Journal of Magnetic Resonance 249 (2014) 53–62

Contents lists available at ScienceDirect

Journal of Magnetic Resonance journal homepage: www.elsevier.com/locate/jmr

Simultaneous 19F–1H medium resolution NMR spectroscopy for online reaction monitoring Nicolai Zientek a, Clément Laurain a,b, Klas Meyer a, Matthias Kraume c, Gisela Guthausen d, Michael Maiwald a,⇑ a

BAM Federal Institute for Materials Research and Testing, Richard-Willstaetter-Straße 11, 12489 Berlin, Germany École Nationale Supérieure de Chimie de Lille, Avenue Mendeleiev CS 90108, 59652 Villeneuve D’ascq Cedex, France Department of Chemical Engineering, Technische Universität Berlin, Straße des 17. Juni 136, MA 5-7, 10623 Berlin, Germany d Pro2NMR, Institute of Mechanical Process Engineering and Mechanics and Institute of Biological Interfaces, KIT, Adenauerring 20 b, 76131 Karlsruhe, Germany b c

a r t i c l e

i n f o

Article history: Received 7 August 2014 Revised 3 October 2014 Available online 18 October 2014 Keywords: NMR 1 H 19 F Medium-resolution NMR Online NMR Quantitative NMR Reaction monitoring Data processing Process analytical technology

a b s t r a c t Medium resolution nuclear magnetic resonance (MR-NMR) spectroscopy is currently a fast developing field, which has an enormous potential to become an important analytical tool for reaction monitoring, in hyphenated techniques, and for systematic investigations of complex mixtures. The recent developments of innovative MR-NMR spectrometers are therefore remarkable due to their possible applications in quality control, education, and process monitoring. MR-NMR spectroscopy can beneficially be applied for fast, non-invasive, and volume integrating analyses under rough environmental conditions. Within this study, a simple 1/1600 fluorinated ethylene propylene (FEP) tube with an ID of 0.0400 (1.02 mm) was used as a flow cell in combination with a 5 mm glass Dewar tube inserted into a benchtop MR-NMR spectrometer with a 1H Larmor frequency of 43.32 MHz and 40.68 MHz for 19F. For the first time, quasi-simultaneous proton and fluorine NMR spectra were recorded with a series of alternating 19 F and 1H single scan spectra along the reaction time coordinate of a homogeneously catalysed esterification model reaction containing fluorinated compounds. The results were compared to quantitative NMR spectra from a hyphenated 500 MHz online NMR instrument for validation. Automation of handling, pre-processing, and analysis of NMR data becomes increasingly important for process monitoring applications of online NMR spectroscopy and for its technical and practical acceptance. Thus, NMR spectra were automatically baseline corrected and phased using the minimum entropy method. Data analysis schemes were designed such that they are based on simple direct integration or first principle line fitting, with the aim that the analysis directly revealed molar concentrations from the spectra. Finally, the performance of 1/1600 FEP tube set-up with an ID of 1.02 mm was characterised regarding the limit of detection (LOQ (1H) = 0.335 mol L1 and LOQ (19F) = 0.130 mol L1 for trifluoroethanol in D2O (single scan)) and maximum quantitative flow rates up to 0.3 mL min1. Thus, a series of single scan 19F and 1H NMR spectra acquired with this simple set-up already presents a valuable basis for quantitative reaction monitoring. Ó 2014 Elsevier Inc. All rights reserved.

1. Introduction Quantitative high-resolution online nuclear magnetic resonance spectroscopy based on the application of online (flow) techniques for process applications is the method of choice for the investigation of complex fluid mixtures since it works online and without sample preparation [1,2]. A major advantage of NMR spectroscopy is that the method features linearity between absolute ⇑ Corresponding author at: BAM Federal Institute for Materials Research and Testing, Richard-Willstaetter-Straße 11, 12489 Berlin, Germany. E-mail address: [email protected] (M. Maiwald). http://dx.doi.org/10.1016/j.jmr.2014.10.007 1090-7807/Ó 2014 Elsevier Inc. All rights reserved.

signal area according to the number of spins and sample concentration, which makes it an absolute analytical comparison method being independent on the matrix. Thus it is a promising quantitative technique without elaborate calibration. Various theoretical and experimental papers have dealt with quantitative NMR spectroscopy, e.g., with respect to continuous flow [3,4], in agriculture [5], for trace components [6] or process engineering or pharmaceutical chemistry applications [7–12]. The increasing performance of compact MR-NMR spectrometers with permanent magnets allows employing NMR spectrometers in an industrial environment without high maintenance requirements and without the need for cryogenic liquids [13–15].

54

N. Zientek et al. / Journal of Magnetic Resonance 249 (2014) 53–62

Field strengths are usually below 1.88 T (80 MHz proton frequency) for the observation of 1H. 19F and 31P nuclei are likewise observable with MR-NMR instruments with acceptable sensitivity – and even 13C. Magnetic fields with small stray fields are typically produced by a permanent magnet for employment in direct proximity of production plants. We decided to use the term medium resolution NMR spectroscopy to address NMR spectroscopy with permanent magnet systems to distinguish them from high-resolution NMR spectroscopy conducted with cryo magnets. Some authors use the term low field NMR spectroscopy which, in our eyes, is confusing as the spectral resolution is often not mentioned and sometimes, also relaxation time distributions are considered as spectroscopy. MR-NMR spectra can only be derived if sufficient permanent magnetic field homogeneity (shim) is provided. MR-NMR spectroscopy provides physical information like chemical shift and coupling constants as an alternative ‘‘physical axis’’ compared to conventional methods like optical spectroscopy. Therefore, MRNMR spectroscopy extends the accessible information from technical processes especially for, e.g., aromatics-to-aliphatic conversions or isomerisations where conventional methods fail due to only minor changes in functional groups. Commercial NMR instruments with a flow-through cell, possessing a good lineshape, signal-to-noise-ratio, and sufficient robustness are still under development. Since only recently promising MR-NMR benchtop instruments with acceptable performance came to market, a process integrated instrument could accordingly be developed on the basis of presently emerging laboratory instruments. Within this study, a different approach from [16–22] was realised: A simple 1/1600 FEP tube with an ID of 0.0400 (1.02 mm) was used as a flow probe. The thermal isolation between the flow tube and the NMR probe was realised with a simple 5 mm glass Dewar tube inserted into a benchtop MR-NMR spectrometer. In order to gain the highest possible S/N ratio in a flowing fluid, its flow rate has to be limited with respect to a sufficiently long effective polarisation length, otherwise, non-equilibrium effects have to be considered [17]. The adaption of commercially available devices for online measurements additionally requires considerations on flow effects and optimisation of flow probes in respect to sensitivity and selectivity. It should be pointed out that for technical applications special importance lays on the magnetic field geometry in the pre-magnetisation volume, where the sample is magnetised completely before it enters the active volume of the MR-NMR probe. In general, this difficulty can be circumvented by the design of a bypass system with reduced flow in the NMR probe or triggered stop flow, if the effective polarisation length turns out to be very short. Last but not least, automated data analysis of the MR-NMR spectra is an emerging field of interest, since process monitoring produces an immense quantity of data which require automated signal processing and data analysis tools. Even though big progress has been made in data processing within the last decade [25], a practical and user-friendly processing of immense number of spectra remains challenging. In addition to the challenge to prepare large series of data, manual preparation comes with the disadvantage of user-introduced uncertainties. Conventional spectral data processing software, which comes with the NMR instruments or is available as manufacturer-independent toolbox, offers automatic preprocessing procedures. However, the results are often not satisfying for technical mixtures with respect to precision along a time series of spectra and therefore cannot be used for an online data pretreatment. For quantitative analysis of the NMR data different approaches are known. The traditional approach is direct integration of lines which can be correlated with the corresponding nuclei involved in the reaction. This approach suffers from imperfect baselines

and phasing and induced errors in the case of overlapping lines. In a more advanced approach, lines are modelled by analytical functions such as a Lorentz-Gauss function. The spectra are composed by the model lines either by applying statistical tools or using prior knowledge. Together with prior knowledge, also complex spectra have been analysed in this way, and the quantity of each component can be extracted. These line shape fittings are known as spectral deconvolution or hard modelling as they rely on physicochemical information which is sample and substance specific. A modification of the hard modelling is available: indirect hard modelling (IHM) [23,24], which considers each line separately, but allows restraints in the manifold. Reaction monitoring, involving prior knowledge about possible reaction mechanisms are a predestinated application field for IHM. The later approach using prior knowledge about the chemical reaction and the multiplet structure was found to work also for rather signal-to-noise limited MR-NMR spectra [17]. Spectral dispersions of MR-NMR spectra are significantly reduced compared to high field instrumentation due to the relatively low field strengths so that obviously various data analysis tools and even chemometric models have been applied to enable a multi-component or multi-parameter correlation of sample properties. Examples for the application of PLS-R to NMR data are available for high-field [25] (and references therein) as well as for MR-NMR spectroscopy [18], including comparisons with classical data analysis. In contrast to Fourier-transformed spectra, the direct exponential curve resolution algorithm (DECRA) was applied to time-domain data [26]. An interesting approach is to reduce the calibration effort, which could be explored also for NMR data [27]. All data analysis methods have in common, that a constant signal quality is an important prerequisite for success – maybe even more important than the highest possible magnetic field. Therefore, an external reference was used in [18]. Beyond this, chemometric models require fixed chemical shifts.

2.

19

F NMR spectroscopy

Because of its favourable nuclear properties and high abundance, 19F NMR measurements are almost equally fast than 1H NMR experiments (84% relative sensitivity compared to protons) and 19F is probably the third most studied nucleus [28,29]. Since fluorine is the most electronegative element, physical and chemical properties of molecules, such as their acidity and basicity, can tremendously be influenced by fluorine atoms placed in the molecule [30]. Therefore, fluorine is contained in a large number of organo-chemicals, polymers and pharmaceutical products. Acquisition of fluorine NMR can also be beneficially used in reaction monitoring especially when proton spectra show overlapping signals. The range of the chemical shift extends over a range from approx. 0 to 200 ppm compared to 0–10 ppm for protons and is thus less likely for signal overlapping. Furthermore the repetition delay between two measurements is smaller for fluorine since T1 relaxation times are often shorter. In the current technical realisation, it is not required to change the tuning of the probe or the shim to measure fluorine and protons subsequently, therefore measurement of both nuclei can be easily done one after another. 19 F information from semi parallel acquisitions present could also improve chemometric models for NMR data evaluation, since additional and physically independent information is provided which can be used to improve models and streamline calibrations. In this paper, the quasi-simultaneous acquisition of 19F and 1H NMR spectra is applied to the study of an esterification reaction with the aim of an accurate monitoring of the kinetics of the reaction. MR-NMR data are analysed and compared to high field data

55

N. Zientek et al. / Journal of Magnetic Resonance 249 (2014) 53–62

for obtaining an impression about the current possibilities of MRNMR with respect to chemical reaction and process analytics.

titative and independent flow rates, both in the HR-NMR probe (0.1–1.5 mL min1) as well as in the MR-NMR probe (0.1– 0.3 mL min1), in addition to a fast bypass line back to the reactor. The latter was introduced to achieve a rapid sample transfer from the reactor to the four-way split in the vicinity of the NMR instruments. The flow rates of the recycling bypass line and through the HR-NMR spectrometer were adjusted by variable back pressure regulators V21 and V22 (P-446, 1.72 MPa, IDEX Health&Science, Oak Harbor, WA, USA), respectively. The flow rate through the MR-NMR spectrometer was controlled by a needle valve V23 (P788, IDEX Health&Science). In order to observe the pressure prior to the HR-NMR an ultra low volume pressure transducer (XTM190, 3.5 MPa, Kulite Semiconductor, Leonia, NJ, USA) was used. Another pressure transducer integrated into the pump was used for an automated pump switch off. Since dust particles or precipitating solids are particularly dangerous for the set-up as they can block the tubing, further care was taken using filters F2 and a pressure relief valve (V24, 1.7 MPa, U-456, IDEX Health&Science). Mass flow rates were determined with a balance close to the reactor. For this purpose the flow was temporarily routed on the balance via the back-leading tubing to the reactor. For hyphenating, polyether ether ketone (PEEK) tubing was used for the connections between the reactor and the NMR. The distance from the reactor (C1) to the NMR instruments was 3.5 m. A choice of tubing with small inner diameter (ID) results in short delay times for sample transfer, but high pressure drops. The tubes leading to the NMR were typically 0.0200 ID (0.50 mm), but the return tubing back to C1 was wider with 0.0300 ID (0.75 mm). PEEK tubing was chosen in most cases because of its good mechanical properties and chemical resistance. All lines were liquid thermostated at the reaction temperature of 35 °C in C1. For this purpose, the lines were mounted inside insulated silicon tubing filled with heat transfer fluid which was connected to the cryostat via tees (see Fig. 2). Most parts in contact to the solution were also thermostated. The total hold-up of the system described (reactor–NMR–reactor) including filters, tubing, pump, valves, pressure transducer, and the NMR flow cell was kept below 5 mL and therefore small ( 30 s) and stopped flow. In the case of tr ? 1, which is conform with the pulse repetition time tr  5 T1 in experiments, the normalised signal S/S0 (corresponding to M/M0) was constant at low volume flow rates (Fig. 5, right). It is expected, the normalised signal S/S0 starts to decrease with increasing volume flow rate V_ from a certain point where spins enter the coil region without fully developed magnetisation (indicated as (iv) in Fig. 5. However, Fig. 5 exhibits a slight increase instead for reasons given below). Surprisingly, a relatively high maximal flow rate of 1 V_  0:3 mL min was experimentally found. The length of the coil Lc was determined to 7.5 mm (half width) with a flat bottomed 5 mm NMR tube filled to 1 mm with olive oil and successively moved through the active region of the MR-NMR spectrometer while plotting the proton intensity. Lpol was assumed to approximately 100 mm, represented in the curve simulation (ii) according to Eq. (2) in Fig. 5. The abovementioned experimental signal increase is due to temperature decrease and density increase for flow rates larger than 0.3 mL min1, indicated as guide for the eye (i). The temperature was derived from the chemical shift differences of the –OH and the –CH3 signal of methanol [38]. The temperature decrease

59

at higher flow velocities can be explained by the temperature difference between the non-thermostated syringe reservoir at 25 °C and the thermostated tubing of the inlet line at 28.5 °C, which was brought to magnet temperature to have adiabatic conditions. Preliminary experiments proved that heat flow and radiation cannot completely be neglected for very low flow rates, which affect the signal intensities by density effects. Since T1 slightly increases with temperature [17], this represents a further effect of improved saturation at lower temperatures. Fig. 5 also shows the half widths of the –CH3 signal as a function of the mean flow velocity. At the maximum quantitative volume 1 flow rate of V_  0:3 mL min the half width is already increased by about 20% (indicated with (iii) in Fig. 5). Hence, the signal shape is a practical indicator for quantitative analysis of spectra at these flow velocities. 4.4. Automated data pre-treatment The automated phasing method turned out to be very efficient correcting the phase of a big set of even noisy NMR spectra in short time. In Fig. 6: the superimposed spectra of a kinetic experiment are shown. In Fig. 6a the spectra are corrected by an automatic phase correction method provided by spectra evaluation program. In Fig. 6b the phase is optimised by the minimum entropy method. In comparison to the above mentioned classically phased spectra, the improvement in the spectral alignment is apparent. However, 1H MR-NMR spectra came up with smaller phase drifts or phase jumps during the course of the reaction compared to the 19F MR-NMR spectra. The effect is further conferred in the discussion. 4.5. Phase stability In order to discover the origin of the largely different S/N of 1H and 19F, spectra of neat 2,2,2-trifluoroethanol were observed for a series of shortly repeated single measurements. Presently, the instrument software performs an automatic phase correction after each acquisition, presumably behaving non-beneficial to the phase stability. The integrals of the real part of these phase corrected complex conjugated spectra were compared to magnitude spectra for both nuclei. The integrals were divided by the number of nuclei (19F/3 trifluoro ethyl triplet at 75.0 ppm; 1H/2, methylene quartet at 4.08 ppm) and the mean value of 60 spectra normalised to 1 with a fixed factor. It is well known that, due to the broad signal shape, magnitude spectra are less suited for quantification than auto-phased absorption signals. However, the comparison allows an analysis of influencing factors. Fig. 7 clearly demonstrates that scatter is predominantly phase induced. We expect that the instrument originally behaves much more phase stable, i.e., with as high precision as found for the magnitudes, when the automatic phase correction could be omitted, e.g., by user settings. It is observed about twice bigger for 19F compared to 1H, which is due to the sensitivity of the 19F channel. This is indicated by the larger pulse length required for a 90° flip angle. Hence, the auto-phased 19F MR-NMR spectra present a larger uncertainty after the automated phase correction, as can be seen from Fig. 8. 4.6. Esterification reaction Fig. 8 (and the enlarged Fig. 9) shows the experimental, time resolved molar concentrations of the molecular moieties from the NMR spectra. The time axis was corrected for the transfer time difference between reactor and MR- and HR-NMR, respectively. In order to plot the molar concentrations, a set of calibration mixtures was prepared and repeatedly measured and analysed by all methods. Original calibration data is presented in the Supplementary

60

N. Zientek et al. / Journal of Magnetic Resonance 249 (2014) 53–62

Fig. 7. Phase scatter of 1H and 19F MR-NMR signals for repeated measurements of 2,2,2-trifluoroethanol (1H/2: quartet at 4.08 ppm; 19F/3: triplet at 75.0 ppm). Comparison of absorption signals vs. magnitude signals, mean normalised to 1 with a fixed factor.

Material. Figs. 8 and 9 show the components TFE = 2,2,2-trifluoroethanol in blue, AA = acetic acid in green, and TFEA = 2,2,2-trifluoroethyl acetate in red. Whereas all data analysis was done by the direct integration method (DI), every 5th HR-NMR spectrum was also analysed using line fitting (LF), represented with symbols with black circles in Figs. 8 and 9. Since molar concentrations were determined, it was possible to balance the involved nuclei with (magenta) and without (black) the OH/H signal in order to analyse the stability and scatter of the balanced data, represented with black symbols in Fig. 8. The calibration curves (plotting signal area Ai vs. molar concentration ci/mol L1 for each species i, taking into account the number of contributing nuclei) allow for analysis and comparison of analytical sensitivities and blank corrections. However, LOD and LOQ values were not derived, since the calibrations cover compositions

over the full range, whereas only lower concentrations are reasonable for this reaction as well as for determination of the detection and quantification limits, as presented above. On a first view, all results are very soundly comparable over the course of the reaction and between the three different methods and the different data analysis methods. However, these results are already obtainable with direct integration – the simplest method of analysis. Since for the 1H MR-NMR spectra, the quartets of the methylene group overlap, only the non-overlapping two signals of each quartet were integrated. Also, the methylene singlet signals overlap for all 1H spectra, and integration with 64-fold the half width was not possible. Therefore, the signals were integrated to the minimum between both signals. As can be seen from the figure, this results in an over-estimate for the smaller product signals. During the reaction 2,2,2-trifluoroethanol and acetic acid signals decrease, whereas 2,2,2-trifluoroethyl acetate increases to the same extent. Since the product water is formed at equimolar ratio during the reaction but not quantified, the proton balance signals go down. If the OH/H signal is included into the balance, it agreeably follows a straight horizontal line. In case of fluorine only TFE and TFEA are quantified so that the equimolar reaction balance is fulfilled. Any deviations would allow following signal losses of undetected by-product formation as well as temperature, density, and partial excess volume effects, which are not observed here. Initiation of the reaction with sulphuric acid can be observed as little jump down for the reactant and balance signals for all substances present at this time as well as temporary outliers due to distorted NMR signals. Most interesting is the low scatter of the integral along the reaction propagation. It comes out, that MR-NMR proton signals exhibit only little more scatter than HR-NMR signals, which is a very good result for single scan spectra taken from the small sample volume. Compared to 1H, 19F MR-NMR scatter is about two-fold larger as stated above. In general, the good stability over the whole observation time under technical condition has to be pointed out

balance OH/H

1 1

1

H 500 MHz H 43 MHz

H 500 MHz H 43 MHz

1

balance 1

H 500 MHz (DI) H 43 MHz F 40 MHz

1

1

H 500 MHz (LF)

19

19

F 40 MHz

balance

1

H 500 MHz (DI) H 43 MHz

1

1

H 500 MHz (DI) H 43 MHz F 40 MHz

1

1

H 500 MHz (LF)

AA TFE

1

H 500 MHz (LF)

19

TFEA

Fig. 8. Course of the reaction monitored by MR and HR-NMR spectroscopy and data analysis by the direct integration method (DI) after calibrating signal areas to molar concentrations. The time axis has been corrected to the transfer times. (Legend: s 1H HR-NMR;  1H MR-NMR; h 19F MR-NMR; TFE = 2,2,2-trifluoroethanol (blue); AA = acetic acid (green); TFEA = 2,2,2-trifluoroethyl acetate (red), balance = summarised signal for each NMR method without OH/H signal (black), OH/H = summarised signal for each NMR method including OH/H signal (magenta); every 5th HR-NMR spectrum was also analysed using line fitting (LF): symbols with black circles). (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

61

N. Zientek et al. / Journal of Magnetic Resonance 249 (2014) 53–62

1

H 500 MHz (DI) H 43 MHz F 40 MHz

1

1

H 500 MHz (LF)

19

TFE

1

H 500 MHz (DI) H 43 MHz

1

H 500 MHz (DI) H 43 MHz F 40 MHz

1

1

1 1

H 500 MHz (LF)

H 500 MHz (LF)

AA

TFEA

19

Fig. 9. Course of the reaction monitored by MR and HR-NMR spectroscopy – detail of Fig. 8.

very positively, e.g., for the balances. Balances are constantly but only slightly overestimated for MR-NMR spectroscopy. Also the quantitative results are respectable since all methods fairly overlap. Little deviations are expected to be caused by the simple data analysis methods. Signals from direct integrations (especially 1H signals) might be slightly underestimated in comparison to line fittings. Additional uncertainties arise from the changes of the calibration spectra compared to the spectra in the technical mixture. However, since a detailed survey would inflate this paper, it will be subject of detailed further analysis – also using additional methods like PLS-R or IHM. Severe shifts of signal positions in the order of 0.01 ppm and changes in signal line widths were observed due to changing mixture matrix, which might present an obstacle to some data analysis methods.

5. Conclusion and outlook The performance of the 1/1600 FEP tube set-up in combination with a 5 mm Dewar tube is suitable for reaction monitoring applications and presents a valuable basis. The combined quasi-simultaneous proton–fluorine NMR spectra recorded with a series of alternating 19F and 1H single scan spectra yielded valuable information of the reacting mixture over the whole concentration range. The results were compared with quantitative HR-NMR spectroscopy, and a very good agreement was observed, which verifies the assumptions made for data analysis and the experimental basis. Detailed dynamic analysis of the flow behaviour in the flow tube revealed a relatively high maximal flow rate of 1 V_  0:3 mL min , up to which a complete magnetisation buildup is observed. Actually, for the observed esterification reaction a maximum flow rate up to 0.3 mL min1 was used. This behaviour makes triggered stop flow unnecessary and is most practical for an online set-up and operation under technical requirements, e.g., with little variations in flow rate. We think that this is especially due to the Halbach magnet construction, which presents a strong magnetic field for the inflowing fluid along the y-axis. For the magnet used in the present investigations, an effective

polarisation length of approximately 100 mm can be estimated, which is most favourable. Due to the small remaining active volume of the 1/1600 tube setup, which drastically reduces the effective volume of the MR-NMR probe and the analytical sensitivity (factor of 16), limits of detection for 1H and 19F are comparably low. For the reason of compromises to 19F tuning, which, e.g., can be read from the pulse times, especially fluorine suffers from reduced sensitivity of about half of the theoretical value. However, an improved phase scatter performance could be achieved by omitting any automatic phase correction by the instrument software after single acquisitions, e.g., by offering this as an option. If not highly concentrated mixtures are analysed like in this study, an alternative NMR flow cell design with optimised active volume promises LODs, which could be increased up to the physical limit. Assuming an ID of 3 mm LODs could be increased to approach the current performance of a 5 mm standard probe at 500 MHz by only a factor of 140. By designing an alternative NMR flow cell with optimised shape within the active region of the NMR, quantitative flow rates of even over 1 mL min1 could be realised in the future with only little diameter increase in the active volume. One of the most important prerequisites for technical applications is the capability for shimming on protons (of technical mixtures), which performed very well and was sustainably kept over a sufficiently long time. If strong line shape distortions occur during the reaction shimming can also be performed on the reacting mixture. Therefore the reaction has to be slow to guarantee a reasonable time resolution of the reaction course. All data analysis methods in general and especially chemometric models require a constant signal quality over the duration of the experiment, i.e., line shape and amplitude sensitivity. Finally it turns out that these factors are important prerequisites for a successful data analysis and even more important than the highest spectral dispersion. A unified data structure, not only for the raw data but also for switching between modular data preparation and data analysis tools is preferable for future developments since it smoothes the way for automated applications. The next steps towards system integration of MR-NMR spectrometers for field applications will be an explosion safe set-up,

62

N. Zientek et al. / Journal of Magnetic Resonance 249 (2014) 53–62

which is easily achievable since typical MR-NMR instruments come without need for cooling gas. The operation of all units of the set-up – from the spectrometer to data preparation and analysis modules – via a Programmable Logic Controller (PLC) is highly desirable. Besides, automated switching between different sample lines as well as for cleaning and rinsing has to be implemented in an automated set-up. Likewise standards have to be brought into the spectrometer. The PLC might also regulate internal control functions such as switching off functions or leakage control, which would have to be implemented.

[13]

[14]

[15]

[16]

Acknowledgments [17]

The authors thank Deutsche Forschungsgemeinschaft (Germany) for support by projects MA 2292/1-1 and GU1123/2-1. The authors are very grateful for the uncomplicated support from Magritek (Germany and New Zeeland), especially for triggering the instrument. Further, helpful discussions with Franz Dalitz are gratefully acknowledged. MM wants to thank Ulrich Panne for supporting this project. GG also acknowledges the Deutsche Forschungsgemeinschaft for financial support of the instrumental facility Pro2NMR.

[18]

[19]

[20] [21]

Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.jmr.2014.10.007.

[22]

References [23] [1] M. Maiwald, H.H. Fischer, Y.K. Kim, K. Albert, H. Hasse, Quantitative highresolution on-line NMR spectroscopy in reaction and process monitoring, J. Magn. Reson. 166 (2004) 135–146, http://dx.doi.org/10.1016/ j.jmr.2003.09.003. [2] M. Maiwald, O. Steinhof, C.J. Sleigh, M.A. Bernstein, H. Hasse, Quantitative high-resolution online NMR spectroscopy in pharmaceutical reaction and process monitoring, in: I. Wawer, B. Diehl, U. Holzgrabe (Eds.), NMR Spectroscopy in Pharmaceutical Analysis, Elsevier, Amsterdam, 2008, pp. 471–491. http://dx.doi.org/10.1016/B978-0-444-53173-5.00020-2. [3] J.F. Haw, T.E. Glass, H.C. Dorn, Conditions for quantitative flow FT 1H NMR measurements under repetitive pulse conditions, J. Magn. Reson. 49 (1982) 22–31, http://dx.doi.org/10.1016/0022-2364(82)90292-X. [4] L.F. Gladden, Industrial applications of nuclear magnetic resonance, Chem. Eng. J. 56 (1995) 149–158, http://dx.doi.org/10.1016/B978-0-444-53173-5.000202. [5] G. Maniara, K. Rajamoorthi, S. Rajan, G.W. Stockton, Method performance and validation for quantitative analysis by 1H and 31P NMR spectroscopy. Applications to analytical standards and agricultural chemicals, Anal. Chem. 70 (1998) 4921–4928, http://dx.doi.org/10.1021/ac980573i. [6] T. Nakayama, Y. Fujiwara, Computer compensation for nuclear magnetic resonance quantitative analysis of trace components, Anal. Chem. 54 (1982) 25–28, http://dx.doi.org/10.1021/ac00238a010. [7] A.M. Wolters, D.A. Jayawickrama, C.K. Larive, J.V. Sweedler, Insights into the cITP process using on-line NMR spectroscopy, Anal. Chem. 74 (2002) 4191– 4197, http://dx.doi.org/10.1021/ac025585p. [8] M. Maiwald, H.H. Fischer, Y.-K. Kim, H. Hasse, Quantitative on-line highresolution NMR spectroscopy in process engineering applications, Anal. Bioanal. Chem. 375 (2003) 1111–1115, http://dx.doi.org/10.1007/s00216002-1723-y. [9] M.A. Bernstein, M. Stefinovic, C.J. Sleigh, Optimising reaction performance in the pharmaceutical industry by monitoring with NMR, Magn. Res. Chem. 45 (2007) 564–571, http://dx.doi.org/10.1002/mrc.2007.abs. [10] M. Khajeh, A. Botana, M.A. Bernstein, M. Nilsson, G.A. Morris, Reaction kinetics studied using diffusion-ordered spectroscopy and multiway chemometrics, Anal. Chem. 82 (2010) 2102–2108, http://dx.doi.org/10.1021/ac100110m. [11] D.A. Foley, C.W. Doecke, J.Y. Buser, J.M. Merritt, L. Murphy, M. Kissane, S.G. Collins, A.R. Maguire, A. Kaerner, ReactNMR and ReactIR as reaction monitoring and mechanistic elucidation tools: the NCS mediated cascade reaction of a-thioamides to a-thio-b-chloroacrylamides, J. Org. Chem. Am. Chem. Soc. 76 (2011) 9630–9640. http://dx.doi.org/10.1021/jo201212p. [12] J.M. Merritt, J.Y. Buser, A.N. Campbell, J.W. Fennell, N.J. Kallman, T.M. Koenig, H. Moursy, M.A. Pietz, N. Scully, U.K. Singh, Use of modeling and process analytical technologies in the design of a catalytic amination reaction: understanding oxygen sensitivity at the lab and manufacturing scales, Org.

[24]

[25]

[26]

[27] [28]

[29]

[30] [31]

[32] [33] [34]

[35]

[36]

[37]

[38]

Process Res. Dev. Am. Chem. Soc. 18 (2013) 246–256. http://dx.doi.org/ 10.1021/op400226m. F. Dalitz, M. Cudaj, M. Maiwald, G. Guthausen, Process and reaction monitoring by low-field NMR spectroscopy, Prog. Nucl. Magn. Reson. Spectrosc. 60 (2012) 52–70, http://dx.doi.org/10.1016/j.pnmrs.2011.11.003. S.K. Küster, E. Danieli, B. Blümich, F. Casanova, High-resolution NMR spectroscopy under the fume hood, Phys. Chem. Chem. Phys. 13 (2011) 13172–13176, http://dx.doi.org/10.1039/C1CP21180C. S.S. Zalesskiy, E. Danieli, B. Blümich, V.P. Ananikov, Miniaturization of NMR systems: desktop spectrometers, microcoil spectroscopy, and ‘‘NMR on a Chip’’ for chemistry, Biochem. Ind. Chem. Rev. 114 (2014) 5641–5694, http:// dx.doi.org/10.1021/cr400063g. F. Dalitz, M. Maiwald, G. Guthausen, Considerations on the design of flow cells in by-pass systems for process analytical applications and its influence on the flow profile using NMR and CFD, Chem. Eng. Sci. 75 (2012) 318–326, http:// dx.doi.org/10.1016/j.ces.2012.03.042. F. Dalitz, L. Kreckel, M. Maiwald, G. Guthausen, Quantitative mediumresolution NMR spectroscopy under non-equilibrium conditions, studied on the example of an esterification reaction, Appl. Magn. Reson. 45 (2014) 411– 425, http://dx.doi.org/10.1007/s00723-014-0522-x. G. Guthausen, A. Garnier, R. Reimert, Investigation of hydrogenation of toluene to methylcyclohexane in a trickle bed reactor by low-field nuclear magnetic resonance spectroscopy, Appl. Spectrosc. 63 (2009) 1121–1127, http:// dx.doi.org/10.1366/000370209789553156. A. Nordon, C.A. McGill, D. Littlejohn, At-line process analysis by low-field medium-resolution NMR spectrometry, Spectrosc. Eur. 13 (2001) 10–13. DOI: –. A. Nordon, C.A. McGill, D. Littlejohn, Process NMR spectrometry, Analyst 126 (2001) 260–272, http://dx.doi.org/10.1039/B009293M. C.A. McGill, A. Nordon, D. Littlejohn, Comparison of in-line NIR, Raman and UV–visible spectrometries, and at-line NMR spectrometry for the monitoring of an esterification reaction, Analyst 127 (2002) 287–292, http://dx.doi.org/ 10.1039/B106889J. A. Nordon, A. Diez-Lazaro, C.W.L. Wong, C.A. McGill, D. Littlejohn, M. Weerasinghe, D.A. Mamman, M.L. Hitchman, J. Wilkie, Consideration of some sampling problems in the on-line analysis of batch processes by low-field NMR spectrometry, Analyst 133 (2008) 339–347, http://dx.doi.org/10.1039/ b714266h. E. Kriesten, F. Alsmeyer, A. Bardow, W. Marquardt, Fully automated indirect hard modeling of mixture spectra, Chemometr. Intell. Lab. 91 (2008) 181–193, http://dx.doi.org/10.1016/j.chemolab.2007.11.004. E. Kriesten, D. Mayer, F. Alsmeyer, C.B. Minnich, L. Greiner, W. Marquardt, Identification of unknown pure component spectra by indirect hard modeling, Chemometr. Intell. Lab. 93 (2008) 108–119, http://dx.doi.org/10.1016/ j.chemolab.2008.05.002. T.M. Alam, M.K. Alam, Chemometric analysis of NMR spectroscopy data: a review, Ann. Rep. NMR Spectrosc. 54 (2005) 41–80, http://dx.doi.org/10.1016/ S0066-4103(04)54002-4. A. Nordon, P.J. Gemperline, C.A. McGill, D. Littlejohn, Quantitative analysis of low-field NMR signals in the time domain, Anal. Chem. 73 (2001) 4286–4294, http://dx.doi.org/10.1021/ac0102866. R. Marbach, Multivariate calibration, selectivity, and the SBC method, Chem. Ing. Tech. 82 (2010) 453–466, http://dx.doi.org/10.1002/cite.200900174. R.K. Harris, E.D. Becker, S.M. Cabral de Menezes, et al., NMR nomenclature: nuclear spin properties and conventions for chemical shifts: IUPAC recommendations 2001, Pure Appl. Chem. 73 (2001) 1795–1818. http:// dx.doi.org/10.1351/pac200173111795. W.S. Brey, M.L. Brey, Flurine-19 NMR, in: R.K. Harris, R.E. Wasylishen (Eds.), Encyclopedia of NMR, second ed., Wiley, Chichester, 2012, pp. 1608–1619. http://dx.doi.org/10.1002/9780470034590. W.R. Dolbier Jr., Guide to Fluorine NMR for Organic Chemists, Wiley, Chichester, 2009. R.L. Haner, W. Llanos, L. Mueller, Small volume flow probe for automated direct-injection NMR analysis: design and performance, J. Magn. Reson. 143 (2000) 69–78, http://dx.doi.org/10.1006/jmre.1999.1983. http://www.magritek.com/products/spinsolve (accessed 23.07.14). Magritek GmbH, personal communication. L. Chen, Z. Wenig, L. Goh, et al., An efficient algorithm for automatic phase correction of NMR spectra based on entropy minimization, J. Magn. Reson. 158 (2002) 164–168, http://dx.doi.org/10.1016/S1090-7807(02)00069-1. T.D.W. Claridge, J.E. Baldwin, F.R.S. Williams, R.M. Williams (Eds.), HighResolution NMR Techniques in Organic Chemistry, Pergamon, Bd. 19. Amsterdam, 1999. ISBN: 0-08-042799-5. J. Mocak, A.M. Bond, S. Mitchell, G. Scollary, A statistical overview of standard (IUPAC and ACS) and new procedures for determining the limits of detection and quantification: application to voltammetric and stripping techniques, Pure Appl. Chem. 69 (1997) 297–328, http://dx.doi.org/10.1351/pac199769020297. DIN 32645:2008-11, Deutsche Industrienorm: Chemical Analysis – Decision Limit, Detection Limit and Determination Limit under Repeatability Conditions – Terms, Methods, Evaluation, Beuth Verlag, Berlin, 2008. C. Amman, P. Meier, A.E. Merbach, A simple multinuclear NMR thermometer, J. Magn. Reson. 46 (1982) 319–321, http://dx.doi.org/10.1016/00222364(82)90147-0.