Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

Chemical Engineering Journal 210 (2012) 33–44

Contents lists available at SciVerse ScienceDirect

Chemical Engineering Journal


journal homepage: www.elsevier.com/locate/cej

Response surface-optimized Fenton’s pre-treatment for chemical precipitation


of struvite and recycling of water through downstream nanofiltration
Ramesh Kumar, Parimal Pal ⇑
Environment and Membrane Technology Laboratory, Chemical Engineering Department, National Institute of Technology, Durgapur 713 209, India

h i g h l i g h t s

" Optimization of Fenton’s process done by response surface methodology (RSM).


" 98% recovery of ammonia as struvite fertilizer achieved in chemical precipitation.
" Integration of flat sheet cross flow membrane modules turned water reusable.
" Novel approach produced struvite free from cyanide and phenol contamination.

a r t i c l e i n f o a b s t r a c t

Article history: Experimental investigations were carried out on a new membrane-integrated hybrid treatment scheme
Received 21 May 2012 for removal of hazardous cyanide and phenol from coke wastewater and chemical precipitation of ammo-
Received in revised form 8 August 2012 nia as a value-added struvite by-product. Fenton’s pre-treatment with response surface optimization
Accepted 14 August 2012
ensured very effective removal of cyanide and phenol facilitating subsequent chemical precipitation of
Available online 28 August 2012
over 98% ammonia as pure struvite fertilizer. Integration of micro and nano-membrane filtration with
these chemical treatment units raised water quality to reusable level. Largely fouling-free flat sheet cross
Keywords:
flow membrane modules could remove over 96–98% of the chemical contaminants while still yielding a
Chemical precipitation
Response surface optimization
high flux of 116 l/h/sq m of the membrane surface at a transmembrane pressure of only 15–16 bars. The
Struvite recovery new integrated scheme embracing the concept of zero emission and value addition is a novel approach in
Recycling continuous treatment of industrial wastewater for recycling.
Zero emission Ó 2012 Elsevier B.V. All rights reserved.

1. Introduction generated, biological treatments like activated sludge process, nitri-


fication, denitrification and similar treatments are often considered
In the iron and steel making industries around the world, mil- as low cost options. But high concentration of certain pollutants like
lions of tonnes of metallurgical coke are used in the blast furnace cyanide and phenol often leads to failure of biological treatment
and the coke plants produce this coke by carbonization of coal. After plants [1]. Though individual biodegradation of phenol, cyanide
carbonization, enormous quantity of water is used for quenching (CN), thiocyanate (SCN) and ammonia–nitrogen (NHþ 4  N) has
the hot coke mass and for washing the coke-oven gases. Liquid efflu- been found to be effective, their combined presence in a wastewater
ents thus generated during coke making are heavily polluted with stream, however, turns successful treatment a difficult task [2,3].
cyanide, phenol, ammonia and varieties of other organic as well as Metal-complexes like ferricyanide are very difficult to degrade
inorganic pollutants that can be broadly expressed in terms of through biological methods, so it remains in the effluent [4].
COD (Chemical Oxygen Demand) and BOD (Biochemical Oxygen Conventional methods like caustic chlorine treatment can convert
Demand). This wastewater on being discharged from industry cyanide into cyanate but this is also toxic. Moreover, chlorination
pollutes hundreds of kilometer of river bodies affecting both aquatic also produces secondary by-products such as trihalomethanes and
lives and human health seriously. Effective treatment of coke chloro-phenols, which are highly toxic and suspected carcinogens
wastewater has for long remained a challenge to the scientific and are also less effective for the removal of iron cyanide [5].
community. Considering the enormous quantity of the wastewater Simultaneous presence of phenol and cyanide opposes each other’s
removal [6]. The other physico-chemical processes that have been
⇑ Corresponding author. Tel.: +91 343 2755055, mobile: +91 9434469750; fax: studied include stripping of ammonia after removal of cyanide
+91 343 2754078. and phenol. But this adds to air pollution and is often expensive
E-mail addresses: parimalpal2000@yahoo.com, parimalpaul1999@yahoo.in due to involvement of chemical feed, stripping tower, pump and
(P. Pal).

1385-8947/$ - see front matter Ó 2012 Elsevier B.V. All rights reserved.
http://dx.doi.org/10.1016/j.cej.2012.08.036
34 R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44

Nomenclature

Jw the pure water permeability F Faraday’s constant


DP the difference in applied pressure across the membrane R the gas constant
rp pore radius T temperature
d the thickness of the membrane W the electric potential
l viscosity of fluid Ki the distribution coefficient of solute i
Cis feed side concentration of solute Jv the volume flux and can be estimated based on the
Cip permeate side concentration of solute membrane area
Ji solute flux i Ci the bulk (or feed) concentration of solute i
Di diffusivity of solute i WD Donnan potential
ci the concentration of solute i at the surface of the mem- Ws electrical potential of the solution
brane Wm electrical potential of the membrane
x mole fraction for solute i Ci initial concentration of the pollutants
zi the valance of solute i Cf residual concentration in the permeate side

liquid spray system, forced air and carbonization system [7]. Ad- limit) cannot be determined, but in RSM with the help of central
vanced oxidation processes (AOPs) using ultraviolet (UV) radiation composite design a fixed range can be taken. It is very difficult in
in combination with various oxidants (viz. Fenton’s reagent) have case of classical factorial design to develop a mathematical equa-
also been used for the treatment of effluents containing toxic com- tion for interaction of parameters which is possible in case of
pounds like cyanide, phenol, etc. which are refractory in nature and RSM. RSM can provide optimized value in less number of experi-
difficult to be oxidized completely by conventional methods [8]. Ul- ments for large number of variable parameters which is not possi-
tra-violet radiation (UV), Ozone (O3), hydrogen peroxide (H2O2) and ble in case of classical factorial optimization. Quite a few studies
their various combinations as well as photo-Fenton methods [9] [15–17] have attempted to recover nutrients in the form of struvite
broadly fall in the category of advanced oxidation process. However, (magnesium ammonium phosphate) from wastewater in order to
the major drawbacks of the AOPs are the high cost of the chemical add economy to the process. But in most of these cases, studies
reagents and expenses on operation and maintenance of special were on municipal wastewater or poultry wastewater and were lar-
light sources such as ultra-violet lamps [10]. In the recent years, gely in batch mode. Coke wastewater containing so many hazard-
membrane-integrated approaches have been adopted for treatment ous pollutants has not received adequate attention in this
of industrial wastewater to achieve high quality of effluent [11,12]. context. In complex steel making industry, a treatment plant of
Though biological treatment plants have most widely been stud- coke wastewater often fails to attract desired attention as it does
ied and practised in handling coke wastewater, such plants often not directly produce any product and the conventional treatment
fail due to inadequate attention that largely result from non-reve- steps are time-consuming. In this fast changing world, there is ‘no
nue activities in waste treatment plants. Thus, in the present article, time for waste’ and hardly the existing treatment schemes are in
a novel approach has been adopted to address these issues. The first tune with the ‘zero emission’ concept. Thus there needs to be inno-
and the most important step is the optimization of Fenton’s meth- vative approaches in turning waste treatment plants economically
od. In conventional optimization study, effect of one governing attractive, technologically innovative and operationally interesting
parameter on the response of interest is normally found out keep- with incorporation of ‘zero emission’ concept and value addition to
ing other parameters constant. But in a real situation, there may chemical compounds present as pollutant. The present integrated
be mutual interactions of the control variables and such interac- treatment scheme is an attempt in simultaneously turning waste-
tions cannot be taken care of in conventional optimization study. water reusable while converting chemical pollutant into value-
Response surface methodology is an experimental strategy that added product. Such a scheme is expected to result in zero emission
develops a fairly realistic functional relationship between the re- and enhanced economic viability towards sustainable management
sponse of interest and the controlling or governing variables of water resource.
through empirical statistical modeling. It takes care of mutual inter-
actions of the variables through introduction of the interaction 2. Theoretical
terms in the developed model. Though other approaches are also
there for process optimization, there are always limitations of each Cyanide and phenol can be degraded by chemical method (Fen-
approach in matters of involvement of controlling variables, num- ton’s reagent) and ammonia may be precipitated out by addition of
ber of experiments, practical difficulty of conducting extremely magnesium and phosphate salts. The principle of the Fenton’s pro-
large number of experiments, probability of introduction of large cess is the catalytic cycle of the reaction between iron (catalyst)
errors through inadequate number of experiments, etc. Considering and hydrogen peroxide (oxidant) to produce hydroxyl radicals
the number of variables involved and region of response in the [18]. Hydrogen peroxide is used in broad spectrum for the chemi-
present investigation, the response surface optimization approach cal oxidation of the dissolved organic and inorganic compounds
of Design Expert Software (Design-ExpertÒ Version 8.0.6) was used present in the industrial effluents as H2O2 is strong oxidizing agent
in the experimental strategy. The conventional technique for the (oxidation potential 1.77). The hydroxyl radical (OH) can destroy
optimization of multivariate system usually defines one factor at structure of organic synthesis because of strong oxidation. Hydro-
a time. RSM is very useful for developing, improving and optimizing gen peroxide is a very weak acid and liberates molecular oxygen in
processes by utilizing its mathematical and statistical tools [13]. It the presence of trace alkali. Iron acts as a catalyst and its addition
has unique ability to generate the relationships between the re- results in a positive impact on the removal of pollutants. However,
sponse and the independent variables with a minimum number increasing the iron concentration may also increase downstream
of trials according to special experimental design. Moreover, RSM problems associated with the metal’s removal. Fenton’s treatment
is less expensive and faster than the classical methods [14]. In clas- has two distinct stages [19] namely Fenton’s oxidation (OH
sical factorial design, range of a parameter (upper limit and lower generation) and Fenton’s coagulation. Ferric coagulation following
R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44 35

oxidation results into sludge. OH radical generation in Fenton’s Ionic transport through nanofiltration membranes has widely
treatment takes place following the reaction below: been explained by models based on Donnan equilibrium theory
that states the equality of electrochemical potential (W) of the
H2 O2 þ Fe2þ ! Fe3þ þ OH þ OH ð1Þ solutions on either side of the solution–membrane interface. The
2+ concentration at the surface of the membrane can be estimated
There is a possibility of auto regeneration of Fe in this system
that acts as catalyst. using the Donnan equilibrium as shown in the following equation:
 
Feþ3 þ H2 O2 ! HO2 þ Feþ2 þ Hþ ð2Þ F
WD ¼ Wm  Ws or C i ¼ C i exp zi WD ð10Þ
RT
Dosing of the hydrogen peroxide concentration results in the in-
creased availability of the hydroxyl radicals (OH) in the solution
where Ci is the bulk (or feed) concentration of solute i; Donnan po-
increasing the rate of oxidation of cyanide compound which is first
tential (WD) is the difference between the electrical potential of the
oxidized to cyanate and then further oxidized to ammonium and
solution (Ws) and the electrical potential of the membrane (Wm).
carbonate ions through the reactions below:
The Nernst-plank equation coupled with the Donnan Equilibrium
CN þ H2 O2 ! CNO þ H2 O ð3Þ has been shown to accurately predict the rejection of various salts
by nanofiltration and reverse osmosis membrane [23,24]. Eqs.
H2 O2
CNO þ 2H2 O ! NHþ4 þ CO2
3 ð4Þ (10) and (11) show that solute concentrations at the surface of
Phenol can be degraded following the reaction with hydrogen per- the membrane as well as charge on the membrane surface deter-
oxide in the presence of iron as a catalyst as below: mine retention or passage of the solutes. If the ion is neutrally
charged, then the concentration at the membrane surface is unaf-
C6 H6 O þ 14H2 O2 ! 6CO2 þ 17H2 O ð5Þ fected by the charge of the membrane [23,24] and separation lar-
Phenol oxidation produces cyclic intermediates hydroquinone gely follows steric mechanism.
(HQ), and catechol (CTL) and a strong brown color is obtained in At pH values greater than 7.0 most of the polyamide composite
presence of H2O2 that could be due to quinine condensation [20]. nanofiltration membranes possess negative zeta potential [25].
This ammoniacal nitrogen was removed by chemical precipita- Due to the charged nature of the NF membrane, solutes with an
tion as magnesium ammonium phosphate (MAP) well known as opposite charge compared to the membrane (counter-ions) are at-
struvite. MAP is a white crystal substance consisting of equal molar tracted, while solutes with a similar charge (co-ions) are repelled.
concentrations of magnesium, ammonium and phosphorus. MAP
precipitation takes place following the chemical reaction is ex-
3. Experimental
pressed in the following equation:
Mgþ2 þ NHþ4 þ PO3
4 þ 6H2 O () MgNH4 PO4  6H2 O #; 3.1. Raw wastewater and characterization
pK s ¼ 12:6 ð6Þ
Nanofiltration membranes are in between ultrafiltration and re- Raw wastewater (influent) used in the experiments was col-
verse osmosis membranes in separation performance that can be lected from a coke oven plant (Durgapur Projects Limited, India).
predicted in terms of pure water flux and solute flux which are Table 1 presents the range and their mean value of major parame-
uncoupled. Separation of the solutes from the solution depends ters of this wastewater.
both on steric (sieving) mechanism as well as Donnan (electro-
static) exclusion principle [21] mechanisms. The steric mechanism 3.2. Materials
may be explained by continuum hydrodynamic models of Ferry
[22] whereas Donnan mechanism is well explained by extended All the chemical reagents like hydrogen peroxide, ferrous sul-
Nernst–Planck model. fate, magnesium chloride, sodium biphosphate, hydrogen chloride,
The pure water flux may be related to pressure using the sodium hydroxide, cyanide standard solution (1000 mg/L), Metha-
Hagen–Poiseuille equation, as shown in (Eq. (8)) [23]. nol (Sigma–Aldrich, WI, USA) for calibrating phenyl column and
r2p Dp phenol standard (Merck) were from Merck (Germany). All working
Jv ¼ ð7Þ solutions of required concentration were prepared using deionised
8ld
water from Milli-Q purification system (Waters, MA, USA). Thin
According to Eq. (8), increasing the pressure will increase the film composite polyamide nanofiltration (NF1, NF2, NF3, NF20)
pure water flux, while the solute flux is proportionally related to as well as poly vinylidene fluoride (PVDF) microfiltration mem-
the solute concentration gradient across the membrane. In com- branes were procured from Sepromembranes Inc. and Membrane
puting effective pressure difference between two sides of the solution (USA) respectively. The polyamide composite nanofiltra-
membrane, osmotic pressure may be taken care of using Van’t Hoff tion membranes of 165 lm thicknesses were all very effective in
equation Eq. (9). rejection of bivalent salts while having poor retention for sodium
X
Dp ¼ RT ðC is  C ip Þ ð8Þ
Table 1
where Cis, Cip are, respectively feed and permeate side concentra- Characteristics of industrial wastewater (coke-oven).
tions of the solute.
Parameter Concentration range (mg/L) Mean value (mg/L)
Solute flux is defined using in the extended Nernst–Planck
equation as shown in the following equation: Cyanide 110–138 124
    Phenol 122–202 162
dci F dw Ammonia 1720–2160 1940
J w ¼ ½K i ci J v   Di  D i zi c i ð9Þ
dx RT dx COD 3436–4276 3856
TDS 20078–28162 24120
The above extended Nernst–Planck equation describes the flux Conductivity (mS/cm) 16.06–22.18 19.12
which is the sum of the fluxes due to convection, diffusion (as a Salinity 8.35–15.05 11.7
Oil and grease 40.5–61.5 51
function of concentration gradient across the membrane) and elec-
pH 8.5–9.5 9.0
tromigration (as a function of electrical charge gradient).
36 R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44

chloride as per manufacturers’ data and detail characteristics pre- temperature for duration of 1 h. Investigations on degradation of
sented in Table 2. phenol and cyanide following Fenton’s process were done in con-
tinuous system subsequently with the optimized parameters. The
3.3. Experimental set-up pH of the treatment medium was adjusted by addition of sulfuric
acid (H2SO4, 2 M) and sodium hydroxide (NaOH, 1 M). After Fen-
The experimental set-up as shown in Fig. 1 consists of a 30 l ton’s treatment, coke wastewater was passed to the next continu-
capacity continuous stirred tank reactor (CSTR) for Fenton’s treat- ous stirred tank reactor (CSTR) where magnesium and phosphate
ment, a 30 l capacity CSTR for precipitation of MAP (Magnesium compounds were added at optimum doses as described in results
Ammonium Phosphate), flat sheet cross flow microfiltration and Section 4.2. The slow releasing fertilizer magnesium ammonium
nanofiltration membrane modules and circulation pumps along phosphate (MAP) which is popularly known as struvite resulted
with flow meters and pressure gauges. A circulation bath with from the reactions of ammonia of wastewater and the added mag-
temperature controller (PolyScience, USA) was attached to the feed nesium and phosphate compounds. MAP got largely settled in the
tank to maintain constant temperature. The effective filtration area reactor itself as well as in the two downstream settling tanks.
of each membrane module was100 cm2. Clean overflow water from the first settling tank was microfiltered
under low pressure (around 2 bars) in the flat sheet cross flow
microfiltration membrane module (with effective filtration surface
3.4. Experimental procedure
of 100 cm2 each module). Microfiltered water was subsequently
passed to a second nanofiltration membrane module of similar
3.4.1. Response surface optimization and continuous mode treatment
type via a second settling tank (for final precipitation of the resid-
Fenton’s process for cyanide and phenol removal was optimized
ual fertilizer granules) where almost all the charged as well as non-
through rigorous experimentation using response surface method-
charged particles got separated out from the stream following
ology (RSM) of Design Expert software (Design-ExpertÒ Version
Donnan-steric mechanisms of nanofiltration [28]. Finally treated
8.0.6; serial 9 l no 7020–7114, State-ease, Inc.) and the optimum
water was collected as permeate of the nanofiltration membrane
concentrations for the reagents (FeSO47H2O and H2O2) and the
module that was operated at around 15–16 bars pressure.
optimum pH were determined. There is hardly any reported study
on the optimization of this process specifically on simultaneous
degradation of phenol and cyanide. RSM is a statistical method that 3.5. Analytics
uses experimental data generated from specific experimental de-
sign to model and optimizes any process in which several variables Analysis for cyanide, ammonia and pH was done in Orion 4 Star
influence the desired response [13]. Three major steps are involved pH ISE Bench top Ion Meter of Thermo Electron Corporation (USA).
in optimization viz., performing the statistical designed experi- The electrodes were first calibrated using the previously prepared
ments, estimating the coefficients in a mathematical model and standard of respective compounds. The phenol content was deter-
predicting the responses and examining the adequacy of the model mined by HPLC (Agilent Technologies 1200 series, USA) with
[26]. RSM helps to enumerate the relationships between output Zorbax SB-Phenyl column (Germany) having mobile phase metha-
variables called responses (Y) and input variables called factors nol: water (70:30) at flow rate 1 mL/min, residence time of 3.5 min
(Xis) [13]. and injection volume of 5 ll. Chemical Oxygen Demand (COD) was
RSM design called Central Composite Design (CCD) was applied measured by COD Vario Tube Test MR (0–1500 mg/l) of COD ana-
in this study. It is suitable for fitting a quadratic surface and it helps lyzer of LoviBond. Total dissolved solids (TDSs), conductivity and
to optimize the effective parameters with a minimum number of salinity were measured by InoLab Cond 720, with electrode Tetra-
experiments, as well as to analyze the interaction between param- Con 325 (WTW, Germany). Oil and grease were determined follow-
eters [27]. Generally, the CCD consists of a 2k, 2 k and kc factorial ing the standard procedures described in the standard methods
runs, axial runs and central run respectively [27]. The central point [29]. During nanofiltration percentage removal of pollutants were
helps to evaluate the experimental errors and reproducibility of calculated using the initial concentration (Ci) of the pollutants in
the data. Total number of experiments with different combinations the feed sample and the residual concentration (Cf) in the permeate
of three variables (concentrations of iron salt in the range 1.00– side respectively using Eq. (12) as given below.
3.75 g/L, hydrogen peroxide in the 1.50–5.50 g/L range and pH
 
3.0–10.0) was 20 (2k + 2k + kc), where k denoted the number of Cf
variables and kc (=6), the number of central points for a three % removal of pollutants ¼ 1   100 ð11Þ
Ci
variables system. Degradation experiments were carried out batch-
wise in 250 ml conical flasks with a working volume 100 ml in
each case in incubator shaker (speed 160 rpm) at ambient
3.6. SEM-EDS, XRD, FT-IR, TGA analysis for struvite

Table 2 The collected precipitates were washed with pure water repeat-
Features of the nanofiltration membranes (Sepro, USA) used. edly, air dried for 48 h, and then analyzed by X-ray diffraction
Characteristics Membranes (XRD, D/max-RB, Rigaku, Japan) and Fourier transform infra-red
(FT-IR) spectroscopy (Nicolet iS10, Thermo Fischer Scientific,
NF1 NF2 NF3 NF20
USA). For FT-IR study, 40 mg of KBr (MERCK) was mixed well with
Geometry Flat-sheet Flat-sheet Flat-sheet Flat-sheet 2 mg of finely ground sample for the preparation of transparent
Material Polyamide Polyamide Polyamide Polyamide
Thickness (lm) 165 165 165 165
pellets, for the determination of functional groups. Scanning elec-
tron microscopy with energy dispersive X-ray analysis (SEM-EDS,
Solute rejection (%)
MgSO4 99.5 97 98 98
S-3000, Hitachi Japan) study of MAP was done to get its crystalline
NaCl 90.0 50 60 35 structure, size of the crystal and surface composition of precipi-
pH 2–11 2–11 2–11 2–11 tates. SEM photograph was taken with scanning electron micro-
Maxim Temp (°C) 50 50 50 50 scope at the required magnification at room temperature. TGA
Maxim pressure (bar) 83 83 83 83
and DTGA analysis of struvite was done (DTG-60H, Shimadzu,
Pore radius (nm) 0.53 0.57 0.55 0.54
Japan) to get the heating profile at 10 °C/min.
R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44 37

Fig. 1. Schematic diagram of the membrane-integrated hybrid treatment process for coke wastewater.

4. Results and discussion data, and square root was chosen (being the best fit in this case)
in both cases (phenol and cyanide removal (%)). From fit summary
4.1. Fenton oxidation for the degradation of cyanide and phenol section in design, the model F-values as obtained for cyanide re-
moval (145.93) and phenol removal (57.86) implied that models
Phenol and cyanide were simultaneously removed from the were significant. Value of P (0.0001) in both cases being less than
coke wastewater by Fenton’s reagent (FeSO47H2O and H2O2) by 0.0500 also indicates that the model terms are significant. The final
optimizing the different concentration hydrogen peroxide, ferrous regression equation made by analysis of variance (ANOVA) shows
sulfate and pH. As hydrogen peroxide is a strong oxidizing agent the empirical relationship among the target variables (cyanide
(oxidation potential 1.77) and dissociate into hydroxyl radicals in and phenol removal) and the three operating conditions or vari-
solution that oxidize cyanide compounds to cyanate and further ables. The equation in terms of coded factors is represented by
to ammonia and carbonate ions. This leaves little chance for cya- Eqs. (13) and (14) and the statistical parameters obtained from
nide to get converted into HCN. the ANOVA for these two regression models are listed in Table 4.

Sqrt ðcyanide removal %Þ ¼ þ9:94 þ 1:22  H2 O2 þ 1:25


4.1.1. Results of response surface optimization study using design
 iron salt þ 1:33  pH  0:46
expert software
Experiments were planned and conducted following central  H2 O2  iron salt  0:23
composite design (CCD), considering the minimum and maximum  H2 O2  pH þ 0:33
levels for hydrogen peroxide (1.50–5.50 g/L), iron salt (1.00–3.75
g/L) and pH (3.0–10.0) and the experimental results were incorpo-  iron salt  pH
rated in that design (Table 3). These ranges of the variables param-  0:75ðH2 O2 Þ2
eters were fixed based on the data as reported in the literature in
various studies of phenol and cyanide degradation by Fenton’s
 0:67 ðiron saltÞ2
reagent. In determining the interrelationships of those variable,  2:03 ðpHÞ2 ð12Þ
second-order polynomial equation was fitted to the experimental

Table 3
Experimental response under suggested operating conditions.

STD Run H2O2 (g/L) Iron salt (g/L) pH Response 1 cyanide removal (%) Response 2 phenol removal (%)
14 1 3.50 2.38 12.39 36 12
18 2 3.50 2.38 6.50 99 98
1 3 1.50 1.00 3.00 7 59
15 4 3.50 2.38 6.50 99 98
19 5 3.50 2.38 6.50 99 98
3 6 5.50 1.00 3.00 26 95
6 7 1.50 3.75 10.00 63 34
13 8 3.50 2.38 0.61 4 99
16 9 3.50 2.38 6.50 99 98
12 10 6.86 2.38 6.50 99 98
9 11 3.50 0.06 6.50 28 31
5 12 1.00 1.50 10.0 25 9
4 13 5.50 3.75 3.00 51 100
11 14 0.14 2.38 6.50 33 28
20 15 3.50 2.38 6.50 99 98
10 16 3.50 4.69 6.50 99 98
8 17 5.50 3.75 10.0 98 46
17 18 3.50 2.38 6.50 99 98
7 19 5.50 1.00 10.00 88 25
2 20 1.50 3.75 3.00 35 78
38 R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44

Table 4 were not amongst the 20 experiments previously designed by


Statistical parameters obtained from the analysis of variance for the regression CCD and were rather assumptions, experiments with selected
models.
criteria were performed in shaker flask level. In the first case, ob-
Response R2 Adj. R2 CV (%) S.D AP tained cyanide and phenol removal was 99.5% and 98.6% and in
Cyanide removal 0.9924 0.9856 4.23 0.32 36.11 the second case 99.7% and 100%, respectively. In both the cases,
Phenol removal 0.9812 0.9642 5.79 0.46 23.98 the predicted responses were in close agreement with the actual
AP: adequate precision; SD: standard deviation; CV: coefficient of variance.
experimental values. Solution (b) i.e. at pH 6.09, H2O2 5.45 g/L
and iron salt 1.91 g/L was accepted as the complete removal of cya-
nide and phenol with minimum concentration of iron salt which
Sqrt ðphenol removal %Þ ¼ þ9:90 þ 0:97  H2 O2 þ 1:02 reduced the sludge generation during treatment and avoided iron
contamination in final effluent. The optimum values thus obtained
 iron salt  1:92  pH  0:24 were used in the continuous process to control the doses of the re-
 H2 O2  iron salt þ 0:40 agent. Fig. 4 shows the HPLC result of phenol concentration before
the Fenton’s treatment and after the treatment with optimized
 H2 O2  pH  0:035
concentration. Thus after chemical pre-treatment most of the
 iron salt  pH phenolic compounds were found to get degraded, whatever inter-
 0:78 ðH2 O2 Þ2 mediates were produced during chemical pre-treatment were
completely removed by nanofiltration membrane (as in Section
 0:83ðiron saltÞ2  1:15 ðpHÞ2 ð13Þ 4.3) as indicated in the final analysis.
The quality of the model developed was evaluated based on the
correlation coefficient R2 and standard deviation value. The closer 4.2. Removal of ammonia through chemical precipitation as struvite
the R2 value to unity the smaller the standard deviation implying by-product
more accurate response that could be predicted by the model.
The R2 value for cyanide and phenol removal were found to be After Fenton’s treatment, ammonia precipitated out as magne-
99.24% and 98.12% of respectively. The value of adjusted determi- sium ammonium phosphate (MAP) from Fenton’s treated effluent
nation coefficient was within reasonable agreement with the pre- in the struvite by-product unit. The effluent from the Fenton’s
dicted R2 in both responses (cyanide and phenol) and it implies treatment unit contained a high concentration of ammonia
2
that the model was significant. Effects of H2O2, iron salt and pH (2135 mg/L) due to conversion of cyanide to NHþ 4 and CO3 . In
were highly significant as the P values <0.0001 in both cases. The the first step of MAP precipitation, three different combinations
high significance of the model was also established in the plot of of the chemicals MgCl26H2O, Na2HPO412H2O, MgO, H3PO4, Ca(H2-
calculated values against the experimental values of cyanide and PO4)2H2O and MgSO47 H2O at different molar ratios of Mg2+ and
phenol removal % (Fig. 2). Clustering of points around the diagonal PO3
4 with respect to NH3 concentration were investigated to find
line indicated capability of the model to predict the experiments in out the best combinations for maximum precipitation of ammonia
both cases. The coefficient of variance (CV) should not be larger as struvite. The experimental results as shown in Fig. 5 indicate
than 10% and is defined as the percent ratio between the standard that the combination of MgCl2.6H2O, Na2HPO4.12H2O and ammo-
error of the estimate and the mean value of the observed responses nia was the best when used in the molar ratio of 1:1:1. Struvite
and this is a measure of the reproducibility of the model [30]. Thus, precipitation follows Eq. (15):
low CV values in both models (cyanide removal 4.23% and phenol
removal 5.79%) indicated a good reproducibility. Adequate preci- MgCl2  6H2 O þ NHþ4 þ Na2 HPO4  12H2 O
sion is a measure of the range in predicted response relative to () MgNH4 PO4  6H2 O # þ2NaCl ð14Þ
its associated error or, in other words, a signal-to-noise ratio. Its
desired value is 4 or greater [31]. In the present study, adequate However, excess dose of either Mg or2+
PO3
4 did not result in
precision values (cyanide removal 36.11 and phenol removal any significant increase in ammonia removal. The combination of
23.98) were achieved. MgCl26H2O and Na2HPO412H2O was the most efficient in the
Fig. 3 presents the response surface modelling in a three dimen- ammonium removal, but this also led to high salt concentration
sional representation reflecting the effects of hydrogen peroxide, in the effluent in comparison to other two combinations of chem-
iron salt and pH on the cyanide and phenol removal after 1 h of icals as evident from Eq. (15). The salts produced at this stage were
reaction. As a general trend, it was observed that effects of hydro- subsequently removed by nanofiltration membrane (as discussed
gen peroxide and iron salt on removal of cyanide and phenol was in Section 3.4).
pH-dependent. At low pH, cyanide degradation by hydrogen perox- pH exhibits significant role in the struvite production which
ide and iron salt is very difficult due to presence of cyanide mainly ultimately leads to ammonia removal. In terms of thermodynamic
as HCN gas. At high pH, cyanide present as CN ions, so it reacts equilibrium, hydrogen ion is released into the solution when stru-
easily with H2O2 and iron salt [32]. Low levels of pH favoured vite is formed, resulting in a decrease in pH. At low pH, hydrogen
reduction of phenol which could be traced to generation of hydro- ion in reaction solution would inhibit struvite formation resulting
xyl radicals. in reduced removal of NHþ 4  N. Under high pH values, other lower

During design of experiments, criteria were selected for optimi- precipitates like Mg3(PO4)2 are formed instead of MAP, resulting in
zation of H2O2 and pH (in the already selected range) at minimum decreased removal of NHþ 4  N [15]. A pH value of 9.0 has to be

concentration of iron salt for maximum removal of cyanide and maintained during the reaction by adding NaOH and HCl for better
phenol. Some optimized solutions with different criteria were sug- yield of struvite.
gested by the software. From those suggested solutions, two The content of struvite in precipitates has been confirmed
acceptable solutions were taken as follows: (a) at pH 6.50, H2O2 through SEM image, SEM-EDS, FT-IR, XRD and TGA analysis as
3.51 g/L and iron salt 2.38, while the expected cyanide and phenol shown in the Fig. 6a–e. The FT-IR analysis showed that the infrared
removal was 99% and 98.23%; (b) at pH 6.09, H2O2 5.45 g/L and iron spectrum of the precipitate was close to that of the MAP as re-
salt 1.91 g/L, while the expected 99.002% and 100% cyanide and ported elsewhere [33]. At 3700–2500 cm1 water-stretching broad
phenol removal. As the selected optimum criteria for variables band was observed which indicated the presence of crystalline hy-
drate. The water–phosphate hydrogen bonding was attributed to
R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44 39

Fig. 2. Distribution of experimentally determined values versus statistically predicted values of (A) cyanide removal (%) and (B) phenol removal (%).

2350 cm1 band. The 1683 cm1 band was attributed to the water 250 °C and 51% corresponds to the following decomposition for
HOH bending region. The 1436 cm1 band were characteristics the struvite:
ammonium band. The XRD patterns and elemental profile gener-
MgNH4 PO4  6H2 O ! MgHPO4 þ NH3 " þ 6H2 O " ð15Þ
ated by SEM-EDS showed that the precipitates were close to that
of the MAP as reported elsewhere [33,34]. The TGA and DTGA anal- DTGA curve of the struvite shows single peak which is attained
ysis was done for struvite at 10 °C/min in presence of nitrogen gas. at temperature 103 °C due to simultaneous loss of both ammonia
These data indicate that mass loss begins at a temperature around and water molecules which indicates that precipitate material is
55 °C and is essentially complete when temperature reached above struvite.
40 R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44

Fig. 3. Response surface plot showing the removal (%) of cyanide (A–C) and phenol (D–F) with variable parameters pH, iron salt and hydrogen peroxide.

Struvite formed during chemical precipitation reaction, were 4.3. Nanofiltration for the removal of trace chemical contaminants: the
separated by PVDF microfiltration membrane fitted in the final polishing step towards recycling
membrane module in the cross-flow mode. The recovery efficiency
of the used membrane module for MAP was more than 99%. The Two stages of chemical treatment for the removal of cyanide,
recovered struvite can be used as a slow releasing fertilizer for phenol and ammonia resulted in increased values of TDS, salinity
agricultural use. Table 5 represents the major characteristics of and conductivity due to introduction new ionic species and salts
wastewater after chemical treatments (Fenton’s and MAP precipi- in the solution. During Fenton’s treatment OH radicals were pro-
tation) and its effluent acts as influent for membrane separation. duced which alone might not be sufficient to degrade all types of
R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44 41

Fig. 4. HPLC–UV graph for the raw coke wastewater and after treatment.

during nanofiltration. At pH values greater than 7.0, most of the


polyamide composite nanofiltration membranes possess negative
zeta potential [25]. Due to the charged nature of the NF membrane,
solutes with an opposite charge compared to the membrane (coun-
ter-ions) are attracted, while solutes with a similar charge (co-
ions) are repelled. Apparent pore size of polyamide NF membranes
can also vary with solution pH. At higher pH values, addition of so-
dium hydroxide leads to an increase in osmotic pressure and ionic
strength, thus reducing the membrane permeability and increasing
rejection. Moreover, functional groups such as carboxyl, hydroxyl
groups were present at the surface of the membrane become
deprotonated at high pH. High pH favors an increased thickness
of diffuse double layer of charged functional groups over the sur-
face of membrane thus reduces the apparent pore size and resulted
in greater rejection charged solutes [35].
The flux data as presented in Fig. 7a and b shows that permeate
flux increased with increase of transmembrane pressure from 5 to
16 bars as well as cross-flow rate from 350 to 800 liters per hour
(LPH) for all nanofiltration membranes. On running the cross-flow
filtration system under various transmembrane pressure and
cross-flow rates, it is possible to find out the critical flux, i.e. the
3
Fig. 5. Effect of different Mg2+: NHþ þ
4 : PO4 molar ratios on NH4  N removals and flux below which minimum fouling occurs. NF2 membrane yielded
selection of compounds at pH 9.0. the highest flux due to its large pore radius (0.57 nm) whereas the
NF1 membrane (0.53 nm) yielded the lowest flux as permitted by
chlorophenols and their degraded intermediates. This is possibly its smallest pore radius. Cross-flow rate also plays a crucial role
the reason of high COD after two stages of chemical treatment. in reducing membrane fouling and optimization of minimizing
Nanofiltration played a significant role in removal of these pheno- membrane area requirement. Increase of cross flow rates, flux in-
lic intermediate compounds (CODs), ionic species and salts in the creases; this is because of greater convective force that minimizes
final stage of treatment of coke wastewater. Four different polyam- cake formation and concentration polarization effects. From above
ide composite nanofiltration membranes (NF1, NF2, NF3 and NF4) experiments, an optimum operating pressure and cross flow rate of
fitted in a largely fouling-free cross flow module were examined 16 bars and 800 LPH were arrived respectively, under which the
42 R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44

Fig. 6. Surface characterization analysis of struvite (MAP precipitate) obtained during the chemical treatment of ammonia. (a) scanning microscopy analysis of struvite; (b)
energy dispersive X-ray analysis of struvite; (c) Fourier transform infrared spectroscopy analysis of struvite; (d) XRD pattern of the struvite crystel; (e) TGA and DTGA curves
for struvite for heating rate 10 °C min1.

present system could be operated under a stable condition. Mod- Solution-diffusion mechanism refers to diffusion of solution
ules should be operated at these critical levels of cross flow rates through nanofiltration membrane. In solution diffusion mecha-
which permit maximization of flux for a given membrane surface nism, solute flux and solvent flux are uncoupled and hence in-
area. crease of solvent flux with increase of transmembrane pressure
Fig. 8 exhibits trend in COD rejection with increase of applied does not result in commensurate increase of solute flux. Rather in-
pressure. 97.5% COD could be reduced by NF1 membrane showing crease of solvent flux obstructs the transport of solute though the
the best performance compared with the other three investigated membrane. Increase in transmembrane pressure leads to increase
membranes. With increase in the pressure, COD rejection increased in solvent flux. That means solutes rejection will increase with
for all the membranes. The solution diffusion mechanism that ap- increase in transmembrane pressure. Solute flux in this case
plies to nanofiltration plays the major role in the COD rejection. refers to l/m2/h of solutes (COD) that permeates out through the
R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44 43

Table 5
Characteristics of wastewater after chemical treatment.

Parameter (mg/L) Influent Effluent after chemical treatment (mg/L)


Cyanide 124 0.42
Phenol 162 ND*
Ammonia 1940 48
COD 3856 2270
TDS 24120 37800
Conductivity (mS/cm) 19.12 38.56
Salinity 11.7 25.9
Oil and grease 51 3.5
pH 9–9.5 8.5–9.0
*
Not Detected.

Fig. 8. Effect of applied pressure on COD rejection percentage for all the four
membranes. Operating conditions: transmembranes 16 bar, cross flow rate 800 L/H,
pH 9.0 and temperature 35 °C.

Fig. 9. Effect of applied pressure on TDS, salinity and conductivity rejection


percentage for NF1 membrane. Operating conditions: transmembranes 16 bar,
cross flow rate 800 L/H, pH 9.0 and temperature 35 °C.

Table 6
Characteristics of wastewater after final treatment (membrane separation).

Parameter (mg/L) Influent Final effluent limit Permissible


Fig. 7. Effect of (a) transmembrane pressure (b) cross-flow rate on Flux throughF1, (mg/L) (mg/L)
NF2, NF3 and NF20 membrane. Operating conditions: pressure range 5–16 bar and Cyanide 0.42 ND* <0.1
cross flow rate 300–800 L/H. Phenol ND* ND* <0.5
Ammonia 48 22 <30
COD 2270 102 <250
nanofiltration membrane. Solvent flux is defined as liter of aqueous TDS 37800 1360 –
Conductivity (mS/cm) 38.56 0.9254 –
solvent per meter2 of membrane surface per hour. Salinity 25.9 0.8547 –
Fenton oxidation and MAP precipitation resulted in an effluent Oil and grease 3.5 ND* <10
with 37,800 mg/l TDS, 25.9 salinity and 38.56 mS/cm conductivity pH 8.5–9.0 7.5–8.5 –
reflecting an increase of 56.7%, 121.36% and 73.6% of these three *
Not Detected.
parameters as compared to those in initial raw wastewater. High
salinity resulted from formation of NaCl as a by-product during
MAP precipitation by MgCl26H2O and Na2HPO412H2O (Eq. (15)).
Effect of transmembrane pressure on the rejection of TDS, salinity charge. The high TDS, salinity and conductivity was due to the
and conductivity by NF1 membrane was determined with the charged inorganic ions like bicarbonate ions, Cl, SO24 , etc. and
microfiltrate feed. Fig. 9 shows that with increase of transmem- when these solutes were rejected in contact with negatively
brane pressure from 5 to 16 bars, rejection increased from 73.4% charged NF1 membrane following Donnan exclusion mechanism.
to 96.5%, 78.2% to 96.4% and 76.1% to 97.6% for TDS, salinity and The uncoupling of solute and solvent flux (solution diffusion
conductivity, respectively by NF1membrane. The pH of the solu- mechanism) also plays major role in the removal of TDS, salinity
tion was 9.0, and at this pH, NF1 membrane exhibited negative and conductivity. After MAP precipitation, the remaining ammonia
44 R. Kumar, P. Pal / Chemical Engineering Journal 210 (2012) 33–44

(48 mg/l) was effectively rejected by NF1 membrane (22 mg/l after [12] C. Korzenowski, M. Minhalma, A.M. Bernhandes, J.Z. Ferreira, Nanofiltration for
the treatment of coke plant ammoniacal wastewaters, Sep. Purif. Technol. 76
NF). In case of ammonia, pore size was effective. The used mem-
(2011) 303–307.
branes are not regenerated, but are only rinsed with water when [13] J.N. Sahu, J. Acharya, B.C. Meikap, Response surface modeling and optimization
flux falls drastically. Normally when rinsing with water or chemi- of chromium (VI) removal from aqueous solution using tamarind wood
cal cleaning fails to revive the flux, the membrane is discarded. Ta- activated carbon in batch process, J. Hazard. Mater. 4 (2009).
[14] J.P. Silva, S. Sousa, I. Goncalves, J. Porter, S. Ferreira Dias, Modelling adsorption
ble 6 shows the major features of wastewater after membrane of acid orange 7 dye in aqueous solutions to spent brewery grains, Sep. Purif.
separation and its comparison with emission limit values for Technol. 40 (2004) 163–170.
wastewater and treated effluent discharged to surface water follow [15] N.A. Booker, A.J. Priestley, I.H. Fraser, Struvite formation in wastewater
treatment plants: opportunities for nutrient recovery, Environ. Technol. 20
by most of the countries. (1999) 777–782.
[16] J.D. Doyle, S.A. Parsons, Struvite formation, control and recovery, Water Res. 36
(16) (2002) 3925–3940.
5. Conclusion
[17] E.Y. Yazici, H. Deveci, I. Alp, Treatment of cyanide effluents by oxidation and
adsorption in batch and column studies, Karadeniz Technical University 61080
The study demonstrates that for recycling water, there needs to Trabzon, Turkey, J. Hazard. Mater. 166 (2009) 1362–1366.
[18] G.F. IJpelaar, R.T. Meijers, R. Hopman, J.C. Kruithof, Oxidation of herbicides in
be a paradigm shift in treatment regime for industrial wastewater.
groundwater by the Fenton process: a realistic alternative for O3/H2O2
Conventional treatment plants could be turned operationally fast, treatment?, Ozone Sci Eng. 22 (2000) 607–616.
economically more viable and environmentally more benign by [19] S.F. Kang, H.M. Chang, Coagulation of textile secondary effluents with Fenton’s
appropriate integration of membrane-based treatment with con- reagent, Water Sci. Technol. 36 (12) (1997) 215–222.
[20] A. Santos, P. Yustos, S. Rodriguez, E. Simon, F. Garcia-Ochoa, Abatement of
ventional chemical treatment. However, chemical treatment units phenolic mixtures by catalytic wet oxidation enhanced by Fenton’s pre-
should be operated under well optimized conditions where rigor- treatment: effect of H2O2 dosages and temperature, J. Hazard. Mater. 146
ous experimentation using response surface methodology could (2007) 595–601.
[21] H. Strathmann, in: W.S. Ho, K.K. Sirkar (Eds.), Membrane Handbook, Van
be of great help. Future studies should explore raising scale up con- Nostrand Reinhold, 1992.
fidence through modelling and simulation with varieties other [22] J.D. Ferry, Ultrafilter membranes and ultrafiltration, Chem. Rev. 18 (1936)
membranes that could be more effective in separation while ensur- 373–455.
[23] S. Bandini, D. Vezzani, Nanofiltration modelling: the role of dielectric exclusion
ing prolonged service without significant membrane fouling. in membrane characterization, Chem. Eng. Sci. 58 (15) (2003) 3303–3326.
[24] M.D. Afonso, G. Hagmeyer, R. Gimbel, Streaming potential measurements to
Acknowledgment assess the variation of nanofiltration membranes surface charge with the
concentration of salt solutions, Sep. Purif. Technol. 22–23 (1–3) (2001) 529–
541.
Authors are thankful to the Department of Science and Technol- [25] A.I. Schafer, D.I. Nghiem, T.D. Waite, Removal of natural hormone estrone from
ogy, Government of India for financial support under DST-FIST aqueous solutions using nanofiltration and reverse osmosis, Environ. Sci.
Technol. 37 (1) (2003) 182–188.
program.
[26] R.H. Myers, D.C. Montgomery, Response Surface Methodology: Process and
Product Optimization Using Designed Experiments, first ed., John Wiley and
References Sons Inc., New York, 1995.
[27] B.H. Hameed, I.A.W. Tan, A.L. Ahmed, Optimization of basic dye removal by oil
[1] M.K. Ghose, S. Roy, Status of water pollution due to coke oven effluent: a case palm fibre based activated carbon using response surface methodology, J.
study, J. Publ. Health Eng. 3 (1996) 1–9. Hazard. Mater. 158 (2008) 324–332.
[2] A. Akcil, A.G. Karahan, H. Ciftci, O. Sagdic, Biological treatment of cyanide by [28] R. Kumar, P. Bhakta, S. Chakraborty, P. Pal, Separating cyanide from coke
natural isolated bacteria (Pseudomonas sp.), Min. Eng. 16 (2003) 643–649. wastewater by cross flow nanofiltration, Sep. Sci. Technol. 46 (2011) 2119–
[3] C. Staib, P. Lant, Thiocyanate degradation during activated sludge treatment of 2127.
coke-ovens wastewater, Biochem. Eng. J. 34 (2007) 122–130. [29] American Public Health Association (APHA), Standard Methods for the
[4] D. Park, Y.M. Kim, D.S. Lee, J.M. Park, Chemical treatment for treating cyanides- Examination of Water and Wastewater, 19th ed., APHA–AWWA–WEF,
containing effluent from biological cokes wastewater treatment process, Washington, DC, USA, 1995.
Chem. Eng. J. 143 (2008) 141–146. [30] Q. Beg, V. Sahai, R. Gupta, Statistical media optimization and alkaline protease
[5] Z. Shen, B. Han, S.R. Wickramasinghe, Cyanide removal from industrial production from Bacillus mojavensis in a bioreactor, Process Biochem. 39
praziquantel wastewater using integrated coagulation-gas-filled membrane (2003) 203–209.
absorption, Desalination 195 (2006) 40–50. [31] A.A.L. Zinatizadeh, A.R. Mohamed, A.Z. Abdullah, M.D. Mashitah, I.M. Hasnain,
[6] G. Banerjee, Phenol and thiocyanate based wastewater treatment in RBC G.D. Najafpour, Process modeling and analysis of palm oil mill effluent
reactor, ASCE 122 (1996) 941–948. treatment in an up-flow anaerobic sludge mixed film bioreactor using
[7] T.F. Edgar, Coal Processing and Pollution Control, Gulf Publishing Co., Houston, response surface methodology (RSM), Water Res. 40 (2006) 3193–3208.
2000. [32] M. Sarla, M. Pandit, D.K. Tyagi, J.C. Kapoor, Oxidation of cyanide in aqueous
[8] P. Ning, H. Bart, Y. Jiang, A. de Haan, C. Tien, Treatment of organic pollutants in solution by chemical and photochemical process, J. Hazard. Mater. B116
coke plant wastewater by the method of ultrasonic irradiation, catalytic (2004) 49–56.
oxidation and activated sludge, Sep. Purif. Technol. 41 (2005) 133–139. [33] S. He, Y. Zhang, M. Yang, W. Du, H. Harada, Repeated use of MAP
[9] J.J. Pignatello, Dark and photo assisted iron (3+)-catalyzed degradation of decomposition residues for the removal of high ammonium concentration
chlorophenoxy herbicides by hydrogen peroxide, Environ. Sci. Technol. 26 from landfill leachate, Chemosphere 66 (2007) 2233–2238.
(1992) 944–951. [34] C.C. Wang, X.D. Hao, G.S. Guo, M.C.M. van Loosdrecht, Formation of pure
[10] S. Esplugas, J. Giménez, S. Contreras, E. Pascual, M. Rodríguez, Comparison of struvite at neutral pH by electrochemical deposition, Chem. Eng. J. 159 (2010)
different advanced oxidation processes for phenol degradation, Water Res. 36 280–283.
(2000) 1034–1042. [35] A. Braghetta, F.A. DiGiano, W.P. Ball, Nanofiltration of natural organic matter:
[11] M. Minhalma, M.N. de Pinho, Integration of nanofiltration/steam stripping for Ph and ionic strength effects, J. Environ. Eng. 123 (1997) 628–641.
the treatment of coke plant ammoniacal wastewaters, J. Membr. Sci. 242
(2004) 87–95.

You might also like