Cyanide removal from industrial wastewaters using gas membranes

Journal of Membrane Science 257 (2005) 171–181

Cyanide removal from industrial wastewaters using gas membranes Binbing Hana , Zhisong Shena,b , S. Ranil Wickramasinghea,∗ b

a Department of Chemical Engineering, Colorado State University, Fort Collins, CO 80523-1370, USA Department of Environmental Engineering, Jiangsu Institute of Microbiology, Wuxi 214063, Jiangsu Province, China

Received 12 February 2004; received in revised form 23 June 2004; accepted 25 June 2004 Available online 3 February 2005

Abstract Results are presented for the removal of cyanide from four industrial wastewaters using hollow fiber gas membranes in a pilot plant. The plant was operated in batch mode using 1000 L of feed solution. The plant contained 10 hollow fiber modules with total membrane surface area of 180 m2 . The strip stream consisted of 10% NaOH. The overall mass transfer coefficient for cyanide has been determined experimentally and found to agree well with predictions using empirical correlations. Both the feed and membrane mass transfer coefficients contribute to the overall mass transfer coefficient. The strip side mass transfer coefficient may be ignored. Real wastewaters often contain other volatile species. These volatile species will also transfer to the strip solution. If the volatile species does not react with the strip stream, all three individual mass transfer coefficients contribute to the overall mass transfer coefficient. Good agreement between the experientially determined overall mass transfer coefficient and the predicted values is obtained for other volatile species. The osmolarity of a wastewater may be different to that of the strip solution. Consequently, water vapor transport due to osmotic distillation will occur from the lower to higher osmolarity solution. The effects of osmotic distillation should be considered when sizing the feed and strip tanks. © 2004 Elsevier B.V. All rights reserved. Keywords: Membrane distillation; Microporous and porous membranes; Water treatment; Osmotic distillation; Cyanide removal

1. Introduction A microporous hydrophobic membrane may be used to separate two aqueous streams resulting in an immobilized gas–liquid interface at both membrane surfaces. The membrane pores will be air-filled. These ‘gas membranes’ may be used to conduct non-dispersive adsorption or stripping at both gas–liquid interfaces. If a volatile species is present in one of the aqueous streams and not the other, it will desorb from the first solution into the air-filled pores, diffuse through the air-filled pores and absorb into the second solution. Consequently, stripping takes place at the first gas–liquid interface while absorption takes place at the second. In isothermal batch operations, the volatile species will transfer from the ∗ Corresponding author. Present address: Max-Plank-Institut f¨ ur Dynamik komplexer technischer Systeme, Sandtorstraße 1, 39106 Magdeburg, Germany. Tel.: +1 970 491 5276; fax: +1 970 491 7369. E-mail address: [email protected] (S.R. Wickramasinghe).

0376-7388/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2004.06.064

feed to the strip solution till its activity in both aqueous solutions is the same. When this occurs, the partial pressure of the volatile species in equilibrium with both aqueous solutions will be the same; consequently, there is no driving force for further mass transfer. Previous studies have investigated the use of a reactive strip solution. Here, the volatile species reacts with a component in the strip solution. Thus, the activity and the partial pressure of the volatile component in equilibrium with the strip solution are greatly lowered allowing for enhanced removal from the feed solution. Imai et al. [1] studied the separation of NH3 and I2 from an aqueous solution into H2 SO4 and NaOH solutions, respectively. Zhang and Cussler [2,3] studied the removal of Br2 from brine into a NaOH strip solution. Semmens et al. [4] studied ammonia removal from water into a H2 SO4 strip solution. Gas membrane-based separations are closely related to osmotic distillation processes. Osmotic distillation also uses a microporous hydrophobic membrane to separate two aqueous

172

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

Table 1 Three gas membrane processes that can occur simultaneously during cyanide removal from real wastewaters Process

Species removed

Driving force

Non-reactive strip

Volatile species generally present at very low concentration in feed Volatile species generally present at very low concentration in feed Solvent, water for aqueous feed streams

Partial pressure difference, transport ceases when activity in both aqueous streams is the same Partial pressure difference, transport usually continues as long as the reactive component in the strip solution is present in excess Water vapor partial pressure difference between feed and strip solutions. Transport is from the solution of lower to higher osmolarity

Reactive strip Osmotic distillation

streams. If the two aqueous streams have different osmolarities, the water vapor partial pressure will be different leading to a water vapor pressure gradient. The osmotic pressure will drive water vapor from the less concentrated aqueous stream to the more concentrated aqueous stream [5]. Table 1 summarizes the important features of these three gas membrane processes: non-reactive strip, reactive strip and osmotic distillation. Here, we consider cyanide removal from real wastewater streams. Our results indicate that all three gas membrane processes often occur simultaneously. Cyanide is used in many industries. For example, in the electroplating industry, it is extensively employed to hold metallic ions such as zinc and cadmium in aqueous solutions. However, due to its high toxicity, removal of cyanide from liquid waste streams is essential [6]. According to the US Environmental Protection Agency (USEPA), the final cyanide concentration in wastewaters must be reduced to less than 5.0 mg L−1 for facilities discharging less than 10,000 gal/day or 1.9 mg L−1 for facilities discharging more than 10,000 gal/day, respectively [7]. Further, recovery and reuse of cyanide from wastewater streams may result in significant economical benefits. Several different cyanide removal processes have been described. Table 2 lists 10 of the more common methods of cyanide removal [6,8–27]. Among them, alkaline chlorination [6] is the most commonly used in practice. However, this process suffers from several problems such as less effective removal of iron cyanides; cyanide cannot be recovered and reused; and more severely, chloramines and free chlorine remain in solution leading to the production of secondary contaminants. The use of gas-filled microporous membranes to remove and recover cyanide from wastewaters, can overcome all these problems [20–27]. Fig. 1 is a schematic diagram of a gas-filled microporous membrane for cyanide removal. A hydrophobic microporous membrane, e.g. polypropylene (PP) or polytetrafluoroethylene (PTFE), is used to separate the two aqueous streams. The membrane pores remain gas-filled as long as the pressure difference between these two aqueous phases is less than the breakthrough pressure. The wastewater containing cyanide flows on one side of the membrane while the reactive strip solution containing NaOH flows on the other side. The protonated form of cyanide, or prussic acid (HCN), is a weak acid with a pKa of 9.31. Since HCN is volatile, it will vaporize, diffuse across the gas-filled pores and into the strip solution. In the reactive strip solution

it will react to form NaCN: HCN + NaOH ↔ NaCN + H2 O

(1)

The reaction between HCN and NaOH is very rapid, thus the HCN concentration in the strip solution is essentially zero. Consequently, HCN will continue to transfer from the feed to the strip solution providing there is excess base present in the strip solution. The use of gas-filled membrane pores offers a number of advantages. The cyanide can be recovered and reused; no secondary pollutants are produced; the energy and chemical requirements are low and the equipment is simple to operate [20–27]. Most previous studies have investigated cyanide removal in laboratory scale experiments using artificial wastewaters [21–26]. Short and co-workers [20,24] reported results of a pilot scale study. However, they used an ion exchange process to remove cyanide from the wastewater. Gas-filled membranes were used to regenerate the reagent and the ion exchange resin. Cyanide removal results using artificial wastewaters are quite different to those for real wastewaters. As shown in Fig. 1, other volatile components present in the feed can also pass through the membrane pores and into the strip solution. If these volatile species do not react with a component of the strip solution, in batch operation, transfer will stop once the activity in the two solutions is equalized. Further, differences in osmolarity between the two aqueous streams will cause transfer of water vapor through the membrane pores to the solution with higher osmolarity.

Fig. 1. Illustration of HCN removal by gas-filled membrane absorption.

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

173

Table 2 Cyanide removal methods Method

Description

Comments

Alkaline chlorination

Cyanide is oxidized to cyanate using chlorine or hypochlorite

Sulfur oxidation

Cyanide is oxidized to cyanate using sulfur dioxide or ferrous sulfate and air in the presence of copper ions. Sulfuric acid formed is neutralized with lime

Hydrogen peroxide oxidation

Cyanide is oxidized to cyanate using hydrogen peroxide in the presence of copper ion

Acidification–volatilisation–recovery (AVR)

The pH of a cyanide solution is lowered by addition of sulfuric acid so that HCN gas is formed. This gas can then be absorbed into a NaOH solution

Activated carbon adsorption Ion exchange

Cyanide is removed by activated carbon adsorption Cyanide is removed by anion exchange resin

Most commonly used; less effective for iron cyanides; cannot recover cyanide; remaining chloramines and free chlorine lead to secondary contamination Reaction is slow at low temperatures leading to larger tank volumes; generally does not remove thiocyanate, cyanate, or ammonia; cyanate can be transformed into toxic ammonia by microbial action Hydrogen peroxide is hazardous and expensive. Requires specialized equipment increasing the total capital cost. The treatment process generates ammonia, which is toxic to fish Cyanide can be recovered; more complex and hazardous than other processes. Sealed mixing vessels and packed columns are required. The economics may vary depending on the value of the recovered cyanide Regeneration of activated carbon is difficult

Ozonation

Cyanide is oxidized using ozone

Photochemical destruction

Cyanide is destroyed using ultraviolet (UV) irradiation Microbial transformation of cyanide to ammonia or nitrate

Microbiological degradation

Gas-filled membranes

HCN transfer through a gas-filled hydrophobic microporous membranes to striping solution containing NaOH

Results are presented here for four real wastewater streams. Two of the wastewaters contained other volatile compounds (acrylonitrile or CH2 CHCN, and benzonitrile or C6 H5 CN) while the other two wastewater streams contained no other volatile components. Cyanide removal was studied using a 1000 L pilot scale plant operated in batch mode. The plant used 10 hollow fiber modules with a total membrane surface area of 180 m2 . Experimental results obtained here indicate that the transfer of other volatile components as well as water vapor is very likely to occur when using real wastewaters.

2. Theory The flux, J, of HCN from the feed to strip solution may be described by the following equation: J = K([HCN]F,t − [HCN]S,t ) ≈ K[HCN]F,t

(2)

Regeneration of resin is difficult since there are cyanide complexes besides free cyanides Not effective for removing the iron-cyanide complex Usually inadequate by itself and requires chemical treatment Avoids using toxic or hazardous chemicals. May not be possible to treat wastewaters containing high concentrations of cyanide. Process may be adversely affected by cold temperatures. Capital costs may be higher than for the oxidation processes. System response to a sudden change in cyanide or nutrient concentration may be sluggish Cyanide can be recovered. No secondary pollutants produced. Energy and chemical requirements are low. Simple operation

where K is the overall mass transfer coefficient and [HCN]F,t and [HCN]S,t are the HCN concentration in the bulk feed and strip solutions at time t, respectively. It is assumed that excess base is present, thus the HCN concentration in the strip solution, [HCN]S,t , is always essentially zero. Eq. (2) indicates that the HCN flux may be predicted if the overall mass transfer coefficient and the bulk HCN concentration in the feed solution are known. A mass balance around the feed reservoir results in the following differential equation: d([HCN]F,t VF,t ) = −KA([HCN]F,t − [HCN]S,t ) dt ≈ −KA[HCN]F,t

(3)

The left-hand side of Eq. (3) gives the rate of change of HCN in the bulk feed. VF,t is the volume of the feed solution at time t. The right-hand side of Eq. (3) gives the rate of transfer of HCN from the feed to the strip solution, where K is the overall mass transfer coefficient of HCN and A is the membrane

174

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

surface area. Again, since it is assumed there is excess base, [HCN]S,t is zero. In deriving Eq. (3), it is also assumed that the recirculation rate is fast relative to the rate of mass transfer [5]; consequently, the HCN concentration in the hollow fiber modules does not change significantly in one pass. If the osmolarities of the feed and strip solutions are different, water vapor will be transferred into the solution with higher osmolarity leading to a change of volume. The change in feed solution volume may be described by: VF,t = VF,0 + αt

(4)

where VF,0 is the initial feed volume, α is the average rate of water vapor transfer and t is the time. Usually, α is small therefore the volume of the feed solution in the hollow fibers does not change significantly in one pass. Substituting Eq. (4) into Eq. (3) and then integrating Eq. (3) from t = 0 to t gives:


 [HCN]F,t αt KA + α ln ln 1 + =− (5) [HCN]F,0 α VF,0 For |αt/VF,0 | < 1,  

αt αt 2 αt = − + ··· ln 1 + VF,0 VF,0 VF,0

(6)

Thus, plotting ln([HCN]F,0 /[HCN]F,t ) versus t should yield a straight line with slope (KA + α)/VF,0 . In the absence of water vapor transport, α = 0, and Eq. (7) is identical to the equation used in previous cyanide removal studies using artificial wastewaters [21–27]. In aqueous solution, HCN can dissociate as follows: pKa = 9.31

(8)

If the pH of the aqueous feed solution is much less than 9.31, HCN will be the dominant component. However, if the pH of the feed is more than 8.5, less than 85% of the cyanide will be present as HCN. Since HCN is the species that diffuses through the membrane pores, if the pH of the feed is higher than 8.5, the actual undissociated HCN concentration in the feed should be used. As analytical methods determine the total cyanide present [27], it is necessary to express the undissociated HCN concentration in terms of the total cyanide, [CN]T , present as: [CN]T = [HCN] + [CN− ]

(9)

Further, [H+ ][CN− ] = 10−9.31 [HCN]

[HCN] = [CN]T

(10)

10−pH 10−pH + 10−9.31

(11)

Eq. (11) should be substituted into Eq. (7) if the pH of the feed is above 8.5. For a non-reactive strip solution, the concentration of the solute in the strip solution cannot be ignored. Eq. (3) is rewritten below for a volatile component (VC) which does not react with the strip solution. For simplicity, the effects of water vapor transport are ignored, i.e. the feed and strip volumes, VF and VS are assumed to be constant: VF

d([VC]F,t ) = −KA([VC]F,t − [VC]S,t ) dt

(12)

where [VC]F,t and [VC]S,t are the concentration of volatile component (VC) in the feed and strip solutions, respectively. Initially, the strip solution is free of VC. Consequently, the concentration of VC in the strip solution is given by: [VC]S,t =

Further, for small αt/VF,0 , the series expansion in Eq. (6) may be truncated after the first term and substituted into Eq. (5) resulting in:

 [HCN]F,t KA + α ln =− t (7) [HCN]F,0 VF,0

HCN ↔ H+ + CN− ,

Combining Eqs. (9) and (10) leads to:

VF [VC]F,0 − VF [VC]F,t VS

(13)

where [VC]F,0 is the initial concentration of VC in the feed solution. Substituting Eq. (13) into Eq. (12) yields: VF

d([VC]F,t ) KA =− {(VF + VS )[VC]F,t − VF [VC]F,0 } dt VS (14)

Integrating (14) subject to the condition that at t = 0, [VC]F,t=0 = [VC]F,0 and [VC]S,t=0 = 0, gives:  
 VS [VC]F,0 VF + VS ln = KAt (VS + VF )[VC]F,t − VF [VC]F,0 VS V F (15) Thus, plotting ln{VS [VC]F,0 /((VS + VF )[VC]F,t − VF [VC]F,0 )} versus t should result in a straight line with slope ((VF + VS )/VS VF )KA. The overall mass transfer coefficient, K, is given by: 1 1 1 1 = + + K kF kM kS

(16)

where kF , kM and kS are the three individual mass transfer coefficients that describe the transfer of volatile components across the feed side concentration boundary layer, through the membrane and across the strip side concentration boundary layer, respectively [23]. In the studies conducted here, the feed stream always flowed inside the hollow fibers. Previous investigators have shown that the lumen side mass transfer coefficient may be estimated from the equation:
2 1/3 kF d d vF = 1.64 (17) DF lDF where d and l are the inside diameter and the length of the hollow fibers, respectively; DF is the diffusion coefficient of the

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

volatile species in water and vF is the velocity of the liquid phase through the hollow fiber lumen. Eq. (17) was originally developed by L´evˆeque [28] for heat transfer in tubes. Though numerous assumption are involved in the derivation of Eq. (17), see Castino and Wickramasinghe [29], the mass transfer analogue has been used by Kenfield et al. [23], Shen et al. [25–27] and Yang and Cussler [30] to predict the mass transfer coefficient inside the fibers. The membrane mass transfer coefficient may be predicted by the equation: kM =

εDM H δτ

(18)

where ε and δ are the void fraction and wall thickness of the hollow fiber membrane, τ is the tortuosity of the membrane pores, DM is the diffusion coefficient of the volatile species in the gas within the membrane pores and H is the partition coefficient which relates the concentration of volatile components in the gas phase to that in the liquid phase [23]. Since the porosity and thickness of the hollow fibers are constant in the experiments conducted here, changes in kM will be due to changes in the partition coefficient and diffusion coefficient of the volatile components. The tortuosity factor accounts for the pore geometry. Tortuosity factors ranging from 2 to 12 have been reported [3,31]. For non-reactive strip solutions, the following mass transfer correlation has been derived for flow outside and parallel to hollow fibers [30]:   kS de de 0.93 0.33 = 1.25 Re Sc DS l

(19)

where kS , de and DS are the mass transfer coefficient of the solute in the strip solution, the equivalent diameter for flow on the shell side and the diffusion coefficient of the solute in the strip solution. The equivalent diameter is defined as 4 × (cross-sectional area for liquid flow)/wetted perimeter. The Reynolds and Schmidt numbers (Re and Sc) are defined as vS de /ν and ν/DS , where vS is the average strip velocity and ν is the kinnematic viscosity. The average strip velocity was calculated by dividing the strip flow rate by the actual crosssectional area for shell side flow. Eq. (19) has been used for Reynolds number up to about 1000. The strip side mass transfer coefficient for a reactive strip solution is more difficult to predict. Cussler [31] has shown that the form of the mass transfer correlation depends on whether the reaction, for example, the reaction between HCN and NaOH, is fast or instantaneous. Astarita et al. [32] have shown that the presence of a chemical reaction may significantly accelerate the rate of mass transfer. For cyanide removal, Kenfield et al. [23] and Shen et al. [22] have shown that in the presence of excess NaOH, the strip side mass transfer coefficient is much larger than the feed side mass transfer coefficient. Thus, for cyanide removal, kS may be ignored

175

and the overall mass transfer coefficient is given by: 1 1 1 = + K kF kM

(20)

Unlike the transport of a volatile component, transport of water vapor involves transport of the solvent. Consequently, mass transfer resistances in the feed and strip solutions due to the presence of a concentration boundary layer will be negligible. For water vapor transport, only the membrane resistance needs to be considered. Consequently, 1 1 = K kM

(21)

We shall make use of these equations when analyzing the experimental data.

3. Experimental Four different cyanide containing wastewaters: acrylonitrile, benzonitrile, caffeine, and praziquantal were investigated. Table 3 gives further details of these wastewater streams. The strip solution consisted of 10% NaOH. Hydrophobic microporous polypropylene hollow fibers, 450 ␮m o.d., average wall thickness 60 ␮m, pore size 0.05–0.2 ␮m, and average porosity 35%, were used in this study. Table 4 gives further details of the hollow fibers. These hollow fibers were packed into module housings of two different dimensions as described in Table 5. The pilot scale experimental set up is shown in Fig. 2. The cyanide containing wastewater is pumped from the reserve tank into the wastewater feed tank through a polypropylene fine filter (pore size less than 5 ␮m). The strip solution is also pumped from a reserve tank through a polypropylene fine filter and then into the strip tank. These two fine filters were used to remove particles which may be present in the wastewater and strip solutions. From the feed tank, the wastewater is pumped through the lumen side of the hollow fiber modules at 15 m3 h−1 , while the strip solution is introduced into the shell side of the modules at the same flow rate. The strip and feed solutions are recycled back to their respective tanks. By assuming the density and viscosity of the feed solution are 103 kg m3 and 10−3 Pa s, respectively, the calculated Reynolds number in the lumen side of the hollow fiber modules is about 100. Consequently, laminar flow is assumed. Similarly, the calculated Reynolds number in the shell side of the hollow fiber modules is also about 100. Thus, Eq. (19) is applicable. The volume of the feed and strip solutions was 1000 and 500 L, respectively. The temperature was 18 ± 2 ◦ C. The hollow fiber membrane unit contained 20 modules divided into two groups of 10 housed in a stainless steel frame 2000 mm × 900 mm × 1500 mm. At any given time, one group of 10 modules was in operation while the other group was being cleaned or in standby mode.

176

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

Table 3 Details of the wastewaters tested Wastewater

[CN− ] (mg L−1 )

[CN− ] (mol L−1 )

Color

Acrylonitrile

500–1500

0.019–0.058

Brown

Benzonitrile Caffeine Praziquantal

3000–4000 500–2000 2000–3500

0.115–0.154 0.019–0.077 0.077–0.135

Light yellow Colorless Reddish

a

pH

Salinity (%)

6–7

1–2

9.5–10 2–3 9.5–10

Water vapor pressure at 20 ◦ C (Pa)a 2290

25–30 <1 <1

1600 ∼2330 ∼2330

Other major volatile component

Comments

CH2 CHCN

CH2 CHCN concentration around 5200 mg L−1

C6 H5 CN, NH3 , C6 H6 None None

Vapor pressure of deionized water at 20 ◦ C is 2333 Pa.

Fig. 2. Pilot scale experimental set up.

Table 4 Details of the hollow fibres Outside diameter (␮m) Wall thickness (␮m) Pore size (␮m) Porosity (%) Gas penetrability (cm3 /cm2 s cmHg) Breakthrough pressure (MPa) Manufacturer

450 50–70 0.05–0.2 30–40 7.2 0.25 Hangzhou Hualu Membrane Engineering Co. Ltd., Zhe Jiang Province, PR China

During operation, 30–50 mL samples of the wastewater and strip solution were removed and the cyanide concentration was determined. The cyanide concentration was measured using silver nitrate titration for concentrations

above 10 mg L−1 , and colorimetric analysis for concentrations below 10 mg L−1 [33]. In addition, the pH in both solutions was also determined. Additional laboratory scale tests were conducted using smaller hollow fiber modules which contained 200 polypropylene fibers potted in a glass tube (see Table 5). The effective fiber length was 18 cm, the membrane area 452 cm2 and the packing density 30%. The experimental set up for the laboratory scale experiments was similar to that used at the pilot scale. The volume of both the feed and strip solutions was 500 mL. The flow rate of both solutions through the hollow fiber module was 180 mL min−1 . Laboratory scale modules were used to determine the water flux across the membrane due to osmotic distillation. The feed stream in these experiments consisted of deionized

Table 5 Details of hollow fiber modules No.

Hollow fiber

Module

Housing material

Comments

Used in pilot scale experiments. 10 modules are used in this study in parallel Used in laboratory scale experiments

Length (cm)

Number

Internal diameter (cm)

Length (cm)

A

82.5

16000

8

112

ABS

B

18

32

Glass

200

1.2

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

177

Fig. 3. Determination of the overall mass transfer coefficient of cyanide for the four wastewaters.

Fig. 5. Determination of the overall mass transfer coefficient of acrylonitrile and cyanide for the acrylonitrile wastewater.

water. A range of NaOH concentrations was used in the strip solution. The change in volume of the feed solution was determined as a function of time. Laboratory scale experiments were also conducted in order to determine the rate of acrylonitrile transfer from the feed to the strip solution. In these experiments, the feed consisted of the acrylonitrile wastewater while the strip solution was 10% NaOH. Acrylonitrile does not react with the strip solution.

does not react with the strip solution, transfer stops once the concentration in the feed and strip solutions is equal. Cyanide on the other hand continues to transfer to the strip solution. Results for cyanide and acrylonitrile transfer are shown in Fig. 5 by filled diamonds and open triangles, respectively. Results for cyanide are plotted as ln([HCN]F,0 /[HCN]F,t ) versus t and should be read using the left-hand side y-axis. Results for acrylonitrile are plotted as ln{VS [VC]F,0 /((VS + VF )[VC]F,t − VF [VC]F,0 )} versus t (see Eq. (15)) and should be read using the right-hand side y-axis. For cyanide, the results fall on a straight line, the slope of which is (KA + α)/VF,0 . For acrylonitrile, the results for short times are fitted to a straight line which has a slope equal to ((VF + VS )/VS VF )KA (see Eq. (15)). As the acrylonitrile concentration in the strip solution approaches that in the feed solution, the rate of transfer will become very slow. Since the rate of change in acrylonitrile concentration in the feed solution will be very small, using data for longer times to calculate the mass transfer coefficient will lead to large errors. Thus, in Fig. 5, only the initial period is used for determining the acrylonitrile mass transfer coefficient.

4. Results Fig. 3 gives results obtained from the pilot plant for the four different industrial wastewaters plotted as ln([HCN]F,0 /[HCN]F,t ) versus t. As can be seen, the data fall on four different straight lines. As indicated by Eq. (7), the slope is equal to (KA + α)/VF,0 . Fig. 4 shows the variation of cyanide and acrylonitrile concentration with time. Results are given for laboratory scale experiments for the acrylonitrile wastewater. Since acrylonitrile

Fig. 4. Variation of acrylonitrile and cyanide concentration in the acrylonitrile wastewater.

178

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

Fig. 6. Experimental and predicted water vapor flux.

Fig. 6 gives the water vapor flux as a function of partial pressure difference. A semi-logarithmic scale is used as the water vapor flux varies over two orders of magnitude. Open squares indicate experimental results while filled squares indicate predicted fluxes using Eq. (18) and converting the partial pressure difference to a concentration difference by the ideal gas law. In Eq. (18), the tortuosity was used as a fitting parameter. A tortuosity of 9 gave the best fit to the data. Zhang and Cussler [3] and Cussler [30] indicate that tortuosity factors between 2 and 12 have been reported. Table 6 gives the water vapor flux for the laboratory scale experiments and the pilot plant. For the laboratory scale experiments, the water vapor flux obtained for a 10% NaOH strip stream is given. For the pilot plant, the water vapor flux for the caffeine feed stream is given. Since these wastewaters had very low salinity, the water vapor pressure is similar to that of deionized water (see Table 3). As can be seen, water vapor fluxes for the laboratory scale experiments and the pilot plant are in close agreement.

Table 6 also gives values of α (water vapor flux multiplied by the membrane area). The fourth column of Table 6 gives values of KA, where K is the overall mass transfer coefficient of cyanide in the caffeine wastewater obtained from Fig. 3 assuming α can be ignored; and A is the total membrane surface area. The fifth column gives the ratio of KA to α. As can be seen, α is more than 100 times smaller than KA. Thus, ignoring the change in volume of the feed solution due to the water vapor flux will have little effect on the experimentally determined mass transfer coefficient. The experimentally determined mass transfer coefficients are given in Table 7. Results are given for the four wastewaters obtained from pilot plant runs. The mass transfer coefficients for cyanide, acrylonitrile and water vapor obtained from the laboratory scale experiments are also included. In Table 6, α is taken to be zero. Further, for the benzonitrile and praziquantal wastewaters, as the pH was higher than 8.5, Eq. (11) was used to determine the actual HCN concentration present. Mass transfer coefficients obtained from pilot plant runs and laboratory scale experiments are in good agreement.

Table 6 Comparison of KA and α Pilot scale Laboratory scale

Water vapor flux (J, m s−1 )

Volume change (α, m3 s−1 )

KA of cyanide (m3 s−1 )

KA/α

Comments

1.85 × 10−8

3.3 × 10−6

1.96 × 10−3 3.8 × 10−7

590 640

For caffeine wastewater NaOH concentration in the strip side was 10%

1.72 × 10−8

7.78 × 10−10

Table 7 Individual and overall mass transfer coefficients Component

kF

kM

kS

KC

KE

((KE − KC )/KE ) × 100%

Comments

HCN

2.31 × 10−5

3.53 × 10−5 3.30 × 10−5

– –

1.39 × 10−5

1.11 × 10−5 1.09 × 10−5 1.03 × 10−5 0.61 × 10−5 0.47 × 10−5

−25 −6 −13 −90 −147

Laboratory Pilot, praziquantal Pilot, caffeine Pilot, benzonitrile Pilot, acrylonitrile

Acrylonitrile

2.86 × 10−5

2.67 × 10−5

9.57 × 10−6

5.65 × 10−6

6.08 × 10−6

7

Laboratory

–

2.25 × 10−7

–

2.25 × 10−7

2.47 × 10−7

9 −3

Laboratory Pilot

Water vapor

1.80 × 10−5

1.16 × 10−5

2.18 × 10−7

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

Calculated individual and overall mass transfer coefficients are also given in Table 7. A tortuosity factor of 9 was used in all cases. As can be seen, except for the benzonitrile and acrylonitrile wastewaters which contain other volatile components, the percentage error between the experimental and calculated overall mass transfer coefficients ranged from 3 to 25% indicating good agreement. 5. Discussion For wastewaters containing only cyanide as the volatile component, the experimentally determined overall mass transfer coefficients for cyanide are in good agreement with the predicted mass transfer coefficients. Further, Table 7 indicates that the individual feed and membrane mass transfer coefficients are similar in magnitude. Consequently, both contribute to the overall mass transfer coefficient. Unlike many other membrane separation processes, providing the membrane is hydrophobic, the membrane material has no effect on the rate of mass transfer. Only the membrane porosity and tortuosity affect the rate of mass transfer. Shen et al. [27] show that the diffusion and partition coefficient of the transported species depend on the stagnant gas that fills the membrane pores which in turn can lead to different rates of transport. In the experiments conducted here, the pores are assumed to be air-filled. The results obtained here indicate that other volatile species will also be transferred into the strip solution. Further, the transfer of other volatile components decreases the rate of cyanide transfer. The transfer of acrylonitrile may be predicted using mass transfer correlations that are available in the literature. As acrylonitrile does not react with the strip solution, transfer of acrylonitrile ceases when the concentration in the feed and strip solutions is equalized. Osmotic distillation can lead to water vapor transport from the aqueous solution of lower osmolarity to the one of higher osmolarity. For the four wastewaters studied here, transfer was from the feed to strip solution except for the benzonitrile wastewater where the direction of water vapor transport was reversed. The benzonitrile wastewater had high salinity leading to a significantly lowered water vapor pressure. A simple method to account for the change in feed volume, when determining the overall mass transfer coefficient (see Eq. (7)), is presented. The results obtained here indicate this effect is insignificant. However, osmotic distillation will lead to changes in the feed and strip volumes which need to be considered when sizing the feed and strip tanks. The presence of other volatile components such as acrylonitrile, could lead to multicomponent diffusion effects in the liquid and gas phases. Multicomponent diffusion may be described by generalizing the binary diffusion equation to an n-component system as [33]: −Ji =

n−1  j=1

Dij ∇cj

(22)

179

where the Dij are multicomponent diffusion coefficients. The diagonal terms (the Dij,i=j ) are called the “main term” diffusion coefficients while the off diagonal terms (the Dij,i=j ) are called the “cross-term” diffusion coefficients. In general, the diffusion coefficients are not symmetric, i.e., Dij = Dji . However, estimating these multicomponent diffusion coefficients is limited to some very simple cases, such as ideal gas mixtures. For most of systems encountered in practice, such as the systems studied here, estimation of the multicomponent diffusion coefficients is difficult. According to empirical rules, multicomponent diffusion will be important in cases where, the components show strong thermodynamic interactions, have very different molecular weights, the solutions are not dilute, or one solute gradient is much larger than the others [34]. In the gas phase, Thomas indicates that interactions between cyanide and acrylonitrile and water vapor may be possible [35,36]. In the liquid phase for the acrylonitrile wastewater, the average initial cyanide concentration was 562 mg L−1 while that of acrylonitrile was 5212 mg L−1 , about 10 times higher. Thus, the transfer of acrylonitrile may affect the rate of transfer of cyanide. The mass transfer coefficients for benzonitrile and acrylonitrile are much lower than predicted. Similar deviations between predicted and experimental values have been observed in the past for artificial wastewaters [4,23]. For the real wastewaters, investigated here, it was assumed that the hollow fiber pores were air-filled. The transfer of other volatile species that do not react with the strip solution could result in a significant fraction of the pore volume being no longer air-filled. This could lead to altered diffusion and partition coefficients for cyanide resulting in a lower mass transfer coefficient within the membrane pores [27]. In addition, Table 3 indicates that both the benzonitrile and acrylonitrile wastewaters were highly colored. Insoluble particulate matter could block the membrane pores again leading to lower rates of mass transfer. The analysis we present here could be used when designing a pilot scale facility based on laboratory scale data. The membrane tortuosity was used as an adjustable parameter to fit the predicted water vapor flux data to the experimentally determined fluxes. The same tortuosity factor gave good agreement between the experimental and plant data for cyanide and acrylonitrile fluxes. Since tortuosity factors depend on membrane morphology, they are difficult to predict. However, the tortuosity factor determined here lies within the range of previously reported values. In designing a pilot scale facility, it is important to include the effects of transport of other volatile components and water vapor.

6. Conclusion Cyanide removal from wastewater streams using gas membranes has been studied in a pilot scale plant. The strip stream consisted of 10% NaOH. The feed side and membrane mass transfer coefficients both contribute to the overall mass

180

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181

transfer coefficient. The overall mass transfer coefficient for cyanide can be predicted using empirical correlations from the literature. Other volatile species present in the feed solution will also transfer to the strip solution. Transfer of other volatile species could affect the rate of cyanide transfer. Water vapor transport due to osmotic distillation will occur if the osmolarity of the aqueous feed and strip solutions is different. The effects of osmotic distillation should be considered when sizing the strip and feed tanks. Nomenclature A c d de D H J k K l t v V

membrane surface area concentration inside diameter of hollow fibers equivalent diameter for flow in the shell side diffusion coefficient partition coefficient flux individual mass transfer coefficient overall mass transfer coefficient length of hollow fibers time velocity volume

Greek letters α average rate of water vapor transfer δ membrane thickness ε membrane porosity ν kinnematic viscosity of liquid τ tortuosity of membrane pores Subscripts C calculated E experimental F feed i, j component M membrane S strip solution t time 0 initial Dimensionless numbers Re Reynolds number, vde /ν Sc Schmidt number, ν/D

References [1] M. Imai, S. Furusaki, T. Miyauchi, Separation of volatile materials by gas membranes, Ind. Eng. Chem. Process Des. Dev. 21 (1982) 421.

[2] Q. Zhang, E.L. Cussler, Hollow fiber gas membranes, AIChE J. 31 (1985) 1548. [3] Q. Zhang, E.L. Cussler, Microporous hollow fibers for gas absorption. II. Mass transfer across the membrane, J. Membr. Sci. 23 (1985) 333. [4] M.J. Semmens, D.M. Foster, E.L. Cussler, Ammonia removal from water using microporous hollow fibers, J. Membr. Sci. 51 (1990) 127. [5] K.K. Sirkar, Other new membrane processes, in: W.S.W. Ho, K.K. Sirkar (Eds.), Membrane Handbook, van Nostrand Reinhold, New York, NY, 1992. [6] U.S. Environmental Protection Agency, Treatment of cyanide heap leaches and tailings, Washington, DC, EPA 530-R-94-037, 1994. [7] U.S. Environmental Protection Agency, Code of federal regulations: effluent guidelines and standards; electroplating point source category pretreatment standards for existing sources; final rule and amendments, Washington, DC, Title 40 CFR Part 413, 1981. [8] S.Q. Hassan, M.P. Vitello, M.J. Kupferle, D.W. Grosse, Treatment technology evaluation for aqueous metal and cyanide bearing hazardous wastes (F007), J. Air Waste Manage. 41 (1991) 710. [9] D. Pak, W. Chang, Oxidation of aqueous cyanide solution using hydrogen peroxide in the presence of heterogeneous catalyst, Environ. Technol. 18 (1997) 557. [10] P.A. Riveros, D. Koren, V.M. McNamara, J. Binvignat, Cyanide recovery from a gold mill barren solution containing high levels of copper, CIM Bull. 91 (1998) 73. [11] R.G. Kunz, J.F. Giannelli, Activated carbon adsorption of cyanide complexes and thiocyanate ion from petrochemical wastewaters, Carbon 14 (1976) 157. [12] N. Adhoum, L. Monser, Removal of cyanide from aqueous solution using impregnated activated carbon, Chem. Eng. Process. 41 (2002) 17. [13] L. Monser, N. Adhoum, Modified activated carbon for the removal of copper, zinc, chromium and cyanide from wastewater, Sep. Purif. Technol. 26 (2002) 137. [14] E. Goldblatt, Recovery of cyanide from waste cyanide solutions by ion exchange, Ind. Eng. Chem. 48 (1956) 2107. [15] H. Kurama, T. Catalsarik, Removal of zinc cyanide from a leach solution by an anionic ion-exchange resin, Desalination 29 (2000) 1. [16] M.D. Gurol, W.M. Bremen, Kinetics and mechanism of ozonation of free cyanide species in water, Environ. Sci. Technol. 19 (1985) 804. [17] F. Novak, G. Sukes, Destruction of cyanide waste-water by ozonation, Ozone Sci. Eng. 3 (1981) 61. [18] B.R. Kim, D.H. Podsiadlik, E.M. Kalis, J.L. Hartlund, W.A. Gaines, Photochemical destruction of cyanide in landfill leachate, J. Environ. Eng. ASCE 124 (1998) 1108. [19] Y.B. Patil, K.M. Paknikar, Biodetoxification of silver-cyanide from electroplating industry wastewater, Lett. Appl. Microbiol. 30 (2000) 33. [20] A.E. Short, The GM-IX process: a pilot plant for recovering zinc cyanides from plating rinsewaters, M.S. thesis, University of Minnesota, Minneapolis, MN, 1990. [21] M.J. Semmens, C.F. Kenfield, R. Qin, A gas membrane-ion exchange process for cyanide recovery, Met. Finish. 11 (1987) 47. [22] Z. Shen, J. Huang, G. Qian, Recovery of cyanide from wastewater using gas-filled membrane absorption, Water Environ. Res. 69 (1997) 363. [23] C.F. Kenfield, R. Qin, M.J. Semmens, E.L. Cussler, Cyanide recovery across hollow fiber gas membranes, Environ. Sci. Technol. 22 (1988) 1151. [24] A.E. Short, S.F. Haselmann, M.J. Semmens, The GM-IX process: a pilot study for recovering zinc cyanides, J. Environ. Sci. Health, Part A: Toxic/Hazard. Subst. Environ. Eng. 32 (1997) 215.

B. Han et al. / Journal of Membrane Science 257 (2005) 171–181 [25] Z. Shen, J. Huang, L. Sun, The treatment of praziquantal wastewater using coagulation and membrane separation integrated process, Chin. J. Environ. Pollut. Technol. 12 (1999) 101. [26] Z. Shen, G. Qian, M. Wang, J. Chen, N. Wang, The thermal efficiency in the gas membrane process, Chin. J. Membr. Sci. Technol. 18 (1998) 46. [27] Z. Shen, B. Han, S.R. Wickramasinghe, Cyanide removal from wastewater using gas membranes: pilot-scale study, Water Environ. Res. 76 (2004) 15. [28] A. L´evˆeque, Les lois de la transmission de chaleur par convection, Ann. Mines 13 (1928) 201. [29] F. Castino, S.R. Wickramasinghe, Washing frozen red blood cell concentrates using hollow fibres, J. Membr. Sci. 110 (1996) 169. [30] M.-C. Yang, E.L. Cussler, Designing hollow fiber contactors, AIChE J. 32 (1986) 1910.

181

[31] E.L. Cussler, Diffusion: Mass Transfer in Fluid Systems, Cambridge University Press, New York, NY, 1984. [32] G. Astarita, D.W. Savage, A. Bisio, Gas Treating with Chemical Solvents, Wiley, New York, NY, 1983. [33] American Public Health Association (APHA), Standard Methods for the Examination of Water and Wastewater, 20th ed., APHA, Washington, DC, 1998. [34] E.L. Cussler, Multicomponent Diffusion, American Elsevier Scientific Publishing Company Inc., New York, NY, 1976. [35] R.K. Thomas, Hydrogen bonding in the gas phase: the thermodynamic properties of hydrogen fluoride-ether complexes and their far infrared spectra, Proc. R. Soc. Lond. A: Mater. 322 (1971) 137. [36] R.K. Thomas, Hydrogen bonding in the gas phase: the infrared spectra of complexes of hydrogen fluoride with hydrogen cyanide and methyl cyanide, Proc. R. Soc. Lond. A: Mater. 325 (1971) 133.