Gassing Crystallization at Different Scales: Potential to Control

Jan 30, 2017 - In this work the linear cooling profiles of experiments with Tsat = 39.6 °C ( 160 gSA/kgwater) and those with Tsat = 30 °C ( 106 gSA/...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/crystal

Gassing Crystallization at Different Scales: Potential to Control Nucleation and Product Properties Tobias Kleetz, Gordon Paẗ zold, Gerhard Schembecker, and Kerstin Wohlgemuth* TU Dortmund University, Laboratory of Plant and Process Design, Emil-Figge-Straße 70, 44227 Dortmund, Germany S Supporting Information *

ABSTRACT: Gassing crystallization is an induced nucleation process during batch cooling crystallization with the aim to control the nucleation step and thus product crystal properties. All previous studies have been made at lab scale and show that the metastable zone or the supersaturation at which gassing is started is crucial for the success of gassing crystallization. Since the metastable zone width depends on many factors, the purpose of this paper was to verify the hypothesis that, especially for parameter combinations which result in broad metastable zone widths, the success of gassing crystallization is independent of crystallizer scale and geometry. The studies were made for the substance system succinic acid/water in a 1 L lab and a 30 L pilot scale crystallizer. The effect of gassing on the metastable zone width and the median diameter was evaluated for varying process parameters (saturation concentration, gassing supersaturation, cooling rate, and stirrer speed) and compared to normal cooling crystallization. After the application of gassing, metastable zone widths were narrower, median diameters were bigger, and reproducibility was enhanced. We found that for process parameters which resulted in broad metastable zone widths the effect of gassing on the median diameter was largest, independent of crystallizer scale and geometry. Gassing crystallization induces nucleation and affects product crystal properties, which works best for process conditions resulting in broad metastable zone widths.

1. INTRODUCTION The efficiency of process steps after crystallization depends in particular on the properties of the crystal product.1−3 An in situ way to affect crystal product properties is to control either nucleation or crystal growth or both.3,4 This work focuses on the control of the nucleation step. Here, gassing crystallization is used as an alternative to common nucleation control strategies like seeding or the application of ultrasound. Gassing in the metastable zone during batch cooling crystallization can be used to induce nucleation.4−6 It has been developed from sonocrystallization as cavitation bubbles from ultrasound were replaced by gas bubbles of synthetic air. Wohlgemuth et al. found that gassing has a similar effect on the metastable zone width (MZW) and the crystal size distribution (CSD). They concluded that for both gassing and sonocrystallization the nucleation mechanism must be a heterogeneous one where the surface of the gas or cavitation bubble reduces the Gibb’s free energy and acts as a nucleation center.4−6 The gassing process is affected by three gassing parameters: gas volume flow V̇ gassing, gassing duration tgassing, and the supersaturation at which gassing is started Δcgassing. Previous work focused on the investigation of gassing crystallization at lab scale.5−7 Using the model system succinic acid−water, it has been shown that for linear cooling profiles it was possible to control the median diameter of product crystals in a range of 100 μm by the selection of Δcgassing.7 Independent of the cooling profile applied, the median diameter of product crystals © 2017 American Chemical Society

has been enhanced always by the application of gassing compared to normal cooling crystallization without gassing. The effect of gassing can be traced back to the amount of nuclei induced. Compared to normal crystallization with spontaneous nucleation, gassing induces a lower amount of nuclei at lower supersaturation. These nuclei degrade supersaturation by growth, which results in product crystals with a bigger median diameter.7 Similar effects were reported regarding induction time measurements investigating the influence of gassing parameters. The induction time at constant temperature describes the duration of the formation of nuclei and their growth to a detectable size and can be used as a measure of the amount of nuclei induced.8−10 By the application of gassing, induction time was decreased remarkably compared to normal cooling crystallization. Again, the selection of the gassing supersaturation allowed an influence on the induction time, which was shorter for gassing at high supersaturation where a high driving force for nucleation led to a comparatively high amount of nuclei induced faster degradation of supersaturation.11 A comparison of the CSDs from linear cooling after induction time measurement and with a constant linear cooling profile showed that only during constant cooling was an effect Received: August 29, 2016 Revised: January 29, 2017 Published: January 30, 2017 1028

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

⎡ g ⎤ csat⎢ SA ⎥ = 29.615 × exp(0.0426 × T[°C]) ⎢⎣ kg water ⎥⎦

of gassing on the CSD generated. If temperature continued to decrease right after gassing, supersaturation was degraded directly. By this, nucleation showers were avoided or supersaturation at which nucleation showers occurred was decreased. Consequently, the amount of nuclei induced was less for these experiments, leading to product crystals with bigger median diameter. This nucleation behavior was observed during analysis of supersaturation profiles of normal cooling crystallizations as well as gassing crystallizations.11 From these findings it can be derived that gassing crystallization has an effect on product properties if supersaturation at which nucleation occurs is low compared to the experiment without gassing. A measure for the supersaturation at which nucleation occurs is the MZW, which can be converted into a difference between saturation temperature and nucleation temperature via the solubility line.10 Besides induced nucleation processes, there are many other process parameters which affect the MZW during cooling crystallization.12−14 The most important ones are the stirrer speed, the concentration of the solute, and the cooling rate, and also the scale, the geometry, and the surface material of the crystallizer.12−14 All previous studies show that the metastable zone or the supersaturation at which gassing is started, respectively, is crucial for the success of gassing crystallization.6 Since the MZW depends on many factors, the purpose of this paper is to validate that gassing crystallization is not a lab scale phenomenon only but is successful especially for parameters which result in a broad MZW. Therefore, the success of gassing is independent of the crystallizer scale or geometry unless a sufficiently broad MZW is available. The effect of gassing crystallization in comparison to normal cooling crystallization is therefore evaluated with respect to enlarging the median diameter of product crystals, which is often the objective of crystallization optimization processes.15−18 Other product property parameters like width of CSD or degree of agglomeration had already been discussed in previous work.7,11 Quantitative relationships between the effect of gassing at lab and pilot scale crystallizers are not the focus of this paper. This paper is organized as follows: First, the effect of process parameters on MZW is investigated on the lab scale. Gassing crystallizations are compared to normal cooling crystallizations to investigate the relationship between MZW and the enlarging effect on product crystals in the case of gassing crystallization. Second, studies in a 30 L pilot plant are performed. The effect on the MZW for different saturation concentrations is determined for normal cooling crystallization processes first. Subsequently, the effect of gassing crystallization is validated for the parameter combinations which result in a broad and narrow MZW during normal cooling crystallization.

(1)

2.2. Experimental Setup. Lab Scale. The setup for lab scale experiments was identical to the one used in previous publications.7 As crystallizer a 1 L glass reactor system (LabMax, Mettler Toledo) with a tempering jacket and silicone oil as tempering medium was used. The stirrer was a 45° pitched blade stirrer of stainless steel with a diameter of 5 cm. The crystallizer was equipped with a temperature probe, an ATR-FTIR probe, an FBRM probe, and a gassing ring. Mettler Toledo software (Icontrol 5.0, IC IR 4.2) was used to control stirrer speed, temperature of the solution, and to monitor succinic acid concentration online. Therefore, a model for the prediction of succinic acid concentration from IR spectra using partial least-squares regression was applied, using ICQuant from Mettler Toledo. The FBRM signal was not evaluated further in this work, although the probe was present in the crystallizer during the experiments. The gassing ring was made of stainless steel and had an inner diameter of 50 mm and 24 holes with a diameter of 0.5 mm drilled into the upper side. The gassing ring was placed above the stirrer and had a distance of 100 mm to the liquid surface. Thus, deformation of gas bubbles was inhibited and reproducibility enhanced. The synthetic air gas flow was controlled with a needle valve. Additionally, synthetic air was heated and saturated in a water bath with the same temperature as the solution before entering the crystallizer during gassing. This should prevent evaporation of the water into the gas. Pilot Scale. As crystallizer a glass pilot plant reactor with a filling volume of 30 L was used (Gr30, Büchi Glas Uster). A schematic view of the crystallizer is shown in Figure 1. The impeller stirrer had three

Figure 1. Left: Schematic view of the pilot plant crystallizer (30 L) with stirrer, baffle, and gassing unit. Right: top view of the three-blade impeller stirrer. curved blades and was made of stainless steel. The temperature probe placed in the crystallizer worked as a baffle also. For temperature control, a thermostat (Unistat 510w, Huber) was used; as tempering medium silicon oil was used in the double jacket. The crystallizer was equipped with a self-constructed gassing unit which is shown in Figure 2. The gassing unit consisted of three Teflon plates integrated into frames of stainless steel and was designed with the purpose to create a bubble surface area as large as possible. Therefore, each plate had 200 holes with a diameter of 0.5 mm drilled into the upper side, similar to the gassing ring from the lab scale. All three plates were placed on a framework which was placed directly above the stirrer and had a distance of 200 mm to the liquid surface.

2. MATERIALS AND METHODS 2.1. Investigated System. Succinic acid (purchased from Wittich Umweltchemie GmbH) was used as solute with a purity higher than 99.5%. Water (Ultrapure, 0.05 μS/cm, Millipore) was used as solvent. The solubility of succinic acid in water has been measured in our laboratory for temperatures between 0 and 40 °C at atmospheric pressure. Equation 1 shows a correlation for the solubility of succinic acid in water fitted to our data.7 Synthetic air (Air Liquide, >99.99%) stored in a gas bottle was used for gassing.; 1029

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

profile to Tfinal was started. In this work cooling profiles with cooling rates κ of 0.1, 0.25, and 0.4 K/min were applied. In the case of gassing crystallization experiments, gassing was started at Δcgassing = 5.3 gSA/ kgwater or Δcgassing = 9.5 gSA/kgwater for lab scale experiments and Δcgassing = 5.3 gSA/kgwater for the pilot plant. Gassing should be executed only in a crystal clear solution, which is guaranteed at the supersaturations chosen. The gassing duration for both scales was tgassing = 20 s and the gas volume flow V̇ gassing = 200 L/h for lab scale and V̇ gassing = 3400 L/h for pilot scale, respectively. During gassing, the stirrer was turned off which resulted in less turbulence and allowed the bubbles to rise straight to the top without deformation. When Tfinal was reached, the crystals were first harvested from the crystallizer and then separated from the mother liquor. For the lab scale experiments crystals were filtered with a funnel filter, filter paper (pore size 2 μm), and a vacuum pump (Mini diaphragm vacuum pump VP 86, VWR). For solid liquid separation in the pilot scale, a nutschtype filter (30 L, Büchi Glas Uster) with a diameter of 300 mm with a PTFE membrane (pore size 50 μm, Büchi Glas Uster) was used. The suspension was transported via vacuum with a vacuum pump (MD12, Vaccubrand). After solid liquid separation, the wet product crystals from pilot scale experiments were divided into 12 parts of equal size manually before drying. One 12s of the product crystals from pilot scale and all product crystals from lab scale were dried in a fluidized bed dryer (TG200, Retsch) with a volume flow of 45 L/h at 60 °C for 1 min. Therefore, the fluidized bed dryer was equipped with a 6 L vessel for the pilot scale product crystals and with three 0.3 L vessels for the lab scale product crystals. Drying procedure was repeated at least five times until constant weight of the sample was reached. The dry product crystals of each experiment were divided into eight samples of equal mass using an automated sample divider (Rotary sample divider laborette 27, Fritsch). This procedure was repeated until a sample of approximately 5 g was left, which was then analyzed. The dry product crystals were analyzed with respect to the volumetric crystal size distribution (CSD) and its characteristic values. The median crystal diameter is represented by d50 and the width of the crystal size distribution by the difference between d90 and d10. Therefore, laser diffraction (LS 13 320 laser diffraction particle size analyzer, Beckman Coulter) with a Tornado Dry Powder System was used. In order to evaluate measuring errors, all experiments have been performed twice unless otherwise mentioned. 2.4. Metastable Zone Width Measurement. The metastable zone width for lab scale experiments was measured recording the drop of the concentration of succinic acid measured with the ATR-FTIR probe. In this context, MZW is defined as the difference of the saturation temperature and the temperature at which concentration is decreasing. Therefore, concentration was plotted over time and the

Figure 2. Self-constructed gassing unit consisting of three Teflon plates integrated into frames. Compressed and filtered air was used for gassing. The air supply was connected to each plate individually and the volume flow could be controlled with a needle valve. Different from the lab scale experiments, the air was not heated and saturated in a water bath. Experiments showed that there was no considerable effect of gassing on the temperature of the solution and gassing experiments with and without saturated air did not show any difference.6 2.3. Execution of Crystallization Experiments. The experimental procedures for experiments in lab and in pilot scale were similar. Figure 3 shows exemplary temperature profiles for experiments with a saturation temperature of Tsat = 39.6 °C and a final temperature of Tfinal = 17 °C. In order to provide identical conditions, every experiment started with a preparation phase (t < 0 min) in which succinic acid crystals were dissolved at 10 K above Tsat for 1 h. The preparation phase provided identical conditions for every experiment. In this work the linear cooling profiles of experiments with Tsat = 39.6 °C (≜ 160 gSA/kgwater) and those with Tsat = 30 °C (≜ 106 gSA/kgwater) ended at Tfinal = 17 °C. The minimum stirrer speed which was necessary to suspend all crystals at Tfinal was chosen according to the 1 s criterion, which ensures that no crystal lasts longer than 1 s at the bottom of the vessel. The residence time of the crystals at the vessel bottom was evaluated qualitatively. The stirrer speed had to be at least Rmin = 300 1/min for the lab crystallizer and Rmin = 60 1/min for the pilot plant crystallizer. After the preparation phase a linear cooling

Figure 3. Exemplary experimental procedure for crystallization experiments with preparation phase (black line) and linear cooling ramps with different cooling rates (gray lines). 1030

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

Figure 4. Exemplary determination of MZW with GLV and concentration measurement during a linear cooling profile with Tsat = 30 °C in 1 L lab scale experiment. The start of the measurement (t = 0 min) refers to Tsat. ⎛ dc ⎞ log(κ ) = n log(MZW) + log(k b) + (n − 1)log⎜ sat ⎟ ⎝ dT ⎠

intersection of two balance lines was calculated. One balance line represented the constant concentration part and a second one represented the part of decreasing concentration after nucleation. The time of the intersection correlated with the temperature of the solution, which enabled the direct determination of MZW. Since an ATR-FTIR probe was not available for pilot scale experiments, MZW was measured with an offline video analysis. Therefore, a camera (Yi Sport Cam, Xiaomi) in a waterproofed case was placed directly below the liquid surface in the crystallizer during the experiment. After the experiment, images taken were analyzed regarding the gray level value (GLV) of the solution. The GLV correlated with the turbidity of the solution and therefore could be used as a measure for the MZW. Similar to the determination from concentration data, GLV was plotted over time and two balance lines were drawn: one for the part of constant GLV and another one for increasing GLV after nucleation. From the time corresponding to the intersection, the corresponding temperature from temperature−timeprofile and thus the MZW was calculated. 2.5. Calculation of Nucleation Kinetics. Nucleation kinetic parameters from MZW data were determined following the classical nucleation approach by Nyvlt.19 The primary nucleation rate J can be expressed by eq 2 at which kb represents the mass nucleation rate constant, n the apparent order of nucleation, and Δc the absolute supersaturation. J = k b × Δc n

In order to calculate the nucleation kinetic parameters kb and n it was necessary to determine the MZW represented by ΔTmax for different cooling rates κ, while other process parameters maintained constant.

3. RESULTS AND DISCUSSION 3.1. Validation of GLV-Method. Metastable zone width was determined with respect to the concentration drop using ATR-FTIR spectroscopy in 1 L scale. For the 30 L pilot plant, ATR-FTIR spectroscopy was not available so the MZW was determined with respect to GLV data which corresponds to the turbidity of the solution. Figure 4 displays the concentration and GLV profile of a normal cooling crystallization with spontaneous nucleation on 1 L lab scale. After a time of constant concentration, the concentration dropped indicating prior nucleation. Similarly, the GLV value was constant at the beginning, and then increased due to crystals present clouding the solution and thus also indicating nucleation. The resulting times at which nucleation occurred, tnuc, the corresponding nucleation temperatures, Tnuc, and the MZWs are summarized in Table 1. The results were nearly identical and proved that GLV measurement is a suitable method to determine the MZW.

(2)

In the case of cooling crystallizations, the rate of supersaturation generation corresponds to the cooling rate κ via eq 3 at which dcsat is dT the slope of the solubility curve at a given csat. The relationship between maximum supercooling (MZW) and maximum supersaturation (Δcnuc) is given by eq 4 dc dΔc = κ sat dt dT

(3)

⎛ dc ⎞ Δcnuc = ⎜ sat ⎟MZW ⎝ dT ⎠

(4)

Table 1. Nucleation Time tnuc, Nucleation Temperature Tnuc, and MZW Gained from Concentration Profile and GLV Profile tnuc [min] Tnuc [°C] MZW [K]

Assuming that the rate of supersaturation generation is proportional to the nucleation rate J at the beginning of nucleation, eq 5 can be derived from the combination of eqs 2, 3, and 48,19−21

⎛⎛ dc ⎞ ⎞n dc k b × Δc n = k b⎜⎜ sat ⎟MZW⎟ = κ sat ⎝ ⎠ ⎝ dT ⎠ dT

(6)

concentration profile

GLV profile

16.29 23.78 6.22

16.17 23.82 6.18

After this validation, the GLV measurement was applied to the 30 L plant, where the resulting GLV profiles appeared identical and were suitable to determine the MZW. 3.2. Lab Scale Experiments. The effect of process parameters and gassing crystallization on MZW and d50 was investigated varying one process parameter (saturation concentration, gassing supersaturation, cooling rate, stirrer

(5)

Subsequently, the parameters kb and n can be derived from a log plot of κ versus MZW (eq 6) 1031

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

Figure 5. MZW (left) and d50-values (right) for experiments with and without gassing: (A) varying saturation concentration, (B) varying gassing supersaturation, (C) varying cooling rate, and (D) varying stirrer speed. Constant process parameters are given in the top of each part. Width of CSD is provided as Supporting Information (Table S1).

controlled induction of nuclei led to a better reproducibility in almost all cases. In order to compare the effect of gassing on d50 in a fair way we introduce a new measure, Δd50. Therefore, the differences of the d50-values from experiments with and without gassing have been divided by the theoretical yield, given as the concentration difference (eq 7). Low values of Δd50 mean that gassing has a small effect on the d50, whereas higher values characterize a big effect. The resulting values for Δd50 are given in Figure 5 (righthand side) also.

speed) while the others remained constant. The resulting MZWs and d50-values are displayed in Figure 5. Figure 5 is divided into four parts: (A) shows the results for varying saturation concentration, (B) for varying gassing supersaturation, (C) for varying cooling rate, and (D) for varying stirrer speed. For comparison and better comprehension in addition to each gassing experiment (hatched bars) the results of the experiments without gassing are shown (solid bars). For all investigations, the application of gassing induced stable nuclei at much lower supersaturation, which could degrade supersaturation at an earlier point of the cooling process resulting in narrower MZW and bigger d50-values compared to normal cooling crystallization. Moreover, the

Δd50 = 1032

d50,with gassing − d50,without gassing csat − c Tfinal

(7) DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

Figure 6. Log plot of MZW data versus cooling rate κ for experiments with and without gassing. Constant process parameters: c* = 106 gSA/kgwater, R = 300 1/min. Gassing parameters: Δcgassing = 9.6 gSA/kgwater, tgassing = 20 s, V̇ gassing = 200 L/h.

chose a low csat again, because the effect of gassing should be higher for the broader MZW. The MZWs were broader for experiments with higher cooling rates. The broader MZW for higher cooling rates resulted from the kinetic inhibition of the formation of a solid phase from a solution. Median diameters were smaller for higher cooling rates because MZWs were broader for those experiments. This correlated with a higher supersaturation at nucleation, which resulted in more nuclei which grew in sum to product crystals of smaller size. The relationship between MZWs and d50 shown for experiments with and without gassing was confirmed for varying cooling rate also. Figure 5D shows the results of MZW and d50 for experiments with and without gassing and varying stirrer speed. The MZWs for stirrer speed R = 500 1/min were lower. That was expected because higher stirrer speed caused better mixing enhancing the probability of succinic acid molecule cluster collisions and the formation of stable nuclei leading to a narrower MZW. The smaller d50 for experiments with broader MZWs were in accordance with results shown before. For this parameter setup Δd50 was bigger for the experiments with lower stirrer speed showing a broader MZW of the experiments with and without gassing also. This supports our hypothesis once more. If we compare normal and gassing crystallization, the effect on Δd50 was the biggest for the experiment with the highest cooling rate (Figure 5 C, κ = 0.4 K/min). Here, the MZW for the normal cooling crystallization was the broadest and consequently had the greatest possibility to affect d50. When the MZW of the normal cooling crystallization was narrow, Δd50 was low, meaning that gassing had a small impact on d50 only. Nevertheless, gassing as an induced nucleation process during cooling crystallization is a promising technology to affect the MZW, and with this, d50. Especially for process conditions resulting in broad MZW, where crystallizations are often difficult to handle and the potential of gassing crystallization is greatest. Control of Nucleation Kinetic Parameters by Gassing. In order to show the feasibility of gassing to control nucleation for cooling crystallization quantitatively, nucleation kinetic parameters have been calculated. As the data basis, the MZWs for varying cooling rates κ for experiments with and without gassing (Figure 5C) were used. Figure 6 shows a log plot of the cooling rate κ versus the MZW. The result of the experiments with and without gassing

Figure 5A shows the results of MZW and d50 for experiments with and without gassing and varying saturation concentration. For lower csat MZWs were broader for experiments with and without gassing because fewer succinic acid molecules were present in the solution. Since nucleation is a statistical process, solutions with low csat needed a higher driving force to form stable nuclei. Moreover, the theoretical yield was lower because the final concentration was identical for every experiment. Accordingly, the d50 of experiments at low csat were smaller compared to those with higher csat. However, by the application of gassing it was possible to enhance d50 remarkably independent of saturation concentration. It was possible to gain similar d50 for experiments with gassing at low csat = 106 gSA/kgwater and experiments without gassing with csat = 160 gSA/kgwater. Δd50, which represents the effect of gassing on d50, was bigger at low csat although the absolute difference in d50 was lower. This leads to the conclusion that the broader the MZW of an experiment without gassing the bigger the effect of gassing on d50. Figure 5B shows the results of MZW and d50 for experiments with and without gassing and varying gassing supersaturation. For gassing experiments at csat = 160 gSA/kgwater we published a linear relationship between Δcgassing and d50, where gassing at lower Δcgassing resulted in bigger d50.7 For gassing experiments with low csat we expected that the relationship could be nonlinear. Especially for low Δcgassing, gassing could result in smaller d50 because the amount of nuclei induced might be too low to degrade supersaturation present, causing additional nucleation during cooling. Therefore, we investigated the effect of gassing at csat = 106 gSA/kgwater. During gassing at Δcgassing = 9.5 gSA/kgwater more nuclei were induced in comparison to Δcgassing = 5.3 gSA/kgwater, which resulted in a narrower MZW. More nuclei induced should result in smaller d50, but the d50 was bigger for gassing at Δcgassing = 9.5 gSA/kgwater. As expected, during gassing at Δcgassing = 5.3 gSA/kgwater and csat = 106 gSA/ kgwater the amount of nuclei induced was insufficient to degrade supersaturation completely. Consequently, additional nucleation during cooling led to a higher amount of nuclei which resulted in a smaller d50. Unfortunately, this effect could not be observed as shoulders in the corresponding CSDs. In conclusion, for investigating the effect of gassing at csat = 106 gSA/kgwater, a bigger effect of gassing can be expected if gassing is operated at higher Δcgassing which is expressed by Δd50 also. Figure 5C shows the results of MZW and d50 for experiments with and without gassing and varying cooling rate. Here, we 1033

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

could degrade supersaturation. This led to bigger d50 values after the application of gassing, although Δd50 is very low for the higher csat. The controlled induction of nuclei resulted in enhanced reproducibility of gassing crystallization experiments. Different from the lab scale (Figure 5A), MZW of the experiment with csat = 160 gSA/kgwater was comparatively narrow. Since the MZW depends on various factors like process parameters or crystallizer surface material, as well as stirrer type and material, the reason for the difference observed could not be identified completely. We assumed that a higher temperature gradient between bulk solution and heating jacket was the reason for early nucleation and thus a narrower MZW. From the narrower MZWs it followed that the differences in the d50values for experiments with and without gassing at csat = 160 gSA/kgwater were very low which was expressed also by a low Δd50. Consequently, it was not possible to create a sufficient difference in the amount of nuclei induced by gassing compared to spontaneous nucleation of a normal cooling crystallization. For lower csat the effect expressed in Δd50 was on a similar order of magnitude as the lab scale experiments. For both csat investigated, the d50-values were bigger for experiments with and without gassing compared to those of the lab scale experiments (Figure 5A). Nucleation is supposed to occur preferably at the surfaces like the crystallizer wall. The ratio of the surface area of the crystallizer wall to the liquid volume was smaller for the pilot scale which resulted in relatively fewer nuclei created. Fewer nuclei led to comparatively bigger d50. It can be concluded that gassing affects MZW for the pilot scale also. The MZW correlates directly with the amount of nuclei created and thus with the d50 of the final product. The experiments at pilot scale demonstrate that the potential of gassing crystallization is greatest if process conditions lead to broad and hardly controlled MZWs.

can be described by a linear regression function which is also given in Figure 6. By applying classical nucleation theory formulas (section 2.5), the apparent nucleation order n can be determined from the slope of the regression functions while the nucleation rate constant kb is calculated from the intercept with the y-axis. The results are summarized in Table 2. Gassing especially enhances Table 2. Nucleation Kinetic Parameters for Experiments without and with Gassing without gassing with gassing

n [−]

kb [−]

3.29 2.82

0.0038 0.0174

the nucleation rate constant for several factors, while the nucleation order remains in a similar order of magnitude. This indicates a change in the nucleation mechanism. The results demonstrate the effect of gassing on nucleation qualitatively for the first time. 3.3. Pilot Scale Experiments. The aim of pilot plant experiments was to show that the effect of gassing crystallization depends on the MZW mainly and does not exist for lab scale experiments exclusively. The results of the previous section showed that different MZW could be expected for different csat, which was additionally a parameter easy to adjust. The resulting MZWs and d50 values for experiments with and without gassing at csat = 106 gSA/kgwater and csat = 160 gSA/kgwater are shown in Figure 7. Figure 7 shows the results of MZW and d50 for experiments with and without gassing and varying saturation concentration. More or less, the findings during lab scale experiments were reflected here. The MZWs for experiments at low csat were broader and could be narrowed by gassing. The median diameter was lower for low csat and enhanced by gassing. Reproducibility of gassing crystallization experiments was always good and better in comparison to normal cooling crystallization. MZWs were narrower for higher csat, because more succinic acid molecules were present in the solution enhancing the probability of the formation of stable nuclei. Product crystals at higher csat had bigger d50 because MZWs were narrower leading to fewer nuclei and the theoretical yield was higher for those experiments causing crystals to grow to bigger sizes. Induced nucleation by gassing led to narrower MZWs because nuclei were present at an earlier point of the cooling process and

4. CONCLUSION The purpose of this paper was to validate the hypothesis that especially for parameter combinations which result in broad MZW the success of gassing crystallization is independent of crystallizer scale and geometry. Therefore, the effect of gassing on the median crystal diameter was evaluated in comparison to normal cooling crystallization. At first, lab scale experiments were performed, varying process parameters (saturation concentration, gassing supersaturation, cooling rate, and stirrer speed) which all affected MZW. Then, the effect of gassing on the median diameter was evaluated for pilot scale experiments

Figure 7. MZWs (left) and d50 (right) for experiments with and without gassing with varying saturation concentration. Constant process parameters: R = 60 1/min, κ = 0.25 K/min, Δcgassing = 5.3 gSA/kgwater, tgassing = 20 s, V̇ gassing = 3400 L/h. Width of CSD is provided in Supporting Information Table S2. 1034

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035

Crystal Growth & Design

Article

cTfinal

for varying saturation concentrations, which ensured different MZWs also. To quantify the effect of gassing crystallization we introduced Δd50 as a new measure. For lab scale experiments, gassing caused narrower MZWs, d50 values were bigger, and reproducibility was enhanced. Comparing experiments with and without gassing, we found that especially for process conditions resulting in broad MZWs, at which crystallizations are often difficult to handle, our measure Δd50 was largest showing the great potential of gassing crystallization. The calculation of nucleation kinetic parameters for experiments without and with gassing underlined the effect of gassing quantitatively. For pilot scale experiments, we found more or less the same relationship between MZW and the effect of gassing on d50. This verified our hypothesis that, independent of crystallizer scale and geometry, gassing crystallization works best for process conditions which result in broad and hardly controlled MZWs. However, a relationship between the quantitative effect of gassing between lab and pilot scale cannot be drawn yet and should be the object of future investigations.



d50 Δd50 J kb n R Rmin T Tfinal Tnuc Tsat tgassing tnuc V̇ gassing

Greek Letters

κ cooling rate [K/min]



ASSOCIATED CONTENT

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.cgd.6b01280. Tables S1 and S2 are provided including data about width of CSD d90−d10 for the experiments presented in this paper (PDF)

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Tel.: +49 (0)231 755 3020. Fax: +49 (0)231 755 2341. ORCID

Kerstin Wohlgemuth: 0000-0001-7914-4303 Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work is funded by the Ministry of Innovation, Science and Research of the German Federal State of North RhineWestphalia (NRW) and by TU Dortmund University through a scholarship from the CLIB-Graduate Cluster Industrial Biotechnology (CLIB2021).



REFERENCES

(1) Aamir, E.; Nagy, Z.; Rielly, C. Chem. Eng. Sci. 2010, 65, 3602− 3614. (2) Barrett, P.; Smith, B.; Worlitschek, J.; Bracken, V.; O’Sullivan, B.; O’Grady, D. Org. Process Res. Dev. 2005, 9, 348−365. (3) Ulrich, J.; Frohberg, P. Front. Chem. Sci. Eng. 2013, 7, 1−8. (4) Wohlgemuth, K. In Induced Nucleation Processes during Batch Cooling Crystallization; München, 2012. (5) Wohlgemuth, K.; Ruether, F.; Schembecker, G. Chem. Eng. Sci. 2010, 65, 1016−1027. (6) Wohlgemuth, K.; Kordylla, A.; Ruether, F.; Schembecker, G. Chem. Eng. Sci. 2009, 64, 4155−4163. (7) Kleetz, T.; Braak, F.; Wehenkel, N.; Schembecker, G.; Wohlgemuth, K. Cryst. Growth Des. 2016, 16, 1320−1328. (8) Lenka, M.; Sarkar, D. J. Cryst. Growth 2014, 408, 85−90. (9) Kubota, N. J. Cryst. Growth 2008, 310, 629−634. (10) Beckmann, W. In Crystallization; Wiley-VCH: Weinheim, 2013. (11) Kleetz, T.; Funke, F.; Sunderhaus, A.; Schembecker, G.; Wohlgemuth, K. Cryst. Growth Des. 2016, 16, 6797. (12) Mitchell, N. A.; Frawley, P. J. J. Cryst. Growth 2010, 312, 2740− 2746. (13) Liang, K.; White, G.; Wilkinson, D.; Ford, L. J.; Roberts, K. J.; Wood; Will, M. L. Cryst. Growth Des. 2004, 4, 1039−1044. (14) Akrap, M.; Kuzmanić, N.; Prlić-Kardum, J. J. Cryst. Growth 2010, 312, 3603−3608. (15) Bolaños-Reynoso, E.; Sánchez-Sánchez, K. B.; Urrea-García, G. R.; Ricardez-Sandoval, L. Ind. Eng. Chem. Res. 2014, 53, 13180−13194. (16) Choong, K. L.; Smith, R. Chem. Eng. Sci. 2004, 59, 313−327. (17) Shi, D.; El-Farra, N. H.; Li, M.; Mhaskar, P.; Christofides, P. D. Chem. Eng. Sci. 2006, 61, 268−281. (18) Lang, Y.-d.; Cervantes, A. M.; Biegler, L. T. Ind. Eng. Chem. Res. 1999, 38, 1469−1477. (19) Nyvlt, J. J. Cryst. Growth 1968, 3−4, 377−383. (20) Barrett, P.; Glennon, B. Chem. Eng. Res. Des. 2002, 80, 799−805. (21) Nývlt, J.; Rychlý, R.; Gottfried, J.; Wurzelová, J. J. Cryst. Growth 1970, 6, 151−162.

S Supporting Information *



concentration at final temperature of the cooling profile [gSA/kgwater] median crystal diameter [μm] measure for the effect of gassing on the d50-value [μm*kgwater/gSA] nucleation rate [1/m3 s] mass based nucleation rate constant [-] apparent order of nucleation [-] stirrer speed [1/min] minimum stirrer speed [1/min] temperature [°C] final temperature of the cooling profile [°C] temperature at which nucleation occurred [°C] saturation temperature [°C] duration of the gassing process [s] time at which nucleation occurred [min] gas volume flow [L/h]

NOTATION

Abbreviations

ATR-FTIR attenuated total reflectance Fourier transform infrared spectroscopy CSD crystal size distribution FBRM focused beam reflectance measurement GLV gray level value MZW metastable zone width SA succinic acid Symbols

Δcgassing supersaturation, at which gassing is started [gSA/ kgwater] Δcnuc supersaturation, at which nucleation occurs [gSA/ kgwater] csat saturation concentration [gSA/kgwater] 1035

DOI: 10.1021/acs.cgd.6b01280 Cryst. Growth Des. 2017, 17, 1028−1035