a benzene production plant
• qualitative and quantitative analysis of polyether polyols
• quantitative analysis of the esterification reaction of crotonic acid and 2-butanol
• qualitative analysis of the esterification reaction of itaconic acid and 1-butanol

At-line ¹⁹F NMR spectrometry

• quantitative analysis of the dehydroxylation of tetrafluorohydroquinone
• quantitative analysis of the aminolysis of 4-fluorobenzoyl chloride
• quantitative analysis of an Avecia process

On-line NMR spectrometry

• quantitative analysis of the esterification reaction of crotonic acid and 2-butanol

3 Data Analysis

• PLS analysis of spectral data
• Analysis of low-field NMR signals in the time domain using the continuous wavelet transform
• Analysis of low-field NMR signals in the time domain using a modification of the generalised rank annihilation method (FID-GRAM)
• Analysis of low-field NMR signals in the time domain using a modification of the direct exponential curve resolution algorithm (FID-DECRA)

4 Deliverables

• Six IMB reports have been issued and a further six are in preparation
• One paper has been published, one paper is in press, one paper has been submitted for publication and a further six papers are in preparation
• Oral presentations have been given at a number of conferences including IFPAC 2000 (invited) and 2001, FACSS 1999, CAC 2000 and Pittcon 2001. An invited presentation will also be given at RSC Annual Conference 2001.

5 Objectives versus Achievements

With the exception of the final objective, all the objectives were realised during the project.

• A thorough review of the literature was prepared (CPACT project 3 IMB report, document ref. 98/P3/1), which was subsequently submitted to ‘Analyst’ for publication (A. Nordon, C.A. McGill and D. Littlejohn, Analyst, 2001, 126, 260).
• A low-field high-resolution NMR spectrometer was developed by Resonance Instruments.
• A number of samples were obtained from the member companies to assess the potential use of at-line and on-line NMR spectrometry. However, the model esterification reactions conducted within CPACT project 1 were also used to evaluate on-line NMR spectrometry.
• Low-field NMR signals were analysed in the frequency domain using standard chemometric techniques. However, methods for analysis of signals in the time domain were developed in collaboration with CPACT project 2 (wavelet transform) and MCEC (FID-GRAM and FID-DECRA).
• Comparisons of NMR spectrometry with other process analytical techniques, such as NIR, Raman and UV-visible spectrometries, were made using the model esterification reactions conducted within CPACT project 1. The reactor facility at University of Strathclyde was used extensively for this purpose.
• Participation in plant-based trials was sought from the CPACT member companies, but this objective was not realised.

Deliverables

Review of Flow and Process NMR document
Document Ref: 98/P3/1
Issued: 13 October 1998

Executive Summary
NMR spectrometry is a widely used analytical technique for structural elucidation and identification of chemical species in the analysis of many different materials including organic chemicals, inorganic complexes and large biological molecules. In this context, the instrumentation used comprises a high-field superconducting magnet (typically 200-600 MHz), and so detailed chemical shift and coupling constant information may be extracted from the high-resolution spectra obtained. At the moment NMR spectrometry is not commonly used in process analytical chemistry. Indeed, NMR spectrometry has been mentioned for the first time in the most recent of a number of reviews of process analytical chemistry, but only then within a miscellaneous techniques section. This is partly due to NMR techniques being rather complicated and the signals obtained can be difficult for non-NMR specialists to understand. NMR instruments can be expensive and the technique is much less sensitive than others, e.g. IR spectroscopy.

However, NMR spectrometry has the potential of being a very useful technique in process environments as it is non-destructive and does not require the measurement probe to be inserted into the process liquors, which avoids fouling. Another problem that was faced when NMR spectrometry was initially applied in at-line or on-line process analysis was that high-field laboratory instruments were simply moved into the plant. Process operators found problems in the calibration and maintenance of these instruments. Recently, small, dedicated low-field (<60 MHz) NMR systems, based on permanent magnet technology, have been developed, which in turn has led to more reports of at-line and on-line process applications.

As interest in the use of NMR spectrometry in process applications is increasing, a review of the subject has been prepared. The report covers various aspects of the technique that are relevant to NMR analysis of static or flowing materials (e.g. foods, chemicals and polymers). The review begins with a discussion of the factors that affect NMR measurements of flowing samples. This is followed by sections on low-field NMR applications, on-line process applications and the use of magnetic resonance imaging in industry. Brief details are given on current commercial instrumentation. The final section describes chemometric and data analysis procedures used to acquire chemical and process information from NMR measurements, especially when overlap of resonances occur, as is often encountered in low-field NMR spectra.

The Effect of Resolution and Signal-to-noise Ratio of Simulated Spectra on the Predictive Capabilities of a PLS Calibration Model
Document Ref: 99/P3/1
Issued: 29 September 1999

Summary
Techniques that are used, or are of potential use, in process analysis tend to yield spectra which are poorly resolved and less easy to interpret than signals produced by laboratory-based techniques. To extract the desired information from overlapping signals, chemometric procedures, such as partial least squares (PLS), are commonly used. In this study, spectra have been simulated for a two component mixture in order to examine the factors that affect the predictive ability of PLS. The factors considered include resolution, signal-to-noise ratio and lineshape. Although a decrease in predictive accuracy was observed for poorly resolved signals, the signal-to-noise ratio was found to have the greatest influence on accuracy. Whether the lineshape was Gaussian or Lorentzian had no effect on the predictive ability of PLS. It should be possible to use PLS to predict concentrations with a high degree of accuracy from data acquired using a wide range of analytical techniques, irrespective of line shape and resolution, if the signal-to-noise ratio is of a sufficient level.

Potential Applications of At-Line Process Analysis by 1H NMR Spectrometry at 28.8 MHz
Document Ref: 99/P3/2
Issued: 15 October 1999

Summary
The development version of a bench-top low-field NMR spectrometer, operating at 28.8 MHz for ¹H, has been evaluated for use in process analysis. The instrument incorporates a permanent magnet, and is smaller and more robust than high-field NMR spectrometers operating at 200-600 MHz for ¹H. Areas of potential application assessed include determination of acid concentration, individual component concentration in hydrocarbon mixtures and alkyl chain length, and the monitoring of a batch reaction.

It has been demonstrated that acid concentration can be determined with a precision of

+ 0.2 M using low-field NMR, except where the level of iron (III) (and/or other paramagnetic species) is in excess of 500 µg mL^-I. Through the use of NMR spectrometry in combination with partial least squares (PLS) calibration, the composition of multi-component mixtures can be analysed quantitatively. The results obtained for mixtures of cyclohexane, hexene and toluene were of a comparable degree of accuracy to those obtained using NIR spectrometry. It was also possible to determine the concentration of individual species present in an esterification reaction, thus enabling the extent of reaction to be monitored. The ester concentrations predicted from the NMR spectra were in good agreement with the concentrations obtained using GC. Qualitative analysis of samples taken at various stages in a benzene production process was also possible using low-field ¹H NMR spectrometry. The removal of non-aromatic C6 species, at different stages in the process, was identified. The degree of ethoxylation in a range of nonylphenol ethoxylate samples was determined from integration of signals present in the low-field ¹H NMR spectrum and the results were in good agreement with the values obtained using high-field ¹H NMR spectrometry

At present, the precision of the results obtained by low-field process NMR spectrometry is limited by the instrumentation, although accuracy is comparable to both NIR and GC. However, developments are currently underway in this area, which should then enable both the accuracy and the precision of analyses performed with the Resonance Instruments' spectrometer to rival that of more established process analytical techniques.

Consideration of the effect of flow for on-line measurements by NMR spectrometry at 28.8 MHz
Document Ref: 99/P3/3
Issued: 15 October 1999

Summary
High- field NMR spectrometers found in most laboratories are unsuitable for the rigours of a process environment. However, on-line process NMR may be feasible through the use of low-field instrumentation based on a permanent magnet, incombination with chemometric techniques. In the work described here, a bench-top low-field NMR spectrometer operating at 28.8 MHz for ¹H has been used to monitor samples on-line. The decrease in signal intensity observed with increasing flow rate has been reduced by designing a flow cell that incorporates a premagnetisation region. Concentrations of toluene, cyclohexane and hexene in streams flowing at 25 mL min^-1 were determined with a similar degree of accuracy to that achieved for static samples when the premagnetisation flow cell was used and the ¹H spectra were analysed with PLS calibration.

Using a simple sampling loop, it has also been possible to acquire on-line ¹H NMR spectra of an esterification reaction conducted in a 5 L batch reactor. Through the use of suitable calibration models and a more advanced sampling loop, it should be possible to analyse quantitatively batch reactions with on-line NMR spectrometry. At present, the precision of the results obtained by process NMR is limited by the instrumentation. However, developments are currently underway in this area, which should then improve both the accuracy and precision of NMR at low frequency for both static and flowing samples.

Preliminary investigation of quantitative process analysis by low-field ¹⁹F NMR spectrometry using a model reaction
Document Ref: 99/P3/4
Issued: 20 December 1999

Summary
The paper describes the use of a small, robust spectrometer with a permanent magnet for the measurement of ¹⁹F NMR spectra of mixtures of tetrafluorohydroquinone, 1,2,4,5-tetrafluorobenzene and 2,3,5,6-tetrafluorophenol. Partial least squares regression and variable selection were used to determine the concentration of the three analytes present at levels up to 10% m/v. It was not necessary to use the full spectrum to obtain the optimum standard errors of prediction (SEP). When 10 Hz line broadening was applied to the data, the signal-to-noise ratio increased while the resolution decreased. In general, this lowered the SEP values and detection limits were obtained for the three components, emphasising the importance of acquiring good signal-to-noise data. The SEP values obtained with 10 Hz line broadening were 0.038% m/v, 0.074% m/v and 0.164% m/v for tetrafluorohydroquinone, 1,2,4,5-tetrafluorobenzene and 2,3,5,6-tetrafluorophenol, respectively. The detection limits obtained for tetrafluorohydroquinone, 1,2,4,5- tetrafluorobenzene and 2,3,5,6-tetrafluorophenol were 0.01 % m/v, 0.03% m/v and 0.11% m/v, respectively. The use of l^st and 2^nd derivatives of the ¹⁹F NMR data did not improve the quality of the calibration. Although the accuracy of the NMR results was good, the precision of the measurements (1-5%) was poorer than that obtained by near-infrared spectrometry (0.1-3%). This was due to limitations in the NMR development instrument which are presently being removed.

Quantitative Analysis of Low-field NMR Signals in the Time Domain
Document Ref: 01/P3/1
Issued: 26 February 2001

Quantitative analysis of low-field NMR signals in the time domain

Alison Nordon,¹ Paul J. Gemperline,² Colin A. McGill¹ and David Littlejohn,^1
1 Department of Pure and Applied Chemistry/CPACT, University of Strathclyde, 295 Cathedral Street, Glasgow, G1 1XL, UK.
² Department of Chemistry/MCEC, East Carolina University, Greenville, NC, USA.

Background
Two methods, based on adaptations of the Generalised Rank Annihilation Method (GRAM) and the Direct Exponential Curve Resolution Algorithm (DECRA), have been developed which allow direct quantitative analysis of the time domain signal, i.e. free induction decay (FID), without the need for construction of calibration models.

Experimental
The two methods, termed FID-GRAM and FID-DECRA, were written in Matlab version 5.3 (Mathworks Inc., MA, USA). The performance of each method was investigated using simulated data. FID-DECRA was also used to analyse ¹⁹F NMR data, which was obtained using a NMR spectrometer (operating frequency of 29 MHz for ¹H) manufactured by Resonance Instruments. Spectral data, i.e. after Fourier transformation, were also analysed using partial least squares (PLS) for comparison.

Results
Both FID-GRAM and FID-DECRA can be used to analyse NMR signals without the need for the application of a phase correction to the data. However, as FID-DECRA does not require a reference sample for quantification, this procedure is preferable. Application of FID-DECRA to ¹⁹F NMR data gave concentrations comparable to those derived from PLS analysis of spectral data (typical spectrum shown in figure). The table shows the results obtained for tetrafluorohydroquinone. However, only a single calibration sample was employed with FID-DECRA, whereas ten calibration samples were required for PLS.

Conclusions
As FID-DECRA does not require phase correction nor construction of multivariate calibration models, it is likely to be particularly useful in process NMR spectrometry.

Determination of ethylene oxide content of polyether polyols by low-field ¹H NMR spectrometry
Document Ref:01/P3/3
Issued: 27 September 2001

Background
Methods have been developed and compared for the analysis of a glycerol-based polyether polyol using a low-field, medium-resolution NMR spectrometer.

Experimental
Twenty samples of a glycerol-based polyether polyol with an average relative molecular mass of 5000 and an EO content of 14.8-15.9 % w/w were analysed at-line using a low-field NMR spectrometer (29 MHz for ¹H; Resonance Instruments). Two methods of obtaining signal areas were compared, spectral integration (frequency domain) and a direct exponential curve resolution algorithm for FID analysis (time domain), known as FID-DECRA. Both methods were used to calculate directly the EO content of the samples and the results were compared to those of high-field ¹³C NMR spectrometry.

Results
FID-DECRA offered advantages over spectral integration as phase correction was not required and a 2-fold lower RSD was achieved. Even so, direct analysis gave a negative bias in the EO concentration (c.f. the ¹³C NMR results).

When a univariate calibration model, based on the ¹³C NMR results for some of the samples, was used (see table) more accurate (average % error of 0.5 %) and precise (average RSD of 0.6 %) concentrations of EO were obtained with FID-DECRA.

Conclusions
FID-DECRA has accuracy and precision advantages over spectral integration for the at-line determination of % EO content of polyether polyols. The procedure is also easier to automate, which desirable in process analysis.

Comparison of calibration methods for the monitoring of a fluorobenzene batch reaction using low-field ¹⁹F NMR, ¹H NMR, NIR and Raman spectrometries
Document Ref: 01/P3/4
Issued: 27 September 2001

Background
The suitability of different process spectrometry techniques has been assessed, in terms of calibration requirements, accuracy and precision, for the at-line monitoring of the sulphonation of fluorobenzene.

Experimental
Mixtures of fluorobenzene, 4-fluorobenzenesulphonyl chloride and 4-fluorophenyl sulphone were prepared in dichloromethane to represent the reaction shown. Samples were analysed by low-field ¹H and ¹⁹F NMR (Resonance Instruments), NIR (Bomem) and Raman (Kaiser) spectrometries.

Results

Partial least squares (PLS) calibration was required to analyse the spectra obtained by NIR spectrometry and low-field (29.1 MHz) ¹H NMR spectrometry. The low-field (27.4 MHz) ¹⁹F NMR spectra contained well-resolved signals for the three fluorine-containing compounds and univariate calibration was adequate. The Raman spectra of two of the compounds exhibited fluorescence and so this technique was not considered suitable for monitoring the reaction. The most accurate and precise results were obtained using PLS analysis of NIR data and univariate analysis of ¹⁹F NMR data. Results of poor accuracy were obtained using PLS analysis of ¹H NMR data as spectra for two of the compounds were too similar.

Conclusions
NIR and ¹⁹F NMR spectrometries are suitable for the monitoring of the sulphonation of fluorobenzene. Analysis of the NIR data required construction of PLS calibration models, whereas univariate calibration was satisfactory for ¹⁹F NMR spectra. Hence, low-field ¹⁹F NMR spectrometry may offer a simpler alternative method of analysis to NIR spectrometry for the monitoring of fluorocarbon processes.