Use of Raman spectroscopy to characterize hydrogenation reactions

Article abstract of DOI:10.1021/op0600355

Raman spectroscopy was used to characterize hydrogenation reactions involving single-step and two-step processes. The Raman technique was shown to be well-suited for endpoint determination as well as process optimization. In this investigation, hydrogenation of cyclohexene to produce cyclohexane was used as a model system. Conditions were varied to determine the effect of catalyst loading, solvent ratios, and reactant concentrations. Four catalysts were evaluated. The kinetic profiles of each reaction process were determined for each of the catalysts. In one case, a side reaction leading to an intermediate was observed for the hydrogenation reaction when run under hydrogen-starved conditions. After these cyclohexene hydrogenations were characterized, Raman spectroscopy was applied to the conversion of carvone to tetrahydrocarvone and the hydrogenation of 2-(4-hydroxyphenyl) propionate. Raman was used to characterize the kinetics of these reactions and was also used to prove that two-step hydrogenation mechanisms occurred in each. Raman was shown to be useful for process understanding, process optimization, process monitoring, and endpoint determination. Accomplishment of these goals leads to better process controls upon transfer of the procedure to a process environment. This ultimately leads, in turn, to the mitigation of risk of making out-of-specification product in manufacturing.

Full text of DOI:10.1021/op0600355

Organic Process Research & Development 2006, 10, 927−933
Use of Raman Spectroscopy to Characterize Hydrogenation Reactions
Venkat S. Tumuluri,† Mark S. Kemper,‡ Anjaneyulu Sheri,§ Seoung-Ryoung Choi,§ Ian R. Lewis,‡
Mitchell A. Avery,§ and Bonnie A. Avery*,†
Department of Pharmaceutics, UniVersity of Mississippi, UniVersity, Mississippi 38677, U.S.A., Kaiser Optical Systems,
Ann Arbor, Michigan, U.S.A., and Department of Medicinal Chemistry, UniVersity of Mississippi,
UniVersity, Mississippi 38677, U.S.A.
Abstract:
conversions include transformation of nitro compounds to
amines, functional group deprotection procedures, and other
processes.3,4
Raman spectroscopy was used to characterize hydrogenation
reactions involving single-step and two-step processes. The
Raman technique was shown to be well-suited for endpoint
determination as well as process optimization. In this investiga-
tion, hydrogenation of cyclohexene to produce cyclohexane was
used as a model system. Conditions were varied to determine
the effect of catalyst loading, solvent ratios, and reactant
concentrations. Four catalysts were evaluated. The kinetic
profiles of each reaction process were determined for each of
the catalysts. In one case, a side reaction leading to an
intermediate was observed for the hydrogenation reaction when
run under hydrogen-starved conditions. After these cyclohexene
hydrogenations were characterized, Raman spectroscopy was
applied to the conversion of carvone to tetrahydrocarvone and
the hydrogenation of 2-(4-hydroxyphenyl) propionate. Raman
was used to characterize the kinetics of these reactions and was
also used to prove that two-step hydrogenation mechanisms
occurred in each. Raman was shown to be useful for process
understanding, process optimization, process monitoring, and
endpoint determination. Accomplishment of these goals leads
to better process controls upon transfer of the procedure to a
process environment. This ultimately leads, in turn, to the
mitigation of risk of making out-of-specification product in
manufacturing.
With any synthetic reaction carried out on an industrial
scale, it is useful to perform in-process checks in order to
ascertain the progress of the reaction and know when the
endpoint has been reached. In the case of hydrogenations,
many of the reactions are inherently highly exothermic.
Ineffective heat removal can result in thermal runaway
leading to an explosion.5 In these cases, monitoring of the
progress of the reaction to avoid such events is essential.
Real-time feedback through a monitoring scheme that allows
for process adjustments is preferred.
In the pharmaceutical industry, active ingredient synthesis
is often referred to as primary processing. The Process
Analytical Technology (PAT) initiative proposed by the
United States Food and Drug Administration (USFDA) has
produced much activity in the realm of process chemistry
amongst pharmaceutical companies.6 The premise of this is
that better understanding necessarily leads to the ability to
impose better controls. In turn, a better control regime
mitigates the risk of making product that is out-of-specifica-
tion. Such a result provides benefits for both the consumer
and the manufacturer. In the case of the manufacturer,
benefits are realized by a reduction of scrapped materials
and wasted production capacity as well as the reduction or
elimination of costly product recalls. For the consumer, the
benefit is the assurance that the product they use was
manufactured to the highest standards for purity, safety, and
efficacy as the processes are monitored in real time.
Vibrational spectroscopic techniques such as Fourier
transform infrared (FT-IR) spectroscopy, near-infrared (NIR)
spectroscopy and Raman spectroscopy are extremely valuable
tools for many process situations. Various groups have
experimented with NIR in the area of reaction monitoring.
For example, hydrogenation of itaconic to methyl succinic
acid was carried out by Wood et al.7 Fermentation was
monitored by Lendl et al.8 Pharmaceutical companies have
Introduction
Hydrogenation reactions are ubiquitous in chemical
manufacturing processes. In the pharmaceutical industry, they
are quite commonly used in multistep procedures employed
in the synthesis of active pharmaceutical ingredients (APIs).1
In fact, hydrogenation reactions account for about 10% to
20% of all reactions employed by API manufacturers.1 This
class of reactions is also used in the chemical industry for
general synthetic procedures.
Hydrogenations can be performed to accomplish several
types of chemical transformations, but they are often used
to convert olefinic bonds to aliphatic bonds.2 Other common
(2) Dyson, P. J.; Zhao, D. Hydrogenation. Multiphase Homogeneous Catalysis
2005, 2, 494-511.
* To whom correspondence should be addressed. E-mail: bavery@olemiss.edu.
Dr. Bonnie A. Avery, Associate Professor, 107 Faser Hall, Department of
Pharmaceutics, University of Mississippi, University, MS 38677. Telephone:
662-915-5163. Fax: 662-915-1177.
(3) An Leeuwen, P. W. N. M.; Van Koten, G. Catalysis 1993, 79, 199-248.
(4) Jacobson, S. E. Catalytic hydrogenolysis of organic thiocyanates and
disulfides to thiols. PCT Int. Appl. 1997, 17 pp.
(5) Rylander P. N. Catalysts, Reactors, and Reaction Parameters. Hydrogena-
tion methods; Academic Press: Orlando, FL, 1985.
(6) Guidance for Industry PAT sA Framework for Innovative Pharmaceutical
Development, Manufacturing, and Quality Assurance. Center for Drug
Evaluation and Research, Division of Food and Drug Administration. 2003
(7) Wood, J.; Turner, P. Appl. Spectrosc. 2003, 57 (3), 293-8.
† Department of Pharmaceutics, University of Mississippi.
‡ Kaiser Optical Systems.
§ Department of Medicinal Chemistry, University of Mississippi.
(1) Pavlenko, N. V.; Tripolskii, A. I. L. V. Res. Chem. Kinetics 1995, 207-
258.
10.1021/op0600355 CCC: $33.50 © 2006 American Chemical Society
Published on Web 07/14/2006
Vol. 10, No. 5, 2006 / Organic Process Research & Development
•
927

been actively working on using NIR to monitor syntheses9
and also reaction completion in a closed loop hydrogenator.10
In relation to NIR, Raman spectroscopy is fairly new to the
area of reaction monitoring and particularly a convenient tool
for real time process analysis.11,12 Because Raman spectrom-
eters such as NIR can be coupled to fiber optics, in situ
sampling is possible. Raman is suitable for process monitor-
ing for many reasons but chiefly because remote analysis is
possible. Researchers have implemented this technology to
monitor chemical reactions13 as well as the synthesis14,15 of
compounds. Some more successful applications for the use
of Raman in process analyses have been reported in the
literature.16-20
For hydrogenation processes, Raman spectroscopy can be
a very valuable tool to characterize kinetics and endpoints.
This is especially true for conversion of olefins because
Raman is very sensitive to the CdC double bond stretching.21
Often such reactions can be followed by monitoring the
disappearance of this band in the 1560-1680 cm-1 shift
region. Raman spectroscopy has historically been applied
for many applications involving olefins. It has been dem-
onstrated as the preferred spectroscopic method for studying
the microstructure of CdC bonds in polybutadiene.22 In
addition, the technique has been demonstrated as a means
for monitoring the hydrogenation of polybutadiene using a
number of catalysts.23
Scheme 1. Conversion of cyclohexene to cyclohexane
spectral and corresponding intensity profiles by matrix
decomposition methods. These initial estimates are neither
physically meaningful nor typically the pure spectra of the
components present in the sample. However, they can be
transformed into physically meaningful solutions that rep-
resent the spectra of pure components by self-modeling
methods.26 The ability of factor analysis to reduce and resolve
data matrices of chemical mixtures into pure component
spectra and individual concentration profiles has been
accomplished in several cases. These mixture analyses
provide an estimation of the number of chemical species
present, the identification of these species, and the determi-
nation of their concentrations. A very successful technique
in analyzing vibrational spectroscopy data files is multivariate
curve resolution (MCR), a group of techniques which intend
the recovery of response profiles (spectra, pH profiles, time
profiles, elution profiles) of the components in an unresolved
mixture obtained in evolutionary processes when no prior
information is available about the nature and composition
of these mixtures.
The work described below represents an investigation of
the capability of Raman spectroscopy to monitor hydrogena-
tion reactions of interest to the pharmaceutical community.
The utility of MCR as a data analysis tool was also
demonstrated in this work. This study uses model compounds
to highlight the potential benefits of Raman spectroscopy as
an in situ PAT tool for the study of hydrogenations, which
is an important class of industrial reactions.
For processes involving olefins, regions other than the
CdC stretch can also be monitored depending on the
situation. In some cases, however, distinct bands are difficult
to find. This is especially true when more than one species
are involved in the reaction dynamics, such as the case when
intermediates are produced that are Raman active. In such
cases, curve deconvolution techniques such as Multivariate
Curve Resolution (MCR) can be valuable for the successful
extraction of significant information from Raman data.24-26
This approach begins by generating initial estimates of
Experimental Section
Materials. Platinum, 5 wt % on activated carbon, pal-
ladium, 5 wt % on carbon, rhodium, and Rainey Nickel were
used for the hydrogenation reactions. These catalysts along
with the reactants carvone, cyclohexane, and cyclohexene
were obtained from Sigma-Aldrich Inc. (St. Louis, MO).
The compound 2-(4-hydroxyphenyl) propionate was pre-
pared in the medicinal chemistry department at the University
of Mississippi according to the published procedure.27
Hydrogenation Reactions and Products. Conversion
of Cyclohexene to Cyclohexane. The first model system
explored was that of the simple conversion of cyclohexene
to cyclohexane (Scheme 1). Several catalysts were used for
these reactions including palladium on carbon, platinum on
carbon, rhodium complex, and Rainey nickel. A 100-mL
three-necked flask was used for all reactions. The flask was
covered with Al foil to avoid ambient light interference with
the Raman scatter. One neck was stoppered with a serum
(8) Gunta, M.; Josef, D.; Josefa, R. B.; Erwin, R.; Bernhard, L. Appl. Spectrosc.
2004, 58 (7), 804-810.
(9) Wiss, J.; La¨nzlinger, M.; Wermuth, M. Org. Process Res. DeV. 2005, 9
(3), 365-371.
(10) Ward, H. W., II; Sekulic, S. S.; Wheeler, M. J.; Taber, G.; Urbanski, F. J.;
Sistare, F. E.; Norris, T.; Aldridge, P. K. Appl. Spectrosc. 1998, 52 (1),
17-21.
(11) Lewis, I. R. Process Raman Spectroscopy. In Handbook of Raman
Spectroscopy; Lewis, I. R., Edwards, H. G. M., Eds.; Marcel Dekker: New
York, 2001.
(12) Adar, F.; Geiger, R.; Noonan, J. Appl. Spectrosc. ReV. 1997, 32, 45-101.
(13) Fletcher, P. D.; Haswell, S. J.; Zhang, X. Electrophoresis 2003, 24 (18),
3239-45.
(14) Lee, M.; Kim, H.; Rhee, H.; Choo, J. Bull. Korean Chem. Soc. 2003, 24
(2), 205.
(15) Svensson, O.; Josefson, M.; Langkilde, F. W. Eur. J. Pharm. Sci. 2000, 11
(2), 141-55.
(16) Gervasio, G. J.; Pelletier, M. J. AT-Process 1997, 3, 7-11.
(17) Lipp, E. D.; Grosse, R. L. Appl. Spectrosc. 1998, 52, 42-46.
(18) Al-Khanbashi, A.; Dhamdhere, M.; Hansen, M. Appl. Spectrosc. ReV. 1998,
33, 115-131.
(19) Bauer, C.; Amram, B.; Agnely, M.; Charmot, D.; Sawatski, J.; Dupuy, D.;
Huvenne, J.-P. Appl. Spectrosc. 2000, 54, 528-535.
(20) Wethman, R.; Ray, C.; Wasylyk, J. Am. Pharm. ReV. 2005, 8 (6), 57-63.
(21) Linn-Vien, D.; Colthup, N. B.; Fateley, W. G.; Grasselli, J. G. The Handbook
of Infrared and Raman Characteristic Frequencies of Organic Molecules;
Academic Press: San Diego, 1991.
(24) Budevska, B. O.; Sum, S. T.; Jones, T. J. Appl. Spectrosc. 2003, 57 (2),
124-131.
(25) Tauler, R.; Smilde, A. K.; Henshaw, J. M.; Burgess, L. W.; Kowalski, B.
R. Anal. Chem. 1994, 66, 3331-3344.
(22) Edwards, H. G. M.; Johnson, A. F.; Lewis, I. R.; McLeod, M. A. Polymer
1993, 34, 3184-3195.
(23) Frankland, J. A.; Edwards, H. G. M.; Johnson, A. F.; Lewis I. R.;
Poshyadida, S. Spetrochim. Acta, Part A 1991, 47, 1511-1524.
(26) Schoonover, J. R.; Zhang, S. L.; Johnston, C. T. J. Raman Spectrosc. 2003,
34, 404-412.
(27) Chin, C. S.; Lee, B.; Moo, J.; Song, J.; Park, Y. Bull. Korean Chem. Soc.
1995, 16 (6), 528-33.
928
•
Vol. 10, No. 5, 2006 / Organic Process Research & Development

Table 1. Conditions used for cyclohexene reactions
Table 2. Conditions used for carvone reactions
amounts
amounts
catalyst
(% w/w)
cyclohexene
(M)
methanol
(mL)
catalyst (Pd/C)
(% w/w)
carvone
(mmol)
methanol
(mL)
entry
entry
1
2
3
4
5
6
7
8
Pd/C 12
Pd/C 8
Pd/C 7
0.48
0.48
0.49
0.48
0.48
0.58
0.58
0.48
0.48
0.71
0.48
20
30
30
30
20
20
25
30
20
20
30
1
2
3
4
3.5
7
21
5
20
10
10
10
30
30
20
30
Pd/C 8
Pt/C 12
Pt/C 12
Pt/C 8
Pt/C 7.75
Rd/C 12
Rd/C 8
Rd/C 6.7
Scheme 3. Hydrogenation of 2-(4-hydroxyphenyl)
propionate
9
10
11
Raman Instrumentation and Data Analysis. Raman
measurements were accomplished using a RamanRxn1
dispersive Raman analyzer (Kaiser Optical Systems, Inc.,
Ann Arbor, MI). The instrument was equipped with either
an Mk II or an MR Probe fiber optic probe head fitted with
a short focal length immersion optic for sampling. The
immersion optic was either 6 in. long with a 1/4-in. diameter
Scheme 2. Conversion of carvone to dihydrocarvone to
tetrahydrocarvone
1
or 12 in. long with a /2-in. diameter. The immersion optic
cap, and the immersion probe was fitted through the cap into
the reaction flask. The other two necks were closed with
serum caps. In general, the system was charged with 0.1 g
of the catalyst and 20 mL of methanol as the solvent. The
system was initially evacuated and then filled with hydrogen
using a balloon. The mixture was stirred magnetically making
sure the catalyst was suspended in methanol and not clumped
to one of the sides of the flask. 0.82 g of cyclohexene was
added after 5 min. The solution was mixed using a magnetic
stir bar. Hydrogen was added continuously to the system
from the balloon throughout the course of the reaction. The
reactions were started at ambient temperature and pressure.
The pressure inside was maintained from the balloon.
Spectral data were acquired in 3-min intervals using 200
exposures and 1 accumulation.
was inserted directly into the reaction vessel for real-time
monitoring. A 785-nm laser with 110 mW of power at the
sample was employed for excitation. Spectra were collected
using 60-s exposures over a spectral range from 150 to 3450
cm-1 at a spectral resolution of approximately 4 cm-1.
HoloReact software (Kaiser Optical Systems) was used
for data analysis, and Excel (Microsoft Corp., Redmond,
WA) was used for simple curve fitting for kinetic models.
Grams software (Thermo Electron, Salem, NH) was used
for some data plots and manipulation. Data pretreatments
included combinations of baseline correction using Pearson’s
method,28 peak normalization, and/or window clipping. For
the cyclohexene conversions, a simple area-under-the-curve
measurement was used for trend plotting. The peak at 802
cm-1 was used to determine cyclohexane appearance, and
the band at 823 cm-1 was used to follow cyclohexene
disappearance. Quantitative conversions for kinetic modeling
were based on the known starting concentrations of cyclo-
hexene in each case.
The effect of catalyst and solvent on the reaction rates
and the rate constants were also examined. Table 1 indicates
the concentrations of the catalyst, cyclohexene, and the
amount of methanol used in the reactions.
Reduction of Carvone to Tetrahydrocarvone. The
applicability of Raman monitoring to the reduction of
carvone (Scheme 2) was investigated. 21% w/w of the
catalyst (Pd/C) was taken in 20 mL of the solvent (methanol).
1.5 g of carvone was added to the above mixture. The
reaction setup used was similar to that used for the cyclo-
hexene hydrogenations. Different conditions were also used
like varying the amounts of catalyst, methanol, and carvone.
Table 2 indicates the concentrations of each of these used
in the reactions.
Reduction of 2-(4-Hydroxyphenyl) Propionate. The
reduction of 2-(4-hydroxyphenyl) propionate (Scheme 3) was
also studied using in situ Raman spectroscopy. 10 mmol
(1.802 g) of the reactant was dissolved in methanol (25 mL).
5 mol % of Rhodium black (51 mg) catalyst was added. The
reaction setup used was similar to that used for the cyclo-
hexene and carvone hydrogenations.
A multivariate curve resolution24-26 routine was applied
to all of the cyclohexene reaction data. The reaction data
were assessed for the formation of intermediate species. In
the carvone and 2-(4-hydroxyphenyl) propionate reductions,
MCR was used as the primary analysis technique due to the
expected complexity of the reaction.
Results and Discussion
Cyclohexene Hydrogenations. The conversion of cy-
clohexene to cyclohexane was used as a model to confirm
the potential use of Raman to monitor hydrogenation
reactions under a variety of conditions. Figure 1 displays
the spectral data for various time points in the process of
one of the hydrogenation reactions. The region around the
bands of interest between 650 and 950 cm-1 are highlighted.
(28) Pearson, G. A. J. Magn. Reson. 1997, 27, 265.
Vol. 10, No. 5, 2006 / Organic Process Research & Development
•
929

Figure 1. Raman spectra extracted from the process of
cyclohexene reduction to cyclohexane.
Figure 3. Kinetics of cyclohexene in the presence of rhodium,
Pt/C, and Pd/C.
Figure 2. Kinetic profile for the disappearance of cyclohexene
with Pd/C as the catalyst.
The resolved band for cyclohexane at 802 cm-1 clearly
increases with time, while the resolved band for cyclohexene
at 823 cm-1 clearly decreases with time. The band at 802
cm-1 corresponds to the CH2 deformation plus ring breathing
vibration of cyclohexane, and the one at 823 cm-1 corre-
sponds to a similar vibrational mode of cyclohexene.
The trend of cyclohexene disappearance and cyclohexane
formation can easily be followed qualitatively using the areas
under the curves of the unique peaks previously noted (Figure
2). The advantage of using a band area ratio method for two
unique peaks is that any baseline fluctuation will be
effectively accounted for in the analysis. In the example
shown, the reaction is completed between 1.5 and 2 h. It is
noteworthy that if endpoint determination is all that is
desired, this simple analysis is sufficient. Based on the initial
concentration of cyclohexene, the intensity of the 802 cm-1
band was used to construct a quantitative model for the
kinetic calculations.
Examples of the plots describing the kinetics of the
cyclohexene reduction in the presence of Rh, Pt/C, and Pd/C
can be viewed in Figure 3. Zero-order models provided the
best fit for each set of reaction data. This is not unusual for
hydrogenation reactions.29 Cyclohexene reactions in the
presence of Rh, Pt/C, and Pd/C gave rate constants of 0.310,
0.504, and 0.842 mol L-1 h-1. The cyclohexene reactions
were run in duplicate with each of the catalysts. From the
data, it appears that Pd/C and Pt/C were similar in promoting
efficient reactions and were superior to Rh in this regard.
Figure 4. First-order kinetics of cyclohexene with Rainey
Nickel as the catalyst.
In the cyclohexane reaction, Raman spectroscopy pro-
vided information to allow the choice of optimal catalyst.
Rainey Nickel was also employed as a catalyst for this
reaction. However, the mechanism using this catalyst is
proposed to be different, as a first-order plot shown in Figure
4 appeared to model the data better than a zero-order plot.
This suggests that the reaction in the presence of Rainey Ni
was different from the other three.
Another aspect of analytical data collection during process
development is the ability to understand processes better.
This leads to the ability to better control processes in
production. In one case using Pd/C as a catalyst, it was noted
that there was a 15 to 20 minute delay before cyclohexane
began to appear and cyclohexene began to disappear sug-
gesting a significant reaction induction period.
This case was investigated further using MCR data
analysis. Data were pretreated using a baseline correction
according to Pearson’s method. A four-factor analysis
indicated that an unanticipated side reaction might have
occurred during the induction delay. The MCR results are
shown in Figure 5. The first trend profile is consistent with
cyclohexane formation (with the allowance for solid clump-
ing on the optic), and the plot of the factor “spectrum” is
similar to the cyclohexane spectrum. Catalyst clumping on
the window of the optic can be a common problem with
(29) Paseka, I. J. Catal. 1990, 121, 349-55.
930
•
Vol. 10, No. 5, 2006 / Organic Process Research & Development

Figure 5. MCR data analysis for Raman spectra for the cyclohexene reduction using Pd/C as the catalyst.
heterogeneous catalysts and points out the importance of the
positioning of the probe relative to the stir bar. This mostly
occurred when the magnetic stir bar was not in the center
forming a small vortex. This caused the peak intensities to
go down due to obscuration of the probe window by the
catalyst. This was evident from the particle aggregation on
the optic. Care was taken to cover the hole drilled in the
serum cap for inserting the probe.
can prove to be invaluable, as it would have been extremely
difficult to obtain this level of understanding in the absence
of real-time chemical monitoring. In this experiment the
intermediate was benign and did not represent a safety
hazard. However the production of unstable intermediates
is not uncommon during hydrogenations. The ability to
observe, track, and generate kinetic information on the
formation and concentration of intermediates is particularly
valuable when designing and operating a production-scale
reactor.
Not surprisingly, it was found that the kinetic profiles
changed when the ratios of the solvent, catalyst, and reactant
were altered. Any changes in these ratios caused a significant
change in the overall kinetics of the reaction. For example,
the reduction in Pd/C catalyst and reactant relative to overall
solvent caused a significant decrease in the rate of reaction.
This can be assigned to a simple dilution effect with a
reduction in the availability of hydrogen at the catalyst site
for hydrogenation. Another noteworthy observation was that
the spectral signal seemed to decrease in magnitude if the
reaction was allowed to proceed for a long time. This could
be due to the catalyst physically adhering to the window of
the immersion optic, as was observed in some cases when
the probe was removed from the reaction vessel. This
understanding is valuable to the process engineer responsible
for implementing the on-line analyzer. In the design of the
production implementation, catalyst adherence could be
accounted for by employment of a solvent wash step, an
agitator, better positioning of the optic face relative to the
flow, etc.
The second MCR factor data are similar to the spectral
features of cyclohexene, and therefore the second profile can
be assigned to the disappearance of cyclohexene. The fourth
profile remains relatively constant throughout the reaction
(with allowance for solid clumping on the probe that affected
overall Raman intensity) and is attributed to the methanol
solvent. The third trend plot is of greatest interest. This profile
begins at a finite level and then begins to disappear
immediately. This transient material seems to be completely
depleted in 30 to 45 min. This roughly corresponds to the
induction time found in the simple area-under-the curve plots
(Figure 2) for this reaction. The factor plot in this case is
quite noisy and, thus, indicates a low level of spectral
variation responsible for the composition of the factor.
The presence of a factor that seems to complement the
induction time suggests that it could represent real chemical
information. Because of the noise inherent in the factor plot,
it is difficult to establish an identity for this potential
intermediate with complete certainty. The factor plot appears
to contain some features that may be assigned to solvent
peaks. However based on the most intense features observed
in the MCR plot (peaks at 827, 1073, 1333, and 1593 cm-1)
and the general knowledge of the chemistry, we have
postulated that this is a material related to a transient
intermediate involving a cyclohexene-Pd complex. Such
information can be gained in real time from the combination
of Raman spectroscopy and MCR analysis. Such knowledge
Carvone Hydrogenation. The kinetic profile of the two-
step reduction of carvone to tetrahydrocarvone is shown in
Figure 6. This reaction30 is known to proceed through a two-
(30) Coche, L.; Ehui, B.; Limosin, D.; Moutet, J. C. J. Org. Chem. 1990, 55
(23), 5905-10.
Vol. 10, No. 5, 2006 / Organic Process Research & Development
•
931

Figure 6. MCR data analysis profiles for carvone reduction.
Figure 7. Kinetics of carvone reduction.
step mechanism. Because the CdC bond endo to the ring is
resonance stabilized with the carbonyl group, it is expected
that the double bond exo to the ring will be hydrogenated
selectively before the resonance-stabilized double bond. The
reduction of carvone in the presence of various catalysts has
been studied by several researchers.31,32
To monitor the reduction of carvone, the bands between
the 452-498 cm-1 and 1134-1194 cm-1 shift were chosen.
These regions were considered to account for the symmetric
CsCdC stretch vibration which occurs in the region near
1130 cm-1. The data were clipped, and a baseline correction
using Pearson’s method was performed. The data were then
normalized to the band at 1082 cm-1 prior to MCR analysis.
Three factors were generated using MCR.
The first profile shown in Figure 6 is consistent with a
substance in solution at the beginning of the process.
According to the profile, this material is consumed over the
course of about 7-8 h. This represents the starting material
carvone. A second profile is consistent with a substance that
is absent until approximately 5-6 h after the start of the
hydrogenation. This material then increases in level through
about 16 h after the commencement of the reaction. This
represents the ultimate product tetrahydrocarvone. The third
profile represents the intermediate dihydrocarvone as it
increases in level after the beginning of the hydrogenation
peaking in concentration approximately 4 h into the reaction
procedure. It then decreases through the remainder of the
time that the process was monitored. The rate constant for
carvone reduction in the presence of Pd/C as the catalyst
was calculated to be 0.150 mol L-1 h-1 (Figure 7).
Figure 8. Kinetic profile of 2-(4-hydroxyphenyl) propionate
reduction.
2-(4-Hydroxyphenyl) Propionate Hydrogenation. The
kinetic profile for the reduction of 2-(4-hydroxyphenyl)
propionate is shown in Figure 8. The MCR analysis was
performed using the spectral region from 250 to 1700 cm-1,
and the resulting profiles (Figure 9) suggested the formation
of an intermediate product, which was confirmed off-line
by TLC. The intermediate detected is likely the compound
with a partially hydrogenated ring. The initial hydrogenation
step is the most difficult and can be accomplished only
through the facilitation of the phenolic group.33 This hydro-
genation reaction is similar to the reduction of phenol to
cyclohexanol, which has been looked at by Gonzalez-Velasco
et al. and others.34,35 This initial reduction phase likely
produces the observed intermediate prior to more rapid
reduction to the ultimate product. As can be seen from Figure
9, there appears to be a buildup of intermediate that is
detectable. This seems accurate as the result is confirmed
by TLC. This may suggest that the kinetics of the second
step may not be as fast relative to the subsequent reduction
as anticipated. The rate constant for the reduction of the
propionic acid derivative was found to be 0.247 mol L-1
h-1. Although the product can be detected, the profile for
this material appears to be much noisier than the profiles
for the reactant and intermediate because the product no
longer has an aromatic functional group. This sort of
conversion often will reduce the propensity for Raman
scatter.
MCR analysis of the data leads to a three-factor result
that is consistent with the proposed two-step reaction
mechanism. The analysis clearly showed three profiles that
are assigned to (a) the loss of carvone (the reactant), (b) the
formation of dihydrocarvone (an intermediate), and (c) the
subsequent formation of the required product (tetrahydro-
carvone). The profiles suggest that tetrahydrocarvone is only
formed by hydrogenation of the intermediate. The results
are consistent with the literature31 which suggests the
existence of an intermediate as the reduction of carvone
proceeds. The reduction of carvone to tetrahydrocarvone is
very common unlike the reduction of carvone to dihydro-
(33) Griffin, K. G.; Hawker, S.; Johnson, P.; Palacios-Alcolado, M. L.; Calyton,
C. L. Chem. Ind. (Dekker) 2003, 89, 529-35.
30
carvone, which is very selective.
(34) Gonzalez-Velasco, J. R.; Gutierrez-Ortiz, J. I.; Gonzalez-Marcos, J. A.;
Romero, A. Reaction Kinetics and Catalysis Letters 1986, 32 (2), 505-12.
(35) Mahata, N.; Vishwanathan, V. Journal of Molecular Catalysis A: Chemical
1997, 120 (1-3), 267-270.
(31) De Miguel, S. R.; Roman-Martinez, M. C.; Cazorla-Amoros, D.; Jablonski,
E. L.; Scelza, O. A. Catalysis Today 2001, 66 (2-4), 289-295.
(32) Cerveny, L. Chemical Engineering Communications 1989, 83, 31-63.
932
•
Vol. 10, No. 5, 2006 / Organic Process Research & Development

Figure 9. Kinetic profile of 2-(4-hydroxyphenyl) propionate reduction suggesting the detection of an intermediate.
This process was the most complex among the reactions
investigated and represents a case in which process under-
standing is critical before scale-up. Understanding the
conditions under which the production of the intermediate
occurs and the kinetics and thermodynamics of the overall
process are critical to successful transfer to a production
environment. Raman spectroscopy was demonstrated to help
with this critical level of understanding.
The results of this work suggest that a real-time in situ
Raman spectroscopic tool is an attractive alternative to
traditional off-line laboratory-based analytical techniques
such as thin-layer chromatography (TLC), high-pressure
liquid chromatography (HPLC), or nuclear magnetic reso-
nance (NMR) spectroscopy. In comparison to these tech-
niques, in situ Raman analysis offers the benefits of faster
analysis time and faster feedback to the operator. It also
facilitates averting the requirement to sample the process thus
avoiding potential exposure of production workers to hazard-
ous reagents, minimizing contamination, and avoiding sam-
pling issues including nonrepresentative sampling.
In this work Raman spectroscopy has been demonstrated
as a real-time, in situ tool for monitoring hydrogenation
reactions. Hydrogenation reactions are an important class of
reactions that are often employed during the complex
multistage synthesis of active pharmaceutical ingredients. The
understanding and ability to optimize stages within an active
pharmaceutical substance’s production may be considered
under the umbrella of the US FDA’s PAT initiative. The
PAT initiatives goals include (1) increasing process under-
standing, (2) facilitating process monitoring, and (3) provid-
ing a mechanism for process optimization and control in
order to guarantee the maintenance of product quality. Using
analytical tools to understand processes during the develop-
ment phase leads to the ability to design more robust
processes that should result in a reduced risk of producing
an out-of-specification product. In summary, using the
hydrogenation of model compounds Raman spectroscopy has
been demonstrated as a tool capable of meeting and ac-
complishing many of the PAT initiative’s goals.
Conclusions
In this work, the use of Raman spectroscopy as an
analytical tool for monitoring hydrogenation processes has
been demonstrated. The results of the Raman spectroscopic
analyses led to increased reaction understanding for these
processes. In one process, a side reaction was observed that
would not have been detected without the use of in situ
spectroscopic analysis. In two other reactions, analyses of
the Raman spectra lead to the confirmation of the presence
of reaction intermediates. In addition to the observation of
intermediates, kinetic profiles of each reaction were also
elucidated and, in the case of the cyclohexene reduction, the
optimal catalyst was identified. The ability to observe and
quickly identify intermediates in situ and in real time is very
important to the industrial organic chemist. Elucidating
reaction mechanisms can be crucial to safely scaling-up a
new synthesis and can be important when controlling the
overall process. In addition, Raman spectroscopy was
demonstrated as a tool for reaction endpoint determination
that, if utilized in production, would translate into efficient
reactor usage, which is a quantifiable business benefit.
The knowledge gathered regarding the kinetics of the
reactions led to further enhanced reaction understanding. This
understanding could be used for process optimization to
identify process excursions and allow for timely adjustments
to maintain the desired outcome. These adjustments could
include varying the reaction conditions such as temperature,
stirring rate, reagent addition, or reaction time.
Received for review February 14, 2006.
OP0600355
Vol. 10, No. 5, 2006 / Organic Process Research & Development
•
933