APPLIED RHEOLOGY FOR CHARACTERIZATION OF NANOFIBER BASED FILTERS

Transcription

1 APPLIED RHEOLOGY FOR CHARACTERIZATION OF NANOFIBER BASED FILTERS Martin Zatloukal 1, Wannes Sambaer 1 and Dusan Kimmer 2 1 Polymer Centre, Faculty of Technology, Tomas Bata University in Zlin, Vavreckova 275, Zlin, Czech Republic 2 SPUR a.s., T. Bati 299, Zlin, Czech Republic Abstract Full 3D polydisperse particle filtration modeling at low pressures has been performed for a polyurethane nanofiber based filter prepared via electrospinning process in order to more deeply understand the filter clogging and the cake formation. In this work, realistic SEM image based 3D filter model, transition/free molecular flow regime, Brownian diffusion, aerodynamic slip, particlefiber and particle-particle interactions together with a Euclidian distance map based methodology to calculate the pressure drop have been utilized. Model predictions have been compared with relevant experimental data in order to validate the used assumptions, methodologies and numerical scheme. The effect of particle-particle as well as particle-fiber interactions on the nanofiber based filter efficiency, pressure drop and the quality factor during the filter clogging has been investigated in more detail. Introduction Fibrous based air filters are widely used in many applications in area of pharmaceutical, medical, biology, food, semi-conductor industries and therefore it is not surprising that both experimental and theoretical studies dealing with their basic characteristics (such as the pressure drop and particle collection efficiency). Nowadays, nanofiber nonwoven based filters becomes of high interest due to their ability to reach high filtration efficiency for ultrafine particles with a low pressure drop due to aerodynamic slip around the nanofibers [1-8]. Unfortunately, till now, the detailed theoretical 3D analysis of the nanofiber based filters clogging has not been performed yet and thus, the full understanding of the filtration cake formation on the nanofiber based filters and its role on the final filter efficiency is not fully understood. In order to provide a better understanding of the nanofiber based filters loading process with respect to pressure drop and filtration efficiency evolution, a realistic SEM image based 3D structure model of the filter, transition/free molecular flow regime, Brownian diffusion, particle-fiber interactions, aerodynamic slip and sieve has been utilized in this work. Experimental Materials A polyurethane solution in dimethylformamide (DMF) based on 4,4 methylenebis(phenylisocyanate) (MDI), poly(3-methyl-1,5-pentanediol)-alt-(adipic, isophtalic acid) (PAIM) and 1,4 butanediol (BD) was synthesized in molar ratio 9:1:8 at 90 C for 5 hours (per partes way of synthesis starting with preparation of prepolymer from MDI and PAIM and followed by addition of BD and remaining quantity of MDI). The prepared solution was suitable for electrospinning process and had a PU concentration of 24.1 wt%, viscosity of 1.45 Pa.s and conductivity of 157 µs/cm (adjusted by 3 parts of citric acid and 1 part of borax). Nanofiber Layers Preparation by Electrospinning Process In this work, a commercially available NanoSpider TM machine (Elmarco s.r.o. Liberec, Czech Republic, equipped with patented (PCT/CZ2010/000042) rotating electrode with 3 cotton cords spinning elements was used to prepare nanofiber nonwoven based filter sample from polyurethane solution. The experimental conditions were: relative humidity 25 %, temperature 23.5 C, high voltage applied into PU solution 75 kv, distance between electrodes 210 mm, rotational electrode speed 7 rpm and speed of polyester nonwoven supporting textile collecting nanofibers was 0.4 m/min. Samples for further analyses were taken from the middle part of the 40 cm wide fabric coated with layer of nanofibers. Nanofiber Filter Characterization A nanofiber based filter, prepared via electrospinning process, has been characterized by Scanning Electron Microscope technique (Vega 3, Tescan, Czech Republic), as visible in Figure 1a. The obtained SEM picture has consequently been used for the fiber diameter/pore size distribution determination by using recently proposed digital image analysis technique [9]. The filter has an average fiber diameter and average pore size equal to 133 nm and 156 nm, respectively, thickness (3.59 µm) and mass area (0.54 g/m²). The electric potential generated on the surface of the filter was measured by electrostatic fieldmeter Simco, model FMX-003, the Netherlands. It SPE ANTEC Anaheim 2017 / 177

2 has been found that the surface charge potential of the filter was practically negligible, equal to 70 V, i.e. there are practically no electrostatic forces, which could affect its filtration characteristics. 1a) 1b) 5µm Perspective View Theoretical All calculations performed in this work have been obtained with in-house developed software (UTBsoft Filtration) which has been ran on a PC utilizing Windows 7 Professional with the following configuration: Intel Core I7-980X processor equipped with 6 cores able to run 12 threads, having each a clock speed of 3.33 GHz, and 24 GB DDR3 memory and a NVIDEA Quadro 2000 graphical card. The utilized filtration model is provided in [4-5]. In this work, the fully sticking (single touching point) boundary condition was applied (similarly to other theoretical studies, which can be justified for the 10nm- 5µm particle size range as showed experimentally by Wang and Kasper [10]. Under this condition, dendrites (formed by the accumulated particles during the filtration process) cannot bend or break down. Evaluation of Model Prediction Capabilities for Clean Filter The filtration efficiency of the filter is theoretically determined by the number of particles, which can be captured by the filter [3]: Nin N E N in out, (2) S1 Top View Side View Figure 1. Manufactured nanofiber based nonwoven filter. 1a) SEM image. 1b) Full 3D structure model of the manufactured filter. The filtration efficiency for the prepared nanofiber based filter was determined experimentally in the ultrafine particle range Note that the filtration efficiency and pressure drop measurements were repeated four times and the following filtration efficiency characteristics of the filter were experimentally determined: the Maximum Penetration Pore Size (MPPS) 70 nm, the minimum filtration efficiency %, the average pressure drop Pa and the quality factor 27.5 kpa -1, which is defined according to Eq. 1. QF ln( E) p (1) where p is the pressure drop and E is the filtration efficiency at MPPS. where N in and N out are the numbers of entering and exiting particles, respectively. Twenty different groups of identical particles having diameter linearly varying from 20 to 500 nm on a logarithmic scale were injected in the large modeled clean filter inlet having dimensions of 7.2 µm on 7.2 µm. Polyurethane nanofiber 3D structure model (see Figure 1b), generated from the SEM image, was utilized in the filtration modeling. The comparison between the predicted and experimentally determined filtration efficiency, pressure drop and quality factor data for the tested filter samples is provided in Figure 2. It is visible, that firstly, the model predictions are in a fairly good agreement with the corresponding experimental data, which justifies the used model assumptions. Deposition Dynamics vs. Filtration Characteristics In order to understand deposition dynamics for nanofiber based filter, the 3D structure model of the filter depicted in Figure 1b was randomly penetrated with particles, having the particle size distribution the same as the nebulized solution used in the experimental filtration test, considering both particle-fiber and particle-particle interactions (see Figure 3). Calculated pressure drop as the function of the depth direction is provided in Figure 4. The closer analysis revealed existence of four regions with different pressure gradients entitled here as A, B, C and D. The region A describes top cake active layer open to the ambient air at which particle attachment takes predominantly the place. SPE ANTEC Anaheim 2017 / 178

3 2a) average pore size diameter (156 nm in this case). In numbers, the overall packing factor for region D (determined as the volume fraction occupied by particles and fibers) is whereas the packing factor in region B is Note that the packing factor of the clean filter (i.e. the volume fraction occupied by fibers only) is b) Figure 3. Side view of predicted particles deposit. Figure 2. Filtration model predictions vs. experimental data. 2a) Filtration efficiency. 2b) Pressure drop and quality factor. Due to the openness of this region the pressure gradient is the smallest here. The region B represents consolidated layer of the cake characterized by a constant mean particles density where the probability that the new particle arrives from the surrounding gas is minimal. Thus, the pressure gradient in region B is higher than in region A. In region C (near-fiber wall region), the particle density is the highest due to presence of nanofibers having high collecting surface area. As the result, the pressure gradient in region C is the highest. It is interesting that even if the near-fiber wall region is very thin, it consumes considerable amount of pressure. The region D consists of the clogged filter at which particle density decreases with the filter depth. The pressure gradient in region D is comparable with region B due to structural similarity in both regions leading to equivalent resistance against the air flow. In this particular case, the overall structural similarity of D region (containing low number of nonconsolidated particles) with region B (containing high number of consolidated particles) is reached by the presence of nanofiber based filter having very small Figure 4. Calculated total and local pressure drop as the function of the depth direction for top cake active layer (region A), consolidated layer of the cake (region B), near-fiber wall region (region C) and clogged filter region (region D). The role of Particle-Fiber Friction Coefficient In order to more deeply understand the particle-fiber friction coefficient effect on the cake build up, filtration efficiency, pressure drop and quality factor of the nanofiber based filters, its value was varied in the utilized filtration model. The 3D filter model was randomly penetrated with particles, with an experimentally defined particle size distribution. The filtration efficiency, the pressure drop and quality factor (calculated by Eq. 1) was recorded during the filter clogging as a function of SPE ANTEC Anaheim 2017 / 179

4 the captured particle mass for different particle-fiber frictions and the results are depicted in Figure 5a. Clearly, the increase in the fiber-particle friction coefficient leads to higher quality factor. On the other hand, the filter samples with higher fiber-particle friction coefficient shows much higher quality factor reduction with the increased collected particle mass in comparison with the filter samples having lower fiber-particle friction coefficient. These theoretical conclusions are supported by the recent experimental data showing that roughness structure of nanofibers, created by nano protrusions, shows more enhances filtration properties (the quality factor 30.8 kpa -1 ) than circular cross section nanofibers (the quality factor 21.8 kpa -1 ) for given polymer type and air filtration conditions [11]. In order to understand these findings in more detail, pressure and packing profiles for two different particle-fiber friction coefficients (0 and 0.033) were generated and plotted in Figure 5b and Figure 5c, respectively. As it can be seen, particle-fiber friction coefficient influences pressure and packing profiles only at the near-fiber wall region C and clogged filter region D. In more detail, increase in friction coefficient increases the pressure drop and packing factor in region C considerably. On the other hand, in region D, the trend is switched and the effect of friction coefficient on the pressure drop is very small due to the fact that the packing factor of the particles is one order of magnitude lower than packing factor of the clean filter, i.e. the variation of the particle packing factor due to change in friction coefficient has only a small effect on the pressure drop here. This suggests that fiber-particle friction coefficient controls the overall pressure drop predominantly via the packing factor at the near-fiber wall region C at which the cake build up is initiated. In this particular case, increase of particle-fiber friction coefficient from 0 to increases the overall pressure drop by 15.6%. Conclusions Polyurethane nanofiber based filter has been prepared by electrospinning process having average fiber diameter and average pore size equal to 133nm and 156 nm, respectively, thickness 3.59 μm, mass area 0.54 g/m 2, maximum penetration pore size 70 nm, minimum filtration efficiency %, average pressure drop Pa and quality factor 27.5 kpa -1. The realistic SEM image based 3D filter model, transition/free molecular flow regime, Brownian diffusion, aerodynamic slip, particlefiber and particle-particle interactions have been utilized for the prepared polyurethane nanofiber based filter in order to more deeply understand the filtration cake formation and its role on the final filter efficiency and the pressure drop. Model predictions have been found to be in a fairly good agreement with the performed filtration efficiency, pressure drop and quality factor measurements for the clean filter, which justifies the used assumptions and numerical scheme for the given filter type and filtration conditions. 5a) 5b) 5c) Figure 5. Theoretically predicted effect of fiber-particle friction coefficient during filter clogging. 5a) Quality factor. 5b) Pressure drop. 5c) Packing factor. SPE ANTEC Anaheim 2017 / 180

5 Based on the performed theoretical analysis, the following conclusion remarks can be formulated. Increase in the fiber-particle friction coefficient leads to filter quality factor increase due to the creation of highly packed and thin cake layer occurring just at the top filter nanofibers, which causes much higher increase in the filtration efficiency in comparison with the pressure drop. On the other hand, filters with higher fiber-particle friction coefficients show much higher quality factor reduction with the increased collected particle mass than filters having lower fiber-particle friction coefficients. Acknowledgments The authors wish to acknowledge the Grant Agency of the Czech Republic (grant No S) for the financial support. References 1. Sun, B., Long, Y.Z., Zhang, H.D., Li, M.M., Duvail, J.L., Jiang, X.Y., Yin, H.L., Progress in Polymer Science, 39(5), (2014). 2. Hosseini, S.A., Tafreshi, H.V., Chem. Eng. Sci., 65, (2010). 3. Maze, B., Tafreshi, H.V., Wang, Q., Pourdeyhimi, B., J. Aerosol Sci. 38(5), (2007). 4. Sambaer, W., Zatloukal, M., Kimmer, D., Chem. Eng. Sci., 66, (2011). 5. Sambaer, W., Zatloukal, M., Kimmer, D., Chem. Eng. Sci., 82, (2012). 6. Cai, R.-R., Zhang, L.-Z., Yan, Y., Applied Thermal Engineering, 111, (2017). 7. Zhao, X., Wang, S., Yin, X., Yu, J., Ding, B., Scientific Reports, 6, art. no (2016). 8. Liu, B., Zhang, S., Wang, X., Yu, J., Ding, B., Journal of Colloid and Interface Science, 457, (2015). 9. Sambaer, W., Zatloukal, M., Kimmer, D., Polym. Test. 29, (2010). 10. Wang, H.C., Kasper, G., J. Aerosol Sci., 22(1), (1991). 11. Wang, N., Si, Y., Wang, N., Sun, G., El-Newehy, M., Al-Deyab, S.S., Ding, B., Separation and Purification Technology, 126, (2014). SPE ANTEC Anaheim 2017 / 181