In-line monitoring of cocrystallization process and quantification of carbamazepine-nicotinamide cocrystal using Raman spectroscopy and chemometric tools

Article abstract of DOI:10.1016/j.saa.2017.02.045

A cocrystallization process may involve several molecular species, which are generally solid under ambient conditions. Thus, accurate monitoring of different components that might appear during the reaction is necessary, as well as quantification of the final product. This work reports for the first time the synthesis of carbamazepine-nicotinamide cocrystal in aqueous media with a full conversion. The reactions were monitored by Raman spectroscopy coupled with Multivariate Curve Resolution - Alternating Least Squares, and the quantification of the final product among its coformers was performed using Raman spectroscopy and Partial Least Squares regression. The slurry reaction was made in four different conditions: room temperature, 40?°C, 60?°C and 80?°C. The slurry reaction at 80?°C enabled a full conversion of initial substrates into the cocrystal form, using water as solvent for a greener method. The employment of MCR-ALS coupled with Raman spectroscopy enabled to observe the main steps of the reactions, such as drug dissolution, nucleation and crystallization of the cocrystal. The PLS models gave mean errors of cross validation around 2.0 (% wt/wt), and errors of validation between 2.5 and 8.2 (% wt/wt) for all components. These were good results since the spectra of cocrystals and the physical mixture of the coformers present some similar peaks.

Full text of DOI:10.1016/j.saa.2017.02.045

Accepted Manuscript
In-line monitoring of cocrystallization process and quantification
of carbamazepine-nicotinamide cocrystal using Raman
spectroscopy and chemometric tools
Frederico L.F. Soares, Renato L. Carneiro
PII:
DOI:
Reference:
S1386-1425(17)30159-2
doi: 10.1016/j.saa.2017.02.045
SAA 14967
Spectrochimica Acta Part A: Molecular and Biomolecular
Spectroscopy
To appear in:
Received date:
Revised date:
Accepted date:
6 July 2016
19 February 2017
21 February 2017
Please cite this article as: Frederico L.F. Soares, Renato L. Carneiro , In-line monitoring
of cocrystallization process and quantification of carbamazepine-nicotinamide cocrystal
using Raman spectroscopy and chemometric tools. The address for the corresponding
author was captured as affiliation for all authors. Please check if appropriate. Saa(2017),
doi: 10.1016/j.saa.2017.02.045
This is a PDF file of an unedited manuscript that has been accepted for publication. As
a service to our customers we are providing this early version of the manuscript. The
manuscript will undergo copyediting, typesetting, and review of the resulting proof before
it is published in its final form. Please note that during the production process errors may
be discovered which could affect the content, and all legal disclaimers that apply to the
journal pertain.

ACCEPTED MANUSCRIPT
In-line monitoring of cocrystallization process and quantification of
Carbamazepine-Nicotinamide cocrystal using Raman spectroscopy
and chemometric tools
Frederico L. F. Soares, Renato L. Carneiro*
Federal University of São Carlos, Department of Chemistry, Rod. Washington Luis km
235, Zip Code 13565-905, São Carlos, SP, Brazil.
* Renato Lajarim Carneiro (Correspondent author)
renato.lajarim@ufscar.br
Phone: +55 16 33519366
1

ACCEPTED MANUSCRIPT
Abstract
A cocrystallization process may involve several molecular species, which are generally
solid under ambient conditions. Thus, accurate monitoring of different components that
might appear during the reaction is necessary, as well as quantification of the final
product. This work reports for the first time the synthesis of carbamazepine-
nicotinamide cocrystal in aqueous media with a full conversion. The reactions were
monitored by Raman spectroscopy coupled with Multivariate Curve Resolution -
Alternating Least Squares, and the quantification of the final product among its
coformers was performed using Raman spectroscopy and Partial Least Squares
regression. The slurry reaction was made in four different conditions: room temperature,
40°C, 60°C and 80°C. The slurry reaction at 80°C enabled a full conversion of initial
substrates into the cocrystal form, using water as solvent for a greener method. The
employment of MCR-ALS coupled with Raman spectroscopy enabled to observe the
main steps of the reactions, such as drug dissolution, nucleation and crystallization of
the cocrystal. The PLS models gave mean errors of cross validation around 2.0 (%
wt/wt), and errors of validation between 2.5 – 8.2 (% wt/wt) for all components. These
were good results since the spectra of cocrystals and the physical mixture of the
coformers present some similar peaks.
Keywords: Carbamazepine-nicotinamide cocrystal, In-line monitoring, Partial least
squares, Multivariate curve resolution, Cocrystallization.
2

ACCEPTED MANUSCRIPT
1. Introduction
Cocrystals are mixed crystals of two or more compounds (which are neither solvates
nor simple salts) that when combined forms a unique crystallographic pattern, different
from its precursors[1, 2]. Pharmaceutical cocrystals are a class of crystals of which at
least one active principle ingredient (API) is present in the crystallographic motif[3, 4].
This class of crystals presents several properties distinct from its coformers, becoming a
material of high interest for the pharmaceutical industry[5].
Solubility, thermal and mechanic stability and bioavailability are some of the
characteristic that can be altered by the addition of coformer molecules in the
crystallographic pattern of the API[6]. The solubility enhancement and high chemical
stability are advantageous features for a pharmaceutical compound with polymorphic
forms and low solubility[7].
Carbamazepine (CBZ) is a carboxamide highly used in the treatment of epilepsy and
trigeminal neuralgia[8]. Although it has been used as medicament for more than 30
years, it presents low solubility, a treatment with high doses becoming necessary[9].
Four crystallographic forms of the drug are well-known, where, Form III (CBZ III) is
the most thermodynamically stable at room temperature, which is used in the
commercial tablets. Under normal conditions, CBZ III can be hydrated by atmospheric
moisture. The hydrated form of carbamazepine (CBZ DH) presents dissolution
properties and aqueous solubility lower than its anhydrous form, which is a problem for
pharmaceutical industries[10, 11].
The carboxamide group of carbamazepine can originate a homosynthon interaction in
the crystallographic motif of this molecule, generating dimers inside the pure
crystal[12]. Therefore, it is possible to use homosynthon and heterosynthon interactions
3

ACCEPTED MANUSCRIPT
on the carboxamide group with different organic compounds with high water solubility
in order to enhance the dissolution features of this API.
Two of the most studied coformers used to crystallize with carbamazepine are
saccharin and nicotinamide[13]. Nicotinamide is a vitamin belonging to the B vitamin
group; it is widely used in biochemical systems for NAD (Nicotinamide adenine
dinucleotide) and NADP (Nicotinamide adenine dinucleotide phosphate) formation,
which are used as coenzymes in several biochemical reactions.
In the crystallographic motif of carbamazepine-nicotinamide cocrystal, it is possible
to observe the presence of a carbamazepine dimer surrounded by nicotinamide dimers.
Carbamazepine organizes in a boat form and the cocrystal presents a monoclinic
structure. Carbamazepine-nicotinamide cocrystal present a solubility rate 152 times
higher than the pure API (at 23 – 25 °C)[13] .
Among the methods used to crystalize carbamazepine-nicotinamide, the slow
evaporation method is the most used and studied in the field. However, the slow
evaporation method can only be used when the desired cocrystal is present in an
equivalent stoichiometric ratio, i.e. both initial substrates must be at the same amount in
the cocrystal form. Furthermore, the slow evaporation method uses hazardous solvents,
such as: methanol[14], acetyl acetate[15] and DMSO[10].
The slurry conversion procedure presents several advantages when compared to
traditional cocrystallization methods, such as: development of in-situ analysis, easily
transferable to industrial scales and environmentally friendly[16]. In 2006, Rodríguez-
Hornedo et. al. studied the crystallization reaction of carbamazepine-nicotinamide by
slurry conversion, using ethanol and water as solvents, at room temperature[17]. With a
Raman probe, the authors observed the cocrystal formation by the shifts on one peak of
the spectra, relative to the pure carbamazepine spectra.
4

ACCEPTED MANUSCRIPT
The reaction monitoring enables better control of the reaction, with more accurate
responses (in real time) and decreasing the reprocess[18-21]. However, there are some
limitations when univariate analysis is employed in the reaction monitoring. With
multivariate analysis it is possible to build models in the presence of interferents, to
analyze more than one component present in the reaction mixture and to obtain more
robust and accurate responses, which can decrease the errors involved in the process
analysis and in the final product[22, 23].
The chemometric tools most used in the process analysis and reaction monitoring
are[23]: principal component analysis (PCA), partial least squares (PLS) and
multivariate curve resolution (MCR-ALS). PLS enables to correlate experimental data
with several different material properties, and thus, quantify some product
characteristics, for example API quantification[24], excipients and drugs
granulation[25] and monitoring of polymerization reaction[26]. It is a chemometric tool
widely used to quantification experiments where the interferent overlay the signal of the
analyte.
MCR recovers concentration profiles as a function of time, which allows the
monitoring of initial substrates in the final products[27, 28], and different from the
PCA, it recovers a spectral information more similar to the real spectra. It is widely used
in in-situ systems, where it is possible to get data relative to the compounds present in
the reaction in real time, and then observe the formation of the final product, either in
chemical reactions[29] or in physical processes[30].
Thus, the aims of this work were: to develop a method to synthesize the carbamazepine-
nicotinamide cocrystal by slurry reaction using a greener method (with water as solvent); to
monitor the reaction at different temperatures using in-line Raman spectroscopy and MCR-
ALS; and to develop PLS models to quantify simultaneously the cocrystal and its
coformers.
5

ACCEPTED MANUSCRIPT
2. Materials and Methods
2.1 Synthesis of CBZ-NCT “standard” cocrystal
The synthesis of the CBZ-NCT “standard” cocrystal was achieved by following the
method developed by Nehm et. al.[15]. The synthesis was based on the crystallization
reaction, also known as solvothermal synthesis, where equimolar amounts of API and
coformer are added in an organic solvent (in this case, ethyl acetate), and then heated
until total dissolution of the initial substrates. The cocrystal is obtained by slow
evaporation of the solvent.
Carbamazepine was from Jubilant Life Science Limited and nicotinamide from Aarti
Drugs Limited. Ethyl acetate was from Quemis, P.A. grade. The “standard” cocrystal
was characterized by Raman spectroscopy and powder X-ray diffraction (PXRD),
evidencing the high pureness of the cocrystal.
2.2 CBZ-NCT cocrystal quantification among its precursors
A ternary mixture design was used to prepare the samples of the calibration curve.
Different amounts of carbamazepine, nicotinamide and cocrystals were weighed, such
that the total weight for each point of the calibration curve was around 50 mg. The
samples were then homogenized using a vortex mixer and analyzed by Raman
spectroscopy. The validation samples were made by preparing five samples with known
amounts of each component and prepared following the procedure of the calibration
curves. This kind of methodology is largely used to validate PLS quantification models
[31-34]. Table 1 shows the mass of each component for each sample.
In order to evaluate the models error, it was used two frequently used measurements
for error calculation in multivariate analysis: the values of root means square error of
6

ACCEPTED MANUSCRIPT
cross validation (RMSECV) and root mean square error of validation (RMSEV)
calculated by Equation 1.
푅푀푆퐸 = √푛1 푖=1(푦푖 − 푦̂푖)2
(Equation 1)
푛
∑
It is based on the difference between the estimated value and the real value of a
model, where n is the total samples of calibration for RMSECV, and total samples of
validation for RMSEV; 풚풊 is the real value weighted for the analyte in the sample; and
풚̂풊 the estimated value obtained from the calibration regression model.
2.3 Synthesis of cocrystal by slurry conversion and Raman monitoring
For the slurry conversion synthesis, 2 mmol of carbamazepine (472.5 mg) was
weighed in a beaker. In a different recipient, 2 mmol of nicotinamide (244.3 mg) and
1.5 mL of ultrapure water were placed. This solution was added to the carbamazepine
powder and the reaction was monitored by Raman spectroscopy. The reaction was
heated using an IKA heating plate, model C-MAG HS 7, equipped with electronic
thermometer and magnetic stirrer.
For all analyses an i-Raman BWS 415-785H was used, (B&W Tek, Inc., Newark, DE,
USA), with red laser of 785 nm and spectral resolution of 3.5 cm-1. For the in-line
monitoring, a B&W Tek Raman probe BAC-100 was used, placed above the reaction
system. The spectra were obtained continuously using an integration time of 120 s. The
spectral ranges were 500 to 1666 cm-1 (Raman Shift), with a laser power of 224 mW.
The first spectra obtained was relative to the pure spectra of carbamazepine (time = 0
h). After this, the nicotinamide solution was added into the beaker and the heating
started. The reaction was performed in four different temperatures: room temperature
(23 °C ± 1 °C), 40 °C (± 2 °C), 60 °C (± 3 °C) and 80 °C (± 3 °C); this last one being
the reaction described in the literature as the temperature of polymorphic transition of
7

ACCEPTED MANUSCRIPT
dihydrate CBZ into the Form III CBZ[10]. The reactions were monitored for five hours,
the first three hours being under heating (heating stage) and the last two hours without
heating (cooling stage).
2.4 Raman spectroscopy for quantification of final product
After the synthesis, the dried final product was characterized and quantified. The
parameters for the Raman analysis were: integration time of 60 s; laser power from 224
mW to 320 mW; a B&W Tek BAC151 microscope coupled to the Raman system with
20x objective lens and spectral range between 285 and 1708 cm-1 (Raman shift). The
calibration samples (Table 1) and the products from the different syntheses were
manually compacted using a 13 mm die to obtain packed powder, improving the spectra
quality. Due to the microheterogeneity of the analyzed samples, ten spectra were
obtained from each calibration and validation sample at random points, in order to get a
better representativeness of the whole sample, avoiding subsampling problems.
2.5 X-ray powder diffraction (XRD)
To characterize the “standard” cocrystal and the final products of the slurry reactions
an X-ray powder diffraction Rigaku Multiflex diffractometer was used with Cu source
(0.154 nm), current of 30 mA and voltage of 40 kV. The samples were scanned from 5°
to 45° (2θ) at 0.2° min-1 and step size of 0.02°.
2.6 Chemometric methods
To quantify the cocrystal of CBZ-NCT, the iPLS routine present in PLS_toolbox 6.2
(Eigenvector Research Inc., WA, USA) was used, available for Matlab® 2011a
(Mathworks Inc., MA, USA). The samples were normalized by area, and the average of
8

ACCEPTED MANUSCRIPT
the triplicate analysis was used in the iPLS routine. The first derivative was used in the
models. Different interval ranges of Raman spectra for iPLS were also used, in order to
obtain the lowest validation and cross-validation errors.
For in-line monitoring, the first derivative spectra were used, which were normalized
by the area and filtered by PCA. By applying the first derivative, it can be observed
slightly changes in the acquired spectra that were not too evident in the unmodified
spectra. The PCA filter intends to reduce the instrumental noise without losing spectral
information, using the recovered matrix acquired by a product of scores and loadings
that only retains principal components related to meaningful data. In this case, 6
principal components were used in order to ensure all significant information obtained
during the monitoring. This routine ensures lower residuals in MCR-ALS analysis, and
therefore a better value of lack of fit. For the MCR-ALS, the ALS routine was used,
available on the MCR GUI 2.0 toolbox[35]. The number of components was determined
using the SVD (single value decomposition) function on MATLAB. The initial
estimative of the pure spectra was obtained using the PURE routine, also available on
the toolbox. The MCR-ALS was performed using the non-negativity constraint only for
the concentration profile, since it was used the first derivative for the spectra in order to
improve the resolution of MCR-ALS. The pure spectra recovered by MCR-ALS were
integrated after the deconvolution.
3. Results and discussion
3.1 Characterization of the “standard” cocrystal
Supplementary Figure 1 presents the X-ray diffractograms of carbamazepine Form
III, dihydrate carbamazepine, nicotinamide and synthesized carbamazepine-
nicotinamide cocrystal. The crystallographic pattern of CBZ-NCT indicate the
9

ACCEPTED MANUSCRIPT
formation of a crystallographic motif distinct from its coformers. The main peaks are:
6.70°, 8.92°, 10.26°, 13.40° and 20.50° 2θ for the cocrystal; 11.34°, 14.80°, 19.06°,
19.54°, 19.92° and 22.24° 2θ for nicotinamide; 6.08°, 8.84°, 12.22°, 15.12°, 18.42°,
18.88°, 19.44°, 24.12° and 24.68° 2θ for the dihydrate and; 10.08°, 13.00°, 15.24°,
24.82° and 27.54° 2θ for CBZ Form III. The diffractograms obtained were in a very
good agreement with the literature[10, 36, 37].
Supplementary Figure 2 presents the Raman spectra of carbamazepine-nicotinamide
and of its coformers. In the Raman spectrum of CBZ-NCT characteristic peaks appear
common to the pure carbamazepine spectrum, with some slight differences that enabled
to characterize and distinguish their spectra. In the region between 1550 – 1630 cm-1,
relative to C=C stretch and N-H bend, it is possible to observe a more intense peak
around 1600 cm-1 in the cocrystal spectrum, concerning the N-H secondary amide bend
presented in both nicotinamide and carbamazepine molecules. This vibration is accounts
for the homosynthon interaction existing between the carbamazepine dimers and
nicotinamide amide in the crystallographic motif of the cocrystal. Changes in
nicotinamide spectrum were evident, mainly regarding the extinction of the peak around
1675 cm-1 relative to C=O stretch of amides, that occurs due to the homosynthon
interaction between the nicotinamide and carbamazepine molecules.
Another characteristic region on the cocrystal spectrum was about 1035 cm-1. In the
carbamazepine spectrum it is possible to observe two different peaks concerning the 7-
membered ring, which contains the carbamate group, in the region of 1040 cm-1 and
1030 cm-1. In nicotinamide it is possible to observe a peak around 1040 cm-1 originated
from the pyridine ring breathing, this being the most intense peak of this molecule.
However, in the cocrystal spectrum, besides the carbamazepine peaks, a third peak was
observed in an intermediary Raman shift, close to 1035 cm-1, relative to an asymmetric
10

ACCEPTED MANUSCRIPT
breathing of the aromatic ring with the heteroatom. This is the most intense peak in the
cocrystal spectrum. The spectra obtained was in agreement with the literature[10, 15].
Owing to the results obtained in the characterization, the cocrystal synthesized by the
solvothermal method was used as a standard in the calibration curve to quantify the
cocrystal among its coformers as well as a comparison material for the products
obtained in the slurry reactions.
3.2 Quantification of carbamazepine-nicotinamide cocrystal by Raman spectroscopy
and iPLS
For the calibration and validation data sets, the average of 10 spectra acquired in
random points of the samples was used. Due to the differences in the Raman scattering
between the samples, the spectra were normalized by the area, so that all components of
the mixture presented the same relative intensity, i.e., all samples had the same
importance for calibration and validation models. The iPLS algorithm was used in order
to find different spectral intervals which provided more linear information for each
analyte.
Table 2 presents the summarized data for the PLS regression models using Raman
spectroscopy. The values of root mean square error of validation (RMSEV) and cross
validation (RMSECV), enable to estimate the model efficiency for quantification of the
compounds. The lower the RMSECV and RMSEV values, the lower the error to
quantify that substance. For the cocrystal quantification, the values of RMSECV and
RMSEV were 2.0% and 3.5% (wt/wt), respectively, which are low error values for
quantification of crystalline states.
The selected models showed a good linearity and low data dispersion, indicating a
small difference between the real and predicted values for each sample. All components
11

ACCEPTED MANUSCRIPT
presented a coefficient of determination of cross validation (R² CV) and calibration (R²
cal) higher than 0.99 and a coefficient of determination of validation (R² val) higher
than 0.93.
It is inferred by the acquired data that it is possible to quantify the carbamazepine-
nicotinamide cocrystal among its coformers from a ternary mixture of these components
using Raman spectroscopy and iPLS.
3.3 Monitoring of the cocrystal synthesis in aqueous solution
3.3.1 Reaction at room temperature
During the reaction at room temperature we obtained 151 spectra during a 5 hour
period. At the end of reaction, these spectra were organized in a matrix with 151 rows
(time) and 751 columns (wavelengths). The spectra were pretreated using first
derivative, normalization by area and filtered by PCA using 6 principal components.
MCR-ALS was performed to obtain the concentration profiles which enable to observe
the reaction pathway.
Figure 1a presents some spectra obtained during the room temperature monitoring.
The regions that showed more significant changes during the reaction were 500 cm-1,
700 cm-1, 1020 cm-1 and 1600 cm-1. All acquired spectra exhibited a spectra profile
similar to the carbamazepine spectrum. This occurs because the Raman scattering of
carbamazepine is more intense than that of nicotinamide, furthermore, carbamazepine
was present in solid-state during the whole reaction, while nicotinamide was in a diluted
solution, which also contributes for low Raman scattering.
At the beginning of the reaction, only the spectrum of carbamazepine Form III can be
observed, which is the commercial form of this drug. This spectrum can be
characterized by two peaks relative to the 7-member ring molecule (characteristic peaks
12

ACCEPTED MANUSCRIPT
at 1020 and 1040 cm-1). After 4 minutes of reaction, the acquired spectrum was similar
to the initial spectrum of the reaction. However, at 6 minutes the Raman spectrum was
related to the dihydrate form of carbamazepine, characterized by the inversion in the
intensity of the peaks at 1020 and 1040 cm-1, and the disappearance of the 535 cm-1
peak. At the end of the reaction, no significant change was available, and the final
spectrum was the same as the dihydrate form of carbamazepine spectrum.
Figures 1b and 1c shows the concentration and spectral profile recovered by the
MCR-ALS model, respectively. The concentration profile demonstrated that at room
temperature a quick polymorphic transition is observed. The first three spectra acquired
for the reaction monitoring (times equal to 0, 2 and 4 minutes) were related to the
carbamazepine form III spectrum. After 6 minutes of reaction, all carbamazepine
presented in the reaction turned into the dihydrate form of CBZ. Dihydrate CBZ has
lower water solubility than CBZ III, inhibiting the solubilization of this drug, even
partially in the aqueous media, to interact with nicotinamide molecule. Therefore, at
room temperature no evidence of a cocrystal was observed.
It can be observed in the Figure 1c that the recovered spectra present identical spectral
profiles of pure carbamazepine Form III and dihydrate form spectra, an expected fact,
since it is well known that carbamazepine presents a polymorphic transition for its
hydrated forms in the presence of humidity, what is one of the biggest issues regarding
carbamazepine handling in pharmaceutical industries[11]. The model generated by
MCR-ALS presented good agreement with the experimental data, with a percentage of
explained variance of 99.65% (R² = 0.99) and a low lack of fit, 5.89 %.
3.3.2 Reaction at 40 °C
13

ACCEPTED MANUSCRIPT
Figure 2a presents some spectra obtained during the 40 °C reaction monitoring. As in
the room temperature reaction, the spectra presented only characteristic peaks relative to
the carbamazepine spectra. With 4 minutes of reaction, it is possible to observe a mixed
spectrum between Form III and the dihydrate form of carbamazepine, where relative
intensity of peaks 1020 cm-1 and 1040 cm-1 were altered, and the double peak at 530 cm-
1
starts to decreased. With 6 minutes of reaction, the acquired spectrum is closely
similar to the hydrate form spectrum. At the end of the reaction, it is only possible to
observe the pure spectrum of CBZ dihydrate.
The concentration and spectral profile recovered by MCR-ALS are presented in
Figure 2b and 2c, respectively. The concentration profile shows the polymorphic
transition of CBZ III to the dihydrate form of the drug. After 8 minutes of reaction, all
CBZ III was converted in CBZ DH making the CBZ-NCT crystallization impossible.
By the spectral profile it can be observed that one of the components presented a Raman
spectrum identical to the carbamazepine Form III, while the other component presented
a spectrum similar to the dihydrate carbamazepine. The recovered data showed a good
adjustment related to the experimental data, presenting 99.46 % of explained variance,
with a small lack of fit (7.38 %).
3.3.3 Reaction at 60 °C
Figure 3a shows four different spectra acquired during the 60 °C reaction. At the
beginning the spectrum obtained is related to the Form III carbamazepine, however, at
80 minutes of reaction it is possible to observe a slight difference between the intensity
of the peaks in the 1000 – 1040 cm-1 region, this alteration can indicate both a
polymorphic transition to dihydrate and cocrystallization. At 150 minutes of reaction,
the spectrum is similar to the cocrystal, with the presence of a few selective peaks, such
14

ACCEPTED MANUSCRIPT
as the main peak around at 1035 cm-1(instead of 1040 cm-1) and the changes in the
relative intensity of the peaks presented at 750 – 800 cm-1. At the end of the reaction,
the spectrum was identical to the cocrystal spectra.
The concentration and spectral profiles obtained by MCR-ALS are presented in
Figures 3b and 3c. The spectral profile presents two distinct spectra, one of the
recovered spectrum is related to the CBZ Form III, and the other spectrum shows as
identical to the CBZ-NCT cocrystal spectrum. At the beginning of the reaction, there is
only pure carbamazepine. In 60 minutes, no significant changes can be observed. In the
previous reactions, the carbamazepine undergoes a polymorphic transition in 6 minutes
of reaction, the preservation of spectral profile in over an hour of reaction suggests that
at 60 °C the dihydrate form of carbamazepine is not stable, even in aqueous solution.
At close to 80 minutes of reaction, the relative concentration of CBZ Form III started
to decrease and the intensity of the cocrystal spectrum increased. After two hours of
reaction, the cocrystal is predominant.
The modeled data showed a good agreement with the experimental data, with a
percentage of explained variance of 96.62 % and lack of fit of 14.88 %. The reaction at
60 °C enabled a high conversion of the initial reactants to cocrystal.
3.3.4 Reaction at 80 °C
Even with a high conversion at 60 °C, a reaction at 80 °C was conducted, which is the
temperature described in the literature[11] as needed for the polymorphic transition of
CBZ DH to Form III. As noted in the reactions at 40 °C and at room temperature, the
CBZ DH inhibits the cocrystallization reaction with nicotinamide.
Figure 4a presents some spectra obtained during the reaction monitoring. With 170
minutes of reaction, changes can be noticed in the relative intensities of the initial
spectrum in four distinct regions: 520 – 550 cm-1, 740 – 800 cm-1, 1000 – 1050 cm-1 and
15

ACCEPTED MANUSCRIPT
between 1550 – 1650 cm-1. At the out of plane bend of the aromatic ring region
(between 500 and 550 cm-1), it was possible to observe an enhancement in the intensity
of the peak at 530 cm-1, when compared to the initial spectrum. At 740 – 800 cm-1, there
was an emergence of a peak close to 780 cm-1, and a clear alteration in the relative
intensities of the peaks in this region. In the aromatic ring C-H bend region (1000 –
1050 cm-1), there was a shift of a peak and appearance of a shoulder, that is an
indicative that two or more peaks were overlapped in this area. Between 1550 and 1650
cm-1, the relative intensity among the peaks was different from the expected for the
three evaluated forms, DH, Form III and cocrystal.
Right after the heating stage, at 210 minutes of reaction, the acquired spectrum is
quite similar to the 170-minute spectrum. However, a decrease can be noticed in the
shoulder close to 1045 cm-1 and an increase in the peak around 1035 cm-1, characteristic
of the cocrystal spectrum. At the end of the reaction, only a spectrum identical to the
CBZ-NCT spectrum can be seen.
When the PCA filter was executed for the previous reactions, only two principal
components were significant, which were related to the carbamazepine Form III and
dihydrate CBZ (for room temperature and 40 °C) or cocrystal (for 60 °C). In the
reaction at 80 °C, it was possible to observe three principal components that were
significant for the model. Thus, in the 80 °C reaction, the MCR-ALS routine was
performed using 3 components. The concentration profile and spectral profile recovered
by the model were presented in Figures 4b and 4c. The first recovered spectrum was
related to the pure spectrum of Form III, the second one looks like a mixture of
cocrystal and the spectrum of amorphous carbamazepine, and the last was the CBZ-
NCT spectrum.
16

ACCEPTED MANUSCRIPT
Figure 4b displays the concentration profile of the reaction. At the beginning of the
reaction, only the Form III spectrum remains. After 150 minutes the Form III is
consumed and a mixture of amorphous carbamazepine and cocrystal is observed. This
mixture is maintained until the cooling stage. When the solution is cooled, the crystals
start to nucleate as a cocrystal form, resulting in almost full conversion of the reactants
in the desired product. At the end of the reaction it is possible to observe CBZ III
concentration values above zero, this occurs due to the lack of fit of the model and the
high similarity of the spectra.
Some studies have reported that the cocrystallization mechanism of carbamazepine
combined with nicotinamide in aqueous solution occurs firstly with a polymorphic
transition of crystallized carbamazepine (usually in Form III) to the amorphous form.
The amorphous form presents a higher dissolution rate and water solubility than other
polymorphs[38]. Therefore, the amorphous CBZ solubilize in the nicotinamide solution
and due to the homosynthon interaction the coprecipitation occurs. This study also
corroborates with the described mechanism evidencing a recovered spectrum which is a
mixture of cocrystal and amorphous CBZ, characterized by the shoulder at 1040 cm-1
and the relative intensity in the region of 1550 - 1650 cm-1, different from the other
CBZ forms.
The recovered data showed a good suitability with the experimental data, obtaining
the lowest error values with 99.51 % of explained variance and lack of fit of 6.97%. The
reaction at 80 °C enabled a better understanding of the cocrystallization mechanism as
well as the almost full conversion of the initial reactants into the cocrystal, this fact
being a novelty in aqueous solution.
3.4 Characterization and quantification of the final products
17

ACCEPTED MANUSCRIPT
In order to evaluate the final product of each reaction, the materials obtained at the
end of the reactions were characterized by powder X-ray diffraction and Raman
spectroscopy. Supplementary Figure 3 presents the diffractograms of the final products
of each reaction. The crystallographic patterns of the reactions at room temperature and
at 40 °C presented a pattern similar to the dihydrate form of carbamazepine,
characterized by the main peaks at 8.84° and 12.22° 2θ. The diffractograms of the
reactions at 60 °C and 80 °C showed pattern analogous to the cocrystal diffraction
pattern, mostly characterized by the first four diffraction peaks at: 6.68°, 8.92°, 10.22°,
13.38° 2θ; and the main peak at 20.46° 2θ. However, in the 60 °C reaction, it is possible
to observe a small peak at 12.28° 2θ, indicating a small contamination of CBZ DH in
the sample. Moreover, the peaks presented in the 60 °C diffractogram related to the
CBZ-NCT are larger than in the 80 °C diffractogram. The broadening of the peaks can
be related to the crystal size, thus the crystals obtained at the 60 °C reaction probably
are smaller than those at 80 °C.
As both temperatures present the CBZ-NCT formation, and considering that at 60 °C
the crystals are smaller, it can infer that 60 °C is the optimum temperature for CBZ-
NCT nucleation, once it is found a good conversion (good number of crystal) but with
small particle size. While at 80 °C, the nucleation process has occurred and the crystals
starts to growth.
Supplementary Figure 4 exhibit the Raman spectra of the final products of each
reaction temperature. The acquired spectra at room temperature and 40 °C reactions are
similar to the CBZ DH spectrum, characterized by the inversion of the relative intensity
of peaks between 1000 and 1040 cm-1, and the single peak at the region of 545 cm-1.
The Raman spectra of the products obtained at 60 °C and 80 °C are similar to the
cocrystal spectrum. The double peak at 1020 cm-1 and the intensity relation of the peaks
18

ACCEPTED MANUSCRIPT
between 1550 and 1650 cm-1 could characterize the acquired spectra. However, the
spectrum of the 60 °C reaction presents a shoulder at 1030 cm-1, and different relation
of intensities in the region of 1550 – 1650 cm-1. These results, coupled with the X-ray
diffractograms, indicates that the 60 °C reaction presents a small contamination of CBZ
DH, while the 80 °C reaction achieved a full conversion of the final products.
In order to evaluate the conversion of each reaction, the samples were analyzed using
the iPLS regression models developed in the section 3.2. Table 3 presents the
percentage of each component in the final products of the reactions. The room
temperature and 40 °C reactions resulted in a full conversion of Form III to CBZ DH.
The reaction at 60 °C presented a high conversion of the initial reactants to cocrystals,
although with a small amount of unreacted CBZ DH. The reaction at 80 °C was the only
experiment which enabled a full conversion of carbamazepine and nicotinamide into the
CBZ-NCT cocrystal.
4. Conclusion
It was possible to perform the cocrystallization reaction of carbamazepine-
nicotinamide cocrystal using a slurry reaction method, with an environmentally friendly
solvent in two different temperatures (at 60 °C, and 80 °C), with full conversion of the
initial substrates at 80 °C.
iPLS regression shows to be robust and reliable to quantify the CBZ-NTC cocrystal
among its coformers using Raman spectroscopy, which is very important to verify the
quality of the final product. The models gave validation errors lower than 8.2 %, which
is an acceptable value for quantification of different crystalline structures of the same
molecules, due to the high similarity of the spectra.
19

ACCEPTED MANUSCRIPT
The in-line monitoring using Raman spectroscopy and MCR-ALS shows to be a
powerful combination to in-situ evaluation of crystallization process. The recovered
concentration profiles of each reaction showed high correlation with the final results
found by Raman and XRD.
The reactions at room temperature and 40 °C showed a quick polymorphic transition
from Form III to the dihydrate. Although it was not the main focus of this work, the
results showed the great importance of the control of the drug to prevent undesirable
changes during the tablet manufacturing. In the reaction at 60 °C, there was a small
formation of the dihydrate form, but CBZ-NCT cocrystals were mainly yielded. At 80
°C it was possible to identify the presence of amorphous carbamazepine corroborating
the proposed cocrystallization mechanism. The amorphous form interacts with the
nicotinamide present in solution and forms the cocrystal. The reaction at 80 °C also
allowed a cocrystallization process with several advantages, such as: high final product
purity (conversion higher than 98%), easy to control and environmentally friendly.
Acknowledgements
The authors would like to thank the financial supports from the Coordination for the
Improvement of Higher Education Personnel (CAPES), 472948/2013-0 from National
Council for Scientific and Technological Development (CNPq) and grant #2010/16520-
5, São Paulo Research Foundation (FAPESP).
Bibliography
[1] P.M. Bhatt, Y. Azim, T.S. Thakur, G.R. Desiraju, Co-Crystals of the Anti-HIV
Drugs Lamivudine and Zidovudine, Cryst. Growth Des., 9 (2009) 951-957.
20

ACCEPTED MANUSCRIPT
[2] M. Herrmann, U. Forter-Barth, H. Krober, P.B. Kempa, M.D. Juez-Lorenzo, S.
Doyle, Co-Crystallization and Characterization of Pharmaceutical Ingredients, Part.
Part. Syst. Char., 26 (2009) 151-156.
[3] A.D. Bond, What is a co-crystal?, CrystEngComm, 9 (2007) 833-834.
[4] M. Habgood, M.A. Deij, J. Mazurek, S.L. Price, J.H. ter Horst, Carbamazepine Co-
crystallization with Pyridine Carboxamides: Rationalization by Complementary Phase
Diagrams and Crystal Energy Landscapes, Cryst. Growth Des., 10 (2010) 903-912.
[5] J.C. Burley, A. Alkhalil, M. Bloomfield, P. Matousek, Transmission Raman
spectroscopy for quality control in model cocrystal tablets, Analyst, 137 (2012) 3052-
3057.
[6] S. Vijayalakshmi, S. Kalyanaraman, Non linear optical analyses of hexamine:
Phenol cocrystals based on hydrogen bonding: A comparative study, Spectrochimica
Acta Part A: Molecular and Biomolecular Spectroscopy, 120 (2014) 14-18.
[7] C.B. Aakeroy, S. Forbes, J. Desper, Using Cocrystals To Systematically Modulate
Aqueous Solubility and Melting Behavior of an Anticancer Drug, J. Am. Chem. Soc.,
131 (2009) 17048-17049.
[8] A. Shayanfar, S. Velaga, A. Jouyban, Solubility of carbamazepine, nicotinamide and
carbamazepine-nicotinamide cocrystal in ethanol-water mixtures, Fluid Phase Equilibr.,
363 (2014) 97-105.
[9] M.B. Hickey, M.L. Peterson, L.A. Scoppettuolo, S.L. Morrisette, A. Vetter, H.
Guzman, J.F. Remenar, Z. Zhang, M.D. Tawa, S. Haley, M.J. Zaworotko, O.
Almarsson, Performance comparison of a co-crystal of carbamazepine with marketed
product, Eur. J. Pharm. Biopharm., 67 (2007) 112-119.
[10] S.G. Fleischman, S.S. Kuduva, J.A. McMahon, B. Moulton, R.D.B. Walsh, N.
Rodriguez-Hornedo, M.J. Zaworotko, Crystal engineering of the composition of
21

ACCEPTED MANUSCRIPT
pharmaceutical phases: Multiple-component crystalline solids involving carbamazepine,
Cryst. Growth Des., 3 (2003) 909-919.
[11] L.E. McMahon, P. Timmins, A.C. Williams, P. York, Characterization of
dihydrates prepared from carbamazepine polymorphs, J. Pharm. Sci., 85 (1996) 1064-
1069.
[12] G.T.R. Palmore, J.C. Macdonald, The Role of Amides in the Noncovalent
Synthesis of Supramolecular Structures in Solution, at Interfaces and in Solids, in: A.
Greenberg, C.M. Breneman, J.F. Liebman (Eds.) The Amide Linkage: Structural
Significance in Chemistry, Biochemistry, and Materials Science, John Wiley and Sons,
New York, 2000, pp. 291-336.
[13] D.J. Good, N. Rodriguez-Hornedo, Solubility Advantage of Pharmaceutical
Cocrystals, Cryst. Growth Des., 9 (2009) 2252-2264.
[14] K.C. Lee, K.J. Kim, Effect of Supersaturation and Thermodynamics on Co-Crystal
Formation, Chem. Eng. Technol., 34 (2011) 619-623.
[15] S.J. Nehm, B. Rodriguez-Spong, N. Rodriguez-Hornedo, Phase solubility diagrams
of cocrystals are explained by solubility product and solution complexation, Cryst.
Growth Des., 6 (2006) 592-600.
[16] S.L. Childs, K.I. Hardcastle, Cocrystals of chlorzoxazone with carboxylic acids,
CrystEngComm, 9 (2007) 364-367.
[17] N. Rodriguez-Hornedo, S.J. Nehru, K.F. Seefeldt, Y. Pagan-Torres, C.J.
Falkiewicz, Reaction crystallization of pharmaceutical molecular complexes, Mol.
Pharm., 3 (2006) 362-367.
[18] M.J. Bevill, P.I. Vlahova, J.P. Smit, Polymorphic Cocrystals of Nutraceutical
Compound p-Coumaric Acid with Nicotinamide: Characterization, Relative Solid-State
22

ACCEPTED MANUSCRIPT
Stability, and Conversion to Alternate Stoichiometries, Cryst. Growth Des., 14 (2014)
1438-1448.
[19] K.S. Lee, K.J. Kim, J. Ulrich, In Situ Monitoring of Cocrystallization of Salicylic
Acid-4,4 '-Dipyridyl in Solution Using Raman Spectroscopy, Cryst. Growth Des., 14
(2014) 2893-2899.
[20] Y. Du, Y. Xia, H. Zhang, Z. Hong, Using terahertz time-domain spectroscopical
technique to monitor cocrystal formation between piracetam and 2,5-dihydroxybenzoic
acid, Spectrochimica Acta Part A: Molecular and Biomolecular Spectroscopy, 111
(2013) 192-195.
[21] Y. Du, H. Zhang, J. Xue, H. Fang, Q. Zhang, Y. Xia, Y. Li, Z. Hong, Raman and
terahertz spectroscopical investigation of cocrystal formation process of piracetam and
3-hydroxybenzoic acid, Spectrochimica Acta Part A: Molecular and Biomolecular
Spectroscopy, 139 (2015) 488-494.
[22] W. Chew, P. Sharratt, Trends in process analytical technology, Anal. Methods, 2
(2010) 1412-1438.
[23] A.L. Pomerantsev, O.Y. Rodionova, Process analytical technology: a critical view
of the chemometricians, J. Chemometr., 26 (2012) 299-310.
[24] S. Kim, M. Kano, H. Nakagawa, S. Hasebe, Estimation of active pharmaceutical
ingredients content using locally weighted partial least squares and statistical
wavelength selection, Int. J. Pharmaceut., 421 (2011) 269-274.
[25] S. Poutiainen, S. Matero, T. Hamalainen, J. Leskinen, J. Ketolainen, K. Jarvinen,
Predicting granule size distribution of a fluidized bed spray granulation process by
regime based PLS modeling of acoustic emission data, Powder Technol., 228 (2012)
149-157.
23

ACCEPTED MANUSCRIPT
[26] L. Saerens, L. Dierickx, T. Quinten, P. Adriaensens, R. Carleer, C. Vervaet, J.P.
Remon, T. De Beer, In-line NIR spectroscopy for the understanding of polymer-drug
interaction during pharmaceutical hot-melt extrusion, Eur. J. Pharm. Biopharm., 81
(2012) 230-237.
[27] S.E. Richards, E. Becker, R. Tauler, A.D. Walmsley, A novel approach to the
quantification of industrial mixtures from the Vinyl Acetate Monomer (VAM) process
using Near Infrared spectroscopic data and a Quantitative Self Modeling Curve
Resolution (SMCR) methodology, Chemometr. Intell. Lab., 94 (2008) 9-18.
[28] F.L.F. Soares, R.L. Carneiro, Green Synthesis of Ibuprofen–Nicotinamide
Cocrystals and In-Line Evaluation by Raman Spectroscopy, Crystal Growth & Design,
13 (2013) 1510-1517.
[29] R. Tauler, B. Kowalski, S. Fleming, Multivariate Curve Resolution Applied to
Spectral Data from Multiple Runs of an Industrial-Process, Anal. Chem., 65 (1993)
2040-2047.
[30] T.R.M. De Beer, P. Vercruysse, A. Burggraeve, T. Quinten, J. Ouyang, X. Zhang,
C. Vervaet, J.P. Remon, W.R.G. Baeyens, In-Line and Real-Time Process Monitoring
of a Freeze Drying Process Using Raman and NIR Spectroscopy as Complementary
Process Analytical Technology (PAT) Tools, J. Pharm. Sci., 98 (2009) 3430-3446.
[31] A. Heinz, M. Savolainen, T. Rades, C.J. Strachan, Quantifying ternary mixtures of
different solid-state forms of indomethacin by Raman and near-infrared spectroscopy,
European Journal of Pharmaceutical Sciences, 32 (2007) 182-192.
[32] A. Aina, M.D. Hargreaves, P. Matousek, J.C. Burley, Transmission Raman
spectroscopy as a tool for quantifying polymorphic content of pharmaceutical
formulations, Analyst, 135 (2010) 2328-2333.
24

ACCEPTED MANUSCRIPT
[33] F.L.F. Soares, R.L. Carneiro, Evaluation of analytical tools and multivariate
methods for quantification of co-former crystals in ibuprofen-nicotinamide co-crystals,
Journal of Pharmaceutical and Biomedical Analysis, 89 (2014) 166-175.
[34] R. Caliandro, G. Di Profio, O. Nicolotti, Multivariate analysis of quaternary
carbamazepine–saccharin mixtures by X-ray diffraction and infrared spectroscopy,
Journal of Pharmaceutical and Biomedical Analysis, 78–79 (2013) 269-279.
[35] J. Jaumot, A. de Juan, R. Tauler, MCR-ALS GUI 2.0: New features and
applications, Chemometr. Intell. Lab., 140 (2015) 1-12.
[36] E. Gagniere, D. Mangin, F. Puel, A. Rivoire, O. Monnier, E. Garcia, J.R. Klein,
Formation of co-crystals: Kinetic and thermodynamic aspects, J. Cryst. Growth, 311
(2009) 2689-2695.
[37] Z. Rahman, C. Agarabi, A.S. Zidan, S.R. Khan, M.A. Khan, Physico-mechanical
and Stability Evaluation of Carbamazepine Cocrystal with Nicotinamide, Aaps
Pharmscitech, 12 (2011) 693-704.
[38] K. Seefeldt, J. Miller, F. Alvarez-Nunez, N. Rodriguez-Hornedo, Crystallization
pathways and kinetics of carbamazepine-nicotinamide cocrystals from the amorphous
state by in situ thermomicroscopy, spectroscopy and calorimetry studies, J. Pharm. Sci.,
96 (2007) 1147-1158.
25

ACCEPTED MANUSCRIPT
Figure 1
Figure 1. Reaction at room temperature: a) Raman spectra obtained at four different
times; b) Concentration profile and; c) Integrated Raman spectral profile.
26

ACCEPTED MANUSCRIPT
Figure 2
Figure 2. Reaction at 40 °C: a) Raman spectra obtained at four different times; b)
Concentration profile and; c) Integrated Raman spectral profile.
27

ACCEPTED MANUSCRIPT
Figure 3
Figure 3. Reaction at 60 °C: a) Raman spectra obtained at four different times; b)
Concentration profile and; c) Integrated Raman spectral profile.
28

ACCEPTED MANUSCRIPT
Figure 4
Figure 4. Reaction at 80 °C: a) Raman spectra obtained at four different times; b)
Concentration profile and; c) Integrated Raman spectral profile.
29