Optimization of pulp mill effluent treatment with catalytic adsorbent using orthogonal second-order (Box–Behnken) experimental design

PAPER www.rsc.org/jem | Journal of Environmental Monitoring

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online / Journal Homepage / Table of Contents for this issue

Optimization of pulp mill effluent treatment with catalytic adsorbentusing orthogonal second-order (Box–Behnken) experimental design

Ekaterina V. Rokhina,*a Mika Sillanpaa,b Mathias C. M. Noltec and Jurate Virkutytea

Received 7th July 2008, Accepted 18th September 2008

First published as an Advance Article on the web 3rd October 2008

DOI: 10.1039/b811556g

Novel catalytic adsorbent (ruthenium on carbon) was employed for the treatment of pulp mill effluent

in the presence of hydrogen peroxide. Mathematical model and optimization of the process regarding

the most favorable COD (%), TOC (%) and color (%) removal rates was developed and performed with

experimental design taking into account catalytic adsorption process kinetics. As the initial

experimental design, 33�1 half-fractional factorial design (H-FFD) was accomplished at two levels to

study the significance of the main effects, such as catalytic adsorbent (g l�1) and hydrogen peroxide

(ppm) concentrations using the response surface methodology (RSM). Finally, a four factor-three

coded level central composite design (CCD) with 28 runs was performed in order to fit a second-order

polynomial model. Validation of the model was accomplished by different criteria including coefficient

of determination and the corresponding analysis of variance. The achieved removal rates for TOC

(up to 75%), COD (up to 73%) and color (up to 68%) were observed for the defined optimal conditions:

1g l�1 of ruthenium on carbon, 7 ppm of hydrogen peroxide, pH ¼ 4 and ambient temperature. The

proposed method benefited significantly improved TOC, COD and color removal efficiency,

regenerability and reusability of the catalytic adsorbent and unaltered initial pH of an effluent in

comparison to traditional adsorption or oxidation processes.

Introduction

In Finland, the pulp and paper industry accounts for a major

portion of the country’s economy in terms of value of production

and total wages paid.1 However, the wood pulping and

production of paper products generate a considerable amount of

hazardous pollutants. In general, level of contamination is

mainly characterized by total organic carbon (TOC), chemical

oxygen demand (COD), suspended solids (SS), toxicity, and

colour when untreated or poorly treated effluents are discharged

to receiving waters.

Advanced oxidation processes (AOPs) are widely used as

powerful methods capable of transforming contaminants into

harmless substances.2 AOPs are aimed to generate very reactive

non-selective transient oxidizing species such as the hydroxyl

radicals cOH, which are generally accepted as the dominant

oxidizing species. For effective oxidations of refractory organic

compounds, the hydroxyl radicals must be continuously gener-

ated in situ through chemical, photochemical or sonochemical

reactions due to their instability. The formation of cOH radicals is

aWaste Management Group, Department of Environmental Science,University of Kuopio, P.O. Box 1627, FIN-7021 Kuopio, Finland.E-mail: [email protected]; Fax: +358 153556363; Tel: +358440176196bLaboratory of Applied Environmental Chemistry, Department ofEnvironmental Science, University of Kuopio, Patteristonkatu 1, FIN-50100 Mikkeli, Finland. E-mail: [email protected]; Fax: +358153556513; Tel: +358 153556830cInstitute of Environmental Technology and Energy Economics, TechnischeUniversitat Hamburg–Harburg, Eibendorfer Strabe 40, 21073 Hamburg,Germany. E-mail: [email protected]; Fax: +40 428782315; Tel: +40428782597

1304 | J. Environ. Monit., 2008, 10, 1304–1312

usually accelerated by combining oxidizing agents such as ozone

(O3), hydrogen peroxide (H2O2) and UV radiation.3 However,

during recent decades, considerable attention has been focused on

the further improvement of these processes and the development

of catalytic AOPs methods because of their simplicity, reliability

and high effectiveness.1–4 Currently available methods, such as

catalytic wet oxidation process (CWAO), Fenton andFenton-like

processes or photocatalytic oxidation with TiO2, all have some

technical or economical drawbacks that limit their usefulness.1–4

Adsorption and catalytic oxidation reactions have always been

known as interrelated processes. It is argued that if successfully

developed, a combination of sorption with destruction technique

may overcome many of technical and economical drawbacks and

become a preferable choice for many industries.2,3,5 The work

principle of catalytic adsorbent is based on the assumption that

sorption and reaction sites are fundamentally different entities

and that the chemical reaction does not profit from sorption at

non-reactive sites.6 Centi and colleagues4 explored the use of

mixed oxides as regenerable solid catalytic adsorbents for the

development of an adsorption method to remove the organics

present in the rinse water of the electronics industry at room

temperature and the regeneration of the adsorbent at mild

conditions. Trawczynski7 investigated the effectiveness of noble

metal catalysts surfaces (Pt, Pd, Ru) supported on carbon black

composites (CBC) for the oxidation of phenol solutions using

a fixed-bed reactor working in a trickle-flow regime. Rodas-

Grapaın8 studied CuO–CeO2 sorbent-catalysts in de-SOx type

reactions. In addition, Moon and colleagues9 found that the

Pd/activated carbon powder can be used as an adsorbent for

recovering chloropentafluoroethane (CF3CF2Cl, CFC-115)

and also as a catalyst for manufacturing pentafluoroethane

This journal is ª The Royal Society of Chemistry 2008

http://www.rsc.org/jem

http://dx.doi.org/10.1039/b811556g

http://pubs.rsc.org/en/journals/journal/EM

http://pubs.rsc.org/en/journals/journal/EM?issueid=EM010011

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

(CF3CF2H, HFC-125) from CFC-115. Adsorption-catalytic

method for carbon disulfide removal from air with the use of

(Fe-EDTA/carbon) was experimentally tested by Yegiazarov

and co-authors.10 All these studies showed that catalytic adsor-

bents may be an efficient alternative to conventional treatment

methods because the catalytic component provides active sites

for oxidation of adsorbed substances and also tunes the textural

properties, as well as increases resistance and regenerability of

the adsorbent. Thus, catalytic adsorbents are apt for wastewater

treatment due to their high activity, commercial availability and

broad spectrum of substrates. Generally, three modes of opera-

tion are possible: (i) a two-step cycle consisting of an adsorption

followed by a regeneration step of the adsorbent, (ii) a mode

where sorption and destruction reaction take place simulta-

neously, and (iii) a quasi-simultaneous mode with the destruction

reaction cyclically performed under intensified conditions for the

complete regeneration of the adsorbent.5

Themain aimof thisworkwas to explore the use of rutheniumas

a novel catalytic element embedded in active support with high

sorption capacity—carbon. The resulting product was studied as

regenerable solid catalytic adsorbent for the development of

aprocess to removepersistent organic contaminants frompulpmill

factory effluent. The advantageous use of such catalytic adsorbent

was complemented by mild operational parameters (room

temperature and atmospheric pressure) and the ability of the

process to be performed without the adjustment of initial waste-

water characteristics (pHand temperature). In addition, secondary

aim of the study was to establish the optimum conditions of

the process by considering technological (maximizing removal

efficiencies) and economical (minimizing chemicals consumption)

aspects. It was hypothesized that Box–Behnken experimental

design would serve this purpose well through the collection of

mathematical and statistical methods for analysing and modelling

of the effects of several independent variables and their responses.

Fig. 1 Representative SEM micrographs of (a) fresh (b) calcinated at

400 �C and (c) after treatment with hydrogen peroxide ruthenium on

carbon catalytic adsorbent.

Results and discussion

Characterization of the catalytic adsorbent

Fig. 1 shows a typical SEM image of fresh ruthenium on carbon,

catalytic adsorbent calcinated at 400 �C for 4 h and the sample

after treatment with hydrogen peroxide. This figure clearly shows

the typical mesoporous order of the inner structure of the

material, showing that the addition of hydrogen peroxide did not

affect the catalytic adsorbent structure. When this material was

calcinated, the ordered structure was lost, a worm-like structure

was observed and a long range mesoporous order was not found

even after the calcination process at high temperature (Fig. 1b).

Despite the loss of the long-range mesoporous order, the relative

surface area of catalytic adsorbent was high (625 m2 g�1).

Accordingly, the surface area of the calcinated sample was the

same but the surface area of the catalytic adsorbent after

hydrogen peroxide treatment gradually decreased to 28.1 m2 g�1

as a result of adsorption of the pollutants onto the surface of

ruthenium on carbon. Fig. 2 shows textural properties of the

catalytic adsorbent and its particle size distribution. As can be

seen in Fig. 2a, the N2 adsorption/desorption isotherm belongs to

the type IV of isotherms, which is typical for these solids as well

as for mesostructured materials in general. The pore size


distribution, calculated using the Barrett-Joyner-Halenda (BJH)

method, ranged from 12.5 to 120 A (Fig. 2b). Particle size

distribution was the same for all three samples regardless

experimental conditions and was in the range of 266–1036 nm

with the average particle size of 347 nm as presented in Fig. 2b as

inset. There was no significant improvement observed in

adsorptive properties, physical stability or the overall perfor-

mance of the calcinated sample in comparison to the fresh

catalytic adsorbent. Thus, all the following discussion was only

based on results obtained while using fresh catalytic adsorbent.

Adsorption and catalytic process in the presence of hydrogen

peroxide

Generally, applied process is extremely complex and can

be regarded as two interrelated processes: adsorption and

J. Environ. Monit., 2008, 10, 1304–1312 | 1305


Fig. 2 Textural features of ruthenium on carbon (a) typical N2

adsorption/desorption isotherm, (b) pore size distribution (particle size

distribution as inset).

Table 1 Synergistic index of the catalytic adsorbent–H2O2 system interms of COD, TOC and color removal

COD TOC Color

SI 0.99 0.98 1.0RRads (%) 67.5 68.2 54.6RRox (%) 11.3 10.9 13.3

Table 2 Kinetic parameters of COD, TOC and color removal and theequilibrium adsorption capacities

Parameters g � 10�5/l mg�1min�1 k/min�1 qe/mg g�1

COD 0.069 0.035 0.046TOC 0.89 0.33 1.23Color 0.63 0.56 0.82

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

heterogeneous catalytic oxidation of organic contaminants

present in the pulp and paper effluent. The process investigated

in the current study consists of few steps: the adsorption of

contaminants onto adsorptive sites, decomposition of hydrogen

peroxide over active catalytic sites, and the reactions on the

surface with following desorption of the reaction products

from active catalytic sites. This mechanism is favourable for

cases when reactive species are generated at the surface of the

adsorbent, i.e. in a close vicinity to the adsorbed organic

contaminants.5 The generation of radicals via decomposition of

hydrogen peroxide on catalytically active sites follows the well

known Haber–Weiss mechanism:

Me + H2O2 / Me+ + OH� + cOH

cOH + H2O2 / H2O + cHO2

Short-lived radicals are mobile and react with adsorbed

substrate located nearby, immediately after the formation. This

hypothesis is supported by findings of Kwan and Voelker,6 who

stated that adsorbed substrate is preferentially available for

oxidation by hydroxyl radicals.

1306 | J. Environ. Monit., 2008, 10, 1304–1312

Total efficiency of the process may be expressed as a sum of

removal efficiencies from adsorption and catalytic oxidation in

the presence of hydrogen peroxide, studied separately. Therefore,

the synergism of both processes was assessed by using a syner-

gistic index (SI) expressed below:

SI ¼ RR/(RRads + RRox) (1)

where SI were synergistic indexes in terms of COD, TOC and

colour removal and RRs were the removal rates (%). Table 1

shows synergistic indexes of COD, TOC and colour. For the

catalytic adsorption, removal rates were marginally greater than

those obtained from the separate processes, indicating beneficial

effect of coupled over conventional processes. Another benefit

observed was the increased reusability of ruthenium on carbon

(up to six cycles of catalytic adsorption without the necessary

regeneration) with no significant loss (under 5%) in efficiency,

instead of two cycles with 50% loss in efficiency employing same

processes separately. The most likely reason for that was a direct

oxidation of adsorbed species on the catalytic absorbent surface

with hydrogen peroxide.11 It has been reported that regeneration

of activated carbons saturated with organic compounds may be

performed using reactive methods which also include catalytic

oxidation.12 Therefore, catalytic adsorbents used in the presence

of the oxidizing agent (hydrogen peroxide in a current study),

enhance the decomposition rates of contaminants and enable to

carry the self-regeneration at lower temperatures and lower

residence times in comparison to conventional processes.11

Process kinetics

Process kinetics, demonstrating the solute uptake rate, is one

of the most important characteristics, which represents the

efficiency of the selected catalytic adsorbent and therefore,

determines its potential application. As the process of catalytic

adsorption is rather complex, the proposed kinetics has been

presented as kinetics of main processes studied separately.

Table 2 shows main kinetic parameters of the catalytic adsorp-

tion process: adsorption (g) and oxidation kinetics coefficients

(k) and the equilibrium adsorption capacity (qe). Adsorption

kinetics of COD, TOC and color removal using catalytic

adsorbent was calculated as the initial velocity normalized by the



Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

operating conditions parameters according to Brasquet and

colleages,13 and it was defined by the following equation:

g ¼ V/C0m(dC/dT)t/0 (2)

where V was the initial volume of the solution (l), m was weight

of the catalytic adsorbent (mg), C0 was the initial concentration

of COD, TOC and color (ppm) and C was the concentration

(ppm) of COD, TOC and color at time T (min).

Oxidation kinetics was found to follow the pseudo–first order

and kinetic rate constant was derived from the power law:

d[C]/dt ¼ k[C]n (3)

where C was the concentration of COD, TOC and color at time

(ppm) and n was the reaction order.

The most widely used isotherm equation for modelling of

adsorption data is the Langmuir equation, which is valid for

monolayer adsorption onto a surface with a finite number of

identical sites and is given as:

Qe ¼ KLqmCe/(1 + KLCe) (4)

where KL is the adsorption equilibrium constant including the

affinity of binding sites (g�1), Ce is the concentration of COD,

TOC and color at the equilibrium (ppm), qm is the maximum

adsorption capacity (mg g�1) and qe is the amount of sorbed

COD, TOC and color at the equilibrium (mg g�1). According to

the kinetic evaluation the optimization model was developed in

order to study the interaction of main process parameters taking

into account the adsorptive capacity and the oxidation ability of

novel catalytic adsorbent—ruthenium on carbon.

Experimental design

Preliminary study was accomplished at two coded levels: �1 and

+1 (Table 3) by means of orthogonal second-order (Box–

Behnken) half fractional factorial experimental design. Results

are shown in Table 3 marked with asterisks. With the assistance

of normal probability plot, the influence of main effects—X1and

X2 was assessed in terms of contribution to the removal efficiency

(Fig. 3). The normal probability plot revealed that studied effects

(X1 and X2) represented a sample from normal population since

they formed approximately a straight line. Thus, main effects and

their factor interactions influenced the removal efficiency of

COD, TOC and color in a great extent. 3D surfaces presented in

Fig. 4a, b and c for COD, TOC and color removal, respectively,

showed an improved conversion of all parameters with an

increase in concentrations of both, hydrogen peroxide and

Table 3 Coded levels and range of independent variables for theexperimental design

Variable Symbol

Coded variable level

Low �1 Center 0 High + 1

Ruthenium on carbon/g l�1 X1 1 2 3Hydrogen peroxide (ppm) X2 0 7 15pH X3 4 7 11Temperature/�C X4 20 40 60


ruthenium on carbon. It could be seen that an increase in

hydrogen peroxide concentration at a fixed high level of ruthe-

nium on carbon concentration considerably enhanced the

removal efficiency of COD, TOC and color. While the increase in

the H2O2 concentration at a low level of the catalytic adsorbent,

led to the insignificant changes in the removal rates. In addition,

at a low level of H2O2, increasing in the ruthenium on carbon

concentration produced a little increase in the removal efficiency.

However, the removal efficiencies strongly improved at high

levels of H2O2 concentration. It seems that with low catalytic

adsorbent loading, all the oxidation mechanisms that are based

on the catalytic decomposition of H2O2 slowed down due to

decreased amount of active catalytic sites. The removal rates

consistently reached maximum for high catalytic adsorbent

loading and medium hydrogen peroxide concentration level. The

excess amount of hydrogen peroxide was reported to be detri-

mental for the process due to scavenging of hydroxyl radicals and

also un-reacted hydrogen peroxide may lead to the secondary

contamination of wastewater.12 The proposed model supported

this observation as the significant decrease in the removal rate of

COD, TOC and color was observed at the highest level of H2O2

concentration. High positive regression coefficients indicated

that there was a strong correlation between independent vari-

ables and obtained results. Therefore, the next step of the design

was performed for aliased effect (X1X2) of the main effects versus

confounded (X3 and X4) effects.

The process is rather complex due to the possible effects of

different factors, not considered in the current study, e.g. the

amount of different organic compounds, their possible interac-

tion, etc. Moreover, to ensure the reproducibility of the experi-

ments, the CCD was performed in random order. In general, all

the data were very reproducible. The experimental data used for

fitting the model, are shown in Table 4. The ultimate model

resulting from ANOVA analysis in terms of the coded factors for

the removal efficiency is represented by the following equations:

Y1¼ 61.62 + 18.61X1 + 6.85X2� 10.12X3 + 19.46X4� 13.85X12�

8.86X22 � 2.03X3

2 � 1.71X1X2 + 6.12X1X3 � 7.18 X2X3 (5)

Y2 ¼ 72.33 + 24.67X1 + 9.33X2 � 14.0X3 � 6.0X12 � 20.0X2

2 �1.71X3

2 � 1.08X1X2 + 5.78X1X3 � 4.15 X2X3 (6)

Y3¼ 54.05 + 18.62X1 + 6.95X2� 11.02X3� 16.92X12 + 4.43X2

2�1.13X3

2 � 1.71X1X2 + 5.30X1X3 � 12.6 X2X3 (7)

For the current model only the coefficients for interacting

factors, which have a value of 0.15 or less were included in the

model. Therefore, among the interacting effects, the concentra-

tion of catalytic adsorbent � H2O2 concentration, catalytic

adsorbent � pH and pH � H2O2 concentration were the most

significant ((prob > F) <0.15), whereas those of catalytic adsor-

bent � temperature, H2O2 concentration � temperature and

temperature � pH were insignificant ((prob > F) > 0.2) in vari-

able ranges investigated and therefore were not considered in the

resulting model equations.

The adequacy of fit of models was possible to confirm by

different criteria (Tables 5–7). The F values of 23.4, 25.6 and 24.8

for COD, TOC and color removal models, respectively, implied

J. Environ. Monit., 2008, 10, 1304–1312 | 1307


Fig. 3 Normal probability plot of main effects (X1 and X2) and their interactions at different levels.

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

that the model was significant, and there was only a 0.01%

chance that a model with F value could occur due to noise. Also,

the lack of fit was found to be non essential (the highest 0.0450)

implying that the obtained model adequately represented the

experimental data. The statistical significance of the model was

also validated by the coefficients of the model determinations

(0.93, 096 and 0.95 for COD, TOC and color, respectively),

which revealed that only 4–7% of the variability in the responses

was not explained by the developed models. The values of the

predicted and observed data are given in Table 4, with the

correlation between the observed and predicted values being

satisfactory (r ¼ 0.97). Therefore, the proposed models were

found suitable for predicting the removal efficiencies in the range

of operation conditions investigated.

Assessment of the results indicated that there were three

interacting factors (concentrations of the catalytic adsorbent and

hydrogen peroxide and pH) that significantly affected the

removal efficiencies. Temperature did not affect the process,

although the temperature interval, chosen for the current study

was relatively low (up to 70 �C). However, temperatures higher

than 100 �C already belong to another group of AOPs and such

investigation is beyond the scope of the present work. From the

practical point of view, pH is one of the most important

parameters influencing the wastewater treatment. Acidic pH

(below 5) favors the efficient performance of hydrogen peroxide.

It has been reported that, at pH above 4, the rapid H2O2

decomposition produces molecular oxygen with an insufficient

amount of generated hydroxyl radicals.14 The effect of pH is

extremely important when adsorbing species are capable of

ionizing in response to the prevailing pH.15 In lower pH (pH < 5)

the majority of dissolved organic contaminants change their

charge and their tendency to adsorb increase.12 It is well known

that organic compounds adsorb poorly when they are ionized.

Acidic species adsorb better at low, and basic species adsorb

better at high pH.15 The experiments demonstrated that the

1308 | J. Environ. Monit., 2008, 10, 1304–1312

removal rates of TOC, COD and color significantly improved

with the decrease in pH. These results were in accordance with

Zhang and Chuang15 who studied the adsorption of organics

from kraft pulp mill wastewater on the activated carbon and

polymeric resin and proved that acidic species prevail in this type

of wastewater. However, it has also been reported that low pH

favors leaching of the active metal.16 Nevertheless, the leaching

of ruthenium in a current study was not detected throughout the

experiments.

Fig. 5 denotes the interaction between pH and the main effects

(1 g l�1 of ruthenium on carbon and 7 ppm of H2O2). The results

clearly showed that an increase in the concentration of reagents

at high pH had no significant effect on the removal of COD,

TOC and color. Moreover, the increase in pH was accompanied

by the substantial decrease in the removal rates. Obviously, high

pH favors the formation of carbonate ions, which are effective

scavengers of hydroxyl ions and therefore can reduce the effi-

ciency of the overall process.14 Furthermore, hydrogen peroxide

is unstable in basic solutions and may deliberately decompose to

O2 and H2O, losing some of its oxidation ability.17 With the

decrease in pH the removal rates were observed to increase and

pH 4 was predicted to be the optimum for removal of COD, TOC

and color, which was also in agreement with the experimental

data. Moreover, it was observed that the removal efficiencies for

COD, TOC and color at pH 4 were higher than these at pH 11,

regardless of the operational temperature of the process. The

remarkable increase in Y was already achieved for the average

level of X3 in comparison to high level, demonstrating the

efficient removal of COD, TOC and color at the initial pH 7 of

the wastewater (53%, 51% and 47%, respectively).

Experimental

All experiments were carried out in a glass reactor, equipped with

a magnetic stirrer and a temperature controller. The reaction



Fig. 4 Design expert plot. 3D surface and contour plot of the removal

efficiency showing the interaction effect of Ru on C and H2O2 concen-

trations for (a) COD, (b) TOC, and (c) color removal efficiencies at pH¼7 and ambient temperature.

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

mixture was stirred at a speed of 750 rpm for 120 min to provide

a complete mixing of the solution for uniform distribution

and full suspension of the catalytic adsorbent particles. For

experiments, unless specified otherwise, ruthenium on carbon

(1–3 g L�1) was introduced as a suspension to the reaction

medium with the following (1–15 ppm) H2O2 (30%) addition. To


study the effect of different operation parameters on the

degradation of the contaminants in order to optimize the

process, the reaction was carried over a wide range of pH (1–7,

11) and various temperatures (ambient, 40, 60 �C). Samples were

taken at regular intervals for the subsequent analysis and filtered

through the syringe filter (0.45 mm hydrophilic Millipore filter)

to separate the catalytic adsorbent particles from the solution.

The pH of filtrate was then determined by a pH meter

(3401 WTW, Germany). The progress of contaminants degra-

dation was monitored by measuring total organic carbon

(TOC analyser TOC–5000, Shimadzu Seisakusho Co., Japan),

chemical oxygen demand (COD) according to the closed reflux,

titrimetric method18 and colour. Colour values were obtained by

comparing spectrophotometric readings of samples to those of

platinum cobalt standards.19 The spectrophotometric analyses

were carried out at 465 nm, using a Hach Lange spectropho-

tometer (DR 2800, Germany). Pulp mill effluent was supplied

by Sunila Oy, located in Kotka, Finland. The pulp mill produces

long-fibre bleached softwood pulp. Composition of the pulp

mill wastewater used in this study is presented in Table 8.

Hydrogen peroxide solution (30%, w/w) was purchased from

J.T. Baker (Germany). The pH of aqueous solutions was

adjusted using sodium hydroxide and sulfuric acid solutions

(1%). All the solutions were prepared by using M-Q water

(18 MU cm�1 at 25 �C).

Catalytic adsorbent characterization

Ruthenium 5wt% on carbon was purchased from Aldrich

(Germany). Prior to experiments, catalytic adsorbent samples

were calcinated at 400 �C for 4 h. The textural properties such as

the specific surface area, pore volume and the pore size distri-

bution were determined using a Quantachrome Autosorb system

(Quantachorome instruments, UK) with nitrogen gas as adsor-

bate. Particle size distribution was assessed by means of

a dispersion analyser LUMIsizer (L.U.M. GmbH, Germany).

The filtrate was subjected to ICP–OES (iCAP 6000, UK) analysis

to assess the potential leaching of Ru ions to the solution, as

a result of ruthenium on carbon dissolution with the detection

limit for Ru ions of 0.003151 ppm.

SEM measurements

The surface morphology of the catalytic adsorbent samples was

observed by scanning electron microscopy (SEM) using a Leo

Gemini 1530 apparatus (USA), EDX at 20 kV for the imaging in

in-lens mode at 10 kV.

Experimental design

The classical approach of changing one variable at a time to

study the effects of variables on the response is a time consuming

method particularly for multivariable systems and also when

more than one response is considered. Statistical design of

experiments reduces the number of experiments to be performed,

considers interactions among the variables and can be used for

the optimization of the operating parameters in multivariable

systems.20

J. Environ. Monit., 2008, 10, 1304–1312 | 1309


Table 4 Results of CCD, FFD (marked with asteriks) conditions and the experimental as well as predicted values for COD (%), TOC (%) and color (%)(Y1, Y2 and Y3, respectively)

Run

Coded levels of variables Y1 (%) Y2 (%) Y3 (%)

X1 X2 X3 X4 Exp Pred Exp Pred Exp Pred

1 1 �1 �1 0 62.12 62.00 60.15 60.05 48.23 48.002 0 0 0 0 55.23 55.00 53.67 53.67 43.74 43.003 �1 �1 �1 1 47.00 47.00 45.12 45.12 33.62 33.004 �1 �1 �1 �1 46.12 46.00 46.87 45.87 34.41 34.005 �1 1 �1 0 57.32 57.00 55.01 55.51 49.64 49.006 1 1 �1 �1 72.46 69.20 70.28 68.20 68.01 61.897 1 1 1 1 40.87 38.54 39.57 37.91 35.00 34.308 �1 1 1 1 36.35 37.46 33.22 34.88 33.15 33.709 0 0 1 �1 45.13 45.00 44.29 44.29 39.37 39.0010 0 0 �1 �1 75.12 74.40 72.81 72.77 68.82 66.7811 1 0 �1 �1 60.00 66.93 59.01 64.79 58.69 61.9012 1 �1 1 �1 55.03 55.00 56.83 56.83 49.01 49.0013 0 0 1 0 51.64 51.00 49.75 49.75 36.94 36.0014 0 1 0 0 51.54 50.74 48.92 48.58 36.43 36.4115 �1 �1 1 1 34.23 32.54 33.56 31.90 27.13 26.3016 �1 �1 1 �1 34.67 34.00 33.12 33.12 27.00 27.0017 1 �1 0 1 45.46 45.00 43.25 43.25 37.47 37.0018 1 0 �1 0 77.14 74.87 76.38 73.27 67.84 66.2119 1 �1 1 1 35.36 34.73 33.05 34.09 27.19 27.3520 0 1 1 1 34.98 34.00 33.96 33.96 27.17 27.0021 0 1 �1 �1 64.00 64.34 63.74 63.45 56.73 57.6322 1 0 �1 �1 73.46 66.93 70.19 64.79 65.87 61.9023 1 0 �1 0 73.19 74.87 70.49 73.27 65.00 66.2124 0 1 �1 0 67.45 67.26 67.57 67.91 56.34 55.5925 1 1 �1 �1 67.12 69.20 66.16 68.20 57.24 61.8926 1 �1 1 1 33.13 34.73 33.48 34.09 27.37 27.3527 0 0 0 1 34.25 34.00 32.29 32.29 27.72 27.0028 1 0 0 0 56.76 56.26 55.00 55.34 48.51 47.5929* �1 0 0 �1 44.36 44.00 45.95 45.95 35.63 35.0030* 0 1 0 �1 52.32 51.66 49.18 49.47 45.00 43.3731* 1 �1 0 �1 56.05 54.54 53.48 51.82 49.03 48.3032* �1 �1 0 �1 40.00 41.46 37.71 39.37 33.23 33.7033* 0 0 0 �1 56.97 56.60 55.75 55.79 48.64 49.2234* 1 0 0 �1 53.23 52.14 52.58 52.20 49.00 48.1935* �1 1 0 �1 49.14 47.54 47.00 45.34 38.07 37.3036* 1 1 0 �1 62.63 59.53 61.37 58.63 54.12 51.4637* 1 1 0 �1 55.00 59.53 54.18 58.63 47.67 51.46

Table 5 ANOVA test for the response function Y1 (COD removalefficiency, %)

SourceSum ofsquares DF

Meansquare F value Prob > F

Model 6107.081 9 605.4056 12.9657 0.000070X1 137.136 1 68.56812 3.659755 0.000138X2 60.394 1 30.19695 1.611732 0.000747X3 136.803 1 68.40164 3.650869 0.000402X1

2 38.367 1 19.18327 1.023888 0.000237X2

2 97.437 1 32.47915 1.733542 0.000419X3

2 82.426 1 20.60657 1.099855 0.000166X1X2 160.714 1 40.17858 2.144491 0.007878X1X3 235.531 1 58.88281 3.142811 0.000928X2X3 0.163 1 0.16258 0.008678 0.000564Residual 59.055 16 14.76364 0.787995Lack of fit 51.23 11Pure error 145.36 8Cor. total 6598 25SD 11.6 R squared 0.940475Mean 55.876 Adj R

squared0.89765

CV 23.587 Pred rsquared

0.758493

Press 7654.1 Adeqprecision

10.6543

Table 6 ANOVA test for the response function Y2 (TOC removal effi-ciency, %)



Model 7903.382 9 705.4998 13.9543 0.000045X1 175.169 1 175.169 4.754701 0.000521X2 80.854 1 80.854 3.632214 0.000312X3 145.004 1 145.004 2.690822 0.000500X1

2 27.520 1 27.520 1.887932 0.000399X2

2 87.860 1 87.860 1.542280 0.000707X3


squared0.91235


0.76432


11.345

1310 | J. Environ. Monit., 2008, 10, 1304–1312 This journal is ª The Royal Society of Chemistry 2008

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online


Table 7 ANOVA test for the response function Y3 (color removal effi-ciency, %)



Model 6068.108 9 6068.108 12.8553 0.000060X1 128.441 1 128.441 4.039830 0.000446X2 51.211 1 51.211 1.610743 0.000018X3 83.968 1 83.968 2.641015 0.000248X1

2 53.418 1 53.418 1.680151 0.000943X2

2 54.767 1 54.767 1.148390 0.000831X3


squared0.942147


0.786742


11.8675

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

Half-fractional factorial design (H-FFD)

At the initial stage of the experimental design, a half-fractional

factorial design (33�1), H-FFD was accomplished at three

levels in order to get a general view of the significant factors

and possible noteworthy interactions in the studied range of

variables. Table 3 demonstrates the coded variables and

confounded effects of that design.

Fig. 5 Design expert plot. 3D surface and contour plot of the removal

efficiency showing the interaction effect of Ru on C and H2O2 concen-

trations and various pH for (a) COD, (b) TOC, and (c) color removal

efficiencies.

Central composite design (CCD)

During the next step, a central composite design was employed to

determine the optimal conditions for the critical factors. A

modified central composite experimental design known as the

Box–Behnken design is an independent, rotatable quadratic

design with no embedded factorial or fractional factorial points

where the variable combinations are at the midpoints of the

edges and at the center of the variable space. Among all the

response surface methodology (RSM) designs, Box–Behnken

design requires fewer runs than the others, e.g., 15 runs for

a three-factor experimental design. The actual values of inde-

pendent variables (Xi) were coded assigning the lowest values

listed in Table 3 as �1 and the highest values as +1, according to

the following equation:

xi ¼ Xi � xi/Dxj (8)

where xi was dimensionless value of independent variable, Xi

represented the real value of the independent variable, xi was the

real values of the independent variable at the center point, and

Dxj was the step charge.

The CCD allowed the response surface to be modelled by

fitting a second-order polynomial with a number of experiments

equal to 2f + 2f + n, where f and n were numbers of factors and

center runs, respectively (f ¼ 4 and n ¼ 2). The repetition of the

central runs was carried out to obtain the data on the variation of


the responses in average, the residual variance, and eventually to

evaluate the pure experimental uncertainty. Using a four factor–

three coded level (Table 4) central composite design (CCD),

28 runs were carried out to fit the general model, as following:

Y ¼ b0 +P

biXi +P

biiXi2 +

PPbijXiXj (9)

J. Environ. Monit., 2008, 10, 1304–1312 | 1311


Table 8 Main characteristics of the pulp mill effluent

Parameter Unit Value

COD ppm 510TOC ppm 220Color PtCo 540AOX ppm 5.4BOD7 ppm 8pH 7.1

Publ

ishe

d on

03

Oct

ober

200

8. D

ownl

oade

d by

Sta

te U

nive

rsity

of

New

Yor

k at

Sto

ny B

rook

on

25/1

0/20

14 2

1:07

:59.

View Article Online

where Y was the coded removal efficiency (%) of COD, TOC and

color, and bs were model coefficients calculated from the

experimental data.

The reproducibility of data in a complex process such as the

degradation of contaminants is a very important matter;

therefore, CCD experiments were performed twice and in

randomized order to minimize the effect of uncontrolled factors

and time trends. The average data was used as the final values for

the development of a current model. STATISTICA 7 (USA)

software was used for regression and graphical analysis of data.

Conclusion

Introduced novel catalytic adsorbent benefited higher removal

rates of contaminants regarding COD, TOC and color, better

reusability and regenerability as well as unaltered initial pH of

the wastewater than the traditional adsorbent. The optimum

operational parameters for effective treatment of pulp mill

effluent were determined and substantiated (1 g l�1 of ruthenium

on carbon, 7 ppm of H2O2, pH 4 and ambient temperature).

Removal rates demonstrated were 75, 73 and 68% for COD,

TOC and color, respectively. The valid effective models were

developed and explained using Box–Behnken design and CCD.

The major advantage of the used second-order polynomial

model was found to be its flexibility. It means that if there were

any limitations imposed by the operational and economical

consideration on the values of some factors, it was possible to

calculate new optimum conditions in which the effects of these

limitations would be compensated by the other factors.

1312 | J. Environ. Monit., 2008, 10, 1304–1312

Acknowledgements

Mikkeli Technology Center, Mikkeli University Consortium

(MUC), EU Objective 1 and Academy of Finland (decision

number 212649) are acknowledged for their financial support of

research.

References

1 E. C. Catalkaya and F. Kargi, J. Environ. Manage., 2008, 87, 396–404.

2 H. Nurmesniemi, R. Poykio and R. L. Keiski, Waste Manage., 2007,27, 1939–1948.

3 M. Kitisa, E. Karakayaa, N. O. Yigita, G. Civelekoglua andA. Akcilb, Water Res., 2005, 39, 1652–1662.

4 G. Centi, M. Gotti, S. Perathoner and F. Pinna, Catal. Today, 2000,55, 51–60.

5 A. Georgi and F.-D. Kopinke, Appl. Catal., B, 2005, 58, 9–18.6 W. P. Kwan and B. M. Voelker, Environ. Sci. Technol., 2004, 38,3425.

7 J. Trawczynski, Carbon, 2003, 41, 1515–1523.8 A. Rodas-Grapaın, J. Arenas-Alatorre, A. Gomez-Cortes andG. Dıaz, Catal. Today, 2005, 107–108, 168–174.

9 D. J. Moon, M. J. Chung, K. Y. Park and S. I. Hong, Carbon, 1999,37, 123–128.

10 Yu. Yegiazarov, J. Clark, L. Potapova, V. Radkevich, V. Yatsimirskyand D. Brunel, Catal. Today, 2005, 102–103, 242–247.

11 M. Sheintuch and Y. I. Matatov-Meytal, Catal. Today, 1999, 53, 73–80.

12 Q. U. Jiuhui, J. Environ. Sci., 2008, 20, 1–13.13 C. Brasquet, B. Rousseau, H. Estrade-Szwarckopf and P. Le Cloirec,

Carbon, 2000, 38, 407–422.14 C. W. Jones, Applications of hydrogen peroxide and derivatives, Royal

Society of Chemistry, Cambridge, UK, 1999.15 Q. Zhang and K. T. Chuang, Adv. Environ. Res., 2001, 3, 251–258.16 R. Molina, F. Martinez, J. A. Melero, D. H. Bremner and

A. G. Chakinala, Appl. Catal., B, 2006, 66, 198–207.17 C. Catrinescua, C. Teodosiua, M. Macoveanua, J. Miehe-Brendleb

and R. Le Dredb, Water Res., 2003, 37, 1154–1160.18 Water Analysis, Guidelines for the Determination of Chemical Oxygen

Demand (COD), Finnish Environmental Institute, Helsinki, Finland,SFS 5504, 1988.

19 Method NCASI 253, Analytical Methods for the Determination ofPollutants in Pulp and Paper Industry Wastewater, EPA 821-R-93–01, U.S. Environmental Protection Agency, Office of Water,Washington, DC, 1993.

20 Z. R. Lazic, Design of Experiments in Chemical Engineering, Wiley-VCH, Weinheim, 2004.



Documents

Optimization of pulp mill effluent treatment with catalytic adsorbent using orthogonal second-order (Box–Behnken) experimental design