Anal. Chem. 2003, 75, 622-626
Reaction Monitoring of Aliphatic Amines in Supercritical Carbon Dioxide by Proton Nuclear Magnetic Resonance Spectroscopy and Implications for Supercritical Fluid Chromatography Holger Fischer,† Olle Gyllenhaal,‡ Jo 1 rgen Vessman,‡ and Klaus Albert*,†
Institut fu¨r Organische Chemie, Universita¨t Tubringen, Auf der Morgenstelle 18, D-72076 Tu¨bingen, Germany, and AstraZeneca R&D Mo¨lndal, S-431 83 Mo¨lndal, Sweden
In the recent years, it has repeatedly been stated that amines react with CO2 and can therefore not be chromatographed under supercritical conditions with CO2. The aim of the present work is to elucidate the structural requirements and conditions that can lead to the reaction of an amine analyte with CO2 and, if this occurs, the structure of the formed product. The use of on-line nuclear magnetic resonance (NMR) spectroscopy with a flow probe for supercritical fluid chromatography (SFC) enables the investigation of these unstable analytes in supercritical mediums. Several alkyl-substituted secondary benzylamines and some primary aromatic amines were dissolved in supercritical CO2 and investigated by employing on-line SFC-1H NMR spectroscopy. It was found that the condition of carbamic acid formation depends on the steric properties of the substituents of the amine. A 2-isopropylamino alcohol compound, metoprolol, was also investigated with the setup. No carbamic acid could be detected with the present conditions. For many years there has been an ongoing debate in the chromatographic community using supercritical fluid chromatography (SFC) about the reaction of amines with supercritical CO2 (scCO2) and whether amines can be separated under SFC conditions with CO2 without complications.1 A lot of investigations of the separation behavior of aliphatic amines in SFC have been performed in capillaries and in packed columns. In capillary SFC, aliphatic amines showed a similar retention behavior and peak shape.2,3 Packed column SFC was used for the analysis of metoprolol, an amino alcohol, and analogues with 10% methanol in scCO2 and diol-silica as support. Triethylamine was used as a basic additive to ensure good peak symmetry.4 With a similar mobile-phase composition and porous graphitic carbon as support, * Corresponding author. E-mail:
[email protected]. † Universita ¨t Tubringen. ‡ AstraZeneca R&D Mo ¨lndal. (1) Fields, S. M.; Grolimund, K. J. High Resolut. Chromatogr. Commun. 1988, 11, 727-729. (2) Gyllenhaal, O.; Vessman, J. J. Chromatogr. 1990, 516, 415-426. (3) Baastoe, M. B.; Lundaners, E. J. Chromatogr. 1991, 558, 458-463. (4) Gyllenhaal, O.; Vessman, J. J. Chromatogr. 1999, 839, 141-148.
622 Analytical Chemistry, Vol. 75, No. 3, February 1, 2003
a baseline elevation was observed for the analyte peak metoprolol. The size of the area of this part of the chromatogram correlated with the temperature of the column oven. Chiral separations of similar analytes were reported and proposed to be enhanced by the aid of CO2 and a cyclic transient complex formed with CO2 and the amino alcohol.5 These suggestions were supported by off-line nuclear magnetic resonance (NMR) experiments by bubbling CO2 through an analyte solution in the NMR probe.6 With the increasing use of semipreparative SFC, artifacts have been encountered when isopropylamine is used as a basic additive and the collected fractions are evaporated to dryness.7 The principal reaction of amines with CO2 to carbamic acids is well known and therefore used for absorption of acid gases.8 A mixture of piperazine and methyldiethanolamine patented by BASF9 is widely used in gas absorbers, where the secondary amine piperazine reacts with CO2 to form the corresponding carbamic acid and derivatives. The stability of aminopropyl silica in scCO2 was studied by Bayer et al.10 At 150 bar, 100 °C, and 3-15 h reaction time, the formation of carbamic groups was proposed based on solid-state NMR spectroscopy studies. The new phase showed different and improved selectivity for the analysis of unsaturated triglycerides. Pinkston and Baker developed an ion spray interface for capillary SFC-MS. But due to the presence of CO2 in the ion source, the M + 44 ion observed from didodecylamine could not be unambiguously assigned to corresponding carbamic acid dervatives.11 Leitner et al.12 reported the formation of a white insoluble solid in the iridium-catalyzed enantioselective hydration of imines in (5) Siret, L.; Bargman, N.; Tambute´, A.; Caude, M. Chirality 1992, 4, 252262. (6) Bargman-Leyder, N.; Sella, C.; Bauer, D.; Tambute´, A.; Caude, M. Anal. Chem. 1995, 67, 952-958. (7) Villeneuve, M.; Lefler, J. L. Lecture 3, 10th Int. Symp. SFC SFE; Myrtle Beach, SC, August 19-22, 2001. (8) Bishnoi, S.; Rochelle, G. T. Chem. Eng. Sci. 2000, 55, 5531-5543. (9) Appl, M.; Wagner, U.; Henrici, H. J.; Kuessner, K.; Volkamer, F.; Fuerst, E. U.S. Patent 4336233, 1982. (10) Zhang, S.; Nicholson, G.; Schindler, B.; Bayer, E. 18th Int. Symp. Capillary Chromatogr.; Riva del Garda, Italy, 1996; pp 1785-1790. (11) Pinkston, J. D.; Baker, T. R. Rapid Commun. Mass Spectrom. 1995, 9, 1087. 10.1021/ac020527p CCC: $25.00
© 2003 American Chemical Society Published on Web 01/01/2003
supercritical CO2 dependent upon the substituents of the imine. Due to the insolubility of the product, it could not be verified in the NMR high-pressure cell. The two investigated substances were N-phenyl-1-phenylethylamine and N-benzyl-1-phenylethylamine, in which only the last formed a white solid. Here the formation of a carbamic acid derivative was supposed. An advantage can be that amines are protected as carbamic acids and can later be recovered in their original state. A recent paper from this group discusses the olefin metathesis in supercritical carbon dioxide.13 The systems were investigated by 1H NMR spectroscopy, where signal shifting of amine protons occurred, which can only be explained by the formation of a carbamic acid. The formation itself is described as a reversible process. For monitoring of complex reaction processes and even kinetics, on-line NMR spectroscopy is an excellent method due to the fast and easy transfer to the detector and the high information content provided by NMR spectroscopy.14,15 Whereas in the case of common liquids solvent suppression techniques are required, supercritical CO2 is attractive as a medium for NMR studies due to the absence of absorbing protons.16-21 On the other hand, NMR spectra in supercritical fluids may be dependent on the density of the supercritical medium, which is given by the temperature and pressure of the medium. Furthermore, the sensitivity of the technique is moderate and not sufficient for analysis of trace eluents from SFC. We therefore arranged a stopflow NMR system that made it possible to load adequate amounts of analyte into the SF system and to collect data during a sufficiently long time period. Our present work has the aim to elucidate the structural requirements and conditions that can lead to the reaction of an amine with supercritical CO2 to form carbamic acid and, if this occurs, the structure of the formed product. The supposed reaction scheme is depicted in Figure 1. EXPERIMENTAL SECTION Apparatus. For the NMR measurements, an inverse 1H/13C flow probe (Bruker, Rheinstetten, Germany) equipped with a pressure-stable 120-µL sapphire-titanium flow cell, especially developed for SFC NMR purposes,17-21 was used (Figure 2). The probe can be thermostated with heated air (or nitrogen) up to 353 K and can be used in a pressure range up to 350 bar. Solvent/ analyte supply is provided through stainless steel capillaries to the titanium-sapphire flow cell used for detection. The lock (12) Kainz, S.; Brinkmann, A.; Leitner, W.; Pfaltz, A. J. Am. Chem. Soc. 1999, 121, 6421-6429. (13) Fu ¨ rstner, A.; Ackermann, L.; Beck, K.; Hori, H.; Kock, D.; Langemann, K.; Liebl, M.; Six, Ch.; Leitner, W. J. Am. Chem. Soc. 2001, 123, 9000-9006. (14) Albert, K. Habilitationsschrift, Tu ¨ bingen University, 1988. (15) Albert, K., Ed. On-line LC NMR and Related Techniques; Wiley: Chichester, 2002. (16) Allen, L. A.; Glass, T. E.; Dorn, H. C. Anal. Chem. 1988, 60, 390-394. (17) Albert, K.; Braumann, U.; Tseng, L.-H.; Nicholson, G.; Bayer, E.; Spraul, M.; Hofmann, M.; Dowle, C.; Chippendale, M. Anal. Chem. 1994, 66, 30423046. (18) Braumann, U.; Ha¨ndel, H.; Albert, K.; Ecker, R.; Spraul, M. Anal. Chem. 1995, 67, 930-935. (19) Albert, K.; Braumann, U.; Streck, R.; Spraul, M.; Ecker, R. Fresenius J. Anal. Chem. 1995, 352, 521-528. (20) Braumann, U.; Ha¨ndel, H.; Strohschein, S.; Spraul, M.; Krack, G.; Ecker, R.; Albert, K. J. Chromatogr., A 1997, 761, 336-340. (21) Albert, K. J. Chromatogr., A 1997, 785, 65-83.
Figure 1. Proposed reaction scheme for a secondary benzylamine with supercritical carbon dioxide leading to a carbamic acid.
Figure 2. General layout of the stop-flow SFC NMR system: (a) modifier pump, (b) CO2 pump, (c) CO2 cylinder, (d) cooling bath, (e) GC oven, (f) injection valve, (g) UV detector, (h) Bruker ARX 400 NMR spectrometer, (i) back pressure regulator, (j) control unit, and (k) stop-flow valve.
capillary installed in the center of the flow cell can be filled with deuterated solvent and used for field-frequency stabilization and referencing. Spectra were recorded with a 9.4-T Bruker ARX 400 NMR spectrometer (Bruker). The experimental setup is shown in Figure 3. The SFC piston pump was from Gilson (model M308, Abimed AG, Langenfeld, Germany) and was equipped for SFC purposes with a cooled pump head which delivers liquid CO2 with a flow of up to 10 mL/min and a pressure of up to 400 bar. Cylinders with grade 5.5 CO2 and dipper tubes were obtained from Messer-Griesheim AG (Krefeld, Germany). Modifier, when needed, was delivered by a Bischoff HPLC pump (Leonberg, Germany) and added to the main CO2 stream. The pumps were manually regulated. A Dani 3800 GC oven was used to establish the subor supercritical conditions. Herein 1.5 m of stainless steel loops for temperature equilibration and mixing chamber were included. A six-port injection valve equipped with a 20-µL loop was used to inject samples. Analyte signals were detected at 254 nm with a UV detector (UV VWD 1050, Hewlett-Packard, Waldbronn, Germany) from which the distance to the NMR probe was bridged with 3 m of stainless steel capillaries (0.7-mm i.d.). A Valco valve was inserted halfway down this connection, which made it possible to stop the flow to the NMR probe. The pressure of the system Analytical Chemistry, Vol. 75, No. 3, February 1, 2003
623
right position in the NMR detection cell. For each benzylamine spectrum, four transients were recorded with 32K data points and a spectral width of 6000 Hz. The metoprolol spectrum was recorded with 32 transients, 32K data points, and a spectral width of 7352 Hz. The SFC NMR probe temperature was set to 323 K. All spectra were processed with XWINNMR-Software version 6.0 (Bruker-Franzen).
Figure 3. View of the 1H/13C inverse SFC NMR probe with titanium-sapphire flow cell and lock capillary.
was maintained with a Jasco model 880-81 back pressure regulator (Tokyo, Japan). The UV detector signal was displayed using ChromStar software version 3.3 (Bruker-Franzen, Bremen, Germany), and the back pressure regulator was controlled by the same software using an interface LC 243 M analog pump (Bruker Franzen Analytik, Bremen Germany). Cooling of the CO2 pump head to 2 °C was accomplished by a Haake thermostat G (Karlsruhe, Germany). Reagents and Chemicals. N-Methyl-N-benzylamine, N-isopropyl-N-benzylamine, N-tert-butyl-N-benzylamine, N,N-dimethylN-benzylamine, N-benzylamine, and phenylethylamine were purchased from Fluka (Neu-Ulm, Germany), and deuterated solvents were obtained from Merck (Darmstadt, Germany). Metoprolol tartrate was obtained from the Department of Medicinal Chemistry (AstraZeneca R & D Mo¨lndal, Sweden). It was converted to the free base through extraction with dichloromethane in the presence of a weak sodium carbonate buffer. The organic phase was evaporated to dryness in a stream of dry nitrogen. All solvents were of analytical grade quality. Method. The analytes were dissolved in an appropriate deuterated solvent (10-20%) and submitted to SFC NMR, where they were recorded in sub- or supercritical state. The UV detector trace was observed as well as the NMR signals. Because supercritical CO2 does not contain any proton signals, the change of the signal intensity of the free induction decay (FID) can be used to trap the analyte in the flow cell. By constantly watching the FID of the NMR signal, the advance of the analyte in the NMR probe was awaited. Then the valve downstream to the UV detector was closed to seal the NMR probe and trap the analyte at the 624 Analytical Chemistry, Vol. 75, No. 3, February 1, 2003
RESULTS AND DISCUSSION Referencing of NMR spectra acquired in the sub-/supercritical state is difficult, due to different susceptibilities in the flow tube (sub-/supercritical environment) and in the lock capillary (liquid environment). So only chemical shifts of spectra acquired under the same supercritical conditions are comparable in the common NMR sense. We observed a shift on the δ-scale to higher frequencies of about 1-1.5 ppm going from liquid to sub-/ supercritical conditions. The N-benzylamine derivatives were dissolved in deuterated dichloromethane and injected via the valve to the SFC NMR system. Residence times of analytes after injection were in the 2-min range. Figure 4 shows four spectra recorded in scCO2 at 180 bar and 353 K oven temperature for each derivative. The NMR detection temperature was 323 K. N-tert-Butyl-N-benzylamine, where the amine proton is strongly shielded by the sterical demanding tert-butyl group, shows no formation of carbamic acid. The signal of the amine proton appears at 1.7 ppm (see arrow). In the same way, the amine proton of N-isopropyl-N-benzylamine (see arrow) is shielded to 3.5 ppm by the isopropyl group. The large chemical shift difference of tertbutyl- or isopropyl-substituted amine indicates medium interactions of the amine proton with the solvent environment. The spectrum of N-methyl-N-benzylamine, on the other hand, shows a broad signal at 11.8 ppm (see arrow) indicating a carbonic acid proton, the proton of N-methyl-N-benzylcarbamic acid. Signal from the remaining amine protons, expected between 1 and 6 ppm, could not be detected. It seems that the whole amount of amine has reacted to form carbamic acid. This is in line with one of the early experiments with other less shielded secondary amines and primary amines. Here, the primary amine N-benzylamine was dissolved in dichloromethane and injected into the SFC NMR system. Instead of a transportation of the analyte to the NMR detection cell, immediately after injection the whole system was blocked, presumably by products formed from the analyte and scCO2, which could be N-benzylcarbamic acid or more probably dibenzylurea. Due to the insolubility of the product of this reactive amine in CO2, no spectrum could be obtained. The experiment was repeated several times at different temperature and pressure and also with modifier solvent (10% methanol), but with no sign of significant NMR signals that could be recorded. A similar result was also obtained with phenylethylamine. These observations are consistent with the results published by Leitner et al.12 Figure 5 depicts the carbamic acid formation of the reaction of scCO2 with N-methyl-N-benzylamine at different pressures from 80 to 200 bar. The carbamic acid proton appears at ∼11.5 ppm; see arrows. It can be seen that the reaction to carbamic acid occurred over the whole pressure range. In the same way, the reaction of scCO2 with N-isopropyl-N-benzylamine was investigated. But in contrast to N-methyl-N-benzylamine, no reaction took
Figure 4. NMR spectra of benzylamines recorded in stop-flow SFC mode. Conditions: flow rate of neat carbon dioxide, 10 mL/min during injection; back pressure, 180 bar; oven temperature, 353 K; NMR temperature, 323 K. The arrows indicate the protons of the amine group.
place over the observed pressure range. Only a shift of the amine proton was observed due to the changing solvating power of the supercritical CO2 medium. The basicity of the amines increases in the substituent order methyl, isopropyl, tert-butyl group. In the same way, the stability of the theoretically formed carbamic acid should be increased through the electron donor properties of the substituents. In case that the carbamic acid reaction would become much faster for the more shielded amines and therefore not be observable on the NMR time scale, the line width of the amine signal should change with temperature. In a further experiment, N-isopropylN-benzylamine was measured at two different temperatures (300 and 323 K), but at the investigated temperatures, no line broadening of the amine signal was observed. Therefore, chemical exchange processes were not considered. In Figure 6, two spectra of metoprolol, recorded in chloroform (top) and in the supercritical state (bottom), are depicted. The chemical shift of the spectrum in scCO2 was adjusted to the spectrum recorded for comparison. The spectrum in chloroform was acquired at atmospheric pressure at 293 K. In the spectrum acquired in scCO2 at 100 bar, 353 K oven temperature, and 323 K NMR temperature, small chemical shift changes can be seen compared to the spectrum of the sample dissolved in chloroform. Especially protons in the neighborhood of polar groups (signals 2, 4, 5, 7, 11, and 12) show slightly different chemical shifts. Polar groups are solvated by scCO2 and are most affected by the solvating properties of scCO2. The formation of carbamic acid is not observed in the supercritical spectrum. So the beneficial steric
Figure 5. NMR spectra of N-isopropyl-N-benzylamine recorded in stop-flow SFC mode. Conditions: flow rate of neat carbon dioxide, 10 mL/min during injection; oven temperature, 353 K; NMR temperature, 323 K. The arrows indicate the protons of the amine group.
Analytical Chemistry, Vol. 75, No. 3, February 1, 2003
625
Figure 6. NMR spectra of metoprolol recorded in stop-flow SFC mode and under liquid conditions (for conditions, see text).
protection of the isopropyl side chain prevents the carbamic acid formation of the β-blocker drug metoprolol. Attempts to study the less shielded analogue of metoprolol, with N-methyl instead of N-isopropyl, were futile. The analyte apparently precipitated when the sample solvent was mixed with scCO2. This was supported by the observation of strong signals from the UV detector when 20% methanol in CO2 was introduced. However, the NMR signals from the modifier solvent obstructed any attempts to monitor the signals from the analyte itself. CONCLUSIONS The use of on-line NMR spectroscopy employing a SFC probe enables the investigation of analytes and even reactions in the supercritical medium. The conditions when amines react with supercritical CO2 in the pressure range of 80-200 bar depend on the steric properties of the amine substituents. Primary amines react supposedly immediately to carbamic acids or urea derivatives. But this could not be proved up to now with the SFC NMR technique, because primary amines are rather polar and pure scCO2 is not sufficiently strong as solvent. Secondary amines react to carbamic acids, when they are substituted with smaller groups, such as methyl or ethyl groups. More sterically crowded groups, suxh as isopropyl or tert-butyl groups, can shield the amine proton 626
Analytical Chemistry, Vol. 75, No. 3, February 1, 2003
and prevent it from reacting to form carbamic acid and derivatives. Steric protection of the isopropyl side chain is also observed for the β-blocker drug metoprolol. The reason for the higher baseline in the SFC separation of metoprolol derivatives could not be found. Maybe the porous graphitic carbon support has catalytic activity that promotes interaction. It would be attractive if deuterated polar modifiers could be used to enhance the solubility of the analytes in the fluid. However, most available deuterated polar solvents have deuteriums that are likely to be exchanged with the protons of the amines of interest. Overall, the present investigation provides a clear understanding of the conditions when amines react with supercritical CO2. The steric properties of the amine substituent direct the reaction pathway. ACKNOWLEDGMENT The authors gratefully acknowledge Dr. Ulrich Braumann for technical support (Bruker BioSpin; Rheinstetten, Germany). Received for review August 15, 2002. Accepted November 11, 2002. AC020527P
