ISSN 1451 - 9372(Print) ISSN 2217 - 7434(Online) APRIL ... No2_p229-367_Apr … · N.S. TEHRANI et...

ISSN 1451 - 9372(Print)ISSN 2217 - 7434(Online)APRIL-JUNE 2015Vol.21, Number 2, 229-367

www.ache.org.rs/ciceq

Journal of the Association of Chemical Engineers of Serbia, Belgrade, Serbia

Vol. 21 Belgrade, April-June 2015 No. 2

Chemical Industry & Chemical EngineeringQuarterly (ISSN 1451-9372) is published

quarterly by the Association of ChemicalEngineers of Serbia, Kneza Miloša 9/I,

11000 Belgrade, Serbia

Editor:Vlada B. Veljković

[email protected]

Editorial Office:Kneza Miloša 9/I, 11000 Belgrade, Serbia

Phone/Fax: +381 (0)11 3240 018E-mail: [email protected]

www.ache.org.rs

All the manuscripts are not to be returned

For publisher:Tatijana Duduković

Secretary of the Editorial Office:Slavica Desnica

Marketing and advertising:AChE Marketing Office

Kneza Miloša 9/I, 11000 Belgrade, SerbiaPhone/Fax: +381 (0)11 3240 018

Publication of this Journal is supported by theMinistry of Education and Science of the

Republic of Serbia

Subscription and advertisements make payableto the account of the Association of Chemical

Engineers of Serbia, Belgrade, No. 205-2172-71, Komercijalna banka a.d., Beograd

Computer typeface and paging:Vladimir Panić

Printed by:Faculty of Technology and Metallurgy,

Research and Development Centre of PrintingTechnology, Karnegijeva 4, P. O. Box 3503,

11120 Belgrade, Serbia

Abstracting/Indexing:Articles published in this Journal are indexed inThompson Reuters products: Science Citation

Index - ExpandedTM - access via Web of Science®, part of ISI Web of KnowledgeSM

CONTENTS

Nazila Samimi Tehrani, Ghasem D. Najafpour, Mostafa Rahimnejad, Hossein Attar, Performance of upflow anaerobic sludge fixed film bioreactor for the treatment of high organic load and biogas production of cheese whey wastewater ................................................................ 229

Seyed Majid Ataei Ardestani, Morteza Sadeghi, Babak Beh-eshti, Saeid Minaei, Naser Hamdami, Vibro-fluidized bed heat pump drying of mint leaves with respect to phenolic content, antioxidant activity and color indices ..... 239

Marija S. Petrović, Tatjana D. Šoštarić, Lato L. Pezo, Slavka M. Stanković, Časlav M. Lačnjevac, Jelena V. Miloj-ković, Mirjana D. Stojanović, Usefulness of ANN-based model for copper removal from aqueous solutions using agro industrial waste materials .................. 249

Bore V. Jegdić, Biljana M. Bobić, Miloš K. Pavlović, Ana B. Alil, Slaviša S. Putić, Stress corrosion cracking resist-ance of aluminum alloy 7000 series after two-step aging ................................................................................... 261

Aleksandra Petrovič, Marjana Simonič, The efficiency of a membrane bioreactor in drinking water denitrification........ 269

Dragutin M. Nedeljković, Marija P. Stevanović, Mirko Z. Stije-pović, Aleksandar P. Stajčić, Aleksandar S. Grujić, Jasna T. Stajić-Trošić, Jasmina S. Stevanović, The possibility of application of zeolyte powders for the construction of membranes for carbon dioxide separ-ation .................................................................................... 277

Veselinka Grudić, Jelena Šćepanović, Ivana Bošković, Rem-oval of cadmium (II) from aqueous solution using fermented grape marc as a new adsorbent ........................ 285

Amal Juma Habish, Slavica Lazarević, Ivona Janković-Čas-tvan, Branislav Potkonjak, Đorđe Janaćković, Rada Petrović, The effect of salinity on the sorption of cadmium ions from aqueous medium on Fe(III)-sep-iolite .................................................................................... 295

Shaopeng Guo, Lina Lv, Jia Zhang, Xin Chen, Ming Tong, Wanzhong Kang, Yanbo Zhou, Jun Lu, Simultaneous removal of SO2 and NOx with ammonia combined with gas-phase oxidation of NO using ozone .............................. 305

Jau-Kai Wang, Jir-Ming Char, Optimization study on hardness of gold film through supercritical electroplating process by response surface methodology ...................................... 311

Contents continued Lawrence Koech, Ray Everson, Hein Neomagus, Hilary Rutto, Dissolution kinetics of south african coal fly ash and the development of a semi-empirical model to predict dissolution ............................................................... 319

Zhuoni Hou, Xianrui Liang, Feng Su, Weike Su, Preparative isolation and purification of seven compounds from Hibiscus mutabilis L. leaves by two-step high-speed counter-current chromatography ........................................ 331

Ruifang Zhao, Yulong Wang, Yonghui Bai, Yongfei Zuo, Lun-jing Yan, Fan Li, Effects of fluxing agents on gasi-fication reactivity and gas composition of high ash fusion temperature coal ......................................................... 343

Veselinka V. Grudić, Nada Z. Blagojević, Vesna L. Vukašinović-Pešić, Snežana R. Brašanac, Kinetics of degradation of ascorbic acid by cyclic voltammetry method ................................................................................... 351

Ahmet Ozan Gezerman, Burcu Didem Çorbacıoğlu, Effects of sodium silicate, calcium carbonate and silicic acid on ammonium nitrate degradation and analytical invest-igations of the degradation process on an industrial scale ................................................................................... 359

Chemical Industry & Chemical Engineering Quarterly

Available on line at

Association of the Chemical Engineers of Serbia AChE www.ache.org.rs/CICEQ

Chem. Ind. Chem. Eng. Q. 21 (2) 229−237 (2015) CI&CEQ

229

NAZILA SAMIMI TEHRANI1

GHASEM D. NAJAFPOUR2

MOSTAFA RAHIMNEJAD2

HOSSEIN ATTAR1 1Department of Chemical Engineering, Science and

Research Branch, Islamic Azad University, Tehran, Iran

2Biotechnology Research Lab, Faculty of Chemical Engineering,

Noshirvani University of Technology, Babol, Iran

SCIENTIFIC PAPER

UDC 60:66.02]:628.1/.3

DOI 10.2298/CICEQ131105018T

PERFORMANCE OF UPFLOW ANAEROBIC SLUDGE FIXED FILM BIOREACTOR FOR THE TREATMENT OF HIGH ORGANIC LOAD AND BIOGAS PRODUCTION OF CHEESE WHEY WASTEWATER

Article Highlights • Dairy wastewater was successfully treated in UASFF with HRT of 48 h • The amount of collected biogas was 2.4 L/d. The obtained biomethane had purity of

61% • COD removal rate of 80% for organic loading rate (OLR) of 25.85 g COD/Ld • Among the fitted models, Riccati model fitted in good agreement with the experi-

mental data Abstract

Among various wastewater treatment technologies, biological wastewater treatment appears to be the most promising method. A pilot scale of hybrid anaerobic bioreactor was fabricated and used for whey wastewater treatment. The top and bottom of the hybrid bioreactor known as upflow anaerobic sludge fixed film (UASFF); was a combination of upflow anaerobic sludge blanket (UASB) and upflow anaerobic fixed film reactor (UAFF), respectively. The effects of operating parameters such as temperature and hydraulic retention time (HRT) on chemical oxygen demand (COD) removal and biogas pro-duction in the hybrid bioreactor were investigated. Treatability of the samples at various HRTs of 12, 24, 36 and 48 h was evaluated in the fabricated bio-reactor. The desired conditions for COD removal such as HRT of 48 h and operation temperature of 40 °C were obtained. The maximum COD removal and biogas production were 80% and 2.40 (L/d), respectively. Kinetic models of Riccati, Monod and Verhalst were also evaluated for the living microorg-anisms in the treatment process. Among the above models, Riccati model was the best growth model fitted with the experimental data with R2 of about 0.99.

Keywords: COD, Riccati model, UASFF bioreactor, high organic load, cheese whey wastewater.

In cheese processing plants, large quantities of wastewater are generated. The wastewater consists of high biological oxygen demand (BOD) and chem-ical oxygen demand (COD) with concentrations of 40 and 50–70 g/L, respectively [1]. Dairy wastewaters can have different characteristics, depending on the product obtained (yogurt, cheese, butter, milk, ice cream, etc.). Some other parameters such as waste-

Correspondence: N.S. Tehrani, Department of Chemical Eng-ineering, Science and Research Branch, Islamic Azad Univer-sity, Tehran, Iran. E-mail: [email protected] Paper received: 5 November, 2013 Paper revised: 3 May, 2014 Paper accepted: 5 June, 2014

water management, operating conditions and also types of process cleaning may influence the dairy effluents characterization [2,3].

In process of cheese production, whey is gener-ated from the liquid residue when casein and fat are separated through coagulation of milk. It is a by-pro-duct of cheese or casein, and has several commercial uses, such as concentrated whey protein for food additives and nutritional diets [4]. Whey contains lact-ose (70–75%) and soluble proteins (10–15%) which results in a high COD (50–70 g/L) [5]. High concen-tration of organic matter in whey wastewater such as whey causes serious pollution problems to surround-ings [6,7]. If cheese effluent is discharged to environ-

N.S. TEHRANI et al.: PERFORMANCE OF UPFLOW ANAEROBIC SLUDGE… Chem. Ind. Chem. Eng. Q. 21 (2) 229−237 (2015)

230

ment causes depletion of the dissolved oxygen in the stream that may jeopardize the aquatic life. In addi-tion the contaminated water body in irrigation gradu-ally reduces the land fertilities [8-10].

World annual production of whey is around 115 million tons; approximately 47% of the produced whey is disposed into the environment [11]. Since whey naturally contains lactose and biodegradable organic matter, biological treatment is a practical process [12]. Among biological treatment processes, treatment in ponds, activated sludge plants and anaerobic treat-ment are commonly employed [13]. Anaerobic method for the treatment of whey wastewater is attractive and often catches the attention of researchers because of the presence of high organic content in the waste, low energy requirement of the process, less sludge pro-duction and generation of fuel in form of methane [14]. Dairy wastewater has been extensively treated by coagulation/flocculation and sedimentation proces-ses [3,15]. The main disadvantages of these pro-cesses are high coagulant costs, high sludge pro-duction, and poor removal of COD [16]. Therefore, biological treatment is usually preferred for dairy wastewater treatment [17,18]. Due to high organic content of whey low treatment costs, anaerobic dig-estion processes are recommended [19,20].

Recently, an upflow anaerobic reactor has been successfully employed for treatment of dairy waste-water in full-scale applications [21]. The use of a lab-oratory-scale bioreactor for the treatment of dairy wastewater at an operational temperature of 30 °C was previously investigated [22]. It was found that COD removal varied between 85 and 88% at a hyd-raulic retention time (HRT) of 5.0 h and organic loading rates (OLRs) of 8.5 g COD/Ld. Another lab-oratory-scale investigation resulted in over 97% of COD removal using cheese whey wastewater in an anaerobic treatment process [23]. At a high OLR of 31 g COD/Ld, the UASB reactor treating cheese pro-duction wastewater provided removal of 90% for COD

[24]. There is no sufficient information available in the literature regarding the COD removal from cheese whey wastewater in UASFF hybrid bioreactors. The main objective of this work is to enhance more thor-oughly the treatment of cheese whey wastewater in a hybrid bioreactor. The effects of operating parameters such as temperature and HRT on COD removal and biogas production in the hybrid bioreactor were inves-tigated.

MATERIALS AND METHODS

Wastewater characterization

The cheese whey wastewater used in present work was collected from a local cheddar cheese processing industry. The characteristics of cheese whey wastewater samples are presented in Table 1; the data were compared with Normex (Mexico) [25] and cheese processing wastewater [26]. The col-lected wastewater from Gela factor was obtained from the plant discharge known as cheese whey waste-water, which was homogeneously mixed and divided into small portions and kept at 4 °C for further use. Cheese whey wastewater contains most of the essen-tial nutrients for microbial growth without additional nutrients to fresh substrate.

Hybrid bioreactor setup and start up

The hybrid reactor was constructed from a cyl-indrical pyrex column with internal diameter of 2.76 cm and total height of 160 cm. The total volume of the reactor was 960 ml. The density and specific surface of the packing media were 400 kg/m3 and 500 m2/m3, respectively. The shape of packing used in UASFF is similar to honeycomb shape of polyethylene material with high surface area. A funnel shaped gas separ-ator was used to liberate the generated biogas from the effluent, and then the gas was led to a gas col-lector tank. The gas tank was a cylindrical glass pipe with an internal diameter of 80 mm and height of 1 m. The liberated gas was frequently measured at a def-

Table 1. Characteristics of cheese whey wastewater

Cheese processing wastewater [26]Normex, mg/L [25] Gela, mg/L Parameter

60000 54000-77300 50000-70000 COD - - 27000-36000 BOD

- 3900-58900 55000-65000 TS

2500 - 10000-15000 TSS

830 500-5600 10-20 TKN

280 - 40-60 P-PO4

- - 50000-60000 Lactose

- - 200-400 Alkalinity

- 4.3-8.7 6-6.5 pH


231

ined HRT and the gas volume was recorded with respect to time. The gas tank was initially filled with water which was saturated with methane. The volume of liberated gas was determined by the displacement of water in the gas tank. Figure 1a and b show a schematic diagram and actual of the experimental setup, respectively.

The system was operated at fixed temperatures, 24 and 40 °C with the aid of temperature-controller. The cheese whey wastewater sample was often col-lected from Gela factory (Amol, Iran). A sufficient large storage tank (1 m3) was installed in the settle-ment to collect the cheese whey wastewater from the inlet sewer pipe of the first septic tank. The collected samples of wastewater were stored in the refrigerator at 4 °C to minimize substrate decomposition before any primary treatment. The inoculums for seeding were obtained from anaerobic sludge pound in Gela dairy industrial plant (Amol, Iran).

Bioreactor operation and monitoring

COD and BOD of the raw and treated waste-water were analyzed according to methodology deve-

loped by American Public Health Association (APHA) and American Water and Wastewater Association (AWWA) standards [27]. The withdrawn duplicated samples were analyzed and average value was rec-orded. Biogas production was volumetrically mea-sured by displacement of water in a cylindrical col-umn. Methane content was determined by gas chro-matography, using a GC model 7890A (Agilent Tech-nologies, USA) equipped with TCD detector and packed column Carboxen 1000 was used for the gas analysis.

Kinetic models

Biological models are used to determine the rel-ationships between variables. These models were used to control and optimize the treatment in labor-atory scale UASFF bioreactor. In this study, an unstructured model describing the exponential cell growth was used. In addition, other models were used to obtain the desired model for the microorganism’s growth. Riccati, Monod and Verhalst models were evaluated to define the related kinetic parameters (Table 2).

Figure 1. a) Schematic diagram of the experimental setup; b) actual experimental setup. 1. Feed tank, 2. fee pump, 3. sampling valve,4. UASB reactor, 5. UAFF reactor, 6. outlet stream, 7. gas and liquid separator, 8. exhaust gas valve, 9. settling tank, 10. treated

wastewater, 11. recycle pump, 12. gas exhaust valve, 13. valve connected to vacuum pump, 14. vacuum pump and 15. biogas collector.


232

Table 2. Evaluated kinetic models

Kinetic model Non-linear model Linear model

Riccati μ

μ=

− −

0

01 (1 )

m

m

t

t

m

X eX

Xe

X

-

Monod μμ =+

m

S

SK S

μ μ μ

= +1 1 1S

m m

KS

Verhalst μμ μ= − mm

mX

X

μμ μ= − mm

mX

X

RESULTS AND DISCUSSION

Reactor performance

The COD and pH of the cheese whey waste-water were daily measured for the duration of 34 days. The bioreactor operation and variation of pH during the startup of the system was monitored. Figure 2a shows variations in influent and effluent COD over the period of 34 days. Initial COD has increased in five steps and gradually shifted from 4000 to 14000 mg/L. The COD removal has inc-reased from 15 to 62%. The effluent pH altered when

feed was added due to initial pH of the feed; however most of time, the effluent pH was relatively constant nearly recorded at 6.5. The pH of bioreactor varied from 5.5 to 7 throughout the experiment; there was slight variation in pH which was unnecessary to make any adjustment; this range is suitable for anaerobic microbes. Outlet pH was almost neutral (approxim-ately 6.5), and the performance of bioreactor was high and stable. Steady state conditions were achieved at stable conditions with high COD removal. Fluctu-ations of pH are shown in Figure 2b.

Effect of HRT and OLR in COD removal

After start-up period, COD removal efficiency was examined at different OLR with fixed HRT at 24 °C. OLR was gradually increased with specific HRT from 12 to 48 h. Figure 3a also shows that an inc-rease in HRT and OLR, improved COD removal. Max-imum COD removal efficiency was 59% at HRT of 48 h and OLR of 25.61 (g COD/Ld). An increase in HRT resulted in an increase in contact time between wastewater and granular sludge, and enhanced COD removal. Figure 3b shows the COD removal at var-ious organic loading rate (OLR) and constant tempe-

0

2000

4000

6000

8000

10000

12000

14000

16000

0 10 20 30 40

CO

D (

mg/

L)

Time (d)

COD in

COD out

(a)

4.8

5.3

5.8

6.3

6.8

7.3

0 5 10 15 20 25 30 35 40

Eff

luen

t p

H

Time (d)

(b)

Figure 2. Examination of the wastewater system: a) COD variation and b) pH variation. Experimental conditions at start up: stepwise input of COD from 4248 to 13845 mg/L at room temperature.


233

rature of 40 °C. The COD removal was also increased with increase in HRT and OLR. Maximum obtained COD removal efficiency was 80% at an HRT of 48 h, OLR of 25.85 (g COD/Ld) at 40 °C. It can be seen that an increase in microbial activity at high tempe-rature, removal efficiency was improved from 59 to 80% when the bioreactor temperature was increased from 24 to 40 °C. As Figure 3 illustrates, for each HRT there was a maximum in OLR removal. Beyond the maximum value at desired HRT, the OLR removal decreased due to overloading rate of OLR.

The present data were compared with other work for the biological treatment of cheese whey wastewater. It was found that Frigon et al. [5] have treated cheese whey wastewater by sequential anaer-obic and aerobic steps in a single digester for dur-ation of 4 days; they removed about that 88% of bio-degradable COD. Gavala et al. [26] treated dairy wastewater in UASB for HRT of about 20 days; their reported COD removal was 81-85%. Monroy et al. [28] have also used dairy wastewater treatment in anaerobic filtration; a 70% COD removal was rep-

orted for HRT of 4 days. In present work an 80% of total COD at HRT of 48 h was removed.

Effect of initial COD on effluent pH

The effluent pH varied due to additional fresh feed and its initial pH. Figure 4a and b shows the effluent pH with initial COD at two different tempe-ratures. The pH of effluent increased at high initial COD. For constant initial COD, effluent pH and COD removal increased with an increase in HRT. It showed that for anaerobic bacteria at high HRT, an expected and desired range of pH would be 6.7-7.0. Therefore, COD removal was increased. Because COD removal had an increasing trend at high temperature, the effluent pH was slightly lowered. This was due to high biological activities and increasing trend of COD rem-oval, which resulted in activity of acid formers in the anaerobic process.

Biogas production

The effect of OLR and HRT on the biogas pro-duction rate is shown in Figure 5a and b. Accord-

0%

10%

20%

30%

40%

50%

60%

70%

0 20 40 60 80 100 120 140

CO

D R

emov

al (

%)

OLR (g COD/L.d)

HRT=12 h

HRT= 24 h

HRT= 36 h

HRT= 48 h

(a)

0%

20%

40%

60%

80%

100%

0 20 40 60 80 100 120 140

CO

D R

emov

al (

%)

OLR (g COD/L.d)

HRT=12 h

HRT=24 h

HRT=36 h

HRT=48 h

(b)

Figure 3. Evaluation of COD removal in different OLR: a) at 24 °C, effluent pH 5.5-5.8, influent COD 8430-63240 mg/L; b) at 40 °C, effluent pH 5.8-6.9, influent COD 9247-65211 mg/L.

N.S. TEHRANI et al.: PERFORMANCE OF UPFLOW ANAEROBIC SLUDGE… CI&CEQ 20 (2) 229−237 (2015)

234

5.4

5.6

5.8

6

6.2

6.4

0 10000 20000 30000 40000 50000 60000 70000

Eff

luen

t pH

COD in (mg/L)

HRT=12 h

HRT=24 h

HRT=36 h

HRT= 48 h

(a)

5.5

5.9

6.3

6.7

7.1

7.5

0 10000 20000 30000 40000 50000 60000 70000

Eff

luen

t pH

COD in (mg/L)

HRT= 12 h

HRT= 24 h

HRT= 36 h

HRT= 48 h

(b)

Figure 4. Evaluation of pH effluent in different OLR at: a) 24 and b) 40 °C.

0

0.2

0.4

0.6

0.8

1

1.2

0 50 100 150

Bio

gas

pro

du

ctio

n r

ate

(L/d

)

OLR (g COD/L.d)

HRT=12 hHRT=24 hHRT=36 hHRT=48 h

(a)

0

0.5

1

1.5

2

2.5

3

0 20 40 60 80 100 120 140

Bio

gas

pro

du

ctio

n r

ate

(L/d

)

OLR (g COD/L.d)

HRT=12 h

HRT=24 h

HRT=36 h

HRT=48 h

(b)

Figure 5. Evaluation of biogas production in different OLR: a) at 24 °C, effluent pH 5.5-5.8, influent COD 8430-63240 mg/L; b) at 40 °C, effluent pH 5.8-6.9, influent COD 9247-65211 mg/L.

N.S. TEHRANI et al.: PERFORMANCE OF UPFLOW ANAEROBIC SLUDGE… CI&CEQ 20 (2) 229−237 (2015)

235

ing to Figure 5a at 24 °C, amount of biogas was increased by increasing OLR at a constant HRT. Maximum biogas (1.02 L/d) was produced at OLR of 25.61 (g COD/Ld) and constant HRT 48 h. Figure 5b shows effect of OLR on biogas production rate at 40 °C. Biogas production rate was also increased with increasing OLR. With an increase in OLR, biodeg-radability of the substrate and sufficient microbial community was increased and then the biogas pro-duction will increase [29,30]. It is very important that because of removal efficiency was improved at high temperature (40 °C), the amount of biogas production was also slightly increased at low temperature (24 °C). The amount of 2.40 L/d was obtained at high HRT (48 h), OLR (32.60 (g COD/Ld)) and tempe-rature (40 °C). GC analysis of the biogas at optimum condition showed that almost 61% of biogas was methane and the remaining gas was CO2. The obtained results showed that there is a suitable con-dition for UASFF hybrid bioreactor to operate for max-imum COD removal.

Evaluation of kinetic models

The utilization of organic matters and sludge productions are predicted by various microbial kinetic growth models. Monod and Verhalst models are simple models for anaerobic degradation of organic compounds in cheese whey wastewater. Table 3 summarizes the kinetic parameters for all 3 presented models with non-linear and linear equations. The kinetic data for all of the stated models showed that experimental data may be corresponded with theo-retical concepts and the models were appropriate for predicting the performance of fabricated UASFF bio-reactor. The regression coefficients and the kinetic coefficients were compared. Figure 6a–c indicates the relationship between the cell dry weight with respect to time, the linear model known as Lineweaver-Burke plot and growth rate versus cell dry weight, res-pectively. The experimental data fitted with the Riccati model had R2 of 0.99. The Riccati model was derived from the Riccati equation with the concept of second order inhibition. Therefore, the Riccati model predicts any possible inhibition may exist in the bioreactor [31]. Figures 6b and 6c show the linearized Monod and Verhalst models, respectively. The results of the analysis indicated that the experimental values were in good agreement with the predicted values. The predicted and observed results indicated that there is a high regression coefficient in the three models but in the other hand, predicted µmax obtained with the Riccati model had a high accordance with observed results, so the Riccati model was selected as sug-

gestion kinetic models for the hybrid bioreactor. The kinetic constants and regression coefficients resulted in three models are presented in Table 3. Based on the reported value for µmax the values for Riccati and Verhalst models were very close, while the maximum specific growth rates were higher than the value in the Monod model. Thus, the Riccati and Verhalst models should project faster rate compared to the Monod model.

Table 3. Kinetic parameters in hybrid bioreactor treating cheese whey wastewater

R2KS / mg L-1 µmax / h-1 Model

0.995 - 1.31 Riccati

0.977 7.398 0.807 Monod

0.983 - 1.23 Verhalst

CONCLUSION

The hybrid bioreactor was found to be a suc-cessful biological treatment system, achieving a high COD removal efficiency for the treatment of cheese whey wastewater. The results are summarized as fol-lows:

1. It was concluded that the hybrid system is much faster than the conventional digester and even with upflow anaerobic sludge bed (UASB). The rem-aining untreated COD may need further treatment before disposed to environment.

2. COD removal was examined at two different temperatures (24 and 40 °C). Maximum COD removal (80%) was obtained at HRT of 48 h, OLR of 25.85 g COD/Ld and constant temperature of 40 °C.

3. Maximum biogas production was 2.40 (Ld) that was obtained at HRT of 48 h, OLR of 32.60 (g COD/Ld) and constant temperature of 40 °C.

4. Effluent pH was increased when HRT inc-reased from 12 to 48 h. The optimum pH for anaer-obic microbes was obtained at HRT of 48 h that it is shown a decreasing organic acid and increasing of COD removal. The pH variation shows shift of bio-conversion of organic acid to methane occurred while COD removal increased.

5. When bioreactor temperature increased from 24 to 40 °C the COD removal and biogas production significantly increased.

6. Among the project growth kinetic models, the Riccati model was the best growth model fitted with experimental data.

7. A combination of air flotation and then mem-brane separation processes are suggested for the post treatment and the removal of remaining conta-minants.


236

Acknowledgements The authors acknowledged Noshirnavi Univer-

sity of Technology, Biotechnology Research Lab for providing all necessary facilities to conduct present research.

REFERENCES

[1] D. Orhon, E. Gorgon, G. Grimly, N. Rattan, Water Res. 27 (1993) 625–633

[2] R. Pattnaik, R.S. Yost, G. Porter, T. Masunaga, T. Atta-nandana, Ecol. Eng. 32 (2007) 1-10

[3] A.R. Prazeres, F. Carvalho, J. Rivas, J. Environ. Manage. 110 (2012) 48-68

[4] G.A. Pinto, R.L. Giordano, and R.C. Giordano, Biopro-cess Biosyst. Eng. 32 (2009) 69-78

[5] J.C. Frigon, J. Breton, T. Bruneau, R. Moletta, S.R. Guiot, Bioresour. Technol. 100 (2009) 4156–4163

[6] M. Perle, S. Kimchie, G. Shelef, Water Res. 29 (1995) 1549–1554

[7] J. Rajesh Banu, S. Anandan, S. Kaliappan, I.-T. Yeom, Solar Energy 82 (2008) 812–819

[8] A. Ebrahimi, G.D. Najafpour, M. Mohammadi, B. Hashe-miyeh, Chem. Ind. Chem. Eng. Q. 16 (2010) 175-182

0

1

2

3

4

5

0 5 10 15 20 25 30 35

Cel

l dry

wei

ght

(g/L

)

Time (h)

(a)

0

2

4

6

8

10

12

0 0.2 0.4 0.6 0.8 1 1.2 1.4

1/µ

(1/

h)

1/S (L/g)

(b)

0

0.2

0.4

0.6

0.8

1

1 2 3 4

Gro

wth

rat

e (1

/h)

Cell dry weight (g/L)

(c)

Figure 6. The plotted kinetic models with initial COD = 51714 mg/L, HRT = 48 h: a) Riccati model; b) linearized Monod model; c) Verhalst model.


237

[9] F. Carta-Esobar, J. Pereda-Marin, P. Alvarez-Mateos, F. Romero-Guzman, M. Duran-Barrantes, F. Barriga-Mateos, Biochem. Eng. J. 21 (2004) 183-191

[10] M.I. Gonzalez Siso, Bioresour. Technol. 57 (1996) 1-11

[11] A. Saddoud, I. Hassairi, S. Sayadi, Bioresour. Technol. 98 (2007) 2102

[12] J.Q. Yan, P.H. Liao, K.V. Lo, Biomass 17 (1988) 185-193

[13] A.E. Ghaly, J.B. Pyke, Appl. Chem. Biotechnol. 27 (1991) 217–237

[14] J. Rivas, A.R. Prazeres, F. Carvalho, F. Beltran, J. Agric. Food Chem. 58 (2010) 7871-7877

[15] G. Moody, Filtration Separation 32 (1995) 329-336

[16] S.R. Guiot, L. Van den Berg, Biotechnol. Bioeng. 27 (1985) 800–806

[17] G. Vidal, A. Carvalho, R. Mendez, J.M. Lema, Bioresour. Technol. 74 (2000) 231–239

[18] A. Tawfik, M. Sobhey, M. Badawy, Desalin. 227 (2008) 167–177

[19] K.V. Lo, R.H. Liao, Biomass 10 (1986) 243-249

[20] F. Malaspina, C.M. Cellamare, L. Stante, A. Tilche, Bio-resour. Technol. 55 (1996) 131-136

[21] I.F. Malina, F.G. Pohland, Design of Anaerobic Pro-cesses for the Treatment of Industrial and Municipal Wastes, Technomic, Lancaster, PA, 1992, pp. 119–120

[22] I. Ozturk, V. Eroglu, G. Ubay, I. Demir, Water Sci. Tech-nol. 28 (1993) 77–85

[23] I.Q. Van, K.V. Lo, P.H. Liao, Boil. Wastes 27 (1989) 289–305

[24] J.L. Rico Gutierrez, P.A. Garcia Encina, F. Fdz-Polanco, Bioresour. Technol. 37 (1991) 271–276

[25] S.V. Kalyuzhnyi, E.P. Martinez, J.R. Martinez, Bioresour. Technol. 60 (1997): 59–65

[26] H. Gavala, H. Kopsinis, I. Skiadas, K. Stamatelatou, and G. Lyberatos, J. Agric. Eng. Res. 73 (1999) 59-63

[27] American Public Health Association (APHA) and Amer-ican Water and Wastewater Association (AWWA), Stan-dard methods for the examination of water and waste-water, 17th ed., APHA, Washington D.C., 1989, pp. 5-15

[28] O. Monroy, K. Johnson, A. Wheatley, F. Hawkes, M. Caine, Bioresour. Technol. 50 (1994) 243-251

[29] H. Beyenal, Z. Lewandowski, Water Res. 34 (2000) 528– -538

[30] A.A.L. Zinatizadeh, A.R. Mohamed, A.Z. Abdullah, M.D. Mashitah, M. HasnainIsa, G.D. Najafpour, Water Res. 40 (2006) 3193–3208

[31] G.D. Najafpour, M. Tajallipour, M. Komeili, M. Moham-madi, Afr. J. Biotechnol. 8 (2009) 3590-3596.

NAZILA SAMIMI TEHRANI1

GHASEM D. NAJAFPOUR2

MOSTAFA RAHIMNEJAD2

HOSSEIN ATTAR1

1Department of Chemical Engineering, Science and Research Branch, Islamic

Azad University, Tehran, Iran 2Biotechnology Research Lab, Faculty

of Chemical Engineering, Noshirvani University of Technology, Babol, Iran

NAUČNI RAD

KARAKTERISTIKE BIOREAKTORA SA NEPOKRETNIM FILMOM ANAEROBNOG MULJA NAVIŠE ZA OBRADU VELIKOG ORGANSKOG OPTEREĆENJA I PROIZVODNJU BIOGASA IZ OTPADNE SURUTKE

Među različitim tehnologijama za obradu otpadnih voda, izgleda da biološko prečišćavanje

najviše obećava. Poluindistrijski hibridni anaerobni bioreaktor je korišćen za obradu

otpadne surutke. Vrh i dno hibridnog bioreaktora poznatog kao nepokretni film anearobnog

mulja koji se kreće naviše (UASFF) je kombinacija prekrivača anaerobnog mulјa (UASB) i

bioreaktora sa nepokretnim filmom anaerobnog mulja (UAFF) koji se kreću naviše, redom.

Ispitivani su efekti radnih uslova, kao što su temperatura i hidrauličko retenciono vreme

(HRT), na uklanjanje hemijske potrošnje kiseonika (HPK) i proizvodnju biogasa u hib-

ridnom bioreaktoru. Obradivost uzoraka pri različitim HRT od 12, 24, 36 i 48 h je ocenjena

u hibridnom bioreaktoru. Dobijeni su želјeni uslovi za uklanjanje HPK od HRT = 48 h i

radna temperatura od 40 °C. Maksimalno uklanjanje HPK i produkcija biogasa su 80% i

2,40 dm3/dan, redom. Kinetički modeli Riccati, Monod i Verhalst su primenjeni za rast

mikroorganizama u procesu obrade. Među navedenim modelima, Riccati model je najbolјi

model rasta koeficijentom determinacije R2 = 0,99.

Ključne reči: HPK, Riccati model, UASFF bioreaktor, visoko organsko optere-ćenje, otpadna surutka.





239

SEYED MAJID ATAEI ARDESTANI1

MORTEZA SADEGHI2

BABAK BEHESHTI1

SAEID MINAEI3

NASER HAMDAMI4 1Department of Mechanics of

Agricultural Machinery, Science and Research Branch, Islamic Azad University, Tehran, Iran

2Department of Agricultural Machinery Engineering, College of

Agriculture, Isfahan University of Technology, Isfahan, Iran

3Department of Agricultural Machinery Engineering, Faculty of

Agriculture, Tarbiat Modares University, Tehran, Iran

4Department of Food Science, College of Agriculture, Isfahan University of Technology, Iran

SCIENTIFIC PAPER

UDC 66.047:543.5:615.322

DOI 10.2298/CICEQ131206021A

VIBRO-FLUIDIZED BED HEAT PUMP DRYING OF MINT LEAVES WITH RESPECT TO PHENOLIC CONTENT, ANTIOXIDANT ACTIVITY AND COLOR INDICES

Article Highlights • We study vibro-fluidized bed drying assisted heat pump of mint leaves • Temperature caused a remarkable loss of green color due to chlorophyll degradation • The values of energy activation were within the reported range for food materials Abstract

Due to high porosity and stickiness, good fluidization of mint leaves can be difficult to achieve. In this study, a vibro-fluidized bed dryer assisted heat pump system was designed and fabricated to overcome this problem. The drying experiments were carried out at temperatures of 40, 50 and 60 °C. The quality of the dehydrated samples was assessed based on color indices, antioxidant activity, and total phenolic content. Drying process primarily occurred in falling rate period. The effective coefficient of moisture transfer of the samples was increased with air temperature and varied from 4.26656×10-11 to 2.95872×10-10 m2 s-1 for heat pump drying (HPD) method, and 3.71918×10-11 to 1.29196×10-10 m2 s-1 for none-heat pump drying (NHPD) method. The color indices for tem-peratures of 40 and 50 °C were very close to each other, whereas by inc-reasing temperature to 60 °C, a remarkable loss of green color was observed. The highest phenolic content was found in methanolic extract for HPD at 60 °C, and NHPD at 50 °C contained the lowest amount of phenolic compounds. NHPD treatments showed lower antioxidant activity compared to HPD treat-ments at the same temperature due to the longer drying times.

Keywords: drying kinetics, fluidized bed drying, moisture diffusivity, pharmaceutical plants.

Mint (Mentha spicata) is one of the most impor-tant spices throughout the world and a perennial plant belonging to Lamiaceae family. Its leaves are used for flavoring, tea infusions and spicing. The use of mint leaves in a variety of dishes such as vegetable cur-ries, chutney, fruit salads, vegetable salads, salad dressings, soups, desserts, juices, and sherbets has been reported [1,2].

Processing method influences the volatile oils of medicinal plants considerably. Drying, as one of the oldest methods of food preservation, represents a

Correspondence: S.M.A. Ardestani, Department of Mechanics of Agricultural Machinery, Science and Research Branch, Islamic Azad University, Tehran, Iran. E-mail: [email protected] Paper received: 6 December, 2013 Paper revised: 11 June, 2014 Paper accepted: 23 June, 2014

very important aspect of food processing. The main purpose of drying is to extend product shelf life, mini-mize packaging requirements and reduce shipping weights [3]. Phenolic content, antioxidant activity, and color indices are the main characteristics which must be taken into consideration in terms of drying medi-cinal and spice plants. Hossain et al. [4] assessed the effect of different drying methods on phenolic com-pounds and antioxidant capacity of six Lamiaceae herbs. Among the drying methods tested, air-drying was found to be the best for all samples. Siriamorn-pun et al. [5] evaluated the effect of some drying treat-ments on color, antioxidant activities and carotenoids of marigold flower. However, there is no report reg-arding the effect of drying on total phenolic and anti-oxidant activity of mint leaves.

Among drying methods, fluidized bed drying (FBD) has the advantage of high drying rate due to

S.M.A. ARDESTANI et al.: VIBRO-FLUIDIZED BED HEAT PUMP DRYING… Chem. Ind. Chem. Eng. Q. 21 (2) 239−247 (2015)

240

high rates of heat and mass transfer, and conse-quently high thermal efficiency with uniform and closely controllable temperature in the bed [6]. In the case of FBD, batch mode is preferred for small-scale pro-duction and heat sensitive materials. Therefore, fluid-ized bed technology is widely employed by pharma-ceuticals, foods, fertilizers, and many other chemical industries, where either wet granulation or drying solid materials is a fundamental stage of these industries [7].

In order to reduce energy consumption per unit of product moisture, it is important to investigate dif-ferent methods for enhancing the energy efficiency of the drying operation. Equipping convective hot air dryers with heat pumps has been recognized as an ideal approach for this purpose. A heat pump dryer (HPD) is an economic, environmentally friendly, hygi-enic drying device used to dry food materials [8]. The HPDs available in the market can allow the energy demand to be reduced by 50% [9]. A HPD can be operated over a wide range of temperatures, pro-viding good conditions for drying heat sensitive materials. The technology requires less energy, as the system can recover the latent heat in a closed loop, and it can be independent of ambient weather conditions. Strommen et al. [10] found that HPDs con-sumed 60–80% less energy compared with conven-tional dryers operating at the same temperature.

In spite of the vast applications of FBD, it has not been used for drying pharmaceutical plants, espe-cially leaves products such as mint. This could be due to high porosity of these products, and stickiness of the leaves to each other during fluidization. Moreno et al. [11] reported vibration as a method for improving the quality of fluidization and avoiding problems such as channeling and defluidization. The aims of this study were investigating drying kinetics of mint leaves using a vibro-fluidized bed dryer (VFBD) assisted heat pump system in batch mode, determining mois-ture diffusivity under different drying conditions, and analyzing qualitative parameters including color indi-ces, total phenolic content and antioxidant activity of the dehydrated product.


Sample preparation

Fresh mint leaves were purchased from a local market in Isfahan (central Iran), and stored in a refri-gerator at 5 °C before the drying experiments. The samples without any pretreatment were used for con-ducting tests. The initial moisture content of the mint leaves was determined using AOAC [12] standard method (vacuum drying at 70 °C for 24 h). Drying of

the leaves was finished until the moisture content did not change and the weight of samples became cons-tant.

Vibro-fluidized bed heat pump drying

In preliminary tests [13], due to high porosity and stickiness, the mint leaves could not be fluidized and stacked to each other during fluidization. There-fore, a laboratory vibro-fluidized bed dryer (VFBD) assisted heat pump system (Figure 1) was designed and fabricated at Isfahan University of Technology (IUT) to conduct the experimental part of this study. A 1.5 kW blower (Motogen 90L2A; Motogen Co. Ltd., Tabriz, Iran), 2830 rpm, was coupled to an inverter (Teco 7300 CV, with ±0.01 Hz accuracy; TECO Elec-tric & Machinery Co. Ltd., Taipei, Taiwan) and used for supplying and controlling the airflow rate. The dry-ing air was heated with a 10 kW electric heater. The exit temperature of air from the heater was main-tained constant within ±0.5 °C using a PI control sys-tem.

The air velocity was measured using an air velo-city transmitter (AVT; HK instruments Co. Ltd., Muur-ame, Finland) placed downstream from the heater. The distributor plate was constructed from a circular Plexiglas plate with the thickness of 5 mm, along with 1630 holes of 2 mm diameter.

The cylindrical bed column was made of Plexi-glas with a wall thickness of 5 mm, an internal dia-meter of 140 mm, and the length of 1000 mm. Three sensors (SHT75; Sensirion AG, Staefa, Switzerland) were used to measure the relative humidity and tem-perature at the inlet of the bed as well as locations at 10 and 300 mm above the distributor plate (Figure 1). Three other sensors were located before and after the evaporator, and after the condenser of the heat pump system. A digital balance (Kern 572-57, with ±0.1 g accuracy; Kern & Sohn GmbH, Balingen, Germany) was used to measure the weight of samples during the experimental runs. All the signals from the SHT75 sensors, air velocity sensor, digital balance, heater and energy module were acquired simultaneously every 15 s using the data acquisition board and the LabView software and stored in a desktop computer for subsequent analysis. To evaluate the influence of air temperature on drying curves, the experiments were carried out at 40, 50 and 60 °C. The mint samples were dried under heat pump drying (HPD), and non-heat pump drying (NHPD) treatments.

In order to apply vibration to the drying chamber, a single-phase motor (Motogen CRS 90L2A; Motogen Co. Ltd., Tabriz, Iran), and a 3 mm eccentric mech-anism with 80 Hz frequency of vibration were used.


241

Table 1 presents the information regarding the conditions of the experiments.

Table 1. Some physical characteristics of mint leaves and the conditions of the experiments

Parameter Value

Frequency of vibration, Hz 80

Amplitude of vibration, mm 3

Initial solid (leaves) temperature, °C 22

Inlet air humidity, % 9

Air velocity, m s-1 2

Bed height, mm 150

Bed porosity, % 92

Thickness of leaves, mm 0.187±0.01

Leaves size, mm 20×40±3

Particle density, kg m−3 903

Before each experiment, the unit was run with-out any sample for about 30 min to reach the thermal steady state.

Drying kinetics and moisture diffusivity

The moisture ratio (MR) and drying rate (DR) of mint leaves during drying experiments were cal-culated using Eqs. (1) and (2), respectively [14,15]:

−=−0

e

e

M MMR

M M (1)

+ −= = d1 d 1

d dt t tM MM

DRA t A t

(2)

where M, Me, M0, Mt and Mt+dt are the moisture content at any time, the equilibrium moisture content, the initial moisture content, the moisture content at t and the moisture content at t+dt, respectively, t is the drying time (min), and A is the drying area (m2).

Fick’s second diffusion equation was used to determine the effective coefficient of moisture transfer of the samples. The mint leaves were considered the slab geometry [16]. The equation is expressed as [17]:

( )( ) π

π

∞

=

− = − −

2 2

2 2 201

2 18 1exp

42 1

eff

n

n D tMR

ln (3)

where Deff is the effective coefficient of moisture trans-fer (m2 s-1), l0 is the half thickness of the slab (m) and n is the positive integer. For long drying times, only the first term of Eq. (3) can be used [18]:

ππ

= −

2

2 20

8exp

4effD t

MRl

(4)

1. Blower

2. Electric heater

3. Signal connector board

4. Plenum

5. Distributor plate

6. Evaporator

7. Condenser

8. Heat pump

9. Data acquisition system

10. Vibration system

V = Air velocity sensor

H = Relative humidity sensor

T = Temperature sensor

1

2

3

4

5

67

8

9

10

V

H1&T1

H2&T2

H3&T3

H4 & T4H5 & T5H6 & T6

Figure 1. Schematic diagram of vibro-fluidized bed heat pump drying system.


242

( ) ππ

= −

2

2 20

8ln ln

4effD t

MRl

(5)

The slope (K0) is calculated by plotting ln (MR) versus t according to Eq. (5) to calculate the effective coefficient of moisture transfer for different tempe-ratures:

π=2

0 204effD

Kl

(6)

Activation energy

Temperature and effective coefficient of moist-ure transfer can be related using Arrhenius equation [18] as follows:

( )

= − + 0 exp

273.15a

effE

D DR T

(7)

where D0 is the constant in Arrhenius equation (m2 s-1), Ea is the activation energy (kJ mol-1), T is the tempe-rature of air (°C), and R is the universal gas constant (8.3145 J mol-1 K-1). Equation (7) can be written in the form of:

( )

= − + 0ln ln

273.15a

effE

D DR T

(8)

The slope (KE) is calculated by plotting ln Deff versus T-1 according to Eq. (8) to determine the acti-vation energy as follows:

= aE

EK

R (9)

Color measurements

The color of the mint samples was evaluated before and after drying according to CIE (Commission International de l’Eclairage) [19]. The color values were measured using a spectrophotometer Text Flash (Datacolor Corp., Switzerland). The lightness (L*), redness (a*), and yellowness (b*) were captured for each treatment. The total color difference (ΔE) was calculated using Eq. (10). This index is a single value that takes into account the differences between the L*, a* and b* of the sample and standard. Chroma or strength of color (C*), and hue angle (h*), which are related to a* and b*, were also calculated by Eqs. (11) and (12) [19]:

( ) ( ) ( )Δ = − + − + −2 2 2* * * * * *

0 0 0E L L a a b b (10)

( ) ( )= +2 2* * *C a b (11)

− =

** 1

*tan

bh

a (12)

where L0*, a0* and b0* are the values for fresh mint leaves.

Total phenolic extraction

For phenolic extraction, 6 g of powders was extracted with 200 ml of 80% methanol. The extract-ion was carried out using an orbital shaker (150 rpm) at 25 °C for 24 h. The extracts were filtered through four layers of cheesecloth to remove the solid debris. The extraction was done four times. Total phenolics were determined colorimetrically using Folin-Ciocal-teu reagent as described by Pinelo et al. [20]. In this regard, ten-fold diluted reagent (2.5 ml), 7.5% sodium carbonate (2 ml) and methanolic extract (0.5 ml) were mixed. Then, after heating at 45 °C for 15 min, the absorbance was measured at 765 nm against a blank. The phenolic content was expressed as tannic acid equivalent per gram dry weight of sample.

Antioxidant activity measurements

Free radical scavenging activity of the mint leaves extracts and standard antioxidant was assessed using the 1,1-diphenyl-2-picrylhydrazyl (DPPH) method [21]. Different dilutions of the mint leaves extracts (equi-valent to 50, 100, 300 and 500 ppm) were prepared in methanol. Butylated hydroxytoluene (BHT) was used as standard antioxidant in 1-100 μg ml-1 solution. Five milliliters of a 0.1 mM methanolic solution of DPPH was mixed with 0.1 ml of sample and standard solu-tions separately. Radical scavenging of the extracts was calculated by employing Eq. (13) and using methanol (80%) and DPPH solution (0.1 mM, 5 ml) as a blank and control sample, respectively:

×

% Radial scavenging activity =

sample= control - 100

control

ODOD

OD

(13)


Drying kinetics

The results of variation in moisture ratio versus drying time, and drying rate versus moisture content, as obtained for HPD and NHPD treatments carried out at 40, 50 and 60 °C, are shown in Figures 2 and 3, respectively. As expected, with increasing drying temperature, the drying time was decreased. The constant rate drying period was not detected in drying curves. In other words, drying of mint leaves occurred primarily in falling rate period, indicating that initial


243

mass transfer occurred by diffusion. The same results have been reported for drying mint leaves [16,22].

Figure 2. Drying curve for mint leaves dehydrated at different

air-drying temperatures: a) heat pump drying (HPD) method and b) non-heat pump drying (NHPD) method.

Moisture diffusivity and activation energy

Values of effective coefficient of moisture trans-fer (Deff) for mint leaves dehydrated under HPD and NHPD treatments at various temperatures are shown in Table 2. The values of Deff were increased with inc-reasing air-drying temperature. Effective coefficient of moisture transfer values were varied from 4.25656×10-11 to 2.95872×10-10 m2 s-1 for HPD, and from 3.71918×10-11 to 1.29196×10-10 m2 s-1 for NHPD at temperatures ranging from 45 to 60 °C. It is obs-erved that the difference between the values for HPD and NHPD methods was not remarkable. This is due to the fact that the effective coefficient of moisture transfer is an internal parameter. Similar variations were also observed during drying of mint [22]. Gen-erally, effective coefficient of moisture transfer values for food materials are in the range of 10-9 to 10-11 m2 s-1 [23].

Figure 3. Drying rate curve for mint leaves dehydrated at

different air-drying temperatures: a) heat pump drying (HPD) method and b) non-heat pump drying (NHPD) method.

Table 2. Values of effective coefficient of moisture transfer (m2 s-1) of mint leaves dehydrated at different air-drying temperatures and two drying methods

Temperature, °C Drying method

HPD NHPD

40 4.25656×10-11 3.71918×10-11

50 1.04859×10-10 1.02623×10-10

60 2.95872×10-10 1.29196×10-10

Values of ln Deff versus T-1 for experiments infor-mation are plotted in Figure 4. The activation energy was determined to be 84 kJ mol-1 for HPD and 54.34 kJ mol-1 for NHPD. Park et al. [1] and Doymaz [16] reported the values of 82.93 and 62.96 kJ mol-1 for drying mint leaves, respectively.

Color, total phenolic content and antioxidant activity

The L*, a*, b*, c* and h* values of the samples dried using HPD and NHPD methods at different temperatures are presented in Table 3. Low negative a* value (–7.48) confirms greenness and conse-quently, a high (116.34° > 90°) hue angle, thereby indicating more greenness for fresh mint leaves. L* value was also approximately high (47.71) for fresh sample. L* value is a measure of the color in the light-


244

Figure 4. The relationship between air temperature and effective

coefficient of moisture transfer for mint leaves dehydrated: a) heat pump drying (HPD) method and b) non-heat pump

drying (NHPD) method.

-dark axis. Reduction in this parameter indicates that the samples were turning darker. As observed, there is an improvement in color indices with increasing drying time (using the temperature of 40 °C compared with 60 °C, Table 3). However, the color indices for temperatures of 40 and 50 °C were very close to each other, whereas by increasing temperature to 60 °C, a remarkable loss of green color was observed. This

might be caused by chlorophyll degradation in leaf cells [24]. Previous reports have shown that degra-dation of chlorophyll occurred at temperatures exceed-ing 50 °C in thyme and 60 °C in broccoli juice [25].

The total color difference (ΔE), which is a com-bination of Hunter L* a* b* values, is a colorimetric parameter extensively used to characterize the vari-ation of colors during processing of foods [23]. Less values of ΔE indicate minimum differences between L*, a* and b* of the samples and fresh ones. The samples dehydrated at 40 °C had minimum ΔE, and the samples at 60 °C presented the maximum value. The chroma value indicates the degree of saturation of color and is proportional to the strength of the color. Large changes were found in chroma between fresh and dried mint leaves at 60 °C (Table 3). This reveals lack of stability of green color in mint leaves for this treatment.

For all color indices, a remarkable difference was not observed between HPD and NHPD methods at a constant temperature. Drying of agricultural and food products depends on the temperature and humidity of the drying air. One of the functions of the HPD method is to isolate the humid environment from the drying process, while, drying by NHPD (open- -type) method may prolong the drying duration because of humid air entering into the drying chamber in high humid areas. The present study was con-ducted under low humidity conditions. Therefore, in such arid and semi-arid areas, the use of heat pump system could not establish a meaningful difference in drying times. This is the reason for having the same color parameters under different methods.

Total phenolic content was varied from 30.46 to 67.32 mg tannic acid per 1 g dry weight of the samples. The highest phenolic content was found in methanolic extract of 60 °C for HPD, whereas air temperature of 50 °C for NHPD contained the lowest amount of phenolic compounds (Figure 5). There are similar reports regarding the effect of different drying methods on total phenolic content of plants. Asami et al. [26] assessed the effect of three postharvest treat-

Table 3. Values of L*, a*, b*, c*, h* and ΔE for mint leaves dehydrated at different air-drying temperatures and two drying methods

Treatment L* a* b* c* h* ΔE

HPD, T=40 °C 38.05 -3.37 15.68 16.04 102.14 10.51

NHPD, T=40 °C 39.34 -2.76 16.38 16.61 99.56 9.69

HPD, T=50 °C 39.15 -2.87 16.72 16.96 99.73 9.85

NHPD, T=50 °C 38.66 -3.12 16.50 16.80 100.72 10.14

HPD, T=60 °C 35.79 0.43 13.71 13.71 88.22 14.37

NHPD, T=60 °C 34.83 -0.06 14.20 14.20 90.23 14.89

Fresh mint 47.71 -7.48 15.11 16.86 116.34 0.00


245

ments (freezing, freeze-drying, and air-drying) on total phenolic content of some fruits.

Sejali and Anuar [27] evaluated the effect of three drying treatments (shade, oven-dried at 45 and 70 °C) on total phenolic contents of Neem (Azadi-rachta indica) leaf powder. Similar to the present study, lower temperatures and fine particle size led to higher phenolic content. It was also observed that higher phenolic content in dried oregano could be reached by the reduction of temperature [28].

The antioxidant activity values of samples were also measured at different treatments (Figure 6). High values of IC50 indicate less antioxidant activity. As depicted, antioxidant activity was increased in higher temperatures. All the NHPD treatments showed lower antioxidant activity compared with HPD treatments at the same temperature due to the longer drying time. Two factors can affect the antioxidant activity and major components of plant species: drying tempera-ture and drying time [24]. As increasing the tempera-ture can lead to degrading the antioxidants, the pro-bable reason for increasing antioxidant activity is the

shorter drying time when applying higher tempera-tures in the methods used in the present study. Mad-rau et al. [29] reported similar results in apricot drying. In their study, higher antioxidant activity was obs-erved when temperature was increased from 50 to 70 °C. As the drying time is reduced with increasing temperature, it is concluded that the antioxidant acti-vity of the samples was affected by drying time rather than drying temperature.

CONCLUSIONS

Drying of mint leaves was performed using a laboratory vibro-fluidized bed assisted heat pump system. Values of effective coefficient of moisture transfer were increased as air-drying temperature was increased. But the values between heat pump drying (HPD) and non-heat pump drying (NHPD) methods did not show any remarkable difference. The values of energy activation were within the reported range for food materials. As a result of chlorophyll degradation in leaf cells, increasing temperature to 60

Figure 5. The total phenolic content values for mint leaves dehydrated at different air-drying temperatures and two drying methods.

Figure 6. The antioxidant activity values for mint leaves dehydrated at different air-drying temperatures and two drying methods.


246

°C caused a remarkable loss of green color. How-ever, the values between HPD and NHPD treatments did not show a meaningful difference. It was revealed that the effect of drying time on antioxidant activity of the samples was more than that of temperature. Overall, it was concluded that in arid and semi-arid areas, the use of heat pump system could not establish a considerable difference in drying process.

Acknowledgements

Financial support of this study was received from the Iran National Science Foundation (INSF) (Project #89001230) and Isfahan University of Tech-nology, both of which should be gratefully acknow-ledged.

Nomenclature

a* redness M moisture content at any time (g water/g

dry matter) b* yellowness Me equilibrium moisture content (g water/g

dry matter) C* chroma Mo initial moisture content (g water/g dry

matter) Deff effective coefficient of moisture transfer

(m2 s-1) Mt moisture content at t time (g water/g dry

matter) D0 constant in Arrhenius equation (m2 s-1) Mt+dt moisture content at t+dt time (g water/g

dry matter) Ea activation energy (kJ mol-1) MR moisture ratio (dimensionless) HPD heat pump drying NHPD non-heat pump drying h* hue angle n exponent and positive integer ko, kE slope R gas constant (J mol-1 K-1) l0 slab thickness T temperature (°C) L*

0, a*0, b

*0 values for fresh mint

t drying time (min) DR drying rate (g m-2 s-1) L* lightness

A drying area (m2) ΔE total color difference

REFERENCES

[1] K.J. Park, Z. Vohnikova, F.P.R. Brod, J. Food Eng. 51 (2002) 193–199

[2] A.K. Thompson, Fruits and Vegetables, Blackwell Publishing, Oxford, 2003

[3] M.R. Okos, G. Narsimhan, R.K. Singh, A.C. Weitnauer, in Handbook of Food Engineering, D.R. Heldman, D.B. Lund, Ed., Marcel Dekker Inc, New York, 1992

[4] M.B. Hossain, C.A. Barry-Ryan, B. Martin-Diana, N.P. Brunton, Food Chem. 123 (2010) 85–91

[5] S. Siriamornpun, O. Kaisoon, N. Meeso, J. Funct. Food. 4 (2012) 757–766

[6] A. Topuz, M. Gur, M.Z. Gul, Appl. Therm. Eng. 24 (2004) 1535–1547

[7] F. Portoghese, F. Berruti, C. Briens, Powder Technol. 181 (2008) 169–177

[8] D.C. Wang, G. Zhang, Y.P. Han, J.P. Zhang, X.L. Tian, Energy Convers. Manag. 52 (2011) 2919–2924

[9] Z. Oktay, A. Hepbasli, Energy Convers. Manag. 44 (2003) 1193–1207

[10] I. Strommen, T.M. Eikevik, O. Alves-Filho, K. Syverud, O. Jonassen, Low Temperature Drying with Heat Pumps New Generations of High Quality Dried Products, in the 13th International Drying Symposium, Beijing, 2002

[11] R. Moreno, R. Rios, H. Calbucura, Drying Technol. 18 (2000) 1481–1493

[12] AOAC, Official Method of Analysis (No. 934.06), Association of Official Analytical Chemists, Washington D.C., 1990

[13] S.M. Ataee Ardestani, PhD Thesis, Science and Research Branch, Islamic Azad University, Tehran, 2014 (In Persian).

[14] A. Midilli, Int. J. Energy Res. 25 (2001) 715–725

[15] S. Erenturk, M.S. Gulaboglu, S. Gultekin, Biosyst. Eng. 89 (2004) 159–166

[16] I. Doymaz, J. Food Eng. 74 (2006) 370–375

[17] J. Crank, the Mathematics of Diffusion, 2nd ed., Clarendon Press, Oxford, 1975

[18] A. Lopez, A. Iguaz, A. Esnoz, P. Virseda, Drying Technol. 18 (2000) 995–1006

[19] G. Wyszecki, W.S. Stiles, Color Science: Concepts and Methods, Quantitative Data, and Formulae, John Wiley & Sons, New York, 2000, pp. 130–174

[20] M. Pinelo, M. Rubilar, J. Sineiro, M.J. Nunez, Food Chem. 85 (2004) 267–273

[21] A. Braca, C. Sortino, M. Politi, J. Ethnopharmacology 79 (2002) 379– 381

[22] D.M. Kadam, R.K. Goyal, K.K. Singh, M.K. Gupta, J. Med. Plants Res. 5 (2011) 164–170

[23] A. Maskan, S. Kaya, M. Maskan, J. Food Eng. 54 (2002) 81–88

[24] M. Rahimmalek, S.A.H. Goli, Ind. Crop Prod. 42 (2013) 613–619

[25] C.A. Weemaes, V. Ooms, A.M. Van Loey, M.E. Hend-rickx, J. Agric. Food Chem. 47 (1999) 2404–2409

[26] D.K. Asami, Y.J. Hong, D.M. Barrett A.E. Mitchell, J. Agric. Food Chem. 51 (2003) 1237–1241

[27] S.N.F. Sejali, M.S. Anuar, J. Herbs Spices Med. Plants 17 (2011) 119–131.


247

[28] K. Jaroszynski, A. Figiel, A. Wojdylo, Acta Agrophys. 11 (2008) 81–90

[29] M.A. Madrau, A. Piscopo, A.M. Sanguinetti, A. DelCaro, M. Poiana, F.V. Romeo, A. Piga, Eur. Food Res. Technol. 228 (2009) 441–448.

SEYED MAJID ATAEI

ARDESTANI1

MORTEZA SADEGHI2

BABAK BEHESHTI1

SAEID MINAEI3

NASER HAMDAMI4

1Department of Mechanics of Agricultural Machinery, Science and

Research Branch, Islamic Azad University, Tehran, Iran

2Department of Agricultural Machinery Engineering, College of Agriculture,

Isfahan University of Technology, Isfahan, Iran

3Department of Agricultural Machinery Engineering, Faculty of Agriculture,

Tarbiat Modares University, Tehran, Iran

4Department of Food Science, College of Agriculture, Isfahan University of

Technology, Iran

NAUČNI RAD

UTICAJ SUŠENJA MENTE U VIBRO-FLUI-DIZOVANOM SLOJU UZ KORIŠĆENJE TOPLOTNE PUMPE NA SADRŽAJ FENOLA, ANTIOKSIDATIVNU AKTIVNOST I INDEKSE BOJE

Zbog visoke poroznost i krutosti, lišće mente nisu mogli biti dobro fluidizovani tokom

fluidizacije. U ovom radu je projektovana i izrađena sušara sa vibro-fluidizovanim slojem

da bi bio prevaziđen ovaj problem. Eksperimenti sušenja su izvedeni na temperaturama od

40, 50 i 60 °C. Kvalitet dehidrisanih uzoraka je ocenjen na osnovu indeksa boje, antio-

ksidativne aktivnosti i sadržaja ukupnih fenola. Proces sušenja se prvenstveno dešava u

period opadajuće brzine. Efektivna koeficijent prenosa vlage iz uzoraka se povećava sa

temperaturom vazduha i varira od 4,27×10-11 do 2,96×10-10 m2 s-1 za sušenje pomoću

toplotne pumpe (HPD), a 3,72×10-11 do 1,29×10-10 m2 s-1 za sušenje bez toplotne pumpe

(NHPD). Indeksi boja za temperature od 40 i 50 °C su veoma blizu jedan drugom, a

povećanjem temperature do 60 °C, uočen je izuzetan gubitak zelene boje. Najviši sadržaj

fenola nađen u metanolnom ekstraktu za HPD na 60 °C, a ekstrakt NHPD na 50 °C je

sadržao najniži sadržaj fenolnih jedinjenja. NHPD tretmani su pokazali manju antioksi-

dativnu aktivnost u odnosu na HPD tretmane na istoj temperaturi usled dužeg vremena

sušenja.

Ključne reči: kinetika sušenja, sušenje u fluidizovanom sloju, difuzivnost vlage, medicinsko bilje.





249

MARIJA S. PETROVIĆ1

TATJANA D. ŠOŠTARIĆ1

LATO L. PEZO2

SLAVKA M. STANKOVIĆ3

ČASLAV M. LAČNJEVAC4

JELENA V. MILOJKOVIĆ1

MIRJANA D. STOJANOVIĆ1 1Institute for Technology of Nuclear

and Other Mineral Raw Materials, Belgrade, Serbia

2Institute of General and Physical Chemistry, University of Belgrade,

Belgrade, Serbia 3Faculty of Technology and

Metallurgy, University of Belgrade, Belgrade, Serbia

4Faculty of Agriculture, University of Belgrade, Serbia

SCIENTIFIC PAPER

UDC 66.081:634.21:633.21:54

DOI 10.2298/CICEQ140510023P

USEFULNESS OF ANN-BASED MODEL FOR COPPER REMOVAL FROM AQUEOUS SOLUTIONS USING AGRO INDUSTRIAL WASTE MATERIALS

Article Highlights • Apricot stones and corn cobs are low cost and locally available biosorbents • SOP and ANN models were used for analysing and optimizing of the biosorption

process • Among the input parameters, biosorbent mass has the most significant influence on

biosorption • ANN model yield a bit better fit of experimental data, according to r2 and SOS of both

models • Both materials can be used as biosorbents for the removal of copper ions from

aqueous solution Abstract

The purpose of this study was to investigate the adsorption properties of locally available lignocellulose biomaterials as biosorbents for the removal of copper ions from aqueous solution. Materials are generated from juice production (apricot stones) and from the corn milling process (corn cob). Such solid wastes have little or no economic value and very often present a disposal problem. Using batch adsorption techniques the effects of initial Cu(II) ions concentration (Ci), amount of biomass (m) and volume of metal solution (V) on biosorption efficiency and capacity were studied for both materials, without any pre-treatments. The optimal parameters for both biosorbents were selected depending on the highest sorption capability of biosorbent in removal of Cu(II). Experimental data were compared with second order polynomial regression models (SOPs) and artificial neural networks (ANNs). SOPs showed accept-able coefficients of determination (0.842-0.997), while ANNs performed with high prediction accuracy (0.980-0.986) in comparison to experimental results.

Keywords: biosorption, apricot stones, corn cob, copper ions, SOPs, ANN.

Copper is a heavy metal and the most common pollutant of the environment due to its wide use in different industries including electroplating, azo dye manufacture, engraving, lithography, petroleum refin-ing and pyrotechnics etc. [1,2]. Copper ion concen-trations can approach 100-120 mg L-1 in industrial wastewaters [3]. Although copper is an essential nutrient for living organisms, it is a common and ser-ious environmental pollutant because of its toxicity Correspondence: M.S. Petrović, Institute for Technology of Nuc-lear and Other Mineral Raw Materials, Bulevar Franše d’Eperea 86, 11000 Belgrade, Serbia. E-mail: [email protected] Paper received: 10 May, 2014 Paper revised: 20 June, 2014 Paper accepted: 24 June, 2014

and its bioaccumulation in the food chains [4]. Due to its toxic effect on living organisms, treatment of cop-per contaminated water is necessary. One possible treatment could be the process of biosorption.

Biosorption is a promising technology based on the ability of biological materials (biosorbents) to accumulate metal ions from wastewater by either metabolically mediated processes or physicochemical pathways of uptake [5]. Biosorbents derived from appropriate biomass can be used for the removal and recovery of heavy metal ions from wastewater [6]. Various lignocellulose biomaterials such as: pine cone shell [7], olive stone [8], rice shell [9], cocoa shell [10], orange and banana peel [11], corn cob [12], peach shell [13], barley straw [14] can be used as a

M.S. PETROVIĆ et al.: USEFULNESS OF ANN-BASED MODEL FOR COPPER… Chem. Ind. Chem. Eng. Q. 21 (2) 249−259 (2015)

250

low-cost biosorbents for heavy metal removal from water solution.

The purpose of this work was to investigate the application of locally available agro industrial waste such as apricot stones (Prunus armeniaca L.) and corn cobs (Zea mays L.) as biosorbents for the rem-oval of copper ions from aqueous solutions.

Treatment of contaminated water using biosorp-tion is a complex process due to the influence of dif-ferent variables on biosorption efficiency. The effects of different variables can be evaluated using artificial neural network (ANN) [15]. Recently, second order polynomials models (SOPs) and ANNs have been used jointly for both modelling and optimization purposes in environmental studies [16]. ANN models are recognized as a good modelling tools since they provide the solution to the problems from a set of experimental data, and are capable of handling com-plex systems with nonlinearities and interactions between decision variables. Prediction of biosorption process using ANN has been attempted by many researchers with a reasonably good degree of suc-cess [17-19].

The specific objective in this study was to inves-tigate the effect of amount of biomass (m), initial con-centration (Ci) and volume (V) for two biosorption materials (apricot stones and corn cobs) on sorption of Cu(II). The performance of ANNs was compared with the SOP models, and also with experimental results. The focus of this study was to determine the optimal m, Ci and V for apricot stones and corn cobs, depending on maximum removal efficiency (R) and biosorption capacity (q). Developed empirical models give a reasonable fit to experimental data and suc-cessfully predict sorption characteristics.

METHODS

Biosorbent preparation

Apricot stones were obtained from Juice Factory “Vino Župa” Aleksandrovac and corn cobs were obtained from Maize Research Institute, Zemun Polje, Serbia. Both biomaterials were milled (KHD Humbolt Wedag AG) and <1 mm fraction was chosen for the biosorption tests without any pre-treatment.

Characterization of the biosorbents

Surface morphology of apricot stones and corn cob was determined by SEM-EDX (JEOL JSM-6610LV model). FTIR-ATR spectroscopic analyses were carried out using a Nicolet 380 spectrophoto-meter in the spectral range 4000 to 400 cm-1. FTIR characterization was performed in order to determine chemical functional groups that might be involved in

the biosorption process. The ANKOM 2000 fiber analyzer was used for determination of neutral deter-gent fiber (NDF) content of both biomaterials.

Preparation of stock solution

Stock solutions were obtained by dissolving pre-cise amount of Cu(NO3)2⋅3H2O (p.a. grade) in deion-ized water. Desired solutions of different copper ion concentrations were prepared by appropriate dilution of stock solution. Using a pH meter (Sension MM340), the pH value was adjusted with 0.1 M HNO3 and 0.1 M NaOH solutions. Buffering was not used due to unknown effects of buffer compounds on bio-sorption [20]. All experiments were performed in trip-licate.

Batch experiments

In order to optimize the experimental conditions, the batch studies were performed with different amounts of biomass (from 0.1 to 1.0 g), volumes of metal solution (50, 75 and 100 mL) and initial con-centrations of copper ions (60, 120, 180 and 240 mg L−1). The solutions including the metal ions and bio-sorbent were shaken during optimum contact time in shaker at 250 rpm. Equilibrium of the process was reached at 120 min of contact time [21,22]. After equilibrium was reached, the contents of the flasks were filtered and analyzed for the amount of copper remaining in the solution. The final concentration of Cu(II) ions was measured by atomic absorption spec-trometry (Perkin-Elmer, AAnalyst 300).

Variables affecting the biosorption process were: initial metal-ion concentration, amount of bio-mass, pH, sorption time, particle size, temperature, agitation rate [23]. The initial concentration of Cu(II) ions, amount of biomass and volume of metal sol-utions were three variables which were used as input vectors to train the network. The other variables were kept constant at their optimum values, as previous studies had indicated (pH 4.5, t = 120 min, T = 25 °C) [21,22].

The biosorption capacity of the biosorbent (q, mg g−1) was calculated as:

= −( ) /i eqq C C V m (1)

The removal efficiency of copper ions (R / %) was calculated as:

= −100( ) /i eq iR C C C (2)

where V is solution volume (L), m is mass of the sorbent (g), and Ci and Ceq (mg L-1) are the initial and final (or equilibrium) concentration of the copper ions in the solution, respectively.


251

Mathematical model

Second order polynomial (SOP) model

In order to check the significant effect of the input variables over the output and to justify the use of ANN model by coefficient of determination (r2), it is recommended to perform a statistical analysis of the available data, such as analysis of variance (ANOVA). For each dependent variable where factors were rejected when their significance level was p > 0.05, the SOP models were developed. All the analysis and mathematical modelling were performed using Statistica 10 software [25].

The SOP model estimated the main effect of the process variables on final products properties during the sorption of Cu(II). The independent variables were: m, Ci and V. The influences of independent variables on sorption characteristics of Cu(II), for both sorption materials, are presented in this work. All SOP models were fitted to data collected by expe-rimental measurements. The models of the following form were developed to relate six dependent outputs (Y) to three process variables (X):

β β β

β

= =

= = +

= + + +

+

3 32

0

1 1

3 3

1 1

k k ki i kii ii i

kij i ji j i

Y X X

X X

, k = 6 (3)

where: βk0, βki, βkii, βkij are constant regression coef-ficients; Yk, either equilibrium concentration of Cu(II) (Ceq), removal efficiency (R), biosorption capacity (q) for either apricot stones or corn cobs; while Xk are either sorbent weight (m), initial concentration (Ci) or volume (V). The significant terms in the model were found using ANOVA for each dependent variable.

Artificial neural network (ANN) modelling

According to StatSoft Statistica’s recommend-ations, the database is randomly divided to: training data (60% of data), cross-validation (20%) and testing data (20%). The cross-validation data set was used to test the performance of the network while training was in progress as an indicator of the level of general-ization and the time at which the network has begun to over train. Testing data set was used to examine the network generalization capability.

To improve the behaviour of the ANN, both input and output data were normalized according to Eq. (4):

( ) ( )= − −. min( ) / max( ) min( )i norm i i i ix x x x x (4)

where xi is i-th case, with measured Ceq, R or q, for either apricot stones or corn cobs. Normalized vari-ables gained values in the range of 0 to 1, and have no physical meaning.

In order to obtain good network behaviour, it is necessary to make a trial and error procedure and also to choose the number of hidden layers, and the number of neurons in hidden layer(s). The use of only one layer is advisable, because more layers exacer-bates the problem of local minima [26].

A multi-layer perceptron models (MLP) con-sisted of three layers (input, hidden and output), which is the most common, flexible and general-pur-pose kind of ANN. Such a model has been proven as a quite capable of approximating nonlinear functions [26], which is why it was chosen for this study. The network consists of one layer of linear output neurons and one hidden layer of nonlinear neurons. The MLP neural network learns using an algorithm called “back propagation”. The Levenberg-Marquardt algorithm proved to be the fastest and particularly adapted for networks of moderate size. During this iterative pro-cess, input data are repeatedly presented to the net-work [27].

The first estimation of the number of neurons can be obtained from the following equation [28,29]:

= + + +( 1) ( 1)w n x y n (5)

where x and y represent the number of input and output neurons, respectively, n is the number of neu-rons in the hidden layer and w is the number of weights (connections between layers) in the neural network. W can be taken as the number of training exemplars divided by 10. Some suggestions regard-ing the number of hidden neurons are as follows: this number should be between the sizes of the input and output layers, it should be 2/3 the size of the input layer, plus the size of the output layer, or less than twice the size of the input layer [28].

In this work, the ANN procedure of StatSoft Sta-tistica was used to model the ANN, and the number of hidden neurons, n, varied from 8 to 9. There were x = 3 inputs, y = 6 outputs, and 86 to 96 weight coef-ficients, w (depending on n). The Broyden-Fletcher- -Goldfarb-Shanno (BFGS) algorithm, implemented in StatSoft Statistica’s evaluation routine, was used for ANN modelling. The most common nonlinear acti-vation functions used in StatSoft Statistica ANN cal-culation are: logistic, sigmoid, hyperbolic and tangent functions (also exponential, sine, softmax, Gausian). In most applications, hyperbolic tangent function beh-aves better as compared to the other functions [30].

Coefficients associated with the hidden layer (both weights and biases) are grouped in matrices W1 and B1. Similarly, coefficients associated with the output layer are grouped in matrices W2 and B2. If Y is the matrix of the output variables, f1 and f2 are transfer functions in the hidden and output layers, respect-


252

ively, and X is the matrix of input variables, it is pos-sible to represent the neural network, by using matrix notation, as follows [31]:

= + +1 2 2 1 1 2( ( ) )Y f W f W X B B (6)

Weights (elements of matrices W1 and W2) are determined during the training step which updates them using optimization procedures to minimize the error function between network and experimental outputs [28,29], evaluated according to the sum of squares (SOS) and BFGS algorithm, used to speed up and stabilize convergence [32].

Training, testing and system implementation

After defining the architecture of ANN, the train-ing step is initiated. The training process was rep-eated several times in order to get the best per-formance of the ANN, due to a high degree of vari-ability of parameters. It was accepted that the suc-cessful training was achieved when learning and cross-validation curves (SOS vs. training cycles) approached zero. Testing was carried out with the best weights stored during the training step. Coef-ficient of determination (r2) and SOS were used as parameters to check the performance (i.e., the accu-racy) of the obtained ANNs.

After the best behaved ANN is chosen, the model is implemented using an algebraic system of equations to predict Ceq, R or q, for either apricot stones or corn cobs, using Eq. (6). This step can be easily achieved in some spreadsheet calculus (Mic-rosoft Office Excel, for instance).

Fuzzy synthetic optimization

The optimization procedure was performed using the fuzzy synthetic evaluation (FSE) algorithm implemented in Microsoft Excel 2007, in order to determine the workable optimum conditions for the biosorption of Cu(II). The FSE method was imple-mented using the results of models proposed to rep-resent R or q, for either apricot stones or corn cobs, using Eq. (7). FSE is commonly used technique to solve problems with constraints involving non-linear functions. The method aims to solve a sequence of simple problems whose solutions converge to the solution of the original problem.

Trapezoidal membership function used in this calculation, could be written as:

− ≤ < −= ≤ < − ≤ < − −

,

( , , , , ) , 1

, 1

x aa x u

u aA x a u v b u x v

x vv x b

b v

(7)

where x is whether R or q, for each biosorption material, and the values of a, b, u and v are function parameters. The interval a-b represents the range in which measured values occurred, while u-v is the expected optimal values range for output variables.


Characterization of biosorbents

In order to identify surface morphology of bio-sorbents, apricot stones and corn cobs were char-acterized by SEM-EDX (JEOL JSM-6610 LV electron micrograph). SEM micrographs (Figure 1) showed a

a b

c d

Figure 1. SEM Micrographs of apricot stones: 1000× (a) and 10,000× magnification (b), and corn cob: 1000× (c) and 5000× magnification (d).


253

highly porous surface of apricot stones. At 10,000× magnification, single pores with diameter of 1 μm can be seen. In comparison with apricot stones, surface of corn cob has a low porosity and on the corn cob surface only channels with diameter of 10 μm are vis-ible. Leyva-Ramos et al. have indicated that natural corn cob is an almost nonporous material because its surface area is less than 5 m2 g-1 [12].

The EDX spectra of apricot stones and corn cobs are shown in Figure 2. Both spectra indicate the

presence of potassium, magnesium, phosphorus and calcium. Also, the spectra reveal presence of copper after biosorption treatment, indicating that binding of copper ions to the surface of the material has occur-red.

Analyses of NDF content (neutral detergent fiber) confirmed that both materials, apricot stones and corn cob, are lignocellulosic consisting of 79.05 and 84.74% of lignin, hemicellulose and cellulose, respectively. FTIR analysis of the selected biomat-

Apricot stones

Apricot stones after biosorption

Corn cob

Corn cob after biosorption

Figure 2. The EDX spectra of apricot stones and corn cobs before and after adsorption of metal.


254

erials (Figure 3) showed high content of –COOH and –OH groups. Hydroxyl groups corresponding to carbo-xylic acids in cellulose and lignin are represented by broad, intense peaks at 3420 and 3421 cm-1 assigned to stretching vibrations of inter- and intra-molecular hydrogen bonds of polymeric compounds [33,34]. Furthermore, peaks observed at 2924 cm–1 can be attributed to symmetric and asymmetric C–H stretch-ing vibration of aliphatic acids, and peaks at 1742 cm-1 correspond to stretching vibration of C=O bonds pre-sent in non-ionic carboxyl groups like –COOH. Also, peaks at 1636 and 1647 cm-1 can be assigned to C=O stretching vibrations. Peaks at 1595 and 1605 cm-1 possibly relate to C=C stretching vibration, while 1507 and 1516 cm-1 could indicate presence of aromatic ring with C=C bonds. Two peaks at 1457 and 1424 cm-1 can be ascribed to C–O bond present in carbo-xylic groups, whereas number of peaks between 1376 and 1047 cm-1 can be assigned to C–O bonds in phe-nols. Intense peaks at 1047 and 1044 cm-1 can be ascribed to stretching vibrations of C–O bond present in carboxylic acids and alcohols [35]. It is highly possible that these groups in de-protonated forms are key sites for coordination of heavy metals [36].

Figure 4 summarizes all experimental data and shows the effect of three variables (amount of bio-

mass, volume and initial concentration of metal sol-ution) on biosorption capacity and removal efficiency for both biosorbents. The ratio of amount of biomass and volume of metal solution is a very significant fac-tor for metal removal during biosorption process as it determines equilibrium of sorbent and sorbate in the system [37,38]. With increase of the amount of bio-mass, the biosorption capacity decreased, which can be attributed to overlapping or aggregation of biosorp-tion sites [39,40]. Figure 3 reveals that biosorption capacity increased and removal efficiency decreased with increase of initial concentration of copper ions in solution, which means that the biosorption process is highly dependent on initial concentration of copper ions [38]. Due to saturation of the active sites on the surface of the biosorbents, the biosorption capacity has a tendency to stagnate (the curve becomes flat-tened).

Comparison between biosorption capacities of some adsorbents investigated by other researches is shown in Table 1. The results of present study are compatible with similar biosorbents by its potential for heavy metal removal from aqueous solution.

Variables Ceq, R or q, for either apricot stones or corn cobs, varied significantly, implying that fitting of the experimental data could be performed using SOP

a

b

Figure 3. FTIR Spectra of native apricot stones (a) and corn cob (b).


255

and ANN modelling. Calculation of objective function F has been performed using Eq. (7).

Table 1. The biosorption capacites obtained for sorption of Cu(II) onto different types of agro-industrial wastes

Biosorbent q / mg g-1 Ref.

Pine cone shell 6.81 [7]

Rice shell 2.95 [9]

Cocoa shell 2.87 [10]

Banana peel 4.75 [11]

Orange peel 5.25 [11]

Barley straw 4.64 [14]

Analysis of variance and SOP models

Analysis of variance (ANOVA) was conducted for obtained SOP models, and outputs were tested against the impact of input variables (Table 2). All the variables that entered into the analysis showed significant effects on the outputs, either as linear or square members, or in the form of a product of two variables. According to ANOVA (Table 2), Ceq is mostly affected by the linear term of Ci, statistically significant at p < 0.05 (for both biosorption materials). Linear terms of m and V also affected Ceq, statistically significant at level p < 0.05. The effects of non-linear terms of SOP model was also noticed, but were less prominent than linear terms (quadratic terms of Ci

Figure 4. Effect of amount of biomass, initial concentration and volume of metal solution.


256

and V have been found statistically significant, and also interchange terms m×Ci and m×V). R was almost equally affected by linear terms of m and Ci, statis-tically significant at p < 0.05 level. It was also strongly affected by m, and also V and Ci (at p < 0.05 level). The quadratic term of Ci showed statistically signi-ficant impact on R calculation q was mostly influenced by linear and quadratic terms of m. Both linear and quadratic terms of Ci and V have been found statis-tically significant at p < 0.05 level. All variables con-sidered in the ANOVA analysis were also used for the ANN modelling.

The residual variance, marked as error in Table 2, presents the model disagreement with the expe-rimental values (contributions of other members that are not described in the SOP model). All the deve-loped models showed statistically insignificant devi-ation from the experimental values of the model, which confirmed their suitability. The r2 is also a share of output variability of the system, which was cal-culated based on regression analysis. High r2 values also indicated that the experimental data satisfactorily coincided with the mathematical models.

Analysis revealed that linear, quadratic, as well as interchange terms considerably influenced forming SOP models. It is concluded that all of the obtained models were statistically significant and in agreement with experimental results.

Neurons in the ANN hidden layer

Determination of the appropriate number of hid-den layers and number of hidden neurons in each layer is one of the most critical tasks in ANN design. The number of neurons in a hidden layer depends on the complexity of the relationship between inputs and

outputs. As this relationship becomes more complex, more neurons should be added [30].

The optimum number of hidden neurons was chosen upon minimizing the difference between pre-dicted ANN values and desired outputs, using SOS during testing as performance indicator. Results of Ceq, R and q, for both apricot stones and corn cobs during testing with eight to nine neurons in the hidden layer are presented. Used MLPs are marked accord-ing to StatSoft Statistica’s notation, MLP followed by number of inputs, number of neurons in the hidden layer, and the number of outputs. According to ANN performance, regarding the sum of r2 and SOSs for all variables in one ANN, it was noticed that the optimal number of neurons in the hidden layer is 8 (network MLP 3-8-6), when obtaining high values of r2 and also low values of SOS.

The SOS between the experimental and the network predicted values was used as the iteration termination criterion, as StatSoft Statistica’s default. As soon as the cross-validation SOS starts to inc-rease, the training step is terminated; otherwise, the training step ends after a fixed number of epochs or training cycles [32].

Simulation of the ANNs

Process outputs Ceq, R and q, for both apricot stones and corn cobs can be calculated by the Eq. (6), using matrices W1 and B1, and matrices W2 and B2, which represent the system, incorporating coef-ficients associated with the hidden layer (both weights and biases). Output variables are calculated by apply-ing transfer functions f1 and f2 (from Table 2) in the hidden and output layers, respectively, onto the mat-rix of input variables X using Eq. (6). The algebraic

Table 2. ANOVA calculation; * - significant at p < 0.05 level; ** - significant at p < 0.10; level 95% confidence limit, error terms were found statistically insignificant; df - degrees of freedom

Variable df

Sum of squares

Apricot stones Corn cobs

Ceq R q Ceq R q

m 1 1501.3* 6335.5* 27.1* 340.7* 1538.8* 4.5*

m2 1 106.1* 135.1* 32.1* 19.2 15.3 11.4*

Ci 1 11274.9* 2826.5* 8.6* 14227.1* 386.4* 7.4*

Ci2 1 595.3* 958.8* 7.5* 85.7* 101.5* 4.3*

V 1 977.4* 3022.3* 0.4 139.3* 652.4* 0.3*

V2 1 14.6 2.5 0.8** 183.4* 7.4 10.9*

m×Ci 1 628.2* 1287.8* 2.2* 138.6* 339.2* 0.9*

m×V 1 621.6* 182.9* 23.7* 217.8* 89.8* 0.0

Ci×V 1 14.4 728.1* 0.4 7.6 141.9* 0.2

Error 110 2034.0 1328.5 30.8 1452.4 809.8 7.3

r2 0.996 0.966 0.842 0.997 0.907 0.928


257

system of equations is easily evaluated in a spread-sheet (i.e., Microsoft Excel) to predict Ceq, R and q, for both apricot stones and corn cobs, using Eq. (6), with shown calculated weights and biases matrices.

ANN models were used to predict experimental variables for both biosorption materials (Ceq, R and q). Figure 5 shows simulated curves in comparison with experimental data, for the best network found (MLP 3- -8-6). This network was able to predict reasonably well all process outputs for a broad range of the process variables, shown in Table 2. The predicted values were very close to the desired values in most cases, in terms of r2 value, for both SOP and ANN models. SOSs obtained with ANN models are of the same order of magnitude as experimental errors for Ceq, R and q reported in the literature [32].

SOP models using StatSoft Statistica 10 were also developed, and sum of squares and r2 values are presented in Table 2. It can be seen that these r2 values are much lower than those associated with the ANN model, which is in agreement with other authors [28,29,32]. Although ANN models are more complex (86-96 weights-biases for Ceq, R and q model, for five different ANNs) than SOPs, ANNs performed a bit better because of the high nonlinearity of the deve-

loped system [32]. ANN models gained somewhat better results, compared to SOPs (Table 2), regarding the r2 comparison between experimental and cal-culated outputs. The r2 values between experimental and SOP model outputs for Ceq, R and q, were 0.996, 0.966 and 0.842, respectively (for apricot stones) and 0.997, 0.907 and 0.928, respectively (for corn cobs), while the best ANN model (MLP 3-8-6, No 1) gained: 0.998, 0.987 and 0.927, respectively (for apricot stones) and 0.998, 0.986 and 0.982, respectively (for corn cobs), during the training period.

Fuzzy synthetic optimization

Fuzzy synthetic optimization of the output vari-ables was accomplished in order to find the m, Ci and V that give optimums of R and q. Trapezoidal mem-bership function was used, according to Eq. (7), in which a-b covered the complete interval of obtained output values, and u-v represented the optimal values (Table 3). The optimal parameters, used for FSE evaluation, were given based on our experience, calculating m as 35% of b.

The objective function (F) is the mathematical function whose maximum would be determined, by summing the FSE results for the ANN model, accord-

Figure 5. Target and ANN predicted values for Ceq, R and q (apricot stones and corn cobs).


258

ing to Eq. (6). Each input parameter (m, Ci and V) has the equal influence on the function F:

= +( , , )iF m C V R q (8)

Table 3. Trapezoidal membership function parameters

Parameter Apricot stones Corn cobs

R q R q

a 1.41 2.39 1.44 1.09

b 68.40 9.10 40.92 5.51

u 23.94 3.18 14.32 1.93

v 68.40 9.10 40.92 5.51

The maximum of function F represents the optimal m, Ci and V, and also the optimum of R and q, for both biosorption materials. The values of F were determined using Eq. (8). Values of membership function closer to 1 show the tendency of processing parameters to be optimal. Optimized process para-meters (inputs and outputs) for all groups are shown in Table 4.

Table 4. Optimizing parameters

Parameter Apricot stones Corn cobs

Optimal inputs m Ci V m Ci V

0.3 120 50 0.3 120 50

Optimal outputs R q R q

26.9 5.9 18.9 3.4

CONCLUSIONS

SOP- and ANNs-based models were developed to find adequate parameters (amount of biomass, initial concentration and volume of metal solution) in order to get optimal relation between biosorption efficiency (R) and biosorption capacity (q) from a given set of experimental data (240 experimental runs). As compared to SOP model, the ANN model yielded a bit better fit of experimental data, according to r2 and SOS of both models. The determined opti-mal parameters for both materials were the same: 0.3 g of biomass, initial concentration of 120 mg L-1 and solution volume of 50 mL. At these conditions, the copper removal efficiency was 26.9, 18.9 and bio-sorption capacity was 5.9 and 3.4 mg g-1 for apricot stones and corn cob, respectively. Firstly, study demonstrates the usefulness of ANN-based model for copper biosorption from aqueous solutions using sel-ected agro industrial waste materials. Secondly, the results of present investigation show that apricot stones and corn cob can be used as low cost adsorb-ents for removing the Cu(II) ions from aqueous sol-ution.

Acknowledgement

The authors are grateful to the Ministry of Education, Science and Technological Development of the Republic of Serbia for the financial support of this investigation included in the projects TR 31003 and TR 31055, project cycle 2011-2015.

REFERENCES

[1] A. Özer, D. Özer, A. Özer, Process Biochem. 39 (2004) 2183–2191

[2] J.L. Zuane, R. J. Kiff, Hand book of drinking water quality, second ed., Van Nostrand Reinhold, New York, 1996

[3] F. Ekmekyapar, A. Aslan, Y.K. Bayhan, A. Cakici, J. Hazard. Mater. 137 (2006) 293–298

[4] F.J. Cerino-Córdova, A.M. García-León, E. Soto-Regal-ado, M.N. Sánchez-González, T. Lozano-Ramírez, B.C. García-Avalos, J.A. Loredo-Medrano, J. Environ. Manage. 95 (2012) S77-S82

[5] N. Prakash, S.A. Manikandan, L. Govindarajan, V. Vijaya-gopal, J. Hazard. Mater. 152 (2008) 1268–1275

[6] J. Bayo, Chem. Eng. J. 191 (2012) 278–287

[7] G. Blázquez, M.A. Martín-Lara, E. Dionisio-Ruiz, G. Ten-orio, M. Calero, Ind. Eng. Chem. 18 (2012) 1741–1750

[8] N. Fiol, I. Villaescusa, M. Martinez, N. Miralles, J. Poch, J. Serarols, Sep. Purif. Technol. 50 (2006) 132–140

[9] H. Aydin, Y. Bulut, C. Yerlikaya, J. Environ. Manage. 87 (2008) 37–45

[10] N. Meunier, J. Laroulandie, J.F. Blais, R.D Tyagi, Biores. Technol. 90 (2003) 255–263

[11] A. Annadurai, R.S. Juang, D.J. Lee, Water Sci. Technol. 47 (2002) 185–190

[12] R. Leyva-Ramos, L.E. Landin-Rodriguez, S. Leyva-Ramos, N.A. Medellin-Castillo, Chem. Eng. J. 180 (2012) 113–120

[13] Z. Lopičić, J. Milojković, T. Šoštarić, M. Petrović, M. Mihajlović, Č. Lačnjevac, M. Stojanović, Hem. Iind. 67 (2013) 1007-1015

[14] E. Pehlivan, T. Altun, S. Parlayici, J. Hazard. Mater. 164 (2009) 982-986

[15] G. Annadurai, Y.L. La, F.L. Jiunn, Afr. J. Biotech. 6 (2007) 296–303

[16] F. Geyikci, E. Kılıc, S. Coruh, Ghosh Elevli, Chem. Eng. J. 183 (2012) 53–59

[17] K.H. Chu, Eur. J. Miner. Process. Environ. Prot. 3 (2003) 119–127

[18] K. Yetilmezsoy, S. Demirel, J. Hazard. Mater. 153 (2008) 1288–1300

[19] A.Ghosh, K. Sinha, P.D. Saha, Desalination Water Treat. 51 (2013) 7791-7799

[20] S. Klimmek, H.J. Stan, A. Wilke, G. Bunke, R. Buchholz, Environ. Sci. Technol. 35 (2001) 4283-4288

[21] M. Petrović, T. Šoštarić, M. Stojanović, Z. Lopičić, J. Mil-ojković, M. Mihajlović, in Proceedings of III International Congress “Engineering, Environment and Materials in


259

Processing Industry, Jahorina, Bosnia and Herzegovina, 2013, pp. 340-346

[22] T. Šoštarić, M. Petrović, Z. Lopičić, J. Milojković, Č. Lač-njevac, M. Stojanović, in Proceedings of III International Congress “Engineering, Environment and Materials in Processing Industry, Jahorina, Bosnia and Herzegovina, 2013, pp. 323-327

[23] M. Torab-Mostaedi. Chem. Ind. Chem. Eng. Q. 19 (2013); 79-88

[24] N. Chubar, J.R. Carvalho, M.J.N. Correia, Colloids Surfaces A: Physicochem. Eng. Aspects 230 (2003) 57– -65

[25] Statistica (Data Analysis Software System), v.10, StatSoft, Inc, Tulsa, OK, USA, 2010, (http://www.statsoft.com)

[26] M. Arsenović, Z. Radojević, S. Stanković, Ž. Lalić, L. Pezo, Ceram. Int. 39 (2013) 1667-1675

[27] S. Grieua, O. Faugeroux, A. Traoré, B. Claudet, J.L. Bodnar, Energy Build. 43 (2011) 543–554

[28] L.L. Pezo, B.Lj. Ćurčić, V.S. Filipović, M.R. Nićetin, G.B. Koprivica, N.M. Mišljenović, Lj.B. Lević, Hem. Ind. 67 (2013) 465-475

[29] I.C. Trulea, A.L. Raoult-Wack, G. Trystram, Food Sci. Technol. Int. 3 (1997) 459–465

[30] B.J. Taylor, Methods and Procedures for the Verification and Validation of Artificial Neural Networks, Springer Science and Business Media, Inc., New York, 2006

[31] T. Kollo, D. von Rosen, Advanced Multivariate Statistics with Matrices, Mathematics and its applications, Vol. 579, Springer, Dordrecht, 2005.

[32] L.A. Basheer, M. Hajmeer, J. Microbiol. Meth. 43 (2000) 3–31

[33] R. Gnanasambandam, A. Protor, Food Chem. 68 (2000) 327-332

[34] M. Iqbal, A. Saeed, S.I. Zafar, J. Hazard. Mater. 164 (2009) 161-171

[35] G. Guibaud, N. Tixier, A. Bouju, M. Baudu, Chemosphere 52 (2003) 1701-1710

[36] R. Ashkenazy, L. Gottlieb, S. Yannai, Biotechnol. Bioeng. 55 (1997) 1-10

[37] M.A. Hanif, R. Nadeem, H.N. Bhatti, N.R. Ahmed, T.M. Ansari, Ni(II) biosorption by Cassia fistula (golden shower) biomass, J. Hazard. Mater. 139 (2007) 345-355

[38] N. Ahalya, R.D. Kanamacdi, T.V. Ramachandra, Eur. J. Biotechnol. 8 (2005) 258-264

[39] A.A. Pal, S. Ghosh, A.K. Paul, Bioresour. Technol. 97 (2006) 1253–1258

[40] S. Kahraman, D. Asma (Hamamci), S. Erdemoglu, O. Yesilada, Eng. Life Sci. 5 (2005) 72–77.

MARIJA S. PETROVIĆ1

TATJANA D. ŠOŠTARIĆ1

LATO L. PEZO2

SLAVKA M. STANKOVIĆ3

ČASLAV M. LAČNJEVAC4

JELENA V. MILOJKOVIĆ1

MIRJANA D. STOJANOVIĆ1

1Institut za tehnologiju nuklearih I drugih mineralnih sirovina, Beograd

2Institut za opštu i fizičku hemiju, Univerzitet u Beogradu, Beograd 3Tehnološko-metalurški fakultet,

Univerzitet u Beogradu, Beograd 4Poljoprivredni fakultet, Univerzitet u

Beogradu, Beograd

NAUČNI RAD

PRIMENA ANN MODELA NA PROCES UKLANJANJA BAKRA IZ VODENIH RASTVORA UPOTREBOM AGRO-INDUSTRIJSKOG OTPADA

U radu je ispitana mogućnost upotrebe lokalno dostupnih lignoceluloznih materijala kao

potencijalnih biosorbenata u svrhu uklanjanja jona bakra iz vodenih rastvora. Ispitivani

materijali predstavljaju čvrst otpad koji nastaje nakon prerade kukuruza (oklasak kukuruza)

i nakon prerade voća (koštice kajsije). Ovakav otpad ima malu ekonomsku vrednost a

njegovo odlaganje predstavlja ekološki problem. U šaržnom sistemu ispitani su uticaj

inicijalne koncentracije Cu(II) jona (Ci), količine biomase (m) i zapremine rastvora (V) na

efikasnost biosorpcije i vrednost biosorpcionog kapaciteta. Utvrđeni su optimalni procesni

parametri. Eksperimentalni rezultati su poređeni sa dva modela: SOP (second order poly-

nomial regression models) i ANN (artificial neural networks), pri čemu je SOP model

pokazao prihvatljiv determinacioni koefijent (0,842-0,997), dok je ANN pokazao visoku

tačnost prognoze (0,980-0,986) u odnosu na eksperimentalne rezultate.

Ključne reči: biosorpcija, koštice kajsije, oklasak kukuruza, joni bakra, SOP, ANN.





261

BORE V. JEGDIĆ1

BILJANA M. BOBIĆ1

MILOŠ K. PAVLOVIĆ2

ANA B. ALIL3

SLAVIŠA S. PUTIĆ4 1Institute of Chemistry, Technology

and Metallurgy, University of Belgrade, Belgrade, Serbia

2Welding Institute, Belgrade, Serbia

3Innovation center, Faculty of Technology and Metallurgy,

University of Belgrade, Belgrade, Serbia

4Faculty of Technology and Metallurgy, University of Belgrade,

Belgrade, Serbia

SCIENTIFIC PAPER

UDC 669.715:620.194.2

DOI 10.2298/CICEQ140324024J

STRESS CORROSION CRACKING RESISTANCE OF ALUMINUM ALLOY 7000 SERIES AFTER TWO-STEP AGING

Article Highlights • SCC resistance of one-step and new two-step aged Al-Zn-Mg-Cu alloy was tested • SCC was tested by SSRT and using fracture mechanics method • After new two-step aging the alloy has considerably higher SCC resistance • Mechanical properties of one-step aged and new two-step aged alloy are similar • Two Ea values for process that controls stress corrosion crack rate were obtained Abstract

The effect of one-step and a new (short) two-step aging process on the resist-ance to stress corrosion cracking of an aluminum alloy 7000 series was inves-tigated, using slow strain rate test and fracture mechanics method. The aging level in the tested alloy was evaluated by means of scanning electron micro-scopy and measurements of electrical resistivity. It was shown that the alloy after the new two-step aging is significantly more resistant to stress corrosion cracking. Values of tensile properties and fracture toughness are similar for both thermal states. Processes that take place at the crack tip have been con-sidered. The effect of the testing solution temperature on the crack growth rate on the plateau was determined. Two values of the apparent activation energy were obtained. These values correspond to different processes that control crack growth rate on the plateau at higher and lower temperatures.

Keywords: aluminum alloys, stress corrosion cracking, aging, mech-anical properties.

Stress corrosion cracking (SCC) is a time-dep-endent process that occurs under the influence of residual or imposed tensile stress and specific cor-rosive environment. Localized corrosion (intergra-nular, pitting) usually proceeds to SCC. Mechanical damage on the metal surface may play the role of an initial crack [1-3].

Aluminum alloys 7000 series (Al-Zn-Mg-Cu) are characterized by high strength, but they are prone to SCC. However, the tendency of these alloys to SCC changes depending on the content of alloying ele-ments, mechanical, thermal and thermo-mechanical treatment [3-9]. Precipitation hardening of these alu-minum alloys takes place through the segregation of GP zones that are transformed through the intermed-

Correspondence: B.V. Jegdić, Institute of Chemistry, Techno-logy and Metallurgy, University of Belgrade, Njegoševa 12, Belgrade, Serbia. E-mail: [email protected] Paper received: 24 March, 2014 Paper revised: 28 May, 2014 Paper accepted: 8 July, 2014

iate η' phase into the equilibrium phase MgZn2 [10- -13]. Maximum strength in the aluminum alloys 7000 series is achieved when there is a mixture of GP zones and η' precipitates. In the state of maximum strength, the alloys are prone to SCC and exfoliation corrosion. In the over-aged state, the alloys are char-acterized with a good resistance towards both SCC and exfoliation corrosion, while in the partially over-aged state the alloys show a slightly lower resistance to SCC and high resistance to exfoliation corrosion [1,10,13,14]. Retrogression and re-aging procedures can be also applied in order to preserve high strength and resistance to SCC [15,16].

The presence of copper in the 7000 series alu-minum alloys has a beneficial effect on hardness, increasing the volume fraction of hardening precipi-tates. It was found that copper is incorporated in GP zones, making them more stable even at higher tem-peratures [10,17]. In addition, copper atoms replace zinc atoms in the hardening precipitates, particularly at temperatures above 150 °C, making the precipitate nobler [14,17,18].

B.V. JEGDIĆ et al.: STRESS CORROSION CRACKING RESISTANCE OF Al… Chem. Ind. Chem. Eng. Q. 21 (2) 261−268 (2015)

262

As it was mentioned above, aluminum alloys 7000 series in the over-aged state (after two-step aging) are characterized by high resistance to SCC and exfoliation corrosion. However, the aging time is relatively long, and hardness of the alloys is signific-antly reduced (∼15% compared to the state of max-imum hardness).

The influence of a surrounding environment temperature on the kinetics of the stress corrosion crack growth is of great practical importance, because a large number of SCC processes takes place at increased temperatures. The specific role of copper is not yet fully established for SCC occurrence in aque-ous saline solutions, even at room temperature. It was shown that the crack growth rate on the plateau, vpl, at temperatures below 40 °C is highly dependent on the copper content. No copper content depend-ency was observed at temperatures above 40 °C [19,20].

Based on the results reported [21-25], a shorter two-step aging process can be applied. The new two-step aging process was performed in this work, with shorter aging time. Structural and stress corrosion characteristics of the alloy in the state of maximum hardness as well as in the state after the new two-step aging were investigated in this study. In addition, SCC tests at different temperatures were done in a sodium chloride solution and apparent activation energies were determined.

Examinations of aluminum alloys resistance to SCC have been frequently performed by slow strain rate tests (SSRT) and fracture mechanics tests (FM). Results obtained by SSRT are qualitative in nature, while results obtained by the FM method have a quantitative character.

EXPERIMENTAL

The chemical composition of the tested experi-mental aluminum alloy is given in Table 1. Test samples were cut from rectangular extruded rods (80 mm×30 mm). The rods were obtained by continuous casting of billets and then subjected to homogeniz-ation and extruded.

Heat treatment of the samples was performed according to the following regimes:

• Solution heat treatment at 460 °C/(1 h), quenching in water at room temperature, then arti-ficially aging at 120 °C/(24 h) (one-step aging, indi-cated in this paper with TA).

• Solution heat treatment at 460 °C/(1 h), quenching in water at room temperature, artificially aging at 100 °C/(5 h), and then at 160 °C/(5 h) (two- -step aging, indicated in this paper with TB).

Tensile properties

Tensile properties (tensile strength, Rm, yield strength, Rp0.2, and elongation, A5) were determined using short cylindrical proportional specimens ∅ 8 mm (ASTM B557M). Test specimens were made from the alloy in the TA and TB state. The test specimens were cut in transverse direction from pressed rods. Tests were performed using a universal tensile machine Instron.

Micro structural examinations

Micro structural examinations were performed using scanning electron microscopy (SEM Jeol JSM– -6610LV). Before SEM analysis, the samples were prepared mechanically and electrochemically. The samples were ground using abrasive papers (1200 to 4000 grit) and polished using diamond paste (5/3 and 3/2 µm). The samples were then electrochemically polished in the perchloric acid (HClO4) and etched in Keller’s solution (2 ml HF + 3 ml HCl + 5 ml HNO3 + 190 ml distillated water) for 10 s.

The presence of intermetallic compounds, their number, size and volume fraction were determined with structural analyzer TAS+, for both thermal states of the alloy (TA and TB).

Electrical resistivity

The measurements of electrical resistivity were performed on the TA and TB samples. The method of measurement is described in ASTM B193 standard. Electrical resistivity was measured by a micro ohm-meter in accordance with the manufacturer’s instruct-ions. The value of the measured electrical resistivity (ρ) was recalculated into electrical conductivity (χ = = 1/ρ), as well as into the IACS (%) factor, using the following equation:

χχ

=Cu

100IACS (1)

where χ is the value of electrical conductivity of the tested alloy, and χCu is the electrical conductivity of pure copper (58.34 MS m-1).

Table 1. Chemical composition of tested aluminum alloy (wt.%)

Zn Mg Cu Mn Cr Zr Fe Si Al

7.2 2.15 1.46 0.28 0.16 0.12 0.12 0.05 Rest


263

Slow strain rate test

The susceptibility of the aluminum alloy to SCC was determined by the SSRT method. The index of tendency to SCC (ISCC) was calculated based on the elongation to fracture during testing of identical samples in the corrosive environment and in the air:

ISCC = ASCC/A0 (2)

where ASCC is elongation in the corrosive environ-ment, and A0 is elongation in the air.

For an alloy resistant to SCC in a given cor-rosive environment the value of ISCC → 1, while for an alloy prone to SCC the value of ISCC → 0.

The SSRT was performed at a chosen value of the initial strain rate from the critical interval of strain rates. Test specimens were of a circular cross section (∅ 6 mm) and of 30 mm working length (ASTM E8). Before the test, dimensions of the specimens were carefully measured. Then, the specimens were deg-reased in ethanol and placed in the cell for SCC testing. The tests were done using the Instron tensile machine, in the air at the standard initial strain rate and in the corrosive environment (2 wt.% NaCl + 0.5 wt.% Na2CrO4, pH 3). In this case, the initial strain rate was 6.94×10-6 s-1. The data for the calculation of the ISCC were taken from the obtained stress-strain curve.

Fracture mechanics method

A bolt-loaded double-cantilever-beam (DCB) test specimen was chosen for testing SCC by the FM methodology. The samples were cut in the S-L ori-entation, since aluminum alloys are most sensitive to SCC at this orientation (force action in the short transverse direction).

The thickness of the sample (B) was calculated based on the known fracture toughness (KIC) and the yield strength (R0.2) of the alloy, using the equation:

= 2IC

p0.2

min2.5( )K

BR

(3)

Other dimensions of the DCB specimen were determined in dependence on the sample thickness (B = 25 mm). The starting value of the KI0 was calculated on the basis of the measured value of the crack length (a), the specimen half-height (H) and the given size of the crack opening on the line of the load (VY), according to the following equation (ISO 7539-6):

( )( )

+ +=

+ +

2 3

I0 3 2

3 0,6

4 0.6

YE H H a HV HK

a H aH (4)

The starting crack was formed mechanically on the specimen [26], and the crack length was precisely measured. The SCC testing was done in accordance with the procedure described [10] in the 3.5 wt.% NaCl solution. The increase of the crack length was measured microscopically in the next 150 days. The diagram of time dependence of the crack length was fitted and smoothed with the appropriate curve. That curve is used for calculating and drawing the diagram of the crack growth rate dependence on the KI. At the end of testing, a fracture of the samples was per-formed. On the surface of the fracture, the length of the crack formed mechanically and the total length of the mechanical and stress-corrosion crack were pre-cisely measured. The real value of the stress intensity in the beginning of the SCC test (that approximately corresponds the value of the KIC) and the real value of the stress intensity when the crack practically stops, KISCC, were calculated (Eq. (4)).

In addition, SCC tests were performed in the 3.5 wt.% NaCl solution at different temperatures, in order to determine the apparent activation energy of the process that controls crack grow rate on the plateau. The length of the crack was measured during the time and the vpl was calculated for every chosen tempe-rature.


Tensile properties

Tensile properties of the alloy in the TA and TB state are given in Table 2. The values of tensile properties are rather high and similar for both thermal states. It can be seen that the value of Rp0.2 is slightly higher for the thermal state TB, while the values of Rm and A5 are somewhat higher for the thermal state TA.

Table 2. Tensile properties of tested aluminum alloy

Thermal state Rp0.2 / MPa Rm / MPa A5 / %

TA 560 620 10.5

TB 570 600 9.5

Micro structural examinations

The typical microstructure of the alloy (TA and TB state) at low magnification is shown in Figure 1. It can be seen that the alloy is characterized with oriented structure. Grains are elongated and particles of secondary phases are oriented in direction of deformation. There are large light particles and small dark particles of intermetallic compounds of different size. Chemical composition of these particles was determined using SEM/EDS analysis. Typical chem-


264

ical composition of the intermetallic compounds is presented in Table 3.

Figure 1. Typical SEM microphotograph of the alloy.

Fracture toughness of the alloy is highly dependent on the size and volume fraction of par-ticles of secondary phases [27,28]. The minimum particle size is 0.67 µm, the maximum particle size is 2.12-2.94 µm, the volume fraction is 2.14 vol.% and the particle number per mm2 is 7200 for the alloy in TA and TB state.

Microstructures of the alloy after one-step and after two-step aging at higher magnification are

shown in Figures 2 and 3. After one-step aging (Fig-ure 2a and 2b) the precipitates are smaller with regard to the precipitates after the two-step aging (Figure 3a and 3b). The mixture of η' phase and stable η phase was formed during the two-step aging (TB). It is possible that most of these precipitates were dissolved in the etching solution. Reduced time of the two-step aging (100 °C/(5 h) + 160 °C/(5 h)) compared to the time of the standard one-step aging (120 °C/(24 h)) can explain the difference in the size of strengthening particles.

Electrical resistivity

Values of the IACS factor change during the second step of two-step aging (TB) as is shown in Table 4.

Table 4. Time dependence of IACS factor for tested aluminum alloy during second step of two-step aging

Time / h 0 4 8 12 16 20 24

IACS / % 31.04 35.92 38.57 40.49 41.43 41.63 41.95

The experimentally obtained results show that the TB state has larger conductivity (36.71% IACS) than the TA state (32.56% IACS) and the T0 state

Table 3. Typical chemical composition of intermetallic compounds (at.%)

Element Mg Al Si Cr Mn Fe Cu Zn

Light particles 0.38 83.5 0.00 0.13 1.32 8.98 2.32 1.72

Dark particles 16.9 65.3 13.7 0.00 0.00 0.00 0.43 1.78

Figure 2. SEM microphotographs of the tested aluminum alloy after one-step aging.

Figure 3. SEM microphotographs of the tested aluminum alloy after two-step aging.


265

immediately after homogenization annealing (29.65% IACS). A supersaturated solid solution with a high concentration of vacancies was obtained after quenching. Fields of elastic strains around vacancies cause significant dissipation of electrons [29], so lower values of conductivity are measured for the thermal state T0. During aging, elastic strains around GP zones and semi coherent η' phase cause electron dissipation to a certain extent (thermal state TA). With appearance of the stable, incoherent η phase during two-step aging, elastic strains decrease, and the alloy conductivity increases. Formed precipitates get coarser and conductivity still increases with a pro-longed time of aging (41.95% IACS, after 24 h, Table 4). However, mechanical characteristics of the alloy (hardness) are decreased [10], as well as the resist-ance to SCC and exfoliation corrosion. The resistance to SCC and exfoliation corrosion for aluminum alloys 7000 series was evaluated according to values of electrical conductivity [30]. A detailed model for the electrical conductivity of 7000 aluminum alloys under various aging conditions was presented [31].

Slow strain rate test

The results of SCC testing with SSRT method are presented in Figure 4a. It can be seen that the alloy in the TB state is more resistant to SCC than in the TA state. Index of tendency to stress corrosion cracking ISCC for the alloy after one-step aging is 60.9% and after two-step aging is 98.9%. Processes of crack formation and growth and the final unstable fast fracture are not separated, so the total tendency to stress corrosion cracking is obtained using this method.

A typical experimental stress-strain curve for the alloy in TA state is shown in Figure 4b. Elongation of the specimen in the corrosive environment (NaCl + Na2CrO4) is significantly lower compared to the elong-

ation in the air. Index of resistance to SCC for the alloy in TA and TB (ISCC) state was determined from similar diagrams. The results were obtained in a rel-atively short time (about 10 h per one sample).

Fracture mechanics method

The results of SCC testing by fracture mecha-nics method are presented in Figure 5, as the depen-dence of the stress corrosion crack rate on the KI.

Figure 5. v–KI dependence for the tested aluminum alloy in

TA and TB state.

The value of critical stress intensity factor for stress corrosion crack rate KISCC is 10.87 MPa m1/2 for the alloy in TA state, and 12.87 MPa m1/2 for the alloy in TB state. Higher value of KISCC indicates better resistance to SCC. When the KI is smaller than the KISCC, there is no growth of the stress corrosion crack or the growth rate is too small that can be neglected. In the first stadium (steep part of the curve v-KI), the crack growth rate strongly depends on the KI. In the

Figure 4. a) ISCC values for the tested aluminum alloy in TA and TB state; b) Experimental stress – strain curve for the alloy in TA state.


266

second stadium (plateau of the curve v-KI), the crack growth rate practically does not depend on the KI. The influence of the alloy heat treatment is significant and it reflects in shifting the level of the second stadium to higher values for the alloy in the thermal state TB (Figure 5). The value of crack propagation rate on the plateau (vpl) is 14.4×10-9 m s-1 for the thermal state TA, while it is significantly lower for the thermal state TA (1.16×10-9 m s-1).

In the beginning of the test the value of the stress intensity factor KI0 (that approximately corres-ponds to the value of KIC) was also determined. The value of KI0 for the thermal state TA is 28.25 and 29.25 MPa m1/2 for the thermal state TB. It can be seen that fracture toughness of the alloy is somewhat higher after two-step aging than after one-step aging. The size of intermetallic phases and their volume fraction highly influence the fracture toughness of alu-minum alloys. Fracture toughness KIC of the tested alloy was calculated according to the model of Hahn and Rosenfield [27,28] using the previously mentioned values for the particle size and volume fraction of the secondary phases. Calculated values of KIC are in good agreement with experimental values.

It can be seen that that the alloy is more res-istant to SCC after two-step aging (TB). The differ-ence in SCC resistance is highly reflected on the value of the crack propagation rate on the plateau vpl (Figure 5). The value of the vpl is lower for more than one order of magnitude for the TB state. The structure of the tested alloy after two-step aging is considerably more resistant to SCC than the structure after one-step aging. The maximum strength of the alloy (TA state) is obtained when GP zones and η' precipitates are present in the alloy structure. Local plastic defor-mation at the tip of the stress corrosion crack causes mainly planar slipping when dislocations cut GP zones and smaller particles of η' phase. It comes to the accumulation of dislocations at grain boundaries at the crack tip, which causes increase in local stress, so that SCC starts at lower external stress. This creates favorable conditions for SCC to occur accord-ing to the mechanism of local hydrogen embrittlement or to the mechanism of anodic dissolution.

In the TB state, a great number of GP zones is created in the first step, at lower aging temperature (100 °C). During the second step of aging (160 °C) particles of η' phase are precipitated on GP zones, where they are partially transformed into the stable η phase. In this case, the local plastic deformation at the crack tip is homogeneous. Dislocations cannot succeed in cutting particles of the stable η phase and due to this the dislocations are uniformly distributed inside the grains. There is no local increase in stress

at the grain boundaries, which has a favorable inf-luence to SCC resistance. The presence of the stabile η phase in the thermal state TB has no significant influence on the values of mechanical properties of the alloy (Table 2). Duration of the two-step aging performed in this work (Experimental part) was shorter compared to the time of the standard one-step and standard two-step aging. The values of tensile properties and fracture toughness are similar for the alloy in both thermal states (TA and TB).

Tested aluminum alloy contains 1.46 wt.% Cu (Table 1). Alloying with copper reflects in the elec-trochemical characteristics of the alloy. In the Al-Zn- -Mg alloys, the η phase precipitated on the grain boundary is anodic compared to the grain itself (which is covered with a passive film). These conditions are favorable for intergranular corrosion and SCC to occur. In the tested alloy, copper atoms enter into the solid solution and into the Mg(AlCuZn)2 precipitates, making them nobler [14,17,18]. The precipitates on the grain boundaries are dissolved slower, and the cathodic reaction (hydrogen ion reduction) becomes more difficult [10,14]. In the presence of copper, GP zones are more stable at higher temperatures and the fraction of the strengthening precipitates in the alloy increases. This results in the increase in strength and resistance to SCC [10,17,18]. Accordingly, copper affects the mechanism of plastic deformation (slip-ping) and the electrochemical characteristics of the solid solution and precipitates. The influence of other factors, such as formation of magnesium hydride at grain boundaries [11,32-35] and other possible mechanisms of SCC occurrence are not considered in this paper.

The influence of the test solution temperature on the vpl is shown in Fig. 6. The exponential increase of the vpl with the increase in temperature was noticed, which is expressed by the following equation:

v = v0exp (-Ea/RT) (5)

where Ea is the apparent activation energy of the process that controls the vpl.

If the previous equation is logarithmed, a linear dependence between the logarithm of the vpl and the reciprocal value of the temperature is gained:

ln v = ln v0 – Ea/RT (6)

The dependence of the vpl on the reciprocal value of the temperature is shown in Figure 6. From the slope of the straight line, the apparent activation energy of the process that controls the vpl is obtained. There are two values of the apparent activation energy. One value (Ea = 46.6 kJ mol-1) refers to the temperature interval from 23 to 83 °C, while the other


267

(Ea = 70.4 kJ mol-1) refers to the lower testing tem-peratures, from 3 to 23 °C. These values of apparent activation energies correspond to different processes that control the vpl. This is in accordance with the reported results [10,19].

Figure 6. Temperature dependence of vpl during SCC testing

(TA state).

According to [19,20] crack propagation rate on plateau, vpl, at temperatures above 40 °C is asso-ciated with aluminum hydride formation. The decom-position of aluminum hydride within the crack-tip region leads to significantly enhanced local entry of hydrogen, which facilitates the observed increase of vpl, independent of the copper content. Crack pro-pagation rate on plateau at temperatures below 40 °C is dependent on the availability of hydrogen gener-ated via the electrochemical process. Anodic dissol-ution and grain boundary slipping are possible can-didates for occurrence of SCC processes at temperatures below 40 °C.

CONCLUSIONS

Resistance to SCC of a high strength 7000 series aluminum alloy was tested. The alloy after the new two-step aging shows considerably higher resist-ance to SCC compared to the alloy after one-step aging. The values of tensile properties and fracture toughness are similar for both thermal states. Dur-ation of the two-step aging was very short.

The results obtained by the SSRT method indi-cate total resistance of the alloy to SCC. It was shown that the thermal state TB is more resistant to the SCC than the thermal state TA. The value of ISCC is 98.9% for the thermal state TB and 60.9% for the thermal state TA. The value of KISCC is higher for thermal state TB (12.87 MPa m1/2) compared to thermal state TA

(10.87 MPa m1/2). The value of vpl is lower for more than one order of magnitude for the two-step aged alloy (1.16×10-9 compared to 14.4×10-9 m s-1 for the thermal state TA). Two values of apparent activation energy for the process that controls the vpl were obtained.

Acknowledgement

This work was co-financed from the Ministry of Education, Science and Technological Development of the Republic of Serbia through projects TR 34028 and TR 34016.

REFERENCES

[1] H.R. Jones, in Corrosion: Fundamentals, Testing, and Protection, Vol 13A, D.S. Cramer, B.S. Covino, Ed., ASM International, Materials Park, OH, 2003, p. 346

[2] D. Landolt, Corrosion and Surface Chemistry of Metals, EPFL Press, Lausanne, 2007, p. 491

[3] W.R. Herberg, Deformation and Fracture Mechanics of Engineering Materials, John Wiley. and Sons, New York,, 1983, p. 329

[4] O.M. Speidel, Metall. Mater. Trans., A 6 (1975) 631–651

[5] F.B. Brown, Stress Corrosion Cracking Control Mea-sures, NACE monography 156, NACE, Houston, TX, 1977, p. 23

[6] J. Moran, in Corrosion: Fundamentals, Testing, and Pro-tection, Vol 13A, D.S. Cramer, B.S. Covino, Ed., ASM International, Materials Park, OH, 2003, p. 275

[7] S.G. Peng, H.K. Chen, Y.S. Chen, C.H. Fang, Mater. Corros. 63 (2012) 254–258

[8] J. Onoro, Mater. Corros. 61 (2010) 125–129

[9] S.C. Paglia, V.K. Jata, G.R. Buchheit, Mater. Corros. 58 (2007) 737–750

[10] W.A. Thompson, M.I. Bernstein, in Advances in Corrosion Science and Technology, Vol 7, G.M. Fontana, W.R. Staehle, Ed., Plenum Press, New York, 1980, p. 53

[11] M. Bobby, P. Kannan, S. Bala, S.V. Raja, in Stress Cor-rosion Cracking, Theory and Practice, S.V. Raja, T. Shoji, Ed., Woodhead Publishing, Oxford, 2011, p. 307

[12] S.V. Sinjavskij, D.V. Valjkov, D.V. Kalinin, in Corrosion and Protection of Aluminium Alloys, Metalurgija, Moskva, 1986, p. 43

[13] B.V. Jegdić, Sci.-Tech. Rev. 13 (2003) 19–24

[14] T.J. Summerson, D.O. Sprowls, in: Aluminum Alloys, their Physical and Mechanical Properties, EMAS, Cradley Heath, 1986, p. 1576

[15] Y.P. Xiao, Q.L. Pan, W.B. Li, X.Y. Liu, Y.B. He, Mater. Des. 32 (2011) 2149–2156

[16] M. Angappan, V. Sampath, B. Ashok, V.P. Deepkumar, Mater. Des. 32 (2011) 4050–4053

[17] E.J. Hatch, Aluminum: Properties and Physical Metal-lurgy, ASM International, Materials Park, OH, 1984, p. 242


268

[18] F.L. Mondolfo, Aluminum Alloys: Structure and Pro-perties, Butterworths, London, 1976, p. 842

[19] N.J. Henry Holroyd, M.G. Scamans, Metall. Mater. Trans., A 42 (2011) 3979–3998

[20] N.J. Henry Holroyd, G.M. Scamans, Metall. Mater. Trans., A 44 (2013) 1230–1253

[21] G. Sha, A. Cerezo, Acta Mater. 52 (2004) 4503–4516

[22] G. Sha, A. Cerezo, Acta Mater. 53 (2005) 907–917

[23] L.B. Ou, G.J. Yang, Y.M. Wei, Metall. Mater. Trans., A 38 (2007) 1760–1773

[24] T. Marlaud, A. Deschamps, F. Bley, W. Lefebvre, B. Baroux, Acta Mater. 58 (2010) 248–260

[25] P.S. Lynch, Metall. Mater. Trans., A 44 (2013) 1209–1229

[26] B. Phull, in Corrosion: Fundamentals, Testing, and Pro-tection, Vol 13A, D.S. Cramer, B.S. Covino, Ed., ASM International, Materials Park, OH, 2003, p. 575

[27] G.T. Hahn, A.R. Rosenfield, Metall. Mater. Trans., A 6 (1975) 635–668

[28] R.H. Van Stone, J.A. Psioda, Metall. Mater. Trans., A 6 (1975) 668–670.

[29] I.I. Novikov, Theory of Heat Treatment of Metals, Mir Publishers, Moscow, 1978, p. 298

[30] J.R. Davis, Corrosion of Aluminum and Aluminum Alloys, ASM International, Materials Park, OH, 1999, p. 99

[31] J.M. Starink, M.X. Li, Metall. Mater. Trans., A 34 (2003) 899–911

[32] A.G. Young, R.J. Scully, Metall. Mater. Trans., A 33 (2002) 101–115.

[33] R.K. Cooper, G.R. Kelly, Corros. Sci. 49 (2007) 2636– -2662

[34] G.R. Song, K.M. Tseng, J.B. Zhang, J Liu, H.Z. Jin, S.K. Shin, Acta Mater. 44 (1996) 3241–3248

[35] D.T. Burleigh, Corrosion 47 (1991) 89–97.

BORE V. JEGDIĆ1

BILJANA M. BOBIĆ1

MILOŠ K. PAVLOVIĆ2

ANA B. ALIL3

SLAVIŠA S. PUTIĆ4

1Institut za hemiju, tehnologiju I metalurgiju, Univerzitet u Beogradu,

Beograd 2Institut za zavarivanje, Beograd

3Inovacioni centar Tehnološko-metalurškog fakulteta, Beograd

4Tehnološko-metalurški fakultet, Univerzitet u Beogradu, Beograd

NAUČNI RAD

OTPORNOST ALUMINIJUMSKE LEGURE SERIJE 7000 PREMA NAPONSKOJ KOROZIJI POSLE DVOSTEPENOG TERMIČKOG TALOŽENJA

Ispitivan je uticaj jednostepenog i novog (kratkotrajnog) dvostepenog termičkog taloženja

na otpornost prema naponskoj koroziji aluminijumske legure serije 7000, primenom

metode male brzine deformacije i metode mehanike loma. Stepen starenja ispitivane

legure je procenjen primenom skening elektronske mikroskopije i na osnovu merenja elek-

trične otpornosti. Pokazano je da je legura posle novog dvostepenog termičkog starenja

znatno otpornija prema naponskoj koroziji. Vrednosti zateznih karakteristika i žilavosti loma

su slične za oba termička stanja legure. Razmatrani su procesi koji se odvijaju na vrhu

naponsko-korozione prsline. Određen je uticaj temperature rastvora za ispitivanje na

brzinu rasta prsline na platou. Određene su dve vrednosti prividne energije aktivacije. Ove

vrednosti odgovaraju različitim procesima koji kontrolišu brzinu rasta prsline na platou, na

visokim i niskim temperaturama.

Ključne reči: aluminijumske legure, naponska korozija, starenje, mehaničke osobine.





269

ALEKSANDRA PETROVIČ MARJANA SIMONIČ

Faculty of Chemistry and Chemical Engineering, University of Maribor,

Maribor, Slovenia

SCIENTIFIC PAPER

UDC 628.1/.16:66.095.82

DOI 10.2298/CICEQ131129026P

THE EFFICIENCY OF A MEMBRANE BIOREACTOR IN DRINKING WATER DENITRIFICATION

Article Highlights • High nitrate removal efficiencies (more than 90%) were obtained at flow rates below

4.8 L/h • The denitrification efficiencies were highly dependent on the the hydraulic retention

time • The highest specific denitrification rate was achieved at 0.2738 g/L NO3/(g/L MLSS d) • The maximum reactor removal capacity was calculated at 8.7472 g NO3/m

3 h Abstract

The membrane bioreactor (MBR) system was investigated regarding its nitrate removal capacity from drinking water. The performance of a pilot-scale MBR was tested, depending on the operational parameters, using sucrose as a carbon source. Drinking water from the source was introduced into the reactor in order to study the influence of flow rate on the nitrate removal and denit-rification efficiency of drinking water. The content of the nitrate was around 70 mg/L and the C/N ratio was 3:1. Nitrate removal efficiencies above 90% were obtained by flow rates lower than 4.8 L/h. The specific denitrification rates varied between 0.02 and 0.16 g/L NO3/(g/L MLSS d). The efficiencies and nitrate removal were noticeably affected by the flow rate and hydraulic reten-tion times. At the maximum flow rate of 10.2 L/h still 68% of the nitrate had been removed, while the highest specific denitrification rate was achieved at 0.2738 g/L NO3/(g/L MLSS d). The maximum reactor removal capacity was calculated at 8.75 g NO3/m

3 h.

Keywords: capacity, denitrification, drinking water, efficiency, mem-brane bioreactor, sucrose.

Drinking water sources contaminated by nitrate and nitrite represent one of the greater environmental concerns and risks for human health. High standards for drinking water are recommended by the Council of European Communities, a maximum of 11.3 mgNO3-N/L for nitrate, and 0.03 mgNO2-N/L for nitrite [1,2]. If denitrification efficiency is studied, accurate standard methods must be chosen in order to gain reliable analyses of water samples [3-6]. Several methods (physical, chemical, physico-chemical, and biological) for drinking water treatment have been examined to date. In the process of biological denitrification, which

Correspondence: A. Petrovič, Faculty of Chemistry and Chem-ical Engineering, University of Maribor, Smetanova 17, SI-2000 Maribor, Slovenia. E-mail: [email protected] Paper received: 29 November, 2013 Paper revised: 25 January, 2014 Paper accepted: 8 July, 2014

is considered as one of the more cost-effective and friendly methods in relation to the environment [7,8], bacteria use nitrate as an electron acceptor for res-piration under anoxic conditions [1,9]. An external car-bon source, i.e., an electron donor, should be ensured for the activities of microorganisms during heterotrophic denitrification [10]. The presence of cer-tain carbon sources importantly effects the denitri-fication rate and COD demand. The efficiency of hete-rotrophic denitrification is usually relatively high; nevertheless, the nitrate and nitrite accumulation have been found to inhibit complete nitrate removal [10-12].

Membrane processes could efficiently remove several contaminants from the contaminated water: organic and inorganic matter, bacteria, various sus-pended solids and residue carbon sources [13]. The membrane bioreactor (MBR) is a combined system of biological treatment and membrane filtration that offers several advantages: high effluent quality, low

A. PETROVIČ, M. SIMONIČ: THE EFFICIENCY OF A MEMBRANE BIOREACTOR… Chem. Ind. Chem. Eng. Q. 21 (2) 269−275 (2015)

270

sludge production, lower needs of chemical agents, and high denitrification rates [14-16].

On the other hand, one of the major problems associated with conventional denitrification systems and MBRs is the problem of the direct contact between the denitrifying organisms and the water [17]. An interesting alternative for solving this problem is the membrane bioreactor/membrane contactor sys-tem (MBR/MC). In contrast with other MBR systems, this system uses a special membrane for separating the biomass from the water and therefore reducing the risk of induced microbial contamination [18]. A bench-scale microporous membrane bioreactor has also been developed for this purpose [17]. However, in conventional processess for drinking water treat-ment a process of disinfection is usually required that could increase the operational costs. Despite the above mentioned weaknesses the potential of a membrane bioreactor for nitrate removal in drinking water has been investigated by many other researchers [2,15,19-21].

The process highly depends on the operating conditions, particularly the inflow nitrate concentration and the flow rate. Other important factors in the rem-oval of pollutants by the MBR system are the hyd-raulic retention and sludge retention times [15,22]. However, due to the formation of filter cake, mem-brane separation efficiency decays over time [23]. Recently, studies concerning membrane bioreactors and denitrification have mostly been related to the operation parameters such as pH and oxygen levels [12,24], as well as the different types and effects of external carbon sources [10,19]. Moreover, some investigations have focused on the kinetic properties of the process [1,9,12], and some on the modelling of the process [1,14,16]. As the MBR operation is asso-ciated with energy consumption and operational costs, many researches have examined the fouling itself and the factors influencing membrane fouling [16,25]. However, applications of MBRs in drinking water treatment have been very limited regarding nit-rate removal capacity calculations.

The basic purpose of this research was to per-form a series of experiments in order to study the efficiency of nitrate removal from drinking water. Denitrification is well documented in the literature, yet the novelty of present paper is in high process effi-ciencies using cost effective sucrose in comparison with more expensive carbon sources, such as etha-nol, methanol, glucose, etc. The influences of flow rates and hydraulic retention time on the biological process in MBR were studied. The specific denitri-fication rate and the nitrate removal capacity of the reactor were calculated.

EXPERIMENTAL

Pilot plant and operational conditions

Drinking water enriched with nitrates, with suc-rose as the carbon source, was introduced into a Zenon ZeeWed 10 membrane bioreactor. A series of experiments were carried out in order to study the efficiency of heterotrophic denitrification from drinking water samples. The inflow rates gradually increased from 0.6 up to 10.2 L/h. The steady-state conditions were ensured at each inflow rate. A submerged hol-low-fibre membrane with a pore size of 0.04 µm was used in the experiments. The membrane had 0.93 m2 of active surface area (A) and consisted of polyvinyl-idene difluoride (PVDF) material. The mixing within the reactor was provided by a stirrer and by constant backflow. The nitrate concentration varied between 68.5 and 75.6 mg/L in the inflow, whilst the C/N ratio was constant at 3:1. The pH level and the tempera-ture within the reactor were controlled continuously by using an automatic control unit. The operating tempe-rature of the 60 L reactor was within the range of 23.7-29.0 °C, with a pH value of around 8.0. Anoxic conditions within the reactor were ensured by flushing with nitrogen gas. The denitrifying culture was taken from a local wastewater treatment plant and adapted for 20 days under anoxic conditions.

Analytical procedures

The samples for analyses were taken from inf-luent and efluent, and were collected daily. Several measurements were obtained in order to follow the denitrification efficiency: temperature, pH, the con-centration of nitrate and nitrite ions, the chemical oxygen demand (COD), and the concentration of acti-vated sludge (mixed liquor suspended solids, MLSS). The average values from this series of measurements were calculated at different flow rates. All the mea-surements were performed in accordance with ISO standards. Analyses for nitrate and nitrite concentra-tions were conducted spectrophotometrically [4,5] by using an Agilent 8453 UV-Visible spectrophotometer at 324 and 540 nm wavelengths, respectively. The chemical oxygen demand (COD) was determined by a volumetric method using titration with KMnO4 (determination of permanganate index) [3]. MLSS concentrations were measured occasionally accord-ing to standard methods for the examination of water and wastewater [6].

Calculations

The nitrate removal efficiency, R (%), and the specific denitrification rate, S (g/LNO3/(g/L MLSS d)), were determined in order to evaluate the suitability of


271

the membrane bioreactor for drinking water denitri-fication. Moreover, the MBR system’s nitrate removal capacity, EMBR (g/(m2 h)), and the reactor’s nitrate removal capacity, ER (g/(m3 h)), were calculated. The equations used for the calculations were taken from the literature [23].

The MBR system nitrate removal capacity (EMBR) is defined as the product of the permeate flux (JP = q/A) and the difference between the average inflow and outflow nitrate concentrations:

γ γ= −MBR P 3 0 3 t( (NO ) (NO ) )E J (1)

Reactor nitrate removal capacity, ER, (Eq. (3)) is a function of the dilution rate, D (h–1), (Eq. (2)), which is defined as the quotient of the influent flow rate (q) and the bioreactor volume (V).

= /D q V (2)

γ γ= −R 3 0 3 t( (NO ) (NO ) )E D (3)

A specific denitrification rate was calculated according to Eq. (4):

γ γ= −3 0 3 t( (NO ) (NO ) ) / ( * )S MLSS HRT (4)

where γ(NO3)0 and γ(NO3)t are the average nitrate concentrations (g/L) in the inflow and outflow during each experiment, and HRT is the hydraulic retention time (h).

The nitrate removal efficiency, R (%), was deter-mined as:

γ γ γ= −3 0 3 t 3 0100( (NO ) (NO ) ) / ( (NO )R (5)


Nitrate removal efficiency and the specific denitrification rate

Eight experiments were performed, each accord-ing to the selected flow rate. The experimental results are gathered in Table 1. The results represented herein were obtained when reaching the steady-state at the selected flow rate. The average nitrate and nitrite concentrations at the inflow (marked as “in”) and at the outflow (marked as “out”) are shown in Figure 1. It can be seen that a relatively high level of MBR efficiency was achieved at flow rates lower than 4.8 L/h as the outflow nitrate concentrations were lower than 7 mg/L. The nitrite concentrations within the inflow were below the permitted limit value (0.5 mg/L) throughout the entire experiment. The nitrite concentrations remained below 0.5 mg/L at flow rates lower than 4.8 L/h at the outflow, whilst at higher flow rates the measurements exceeded the threshold, and the denitrification process was less effective. The mic-roorganisms had insufficient time to complete the reaction due to the higher water-flow through the

Table 1. Experimental results depending on the inflow rates

Exp. No. q / L h-1 JP / L (m2 h)-1 HRT / h MLSS / g L-1 COD / mg L-1

1 0.6 0.6452 100.0 0.7731 0.51

2 1.2 1.2903 50.0 0.7693 0.77

3 1.8 1.9355 33.3 0.7983 0.79

4 2.4 2.5806 25.0 0.8312 0.96

5 3 3.2258 20.0 0.6031 1.07

6 4.8 5.1613 12.5 0.8090 1.12

7 7.2 7.7419 8.3 0.9327 3.10

8 10.2 10.9677 5.9 0.7668 7.02

Figure 1. Average nitrate (a) and nitrite (b) concentrations measured at the inflow and at the outflow, depending on the flow rate (for eight different experiments).


272

reactor and therefore lower hydraulic retention time. Consequently, the concentrations of nitrates and nit-rites increased at the outflow.

The highest removal efficiency, R (%), was achieved at the lowest inflow rates (Figure 2). When the flow rate varied within the range from 0.6 to 4.8 L/h, the removal efficiency was above 90%, whilst at the lowest inflow rate used the efficiency was even 97.67%. The hydraulic retention time (HRT) dec-reased with any further increase of the inflow and the removal efficiency consequently decreased. This is in accordance with the findings of previous studies, which confirmed that higher HRT values usually result in better removal performance and lower values indi-cate higher organic loading rates [22]. On the other hand, short HRT requires a high trans-membrane flux and could influence the quality of the treated effluent [15]. The experiments performed during this research indicated that denitrification efficieny (nitrate removal) was also affected by the oxygen and pH variations, as well as the inhibitive effect of accumulated nitirite at higher flow rates that could not be neglected even if the amounts were very low. Nevertheless, sufficient results were obtained at the highest inflow of 10.2 L/h (dilution rate 0.17 h-1, HRT 5.9 h), namely 68.08% of the nitrate was removed. Although the nitrate was incompletely removed, the efficiencies obtained from this study can be compared with the results of other studies using extractive denitrifying MBR [21]. The nitrate removal efficiency was above 99% with an ini-tial nitrate concentration of 200 mgNO3/L using methanol as substrate, and denitrification rates were obtained of up to 1.1 g/(m2 d). In another research performed by two-stage anoxic/oxic biofilm MBR with ethanol (C/N ratio = 1.4-2.5) and commercially avail-able Biocontact-N biocarriers (to enable immobil-isation), the nitrate conversions were also very high. In addition, no nitrite formation was observed during the process, whereas the influent nitrate concen-

tration was equal to 150 mg/L (HRT 2.5 h) [20]. Similarly, a varying degree of nitrate reduction, from 96% to complete removal, was found in the denit-rification of water with 60 mg/L NO3 in the inflow [7].

Significant impact on the specific denitrification rate was observed by the variations of flow rates (dil-ution rates). The maximum specific denitrification rate was acquired at the highest inflow rate of 10.2 L/h (Figure 2) and reached the value of 0.2738 g/L NO3/(g/L MLSS d). According to the calculated values and Eq. (4), the specific denitrification rate increases by decreasing the HRT. The higher the dilution rate, the higher the denitrification rate.

Similar results were achieved during the research performed on ground water, as performed by Butti-glieri [19], using ethanol as the carbon source (C/N ratio = 2.2, HRT 19-37 h), whereas the inflow nitrate concentration of 30 mg/L allowed maximum nitrate removal rates of between 0.36 and 0.48 gNO3/(gTSS d). It was reported that the bench scale MBR was able to offer nitrate removals of up to 98.5%, which is close to the values obtained herein, whilst the specific denit-rification rates achieved were lower, up to 0.02 d-1 [23]. There are several ways of explaining the differences in the above-mentioned results. The type and amount of external carbon source might signific-antly affect the microbial growth and its activity, which could be further reflected in lower nitrate removal and denitrification rates [10].

The concentrations of mixed-liquor suspended solids (MLSS) within the membrane bioreactor varied at between 0.6031 and 0.9327 g/L. This value was lower by half when compared to study of nitrate rem-oval from synthetic groundwater prepared by lake water and performed within same type of MBR, where anoxic sludge concentrations were determined of between 1.6 and 2 g/L [19]. Although the MLSS con-centration was low, the problem of fouling was obs-erved during the final phase of the MBR operation

Figure 2. Specific denitrification rate and removal efficiency depending on the flow rate.


273

and the transmembrane pressure increased. Since the denitrifying culture accumulates on the membrane surface where it forms filter cake, the membrane separation efficiency decays over time. One of the reasons that has been reported to cause membrane fouling in MBRs is decreasing HRT [22]. The mem-brane fouling could afterwards lead to increased energy consumption and operating costs [13]. There-fore, appropriate measures need to be ensured to prevent membrane fouling. It could be additionaly affected by sludge retention time and the character-istics of activated sludge [22]. Based on these facts, it can be concluded that the problem of fouling during our experiments could have been caused by dec-reased HRT, since the MLSS concentration was low and did not increase significantly.

The biological denitrification of drinking water performed by heterotrophic micro-organisms requires an aditional carbon source, although residual carbon sources in effluent can cause many problems during drinking water treatment [9]: fast growth of bacteria, formation of by-products within the treated water, and high COD values in the efluent [11]. In this experiment the values of COD measurements in the effluent (Table 1) were below 1 mg O2/L at lower inflow rates, while the maximum value of 7.02 mg O2/L, which exceeds the threshold value of 5 mg O2/L was achieved at the highest inflow rate. However, the concentrations at the inflow were around 95 mg O2/L due to the added carbon source, i.e., sucrose. Although sucrose in studies on denitrification has been rarely used as a carbon source, the presented study proved it to have quite good potential for nitrate removal effi-ciency.

Reactor removal capacity

In order to facilitate the evaluation of the expe-rimental results and to easily compare the tested

MBR system, it was necessary to calculate the reac-tor’s removal and the MBR system’s removal capa-cities. The reactor removal capacity (ER) of the pilot- -scale MBR was calculated in accordance with Eq. (3). A maximum reactor removal capacity of 8.75 g/(m3 h) was achieved at the highest dilution rate (Fig-ure 3). This could be explained by the fact that these two parameters are linearly correlated. Therefore, the higher the dilution rate, the higher the reactor removal capacity. The lowest capacity achieved by MBR was below 1 g/(m3.h). The study on experimental results demonstrated that the capacity not only depends on nitrate removal and the volume of the reactor, but also on the process conditions, such as flow rate, HRT, etc. However, according to Figure 3, it appears that the best performances of the reactor system can be obtained if the flow rate is higher than 4.8 L/h. But it must be considered that at higher flow rates denit-rification was incomplete due to the lower HRTs, and therefore the nitrate removal efficiencies were much lower. In addition, the concentration of the effluent COD at the highest flow rate was above the value limit. Therefore, operation at a flow higher than 4.8 L/h is unreasonable. The optimal reactor removal capacity (ER) that could be accepted regarding the above mentioned facts for these experiments was achieved at 5.29 g/(m3 h). The reactor removal capa-cities obtained herein are up to eight times higher due to the more than ten times higher inflow rate, com-pared with the literature [23]. Moreover, in this research the active surface area of the membrane and the working volume of the reactor, were consider-ably higher. Otherwise, the removal capacity during nitrate removal from the ground water using methanol as the carbon source and at the inflow a nitrate con-centration of 60 mg/L, was higher than that calculated in the presented study (29.2-70.8 gNO3/(m

3 h)) [7].

Figure 3. The reactor removal capacity and MBR system removal capacity depending on the flow rates.


274

Finally, the MBR system nitrate removal cap-acity (EMBR) was calculated according to Eq. (1). The MBR removal capacity is highly influenced by the type of filtration membrane and consequently by the active surface area of the membrane. The lowest value 0.04 g/(m2.h) was obtained at an inflow rate of 0.6 L/h, whilst the maximum value 0.56 g/(m2.h) was cal-culated at the highest inflow rate (dilution rate).

The capacities for both the reactor and MBR system increased linearly with the flow rate (see Fig-ure 3). However, by increasing the flow rate the cap-acity can increases but this could have a negative inf-luence on the nitrate removal efficiency, as shown above. Similarly as for the reactor removal capacity, it could be concluded that optimal MBR system cap-acitiy was achieved at a flow rate of 4.8 L/h at 0.34 g/(m2 h). It was found also that the ratio of MLSS and the initial nitrate concentration at the inflow, had had insignificant impact on the capacity and therefore with the lower mass ratio higher specific denitrification rates could be achieved [23]. The MBR system rem-oval capacities calculated during this experiment were close to the results obtained in the above-men-tioned paper, just slightly lower efficinces were deter-mined due to the lower permeate flux and the differ-ences in the specifications of the membrane module.

However, efficient denitrification in MBR using sucrose as the carbon source and at C/N ratio at 3:1 was visible at flow rates lower than 4.8 L/h (dilution rate 0.08 h-1). Nitrate removal efficiency exceeding 90% could be expected under these operational con-ditions. The drawback is that the removal capacities at low flow rates were low. In contrast, removal cap-acities were very high at above 4.8 L/h, but incomp-lete denitrification results in the production of nitrite and increase of COD. Therefore, the presented MBR system is beneficial for application within small water systems that operate at low flow rates.

CONCLUSIONS

The suitability of the membrane bioreactor for drinking water denitrification was investigated in this research. It was found that:

- Denitrification of drinking water in the pilot-scale MBR by using sucrose as the carbon source successfully removed nitrates from the drinking water.

- High nitrate removal efficiencies (more than 90%) were obtained at flow rates below 4.8 L/h.

- The denitrification efficiencies were highly dependent on the operational conditions, especially the flow rate, and therefore the hydraulic retention time.

- At lower HRTs, denitrification was incomplete and an accomulation of nitrite was observed.

Acknowledgements

The authors would like to acknowledge the Slo-venian Research Agency for the financial support (Project No. 1000-11-310131).

REFERENCES

[1] S. An, H. Stone, M. Nemati, J. Hazard. Mater. 190 (2011) 686-693

[2] E.J. McAdam, S.J. Judd, Desalination 196 (2006) 135-148

[3] ISO 8467:1993, International Organization of Standard-ization, Geneve, 1993

[4] ISO 7890-1:1986, International Organization of Standard-ization, Geneve, 1986

[5] ISO 6777:1984, International Organization of Standard-ization, Geneve, 1984

[6] APHA 1995, American Public Health Association, Wash-ington, DC, 1995

[7] E. Wąsik, J. Bohdziewicz, M. Błaszczyk, Process Bio-chem. 37 (2001 57-64

[8] K.A. Karanasios, I.A. Vasiliadou, S. Pavlou, D.V. Vayenas, J. Hazard. Mater. 180 (2010) 20-37

[9] I.A. Vasiliadou, S. Pavlou, D.V. Vayenas, Process Bio-chem. 41 (2006) 1401-1408

[10] M.A. Gómez, J. González-López, E. Hontoria-Garcı́ a, J. Hazard. Mater. 80 (2000) 69-80

[11] Y. Zhao, B. Zhang, C. Feng, F. Huang, P. Zhang, Z. Zhang, Y. Yang, N. Sugiura, Bioresour. Technol. 107 (2012) 159-165

[12] C. Glass, J. Silverstein, Water Res. 32 (1998) 831-839

[13] J.-Y. Tian, H. Liang, J. Nan, Y.-L. Yang, S.-J. You, G.-B. Li, Chem. Eng. J. 148 (2009) 296-305

[14] H. Sanaeepur, O. Hosseinkhani, A. Kargari, A. Ebadi Amooghin, A. Raisi, Desalination 289 (2012) 58-65

[15] X.-Y. Li, H.P. Chu, Water Res. 37 (2003) 4781-4791

[16] M.F.R. Zuthi, H.H. Ngo, W.S. Guo, Bioresour. Technol. 122 (2012) 119-129

[17] B.O. Mansell, E.D. Schroeder, Water Research 36 (2002) 4683-4690

[18] M. Fabbricino, L. Petta, Desalination 210 (2007) 163-174

[19] G. Buttiglieri, F. Malpei, E. Daverio, M. Melchiori, H. Nieman, J. Ligthart, Desalination 178 (2005) 211-218

[20] M. Ravnjak, J. Vrtovšek, A. Pintar, Bioresour. Technol. 128 (2013) 804-808

[21] S.J. Ergas, D.E. Rheinheimer, Water Res. 38 (2004) 3225-3232

[22] N. Fallah, B. Bonakdarpour, B. Nasernejad, M.R. Alavi Moghadam, J. Hazard. Mater. 178 (2010) 718-724

[23] A. Nuhoglu, T. Pekdemir, E. Yildiz, B. Keskinler, G. Akay, Water Res. 36 (2002) 1155-1166

[24] J. Oh, J. Silverstein, Water Res. 33 (1999) 1925-1937

[25] F. Meng, S.-R. Chae, A. Drews, M. Kraume, H.-S. Shin, F. Yang, Water Res. 43 (2009) 1489-1512.


275

ALEKSANDRA PETROVIČ

MARJANA SIMONIČ

Faculty of Chemistry and Chemical Engineering, University of Maribor,

Maribor, Slovenia

NAUČNI RAD

EFIKASNOST MEMBRANSKOG BIOREAKTORA ZA DENITRIFIKACIJU VODE ZA PIĆE

U radu je ispitan kapacitet membranskog bioreactora (MBR) za uklanjanje nitrata iz pijaće

vode. Testirane su mogućnosti poluindustrijskog postrojenja MBR u zavisnosti od ope-

rativnih parametara koristeći saharozu kao izvor ugljenika. Pijaća voda iz izvora je dove-

dena u reaktor da bi se analizirao uticaj protoka na efikasnost denitrifikacije i uklanjanja

nitrata iz pijaće vode. Sadržaj nitrata je bio oko70 mg/l, a odnos C/N je bio 3:1. Efikasnost

uklanjanja nitrata je oko 90% pri protoku manjem od 4,8 l/h. Specifična brzina denitrifikacije

je bila u opsegu od 0,02 do 0,16 g/l NO3/(g/L MLSS d). Efikasnost uklanjanja nitrata je

značajno zavisila od protoka i hidrauličkog vremena zadržavanja. Pri maksimalnom

protoku od 10,2 L/h moguće je uklanjanje do 68% nitrata, dok je ostvarena najveća

specifična brzina denitrifikacije od 0,2738 g/L NO3/(g/L) MLSS d). Izračunato je da je

maksimalni kapacitet uklanjanja nitrata 8,75g NO3/m3 h.

Ključne reči: kapacitet, denitrifikacija, pijaća voda, efikasnost, membranski bio-reaktor, saharoza.





277

DRAGUTIN M. NEDELJKOVIĆ1

MARIJA P. STEVANOVIĆ2

MIRKO Z. STIJEPOVIĆ3

ALEKSANDAR P. STAJČIĆ1

ALEKSANDAR S. GRUJIĆ1

JASNA T. STAJIĆ-TROŠIĆ1

JASMINA S. STEVANOVIĆ1 1University of Belgrade, Institute of

Chemistry, Technology and Metallurgy, Belgrade, Serbia

2TENT doo TE “Morava”, Svilajnac, Serbia

3Department of Chemical Engineering, Texas A&M

University at Qatar, Education City, Doha, Qatar

SCIENTIFIC PAPER

UDC 66:628.385:549.67:54-126

DOI 10.2298/CICEQ130924025N

THE POSSIBILITY OF APPLICATION OF ZEOLYTE POWDERS FOR THE CONSTRUCTION OF MEMBRANES FOR CARBON DIOXIDE SEPARATION

Article Highlights • Carbon dioxide separation from the waste gases • Application of mixed matrix membranes for gas separation • Membranes based on polymer matrix and inorganic zeolyte powder • Polymers with PEO groups and zeolytes with two-dimensional pores were tested Abstract

The aim of this study was to construct a polymer-based mixed matrix mem-brane that could be used for waste gases treatment. Therefore, high per-meability for the carbon dioxide and low permeability for other gases com-monly present in the industrial combustion waste gases (nitrogen, oxygen, hydrogen and methane) are essential. These membranes belong to the group of dense composite membranes, whose separation is based on the solution-diffusion mechanism. In this paper, feasibility of the application of poly(ethyl-ene oxide)-copoly(phtalamide) was tested. In order to enhance the per-meability of carbon dioxide, three different zeolites with two-dimensional pores (IHW, NSI and TER) were incorporated, and in order to improve compatibility between the inorganic particles and polymer chains, n-tetradecyldimethyl-amonium bromide (NTAB) was added. The best results in carbon dioxide/hyd-rogen selectivity were obtained with the membrane constructed with PEBAX 1657 and surface treated zeolites, while better results concerning selectivity were gained with membranes based on Polyactive.

Keywords: mixed matrix membrane, zeolite, carbon dioxide separation, polymer matrix, membrane selectivity.

Global warming has emerged as one of the most serious problems in chemical and environmental engineering in the recent decades, with carbon diox-ide being the main atmospheric pollutant. Carbon dioxide is emitted through various processes that inc-lude combustion (mining, power plants, transport, ind-ustrial facilities). As coal and petrol, the main sources of both energy and carbon dioxide, have no feasible alternative at the global scale, the main goal of the environmental analysis is to reduce the emission of carbon dioxide. Currently, the most common proce-

Correspondence: D.M. Nedeljković, Institute of Chemistry, Technology and Metallurgy (ICTM), University of Belgrade, Njegoševa 12, 11000 Belgrade, Serbia. E-mail: [email protected] Paper received: 24 September, 2013 Paper revised: 17 April, 2014 Paper accepted: 9 July, 2014

dures for CO2 removal include adsorption and cryo-genic processes [1-3]. The United Nations introduced a plan to gradually reduce CO2 emission in the follow-ing years (United Nations Framework Convention on Climate Change (UNFCCC, colloquially known as the Kyoto protocol) [4]. Carbon dioxide membranes based on the solution-diffusion mechanism have high potential for research and development, especially in the small-to-medium scale facilities with moderate requirements concerning the purity of the products [5]. Research in this field of membrane development has rapidly expanded in last three decades, with var-ious polymers developed as the main component of the membrane [6-10]. The suitable polymer should contain units that would enhance the solubility of the carbon dioxide, keeping at the same time low solu-bility of the other gasses, thus creating high perm-eation selectivity. Poly(ethylene oxide) (PEO) has

D.M.NEDELJKOVIĆ et al.: THE POSSIBILITY OF APPLICATION OF ZEOLYTE… Chem. Ind. Chem. Eng. Q. 21 (2) 277−284 (2015)

278

been proven to be potentially suitable for this applic-ation [11]. The negative property of pure PEO is its strong tendency to crystallize, which has a negative effect on the overall permeability of the membrane [12]. Instead of pure PEO, polymers that contain EO units can be used for this purpose. A polymer that is commercially available by Arkema (formerly Atotech) under the name PEBAX has the structure of poly-(amide-b-ether) and can be used as a good alter-native material for this purpose [13]. PEBAX belongs to the group of thermoplastic elastomers (Figure 1a). As a second choice, the polymer under commercial name Polyactive (supplied by IsoTis OrthoBiologics) was tested (Figure 1b).

PA stands for the polyamide hard block, and usually is nylon-6 or nylon-12, while the PE stands for the soft, amorphous polyether block (polyethylene oxide (PEO) or polytetramethylene oxide (PTMO)) [14].

As it can be seen from Figure 1b, Polyactive consists of polyethylene glycol (PEG) and polybutyl-ene terephthalate (PBT). The ratio PEG:PBT is 77:23 (wt.%) with PEG of molecular weight of 1500 g/mol.

The properties of the polymers (chemical, phy-sical and mechanical) can be easily modelled by the variation of the molar ratio of the blocks [15]. Accord-ing to the previously reported research, both Pebax and Polyactive turned out to be the promising mater-ials for acid gas treatment [16-19]. These polymers, when applied for the construction of the membrane, have also shown high selectivity of carbon dioxide versus both nitrogen and oxygen. The theoretical exp-lanation of the high selectivity is that ester and ether groups show strong affinity to the carbon dioxide sol-ution. The other reason for high selectivity versus nitrogen is the polarizability of the carbon dioxide (as well as the sulphur dioxide) in the presence of PEO

segments [17]. The possible solution for the increase of permeability and selectivity is the construction of a mixed matrix membrane that consists of a polymer matrix and inorganic powder. The matrix is usually made of the polymer that contains PE blocks, and the dispersed phase is inorganic particles [18-19]. The dispersed particles can be zeolytes, carbon molecular sieves or other nanoparticles. The presence of these fillers improves selectivity and permeability comparing to the membranes made of the pure polymer due to their inherent separation characteristics. However, addition of the charged inorganic particles in the poly-mer matrix can cause problems with dispersion, agglomeration und uneven distribution of the par-ticles. Fragility, as one of the main problems of the inorganic membrane is avoided due to the flexibility and elasticity of the polymer.

The first attempts of the mixed matrix mem-branes permeability improvement were reported 30 years ago, when the diffusion time lag of the carbon dioxide and methane was discovered [20]. Authors have observed that addition of the zeolite increases the time lag, but has apparently no effect on the steady-state permeation [21].

EXPERIMENTAL

The Pebax and Polyactive polymers were sup-plied by Arkema and IsoTis OrthoBiologics, respect-ively. The polymers were supplied in the form of pow-der and used as received. Three different zeolites with two-dimensional pores were used in this experi-ment. Their properties are compiled in Table 1. The average specific surface of the zeolite was 500 m²/g.

Solvents (ethanol and chloroform), zeolite and n-tetradecyltrimethylammonium bromide (NTAB) were

Figure 1. Structural formula of a) Pebax co-polymer; b) Polyactive co-polymer.


279

supplied by ABCR. All of the chemicals and used as received. The aim of the NTAB addition was to pro-vide good contact between electrically charged zeo-lyte particles, and highly hydrophobic polymer matrix. The electrically charged end and long, normal hydro-carbon chain should interact with the aforementioned components of the composite membrane, respect-ively.

Table 1. Properties of different types of zeolite used for the construction of the membrane; 2D channel system dimension

Framework type code Pore size, nm

IHW 0.37 X 0.35

NSI 0.26 X 0.24

TER 0.52 X 0.47

The first step in the membrane preparation pro-cedure was the solution of the polymer in the suitable solvent. For the Pebax membranes, the solvent was mixture ethanol/distilled water (70/30 mass ratio). The solvent for the Polyactive membranes was chloro-form. The Pebax was dissolved at the 80 °C under reflux, while Polyactive was dissolved at the room temperature, both of them for two hours. The zeolyte powder was at the same time dissolved in the same solvent as the polymer, and the additive was added (if the samples were made with additive). The homogen-ization of the zeolyte solution was done by ultrasound mixing with a titanium head. The duration of ultra-sound mixing was five minutes in order to avoid con-tamination of the solution by titanium nanoparticles detached from the head. The zeolyte solution and polymer solution were mixed and stirred overnight at the same temperature as the respective polymers. Overnight stirring was necessary in order to eliminate possible clusters and agglomerates of the zeolyte powder formed in the solution. This procedure resulted in viscous solution that was casted to the Teflon surface, covered with non-woven textile in order to protect sample from dust and any other unwanted particles and left overnight at room temperature and ambient pressure to dry. Teflon was used in order to avoid stitching of the membrane to the drying surface. If the drying process was too fast, it would result in the formation of bubbles and thus, bad permeation properties of the membrane. The viscosity of the solution had to be kept at an optimal value, which was determined empirically. If the viscosity is too low, the sedimentation velocity is too high, and the resulting membrane will have an uneven distribution of the par-ticles trough the volume. On the other hand, if the viscosity is too high, casting and drying processes are dominated by the surface tension, resulting in a mem-

brane with uneven thickness. After drying at room temperature, the material was placed on a high vacuum line in order to remove any traces of the resi-dual solvent.

The gas permeability measurements were car-ried out by the time lag method. The, diffusivity (D), permeability (P) and selectivity (α) were determined by the equations [22-24]:

−= =

Δ − +2 1

2 1

( )

( ( ) / 2)p p p

f p p

V l p pP DS

ART t p P p (1)

θ=

2

6

ld (2)

α = =/A A A

A BB B B

P D SP D S

(3)

where Vp stands for the constant permeate volume, l for the thickness of the membrane, A for the area of the membrane, R for the universal gas constant, Δt for the time that permeate pressure needs to increase from value pp1 to value pp2, pf for the feed pressure and D for the diffusion coefficient. In Eq. (2), d stands for the intercept on the time axis when pressure on the permeate side is presented versus time. This curve has a parabolic shape which turns into a straight line once the steady state is obtained. Extra-polation of the steady-state line to the x-axis gives the parameter d. The solubility can be calculated as the ratio between permeability and diffusivity. The sol-ution-diffusion model was used for the analysis of the gas transport properties of the membranes [25]. The selectivity of the membrane for the gas A versus gas B was defined as the ratio of their permeabilities.

Before the permeability measurements, the membrane was kept at high vacuum for 30 min in order to remove any traces of humidity that could have penetrated at the ambient conditions. After the drying, the gas that was measured was applied at one side of the membrane. The other side of the mem-brane was evacuated, causing the pressure gradient as the driving force for the diffusion and gas per-meation. The pressure as the function of time was measured at the low pressure side of the membrane, and the permeation properties were calculated by equations (1)-(3). Due to security measures in order to avoid mixing of the flammable gases, the order of gases for the measurement was: hydrogen, nitrogen, oxygen, carbon dioxide. Between measurements of different gases, the membrane and the whole equip-ment was kept under high vacuum for 15 min [26].


280


Four different series of the membranes were constructed, two with each polymer. Two series were made of pure polymer and zeolyte, without the addi-tion of the filler. In the second series, NTAB was added to the polymer and zeolyte. The weight ratios were calculated as the mass ratio versus overall mass of the membrane. The compositions of the membranes are given in Table 2.

Table 2. The composition and appearance of the membranes made without the additive

Series Polymer Additive

I Pebax -

II Pebax NTAB

III Polyactive -

IV Polyactive NTAB

Initially, the appearance of the membranes was assessed visually. If the membrane is made properly, it should be smooth, transparent or pale, flat without pinholes or visible damages. If the membrane is not transparent, that indicates that the light transmitted at the polymer-zeolyte surface and the contact between them is bad. Rough surface indicates uneven dis-tribution of the particles trough the volume of the membrane, while the self-rolling of the membrane indicates the sedimentation of the particles, and thus uneven distribution of the zeolyte.

The composition and evaluation of the mem-branes from the series I and series II are given in Table 3.

Table 3. The composition and appearance of the membranes of Series I and II

Membrane No. Zeolite filler Filler, % Additive, % Appearance

I-1 IHW 22 - White

I-2 IHW 22 - White

I-3 NSI 22 - Transparent

I-4 NSI 22 - White

I-5 TER 22 - White areas

I-6 TER 22 - Transparent

II-1 - - 3.3 Transparent

II-2 IHW 22 3.3 White

II-3 IHW 22 3.3 White

II-4 NSI 22 3.3 Transparent

II-5 NSI 22 3.3 Transparent

II-6 TER 22.5 2.2 Transparent

II-7 TER 23 1.1 Transparent

As it is obvious from Table 3, both NSI and TER types of zeolyte could be used for the construction of the membrane. Although transparent membranes were not obtained in all of the samples, their con-struction with NTAB was attempted. The explanation of the white spots in the membranes constructed with IHW can be that the zeolyte particles agglomerate. This agglomeration comes as the consequence of strong electrostatic forces between the zeolyte par-ticles, which are stronger than the viscosity of the polymer solution. Areas of the different colour indicate a non-stationary drying process that causes rapid local variations in viscosity of the solution, and there-fore, the agglomeration was allowed is some areas of the membrane. To check the possible agglomeration and distribution of the zeolytes in the polymer, an SEM image of the sample I-6 was obtained (Figure 2). In general, good distribution of the zeolyte particles is present in this system. Agglomeration is still visible as the white cluster in the upper right part of the figure.

Figure 2. SEM image of the sample I-6.

The compatibility of the polymer and filler was tested by the construction of the membrane that was solely made of polymer and additive. A transparent membrane was obtained, and therefore this filler was taken as the compatibilization additive.

Analyzing the appearance of the membranes from the series II, it is obvious (Table 3) that both NSI and TER have shown good compatibility with Pebax in the presence of NTAB as the additive. Therefore, these two zeolytes were used for the construction and measurement of the permeation properties of gases. The IHW zeolyte could not be used for this purpose due to the bad polymer-zeolyte contact (sample II-2) and agglomeration of the zeolyte particles (sample II-3). An SEM image was obtained (Figure 3) in order to check the microstructure of the membrane. Com-parison of the results for the series I and II shows sig-nificant improvement in the zeolyte-polymer contact.


281

Figure 3. SEM image of the sample II-6.

Comparing Figures 2 and 3, it can be observed that the latter shows improved distribution of the zeolyte. Both of the samples show good particle dis-tribution, but in the case of sample II-6, agglomeration is nearly avoided. Only partial formation of the cluster is visible (bottom part of the Figure 3 in the middle). However, it can bee seen that the polymer is present between the particles of the cluster.

The membranes based on the Polyactive based polymer (series III and IV) were constructed in the manner analogous to the Pebax based membranes (series I and II). The main difference in the procedure is that chloroform was used rather than ethanol/water mixture for Polyactive and Pebax polymers, respect-ively. Due to the low boiling point of the chloroform, the removal of residual solvent from the membrane was easier. Tetrahydrofuran (THF) can be used for this purpose as an alternative. The amounts of poly-mers, zeolytes and additives were analogous as for the membranes of the series I and II. The data of

composition and appearance of the membranes are presented in Table 4.

The sample II-1 was constructed solely from the Polyactive polymer. The purpose of this sample was to compare its properties with the properties specified by the supplier. The measured results slightly differ from the specification. This difference can be attri-buted to the eventual error in the measurement, or to the variations in the different polymer batches. Simi-larly to the Pebax based membranes, the IHW filler is not compatible in the series without additive (samples III-2 and III-3) and with additive (samples IV-2 and IV-3). As in the case of the Pebax based membranes, both NSI and TER have been proved as good fillers, resulting in smooth, transparent membranes without visible pinholes or other damage. However, com-paring the appearance of the membranes based on Pebax and Polyactive containing NSI and TER, it might be seen that results obtained with the Poly-active are not as good as the results obtained with Pebax. Although the IHW cannot be used for the con-struction of the membrane with any of tested poly-mers, different behaviour with two different polymers was observed. While the samples with IHW from series I and II (prepared with Pebax) are white, the samples from series III and IV contain white spots. This behaviour indicates different types of the behaviour of the zeolyte particles in the presence of different polymers, and thus, difference in the struc-ture of the samples made with the same powder, but different polymer. The samples with white spots show the agglomeration of zeolyte particles, and white membrane or white areas on the membrane show bad contact between the zeolyte particles and poly-mer chains.

Table 4. The composition and appearance of the membranes of Series III and Series IV

Membrane number Porous filler Filler, % Additive, % Appearance

III-1 - - - Transparent

III-2 IHW 22 - White spots

III-3 IHW 22 - White spots

III-4 NSI 22 - Transparent

III-5 NSI 22 - White areas

III-6 TER 22 - Transparent

III-7 TER 22 - White spots

IV-1 - - 3.3 Transparent

IV-2 IHW 22 3.3 White spots

IV-3 IHW 22 3.3 White areas

IV-4 NSI 22 3.3 White spots

IV-5 NSI 22 3.3 Transparent

IV-6 TER 22 3.3 Transparent

IV-7 TER 22 3.3 White


282

The sample IV-1 was made, in the analogue manner as the sample II-1, in order to check the possibility of the application of the NTAB in the Poly-active matrix. As the experiment resulted in a trans-parent membrane without visible spots, it was con-cluded that NTAB could be used as an additive for the membranes made with Polyactive. Similarly to the Pebax membranes, both NSI and TER have shown good results, and again, better results were gained in the presence of NTAB. The amount of filler was determined as the optimal value of 22 wt.% versus overall mass of the membrane. If the concentration of the filler is lower, the permeation properties of the membrane are negatively affected. If the concentra-tion is higher, the agglomeration is enhanced due to the electrostatic forces between the particles of the zeolyte powder. The amount of the additive is deter-mined in the similar manner. If the concentration of the additive is too low, not all of the particles of the zeolyte could be covered. On the other hand, if the concentration of the additive is too high, the zeolyte particles precipitate, and that precipitation could be attributed to the electrostatic forces. The SEM image

of sample IV-6 is shown in Figure 4. Comparing this sample with the previous, it can be seen that the polymer completely surrounds the zeolyte particles and that no agglomeration is present.

For the permeability measurements, all of the transparent samples without visible damage were used, regardless of whether they were made with or without additive. Prior to the measurement of the membranes, the permeability of samples II-1 and IV-1 was measured in order to compare permeability of the pure polymer and permeability of the polymer with additive. The obtained results clearly indicate that dispersion of the NTAB does not influence the per-meability of the pure polymer. The results of perme-ability and selectivity measurements are given in Table 5.

It should be noted that the usual unit for the gas permeability of the membrane in the membrane research community is a Barrer. One Barrer is the permeability of 1 cm3 of a gas under the standard pressure and temperature conditions, trough the 1 cm2 of the area and 1 cm of the thickness driven by the pressure gradient of 1 cmHg in 1 s divided by the fac-tor of 10-10.

Analyzing the permeability data presented in Table 5, it is obvious that all of the membranes that appeared transparent showed good and comparable permeability and diffusivity properties. Comparison of the obtained results indicates that permeability for carbon dioxide is slightly better in cases of mem-branes constructed with Pebax (series I and II), but the selectivity is better in the case of Polyactive (series III and IV). Therefore, the choice of polymer for the construction of the membrane is determined by the requirements concerning the capacity of equip-ment (higher flux requires higher permeability, and therefore Pebax is the preferred polymer) and purity of the products (higher requirements for purity need

Figure 4. SEM image of the sample IV-6.

Table 5. The results of the permeability measurement of the membranes

Membrane No. Thickness, μm P (CO2), Barrer α (CO2/H2) α (CO2/O2) α (CO2/N2)

I-3 187 70 8.1 20 52

I-6 152 110 8.7 19 48

II-3 134 112 8.4 17 55

II-4 205 120 9.1 18.6 60

II-5 268 95 8.8 19 59

II-6 254 105 9.6 21 57

III-4 189 92 8.2 19.5 61

III-6 165 89 8.5 21 63

III-7 176 97 9.9 20 56

IV-5 171 91 10.8 21.5 60.4

IV-6 231 99 9.9 19.7 58.4


283

higher selectivity, and Polyactive is in that case the better choice).

CONCLUSIONS

In this paper, the possibility of the construction of a mixed matrix membrane based on a polymer matrix and surface treated inorganic powder was examined. Two different types of polymers, three dif-ferent types of two-dimensional zeolytes and one additive were used for the construction of the mem-brane. Preliminary optical testing showed that not all of the combinations are suitable for the construction of the membrane. In the case of Pebax-based mem-branes, NSI and TER have shown good compatibility between the inorganic powder and polymer chains. Good contact and distribution of the particles could not be provided in the case of the IHW zeolyte. The NTAB has improved the compatibility between the inorganic zeolyte particles and polymer chains without affecting the permeability properties of the membrane. Therefore, it is reasonable to conclude that NTAB is a good foundation for future research in the field of gas separation membranes. The main challenge in future research would therefore be the downsizing of the membrane thickness to values lower than 100 µm. The other goal of the research is to test the possibility of application, permeability and selectivity in wet conditions.

Acknowledgement

The authors would like to acknowledge the financial support of the Ministry of Education, Science and Technological Development of the Republic of Serbia, through research projects TR 34011 and III 45019.

REFERENCES

[1] A.B. Rao, E.S. Rubin, Environ. Sci. Technol. 36 (2002) 4467–4475

[2] U. Desideri, R. Corbelli, Energy Convers. Manage. 39 (1998) 857–867

[3] A. Meisen, S. Xiaoshan, Energy Convers. Manage. 38 (1997) 37–42

[4] The United Nations Framework Convention on Climate Change, Kyoto, 1997

[5] W.J. Koros, G.K. Fleming, J. Membr. Sci. 83 (1993) 1–80

[6] S.P. Nunes, K.-V. Peinemann, Wiley-VCH Verlag GmbH, Berlin, 2006, p. 53

[7] K. Ghosal, B.D. Freeman, Polym. Adv. Technol. 5 (1994) 673–697

[8] R.W. Baker, McGraw-Hill, New York, 2000, p. 301

[9] R.W. Baker, Ind. Eng. Chem. Res. 41 (2002) 1393–1411

[10] S. Matteucci,Y. Yampolskii, BD. Freeman, I. Pinnau, John Wiley & Sons, Chichester, 2006, p. 1

[11] H. Lin, B.D. Freeman, J. Mol. Struct. 739 (2005) 57–74

[12] H. Lin, B.D. Freeman, J. Membr. Sci. 239 (2004) 105–117

[13] L.A. Utracki, Polym. Eng. Sci. 35 (1995) 2–17

[14] G. Deleens, N.R. Legge, G. Holder, H.E. Schroeder, Hanser Publishers, New York, 1987, p. 215

[15] M. Yoshino, K. Ito, H. Kita, K.-I. Okamoto, J. Polym. Sci., B: Polym. Phys. 38 (2000) 1707–1715

[16] V. Bondar, B.D. Freeman, I. Pinnau, J. Polym. Sci., B: Polym. Phys. 37 (1999) 2463–2475

[17] J.H. Kim, S.Y. Ha, Y.M. Lee, J. Membr. Sci. 190 (2001) 179–193

[18] A. Car, C. Stropnik, W. Yave, K.-V. Peinemann, J. Membr. Sci. 307 (2008) 88-95

[19] A. Car, C. Stropnik, W. Yave, K.-V. Peinemann, Adv. Funct. Mater. 18 (2008) 2815–2823

[20] D.R. Paul and D.R. Kemp, J. Polym. Sci. Polym. Phys. 41 (1973) 79–93

[21] S. Kulprathipanja, R.W. Neuzil, N.N. Li, US patent 4740219, 1988

[22] A.M. Shishatskii, Y.P. Yampolski, K.-V. Peinemann, J. Membr. Sci. 112 (1996) 275–285

[23] J. Qiu, J.-M. Zheng, K.-V. Peinemann, Macromolecules 40 (2007) 3213–3222

[24] J. Crank, The Mathematics of Diffusion, 2nd ed., Oxford University Press, Oxford, 1980, second edition

[25] G. Wijmans, R.W. Baker, J. Membr. Sci. 107 (1995) 1–21

[26] D. Nedeljković, A. Stajčić, A. Grujić, J. Stajić-Trošić, M. Zrilić, J. Stevanović, S. Drmanić, Dig. J. Nanomat. Biost. 7 (2012) 269-782.


284

DRAGUTIN M.NEDELJKOVIĆ1

MARIJA P. STEVANOVIĆ2

MIRKO Z. STIJEPOVIĆ3

ALEKSANDAR P. STAJČIĆ1

ALEKSANDAR S. GRUJIĆ1

JASNA T. STAJIĆ-TROŠIĆ1

JASMINA S. STEVANOVIĆ1

1Institut za hemiju, tehnologiju i metalurgiju (IHTM) – Centar za mikroelektronske tehnologije,

Univerzitet u Beogradu, Njegoševa 12, 11000 Beograd, Srbija

2TENT doo TE “Morava”, Svilajnac, Srbija

3Department of Chemical Engineering, Texas A&M University at Qatar,

Education City, Doha, Qatar

NAUČNI RAD

MOGUĆNOST PRIMENE ZEOLITNIH PRAHOVA ZA KONSTRUKCIJU MEMBRANA ZA IZDVAJANJE UGLJEN-DIOKSIDA

Cilj ovog rada je bio da se konstruiše neporozna kompozitna membrana bazirana na

polimernom matriksu koja može da bude upotrebljena za tretman otpadnih gasova. Za ovu

svrhu, neophodno je da membrana ima visoku permeabilnost za ugljen-dioksid i nisku

permeabilnost za druge gasove koji ne najčešće sreću u produktima sagorevanja (azot,

kiseonik, vodonik, metan). Ove membrane pripadaju grupi neporoznih membrana i meha-

nizam separacije gasova je baziran na rastvorljivosti i difuziji. U ovom radu, testirana je

mogućnost primene poli(etilenoksida)-kopoli(ftalamida). Da bi se povećala permeabilnost

ugljen-dioksida, dodavana su tri različita zeolitna praha, a da bi se poboljšala kompa-

tibilnost neorganskih čestica i polimernih lanaca, dodat je n-tetradeciltrimetilamonijum-

-bromid (NTAB). Ispitivani zeoliti pripadaju grupi sa dvodimenzionalnim porama (IHW, NSI

i TER). Najbolji rezultati u separaciji ugljen-dioksida i voidonika su postignuti kod mem-

brana baziranim na polimeru PEBAX 1657 i zeolitima uz dodatak aditiva. Sa druge strane,

bolja selektivnost je postignuta kod membrana baziranih na Polyactive polimeru.

Ključne reči: neporozne membrane, zeolitni prahovi, separacija ugljen-dioksida, polimerni matriks, selektivnost membrane.





285

VESELINKA GRUDIĆ JELENA ŠĆEPANOVIĆ

IVANA BOŠKOVIĆ

Faculty of Metallurgy and Technology, University of

Montenegro, Podgorica, Montenegro

SCIENTIFIC PAPER

UDC 66.081:546.48-128:663.26

DOI 10.2298/CICEQ140418027G

REMOVAL OF CADMIUM (II) FROM AQUEOUS SOLUTION USING FERMENTED GRAPE MARC AS A NEW ADSORBENT

Article Highlights • The sorption process is described by the Langmuir isotherm • The grape marc represents an alternative to expensive sorbents • FTIR and EDS analysis show that the sorption of Cd(II) is a complex process Abstract

Biosorption of cadmium from aqueous solutions using fermented grape marc, as well as influence of the most important factors, such as: contact time, gra-nulation of biosorbent and the initial concentration of metal ions, is investigated in this paper. The equilibrium sorption of cadmium ions is achieved after 15-20 min, depending on the initial concentration of metal ions. Such a short time needed to achieve the equilibrium indicates that mass sorption is the dominant process. Langmuir, Freundlich and Dubinin-Radushkevich isotherms were used to describe the equilibrium sorption process, and the Langmuir model was found to be the most convenient. Maximum of sorption capacity is 20 mg g-1. EDS spectrum analysis showed that the process of ion exchange is one of the main sorption mechanisms. Minor changes observed in the FTIR spectrum of grape marc after the sorption of Cd(II) ions indicate the formation a bond between metal ions and partially ionized carboxyl and phenol groups from the biomass. The results of this study confirmed that fermented grape marc, due to its porous structure and characteristic chemical composition, is a good sorption material.

Keywords: biosorption, grape marc, isotherm, cadmium ions.

The presence of heavy metal ions in environ-ment is a growing issue. Therefore, the treatment of wastewaters and the removal of heavy metals are very important from the aspect of protection of human health and of environmental protection. Among heavy metals, cadmium is one of the most toxic, carcino-genic and bioaccumulative [1].

Different methods have been used for removing heavy metals from wastewater: ion exchange, che-mical precipitation, ultrafiltration, reverse osmosis, electrochemical processes, etc. [2,3]. The main dis-advantages of these methods are low selectivity, high costs and other technical constraints, and it is, there-fore, necessary to find an economical alternative. Correspondence: V. Grudić, Faculty of Metallurgy and Tech-nology, University of Montenegro, Džordža Vašingtona bb, 81000 Podgorica, Montenegro. E-mail: [email protected] Paper received: 18 April, 2014 Paper revised: 23 June, 2014 Paper accepted: 9 July, 2014

Biosorption of heavy metal ions is a new method that proved to be very effective for their removal from aqueous solutions [4]. The main advantages of this method are: high selectivity, low cost (based on availability and price of natural biosorbents) and high efficiency at low concentrations of heavy metals [5].

Many biological materials are used as bio-sorbents for the removal of heavy metal ions: pome-granate peel [6], the bark of walnut, hazelnut and almond [7], the remains of sugarcane [8], olive leaf [9], lemon zest, bananas and oranges [10], etc.

Grape marc is a biological cellulosic material with a multilayer structure whose layers are con-nected systems of channels and pores that allow relatively high specific surface area necessary for sorption. Metal ions can easily penetrate through the pores and channels where they adsorb on the surface of a number of internal active centers. The chemical composition of grape marc, which prevails phenolic compounds, is of great importance for the sorption of

V. GRUDIĆ et al.: REMOVAL OF CADMIUM (II) FROM AQUEOUS SOLUTION… Chem. Ind. Chem. Eng. Q. 21 (2) 285−293 (2015)

286

heavy metals. The goal of this research was to inves-tigate specific biosorbent properties of grape marc remains from wine production of the “Plantaže 13. Jul” company in Podgorica.

EXPERIMENTAL

Sorbent

The biosorbent was used in its original form or in the form of its various modifications in order to achieve better sorption effectiveness. The original form was modified by physical, chemical and thermal treatments. The grape marc remains from the wine production was subjected to leak through the press under a pressure of 2×105 Pa. The resulting grape marc was washed with distilled water, then rinsed with dilute hydrochloric acid, and again washed with distilled water to remove Cl–. Biosorbent was then dried at ambient temperature and at 60 °C to constant weight. The granulation was performed on a steel laboratory blender (ITNMS, Laboratory for preparation of raw materials, Belgrade) and the obtained bio-sorbent was sifted through a standard steel sieve to obtain fractions of appropriate particle size: <0.5 mm, 0.5-1 mm, 1-2 mm, >2 mm. Before doing any exam-ination, biomass was again treated with dilute HCl, rinsed with distilled water to remove Cl- ions and dried at 60 °C to constant mass.

Materials

All chemicals used in this experiment were p.a. grade. The initial concentration of cadmium ions was 1000 mg dm-3 obtained by dissolving the appropriate amount of Cd(NO3)2·4H2O in distilled water. The tests were performed in the concentration range of cad-mium ions from 20-400 mg dm-3, and solutions of specified concentrations were obtained by successive dilution of the initial solution.

Analysis

Method of flame atomic absorption spectro-photometry was performed using AA-6800 instrument (Shimadzu, Japan) at a wavelength of 228.8 nm. The pH value was measured using a pH-meter (WTW inoLab pH 720). Fourier transform infrared spectro-scopy (FTIR) was used to identify different chemical functional groups present in the grape marc in the wavelength range 4000-400 cm-1.

Sorption experiments

The sorption experiments were performed using a mixture of 0.1 g of grape marc and 100 cm3 of sol-ution that contains various concentrations of cadmium ions. In order to find the optimal conditions for

sorption, pH values (3-6), contact time (3-120 min), the concentration of cadmium ions (20-400 mg dm-3) and sorbent particle size were varied. The mixture was shaken in a Heidolph reciprocal shaker 130 rpm. The filtration and stabilization of filtrate using 1 cm3 of concentrated HNO3 were done before spectrophoto-metric analysis.

The amount of metal ion adsorbed per unit mass of the biosorbent was calculated as [11]:

−= o ec cq S

V (1)

where c0 is the initial metal ion concentration (mg dm-3), ce is the equilibrium metal ion concentration (mg dm-3), V is the volume of metal ion solution (dm3) and S is the mass of biosorbent (g).

Percent removal (% R) of metal ions was cal-culated from the following equation [12,13]:

−=% 100o e

o

c cR

c (2)


The influence of initial pH

Increase in the percentage removal of cadmium ions with increasing pH (Table 1) indicates that the investigated sorption process largely performed through the ionic changes [13,14]. The efficiency of removal of cadmium ions increase with decreasing grain size of grape husk for constant pH value. At pH 3, 44% of total Cd in solution was removed by fine grain sorbent while only 40% was removed by larger grain of biosorbent. Percentages of removed Cd ions at pH 5 for mentioned grain size are 97.0 and 86.6%, respectively. At pH > 5, percentage of removal Cd ions is slightly decreased, because Cd(II) ions are transformed to Cd(OH)+ or Cd(OH)2. For these rea-sons, the concentration of Cd (II) ions in the solution and thus amount of adsorbed metal are reduced.

Table 1. Percentage removal of cadmium ions at different initial pH value

Initial pH value Grain size

Fine Large

3 44 40

4 85 75

5 97 86.6

6 96 85.6

The best results obtained at pH 5 can be explained by decreasing quantity of adsorbed H+ ions which reject Cd(II) ions with increasing pH solution value. The influence of pH on the adsorption of metal


287

ions is significantly related to the presence of different functional groups (phenol, aldehyde, carboxylic) on the biosorbent surface [15]. These functional groups have a great affinity for metal ions, which are bounded to the surface in complex ion [16,18-20].

The influence of biosorbent average grain size on removal Cd (II) ions

The biosorbent was fragmentized using a steel mill, and after screening four fractions (0.5 mm, 0.5-1 mm, 1-2 mm, >2 mm) were obtained. It is evident that the absorption of Cd(II) ions is enhanced with dec-reasing biosorbent average grain size (Table 2) due to contact surface enlargement [21,22]. The best results were obtained for granulation of 0.5 mm. How-ever, the fraction with the largest grain size (unground grape marc) achieved the removal percentage of 86.6%, which is significant in terms of future appli-cations from the perspective of technology simpli-fication and cost reduction.

Table 2. Percentage removal of cadmium ions for different bio-sorbent average grain size

Biosorbent average grain size, mm

0.5 1.0 2.0 2.5

Percentage removal of cadmium ions, %

97.0 96.6 96.0 86.6

Influence of initial concentration on the equilibrium of sorption processes

Investigation of adsorption at different initial sorbate concentrations (20-400 mg dm-3) enabled construction of sorption isotherms, based on which the maximum sorption capacity of grape marc was determined. Dependence of the sorption quantity on equilibrium concentration of metal ions in solution is shown in Figure 1.

Figure 1. Adsorption isotherm of Cd(II) on grape marc at

25.0±0.5 °C and pH 5.0.

The obtained isotherm belongs to L-type iso-therms, and is characterized by a much higher slope

in the range of lower metal concentrations, while at high concentrations the dependence slope decreases and tends to plateau. The maximum quantity of sor-bed metal was about 20 mg g-1. Three different iso-therms were used in order to determine which one best matched the experimental results.

The Langmuir isotherm assumes monolayer adsorption on a uniform surface with a conclusive number of adsorption sites [22,23]. The linear form of the Langmuir isotherm model is described as:

= +1e e

e m L m

c cq q K q

(3)

where KL is Langmuir equilibrium constant, related to the energy of sorption (dm3 mg-1) and qm (mg g-1) is the maximum sorption capacity of the sorbent in a complete monolayer of sorbate (mg g-1). The values of Langmuir parameters qmax and KL were calculated from the slope and intercept of the linear plot of ce/qe vs. ce as shown in Figure 2a. The values of qmax, KL and regression coefficient R2 are presented in Table 3.

Freundlich isotherm model applies to adsorption on heterogeneous surfaces with the interaction between adsorbed molecules. The linear form of the Freundlich isotherm model is described as:

= + 1ln ln lne F eq K c

n (4)

where KF is the Freundlich constant related to the bonding energy and 1/n is the heterogeneity factor and n (g dm-3) is a measure of the deviation from linearity of adsorption.

Freundlich equilibrium constants were deter-mined from the slope and intercept of the linear plot of ln qevs ln Ce as shown in Figure 2b. The values of the constants of the Freundlich isotherm are given in Table 3.

The obtained values of R2 show that the Lang-muir isotherm (with R2 > 0.99) has the best agreement with experimental results. Additionaly, the experi-mentally determined values of the sorption capacity of grape marc were close to its theoretical maximum. Such behavior was found for the sorption of heavy metals on the similar biosorbent [22,23]. The most important role in the sorption is assigned to oxygen functional groups, primarily the carbonyl and phenol, from the grape marc. In fact, oxygen electrons pair enables the sorption of positively charged Cd(II) ions. The sorption capacity values of grape marc are close to sorption capacity of other biosorbents (Table 4).

The essential characteristics of the Langmuir isotherm parameters can be used to predict the affi-nity between the sorbate and sorbent using separ-

V. GRUDIĆ et al.: REMOVAL OF CADMIUM (II) FROM AQUEOUS SOLUTION… CI&CEQ 21 (0) 000−000 (2015)

288

Figure 2. Isotherms for sorption of Cd(II) on grape marc: a) Langmuir isotherm; b) Freundlich isotherm; c) D-R isotherm at 25.0±0.5 °C and pH 5.0.

Table 3. Langmuir and Freundlich isotherm constants for bio-sorption of Cd(II) ions on grape marc

Langmuir constant Freundlich constant

qm / mg g-1 KL / dm3 mg-1 R2 KF n R2

22.7 0.028 0.999 1.179 1.776 0.957

Table 4. The sorption capacities of different sorbents for Cd (II)

Biosorbent qm / mg g-1 Reference

Waste cork 2.4 25

Loquat leavs 48.78 26

Red alga 53.1 13

Ceratonia siliqua bark 14.27 27

Pomelo peel 21.83 28

Untreated coffee grou 15.65 29

Rice hus 8.58 30

Maize leaf 10.18 31

Lagenaria vulgaris 11.25 32

Grape marc 20 This work

ation factor or dimensionless equilibrium parameter, RL expressed as in the following equation [32]:

=+

1

1L

L oR

K c (5)

The value of RL indicated the type of Langmuir isotherm can be irreversible (RL = 0), favorable (0 < RL < 1), linear (RL = 1) or unfavorable (RL > 1).

The RL was found to be 0.082-0.641 for con-centration of 20-400 mg dm-3 of Cd(II) ions. They are in the range of 0-1, which indicates favorable bio-sorption. In support of this conclusion is the fact that the experimentally obtained isotherm is convex, which is characteristic of favored sorption.

The Dubinin-Radushkevich (D-R) model, which does not assume a homogenous surface or a cons-tant biosorption potential as the Langmuir model, was also used to test the experimental data in order to determine the nature of the sorption process [33]. The Dubinin-Radushkevich isotherm is given by the equation:

βε= − 2ln lne mq q (6)


289

where β is the coefficient associated with the free energy biosorption (mol2 J-2) and ε is Polanyi’s poten-tial, given by the equation:

ε = +

11

oRT

c (7)

The D-R isotherm parameters β and qm were obtained from the slope and intercept of the plot of ln qe vs. ε2 (Figure 2c). The free energy of biosorption, E, was calculated from the β value using the following equation:

β= 1

2E (8)

For Cd (II) ions biosorption on grape marc, the energy biosorption value was of 10.7 kJ mol-1, indi-cating a chemisorption process and that one of the mechanisms of biosorption process is ion exchange [34].

The influence of contact time on removal of Cd (II) ions

The influence of contact time for different concentration sorption capacity was investigated. The results are shown in Figure 3.

Figure 3. The effect of contact time on Cd(II) sorption on grape marc. The initial solution concentrations of Cd (II) ions (mg g-1)

were: 20 (♦), 50 (■): 100 (▲), 200 (X) 400 (○).

According to the results, the investigated sorp-tion process consists of two phases. The first one is characterized by fast sorption and represents a con-sequence of significant difference in concentration of metals in solution and on the sorbent surface as well as of large number of available active sites. The second phase is slower and occurs until the equilib-rium state. The time to achieve the equilibrium was 15 min for lower concentrations of 20 and 50 mg dm-3 and 20 min for higher concentrations of 100-400 mg dm-3. The short time needed to achieve the equilib-rium indicates that the dominant process is mass sorption.

The percentage of removal of cadmium ions largely depends on the concentration. With increasing the initial concentration, the percentage of removal of Cd(II) ions decreased from 85% at 20 mg dm-3 to 50% at a 400 mg dm-3. This behavior is due to limited num-ber of active sites on the surface for sorption process [35].

EDS Analysis

EDS spectrum analysis before the sorption process (Figure 4a) indicated that the carbon and oxygen are two main elements in the composition of biosorbent surface along with the other elements present in trace amounts. Registered peaks of alkali (K) and alkaline earth metals (Ca) are derived from the raw biosorbent which contains these metals. The minor amounts of copper are probably due to fertil-ization and after purification this metal is almost com-pletely removed from biomass.

EDS spectrum analysis after the sorption process (Figure 4b) shows that the K peaks are almost absent, indicating that these ions were sub-stituted by Cd(II) ions during the process of ion exchange, which is one of the main sorption mech-anisms.

FTIR Analysis

The FTIR analysis enables identification of the type of functional groups on the surface of the sample that could participate in the binding of metal ions, as well as their features before and after the sorption process [19,36]. The FTIR spectra were obtained using Omic software and presented in Figures 5 and 6.

The broad intense band at 3324.4 cm-1 (which typically occurs in the range of 3200 to 3600 cm-1) belongs to valence vibration of O–H group. The not-iceable peak at 2922 cm-1 indicates the symmetric or asymmetric C–H valence vibration of aliphatic acids, while the peak registered at 1743.5 cm-1 further indi-cates the valence vibration of –C=O connection that comes from the unionized carboxyl group and may correspond to protonated form or the corresponding ester (–COOH, –COOCH3) [37].

Also, the band at 1032.6 cm-1 can be connected with the existence of valence vibration of –C–O of alco-hols or carboxylic acids [38]. FTIR spectrum of grape marc clearly indicates the presence of carbonyl and hydroxyl groups, which are the most important sites for the binding of cadmium ions.

The obtained results show that there are no significant changes in the spectra before and after the sorption process. The slightly decrease in intensity and shift of the peak at 3324.4 cm-1 indicates dec-reasing content of free hydroxyl groups of the sorbent.


290

Also, slightly decreasing in intensity and shift of the noticeable peaks at 1605.2 and 1032.6 cm-1 indicates binding heavy metal and –C=O group (Table 5). The carbonyl groups have the highest probability for bind-

ing investigated metal ions. Since the partially ionized carboxyl and phenol groups are the most responsible for the sorption of metals, minor changes observed in the FTIR spectrum of grape marc after the sorption

Figure 4. EDS Spectra of grape marc: a) before biosorption; b) after biosorption.

Figure 5. FTIR spectrum of biosorbent before the sorption process.


291

Figure 6. FTIR spectrum of biosorbent after the sorption process.

Table 5. Wavenumbers (cm-1) and functional groups from FTIR spectra before and after sorption

Grape marc Functional group

Native fermented Fermented, after sorption

3324.4 3314.4 valence vibration of O–H group

2922 2922 symmetric and asymmetric C–H valence vibration of aliphatic acids

1743.5 1743.6 –C=O valence vibration of carboxyl acids and corresponding esters

1605.2 1605.0 –C=O valence vibration (acids, esters)

1032.6 1031.7 –C–O valence vibration

process can be explained by the formation a bond between the metal ions and these groups [39]. It is assumed that the binding of metal cations to the aforementioned functional groups leads to the release of protons, because of their high affinity for Cd (II) ions. The release of protons is evident by decreasing pH of the solution during the sorption.

CONCLUSION

Effective removal of Cd (II) ions from aqueous solutions using grape marc largely depends on the experimental conditions such as pH value, initial con-centration, contact time and biosorbent grain size. The optimal parameters of the metal ions sorption are: pH 5.0 and biosorbent grain size < 0.5 mm. The equilibrium of the investigated process is best des-

cribed by the Langmuir isotherm. The sorption pro-cess is fast and the maximum capacity of 20 mg g-1 is reached after 20 min.

Based on the results of FTIR and EDS analysis, it was found that sorption of Cd (II) in the grape marc complex process is followed by electrostatic attract-ion, ion exchange and formation of complex com-pounds [18,19].

The results of this study show that the biosor-bent based on grape marc can be recommended as a very efficient and cost-effective mean to remove heavy metals from both natural and waste waters. The biosorbents represent an alternative to expensive sorbents and technologies used in the purpose of purification of aqueous solutions from heavy metals.


292

REFERENCES

[1] M.P.Wealkes, J.Inorg. Biochem. 79 (2000) 241 – 2443

[2] O. J.Esalah, M.E.Weber, J.H.Vera, Can. J. Chem. Eng. 78 (2003) 948–954

[3] LM. Canet, M. Ilpide. P. Seat, Sep. Sci. Technol. 37 (2002) 1851–1860

[4] J. Febrianto, A. N. Kosasih, J. Sunarso, Y. Ju, N. Indra-swati,S. Ismadji, J. Hazard. Mater. 162 (2009) 616-645

[5] V. Stanković, D. Božić, M. Gorgievski, G. Bogdanović, Chem. Ind. Chem Eng. Q. 15(4) (2009) 237−249

[6] E.-S.Z. El-Ashtoukhy, N.K. Amin, O. Abdelwahab, Desali-nation 223 (2008) 162–173

[7] T.N. Gamze, B. Mesci, Environ. Protection Eng. 37 (2011) 143-161

[8] F.W. Sousa, M.J. Sousa, R.N. Isadora I.R.N. Oliveira, A.G. Oliveira, R.M. Cavalcante, P.B.A. Fechine, V.O.S. Neto, D. de Keukeleire, R.F. Nascimento, J. Environ. Manage. 90 (2009) 3340-3344

[9] A.M. Awwad, A.M. Farhan, Amer. J. Chem. 2 (2012) 238- -244

[10] K. Kelly-Vargas, M. Cerro-Lopez, S. Reyna-Tellez, E.R. Bandala, J.L. Sanchez-Salas, Phys. Chem. Earth 37–39 (2012) 26–29

[11] V.B.H. Dang, H.D. Doan, T. Dang-Vu, A. Lohi, Biores. Technol. 100 (2009) 211–219

[12] Y. Hannachi, Holist, Approach Environ. 2 (2012)3 93-109

[13] V.J. Inglezakis, M.D. Loizidou, H.P. Grigoropoulou, J. Colloid Interface Sci. 261(1) (2003) 49-54

[14] Y. Sag, A. Kaya, T. Kutsal, Hydrometallurgy 50 (1998) 297-314

[15] W. EShin, M.R. Rowell, Chemosphere 60 (2005) 1054- -1061

[16] H. Ucun, O. Aksakal, E. Yildiz , J. Hazard. Mater. 161 (2009) 1040-1045

[17] M. Horsfall, Jr., A.I. Spiff , Afr. J. Biotechnol. 4(2) (2005) 191-196

[18] C.G. Panda, K.S. Das, K.A. Guha, Colloids Surfaces., B 62 (2008) 173-179

[19] M. Iqbal, A. Saeed , N. Akhtar, Bioresour. Technol. 81(2) (2002) 151–153

[20] M. Dundar, C. Nuhoglu, Y. Nuhoglu, J. Hazard. Mater. 151 (2008) 86-95

[21] R. Nadeem, M.A. Hanif, F. Shaheen, Sh. Perveen, M.N. Zafar, T. Iqbal, J. Hazard. Mater. 150 (2008) 335-342

[22] V.B.H. Dang, H.D. Doan, T. Dang-Vu, A. Lohi, Bioresour. Technol. 100(1) (2009) 211–219

[23] P. Miretzky, A.F. Cirelli, J. Hazard. Mater. 180(1-3) (2010) 1-19

[24] M. López-Mesas, E.R. Navarrete, F. Carrillo, C. Palet,, Chem. Eng. J. 174 (2011) 9-17

[25] A.M. Awwad, N.M. Salem, J. Saudi Chem. Soc., 2011

[26] A.M. Farhan, N.M. Salem, A.L. Ahmad, A.M. Awwad, Am. J. Chem. 2(6) (2012) 335-342

[27] W. Saikaew, P. Kaewsarn, W. Saikaew, Eng. Technol. 56 (2009) 287–291

[28] N. Azouaou, Z.Z. Sadaoui, A. Djaafri, H. Mokaddem, J. Hazard. Mater. 184 (2009) 126–134

[29] U. Kumar, M. Bandyopadhyay, Bioresour. Technol. 97 (2006) 104–110

[30] N.A.A. Babarinde, J.O. Babalola, S.O. Sanni, Int. J. Phys. Sci. 2 (2007) 207–211

[31] D. Stojanović, D. Bojić, J. Mitrović, T. Anđelković, M. Radović, A. Bojić, Chem. Ind. Chem. Eng. Q. 18(4) (2012) 563−576

[32] A.Ozer, D. Ozer, Process Biochem. 39 (2004) 2183–2191.

[33] N.D.Huston, R.T. Yang, Adsorption 3 (1997) 185-189

[34] F. Helfferich, Ion Exchange, McGraw Hill, New York, 1962, p. 166

[35] A. Khan Tabrez, S. Vedvati, A. Imran, J. Environ. Protect. Sci. 3 (2009) 124–132

[36] C.G. Panda , K.S. Das, K.A.Guha, Biointerfaces 62 (2008) 173-179

[37] L. Yao, Z.F. Ye, M.P. Tong, P. Lai , J.R. Ni, J. Hazard. Mater. 165 (2009) 250-255

[38] F.T. Li, H. Yang, Y. Zhao, R. Xu , Chin. Chem. Lett. 18 (2007) 325-328

[39] G. Guibaud, N. Tixier, A. Bouju, M. Baudu, Chemosphere 52 (2003) 1701-1710.


293

VESELINKA GRUDIĆ

JELENA ŠĆEPANOVIĆ

IVANA BOŠKOVIĆ

Faculty of Metallurgy and Technology, University of Montenegro, Podgorica,

Montenegro

NAUČNI RAD

UKLANJANJE KADMIJUM (II) JONA IZ VODENOG RASTVORA POMOĆU FERMENTISANE KOMINE GROŽĐA KAO NOVOG SORBENTA

U ovom radu je ispitivana biosorpcija kadmijuma iz vodenih rastvora korišćenjem fermen-

tisane komine grožđa i uticaj najvažnijih faktora, kao što su: kontaktno vrijeme, granulacija

biosorbenta i početna koncentracija metalnih jona. Ravnoteža procesa sorpcije jona

kadmijuma se postiže nakon 15-20 min u zavisnosti od početne koncentracije metalnih

jona.Tako kratko vrijeme potrebno da se postigne ravnoteža ukazuje na to da je transport

mase dominantan proces. Za opisivanje ravnoteže sorpcionog procesa korišćene su

Langmuir, Freundlich i Dubinin-Radushkevich izoterme, pri čemu je Langmuir model

najpogodniji. Maksimalni sorpcioni kapacitet je 20 mg g -1. Analiza EDS spektra je poka-

zala da je proces jonske razmjene jedan od glavnih mehanizama sorpcije. Manje promjene

uočene u FTIR spektru komine grožđa nakon sorpcije Cd (II) jona ukazuju na formiranje

veza između jona metala i djelimično jonizovane karboksilne i fenolne grupe iz biomase.

Rezultati istraživanja pokazuju da je fermentisana komina groždja, zahvaljujući svojoj

poroznoj strukturi i karakterističnom hemijskom sastavu, dobar sorbent.

Ključne reči: biosorpcija, komina groždja, izoterma, joni kadmijuma.





295

AMAL JUMA HABISH1

SLAVICA LAZAREVIĆ1

IVONA JANKOVIĆ-ČASTVAN1

BRANISLAV POTKONJAK2

ĐORĐE JANAĆKOVIĆ1

RADA PETROVIĆ1 1Faculty of Technology and

Metallurgy, University of Belgrade, Belgrade, Serbia

2Institute of Chemistry, Technology and Metallurgy, University of

Belgrade, Belgrade, Serbia

SCIENTIFIC PAPER

UDC 66.081:546.48-128:628:544

DOI 10.2298/CICEQ140130028H

THE EFFECT OF SALINITY ON THE SORPTION OF CADMIUM IONS FROM AQUEOUS MEDIUM ON Fe(III)-SEPIOLITE

Article Highlights • Cadmium ions sorption from saline waters onto Fe(III)-sepiolite was studied • Sorption capacities in saline waters were lower than in distilled water • The bonds sorbate-sorbent were stronger in distilled water than in saline waters • Sorption was well described by Sips isotherm and pseudo-second order kinetic

models • Desorption studies indicated strong sorbate-sorbent bonds, i.e., chemisorption Abstract

In this study, the sorption of cadmium ions onto sepiolite modified with hyd-rated iron(III) oxide, Fe(III)-sepiolite, has been investigated in natural seawater, artificial seawater, aqueous solution of NaCl of the same ionic strength as the seawater and distilled water. The sorption experiments were performed as a function of the initial solution pH value, initial metal concentration, and equi-libration time, using the batch method. The equilibrium sorption data were analyzed by the Langmuir, Freundlich and Sips isotherm models and the kinetics of sorption was analyzed using the pseudo-first-order and the pseudo-second-order kinetic models. The maximum sorption capacity and the strength of the sorbate-sorbent bonds at initial pH 7 were found to decrease in the following order: distilled water > NaCl solution > artificial seawater > natural seawater. The values of parameter nS in the Sips model, which fitted the equi-librium sorption results best, showed that heterogeneity of the sorbent surface was the highest in distilled water and the lowest in natural seawater. The sorption kinetic data fitted well with the pseudo-second-order kinetic model, which suggests that the rate-limiting step in Cd2+ sorption onto Fe(III)-sepiolite could be chemisorption. The low desorption percentage in both distilled water and 0.001 M HNO3 indicated that sorption occurred mainly by chemisorption mechanisms.

Keywords: chemisorption, Cd2+, modified sepiolite, seawater, modeling, desorption.

The contamination of natural waters by toxic heavy metals through the discharge of industry is a significant environmental problem. In particular, cad-mium is a highly toxic element at relatively low dos-ages and causes serious health problems to human, animals and aquatic life [1,2]. Cadmium is released into natural waters mainly through a number of ind-

Correspondence: R. Petrović, Faculty of Technology and Metallurgy, University of Belgrade, Karnegijeva 4, 11000 Bel-grade, Serbia. E-mail: [email protected] Paper received: 30 January, 2014 Paper revised: 9 July, 2014 Paper accepted: 29 July, 2014

ustries such as smelting, metal plating and mining, nickel-cadmium battery manufacturing, phosphate fertilizers, pigments, stabilizer and alloys production [1,3]. To avoid significant toxic effects on aquatic eco-systems, a cost-effective method to remove Cd2+ from natural waters is needed. The sorption is an appro-priate technology for cadmium removal from natural water systems, especially using low-cost sorbents such as zeolite [4], bentonite [5,6] and sepiolite [7,8].

Sepiolite is fibrous hydrated magnesium silicate clay with a unit cell formula Mg8Si12O30(OH)4 (OH2)4⋅8H2O. In some aspects, sepiolite is similar to other 2:1 trioctahedral silicates, such as talc (two layers of tetrahedral silica and a central octahedral

A.J. HABISH et al.: THE EFFECT OF SALINITY ON THE SORPTION OF Cd2+… Chem. Ind. Chem. Eng. Q. 21 (2) 295−303 (2015)

296

magnesium layer), but it has discontinuities and inversion in silica sheets, which give rise to structural channels filled with water molecules and running along the c-axis of fibrous particles. Due to its unique structure with micro channels and high surface area, sepiolite has received considerable attention as a sor-bent for organic and inorganic compounds from water, as well as a support for catalysts [9-14].

Recently, research interest has increased in term of use of iron oxides for modification of clay minerals to improve their sorption capacity [10-13]. These systems, which are characterized by the pre-sence of additional active sites, were applied as effective sorbents for metal ions removal from aque-ous solutions.

A number of sorbents have shown high capa-cities for the removal of Cd2+ from aqueous solutions. However, these sorption capacities have been deter-mined in deionized water, a simple matrix that fails to account for the effects of metal ion speciation, com-plexation, competing ions, and fouling of the sorbent materials by organic molecules often present in real waters [15]. In this study, the cadmium sorption on sepiolite modified with hydrated iron(III) oxide was investigated by using saline waters as an aquatic matrix. The cadmium sorption from different types of saline waters (natural seawater, artificial seawater and aqueous solution of NaCl of the same ionic strength as seawater) was compared with the sorp-tion from distilled water. The influence of the initial solution pH value, initial Cd2+ concentration and the equilibration time on the amount of Cd2+ sorbed on the modified sepiolite and the capability of the Lang-muir, Freundlich and Sips isotherm models to fit the experimental sorption data, were investigated.

EXPERIMENTAL PROCEDURE

Materials

Sorbent

The natural sepiolite used for the modification with hydrated iron(III) oxide was sampled from Andrici (Serbia). The chemical composition, specific surface area, pore size, X-ray diffraction and FTIR analyses of the sepiolite were reported previously [14]. The Fe(III)-sepiolite was prepared by mixing 20.0 g of sepiolite, 200 cm3 of freshly prepared 0.5 mol/dm3 FeCl3 aqueous solution, and 360 cm3 of 1 mol/dm3 NaOH aqueous solution [11]. The addition of NaOH solution was rapid and with stirring. The suspension was diluted to 1 dm3 by distilled water and was kept in closed polyethylene flask at 70 °C for 48 h. Then, the precipitate was centrifuged and washed by distilled water until it was Cl¯-free. The Fe(III)-sepiolite powder

had a dark red color, and its elemental analysis rev-ealed that the content of iron reached 22.2 wt.% [11]. The characterization of the modified sepiolite showed that Fe(III)-sepiolite maintained the basic structure of the natural sepiolite, which contained pure sepiolite, without impurities [11]. The presence of new crystal-line Fe phases was not observed by X-ray diffraction analysis, indicating that an amorphous Fe compound was formed. The pHpzc of Fe(III)-sepiolite (8.5±0.1) was higher than the pHpzc of the natural sepiolite (7.4±0.1), which means that the basicity of the sepio-lite surface was increased [11]. In addition, it was showed that modification of the sepiolite with hyd-rated iron(III) oxide caused negligible change of the specific surface area (from 280.0 to 285.6 m2/g) and small decrease of the pore volume (from 0.311 to 0.227 cm3/g) [12,13].

Types of water

The four types of aquatic systems were used for Cd2+ solutions preparation:

1. Natural seawater (NSW) obtained from Greece, which contained Cl¯ (564.13 mmol/dm3), Na+ (488.9 mmol/dm3), K+ (8.74 mmol/dm3), Mg2+ (45.22 mmol/ /dm3), Ca2+ (8.56 mmol/dm3), SO4

2- (27.91 mmol/dm3) and Br- (0.62 mmol/dm3); the concentration of the major cations and anions in NSW was determined on a Metrohm ion chromatography instrument, 861 Adv-anced Compact IC MSM II.

2. Artificial seawater (ASW), prepared according to analysis of the seawater; the quantities of salts used for the artificial seawater preparation were: 488.9 mmol/dm3 NaCl, 8.74 mmol/dm3 KCl, 25.88 mmol/dm3 MgCl2, 19.34 mmol/dm3 MgSO4, 8.56 mmol/dm3 CaSO4 and 0.62 mmol/dm3 NaBr.

3. Distilled water (DW); 4. Aqueous solution of NaCl (NCS) of the same

ionic strength as the seawater, prepared by dissol-ution of 40.2 g NaCl(s) in 1 dm3 of distilled water; ionic strength of NSW, I, was calculated according to the equation:

= 1

2 i iI c z (1)

where: ci is the concentration of the ion and zi is the charge of the ion.

Sorbate

The Cd2+ solutions were prepared by using cad-mium nitrate (Cd(NO3)2·4H2O) produced by Zorka, Šabac (Serbia).

Sorption experiments

The batch equilibration method was used to investigate the removal of Cd2+ from the solutions by


297

Fe(III)-sepiolite at 25±1 °C. The equilibrations of samples with 0.04 g Fe(III)-sepiolite and 20 cm3 of the solution were performed in a thermostated water bath with shaking. The sorption experiments were per-formed as a function of the initial metal concentration, the initial solution pH and the equilibration time. The initial Cd2+ concentrations as well as the concentra-tion of Cd2+ after the sorption were determined using atomic absorption spectroscopy (AAS, Perkin-Elmer 730; detection limit for Cd2+ was 0.3 mg/dm3). All the sorption studies were repeated twice; the reported values are the average of two measurements.

Influence of the initial solution pH

The influence of the initial solution pH value (pHi) on the sorption capacity was investigated using Cd2+ solution of 50 mg/dm3 concentration. The start-ing pH value was adjusted by HNO3 or KOH at 4, 5, 6, 7 and 8. The suspensions were equilibrated for 24 h and then filtrated through a filter paper (retention of particles larger than 11 μm). The final solution pH (pHf) was measured using a pH meter InoLab WTW series pH 720.

The amount of cadmium sorbed per unit mass of Fe(III)-sepiolite at equilibrium, qe (mg/g), was calcul-ated using the equation:

−= ee

oc cq V

m (2)

where co and ce are the initial and the equilibrium concentrations (mg/dm3), m is the mass of the sor-bent (g), and V is the volume of the solution (dm3).

Sorption isotherms

Sorption isotherms were determined using Cd2+

solutions of different concentration (5, 10, 20, 30, 45, 75, 100 and 150 mg/dm3) at pHi 7.0±0.1. The samples were equilibrated for 24 h. The amount of Cd2+ sorbed per unit mass of Fe(III)-sepiolite at equilibrium, qe (mg/g), was calculated using Eq. (2).

Sorption kinetics

The kinetic experiments were performed at the Cd2+ initial concentration of 50 mg/dm3 (co), at the ini-tial pH 7.0±0.1, for the contact time of ½, 1, 2, 4, 8, 16 and 24 h.

The quantity of Cd2+ ions sorbed per unit mass of Fe(III)-sepiolite after the period of time t (qt) was calculated according to the equation:

qt = −o tc c

Vm

(3)

where ct is the concentration of Cd2+ after the period of time t.

Desorption studies

Desorption studies were performed with the sor-bent samples obtained by the sorption from the sol-utions in DW, NCL, ASW and NSW at the initial con-centration of 150 mg/dm3. The loaded sorbent samples were dried at 100 °C for 24 h and used in batch desorption experiments. Test flasks were filled with 20 cm3 of desorbing solution (distilled water or 0.001 M HNO3), and 0.04 g of the loaded sorbent sample was added. After an equilibration for 24 h in a thermo-stated water bath with shaking at a temperature of 25 °C, the dispersions were filtered and the filtrates were analyzed to check the desorbed Cd2+, using AAS. The desorption efficiency was defined as the ratio between the amount of Cd2+ desorbed and the amount of Cd2+ sorbed on the sorbent. In addition, concentration of Fe3+ in the filtrates after desorption, as well as in sol-ution after the sorption at the initial concentration of 150 mg/dm3, was determined in order to check sta-bility of hydrated iron(III) oxide on the sepiolite sur-face.


Influence of initial solution pH on the sorption capacity

Metal sorption from aqueous solutions can be greatly affected by the solution pH that influences not only the binding sites (e.g., degree of protonation or deprotonation of functional groups at the sorbent sur-face), but also the metal chemistry (e.g. speciation and precipitation). At pH below 8, Cd2+ is a dominant and soluble form, while precipitate is formed at higher pH values [16]. In general, the precipitate formation depends on the Cd2+ concentration, other present cat-ions concentration, types and concentrations of anions, temperature, and presence of solids.

The influences of the initial solution pH value on the sorption capacity of Fe(III)-sepiolite in distilled water, seawater, artificial seawater and NaCl solution are presented in Figure 1. The effect of pHi was studied in the pH range from 4 to 8 in order to avoid the occurrence of the precipitation of cadmium hyd-roxide at higher pH values. According to the value of the solubility product constant of Cd(OH)2 and Cd2+

initial concentration, Cd(OH)2 precipitation starts at pH > 8.5. The final pH values, pHf, for each sample are also given in Figure 1.

According to Figure 1, the sorption capacity gen-erally increased with the initial pH increase for all types of water systems that can be explained by decrease in number of H+ that compete with Cd2+ for the sorption sites. The values of pHf show that H+ ions associated with the surface functional groups of Fe(III)-sepiolite at pHi ≤ 7 (pHf > pHi), while at pH 8 they were released into the solution due to disso-


298

ciation of the functional groups (pHf < pHi). Thus, the pH value at which the surface charge changed from positive to negative and vice versa is between pH 7 and pH 8 for all types of water, which is lower than the pHpzc of Fe(III)-sepiolite (8.5±0.1). This indicates the specific sorption of Cd2+, i.e., formation of inner-sphere complexes with the surface functional groups. Such bonds are stronger than electrostatic interactions between charged surface and oppositely charged ions (outer-sphere complexes). As pHi decreased from 7 to 4, more H+ were bonded to the surface functional groups, leaving a smaller number of groups for Cd2+ binding, so the sorption capacity decreased with the pHi decrease. In addition, as pHi was lower, the negative charge of the surface became lower and the positive charge higher, which also decreased the sorption of positively charged cadmium ions by elec-trostatic interactions.

3 4 5 6 7 8 90

7

14

21

7.687.907.62

7.527.557.44

7.217.17

7.24

6.947.02

7.05

6.846.946.97

7.287.24

DW NSW ASW NCSq e

(mg

Cd2+

/g F

e(II

I)-s

epio

lite)

pH

7.037.14

7.20

Figure 1. The effect of the initial pH on Cd2+ sorption by Fe(III)-sepiolite from DW, NCS, ASW and NSW for the initial Cd2+ concentration of 50 mg/dm3 (the numbers in the figure

indicate the final pH values).

The sorption capacity of Fe(III)-sepiolite is sig-nificantly higher in the case of sorption from distilled water than from the saline waters, which can be explained by competitive effect of ions from saline waters for active sites at the surface of the sorbent. NSW and ASW contain higher-valence ions (Ca2+ and Mg2+), which have more competitive effect than monovalence ions, so sorption capacity was lower in NSW and ASW than in NaCl solution. Moreover, high concentration of ions in water, like in all types of saline waters, caused decreasing of surface charge of the sorbent [17] and consequently, sorption capacity decreasing. Comparison of the sorption capacity of Fe(III)-sepiolite in different types of water was done fur-ther by sorption isotherm determination at pHi 7.0±0.1.

Sorption isotherms studies and modeling

The experimental sorption isotherms as the dependence of mass of Cd2+ sorbed per unit mass of Fe(III)-sepiolite at equilibrium on the Cd2+ equilibrium concentration, for distilled water, natural seawater, artificial seawater and NaCl solution at pHi 7.0±0.1 are presented in Figure 2. The equilibrium sorption data were fitted to the Langmuir, Freundlich and Sips isotherms, and the model fits are presented with the experimental data in Figure 2. The sorption isotherms constants, determined by non-linear regression anal-ysis using the Easy Plot, are summarized in Table 1, where: qm is the maximum adsorption capacity, KL is the Langmuir constant related to the energy of ads-orption, Kf is the Freundlich constant related to the adsorption capacity, n is the dimensionless sorption intensity parameter, Ka is the Sips equilibrium cons-tant and nS is the index of heterogeneity.

According to Figure 2, the Fe(III)-sepiolite showed the highest sorption capacity for Cd2+ in dis-tilled water, then in NaCl solution, followed by sorp-tion capacity in artificial seawater and the lowest in natural seawater. Such order of sorption capacity, which is the same as in the investigation of the effect of pHi on Cd2+ sorption, confirmed strong influence of salinity on the sorption properties of Fe(III)-sepiolite. Salinity was also observed to limit performance of dif-ferent organic and inorganic chemisorbents for the collection of a diverse range of fission and activation product elements of the nuclear fuel cycle including Co, Zr, Nb, Ru, Ag, Te, Sb, Ba and Cs [18]. Salinity dependence was most significant for hard cations, i.e., Cs+ and Ba2+. Likewise, sorption capacities of three nanoporous sorbents containing chelating diamine functionalities [15] for Cu2+ were lower in sal-ine water than in river water.

The saline waters used in this research had the same ionic strength, but Fe(III)-sepiolite had lower sorption capacity in NSW and ASW than in NaCl sol-ution because Ca2+ and Mg2+ from seawaters com-pete with Cd2+ more effectively than Na+. The similar results were observed for the sorption of Cd2+ onto amorphous hydrous manganese dioxide [19], where sorption has been studied in 0.5 M NaCl solution and major ion seawater: sorption capacity was lower in seawater than in 0.5 M NaCl solution. Lower sorption capacity of Fe(III)-sepiolite in the NSW than in ASW may result from the presence of natural organic mat-ter (NOM) and other ions that are not present in ASW. NOM might have occupied some binding sites of Fe(III)-sepiolite, which led to the reduction of Cd2+ sorption. It was shown also that the presence of nat-ural organic matter had little effect on the ability of dif-ferent chemisorbents to extract target elements [18].

A.J. HABISH et al.: THE EFFECT OF SALINITY ON THE SORPTION OF Cd2+… CI&CEQ 20 (2) 295−303 (2015)

299

0 20 40 60 80

0

10

20

30

40

50

q e (m

g C

d2+/g

Fe(

III)

-sep

iolit

e)

DW Langmuir Freundlich Sips

ce (mg Cd2+/dm3 )

0 20 40 60 800

10

20

30

ce (mg Cd2+/dm3 )

q e (m

g C

d2+/g

Fe(

III)

-sep

iolit

e)

NCS Langmuir Freundlich Sips

(a) (b)

0 20 40 60 80 100 1200

5

10

15

20 ASW Langmuir Freundlich Sips

ce (mg Cd2+/dm3 )

q e (m

g C

d2+/g

Fe(

III)

-sep

iolit

e)

0 20 40 60 80 1000

2

4

6

8

10

12

q e (m

g C

d2+/g

Fe(

III)

-sep

iolit

e) NSW Langmuir Freundlich Sips

ce (mg Cd2+/dm3 )

(c) (d)

Figure 2. The sorption isotherms for Cd2+ onto Fe(III)-sepiolite in DW (a), NCS (b), ASW (c) and NSW (d) at pHi 7.

Table 1. Langmuir, Freundlich and Sips isotherms constants for the sorption of Cd2+ onto Fe(III)-sepiolite at pHi 7

Model Type of water

DW NCS ASW NSW

Langmuir model

=+m L e

eL e1

q K cq

K c

KL / dm3 mg-1 0.349 0.0784 0.0289 0.0257

qm / mg g-1 41.1 26.4 17.7 14.3

R2 0.888 0.976 0.949 0.979

Freundlich model

= 1/e f e

nq K c

Kf / mg(1-1/n)dm3/n g-1 12.3 4.20 1.35 0.804

n 3.22 2.49 1.98 1.76

R2 0.938 0.983 0.924 0.974

Sips model

=+

S

S

m a ee

a e

( )

1 ( )

n

n

q K cq

K c

qm / mg g-1 79.3 44.0 18.1 19.1

Ka / (dm3 mg−1)nS 0.181 0.0836 0.0299 0.0279

nS 0.457 0.602 0.978 0.821

R2 0.948 0.988 0.949 0.982

The study of equilibrium sorption isotherms provides important data for understanding the mech-anism of the sorption. The different sorption isotherms are characterized by their constants, the values of

which suggest the surface properties and affinity of the sorbent to the sorbate. The Langmuir model [20,21] assumes monolayer sorption at specific homo-geneous sites, without any interactions between sor-


300

bed species. The Freundlich equation [21,22] des-cribes sorption (possibly multilayer in nature) on a heterogeneous surface consisting of non-identical and energetically non-uniform sites. A combined form of Langmuir and Freundlich expressions, the Sips model [21,23], deduced for predicting the heterogen-eous sorption systems and circumventing the limit-ation of the rising sorbate concentration associated with the Freundlich isotherm model. At low sorbate concentrations (when the sorbate content is much lower than the sorbent capacity), the model reduces to Freundlich isotherm, while at high concentrations (when the sorbate content is higher than the sorbent capacity), it predicts a monolayer sorption capacity, characteristic of the Langmuir isotherm.

The results presented in Table 1 show that the Sips model is the best model to explain the sorption behavior of Cd2+ onto Fe(III)-sepiolite, but also both Langmuir and Freundlich fit the results well.

The maximum sorption capacity obtained by Langmuir equation, qm, and the affinity parameter, KL, were found to decrease in the following order: DW > > NCS > ASW > NSW. The higher values of KL imply stronger sorbate-sorbent interactions, so the strong-est bonds were formed in distilled water and the weakest in natural seawater [21]. The Freundlich con-stant Kf, which is an approximate indicator of adsorp-tion capacity, and intensity parameter n, which is an indicator of the strength of sorption [21], decrease in the same order as qm and KL.

The values of the maximal sorption capacity obtained by the Sips model were higher than those obtained by the Langmuir model. The Sips constant Ka, similarly to the Langmuir constant KL, could be regarded as representative of the strength of sorption: higher value of Ka implies stronger bonds between sorbate species and sorbent active sites [21,23]. According to the values of Ka in Table 1, the strength of sorbate-sorbent bonds decreases in the order: DW > > NCS > ASW > NSW. The parameter nS in Sips model is the heterogeneity factor, which shows the deviation from the Langmuir model. The value of nS for a homogeneous material is 1, and it is less than one for heterogeneous materials [24,25]. According to the values of nS for Cd2+ sorption onto Fe(III)-sepiolite (Table 1), the heterogeneity of the sorbent is the high-est in distilled water, lower in NaCl solution and the lowest in seawaters. Heterogeneity of a sorbent sur-face is a result of existence of different active sites for sorption. In the case of sorption from NaCl solution, it can be supposed that Na+ and Cl- blocked some active sites for Cd2+ sorption at the surface of Fe(III)- -sepiolite and thus the surface had less different act-ive sites than in the case of the sorption from distilled

water, meaning lower heterogeneity of the surface. Having in mind that sorbate-sorbent bonds were stronger in distilled water than in NaCl solution, it can be supposed that Na+ and Cl– blocked active sites of higher energy, leaving active sites of lower energy for Cd2+ sorption. In seawaters, where there were Ca2+ and Mg2+, the heterogeneity was even lower and sor-bate-sorbent bonds were weaker than in NaCl sol-ution.

The results of the isotherm modeling showed that Cd2+ sorbed onto heterogeneous surface of Fe(III)- -sepiolite until the monolayer is formed, but sorption capacity and surface heterogeneity were highly inf-luenced by the presence of ions in saline waters.

Sorption kinetics and modeling

The influence of the contact time t on the amount of Cd2+ sorbed onto Fe(III)-sepiolite, qt, is shown in Figure 3. It can be seen that the removal of Cd2+ from different water systems by sorption onto Fe(III)-sepiolite is comprised of three steps: the rapid sorption in the first 0.5 h, then slightly decreased sorption, followed by its slow increase after 2 h, until the equilibrium was reached. The high number of active sites available at the beginning of the sorption may explain the fast initial Cd2+ uptake at the Fe(III)- -sepiolite surface. However, some rapidly sorbed ions were desorbed in the next step. After that, a slow increase of the amount of sorbed Cd2+ can be obs-erved, since the most sorption sites were occupied and Cd2+ concentration in the solutions decreased; thus, the sorption became less efficient.

0 250 500 750 1000 1250 1500 17500

8

16

24 DW NSW ASW NCS

q t (m

g C

d2+/g

Fe(

III)

-sep

iolit

e)

t (min)

Figure 3. The effect of contact time on Cd2+ sorption onto Fe(III)-sepiolite in DW, NCS, ASW and NSW at pHi 7, for initial

Cd2+ concentration of 50 mg/dm3.

The kinetics of Cd2+ sorption onto Fe(III)-sepio-lite was analyzed using the pseudo-first-order equa-tion proposed by Lagergren [26] and the pseudo-


301

second-order kinetic model proposed by Ho and McKay [27], and the results are presented in Table 2. Accordingly, the pseudo-second-order equation is the model that best describes the experimental data, since the qe values estimated by this model are in good agreement with the experimental results and the values of the correlation coefficients are close to 1. The fact that the pseudo-second-order model gave better agreement with the experimental data than the pseudo-first-order model is not surprising, since this seems to be a generally prevalent observation in heavy metal sorption studies [27,28]. Further, this trend suggests that the rate-limiting step in heavy metal sorption is chemisorption, which involves val-ence forces through sharing or exchange of electrons between sorbent and sorbate, complexation, coordi-nation and/or chelation, rather than physisorption [28]. As the correlation coefficients for the pseudo-second-order model are high (Table 2), it may be assumed that the sorption of Cd2+ onto Fe(III)-sepiolite occurs mainly through chemisorption mechanisms, e.g., spe-cific sorption, as it was shown during investigation of pHi influence on the sorption capacity, and ion exchange of Mg2+ from the sepiolite structure by Cd2+

from the solutions [14]. Similarly, sorption of Ni2+ [11] and Co2+ [12] from solution in distilled water onto Fe(III)-sepiolite was mainly governed by chemisorp-tion and was best described by the pseudo-second-order model.

Kinetic analysis confirmed that sorption capacity of Fe(III)-sepiolite is the highest in distilled water, and that capacity in saline waters (natural and artificial) with higher valence ions present, is lower than in NaCl solution. Salinity influenced not just sorption capacity, but also rate of the sorption: according to the values of the kinetic constant, the rate of sorption was the highest in natural seawater, lower in artificial and distilled water and the lowest in NaCl solution.

Desorption studies

Desorption studies were conducted to better understand the mechanism of Cd2+ sorption, i.e., to investigate the strength of sorbate-sorbent bonds. The results of desorption studies are given in Table 3.

Table 3. Percentage desorption of Cd2+ from Fe(III)-sepiolite loaded in different types of water

Fe(III)-sepiolite loaded in:

Percentage desorption in:

Distilled water 0.001 M HNO3

DW 0.36 2.54

NCL 0.40 2.80

ASW 2.24 3.61

NSW 3.30 4.67

Very low desorption percentage in both desorb-ing media suggests that the sorption of Cd2+ onto Fe(III)-sepiolite occurred significantly via chemisorp-tion mechanisms, as it was found according to dependences pHf-pHi and kinetics modeling. Desorp-tion was slightly more intense in 0.001 M HNO3 than in distilled water, as it was expected, because the acid solution is a stronger desorbing agent than water due to higher concentration of H+. In both desorbing media, percentage desorption was the lowest for the sorbent loaded in DW and the highest for the sorbent loaded in NSW (Table 3). These results indicate that the strongest sorbate-sorbent bonds were formed in

DW, than in NCL, followed by ASW and the weakest in NSW, as it was found according to isotherm mod-eling.

The concentration of Fe3+ in solution after sorp-tion, as well as after desorption in both distilled water and 0.001 M HNO3, was below detection limit of the AAS (≤ 0.3 mg/dm3), which indicate high stability of hydrated iron(III) oxide on the sepiolite surface, i.e. low Fe3+ releasing from Fe(III)-sepiolite, even in acid solution.

Table 2. Kinetic parameters for Cd2+ sorption onto Fe(III)-sepiolite at pHi 7, for the initial Cd2+ concentration of 50 mg/dm3 (k1 is the rate constant of pseudo-first-order model; and k2 is the rate constant of the pseudo-second-order model)

Kinetics Type of water

DW NCS ASW NSW

Pseudo-first-order

− = − 1e elog ( ) log

2.303tk

q q q t

k1 / min-1 13.772 21.913 10.840 0.0085

qe / mg g-1 8.2 7.7 3.6 0.19

R2 0.888 0.886 0.841 0.530

Pseudo-second-order

= +2

e2 e

1 1

t

tt

q qk q

k2 / g mg-1·min-1 0.0016 0.00035 0.0029 0.0123

qe / mg g-1 25.0 14.3 7.7 7.7

R2 0.991 0.991 0.995 0.998


302

CONCLUSIONS

In this study, sepiolite modified with hydrated iron(III) oxide, Fe(III)-sepiolite, was used as a sorbent for Cd2+ from different types of water and aqueous solutions. It was shown that the sorption capacity was highly influenced by the ions presented in aqueous media: the highest capacity was in distilled water, where there were no other ions except Cd2+, than in NaCl solution, followed by artificial seawater and the lowest capacity was in natural seawater. The equ-ilibrium sorption data modeling with the Langmuir, Freundlich and Sips models showed that the Sips model was the best to explain the sorption behavior of Cd2+ onto Fe(III)-sepiolite at initial pH 7, which indi-cates the sorption onto heterogeneous surface until the monolayer is formed. According to the values of Sips parameters, it was concluded that the strength of sorbate-sorbent bonds and heterogeneity of the sorb-ent surface decreased in the following order: distilled water > NaCl solution > artificial seawater > natural seawater. The sorption kinetic data fitted well with the pseudo-second-order kinetic model, which suggested that the rate-limiting step in Cd2+ sorption onto Fe(III)- -sepiolite could be chemisorption. The desorption studies in distilled water and 0.001 M HNO3, as well as dependence of final pH on initial pH, suggested that the sorption of Cd2+ onto Fe(III)-sepiolite occurred significantly via chemisorption mechanisms.

Acknowledgements

The authors wish to acknowledge the financial support for this research from the Ministry of Edu-cation, Science and Technological Development of the Republic of Serbia through the project III 45019.

REFERENCES

[1] E. Gutierrez-Segura, M. Solache-Rios, A. Colin-Cruz, C. Fall, J. Environ. Manage. 97 (2012) 6-13

[2] A. Maleki, A.H. Mahvi, M.A. Zazouli, H. Izanloo, A.H. Barati, Asian J. Chem. 3 (2011) 1373-1376

[3] H.K. Boparai, M. Joseph, D.M. O'Carroll, J. Hazard. Mater. 186 (2011) 458-465

[4] S. Wang, Y. Peng, Chem. Eng. J. 156 (2010) 11-24

[5] D. Zhao. S. Chen, S. Yang, X. Yang, S. Yang, Chem. Eng. J. 166 (2011) 1010-1016

[6] E. Alvares-Ayuso, A. Garcia-Sanchez, Clays Clay Miner. 51 (2003) 475-480

[7] S. Kocaoba, Desalination 244 (2009) 24-30

[8] M.F. Brigatti, C. Lugli, L. Poppi, Appl. Clay Sci. 16 (2000) 45-57

[9] I. Kuncek, S. Sener, Ultrason. Sonochem. 17 (2010) 250-257

[10] E. Eren, H. Gumus, Desalination 273 (2011) 276-284

[11] S. Lazarevic, I. Jankovic-Castvan, V. Djokic, Z. Radova-novic, Dj. Janackovic, R. Petrovic, J. Chem. Eng. Data 55 (2010) 5681-5689

[12] S. Lazarevic, I. Jankovic-Castvan, B. Potkonjak, Dj. Janackovic, R. Petrovic, Chem. Eng. Process. 55 (2012) 40-47

[13] S. Lazarevic, I. Jankovic-Castvan, A. Onjia, J. Krstic, D. Janackovic, R. Petrovic, Ind. Eng. Chem. Res. 50 (2011) 11467-11475

[14] S. Lazarevic, I. Jankovic-Castvan, D. Jovanovic, S. Milo-njic, D. Janackovic, R. Petrovic, Appl. Clay Sci. 37 (2007) 47-57

[15] W. Chouyyok, Y. Shin, J. Davidson, W.D. Samuels, N.H. Lafemina, R.D. Rutledge, G.E. Fryxell, Environ. Sci. Technol. 44 (2010) 6390-6395

[16] V.A. Nazarenko, V.P. Antonovich, E.E. Neveskaya, Hyd-rolysis of metal ions in diluted solutions, Atomizdat, Moscow, 1979

[17] J.C. Crittenden, R.R. Trussell, D.W. Hand, K.J. Howe, G. Tchobanoglous, Water Treatment: Principles and Design, John Wiley & Sons. Inc., Hoboken, NJ, 2005

[18] B.E. Johnson, P.H. Santschi, R.S. Addleman, M. Douglas, J.D. Davidson, G.E. Fryxell, J.M. Schwantes, Appl. Radiat. Isotopes 69 (2011) 205–216

[19] S.S. Tripathy, J.-L. Bersillon, K. Gopal, Desalination 194 (2006) 11–21

[20] I. Langmuir, J Am. Chem. Soc. 40 (1918) 1361-1403

[21] K.Y. Foo, B.H. Hameed, Chem. Eng. J. 156 (2010) 2-10

[22] H. Freundlic, Z. Phys. Chem. 57 (1906) 385-470

[23] V. Hernández-Montoya, M.A. Pérez-Cruz, D.I. Mendoza-Castillo, M.R. Moreno-Virgen, A. Bonilla-Petriciolet, J. Environ. Manage. 116 (2013) 213-221

[24] E. Turiel, C. Perez-Conde, A. Martin-Esteban, Analyst 128 (2003)137-141

[25] R.J. Umpleby, S.C. Baxter, Y. Chen, R.N. Shah, K.D. Shimizu, Anal. Chem. 73 (2001) 4584-4591

[26] S. Lagergren, K. Sven, Vetenskapsakad. Handl. 24 (1898) 1-39

[27] Y.S. Ho, G. Mckay, Process Biochem. 34 (1999) 451-465

[28] S.O. Lesmana, N. Febriana, F.E. Soetaredjo, J. Sunarso, S. Ismadji, Biochem. Eng. J. 44 (2009) 19-41.


303

AMAL JUMA HABISH1

SLAVICA LAZAREVIĆ1

IVONA JANKOVIĆ-ČASTVAN1

BRANISLAV POTKONJAK2

ĐORĐE JANAĆKOVIĆ1

RADA PETROVIĆ1

1Tehnološko-metalurški fakultet, Univerzitet u Beogradu, Beograd,

Srbija 2Institut za hemiju, tehnologiju i

metalurgiju, Univerzitet u Beogradu, Beograd, Srbija

NAUČNI RAD

UTICAJ SALINITETA VODE NA SORPCIJU JONA KADMIJUMA NA Fe(III)-SEPIOLITU

U ovom radu je proučavana sorpcija jona kadmijuma iz prirodne morske vode, labo-

ratorijski pripremljene morske vode, vodenog rastvora NaCl iste jonske jačine kao morska

voda i destilovane vode, na sepiolitu modifikovanom hidratisanim gvožđe(III)-oksidom.

Eksperimenti su izvedeni u šaržnim uslovima, pri različitim početnim pH vrednostima,

početnim koncentracijama jona kadmijuma i vremenima uravnotežavanja. Rezultati sorp-

cije u ravnotežnim uslovima su analizirani primenom modela Langmira, Frojndliha i Sipsa,

a kinetika sorpcije je analizirana kinetičkim modelima pseudo-prvog i pseudo-drugog reda.

Utvrđeno je da maksimalni adsorpcioni kapacitet i jačina veza sorbat-sorbent pri početnoj

pH 7 opadaju u nizu: destilovana voda > NaCl rastvor > laboratorijski pripremljena morska

voda > prirodna morska voda. Vrednosti parametra ns u modelu Sipsa, koji najbolje opisuje

eksperimentalne rezultate, pokazuju da je heterogenost površine sorbenta najveća u

destilovanoj vodi i najmanja u prirodnoj morskoj vodi. Rezultati ispitivanja kinetike sorpcije

se bolje fituju modelom pseudo-drugog reda, što sugeriše da stupanj koji određuje brzinu

sorpcije može biti hemisorpcija. Mali stepen desorpcije i u destilovanoj vodi i u 0.001 M

HNO3 je potvrdio da se joni kadmijuma vezuju za modifikovani sepiolit uglavnom hemi-

sorpcijom.

Ključne reči: hemisorpcija, Cd2+, modifikovani sepiolit, morska voda, modelo-vanje, desorpcija.





305

SHAOPENG GUO1

LINA LV1

JIA ZHANG1

XIN CHEN2

MING TONG2

WANZHONG KANG2

YANBO ZHOU1

JUN LU1 1Key Laboratory of Coal

Gasification and Energy Chemical Engineering of Ministry of

Education, East China University of Science & Technology, Shanghai,

P. R. China 2SINOPEC Ningbo Engineering

Co., Ltd., Ningbo, P. R. China

SCIENTIFIC PAPER

UDC 66.081.2:546.214:54

DOI 10.2298/CICEQ140618029G

SIMULTANEOUS REMOVAL OF SO2 AND NOx WITH AMMONIA COMBINED WITH GAS-PHASE OXIDATION OF NO USING OZONE

Article Highlights • Simultaneous removal of SO2 and NOx was achieved by oxidation of NO with O3 and

ammonia absorption • The appearance of SO2 in fuel gas has little impact on the oxidation of NO • The O3/NO molar ratio is the most important factor in the ozone oxidation process • Increasing of O3/NO mole ratio and SO2 concentration are favorable to recycling the

byproducts Abstract

A process for simultaneous desulfurization and denitrification is proposed, con-sisting of ozone as the oxidizing agent for NO and ammonia solution as the absorbent. The results showed that the presence of SO2 and the concentration changes of NO and SO2 have little impact on the oxidation of NO, the oxidation efficiency of NO can achieve over 90% when the molar ratio of O3/NO is 1.0. The presence of NOx had little effect on the absorption of SO2, while an appropriate increase of SO2 concentration favorably affected NOx absorption. The removal efficiency of SO2 and NOx reached 99.34 and 90.01% at pH 10, flow rate 0.95 Nm3/h, n[O3]/n[NO] 1.0, initial SO2 concentration 2000 mg/Nm3, initial NO concentration 200 mg/Nm3, ammonia concentration 0.3%, oxygen content of the simulated flue gas 12%, oxidation reaction temperature 423 K and absorption reaction temperature 298 K in the experimental system.

Keywords: ozone, nitrogen oxides, sulfur dioxide, ammonia, simul-taneous absorption.

Sulfur dioxide (SO2) and nitrogen oxides (NOx) are the most abundant air pollutants during coal com-bustion. These pollutants have brought about signi-ficant effects on both the environment and human health in China [1,2]. Many technologies have been used to reduce the emission of sulfur dioxide (SO2) and nitrogen oxides (NOx), among which wet flue gas desulfurization (WFGD) and selective catalytic reduc-tion (SCR) are regarded as the most effective techno-logies for SO2 and NOx removal, respectively. How-ever, the individual treatment technology may lead to high investment and operating costs. To overcome

Correspondence: J. Lu, Key Laboratory of Coal Gasification and Energy Chemical Engineering of Ministry of Education, East China University of Science & Technology, No. 130 Meilong Road, Shanghai 200237, P. R. China. E-mail: [email protected] Paper received: 18 June, 2014 Paper revised: 16 August, 2014 Paper accepted: 25 August, 2014

this problem, many technologies such as nonthermal plasma, electron beam irradiation [3-5] and adsorp-tion process [6-15] have been developed for simul-taneous removal of SO2 and NOx, but only few com-mercial applications have been reported until now.

As an efficient gas phase oxidant with advent-ages of selectivity, high oxidation efficiency, fast oxi-dation speed and non-pollution decomposition pro-ducts, ozone can easily oxidize NO into high order nitrogen species such as NO2, NO3 and N2O5 etc., which are highly soluble in water [16,17]. Wang et al. [18] have confirmed that it is possible to achieve about 97% of NO removal efficiency and nearly 100% of SO2 removal efficiency through ozone oxidation and Ca(OH)2 absorption. Young et al. [19] obtained NOx removal efficiency of about 95% and SO2 rem-oval efficiency of 100% with an ozonizing chamber and an absorber containing Na2S solution. In addition, the ammonia-based wet flue gas desulfurization pro-

S. GUO et al.: SIMULTANEOUS REMOVAL OF SO2 AND NOx… Chem. Ind. Chem. Eng. Q. 21 (2) 305−310 (2015)

306

cess has attracted much attention in China recently due to its high desulfurization efficiency, useful bypro-ducts, no secondary pollution and lower costs [20-22]. Therefore, ammonia-based wet flue gas desulfur-ization process combined with ozone oxidation is a promising technology for simultaneously desulfuriz-ation and denitrification and it is worthy of further research.

In this paper, a flue gas treatment process by utilizing ozone as oxidant and ammonia solution as absorbent was established to achieve simultaneous removal of NOx and SO2. The oxidation of NO by ozone was studied and the performance of this com-bined treatment process was investigated under dif-ferent operation parameters.

EXPERIMENTAL

As shown in Figure 1, the experimental appar-atus for simultaneously desulfurization and denitri-fication is a set of self-made equipment. Ozone was produced by the ozone generator (Qingdao Guolin Industry Co., Ltd, Qingdao, China). The gas reactor was made of stainless steel, 32 mm in inner diameter and 700 mm in length, which was inserted into a tube type resistance furnace with temperature control. The bubbling reactor was a glass cylindrical vessel with a total volume of 1.4 L. The simulated flue gas was sup-plied by air and compressed gas cylinders filled with N2, NO and SO2. The flow rates of all gases were controlled by gas volume flow meters. The mixture of air, N2, NO and SO2 reacted with O3 in the reactor, then the gas mixture went through the cooling pipe and went into the bubbling reactor filled with ammonia solution as absorbent. When the absorption was completed, the gas reactor was purged with nitrogen for 10 min.

The concentration of O3 was measured accord-ing to iodometric method (CJ/T 3028.2-1994). The concentrations of NO, NO2 and SO2 were detected by flue gas analyzer Optima 7 (MRU, Germany). The flow rate and oxygen content of the simulated flue gas were fixed to 0.95 Nm3/h and 12%, respectively. The mass fraction of ammonia in the absorption solution was 0.3%. The oxidation of NO was conducted at 150 °C. The NO oxidation efficiency was investigated with the mole ratios of O3/NO changing from 0.3 to 1.5, the initial concentration of NO ranging from 200 to 800 mg/N m3 and initial SO2 concentration varying from 1000 to 3000 mg/Nm3 in the oxidation process. In the absorption process, the removal efficiency of SO2, NOx were analyzed in different experimental con-ditions which are mentioned above.

RESULTS AND DISCUSSIONS

Influence of O3/NO molar ratio

In this section, the oxidation efficiency of NO, SO2 and NOx removal efficiency were studied in dif-ferent O3/NO mole ratio. The initial concentrations of NO and SO2 were fixed at 500 and 2000 mg/N m3.

Figure 2 shows the oxidation efficiency of NO both in presence and absence of SO2 under different O3/NO mole ratios. From Figure 2 it can be seen that the NO oxidation efficiency gets higher with the inc-rease of O3/NO mole ratio up to 1.0, the oxidation reaction happens quickly as following reaction (1):

+ → +3 2 2NO O NO O (1)

The oxidation efficiency exceeds 90% and many other reactions take place at the same time when the O3/NO mole ratio gets higher than 1.0, mainly inc-luding reactions (2)–(5):

Figure 1. Schematic of the experimental apparatus; 1 - air pump; 2 - gas volume flow meter; 3 - gas reactor; 4 - cooling pipe; 5 - ozone generator; 6 - bubbling reactor; 7 - magnetic stirring constant-temperature water bath; 8 - rubber stopple; 9 - aerator; 10 - magnetic

stirrer; 11 - Ph meter; a, b, c, d - sample connection.


307

→2 3 3 2NO + O NO + O (2)

→2 3 2 5NO + NO N O (3)

→2 5 2 3N O NO + NO (4)

→3 2NO + NO 2NO (5)

Figure 2. The influence of O3/NO mole ratio on NO oxidation efficiency (flow rate: 0.95 N m3/h; oxygen content of the simulated flue gas: 12%; reaction temperature: 423 K).

The oxidation efficiencies of NO are slightly lower when NO coexists with SO2, the reason is that SO2 can be oxidized by O3 as the following reaction:

→2 3 2 3SO + O O + SO (6)

However, the decreases are limited and the NO oxidation efficiency can still reach 90% in pre-sence of SO2 when n[O3]/n[NO] is 1.0. Besides, the SO2 oxidation efficiency is only about 5% during the experiment, it can be concluded that the present ozonizing method can successfully be used for the oxidation of NO, the coexistence of SO2 has little impact on oxidation of NO. A similar result was rep-orted by Wang et al. at a lower temperature 373 K [18].

It can be seen from Figure 3 that the SO2 rem-oval efficiency is always nearly 99%, which indicates that the SO2 absorption in to ammonia solution is almost unaffected by the existing of NOx. NOx removal efficiency increases from 62.1 to 86.7% as O3/NO mole ratio increases from 0.5 to 1.5. Obvi-ously, the increasing of O3/NO molar ratio is favorable to the absorption of NOx. The increasing molar ratio of O3/NO increased NOx oxidation rate and accelerated NOx dissolution, which was helpful for the NOx rem-oval [23,24]:

→2 2 3 4 33NO +H O + 2NH 2NH NO + NO (7)

→2 2 3 4 2NO + NO + H O +2NH 2NH NO (8)

→2- + - 2-2 3 2 2 42NO +SO +H O 2H + 2NO +SO (9)

→- + - 2-2 3 2 2 42NO +HSO +H O 3H + 2NO +SO (10)

Figure 3. The influence of O3/NO molar ratio on SO2 and NOx removal efficiency (flow rate: 0.95 Nm3/h; initial SO2 con-

centration: 2000 mg/Nm3; initial NO concentration: 500 mg/Nm3; ammonia concentration: 0.3%; oxygen content of the simulated flue gas: 12%; oxidation reactor temperature: 423 K; absorption

reaction temperature: 298 K; absorption solution pH value: 10.0).

Additionally, the concentration of SO42– and NO3

– in the solution was tested at pH 5.5. The content of SO4

2– goes up from 53.3 to 76.4% as the O3/NO mole ratio increases from 0.5 to 1.5 and the content of NO3

– keeps a relatively stable level around 10%. More NO2 can be generated as the molar ratio of O3/NO grow, the reactions (9) and (10) will be promoted to produce more SO4

2–. In ammonia-based wet flue gas desulfur-ization process, a great amount of air is needed to oxidize (NH4)2SO3 and NH4HSO3 into (NH4)2SO4, the byproduct. Due to this cause, increasing the O3/NO molar ratio, the energy consumption of air supply will be greatly reduced in the ammonia-based simul-taneously desulfurization and denitrification process.

Influence of initial NO concentration

Figure 4 shows the effect of initial NO concen-tration on the ozonation process. The increase of the initial NO concentration results in increasing NO oxi-dation efficiency, because higher initial NO concen-tration can speed up the oxidation rate when the reac-tions begin. When the mole ratios of O3/NO are 1.0 and 1.2, the NO oxidation efficiency rises to over 90% and keeps steady in spite of initial NO concentration changing. As long as controlling the mole ratio of O3/NO above 1.0, a favorable oxidation effect can be obtained regardless of initial NO concentration. This


308

result is very helpful for practical engineering appli-cations.

Figure 4. The influence of initial NO concentration on NO oxidation efficiency (flow rate: 0.95 N m3/h; initial SO2 concentration: 2000 mg/Nm3; oxygen content of the

simulated flue gas: 12%; reaction temperature: 423 K).

The experimental results indicating the influence of initial NO concentration on SO2 and NOx removal efficiency are shown in Figure 5. The SO2 removal efficiency increases from 99.34 to 99.86% when initial NO concentration increases from 200 to 800 mg/N m3 and the pH value of absorption solution was 10.0. The possible reason is that an appropriate increase in initial NO concentration promotes the reaction between NO2 and SO3

2–, but the increase of desulfur-ization efficiency is very limited, which indicates the reactions (11)-(12) are the main reactions for SO2 removal. As NOx concentration increases, the reac-tion contact time and the chemical equivalent of abs-orbent gradually cannot meet the requirements of NOx absorption, the NOx removal efficiency declines form 90.01 to 71.95%. However, Wei et al. [5] draw a con-clusion that NOx concentration has little effect on denitrification when they used microwave reactor with NH4HCO3 and zeolite to study the simultaneous rem-oval of SO2 and NOx from flue gas. The possible reason may be that the chemical equivalent of abs-orbent has little change in the range of initial NO concentration from 120 to 200 mg/N m3. Jin et al. [6] used aqueous chlorine dioxide solution as absorbent and NOx removal efficiency only reached 66-72%, which proves ozone is more efficient than chlorine dioxide:

→3 2 2 4 2 32NH + H O + SO (NH ) SO (11)

→4 2 3 2 2 4 3(NH ) SO + SO +H O 2NH HSO (12)

The content of SO42- in absorption solution

increases from 60.69 to 73.01% at pH 5.5 as the NO initial concentration increases from 200 to 800 mg/N m3. Though the denitrification efficiency declines with the increase of NO concentration, more amount of NOx in quantity is absorbed into solution which can boost the reactions 9 and 10, resulting in the inc-reases of the SO4

2– content in absorption solutions.

Figure 5. The influence of initial NO concentration on SO2 and NOx removal efficiency (flow rate: 0.95 Nm3/h; n[O3]/n[NO]:

1.0; initial SO2 concentration: 2000 mg/Nm3; ammonia concentration: 0.3%; oxygen content of the simulated flue gas: 12%; oxidation reactor temperature: 423 K; absorption reaction

temperature: 298 K; absorption solution pH value:10.0).

Influence of initial SO2 concentration

The influence of initial SO2 concentration on NO oxidation efficiency was investigated by varying initial SO2 concentration from 1000 to 3000 mg/N m3. As can be seen in Figure 6, the increasing of initial SO2 concentration has a slight inhibition on the oxidation of NO with the NO oxidation efficiencies declining from 96.4 to 90.5%. The oxidation efficiencies of SO2 always maintain at around 5% under different initial SO2 concentrations, which means that the oxidation reaction between SO2 and O3 is extremely weak in comparison with the oxidation of NO by ozone in system. Combined with the previous results, we can draw a conclusion that the appearance of SO2 and the concentration changes of NO and SO2 have little impact on the oxidation of NO, the O3/NO mole ratio is the most important factor in the ozone oxidation process for desulfurization and denitrification, keeping the O3/NO mole ratio above 1.0, the oxidation effi-ciency of NO can achieve over 90% within the expe-riment conditions.

Figure 7 illustrates the influence of initial SO2 concentration on SO2 and NOx removal efficiency. It is evident that the desulfurization efficiencies under dif-


309

ferent initial SO2 concentrations are all close to 100%, the denitrification efficiency increases with rising initial SO2 concentrations. Fang et al. [10] draw an opposite conclusion that NOx removal efficiency sharply decrease with the increment of SO2 concentration. The reason may be that the increase of initial SO2 concentration can generate more SO3

2– in absorption solution, reactions (9) and (10) are promoted and more NO2 will be assimilated. An appropriate inc-rease of SO2 concentration does not cause the abs-orption capacity to decline sharply in our experimental conditions. This is why an appropriate increase of initial SO2 concentration can contribute to the NOx absorption.

Figure 6. The influence of initial SO2 concentration on NO oxidation efficiency (flow rate: 0.95 Nm3/h; initial NO

concentration: 500 mg/N m3; n[O3]/n[NO]: 1.0; oxygen content of the simulated flue gas: 12%; reaction temperature: 423 K).

Figure 7. The influence of initial SO2 concentration on SO2 and NOx removal efficiency (flow rate: 0.95 Nm3/h; n[O3]/n[NO]: 1.0; initial NO concentration: 500 mg/Nm3; ammonia concentration: 0.3%; oxygen content of the simulated flue gas: 12%; oxidation reactor temperature: 423 K; absorption reaction temperature:

298 K; absorption solution pH value: 10.0).

CONCLUSIONS

In this paper, simultaneous removal of SO2 and NOx with ozone as oxidant and ammonia as absorber was investigated. In the ozonation part, the appear-ance of SO2 and the concentration changes of NO and SO2 have little impact on the oxidation of NO, the O3/NO mole ratio is the most important factor in the ozone oxidation process. The oxidation efficiency of NO can achieve over 90% within the experiment con-ditions and the conversion of NO is almost complete when the mole ratio of O3/NO is 1.0. In the absorption part, the SO2 absorption into ammonia is almost unaf-fected by the existing of NOx, an appropriate increase of initial SO2 concentration can contribute to the NOx absorption. The removal efficiency of SO2 and NOx was about 99.34 and 90.01% at pH 10, flow rate 0.95 N m3/h, n[O3]/n[NO] 1.0, initial SO2 concentration 2000 mg/N m3, initial NO concentration 200 mg/N m3, ammonia concentration 0.3%, oxygen content of the simulated flue gas 12%, oxidation reactor tempera-ture 423 K and absorption reaction temperature 298 K. Increases of O3/NO mole ratio and initial SO2 con-centration could elevate the content of SO4

2– in abs-orption solutions, the energy consumption of air sup-ply for oxidation of SO3

2– will be greatly reduced.

Acknowledgement

The work was supported by Sinopec Ningbo Engineering Co. Ltd.

REFERENCES

[1] Z.G. Shen, X. Chen, M. Tong, S.P. Guo, M.J. Ni, J. Lu, Fuel 105 (2013) 578-584

[2] S.P. Guo, X. Chen, M. Tong, Z.G. Shen, Y.B. Zhou, R. Zhang, J. Lu, Asian J. Chem. 25 (2013) 5381-5384

[3] J.H. Ye, J. Shang, H. Song, Q. Li, T. Zhu, Chem. Eng. J. 232 (2013) 26-33

[4] J.S. Chang, K. Urashima, Y.X. Tong, W.P. Liu, H.Y. Wei, F.M. Yang, X.J. Liu, J. Electrostat. 57 (2003) 313-323

[5] Z.S. Wei, Z.H. Lin, H.J. Niu, H.M. He, Y.F. Ji, J. Hazard. Mater. 162 (2009) 837-841

[6] D.S. Jin, B.R. Deshwal, Y.S. Park, H.K. Lee, J. Hazard. Mater. 135 (2006) 412-417

[7] Y. Zhao, T.X. Guo, Z.Y. Chen, Y.R. Du, Chem. Eng. J. 160 (2010) 42-47

[8] Y. Xie, Y. Chen, Y.G. Ma, Z.L. Jin, J. Hazard. Mater. 195 (2011) 223-229

[9] Y. Zhao, Y.H. Han, T.Z. Ma, T.X. Guo, Environ. Sci. Technol. 45 (2011) 4060-40645

[10] P. Fang, C.P. Cen, X.M. Wang, Z.J. Tang, Z.X. Tang, D.S. Chen, Fuel Process. Technol. 106 (2013) 645-653

[11] W.Y. Sun, S.L. Ding, S.S. Zeng, S.J. Su, W.J. Jiang, J. Hazard. Mater. 192 (2011) 124-130


310

[12] W.Y. Sun, Q.Y. Wang, S.L. Ding, S.J. Su, J. Hazard. Mater. 228 (2013) 700-707

[13] I. Liémans, D. Thomas, Energy Procedia 37 (2013) 1348- -1356

[14] W. Nimmo, A.A. Patsias, E. Hampartsoumian, B.M. Gibbs, P.T. Williams, Fuel 83 (2004) 149-155

[15] J. Ding, Q. Zhong, S.L. Zhang, F.J. Song, Y.F. Bu, Chem. Eng. J. 243 (2014) 176-182

[16] C.L. Sun, N. Zhao, Z.K. Zhuang, H.Q. Wang, Y. Liu, X.L. Weng, Z.B. Wu, J. Hazard. Mater. 274 (2014) 376-383

[17] K. Skalska, J.S. Miller, S. Ledakowicz, Chem. Eng. Pro-cess. 61 (2012) 69-74

[18] Z.H. Wang, J.H. Zhou, Y.Q. Zhu, Z.C. Wen, J.Z. Liu, K.F. Cen, Fuel Process. Technol. 88 (2007) 817-823

[19] Y.S. Mok, H.J. Lee, Fuel Process. Technol. 87 (2006) 591-597

[20] S.P. Guo, J.H. Wang, X. Chen, J. Zhang, R. Zhang, J. Lu, Asian J. Chem. 26 (2014) 69-74

[21] X. Gao, H.L. Ding, Z.L. Wu, Z. Du, Z.Y. Luo, K.F. Cen, Energy Fuels 23 (2009) 5916-5919

[22] X. Gao, Z. Du, H.L. Ding, Z.L. Wu, H. Lu, Z.Y. Luo, K.F. Cen, Energy Fuels 24 (2010) 5876-5882

[23] Y. Jia, D.Q. Du, X.X. Zhang, X.L. Ding, Q. Zhong, Korean J. Chem. Eng. 30 (2013) 1735-1740

[24] X. Gao, H.L. Ding, Z. Du, Z.L. Wu, M.X. Fang, Z.Y. Luo, K.F. Cen, Appl. Energ. 87 (2010) 2647-2651.

SHAOPENG GUO1

LINA LV1

JIA ZHANG1

XIN CHEN2

MING TONG2

WANZHONG KANG2

YANBO ZHOU1

JUN LU1

1Key Laboratory of Coal Gasification and Energy Chemical Engineering of

Ministry of Education, East China University of Science & Technology,

Shanghai, P. R. China 2SINOPEC Ningbo Engineering Co.,

Ltd., Ningbo, P. R. China

NAUČNI RAD

ISTOVREMENO UKLANJANJE SO2 I NOX POMOĆU AMONIJAKA I OKSIDACIJE NO U GASNOJ FAZI KORIŠĆENJEM OZONA

Predložen je proces za istovremeno odsumporavanje i denitrifikaciju sa ozonom kao

oksidacionim sredstvom za NO i amonijačnim rastvorom kao apsorbentom. Rezultati

pokazuju da prisustvo SO2 i promene koncentracija NO i SO2 malo utiču na oksidacije NO.

Efikasnost oksidacije NO može iznositi i preko 90 % pri molskom odnosu O3/NO = 1,0.

Prisustvo NOx neznatno utiče na na apsorpciju SO2. Odgovarajuće povećanje koncen-

tracije SO2 dovodi do bolje apsorpcije NOx. U eksperimentalnom sistemu efikasnost ukla-

njanja SO2 i NOx dostiže vrednost 99,34 i 90,01%, redom, pri pH 10, protoku 0,95 N m3/h,

n[O3]/n[NO] 1,0, početnoj koncentraciji SO2 2000 mg/N m3, početnoj koncentraciji NO 200

mg/N m3, koncentraciji amonijaka 0,3%, sadržaju kiseonika u modelu dimnog gasa 12%,

temperaturi reakcije oksidacije 423 K i temperaturi reakcije apsorpcije 298 K.

Ključne reči: Ozon, azotovi oksidi, amonijak, istovremena apsorpcija.





311

JAU-KAI WANG JIR-MING CHAR

Department of Applied Chemistry & Material Science, Fooyin

University, Ta-Liao Hsiang, Kaohsiung City, Taiwan, R.O.C.

SCIENTIFIC PAPER

UDC 544.6:546.59:62

DOI 10.2298/CICEQ140618029G

OPTIMIZATION STUDY ON HARDNESS OF GOLD FILM THROUGH SUPERCRITICAL ELECTROPLATING PROCESS BY RESPONSE SURFACE METHODOLOGY

Article Highlights • The nanometer size gold film was developed by supercritical electroplating • The hardness of deposited gold film can be for industrial application of soft and hard

gold • A theoretical approach was examined by statistical experimental method Abstract

A non-cyanide gold bath has been used to deposit gold film on a brass substrate through electroplating process using supercritical carbon dioxide emulsion. The hardness of the deposited gold film was considered as a response variable to optimize the process parameters of electroplating oper-ation by statistical experimental methods. Effects of current density, pressure temperature, and chemical composition of the solution were investigated to select the optimal operation factors. Scanning electron microscopy and micro--hardness testing were applied to determine the characteristics of the metallic film. The screening of significant variables was examined by a 25-1 fractional factorial design with V resolution method. The experimental results showed that the significant variables affecting the deposition of gold film were current density, pressure and temperature. Based on Box-Behnken design and res-ponse surface methodology (RSM), a regression model was built by fitting the experimental results with a polynomial equation. The optimal operating vari-able conditions can be searched at a specified hardness for industrial hard and soft gold application ranged from 83.8 to 157.7 HV.

Keywords: electroplating; gold; optimization; statistical experimental method.

Due to its remarkable characteristics in terms of chemical and electrical properties, electroplated gold classified into soft gold and hard gold has been widely used in the electronics industry [1]. Hard gold is used on electrical connectors and contacts requiring resist-ance to mechanical wear as well as low electrical contact resistance. On the other hand, the gold applied for bump must be sufficiently soft so that it is easily deformable to accommodate small variations in thickness [2]. However, a common problem exists during electroplating, which is that the electric current

Correspondence: J.-K. Wang Department of Applied Chemistry & Material Science, Fooyin University, 151 Chin-Hsueh Rd, Ta-Liao Hsiang, Kaohsiung City, 831 Taiwan, R.O.C. E-mail: [email protected] Paper received: 16 December, 2013 Paper revised: 27 July, 2014 Paper accepted: 28 August, 2014

also causes the dissociation of water in addition to the electrolysis of metal ions, resulting in hydrogen to be released at the cathode. The formation of hydrogen may create several defects of deposited gold film owing to the pinholes effect. This is an important problem for industrial application that has recently been investigated by other researchers [3-4].

Supercritical carbon dioxide (Sc-CO2) has rec-eived much attention as an alternative to harmful organic solvents used for extraction, separation, reac-tion, and for many other processes. The low viscosity, high diffusivity and zero surface tension of Sc-CO2 has been exploited in a variety of impregnation pro-cesses [5]. Recently, plating technology with Sc-CO2 has attracted special attention because Sc-CO2, in particular, can transport the solute into fine nano-meter-space of the materials and clean even integ-

J.-K. WANG, J.-M. CHAR: OPTIMIZATION STUDY ON HARDNESS OF GOLD… Chem. Ind. Chem. Eng. Q. 21 (2) 311−317 (2015)

312

rated circuits without shrinking or causing other harm due to the interface tension that exists between liquids and gases. Electroplated films obtained in emulsions composed of Sc-CO2 and electroplating solutions have a uniformity and hardness, superior to those of films obtained using conventional electro-plating methods [6]. While the several advantages of this new technique in plating have been well des-cribed, many studies have attempted to explain the mechanism by which electroplating within Sc-CO2 creates such excellent films [7]. However, in gold-con-sumed industrial processes, the operation cost of electroplating system is of critical importance because expensive gold chemicals affect bath formulation and volumetric productivity. A combination of variables generating a certain optimum response can be ident-ified through factorial design and the use of regres-sion methodology [8]. This pattern is designed by using statistical methods to yield the most information by a minimum number of experiments.

In this work, a non-cyanide system for gold elec-troplating was designed to investigate whether nano-meter-scaled gold film could be achieved using super-critical carbon dioxide processing. Furthermore, the surface response method combined with Box-Behn-ken design was applied to deal with the preparatory conditions of supercritical electroplating process in order to obtain suitable hardness of metal films. The aim of this paper was also to elucidate a simple model of electroplating conditions to control the hard-ness of gold film by supercritical electroplating pro-cess through statistical experimental method.


Materials

The electroplating solution, which is usually referred to as a non-cyanide gold plating bath, con-sisted of sodium gold sulfite (Na3Au[SO3]2, 0.08 mol/L), ammonium sulfite ([NH4]2SO3, 0.5mol/L), diso-dium dihdrogenethylenediamine tetraacetate, (0.01 mol/L) and potassium oxalate, (K2C2O4, 0.01 mol/L). All chemical reagents used in this work with a min-imum purity of 99.9% were purchased from Toshin Yuka Kogyo Co., Ltd. Carbon dioxide with a minimum purity of 99.9% was purchased from Yun Shan Co. Ltd. A non-ionic block copolymer - poly(ethylene oxide)-poly(propylene oxide) (HO(CH2CH2O)100CH3

(CHCH2O)30 (CH2CH2O)100H, 0.001 mol/L), PEOPPO was obtained from Serva AG, Heidelberg, Germany and employed as a surfactant in our experiments. The anode was a 99.99% purity platinum plate with a size of 20×20 mm2 and the cathode was a brass substrate of the same size. The brass substrate, which was composed of 65.4% of Cu and 34.6% of Zn, had a Vickers hardness value of 112.5 HV. Before the plat-ing reaction, both the anode and the brass substrate were degreased by dipping successively in a 10 wt.% NaOH and a 20 wt.% HCl and rinsing in de-ionized water.

Experimental apparatus

A high-pressure experimental apparatus were fabricated by ourselves and its outline was shown in Figure 1, was used for electroplating. The tempera-ture variation of each run was observed to be less than 1.0 °C. The maximum working temperature and

Figure 1. Schematic representation of the apparatus used for a batch electroplating reaction with emulsion of a CO2, surfactant and the

electroplating solution. The parts of the apparatus are labeled as follows: a) CO2 cylinder; b) cooler; c) high pressure pump; d) temperature-controlled air bath; e) reactor with magnetic stirrer; f) trap; g) programmable power supply; h) gas meter;

BPR: back-pressure regulator; PI: pressure indicator; TI: temperature indicator; V: valve.


313

the maximum pressure were 90 °C and 250 bar, respectively. The integrated electroplating cell that had a volume of 200 ml was a stainless steel 316 vessel in a temperature-controlled air bath with an agitator. Both the anode and the cathode were attached using platinum wires to the reactor and were connected to a programmable power-supply; model YPP15030, manufactured by Yamamoto-ms Co., Ltd. A typical electroplating reaction was performed in a constantly agitated ternary system of Sc-CO2, the electroplating solution and a surfactant. The 100 ml gold electroplating solution and the surfactant both were put in a high-pressure cell. CO2 was introduced to the high-pressure cell using a pump and pres-surized to a predetermined pressure. The ternary sys-tem was then constantly agitated using a cross-mag-netic stirrer bar at a speed of 400 rpm under a desir-able constant temperature. The bulk electroplating solution commenced after 30 min of agitation and the entire electroplating reaction for each run was carried out for same amount of electric charge at various operation conditions. Based on Faraday’s law, the electroplating time was executed from 0.5 to 1.5 h depending on the conditions of current density and the thickness deposited gold film was obtained around 5.0±0.3 µm at the same quantity of electricity.

Analysis

The microscopic images of gold deposited film were obtained using a Hitachi S-4700I High-Resol-ution Scanning Electron Microscope & Energy Dis-persive spectrometer. The surface features and aver-age surface roughness values (Ra) were measured using atomic force microscopy (Veeco CP-II, Thermo-Microscopes Co. Ltd, USA). Micro-hardness values were measured using an AKASHI Vickers AVK-C2

hardness machine, with a weight of 50 g. For each sample, considering that the gold films from both the front and back of the brass substrate were also given, each hardness value is actually based on 15 mea-surements [9].


Surface observation

Gold electroplating was operated at a current density of 0.3 A/dm2 in a constantly agitated ternary system of Sc-CO2 (100 ml), and the electroplating solution (100 ml) and the surfactant (0.01 mol/L to the electroplating solution) under conditions 55 °C and 101 bar. The SEM analysis in Figure 2 showed that the size of grains in deposited gold film taken from supercritical plating is in the range of 100 nm. There-fore, a nanometer order electroplating is possible by a hybrid of traditional direct current (DC) electroplating and Sc-CO2 techniques with an emulsion. It involves the electrochemical reaction in emulsions formed by Sc-CO2 and aqueous electrolyte with surfactant PEOPPO. The emulsion particle size ranges typically from several nanometers to several millimeters and can be controlled with surfactants to within a relatively narrow size distribution [10]. When the electroplating solution comes in contact with the cathode, the nuc-leation and the crystal growth can occur. However, when the Sc-CO2 comes in contact with the cathode, the nucleation and the crystal growth cannot occur [11]. This result indicates that Sc-CO2 plays an impor-tant role in increasing the quality of plated film higher because Sc-CO2 electroplating solution emulsion has performed characteristics with a higher diffusion coef-ficient, lower viscosity and the surface tension. This makes the deposited atoms on the cathode move

Figure 2. SEM analysis of deposited gold film obtained at 55 °C, 0.3 A/dm2 and 101 bar.


314

more easily into the depressions of place than when using only electroplating solution; thus, films of excel-lent characteristics were obtained [12].

In addition, the average hardness of deposited gold films was 125.2 HV by supercritical electroplat-ing, which is much higher than the hardness 85.6 HV of the gold film produced from the electroplating solution only. Since the hardness value of this plated film in the emulsion has been increased over 50%, there is a possibility that an impurity could have been included in the gold matrix and influence the metal structure. In order to confirm that the plated film is composed of pure gold, we performed measurements using an energy dispersive spectrometer of as-dep-osited 5 µm thick coating made from the emulsion of supercritical CO2 at 101 bar, 55 °C and 0.3 A/dm2 (Figure 3). It can be observed from Figure 3 that the purity of deposited gold film is over 99.9%. The studies on the morphology of deposited gold film per-formed by other investigators [13-14] show that the grain sizes caused by the inhibition of crystal growth as well as inclusions with incorporated impurities play the most important role in determining critical physical properties of hard gold. Thus, we could indicate that the higher hardness of the plated film produced by our method originated from the small grains of the gold and the small grain strengthening that occurs in our system.

Screening the significant variables

Since various parameters potentially affect the supercritical plating process [15], the optimization of experimental conditions represents a critical step in the development of a supercritical gold plating method for hard and soft gold in industrial application. These parameters make it difficult to select the required conditions for subsequent reliable quantif-ications. Several studies have been conducted rec-ently regarding electroplating technology using dense

CO2, such as a supercritical fluid transported chem-ical deposition within a nanometer-scale casting. Electrochemical studies on Sc-CO2 have also been reviewed by other investigators in the past few years [16]. However, there have been no reports within the literature of Sc-CO2 being applied to their practical applications, because of complex of theory and mech-anism. In order to have a better understanding on the role of each process parameter, interactions among process parameters and optimization of process para-meters as well as responses, a statistical analysis is essential. In order to clarify the hardness values of deposited films, an experimental statistical method was applied to designs and analysis of experimental results to deal the preparatory conditions of nano-meter-sized gold film from supercritical electroplating process.

In this study, Design Expert software (version 7.1, Stat-Ease Inc., Minneapolis, MN, USA, 2008) was used for experimental designs and analysis of experimental results. Response surface methodology (RSM) is an empirical modeling technique used to evaluate the relationship between a set of the con-trolled experimental variables and the measured res-ponses. A prior knowledge and understanding of the process and process variables under investigation are necessary for achieving a more realistic model. The effect of five parameters including current density (X1), pressure (X2), temperature (X3), the concentra-tion of gold (X4) and surfactant (X5) were investigated using a 25−1 resolution V fractional factorial design to determine the significant variables that affected hard-ness of deposited gold film. Each variable had two levels to be examined for a high (+1) and low level (−1). The high and low levels selected represented the extremes of normal operating ranges. Table 1 shows the variables and levels in detail.

In variable screening, a 25-1 fractional factorial design (FFD) with V resolution was applied to test the

Figure 3. EDS analysis of deposited gold film obtained at 55 ºC, 0.3 A/dm2 and 101 bar.


315

significant variables. The screen experimental design and results are shown in Table 2. Depending on the operating conditions, the hardness of gold film varied from 81.8 to 158.8 HV. Results from analysis of variance are shown in Table 3. Based on experi-mental statistical analysis for F-test method, it can be seen from Table 3 that the model F-value of 43.64 implies the model is significant for quality of deposited gold films navigating the design space. In this case the values of “Prob > F” less than 0.100 indicate the model terms are significant, it is proven that current density, pressure and temperature are significant model terms, whereas the concentration of gold and surfactant did not exert significantly effect on hard-ness of deposited gold film because the values of “Prob > F” greater than 0.100. The three significant variables – current density, temperature and pressure – will be further investigated for the following opti-mization process.

Table 2. 25-1 fractional factorial design and experimental results

Trial no.

Variables Hardness, HV

X1 X2 X3 X4 X5

1 -1 -1 -1 -1 1 103.3

2 1 -1 -1 -1 -1 94.4

3 -1 1 -1 -1 -1 157.5

4 1 1 -1 -1 1 130.9

5 -1 -1 1 -1 -1 88.5

6 1 -1 1 -1 1 83.4

7 -1 1 1 -1 1 138.1

8 1 1 1 -1 -1 120.9

9 -1 -1 -1 1 -1 90.7

10 1 -1 -1 1 1 81.8

11 -1 1 -1 1 1 158.8

12 1 1 -1 1 -1 133.3

13 -1 -1 1 1 1 93.3

14 1 -1 1 1 -1 82.4

15 -1 1 1 1 -1 143.2

16 1 1 1 1 1 122.6

Table 3. Analysis of variance for screening experiments

Source Sum of squares

Degree of freedom

Mean square

F-Value Prob.>F

Model 10597.89 5 2119.58 43.64 <0.0001

X1 948.64 1 948.64 19.53 0.0013

X2 9273.69 1 9273.69 190.93 <0.0001

X3 368.64 1 368.64 7.59 0.0203

X4 6.76 1 6.76 0.14 0.7169

X5 0.16 1 0.16 0.0033 0.9554

Residual 485.70 10 48.57

Cor Total 11083.59 15

Optimization

Based on Box-Behnken design for 3 variables, a set of 17 experiments was carried out [18] and expe-rimental results are shown in Table 4. The experi-mental results were analyzed using statistical methods appropriate to the experimental design used. Design Expert 7.1 was used to analyze the experimental results. According to the RSM methodology, a poly-nomial model was used to fit the independent vari-ables using the following equation:

= + + + ++ + +

0 1 1 2 2 3 3

12 1 2 13 1 2 23 2 3

Y B B x B x B x

B x x B x x B x x (1)

where Y is the response (hardness of deposited gold film), xi are the variables, B0 is the constant coef-ficient, Bi and Bii refer to the coefficients of linear and interaction terms.

Table 4. Box-Behnken design and experimental results

Trial no. Variable Hardness, HV

X1 X2 X3 Measured Predicted

1 -1 -1 0 93.3 88.7

2 1 -1 0 83.8 87.1

3 -1 1 0 157.7 155.7

4 1 1 0 123.5 129.9

5 -1 0 -1 120.2 123.1

6 1 0 -1 125.2 120.0

7 -1 0 1 119.7 121.2

8 1 0 1 103.2 99.1

9 0 -1 -1 90.8 89.6

10 0 1 -1 157.3 153.4

11 0 -1 1 85.6 86.1

12 0 1 1 135.5 132.3

13 0 0 0 115 115.4

14 0 0 0 115 115.4

15 0 0 0 115 115.4

16 0 0 0 114 115.4

17 0 0 0 113 115.4

Table 1. Variables and their levels for 25-1 fractional factorial design

Variable Code level

-1 0 +1

X1: Current density, A/dm2 0.1 0.3 0.5

X2: Pressure, bar 1.0 101 201

X3: Temperature, °C 45 55 65

X4: Gold ion conc., mol/L 0.06 0.08 0.10

X5: Surfactant conc., mol/L 0.01 0.02 0.03


316

In the study, the hardness of deposited gold film data was processed for Eq. (1) including ANOVA to obtain the interaction between the process variables and the response. The quality of the fit of polynomial model was expressed by the coefficient of deter-mination, R2, and statistical significance was checked by the F-test in the program [19-20]. The regression analysis gave the following regression model:

= − + −− − + −

1 2

3 1 2 1 3 2 3

119.92 10.85 19.23

7.38 8.90 0.75 3.30

Y x x

x x x x x x x (2)

where x1 is the coded variable of current density, x2 is the coded variable of pressure and x3 is the coded variable of temperature.

The R2 (determination coefficient) of the regres-sion equation obtained from analysis of variance is 0.9260 (a value > 0.75 indicates aptness of the model), which means that the model can explain 92.60% variation in the response. Compared the measured and predicted hardness of deposited gold film in Table 4, it can be observed that most of the standard residuals should lie in the interval of 2.55 with respect to its observed response. Therefore, the predicting response surface equation confirms that the equation gives a reasonable fitting to the experi-mentally observed data.

Based on analyzing the Eq. (2) with statistical experimental method, the optimal condition for deposited gold film should be obtained for industrial application of hard and soft gold navigating the design space. For example, the operation condition should be controlled at 128 bar, 54.3 °C and 0.102 A/dm2 if the required hardness of hard gold application needs to be 125 HV. If the designed hardness of gold film was 90 HV for soft gold application, the optimal con-dition could be executed at 75.2 bar, 52.3 °C and 0.44 A/dm2. Confirmation experiments were performed and the results showed that the difference between the predicted values and the measured values is within 15%. Statistical optimization method overcomes the limitations of classic empirical methods and is proven to be a powerful tool for the optimization of gold film deposition.

CONCLUSIONS

High quality gold films with fine grains have been developed using a new electroplating method involving the emulsion of a supercritical carbon dioxide, an electroplating solution and a surfactant, PEOPPO. The films plated using this method have a uniformity of the surface and Vickers hardness better than the results obtained by using conventional elec-troplating methods. Analysis of the composite mater-

ials by SEM allowed the measurement of grain size in neighborhood of 100 nm. It is also highlighted that the developed medium showed a nanometer-sized gold film with hardness ranging from 83.8 to 157.7 HV has been obtained for industrial application of soft and hard gold by combining the optimal settings of those variables. Data from the present investigation have shown that gold film deposition is dependent mainly on current density, pressure and temperature. On the basis of Box-Behnken design, using response surface methodology, a theoretical approach calculated from numerical calculation is in agreement with the expe-rimental data.

Acknowledgment

The authors sincerely appreciate the financial support of the Nation Science Council of the Republic of China (99-2221-E-242 -007) for this work.

REFERENCES

[1] A. Cabanas, D.P. Long, J.J. Watkins, Chem. Mater. 16 (2004) 2028-2033

[2] T. Osaka, A. Kodera, T. Misato, T. Homma, Y. Okinaka, O. Yoshioka, J. Electrochem. Soc. 144 (1997) 3462-3469

[3] M.M. Rahman, M. Sone, H. Uchiyama, M. Sakurai, S. Miyata, T. Nagai, Y. Higo, H. Kameyama, Surf. Coat. Tech. 201 (2007) 7513-7518

[4] H. Yoshida, M. Sone, H. Wakabayashi, H. Yan, K. Abe, X.T. Tao, A. Mizushima, S. Ichihara, S. Miyata, Thin Solid Films 446 (2004) 194-199

[5] S.-T. Chung, W.-T. Tsai, J. Electrochem. Soc. 156 (2009) D457-D461

[6] R. Khanum, T.F.M. Chang, T. Sato, M. Sone, ECS Electrochem. Lett. 2 (2013) D43-D44

[7] D. Kim, J. Kim, G.L. Wang, C.C. Lee, Mater. Sci. Eng., A 393 (2005) 315-319

[8] M.-S. Kim, J.-Y. Kim, C.-K. Kim, Chemosphere 58 ( 2005) 459-465

[9] A. Mizushima, M. Sone, H. Yan, T. Nagai, K. Shigehara, S. Ichihara, S. Miyata, Surf. Coat. Tech. 194 (2005) 149- -156

[10] S. Sonsuzer, S. Sahin, L. Yilmaz, J. Supercrit. Fluid. 30 (2004) 189-2147

[11] Y.G. Li, W. Chrzanowski, A. Lasia, J. Appl. Electrochem. 26 (1996) 843-852

[12] B. Bozzini, P.L. Cavallotti, G. Giovannelli, Metal Finish. 4 (2002) 50-60

[13] C.C. Lo, J.A. Augis and M.R. Pinnel, J. Appl. Phys. 50 (1979) 6887

[14] V.L.D.L. Santos, D. Lee, J. Seo, F.L. Leon, D.A. Busta-mante, S. Suzuki, Y. Majima, T. Mitrelias, A. Ionescu, C. H.W. Barnes, Surface Sci. 603 (2009) 2978–2985

[15] B. Wong, S. Yoda, S.M. Howdle, J. Supercrit. Fluid. 42 (2007) 282-287


317

[16] Y.G. Li, A. Lasia, J. Appl. Electrochem. 26 (1996) 853- -863

[17] H. Yoshida, M. Sone, A. Mizushima, K. Abe, X. T. Tao, S. Ichihara, Chem. Lett. 11 (2002) 1086-1087

[18] A. Berenjian, N. L.-C. Chan, R. Mahanama, A. Talbot, H. Regtop, J. Kavanagh, F. Dehghani, Mol. Biotechnol. 54 (2013) 371-378

[19] Y. Bayat, S. M. Pourmortazavi, J. Supercrit. Fluid. 72 (2012) 248-254

[20] B.K. K¨orbahti, M.A. Rauf, Chem. Eng. J. 136 (2008) 25- -30.

JAU-KAI WANG

JIR-MING CHAR

Department of Applied Chemistry & Material Science, Fooyin University,

Ta-Liao Hsiang, Kaohsiung City, Taiwan, R.O.C.

NAUČNI RAD

OPTIMIZACIJA TVRDOĆE FILMA OD ZLATA DOBIJENOG SUPERKTIČNIM GALVANOTEHNIČKIM PROCESOM METODOM ODZIVNE POVRŠINE

U radu je korišćeno necijanidno zlatno kupatilo za galvanotehničko dobijanje zlatnog filma

na supstratu od mesinga koristeći emulziju superkritičnog ugljen dioksida. Tvrdoća

nanešenog zlatnog filma, kao zavisne promenljive, iskorišćena je u optimizaciji procesnih

galvanotehničkih parametara statističkim eksperimentalnim metodama. U cilju nalaženja

optimalnih uslova analiziran je efekat gustine struje, pritiska, temperature i sastava

rastvora. Za određivanje karakteristike metalnih filmova korišćeni su elektronski mikroskop

i mikro ispitivač tvrdoće. Ovo ocenjivanje značajnih promenljivih je izvršeno u skaldu sa

parcijalnim faktorijelnim planom 25-1 i V rezolucionom metodom. Eksperimentalni rezultat

pokazuje da su značajne promenljive koje imaju uticaj na nanošenje zlatnog filma: gustina

struje, pritisak i temperatura. Korišćenjem Box–Behnken dizajna i metode odzivne povr-

šine, razvijen je regresioni model fitovanjem eksperimentalnih rezultata sa polinomnom

jednačinom. Na osnovu njega moguće je određivanje optimalnih operativnih uslova za

definisanu tvrdoću u opsegu od 83,8 do 157,7 HV za meke zlatne filmove i tvrde

industrijske filmove.

Ključne reči: galvanotehnički proces, zlato, optimizacija, statistička eksperi-mentalna metoda.





319

LAWRENCE KOECH1

RAY EVERSON2

HEIN NEOMAGUS2

HILARY RUTTO1 1Department of Chemical

Engineering, Vanderbijlpark Campus, Vaal University Of

Technology, Vanderbijlpark, South Africa

2Department of Chemical and Minerals Engineering,

Potchefstroom Campus, North West University, Potchefstroom,

South Africa

SCIENTIFIC PAPER

UDC 66.094.522.091.8:662.613.1

DOI 10.2298/CICEQ140423032K

DISSOLUTION KINETICS OF SOUTH AFRICAN COAL FLY ASH AND THE DEVELOPMENT OF A SEMI-EMPIRICAL MODEL TO PREDICT DISSOLUTION

Article Highlights • Dissolution kinetics of fly ash using adipic acid at constant pH was investigated • A semi-empirical model was developed to predict dissolution of fly ash • The sorbent before and after dissolution was characterized XRF, FTIR, BET surface

area and SEM Abstract

Wet flue gas desulphurization (FGD) is a crucial technology which can be used to abate the emission of sulphur dioxide in coal power plants. The dissolution of coal fly ash in adipic acid is investigated by varying acid concentration (0.05–-0.15 M), particle size (45–150 µm), pH (5.5–7.0), temperature (318–363 K) and solid-to-liquid ratio (5–15 wt.%) over a period of 60 min which is a crucial step in wet (FGD). Characterization of the sorbent was done using X–ray fluores-cence (XRF), X–ray diffraction (XRD), Furrier transform infrared (FTIR), scan-ning electron microscope (SEM) and Branauer-Emmett-Teller (BET) surface area. BET surface area results showed an increase in the specific surface area and SEM observation indicated a porous structure was formed after dissol-ution. The experimental data was analyzed using the shrinking core model and the diffusion through the product layer was found to be the rate limiting step. The activation energy for the process was calculated to be 10.64 kJ/mol.

Keywords: flue gas desulphurization, coal fly ash, dissolution, shrinking core model, activation energy.

The increase in the use of coal in thermal power plants has led to an increase in production of waste such as coal fly ash. Coal fly ash is an industrial by-product generated during coal combustion for production of energy. It is collected before flue gas reaches the chimney using either electro-static precipitators, bag filters or cyclones. It is considered hazardous because it contains elements such as boron, vanadium, chromium and arsenic which are harmful to the environment [1,2]. Fly ash is largely used in concrete as cement replacement and making geo-polymers. However, most of the fly ash produced is disposed in ash ponds or landfills [3,4].

Correspondence: H. Rutto, Department of Chemical Engineer-ing, Vanderbijlpark Campus, Vaal University Of Technology, Private Bag X021, Vanderbijlpark, 1900 South Africa. E-mail: [email protected] Paper received: 23 April, 2014 Paper revised: 14 August, 2014 Paper accepted: 9 September, 2014

Because fly ash is a waste material, it is eco-nomical to be used as a partial substitute or as a reagent in flue gas desulphurization (FGD). It is mainly composed of SiO2, Al2O3, CaO and Fe2O3 [5]. It can therefore be utilized in FGD processes as a supplement to act as a source of Ca2+, Al3+ and Si4+ which can improve the total SO2 removal efficiency in both wet and dry flue gas desulphurization systems [6]. Studies have shown that sorbents prepared from fly ash exhibit improved reactivity towards SO2 and also improved sorbent utilization. The amorphous silica in fly ash reacts with hydrated lime to form cal-cium silicate hydrates in the presence of water (poz-zolanic reaction). The calcium silicate hydrates attach to each other forming large particles with more porous structure which leads to an increase in sorbent sur-face area and improved pore volume [7-9].

Fly ash dissolution is dependent on its chemical composition and the dissolution process variables such as temperature, particle size, acid concentration,

L. KOECH et al.: DISSOLUTION KINETICS OF SOUTH AFRICAN COAL FLY ASH… Chem. Ind. Chem. Eng. Q. 21 (2) 319−330 (2015)

320

pH and solid-to-liquid ratio. A study by Kashiwakura [10] on the dissolution of solenoid from coal fly ash particles found out that it is dependent on the pH and acid concentration. Brouwers and Van [11] did a theoretical study on the dissolution of pulverized powder coal fly ash. A shrinking core model was developed for the outer and inner region to explain the behavior of the reaction rate constant as the solid changes. It was reported that the reactivity of fly ash corresponds to the silica content and the outer region is less reactive than the inner region. Pietersen [12] observed a significant increase in dissolution rate of fly ash with increase in the reactant pH (NaOH sol-ution) and also increase in the ambient temperature. A study by Tanaka and Fujii [13] showed that fly ash dissolution is greatly affected by the presence of Si4+ and Al3+. The presence of these ions in solution increases with increase in dissolution period and is accelerated with increased stirring speed. Si4+ and Al3+ are reagents for pozzolanic reaction which takes place in presence of water and leads to formation of products with high surface area.

This study looks into the possibility of using fly ash in wet flue gas desulphurization with a focus on its dissolution kinetics in adipic acid using a pH stat apparatus. It is an improvement to the previous work [14] on dissolution of fly ash for wet FDG process. The extend of dissolution of fly ash is determined by analyzing the amount calcium ions leached into sol-ution which is the most active component during sorb-ent–SO2 reaction. Using the experimental data to fit into the shrinking model, a semi-empirical model to describe the dissolution of calcium ions from fly ash was developed. The sorbent before and after dissol-ution was characterized using BET surface area, XRD and SEM.

MATERIALS AND EXPERIMENTAL METHODS

Materials

Coal fly ash was obtained from a coal-fired ther-mal power plant. XRF results showed that the che-mical composition of the studied fly ash in wt.% consisted of: 49.71 SiO2, 32.12 Al2O3, 10.52 CaO, 3.89 Fe2O3, 1.89 TiO2, 0.15 H2O and 1.92 loss on ignition. Coal fly ash was crushed using a ball mill and sieved to different particle sizes using shaking screen sieves. Adipic acid and Calcium ion standards for the AAS were supplied by CJ Labs.

Experimental

A given amount of coal fly ash was added to the reactor vessel containing a solution. The temperature, solid-to-liquid ratio, particle size and acid concen-

tration were varied at constant stiring speed of 200 rpm. This was done using a temperature controlled magnetic stirrer. The pH of the reaction mixture was determined using a pH electrode dipped into the solution and connected to a pH controller. When the pH exceeds the set value, the pump is activated to add acid to the reaction vessel and lowers the pH value to the set point.

The sample was then removed, filtered and prepared for analysis of calcium ions using atomic absorption spectrophotometer (AAS). Three calcium standard solutions (1, 5 and 10 ppm) were first pre-pared from the calcium ion standards for the AAS (1000 ppm). The standard solution was used to cal-ibrate the AAS machine before analyzing the samples. The AAS machine atomizes the samples in the flame, through which radiation of a chosen wavelength (using a hollow cathode lamp) is sent. The amount of absorbed radiation is a quantitative measure for the concentration of the element to be analyzed. The gas mixture of acetylene and nitrous-oxide was used in the AAS instrument.

The dissolution fraction was calculated as:

= Calcium ions in solution

Total amount of calciuim ions in the original sampleX

Characterization techniques

Qualitative and quantitative analysis of the stu-died fly ash before and after dissolution was done using XRD. A back loading preparation method was used for XRD analysis. Two samples were scanned after addition of 20% Si for qualitative determination of amorphous content. It was analyzed with a PANalytical Empyrean diffractometer with PIXcel detector and fixed with Fe filtered CoKα radiation. X’pert Highscore Plus software was used to identify the phases present in the samples.

The functional groups present in the samples were determined using FTIR analysis. The analysis determined using a Perkin Elmer spectrum (400FT- -IR/FT-NIR) machine equipped with a universal atte-nuated total reflectance (ATR) accessory. There was no sample preparation required for the instrument. The samples were scanned at a range of 4000 to 650 cm-1.

The surface area analysis was conducted using Micrometrics 2020 porosity analyser. The samples were degassed at 150 °C under vacuum condition for 24 h using nitrogen gas. The specific surface area was determined using the Brunauer-Emmett-Teller (BET) method. The microporous volume and area were obtained using the Barrett-Joyner-Halenda (BJH) procedure.


321

The morphological structure of the samples at different dissolution periods was studied using scan-ning electron microscopy (SEM). The samples were sprinkled on an adhesive carbon tape and were met-allized using gold before the analysis. The images of the samples were recorded at various magnifications.


Mechanism for dissolution of fly ash in adipic acid

Coal fly ash mainly consists of SiO2, Al2O3, CaO Fe2O3 and TiO2. These are chemical components that will be affected as dissolution takes place. The mech-anism for dissolution of SiO2, Al2O3, CaO Fe2O3 and TiO2 in adipic acid is therefore shown in Eqs. (2)-(6).

Adipic acid dissociates into solution to form adipate ions:

→ 2- +6 10 4(aq) 6 8 4(aq) (aq)C H O C H O + 2H (1)

The dissolution of SiO2, Al2O3, CaO, Fe2O3 and TiO2 is by adipate complexation to form their res-pective adipate complexes [10]:

→

→

2- +2(s) 6 8 4(aq) (aq)

6 8 4 2(aq) 2 (l)

SiO + 2C H O + 4H

Si(C H O ) + 2H O (2)

→

→

2- +2 3(s) 6 8 4(aq) (aq)

2 6 8 4 3(aq) 2 (l)

Al O + 3C H O + 6H

Al (C H O ) + 3H O (3)

→

→

2- +2(s) 6 8 4(aq) (aq)

6 8 4 2(aq) 2 (l)

TiO + 2C H O + 4H

Ti(C H O ) + 2H O (4)

→

→

2-(s) 6 8 4(aq) 2 (l)

-6 8 4 (aq) (aq)

CaO + C H O + H O

Ca(C H O ) + 2OH (5)

→

→

2-2 3(s) 6 8 4(aq) 2 (l)

-2 6 8 4 3(aq) (aq)

Fe O + 3C H O + 3H O

Fe (C H O ) + 6OH (6)

Dissociation of silicon, aluminium, titanium, iron and calcium adipate in the bulk will occur due to supersaturation [15]. This leads to formation of sili-con, aluminium, titanium, iron and calcium ions in solution with adipate ions [10].

⎯⎯→←⎯⎯ 4+ 2-6 8 4 2(aq) (aq) 6 8 4(aq)Si(C H O ) Si + 2C H O (7)

⎯⎯→←⎯⎯ 3+ 2-2 6 8 4 3(aq) (aq) 6 8 4(aq)Al (C H O ) 2Al + 3C H O (8)

⎯⎯→←⎯⎯ 4+ 2-6 8 4 2(aq) (aq) 6 8 4(aq)Ti(C H O ) Ti + 2C H O (9)

⎯⎯→←⎯⎯ 2+ 2-6 8 4(aq) (aq) 6 8 4(aq)CaC H O Ca + C H O (10)

⎯⎯→←⎯⎯ 3+ 2-2 6 8 4 3(aq) (aq) 6 8 4(aq)Fe (C H O ) 2Fe + 3C H O (11)

The presence of CaO in the coal fly ash is important because Ca2+ is the most reactive ion during chemo-sorption reaction in flue gas desulph-urization. Apart from dissolution by adipate complex-ation (Eq. (5)), hydrogen complexation can also cause CaO dissolution releasing calcium ions into solution [16].

CaO dissolution by hydrogen complexation:

→+ 2+ -(s) (aq) (aq) (aq)CaO + H Ca + OH (12)

Calcium ions from supersaturation (Eq. (10)) can be utilized more in pozzolanic reaction with fly ash being the source of silica and alumina to form aluminosilicate complex compounds [17].

Pozzolanic reaction:

→

→ ⋅

2+ -(aq) 2(s) 2 3(s) (aq)

2 2 8 2 (s)

Ca + 2SiO + Al O + OH

CaAl Si O H O (13)

Calcium aluminosilicate hydrate is formed from the pozzolanic reaction which increases the surface area of the sorbent and hence improve the rate of SO2 absorption in flue gas desulphurization.

XRD analysis

XRD analysis was used to determine the structural changes of coal fly ash after dissolution has taken place. The results are represented in Figure 1. The diffraction peaks exhibited by raw fly ash shows that it is mainly composed of quartz (SiO2)-Q, mullite (Al6Si2O13)-M and silicon (Si)-S. The silicon and quartz peaks diminish after 30 min and more after 60 min dissolution period. This shows that dissolution had an effect on these compounds and they were leached into solution.

A new diffraction peak appears at 2θ = 33° in the sorbents after dissolution. This can be identified as anorthite (CaAl2Si2O8). Dissolution of fly ash leads to Si4+ and Al3+ being leached out of the amorphous aluminosilicate of fly ash. Leaching of these ions (Si4+ and Al3+) increases with prolonged dissolution. A poz-zolanic reaction occurs and anorthite is formed in the sorbents after dissolution [18]. Pozzolanic reaction leads to the formation of sorbents with high surface area which can eventually improve SO2 absorption capacity during flue gas desulphurization.

BET specific surface area

The nitrogen adsorption desorption isotherm plot for fly ash before and after dissolution is illustrated in Figure 2a. The figure clearly shows that the sorbents have an adsorption isotherm of type II according the IUPAC classification [19]. This indicates that porosity of fly ash sorbent after dissolution was in the meso-

L. KOECH et al.: DISSOLUTION KINETICS OF SOUTH AFRICAN COAL FLY ASH… CI&CEQ 21 (2) 319−330 (2015)

322

Figure 1. XRD patterns for raw fly ash and sorbents after 30 and 60 min dissolution period (S - silicon, Q - quartz, M - mullite, A - anorthite).

Figure 2. Adsorption-desorption isotherm plot (a) and FTIR spectra (b) for fly ash at different dissolution periods.


323

pore range. The mesopore range is an effective zone for sulphation reaction between calcium ions and SO2 during flue gas desulphurization [7].

The specific surface area increases significantly with prolonged dissolution period (from 0.3669 to 6.9501 m2/g). The increase in surface area is attri-buted to the products of pozzolanic reaction which yields complex aluminosilicate compounds which are responsible for increased specific surface area in the sorbent [20]. The increase in the specific surface area shows that the formation of calcium aluminosilcate complex continued to change the structure of the particles into a more porous form [8].

FTIR analysis

The chemical heterogeneity of fly ash at differ-ent dissolution periods is shown in Figure 2b. From the diagram, the samples exhibited overlapping absorption bands between 1200 and 900 cm-1. These overlapping bands indicate the presence of quartz and mullite in the samples [21] and it is also due to the asymmetrical stretching of amorphous alumino-silicate formed by the reaction of extracted Si4+ and Al3+ ions. The series of bands representing quartz appear at 1084 and 796 cm-1.

The bands appearing at 1410, 1440 and 1582 cm-1 are assigned to the asymmetrical stretching vibration of CaO band [22]. The depleted peaks of CaO band is due to the effect of dissolution after which CaO goes into solution. This is excellent agree-ment with the results from XRD analysis.

Surface morphology

Figure 3 shows the surface morphology of fly ash samples at different dissolution periods. Raw fly ash mainly consists of particles that are rounded and spherical in shape with smooth surfaces. The smooth

surface in the fly ash is due to aluminosilicate particles that are formed as a result of transformation of mineral particles during coal combustion [23].

The sorbents after exposure to dissolution had their surfaces relatively rough and porous compared to raw fly ash. The sorbent also exhibited deformation of the smooth surfaces after dissolution; this was observed in the sorbent after 60 min dissolution period. This indicates that the aluminosilicate com-pounds in particles of fly ash were extracted into sol-ution [24]. The agglomerated particles in the sorbent after dissolution is attributed to the formation of cal-cium aluminosilicate compounds. This is a product of pozzolanic reaction during dissolution as shown in Eq. (13). The porous structure combined with agglo-merated particles results in an increase in the specific surface area of the sorbents which can improve the SO2 absorption capacity in flue gas desulphurization [25].

Effect of reaction variables

Effect of solid-to-liquid ratio

The effect of solid-to-liquid ratio on dissolution of fly ash in adipic acid was done in the range of 5-15 wt.%. The temperature, pH, particle size and acid concentration were kept constant at 60 °C, 5.5, 45 µm and 0.1 M, respectively. The experimental results are represented in Figure 4a. It is evident that the conversion of calcium ions into solution is higher at lower solid-to-liquid ratio compared to higher solid-to-liquid ratio within the same dissolution period. This is attributed to the decrease in the fluid reactant per unit weight of the solid as solid-to-liquid ratio increases.

Effect of acid concentration

To investigate the effect of acid concentration on dissolution of fly ash, different experiments were per-

60 minutes Raw fly ash

Figure 3. SEM micrographs for raw fly ash and sorbent after 60 min dissolution period.


324

Figure 4. Effect of solid-to-liquid ratio (a), acid concentration (b), particle size (c), pH (d) and temperature (e) on the dissolution of fly ash.

formed at a range of 0.05-0.15 M adipic acid. The temperature, pH, particle size and solid-to-liquid ratio were kept constant at 60 °C, 5.5, 45 µm and 10 wt.%, respectively. Figure 4b shows the experimental results and it is clear that the conversion of fly ash increases with increase in acid concentration. Inc-rease in acid concentration leads to an increase in H+ activity in the liquid film therefore enhancing reaction on the solid surface.

Effect of particle size

Four different size fractions were used to inves-tigate the effect of particle size on the dissolution rate of coal fly ash. The average size fractions used were from 45-150 µm. The temperature, pH, solid-to-liquid ratio and acid concentration were kept constant at 60

°C, 5.5, 10 wt.% and 0.1 M, respectively. The expe-rimental results are represented in Figure 4c, which shows that by reducing the particle size of fly ash, it significantly improved the conversion of calcium ions into solution as compared to larger particle size over the same dissolution period. This is because finer particles have higher surface area, thus enhancing dissolution.

Effect of pH

The effect of pH on dissolution of fly ash was studied in the range of 5.5-7.0. The temperature, solid-to-liquid ratio, particle size and acid concen-tration were kept constant at 60 °C, 10 wt.%, 45 µm and 0.1 M, respectively. Figure 4d represents the experimental results. It is evident that dissolution rate


325

increases with decrease in pH. This is because the increase in pH causes an increase in the apparent mass transfer coefficient.

Effect of temperature

The effect of temperature on the dissolution rate of fly ash was performed in the range between 318- -363 K. The soilid to liquid ratio, pH, particle size and acid concentration were kept constant at 10 wt.%, 5.5, 45 µm and 0.1 M, respectively. The experimental results are depicted in Figure 4e and it can be seen that the conversion of calcium ions is enhanced at higher temperatures than at lower temperatures. High temperatures accelerate the reaction rate because there is an increase in energy which results in more collision between reacting molecules which speeds the reaction.

Dissolution kinetics

The dissolution kinetics of fly ash in adipic acid was studied using the shrinking core model for solid-liquid system. The model considers the reaction of the reactants at the surface of the solid particles which results in both aqueous and solid particles [26]. The unreacted core of the particle reduces in size as the reaction proceeds, with more solids and aqueous products being formed [27]. This model considers the following steps in series:

1. Diffusion of the fluid reactant through the film surrounding the particle to the surface of the solid;

2. Penetration and reaction of the fluid reactant through the layer of ash to the surface of the unreac-ted core;

3. Fluid-solid surface chemical reaction at the reaction surface.

It is considered that the slowest step is the rate controlling step. From the above reaction steps, a heterogeneous system is considered to be controlled by: film diffusion, product layer diffusion or chemical reaction at the surface of the core of the unreacted particle. [28]. These steps can be integrated and written as follows:

Film diffusion equation,

ορ= =3 l A

B

bk CX t kt

R (14)

Chemical reaction control,

ορ− − = =

131 (1 ) s A

rB

bk CX t k t

R (15)

Product layer diffusion,

ορ+ − − − = =

23 6

1 2(1 ) 3(1 ) e Ad

B

bD CX X t k t

R (16)

The experimental data was fitted into the shrinking core model using Eqs. (15) and (16). The fluid media used for this study is liquid in nature and it is therefore considered that mass transfer across the fluid film will have least effect on the system. There-fore the fluid film diffusion step will not be controlling for this case [29]. The apparent rate constants for chemical reaction and product layer diffusion models were obtained by plotting the left side of Eqs. (15) and (16) with the reaction time. The apparent rate cons-tants from the plots and their correlation coefficients are represented in Table 1.

According to Table 1, the product layer diffusion model had the highest regression coefficients. The linear relationship between

+ − − −231 2(1 ) 3(1 )X X

and the reaction time is shown in Figure 5a-e for solid-to-liquid ratio (a), acid concentration (b), particle size (c), pH (d) and temperature (e). This therefore shows that the equation for the dissolution kinetics for this process follows the product layer diffusion model and therefore can be written as follows:

= + − − −231 2(1 ) 3(1 )dk t X X (17)

To include all the reaction parameters, a semi-empirical model can be written as:

ο

− =

aEbRTa c d

dS

k K C D P eL

(18)

Combining Eqs. (17) and (18) yields:

( ) ( )

ο

−

+ − − − =

=

231 2 1 3 1

aEbRTa c d

X X

SK C D P e t

L

(19)

where C is acid concentration, S/L is solid-to-liquid ratio, D is particle size and P is pH.

The values of the constants a, b, c and d are the reaction orders with respect to each parameter. Their values were obtained by plotting the natural logarithm of the reaction rate constants against natural log-arithm of their respective parameter values. Their plots are represented in Figure 6a-d for solid-to-liquid ratio, acid concentration, particle size and pH, res-pectively.

The Arrhenius plot was used to evaluate the activation energy for the dissolution process. Accord-ing to the Arrhenius plot (illustrated in Figure 6e), the intercept was found to be 3.5745 and the activation energy evaluated from the slope was 10.64 kJ/mol.


326

The value of the activation energy shows that the dissolution of coal fly ash in adipic acid is a product layer diffusion controlled process. When the activation energy is below 20 kJ/mol, product layer diffusion is usually the rate controlling step [28,30-32].

A semi-empirical model for this process can therefore be written as follows:

( ) ( )− −

− −

+ − − − =

=

23

0.86210.5592 0.8055 1.196

1 2 1 3 1

3.5745aE

RT

X X

SC D P e t

L

(20)

CONCLUSION

The findings of this study show signigicant effects of the process variables on the dissolution of fly ash in adipic acid. It was found that the dissolution rate increases with increase in temperature and acid concentration, but decreases with increase in particle

size, pH, and solid-to-liquid ratio. The pozzolanic reaction resulted in the formation of anorthite as seen in XRD analysis. This also contributed to a significant increase in the specific surface area of sorbent (0.3669 to 6.9501 m2/g) as observed in the BET surface area analysis. The formation of aggregates and rough surfaces was observed on the surface of the sorbent after dissolution using SEM. The dis-solution of fly ash was found to follow the shrinking core model with product layer diffusion model being the rate limiting step. The product layer being reaction products such as anorthite and other compounds. The semi-empirical model describing the dissolution of coal fly ash can be represented as follows:

( ) ( )− −

− −

+ − − − =

=

23

0.86210.5592 0.8055 1.196

1 2 1 3 1

3.5745aE

RT

X X

SC D P e t

L

Table 1. Dissolution rate constants and their correlations coefficients

Process variable

Surface chemical reaction Product layer diffusion:

− − =131 (1 ) rX k t + − − − =

231 2(1 ) 3(1 ) dX X k t

Kr / min-1 R2 Kd / min-1 R2

Temperature, K

318 0.0016 0.8640 0.00049 0.9729

333 0.0019 0.9100 0.00061 0.9961

348 0.0020 0.9096 0.00073 0.9993

363 0.0021 0.9070 0.00080 0.9989

Solid-to-liquid ratio, wt.%

5 0.0031 0.9129 0.00168 0.9934

10 0.0021 0.8908 0.00080 0.9910

12.5 0.0020 0.8984 0.00077 0.9971

15 0.0019 0.9100 0.00065 0.9961

Concentration, mold m-3

0.05 0.0015 0.8973 0.00040 0.9804

0.075 0.0016 0.8839 0.00050 0.9837

0.1 0.0019 0.9100 0.00065 0.9961

0.15 0.0020 0.8975 0.00072 0.9935

Particle size, µm

45 0.0019 0.9100 0.00065 0.9961

63 0.0016 0.8400 0.00050 0.9685

75 0.0014 0.8874 0.00036 0.9861

150 0.0011 0.9175 0.00025 0.9914

pH

5.5 0.0019 0.8060 0.00065 0.9556

6 0.0018 0.8242 0.00059 0.9698

6.5 0.0017 0.8091 0.00054 0.9519

7 0.0016 0.8342 0.00049 0.9744


327

Figure 5. Linear relationship showing variation of 32

)1(3)1(21 XX −−−+ with the reaction time for different solid-to-liquid ratio

(a), acid concentration (b), particle size (c), pH (d) and temperature (e).


328

Figure 6. Variation of -ln Kd with -ln S/L (a), -ln C (b), -ln D (c), -ln P (d) and 1/T (e) (Arrhenius plot).


329

The activation energy for the process was determined to be 10.64 kJ/mol.

Nomenclature

FGD – Flue gas desulphurization b – Stoichiometric coefficient Kl – Mass transfer coefficient (m min-1) CA – Bulk concentration (mol cm-3) ρB – Sorbent molar density (kg mol m-3) Ro – Initial particle radius (m) t – Reaction time (min) De – Product layer effective diffusion coefficient (m2

min-1) Kd – Product layer reaction rate constant (min-1) Ks – Surface reaction rate constant (m min-1) Ea – Activation energy (kJ mol-1) R – Universal gas constant (kJ mol-1 K-1) T – Temperature (K) C – Acid concentration (mol dm-3) S/L – Solid-to-liquid ratio (g ml-1) D – Particle size (µm) P – pH

REFERENCES

[1] S. Mishra, S. Langwenya, B. Mamba, M. Balakrishnan, Phys. Chem. Earth 35 (2010) 811-814

[2] J. Temuujin, R. Williams, A. Van Riessen, J. Mater. Process. Technol. 209 (2009) 5276-5280

[3] T. Bakharev, Cement Concrete Res. 35 (2005) 1224- -1232

[4] J. Swanepoel, C. Strydom, Appl. Geochem. 17 (2002) 1143-1148

[5] F. Goodarzi, Fuel 85 (2006)1418-1427

[6] H. Song, J. Park, Environ. Technol. 22 (2001) 1001-1006

[7] D. Ogenga, Z. Siagi, M. Onyango, M. Mbarawa, Front. Chem. Sci. Eng. 3 (2009) 46-51

[8] R. Lin, S. Shih, Powder Technol. 131 (2003) 212-222

[9] W. Jozewicz, G.T. Rochelle, Environ. Prog. 5 (1986) 219- -224

[10] S. Kashiwakura, H. Ohno, Y. Kumagai, H. Kubo, K. Mat-subae, T. Nagasaka, Chemosphere 85 (2011) 598-602

[11] H. Brouwers, R. Van Eijk, J. Mater. Sci. 37 (2002) 2129- -2141

[12] H.S. Pietersen, A.L. Fraay, J.M. Bijen, MRS Online Proc. Libr. 89 (1989)

[13] H. Tanaka, A. Fujii, Adv. Powder Technol. 20 (2009) 473- -479

[14] L. Koech, R. Everson, H. Neomagus, H. Rutto, in Proceedings of International Conference on Chemical, Mining and Metallurgical Engineering, Johannesburg, South Africa, 2013, p. 233

[15] S.D. Rocha, M.B. Mansur, V.S. Ciminelli, J. Chem. Technol. Biotechnol. 79 (2004) 816-821

[16] G. Dube, P. Osifo, H. Rutto, Clean Technol. Environ. Policy 16 (2014) 891-900

[17] N. Karatepe, A. Mericboyu, S. Küçükbayrak, Energy Sources 20 (1998) 505-511

[18] K.T. Lee, S. Bhatia, A.R. Mohamed, J. Mater. Cycles Waste Manage. 7 (2005) 16-23

[19] H. Hadjar, B. Hamdi, M. Jaber, J. Brendlé, Z. Kessaissia, H. Balard, J. Donnet, Microporous Mesoporous Mater. 107 (2008) 219-226

[20] P. Adamiec, J. Benezet, A. Benhassaine, Particuology 6 (2008) 93-98

[21] M. Criado, A. Fernández-Jiménez, A. Palomo, Micro-porous Mesoporous Mater. 106 (2007)180-191

[22] K. Thiruppathi, S. Barathan, N. Anandhan, G. Sivakumar, Appl. Phys. Res. 59 (2009) 1-9

[23] S.M. Nyale, O.O. Babajide, G.D. Birch, N. Böke, L.F. Petrik, Procedia Environ. Sci. 18 (2013) 722-730

[24] Z. Sarbak, M. Kramer-Wachowiak, Powder Technol. 123 (2002) 53-58

[25] E. van der Merwe, C. Strydom, J. Therm. Anal. Calorim. 84 (2006) 467-471

[26] W. Hsu, M. Lin, J. Hsu, Int. J. Chem. Biol. Eng. 2 (2009) 4-8

[27] N. Demirkıran, Braz. J. Chem. Eng. 141 (2008) 180-186

[28] O. Levenspiel, Chemical Reaction Engineering, 3rd ed., John Wiley and Sons, New York, 1972, p. 583

[29] P. Maina, M. Mbarawa, Energy Fuels 25 (2011) 2028- -2038

[30] R.H. Limo, C. Enweremadu, Chem. Ind. Chem. Eng. Q.. 17 (2011) 459-468

[31] Z.I. Zafar, Chem. Eng. J. 141 (2008) 233-241

[32] J. Zhao, Y. Li, K. Han, S. Niu, C. Lu, J. Chem. Eng. Jpn. 46 (2013) 827-832.


330

LAWRENCE KOECH1

RAY EVERSON2

HEIN NEOMAGUS2

HILARY RUTTO1 1Department of Chemical Engineering,

Vanderbijlpark Campus, Vaal University Of Technology,

Vanderbijlpark, South Africa 2Department of Chemical and Minerals Engineering, Potchefstroom Campus,

North West University, Potchefstroom, South Africa

NAUČNI RAD

KINETIKA RASTVARANJA LETEĆEG PEPELA JUŽNOAFRIČKOG UGLJA I RAZVIJANJE POLUEMPIRIJSKOG MODELA ZA PREDVIĐANJE RASTVARANJA

Mokra desulfurizacija dimnog gasa (FGD) predstavlja presudnu tehnologiju, koja se može

koristiti za smanjenje emisije sumpor-dioksida u elektranama na ugalj. Rastvaranje letećeg

pepela uglja u adipinskoj kiselini je praćeno promenom koncentracije kiseline (0,05-0,15

M), veličine čestice (45-150 mm), pH (5.5-7.0), temperature (318-363 K) i odnosa čvrsto-

tečno (5-15 mas.%.) tokom perioda od 60 min, što je ključni korak u mokrom postupku.

Karakterizacija sorbenta je urađena pomoću X-fluoroscentne analize (XRF), X-difrak-

cionae analize (XRD), Furrier infracrvene analize (FTIR), skening elektronske mikroskopije

(SEM) i Brunauer-Emmett-Teller metode (BET) za specifičnu površinu. BET rezultati

pokazuju povećanje specifične površine, dok SEM analiza ukazuje na to da je nakon

rastvaranja formirana porozna struktura. Eksperimentalni podaci su analizirani korišćenjem

modela neizreagovanog jezgra i pokazalo se da je stupanj koji određuje brzinu reakcije

upravo difuzija kroz sloj produkta. Izračunato je da aktivaciona energija za proces iznosi

10,64 kJ/mol.

Ključne reči: desulfurizacija dimnog gasa, leteći pepeo, rastvaranje, model neiz-reagovanog jezgra, aktivaciona energija.





331

ZHUONI HOU1,2

XIANRUI LIANG1

FENG SU1

WEIKE SU1 1Key Laboratory for Green

Pharmaceutical Technologies and Related Equipment of Ministry of

Education, College of Pharmaceutical Sciences, Zhejiang

University of Technology, Hangzhou, China

2College of Life Sciences, Zhejiang Sci-Tech University, Hangzhou,

China

SCIENTIFIC PAPER

UDC 58:543.544:54

DOI 10.2298/CICEQ140522034H

PREPARATIVE ISOLATION AND PURIFICATION OF SEVEN COMPOUNDS FROM Hibiscus mutabilis L. LEAVES BY TWO-STEP HIGH-SPEED COUNTER- -CURRENT CHROMATOGRAPHY

Article Highlights • The combination of RSM and HSCCC is applied on separating and purifying rutin for the

first time • HSCCC firstly separates steppogenin and genistein from Hibiscus mutabilis L. leaves • Accurate NMR data of seven compound separated from Hibiscus mutabilis L. leaves

is given Abstract

Seven compounds from Hibiscus mutabilis L. leaves were successfully achieved by two-step high-speed counter-current chromatography with a two-phase solvent system composed of n-butanol-ethyl acetate-water (1:6:9 volume ratio) and n-hexane-ethyl acetate-methanol-water (3:5:3:5 volume ratio). The critical experimental parameters of first-step separation were opti-mized with response surface methodology as follows: flow rate was 1.1 mL/min, revolution speed was 800 rpm and temperature was 30 °C. Under the optimal conditions, around 5.0 mg of salicylic acid, 13.6 mg of rutin, 5.5 mg of genistein were obtained in 100 mg crude sample. Then, 9.2 mg of potengrif-fioside A, 4.7 mg of kaempferol 3-O-rutinoside, 3.0 mg of steppogenin and 2.5 mg of emodin were obtained by second-step separation. The purities of the seven compounds determined by UPLC were 96.2, 93.8, 95.4, 94.3, 98.0, 94.1 and 90.8%, respectively. Their chemical structures were identified by electron spray ionization mass spectroscopy (ESI-MS) and 1H, 13C nuclear magnetic resonance (NMR). Furthermore, compound steppogenin and genistein were first reported from Hibiscus mutabilis L. The purification method was simple, efficient and evaded tedious separation process.

Keywords: Hibiscus mutabilis L., high-speed counter-current chromato-graphy, rutin, genistein, kaempferol-3-O-rutinoside, steppogenin.

Influenza is a serious public health problem due to its infectivity and fatality. It causes about 3-5 million cases of severe illness and from 250,000 to 500,000 deaths throughout the world every year. Hibiscus mutabilis L. leaves, whose extracts exhibited potent anti-flue activity [1,2], are used as raw material in CFDA approved drugs to treat influenza, namely fur-

Correspondence: W. Su, Key Laboratory for Green Pharma-ceutical Technologies and Related Equipment of Ministry of Education, College of Pharmaceutical Sciences, Zhejiang Uni-versity of Technology, Hangzhou 310014, China E-mail: [email protected] Paper received: 22 May, 2014 Paper revised: 10 August, 2014 Paper accepted: 15 September, 2014

ong anti-flu tablet, compound furong tincture and fupu anti-flu granules [3,4]. A few studies have established that the extract of H. mutabilis L. leaves is composed of flavonoids and phloretin [5,6], and antioxidant and radical scavenging activities of these compounds are possible modes of actions. Thus, separation and analysis of flavonoids compounds from the extract of H. mutabilis L. leaves became a significant research topic. Solvent extraction methods have been employed to extract flavonoids of H. mutabilis L. leaves and the effects of extraction were compared by using different solvent systems [6,7]. After extracting, Yao et al. (2003) isolated and purified 10 components from the original extraction by silica gel column chro-

Z. HOU et al.: PREPARATIVE ISOLATION AND PURIFICATION… Chem. Ind. Chem. Eng. Q. 21 (2) 331−341 (2015)

332

matography and C18 reversed phase chromato-graphy [5]. However, conventional methods available for separation and purification of these compounds involve solvent extraction method and column chro-matography are tedious and involve multiple steps. Furthermore, column chromatography often leads to the loss of activity and low recovery of target com-pounds due to its irreversibly absorptive effect of the solid matrix during the isolation and purification pro-cedures [8-10]. Therefore, it is necessary to establish a rapid and highly efficient method for separation and purification of flavonoids from the extract of H. mut-abilis L. leaves.

High-speed counter-current chromatography (HSCCC), a continuous liquid-liquid partition chroma-tography based on partitioning of compounds between

two immiscible liquid phases, conquers the disadv-antage of irreversible adsorptive loss of samples onto the solid support matrices, shortens the separation time, and has been successfully applied to separate and purify various natural products [11-14]. This tech-nique is firstly used for separation of compounds from H. mutabilis L. leaves and the separation condition parameters are optimized by response surface methodology (RSM).

The findings of the current research show that seven compounds from the Chinese medicinal plant H. mutabilis L. leaves is effectively separated and purified by HSCCC. All of the isolated compounds including steppogenin (1), salicylic acid (2), rutin (3), genistein (4), potengriffioside A (5), kaempferol 3-O-rutinoside (6) and emodin (7), Figure 1, are identified

Figure 1. Structures of compounds isolated from leaves of Hibiscus mutabilis L.


333

by ESI-MS, 1H-NMR and 13C-NMR spectra. Further-more, flavonoid steppogenin (1) and isoflavone geni-stein (4) are firstly isolated from H. mutabilis L. leaves.

EXPERIMENTAL

Apparatus

The semi-preparative HSCCC instrument used in this study was a Model TBE-300A HSCCC (Tauto Biotechnique Company, Shanghai, China) with a multilayer coil planet centrifuge equipped with a poly-tetrafluoroethylene (PTFE) preparative coil (diameter of tube 1.6 mm; total volume 280 mL). A manual injection valve with a 20 mL sample loop was used to introduce the sample into the coil system. The two- -phase solvent system was delivered by a model TBP-50A pump (Tauto Biotechnique, Shanghai, China). The β value of the preparative column varied from 0.5 at the internal terminal to 0.8 at the external terminal (β = r/R, where r is the distance from the coil to the holder shaft, and R is the revolution radius or the distance between the holder axis and central axis of the centrifuge). The rotation speed is adjustable from 0 to 1000 rpm. The continuous monitoring of the effluent was operated with a Model UVD-200UV Monitor, a multi-wavelength UV-vis monitor for simul-taneous monitoring of up to four wavelengths at 254, 280, 340 and 365 nm (Shanghai Jinda Biochemy Apparatus Co. Ltd., Shanghai, China). An HX-1050 constant temperature circulating implement (Beijing Boyikang Lab Instrument Co. Ltd., Beijing, China) was employed to control the separation temperature. The data were collected with a model V4.0 chroma-togram workstation (Shanghai Jinda Biochemy Appa-ratus Co. Ltd., Shanghai, China).

Purity analysis was performed on a Waters Acquity UPLC™ system (Waters, Milford, MA, USA) equipped with a binary solvent delivery pump, a ther-mostated column compartment, an auto sampler and a VWD detector, and connected to Waters Empower software.

The structure was identified with an electrospray ionization mass spectrometer (ESI/MS, Finnigan Advantage LCQ, Thermo, USA) and a Varian INOVA 400-MHz-FT-nuclear magnetic resonance (NMR) spectrometer (Varian, CA, USA).

Reagents and materials

Petroleum ether (60-90 °C), ethyl acetate and ethanol for preparation of crude samples were all of industrial grades (Tingting Chemical Industry Co. Ltd., Zhejiang, China). n-hexane, ethyl acetate, n-butanol (Gaojing Fine Chemical Industry Co. Ltd., Zhejiang, China), methanol (Damao Chemical Reagent Factory,

Tianjin, China) used for HSCCC separation and the preparation of crude extracts were all of analytical grade. Methanol used for UPLC analyses was of chromatographic grade (Merck KGaA, Darmstadt, Germany). Formic acid used for UPLC analyses and solvent systems selected was of chromatographic grade (Dima Technology Inc., CA, USA). The water used in the experiment was produced by Barnstead TII super Pure Water System (Thermo Fisher Sci-entific, MA, USA). The dried leaves of Hibiscus mut-abilis L. were purchased from Zhejiang Chinese Medical University Chinese herbal pieces factory.

Preparation of the crude sample

The dried leaves of H. mutabilis L. (7.5 kg) were powdered and added to 100% ethanol, mixed tho-roughly, and then settled for 72 hours at room tempe-rature. The supernatant fluid was filtered and con-centrated to dryness by rotary evaporator under red-uced pressure and further dried under vacuum. The residue was extracted five times with petroleum (b.p. 60-90 °C) and petroleum (b.p. 60-90 °C)-ethyl acetate (1:3 volume ratio) successively; a total of 70 g of pet-roleum extract, 22 g of petroleum-ethyl acetate ext-ract and 60 g residue were obtained. The petroleum-ethyl acetate extracts were further subjected to AB-8 macroporous resin (Zhejiang Zhengguang Industrial Co., Ltd., Zhejiang, China) by stepwise elution with aqueous ethanol (30, 50, 80 and 100 vol.%). After TLC and UPLC analysis, the fraction eluted by 50% aqueous ethanol (6.8 g) were evaporated to dryness under reduced pressure and stored for subsequent HSCCC separation as crude sample.

Determination of partition coefficients (K) and selection of two-phase solvent system

Several two-phase solvent systems were eval-uated for HSCCC separation. The selection of two-phase solvent depended on the partition coefficients (0.5 < K < 2.0). The K values were determined as follows: about 1.0 mg crude sample was added into a 10 mL test tube to which 2 mL of each phase of pre-equilibrated two-phase solvent system was previously added. The tube was shaken completely to distribute the sample in the two phases thoroughly. Then, an equal volume (1 mL) of the upper and lower phase was transferred and evaporated, respectively. The residue was dissolved in 1 mL of UPLC mobile phase for UPLC analysis. The peak area of the upper phase was recorded as AU and that of the lower phase as AL. The K value was calculated according to the equa-tion, K = AU/AL.


334

Preparation of solvent system and sample solution

Two solvent systems composed of ethyl ace-tate-n-butanol-water (6:1:9 volume ratio) and n-hex-ane-ethyl acetate-methanol-water (3:5:3:5 volume ratio) were used for HSCCC separation. The pre-paration of each two-phase solvent system was per-formed in a separatory funnel according to the volume ratios and thoroughly equilibrated by shaking at room temperature until there were two clearly separated phases. Both phases were separated and degassed by using an ultrasonic bath for 45 min.

Sample solution of the first HSCCC separation was prepared by dissolving 100 mg crude sample in 20 mL solvent mixture of upper phase and lower phase (1:1 volume ratio) of ethyl acetate-n-butanol-water (6:1:9 volume ratio). The second one was pre-pared by dissolving 42 mg residue of the first HSCCC separation in 20 mL solvent mixture of upper phase and lower phase (1:1 volume ratio) of n-hexane-ethyl acetate-methanol-water (3:5:3:5 volume ratio).

HSCCC Separation procedure

Each HSCCC was performed as procedure below in the whole separation: the multilayer coiled column was first filled with the upper phase as the stationary phase; then the lower mobile phase was firstly pumped into the head of the column at flow rate of 1.5 mL/min. In the meantime, the HSCCC appa-ratus was rotated at work revolution speed of 850 rpm, and the system was placed at steady tempe-rature of 25 °C. After the mobile phase emerged at the tail outlets and the liquid-liquid equilibrium was established in the column, samples were injected into the injection valve. The effluent was continuously monitored with a UV detector at 254 nm and the peak fractions were collected according to the chroma-togram. After target compounds were collected, the centrifuge was stopped and stationary phase was pumped out of the column with pressured nitrogen and collected in a graduated cylinder to measure the retention volume.

UPLC Analysis, mass spectrometry and 1H- and 13C-NMR identification of the fractions

The crude sample and each purified fraction separated by HSCCC were analyzed by UPLC. The UPLC analysis was performed with an ACQUITY UPLC® BEH C18 column (1.7 μm, 50 mm×2.1 mm i.d., Waters, USA). The mobile phase consisted of A (0.01% formic acid aqueous solution) and B (metha-nol), which was programmed as follow: 0-3 min, 10- -15% B; 3-5 min, 15-20% B; 5-20 min, 20-40% B; 20-30 min, 40-60% B; 30-35 min, 60% B. The detec-tion wavelength was set at 254 nm and flow rate was

0.15 mL/min. Figure 2 shows the chromatograms of the crude sample and peak fractions separated by HSCCC under the optimized UPLC condition.

The purified fraction of hibiscus mutabilis L. obtained from the semi-preparative HSCCC separ-ation was analyzed by ESI-MS, 1H- and 13C-NMR, respectively. 1H- and 13C-NMR spectra were mea-sured with tetramethylsilane (TMS) as internal stan-dard. All the experiments were run under room tem-perature.


Optimization of two-solvent system

An ideal two-phase solvent system was critical for HSCCC separation, which mainly involved four aspects listed below [15]. First, sufficient stationary phase should be retained while the mobile phase passes through the system. Second, for satisfactory retention of the stationary phase, the settling time of the solvent systems should be considerably less than 30s. Third, K of the target compound should be close to 1 for ensuring the retention time of target com-ponent and an acceptable K value is in the range of 0.5-2. Ultimately, the separation factor between two components (α = K2/K1, K2 > K1) should be higher than 1.5 in semi-preparative HSCCC equipment.

Several groups of solvent systems based on ethyl acetate-methanol-water, ethyl acetate-n-bu-tanol-water, ethyl acetate-n-butanol-water-formic acid and n-hexane-ethyl acetate-methanol-water were evaluated in this experiment and the partition coef-ficients were measured as listed in Table 1. All the tested solvent systems failed to provide suitable ranges of K values for all the target compounds in a single run. For separation of compounds in the crude sample, the ethyl acetate-n-butanol-water (6:1:9 volume ratio) system offered acceptable K values for compounds 2 (0.45), 3 (0.79) and 4 (1.35). Never-theless, compounds 1 and 5-7 resulted in a poor sol-ubility in aqueous phase. Formic acid was added to shorten the separation time, but no remarkable effect was observed. Besides, various solvent systems based on n-hexane-ethyl acetate-methanol-water were conducted for partition coefficient tests. Com-paring with different ratios, n-hexane-ethyl acetate-methanol-water with volume ratio at 3:5:3:5 gave suitable partition coefficients for compounds 1 (1.44), 5 (0.87) and 6 (0.95). Furthermore, ethyl acetate- -methanol-water system with volume ratio at 2:1:3 resulted in unsatisfied solubility for compounds 1 and 2, although K values of other target compounds had potential to be modified. Ultimately, ethyl acetate-n- -butanol-water (6:1:9 volume ratio) and n-hexane-


335

Figure 2. UPLC Chromatograms of the crude sample and HSCCC fractions from leaves of Hibiscus mutabilis L.; column: ACQUITY UPLC® BEH C18 column (1.7 μm, 50 mm×2.1 mm i.d., Waters, USA); the mobile phase consisted of A (0.01% formic acid aqueous

solution) and B (methanol), which was programmed as follow: 0-3 min, 10-15% B; 3-5 min, 15-20% B; 5-20 min, 20-40% B; 20-30 min, 40-60% B; 30-35 min, 60% B; flow rate: 0.15 mL/min; detection: 254 nm. A) crude sample of leaves of H. mutabilis L., B) steppogenin (compound 1), C) salicylic acid (compound 2), D) rutin (compound 3), E) genistein (compound 4), F) potengriffioside A (compound 5),

G) kaempferol 3-O-rutinoside (compound 6) and H) emodin (compound 7).


336

Table 1. The K values of target compounds measured in different solvent systems

Solvent system Volume ratio Compound

1 2 3 4 5 6 7

Ethyl acetate-methanol-water 2:1:3 - - 0.17 0.25 7.47 7.10 ∞

ethyl acetate-n-butanol-water 6:1:9 ∞ 0.45 0.79 1.35 ∞ ∞ ∞

ethyl acetate-n-butanol-water-formic acid 6:1:9:0.009 ∞ 0.37 0.61 0.97 246.20 ∞ ∞

n-Hexane-ethyl acetate-methanol-water 3:5:3:5 1.44 - - - 0.87 0.95 ∞

ethyl acetate-methanol-water (3:5:3:5 volume ratio) were used for HSCCC separation of seven com-pounds in two separation runs.

In the first HSCCC separation, three fractions 2-4 were produced and collected (compound 2: 5.0 mg with 96.2% UPLC purity, collected between 126 and 136 min; compound 3: 13.6 mg with 93.8% UPLC purity, collected between 174 and 193 min; com-pound 4: 5.5 mg with 95.4% UPLC purity, collected between 207 and 228 min), respectively (Figure 3).

In the second HSCCC separation, three frac-tions, 1, 5 and 6, were produced and collected (com-pound 5: 9.2 mg with 94.3% UPLC purity, collected between 115 and 128 min; compound 6: 4.7 mg with 98.0% UPLC purity, collected between 135 and 147 min; compound 1: 3.0 mg with 94.1% UPLC purity, collected between 151 and 157 min), respectively (Figure 4). Moreover, 2.5 mg compound 7 with 90.8% UPLC purity was obtained from the residue of the second HSCCC separation.

Optimization of HSCCC procedure with RSM

To obtain a best purity of rutin (compound 3), a three-level and three-variable of Box-Behnken design (BBD) was applied to optimizing the first-step separ-ation experimental parameters. Three independent variables investigated were: separation temperature (20, 25, 30 °C, X1), mobile phase flow rate (1.0, 1.5 and 2.0 mL/min, X2) and revolution speed of the multilayer coiled column (800, 850 and 900 rpm, X3), while the response was the purity of rutin (Y). Seven-teen experiments were designed to explore the vari-ables of separation parameters that affect the purity of rutin. Crude sample of 100 mg was used for all experiments. The experimental design is presented in Table 2, along with the obtained experimental res-ponses studied in each experiment. For predicting the optimal point, an empirical second-order polynomial model was established to perform relationship between independent variables and responses.

Figure 3. HSCCC Chromatograms of the first separation. HSCCC Conditions: solvent system: ethyl acetate-n-butanol-water (6:1:9

volume ratio); stationary phase: upper phase. Flow rate: 1.11 mL/min, revolution speed: 800 rpm, temperature: 30 °C; stationary phase: upper organic phase; detection wavelength: 254 nm; sample size: 100 mg.


337

Figure 4. HSCCC Chromatograms of the second separation. HSCCC Conditions: solvent system: n-hexane-ethyl acetate-methanol-water (3:5:3:5 volume ratio); stationary phase: upper phase. Flow rate: 1.5 mL/min, revolution speed: 850 rpm,

temperature: 25 °C; stationary phase: upper organic phase; detection wavelength: 254 nm; sample size: 42 mg.

Table 2. Box-Behnken design matrix of independent variables and their corresponding responses

Run Separation temperature

X1 / °C Flow rate of the mobile phase

X2 / mL min-1 Revolution speed

X3 / rpm Purity of rutin

Y / %

1 25 (0)a 1.5 (0) 850 (0) 78.94

2 25 (0) 1.5 (0) 850 (0) 79.52

3 25 (0) 1.5 (0) 850 (0) 77.92

4 25 (0) 2 (1) 900 (1) 73.65

5 25 (0) 1.5 (0) 850 (0) 81.52

6 20 (-1) 1.5 (0) 900 (1) 85.73

7 30 (1) 2 (1) 850 (0) 67.77

8 25 (0) 1.0 (-1) 800 (-1) 72.87

9 25 (0) 1 (-1) 900 (1) 64.26

10 20 (-1) 2 (1) 850 (0) 75.05

11 30 (1) 1.5 (0) 900 (1) 72.33

12 20 (-1) 1.5 (0) 800 (-1) 60.57

13 25 (0) 1.5 (0) 850 (0) 78.82

14 30 (1) 1.5 (0) 800 (-1) 82.87

15 30 (1) 1 (-1) 850 (0) 79.67

16 20 (-1) 1.0 (-1) 850 (0) 61.13

17 25 (0) 2.0 (1) 800 (-1) 58.33 aActual value out of bracket and coded value in bracket


338

Experimental data analysis and quadratic model building were conducted using the Design Expert software V8.0.6.1 Trial (State-Ease Inc., Minneapolis, MN, USA). Analysis of variance (ANOVA) and model residuals were used to check the validity of the mathematical models. Quality of the fitted model was expressed with the coefficient of determination R2, and its statistical significance was checked by the F-test. According to these experimental results, the second-order polynomial model was established as follows:

− −

= − + − + −− − + −

− × − × − ×

1 2 3

1 2 1 3 2 3

3 2 2 3 21 2 3

1704.13 35.06 40.42 3.17

2.58 0.04 0.24

6.83 10 33.07 1.52 10

Y X X X

X X X X X X

X X X

The analysis of variance about model is shown in Table 3. The significance of model was determined using the t test. The coefficient of determination (R2) of the model was 0.9803, demonstrating that the relationship between those parameters chosen was

represented adequately by the model. The F value of 38.72 and p value less than 0.0001 implied that the model was significant. The “Lack of Fit F value” of 2.76 and results of error analysis indicated that lack of fit was insignificant (p > 0.05), meaning an agreement between the experimental results and the theoretical values predicted by the polynomial model. The coefficient of variation (C.V.) of less than 5% sug-gested that the model was reproducible. The pre-dicted residual sum of squares (PRESS) of 250.39 showed that the model fitted each point in the design. According to the p value, all variables were significant except for X2 and X1

2, and interaction effect between any two variables had statistically significant effects on purity of rutin. The full model was made three dimensional and contours were plotted to predict the relationships between the independent variables and the dependent variables (Figure 5).

It can be seen that the maximum response was located in the temperature range of 28-30 °C and flow

Table 3. Analysis of variance (ANOVA) for response surface quadratic model

Source Mean squares Dfa Sum of squares F value p value

Model 122.65 9 1103.87 38.72 0.0001

X1 50.80 1 41.50 16.04 0.0052

X2 1.22 1 1.22 0.39 0.5538

X3 56.87 1 56.87 17.95 0.0039

X1X2 166.67 1 166.67 52.61 0.0002

X1X3 318.62 1 318.62 100.58 0.0001

X2X3 143.16 1 143.16 45.19 0.0003

X12 0.12 1 0.12 0.039 0.8495

X22 287.85 1 287.85 90.87 0.0001

X32 60.74 1 60.74 19.18 0.0032

Lack of fit 4.98 3 4.98 2.76 0.1762

Pure error 1.81 4 7.23

R2 0.9803 C.V.% 2.42

Adjusted R2 0.9550 PRESS 250.39 aDegrees of freedom

Figure 5. Response surface plots for the optimization of HSCCC process. A) Effect of temperature and flow rate on purity of rutin at a constant revolution speed of 850 rpm; B) influence of revolution speed and temperature on purity of rutin at a definite flow rate of

1.5 mL/min; C) interaction between revolution speed and flow rate at a fixed temperature of 25 °C.


339

rate range of 1.2-1.4 mL/min when the constant rev-olution speed was 850 rpm (Figure 5A). The maximum response occurred at two areas: the rev-olution speed range of 800-850 rpm and temperature range of 26-30 °C; the revolution speed range of 850- -900 rpm and temperature range of 20-24 °C when the definite flow rate was 1.5 mL/min (Figure 5B). The response was increased with the increase in flow rate and revolution speed, and then declined with further increase in both parameters when a fixed tempe-rature was 25 °C (Figure 5C).

The optimum HSCCC conditions were obtained according to the model equation as follow: tempe-rature was 30 °C; flow rate was 1.11 mL/min; revol-ution speed was 800 rpm (the reason that no higher temperature and lower revolution speed were tested further lied in avoiding sample destroyed and sta-tionary phase loss).

Structure identification

The structure identification of compounds was performed with ESI-MS, 1H- and 13C-NMR. Data of each compound were given as follows:

Compound 1. The ESI-MS of compound 1 yielded ions with m/z 311 [M+Na]+. 1H-NMR (DMSO-d6, 400 MHz, δ / ppm): 12.22 (1H, s, 5-OH), 9.79 (2H, br s, 7-OH, 2′-OH), 7.22 (1H, d, J = 8.2 Hz, H-6′), 6.39 (1H, d, J = 2.1, H-5′), 6.31 (1H, dd, J = 2.2, 8.4 Hz, H-3′), 5.90 (2H, s, H-6, H-8), 5.62 (1H, dd, J = = 2.6, 13.0 Hz, H-2), 3.31 (1H, dd, J = 13.2, 17.0 Hz, H-3), 2.63 (1H, dd, J = 2.8, 17.1 Hz, H-3); 13C-NMR (DMSO-d6, 100 MHz, δ / ppm): 196.8 (C-4), 167.3 (C-7), 163.7 (C-5), 163.6 (C-9), 158.8 (C-2′), 155.5 (C-4′), 128.4 (C-6′), 115.6 (C-1′), 106.6 (C-5′), 102.7 (C-3′), 101.7 (C-10), 96.0 (C-6), 95.2 (C-8), 74.0 (C-2), 41.3 (C-3); the chemical data displayed above were consistent with steppogenin. These findings confirmed that the compound is steppogenin [16].

Compound 2. The ESI-MS of compound 2 yielded ions with m/z 161 [M+Na]+. 1H-NMR (DMSO-d6, 400 MHz, δ / ppm): 7.76–7.79 (1H, m, H-6), 7.47– –7.52 (1H, m, H-4), 6.93–6.95 (1H, m, H-3), 6.89–6.91 (1H, m, H-5); 1C-NMR (DMSO-d6, 100 MHz, δ / ppm): 171.4 (C-7), 160.7 (C-1), 135.3 (C-5), 129.9 (C-3), 118.9 (C-4), 116.8 (C-6), 112.7 (C-2); the chemical data displayed above were consistent with salicylic acid. These findings confirmed that the compound is Salicylic acid [17,18].

Compound 3. the ESI-MS of compound 3 yielded ions with m/z 633 [M+Na]+. 1H-NMR (CD3OD, 400 MHz, δ / ppm): 7.66 (1H, d, J = 2.0 Hz, H-2′), 7.62 (1H, dd, J = 8.4, 2.4 Hz, H-6′), 6.87 (1H, d, J = = 8.4 Hz, H-5′), 6.40 (1H, d, J = 2.0 Hz, H-8), 6.21 (1H, d, J = 2.0 Hz, H-6), 5.11 (1H, d, J = 9.6 Hz,

H-1′′), 4.52 (1H, d, J = 1.6 Hz, H-1′′′); 1C-NMR (CD3OD, 100 MHz, δ / ppm): 179.1 (C-4), 166.5 (C-7), 162.7 (C-5), 158.3 (C-2, C-9), 149.6 (C-3′), 145.6 (C-4′), 135.9 (C-3), 123.4 (C-6′), 122.9 (C-1′), 117.5(C-2′), 115.9 (C-5′), 104.6 (C-10), 102.3 (C-1′′), 99.8 (C-6, C-1′′′), 94.7 (C-8), 78.1 (C-5′′), 77.2 (C-3′′), 75.7 (C-2′′), 73.9 (C-4′′′), 72.2 (C-3′′′), 72.1 (C-2′′′), 71.8 (C-4′′), 69.7 (C-5′′′), 68.5 (C-6′′), 18.0 (C-6′′′); the chemical data displayed above were consistent with rutin. These findings confirmed that the compound is rutin [19].

Compound 4. The ESI-MS of compound 4 yielded ions with m/z 293 [M+Na]+. 1H-NMR (DMSO-d6, 400 MHz, δ / ppm): 12.99 (1H, s, 5-OH), 8.34 (1H, s, H-2), 7.41 (2H, d, J = 8.4 Hz, H-2′, H-6′), 6.85 (2H, d, J = 8.6 Hz, H-3′, H-5′), 6.40 (1H, d, J = 2.2 Hz, H-8), 6.25 (1H, d, J = 2.1 Hz, H-6); 13C-NMR (DMSO-d6, 100 MHz, δ / ppm): 180.3 (C-4), 164.8 (C-7), 162.2 (C-5), 157.8 (C-4′), 157.6 (C-9), 154.0 (C-2), 130.3 (C-2′), 130.3 (C-6′), 122.4 (C-1′), 121.4 (C-3), 115.3 (C-3′), 115.2 (C-5′), 104.5 (C-10), 99.2 (C-6), 93.9 (C-8); the chemical data displayed above were con-sistent with genistein. These findings confirmed that the compound is genistein [20,21].

Compound 5. The ESI-MS of compound 5 yielded ions with m/z 617 [M+Na]+. 1H-NMR (CD3OD, 400 MHz, δ / ppm): 7.98 (2H, br.d, J = 8.8 Hz, H-2′, H-6′), 7.30~7.42 (3H, m, H-3′′′, H-6′′′, H-8′′′), 6.81 (2H, br.d, J = 8.8 Hz, H-3′, H-5′), 6.79 (2H, br.d, J = = 8.4 Hz, H-5′′′, H-9′′′), 6.29 (1H, d, J = 2.0 Hz, H-8), 6.11 (1H, d, J = 2.0 Hz, H-6), 6.07 (1H, d, J = 16.0 Hz, H-2′′′), 5.22 (1H, d, J = 7.6 Hz, H-1); 1C-NMR (CD3OD, 100 MHz, δ / ppm): 179.0 (C-4), 168.5 (C-1′′′), 165.6 (C-7), 161.2 (C-5), 160.8 (C-4′, C-7′′′), 159.1 (C-9), 158.0 (C-2), 146.3 (C-3′′′), 134.9 (C-3), 132.0 (C-2′, C-6′), 130.9 (C-5′′′, C-9′′′), 126.9 (C-4′′′), 122.5 (C-1′), 116.6 (C-6′′′, C-8′′′), 115.9 (C-3′, C-5′), 114.6 (C-2′′′), 103.9 (C-10, C-1′′), 99.9 (C-6), 94.8 (C-8), 77.9 (C-3′), 75.6 (C-2′′, C-5′′), 71.7 (C-4′′), 64.3 (C-6′′); the chemical data displayed above were consistent with potengriffioside A. These findings confirmed that the compound is potengrif-fioside A [22].

Compound 6. The ESI-MS of compound 6 yielded ions with m/z 617 [M+Na]+. 1H-NMR (CD3OD, 400 MHz, δ / ppm): 8.05 (2H, d, J = 8.8 Hz, H-2′, H-6′), 6.88 (2H, d, J = 8.8 Hz, H-3′, H-5′), 6.40 (1H, d, J = 2.0 Hz, H-8), 6.20 (1H, d, J = 2.0 Hz, H-6), 5.13 (1H, d, J = 7.2 Hz, H-1′′), 4.51 (1H, d, J = 1.6 Hz, H-1′′′), 3.80 (1H, br d, J = 10.0 Hz, H-6), 3.63 (1H, dd, J = 3.2, 1.6 Hz, H-2′′′), 3.52 (1H, dd, J = 9.2, 3.6 Hz, H-3′′′), 3.45~3.47 (1H, m, H-5′′′), 3.42~3.44 (1H, m, H-2′′′), 3.40~3.41 (1H, m, H-3′′), 3.37~3.39 (1H, m, H-6′′), 3.32~3.36 (1H, m, H-5), 3.27~3.28 (1H, m,


340

H-4′′′), 3.23~3.26 (1H, m, H-4′′), 1.13 (3H, d, J = 6.0 Hz, H-6′′′); 1C-NMR (CD3OD, 100 MHz, δ / ppm): 179.1 (C-4), 166.7 (C-7), 162.7 (C-5), 161.2 (C-2), 159.1 (C-4′), 158.3 (C-9), 135.3 (C-3), 132.2 (C-2′, C-6′), 122.6 (C-1′), 117.0 (C-3′, C-5′), 104.4 (C-10), 102.3 (C-1′′), 99.8 (C-6), 94.8 (C-8), 78.1 (C-3′′), 77.1 (C-5′′), 75.7 (C-2′′), 73.8 (C-4′′′), 72.2 (C-3′′′), 72.0 (C-2′′′), 71.4 (C-3′′), 69.7 (C-5′′′), 68.5 (C-6′′), 18.0 (C-6′′′); the chemical data displayed above were consistent with kaempferol 3-O-rutinoside. These findings confirmed that the compound is kaempferol 3-O-rutinoside [23].

Compound 7. The ESI-MS of compound 7 yielded ions with m/z 269 [M-H]-. 1H-NMR (DMSO-d6, 400 MHz, δ / ppm): 12.07 (1H, s, 1-OH), 11.99 (1H, s, 8-OH), 11.36 (1H, s, 3-OH), 7.49 (1H, s, H-5), 7.17 (1H, s, H-7), 7.12 (1H, d, J = 2.4 Hz, H-4), 6.59 (1H, d, J = 2.0 Hz, H-2), 2.42 (1H, s, CH3);

1C-NMR (DMSO-d6, 100 MHz, δ / ppm): 187.8 (C-9), 180.9 (C-10), 165.1 (C-3), 164.0 (C-1), 161.0 (C-8), 147.9 (C-6), 134.8 (C-4a), 132.6 (C-10a), 123.8 (C-7), 120.2 (C-5), 113.2 (C-8a), 108.7 (C-9a), 108.5 (C-4), 107.7 (C-2), 21.4 (CH3); the chemical data displayed above were consistent with emodin. These findings con-firmed that the compound is emodin [24].

CONCLUSION

Response surface methodology is a combi-nation of statistical and mathematical techniques. The main advantage of RSM is the reduced number of experiments that are required to evaluate multiple parameters and their interactions. The present paper describes efficient separation and purification rutin from Hibiscus mutabilis L. leaves by the combination of RSM and HSCCC. The results fully demonstrated that the optimization using RSM would get more rea-sonable parameters for HSCCC than the classical one-variable-a-time optimization because it could sci-entifically evaluate the interaction between the vari-ables. Consequently, high purity rutin was obtained with the optimized conditions. Otherwise, it is the first time to apply HSCCC on H. mutabilis L. leaves separ-ation. Compared with previous studies, HSCCC shortens the separation time and firstly separates steppogenin and genistein from H. mutabilis L. leaves. In summary, the combination of RSM and HSCCC was successfully applied to optimize HSCCC para-meters and supply more information from a small number of experiments. As a result, the purity of pro-duct was dramatically improved. This method has good potential in separation and purification of effect-ive compounds from natural product.

REFERENCES

[1] L. Zhang, C.Z. Zhou, J. Changchun Univ. Tradit. Chin. Med. 29 (2013) 28-30

[2] Y.F. Sun, Ms. Thesis, Shandong University of Traditional Chinese Medicine, 2011, p. 42-45

[3] Z.C. Lin, C. Wang, W.Z. Zhu, X.Y. Wang, Y.D. Zhong, H.G. Wen, W.B. Yu, Z.Z. Ye, China Tradit. Patent Med. 9 (1982) 26-28

[4] L. Zhang, C.Z. Zhou, Guide Chin. Med. 11 (2013) 453- -455

[5] L.J. Yao, Y. Lu, Z.N. Chen, Chin. Tradit. Herb. Drugs 34 (2003) 201-203

[6] L.Y. Yao, G.Y. Wang, L.P. Wang, Chin. Tradit. Pat. Med. 22 (2000) 827-829

[7] X.B. Fu, Y.F. Sun, P. Zhang, C.Z. Zhou, Food Drug 14 (2012) 256-259

[8] J.X. Zheng, Y. Zheng, L. Zhang, S.Q. Zhao, Y.X. Fang, K. Zhang, Lishizhen Med. Mater. Med. Res. 23 (2012) 3006- -3007

[9] Y.M. Zhao, H.J. Zou, P.F. Zhu, Y.M. Xue, S. Ma, Q. Zhao, J. Yunnan Univ. Tradit. Chin. Med. 35 (2012) 7-9

[10] X.P. Dai, Q. Huang, B.T. Zhou, Z.C. Gong, Z.Q. Liu, S.Y. Shi, Food Chem. 139 (2013) 563-570

[11] Q. Wang, X.L. Cheng, H. Li, X.Y. Qin, C.Y. Ge, R. Liu, L.W. Qi, M.J. Qin, J. Pharm. Biomed. Anal. 75 (2013) 25-32

[12] M.J. Simirgiotis, G. Schmeda-Hirschmann, J. Borquez, E.J. Kennelly, Molecules 18 (2013) 1672-1692

[13] 13. P.Q. Kuang, D. Song, Q.P. Yuan, X.H. Lv, D. Zhao, H. Liang, Food Chem. 136 (2013) 309-315.

[14] M.I. Fernandez-Marin, R.F. Guerrero, M.C. Garcia-Par-rilla, B. Puertas, T. Richard, M.A. Rodriguez-Werner, P. Winterhalter, J.P. Monti, E. Cantos-Villar, Food Chem. 135 (2012) 1353-1359

[15] Y Ito, J. Chromatogr., A 1065 (2005) 145-168

[16] Z.P. Zheng, K.W. Cheng, J.T.K. To, H.T. Li, M.F. Wang,, Mol. Nutr. Food Res. 52 (2008) 1530-1538

[17] G. Vermeersch, J. Marko, B. Cartigny, F. Leclerc, P. Roussel, M. Lhermitte, Clin. Chem. 34 (1988) 1003-1004

[18] M.M.A. Hassan, M.U. Zubair, Specgtrosc. Lett. 15 (1982) 533-542

[19] X.W. Yang, M.H. Han, Y.P. Jin, J. Chin. Med. Mater. 30 (2007) 797-800

[20] A. Caligiani, G. Palla, A. Maietti, M. Cirlini, V. Brandolini, Nutrients 2 (2012) 280-289

[21] A.P. Mazurek, J.Cz. Dobrowolski, J. Sadlej, E. Bednarek, L. Kozerski, J. Mol. Struct. 520 (2000) 45-52

[22] H.J. Zhong, J.J. Chen, H.Y. Wang, S.D. Luo, Chin. Tradit. Herb. Drugs. 31 (2000) 488-490

[23] M. Juan-Badaturuge, S. Habtemariam, M.J.K. Thomas, Food Chem. 125 (2011) 221-225

[24] K. Danielsen, D.W. Aksnes, G.W. Francis, Magn. Reson. Chem. 30 (1992) 359-363.


341

ZHUONI HOU1,2

XIANRUI LIANG1

FENG SU1

WEIKE SU1

1Key Laboratory for Green Pharmaceutical Technologies and

Related Equipment of Ministry of Education, College of Pharmaceutical

Sciences, Zhejiang University of Technology, Hangzhou, China

2College of Life Sciences, Zhejiang Sci-Tech University, Hangzhou, China

NAUČNI RAD

PREPARATIVNA IZOLACIJA I PREČIŠĆAVANJE SEDAM JEDINJENJA IZ LIŠĆA Hibiscus mutabilis L. DVOSTEPENOM HROMATOGRAFIJOM BRZE IZMENE STRUJE

Sedam jedinjenja iz lišća Hibiscus mutabilis L. je uspešno izolovano dvostepenom hroma-

tografijom brze izmene struje koristeći dvofazni sistem rastvarača koji se sastojao iz

n-butanol-etil-acetat–voda (zapreminski odnos 1:6:9) i n-heksan-etil-acetat-metanol-voda

(zapreminski odnos 3:5:3:5). Kritični eksperimenatalni parametri za prvi stepen odvajanja

optimizovani metodom odzivne površine su sledeći: protok 1,1 ml/min, 800 rpm i tempe-

ratura 30 °C. Pod optimalnim uslovima dobijeno je oko 5,0 mg salicilne kiseline, 13,6 mg

rutina i 5,5 mg genisteina iz 100 mg sirovog uzorka. Onda je drugom separacijom dobijeno

9,2 mg potengrifiozida A, 4,7 mg kaempferol 3-O-rutinozida, 3,0 mg stepogenina i 2,5 mg

emodina. Čistoća ovih sedam jedinjenja određena je UPLC hromatografijom bila je: 96.2%

za salicilnu kiselinu, 93.8% za rutin, 95.4% za genistein, 94.3% za potengrifiozid, 98.0% za

kaempferol 3-O-rutinozid, 94.1% za stepogenin i 90.8% za emodin. Njihova hemijska

struktura određena je pomoću ESI-MS i pomoću 1H- i 13C-NMR. Za stepogenin i genistein

ovo je prvi podatak da su izolovani iz H. mutabilis L. Razvijena metoda je jednostavna,

efikasna i pomalo mučan separacioni proces.

Ključne reči: Hibiscus mutabilis L.; hromatografija brze izmene struje; rutin; genistein, kaempferol-3-O-rutinozid, stepogenin.





343

RUIFANG ZHAO YULONG WANG

YONGHUI BAI YONGFEI ZUO LUNJING YAN

FAN LI

State Key Laboratory Breeding Base of Coal Science and

Technology Co-founded by Shanxi Province and the Ministry of

Science and Technology, Taiyuan University of Technology, Taiyuan,

China

SCIENTIFIC PAPER

UDC 662.764:66

DOI 10.2298/CICEQ140614035Z

EFFECTS OF FLUXING AGENTS ON GASIFICATION REACTIVITY AND GAS COMPOSITION OF HIGH ASH FUSION TEMPERATURE COAL

Article Highlights • NBFA and CFA were used to decrease AFT of particular coal • The AFT of two coals was the minimum when 5 wt.% of CFA or NBFA was added • Gas mole ratio of H2/CO is adjusted by controlling fluxing agent amount Abstract

A Na-based fluxing agent Na2O (NBFA) and a composite fluxing agent (mixture of CaO and Fe2O3 with mass ratio of 3:1, CFA for short) were used to decrease the ash fusion temperature of the Dongshan and Xishan coal from Shanxi of China and make these coal meet the requirements of the specific gasification process. The main constituents of the fluxing agents used in this study can play a catalyst role in coal gasification. Thus, it is necessary to understand the effect of fluxing agents on coal gasification reactivity and gas composition. The results showed that the ash fusion temperature of the two coal used decreased to the lowest point due to the eutectic phenomenon when 5 wt.% of CFA or NBFA was added. Simultaneously, the gas molar ratio of H2/CO changed when CFA was added. A key application was thus found where the gas molar ratio of H2/CO can be adjusted by controlling the fluxing agent amount to meet the synthetic requirements for different chemical pro-ducts.

Keywords: ash fusion temperature, fluxing agents, coal gasification, reactivity, gas composition.

The gasification technology is now developing towards high temperatures and pressures with slag-tapping and coal with high ash fusion temperature (AFT) account for a significant portion of the world coal reserves, especially in China where about 50% of the coal reserves are not suitable for slag-tap gasi-fication [1]. So, it is important to decrease the AFT to about 50-100 K lower than the gasification tempe-rature to meet the requirement of the slag-tapping technology for specific coal.

Currently, adding fluxing agents is one of the most effective ways that can decrease the AFT of a

Correspondence: F. Li, State Key Laboratory Breeding Base of Coal Science and Technology Co-founded by Shanxi Province and the Ministry of Science and Technology, Taiyuan University of Technology, Taiyuan 030024, China. E-mail: [email protected] Paper received: 14 June, 2014 Paper revised: 3 September, 2014 Paper accepted: 23 September, 2014

particular coal and the fluxing agents, such as Fe2O3, CaO and Na2O frequently used, are also the main constituents of the coal ash [2]. Hattingh et al. [3] found the catalytic effect of the inorganic constituents such as the catalytically active species K+, Na+, Ca2+, Mg2+ and Fe3+ is one of the major factors that controls the reactivity of the coal. According to previous research results [4], fluxing agents also have sig-nificant impacts on char reactivity and gas compo-sition due to their catalysis. Karimi et al. [5] dis-covered that gasification reactivity of coal and chars are a function of porosity (hence the surface area), crystal structure of the fixed carbon, as well as the catalytic effect of the ash. Skodras and Sakellaro-poulos [6] deemed that the catalytic activity of inor-ganic minerals is dependent on their concentration, dispersion and chemical form in the coal matrix. Wang et al. [7] investigate that the addition of cat-alysts not only increases the gasification rate but also

R. ZHAO et al.: EFFECTS OF FLUXING AGENTS ON GASIFICATION… Chem. Ind. Chem. Eng. Q. 21 (2) 343−350 (2015)

344

increases the H2 production, which means the catal-ysts can not only change the gasification reactivity but will impact on the gas composition. Sakawa et al. [8] proposed that the “alkali index” could quantify the catalytic effect of the inherent inorganic constituents of coal and found that the CO2 gasification rate inc-reased as the alkali index increases. Once the tempe-rature went higher than the coal ash softening tempe-rature, the catalytic effect would drop off for the min-eral particles in the coal ash was transferred from high activity small particles to relatively less activity agglomerates. Tang et al. [9] came to the conclusion that when the temperature was increased up to 1823 K, the influence of fluxing agent on Pingdingshan coal reactivity almost disappeared because of the fusion and agglomeration of fluxing agent at higher tempera-ture.

Syngas generated from coal gasification often has a relatively low hydrogen content and the gas mole ratio of H2/CO is in the range of approximately 0.7-1.1. For Fischer-Tropsch synthesis of liquid fuels or methanol, the required H2/CO ratios is 2.0 or higher. In order to obtain the required H2/CO ratios, the water–gas shift reaction is indispensable. How-ever, this would lead to expensive expenditure on both management and maintenance, so once the H2/CO ratio could be adjusted by the way of adding a fluxing agent, it would achieve considerable benefit to the whole production technology. In addition, both the operating temperature and pressure of the commonly used commercial entrained-bed gasifier are extremely high to ensure a high carbon conversion, which has brought many challenges to the gasifier materials. For instance, almost all gasifier users identify the nozzles used in gasifiers is one of the weakest links in the process for achieving high on-stream availability fac-tors. Because of high temperature, the gasifier refrac-tory brick often fails [10]. Considering that the fluxing agent can change the reaction pathway, reduce the reaction activation energy and accelerate the reaction rate, it can indirectly influence the gasification reacti-vity and the gas composition. Furthermore, it can be used in catalytic fields to facilitate energy conserva-tion and emission reduction.

The majority of the previous published work [11,12] firstly took a demineralization (acid treatment) and impregnation method to study the catalytic acti-vity of minerals in coal gasification. Treated in acid solutions, cations such as calcium that combined with carboxyl functional groups in the organic structures are replaced by hydrogen atoms of the oxygen-con-taining functional groups [13-14]. Therefore, the rem-oval of mineral matter from coal by acid treatment is

inevitably accompanied by structural changes, which would definitely change the physical property, and thus, the gasification behavior. In order to avoid the impacts caused by acid treatment and the impreg-nation caused, the coal used in this study is raw coal without any destructive pretreatment.

Based on the above-mentioned discussion, the fluxing agent can not only decrease the AFT, but also act as a catalyst. The aim of this work is firstly to optimize fluxing agents on the base of previous work and give the mechanism of decreasing AFT briefly, then focus on the impacts that imposed on gasifi-cation reactivity and the gas composition, which is of great significance to the synthesis of coal chemical products.

EXPERIMENTAL

Material

Dongshan and Xishan coal with high ash fusion temperature were chosen for this study. The ash fusion temperature is more than 1773 K in a reducing atmosphere. The properties of coal samples are shown in Tables 1 and 2. Single fluxing agent inc-luding Fe2O3, CaO and Na2O from the decomposition reaction of Na2CO3 and all chemicals applied were analytical grade. The composite fluxing agent was obtained by mechanical mixing of CaO and Fe2O3 at different mass ratios of 1:1, 2:1, 3:1 and 4:1.

Table 1. Proximate analysis (air dried basis) of the samples (%)

Sample Moisture Ash Volatiles Fixed carbon

Dongshan 0.9 8.7 12.0 78.4

Xishan 2.2 9.2 23.3 65.3

Table 2. Chemical composition of ash (wt.%)

Sample Component

SiO2 Al2O3 Fe2O3 CaO MgO TiO2 SO3 K2O Na2O

Dongshan 43.36 33.70 13.08 3.13 0.71 1.28 0.75 0.35 0.10

Xishan 48.78 40.64 5.06 0.17 0.13 1.70 0.88 0.23 1.30

Procedure and detection method

The gasification process was conducted in a high temperature simultaneous thermal analysis NETZSCH-STA449F3, which consisted of a thermo balance, vapor generator, temperature controller, vapor furnace and high temperature furnace. The weight and temperature precision were 0.01 mg and 0.01 K, respectively. Measuring mass profiles of char samples at high temperature during pyrolysis and gasification reaction with CO2 and H2O was thus possible. After gasification, the produced gas was


345

analyzed by gas chromatography with a thermal conductivity detector and flame ionization detector. The experimental conditions were as follows: coal sample: 15 mg, gasifying agent: steam (30 ml/min), gasification temperature: 1173, 1223, 1273, 1323 and 1373 K, protecting gas: Ar (20 ml/min), purge gas: Ar (20 ml/min), heating rate: 10 K/min, char residence time: 30 min.

Dongshan and Xishan coal were mixed with 5 wt.% Na-based fluxing agent Na2O (NBFA) or mixture of CaO and Fe2O3 with a mass ratio of 3:1 (CFA) and then gasified isothermally at 1173, 1223, 1273, 1323, and 1373 K at atmospheric pressure. The coal sample that had been mixed with the fluxing agent was fed into the plate from the top of the sample carrier in the thermogravimetric analysis (NETZSCH STA449F3). It was heated up to the desired tem-perature and kept for 30 min to prepare the char. The gasification agents, namely steam were injected and reacted with the char for 16 min. Product gases during char gasification process were collected by a gas bag every two minutes as the beginning of gasi-fication and the chemical species of CO, H2, CH4 and CO2 gases were analyzed by the GC to monitor the gas composition.

The coal ash that passed through a 100-mesh miller sieve, was put into a cup and homogenized by 10% dextrin solution. It was then placed in the mold and pressed evenly to shape an ash cone or pyramid. The sample was dried in air and the AFT was measured from room temperature to 1773 K with a SJHR-3 Smart Ash Fusion Analyzer in an oxidizing atmosphere. After the fusion test, the ash samples were taken out from the ash fusion analyzer and cooled down with ice water. The samples were then dried and stored for XRD analyses. The crystal struc-

ture of the sample was analyzed by X-ray diffraction (XRD) using CuKα radiation with the condition of 40 kV, 40 mA. The scanning rate was 4°/min and the scanning range was 10-80°.


Effect of single alkaline oxides fluxing agent on AFT

The AFT of a certain coal mainly depends on the contents of oxides in the coal ash. In general, the AFT goes higher with the contents of acid oxides inc-reased and lower with the contents of alkaline oxides increased reversely [15-17]. From Figure 1 it can be seen that the fusing impact of the three fluxing agents on the AFT of Dongshan ash is dramatically different, and the order of the fluxing effect is Na2O > CaO > > Fe2O3, when single fluxing agent was added to the Dongshan coal with different mass fractions. The reasons may be explained as follows: Na2O reacts with minerals in the coal ash to produce nephelite with low melting point. The reaction of nephelite, quartz and mullite in the coal ash would lead to eutectic phenomenon at about 1373 K [18,19].

When an Fe-based fluxing agent such as Fe2O3 was added, the AFT of the coal ash in a weak red-ucing atmosphere was found to be about 20-150 K lower than that in the oxidizing atmosphere [20]. The Fe2O3 can be reduced to FeO and then will react with aluminium silicates in the coal ash at high tempera-ture to form the mixture of substances with low melt-ing point [21], such as Fe2SiO4 and Fe2Al2O4. For CaO, Figure 1 shows that the AFT of the Dongshan- -CaO mixtures gradually decreases with the increase of the CaO content until it reaches approximately 5 wt.%, when the fluid temperature is a minimum. After that, it increased gradually with the increase of the

3 4 5 6 7 81300

1400

1500

1600

1700

1800

DS+Fe2O3

XS+Fe2O3

DS+CaO XS+CaO DS+Na2OXS+ Na2O

Flu

id t

emp

erat

ure

(K)

Content of fusing agent (%)

Figure 1. Effects of Na-based, Fe-based and Ca-based fluxing agents on the AFT of Dongshan and Xishan coal.


346

CaO content. This is due to the formation of mullite (3Al2O3⋅2SiO2) which produced by the reaction of SiO2, Al2O3 and CaO at 1273 K [22,23]. It undergone a gradual decomposition process with the tempera-ture rise. At about 1473 K, anorthite appeared accom-panied by a small amount of mullite [24].

The same principle is also applicable to the Xishan coal. Na2O, as an ideal fluxing agent, could decrease the AFT of Dongshan and Xishan coal, and a maximum of AFT reduction occurred with an ideal amount of 5 wt.% Na2O. The relationship between the char reactivity and the catalytic effects of NBFA was also investigated.

Effect of composite fluxing agent on AFT

Commonly, there is not enough Fe2O3 and CaO in the coal ash to form the quaternary mixture system of SiO2-Al2O3-Fe2O3-CaO to decrease the AFT. A synergistic effect is shown in Figure 2 when different mass ratios of CaO and Fe2O3 were used to decrease the AFT, and the optimum ratio of CaO and Fe2O3 is 3:1. Figure 2 shows that the effect of decreasing the AFT of the composite fluxing agents is much better than that any of the single fluxing agents of calcium or iron when the additive amount of the composite fluxing agents is higher than 5%. The same principle is also applicable to the Xishan coal when a mixture of CaO and Fe2O3 with mass ratio of 3:1 was added. From these results the Dongshan and Xishan coal with a 5 wt.% CFA will be used to investigate the rel-ationships between the char reactivity and the catal-ytic effects.

The mechanism of decreasing the Dongshan coal AFT by CFA

Coal ash is a complex mixture of crystalline and non-crystalline materials. The crystalline transform-ation process of the newly generated mineral matter

at high temperature can determine the AFT to a large extent. Figure 3 clearly shows the main components of the Dongshan coal ash at different temperatures with and without the addition of CFA. The results indicate that the main mineral matter consists of quartz, hematite, mullite and gehlenite in the raw coal ash, while there is ematite, anorthite and esseneite in the ash with 5 wt.% CFA. The comparison between samples (a) and (b) in Figure 3 shows that the main mineral matter in the Dongshan raw coal ash is SiO2, 3Al2O3⋅2SiO2 and Fe2O3 at the temperature ST (soften temperature), which are all high melting point oxides and that lead to a high AFT of a certain coal. In the Dongshan coal ash with 5 wt.% CFA added, the main mineral matter in the coal ash are some low melting point ternary or quaternary mixtures of SiO2- -Al2O3-Fe2O3-CaO resulting in relatively low AFT min-erals, such as anorthite (CaO⋅Al2O3⋅2SiO2) [25] and esseneite (2CaO⋅Fe2O3⋅Al2O3⋅2SiO2) [26].

Generally speaking, with a fluxing agent added to the coal, not only the generation of high melting point silicate and mullite were inhibited, but the production of low-melting eutectics were formed and this is the essential reason why CFA was used to reduce the AFT.

Effects of CFA and NBFA on char reactivity

Carbon conversion of the Dongshan char versus time at 1173 and 1223 K with and without the addition of 5 wt% fluxing agent is given in Figure 4, showing that the gasification reactivity of the Dongshan char changed remarkably with the addition of the ideal flux-ing agents CFA and NBFA. With addition of 5 wt% CFA and NBFA, the carbon conversion of the Dong-shan char increased nearly 8 times for NBFA, but slightly for CFA. The catalytic ability of the NBFA is much more evident than that of the CFA at both 1173 and 1223 K. The reactivity order is raw char+5%

3 4 5 6 7 81550

1600

1650

1700

1750

a: Dongshan coal

Content of fusing agent (% )

Flu

id t

emp

erat

ure

(K)

CaO Fe2O3

CaO/Fe2O3=2:1 CaO/Fe2O3=3:1 CaO/Fe2O3=4:1

3 4 5 6 7 81550

1600

1650

1700

1750 b: Xishan coal

Flu

id t

empe

ratu

re (K

)

C ontent of fusing agent (% )

C aO Fe2O 3

C aO /Fe2O 3=2:1 C aO /Fe2O 3=3:1 C aO /Fe2O 3=4:1

Figure 2. Effect of composite fusing agents on the AFT of Dongshan and Xishan coal.


347

NBFA > raw char+5% CFA > raw char. The presence of the alkaline cations such as K+, Na+, Ca2+, Mg2+ and Fe3+ on the surface of the char can increase the gasification reactivity. Adanez and Diego [27] dis-covered that the raw char conversion rate was much higher than for the demineralized chars. Wang et al. [28,29] also proved the existing of the catalytic effects of alkali and alkaline metals on steam gasification.

As can be seen from Figure 4, the carbon con-version in the presence of NBFA did not increase to such a large extent with increasing temperature, while the carbon conversion increased noticeably in the presence of CFA, which means the activity effect of CFA at high temperature is much better. When the

temperature is higher than 1124 K (the melting point of Na2CO3), NBFA could react with minerals to pro-duce sodium salt and the decomposition and volatil-ization tendency would decrease once the tempera-ture goes higher than 1273 K. The NBFA is more appropriate for use at low-temperature gasification. Because of the thermal stability at high temperature, CFA can be used as the catalyst of high temperature char catalytic gasification reaction.

Effects of fluxing agents on gas composition

In order to evaluate the impact of a fluxing agent on gas composition, gasification reaction and water- -gas shift reaction during steam gasification, a para-

10 20 30 40 50 60 70 80

1088 K-a

1088 K-b

1323 K-b

1323 K-a

1523 K-b

1523 K-a

ST-b

ST-a

•

•

2Theta/degree

Figure 3. XRD diagram of Dongshan coal ash with 5 wt.% CFA ( SiO2; Fe2O3; CaO·Al2O3·2SiO2; 2CaO·Al2O3·SiO2; 3Al2O3·2SiO2; 2CaO·Fe2O3·Al2O3·2SiO2; CaO); a – Dongshan coal ash; b – Dongshan+5% CFA.

0 4 8 12 160

20

40

60

80

100

Car

bon

con

vers

ion

(%

)

Reaction time (min)

none(1173K)

5% CFA(1173K)

5% NBFA(1173K)

none(1233K)

5% CFA(1233K)

5% NBFA(1233K)

Figure 4. Effect of fluxing agents on Dongshan steam gasification reactivity.


348

meter K value was introduced for an ideal case; namely, only reactions (1) and (2) occurred.

2 2C + H O = CO + H (1)

2 2 2CO + H O = CO + H (2)

Clemens et al. [30] discovered that the exist-ence of calcium iron could promote the water-gas shift reaction to a equilibrium during steam gasific-ation and then influence the gas composition, which means the metal iron has catalytic effects on the water-gas shift reaction and gasification reaction. Before evaluation, the following equations are defined as follows:

CO concentration: = −COn Rg Rs (3)

H2 concentration: = +2Hn Rg Rs (4)

CO2 concentration: =2COn Rs (5)

K value: + −= = =2 2

2 2

CO H 1 CO %

2 CO 2CO %

Rg n nK

Rs n (6)

where Rg is the produced gas concentration during steam gasification reaction, and Rs is the produced gas concentration during the water-gas shift reaction.

Assuming that the gasification reaction and water-gas shift reaction took place simultaneously, the Rg and Rs could reflect the reaction rate of gas-ification and water-gas shift, respectively, and from the K expression, once the concentration of CO2 is given, it can be easily used to predict the K-value and indirectly to evaluate the impact of catalysts on gas composition, gasification reaction and water-gas shift reaction during steam gasification. In theory, the larger the K value, the faster the gasification rate is. Table 3 shows the impacts that CFA and NBFA imposed on K value and CO2 concentration in the gas. Obviously, the K of Dongshan raw coal is relati-vely low compared with that of 5 wt.% CFA and NBFA added.

Effects of CFA on gas composition

Gas compositions of Dongshan coal with and without CFA are shown in Figure 5. For the Dongshan coal, the H2/CO molar ratio firstly increases slowly and then decreases quickly with rising temperature. When the temperature goes higher than 1223 K, the increase of the H2/CO molar ratio is inhibited to some extent and the total yield of syngas increase rapidly until the temperature is higher than 1273 K. In Figure 5, it is shown that the H2/CO mole ratio interval

Table 3 Effects of CFA and NBFA on K value and CO2 concentration in the gas

T / K Dongshan raw coal Dongshan coal with 5 wt.% CFA Dongshan coal with 5 wt.% NBFA

CO2 (vol.%) K CO2 (vol.%) K CO2 (vol.%) K

1373 12.8 3.41 15.81 2.66 12.18 3.61

1323 21.52 1.82 20.25 1.97 15.31 2.77

1273 23.2 1.66 23.1 1.66 22.25 1.75

1223 24.15 1.57 28.31 1.27 29.61 1.19

1173 56.56 0.38 39.74 0.76 41.59 0.7

1150 1200 1250 1300 1350 1400

1.5

2.0

2.5

3.0

3.5

H2+

CO

(V%

)H

2/CO

Temperature(K)

H2/CO(withCFA)

H2/CO(without CFA)

H2+CO(with CFA)

H2+CO(without CFA)

40

50

60

70

80

90

Figure 5. Gas composition of Dongshan coal with and without CFA.


349

broadened from 1.85-2.12 to 1.50-3.75 without and with the addition of CFA, which is of great significance to meet the H2/CO molar ratio of synthetic require-ments for different coal chemical products, such as alcohols, alkanes and ketones. Besides that, with CFA added, the tendency that the increase of H2/CO molar ratio is inhibited sharply by temperature. It means the CFA can accelerate the water-gas shift reaction rate and decelerate steam gasification rate and then influence the gas composition.

CONCLUSIONS

NBFA and CFA were used to decrease ash fusion temperature of the Dongshan and Xishan coal and gasification reactivities and gas compositions of two coal with NBFA and CFA were investigated. The results showed NBFA and CFA were not only two kinds of ideal fluxing agents which can be effectively used to reduce the AFT but improve gasification reac-tivity. The main conclusions are summarized as follows:

1. The AFT of the Dongshan and Xishan coal both decreased to under 1623 K from higher than 1773 K with 5 wt.% CFA and NBFA added, and the reason can be attributed to a eutectic phenomenon.

2. Coal char reactivity can be enhanced by both CFA and NBFA, but that role of CFA is prominently at higher temperature, and the catalytic ability of the NBFA is much more evident at low temperature.

3. The introduction of CFA makes the gas molar ratio of H2/CO change obviously with operation tem-perature increase, which could meet the synthetic requirements for different chemical products. To Dongshan coal, with the addition of 5 wt.% CFA, the range of H2/CO mole ratio is extended from 1.5 to 3.75, which is extremely useful as Fischer-Tropsch synthesis gas from fluid beds gasifier.

Acknowledgments

The authors gratefully acknowledge the National National High Technology Research and Develop-ment Program of China, 863 Program (No. 2015AA050503).

REFERENCES

[1] K. Liu, C. Song, V. Subramani, Hydrogen and Syngas Production and Purification Technologie. John Wiley & Sons, Hoboken, NJ, 2010, pp. 206

[2] J.C. van Dyk, L.L. Baxter, J.H.P. Van Heerden , R.L.J. Coetzer, Fuel 84 (2005) 1768-1777

[3] B.B. Hattingh, R.C. Everson, H.W. Neomagus, J.R. Bunt, Fuel Process. Technol. 92 (2011) 2048-2054

[4] S. Wu, X. Zhang, J. Gu, Y. Wu, J. Gao, Energy Fuels 21 (2007) 1827-1831

[5] A. Karimi, N. Semagina, M. R. Gray, Fuel 90 (2011) 1285-1291

[6] G. Skodras, G.P. Sakellaropoulos, Fuel Process. Technol. 77 (2002) 151-158

[7] J. Wang, K. Sakanishi, I. Saito, T. Takarada, K. Morishita, Energy Fuels 19 (2005) 2114-2120

[8] M. Sakawa, Y. Sakurai, Y. Hara, Fuel 61 (1982) 717-720

[9] L.H. Tang, N. Zhang, F.M. Wang, X.D. Zhu, D.J. Jiang, Z.B. Zhu, J. Fuel Chem. Technol. 31 (2003) 209-213

[10] Z.H. Dai, X. Gong, X.L. Guo, H.F. Liu, F.C. Wang, Z.H. Yu, Fuel 87 (2008) 2304-2313

[11] L.R. Radovic, P.L. Walker, R.G. Jenkins, Jr., J. Catal. 82 (1983) 382-394

[12] Z.H. Wu, Y. Sugimoto, H. Kawashima, Fuel 82 (2003) 2057-2064

[13] P.L. Walker, Jr., Fuel 63 (1984) 1214-1220

[14] E.J. Hippo, R.G. Jenkins, P.L. Walker, Fuel 58 (1979) 338-344

[15] Y. Ninomiya, A. Sato, Energy Conv. Manage. 38 (1997) 1405-1412

[16] X.L. Guo, Z.H. Dai, X. Gong, X.L. Chen, H.F. Liu, F.C. Wang, Fuel Process. Technol. 88 (2007) 451-459

[17] J.C. van Dyk, S.A. Benson, M.L. Laumb, F.B. Waanders, Fuel 88 (2009) 1057-1063

[18] F. Gao, Y.J. Ma, J. Fuel Chem. Technol. 40 (2012) 1161- -1166

[19] J. Mao, M.H. Xu, F. Li, J. Huazhong Univ. Sci. Technol. 31 (2003) 59-62

[20] V.R. Nigel, W. Fraser, W. Jim, Fuel 81 (2002) 673-681

[21] Q.Q. Wang, P.J. Zeng, Coal Convers. 20 (1997) 32-37

[22] J.C. van Dyk, F.B. Waanders, J.H.P. van Heerden, Fuel 87 (2008) 2735-2744

[23] J.B. Li, B.X. Shen, H.X. Li, J.G. Zhao, J.M. Wang, J. Fuel Chem. Technol. 37 (2009) 262-265

[24] J.B. Li, B.X. Shen, L.H. Xun, J.G. Zhao, J.M. Wang, Coal Conver. 37 (2009) 262-265

[25] J.C. van Dyk, F.B. Waanders, K. Hack, Fuel 87 (2008) 2388-2393

[26] A. Bhattacharjee, H. Mandal, M. Roy, J. Kusz, M. Zubko, P. Gütlich, J. Magn. Magn. Mater. 322 (2010) 3724-3727

[27] J. Adanez, L.F. Diego, Int. Chem. Eng. 33 (1990) 656- -662

[28] J. Wang, M.Q. Jiang, Y.H. Yao, Y.M. Zhang, J.Q. Cao, Fuel 88 (2009) 1572-1579

[29] J. Wang, M.Q. Jiang, Y.H. Yao, Y.M. Zhang, J.Q. Cao, Fuel 89 (2010) 310-317

[30] A.H. Clements, L.F Damiano, T.W. Matheson, Fuel 77 (1998) 1017-1020.


350

RUIFANG ZHAO

YULONG WANG

YONGHUI BAI

YONGFEI ZUO

LUNJING YAN

FAN LI

State Key Laboratory Breeding Base of Coal Science and Technology Co-

founded by Shanxi Province and the Ministry of Science and Technology,

Taiyuan University of Technology, Taiyuan, China

NAUČNI RAD

UTICAJ AGENASA ZA FLUKSOVANJE NA REAKTIVNOST GASIFIKACIJE I SASTAV GASA KOD UGLJA SA VISOKOM TEMPERATUROM FUZIJE PEPELA

Agensi za fluksovanje na bazi natrijuma Na2O (NBFA) i kompozitni (mešavina CaO i Fe2O3

sa masenim odnosom 3:1; CFA) su korišćeni za smanjenje temperature fuzije pepela

Dongshan i Xishan uglja iz Shanxi pokrajine u Kini, kako bi ovaj ugalj ispunio zahteve

specifičnog procesa gasifikacije. Glavni sastojci agenasa za fluksovanje korišćeni u ovom

radu mogu imati i ulogu katalizatora u procesu gasifikacije uglja. S tim u vezi, neophodno

je razumeti uticaj agenasa za fluksovanje na reaktivnost gasifikacije uglja i sastav gasa.

Rezultati su pokazali da se temperatura fuzije pepela dva uglja smanjuje do najniže tačke

zbog eutektičkog fenomena kada je dodato 5 mas.% CFA ili NBFA. Istovremeno se sa

dodatkom CFA menja i molski odnos H2/CO. Ključno je pronaći kada se molski odnos

H2/CO može podesiti kontrolom količine agensa za fluksovanje, kako bi se ispunili sinte-

tički zahtevi za različite hemijske produkte.

Ključne reči: temperatura fuzije pepela, agensi za fluksovanje, gasifikacija uglja, reaktivnost, sastav gasa.





351

VESELINKA V. GRUDIĆ NADA Z. BLAGOJEVIĆ

VESNA L. VUKAŠINOVIĆ-PEŠIĆ

SNEŽANA R. BRAŠANAC

Faculty of Metallurgy and Technology, Podgorica,

Montenegro

SCIENTIFIC PAPER

UDC 543.552:547.475.2:66:635.64

DOI 10.2298/CICEQ140712037G

KINETICS OF DEGRADATION OF ASCORBIC ACID BY CYCLIC VOLTAMMETRY METHOD

Article Highlights • Reduced concentration of AA increases with the increase of temperature and storage

time • The kinetics of the oxidation reaction corresponds to the reaction of first order • Substances present in peppers reduced degradation of AA Abstract

Cyclic voltammetry was used to examine the kinetics of degradation of ascor-bic acid (AA) at different temperatures. It has been shown that the reduction of the concentration of AA in all temperatures follow the kinetics of the first order reaction. The rate constant of the oxidation reaction increases with tempera-ture as follows: 5×10-5, 2×10-4, 1×10-3, and 3×10-3 min-1 at temperatures of 25, 35, 65 and 90 °C, respectively. The temperature dependence of the rate cons-tant follows the Arrhenius equation, and the value of activation energy of the reaction degradation is 48.2 kJ mol-1. The effect of storage time at a tempera-ture of 90 °C on AA content in fresh juice of green peppers was investigated. It was shown that AA oxidation reaction in the juice is also the first order reac-tion, while the lower rate constant in relation to the pure AA (5×10-3 min-1) indi-cates the influence of other substances present in the peppers.

Keywords: ascorbic acid, kinetics, cyclic voltammetry, green peppers.

The reduction of food quality with storage time is becoming an increasing problem because of the loss of nutrients and vitamins [1]. Vitamin C, known as AA, is a water soluble vitamin. It is a powerful antioxidant because of its reducing properties and is involved in many biological reactions in the human body [2]. Therefore, it is recognized as one of the most important vitamins and its recommended daily intake is 75-90 mg for an adult [3]. However, the human body is unable to synthesize vitamin C [4], and insufficient intake of this vitamin causes numerous abnormalities [2].

AA is very unstable and is readily oxidized under the influence of light, heat, oxygen [4] and storage time [5,6]. The mechanism of thermal degradation of AA is quite complex and not fully understood. In the presence of oxygen, AA easily oxidizes to dehydro-ascorbic acid (DHA) via its monoanion [7] and the rate

Correspondence: V.L. Vukašinović-Pešić, Faculty of Metallurgy and Technology, Cetinjski put bb, 20000 Podgorica, Monte-negro. E-mail: [email protected] Paper received: 12 July, 2014 Paper revised: 19 September, 2014 Paper accepted: 8 October, 2014

at which DHA is formed is approximately first order with respect to the concentration of AA, oxygen and metal catalysts. Several authors indicate a negative effect of the oxygen on the quality of fruit juices, which is connected to the reduction of AA [8,9], increased browning [10] and the growth of aerobic bacteria [11]. However, while AA is readily oxidized to the DHA, the loss of its vitamin properties occurs after hydrolysis of DHA to form 2,3-diketogulonic acid [12]. Decarbox-ylation of diketogulonic acid may result in the form-ation of xylosone and 3-deoxypentaone. The first reaction product (xylosone) is further degraded to form reductones and ethylglyoxal, and 3-deoxy-pentaone degrades into furfural and 2-furancarboxylic acid [7]. These components may combine with amino acids to form brown pigments [12]. A number of stu-dies suggest that the browning of juice during storage can be associated with the degradation of AA [13-15]. Limacher et al. [16] showed that the main product of heating the dry AA is furan, while we got smaller amounts of furan in the cooking under pressure. The most probable intermediates of AA degradation are 2-deoxyaldoteroses, 2-furoic acid and 2-furaldehyde. The formation of furan as the main products of ther-

V.V. GRUDIĆ et al.: KINETICS OF DEGRADATION OF ASCORBIC ACID… Chem. Ind. Chem. Eng. Q. 21 (2) 351−357 (2015)

352

mal decomposition of AA during baking, cooking or pyrolysis, was confirmed in several studies [17-19].

The importance of vitamin C in human meta-bolism incited numerous studies to determine the content of this vitamin in fruits, vegetables and juices. They used primarily spectrophotometric [20-22], titri-metric and electrochemical methods [22-28]. A num-ber of papers devoted to the quantitative determin-ation of vitamin C in different samples of food and beverages by cyclic voltammetry indicated that this method has a number of advantages compared to other methods, being simple, inexpensive, accurate and selective.

The kinetics of degradation of vitamin C was also studied using titrimetric and spectrophotometric methods [29-35]. In the present study, we used cyclic voltammetry method to examine the kinetics of deg-radation of AA. The temperature and storage time were the varying parameters. The results showed that cyclic voltammetry presents a fast, simple, inexpen-sive and selective method of investigation of kinetics of AA degradation.

EXPERIMENTAL

All the cyclic voltammetric measurements were performed using a Potentiostat & Galvanostat Model 273 coupled with a Pentium IV personal computer. Pt disk (1 mm) sealed in a Teflon plate (2 mm) was used as working electrode. Pt foil, with the surface of 1 cm2, was used as an auxiliary electrode, and a saturated calomel electrode (SCE) was the reference electrode.

The measurements were performed in solutions with ascorbic acid concentration of 10 mmol dm-3 obtained by dissolving the appropriate amount of pure acid (Merck) in distilled water. The solutions were

thermostated at the appropriate temperature (25, 35, 65 and 90 °C), and at specified time intervals the amounts of 25 ml were taken, cooled to room tem-perature, and the I-E curves were recorded. In all experiments, KCl (Merck) was used at a concen-tration of 0.34 mol dm-3 as an auxiliary electrolyte.

The calibration curve was obtained by recording the voltammograms of AA solutions in the concen-tration range 2-10 mmol dm-3, obtained by successive dilution of the stock solution (10 mmol dm-3).

The pepper juice obtained by squeezing fresh green pepper was then filtered and thermostated at 90 °C. I-E curves were recorded in the same way as in the case of pure AA.


Figure 1a shows the I-E curves obtained in the AA solutions with the concentration range of 2-10 mmol dm-3, with the addition of 0.34 mol dm-3 of KCl solution as an auxiliary electrolyte. The anodic peaks at 0.45 V versus SCE is attributed to the oxidation of AA to DHA, in accordance with literature data [23-28].

Based on Figure 1a, a calibration curve was drawn to show the dependence of the current density of the anodic current peak on the concentration (Figure 1b). The anodic peaks at 0.45 V were found to vary linearly with ascorbic acid concentrations. The high value of the regression coefficient (R2 = 0.9968) confirms the validity and legitimacy of cyclic voltam-metry for quantitative determination of AA in different samples.

The kinetics of the degradation reactions of AA at different temperatures are shown in Figure 2.

Figure 2 shows a decrease in oxidation current peak of solution with time of storage at all investi-gation temperatures. Also, at high temperatures,

a) b)

Figure 1. a) Cyclic voltammograms obtained on Pt electrode for different AA concentrations, expressed as mmol dm-3: 2(1), 4(2), 6(3), 8(4) and 10(5) in 0.34 mol dm-3 KCl, t = 25 °C; b) the calibration line presenting the dependence of peak current of anodic AA oxidation

on AA concentration in the 0.34 mol dm-3 KCl solution at t = 25 °C.


353

shorter period of time is required for the reduction of the concentration of AA during storage compared to the time at the room temperature. Thus, for example, at 90 °C the reduction in the concentration of AA after 120 min is 33%, and after 300 min it is 70%. At room temperature, the ascorbic acid degradation occurs much more slowly: after 6 days the reduction of the concentration is 32% and after 12 days, the percent-age is 56%. This fact indicates the significant impact of heat on the stability of AA.

During storage the solution of AA visually changes colour: a colourless solution became yellow, and the colour became more intensive with time. The change in the colour of the solution is accompanied by the change in voltammograms. Namely, on the voltammograms shown in Figure 2, the presence of

the second anodic peak current is observed after a certain storage time of AA solution. According to the literature data, the change of the colour of the solution, or the presence of second current peak may be explained by the further oxidation of DHA, namely the formation of furan-type compounds, resulting from thermal degradation of AA under aerobic conditions [9-19].

The dependences of ln (cτ/c0) as a function of time (Figure 3) have high values of R2, indicating that the kinetics of the degradation of AA at all tested temperatures can be described by first-order kinetic model:

τ=d

dnc

kc (1)

c) d)

Figure 2. Cyclic voltammograms obtained on a Pt electrode in the solutions of 10 mmol dm-3 AA in 0.34 mol dm-3 KCl, at the temperature of 90 (a), 65 (b), 35 (c) and 25 ° C; d) during storage; the current maxima decreased as time increased in the order of

times: a): 0 (1), 20 (2), 40 (3), 90 (4), 120 (5), 180 (6), 270 (7), 300 min (8), b): 0 (1), 120 (2), 240 (3), 480 (4), 720 (5), 1200 (6), 1320 min (7), c): 0 (1), 12 (2), 24 (3), 36 (4), 48 (5), 56 (6), 80 (7), 150 (8), 160 h (9), d): 0 (1), 66 (2), 96 (3), 144 (4), 160 (5), 215 (6),

260 (7), 280 (8), 450 h(9).


354

where c is the concentration, k is the rate constant, and n is the order of reaction. The degradation reac-tion is first order if n = 1, and by integration of Eq. (1) one obtains the equation of the first order reaction as follows:

τ τ

= − 0

lnc

kc

(2)

The dependence of ln (cτ/c0) as a function of time is the curve whose slope determines the rate constant of the investigation reaction, where c0 and cτ are the concentrations of ascorbic acid at the begin-ning of the reaction and after time τ, respectively.

The dependence of reaction rate constant on the temperature is given by the Arrhenius equation:

( )= 0 exp aE RTk A (3)

where Ea is the activation energy, R is the universal gas constant, T is temperature and A0 is a constant.

The first-order degradation reaction of AA is in accordance with the works of other authors [30-34,36- -38].

From the slope of the straight line, the values of the rate constants of the oxidation reaction at different temperatures were calculated. The half-life of the reaction (τ1/2), i.e., the time required to reduce the concentration of AA in half compared to the initial

value is calculated based on the value of the rate constant as 0.693/k [31]. The calculated values of k and τ1/2 at different temperatures are shown in Table 1. With the increase in temperature the rate constant of the reaction increases, while the reaction half-life decreases. Thus, a temperature rise of 25 to 90 °C can cause an increase of the rate constant of 5×10-5 to 3×10-3 min-1, while the half-life of the reaction dec-reases from 231 h to 231 min within the same tempe-rature range.

Table 1. The kinetic parameters of the degradation reactions AA

t / °C k / min-1 τ1/2 / min R2

25 5×10-5 13860 0.996

35 2×10-4 3465 0.996

65 1×10-3 693 0.999

90 3×10-3 231 0.998

Figure 4a shows cyclic voltammograms obtained at different concentrations of AA in Tafel coordinates E = f (log j). These data were used for drawing the dependency logj = f(log c) (Figure 4b). The slope of this dependence is 1.113 and this confirms the previous observation that the kinetics of the AA degradation reaction can be described by a first-order kinetic model.

c) d)

Figure 3. The logarithm of concentration as a function of time for spontaneous degradation reaction of AA at different temperatures: a) 90, b) 65, c) 35 and d) 25 °C, indicating a first-order reaction.


355

The Arrhenius plot of the AA degradation reac-tion is shown in Figure 5. The slope of the line leads to the value of activation energy of the reaction, which is 48.204 kJ mol-1.

2.7.10-3 2.9.10-3 3.1.10-3 3.3.10-3 3.5.10-3

T-1 / K-1

Figure 5. Arrhenius plot for the thermal degradation reactions of AA in 0.34 mol dm-3 KCl solution, as determined by cyclic

voltammetry.

Cyclic voltammograms obtained in fresh pepper juice with the addition of 0.34 mol dm-3 KCl on the Pt electrode at different times of storage at 90 °C are shown in Figure 6a. Based on the calibration curve (Figure 1a), the concentration of AA in the pepper juice is 5.38 mmol dm-3.

As well as in the pure AA, in the pepper juice the reduction of anode current peak corresponding to the oxidation of AA indicates a decrease of AA content in the juice over time. The straight-line dependence ln (cτ/c0) as a function of storage time (Figure 6b) with R2 value of 0.998 indicates that the reaction kinetics can be described by the first-order kinetic model. From the obtained linear dependence the rate constant of the degradation reaction of AA in pepper juice is cal-culated (5×10-3 min-1), which corresponds to the half time of the reaction of 138.6 min.

The lower value of the rate constant of the oxi-dation reaction of AA in pepper juice in comparison to pure AA can be explained by the presence of various

a)

b)

Figure 4. a) The Tafel plot of the dependence of anodic current maxima on potential of the voltammograms of AA oxidation on Pt surface in the solutions 0.34 mol dm-3 KCl, for the concentration of AA, of 2(1), 4(2), 6(3), 8(4) and 10(5) mmol dm-3 ; b) The dependence log j

versus log c of anodic oxidation of AA, in the solutions 0.34 mol dm-3 KCl, at temperature of 25 °C, yielding the reaction order.

a) b)

Figure 6. a) Cyclic voltammograms obtained on Pt electrode in fresh pepper juice in 0.34 mol dm-3 KCl at 90 °C. The anodic peaks decrease with time storage in order of times: 0 (1), 30 (2), 60 (3), 120 (4), 150 (5), 200 min (6); b) the dependence of ln (cτ/co) on time of

the reaction of AA degradation in 0.34 mol dm-3 KCl solution at 25 °C in the pepper juice indicating the first-order kinetics.


356

substances in the juice, which stabilize AA, slowing its degradation [31,35].

CONCLUSION

The paper proposes a fast and simple method for examination of the kinetics of degradation of AA. The experimental studies in this paper have shown that the degradation of AA, whether pure or contained in the pepper juice, increases with the increase of temperature and storage time. The kinetics of the oxi-dation reaction in both cases corresponds to the reaction of first order and the rate constant of the oxi-dation reaction increases with temperature. The lower value of the rate constant in the case of pepper juice in comparison to pure AA indicates a significant effect of the substances present in the pepper on the stab-ility of AA.

REFERENCES

[1] B. M. Laing, D. L. Schlueter, T. P. Labuza, J. Food Sci. 43 (1978) 1440-1443

[2] C.S. Erdurak-Kiliç, B. Uslu, B. Dogan, U. Ozgen, S.A. Ozkan, M. Coskun, J. Anal. Chem. 61 (2006) 1113-1120

[3] K.L. Mahan, S. Escozz – Stump, Krause,s food, nutrition and diet therapy. 11th ed. Saunders, New York, 2004, p. 1321

[4] G.L. Robertson, C.M.L. Samaniego, J. Food Sci. 51 (1986) 184-187

[5] L.R. Gordon, M.C. Samaniego-Esguerra, J. Food Qual. 13 (1990) 361-374

[6] S.J. Padayatty, R. Daruwala, Y. Wang, P.K. Eck, J. Song, W.S. Koh, M. Levine, In: E. Cadenas and L. Packer (Eds.), Handbook of Antioxidants Second Edition, Marcel Dekker, New York, 2002

[7] K. Zerdin, M.L. Rooney, J. Vermue, Food Chem. 82 (2003) 387-395

[8] J.F. Kennedy, Z.S. Rivera, L.L. Lloyd, F.P. Warner, K. Jumel, Food Chem. 45 (1992) 327-331

[9] O. Solomon, U. Svanberg, A. Sahlstrom, Food Chem. 53 (1995) 363-368

[10] S. Meydev, I. Saguy, I. Kopelman, J. Agric. Food Chem. 25 (1977) 602-604

[11] M.N.U. Eiroa, V.C.A. Junqueira, F.L. Schmidt, J. Food Protect. 62 (1999) 883-886

[12] S.R. Tannebaum, M.C. Archer, V.R. Young, Vitamins and minerals, In: O.R. Fennema (Ed.), Food Chem (2nd ed.), Marcel Dekker, New York, 1985, p. 488

[13] K. Clegg, J. Sci Food Agric. 17 (1966) 546-549

[14] K. Clegg, A. D. Morton, J. Sci. Food Agric. 16 (1965) 191- -198

[15] D.J. Trammell, D.E. Dalsis, C.T. Malone, J. Food Sci. 51 (1986) 1021-1023

[16] J. Limacher, B. Kerler, Conde-Petit, I. Blank, Food Addit. Contam. 24 (S1) (2007) 122-135

[17] C.P. Locas, V.A. Yaylayan, J. Agric. Food Chem. 42 (2004) 6830-6836

[18] A. Becalski, S. Seaman, J. AOAC Int. 88 (2005) 102-106

[19] J. Mark, Ph. Pollien, Ch. Lindinger, I. Blank, T. Mark, J. Agric. Food Chem. 54 (2006) 2786-2793

[20] K. Shrivas, K. Agrawal, D. Kumar Patel, J. Chin. Chem. Soc. 52 (2005) 503-506

[21] M. Salkić, R. Kubiček, Eur. J. Sci. Res. 23 (2008) 351- -360

[22] A. Selimović, M. Salkić, A. Selimović, Int. J. Basic Appl. Sci. 11 (2011) 125-131

[23] M. Ogunlesi, W. Okiei, L. Azeez, V. Obakachi, M. Osun-sanmi, G. Nkenchor Int. J. Electrochem. Sci. 5 (2010) 105-115

[24] J.R. Esch, J.R. Friend, J.K. Kariuki, Int. J. Electrochem. Sci. 5 (2010) 1464-1474

[25] W. Okiei, M. Ogunlesi, L. Azeez, V. Obakachi, M. Osun-sanmi, G. Nkenchor, Int. J. Electrochem. Sci. 4 (2009) 276-287

[26] A.M. Pisoschi, A. Pop, G.P. Negulescu, A. Pisoschi, Molekules 16 (2011) 1349-1365

[27] J. Wawrzyniak, A. Ryniecki, W. Zembrzuski, Acta Sci.. Pol., Technol. Aliment. 4 (2005) 5-16

[28] H. Ernst, M. Knoll, Anal. Chim. Acta 449 (2001) 129-134

[29] N.P. Bineesh, R.S. Singhal, A. Pandit, J. Sci. Food Agric. 85 (2005) 1953-1958

[30] A. Steškova, M. Morochovičova, E. Leškova, J. Food Nutr. Res. 45 (2006) 55-51

[31] N. Matei, A. Soceanu, S. Dobrinas, V. Magearu, Ovidius Univ. Ann. Chem. 20 (2009) 132-136

[32] O.O. Faramade, Afr. Crop Sci. Conf. Proc. 8 (2007) 1813- -1816

[33] H.S. Lee, G.A. Coates, Food Chem. 65 (1999) 165-168

[34] S. Rahmawati, B. Bundjali, Indo. J. Chem. 12 (2012) 291- -296

[35] R. Paul, U. Ghosh, Indian J. Biotechnol. 11 (2012) 309- -313

[36] H.S. Ramaswami, F.R. Van De Voort, S. Ghazala, J Food Sci. 54 (1989) 1322-1326

[37] M.A.J.S. Van Boekel, J. Food Sci. 61 (1996) 477-489

[38] A. Anbessie, K. Siraj, S.A. Kitte, Adv. Sci. Eng. Med. 12 (2013) 1312-1315.


357

VESELINKA V. GRUDIĆ

NADA Z. BLAGOJEVIĆ

VESNA L. VUKAŠINOVIĆ-PEŠIĆ

SNEŽANA R. BRAŠANAC

Metalurško-tehnološki Fakultet, Podgorica, Crna Gora

NAUČNI RAD

ISPITIVANJE KINETIKE DEGRADACIJE ASKORBINSKE KISELINE CIKLIČNOM VOLTAMETRIJSKOM METODOM

U radu je predložena brza, jednostavna, jeftina i selektivna metoda ciklične voltametrije za

ispitivanje kinetike degradacije askorbinske kiseline (AA) na različitim temperaturama.

Pokazano je da smanjenje koncentarcije AA na svim temperaturama prati kinetiku reakcije

I reda. Izračunate vrijednosti konstante brzine ispitivane reakcije oksidacije rastu sa poras-

tom temperature i iznose 5×10-5; 2×10-4; 1×10-3 i 3×10-3 min-1 na temperaturama 25, 35, 65

i 90 °C, redom. Temperaturna zavisnost konstante brzine slijedi Arenijusovu jednačinu, a

vrijednost aktivacione energije ispitivane reakcije degradacije iznosi 48,2 kJ mol-1. Takođe

je ispitan i uticaj vremena stajanja na temperaturi od 90 °C na sadržaj AA u svježem soku

zelene paprike. Pokazano je da je reakcija oksidacije AA u soku zapravo reakcija prvog

reda, dok niža vrijednost konstante brzine u odnosu na čistu AA (5×10-3 min-1) ukazuje na

uticaj supstanci prisutnih u paprici na stabilnost AA.

Ključne reči: askorbinska kiselina, kinetika, ciklična voltametrija, zelena paprika.





359

AHMET OZAN GEZERMAN BURCU DIDEM

ÇORBACIOĞLU

Yildiz Technical University, Department of Chemical

Engineering, Faculty of Chemical and Metallurgical Engineering,

Istanbul, Turkey

SCIENTIFIC PAPER

UDC 546.17:543.544:66

DOI 10.2298/CICEQ140705038G

EFFECTS OF SODIUM SILICATE, CALCIUM CARBONATE AND SILICIC ACID ON AMMONIUM NITRATE DEGRADATION AND ANALYTICAL INVESTIGATIONS OF THE DEGRADATION PROCESS ON AN INDUSTRIAL SCALE

Article Highlights • Different compounds were added to prevent the degradation of ammonium nitrate • Multiple instrumental analyses were used to monitor the degradation process • We describe a method to limit the degradation properties of sulfuric acid Abstract

Ammonium nitrate is an inorganic chemical that has numerous applications in different industries. However, various problems are associated with both the production and subsequent storage of ammonium nitrate, including caking, degradation, unwanted phase transition, and recrystallization. Although several methods have been developed to attempt to solve these problems, many of them fail to work in practice. In this study, different compounds including silicic acid and sodium silicate were added to slow the progress of or to prevent the degradation of ammonium nitrate. Multiple instrumental analyses such as ion chromatography and scanning electron microscopy were used to monitor the degradation process.

Keywords: ammonium nitrate, caking, degradation, ion chromato-graphy, downstream process.

The fertilizer industry has developed in many ways according to the requirements of the agricultural industry in the 20th and 21st centuries. However, these advances have created additional problems owing to increased usage conditions, including reactions of chemicals used during production processes and technical problems associated with fertilizer storage and manufacturing methods intended to reduce costs. Currently, many universities, institutes, and research and development centers are working to find solutions to these problems of fertilizer usage.

Simultaneously, various chemical compositions have been developed to improve the chemical pro-perties of fertilizers, including their chemical stability. For example, the reaction of calcium silicate, which is

Correspondence: A. Ozan Gezerman, Yildiz Technical Univer-sity, Department of Chemical Engineering, Faculty of Chemical and Metallurgical Engineering, Istanbul, Turkey. E-mail: [email protected] Paper received: 5 July, 2014 Paper revised: 24 October, 2014 Paper accepted: 28 October, 2014

obtained from an aqueous solution of sodium silicate and calcium chloride, with ammonium nitrate pro-duced a fertilizer salt with good chemical stability [1]. In other studies, the addition of calcium silicate was observed to increase the nutriment properties of ammonium nitrate fertilizer [2]. In a similar study, C1- -C6 alcanoic acid or sodium silicate salts of this acid were sprayed on ammonium nitrate fertilizer [3], and in another study, Joseph et al. used sodium silicate salts to solve the degradation problem by adding 20% of a sodium silicate solution to ammonium nitrate during the production process [4]. Rinkenbach also applied sodium silicate to prevent the degradation of ammonium nitrate by treating 100 parts ammonium nitrate with 0.5–5 parts sodium silicate at 130 °C [5]. Adam investigated methods for increasing the nutrient value of soil by using an ammonium nitrate solution and calcium silicate [6]. Another study applied a poly-meric solution containing sodium silicate to coat the surface of ammonium nitrate particles to improve the degradation properties [7].

A. OZAN GEZERMAN, B.D. ÇORBACIOĞLU: EFFECTS OF SODIUM SILICATE… Chem. Ind. Chem. Eng. Q. 21 (2) 359−367 (2015)

360

The most commonly used chemical reducing agents in fertilizer production are chrome, zinc, iron(II) sulfate, salicylic acid, and silicic acid [8]. Silicic acid (Si(OH)4) is a weak acid employed in the fertilizer pro-duction process to increase the nutrient value and as a reducing agent [8]. The Kjeldahl manufacturing method is an example of the use of this reagent. In this method, sodium thiosulfate (Na2S2O3), salicylic acid, or silicic acid reagents are added during anal-ysis to account for nitrite and nitrate. Gezerman et al. added silicic acid and a calcium lignosulfonate sol-ution to an ammonium nitrate melt to test the degra-dation process [9].

The reducing agent Na2S2O3 reacts with H2SO4 to generate sulfur dioxide (SO2), preventing the formation of double salts. In this process, 1 g Na2S2O3.5H2O consumes 0.5 mL H2SO4. If silicic acid and salicylic acid are used, these reagents also act as reducing agents for H2SO4, generating CO2 and water. During this process, 1 g salicylic acid or 1 mL silicic acid consumes 5.6 mL H2SO4.

Known patented processes for the dilution of fertilizer solution involve the addition of calcium car-bonate [10]. However, we are not aware of any patented or otherwise published processes for limiting the degradation properties of sulfuric acid, which is a required processing component. In this paper, we describe a method for the dilution of fertilizer solution with a mixture of calcium carbonate, sodium silicate, and silicic acid to limit the degradation properties of sulfuric acid and produce a more stable ammonium nitrate fertilizer solution.

The vacuum concentration method [10], which is the most commonly used method in the fertilizer

industry, was used to investigate the effects of the different chemical additives. According to this method, 70–85% concentrated ammonium nitrate solution, formed through the reaction between 55% HNO3 and gaseous anhydrous ammonia, can be concentrated up to 97% (Figure 1). Upon obtaining the maximum concentration of ammonium nitrate, calcium carbo-nate, sodium silicate, and silicic acid were added to the bulk solution. In fertilizer manufacture processes in the literature, the addition of calcium carbonate is a known method for the dilution of ammonium nitrate solution.

On an industrial scale, sodium silicate, which is readily available and can be used as an alternative to or together with calcium carbonate, is used in fertilizer production. The addition of silicic acid to a fertilizer solution with a dilution material such as calcium car-bonate and sodium silicate, and its effects on the amount of sulfuric acid, are investigated analytically in this study for the first time.

Main causes of fertilizer caking and degradation

The mechanism of fertilizer caking involves the formation of contact points between particles in the bulk fertilizer. These contact points are formed by three contact forces: phase contact, adhesive contact, and surface diffusion [11]. Phase contact refers to the formation of crystal bridges between fertilizer par-ticles, and is commonly believed to be the main cause of fertilizer caking problems [11]. Crystal bridges occur through processes such as recrystallization, dissolution, and heating during the reaction to pro-duce the product. Adhesive contact arises from mole-cular movements between contacting surfaces, and is

Ammonia ( NH3, 22 ° C, 8 kg/cm2)

Figure 1. Ammonium nitrate production process (concentrated method by vacuum).


361

more commonly referred to as the van der Waals force. This type of contact typically occurs because of the plasticity of fertilizer particles and pressure in the bulk. Fertilizer particles participating in adhesive con-tact can be destroyed by small forces during handling. Surface diffusion occurs because of the formation of salt complexes from ion hydrates, which results in surface diffusion processes that promote the self-adhesion of fertilizer particles [11].

Besides these, another factor that causes caking of fertilizer particles is the water activity between the particles. Water activity or deliquescence refers to the dissolution of a material by water abs-orption at high water activity. If the relative humidity under the storage conditions increases, water con-densates around the solid particles, causing them to dissolve. Because of the increased surface contact, cohesion between the particles is initiated, and con-sequently, caking progresses. Caking, which con-tinues to proceed by the absorption of water, results in deliquescence [12].

Some materials such as nitrogenous fertilizers absorb a small amount of water via hydrogen bonding at low relative humidity, although they exist as liquid solutions. When the relative humidity increases, the solid particles in the condensate layer are dissolved to result in a saturated solution. If relative humidity increases, because of vapor condensation, a greater amount of melt solution is formed. This sequence of processes continues until complete deliquescence. Deliquescence can accelerate physical changes in the particles and chemical reactions between the particles [12].

Several methods for addressing caking prob-lems have been developed thus far. Some reports describe improvements in only the physical proper-ties, whereas others suggest improvements in fertil-izer nutrient values. The purpose of the present study was to develop methods to prevent caking by chang-ing the physicochemical properties of fertilizer par-ticles to inhibit the contact phenomena described above. Through the addition sodium silicate, the vis-cosity and surface tension of NH4NO3 was expected to decrease, reducing the van der Waals forces between NH4NO3 molecules, and thus, preventing adhesive contact. Furthermore, the addition of silicic acid was expected to prevent salt complex formation, thereby inhibiting surface diffusion processes and preventing the caking of the fertilizer.

EXPERIMENTAL

Materials

Anhydrous ammonia (99.9%), nitric acid (55%), sodium silicate, silicic acid, calcium carbonate and

sulfuric acid were obtained from Merck, Istanbul, of the highest grade available, and used without further purification.

Ammonium nitrate production process

Anhydrous NH3 is pressurized in a vaporizer to 5.6 bar, and is then allowed to react with 55% HNO3. The resulting solution in the reactor is 80% NH4NO3. This melt is brought to the required concentration by evaporation exchangers. Sodium silicate solution can be added during the production process in one of the two evaporation stages. In the present study, sodium silicate was added in the first evaporation stage, as described in the literature [9]. Calcium ammonium nitrate (CAN) containing 26% nitrogen was prepared by using 250 kg CaCO3/t NH4NO3 and NH4NO3 con-taining 33% nitrogen was prepared using 60 kg CaCO3/t NH4NO3 [10]. In this process, the reaction between NH4NO3 and CaCO3 produced foam, which was removed by adding H2SO4.

Diluted solutions of sodium silicate were added together with calcium carbonate during the ammo-nium nitrate production process. Sodium silicate, in addition to its dilution properties, changes the prilling process, which operates at a rotation speed of 300 rpm. This was different from the effects of CaCO3, as seen from the sieving analysis results (Tables 1 and 2), and was attributed to the increased prill particle sizes. For the minimization of production costs, as reported by patented studies [4], 20% sodium silicate was added into the reaction mixture.

Experimental methods

Sodium silicate was added to a solution of 84% NH4NO3. Silicic acid was first added to concentrated H2SO4 (1 mL silicic acid/5.6 mL H2SO4), and this sol-ution was then added to the prepared fertilizer sol-ution. Sulfuric acid is used to inhibit the formation of CO2 during dilution of the fertilizer with CaCO3; how-ever, SO4

2– promotes the formation of double salt complexes in solution, so silicic acid is added to H2SO4 to minimize this. Without further production processes, the NH4NO3 fertilizer was mixed with sodium silicate. This was followed by crystallization at a relative humidity of 75–95% and a fertilizer/water ratio of 2:1. The materials used for examination of the effects of internal additives were sodium silicate, sil-icic acid, H2SO4 (98%), anhydrous NH3 (99.9%), HNO3 (55%) and CaCO3.

Electron microscopy analysis

Electron microscopy was performed to inves-tigate how sodium silicate, silicic acid, and sulfuric acid affected the fertilizer surface according to ASTM E986-97 [13] (Figures 2–4). For this analysis, a Carl

A. OZAN GEZERMAN, B.D. ÇORBACIOĞLU: EFFECTS OF SODIUM SILICATE… CI&CEQ 21 (2) 359−367 (2015)

362

Table 1. Screened analysis of final production and final production after one year for ammonium nitrate prill particles, which contain calcium carbonate

Period, months 3.35 mm, % 2.5 mm, % 2.0 mm, % 1.0 mm, % 0.5 mm, % E.A. mm, % Crushing Strength

SO42–

ppm N%

Final production

3.7 42.1 34.6 18.9 0.7 0 2.11 6.6028 26.52

Final pro-duction after one year

4.5 43.0 37.5 13.8 1.2 0 1.96 6.4317 26.33

Table 2. Screened analysis of final production and final production after one year for ammonium nitrate prill particles, which contain calcium carbonate, sodium silicate, and silicic acid

(a) (b)

Figure 2. SEM image of ammonium nitrate prill particle in which: a) calcium carbonate is added (after production) and b) calcium carbonate and sodium silicate are added (after production).

(a) (b)

Figure 3. SEM image of ammonium nitrate prill particle in which: a) silicic acid, calcium carbonate and sodium silicate are added (after production) and b) calcium carbonate is added (one year after production, under storage condition).

Period, months

3.35 mm, % 2.5 mm, % 2.0 mm, % 1.0 mm, % 0.5 mm, % E.A. mm, %Crushing Strength

SO42-

ppm N %

SiO2

ppm

Final production

7.4 50.4 39.5 2.3 0.4 0 2.35 1.0004 26.49 0.1786

Final pro-duction after one year

7.1 50.1 38.4 3.7 0.7 0 2.12 0.9987 26.31 0.1783


363

(a) (b)

Figure 4. SEM image of ammonium nitrate prill particle in which: a) calcium carbonate and sodium silicate are added (one year after production, under storage condition) and b) silicic acid, calcium carbonate and sodium silicate are added (one year after production,

under storage condition).

Zeiss DSM-960A scanning electron microscope was used. The technical properties of this electron microscope are as follows: accelerating voltage, 1–30 kV; useful magnification, 10–30000×; resolution, 70 Å.

Ion chromatography analysis

For detection of the numbers of anions and cat-ions of ammonium nitrate, ion chromatography was performed according to ASTM E1151-93 [14] (Figure 5) with a Shimadzu Prominence HIC-NS instrument, which has a measuring range of 0.01–51200 μS/cm and a flow rate control range of 0.001–5 mL/min.

Screen sieve analysis

Fertilizer prills containing silicic acid, sulfuric acid, and sodium silicate were stored for one year (Table 1) and then subjected to screening analysis with a Vibratory Sieve Shaker AS 200 according to the ASTM E11-09 standard [15].

Analysis of sulfate

During fertilizer particle production, the amounts of sulfate salts require monitoring, since these species cause the degradation of NH4NO3. The detec-tion method, based on the precipitation of barium sulfate (BaSO4) in fertilizer solution consisted of: dil-ute HCl (d20 = 1.18 g/mL), BaSO4 solution (122 g/L) and AgNO3 solution (5 g/L).

Analysis samples were prepared by adding HCl (20 mL) to H2SO4 (50 mL); this solution was diluted with demineralized water to 300 mL. The prepared solution was then boiled. Barium chloride (BaCl2; 20 mg) in water (20 mL) was added slowly, and the solution was boiled for a few minutes. The hot sol-ution was allowed to stand for 1 h. Once this solution became clear, it was filtered, and the resulting pre-cipitate was rinsed multiple times with hot water until no chloride was present in the filtrate, as determined

(a) (b)

Figure 5. Ion chromatogram of ammonium nitrate in which: a) calcium carbonate is added (after production on storage conditions) and b) calcium carbonate, sodium silicate, and silicic acid are added (after production, under storage condition).


364

by the use of AgNO3. The filter paper and precipitate were placed in a porcelain cup in an oven at 500 °C for 1.5 h. The material was then left to cool in an iso-lated place.

Calculations: 1 mg BaSO4 precipitate = 0.137 mg sulfur or 0.343 mg SO3.

Sulfate in fertilizer was determined by:

=

1

2

(%) 0.0137 )V

S W MV

(1)

=3(%) (%)2.5SO S (2)

where W is the mass of BaSO4 in the precipitate, V1 the sample mass (volume), V2 the total volume, and M the sample mass.

Analysis of silicic acid

The reagents used were dilute HCl (50%), con-centrated H2SO4 (98%) and HF (48%).

A sample containing silicate (0.25 g) was placed in a platinum cup and fired in an oven at 500 °C for 1 h. The sample was rinsed twice with 50% HCl (10 mL). Demineralized hot water was then added to the solution to give a volume of 100 mL. Subse-quently, all the water was evaporated in a water bath, and the cup was left in the bath. Next, 8–10 drops of HCl were added, and the sample was rinsed with water (50 mL).

The container with the solution was covered with a cover glass.This solution was then filtered through a filter paper (Whatman No. 42). The filter paper was placed in a platinum cup and 2–3 droplets of concen-trated H2SO4 were added. This filter paper was fired slowly and kept in a muffle oven at 1000 °C for 0.5 h. The residue was cooled and weighed, and then two droplets of concentrated H2SO4 were added. Subse-quently, 48% HF was added to a 25 mL portion of the original volume. Silicium fluoride was vaporized in a conventional oven at 250 °C for 1 h. This solution was kept in a muffle oven at 1000 °C for 0.5 h, and was then cooled and weighed.

Calculations:

2SiO (mg/L) = 100lost weight /sample amount (mL) (3)

Analysis of nitrogen

Fertilizer samples of 7 g were used. The fertil-izer was diluted to 500 mL with demineralized water. A 10 mL aliquot was taken from this solution, to which 20% NaOH (50 mL) was added. The solution was titrated with standard H2SO4.

Calculations:

( ) ( )= −% 50N A F (4)

where A is the volume of NaOH used in titration and F is the dilution factor.


In the ammonium nitrate production process as supported by the EFMA (European Fertilizer Manuf-acturing Association), ammonium nitrate fertilizer is obtained as a result of the reaction between anhyd-rous ammonia (99.9%) and nitric acid (55%). Accord-ing to commercial patents [10], various methods are used during production, and studies often focus on achieving the lowest costs and dilution rates depend-ing on the manufacturers. The ammonium nitrate production process involves reaction parts that have several inputs and parameters. First, anhydrous ammonia is brought to process conditions through oxidation in the presence of a catalyst to form nitric oxide. This nitric oxide is then reacted with air to form nitrogen dioxide, which is mixed with water to give nitric acid. The acid is then reacted with gaseous anhydrous ammonia to yield ammonium nitrate salts. The other processes include the capturing of elec-trical energy and steam from the energy created during the production of nitric acid from ammonia.

In the present study, the method of concen-tration by vacuum is applied, which is the preferred method in industrial ammonium nitrate production, and has the lowest cost and wastewater problems from an ecological viewpoint [10]. In this vacuum concentration method, all suggested chemical addi-tives that are used for dilution and to decrease the caking problem are added to the fertilizer solution after a suitable concentration is reached (99%, Figure 1), to control the concentration levels of the ammo-nium nitrate solution (which is concentrated on two evaporation stages) and to protect against any cor-rosive effect or other process problems for the pro-cess pipeline and equipment. Calcium carbonate is the most commonly used material added for the dil-ution of fertilizer solution. It is added to the fertilizer solution to compensate for the lack of lime.

Silicate is another required chemical material. For addition to fertilizer solution, the reaction between carbonate and silicate must be known and taken into account. In the analysis conducted, no gas production or foaming reaction was observed upon the addition of sodium silicate, which is used to transform silicate material into fertilizer solution, and the electron micro-scopy images showed that the obtained fertilizer prills had a spherical shape. Another chemical used for investigating its behavior in fertilizer solution was sil-icic acid. There have been some works on silicic acid for fertilizer production [9]. Silicic acid is added to fer-


365

tilizer bulk solution to decrease the amount of sulfuric acid, which forms a double salt with nitrate ions. Sul-furic acid is added to the fertilizer solution to consume the carbon dioxide created by the addition of calcium carbonate (which is used for diluting the fertilizer sol-ution). Electron microscopy (Figures 2–4) and screen sieve analysis (Tables 1 and 2) show that the deg-radation tendency decreases upon addition of silicic acid, while the nutriment value increases; the silicic acid also dilutes the solution and reduces the caking tendency. In terms of fertilizer solution, this work is the first investigation on an industrial scale using ins-trumental analyses such as ion chromatography, electron microscopy, and screen sieve analysis.

The storage of ammonium nitrate after pro-duction also presents several challenges, as it is a strong oxidant, and readily absorbs moisture. This process is related to the phase transition during ammonium nitrate production, and degradation can be limited through the addition of chemical additives such as sodium silicate. This compound may be mixed with calcium chloride to precipitate calcium silicate, which is often used in the nitrogenous fer-tilizer industry since it increases the CaCO3 content in soils that do not contain CaCO3 (according to market requirements). This compound is often prepared in the nitrogenous fertilizer industry through precipitation as calcium silicate; this increases the rate of for-mation of calcium silicate in soils that do not contain calcium silicate, according to market requirements. Calcium silicate can be prepared in many ways from sodium silicate, with one of the most commonly used preparation methods involving the reaction between calcium chloride and sodium silicate [16]. The main purpose of this reaction is to determine the calcium content and nutriment requirement of the soil.

In the current study, a better chemical compo-sition is proposed that is different from that using the chemicals typically suggested for limiting the degrad-ation of ammonium nitrate crystals in many patented works. Moreover, a detailed chemical analysis of the degradation progress was performed by using ion chromatography, electron microscopy and sieve anal-ysis. Through sieve analysis, it was possible to obs-erve the effects on the prill particle size of different compounds typically used to dilute the ammonium nitrate solution, including calcium carbonate, sodium silicate, and silicic acid. These results showed that dilution with calcium carbonate during ammonium nitrate production led to a prill particle size that was smaller than that with sodium silicate. The sieve rate increased with an increase in ammonium nitrate deg-radation in the dilution process with the addition of calcium carbonate (Table 1) than with both sodium

silicate and calcium carbonate (Table 2). Thus, better results were obtained with the mixture of sodium silicate and calcium carbonate. Carbonate, which is formed during the addition of calcium carbonate, creates foam and causes the shape of the ammonium nitrate prill to break. The amount of carbonate should be decreased to reduce the foaming reaction. In many patented works, sulfuric acid is added to the reaction bulk to minimize the formation of carbonate during ammonium nitrate production. However, sulfuric acid can form double salts with nitrate after production, depending on the storage conditions, and crystal bridges, which accelerate degradation, may be formed between these double salts. The crystal bridges, in turn, lead to deliquescence through hyd-rogen bonding between the particles. In the current study, silicic acid was used instead of sulfuric acid to minimize the formation of sulfate. Silicic acid was applied to consume any excess sulfuric acid remain-ing from the production process. Our results show that 1 mL silicic acid is capable of removing 5.6 mL sulfuric acid.

The SEM images show that the formation breaking of ammonium nitrate with sodium silicate was less than that with calcium carbonate. Over one year, in bulk ammonium nitrate under 0.28 kg/cm2 pressure [17], the ammonium nitrate surface was more deformed with calcium carbonate than with sodium silicate (Figures 3b and 4b). From these investigations, it was possible to observe the effects of adding only calcium carbonate for the dilution of ammonium nitrate (Figures 2a and 3b), adding cal-cium carbonate and sodium silicate to increase the nutriment value of ammonium nitrate (Figures 2b and 4a), and adding silicic acid to decrease the formation of double salts (Figures 3a and 4b) in the ammonium nitrate production process.

According to the results of these studies, the best surface properties were obtained when ammo-nium nitrate production involved sodium silicate, cal-cium carbonate, and silicic acid. When sodium silicate solution was added with calcium carbonate to the ambient reaction, the observed crushing strength of ammonium nitrate (Table 2) was higher than that of the ammonium nitrate produced when only calcium carbonate was used for dilution (Table 1).

In this study, ion chromatography was also used to investigate the ammonium nitrate degradation process. The main purpose here was to understand the effects of silicic acid on the sulfate salts. In the current work, the sulfate amount in the bulk ammo-nium nitrate was analyzed after one year, and was observed to have decreased. This was attributed to the consumption of excess sulfuric acid by the addi-


366

tion of silicic acid, as discussed above (Figures 5a and 5b). Note that the ion chromatography analysis presented in the current work was conducted for exp-eriments involving the addition of calcium carbonate only, and for the mixture of calcium carbonate, sod-ium silicate, and silicic acid. It was not conducted for the experiments involving the addition of calcium carbonate and sodium silicate, since sodium silicate has an inorganic structure, and hence, will not pro-duce a foaming reaction releasing CO2 as a product. Moreover, CO2 is produced during the reaction between sulfuric acid and silicic acid, which was not used in the experiments involving only calcium car-bonate and sodium silicate.

CONCLUSIONS

In this study, the effects of the addition of cal-cium carbonate, sodium silicate, and silicic acid in the ammonium nitrate production process were inves-tigated. The addition of sodium silicate to the reaction mixture resulted in increased particle sizes compared to those produced with the addition of calcium carbo-nate, which is typically used to dilute ammonium nitrate solutions. The addition of silicic acid resulted in less breaking of the surface morphology, and in the decreased production of double sulfate salts that cause crystal bridges, thus reducing eventual caking problems.

REFERENCES

[1] E.R. Boller, (DuPont), US Patent 1849704 (1932)

[2] L.D. Bryant, (Columbia Southern Chem. Corp.), US Patent 2903349 (1959)

[3] J. Roberts, G. Volgas, (Helena Chemical Co.), US Patent 5725630 (1998)

[4] J.A. Wyler, (Trojan Powder Co .), US Patent 1932434 (1933)

[5] W.H. Rinkenbach, (Trojan Powder Co.), US Patent 2660541 (1995)

[6] P.T. Adam, (Paul Thomas Adam), US Patent 5472475 (1995)

[7] T.J. Blakemore, Y.S. Chen, (Dober Chemical Corp.), US Patent 6878309 B2 (2005)

[8] S. O’Halloran, J. Persson, M. Wennerholm, Handbook for Kjeldahl Digestion, 4th Ed., Foss Publishing, Tokyo,, 2008, p. 18

[9] A.O. Gezerman, B.D. Çorbacioglu, E-J. Chem., 2014,, Article ID 426014, doi: 10.1155/2014/426014

[10] A.E. van Nieuwenhuyse, Production of Ammonium Nitrate and Calcium Ammonium Nitrate, European Fer-tilizer Manufacturers’ Association, Booklet No. 6, Brus-sels, 2000, p.12

[11] H.M. Hashemb, G.F. Malasha, Alexandria Eng. J. 44 (2005) 685–693

[12] L.D. Grant, Chemical and Physical Stability of Powdered Tagatose as Affected by Temperature and Relative Humidity, Auburn University, Auburn, AL, 2010, p.12

[13] ASTM Standard E986-97, Standard Practice for Scanning Electron Microscope Beam Size Characterization, ASTM International, West Conshohocken(1997), DOI: 10.1520/ /E0986-97, www.astm.org

[14] ASTM Standard E1151-93, Standard Practice for Ion Chromatography Terms and Relationships, ASTM Inter-national, West Conshohocken(2011), DOI: 10.1520/ /E1151-93R11, www.astm.org

[15] ASTM Standard E11-09, Standard Specification for Wire Cloth and Sieves for Testing Purposes, ASTM Inter-national, West Conshohocken(2009), DOI: 10.1520/ /E0011-01, www.astm.org

[16] R.P. Allen, (Goodrich Co B F), U S Patent 2204113 (1940)

[17] D.W. Rutland, Manual for Determining Physical Pro-perties of Fertilizer, Reference manual. No. IFDC-R-10, International Fertilizer Development Center, Muscle Shoals, AL, 1986, p. 15.


367

AHMET OZAN GEZERMAN

BURCU DIDEM ÇORBACIOĞLU

Yildiz Technical University, Department of Chemical Engineering, Faculty of

Chemical and Metallurgical Engineering, Istanbul, Turkey

NAUČNI RAD

EFEKTI NATRIJUM-SILIKATA, KALCIJUM-KARBO-NATA I SILICIJUMOVE KISELINE NA RAZGRADNJU AMONIJUM-NITRATA I ANALITIČKA ISPITIVANJA PROCESA RAZGRADNJE NA INDUSTRIJSKOM NIVOU

Amonijum nitrat je neorganska hemikalija sa brojnim primenama u različitim industrijama.

Međutim, različiti problemi su povezani sa proizvodnjom i kasnijim skladištenjem amonijum-

-nitrata, uklјučujući koksovanje, razgradnju, neželјene fazne prelaze i rekristalizaciju. Iako

je razvijeno nekoliko tehnika u cilju rešavanja ovih problema, mnoge od njih ne rade u

praksi. U ovom radu su različita jedinjenja, uklјučujući silicijumovu kiselinu i natrijum-silikat

dodata da bi usporila napredovanje ili sprečila razgradnju amonijum nitrata. Više instru-

mentalnih analiza, kao što su jonska hromatografija i skenirajuća elektronska mikroskopija,

korišćene su za praćenje procesa razgradnje.

Ključne reči: amonijum-nitrat, koksovanje, razgradnja, jonska hromatografija, proces posle reaktora.

ISSN 1451 - 9372(Print) ISSN 2217 - 7434(Online) APRIL ... No2_p229-367_Apr … · N.S. TEHRANI et...

Documents

Transcript of ISSN 1451 - 9372(Print) ISSN 2217 - 7434(Online) APRIL ... No2_p229-367_Apr … · N.S. TEHRANI et...