Crystallization of Glycine by Drowning-Out Combined with Fines

Aug 6, 2012 - ABSTRACT: The DFC process (a combined process of drowning- ... grow further by the dissolution of fines generated by drowning-out...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/crystal

Crystallization of Glycine by Drowning-Out Combined with Fines Dissolution and Cooling Process with in Situ Control using Focused Beam Reflectance Measurement and Attenuated Total Reflection− Fourier Transform Infrared Spectroscopy Jun-Woo Kim and Kee-Kahb Koo* Department of Chemical and Biomolecular Engineering, Sogang University, Seoul 121-742, Korea S Supporting Information *

ABSTRACT: The DFC process (a combined process of drowningout, fines dissolution, and cooling crystallization) was applied for the crystallization of glycine. The DFC process encourages seed crystals to grow further by the dissolution of fines generated by drowning-out. Therefore, glycine crystals obtained through the DFC process have relatively large size and uniform size distribution with high product yield compared with those obtained by conventional crystallization methods. In the present work, an operating strategy of the DFC process with in situ control was proposed. First, antisolvent is injected until generation of fine particles is assumed to be completed. Second, fines dissolution by heating is conducted until fine particles are mostly dissolved. Here, generation and dissolution of fine particles are monitored by focused beam reflectance measurement and attenuated total reflectance−Fourier transform infrared. The main advantage of the in situ controlled approach demonstrated in the present work is no requirement for any predetermined data and a reduction of the operating time compared with the previously proposed DFC process, in which operating profile is predetermined by the mass balance equations.



INTRODUCTION Crystallization is a key purification and separation technique for the extraction of crystalline materials, such as pharmaceuticals, agrochemicals, minerals, and explosives, from a saturated solution.1−5 Cooling and antisolvent addition are typical methods to induce a supersaturation, which is the driving force for crystallization, because of their simplicity. However, cooling crystallization often results in a poor yield when the solubility is reasonably high at low temperature, and crystallization by drowning-out often leads to very fine crystals due to extremely high local supersaturation generated upon addition of antisolvent.6−8 In our previous work, it was found that fines dissolution by heating acts as an effective bridge between cooling crystallization and drowning-out by the following procedure (Figure 1):9 (a) prepare seed crystals in a crystallizer; (b) perform drowning-out by rapid injection of antisolvent to induce fine crystals; (c) dissolve fine crystals by heating; (d) perform cooling crystallization to grow seed crystals; (e) repeat the process from b to d until the desired yield is obtained. This procedure was named the DFC process (a combined process of drowning-out, fines dissolution, and cooling crystallization). In crystallization by the DFC process, it was shown that relatively large sized crystal with uniform crystal size distribution (CSD) and high product yield can be obtained compared with conventional crystallization processes, because fine particles © 2012 American Chemical Society

induced by drowning-out are dissolved into the solution and then dissolved solute contributes to the growth of the remaining seeds by cooling. Complete dissolution of fine particles is a main challenge to obtain uniform CSD in the DFC process. In our previous work, it was made by a simple mass balance model in which the mass of fine particles produced by drowning-out equals that dissolved by heating, and the empirical parameter was employed to consider partial growth (Figure 1b) and dissolution (Figure 1c) of seed crystals.9 However, this approach requires time-consuming trial and error experiments to estimate an empirical parameter. In addition, when the process conditions are changed such as scale-up, those experiments have to be done again because the empirical parameter may be a sensitive function of CSD, fluid dynamics of solution, crystallizer dimension, and so on. In the present work, the DFC process was applied for the crystallization of glycine from a cosolvent of water (solvent) and ethanol (antisolvent) combining in situ techniques such as focused beam reflectance measurement (FBRM), which provides chord length distribution (CLD) of particles, and attenuated total reflection−Fourier transform infrared specReceived: June 25, 2012 Revised: July 24, 2012 Published: August 6, 2012 4927

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design

Article

Figure 1. Schematic diagram of the DFC process. particle size indicator compared with no weighted mean chord length.10,11 Liquid phase concentration of the glycine/water/ethanol system was measured by ATR-FTIR (ReactIR-15 system, Mettler-Toledo, USA). Infrared spectra of the solution were collected in the 4000−650 cm−1 region with a resolution of 4 cm−1 and recorded over 15 s (50 scans). Because of the low penetration depth, typically 2−3 μm, the ATR probe allows for exclusively liquid-phase monitoring in the presence of solid particles.12 Each spectrum of all data was smoothed using a moving-average filter over three points and baseline-corrected. Solubility Measurements and Thermodynamic Modeling. For the solubility measurements, glycine suspension at a given composition of cosolvent and temperature was prepared and maintained with ATR-FTIR monitoring until the glycine concentration reached a near constant value, which was determined as the solubility of glycine. This procedure was repeated at temperature ranging from 293.15 to 335.15 K and ethanol concentration from 0 to 50 wt % on a solute-free basis. For the water (1)/ethanol (2)/glycine (3) system, at a given temperature T, the solubility of a glycine in a single solvent or cosolvent can be calculated using the Schröder−van Laar equation:13

troscopy (ATR-FTIR), which provides solution concentration. That equipment supports the in situ controlled strategy of the DFC process without any predetermined operating profiles. Product quality and process time of the in situ controlled DFC process were compared with the mass balance model-based DFC process and the combined processes of conventional drowning-out and cooling crystallization.



EXPERIMENTAL SECTION

Materials. α-Glycine (99+%) was purchased from Tokyo Chemical Industry (Japan). The triple-distilled water produced by a distillation apparatus (Younglin, Ultra 370 series, Korea) was used as a solvent. Ethanol (99.5+%, DaeJung Chemicals, Korea) was used as an antisolvent and a wash liquid for product. Experimental Apparatus. The experiments were carried out in a 500 mL jacketed glass crystallizer (inner diameter of 12.3 cm and inner height of 14.0 cm). Solution was stirred by a Teflon-coated two straight blade impeller (width of 7.5 cm, height of 1.7 cm, and thickness of 0.3 cm). The agitation speed of 150 rpm was chosen to suspend crystals without any vortex formation during crystallization. Temperature of the solution was measured by a resistance temperature detectors equipped at ATR probe and controlled with two thermostats (Polyscience, model 9710, USA). Each thermostat was maintained with a constant temperature of heating and cooling water. It helps swift temperature change without any complicated process control devices. Ethanol was injected in one-pot using a syringe or fed at constant flow rate using a syringe pump (KD Scientific, KDS 230, USA). The product suspension taken from the crystallizer was filtered over a jacketed glass filter funnel equipped with an aspirator. To prevent undesired crystal growth and nucleation, the temperature of the glass filter funnel was kept the same as that of the crystallizer by a thermostat. The filtered glycine crystals were washed with ethanol three times. Then, the glycine crystals were dispersed onto a Petri dish and dried in a desiccator with silica gel. In Situ Characterization Techniques. The CLDs measured by FBRM (Lasentec, S400A, USA) were grouped into 90 channels from 0.8 to 1000 μm. From CLD data, two values were calculated by the FBRM software as follows: the number of chord length measurements for the whole size range per unit second, total counts/s, is defined as

∑ ni i=1

m

ln γi =

(1)

k

k

m

∑l = 1 Glixl

m

+

∑ j=1

m ⎛ ∑r = 1 xrτrjGrj ⎞ ⎜ ⎟ τ − m ij m ∑l = 1 Gljxl ⎜⎝ ∑l = 1 Gljxl ⎟⎠

xjGij

(4) (5)

where α is the nonrandomness factor of the mixture, m is the number of components (m = 1, 2, and 3 indicate water, ethanol, and glycine, respectively) and τ is assumed to be a temperature-dependent parameter. Here, α12, α13, α23, τ12, and τ21 were obtained from elsewhere,16,17 and τ13, τ31, τ23, and τ32 were obtained by fitting the experimental solubility data using a Nelder−Mead Simplex minimization routine.18 The predicted x3 was converted to the equilibrium concentration of glycine in g/g solvent, c3*, (x3MW3/(x1MW1 + x2MW2)). Here, MW is molecular weight. This solubility will be used in calculation of the amount of injected antisolvent in the mass balance model-based DFC process. Solubility measurement by a gravimetric method was also carried out. At a given temperature, an excess quantity of raw glycine was injected in water or cosolvent of water and ethanol. The slurry was

∑i = 1 niMi 3 ∑i = 1 niMi 2

∑ j = 1 τjiGjixj

Gji = exp(− αjiτji)

where k is the number of channels in FBRM and ni is the number of counts measured in an individual channel. Second, the square weighted mean chord length (SWMCL) is defined as SWMCL =

(3)

where x, γ, and R are the mole fraction, activity coefficient, and gas constant, respectively. Subscript number 3 indicates third component, glycine. Tm and Δmh are the melting temperature and enthalpy of the solute, respectively. In the present work, values of Δmh/R and Tm were determined to be 2109.3 and 714.3 K, respectively, by the parameter estimation using the perturbed-chain statistical association theory with initial values from the group contribution method.14,15 To estimate the activity coefficient, the NRTL model was applied. The activity coefficient in the solution is given by:13

k

total counts/s =

Δm h ⎛ 1 1⎞ − ⎟ ⎜ R ⎝ Tm T⎠

ln(x3γ3) =

(2)

where Mi is the midpoint length of a channel. SWMCL was sensitive toward the larger particles and could be used as a reliable average 4928

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design

Article i

mildly agitated at constant temperature for 2 h and then stabilized without agitation. After good sedimentation of glycine crystals, the upper clear solution was decanted from the solution. The sample solution was carefully evaporated in a desiccator without rapid boiling until constant weights were obtained. Glycine solubility was calculated by the weight difference between initial solution and residual mass. Characterization of Glycine Crystals. Product crystals were observed by a scanning electron microscope (SEM, JSM-6010LA, JEOL, Japan) and optical microscope (OPTIPHOT2-POL, Nikon, Japan). Size of at least 500 crystals was measured by image analysis from the optical microphotographs. Average length of diameters measured at 2° intervals with passing through object’s centroid was used to determine characteristic size of crystals. The crystal structure of glycine was investigated using an X-ray diffractometer (MiniFlex, Rigaku, Tokyo, Japan) operated at 30 kV and 15 mA with graphite-monochromatized Cu Kα radiation (λ = 1.5418 Å). Powder X-ray diffraction (PXRD) data were collected using a rotating flat-plate sample holder over the 2θ range from 15° to 35° with a step size of 0.02° and a scanning rate of 1.0 deg/min under ambient conditions. Typical peaks representing α-glycine are 2θ ≈ 19°, 21°, and 24°, and for β-glycine are 2θ ≈ 18°, 24°, and 28°, and γglycine can be easily distinguished by PXRD peaks at 2θ ≈ 22° and 26°.19−21



mdf, i = adc3*(Thot , c 2,′ i)(m1 +

∑ m2, k) − c3*(Tcool , c2,′ i) k=0

i

(m1 +

∑ m2, k)

(7)

k=0

where the heating temperature, Thot, was fixed at 313.15 K; ad is an empirical adjusting parameter to estimate partial dissolution of seed crystals. For the uniform CSD, all fines must be dissolved, that is, mcf,i equals to mdf,i. Therefore, the amount of injected ethanol may be determined to be i−1

c3*(Tcool , c 2,′ i − 1)(m1 +

i

∑ m2, k) = ac3*(Thot , c2,′ i)(m1 + ∑ m2, k) k=0

k=0

(8)

a = ad /ag

(9)

A suitable empirical parameter, a, is determined by the trial and error experiments in which short process time and uniform CSD are preferred. After fines dissolution, the (i + 1)th cooling is carried out to Tcool for the growth of the existing seed crystals. It should be noted that the first cooling is performed at the internal seeding step. Finally, the DFC cycle is repeated until the desired yield is reached. The in-Situ Controlled DFC Process with FBRM and ATRFTIR. Operation of the DFC process with in situ control is simple as shown in Figure 2. Product yield is estimated using the total amount of injected ethanol and glycine concentration measured by ATR-FTIR. If the product yield is lower than the desired value, ethanol is injected at a rapid feeding rate (57.0 mL/min, in the present work) until the generation of fine particles is expected to be ended. Subsequently the solution is heated to dissolve fines. Here, the complete generation and dissolution of fines can be detected by no increase and decrease in

METHODOLOGY

Combined Processes of Conventional Drowning-out and Cooling Crystallization. Two types of conventional processes were employed for glycine crystallization: one (type I) consists of cooling crystallization and subsequent drowning-out and the other (type II) is drowning-out followed by cooling crystallization. A saturated glycine aqueous solution at 313.15 K was prepared and kept at the temperature of 323.15 K with a stabilization time of 30 min. Then the solution was cooled to 313.15 K with a cooling rate of 0.5 K/min, and the combined process was carried out. Here, cooling crystallization and crystallization by drowning-out were performed to 293.15 K at a constant cooling rate of 0.5 K/min and ethanol concentration of 38.1 wt % (on a solute-free basis) at a constant feeding rate of 3.55 mL/ min, respectively. Preparation of Internal Seed for the DFC Process. In the present work, internal seed crystals for the DFC process were produced by cooling crystallization as follows. A saturated glycine aqueous solution at 313.15 K was prepared, and the solution was kept at the temperature of 323.15 K with a stabilization time of 30 min. Then cooling crystallization was conducted to 293.15 K with a constant cooling rate of 0.5 K/min. The Mass Balance Model-Based DFC Process.9 The beginning of ith cycle in the DFC process is ith drowning-out for the generation of fines by one-pot injection of ethanol. The amount of crystallized fines by ith drowning-out, mcf,i is i−1

mcf, i = ag c3*(Tcool , c 2,′ i − 1)(m1 +

∑ m2, k) − c3*(Tcool , c2,′ i) k=0

i

(m1 +

∑ m2,k) k=0

(6)

where c*3 is the equilibrium concentration of glycine in g/g solvent, which is estimated from eqs 3, 4, and 5. The cooling temperature, Tcool, was fixed at 293.15 K. c′2,i is the concentration of ethanol on a solute-free basis after ith drowning-out; m1 is the amount of water, which was 250 g in the present work, and m2,i is the amount of injected ethanol at the ith drowning-out. Here, m2,0 is equal to 0; ag is an empirical adjusting parameter to estimate partial growth of seed crystals. After the ith drowning-out, ith fines dissolution is carried out by heating. The amount of dissolved fines by the ith fines dissolution, mdf,i, is

Figure 2. Flowchart of the in situ controlled DFC process. 4929

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design

Article

total counts per second. After fines dissolution, cooling crystallization is conducted to 293.15 K. Such cyclic operation is repeated until reaching the desired yield.

Table 2. NRTL Parameters for the Water (1)/Ethanol (2)/ Glycine (3) System



RESULTS AND DISCUSSION Calibration of Solution Concentration by ATR-FTIR. In the present work, 28 experiments without any glycine particles for concentration calibration were carried out, as listed in Table 1. Selecting variables for the calibration were spectra in the Table 1. Glycine/Water/Ethanol Solutions Used for the Calibration of Solution Concentration by ATR-FTIR glycine concn (wt %)

water concn (wt %)

ethanol concn (wt %)

temp range (K)a

33.32 29.39 24.85 19.99 14.29 7.72 0.00 23.01 20.04 16.53 13.03 9.07 4.76 0.00 15.97 13.00 9.11 4.78 0.00 9.95 6.56 3.84 0.00 5.23 2.46 0.00 2.44 0.00

66.68 70.61 75.15 80.01 85.71 92.28 100.00 67.73 70.36 73.55 76.52 80.00 83.81 88.00 63.87 66.20 69.18 72.38 76.02 57.64 59.82 61.53 63.95 49.30 50.73 52.01 39.02 40.00

0.00 0.00 0.00 0.00 0.00 0.00 0.00 9.26 9.60 9.92 10.45 10.93 11.43 12.00 20.16 20.80 21.71 22.84 23.98 32.41 33.62 34.63 36.05 45.47 46.81 47.99 58.54 60.00

348.15−337.32 348.15−321.54 348.15−305.34 348.15−284.04 348.15−278.15 348.15−278.15 348.15−278.15 348.15−327.30 348.15−315.54 348.15−299.18 348.15−282.62 348.15−278.15 348.15−278.15 348.15−278.15 348.15−327.10 348.15−312.71 348.15−295.09 348.15−278.15 348.15−278.15 348.15−327.46 348.15−307.89 348.15−278.15 348.15−278.15 348.15−328.03 348.15−302.25 348.15−278.15 348.15−278.15 348.15−278.15

parameters

values or functions

α12 α13 α23 τ12 τ21 τ13 τ31 τ23 τ32

0.1803 0.05 0.02 (814.903 + 2.0078T)/(RT) (−697.613 + 0.9765T)/(RT) (−8553.003 + 15.566T)/(RT) (−1495.907 + 22.915T)/(RT) (−2399.639 − 23.805T)/(RT) (8336.032 + 50.612T)/(RT)

Figure 3. Measured (solid circles) and calculated solubility (surface) of glycine in cosolvent of water and ethanol using ATR-FTIR.

FTIR in the present work were confirmed to be well-matched with data measured by the gravimetric method. It can be seen that the solubility values of the present work are in good agreement with data reported previously.17,22−24 Solubility data are summarized in Tables S1 and S2 (Supporting Information). Combined Processes of Conventional Drowning-out and Cooling Crystallization. The temperature and solution concentration profiles of type I and II processes are shown in Figure 5. The final state of solutions obtained by those combined processes was the same as that with ethanol concentration of 38.1 wt %. However, crystal size was found to be dependent on the process, because generation of nuclei is influenced by the operation route. Figure 6 shows the SEM images of products obtained by those processes. Average crystal sizes obtained by type I and II processes are 81.3 and 107.2 μm, respectively. In general, during a drowning-out operation, very high supersaturation is immediately generated in the mixing region of solution and antisolvent. Therefore, undesired nucleation easily occurs, and CSD of glycine crystals obtained by the type I and II processes was clearly bimodal as shown in Figure 6 (see also Figure S3, Supporting Information). A Mass Balance Model-Based DFC Process. Using the MATLAB f minsearch solver, eq 8 was solved with a values of 0.9, 1.0, and 1.1. In this calculation, the desired yield was chosen to 85 wt %. It is obvious that the amount of injection water for each step of the drowning-out process decreases with decreasing a, as listed in Table 3. Therefore, as a becomes smaller, more DFC runs are required to reach the desired yield.

a

A cooling rate of 0.5 K/min was applied, and spectra were recorded every 5 min.

1800−800 cm−1 region with a resolution of 4 cm−1, resulting in 251 variables (See Figures S1 and S2, Supporting Information, for infrared spectra of the water, ethanol, and water/ethanol/ glycine solutions). However, only 19 latent variables were used after the partial least-squares regression and leave-one-out cross-validation.12 Root-mean-square error of cross-validation and correlation coefficient (R2) for glycine, water, and ethanol were 0.188, 0.145, and 0.137 wt % and 0.9997, 0.9999, and 0.9999, respectively. All computations for the calibration and estimation of solution concentration were performed using an iC-IR software with Quant-IR module (Mettler-Toledo, USA). Solubility of Glycine in Cosolvent of Water and Ethanol. NRTL parameters are summarized in Table 2. As shown in Figure 3, the NRTL model was found to provide a good description of the solubility of α-glycine measured by ATR-FTIR in the cosolvent of water and ethanol (R2 = 0.9996). Figure 4 shows that calculated solubilities by ATR4930

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design

Article

Figure 6. SEM images of glycine obtained by the (a) type I and (b) type II process.

Table 3. Amount of Injected Ethanol of ith Drowning-out with the Parameter, a amount of injected ethanol, m2,i (g) parameter, a

i=1

i=2

i=3

i=4

0.9 1.0 1.1

24.0 33.4 42.3

31.9 46.9 63.6

44.6 73.8 119.0

68.3

were 79.6, 330.0, and 326.7 μm, respectively. If the significantly large a values are employed, fine particles are not dissolved completely during the fines dissolution steps. Therefore, many fine particles remained and the seeds formed initially cannot grow enough, as can be seen from Figure 7a. On the other hand, when fine particles are dissolved enough, employing a smaller than 1.0, relatively large crystals and uniform CSD are obtained as shown in Figure 7b,c. See also Figure S4 in Supporting Information. Therefore, considering crystal size and total process time, suitable a was chosen to be 1.0. A profile of temperature and solution concentration in the model-based DFC process with a = 1.0 is shown in Figure 8 and the operating profiles versus time are shown in Figure 9. At seed preparation stage in Figure 9, total counts per second is shown to be nearly zero until primary nucleation is induced at the elapsed time of about 0.6 h. After the end of the primary nucleation, total counts per second and SWMCL were about 1500 and 200 μm, respectively. During the crystal growth stage (0.75−0.95 h), however, noticeable change in SWMCL was not observed although glycine concentration was continuously decreased. During the drowning-out process of each DFC run at the operating times of 0.95, 1.20, and 1.53 h, rapid increase in total counts per second and ethanol concentration were observed. On the other hand, SWMCL and glycine concentrations decrease because of intensive nucleation of glycine by injection of ethanol. With subsequent heating processes, total counts per second was shown to return to the initial value of about 1500. This implies that dissolution of fines is satisfactorily accomplished. However, one can notice that excessive heating was conducted after total counts/s returned to the initial value in the second and third DFC runs (1.30−1.36 h and 1.60−1.70 h), as shown in Figure 9. This indicates that a = 1 is suitable for the first DFC run only, but a values should be larger than 1 for the second and third DFC runs to remove excessive heating. If a values for second and third runs were optimized, the process time would be reduced. However, further trial and error experiments are required. The in-Situ Controlled DFC Process. Figure 10 shows the profiles of temperature and solution concentration monitored from the in situ controlled DFC process, and Figure 11 shows the operating profiles versus time of the process. From Figure 11, it can be seen that total counts per second and SWMCL of

Figure 4. The comparison of solubility of glycine with the reported data:17,22−24 (a) without addition of ethanol and (b) at a temperature of 298.15 K.

Figure 5. Temperature and solution concentration profiles of the two combined processes of conventional cooling crystallization and drowning-out. Surface plot indicates the calculated solubility.

If the product quality were the same, a larger a is desirable for the reduction of total process time. Figure 7 shows the SEM images of glycine obtained by the DFC process with various a values. The average sizes of crystal obtained by the mass balance model-based DFC process with a = 1.1, 1.0, and 0.9 4931

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design

Article

Figure 7. SEM images of glycine obtained by the mass balance model based DFC process with a of 1.1 (a), 1.0 (b), and 0.9 (c).

Figure 10. A profile of temperature and solution concentration in the in situ controlled DFC process. Surface plot indicates calculated solubility.

Figure 8. A profile of temperature and solution concentration in the model-based DFC process with a = 1. Surface plot indicates the calculated solubility.

original seeds are about 1500 and 200 μm, respectively. Those results are in good agreement with seed crystals used in the model-based DFC process, as shown in Figure 9. During the drowning-out stage of the DFC runs at 0.95 and 1.38 h (dashed lines in Figure 11), ethanol was injected with

relatively high feeding rate of 57.0 mL/min until the total counts per second did not increase. SWMCL and glycine concentration are shown to decrease due to the intensive nucleation. After the subsequent heating process, total counts per second is shown to decrease and return to the initial value

Figure 9. Operating profiles of temperature, solution concentration, and FBRM measurements of the model-based DFC process with a = 1. Gray regions indicate the unnecessary heating periods. 4932

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design

Article

Figure 11. Operating profiles of temperature, solution concentration, and FBRM measurements of the in situ controlled DFC process.

of 1500 counts/s by dissolution of fines. The final heating temperature is varied for each DFC run, which is different from the model-based DFC process. This indicates that unnecessary heating was avoided after complete dissolution of fine particles, and thus total process time was efficiently reduced: total process times for the model-based and in situ controlled DFC processes were about 55 and 40 min, respectively. Figure 12 shows the SEM image of product obtained by the in situ controlled DFC process. Average crystal size obtained

dissolution stage is carried out until the total counts per second is returned to the target value by temperature cycling with monitoring with FBRM.25−29 However, in the in situ controlled DFC process proposed in the present work, fine crystals are intentionally generated by drowning-out and then those are dissolved to improve the product yield as well as CSD. Compared with the reported DNC processes, advantages of the present process are the relatively high product yield and short operation time with uniform CSD. Polymorphism of Glycine Obtained by the DFC Process. Glycine has at least three polymorphs (α-, β-, and γ-forms).30−33 Although γ-glycine is known to be the thermodynamically stable form among three forms under ambient conditions, α-glycine is typically obtained by crystallization from aqueous solutions.30,31 All glycine crystals obtained in the present work also were α-form as shown in Figure 13. However, it has been reported that a large amount of ethanol, commonly more than 50 wt %, induces nucleation of β-glycine or retards solvent-mediated transformation of β- to αglycine.21,34 In the present work, maximum content of ethanol was 38.1 wt %.

Figure 12. SEM image of glycine obtained by the in situ controlled DFC process.

was 328.0 μm, which is similar to that produced by the modelbased DFC process. In the DFC process, crystal size of product was dependent on the population of initial seeds and total amount of injected antisolvent. However, it is interesting to note that crystal size is independent of operating route of the process including heating temperature, total process time, and the number of DFC runs, because of completey dissolution of undesired fine particles (See Figure S4 in Supporting Information). Direct nucleation control (DNC) proposed by Nagy et al.10,11 was a key inspiration for the development of the present process. The DNC is a repeating strategy consisting of a crystallization stage with cooling or antisolvent addition or both and a dissolution stage with heating or solvent addition or both. When the total counts per second of the system increases during the crystallization stage due to nucleation, a subsequent

Figure 13. PXRD patterns of glycine produced by the (a) type I, (b) type II, (c) mass balance model-based DFC process with a = 1.0, and (d) in situ controlled DFC process. 4933

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934

Crystal Growth & Design



Article

(13) Prausnitz, J.; Lichtenthaler, R.; de Azevedo, E. Molecular Thermodynamics of Fluid-Phase Equilibria; Prentice Hall PTR:Upper Saddle River, NJ, 1999. (14) Held, C.; Cameretti, L. F.; Sadowski, G. Ind. Eng. Chem. Res. 2011, 50, 131−141. (15) Marrero, J.; Gani, R. Fluid Phase Equilib. 2001, 183, 183−208. (16) Ghmeling, J.; Onken, U.; Arlt, W. Vapor−Liquid Equilibrium Data Collection. Aqueous Organic Systems; Chemistry Data Series; Dechema: Frankfurt, Germany, 1981; Vol. 1, Part 1a. (17) Ferreira, L. A.; Macedo, E. A.; Pinho, S. P. Chem. Eng. Sci. 2004, 59, 3117−3124. (18) Nelder, J. A.; Mead, R. Comput. J. 1965, 7, 308−313. (19) Zaccaro, J.; Matic, J.; Myerson, A. S.; Garetz, B. A. Cryst. Growth Des. 2001, 1, 5−8. (20) Allen, K.; Davey, R. J.; Ferrari, E.; Towler, C.; Tiddy, G. J.; Jones, M. O.; Pritchard, R. G. Cryst. Growth Des. 2002, 2, 523−527. (21) Dang, L.; Yang, H.; Black, S.; Wei, H. Org. Process Res. Dev. 2009, 13, 1301−1306. (22) Ramasami, P. J. Chem. Eng. Data 2002, 47, 1164−1166. (23) Yang, X.; Wang, X.; Ching, C. B. J. Chem. Eng. Data 2008, 53, 1133−1137. (24) Nozaki, Y.; Tranford, C. J. Biol. Chem. 1971, 246, 1−7. (25) Doki, N.; Seki, H.; Takano, K.; Asatani, H.; Yokota, M.; Kubota, N. Cryst. Growth Des. 2004, 4, 949−953. (26) Abu Bakar, M. R.; Nagy, Z. K.; Rielly, C. D. Org. Process Res. Dev. 2009, 13, 1343−1356. (27) Woo, X. Y.; Nagy, Z. K.; Tan, R. B. H.; Braatz, R. D. Cryst. Growth Des. 2009, 9, 182−191. (28) Kim, J.-W.; Kim, J.-K.; Kim, H.-S.; Koo, K.-K. Org. Process Res. Dev. 2011, 15, 602−609. (29) Saleemi, A. N.; Steele, G.; Pedge, N. I.; Freeman, A.; Nagy, Z. K. Int. J. Pharm. 2012, 430, 56−64. (30) Han, G.; Chow, P. S.; Tan, R. B. H. Cryst. Growth Des. 2012, 12, 2213−2220. (31) Han, G.; Poornachary, S. K.; Chow, P. S.; Tan, R. B. H. Cryst. Growth Des. 2010, 10, 4883−4889. (32) Dowling, R.; Davey, R. J.; Curtis, R. A.; Han, G.; Poornachary, S. K.; Chow, P. S.; Tan, R. B. H. Chem. Commun. 2010, 46, 5924−5926. (33) Towler, C. S.; Davey, R. J.; Lancaster, R. W.; Price, C. J. J. Am. Chem. Soc. 2004, 126, 13347−13353. (34) Weissbuch, I.; Torbeev, V. Y.; Leiserowitz, L.; Lahav, M. Angew. Chem., Int. Ed. 2005, 44, 3226−3229.

CONCLUSIONS Two different DFC processes, the mass balance model-based process reported previously and the in situ controlled process by ATR-FTIR and FBRM proposed in the present work, were applied for the crystallization of glycine from cosolvent of water and ethanol. In the model-based DFC process, the amount of injected antisolvent at every run of the DFC process is estimated, and the fines induced by drowning-out are redissolved by heating and temperature cycling. On the other hand, the in situ controlled DFC process is a model-free approach. In this process, by monitoring with FBRM, drowning-out and heating are conducted until the generation and dissolution of fine particles, respectively, are assumed to be completed. Herein, the DFC process is repeated until the desired value of product yield is obtained by monitoring with ATR-FTIR. Glycine crystals produced by the proposed process in the present work were found to have similar crystal size to that obtained by the mass balance model-based DFC process. However, advantages of the in situ controlled process are that no predetermined data are required to run the process and the operation time is considerably reduced by eliminating the trivial heating period.



ASSOCIATED CONTENT

S Supporting Information *

Infrared spectra of water, ethanol, and glycine solutions, CSDs of glycine, and measured glycine solubilities. This material is available free of charge via the Internet at http://pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail [email protected], telephone +82-2-705-8680, fax +822-711-0439. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by the Sogang University Foundation Research Grant in 2011.



REFERENCES

(1) Mullin, J. W. Crystallization, 4th ed.; Butterworth Heinemann: Oxford, U.K., 2001. (2) Kim, J.-W.; Kim, J.-K.; Kim., E. J.; Kim., H.-S.; Koo, K.-K. Korean J. Chem. Eng. 2010, 27, 666−671. (3) Damour, C.; Benne, M.; Boillereaux, L.; Grondin-Perez, B.; Chabriat, J.-P. J. Ind. Eng. Chem. 2010, 16, 708−716. (4) Kim, J.-H.; Song, S. M.; Kim, J. M.; Kim, W. S.; Kim, I. H. Korean J. Chem. Eng. 2010, 27, 1532−1537. (5) Saengchan, A.; Kittisupakorn, P.; Paengjuntuek, W.; Arpornwichanop, A. J. Ind. Eng. Chem. 2011, 17, 430−438. (6) Berry, D. A.; Dye, S. R.; Ng, K. M. AIChE J. 1997, 43, 91−103. (7) Beckmann, W. J. Cryst. Growth 1999, 198−199, 1307−1314. (8) Kim, Y.; Haam, S.; Shul, Y. G.; Kim, W.-S.; Jung, J. K.; Eum, H.C.; Koo, K.-K. Ind. Eng. Chem. Res. 2003, 42, 883−889. (9) Kim, J.-W.; Park, D. B.; Shim, H.-M.; Kim, H.-S.; Koo, K.-K. Ind. Eng. Chem. Res. 2012, 51, 3758−3765. (10) Abu Bakar, M. R.; Nagy, Z. K.; Saleemi, A. N.; Rielly, C. D. Cryst. Growth Des. 2009, 9, 1378−1384. (11) Saleemi, A. N.; Rielly, C. D.; Nagy, Z. K. Cryst. Growth Des. 2012, 12, 1792−1807. (12) Cornel, J.; Lindenberg, C.; Mazzotti, M. Ind. Eng. Chem. Res. 2008, 47, 4870−4882. 4934

dx.doi.org/10.1021/cg3008574 | Cryst. Growth Des. 2012, 12, 4927−4934