Membrane-assisted VOC removal from aqueous acrylic latex

June 5, 2017 | Autor: Bridget Ulrich | Categoria: Engineering, Membrane Science, CHEMICAL SCIENCES
Share Embed


Descrição do Produto

Journal of Membrane Science 452 (2014) 426–432

Contents lists available at ScienceDirect

Journal of Membrane Science journal homepage: www.elsevier.com/locate/memsci

Membrane-assisted VOC removal from aqueous acrylic latex Bridget Ulrich a, Timothy C. Frank a,b, Alon McCormick a, E.L. Cussler a,n a b

Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, MN 55455, USA Engineering & Process Science Laboratory, The Dow Chemical Company, Midland, MI 48667, USA

art ic l e i nf o

a b s t r a c t

Article history: Received 4 March 2013 Received in revised form 11 October 2013 Accepted 12 October 2013 Available online 23 October 2013

Volatile organic compounds (VOCs) in model aqueous solutions and in acrylic latex binder used for formulating latex paint can be stripped without foaming by using a nanoporous hydrophobic polypropylene membrane. The stripping gas employed was dry, room temperature nitrogen or 50 1C water-saturated air. For the dry nitrogen stripping gas, fouling was minimal for the hydrophobic polypropylene over several days, in sharp contrast to experiments with hydrophilic membranes. No fouling of the polypropylene membrane was observed for experiments with the water-saturated strip gas. The rate of VOC removal in these experiments depends on mass transfer in the aqueous latex and in the membrane. These results allow estimates of the membrane area to strip VOCs from commercially relevant quantities of acrylic latex paint. & 2013 Elsevier B.V. All rights reserved.

Keywords: Latex paint Volatile organic compounds Porous membranes

1. Introduction Latex paint has an odor. When a consumer enters a freshly painted room, he or she will not only notice the color and smoothness of the painted surfaces, but also the smell. This smell has been described as “chemical” or “like vodka”. Most consumers do not like this smell; it causes headaches for some consumers. While the smell is usually flushed from a well-ventilated room in a day or so, it is still undesirable. Odor-free latex paint would be regarded as superior [1,2]. Ironically, the smell in latex paint is due to trace amounts of volatile organic compounds that are not essential to the paint's function. For typical acrylic latex binders like those used in many paint formulations, the most abundant VOCs include acetone and n-butanol. These compounds, typically present at concentrations between 1 and 2500 ppm, do not affect performance properties like ease of application and hiding power. They are added to facilitate various steps in the paint's manufacture, or they are present as impurities. Their removal would improve the final paint formulation by removing the unpleasant smell. Ordinarily, the removal or “stripping” of trace amounts of low molecular weight organics is easily accomplished by contacting the liquid with air, nitrogen or steam. In a typical batch stripping process, gas or vapor is blown through a sparger to create large numbers of small bubbles. Sometimes the contacting is accomplished using a trayed or a packed tower. The organics transfer from the paint to the gas phase due to favorable liquid–vapor equilibrium partition ratios or relative volatilities. Bubbles of gas rise due to buoyancy, and quickly collect the volatile organics because their large surface area overcomes the liquid's slow mass n

Corresponding author. Tel.: þ 1 612 625 1596; fax: þ 1 612 626 7246. E-mail address: [email protected] (E.L. Cussler).

0376-7388/$ - see front matter & 2013 Elsevier B.V. All rights reserved. http://dx.doi.org/10.1016/j.memsci.2013.10.025

transfer. Because air or steam stripping is so effective, it is a widely used separation process in environmental engineering, and is supported by a large and vibrant literature [3–9]. However, air or steam stripping is difficult to carry out on latex because it is stabilized by large amounts of surfactants. When the latex is sparged with air or steam, the small bubbles desired for fast, efficient removal of VOCs can produce large volumes of relatively stable foam. This foam causes major problems in the processing and packaging of the paint. Reducing the detergent concentration or changing the latex properties may make air or steam stripping easier, but these alterations are believed to compromise other properties of the paint. In this paper we explore the use of membranes as an alternative means to remove VOCs from latex and to avoid foaming. Such a process has been used in other situations [8,9]. In our experiments, a nanoporous, hydrophobic membrane separates flowing latex paint from a stream of water-saturated air. The membrane should stabilize the gas–latex interface, while potentially providing a large interfacial area for rapid VOC removal. The membrane should also suppress bubble formation, and hence eliminate foaming. At the same time, the membrane must not significantly retard mass transfer, because the desire to achieve fast VOC removal motivates this work in the first place. In the following sections, we first describe our experiments with model solutions and with latex. We then report our results, emphasizing the search for the mechanism of VOCs removal. This mechanism is the key to judging if membrane-assisted stripping of latex paint has commercial value.

2. Theory To analyze the removal of organic compounds from this system, we consider a reservoir of liquid containing one typical solute at

B. Ulrich et al. / Journal of Membrane Science 452 (2014) 426–432

concentration c1. In our experiments, this liquid will either be an aqueous solution of dissolved organics or an acrylic latex containing dissolved organics, that is, a surfactant-stabilized emulsion of acrylic polymer and dissolved organics in water. This is the latex binder used in formulating latex paint. The concentrations of dissolved organics are dilute, well below their solubility in water. The liquid is pumped at a volumetric flow L from the reservoir, through a membrane module, and returned to the reservoir with a changed solute concentration c′1. The change in the reservoir concentration c1 with time t is given by V

dc1 ¼ Lðc′1  c1 Þ dt

ð1Þ

where V is the volume of solution in the reservoir. In our experiments, V is nearly constant. We find the concentration c1 by writing a steady state mass balance for the concentration change within the module. Because the liquid and gas in the membrane module are nearly well mixed, their concentrations can be approximated by single values. For the small module used here, this is a good approximation; for cases where these concentrations vary within the module, more complex analyses are available elsewhere [10]. A mass balance over the entire module gives p  L ðc′1  c1 Þ ¼ G 1  0 ð2Þ RT where G is the volumetric flow of gas, and p1 is the solute's partial pressure. A mass balance on the liquid alone yields Lðc1  c′1 Þ ¼ K L Aðc′1  cn1 Þ

ð3Þ

where KL is the overall mass transfer coefficient based on a liquid side driving force, A is the membrane area, and cn1 is the hypothetical liquid concentration if the liquid were in equilibrium with the gas. To define cn1 we use the following: cn1 ¼ kH p1

ð4Þ

427

interface. The flux in the gas is given by j1 ¼

kG ðp  p1 Þ RT 1i

ð10Þ

where kG is the individual mass transfer coefficient in the gas and p1i is the partial pressure of the solute in the gas at the gas– membrane interface. This much is the standard for many analyses of mass transfer [11–13]. The less familiar part is the analysis in the membrane itself. Two types of membrane give different results. The first type, which is considered in this work, is a nanoporous hydrophobic film. In this case, the solute diffuses through the gas-filled pores. The flux across the film is j1 ¼

DG ε ðp  p1 Þ lτRT 1i

ð11Þ

where DG is the diffusion coefficient in the gas, ε and τ are the film's porosity and tortuosity, respectively, and l is the film's thickness. On the other hand, the second type of membrane, which is not studied in this paper, is a nonporous film, which may be either rubbery or glassy. In this case, solute dissolves in the nonporous film and diffuses within the polymer itself. The flux across this type of film is j1 ¼

P ðp  p1 Þ lRT 1i

ð12Þ

where P is the membrane permeability, the product of the solute's solubility in the film and its diffusion coefficient in the polymer film. We now can calculate the overall mass transfer coefficient KL as a function of variables like kL, kG, and l. We do so by using Eq. (9)– (12) to eliminate the unknown interfacial concentrations. For the nanoporous film, the result is    ′ 1 j1 ¼ ½K L ðc′1  cn1 Þ ¼ c 1  k H p1 ð1=kL Þ þ ðlτkH RT=DG εÞ þ ðkH RT=kG Þ ð13Þ

where kH is a type of Henry's law constant for the dissolved solute. We now combine Eqs. (1)–(4) to find   dc1 1 ¼  c1 V ð5Þ ð1=K L A Þ þ ð1= L Þ þ ð kH RT=GÞ dt

For the non-porous film, the result is    ′ 1 c1  kH p1 j1 ¼ ½K L  ðc10  cn1 Þ ¼ ð1=kL Þ þ ðlτkH RT=PÞ þ ðkH RT=kG Þ ð14Þ

This is subject to the condition that the initial concentration is known

Only Eq. (13) is tested experimentally in this paper. The analysis, summarized by Eqs. (7) and (13), yields four predictions which can be tested experimentally. First, a plot of the logarithm of the concentration difference should be linear in time, as suggested by Eq. (7). The second prediction is that the reciprocal of the apparent rate constant 1/K should vary linearly with the Henry's law constant kH , as suggested by combining Eqs. (8) and (13) to get   1 1 A lτ 1 A ¼ þ kH RT þ þ þ ð15Þ K kL L kG G DG ε

t ¼ 0;

c1 ¼ c10

Integrating, we find     c1 1 KAt ¼ exp  ¼ exp  Vðð1=K L AÞ þð1=LÞ þ ðkH RT=GÞÞ V c10

ð6Þ

ð7Þ

where K is an overall apparent rate constant 1 1 A kH ART ¼ þ þ K KL L G

ð8Þ

The apparent rate constant K reflects a combination of three different rate processes. If the liquid flow L and the mass transfer product KLA are both large, then K is just a measure of gas flow G. If the gas flow G and KLA are both large, then K is a measure of L. If both L and G are large, then K represents only the overall mass transfer coefficient KL. The overall liquid mass transfer coefficient KL also has important characteristics. It results from three resistances in series: those in the liquid, across the membrane, and in the gas. The flux j1 out of the liquid is given by j1 ¼ kL ðc′1 

c1i Þ

ð9Þ

where kL is the individual mass transfer coefficient in the liquid and c1i is the concentration in the liquid at the liquid–membrane

The intercept on this plot is a measure of both the mass transfer in the liquid and of the liquid flow. We expect the slope on this plot to include effects of the mass transfer through the membrane, of mass transfer in the gas, and of the gas flow. The third and fourth predictions deal with changes in K caused by changes in the gas and liquid flows. Again, from Eqs. (7) and (13) we expect 1 1 A lτkH RT kH RT kH ART ¼ þ þ þ þ K kL L DG ε kG G

ð16Þ

The third prediction is that K varies with G. Because kG varies non-linearly with G, this variation of K will be nonlinear. However, if the mass transfer in the gas is fast, then the term containing 1=kG will be relatively small, and the reciprocal of K will vary with the reciprocal of G. The fourth prediction is similar, but applies to

428

B. Ulrich et al. / Journal of Membrane Science 452 (2014) 426–432

the variation of K with liquid flow L. From boundary layer theory and from many empirical correlations, we expect the individual coefficient in the liquid kL will vary with the square root of liquid flow L. Thus the reciprocal of K should be linear with L  0:5 . We will test all four of these predictions with the experiments described in the following section.

Table 2 Elution Order and Retention Times of VOCs. Isopropyl alcohol (from Fischer Scientific) was used as an internal standard.

3. Experimental 3.1. Materials

Component

Retention time (min)

Methanol Ethanol Acetone Isoproyl alcohol t-Butanol n-Butanol

3.3 5.2 6.5 7.1 0.2 11.4

An aqueous solution of VOCs was used to model the latex paint for some experiments. Model solutions were prepared by weight from distilled water and reagent grade organics. Methanol, ethanol, acetone, and t-butanol were from Sigma Aldrich, (St. Louis, MO); ethyl acetate and n-butanol were from Fischer Scientific (Pittsburgh, PA). Each VOC was present at 1000 ppm in the model solutions. A sample of acrylic latex was obtained from the Dow Chemical Company (Midland, MI). The viscosity of the latex was between 20 and 50 cP at room temperature. The solids' concentration was 50.3%, and the mean particle diameter was between 0.122 and 0.125 μm. The concentrations of VOCs in the latex are shown in Table 1. 3.2. Analysis VOC concentrations were measured using a Trace GC with an FID. Helium was the carrier gas in an Agilent HP-1 capillary column with a 5.0 μm film thickness. The sample loop was 2.0 mL with a 1:10 split. The injector temperature was 150 1C and the FID temperature was 250 1C. The oven temperature was first held at 35 1C for 7 min, then ramped to 220 1C at 30 1C/min, and then held at 220 1C for 1 min. The elution order and retention times are listed in Table 2. The response of the detector for the VOCs was calibrated for concentrations from 10 ppm to 10,000 ppm. Calibration samples were prepared by charging a 2.0 mL GC vial with 500 μL of a standard solution (containing all six VOCs to be analyzed in deionized water) and 2.0 μL isopropyl alcohol. The data in Fig. 1 show that the response for each VOC is linear. Concentrations of the VOCs in the aqueous model solution were measured by 2.0 μL liquid injections. Samples of model solution were prepared by the same method as described for the calibration samples. Concentrations of VOCs in the latex samples were measured by analyzing their headspace gas. A 1 g sample of latex from the reservoir was placed in a 10 mL vial, and the vials were Table 1 Concentrations of VOCs in Latex Sample. Similar VOCs were used in the model solution. The Henry's law constants are at 25 1C. VOCs in latex sample

Conc. (ppm wt)

Henry's law constant (M/atm)

Methanol Ethanol n-Butanol Ethyl acetate Acetone t-Butanol Benzaldehyde Ethyl propionate Butyl acetate Di-butyl ether Butyl acrylate

127 123 4 25 2500 236 6 38

200 180 100 6.7 30 – 6.3 4.6

35 44 5

3.1 0.17 0.46

Fig. 1. GC-FID calibration data for VOCs in model solution. The response is linear over a range of concentrations from 1 ppm to 10,000 ppm.

crimp-sealed with Teflon-lined septa (Chromtech, Apple Valley, MN). After the vials were left to cool and equilibrate at room temperature overnight, 200 mL of the headspace gas was drawn from the vials and injected into the GC.

3.3. Fouling Several membranes were screened for their flux, fouling behavior and solvent resistance. Flat membrane samples were placed in a 20 cm2 filter holder (Millipore Corp., Billerica, MA; Catalog no. XX4404700) altered to have two inlets and two outlets. Dry, room temperature nitrogen flowed through the top compartment of the cell and latex at 50 1C was cycled through the bottom compartment. The total amount of water and organics was condensed in a cold trap in a liquid nitrogen bath. The flux was determined by dividing the mass of permeate collected by the time and membrane area.

3.4. VOC stripping Stripping of both the model solution and the latex was measured with the apparatus shown in Fig. 2. A flat membrane was placed in the altered Milipore filter holder. Water-saturated air at 50 1C was cycled through the bottom compartment and latex or model solution, also at 50 1C, was cycled through the top. The air was saturated in a humidification column (24 in. tall, 1 in. diameter, 0.16 in. Ace Glass ProPac Packing). Heating tape kept the cell at 50 1C and aluminum foil wrapped around the cell reduced heat loss. Experiments were typically conducted over 2 h, and samples were collected from the reservoir every 20–30 min. Samples of the model solution were 0.5 mL and samples of the latex were 1.0 mL. Apparent rate constants were calculated from Eq. (7).

B. Ulrich et al. / Journal of Membrane Science 452 (2014) 426–432

429

Table 3 Membranes Tested for Fouling with a Dry Nitrogen Strip Gas. The most successful membrane was the nanoporous polypropylene membrane. Material

Fig. 2. VOC stripping apparatus. Measurements of the reservoir concentration versus time were used to explore the mass transfer mechanism.

NF90 NF200 Nylon Polypropylene (PP) Cellulose acetate (CA)

Supplier

Dow Dow GE Osmonics W. L. Gore GE Osmonics

(μm)

Overall flux (g/cm2 h)

– – 0.10 0.02 0.20

0.01 0.01 0.15 0.10 –

Pore size

4. Results In this research we seek to remove the odor from latex binder used in formulated latex paint by mass transfer across a membrane. The odor, caused by low molecular weight organics, is difficult to remove by conventional air stripping because the latex foams easily. However, to be successful: 1. Membrane fouling must be minimal. 2. The rate of VOC removal must be fast compared to that for air stripping. 3. The main resistances for mass transfer must be identified. If we can measure these properties, we can make estimates of the membrane areas needed to process commercial quantities of latex. 4.1. Fouling Membranes tested for fouling with dry nitrogen at 25 1C, as well as the overall flux for each membrane, are reported in Table 3. The cellulose acetate membrane crumbled because it was not resistant to the latex, and the overall fluxes from the Dow membranes were approximately an order of magnitude lower than the other membranes. The hydrophobic polypropylene and hydrophilic nylon were chosen for further fouling studies. The nylon membrane fouled significantly, showing a 73% decrease in flux after 20 h of continuous operation. The flux was restored to 92% of the original value after washing with a cleaning solution with base (sodium hydroxide) and detergent (sodium dodecyl sulfate, SDS). It was not necessary to scrub or peel any latex from the membrane. Thus, while the nylon membrane fouled rapidly, it could be cleaned easily. As Fig. 3 shows, the polypropylene membrane fouled much more slowly than the nylon membrane, with a decrease in flux of only 5% observed after 20 h of continuous operation. Though this membrane was cleaned before it was shut down and left overnight, a film deposited on the membrane, reducing the flux by 69% in subsequent runs. After peeling the film of latex from the membrane, the flux was restored to 91% of the original value. Film formation can be prevented by keeping both compartments of the module moist when the system is shut down. That the hydrophobic polypropylene membrane fouled more slowly than the hydrophilic nylon membrane is the opposite of many membrane experiments which ultrafilter biological macromolecules. There, hydrophobic membranes have more fouling and the hydrophilic ones remain almost pristine [14]. Perhaps the surfactant stabilizes the particles in the latex, causing less fouling on the hydrophobic membrane. We must stress that these results showing fouling are for VOC and water removal with dry nitrogen. For experiments removing VOCs with water-saturated air at 50 1C, no significant fouling of the polypropylene membrane was observed. While our data are

Fig. 3. Fouling of the polypropylene membrane. This membrane fouls more slowly, but is more difficult to clean.

limited, we expect no major fouling with the hydrophilic membrane either if the stripping gas is saturated with water or steam is used at the appropriate pressure to avoid excessive temperatures. 4.2. Stripping of VOCs We turn next to stripping experiments. The natural log of the concentration of the reservoir ln c1 =c10 does change linearly with time, as exemplified by the data in Fig. 4 for acetone both in the model solution and in the latex paint. The slopes from plots like this can be used to calculate apparent rate constants K. The rate constant for acetone mass transfer from model solution is 5.1 70.7  10  4 cm/s, 16% faster than transfer from latex, which is 4.3 7 1.1  10  4 cm/s. However, since the error on the rate constant is over 30%, there is little difference in the speeds of mass transfer under the conditions chosen. The value for mass transfer from the latex is a pleasant surprise. The latex is more than 20 times more viscous than the model solutions, so we would expect solution diffusion and hence mass transfer to be slower. This is not the case, perhaps because the latex is somehow facilitating solute transport. 4.3. Mechanisms To understand the characteristics of VOC removal, we must examine the mechanism for solute stripping in more detail. We can do this by considering how the rate varies with the Henry's Law constant kH , the liquid flow rate L, and the gas flow rate G. Because in our experiments we vary kH more than L and G, its variation is more instructive. Before we begin, we consider the Henry's law constant of a latex in more detail, and in particular how mass transfer is related to kH . The total amount of a volatile solvent which can dissolve in an aqueous latex suspension depends on the partitioning of the solute between the water and the latex particles. Thus the total amount of a hydrophobic solute

430

B. Ulrich et al. / Journal of Membrane Science 452 (2014) 426–432

for our experiments: L 1:7 cm3 =s ¼ ¼ 0:085 cm=s A 20 cm2

ð18Þ

Thus the mass transfer resistance in the liquid is more significant than the resistance from the liquid flow rate, consistent with what normally happens in gas stripping [11–13]. Our estimate of the intercept is 1 A 1 1 ¼ 1100 s=cm þ ¼ þ kL L 10  3 cm=s 0:085 cm=s

Fig. 4. Acetone removal from model solution and from latex. In both experiments the liquid flow L was 1.7 cm3/s and the gas flow G was 82 cm3/s. The rates are equal within experimental error.

ð19Þ

which is close to our experimental value of 1300 s=cm: We next turn to the slope in Fig. 5, which is about 1 s/cm. According to Eq. (15), this slope should contain the terms ½ð1=kG Þ þ ðlτ=DG εÞ þðA=GÞ. The first two terms represent the mass transfer resistances in the water-saturated stripping gas and in the gas-filled nanopores of the membrane, respectively. The third term is the effect of the gas flow removing any evaporated organics from the module. As for our estimate for intercept, we can use the parameters of our experiments to estimate these quantities. Neglecting for the moment any correction for different solute volatilities, we can estimate kG with the following equation [17]: kG ¼

DG 0:1 cm2 =s ¼ 1 cm=s ¼ 0:1 cm lG

ð20Þ

where DG and lG are the diffusion coefficient for a solute in the water-saturated stripping gas and the thickness of the gas boundary layer, respectively. Because the membrane has 30% porosity, is 300 mm thick, and has pores with a tortuosity of about 3, the resistance to mass transfer across the membrane is about DG ε 0:1 cm2 =s ð0:3Þ ¼ 0:03 cm=s ¼ 0:3 cm ð3Þ lτ

ð21Þ

The third term is Fig. 5. VOC stripping from model solution versus Henry's law constant. This plot is consistent with Eq. (15).

can be much more in an aqueous latex than in water alone. However, because the latex particles have a large surface area per volume, we expect that they will quickly reach equilibrium with water. Thus we will correlate our rates both in an aqueous solution and in aqueous latex using the water-to-vapor Henry's law constant. We thus avoid the more complex effective volatility used in solubility studies [15,16]. After we measure K from plots like those in Fig. 4, we can plot 1/K vs. 1=kH as shown in Fig. 5. These results strongly support the theory leading to Eq. (15). Moreover, the magnitude of the intercept and the slope on this plot are consistent with our estimates of the magnitudes of these quantities. More specifically, the intercept should equal [(1/kL) þ(A/L)], or the sum of the resistances from mass transfer in the liquid phase and from the liquid flow rate, respectively. The first term will limit VOC removal if the liquid diffusion coefficient is small, while the second term will be limiting if the liquid flow rate is small. The slope reflects changes in volatility, both due to the membrane and to the mass transfer in the gas phase. We can make estimates of the resistances represented by the intercept in Fig. 5. The liquid mass transfer coefficient kL can be found from the film theory of mass transfer: kL ¼

DL 10  5 cm2 =s ¼ 10  3 cm=s ¼ 0:01 cm lL

ð17Þ

where DL is the diffusion coefficient of a typical solute in the liquid, approximately 10  5 cm2/s, and lL is the liquid boundary layer thickness, typically 0.01 cm [17]. Next we calculate the value of L/A

G 20 cm3 =s ¼ ¼ 0:24 cm=s A 82 cm3

ð22Þ

Adding our results from Eqs. (20)–(22) together we get our estimate for the slope: 1 lτ A 1 1 1 ¼ þ þ ¼ 40 cm=s þ þ kG G 1 cm=s 0:03 cm=s 0:24 cm=s DG ε ð23Þ This rough estimate is the same order of magnitude as our measured value of about 80 s/cm. The success of Fig. 5 may seem surprising because the data on which it is based include different VOCs which have different diffusion coefficients in water and in gas. We should expect that different solutes lead to different values of the mass transfer kL and kG. Further study is needed to better understand the extent to which differences in hydrophilic/hydrophobic character affect mass transfer rates. For example, highly hydrophobic solutes are likely to be strongly associated with latex polymer particles, and may prove somewhat slower to strip from the latex compared to highly hydrophilic solutes that tend to reside in the continuous phase. Similarly, we could expect for the various solutions different partition coefficients and different boundary layer thicknesses around latex particles. The fact that these apparently do not occur affirms our use of a single air–water partition coefficient, as discussed above. Thus in our experiments, removal of VOCs from latex is largely controlled by two resistances. The first expected resistance is that due to solute diffusion in the liquid, represented by 1/kL. The second, unexpected one is mostly due to water-saturated gas in the membrane's pores. To further evaluate the mass transfer mechanism, we examine the effects of varied gas and liquid flow

B. Ulrich et al. / Journal of Membrane Science 452 (2014) 426–432

431

Fig. 6. Mass transfer of acetone versus gas and liquid flow. The data are consistent with the expectation that 1/K varies linearly with G  1 and L  1/2.

rates on VOC removal. As discussed above, we expect plots of 1/K versus G  1 and L  1/2 to be linear. To test this, stripping experiments were conducted with the model solution over a small range of flow rates for the liquid (0.8–1.7 cm3/s) and the stripping gas (80–160 cm3/s). The results from these experiments, shown in Fig. 6, are consistent with the dependence of these variations of K with G and L. However, due to the limited accuracy of the data and the relatively small ranges of G and L, this prediction is not proven.

5. Discussion The results above lead to three important conclusions. First, membrane-assisted stripping avoids latex foaming. This is simply not a problem. The second conclusion is that membrane fouling is minor if the VOCs are stripped with water-saturated gas (or by extension, with steam). This conclusion is more restricted than that on foaming, because we have never continuously run our experiments for more than a few days. To be successful commercially, we believe that any membranes must be operated for about 3 years. Achieving this length of service life may require periodic cleaning, an aspect which we have not considered carefully. This point merits further study. The third conclusion is that the mass transfer is about what is observed in more conventional separation processes. In particular, a nanoporous hydrophobic membrane offers little additional mass transfer resistance. This conclusion is that expected from a wide variety of experiments with many other membrane contactors, including those suggested for gas absorption, liquid–liquid extraction, and differential distillation [18–22]. It is important to have this affirmed experimentally for stripping of VOCs from latex, which is what is done in this paper. However, this positive conclusion should be tempered by other experience using microporous polypropylene membranes. While these membranes often give successful rates for the first few days, the membrane's pores can then wet and the membrane's mass transfer rates can decrease. These slower rates often involve solvents or detergents which facilitate pore wetting. We saw no evidence of these effects here, but we did not look at the longer operating times where they would be expected to occur. If these slower rates do occur, they can sometimes be managed by filling the pores with a hydro-gel like polyvinylalcohol. We can use the experimental results to estimate the membrane area required for latex processing. To make this more specific, imagine that we wish to remove 90% of the VOC content from ten tonnes of latex per day, or a flow of 120 cm3 latex/s. We can imagine carrying out this stripping either across a membrane connected to a stirred tank or in a hollow fiber membrane module. For the stirred tank, the membrane would be flat sheets, which give a small membrane area per latex volume but can be easily

disassembled for cleaning. From a balance on the VOCs in the stirred tank, we obtain c1 1 ¼ 1 þ ðKA=LÞ c10

ð24Þ

c1 1=ð5  10  4 cm=sÞA ¼ 0:1 ¼ c10 120 cm3 =s

ð25Þ

A ¼ 200 m2

ð26Þ

where c1/c10 is the fraction of VOC content remaining, and K is the overall rate constant determined in this work. For a hollow fiber module, the area needed is smaller [16]. A mass balance on the VOCs in fibers surrounded by a high gas flow gives c1 ¼ e  ðK L A=Q Þ c10 4 c1 ¼ 0:1 ¼ e  fð1010 c10

A ¼ 30 m2

ð27Þ cm=sÞA=120 cm3 =sg

ð28Þ ð29Þ

where c1/c10 is again the fraction of VOC content remaining and KL is the overall mass transfer coefficient inferred at high gas flow from this work. This assumes that the latex is flowing within the hollow fibers. Other studies of different module designs suggest that KL can be increased by about five times by having the latex flow outside of and across a bed of hollow fibers [23]. This suggests a strategy for decreasing the membrane area still more, though the hollow fibers will be much more difficult to clean than flat membranes. On this basis we believe this membrane-assisted removal of trace VOCs from latex used for paint has considerable potential. Acknowledgments The authors are indebted to W.A. Arnold, M.L. Trippeer, and Grant A. Von Wald for helpful discussions, and to Andrew Wagner for help with preliminary experiments. This work was supported by the Dow Chemical Company.

Nomenclature A c1 c01

membrane area volatile solute concentration in reservoir, moles per volume volatile solute concentration in module, moles per volume

432

cn1 DG ; DL G kG kH kL K KL L l P p1 R T t V

ε τ

B. Ulrich et al. / Journal of Membrane Science 452 (2014) 426–432

hypothetical volatile solute concentration in liquid in equilibrium with gas diffusion coefficients in gas and liquid respectively gas flow, volume per time individual gas mass transfer coefficient, length per time a Henry's law coefficient, moles per volume per pressure individual liquid mass transfer coefficient, length per time apparent rate constant, length per time overall mass transfer coefficient, length per time liquid flow, volume per time membrane thickness permeability partial pressure of volatile solute gas constant temperature time reservoir volume void fraction in membrane tortuosity in membrane

References [1] D. Stoye, B. Marwald, W. Plehn, Paints and coatings. 1. Introduction, Ullmann's Encyclopedia of Industrial Chemistry, Wiley, 2010. [2] P. Wolkoff, How to measure and evaluate volatile organic compound emissions from building products. A perspective, Sci. Total Environ. 227 (1999) 197–213. [3] J.L. Fleming, Volatilization Technologies for Removing Organics from Water, Noyen data, Park Ridge, NJ, 1990. [4] Y.L. Hwang, G.E. Keller, J.D. Olson, Steam stripping for removal of organic pollutants from water. 1. Stripping effectiveness and stripper design, Ind. Eng. Chem. Res. 31 (1992) 1753–1759. [5] Y.L. Hwang, J.D. Olson, G.E. Keller, Steam stripping for removal of organic pollutants from water. 2. Vapor liquid equilibrium data, Ind. Eng. Chem. Res. 31 (1992) 1760–1768.

[6] R.E. Hinchee, Air Sparging for Site Remediation, CRC Press, Boca Raton, FL, 1994. [7] J.R. Ortiz-Del Castillo, G. Guerrero-Medina, J. Lopez-Toledo, J.A. Rocha, Design of steam-stripping columns for removal of volatile organic compounds from water using random and structured packings, Ind. Eng. Chem. Res. 39 (2000) 731–739. [8] J.G. Crespo, K.W. Boddeker, Membrane Processes in Separation and Purification, Springer Verlag, Berlin, 2010. [9] I. Abou-Nemeh, S. Majumdar, A. Saraf, K.K. Sirkar, L.M. Vane, F.R. Alvarez, L. Hitchens, Demonstration of a pilot-scale pervaporation systems for volatile organic compound removal from a surfactant enhanced aquifer remediation fluid, Environ. Prog. 20 (2001) 64–73. [10] L. Dahuron, E.L. Cussler, Protein extractions with hollow fibres, AIChE J. 34 (1968) 130–136. [11] C.J. Geankoplis, Transport Processes and Separation Process Principles, 3rd ed., Prentice Hall, Upper Saddle River, NJ, 2003. [12] W. McCabe, J.C. Smith, P. Harriot, Unit Operations of Chemical Engienering, 7th ed., McGraw-Hill, New York, NY, 2005. [13] J.D. Seader, E.J. Henley, Separation Process Principles, 3rd ed., Wiley, Hoboken, NJ, 2010. [14] M.A. Shannon, P.W. Bohn, M. Elimelech, J.G. Georgiadis, J.B. Marinas, A.M. Mayes, Science and Technology for water purification in the coming decades, Nature 452 (2008) 301–310. [15] T.C.G. Kibbey, K.D. Pennell, K.F. Hayes, Application of sieve-tray air strippers to the treatment of surfaction-containing wastewaters, AIChE J. 47 (2001) 1461–1470. [16] C. Zhang, G. Zheng, C.M. Nichols, Micellar partitioning and its effects on Henry's law constants of chlorinated solvents in anionic and nonionic surfactant solutions, Environ. Sci. Technol. 40 (2006) 20–214. [17] E.L. Cussler, Diffusion, 3rd ed., Cambridge University Press, New York, NY, 2009. [18] Q. Zhang, E.L. Cussler, Microporous hollow fibers for gas absorpion, J. Membr. Sci. 23 (1985) 321–345. [19] R. Prasad, K.K. Sirkar, Dispersion-free solvent extraction with microporous hollow fibers modules, AIChE J. 34 (1988) 177–188. [20] D. Yang, R.S. Barbero, D.J. Delvin, E.L. Cussler, C.W. Colling, M.E. Carrera, Hollow fibers as structured packing for olefin/paraffin separation, J. Membr. Sci. 279 (2006) 61–69. [21] D. Yang, J.D. Devlin, R.S. Barbero, R. Wright, E.L. Cussler, Hollow fibers as structured packing for olefin/paraffin distillation, in: Proceedings of the World Congress of Chemical Engineering, Montreal, Canada, August 2009. [22] E. Bringas, M.F. San Roman, J.A. Irabien, I. Ortiz, Overview of the mathematical modeling of liquid membrane separation processes in hollow fiber contactors, J. Chem. Technol. Biotechnol. 4 (2009) 153–1614. [23] S.R. Wickramasinghe, J.J. Semmens, E.L. Cussler, Mass transfer in various hollow fiber geometries, J. Membr. Sci. 69 (1992) 235–250.

Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.