Supported-gas-membrane process for removal and recovery of aliphatic amines from aqueous streams

Chemical Engineering Science 141 (2016) 330–341

Contents lists available at ScienceDirect

Chemical Engineering Science journal homepage: www.elsevier.com/locate/ces

Supported-gas-membrane process for removal and recovery of aliphatic amines from aqueous streams Jie He a, Huaqun Liu a, Pengfei Shan a, Kunming Zhang a, Yingjie Qin a,b,n, Liqiang Liu b a b

School of Chemical Engineering & Technology, Tianjin University, Tianjin 300072, China Chembrane Research & Engineering, Inc., Bridgewater, NJ 08807, USA

H I G H L I G H T S








Supported-gas-membrane process firstly to remove and recover aliphatic amines from aqueous solution. A mathematical model was established and solved numerically. Overall mass transfer coefficient was a strong function of polarity of amine. Amine could be enriched for 420 times while 495% of amine was recovered. A long-term operation stability was performed and tested well for 430 days.

art ic l e i nf o

a b s t r a c t

Article history: Received 15 July 2015 Received in revised form 29 October 2015 Accepted 21 November 2015 Available online 28 November 2015

The removal and enrichment of aliphatic amines with low molecular weights from their individual aqueous solutions was investigated theoretically and experimentally via a hollow-fiber supported-gasmembrane (SGM) process. Aqueous solutions containing 200–5000 mg L  1 amine were tested as feed, and an aqueous solution of 10 wt% sulfuric acid was used as an absorbing solution. Amines such as methylamine, dimethylamine, trimethylamine, ethylamine, diethylamine and triethylamine, as well as ammonia, were tested in the SGM process. The experimental data demonstrated that the overall mass transfer coefficient was a strong function of the polarity of the amine or amine/water volatility and was not a simple function of the molecular weight or boiling point of amines; the order was determined to be trimethylamine4 ammonia 4triethylamine 4 diethylamine 4ethylamine 4 dimethylamine 4methylamine. The influences on mass transfer coefficients of the feed-in temperature, feed-in concentration, feed flow rate, and the concentration of NaOH pre-added to the feed were also investigated. Among these operating factors, the feed-in temperature and NaOH concentration were crucial; increasing the feed-in temperature and the NaOH concentration led to a significant increase in the mass transfer coefficient, especially when the amine concentration in the feed was low. Mathematical models incorporating laminar flow, ion and molecular diffusion, dissociation equilibrium and vapor–liquid equilibrium were established and solved numerically. When the surface tension of the feed solution was 4 45 mN m  1, the SGM process demonstrated good stability during a test period of at least 30 days. Greater than 95% of the amine was recovered, and the amine could be enriched by 420 times in the absorbing solution. Thus, this SGM-based separation process is suitable to remove, recover, and concentrate amines from their aqueous solutions. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Amines Removal Supported gas membrane Mass transfer coefficient Ion diffusion Dissociation

1. Introduction Aliphatic amines of low molecular weight are widely used in a variety of industries (Meng et al., 2010; Chang et al, 2010; Wang et

n Corresponding author at: School of Chemical Engineering and Technology, Tianjin University, Tianjin 300072, China. Tel./fax: þ 86 22 27890430. E-mail address: [email protected] (Y. Qin).

http://dx.doi.org/10.1016/j.ces.2015.11.020 0009-2509/& 2015 Elsevier Ltd. All rights reserved.

al., 2011). For instance, methylamine is mainly used in the production of medicines (hormone or caffeine), agricultural chemicals (carbaryl, rogor, and so on) and explosives; dimethylamine is used for synthesis of dimethylformamide, agricultural chemicals, and medicines and is used in tanneries; trimethylamine is used in the production of feed additives, weedicides, and ion exchange resins; ethylamine is used in synthesizing dyes, extraction, emulsification, rubber vulcanizing, medicine production and petroleum refining; diethylamine is used in organic synthesis and as a curing agent for

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

epoxy; triethylamine is mainly used as an organic solvent, inhibitor, catalyst, and preservative, and it is also used for synthetizing dyes. Amine-containing wastewater streams are produced during the operation of amine-related processes; for instance, dimethylamine-containing wastewater is produced from the hydrolysis of dimethylformamide in the artificial leather industry (Chang et al, 2010). With the extensive application of amines and amine derivatives, the environmental impact of amine emission cannot be ignored, especially when increasing amounts of aminecontaining wastewater from chemical plants, refineries and pharmaceutical plants are discharged into surface water and underground water in China. Aliphatic amines of low molecular weight are volatile, more basic than ammonia, more toxic than ammonia, and less biodegradable than ammonia. Several conventional methods (Meng et al., 2010; Zhang et al., 2010; Wang et al., 2011; Chen et al., 2008; Chang et al, 2010; Qu and Chen, 2009) such as ion exchange, chemical precipitation, oxidation, biodegradation, distillation and air stripping have been applied to the treatment of aminecontaining wastewaters. Among these methods, air or steam stripping is the most widely used in the chemical industry; however, it still has problems such as high cost of equipment, high power or steam consumption, and secondary pollution. The supported-gas-membrane (SGM) separation process may provide an alternative technology for amine removal and recovery. The SGM process (or trans-membrane stripping/absorption process) is a technology that combines membrane separation with a conventional absorption and stripping (desorption) process, in which a hydrophobic microporous membrane is placed between a feed, which is an aqueous solution containing a volatile species, and an absorbing solution, which is an aqueous solution usually containing a reactive absorbing agent (Qin and Cabral, 1997; Qin et al., 1996). In such a membrane-based separation process, volatile species evaporate from the aqueous feed at the interface between the feed and the filled gas within the micropores in the membrane wall, diffuse through the gas-filled micropores, and then dissolve into the absorbing solution at the interface between the gas-filled micropores and the absorbing solution. The driving force of this membrane process lies in the difference of the partial pressure of the volatile species over the feed and that over the absorbing solution (Yang and Cussler, 1986). However, the partial pressure of the volatile species over the absorbing solution is usually thought to be zero when the volatile species is acidic and the absorbing agent is basic, the volatile species is basic and the absorbing agent is acidic, or the volatile species is an oxidant and the absorbing agent is a reductant. The membrane used in the SGM process is commonly in the form of microporous hydrophobic hollow fibers, and such types of hollow fibers are usually made from polytetrafluoroethylene, polypropylene (PP) and polyvinylidene fluoride. When a hollow-fiber membrane module—which is usually called a membrane contactor—is used, the feed solution usually flows through the lumen of the hollow fibers while the reactive absorption solution flows though the shell side of the membrane modules; such an operation mode usually leads to a higher overall mass transfer coefficient (Qin et al., 1990; Qin and Cabral, 1997). Currently, research on the applications of membrane-based absorption, membrane-based stripping or the SGM process is quite extensively studied for the removal of volatile and reactive species such as O2, CO2, NH3, H2S, SO2, HCN, NO, Cl2, Br2, and I2 from gaseous or aqueous feed streams to another aqueous solution (Yang and Cussler, 1986; Imai et al., 1982; Yang and Cussler, 1986; Kenfield et al., 1988; Semmens et al., 1989, 1990; Qin et al., 1990, 1996; Qin and Cabral, 1997; Wang et al., 1993; Mackenzie and King, 1985; Titmas and Fluto, 1993). Some of those processes have been commercialized, such as the stripping of O2 or CO2 from ultrapure water. However, until now, there has been no report on

331

the application of membrane-based absorption, membrane-based air stripping or the SGM process to remove amine from a gaseous stream or an aqueous solution. Compared to the above-mentioned conventional methods for removing amine from aqueous solutions, the SGM process using modules made from a porous hydrophobic membrane has several prominent features and benefits (Meng et al., 2010; Zhang et al., 2010; Wang et al., 2011; Chen et al., 2008; Chang et al., 2010; Qu and Chen, 2009). First, the stripping of amine from the aqueous solution and the subsequent absorption of amine in another aqueous solution, which is usually an acidic absorbent solution, occur simultaneously in a singlemembrane contactor by replacing both a stripping (desorption) column and an absorption column; thus, this membrane-based process provides a maximum driving force for amine removal (Meng et al., 2010; Chen et al., 2008). Second, the feed and stripping streams are located in each side of the microporous hydrophobic membrane and cannot influence each other; thus, the flexibility is greatly improved relative to the conventional column (Chang et al., 2010; Qu and Chen, 2009). Third, the membrane module has a high packing density, and the hollow-fiber membrane has a small diameter; therefore, the membrane contactor can supply a much higher surface area per volume compared to conventional columns (Wang et al., 2011; Chen et al., 2008). Fourth, compared to conventional separation contactors, a commercialized membrane contactor with excellent shell-side structure design encounters no back-mixing, wall flow, by pass flow, dead angle, flooding, foaming entrainment, and so on; thus, the overall transfer coefficient of this membrane-based separation process is much higher (Zhang et al., 2010; Wang et al., 2011). Finally, compared to air stripping or steam stripping, the operational cost of the SGM process is much lower because it almost eliminates the consumption of steam and consumes much less electricity to drive the wastewater and absorbing solution to pass through the membrane module (Meng et al., 2010; Chen et al., 2008). In this study, the SGM process for removing, recovering, purifying and concentrating various amines from their respective aqueous solutions was investigated experimentally and theoretically. The effects of operation parameters such as the feed-in temperature, feed-in concentration, and feed-in flow rate as well as membrane properties on the performance of ammine removal were examined and discussed in detail. The mass transfer coefficient, removal/recovery percentage, and enrichment ratio were used to characterize the efficiency of the SGM-based separation process for amine removal. The long-term operational stability of the SGM process was tested for dimethylamine. This work is a significant contribution to the treatment of wastewater streams containing amines.

2. Experimental Ammonia, methylamine, dimethylamine, trimethylamine, ethylamine, diethylamine, trimethylamine, hydrochloric acid (HCl), and sulfuric acid (H2SO4) were all purchased from Jiangtian Chemical Corporation (Tianjin, China). All chemicals used in the experiment were of reagent grade and were employed without further purification. Aqueous amine solutions were prepared by adding a certain quantity of amine to deionized water. The exact concentration of amine was determined by titration with a HCl solution using a titrator (DL28, Mettler Toledo, Columbus, OH, USA) with an error of less than 0.5%. The surface tension of the aqueous feed solution containing different amines at various concentrations was measured by using surface tensiometer (NDJ-8S, Fangrui Instrument, Inc., Shanghai, China). The feed liquid entry pressure (LEP) of the polypropylene (PP)

332

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

Table 1 Characteristics of the hollow fibers and modules. Module no.

Membrane supplier

Ri (mm)

Ro (mm)

Average pore size (μm)

ε

ZM (m)

Shell inside diameter (mm)

n

Fiber packing density (%)

S (m2)

1a 1b 1c 1d 2

Chembrane Chembrane Chembrane Chembrane Membrana

0.19 0.19 0.19 0.19 0.14

0.24 0.24 0.24 0.24 0.19

0.2 0.2 0.2 0.2 0.2

0.40 0.40 0.40 0.40 0.50

0.31 0.31 0.31 0.31 0.31

40 40 40 40 20

1670 1670 1670 1670 684

23.8 23.8 23.8 23.8 24.5

0.618 0.618 0.618 0.618 0.181

Fig. 1. Schematic diagram of experimental apparatus.

porous hydrophobic hollow fiber membrane was measured by using pressure test (Qin et al., 1996). The PP porous hollow fiber membranes used in this study were provided by Chembrane Engineering & Technology, Inc., Tianjin, China, and Membrana GmbH, Wuppertal, Germany. The hollowfiber modules used in the experiment were fabricated by placing the desired number of hollow fibers into the PP shell of ID ¼40 mm or ID ¼20 mm and sealing with epoxy resin on both ends of the shell. The detailed specifications of the hollow fibers and modules are listed in Table 1, where Ri and Ro are the fiber lumen radius and the fiber outside radius, respectively; ε is the porosity of the porous fiber membrane wall; ZM is the effective length of the hollow fibers; n is the number of hollow fibers in a module; and S is the effective surface area based on the lumenside radius of the hollow fibers. The integrity of each new module was tested by filling the module shell side with deionized water and pressuring the shell side with compressed air at 2.5 atm (gauge). This test was performed for at last 8 h, and only the modules that had no leakage would be further used in the experiments of amine removal. The integrity of the used modules was further tested by filling the module shell side with deionized water and pressuring the shell side with compressed air at a gauge pressure of 0.3–1.0 atm (Qin et al., 1996). The set up used in the experiment is shown in Fig. 1, in which T1, T2, T3 and T4 indicate thermometers at the inlet and outlet of the shell side and lumen side of the membrane module, respectively; R1 indicates the reservoir of absorbing solution; R2 and R3 indicate the feed-in and feed-out reservoirs, respectively; B1 and B2 indicate the peristaltic pumps for transporting feed and absorbing solution, respectively; and E1 and E2 indicate the heat exchangers to control the temperature of the feed and absorbing solutions, respectively (Qin and Cabral, 1997; Qin et al., 1996). An aqueous solution (50 L) with a certain amine concentration was supplied as feed. An aqueous solution (2.5 L or 5 L) containing sulfuric acid with a concentration of 10 wt% H2SO4 was used as the absorbing solution. Each solution was contained in a glass vessel with a stirring bar. The temperatures of the feed and absorbing solution at the inlets of the membrane module were maintained

by external heat exchangers, and measured by two Pt-100 sensors with a sensitivity of 70.1 °C. The experiments were conducted in semi-batch mode in most cases, in which the feed solution was passed through the lumen side of the membrane module in one pass, and the absorbing solution was circled between the shell side of the membrane module and the reservoir (Wang et al., 1993). The volumetric flow rate of the feed and absorbing solution was controlled at 15– 45 L h  1 and 50 L h  1, respectively. Once the experiment was started, sample quantities of the feed-in solution and feed-out solution were, respectively, taken from the feed-in reservoir and feed-out reservoir at certain intervals and were analyzed by titrating with HCl solution of a certain concentration using the titrator. Each peristaltic pump used in the study was calibrated before each experiment. Unless a high-enrichment test was performed, the absorbing solution was usually replaced when approximately 30 wt% of the initial concentration of H2SO4 was consumed. To ensure the accuracy of the experimental results, each experiment except the long-term one was repeated more than twice under the same conditions, and their average values were used for the final data demonstrated in this paper. When experiments with a high amine removal/recovery percentage were performed, several identical modules in series were used, and the feed was flowed through the lumen side in one pass while the absorbing solution was circled between the shell side of the membrane module and the reservoir (Qin and Cabral, 1998). When experiments with a high enrichment of amine in the absorbing solution were performed, they still operated with six modules in series, in which the feed solution was passed through the lumen side of the membrane modules in one pass, and the absorbing solution was circled between the shell side of the membrane modules and the reservoir. However, the treated feed in reservoir R3 was reused by restoring the amine concentration near the original value by adding concentrated amine solution to the used feed and returning it to reservoir R2. The highenrichment test was stopped when the pH of the absorbing solution in reservoir R1 increased to a value of 3–5. The concentration of amine in the absorbing solution was measured by the following procedures. The pH value of the absorbing solution was adjusted to approximately 13.6 where only o0.1% of amine was in the dissociated form. Then, the solution was distillated in a glass distillation column with a certain reflux ratio, and nearly 40% of the solution was collected as the distillate. When the reflux ratio was sufficiently high, almost all amine was recovered in the distillate. The distillate was further titrated with HCl solution using the titrator, and then the amine concentration in the original absorbing solution was calculated by mass balance. To ensure that the pH of the feed solution maintained a sufficiently high value to minimize the dissociation of amine, a concentrated solution of NaOH was added to the deionized water in order to maintain a NaOH concentration of 1-10 mM; this dilute basic solution was then used as a solvent to prepare the amine solution. The concentration of amine in the solution also could be titrated with the titrator. Only the concentration of NaOH in the solution must be known beforehand. The amount of NaOH was excluded from total amount of base (Wang et al., 1993).

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

A long-term operational stability test was performed by an intermittent mode; namely, feed solution was also recycled between the module and only one feed reservoir (Qin and Cabral, 1997). The test was continuously run for at least 8 h per day under a low flow rate. Amine was added to the feed solution when nearly 90% of the amine was removed from the feed reservoir and was exchanged with the absorbing solution when nearly 90% of the H2SO4 was consumed or the pH of the absorbing solution was 43. This test was performed for at least one month and then was stopped arbitrarily. For such a long-term test, the leakage problem was also detected occasionally during the test as follows. When amine in the feed was nearly completely removed, a fresh aqueous solution of 10% H2SO4 was used as the absorbing solution, and the SGM process continued to run while maintaining a lumen-side pressure slightly lower than the shell-side pressure; then, the pH value of the feed flowing out from the membrane module was measured. When such an operation was maintained for at least 3 h and no value of pH o5 in the feed was observed, the test ensured that no leakage of the absorbing solution into the feed occurred within the module during regular operation of the SGM process.

3. Theory model 3.1. Dissociation equilibrium and vapor–liquid equilibrium In an aqueous amine solution, dissociation equilibrium can be established as A þH2 O 3 AH þ þ OH  C CA

0 C  CA

0

ð1Þ

at initial

C  CA

333

methylamine, 45 mg L  1 ethylamine or dimethylamine, and so on), the extent of amine dissociation is large; even if the initial amine concentration is 10 mM, the extent of amine dissociation is not negligible. Taking dimethylamine and diethylamine as examples, the extent of dissociation is calculated and listed in Table 3. When an aqueous solution contains both amine and NaOH, dissociation equilibrium of amine can be described by A þ H2 O3 AH þ þ OH  C CA

0

C NAOH

C C A

ð4Þ

at initial

C NAOH þ C  C A

at equilibrium

ðC C A ÞðC NaOH þ C  C A Þ=C A ¼ K b or, CA ¼ C 

ð5Þ

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðC NaOH þK b Þ2 þ 4K b C ðC NaOH þ K b Þ

ð6Þ

2

Then, from Eq. (5), it is known that when the concentration of NaOH, CNaOH, is sufficient high, the extent of dissociation, (C  CA)/ C, is very small. The extent of dissociation of dimethylamine or diethylamine when NaOH exists in the same aqueous solution is also given in Table 3. When the dilute amine aqueous solution achieved vapor–liquid equilibrium and the total pressure was lower than 5 atm, the dilute amine aqueous solution satisfied Henry's law. Henry's constant H of ammonia and amines at 25 °C (Wang et al., 1993; Yaffe et al., 2003; Altschuh et al., 1999; Nirmalakhandan et al., 1997) are expressed in Table 3. Great attention must be paid to ensure that the vapor–liquid equilibrium is consistent with Henry's law only when amines are in the free state.

atequilibrium 3.2. Diffusion of molecular and ions in feed solution

Then, a quadratic equation of one variable can be gained as ðC  C A Þ2 =C A ¼ K b

ð2Þ

Therefore, concentration CA of amine in the form of free molecules can be calculated from the following equation: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K 2b þ 4K b C  K b CA ¼ C  ð3Þ 2 where C is total concentration of amine in the aqueous solution and Kb is its dissociation equilibrium constant. Then, from Eq. (3), it can be seen that when the dissociation equilibrium constant is sufficiently low or the total concentration C is sufficiently high, the extent of dissociation, (CCA)/C, is very small; conversely, when the dissociation equilibrium constant is moderately high or the total concentration C is very low, the extent of dissociation, (C CA)/C, becomes very large. According to the strength of the alkalinity indicated by the Kb value listed in Table 2, amines and ammonia can be ordered as diethylamine4triethylamine4ethylamine4dimethylamine4methylamine4trimethylamine4ammonia. When the total concentration of amine in the aqueous solution is very low (e.g., 1 mM amine, which is equal to an amount of 17 mg L  1 ammonia, 31 mg L  1

Diffusion of amine in the form of free molecules in its dilute aqueous solution is consistent with Fick's law. However, the diffusion of charged ions such as hydroxide ions (OH  ), hydrogen ions (H þ ), and ammonium ions (AH þ ) formed from the dissociation of amine in aqueous solution is not consistent with Fick's law but can be described by the Nernst–Planck equation (Cussler, 1984; Qin and Cabral, 1998). When the pH of the solution is high and the concentration of H þ is too small—and thus its influence on the diffusion of OH  and AH þ can be neglected—the pair of ion AH þ and ion OH  has an equal diffusion coefficient, and the diffusion coefficient of the ion pair OH  -AH þ can be expressed as Table 3 Extent of dissociation of amines at various concentration in aqueous solution. C, 10 mM, C, 1 mM C, 10 mM C, 100 mM C, 1 mM, with 10 mM with 10 mM NaOH NaOH Dimethylamine 49.3% Diethylamine 62.6%

19.6% 27.5%

6.7% 9.7%

4.8% 9.4%

4.6% 8.8%

Table 2 Properties of various amines and ammonia related to the SGM process. Amine species

MA

b.p. (°C)

Kb, 10  4

Dg, (m2 s  1)

H (mol (m3  Pa)  1

Ratio

Ammonia Methylamine Dimethylamine Trimethylamine Ethylamine Diethylamine Triethylamine

17.03 31.06 45.08 59.11 45.08 73.14 101.19

 33.4  6.37 7.07 3.08 16.77 55.73 88.63

0.18 4.26 4.79 0.63 5.01 10.47 5.62

1.80E  05 1.27E  05 1.01E  05 8.39E  06 1.01E  05 7.55E  06 6.25E  06

0.618 0.557 0.544 0.474 0.463 0.384 0.328

1 0.783 0.637 0.608 0.749 0.675 0.654

334

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

and kM can be expressed as (Wang et al., 1993)

(Cussler, 1984): D0 ¼

2DAH þ DOH DAH þ þ DOH  

ð7Þ

kM ¼ Dg

ε 1   τ Ri RTH ln Ro =Ri

ð11Þ

where C0 is the inlet concentration of the volatile base (or acid) in the feed-in solution, C1 is the outlet mixed-cup concentration of the volatile base (or acid) in the feed-out solution, K is the overall mass transfer coefficient, S is the effective area of the membrane module, and Q is the volumetric flow rate of the feed through the membrane module. The overall mass transfer coefficient K can be obtained through experimentation, and K can be further described in the following form (Semmens et al., 1990; Qin and Cabral, 1997; Wang et al., 1993):

where Dg is the molecular diffusion coefficient of amine or ammonia in the gaseous phase; ε is porosity of the fiber membrane wall; τ is tortuosity of the membrane wall; Ri and Ro are fiber lumen radius and fiber outside radius, respectively; R is the universal gas constant; T is the temperature; and H is Henry's constant. Theoretically, if all parameters in Eq. (11) can be exactly calculated or experimentally measured, kM can be calculated. However, ε and τ cannot be measured precisely through experimentation. Therefore, a more accurate value of kM should be obtained by comparing the SGM experimental data to the simulated value by using mathematical modeling and the least square method. Then, the ratio of kM for various amines and that of ammonia can be calculated by using Henry's constant H for ammonia and amines (Wang et al., 1993; Yaffe et al., 2003; Altschuh et al., 1999; Nirmalakhandan et al., 1997) and the gas diffusion coefficient calculated by Eq. (11). The values of MA, Dg, and H of ammonia and amines and the ratio of the amines’ kM and ammonia's kM were calculated at 25 °C and listed in Table 3. The lumen-side mass transfer coefficient kL is a function not only of the properties of the feed flowing in the lumen side but also of the membrane mass transfer coefficient kM when the feed laminarly flows through the lumen side of the fibers (Qin and Cabral, 1997). Therefore, it is impossible to directly obtain the values of kM and kL from Eq. (10) even though the value of K can be obtained from the experimental SGM data. A mass transfer theoretical model similar to that constructed in the literature (Qin and Cabral, 1997; Qin et al., 1996; Wang et al., 1993) is used to describe this SGM process to remove amine and obtain the values of kM and kL.

1 1 1 1 ¼ þ þ K kS kL kM

3.4. Mathematical model to describe the hollow fiber SGM process for separation amine or ammonia from their aqueous solution

where DAH þ and DOH- are the diffusion coefficients of AH þ and OH  respectively. However, no data can be found for the value of DAH þ , and no correlation can be found for ionized amines in the literature. In this paper, it is further simplified that the diffusion coefficient of the ion pair is equal to the diffusion coefficient of free amine molecules. 3.3. Mass transfer coefficient When the SGM process was used to remove volatile base (or acid) from its aqueous solutions by using an acidic (or basic) absorbing solution, respectively, the relationship between the feed-in and feed-out concentration of the volatile base (or acid) can be expressed by Eq. (8) when the concentration of the acid (or base) in the absorbing solution is sufficiently high (Kenfield et al., 1988; Semmens et al., 1990; Qin et al., 1990, 1996; Wang et al., 1993): lnðC 0 =C 1 Þ ¼ KS=Q

ð8Þ

ð9Þ

where kL, kM and kS are the lumen-side, membrane and shell-side mass transfer coefficients, respectively, when a hollow-fiber membrane module is used. In the operation of a hollow fiber-based SGM process, the feed usually flows through the lumen side of the module while the absorbing solution flows through the shell side of the module in order to ensure a higher value of K (Qin et al., 1996; Qin and Cabral, 1997). Because the reaction between amine (or ammonia) and H2SO4 can be considered to be instantaneous, the concentration of H2SO4 in the absorbing solution in the shell side is excessive when compared to that of amine in lumen side. Moreover, the flow rate of the absorbing solution in the shell side is sufficiently high that the dead-end effect in the shell is eliminated, the reaction occurs strictly within the absorbing solution at the outside surface of the hollow fibers, and the concentration of free amine in the shell side can be considered to be zero. Thus, the mass transfer resistance in the shell side of the membrane module can be neglected; consequently, Eq. (9) can be further reduced to: 1 1 1 ¼ þ K kL kM

ð10Þ

The mass transfer coefficient kM of the membrane is a function of the characteristics of the membrane itself and the operation conditions (Qin et al., 1996; Qin and Cabral, 1997). It is supposed that micropores on the wall of the hollow fiber are uniformly distributed along with the axis of hollow fibers, and thus kM remains constant along with the axis of the fibers. When the SGM process is operated at ambient temperature and ambient atmosphere, the mass transportation is dominated by Fickian diffusion

Proper assumptions have been made to construct the mass transfer model during the SGM process (Qin and Cabral, 1997; Qin et al., 1996; Wang et al., 1993). The most essential assumption is that the flow of the feed solution of volatile amine through the lumen side of the hollow fibers is a one-dimensional, steady, fully developed laminar flow. To describe the mass transfer in the hollow-fiber SGM process for separation of amine or ammonia from its aqueous solution, the following assumptions are made in order to describe the fluid flow and species diffusion in the lumen of the hollow fibers and their transport through the fiber wall: (1) the feed flow in the lumen is steadily laminar, and the velocity boundary layer is fully developed when mass transfer occurs; (2) reactions among all of the species in the lumen feed, being much faster than the mass-transfer rate, are assumed to be at local equilibrium; (3) diffusion in the axial direction in the lumen is assumed to be negligible, and the diffusion of molecules in the radial direction is Fickian diffusion; (4) the diffusion of ion pairs, AH þ –OH  , in the radial direction can be described by Fickian diffusion with a diffusion coefficient given by Eq. (6), and the value of the diffusion coefficient of the ion pairs is assumed to be equal to that of free amine molecules; (5) only free amine molecules are treated as volatile components, and Henry's law is effective for them at the gas/liquid interfaces; (6) when NaOH is added to the aqueous amine feed flowing in the lumen side and its concentration is sufficiently high to minimize the disassociation of amine molecules, the influence of NaOH on the volatility and diffusivity of amine molecules is negligible; (7) the absorbing liquid in the shell is an aqueous solution of nonvolatile strong acid, and the preferential pH of the absorbing solution

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

should be smaller than 5 as demanded by post-treatment (thus, the concentration of amine in the molecular state is considered to be zero, and the mass transfer resistance for amine in the shell side does not need to be considered further) (Yang and Cussler, 1986); (8) because the vapor pressure of the amine is usually low, the migration of amine within the micropores may be assumed to be Knudsen's or Fick's diffusion through air; and (9) all physicochemical parameters of the species—such as the reaction equilibrium constants, molecular diffusion coefficients in gas or liquid, and Henry's constants—are treated as only temperature dependent. Therefore, a group of partial differential equations can be used to describe the mass transfer of amine or ammonia in the lumen side of hollow fibers:  2  ∂C A D0 ∂ C A 1 ∂C A ¼ þR ð12aÞ þ r ∂r ∂z 2u½1  ðr=Ri Þ2  ∂r 2  2  ∂C AH þ D0 ∂ C AH þ 1 ∂C AH þ ¼ R þ r ∂r ∂z ∂r 2 2u½1  ðr=Ri Þ2 

ð12bÞ

Then integrating Eqs. (12a) and (12b) can yield:  2  ∂C D0 ∂ C 1 ∂C ¼ þ ∂z 2u½1  ðr=Ri Þ2  ∂r 2 r ∂r

ð12Þ

where C ¼ C A þC AH þ . The corresponding initial and boundary conditions are: z ¼ 0 C ¼ C0

ð13Þ

∂C ¼0 ∂r

ð14Þ

r¼0

∂C kM C A kM ¼ ¼  0 ðC  r ¼ Ri ∂r D0 D

qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi K 2b þ 4K b C K b 2

Þ

ð15Þ

where C0 is the inlet concentration of amine or ammonia in the feed-in solution. When the concentration of amine in the aqueous feed is sufficiently large, i.e., C»Kb and thus CA EC, Eq. (15) can be further reduced to ∂C  kM C A kM ¼ ¼  0C ∂r D0 D

least square method. Then kL can be calculated from Eq. (9), which is actually the average value from the entrance to the exit of the module (Qin and Cabral, 1997; Qin et al., 1996). 3.5. Estimation of gas diffusion coefficient of amine or ammonia Dg and molecular diffusion coefficient of amine in the feed solution D’ The molecular diffusion coefficient of amine or ammonia in the air can be calculated according to the Maxwell–Gilliland equation (Thomas, 1992): 4:36  10  5 T 2 ðM1A þ M1B Þ2 3

Dg ¼

1

1=3 1=3 pðV A þ V B Þ2

ð19Þ

where MA and MB are the molecular weights of amine (or ammonia) and air, respectively; p is the total pressure; and VA and VB are the molecular volumes of amine (or ammonia) and air, respectively. The diffusion coefficient of amine or ammonia in its dilute aqueous solution can be calculated on the basis of the Wilke– Chang equation (Thomas, 1992):  1=2 D0 ¼ 7:4  10  12 φM

T

μV A 0:6

ð20Þ

where φ is the association factor of water, M is the molecular weight of water, μ is the dynamic viscosity of the dilute solution as a function of temperature T, and VA is the molecular volume of the solute amine or ammonia at its normal boiling point. The dynamic viscosity μ of the dilute solution can be considered to be the same as that of pure water. Then, from Eq. (20), the molecular diffusion coefficients D’ of ammonia, methylamine, dimethylamine, trimethylamine, ethylamine, diethylamine and triethylamine in their respective dilute aqueous solutions at 25 °C are calculated as 1.64  10  9 m2 s  1, 1.54  10  9 m2 s  1, 1.18  10  9 m2 s  1, 9.73  10  10 m2 s  1, 1.22  10  9 m2 s  1, 8.90  10  10 m2 s  1 and 7.18  10  10 m2 s  1.

4. Results and discussion ð16Þ

where CA is the concentration of free amine molecules in the aqueous feed, u is the average velocity of feed in the lumen side of the hollow fiber, and D’ is the molecular diffusion coefficient of the volatile amine or ammonia in the feed solution. When NaOH exists in the feed, Eq. (15) can be re-expressed as ∂C kM C A ¼ 0 ∂r Dq ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 0 1 ðC NaOH þ K b Þ2 þ4K b C  ðC NaOH þ K b Þ kM @ A ¼  0 C 2 D

335

r ¼ Ri

ð17Þ

From Eq. (17), it can be seen that when the concentration of NaOH in the aqueous feed solution is sufficient high, i.e., (CNaOH þKb)2»4KbC, Eq. (17) can be reduced to Eq. (16). If a group of values of D’, u, Ri, C0, CNaOH, and kM are given, Eqs. (12)–(17) can be solved by the numerical method; this means that for a certain point in the lumen side, the value of feed concentration C can be obtained. Thus, the mixed-cup concentration Cmix at the feed outlet can be calculated by R Ri  Z Ri  2π ruCdr 4 r 1  ð Þ2 rCdr ð18Þ ¼ 2 C mix ¼ R0 R i Ri Ri 0 0 2π rudr By comparing the simulated value of the outlet mixed-cup concentration to the experimentally obtained data of the feed-out concentration, the real value of kM can be obtained by using the

A series of experiments was performed to investigate the effect of various operating factors, such as amine species, feed-in concentration, feed-in temperature and feed-in flow rate, on stripping of amine by the SGM process. In all experiments employing hollow-fiber membrane modules, the variation rangess of temperature, feed-in concentration and feed-in volumetric flow rate were 15–45 °C, 500–5000 mg L  1 and 15–45 L h  1, respectively. Six aliphatic amines of low molecular weight and ammonia were tested. Among them, dimethylamine (DMA) was the most widely used or frequently occurring in wastewater streams; thus, its removal by the SGM process was emphatically studied. In all stripping tests of amines or ammonia by the SGM process, an aqueous solution of 10 wt% H2SO4 was used as an absorbing solution, and the volumetric flow rate QA of the absorbing solution was maintained at 50 L h  1. Such an initial concentration of H2SO4 and flow rate of the absorbing solution ensured that the mass transfer resistance of amine in the shell side was negligible. The capability of an SGM process to remove amines or ammonia is determined by its removal percentage. When the aqueous feed passed through modules serially in one pass, the removal percentage was defined as RA ¼ ðC 0 C 1 Þ=C 0  100%

ð21Þ

where C0 and C1 were the concentrations at the entrance of the first module in the series and the exit of the last module in the series, respectively.

336

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

Fig. 2. Removal percentage of ammonia and amines vs. operation time without NaOH in the feed solution (Module 2 used, T ¼ 30 °C, C0 E3000 mg L  1, Q¼ 25 L h  1).

Fig. 4. Removal of amines by using three identical modules in series at various feed-in flow rates (Modules 1a, 1b and 1 c used, T ¼30 °C, C0 E3000 mg L  1).

removal percentage of ammonia and amines with operation time was obtained even when the removal of ammonia and amines was up to ca. 85%, which gave 98% removal of ammonia with 96–98% removal of amines at the operation time of 90 min. This could be due to the fact that the existence of NaOH could weaken the disassociation of amine molecules and efficiently increase the removal percentage of amines tested. The experimental results above indicate that the SGM process used for removal of ammonia and amines is feasible. 4.2. Removal of ammonia and amines by using three identical modules in series

Fig. 3. Removal percentage of ammonia and amines vs. operation time with NaOH in the feed solution (Module 2 used, T ¼30 °C, C0 E3000 mg L  1, Q¼ 25 L h  1, CNaOH ¼200 mg L  1).

When the aqueous feed was cycled between a module and a feed reservoir, the removal percentage was defined as R‘A ¼ ðC ‘0  C t Þ=C ‘0  100%

ð22Þ

where C0’ and Ct were the initial concentration and the concentration in the feed reservoir at elapsed time t, respectively. 4.1. Removal of ammonia and amines by SGM process when feed solution cycled between module and reservoir Figs. 2 and 3 show the removal percentage of ammonia and amines vs. operation time without NaOH and with NaOH in the feed solution, respectively. As can be observed, the removal percentage of ammonia and amines quickly increased with the operation time for both feed solutions without NaOH and with NaOH. Nevertheless, as shown in Fig. 2, when the removal of ammonia and amines was up to ca. 85%, a further increase in the removal percentage of amines with operation time became very slow compared to the removal of ammonia for the feed solution without NaOH, which gave 97% removal of ammonia with 90–93% removal of amines at the operation time of 90 min. This could be due to the fact that when the concentration of amine in the aqueous solution was very low, amine dissociated significantly, and only a small part of the amine was in its free and volatile form. On the other hand, as shown in Fig. 3, when a feed solution with a small amount of NaOH, e.g., 200 mg L  1, was used, a higher

The SGM process is usually operated in a continuous mode, and the feed passes the modules in series in order to ensure a high removal percentage of ammonia or amines. The experiments were performed at 30 °C and with a feed-in concentration of approximately 3000 mg L  1. As seen from Fig. 4, trimethylamine was the easiest one to be removed by the SGM process even though it had the highest molecular weight and a low diffusion coefficient in aqueous phase and the gaseous phase; this was consistent with the volatility of amines indicated by their Henry's constants as given in Table 2. From Fig. 4, it could be known that because of the increasing feed-in volumetric flow rate, the retention period of amines in the membrane module decreased. Therefore, the removal of amines decreased with increasing feed-in volumetric flow rate. This is a reasonable conclusion since that ammonia or amine removal percentage can be calculated from Eqs. (9) and (21), in which K increases in a small degree when feed-in flow increases (Qin and Cabral , 1997). As plotted in Fig. 4, when the feed-in volumetric flow rate decreased to 15 L h  1, 495% of amines were removed. It was further observed that amine could be enriched by 4 20 times in the absorbing solution as indicated by the value of amine concentration in the absorbing solution obtained by the pH adjustment–distillation–titration method compared to the inlet concentration of amines in the feed solution. 4.3. Overall mass transfer coefficient of SGM process for amine stripping The overall mass transfer coefficient K of the SGM process for stripping ammonia and amines was calculated from Eq. (8) by using experimental data obtained at various feed-in temperatures. In Fig. 5, the overall mass transfer coefficients of various amines with a feed-in concentration of approximately 3000 mg L  1 and a feed-in volumetric flow rate 45 L h  1 were plotted as a function of the feed-in temperature. It can be seen that K for amines and

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

Fig. 5. Overall mass transfer coefficient of the SGM process for stripping amines at various feed-in temperatures (Module 2 used, C0 E3000 mg L  1, Q¼ 45 L h  1).

Fig. 6. Overall and partial mass transfer coefficients for removal of dimethylamine by the SGM process at various feed-in temperatures (C0 E 3000 mg L  1, Q¼ 45 L h  1).

ammonia obviously increased with increasing feed-in temperature. It can be further seen that overall mass transfer coefficient K was in the order of trimethylamine4 ammonia4triethylamine4 diethylamine 4ethylamine4 dimethylamine4methylamine. This order was not consistent with the order of the amines’ molecular weight or boiling points listed in Table 1. It was concluded that the overall mass transfer coefficients of amines was a strong function of their polarity or their relative volatility (indicated by Henry's law coefficient) as well as their molecular weight (indicated by the diffusion coefficient of the amine in aqueous phase and gaseous phase). This will be further discussed in the next section. 4.4. Partial mass transfer coefficients in the porous membrane and in the aqueous feed flowing in the lumen side The values of the overall mass transfer coefficient were drastically different for different amines, as demonstrated by the values of the overall mass transfer coefficient obtained from the SGM experimental data as shown in Section 4.3. Therefore, the values of partial mass transfer coefficients kM of amine through the porous membrane wall and that through the lumen-side aqueous feed, kL, were calculated using the least square method by comparing the experimental data of the feed-out concentration at the module exit to the simulated values of the outlet mixed-cup concentration obtained by solving partial differential Eqs. (12)– (17). Dimethylamine was taken as an example, and the values of

337

kM and kL are demonstrated in Fig. 6. From Fig. 6, it can be seen that the membrane-wall mass transfer coefficient kM and lumenside feed liquid mass transfer coefficient kL increased significantly with increasing feed-in temperature. The effect of temperature on partial mass transfer coefficients can be explained by the partial pressure of amine, namely, the driving force of the SGM process, increasing with increasing temperature. When the feed-in temperature increased, the gas diffusion coefficient Dg increased and the amine's volatility drastically increased as indicated by the drastic decease of H. Thus, kM increased with increasing temperature according to Eq. (11). Conversely, because the feed solution viscosity decreased and the liquid diffusion coefficient DL increased with increasing temperature, the liquid mass transfer coefficient kL increased with increasing temperature. From Fig. 6, it can be seen that Module 2 provided both a higher K value and a higher kM value than Module 1a used in the experiments. This was because the hollow fibers used in Module 2 had a higher porosity than the hollow fibers used in Module 1a and probably had lower values of tortuosity than the hollow fibers used in Module 1a, thus leading to a higher kM value as expressed by Eq. (10). From Fig. 6, it can also be seen that for Module 1a, the value of K was closer to that of kM, which indicates that for membranes with lower porosity and probably higher tortuosity, the resistance of the porous membrane itself is a dominant resistance compared to the mass transfer resistance in the lumen side; conversely, for fiber membranes with higher porosity and probably lower tortuosity as used in Module 2, the value of K was quite close to the value of kM. It can be further concluded that for both cases mentioned above, the influence of kM on K is important, and according to Eq. (11), the influence of Dg and H on K is significant. Therefore, combining the data shown in Section 4.3, it can be further concluded that the difference of overall mass transfer coefficients for various amines resulted from the difference of their relative volatility in their respective dilute aqueous solutions, as indicated by their Henry's constant, but are less influenced by their molecular weight, which simply influenced the diffusion coefficient of amines in the gaseous phase and the aqueous phase. Unfortunately, the data on Henry's constant listed in Table 2 do not show a significant difference among amines with different molecular weights and different structures, because the values of these Henry's constants were theoretically estimated rather than directly measured from experimental data. 4.5. Influence of feed-in concentration on SGM process Related experiments were conducted to investigate the influence of feed-in concentration on the mass transfer process with different membrane modules and various amine species, especially when the feed-in concentration was very low. To maintain long-term operational stability, the influence of feed-in concentrations on the surface tension of the aqueous amine solution was also measured. 4.5.1. Overall and partial mass transfer coefficients for dimethylamine with a moderately low feed-in concentration The SGM experiments were performed for dimethylamine at a moderately low feed-in concentration of 1000–5000 mg L  1. In Fig. 7, the values of K, kM and kL for dimethylamine in different hollow-fiber modules were compared as a function of feed-in concentration. It was evident from Fig. 7 that for the two types of fibers, the overall and partial mass transfer coefficients were almost independent of feed-in concentration. This can be attributed to the fact that when the feed-in temperature was fixed at a certain value—and thus gaseous diffusion coefficient Dg, liquid diffusion coefficient DL and Henry's constant H changed little with

338

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

Fig. 7. Overall and partial mass transfer coefficients for removal of dimethylamine by the SGM process at various feed-in concentrations (T ¼ 30 °C, Q ¼45 L h  1).

Fig. 8. Overall mass transfer coefficient of ammonia, and various amines vs. feed-in concentration (Module 2 used, T ¼ 30 °C, Q ¼45 L h  1).

feed-in concentrations—kM and kL were almost constant. This is true when feed-in concentration is sufficiently high and dissociation of amine is negligible. 4.5.2. Influence of feed-in concentration on the overall mass transfer coefficient for various amines and ammonia, especially in low feed-in concentration According to the data given in Table 2, all amines tested for the SGM process had a larger dissociation equilibrium constant Kb than ammonia. Therefore, amines may dissociate severely when feed-in concentration of amine was very low, as indicated by Eq. (4). The SGM experiment was thus purposely performed by using a much lower feed-in concentration, for example, using a feed-in concentration below 500 mg L  1. When it was assumed that all of the amine was in its free molecule form, the apparent overall mass transfer coefficient K can still be obtained from experimental data by using Eq. (8). The value of K obtained using such an assumption was plotted in Fig. 8 as a function of feed-in concentration. From Fig. 8, it can be seen clearly that the apparent overall mass transfer coefficient K was almost independent of feed-in concen tration when feed-in concentration was in the range of 1000– 5000 mg L  1, which was consistent with the explanation in the above section. However, when the feed-in concentration was below 1000 mg L  1, the apparent overall mass transfer coefficient K for all amine species except ammonia decreased significantly with decreasing feed-in concentration as demonstrated in Fig. 8.

Fig. 9. Removal percentage of various amines vs. feed-in concentration (Modules 1b used, T ¼ 30 °C, Q¼ 25 L h  1).

This could be attributed to the fact that when the amine concentration in the feed was sufficiently low, a high percentage of amine would exist in the aqueous solution as ionized and nonvolatile ammonium ions, which actually reduced the availability of free amine for stripping by the SGM process. Therefore, it was rational and necessary to ensure a sufficiently high pH to minimize the dissociation of amine and to ensure a high K value and a high removal percentage by adding NaOH to the feed solution in advance. Conversely, when the feed-in concentration was low, the dissociation of amine must be considered, and a simple expression such as Eq. (8) is not sufficient to describe the SGM process for amine removal. A more complicated mathematical model such as that described by Eqs. (12)–(18) must be used. The removal percentage of various amines was further plotted as a function of feed-in concentration in Fig. 9. As a comparison, the simulated values of removal percentage obtained from Eqs. (12)–(18) by using the value of kM experimentally obtained at high feed-in amine concentrations is also given in Fig. 9, as are the simulated values obtained simply by solving Eq. (8) and those experimentally obtained at high feed-in amine concentrations. It can be seen from Fig. 9 that the experimentally obtained amine removal percentage could be well predicted by solving Eqs. (12)– (18) (complicated expression); however, the removal percentage could not be described by the simple expression given by Eq. (7). 4.6. Influence of feed-in volumetric flow rate on SGM process for dimethylamine stripping A series of experiments was performed at various feed-in volumetric flow rates. As plotted in Fig. 10, it can be seen that for both modules, the membrane-wall mass transfer coefficient kM had only a slight improvement; however, the lumen-side feed liquid mass transfer coefficient kL increased more obviously than kM with increasing feed-in volumetric flow rate. It can be said that kM being almost independent of feed-in volumetric flow rate in the lumen side of the membrane module simply confirmed that kM actually represented the features of the porous membrane wall. In addition, an increase in the feed-in volumetric flow rate reduced the mass transfer boundary layer resistance in the lumen side; thus, kL increased to a certain extent. From Fig. 10, it can also be seen that the feed-in volumetric flow rate affected the overall mass transfer coefficient K to a certain degree for Module 2, which was consistent with the conclusion from Section 4 that both the contribution of membrane-wall mass transfer resistance and lumen-side mass transfer resistance to the overall mass transfer coefficient were equally important.

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

339

Fig. 10. Overall and partial mass transfer coefficients of dimethylamine vs. feed-in volumetric flow rate (T ¼ 30 °C, C0 E 3000 mg L  1).

Fig. 12. Overall mass transfer coefficient (a) and removal percentage and (b) for ammonia and amines vs. operation time (Modules 2 used, T ¼30 °C, C0 E3000 mg L  1, Q¼ 45 L h  1). Fig. 11. Surface tension of ammonia and amine vs. their concentration in the aqueous solution at T ¼ 30 °C.

4.7. Measurement of surface tension of aqueous amine solutions Aliphatic amines of low molecular weight are characterized by volatility and apolarity. Thus, it is reasonable to predict that aqueous solutions of amines should have a lower surface tension than aqueous solutions of ammonia. The low surface tension of aqueous feed in the SGM process is a fatal problem because the SGM process would not work normally when the initially hydrophobic porous membrane is wetted by an aqueous solution with low surface tension. Thus, the surface tension of aqueous solutions of various amines at different concentrations was measured at 30 °C. It can be seen clearly from Fig. 11 that the higher the molecular weight of the amine was, the smaller the surface tension was. As can be observed in Fig. 11, the surface tension of various amines decreased with increasing amine concentration. Regarding the longterm use of the SGM process for amine removal, it is suggested that SGM is used to treat aqueous feeds with the surface tension of 445 mN m  1 when the PP porous fibers are used. This is due to the fact that when the aqueous feed solutions such as trimethylamine solution, diethylamine solution and triethylamine solution with the surface tension of 45–50 mN m  1 are used, the liquid entry pressure (LEP) of the feeds measured is still kept at 1.0–1.4 atm at 30 °C, which is typically larger than the inlet pressure of the hollow fiber modulesin-series in industrial application with the value of less than 0.3 atm (Qin et al., 1996). This indicates that when the aqueous aliphatic amine solutions with the surface tension of 445 mN m  1 are used as feed,

the SGM process can be applied for the removal and recovery of aliphatic amines from the aqueous streams, and no leakage and wetting of the porous fibers occurs. However, when the aqueous triethylamine solutions with the concentration of 4000–5000 mg L  1 are used as feed, and such feeds are with the surface tension of less than 40 mN m  1 as shown in Fig. 11, and the LEP of such feeds measured ranges from 0.2 to 0.4 atm, then, the porous fibers in the hollow fiber module are wetted and serious leakage occurrs in a three-day test. Thus, when the aqueous aliphatic amine streams with the surface tension of 445 mN m  1 are used as feed, the removal and recovery of aliphatic amines from the aqueous streams by using SGM process are feasible. 4.8. A long-term operational stability test From the viewpoint of industrial field application, only when the SGM process provides operational stability in a sufficiently long period can it be used for industrial purposes. Fig. 12a and b show the variations of overall mass transfer coefficient K and removal percentage RA for ammonia and amines vs. operation time by using Module 2 in the long-term operational stability test. The long-term operational stability test was performed at 30 °C with a feed-in volumetric flow rate of 45 L h  1 and a feed-in concentration of 3000 mg L  1, and such feeds are with the surface tension of 445 mN m  1 as illustrated in Fig. 11. As seen from Fig. 12a, all the initial values of K for ammonia and amines were high and decreased quickly within a short time of 20– 26 h. Then, the values of K were almost kept stable for the following 700–800 h. Nevertheless, when such a test was carried out for more

340

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

than 1000 h, all the values of K decreased significantly. This is probably due to the biological fouling of the feed without pretreatment of ultrafiltration, which could generate natural surfactants in the aqueous amines streams and then the reduction of the surface tensions of the feed happens in practice. This could lead to the increase of the wettability of the feed to the surface of the porous membrane. Therefore, the pretreatment such as ultrafiltration can be used to remove these microbial agents, and furthermore the modules should be washed periodically by using acid solution to kill these microbial agent. As seen from Fig. 12b, the value of RA of ammonia and amines has an important positive relationship with the value of K, and the initial values of RA for ammonia and amines decreased sharply within 20– 26 h. Then, the values of RA were kept above 20% for the following 700–800 h, and when such a test was stopped after running for more than 1000 h, all the values of RA decreased significantly. Thus, compared to the hydrophobicity and pore size of polypropylene, a hydrophobic porous membrane with higher hydrophobicity and smaller micropore diameter could be more conducive to extend the operation period, e.g., more than 40 days, for the removal of aliphatic amines from aqueous streams by using SGM process. Simultaneously, as analyzed above, the performance of SGM process for Module 2 used can be kept stable for more than 800 h when a variety of aqueous aliphatic amine solutions with the surface tension of 445 mN m  1 are used, which, again, demonstrates that the removal and recovery of aliphatic amines from the aqueous streams by using SGM process are feasible, and demonstrates good stability in the long-term test for SGM process in the removal and recovery of aliphatic amines from the aqueous streams.

5. Conclusion The supported-gas-membrane separation process was investigated experimentally and theoretically for the removal, recovery and enrichment of various amines from their individual dilute aqueous solutions using two types of PP hollow-fiber membranes. Aqueous solutions containing 200–5000 mg L  1 of a single amine were tested as feed, and aqueous solutions of 10 wt% sulfuric acid were used as absorbing solutions. Amines such as methylamine, dimethylamine, trimethylamine, ethylamine, diethylamine and triethylamine, as well as ammonia, were tested for this SGM process. The experimental data demonstrated that the overall mass transfer coefficient was a strong function of the polarity of amine or amine/water volatility and was not a simple function of molecular weight or boiling point of the amines; the order was determined to be trimethylamine4ammonia4 triethylamine4diethylamine4ethylamine4dimethylamine4methylamine. The mathematical models incorporating laminar flow, ion and molecular diffusion, dissociation equilibrium and vapor–liquid equilibrium were established and solved numerically. This model can describe the SGM process for amine removal very well. The feed-in temperature was found to have a strong influence on the efficiency of the separation process because temperature acts on both transportation through the membrane and the kinetics of the reaction. For efficient amine removal, the amine must be in its volatile molecular form, and increasing the feed-in temperature ensures the presence of the free amine in the aqueous solution. The mass transfer coefficients increased moderately with increasing feed-in volumetric flow rate owing to the decreasing boundary layer resistance. The SGM separation process coupled with a reaction in the receiving sulfuric acid side can convert amine to amine salt so that amines can be recycled to avoid secondary pollution and wasted resources. If a target product is enriched in an amine aqueous solution, it can be gained through distillation by adding alkali. In view of that, the overall mass transfers of all amines tested were close to that of ammonia, and the SGM-based stripping process is a highly efficient method to remove, recover, concentrate and purify amines from their individual aqueous solutions. When the surface

tension of the feed solution was 445 mN m  1, the SGM process demonstrated good stability in a test period of at least 30 days. More than 95% of the amine was recovered, and the amine could be enriched by 420 times in the absorbing solution. Thus, this SGM-based separation process is suitable to remove, recover, and concentrate amines from their aqueous solutions.

Nomenclature C0 C0’ C1 Ct Cmix Dg DL DAH þ DOHH n K Kb kL kM kS M MA MB p Q QA R RA RA’ Ri Ro S T u VA VB ZM

inlet concentration of amines in the fiber lumen (mg L  1) initial concentration of amines in the feed reservoir (mg L  1) outlet concentration of amines in the fiber lumen (mg L  1) concentration of amines in the feed reservoir at elapsed time t (mg L  1) mixed-cup concentration at the feed outlet (mg L  1) molecular diffusion coefficient of amines or ammonia in the gaseous phase (m2 s  1) molecular diffusion coefficient of amines or ammonia in the feed solution (m2  1) diffusion coefficient of ion AH þ (m2 s  1) diffusion coefficient of ion OH  (m2 s  1) Henry's constant (mol m  3 Pa  1) number of hollow fibers overall mass transfer coefficient (m s  1) dissociation equilibrium constant of amine lumen-side mass transfer coefficient (m s  1) membrane-wall mass transfer coefficient (m s  1) shell-side mass transfer coefficient (m s  1) molecular weight of water molecular weight of amine or ammonia molecular weight of air total pressure (Pa) volumetric flow rate of the feed through the membrane module (L h  1) volumetric flow rate of absorbing solution (L h  1) universal gas constant (J mol  1 K  1) removal percentage of amine or ammonia through a module or modules in series in one pass removal percentage of amine or ammonia cycled between a module and a feed reservoir fiber lumen radius (mm) fiber outside radius (mm) effective surface area of membrane module (m2) temperature (oC) average velocity in the lumen side of the hollow fiber (m s  1) molecular volume of amine or ammonia molecular volume of air effective length of hollow fibers (m)

Greek letters

ε μ τ φ

porosity of the membrane wall dynamic viscosity of the dilute solution (mPa s) tortuosity of the membrane wall association factor of water

J. He et al. / Chemical Engineering Science 141 (2016) 330–341

Acknowledgment The authors gratefully acknowledge the financial support of Tianjin Municipal Science and Technology Commission, China (Grant no. 11ZCZDSF04700). The authors are grateful to Mr. Dongsheng Cui and Ms. Jianmin Zhao for their sincerely assistance.

References Altschuh, J., Brüggemann, R., Santl, H., Eichinger, G., Piringer, O.G., 1999. Henry's law constants for a diverse set of organic chemicals: experimental determination and comparison of estimation methods. Chemosphere 39, 1871–1887. Chang, Q.L., Li, Q.S., Wang, B.H., 2010. Study on recycling and reusing of industrial wastewater containing dimethylamine. Chin. J. Environ. Eng. 4, 35–38. Chen, S., Sun, D.Z., Zheng, Z.Z., 2008. Method for biologically treating organic wastewater containing high-concentration aniline or amine (ammonia). Chin. Patent 101 (357), 805. Cussler, E.L., 1984. Diffusion. Cambridge University Press, Cambridge, p. 147. Imai, M., Furusaki, S., Miyauchi, T., 1982. Separation of volatile materials by gas membranes. Ind. Eng. Chem. Process Des. Dev. 21, 421–426. Kenfield, C.F., Qin, R., Semmens, M.J., Cussler, E.L., 1988. Cyanide recovery across hollow fiber gas membranes. Environ. Sci. Technol. 22, 1151–1155. Meng, Y.Y., Hu, Q.H., Lu, Q.Z., Xiong, Y.L., 2010. Study on treatment of DMA containing wastewater with ion exchange resin. Environ. Pollut. Control 32, 58–63. Mackenzie, P.D., King, C.J., 1985. Combined solvent extraction and stripping for removal and isolation of ammonia from sour waters. Ind. Eng. Chem. Process Des. Dev. 24, 1192–1200. Nirmalakhandan, N., Brennan, R.A., Speece, R.E., 1997. Predicting Henry's Law constant and the effect of temperature on Henry's Law constant. Water Res. 31, 1471–1481.

341

Qin, R., Zander, A.K., Semmens, M.J., Cussler, E.L., 1990. Separating acetic acid from liquids. J. Membr. Sci. 50, 51–55. Qin, Y., Cabral, J.M.S., 1997. Lumen mass transfer in hollow-fiber membrane processes with constant external resistances. AIChE J. 43, 1975–1988. Qin, Y.J., Cabral, J.M.S., Wang, S.C., 1996. Hollow-fiber gas-membrane process for removal of NH3 from solution of NH3 and CO2. AIChE J. 42, 1945–1956. Qin, Y.J., Cabral, M.S., 1998. Influence of Nernst–Planck diffusion on hollow-fiber mass-transfer processes. AIChE J. 44, 1529–1541. Qu, J.X., Chen, J.Z., 2009. Influence factors of removal of dimethylamine from wastewater by air stripping. Technol. Water Treat. 35, 88–90. Semmens, M.J., Foster, D.M., Cussler, E.L., 1990. Ammonia removal from water using microporous hollow fibers. J. Membr. Sci. 51, 127–140. Semmens, M.J., Qin, R., Zander, A., 1989. Using a microporous hollow fiber membrane to separate VOCs from water. J. Am. Water Work. Assoc. 81, 162–167. Thomas, L.C., 1992. Heat Transfer. Prentice-Hall, Englewoodcliffs, NJ. Titmas, J.A., Fluto, P., 1993. Process and apparatus for removing carbon dioxide and ammonia from wastewater. US Patent 5,190,665. Wang, S.C., Xu, S.C., Qin, Y.J., 1993. Mass transfer in membrane absorptiondesorption of ammonia from ammonia water. Chin. J. Chem. Eng. 1, 160–170. Wang, S.Y., Chen, Y.W., Wei, J.Y., Shen, S.B., 2011. Study on the pretreatment of highly concentrated organic amines wastewater by fenton oxidation reaction. Ind. Water Treat. 31, 25–27. Yaffe, D., Cohen, Y., Espinosa, G., Arenas, A., Giralt, F., 2003. A fuzzy ARTMAP-based quantitative structure-property relationship (QSPR) for the Henry's law constant of organic compounds. J. Chem. Inf. Comput. Sci. 43, 85–112. Yang, M.C., Cussler, E.L., 1986. Designing hollow fiber contactors. AIChE J. 32, 1910–1916. Zhang, Z., Zhou, Y.X., Wang, W.Z., 2010. Study of organic amine wastewater by chemical precipitation method. Environ. Sci. Manag. 35, 63–65.