Nanopowder Fluidization Using the Combined Assisted Fluidization Techniques of Particle Mixing and Flow Pulsation

: In the present study, we report the ﬂuidization behavior of ultraﬁne nanopowder using the assisted ﬂuidization technique of particle mixing, which was further superimposed with the pulsation of the inlet gas ﬂow to the ﬂuidized bed. The powder selected in the present study was hydrophilic nanosilica, which shows strong agglomeration behavior leading to poor ﬂuidization hydrodynamics. For particle mixing, small proportions of inert particles of Geldart group A classiﬁcation were used. The inlet gas ﬂow to the ﬂuidized bed was pulsed with a square wave of frequency 0.1 Hz with the help of a solenoid valve controlled using the data acquisition system (DAQ). In addition to the gas ﬂow rate to the ﬂuidized bed, pressure transients were carefully monitored using sensitive pressure transducers connected to the DAQ. Our results indicate a substantial reduction in the effective agglomerate size as a result of the simultaneous implementation of the assisted ﬂuidization techniques of particle mixing and ﬂow pulsation.


Introduction
Process industries often employ fine and ultrafine powders to enhance surface-based rate processes such as gas-solid-catalyzed reactions and various separation processes [1,2]. An intimate contact between the solid and the fluid phase is an important prerequisite for their efficient utilization. Ultrafine nanopowders, whose size typically ranges from 1 to 100 nm, have a wide range of applications, from in research laboratories to large-scale industrial applications. Their extremely high surface area-to-volume ratio can significantly enhance the processes that depend upon the availability of the surface area. However, the efficacy of nanopowders is often compromised due to their agglomeration behavior, which occurs as a result of strong interparticle van der Waals forces [3][4][5][6][7].
Fluidized beds enhance the interphase mixing, and impart high mass and heat transfer rates while limiting the pressure drop to effective bed weight, even at high fluid velocities. Their hydrodynamics, however, strongly depend on the physical characteristics of the solid particles such as particle diameter and density [1,8,9]. Classified as Geldart group C particles (D p < 30 µm), nanopowders usually show strong agglomeration behavior, which induces non-homogeneities in the bed during their fluidization. This leads to poor interphase mixing, thereby affecting the effectiveness of the fluidized bed [10][11][12][13].
Assisted fluidization techniques are often used to improve the fluidization behavior by providing additional energy to overcome the interparticle forces and promote deagglomeration [14]. Such techniques sometimes involve internal or external vibrations of the fluidized beds using various tapped density was found to be only 50 kg/m 3 , which indicated the highly porous nature of the bed with a void fraction of 0.98. For mixing, we used inert Geldart group A particles (sand) with a size range of 38-75 µm (Sauter mean diameter = 56.5 µm) and a density of 2664 kg/m 3 . Compressed air, at room temperature, was used as fluidizing medium in the experiments.
The experimental setup (Figure 1) consisted of a test section, which was a transparent Perspex column of height 1.5 m and internal diameter 0.07 m. It was preceded by a 0.50 m-long calming section. We used a nylon mesh-covered perforated distributor with a 4% fractional open area to ensure uniform airflow in the test section.
We used hydrophilic fumed SiO2 (Aerosil 200; Evonik Industries, Essen, Germany) in our experiments. Its primary size was 12 nm and it had a surface area of 200 ± 25 m 2 /g. However, due to agglomeration, dry-particle size analysis yielded a wide size distribution, ranging from 2 to 100 µm with a Sauter mean diameter of 20 µm [30]. Though its true density was 2200 kg/m 3 , the tapped density was found to be only 50 kg/m 3 , which indicated the highly porous nature of the bed with a void fraction of 0.98. For mixing, we used inert Geldart group A particles (sand) with a size range of 38-75 µm (Sauter mean diameter = 56.5 µm) and a density of 2664 kg/m 3 . Compressed air, at room temperature, was used as fluidizing medium in the experiments.
The experimental setup (Figure 1) consisted of a test section, which was a transparent Perspex column of height 1.5 m and internal diameter 0.07 m. It was preceded by a 0.50 m-long calming section. We used a nylon mesh-covered perforated distributor with a 4% fractional open area to ensure uniform airflow in the test section. The pressure transients of the nanoparticle bed were recorded using a highly sensitive fastresponse bidirectional differential pressure transducer (Omega PX163-005BD5V; Norwalk, CT, USA) between pressure taps located at distances of 110 and 230 mm above the distributor. A data acquisition system (DAQ) with LabVIEW software (National Instruments, Austin, Texas, USA) was used to record pressure transients at a rate of 100 data/s from the pressure transducers' voltage signals.
The inlet air flow to the fluidized bed was pulsed at a frequency of 0.1 Hz using a solenoid valve (Omega SV 3310; Norwalk, CT, USA). The valve was kept open for 5 s followed by 5 s of closing time, thus completing a cycle of 0.1 Hz. As shown in Figure 2, the 0.1 Hz pulsation allowed complete settling of the bed once the flow was stopped and, moreover, led to complete fluidization when the solenoid valve was open during the flow pulsation [26,40]. The solenoid valve was controlled using a digital I/O signal from the data acquisition system (DAQ) and LabVIEW software.
In the case of flow pulsation, the average pressure drop was carefully computed to avoid the results being affected by the opening and closing of the solenoid. To this end, the initial 2.50 s and the final 0.25 s of the pressure transients during the input flow pulse of 5 s were eliminated, and the The pressure transients of the nanoparticle bed were recorded using a highly sensitive fast-response bidirectional differential pressure transducer (Omega PX163-005BD5V; Norwalk, CT, USA) between pressure taps located at distances of 110 and 230 mm above the distributor. A data acquisition system (DAQ) with LabVIEW software (National Instruments, Austin, Texas, USA) was used to record pressure transients at a rate of 100 data/s from the pressure transducers' voltage signals.
The inlet air flow to the fluidized bed was pulsed at a frequency of 0.1 Hz using a solenoid valve (Omega SV 3310; Norwalk, CT, USA). The valve was kept open for 5 s followed by 5 s of closing time, thus completing a cycle of 0.1 Hz. As shown in Figure 2, the 0.1 Hz pulsation allowed complete settling of the bed once the flow was stopped and, moreover, led to complete fluidization when the solenoid valve was open during the flow pulsation [26,40]. The solenoid valve was controlled using a digital I/O signal from the data acquisition system (DAQ) and LabVIEW software.
Appl. Sci. 2019, 9, 572 4 of 11 data from the remaining 2.25 s interval were used to compute the mean value. The data averaging procedure is discussed in detail elsewhere [27]. The following experimental strategy was followed in the present work: 1. Unassisted fluidization experiments were carried out by first gradually increasing the velocity followed by a gradual defluidization.
2. Before carrying out the fluidization and defluidization experiment, 4.5 vol % Geldart group A particles were added and thoroughly mixed with the nanopowder.
3. The inlet air flow to the fluidized bed was pulsed at 0.1 Hz.
4. Steps 2 and 3 were repeated after increasing the fraction of group A particles to 8.6 vol %. Two Gilmont flow meters of different ranges were used to control the airflow to the fluidized bed. The superficial velocity was gradually increased up to a maximum of 171.5 mm/s to avoid the elutriation of nanoparticles at higher velocities.

Mathematical Equations Used
The average pressure drop and standard deviation were calculated from the pressure drop data using the following equations: The pressure drop in the packed bed is described by the well-known Ergun equation: where 'ε' is the bed void fraction, ' ' is the fluid superficial velocity, 'µ' is the viscosity, and ' ' is the average diameter of the agglomerates of nanoparticles.
Since the fluidized bed pressure drop is equal to the effective weight of the solid in the bed, one can, therefore, write where ' ', ' ', and ' ' are true solid density, fluid density, and bulk density, respectively. In the case of flow pulsation, the average pressure drop was carefully computed to avoid the results being affected by the opening and closing of the solenoid. To this end, the initial 2.50 s and the final 0.25 s of the pressure transients during the input flow pulse of 5 s were eliminated, and the data from the remaining 2.25 s interval were used to compute the mean value. The data averaging procedure is discussed in detail elsewhere [27].
The following experimental strategy was followed in the present work: 1.
Unassisted fluidization experiments were carried out by first gradually increasing the velocity followed by a gradual defluidization.

2.
Before carrying out the fluidization and defluidization experiment, 4.5 vol % Geldart group A particles were added and thoroughly mixed with the nanopowder. 3.
The inlet air flow to the fluidized bed was pulsed at 0.1 Hz.

4.
Steps 2 and 3 were repeated after increasing the fraction of group A particles to 8.6 vol %.
Two Gilmont flow meters of different ranges were used to control the airflow to the fluidized bed. The superficial velocity was gradually increased up to a maximum of 171.5 mm/s to avoid the elutriation of nanoparticles at higher velocities.

Mathematical Equations Used
The average pressure drop and standard deviation were calculated from the pressure drop data using the following equations: The pressure drop in the packed bed is described by the well-known Ergun equation: where 'ε' is the bed void fraction, 'U 0 ' is the fluid superficial velocity, 'µ' is the viscosity, and 'D av ' is the average diameter of the agglomerates of nanoparticles.
Since the fluidized bed pressure drop is equal to the effective weight of the solid in the bed, one can, therefore, write where 'ρ p ', 'ρ f ', and 'ρ b ' are true solid density, fluid density, and bulk density, respectively.
Here, 'M bed ' is the total solid mass in the bed while 'V bed ' is the volume of the solids in the bed.
When the fluidized bed contains a binary mixture of different-sized particles, the average diameter of the particles can be written as where D i is the diameter of the i th component and X i is the fluid-free volume fraction of the i th component.

Results and Discussion
The dependence of the pressure drop on the superficial velocity is reported in Figure 3. The ordinate in the figure is the normalized pressure drop, which was computed by dividing the pressure drop by the effective weight of the bed. During the unassisted fluidization, a significant hysteresis is visible between the fluidization and defluidization. This is caused by initial non-homogeneities due to the presence of cracks, channels, and plugs, which leads to non-uniform flow of air through the bed. The bed is more homogenous during the defluidization. The difference in bed homogeneity usually results in hysteresis as discussed in detail by Gomez-Hernandez et al. [41]. The addition of even a small proportion of Geldart group A particles helps to mitigate the effect of hysteresis, which is further diminished when the flow is pulsed, as seen in Figure 4. The fluidization behavior is clearly further improved with the combination of assisted fluidization techniques of flow pulsation and particle mixing.
Here, ' ' is the total solid mass in the bed while ' ' is the volume of the solids in the bed.
When the fluidized bed contains a binary mixture of different-sized particles, the average diameter of the particles can be written as where is the diameter of the component and Xi is the fluid-free volume fraction of the component.

Results and Discussion
The dependence of the pressure drop on the superficial velocity is reported in Figure 3. The ordinate in the figure is the normalized pressure drop, which was computed by dividing the pressure drop by the effective weight of the bed. During the unassisted fluidization, a significant hysteresis is visible between the fluidization and defluidization. This is caused by initial non-homogeneities due to the presence of cracks, channels, and plugs, which leads to non-uniform flow of air through the bed. The bed is more homogenous during the defluidization. The difference in bed homogeneity usually results in hysteresis as discussed in detail by Gomez-Hernandez et al. [41]. The addition of even a small proportion of Geldart group A particles helps to mitigate the effect of hysteresis, which is further diminished when the flow is pulsed, as seen in Figure 4. The fluidization behavior is clearly further improved with the combination of assisted fluidization techniques of flow pulsation and particle mixing.
Another interesting feature of Figure 3 is that the particle mixing causes a higher pressure drop at low velocities when the bed is not fluidized. The higher the fraction of external particles, the greater the pressure drop, because the addition of external particles lowers the bed void fraction due to the volume-contraction phenomenon, which is reflected in the increase of the pressure drop in the bed [28]. In fact, it was pointed out that a decrease in bed expansion during long-term fluidization of hydrophilic nano-titania at high velocities cannot be attributed to a reduction in the size of agglomerates [42]. Rather, it appears that since high velocities promote greater solid mixing in the bed, smaller agglomerates tend to occupy the interstitial spaces of their large counterparts, thus contributing to the bed volume contraction. The effect of the flow pulsation on the pressure drop profile is more pronounced for a fluidized bed containing a small amount of external particles ( Figure  4a). At higher proportions of external particles, the pressure drop behavior is not affected so much by the flow pulsation (Figure 4b).   It is clear from the pressure drop profiles shown in Figure 4 that hysteresis is present even with particle mixing, especially when higher amounts of external particles are added. The use of flow pulsation, however, eliminates the hysteresis, which is a clear indication of the removal of bed nonhomogeneities.
The minimum fluidization velocity ( ) is an important parameter that characterizes the hydrodynamics of the fluidized bed. At incipient fluidization, the packed pressure drop is equal to the effective bed weight. Using this approach, we determined the minimum fluidization velocity as shown in Figure 5. The results are reported in Table 1. The void fraction at incipient fluidization was computed from the bed expansion data. Clearly, particle mixing lowered the by more than onehalf of the corresponding value obtained using unassisted fluidization. When flow pulsation was introduced, a further substantial reduction in the was seen. Thus, the use of the combined assisted fluidization techniques of particle mixing and flow pulsation helped to reduce the Umf from 118 to 25 mm/s, which is, indeed, a significant reduction. Another interesting feature of Figure 3 is that the particle mixing causes a higher pressure drop at low velocities when the bed is not fluidized. The higher the fraction of external particles, the greater the pressure drop, because the addition of external particles lowers the bed void fraction due to the volume-contraction phenomenon, which is reflected in the increase of the pressure drop in the bed [28]. In fact, it was pointed out that a decrease in bed expansion during long-term fluidization of hydrophilic nano-titania at high velocities cannot be attributed to a reduction in the size of agglomerates [42]. Rather, it appears that since high velocities promote greater solid mixing in the bed, smaller agglomerates tend to occupy the interstitial spaces of their large counterparts, thus contributing to the bed volume contraction. The effect of the flow pulsation on the pressure drop profile is more pronounced for a fluidized bed containing a small amount of external particles (Figure 4a). At higher proportions of external particles, the pressure drop behavior is not affected so much by the flow pulsation (Figure 4b).
It is clear from the pressure drop profiles shown in Figure 4 that hysteresis is present even with particle mixing, especially when higher amounts of external particles are added. The use of flow pulsation, however, eliminates the hysteresis, which is a clear indication of the removal of bed non-homogeneities.
The minimum fluidization velocity (U m f ) is an important parameter that characterizes the hydrodynamics of the fluidized bed. At incipient fluidization, the packed pressure drop is equal to the effective bed weight. Using this approach, we determined the minimum fluidization velocity as shown in Figure 5. The results are reported in Table 1. The void fraction at incipient fluidization was computed from the bed expansion data. Clearly, particle mixing lowered the U m f by more than one-half of the corresponding value obtained using unassisted fluidization. When flow pulsation was introduced, a further substantial reduction in the U m f was seen. Thus, the use of the combined assisted fluidization techniques of particle mixing and flow pulsation helped to reduce the U mf from 118 to 25 mm/s, which is, indeed, a significant reduction. However, when the fraction of external particles was increased to 8.6 vol %, no reduction in the Umf was noticed as compared to the case of 4.5 vol %. Instead, there was an increase, however small. Introducing flow pulsation to the mixed fluidized bed, nonetheless, once again significantly lowered the Umf to 33 mm/s, which is, however, higher than the 25 mm/s achieved in the case of 4.6 vol %.
We further analyzed our pressure drop data using the Ergun equation (Equation 4), which is valid for velocities in the packed bed region when the bed is not fluidized. According to the Ergun equation, the pressure drop (∆P) depends on the bed void fraction (ε) and agglomerate diameter ( ). The bed void fraction (ε) was computed from the bed height using Equation (5). As seen in Table 2, these values are lower than the corresponding incipient values ( ) reported in Table 1. This leaves diameter as the only remaining unknown parameter in the Ergun equation. As shown in Figure 6, we fitted our experimental data with the Ergun equation in order to determine the effective diameter of the agglomerates of nanoparticles present in the bed. The results obtained using regression analysis are reported in Table 2. The R 2 values in most cases indicated an excellent fit with the experimental data (more than 90% in all cases). The effective diameter in the table is the average diameter for both nanoparticles as well as external particles, while the mean agglomerate diameter is the average diameter of agglomerates only, which was calculated using Equation (9). Clearly, particle  However, when the fraction of external particles was increased to 8.6 vol %, no reduction in the U mf was noticed as compared to the case of 4.5 vol %. Instead, there was an increase, however small. Introducing flow pulsation to the mixed fluidized bed, nonetheless, once again significantly lowered the U mf to 33 mm/s, which is, however, higher than the 25 mm/s achieved in the case of 4.6 vol %.
We further analyzed our pressure drop data using the Ergun equation (Equation (4)), which is valid for velocities in the packed bed region when the bed is not fluidized. According to the Ergun equation, the pressure drop (∆P) depends on the bed void fraction (ε) and agglomerate diameter (D av ). The bed void fraction (ε) was computed from the bed height using Equation (5). As seen in Table 2, these values are lower than the corresponding incipient values (ε m f ) reported in Table 1. This leaves diameter as the only remaining unknown parameter in the Ergun equation. As shown in Figure 6, we fitted our experimental data with the Ergun equation in order to determine the effective diameter of the agglomerates of nanoparticles present in the bed. The results obtained using regression analysis are reported in Table 2. The R 2 values in most cases indicated an excellent fit with the experimental data (more than 90% in all cases). The effective diameter in the table is the average diameter for both nanoparticles as well as external particles, while the mean agglomerate diameter is the average diameter of agglomerates only, which was calculated using Equation (9). Clearly, particle mixing helped to decrease the size of nanoparticles agglomerates in the packed bed. The agglomerate diameter decreased by 33.2% and 48.4% with the addition of 4.5 vol % and 8.6 vol % group A particles, respectively. The agglomerate diameter further reduced to 39.6% with the combined effect of 4.5 vol % particle mixing and pulsed airflow as compared to the case of particle mixing alone. However, for the case of 8.6 vol % particle mixing and flow pulsation together, the agglomerate size was further reduced to 50.7 %. This reduction is not as significant as the one seen in the case of 4.5 vol % when flow pulsation was applied.    These results indicate that the use of flow pulsation with particle mixing further enhanced the deagglomeration phenomenon. However, the effect of pulsed flow is more prominent in 4.5 vol % particle mixing than in 8.6 vol % particle mixing. The size reduction can be attributed to the presence of larger micron-sized Geldart group A particles in the bed of nanoparticles. The repeated collision of denser and larger external particles tends to fragment agglomerates of nanoparticles. Moreover, the interagglomerate force equilibrium is disturbed, owing to the presence of different external particles in their midst, which leads to deagglomeration.

Conclusions
The hysteresis effect is reduced by the assisted fluidization technique of particle mixing and almost completely eliminated by the combined application of particle mixing and flow pulsation. Since hysteresis mainly arises due to non-homogeneities present in the bed, it is obvious that flow pulsation in conjunction with particle mixing is quite effective in improving the fluidization hydrodynamics by eliminating various kinds of bed non-homogeneities.
Though the assisted fluidization technique of particle mixing was effective in lowering the U m f , better results were obtained with 4.5 vol % as compared to the case of 8.6 vol %. When the flow was also pulsed for mixed beds, a further substantial reduction was obtained in the value of the U m f . This reduction was, however, higher in the case of 4.5 vol %.
Particle mixing using Geldart group A promoted deagglomeration behavior in the fluidized bed of nanopowder. As much as a 48% size reduction was achieved with 8.6 vol % particle mixing. The assisted fluidization technique of flow pulsation was more effective for 4.5 vol % as compared to the case of 8.6 vol %.