Polymorphism Studied by Lattice Phonon Raman Spectroscopy and

Dec 2, 2008 - de Nantes (IRCCYN), Ecole Centrale de Nantes, 1 rue de la Noë,. BP 92101, 44321 Nantes Cedex 3, France. ReceiVed April 9, 2008; ReVised ...
1 downloads 0 Views 414KB Size
Polymorphism Studied by Lattice Phonon Raman Spectroscopy and Statistical Mixture Analysis Method. Application to Calcium Carbonate Polymorphs during Batch Crystallization C. Carteret,*,† A. Dandeu,‡ S. Moussaoui,§ H. Muhr,‡ B. Humbert,† and E. Plasari‡

CRYSTAL GROWTH & DESIGN 2009 VOL. 9, NO. 2 807–812

Laboratoire de Chimie Physique et Microbiologie pour l’EnVironnement, Nancy UniVersité, CNRS, 405 rue de VandoeuVre, 54600 Villers-le`s-Nancy, France, Laboratoire des Sciences du Ge´nie Chimique-CNRS, Ecole Nationale Supe´rieure des Industries Chimiques-INPL, 1 rue GrandVille, BP 451, 54001 Nancy Cedex, France, and Institut de Recherche en Communications et Cyberne´tique de Nantes (IRCCYN), Ecole Centrale de Nantes, 1 rue de la Noe¨, BP 92101, 44321 Nantes Cedex 3, France ReceiVed April 9, 2008; ReVised Manuscript ReceiVed October 21, 2008

ABSTRACT: This paper investigated the crystallization of calcium carbonate in saline solution at different temperatures using Raman spectroscopy. Application of an advanced mixture analysis algorithm based on Bayesian theory (BPSS) allowed recovery of the Raman spectra of the three pure anhydrous polymorphs (vaterite, aragonite, and calcite). In particular, the low wavenumber range between 50 and 300 cm-1, characteristic of crystallographic structure phonons, clearly distinguished the three structures. Contour plots of the polymorph concentrations during the crystallization process vs temperature and time have been established and discussed. This study demonstrated the ability of low wavenumber Raman spectroscopy combined with BPSS analysis to identify the different polymorphs (known or unknown) during a crystallization process and to assay their relative amounts without calibration. Introduction Polymorphism is the ability of a substance to crystallize in different crystal forms, each of them having the same chemical formula but different stacking of atoms or molecules in the crystal lattice.1,2 Polymorphs exhibit different physical and chemical properties, such as crystal morphology, stability, solubility, bioavailability, melting temperature, hygroscopicity, and chemical reactivity. These differences are key to their end uses with important technical and financial implications in a diverse range of areas. Indeed, crystallinity plays a significant role in pharmaceuticals, agrochemicals, pigments, dyestuffs, foods, explosives geophysics, energy storage, biominerals, and nonlinear and optical applications. The ability to detect the presence of different polymorphs in a sample, to assay their relative amount, and ultimately to obtain the desired form in very high yield is of crucial importance in chemical engineering and technology. Raman scattering (RS) is a powerful method to investigate polymorphs due to its remarkable sensitivity to the crystalline structure of materials.3,4 This performance is improved when considering the lattice phonon region (low wavenumber Raman active vibrational modes below 400 cm-1), which represents the fingerprint of the atomic crystalline lattice. External vibrations or lattice phonons involve indeed relative translations and rotations of the formular units, in the crystallographic cell, as a result of freezing within the crystal of real translations and rotations of each single free unit, while internal modes are those that involve vibration movements of the molecular ions. Because of the very different strengths of the intra- and interionic forces, the lattice modes are expected at lower wavenumbers than the internal modes but with a stronger dependence of the crystal organization. The ability to determine the simultaneous presence of multiple polymorphs in the same sample by distinguishing * To whom correspondence should be addressed. E-mail: cedric.carteret@ lcpme.cnrs-nancy.fr. † Nancy Université, CNRS. ‡ Ecole Nationale Supe´rieure des Industries Chimiques-INPL. § Ecole Centrale de Nantes.

their Raman spectra is a challenge of polymorphs characterization. Successful applications of RS on quantitative determination of polymorph mixtures composition have been reported.4 Usually, a calibration step with standard samples was established to build a regression model. Note that calibration is only valuable for known polymorphs and is inefficient to identify unknown polymorphs. Moreover, this step can be long and difficult because of the syntheses of the standard samples. Moreover, the calibration reduces the operating conditions because samples to be analyzed must be imperatively obtained with the same parameters (shape, purity, etc.) as the standard samples. An additional parameter (shape, particle size, phonon confinement, impurities, additives, pH, etc.) can contribute to the Raman signal and modify it. In this case, the regression model cannot more be used. The resolution of pure component spectra from large complex two-dimensional mixture spectra without any a priori information remains an open problem in the chemical sciences. Such numerical separation of mixture spectra is a rather important issue because it is sometimes the only means to resolve spectra of unknown pure constituents when analytical techniques fail to completely separate multicomponent mixtures or when separation is inherently impossible, for example, in situ reactive studies. In our previous works, we have developed a statistical method of spectral mixture analysis based on Bayesian estimation theory and Markov chain Monte Carlo (MCMC) methods.5,6 This method called BPSS, for Bayesian positive source separation, gave an estimation of the unknown component spectra and the concentrations of the underlying species. Whereas spectroscopic techniques may fail to resolve unknown or unstable individual constituents contained in multicomponent mixtures, Raman spectroscopy combined with BPSS treatment may offer a promising and efficient approach, which would be particularly applicable to polymorphic samples. Calcium carbonate nucleates in three crystalline forms, calcite, aragonite, and vaterite, with rhombohedral, orthorhombic, and hexagonal structures. Calcite is the most thermodynamically stable of the three, followed by aragonite and vaterite. The

10.1021/cg800368u CCC: $40.75  2009 American Chemical Society Published on Web 12/02/2008

808 Crystal Growth & Design, Vol. 9, No. 2, 2009

crystallization of calcium carbonate is a widely occurring process in nature (marble and limestone, coral reef, and shellfish, etc.) as well as an important operation in industry. Calcium carbonate is one of the main components of the scaling that arises in various drainage situations in the chemical industry and in circulating water for heating and cooling in living environments. Calcium carbonate is also used in various industrial fields as additives to medicines, foods, papers, plastics, printing ink, etc. The physical properties of the crystallized product depend largely on the polymorphic composition, so it is necessary to control these polymorphs in a mixture. Several techniques based on infrared spectroscopy (IR), X-ray-diffraction (XRD), or Raman spectroscopy (RS) have been used to determine the composition of CaCO3 polymorph mixtures.7-11 However, contrary to XRD and IR, RS is a nondestructive technique that does not require sample preparation9-11 and offers the potential for in situ measurements4,11 for an in-line polymorphic composition monitoring. The relation between preparation conditions and polymorphism of calcium carbonate has been widely studied.12-21 Factors affecting the polymorphic composition are numerous: temperature, aging time, pH, supersaturation, additives, impurities, etc. In our case, the crystallization process of calcium carbonate is carried out in the 5 M NaCl solution, representative of realistic industrial situations. The main purpose of this research is to study the relation of polymorphs and temperature and to explore favorable conditions for calcite formation in saline solutions. Therefore, experiments were operated in a temperature range between 20 and 70 °C, and an aging of the precipitated mixture was carried out. Low wavenumber Raman spectroscopy combined with BPSS analysis was used to identify the polymorphs during the crystallization process and to estimate their relative concentrations. Experimental Section Raman Spectroscopy. The Raman spectra of dried powders, excited by a laser beam of wavelength 514.5 nm, were recorded with a JobinYvon T64000 Raman spectrometer equipped with a confocal microscope device. The long frontal objective of the microscope (50×, numerical aperture of 0.55) focused the laser beam on the sample and collected the backscattering light. The laser beam power amounted to 2 mW, and the acquisition time was 1 min. The spectral resolution was 3 cm-1, with a wavenumber precision better than 1 cm-1. Wavenumber calibration was obtained by recording the green-emission ray light of an Hg lamp. The homogeneity of each solid was checked by recording Raman spectra at several points, which were randomly distributed throughout the powder. The average of five spectra was considered as the Raman spectrum of each mixture. Raman mixture spectra of all samples were consolidated into a single intensity data matrix for statistical analysis. The Raman spectra of the polymorph mixture were first processed using a background removal approach.22 In this method, the baseline was represented by a polynomial whose parameters were estimated by minimizing a truncated quadratic cost function. This method required the specification of the polynomial order and the threshold of the quadratic cost function truncation. In the framework of our application, the method was applied for each spectrum separately with a fifth order polynomial and a threshold chosen by trial and error. Batch Crystallization Experiments. Calcium chloride and sodium carbonate separately dissolved in sodium chloride solutions of the same concentration (5 M) were rapidly mixed to precipitate calcium carbonate. All solutions were prepared with analytical grade chemicals and deionized water. A 2.5 L thermostatted glass stirred vessel of standard geometry (equipped with four baffles and a Rushton turbine) was used. One hundred milliliters of a solution containing 0.625 M Na2CO3 + 5 M NaCl was instantaneously added to 2.5 L of a solution containing 0.025 M CaCl2 + NaCl 5 M (the precipitation is carried out under stoechiometric conditions). Experiments were conducted at 20, 30, 40, 50, 60, and 70 °C. Calcium selective electrode was interfaced to the batch reactor to determine the solution concentration of calcium

Carteret et al. in real time as the crystallization proceeded. A solid sample was collected (rapidly filtered, washed, and dried) 2 min after the beginning of the experiment to determine the polymorphic composition at the end of the precipitation step. Then, samples were collected at regular time intervals to follow the mixture transformation in the reactor. Preparation of Pure Polymorphs. Vaterite was prepared by mixing simultaneously 250 mL of a 0.1 M solution of Ca(NO3)2 and 250 mL of a 0.1 M solution of Na2CO3 in a 1 L baffled glass stirred vessel at ambient temperature. The suspension was stirred for 20 min at 1200 rpm, and the resulting slurry was filtered through a 0.45 µm membrane, washed with anhydrous ethanol, and left to dry at 100 °C. Aragonite was made by adding 100 mL of 0.625 M Na2CO3 solution to 2.5 L of 0.025 M CaCl2 dissolved in a 5 M NaCl solution in the glass vessel at 60 °C. The suspension was stirred for 2 min, and the resulting slurry was filtered through a 0.45 µm membrane, washed several times with distilled water and anhydrous ethanol, and then left to dry at ambient temperature. Calcite was prepared by adding 100 mL of a 0.625 M Na2CO3 solution to 2.5 L of a 0.025 M CaCl2 solution at ambient temperature. The suspension was stirred for 7 days, and the resulting slurry was filtered through a 0.45 µm membrane, washed several times with distilled water and anhydrous ethanol, and then left to dry at ambient temperature. Mixture Analysis: Bayesian Positive Source Separation. In the mixture analysis method, spectral data sets resulting from observations of multicomponent substances are interpreted as a weighted sum of the unknown pure component spectra. The mixing model assumes that m measured data {D(i,k), k ) 1,..., n}im) 1 are linear combinations of p unknown pure component spectra {S(i,k), k ) 1,..., n}ip) 1. Mathematically, this model is expressed as D(i,k) ) ∑jp) 1C(i,j)S(j,k) + E(j,k), where i ) 1,..., m and j ) 1,..., p, respectively, index the measured spectra and the unknown pure component spectra and the index k corresponds to the spectral variable {λk, k ) 1,..., n}. Each mixing coefficient C(i,j) is proportional to the concentration of the jth pure component in the ith mixture. The additive noise terms {E(i,k), k ) 1,..., n}im) 1 represent the measurement errors and model imperfections. Using matrix notations, this model is written as D ) SC + E where the row vectors of the (m × n) data matrix D contain the m measured spectra, and C is the (m × p) mixing matrix, with its column vectors representing the mixing coefficient profiles of the pure components. S is the (p × n) matrix, with its row vectors containing the p pure component spectra, and E is the (m × n) noise matrix. By assuming a known number of components, the mixture analysis aim is to estimate the pure component spectra and the mixing coefficient profiles from the mixture spectra. As previously introduced, the mixture data were processed using the BPSS approach. This statistical approach addresses the mixture spectra decomposition using a Bayesian inference. The main idea of this approach is to first construct the posterior probability distribution function (pdf) of the pure component spectra and their concentrations. According to Bayes’ theorem, the computation of this pdf needs to specify a statistical model on the mixing model (linear mixing with Gaussian measurement errors) and on the pure spectra and concentrations (appropriate pdf encoding non-negativity and sparseness). A second step is to jointly estimate the pure spectra and the concentrations from this posterior law (pdf) using MCMC methods. The aim of MCMC methods is to generate realizations of a target pdf, which is the posterior law in our case. For more details on this approach, the reader is redirected to refs 5 and 6 (The BPSS code may be kindly supplied to interested persons). In this method, the only parameters to tune are the number of components and MCMC algorithm iterations. Thus, this approach outperforms classical multivariate curve resolution methods in terms of estimation accuracy and simplicity of use.5 It is well-known that quantitative Raman measurements from solids are difficult even when well-characterized references are used for calibration/quantification. Nevertheless, in the absence of any calibration reference, it is possible to obtain a first approximation to the relative concentrations by using BPSS and a mass-balance constraint. It must be emphasized that these are relative concentrations since they are weighted by the intensity of the normalized pure component spectra and not a separate calibration. Because tests on 10 mechanical mixtures of pure polymorphs gave an average error of 7%, we assumed then that the

Calcium Carbonate Polymorphs during Batch Crystallization

Crystal Growth & Design, Vol. 9, No. 2, 2009 809

Figure 2. Raman spectra of synthesized calcite polymorph with, dashed line, and without, solid line, Mg2+ as an impurity. Table 1. Observed Raman Wavenumbers (cm-1) and Relative Intensitya of CaCO3 Polymorphs: Lattice Modes and Carbonate Internal Vibrations Classified as ν1 Symmetric Stretching, ν2 Out-of-Plane Bending, ν3 Asymmetric Stretching, and ν4 In-Plane Bending vibration mode

calcite

aragonite

vaterite

lattice

155 (m) 281 (s)

111 (w) 116 (w) 140 (sh) 152 (s) 178 (w) 190 (vw) 205 (m) 215 (sh) 248 (w) 260 (w) 272 (w) 284 (w)

105 (m) 118 (sh) 148 (vw) 174 (vw) 207 (w) 266 (w) 301 (m) 332 (sh)

internal (ν1)

1086 (vs)

1085 (vs)

internal (ν2) internal (ν3)

1430 (w)

853 (vw) 1462 (vw)

internal (ν4)

712 (m)

1075 (s) 1090 (s) 874 (vw) 1413 (vw) 1465 (vw) 668 (vw) 683 (vw) 740 (w) 750 (w)

Figure 1. Raman spectra of synthesized calcium carbonate polymorphs: (a) external vibrational modes and (b) internal vibrational modes. polymorphic composition in batch experiments was estimable with an error below 10%.

Results and Discussion Pure Synthesized Calcium Carbonate Polymorphs. The Raman spectra of the three synthesized pure polymorphs presented in Figure 1 concurred well with previously reported results.23-26 Wavenumbers are reported in Table 1: Lattice modes appear below 300 cm-1, whereas the ν1, ν2, ν3, and ν4 carbonate internal modes appear around 1075-1090, 850-900, 1430-1600, and 680-750 cm-1, assigned, respectively, to the symmetric stretching, in-plane bending, antisymmetric stretching, and out-of-plane bending modes of the carbonate anion. Most of spectroscopic studies of the CaCO3 polymorphs have been limited to a comparison of the four fundamental vibrations, ν1, ν2, ν3, and ν4, of the CO32- unit in the crystalline lattice. The crystalline structure and Raman spectra of calcite and aragonite are well-documented.23-25 Because of its instability, few vibrational data have been reported for vaterite, and the crystalline structure of vaterite remained contentious until the recent work of Gabrielli et al.26 Because of experimental constraints, Gabrielli et al. have not been able to report the Raman spectra in the lattice modes range. Our spectra in the lattice range presented here complete thus exposed by Gabrielli et al. The strongest Raman signals in all three phases, at 1086, 1085, and 1090 cm-1 for calcite, aragonite, and vaterite,

697 (sh) 702 706 716

a Key: vs, very strong; s, strong; m, medium; w, weak; vw, very weak; and sh, shoulder.

respectively, overlap and cannot be used for analytical purposes. Although ν4 modes exhibit weak intensity, they are often used to characterize the three phases.8,9,11 The lattice modes give numerous signals with medium intensity. Therefore, we focus here our attention on these spectral signatures to distinguish the three polymorphs. Figure 2 exhibits the influence of the presence of magnesium impurity in the solution composition on the spectrum of the same polymorph (calcite). This highlights the inability to use an unique calibration model to follow the polymorphic composition during CaCO3 crystallization in process with different medium compositions. Crystallization Experiments of Calcium Carbonate at Various Temperatures. The batch experiments have been conducted at various temperatures between 20 and 70 °C. During the crystallization process, the calcium concentration in solution has been controlled by using a calcium-controlled electrode, and the precipitated solids were analyzed by Raman spectroscopy. Raman mixture spectra of all samples were consolidated

810 Crystal Growth & Design, Vol. 9, No. 2, 2009

Carteret et al.

Figure 4. CaCO3 polymorphic mixtures at the end of the precipitation (2 min) as a function of temperature. Raman spectra of lattice modes at 20, 40, and 60 °C.

Figure 3. Spectra of the pure synthesized polymorph (continuous line) and estimated spectra by BPSS (dotted line). (a) Vaterite, (b) calcite, and (c) aragonite.

into a single intensity data matrix and were analyzed by the BPSS method. Figure 3 compares the pure component spectra estimated by BPSS with measured spectra of the pure anhydrous polymorphs. The spectral dissimilarities between the estimated and the reference spectra are very small. The only detected polymorphs in batch experiments are the anhydrous crystalline polymorphs (vaterite, aragonite, and calcite). The calcium concentration in the reactor decreases rapidly and reaches a

plateau in less than 1 min. In this fast-reactive crystallization process, the calcium ions and carbonate groups combine into the amorphous CaCO3, the most instable solid-state phase. Then, the initially formed amorphous calcium carbonate (ACC) transforms to anhydrous crystalline polymorphs.15,19 Some Raman spectra, recorded from 50 to 400 cm-1, are presented in Figure 4 that shows qualitatively the influence of temperature on the polymorphic mixture at 2 min. The polymorphic composition is highly dependent on the temperature. At room temperature, vaterite is the major phase while at high temperature aragonite becomes the major product. The concentration profile against the temperature (see Figure 5a) confirms qualitative results observed in Figure 4 because quite pure vaterite precipitated at 20 °C while quite pure aragonite precipitated at 60 °C. Between 20 and 70 °C, polymorphic mixtures with an increasing content of aragonite are observed. Higher temperature promotes the formation of aragonite, while the content of calcite remains very low and is maximal at 40 °C with a content of 10%. Figure 6 shows that some Raman spectra collected during the crystallization process at 40 °C. At the beginning of the experiment, a ternary mixture composed of a majority of vaterite and aragonite and a minority of calcite is observed. Then, after 7 h, the mixture is mainly composed of calcite. In the meantime, a series of vaterite-aragonite-calcite mixtures with increasing content of calcite are observed. The concentration profile at 40 °C (see Figure 5b) confirms qualitative results observed in Figure 4. At 2 min, the ternary mixture is composed of around 60% vaterite, 30% aragonite, and 10% calcite. Then, the vaterite is transformed to the aragonite and calcite. So, by 1 h, a mixture composed of 40% aragonite, 35% calcite, and 25% vaterite is obtained. In this period, the vaterite dissolution feeds the aragonite and calcite growth. After 1 h, the aragonite dissolution occurs due to the important growth rate of calcite. So, after 6 h, vaterite and aragonite are almost totally transformed to the calcite. Contour plots (dressed in Figure 7) of the proportion of the three polymorphs vs the temperature and the aging time give a global view of the whole of the results (the experimental data points are given in the Supporting Information). The time for the complete phase transformation from the metastable phases to calcite is then observed to be highly dependent on temperature. Indeed, the time to obtain 90% of calcite that is around 10 h for temperatures inferior to 40 °C increases strongly for the highest temperatures. During a precipitation process, several polymorphs can be simultaneously produced, and consequently, the end polymor-

Calcium Carbonate Polymorphs during Batch Crystallization

Crystal Growth & Design, Vol. 9, No. 2, 2009 811

Figure 5. Results of the BPSS analysis. (a) Relative concentrations of the three polymorphs at the end of the precipitation step (2 min) against T and (b) relative concentrations of the three polymorphs during the phase transformation at 40 °C.

Figure 7. Contour plots of polymorphic composition as function of time and temperature.

Figure 6. CaCO3 polymorphic mixtures as a function of time during phase transformation at 40 °C. Raman spectra of lattice modes at 2 min and 2, 4, and 7 h.

phic composition depends on the nucleation and crystal growth kinetics of each polymorph. Very often, kinetic factors dominate in such systems, and thermodynamically metastable phases are preferably produced. The stabilities of theses phases (polymorphs, hydrates, or amorphous phases) are manifested as differences in solubility under given conditions. According to

the known Law of Stages formulated by Ostwald, the least stable phase, having the highest solubility, precipitates first and subsequently transforms to the more stable one. Thus, in our case, the formation of calcium carbonate by mixing of two solutions containing, respectively, calcium and carbonate ions takes place during two well-distinguished periods. The first short period (less than 2 min) is the precipitation (nucleation and growth) one. This period is characterized by a drop in the concentration of calcium and provides a mixture of calcium carbonate polymorphs. The second period (a slow process requiring several days to go to the end) represents the phase transformation from the unstable polymorphs (vaterite and aragonite) to the stable one (calcite). In almost cases, because the supersaturation of the stable polymorph is weak, so, only the unstable polymorph would have to be rigorously produced. Consequently the phase transformation to the stable polymorph would be a very long-time process (several years or centuries) due to extremely low nucleation kinetics of the second one. The fact that the phase transformation in our experiments is a

812 Crystal Growth & Design, Vol. 9, No. 2, 2009

relatively fast process (Figure 7) indicates the existence of the stable polymorph, calcite, at the beginning of the second period. The time required to reach the phase transformation is due both to the polymorphic composition at the end of the nucleation/ growth step and to the kinetic of dissolution/growth. Whatever the temperature, the time is higher than 7 h, and calcite is never the predominant polymorph after the first step of precipitation. The key parameter of the phase transformation from the unstable phases to the stable polymorph may be then assumed to be the calcite nucleation. When the initial mixture is binary, the dissolution of the metastable phase feeds directly the growth of calcite, whereas when the mixture is ternary, the dissolution of the vaterite feeds the growth of both aragonite and calcite. Then, the dissolution of aragonite takes place to feed the growth of the calcite. At 20 °C, the vaterite crystallizes alone; the phase transformation does not occur and the vaterite is the single phase after several 10 h. Under industrial conditions, the calcite is the desired product. The content of calcite after the step of precipitation is maximal at 40 °C with a content of 10%. The contour plots (Figure 7) confirm that at this temperature the phase transformation toward the desired polymorph is the fastest. Conclusion The combination of the Raman spectroscopy of lattice modes with mixture analysis Bayesian approach, BPSS, is an efficient novel method to provide insight into the complex chemistry of the polymorphic systems. It is thus possible to achieve the identification of all of the polymorphs (known or unknown) and the measurements of their relative concentrations. The determination of concentration profiles is possible whatever experimental synthesis conditions. The crystallization of calcium carbonate from CaCl2 and Na2CO3 aqueous saline solutions was followed for various temperatures (20-70 °C). At the end of the nucleation step (2 min), three crystalline polymorphs were obtained. At low temperature, the vaterite form is the major phase, whereas above 50 °C, it is aragonite phase. Whatever the temperature (20-70 °C) the calcite phase, at the end of the precipitation time (2 min), is always present, but it corresponds to the minor phase. With time, the unstable polymorphs transform to calcite. The low content of calcite at the end of the precipitation step induces a time higher than 7 h for achieving the phase transformation (to obtain 95% calcite). Raman measurements coupled with BPSS analysis provide a promising solution and simplified approach to monitor and control the solid-phase transformation during crystallization and precipitation processes, testing the influence of various factors like the solution composition, stirring, mixing rate of reactants, supersaturation level, seeding, etc. An interesting industrial

Carteret et al.

application will be the in situ analysis for an online control of the polymorphic composition within a reactor. Acknowledgment. A.D., E.P., and H.M. express their appreciation to the Arkema group, J. M. Bossoutrot and A. F. Blandin. Supporting Information Available: Relative concentrations obtained by Raman/BPSS of the three polymorphs during the phase transformation at 20, 30, 40, 50, 60, and 70 °C. This material is available free of charge via the Internet at http://pubs.acs.org.

References (1) Bernstein, J. Polymorphism in Molecular Crystals; Clarendon Press: Oxford, 2002. (2) Brittain, H. G. Polymorphism in Pharmaceutical Solids; Marcel Dekker: New York, 1999. (3) Gouadec, G.; Colomban, Prog. Cryst. Growth Charact. Mater. 2007, 53 (1), 1–56, and references therein. (4) Fevotte, G. Trans IChemE, Part A, Chem. Eng. Res. Des. 2007, 85 (A7), 906-920, and references therein. (5) Moussaoui, S.; Carteret, C.; Brie, D.; Mohammad-Djafari, A. Chemom. Intell. Lab. Syst. 2006, 81 (2), 137–148. (6) Moussaoui, S.; Brie, D.; Mohammad-Djafari, A.; Carteret, C. IEEE Trans. Sig. Process. 2006, 54 (11), 4133–4145. (7) Vagenas, N. V.; Gatsouli, A.; Kontoyannis, C. G. Talanta 2003, 59, 831–836. (8) Dickinson, S. R.; McGrath, K. M. Analyst 2001, 126, 1118–1121. (9) Kontoyannis, C. G.; Vagenas, N. V. Analyst 2000, 125, 251–255. (10) Dandeu, A.; Humbert, B.; Carteret, C.; Muhr, H.; Plasari, E.; Bossoutrot, J. M. Chem. Eng. Technol. 2006, 29, 221–227. (11) Agarwal, P.; Berglund, K. A. Cryst. Growth Des. 2003, 3, 941–946. (12) Chakrabarty, D.; Mahapatra, S. J. Mater. Chem. 1999, 9, 2953–2957. (13) Dickinson, S. R.; McGrath, K. M. J. Mater. Chem. 2003, 13, 928– 933. (14) Tai, C. Y.; Chen, F.-B. AIChE J. 1998, 44, 1790. (15) Kitamura, M. J. Colloid Interface Sci. 2001, 236, 318. (16) Kitamura, M. J. Cryst. Growth 2002, 237, 2205–2214. (17) Hu, Z.; Deng, Y. J. Colloid Interface Sci. 2003, 266, 359–365. (18) Wei, H.; Shen, Q.; Zhao, Y.; Wang, D.-J.; Xu, D.-F. J. Cryst. Growth 2003, 250, 516–524. (19) Shen, Q.; Wei, H.; Zhou, Y.; Huang, Y.; Yang, H.; Wang, D.; Xu, D. J. Phys. Chem. B 2006, 110, 2994–3000. (20) Han, Y. S.; Hadiko, G.; Fuji, M.; Takahashi, M. J. Eur. Ceram. Soc. 2006, 26, 843–847. (21) Kawano, J.; Shimobayashi, N.; Kitamura, M.; Shinoda, K.; Aikawa, N. J. Cryst. Growth 2002, 237-239, 419–423. (22) Mazet, V.; Carteret, C.; Brie, D.; Idier, J.; Humbert, B. Chemom. Intell. Lab. Syst. 2005, 76 (2), 121–133; http://lsiit-miv.u-strasbg.fr/lsiit/perso/ mazet/backest-en.htm. (23) Porto, S. P. S.; Giordmaine, J. A.; Damen, T. C. Phys. ReV. 1966, 147 (2), 608. (24) Fong, M. Y.; Nicol, M. J. Chem. Phys. 1971, 54 (2), 579–585. (25) Frech, R.; Wang, E. C.; Bates, J. B. Spectrochim. Acta 1980, 36A (10), 915–919. (26) Gabrielli, C.; Jaouhari, R.; Joiret, S.; Maurin, G. J. Raman Spectrosc. 2000, 31 (6), 497–501.

CG800368U