www.icheme.org/journals doi: 10.1205/cherd.


0263–8762/06/$30.00+0.00 # 2006 Institution of Chemical Engineers Trans IChemE, Part A, January 2006 Chemical Engineering Research and Design, 84(A1): 47 – 59

A STUDY ON BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE Process Monitoring Using ATR-FTIR and Product Characterization by Automated Image Analysis
Department of Chemical Technology, Lappeenranta University of Technology, Lappeenranta, Finland


upersaturation as the driving force of crystallization processes is an essential parameter and its effects on the quality of the product should be carefully investigated. In this paper, a systematic study on cooling rate and cooling mode effects on the supersaturation level and, consequently, on the size and shape of the sulphathiazole crystals has been conducted. In situ concentration monitoring was carried out using ATR-FTIR spectroscopy and the size and shape of the crystals produced were measured by automated image analysis. The simultaneous application of these two modern analytical methods was found to provide useful information for the thorough study of the causes and consequences of different crystallization conditions on the outcome of the process. The results obtained in this study showed a significant increase in the average crystal size with decreasing cooling rate when linear cooling profiles where applied. The average size of the sulphathiazole crystals produced by using the programmed cooling was found to be slightly smaller than the size of the crystals obtained by using linear or natural cooling profiles with the same batch time. The repeatability of the experimental procedure was considered to be good since the relative standard deviations between the results of the repeated batches were rather small in all cases. The variations observed in the properties of the crystals produced could be easily and comprehensively explained by the differences observed in the concentration profiles. Keywords: crystallization; supersaturation; ATR-FTIR spectroscopy; crystal characterization; image analysis.

INTRODUCTION The driving force of all crystallization processes is supersaturation. In cooling crystallization, the supersaturation is typically expressed as the concentration of the crystallized material in excess of its solubility at a given temperature. It is well known that the level of supersaturation during the cooling crystallization process influences the properties of the crystalline product obtained. This has been indisputably confirmed by several experimental results presented in the literature (e.g., Myerson et al., 1986; Jagadeshi et al., 1996; Matthews and Rawlings, 1998; Togkalidou et al., 2001a; Srinivasakannan et al., 2002; Ulrich and Strege, 2002; Lewiner et al., 2002). Commonly required properties of a crystalline material are certain crystal size, narrow size distribution and desired crystal
à ¨ Correspondence to: Dr A. Hakkinen, Department of Chemical Technology, Lappeenranta University of Technology, P.O. Box 20, FIN-53851, Lappeenranta, Finland. E-mail: antti.hakkinen@lut.fi

shape. These properties are determined mainly by the processes of nucleation, growth, attrition, breakage and agglomeration. Supersaturation affects nucleation and crystal growth considerably, both of which accelerate as the level of supersaturation is increased. In order to achieve the previously mentioned crystal properties, it is necessary to somehow minimize the effect of the nucleation process and maximize the crystal growth so that most of the material that is dissolved in the solution could accumulate on the surfaces of a small number of crystals. This can be done by controlling the level of supersaturation during the crystallization process. Nucleation typically occurs at rather high supersaturation levels whereas crystal growth can occur and proceed at considerably lower levels of supersaturation (Mersmann, 1999). Based on these considerations, the optimal crystal properties should be achieved if the supersaturation could be kept at a relatively low level throughout the whole crystallization process. One method for influencing the level of supersaturation during cooling crystallization is to control the cooling rate applied. 47

The occurrence of secondary nucleation might set some limits on the size distribution that can be achieved with controlled crystallization because it creates new crystal surfaces on which the available mass of the dissolved material can deposit. 1996. however. Natural cooling refers to a method where the crystallizer is cooled with a constant temperature coolant and this normally results in very high cooling rates at the early parts of the batch. This causes the supersaturation to rapidly exceed the metastable limit thus allowing massive spontaneous nucleation to take place. The obtained simplified profiles were utilized by Mullin ´ and Nyvlt (1971) and by Jones and Mullin (1974) in seeded crystallization experiments performed with aqueous solutions of potassium sulphate and ammonium sulphate and it could be shown in both of these cases that the median crystal size was clearly increased when compared with the size of the crystals obtained using natural cooling. Attrition nucleation refers to the formation of new nuclei as a consequence of crystal breakage. For an unseeded batch cooling crystallization process of a substance with its solubility depending linearly on temperature. the possibility of repeatedly exceeding the metastable limit during the process becomes likely and this again can lead to worsening quality of the product as a consequence of uncontrolled nucleation. who derived an exact theoretical equation that represented the ideal cooling curve based on constant nucleation and growth rates in a seeded solution.. the simplified cooling curve that should be followed in order to maintain approximately constant supersaturation throughout the whole duration of the batch became: (T0 À T)=(T0 À Tf ) ¼ (t=t)4 (1) In situ Concentration Measurement with ATR-FTIR To get real-time information on the supersaturation level throughout the crystallization process. A lot of work has been carried out during the last few decades on finding the optimal cooling profiles that should be followed in order to perform batch cooling crystallization processes in a controlled manner. a stirrer or other crystals in the suspension. 2000). a reliable solute concentration monitoring technique is needed. The ATR-FTIR technique has proved to be a suitable technique for monitoring crystallization processes due to its wide applicability to different systems. b. 2002). 84(A1): 47–13 . 1974). This equation. Consequently. Tf is the final solution temperature. Secondary nucleation can be divided into attrition nucleation and surface nucleation. where T0 is the initial solution temperature. irregularly shaped crystals. The presented experimental results however also showed that the programmed cooling profile was not very effective in eliminating the occurrence of nucleation during the crystallization since all of the obtained crystal products contained a significant amount of fine crystals. If the applied cooling rate is selected correctly. However. since most compounds absorb radiation in the mid-IR range. As the crystallization process then proceeds. Davey and Garside. In addition. was too complex for general use and it was therefore simplified by making some appropriate assumptions. transformation of the spectral data to concentration information is a critical issue in order to obtain reliable results. T is the solution temperature at time t. Part A. the measurement takes place at the interface between the ATR element and a sample and therefore the existing crystals are assumed not to disturb the measurements. This means that at the early stages of the batch. Traditional regression methods can be applied by correlating the heights or areas of specific peaks. In addition. or alternatively a ratio of specific peaks to a concentration of the measured con¨ stituent can be used (Uusi-Penttila and Berglund. the product obtained with natural cooling often consists of a large number of small. it is possible to reduce the high initial supersaturation peak associated with natural cooling and therefore also to diminish the occurrence of unfavourable spontaneous nucleation after the first primary nuclei have been generated. linear and programmed cooling. Linear cooling profile refers to a constant cooling rate throughout the whole crystallization process. However. secondary nucleation is likely to take place and the number of nuclei that are produced by this mechanism will increase with the mass of crystals in the solution (Mersmann. 2006. the surface area of the crystals increases and the cooling rate increases towards the end of the batch. After the first crystals have appeared in the solution. it is not usually necessary in practice to use the exact forms of these optimal equations. even further optimized and more specific theoretical cooling profiles can be found from the literature. Dunuwila and Berglund. 1997. the rate of supersaturation production has to be slow and therefore also the applied cooling rates need to be extremely low. 1996) and it depends on the level of supersaturation. According to Davey and Garside (2000) it is often only required that a reasonable approximation to the appropriate time – temperature profile is achieved. Fevotte. which are natural. when the surface area of the crystals existing in the solution is small. 2002. traditional regression Trans IChemE. 2001a. 2000). 1996. If the cooling rate chosen is too high.48 Cooling Modes ¨ ¨ POLLANEN et al. Consequently. Even though a large number of different. attrition nucleation also decreases the size of the already formed crystals. Lewiner et al. The cooling methods traditionally employed in cooling crystallization processes can be roughly classified into three different types. Too low cooling rates on the other hand can result in uneconomically long batch times. The objective of such programmed cooling profiles is to ensure that the generation rate of supersaturation always matches the available crystal surface area on which the crystal mass formed is to be transferred (Davey and Garside. the solution is usually supersaturated at a considerably faster rate than it is possible to desupersaturate it by the growth on the existing crystals (Jones and Mullin. as was reported by Jones and Mullin (1974). Surface nucleation on the other hand refers to formation of new nuclei on the surfaces of the crystals existing in the solution or in their immediate vicinity (Mersmann. This should lead to an increase in the average crystal size compared to the crystals produced by natural cooling. t is the time and t is the batch time. Chemical Engineering Research and Design. which is caused by collisions between the crystals and the walls of the crystallizer. One of the first papers that dealt with programmed cooling of batch crystallizers was published by Mullin and ´ Nyvlt (1971).

An ideal technique for characterizing crystalline solids would therefore be such. larger and more uniform product crystals can be obtained than when the metastable limit is exceeded (Fujiwara et al. but the impact of concentration or impurity level on the product properties was not considered in that context (Derdour et al. Characterization of the crystals produced was performed with a modern automated image analyser designed especially for size and shape analysis of dry powders.. which the spectral variables typically represent. the process conditions can be optimized to meet the desired product properties.. the principles of which were described above. 2002. 2001a. The cooling rate and mode had a remarkable effect on the supersaturation level during the unseeded crystallization and consequently on the product properties. Current research has 49 been strongly concentrated on the application of closed loop feedback control based on ATR-FTIR concentration measurements.. Part A. be overcome by automated image analysis techniques. it may reduce the noise level. in fact. The results from the studies in which ATR-FTIR was used for concentration measurement during the crystallization experiments using different constant cooling rates with and without seed crystals showed that the metastable range increased with the cooling rate and also with the overall concentration level. to avoid exceeding the metastable limit during ongoing crystallization and to minimize the batch time have been introduced. 84(A1): 47–13 .... PLS reduces the dimensions of the original data and. (2002). (2005). In this work. Profir et al. Concentration measurement using the ATR-FTIR technique in crystallization of a certain polymorphic form of glycine has been investigated (Doki et al.. (2001a. In complex chemical systems. random variation is always present in the spectra. Manual microscopy measurements are very timeconsuming and for that reason the number of particles that can be analysed in a reasonable period of time is often not sufficient to ensure that the statistical significance of the obtained results is high enough. Liotta and Sabesan. 2002).. PLS enables the linear modelling of correlated variables. Alander et al. at the same time. The average size of the product was observed to be largest when seeding was used (Lewiner et al. 2004). that it could simultaneously provide information on the size and the shape of the analyzed crystals.. 2002. The results obtained from microscopy measurements also suffer from operator to operator variability due to the fact that the human operator needs to constantly make selections regarding the examined particles and measured particle dimensions. and for that reason a single peak can provide inaccurate estimates of the concentration level. 2005) is used in this context. The physical and chemical purity of a product is a major issue in crystallization. can be included in the model. (2004). e. 2001b). The overall concentration level increased by seeding as well (Fevotte.. Liotta and Sabesan. Feng and Berglund. Attempts to simultaneously obtain crystals as large as possible. 2004). no single peak can be found to correlate reliably with concentration. b. for example. Controlled crystallization processes with constant supersaturation levels have been compared to ¨ uncontrolled crystallizations in previous studies (Gron et al. (2002).. the concentration levels were measured with ATR-FTIR from the crystallizations using different constant cooling rates as well as natural cooling and programmed cooling modes. 2004). Calderon De Anda et al.e. Consequently. Most of the difficulties related to manual microscopy can. Variations in crystal properties could be easily and comprehensively Trans IChemE. 2003. the measurement of the level of specified impurity simultaneously with the solute concentration measurement using ATR-FTIR has been introduced. Traditionally image analysis has been carried out using manual microscopy. (2005a. 2003). b. 2001b). A calibration routine that includes validation and preprocessing steps for solute concentration ¨ ¨ prediction in cooling crystallization process (Pollanen et al. Multivariate methods. A large number of variables. Faria et al. 2006. 2002). The size of the product crystals decreased as the level of ¨ supersaturation was increased (Gron et al. The results have shown that by maintaining the concentration level under the metastable limit. (2003). a stabilizing advantage rather than a problem (Wold et al. the bands in the IR spectrum from different constituents often overlap each other.BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE methods should not be applied to correlated variables. 2004). The variations in crystal shape in terms of length-towidth ratio were found to be reduced when seeding was applied (Lewiner et al. Currently the only methods that are capable of separating different crystal dimensions from each other and thus generating appropriate shape factors are the ones that are based on imaging techniques. i. 2004). as was expected theoretically. Bernard˚ Michel et al. 2002. however.. In addition. ATR-FTIR has been applied to monitor the concentration level in crystallization of racemic compounds and process conditions that make it possible to separate an enantiomerically pure compound out of a racemic mixture using a cooling crystallization process (Profir et al. Lewiner et al. The information on the concentration of crystallizing substance allows the possibility of investigating the effects of the driving force changes on the obtained product morphology and crystal size distribution (CSD). Pons et al. 2001b. b) and Ferreira et al. The collinearity between variables is. Liotta and Sabesan. Chemical Engineering Research and Design. In addition. Hurley et al. partial least squares (PLS) regression and principal component regression (PCR) calibration models can be applied to overcome these problems and they have previously been applied to solute concentration prediction from crystallization systems (Togkalidou et al.. consequently.. by Lewiner et al. (2003). Crystallization with feedback control with constant supersaturation was found to provide larger crystal sizes than crystallization with constant cooling rate (Liotta and Sabesan. 2002). the absorbencies of specific compounds of interest can be low and. 1983). which is such an arduous method that its utilization in practical applications is rather difficult. Automated Image Analysis The level of supersaturation during the crystallization process affects not only the size but also the shape of the crystals produced. 2001a) and the number of crystals was found to increase with the cooling rate (Lewiner et al.g. The use of different kinds of image analysis techniques for characterizing crystalline solids has been recently reported... 2003. 2002). spectral points. The PLS calibration can be improved by careful data and model validation steps as well as using an appropriate data preprocessing technique.

3.5. 5.7% (w/w).1 m s21. According to Mersmann and Loffelmann (2000). the solubility of the crystallized material in the considered solvent has to be known. The experimental work can be divided into two parts according to the type of cooling profiles applied. The goal of the second part of the crystallization experiments was to investigate the influence of the applied cooling profile. All crystallization experiments in this study were carried out without seed crystals. Solubility Measurements In order to be able to properly control any cooling crystallization process. The experiments with natural and explained by the differences observed in the crystallization conditions that were detected using the in-line concentration measurement data. that the crystallizer could not be cooled according to the temperature profile calculated from equation (1) in an accurate manner in less than 6 h. tip speeds of this magnitude are typical for laboratory scale crystallizers.0 and 2. The main parts of the equipment were the jacketed 4.0 h for the temperature interval from 808C to 258C. Samples that were taken from saturated homogenous solutions at different temperatures were filtered through a 0.A. diameter 160 mm). Part A. cooling profiles was decided to be 1. The experiments were done with four different rates. 84(A1): 47–13 . The crystallizer was supplied with a three-bladed curved blade impeller (diameter 100 mm) and four baffles. which resulted in cooling times of 14. programmable LAUDA RK 8 KP-thermostat unit and ABB BOMEM MB155S spectrophotometer. Selection of the batch time used in these experiments was done based on the limitations caused by the available thermostat unit and the heat transfer characteristics of the equipment. The solution was then heated from 258C to 858C in 2 h and kept at that temperature for 1 h to make sure that all of the added sulphathiazole was dissolved before the beginning of the cooling stage. it was possible to detect the temperature where the first nuclei spontaneously appeared into the solution as the limit of the metastable zone was exceeded.0 dm3 of solvent mixture into the crystallizer and adding a quantity of sulphathiazole such that it corresponded to the previously determined solubility at 808C. The results of these solubility measurements have already been presented ¨ earlier (Hakkinen et al.9. 250 mm.0. Barcelona. At this temperature. The temperature data collected from the experiments with different cooling rates is presented later in this paper in Figure 3. Crystallization Experiments The equipment used in the crystallization experiments is illustrated in Figure 1.2 and 27. the solution was clearly supersaturated but no crystals were yet present in the solution. The same PC was also used to control the thermostat in such a way that the temperature in the crystallizer accurately followed the predefined cooling profiles. 6. The solubility was calculated based on the measured reduction in the mass of the sample during drying. it was decided that the starting temperature of the programmed cooling profile should be well within the metastable zone. Castellbisbal. and it was observed from on those runs. 2003). and also natural. The temperature in the crystallizer was measured with a Pt-100 sensor and was registered on a PC. The solubility of sulphathiazole in the chosen solvent mixture at temperatures ranging from 108C to 808C was determined experimentally using the gravimetrical method. Aspokem Oy. Finland]. MATERIALS AND METHODS The experimental work performed in this study consisted of unseeded laboratory scale batch cooling crystallization experiments carried out using various cooling rates and cooling profiles. Chemical Engineering Research and Design.0 dm3 glass crystallizer (height Figure 1. 99. 10.50 ¨ ¨ POLLANEN et al.. the crystallization experiments were also carried out with natural and programmed cooling.58C h21. which was found to be 758C. Trans IChemE. All measurements were made with three parallel samples in order to ensure adequate accuracy. The crystallized material in all experiments was pharmaceutical grade (European Pharmacopoeia/ United States Pharmacopoeia) sulphathiazole (Industrias GMB S.08C above the detected nucleation temperature. The purpose of the first part of the crystallization experiments was to examine the influence of the applied cooling rate when the cooling was carried out with a constant cooling rate throughout the whole duration of the crystallization. Helsinki. By using data collected during the previous experiment with a linear (6 h) cooling profile. which were made of PTFE. The most relevant part of the results for the current study will also be given later in this paper in Figure 2. Since all experiments in this study were done without seed crystals. The starting point of programmed. In addition to the linear cooling profiles.2 mm-pore-size syringe filter and evaporated to dryness in a vacuum oven.0. Trial runs were made to find the minimum batch time. which corresponded to a ¨ tip speed of 2. 2006. 9. The agitation rate in all experiments was kept constant at 400 rpm. Spain) and the solvent used in the experiments was a 50/50-mixture (w/w) of deionized water and GC-grade n-propanol [purity min. Experimental set-up. The cooling time for the temperature interval from 808C to 258C was therefore chosen to be 6 h in all experiments with different cooling modes.. The initial batches were prepared by introducing approximately 4.

a calibration set was selected..5 h.e. The Q-chart is calculated from the squared prediction error of the residuals. The temperature range for the measurements was from 258C to 808C and the number of measured points was 513. The shape of the element causes the light beam to undergo two internal reflections in the interface of a sample and the element before the attenuated reflection proceeds to the detector. The predicted variable y was the solute concentration. i. Calibration measurements A calibration routine is necessary in order to obtain quantitative information from IR spectra. 84(A1): 47–13 . which resulted in very high initial cooling rates and considerably lower cooling rates at the end of the batch. The crystallizations with the programmed cooling profile were performed using a PC that controlled the thermostat unit in such a way that the temperature in the crystallizer accurately followed the profile calculated from equation (1). The calibration measurements are done in a liquid phase. The MSPC charts are based on a principal component analysis (PCA) model and are applied to the X matrix. the conditions present in true measurements should be covered. Comparison between the parallel batches suggested that considerable batch-to-batch variations did not exist for any of the monitored properties. From the evaluation of the quality of measured calibration data and previous knowledge of the variation of essential process variables that should be covered in a predictive model. concentration range. Equilibrium was detected when the consecutive spectra were constant. A calibration set (86 points) was used to build the stable calibration model and a separate test set was used to select Trans IChemE. the spectral resolution of 16 cm21 was used and each spectrum consisted on average of 20 consecutive scans. These conditions include: mixing conditions. but in a true crystallization process a solid – liquid suspension exists. As a compromise between the quality of the spectrum and the robustness of the measurements. This means that extrapolation is needed when concentration is predicted. the final temperature of the batch was reached in approximately 1. It was. Inc. consequently. the batch time in all experiments with different cooling modes should be the same. The principle of MSPC is described in (MacGregor and Kourti. Chemical Engineering Research and Design. since spontaneous nucleation can take place due to any mechanical or other distraction in the process. This particular validation step tested the applicability of the model to the situation resembling true crystallization where there is a solid –liquid suspension present and its suspension density is changing throughout the process. The spectral data together with temperature and solvent compositions formed the X matrix. The calibration routine includes sets of measurements with known concentrations. the calibration measurements with known concentrations become precarious. To obtain a stable calibration. Calibration sets were measured using the previously mentioned spectrophotometer settings. 2006. Measured calibration data was validated using multivariate statistical process control charts (MSPC). The most relevant wave number range to be used in the model was selected based on contribution charts. however. In practice. temperature range. ATR-FTIR measured solubilities were then compared to the results obtained from gravimetric measurements.BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE programmed cooling profiles were carried out by cooling the solution first from 858C to 758C with the cooling rate of 9. in the case of crystallization measurements the supersaturated stage is unstable. The experiments with the natural cooling on the other hand were performed by rapidly changing the temperature of the cooling medium to a constant temperature of 258C. solvent composition and. Calibration modelling and validation The spectral data was analyzed using Matlab 6. decided that in order to keep the influence of the mixing conditions equal. the time needed to reach equilibrium was approximately 2 h. Within MSPC. of course. reactor geometry.5 from MathWorks. ATR-FTIR Measurements and Calibration Procedure ATR-FTIR device and settings ATR-FTIR measurements were taken using an ABB BOMEM MB155S spectrophotometer equipped with a Dipper 210 ATR immersion probe with a conical ZnSe element manufactured by Axiom Analytical Inc. In those experiments where natural cooling profile was applied. For this reason the calibration 51 measurements were conducted in undersaturated conditions but concentrations as close as possible to the solubility concentration were measured. Part A. 1995). Hotelling’s T 2 charts from score vectors of the PCA analysis are formed and they reveal whether the variation of variables in the plane of extracted PCs to the model is greater than can be explained by random noise. Water was used as a background spectrum.28C h21 after which the chosen cooling profile was started. The mid-IR spectrum from 4000 to 750 cm21 was collected. ATR-FTIR solubility measurements To validate the model. the solubility of sulphathiazole was measured by adding an excessive amount of sulphathiazole to the solvent and mixing this suspension at constant temperature until the equilibrium concentration was reached. A sulphathiazole concentration from 0 to 30 g sulphathiazole/100 g solvent and a solvent composition 0– 100 w% n-propanol was covered in the calibration measurements. and. This was done from 258C to 808C with increments of 58C. All the crystallization experiments performed in this study were conducted at least twice to ensure adequate repeatability. from the data left out from the model of the observations.5 h. Grams32 software was used to collect the spectra. Therefore also the experiments with the natural cooling profile were performed with a batch time of 6 h by maintaining the crystals in the stirred solution at constant temperature of 258C for the remaining 4. However. Variables can be evaluated within the structure (T 2) and residual (Q) space using variable loadings to detect the essential variables that cause a particular sample to represent a different structure than the other samples. Particularly in this type of situation careful data and model validation steps are required in order to obtain a reliable model. The temperature data collected from the experiments with different cooling profiles will be shown later in Figure 5.

52 ¨ ¨ POLLANEN et al. A PLS model was derived. The four samples. The number of PLS components needed in the final predictive model should be decreased by OSC preprocessing. after the overlapping objects had been removed. The completely automated operation of the analyser meant that the measurements were always made routinely and very rapidly (approximately 20 000 particles in 10 min). By comparing the crystal images provided by the PharmaVision and the morphological parameters of the individual crystals. This should also lead to more accurate predictions of a desired property using the model. The volume of each composite sample was approximately 500 ml that corresponded to roughly 1/8 of the total volume of the crystal suspensions.. The optimal conditions for the image analysis can only be achieved when the crystals are dispersed onto the sample plate in a mono-layer. Part A. 2006. were always taken from a mixed suspension in order to reduce the undesired classification due to the sedimentation of the largest crystals. if the appropriate validation procedures were applied. In this context. since otherwise the obtained results may be significantly distorted as a result of the overlapping of the analysed crystals. The formation of a perfect mono-layer is never possible in practice and therefore the influence of the overlapping needs to be eliminated somehow. The camera was then moved across the sample tray in a preprogrammed way by linear actuators and a large set of digitized video images was automatically acquired. Malvern Instruments. it was possible to detect the objects that contained the overlapping crystals and automatically discharge those from the original data. For this reason. for jwj ¼ jqj. This technique was used here to ascertain that the results obtained truly described the dimensions of the individual crystals.15) that separated all the individual crystals from the images and Figure 2. 1983). Two or three sub-samples (approximately 0. Several such derived components can be calculated and the number of components needed in the model is defined using an appropriate validation criterion. RESULTS AND DISCUSSION Prediction Ability of the Model The solubility of sulphathiazole was measured gravimetrically and using ATR-FTIR and is presented in Figure 2. varied between 69 368 and 148 716. Figure 2 shows that there is a good agreement for these two different measurements and therefore the calibration procedure can be expected to give accurate enough predictions for the studied crystallization process. Chemical Engineering Research and Design. the best performing model. The samples for the analyses were prepared by dispersing the dry crystals evenly onto a 100 mm  100 mm sample plate that was placed on a sample tray underneath a video camera. the human operator bias could be eliminated. where w and q are loadings of the X and Y decompositions respectively (Wold et al. The obtained raw video images were processed using PharmaVision 830 software (version 4. 84(A1): 47–13 . The Influence of Cooling Rate on Supersaturation The concentration profiles from crystallizations using different constant cooling rates (Figure 3) illustrate that the overall concentration level tends to increase when the Trans IChemE. The solubility of sulphathiazole in a 50/50-mixture (w/w) of deionized water and n-propanol at different temperatures.2.3– 0. In addition. The data was pretreated using an orthogonal signal correction (OSC) filter.1. The OSC filtering methods remove systematic variation in X uncorrelated to Y by extracting the vector (OSC component) mathematically orthogonal to Y from the data matrix X (Wold et al. The selected model was further validated by comparing ATR-FTIR solubility measurements to gravimetrically measured solubilities using concentrations higher than the calibration measurement points and to the situation where there is a solid – liquid suspension present. determined a set of various morphological parameters separately for each crystal in the sample. The results clearly indicate that calibration in a different physical state (a liquid phase) from the solubility measurement (a solid –liquid suspension) did not decrease the predictive ability of the model. (2005).. it was possible to analyse enough particles in a reasonable period of time to obtain reliable results. The crystals from these composite samples were gently separated from the ¨ solvent with a Buchner funnel after which they were dried in a vacuum oven. Crystal Characterization The samples for crystal characterization were collected immediately after the crystallization experiments from four different locations in the crystallizer by using a vacuum pipe. which were combined to form a composite sample. The dry crystals were then used for crystal size and shape analyses that were performed using an automated image analyser (PharmaVision 830. the selection of the best performing model was essentially based on the root mean squared error of validation (RMSEV) of the external test set. Ltd). A detailed description of the calibration procedure is presented in ¨ ¨ Pollanen et al.4 g each) were taken from the dried composite samples and analysed separately and the crystal data obtained from those analyses were combined to give an average result. The number of crystals analysed from the different samples in this study. PLS finds a score vector t in a column space of X (t ¼ Xw) and a vector u in a column space of Y (u ¼ Yq) that gives the maximal squared covariance: max(u0 t)2 ¼ max(q0 Y0 Xw)2. 1998).

The width of the metastable zone increases with the cooling rate. The reason is that the cooling rate was very low in the beginning of the crystallization process with programmed cooling. There are several possible reasons. Trans IChemE. When the cooling rate is very low. the mass transfer of the solute molecules on the surface of existing crystals is fast enough to release most of the supersaturation and consequently the supersaturation level remains low. Firstly. All analytical measurements and calibration models have certain uncertainty Figure 3. there were few sudden concentration decreases. The predicted concentration level in crystallization using a 27. which causes a sudden release of supersaturation and decreases the concentration rapidly. By this way the authors on the one hand wanted to visualize the concentration changes due to changing temperature together with the solubility curve. It should be noted that the concentration profiles in Figure 5(b) are presented against temperature. but is also illustrated in Figure 4.28C h21 and 5. but on the other hand to illustrate the driving force changes against the cooling profile used. during the ongoing 53 crystallization process after the primary nucleation. Chemical Engineering Research and Design. The width of metastable zone versus the cooling rate for a system at high temperature where the primary nucleation occurs for the first time. the unstable system may cause distractions to ATR-FTIR measurements.58C h21 were used. Part A. which causes the supersaturation level to increase and remain high. the concentration level tends to lie above the metastable limit continuously and new nuclei are generated throughout the whole crystallization process. ‘steps’. (b) concentration profiles and (c) supersaturation profiles. but the supersaturation levels in Figure 5(c) are illustrated against time. with high cooling rates.58C h21 cooling rate is probably too high to be true. concentration and supersaturation profiles from crystallizations using different cooling modes are presented in Figure 5. This is an expected and logical result. The Influence of Cooling Mode on Supersaturation Temperature. This effect was clearly demonstrated with these results. The estimated supersaturation level [Figure 5(c)] for programmed cooling seems to drop below zero. Then new nuclei will be formed. Figure 4. but the system is most likely in the equilibrium state. The concentration profile from the crystallization process using programmed cooling in Figure 5(c) illustrates that the concentration level after the first primary nucleation was as low as the equilibrium concentration from 75 to 72 which corresponds to almost two hours as illustrated in Figure 5(c). which can be seen in Figure 3(b). When the cooling rate increases. ATR-FTIR results for linear cooling profiles: (a) cooling rates used. 2006.BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE cooling rate is increased. Therefore. which actually is not true. When cooling rates of 9. These steps may possibly be caused by an instantaneous exceeding of a metastable limit at these particular points. Secondly. the solubility decrease with decreasing temperature becomes greater than the mass transfer rate onto crystals. 84(A1): 47–13 . the concentration level may be too high to be reliably predicted using the calibration model derived and the model becomes unstable.

the larger crystals seem to commonly exist as rather long and thin plates. These fine secondary crystals are mainly assumed to be attrition fragments of large primary crystals. level. and both of these values have uncertainties. Therefore the negative values obtained in the supersaturation prediction are within the calculated uncertainty level for this type of calibration. As the cooling rate is decreased to 9. The suspension reached the final temperature at t ¼ 1. It can be observed that besides the large primary crystals. ATR-FTIR results for different cooling modes: (a) cooling modes used. 2006. Part A. As can be seen in the Figure 5(c) the supersaturation level increased at the end of the crystallization using programmed cooling as the cooling rate was increased towards the end of the process. however. Crystal Characterization Examples of the optical micrographs of the sulphathiazole crystals obtained from the different experiments performed in this study are presented in Figure 6. the crystals were kept in a mixed suspension at constant temperature and the concentration was close to the solubility concentration. especially when this particular supersaturation value is calculated from two values measured with this technique: solubility and solute concentration. an obvious change in the crystal shape is detected and. the total batch time was fixed and the profile was fitted to that. calibration. with a constant cooling rate of 9. Chemical Engineering Research and Design. Further decrease in the cooling rate appears to make the largest crystals slightly more rounded.28C h21. which is perhaps caused by the increased influence of grinding and breakage of the crystals as a result of collisions in the crystallizer. The images presented show that the shape of the crystals obtained with the fastest cooling rate seems to be highly irregular which may be a result of uncontrolled primary nucleation.28C h21. the solubility of sulphathiazole increases exponentially with respect to temperature and not linearly as was assumed in the approximations concerning the cooling profile described by equation (1). also a significant quantity of finer crystals exist in all samples and the shape of those fine crystals is generally more rounded than the shape of the larger ones. 84(A1): 47–13 .5 h [Figure 5(a)]. (b) concentration profiles and (c) supersaturation profiles. The uncertainty in the concentration level prediction with current calibration technique is approximated to be +1. The suspension cooled down rapidly and practically all the crystals were formed in a few minutes. appears to be slightly larger than with the other profiles and their shape seems to be more needle-like. major changes cannot be detected in the size or shape of the largest crystals compared with the crystals obtained using the linear profile with the same batch time.54 ¨ ¨ POLLANEN et al. These crystals also appear to be more uniform in shape. One reason was that the process conditions were not optimized using the crystallization kinetics and thermodynamic properties when determining the programmed cooling profile but instead. Concentration profiles from crystallization with natural cooling show that the concentration level decreased rapidly at the very moment the temperature of the cooling medium was set to a constant value of 258C. The product crystals from the experiments with the natural cooling profile are clearly smaller than the crystals obtained with the other profiles and their shape appears to be more irregular. In addition. It can be concluded that the programmed cooling did not maintain a constant supersaturation level. in particular. which have been formed as a result of grinding effects caused by the mixing during crystallization.5 h. For the remaining 4. When the crystallization is carried out using the programmed cooling profile. which is probably a result of a decrease in attrition since the majority of the crystals has now been formed during the final parts of the Figure 5.e.00 g solute/100 g solvent. This subject could have been overcame by adjusting the restriction for negative values but the authors wanted to show actual predictions from model which shows also the uncertainty due to Trans IChemE. i. The size of the smaller crystals obtained with programmed cooling.

(b) constant cooling rate of 9. These parameters were readily provided by the PharmaVision-software. For a perfect circle. (c) constant cooling rate of 5. When the cooling rate is decreased. This kind of result can be for several different factors. The crystal shape distributions on the other hand are given as differential roundness distributions by the number of crystals and are shown in Figure 7(b) and Figure 8(b). The crystal length refers to the largest projection of the crystal. since the crystal size distribution with studied cooling rates is rather wide. the roundness equals 1. (e) natural cooling profile and (f) programmed cooling profile. It can be observed that the widest size distribution was obtained when the cooling rate was the highest. the resulting size distributions can be seen to become slightly narrower. The supersaturation profile for the experiments with the highest cooling rate implied that the level of supersaturation was extremely high and it was already suggested earlier that one reason for this may be that spontaneous nucleation has occurred throughout the whole duration of the crystallization. 2006. the crystal samples were analysed using an automated image analyser. For this purpose. and for a needle shaped crystal. who suggested that the final product size can be strongly determined by attrition for crystals above 100 mm. crystallization process.98C h21. It was therefore possible to use this data to examine the relationship between the size and shape of the crystals. The largest size found in all samples was approximately equal and this could imply that the maximum attainable crystal size was limited by the mixing conditions applied in the experiments. The data obtained from the image analyses contained several morphological parameters that were measured separately for each individual crystal. It can also be further speculated that in fact the largest crystals are likely to suffer most from attrition because they have higher collisional probabilities due to their larger area and mass (Matthews and Rawlings. It was decided to describe the size and shape of the crystals by crystal length and roundness. respectively. Spontaneous nucleation has probably occurred repeatedly also Trans IChemE. If this were not the case.58C h21. The Influence of Cooling Rate on the Size and Shape of the Product Crystals Crystal size distributions of the samples obtained from the experiments with different constant cooling rates [Figure 7(a)] show that the cooling rate clearly influenced the size of the produced crystals.28C h21. (d) constant cooling rate of 3. This trend is especially obvious when the relative amounts of the smallest crystals are compared whereas truly significant changes in the size of the largest crystals cannot be observed. 1998). one could expect the largest crystals produced with the lowest cooling rates to be clearly larger than those produced with rapid cooling. Part A.BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE 55 Figure 6. the value of roundness approaches 0. The overall conclusion that can be made from the microscope images is that sulphathiazole crystals obtained from the 50/50-mixture of water and n-propanol tend to grow as elongated rods and that the shape of the crystals produced in this study seems to be strongly influenced by the attrition occurred during the crystallization process. First. the roundness of the crystals was plotted against the corresponding crystal lengths as shown in Figure 7(c) and Figure 8(c). It was already shown in Figure 3(c) that an increase in the cooling rate resulted in an increase in the supersaturation during the crystallization. To obtain numerical information concerning the size and shape of the crystals produced in this study. The crystal size distributions shown in Figure 7(a) and Figure 8(a) are given as differential distributions where the projected surface area of the crystals is plotted against the crystal length. The observed differences in the size distributions can also be explained by variations in the rate of spontaneous nucleation between the different cooling rates. The highest cooling rate resulted in a very wide crystal size distribution which can be explained by massive uncontrolled nucleation. 84(A1): 47–13 . whereas roundness is a measure of the length to width ratio. The shape of sulphathiazole crystals obtained by batch cooling crystallization experiments from a 50/50-mixture of deionized water and n-propanol using (a) constant cooling rate of 27. A similar conclusion has also been made ¨ by Mersmann and Loffelmann (2000).58C h21. Chemical Engineering Research and Design. it can be assumed that attrition nucleation in crystallization of sulphathiazole plays a considerable role.

Part A. The crystal shape distributions in Figure 7(b) show that although differences between the shapes of the crystals produced with different cooling rates are rather small.56 ¨ ¨ POLLANEN et al. The roundness of the crystals seems to increase with increasing batch time. This conclusion can also be justified Trans IChemE. the direction of the observed changes is systematic. when cooling rates of 9. Image analysis results for linear cooling profiles: (a) crystal length distributions. experiments that resulted in the narrowest crystal size distribution. Image analysis results for different cooling modes: (a) crystal length distributions. At the same time. be detected in the supersaturation profile for the experiments with the lowest cooling rate.28C h21 and 5. 84(A1): 47–13 . a population of small irregularly shaped fragments is formed. Figure 8. (b) roundness distributions and (c) crystal shape versus crystal size. Figure 7. Such steps could not. Chemical Engineering Research and Design.58C h21 were used since obvious steps were observed in the supersaturation profiles for those experiments. (b) roundness distributions and (c) crystal shape versus crystal size of the samples. 2006. however. Also this result can be explained by the effect of attrition because it is reasonable to believe that due to attrition the sharp edges of the crystals are smoothed.

58C h a 21 RSD refers to the calculated relative standard deviation between the means of two similar batches.2 mm 112. thus narrowing the size distribution from the upper end. and it can therefore be assumed that the crystal shape is affected by attrition especially with the largest crystals due to their higher collisional probabilities. The first is once again assumed to be attrition.6 mm Average crystal length + RSDa 114.472 0. all crystallization experiments in this study were repeated in order to ensure adequate repeatability.0] 116. which implies that considerable variations between the repeated batches did not exist.200 mm) in all samples are practically equal and significant differences can only be seen when the largest crystals are compared. If this holds true.475 0. that despite the constant low supersaturation. the crystal suspension was mixed for approximately 4. This phenomenon could therefore narrow the distribution from the lower end. The size distribution of the crystals obtained by the programmed cooling profile can be seen to lie between the other two distributions.454 + 3. Observation of the presented distributions reveals that only moderate differences can be observed between the different samples. Trans IChemE.5% 0. however.1% 78. Several reasons may explain these observations. Presumably the size distribution right after the cooling period was even wider than was observed with the fastest cooling rate in Figure 7(a).5 h at a constant temperature and the supersaturation during that period was practically negligible [Figure 5(c)].9 mm + 3.474 + 0.2 mm 93. The supersaturation profiles presented in Figure 5(c) implied that programmed cooling could not maintain the supersaturation constant throughout the whole duration of the crystallization. Part A.4 mm 79. One reason may be the highly unlinear solubility curve of sulphathiazole (Figure 2) and another possible explanation is once again the relatively high rate of secondary nucleation. which is expected to decrease the maximum size of the crystals.6% 95. The results obtained with the different cooling rates implied that sulphathiazole crystals were highly sensitive to mechanical attrition.478 0. Statistical values for the crystal samples obtained with different cooling rates. The temperature profile presented in Figure 5(a) showed that with natural cooling the final temperature of the batch was reached very rapidly.BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE with Figure 7(c) where the crystal shape is plotted against the corresponding crystal size. however.0% 0. It should.456 + 1.3 mm + 2.2 mm + 2. be mentioned here that the chosen batch time probably significantly influenced the properties of the final product. The size and the relative amount of small crystals (. Ostwald ripening may cause the smallest crystals to dissolve and their dissolved mass to be deposited on large crystals.3% Cooling rate 3. Table 1 shows the mean crystal sizes and shape factors. 84(A1): 47–13 . The Influence of Cooling Mode on the Size and Shape of the Product Crystals The crystal size distributions of the samples obtained with different cooling profiles are shown in Figure 8(a). This can be considered a somewhat unexpected result because it was assumed that a programmed cooling profile would have produced the largest crystals. The increase in the supersaturation towards the end of the batch suggests that the cooling rate applied during the final stages was too high and it is therefore likely that spontaneous nucleation has occurred at the end of the batch.5 mm 106. The table also shows the relative standard deviations between the means of the parallel batches that are presented here to quantify the statistical accuracy of the sampling and analysis procedure. An obvious consequence of this is that the number of crystals has probably increased and therefore the maximum attainable crystal size has become smaller.6% in all cases. It is possible that the observed size distribution for crystals produced by natural cooling is considerably narrower than it would have been immediately after the final temperature of the batch was reached. The changes in the crystal properties are assumed to be a consequence of two different processes. considerable changes in the crystal properties during this mixing period have occurred. which were all calculated from the number distributions. it can be concluded that the crystals formed at the beginning of the batch have probably broken already during the initial slow cooling period.443 0.98C h21 5.7 mm 76. On the other hand.5 mm 101. The time required by Ostwald ripening to proceed. When the crystallization was carried out using the natural cooling profile. The number of crystals in the sample Batch Batch Batch Batch Batch Batch Batch Batch 1: 2: 1: 2: 1: 2: 1: 2: 83 115 86 106 125 69 148 131 351 549 247 052 297 368 716 792 Mean crystal length D[1. It can therefore be concluded that the simplifying assumptions made to obtain the equation for the 57 programmed cooling profile were not appropriate for the case of sulphathiazole.451 0.4% 0.3% 103. The calculated relative standard deviations are less than +3. It is assumed. A minimum value for roundness can be found in all cases. The shape of the smallest crystals with all cooling rates clearly tends to approach a sphere and significant decrease in the roundness values can be observed with increasing crystal size. depends strongly on the Table 1. which consequently grow.9% Mean roundness 0.58C h21 9. As was already mentioned earlier.28C h21 27. The narrowest distribution is obtained when the crystallization was carried out using natural cooling and the widest distribution results from the experiments with the linear cooling profile. however. 2006.485 0.460 0. Chemical Engineering Research and Design. Since the batch time was fixed to 6 h.4 mm + 3.482 + 1.3 mm 97. a rather narrow crystal size distribution was obtained.464 Average roundness + RSDa 0. for all batches conducted with different constant cooling rates.

The product properties varied strongly in terms of crystal size and shape and the observed variations could be easily and comprehensively explained by the differences in the concentration profiles.9% 0.458 + 0.Z. and Garside. A minimum value for roundness can again be found in all cases and it can be observed that with natural cooling this minimum value occurs at a clearly smaller crystal size than with the other two profiles. The crystal shape distributions shown in Figure 8(b) show that the differences between the samples crystallized with different cooling profiles are rather large. the differences observed in the properties of the produced crystals were rather small. Chem Eng Sci. REFERENCES ˚ ˚ ¨ Alander.. J.386 Average roundness + RSDa 0. Davey.4 mm 91. 84(A1): 47–13 . by image analysis. Calderon De Anda.S. However. 2005b. As an overall conclusion it seems that the size and shape of sulphathiazole crystals is mostly affected by secondary nucleation. and Rasmuson. J.4% 0. X.455 0.1% in all cases. The number of crystals in the sample Batch Batch Batch Batch Batch Batch 1: 2: 1: 2: 1: 2: 125 69 113 99 124 147 297 368 512 795 545 742 Mean crystal length D[1. 62–73 (Oxford University Press Inc.7% Mean roundness 0.0 mm 88. All of the crystals had been formed before this period and therefore the effect of crystal breakage on their shape is believed to be more remarkable than with the other samples. The product size and shape characterization was done with an automated image analysis technique which provided a method to simultaneously evaluate the size and the shape of the product crystals in a plane of two dimensions.. Statistical values for the crystal samples obtained with different cooling modes. 15(7): 785–797. Uusi-Penttila. 130(1–3): 298–306. The crystals produced with the programmed cooling profile are clearly more elongated than the crystals obtained by linear or natural cooling.2 mm Average crystal length + RSDa 95. Quantification. This can be explained by the previously made assumption that a large fraction of the crystals produced with programmed cooling was probably formed in the final parts of the batch. Trans IChemE. the concentration profiles obtained with the programmed cooling implied that the supersaturation could not be maintained constant which is probably due to the exponential solubility curve of sulphathiazole. The largest roundness values can be found with crystals produced by natural cooling and this implies that their external appearance has been significantly affected by attrition.1% 92. The results presented in this paper demonstrated the affect of cooling rate and mode on the supersaturation level during the unseeded crystallization process. From molecules to crystallizers.4 mm + 3. As was expected. X. J Process Control. USA).. Powder Technol. The relative standard deviations between the means of the parallel batches are less than +3. J.456 + 1. Davey and Garside (2000) have stated that it is nowadays recognized that attrition nucleation is the most significant nucleation mechanism in crystallizers for materials with high or moderate solubility. 2002. Table 2 shows the mean crystal sizes and shape factors for all the batches conducted with different constant cooling modes. K. X. M.7 mm + 1. Wang. of effect of operational conditions on size and shape of precipitated barium sulphate. Lai. It can therefore be concluded that most of the effects caused by changes in cooling conditions could be almost completely overruled by the dominating influence of attrition.. and Vivier.J. 60(4): 1053–1065. Bernard-Michel. The results showed that by measuring an essential process parameter and elaborately characterizing the product. Calderon De Anda. Table 2. Chemical Engineering Research and Design. K.N. New York. M..461 0. 2006. 87(2): 135 –147.3 mm 97. Figure 8(c) shows that the shape of the crystals depends significantly on their size. A. as was already assumed earlier. the conditions leading to certain product properties could be widely and extensively considered and also successively related to known theory of crystallization processes. Part A. and Roberts. Classifying organic crystals via in-process image analysis and the use of monitoring charts to follow polymorphic and morphological changes.451 0. 2005a.58 ¨ ¨ POLLANEN et al.6 mm + 0. The in situ concentration measurements were carried out using an ATRFTIR immersion probe and the concentration predictions were calculated using a PLS calibration model.382 + 1.6 mm 89..C.J. Multi-scale segmentation image analysis for the in-process monitoring of particle shape with batch crystallizers. These crystals have most likely suffered less attrition than the ones formed earlier and their shape is therefore closer to the preferred needle-like shape of the sulphathiazole crystals grown in unmixed conditions.. properties of the solute/solvent combination and it is therefore difficult to estimate the actual effect of it in this case.9 mm 93. This kind of information can be considered essential since it enables the optimization of the process conditions so that the product properties can be more efficiently controlled. which again can be considered as an indicator of rather good repeatability regarding the performed experiments and the sampling and analysis procedure. CONCLUSIONS The effect of different cooling conditions on the supersaturation level and consequently on the outcome of sulphathiazole crystallization was studied.379 0. H. E. B. Characterization of paracetamol agglomerates by image analysis and strength measurement. This statement appears to be valid for sulphathiazole based on the results obtained in this study. R...0] 93.5% Cooling mode Linear Natural Programmed a RSD refers to the calculated relative standard deviation between the means of two similar batches. Pons.Z.3% 88.460 0. The reason for this is assumed to be the long constant temperature mixing period at the end of the batch. 2003. and Roberts.. Although the cooling conditions applied in this study varied considerably. Wang. Chem Eng J.M. 2000. the supersaturation level increased with the cooling rate and a natural cooling mode resulted in a high initial supersaturation peak..

Liotta.. F.N. 237–239(Pt. Ind Eng Chem Res.. Chemometrics Intell Lab Syst. A. and Braatz. Yokota. M.. Puel. C. AIChE J. Cryst Growth Des. and Nyvlt. G.. On-line ATR FTIR measurement of supersaturation during solution crystallization processes. ¨ ¨ ¨ ¨ Hakkinen. F. ¨ Gron.... R... A study on crystallization of oxalic acid in batch cooling crystallizer. J Chem Eng Jpn. Trans IChemE. 8(3): 488– 494. G. D. 1999. 2002.. Martens. D. agglomeration and control. 16(3): 125–129.A.-G. 23(1): 11–15. 76(1): 25 –35. N. J Cryst Growth. L. S. Fevotte. Molecular packing. Chem Eng Trans. 3): 2130– 2135. Chem Eng Sci. and Rasmusson A. 1996. 2001a. Chemical Engineering Research and Design. E. Jones. ` Pons. 44(5): 1119–1127.D. ATR FTIR spectroscopy for in situ measurement of supersaturation. Study of the crystallization of mandelic acid in water using in-situ ATR-IR spectroscopy.. 226(2–3): 348 –362. M. P. 2002.. Wold.. Programmed cooling of batch crystallizers. 128(2–3): 276–286.-P. Feyo de Azevedo. Solute concentration prediction using chemometrics and ATR-FTIR spectroscopy. V. N.A.W. F. and Puel. 59 MacGregor J. K. Jagadeshi.D. 74(A7): 812–820. and Wold H. and Kourti T. Fevotte G. ¨ Wold S. Monitoring and feedback control of supersaturation using ATR-FTIR to produce an active pharmaceutical ingredient of a desired crystal size.A.. L. morphological modeling. Furusjo. Dunuwila. L. Klein. Jones.. Part A. Spectroscopic monitoring of environmentally benign anti-solvent crystallization. 2003. and Fevotte. and Tavare. 2004. 2003. M. M. Lindgren... Some aspects of the importance of metastable zone width and nucleation in industrial crystallizers. Feyo de Azevedo. A.. Statistical process control of multivariate processes.K. Orion Pharma (Finland). P. F. Klein.G. 1974. and Roberts. R.. Kubota. L. V. J. Louhi-Kultanen. ATR-FTIR for determining optimal cooling curves for batch crystallization of succinic acid. 1996..-R. Seki. Myerson. 2004.. Ferreira. and Rao.C. Danielsson. Vasanthakumar. H. Chem Eng Res Des.. 29(5): 865– 873. Chem Eng Sci. 84(A1): 47–13 . ATR-FTIR in monitoring of crystallization processes: Comparison of indirect and direct OSC methods... J Cryst Growth. L. Iyappan. A. 25(4): 925–929.. 1995. J. and Kubota.P. K.S.D. Authelin. 4(4): 711–715. M. Cryst Growth Des. S. M. Cryst Growth Des. M. 1997.S... M.. K. B. H. Lewiner.. R.. and Andrews.. AIChE J. Davidson.. Cryst Growth Des. J. Srinivasakannan. Experimental design and inferential modeling in pharmaceutical crystallization.. Rocha... 1998. D.. and Carvin. 2005. N. Togkalidou. Patel. Doki. and Nystrom.. and image analysis of cyanazine crystals precipitated from aqueous ethanol solutions. Louhi-Kultanen.. 1996. Mersmann. J.. C. 1971. 2(5): 363– 370. Batch crystallization of a photochemical: modeling.... 973: 286– 293. 133(1–3): 54–67.A. Using image analysis to look into the effect of impurity concentration in NaCl crystallization. 29(1): 105 –118. A. Braatz. Pollanen. Control of Engineering Practice. 42(1): 198–206. 2002. H. F.. M. J. Programmed cooling crystallization of potassium sulfate solutions. 56(6): 2069–2084... Org Process Res Dev. and Mullin. 179(1–2): 185– 193. Paracetamol crystallization using laser backscattering and ATR-FTIR spectroscopy: metastability. 166(1–4): 967 –970.B. 2003.. Large and mono-sized product crystals from natural cooling mode batch crystallizer. Morphological analysis of pharmaceutical powders. O. Decker. Chem Eng Res Des. Powder Technol. and Vivier. Lewiner. Finland). Fevotte. Fevotte. S. control and filtration. and Loffelmann. Pons.... and Strege. A.. Supersaturation and nucleation. B. J. Ohman. Improving batch cooling seeded crystallization of an organic weed-killer using on-line ATR FTIR measurement of supersaturation. 2001b. A. Cryst Growth Des. Hakkinen A. 4(5): 949– 953. Solvent selection and batch crystallization. 41(5): 1321–1328..... M. 2(5): 449 –452. H. J. Chow. K.. Orthogonal signal correction of near-infrared spectra. Yokota.W. Hurley. the Graduate School in Chemical Engineering (Finland) and the Academy of Finland are acknowledged for their financial support. 2001a.D. A. 38(4–6): 345 –353. 1983. New perspectives for the on-line monitoring of pharmaceutical crystallization processes using in situ infrared spectroscopy.. Crystallization and precipitation. Faria. 2002. The manuscript was received 13 April 2005 and accepted for publication after revision 7 December 2005. and Puel. H. Togkalidou.. Process control of seeded batch cooling crystallization of the metastable a-form glycine using an in-situ ATR-FTIR spectrometer and an in-situ FBRM particle counter. Mersmann. Chemometrics Intell Lab Syst. An online strategy to increase the average crystal size during organic batch cooling crystallization. J Cryst Growth. F. 2002. 1986. 2000. Crystallization and precipitation: the optimal supersaturation. ¨ ¨ ¨ Pollanen K. E. 214(2): 263–278. Chem Biochem Eng Q.L. 1998.N. Chem Eng Technol... 2002. ¨ Mersmann. Borissova. Calibration and applications on three solute solvent systems..A. Puel. Powder Technol. and Weiping. and Rawlings. and Braatz. 47(1): 160–168. F. R. F. G.S. Int J Pharm... Quantification of the morphology of sucrose crystals by image analysis. S. A.. A. N.J. and Sabesan V. Sato. ACKNOWLEDGEMENTS The National Technology Agency (TEKES. Reinikainen S. A. F.. 83(A4): 331 –338... Delcour. J Cryst Growth.. Matthews.. J. ´ Mullin. N.P. In-process ATR-FTIR spectroscopy for closed-loop supersaturation control of a batch crystallizer producing monosodium glutamate crystals of defined size. L.. Ind Eng Chem Process Des Dev. N. 2(4): 273–279. Antti.. 2002. and Lopes. Johnson. Rocha. P. 44(1– 2): 175–185.G. G. 26(3): 369 –377. H. 3: 931 –936.. The multivariate calibration problem in chemistry solved by the PLS method. H. 3: 403–414.. Fujiwara.. 2005. Powder Technol. 129(1–3): 1–7.F.. F. A. J.B. R.. Asatani. J Cryst Growth. ¨ Uusi-Penttila. and Berglund. Takano. Lewiner. K. Chem Eng Process. T. 2006. Faria. 2001b. Vivier. Ind Eng Chem Res.. ˚ ¨ Profir. Ma. L.BATCH COOLING CRYSTALLIZATION OF SULPHATHIAZOLE Derdour. Chem Eng Sci.. F. Feng. Klein. K. H. and Berglund K. V. 2003. A. and Hammond. Ulrich. Real time evaluation of the concentration of impurities during organic solution crystallization. T. 2004. and Berglund. Lecture Notes in Mathematics. Fujiwara. 231(4): 534–543.G.. Pressure filtration of sulfathiazole crystallized by batch cooling crystallization: the influence of solvent composition.P. and Pailleres-Hubert. and ¨ Nystrom. 2002.

Sign up to vote on this title
UsefulNot useful