Optimization of ultrasound-assisted oxidative desulfurization of model sulfur compounds using commercial ferrate (VI)

Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–2942

Optimization of ultrasound-assisted oxidative desulfurization ofmodel sulfur compounds using commercial ferrate (VI)

Angelo Earvin Sy Choi a, Susan Roces a, Nathaniel Dugos a, Cybelle Morales Futalan b,Shiow-Shyung Lin c, Meng-Wei Wan c,*a Chemical Engineering Department, De La Salle University, Manila, Philippines 1004b Operations Department, Frontier Oil Corporation, Makati, Philippines 1229c Department of Environment Engineering and Science, Chia-Nan University of Pharmacy and Science, Tainan, Taiwan 71710

A R T I C L E I N F O

Article history:

Received 4 April 2014

Received in revised form 31 July 2014

Accepted 2 August 2014

Available online 12 September 2014

Keywords:

Benzothiophene

Box-Behnken Design

Dibenzothiophene

Ferrate

Ultrasound-assisted oxidative

desulfurization

A B S T R A C T

In this study, ultrasound-assisted oxidative desulfurization of benzothiophene and dibenzothiophene

using ferrate and a phase transfer agent was investigated. The effects of ferrate concentration

(100–300 ppm), phase transfer agent (100–300 mg), ultrasonication time (10–30 min), and organic to

aqueous phase ratio (10:30–30:10) on sulfur reduction were examined. A Box-Behnken design was

utilized to evaluate the effects of various parameters and determine optimum conditions in

desulfurization. The level of significance of each parameter and their interaction were analyzed using

the analysis of variance. Results showed the optimal conditions for benzothiophene to be 16.4 min,

122.1 mg PTA, 29.7 mL:10.3 mL and 204.8 ppm ferrate. Meanwhile, optimum desulfurization conditions

for dibenzothiophene could be attained at 29.5 min, 111.6 mg PTA, 16.2 mL:23.8 mL and 245.3 ppm

ferrate. Moreover, sulfur removal of 85.7% and 91.0% in diesel oil was achieved for benzothiophene and

dibenzothiophene using the optimized conditions.

� 2014 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Journal of the Taiwan Institute of Chemical Engineers

jou r nal h o mep age: w ww.els evier . co m/lo c ate / j t i c e

1. Introduction

In the past two decades, rapid technological advancement andincreasing world population have led to a markedly increase in fuelconsumption. Transportation fuels such as diesel oil, jet fuel andgasoline contain high organic sulfur compounds (OSCs), which area major source of air pollution [1]. During combustion, OSCs areconverted into sulfur dioxides (SOx) and particulate matter thatcould endanger public health and the environment [2,3].Moreover, OSCs in fuels could cause corrosion of internalcombustion engines, contribute to air pollution and poisoncatalytic converters [4]. To lessen the environmental and healthimpacts, emission control standards are enforced. For example,sulfur concentration in diesel fuels is set at 10 mg/g in Europe.Meanwhile, USEPA has set the maximum level for sulfurcompounds in diesel and gasoline at 15 parts per million byweight and 30 ppmw, respectively [2,5].

In order to produce low-sulfur fuels, attention has been drawnto technologies such as hydrodesulfurization (HDS), liquid/liquid

* Corresponding author. Tel.: +886 6 266 0615; fax: +886 6 366 2668.

E-mail address: [email protected] (M.-W. Wan).

http://dx.doi.org/10.1016/j.jtice.2014.08.003

1876-1070/� 2014 Taiwan Institute of Chemical Engineers. Published by Elsevier B.V.

extraction, biocatalytic-desulfurization and oxidative desulfuriza-tion [1–4]. At present, HDS is a large-scale chemical methodutilized by industries for removal of sulfur from fuel oils andpetroleum feedstock [6]. It is a conventional hydrotreatmentprocess that utilizes hydrogen in destroying the bonds of sulfur-containing compounds that would produce hydrocarbons andhydrogen sulfide [7,8]. However, the process requires highhydrogen pressure (100–500 psi), high operating temperature(300–400 8C), use of noble metal catalysts, large reactors and a longresidence time, which translates into high operating costs [8–10].In addition, aromatic sulfur compounds such as benzothiophene(BT), dibenzothiophene (DBT) and other dibenzothiophene deri-vatives are resistant to HDS treatment that requires more severereaction conditions [9,11,12].

Oxidative desulfurization (ODS) is one of the most promisingtechnologies for deep desulfurization of fuel oil. In comparison toHDS, this method has several advantages since it could be carriedout under mild conditions (low pressure and temperature),provides high selectivity and does not utilize expensive hydrogenin its operations [13]. The method is composed of oxidizing thesulfur compounds to form sulfones and/or sulfoxides, which arehighly polar and could be easily removed via adsorption, solventextraction or distillation [3,14]. Various oxidants and oxidation

All rights reserved.

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–29422936

systems such as nitric acid [15], ionized ozone [16], H2O2/photocatalysts [17], H2O2/acetic acid [15,18–21], and H2O2/formicacid [21] have been reported.

To further improve desulfurization efficiency, an innovativetechnology called ultrasound-assisted oxidative desulfurization(UAOD) was introduced. The main advantage of UAOD overconventional mixing in ODS is the higher desulfurization ratecaused by smoother dispersion in the system [1,9,22]. The physicaleffect of ultrasonic cavitation leads to the formation of fine emulsionthat promotes mass transfer, while the chemical effect producesradicals through transient collapse of cavitation bubbles, whichaccelerates the reaction [23,24]. The mechanism of oxidizing sulfurcompounds such as BT [25] and DBT [26] using sonication is physicalin nature. This implies that sonication improves oxidation due to themass transfer characteristic by increasing the interfacial areabetween oxidant and oil phase [25,26]. Furthermore, the UAODprocess combines phase transfer catalysis (PTA), transition metalcatalyst (TMC) and ultrasound that would provide selective removalof OSCs by combining selective oxidation, solvent extraction and/orsolid adsorption [2]. There are several studies reported on UAOD,which utilizes different oxidation systems such as H2O2-acetic acid[13], H2O2-phosphotungstic acid [1,9,27], potassium superoxide[28], sodium sulfide [22], H2O2-CF3COOH [29] and H2O2-formic acid[30]. Among the various oxidants, hydrogen peroxide (H2O2) ispreferred due to its high oxidizing efficiency and low cost [31].However, excessive decomposition of H2O2 and longer reactiontimes of the system make it impractical to be applied on an industrialscale [32,33].

Potassium ferrate (K2FeO4) is a well-known oxidant withseveral attractive properties such as high stability, selectivity andhigh oxidizing power. Ferrate (Fe(VI)) is a powerful oxidizing agentwith a reduction potential of +2.20 V in an acidic medium and hasgreater oxidizing ability than O3, Cl2, H2O2 and KMnO4 [34].Moreover, it is an environmentally friendly oxidant, where Fe(VI) isreduced to non-toxic Fe(III) by-products [14]. Liu et al. (2008)investigated the ODS of diesel oil using a Fe(VI)-acetic acid systemwith manganese acetate as the catalyst and various phase transfercatalysts such as hexadecyl trimethyl ammonium bromide, benzyltrimethyl ammonium chloride and tetramethyl ammoniumchloride. Other studies have been performed, which utilized Fe(VI)in the oxidation of sulfur-containing amino acids [35] and Fe(VI)-phosphomolybdic acid in the oxidation of thiophene [36].Currently, there are limited studies on utilizing Fe(VI) in oxidizingsulfur compounds such as BT and DBT. Moreover, there have notbeen any reports on the optimization of the oxidation of diesel oilthat utilize K2FeO4 in a UAOD system. Process optimization isessential in understanding the interdependencies between processvariables in a UAOD system. Moreover, the model would allowreal-time data to be acquired from the process and to makechanges in process parameters in real-time [37].

In this study, sulfur reduction of model compounds such as BTand DBT in a UAOD system using Fe(VI) and PTA was investigated.The effects of Fe(VI) concentration, amount of PTA, ultrasonicationtime and organic to aqueous phase (OP:AP) ratio on sulfurreduction were examined. Optimization studies were carried outby applying Box-Behnken design under the response surfacemethodology to assess statistically important operating parame-ters that affect sulfur removal. Furthermore, optimized conditionswere applied in the desulfurization treatment of diesel oil.

2. Experimental

2.1. Materials

Model sulfur compounds BT (97% purity) and DBT (99% purity)were purchased from Acros Organics (Taiwan) and Alfa Aesar

(Taiwan), respectively. Toluene was obtained from Merck Chemi-cal Company (USA) while glacial acetic acid was purchased fromPanreac (Taiwan). Potassium ferrate (90% purity) and tetraocty-lammonium bromide (TOAB) were acquired from Sigma-Aldrich(Wisconsin, USA). Diesel oil with sulfur content of 1426.8 ppm wassupplied by Taichin Global Company (Taichung, Taiwan).

2.2. Instruments

In this study, a 500-W ultrasound apparatus (Sonic VCX, USA)with amplitude of 40% and 20 kHz frequency, equipped with atitanium probe tip (25 mm diameter and 122 mm length), wasutilized. The sulfur compounds of the feed and products wereanalyzed using a gas chromatograph (Agilent Gas Chromatograph,7890A, California, USA) equipped with fused silica capillary HP-5MS column with thickness of 0.25 mm film (J & W, Scientific, USA)and sulfur chemiluminescence detector (SCD). For BT, the initial GCoven temperature was set at 100 8C for 3 min, heated at a rate of20 8C/min to 180 8C and retained for 3 min. Meanwhile, the columntemperature setting for DBT was set at 150 8C for 1 min and heatedat a rate of 20 8C/min to 280 8C for 1 min. The sulfur content ofdiesel oil was analyzed using a SLFA-2100 X-ray fluorescencesulfur-in-oil analyzer (Horiba).

2.3. UAOD methodology

The 500-ppm stock solution of BT and DBT was prepared bydissolving an appropriate amount of each sulfur compound intoluene. An appropriate volume of model compound or diesel oilwith PTA and a pre-determined volume and concentration of Fe(VI)dissolved in deionized water (aqueous phase) were added in abeaker. The total working volume of the UAOD system is 40 mL.The mixture was irradiated using ultrasound for a specific time atreaction temperature reaching up to 70 � 2 8C. A constant standoffdistance of 1 mm between the ultrasound probe and OP:AP interfacewas maintained. The pH of the UAOD system was adjusted to aninitial pH 4 using 0.1N acetic acid. After cooling, mixture wascentrifuged for 10 min to separate the oil/water mixture. Sulfurcontent was analyzed using GC-SCD and SLFA-2100 for modelcompounds and diesel oil, respectively.

2.4. Statistical analysis

Optimization studies utilized Box-Behnken design (BBD) usingDesign Expert 7.1.4 to determine the effects of process parameterssuch as Fe(VI) concentration, amount of PTA, ultrasonication timeand OP:AP ratio on the removal efficiency of sulfur using UAOD forBT and DBT. In addition, the BBD method was used to determinethe combination of parameters that would yield the optimumsulfur reduction efficiency. The full experimental design of thestudy is comprised of a total of 29 runs for the four parameters tobe tested. In Table 1, the process parameters have been designatedthree levels as �1, 0 and +1 for low, middle and high values,respectively. Furthermore, the statistical approach in BBD gen-erates a 95%-confidence interval that accounts for the error anduncertainty of the outcome of the response. The confidenceinterval implies that there is a 95% chance that the measuredresponse will fall within the specified interval to confirm thegenerated model of BBD.

3. Results and discussion

3.1. UAOD reaction system

The UAOD system is composed of two immiscible phases, anorganic phase containing the sulfur compound or diesel oil and an

Table 1Levels of each variable for the Box-Behnken design in the UAOD system.

Factors Levels

�1 0 1

Ultrasonication time (min), X1 10 20 30

PTA (mg), X2 100 200 300

OP:AP (mL:mL), X3 10:30 20:20 30:10

Ferrate concentration (ppm), X4 100 200 300

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400

Su

lfu

r R

edu

ctio

n(%

)

PTA (mg)

BT

DBT

0

10

20

30

40

50

60

70

80

90

100

0 100 200 300 400

Su

lfu

r R

edu

ctio

n (

%)

Ferrate Concentration (ppm)

BT

DBT

(a)

(b)

Fig. 1. Effect of (a) PTA amount and (b) Fe(VI) concentration on the sulfur reduction

of BT and DBT.

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–2942 2937

aqueous phase that contains the ferrate oxidant, Fe(VI) or (FeO4)2�.There are several reactions involved in a UAOD system, where thefirst step involves ferrate reacting with acetic acid to form a morecomplex, reactive compound, [O3Fe(OH)]� [14]. The complex ferratecompound subsequently binds with the quaternary ammoniumcation (Q+) that acts as the PTA. Due to the lipophilic property of thePTA, the organic phase is mixed with Q+[O3Fe(OH)]�. Uponapplication of ultrasonication, Q+[O3Fe(OH)]� forms an emulsionthat oxidizes organic sulfur compounds. Consequently, BT and DBTform major products of benzothiophene sulfone (BT-O) anddibenzothiophene sulfone (DBT-O), respectively. The use of TOABas PTA in the UAOD system causes several brominated, polar by-products to form such as 3-bromobenzothiophene and 1-bromooc-tane from BT oxidation. Moreover, formation of dibenzo(a,c)fluoren-13-one after desulfurization of DBT solution was observed. The Q+

ion brings the ferrate back in the aqueous phase, where ferratecomplexes are used up and converted to Fe3+.

3.2. Effect of PTA and ferrate concentration

PTAs are surface-active agents that lower surface tension andenhance liquid–liquid interfacial area through emulsification. InFig. 1a, the effect of the amount of PTA on the sulfur reduction of BTand DBT is illustrated. An increase in PTA causes a correspondingincrease in sulfur reduction due to higher amount of oxidanttransferred from aqueous solution to the organic phase. High sulfurreduction of 83.9% and 77.6% was achieved at 200 mg PTA for DBTand BT, respectively. This indicates that a higher amount of PTAimproves formation of emulsion, where two immiscible phases areeffectively mixed that leads to high oxidation activity [14]. However,a decrease in sulfur reduction was observed as the amount of PTAwas further increased from 200 to 300 mg, which is possibly causedby an increase in steric hindrance of the catalytic system. A highamount of PTA indicates high concentration of alkyl groups, whichprevents an effective electrophilic aromatic substitution for theoxidation of BT and DBT [1]. Moreover, thicker and more turbidlayers are formed at 300 mg PTA. This indicates that there is anineffective mass transfer between hydrophobic (oil) and hydrophilic(oxidant) phase, which results to a low sulfur removal [38].

Fig. 1b illustrates the effect of ferrate concentration on thesulfur reduction of BT and DBT. At 100 ppm Fe(VI), sulfur reductionwas observed to be relatively lower in comparison to systems withhigher Fe(VI) concentrations. A low amount of Fe(VI) has limitedoxidizing capability due to insufficient amount of oxidant presentto reduce BT and DBT into sulfones. High sulfur conversion of 79.0%and 86.0% were achieved at 200 ppm Fe(VI) for BT and DBT,respectively. Further increasing Fe(VI) concentration to 300 ppmcauses a drop in sulfur reduction. A high Fe(VI) concentration shiftsthe pH of the mixture to become more basic, hence a decrease inconversion of BT and DBT. A low oxidation activity of Fe(VI) wasexhibited due to its lower redox potential in basic medium [14].

3.3. Effect of ultrasonication time and OP:AP ratio

The effect of varying ultrasonication times (10–30 min) on thesulfur reduction of BT and DBT with FeO4

2�/S (mol/mol) = 3.0 for

BT and FeO42�/S (mol/mol) = 2.0 for DBT is shown in Fig. 2a. In a

UAOD system, emulsions produced by sonication are finer andmore stable than conventional ODS mixing technique, whichfurther enhances the interfacial area available for reaction.Application of ultrasonic irradiation in the system causes anincrease in sulfur reduction rate, which is due to an increase ineffective local concentration of reactive species and an improve-ment in mass transfer in the interfacial region. Hence, high sulfurreduction of 74.0% for BT and 84.1% for DBT was achieved atultrasonication time of 20 min. This illustrates that the increasedtime of ultrasonic irradiation improves the oxidation processthrough biphasic transfer of oxidants [1]. Further increasing theultrasonication time from 20 to 30 min has no significant effect insulfur reduction, which implies that Fe(VI) was spent and degradedat 20 min. Hence, no further oxidation occurred and equilibriumwas attained.

In the oxidation system, OP and AP refer to the volume of modelOSCs and Fe(VI), respectively. In Fig. 2b, the trends in sulfurreduction of BT and DBT were analyzed by varying OP:AP ratiofrom 10:30 to 30:10. A system with a 30:10 ratio indicates a lowervolume of AP in the mixture. This implies that there is less amountof Fe(VI) present that would cause lower oxidation capacity toreduce OSCs in the mixture. The 20:20 ratio provided the highestsulfur reductions of 79.5% and 84.5% for BT and DBT, respectively.On the other hand, a ratio of 10:30, where a higher volume offerrate dilutes the PTA present, which caused a decrease in sulfurreduction. A lower volume of PTA implies that there are lesseremulsion formed and lesser interfacial area for oxidation reactionto occur.

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40

Su

lfu

r R

edu

ctio

n (

%)

Sonication Time (mins)

BT

DBT

0

10

20

30

40

50

60

70

80

90

100

0 10 20 30 40

(%)

Su

lfu

r R

edu

ctio

n

Organic Phase (mL) : Aqueous Phase (mL)

BT

DBT

:40 :30 :20 :10 :0

(a)

(b)

Fig. 2. Effect of (a) ultrasonication time and (b) organic and aqueous phase ratio on

the sulfur reduction of BT and DBT.

Table 2Observed percent (%) sulfur reduction for BT and DBT by Fe(VI) in the UAOD system u

Run Sonication

time (min)

PTA (mg) OP:AP

(mL:mL)

1 10 200 30:10

2 20 100 30:10

3 20 200 20:20

4 10 200 20:20

5 20 100 20:20

6 20 200 10:30

7 20 200 20:20

8 10 300 20:20

9 30 200 10:30

10 30 200 20:20

11 20 200 10:30

12 30 200 30:10

13 20 300 20:20

14 20 200 20:20

15 20 100 10:30

16 10 200 20:20

17 20 200 20:20

18 20 300 20:20

19 20 200 30:10

20 30 200 20:20

21 10 200 10:30

22 20 200 20:20

23 20 200 30:10

24 10 100 20:20

25 20 300 30:10

26 30 300 20:20

27 20 300 10:30

28 20 100 20:20

29 30 100 20:20

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–29422938

3.4. Experimental errors and reproducibility

Experimental errors are composed of systematic and randomerrors that affect the accuracy and precision of a measurement,respectively. Determining the sources of these errors arecarried out in order to ensure the reproducibility of a singledata point. In the UAOD system, sources of systematic errorsmay be due to the inaccurate calibration of sulfur compoundsusing the GC-SCD and SLFA-2100 and use of analytical balancein weighing BT, DBT and ferrate. On the other hand, randomerrors present in the UAOD system comprise of inaccuratelyreading the measurement of the volume in OP and AP usinga graduated cylinder. Eliminating systematic errors and mini-mizing random errors through proper calibration and operationof equipment in turn promotes good reproducibility of results.

3.5. Optimization of UAOD parameters

In the optimization experiments, the effect of processparameters such as ferrate concentration, PTA, ultrasonicationtime and OP:AP ratio on the sulfur reduction of BT and DBT wasexamined using BBD method. Table 2 illustrates the observedresponses of the runs with respect to the percentage (%) sulfurreductions of BT and DBT. The range for the response in %BTreduction was in between 46.0% and 87.7% while response in %DBTreduction was observed to be in the range of 65.0–89.7%.

3.5.1. Analysis of variance

In this study, a quadratic model was utilized in correlatingexperimental data and attainment of the regression equation.From Eq. 1 and 2, the values of the response function coefficientsfor each independent variable (X1, X2, X3, X4) were determinedusing experimental data. The response predictive models withobtained coefficients for percent BT and DBT reduction (Y1, Y2) are

sing BBD experimental design.

Ferrate

concentration (ppm)

%BT

reduction

%DBT

reduction

200 86.8 77.9

200 83.0 69.3

200 77.6 83.9

300 78.1 86.7

300 66.7 88.4

300 82.0 74.8

200 74.1 84.1

200 67.0 88.5

200 61.5 86.2

300 78.4 81.2

100 53.8 82.1

200 75.7 79.6

100 67.7 87.5

200 79.5 84.5

200 46.0 79.3

100 75.5 78.6

200 79.0 86.0

300 71.1 65.0

100 87.7 73.0

100 64.4 89.7

200 59.0 83.7

200 72.4 85.8

300 82.1 73.1

200 71.8 75.7

200 69.8 75.4

200 62.8 77.4

200 72.8 76.2

100 69.1 65.4

200 63.1 87.7

Table 3Results for analysis of variance (ANOVA) using a quadratic model for percent (%) sulfur reduction of BT using Fe(VI) in UAOD system.

Source Sum of

squares

Degrees

of freedom

Mean

square

F-Value p-value

Model 2455.44 14 175.39 14.48 <0.0001

X1-Ultrasonication time 87.19 1 87.19 7.20 0.0178

X2-PTA 10.88 1 10.88 0.90 0.3594

X3-OP:AP 1008.75 1 1008.75 83.29 <0.0001

X4-Ferrate Concn 136.12 1 136.12 11.24 0.0047

X1 X2 5.08 1 5.08 0.42 0.5276

X1 X3 46.96 1 46.96 3.88 0.0691

X1 X4 32.75 1 32.75 2.70 0.1224

X2 X3 398.76 1 398.76 32.93 <0.0001

X2 X4 8.31 1 8.31 0.69 0.4213

X3 X4 285.21 1 285.21 23.55 0.0003

X21 74.77 1 74.77 6.17 0.0262

X22 369.91 1 369.91 30.54 <0.0001

X23 12.45 1 12.45 1.03 0.3278

X24 2.75 1 2.75 0.23 0.6411

Residual 169.56 14 12.11

Lack of Fit 130.17 10 13.02 1.32 0.4236

Pure Error 39.38 4 9.85

Total 2625.00 28

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–2942 2939

given as Eq. 1 and 2:

Y1 ¼ 76:11 � 2:70X1 þ 9:17X3 þ 3:37X4 � 9:98X2X3

� 8:44X3X4 � 3:28X21 � 7:43X2

2 (1)

Y2 ¼ 84:87 þ 0:89X1 � 2:83X3 � 5:80X1X2 � 4:15X1X4

þ 2:29X2X3 � 11:38X2X4 þ 1:86X3X4 þ 2:40X21 � 4:74X2

2

� 5:36X23 � 3:51X2

4 (2)

Tables 3 and 4 summarize the results for the analysis ofvariance (ANOVA), F-values and p-values of the response surfacequadratic model and model terms where the level of confidence isassumed to be 0.05. A high F value indicates that most of thevariation in the response can be explained by the quadratic modelequation while a p-value of less than 0.05 would indicate that themodel and model terms are statistically significant [39]. Highvalues of the coefficient of determination, which are R2 = 0.9354 forBT and R2 = 0.9849 for DBT, indicate that the fit is good andvariation in the observed values can be explained by the models. Inaddition, it also implies that the quadratic model used is highlyreliable in predicting responses. Based on the values of F-test

Table 4Results for analysis of variance (ANOVA) using a quadratic model for percent (%) sulfu

Source Sum of

squares

Degrees

of freedom

Model 1300.57 14

X1-Ultrasonication time 9.46 1

X2-PTA 1.42 1

X3-OP:AP 96.42 1

X4-Ferrate Concn 4.18 1

X1 X2 134.44 1

X1 X3 0.17 1

X1 X4 68.94 1

X2 X3 21.03 1

X2 X4 517.75 1

X3 X4 13.77 1

X21 37.45 1

X22 145.52 1

X23 186.58 1

X24 79.80 1

Residual 19.92 14

Lack of Fit 16.37 10

Pure Error 3.55 4

Total 1320.49 28

(14.48 for BT and 65.29 for DBT) and p-test (<0.0001 for both BTand DBT), it implies that the quadratic model and terms used arestatistically significant.

From Table 3, variables and interactions such as X1, X3, X4, X2X3,X3X4, X2

1 and X22 are significant factors that affect BT reduction.

Moreover, it can be concluded that X3, X2X3 and X22 are found to be

extremely significant (p < 0.0001) in the reduction of BT.Table 4 illustrates that the variables and interactions such as X1,

X3, X1X2, X1X4, X2X3, X2X4, X3X4, X21 , X2

2 , X23 and X2

4 are found to besignificant. In addition, factors X3, X1X2, X1X4, X2X4, X2

2 , X23 and X2

4 arefound to be extremely significant (p < 0.0001) in the reduction of DBT.

In Fig. 3, results show that there is a good agreement observedbetween actual and predicted values on sulfur reduction of BT andDBT. Moreover, the DBT plot is observed to have a better fit incomparison to BT.

3.6. Analysis of responses

To examine the effect of ultrasonication time, PTA, OP:AP ratioand Fe(VI) concentration on the sulfur reduction of BT and DBT,response surface methodology was utilized and 3-D plots weregenerated. From Table 3, factors such as ultrasonication time,

r reduction of DBT using Fe(VI) in UAOD system.

Mean

square

F-Value p-value

92.90 65.29 <0.0001

9.46 6.65 0.0219

1.42 1.00 0.3353

96.42 67.76 <0.0001

4.18 2.94 0.1087

134.44 94.48 <0.0001

0.17 0.12 0.7339

68.94 48.45 <0.0001

21.03 14.78 0.0018

517.75 363.87 <0.0001

13.77 9.68 0.0077

37.45 26.32 0.0002

145.52 102.27 <0.0001

186.58 131.13 <0.0001

79.80 56.08 <0.0001

1.42

1.64 1.84 0.2918

0.89

Fig. 3. Plot of the actual and predicted values of (%) sulfur reduction of (a) BT and (b) DBT.

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–29422940

OP:AP ratio and Fe(VI) concentration have low p-values of 0.0178,<0.0001 and 0.0047, respectively. Moreover, ultrasonication time,OP:AP ratio and Fe(VI) concentration show high F-values of 7.20,83.29 and 11.24. Results indicate that the three factors have asignificant effect on the sulfur reduction of BT, where the OP:APratio provided the most significant effect on the response. FromTable 4, the effect of ultrasonication time (p-value = 0.0219 and F-value = 6.65) and OP:AP (p-value < 0.0001 and F-value = 67.76)were considered to be significant on sulfur reduction of DBT, whichthe OP:AP ratio is found to be extremely significant. In both BT andDBT reduction, the OP:AP ratio was found to have extremelysignificant effect where a 20:20 ratio is favorable for a high sulfurremoval. The oxidation reaction is dependent on the amount ofsulfur present in the OP and amount of Fe(VI) present in the AP inorder to effectively oxidize sulfur compounds. The interactionbetween factors in the conversion of sulfur is shown in theresponse surface graphs of BT and DBT in Figs. 4 and 5, respectively.In Fig. 4, the 3-D response surface plots of sulfur reduction of BT areillustrated as a function of two process parameters: (a) PTA andOP:AP ratio and (b) ultrasonication time and OP:AP ratio. On theother hand, Fig. 5 shows the 3-D surface plots of sulfur reduction ofDBT in terms of the combined effect of (a) ultrasonication time andPTA, (b) ultrasonication time and Fe(VI) concentration, (c) PTA andOP:AP ratio, (d) PTA and Fe(VI) concentration and (e) OP:AP ratioand Fe(VI) concentration.

Fig. 4. 3-D response surface plots of BT of the effect of (a) PTA and OP:AP ratio (X2 = 200 m

3.6.1. Optimization

The optimized conditions for sulfur reduction were obtained byanalyzing the response surface plots and solving the regressionequation. For BT, the predicted sulfur reduction of 88.3% could beattained using an ultrasonication time of 16.4 min, 122.1 mg PTA,an OP:AP ratio of 29.7 mL:10.3 mL and 204.8 ppm Fe(VI). On theother hand, about 91.8% sulfur reduction could be achieved for DBTusing 29.5 min of ultrasonication time, 111.6 mg PTA, an OP:APratio of 16.2 mL:23.8 mL and 245.3 ppm Fe(VI).

The optimal conditions for sulfur reduction of BT and DBT wereapplied to three independent replicates to verify the reliability ofthe quadratic model (Table 5). The experimental results for BT andDBT are in good agreement with the predicted values. This impliesthat the quadratic model is adequate and effective.

3.6.2. Application to real diesel oil

The desulfurization of diesel oil with sulfur content of1428.6 ppm was carried out in a UAOD system using optimumconditions obtained for BT and DBT. The final concentration of BTand DBT in diesel oil was reduced to 204.5 and 128.2 ppm,respectively. Moreover, the sulfur reduction in diesel oil of 85.7%for BT and 91.0% for DBT was attained. The sulfur removal in realdiesel oil is lower in comparison to that of 88.3% and 91.8% sulfurremoval for model compounds BT and DBT, respectively. Diesel oilis a combination of different sulfur compounds that makes it

g, X3 = 20:20) and (b) ultrasonication time and OP:AP ratio (X1 = 20 min, X3 = 20:20).

Fig. 5. 3-D response surface plots of DBT of the effect of (a) ultrasonication time and PTA (X1 = 20 min, X2 = 200 mg), (b) ultrasonication time and Fe(VI) concentration

(X1 = 20 min, X4 = 200 ppm), (c) PTA and OP:AP ratio (X2 = 200 mg, X3 = 20:20), (d) PTA and Fe(VI) concentration (X2 = 200 mg, X4 = 200 ppm) and (e) OP:AP ratio and Fe(VI)

concentration (X3 = 20:20, X4 = 200 ppm).

Table 5Confirmation runs using optimum conditions for sulfur reduction of BT and DBT.

BT Removal DBT Removal

Predicted 88.3% 91.8%

Confidence

Interval

84.2%–92.4% 88.7%–95.0%

Observed: Run 1 88.6% 89.5%

Run 2 86.5% 94.8%

Run 3 89.9% 93.6%

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–2942 2941

difficult for oxidation to occur in the UAOD process as oppose toonly using pure BT and DBT alone. Throughout the study, it wasobserved that DBT is preferentially oxidized over BT by Fe(VI) due tohigher electron density and apparent rate constant value of DBT [40].

4. Conclusions

In this study, utilization of commercial ferrate as an oxidizingagent is effective in the sulfur reduction of BT and DBT using aUAOD system. Results showed that sulfur reduction of OSCs ishighest with moderate increase of ultrasonication time, Fe(VI)

A.E.S. Choi et al. / Journal of the Taiwan Institute of Chemical Engineers 45 (2014) 2935–29422942

concentration, PTA and OP:AP ratio. Response surface methodolo-gy with BBD was utilized in examining the effects of ultrasonica-tion time, Fe(VI) concentration, PTA and OP:AP ratio on thedesulfurization of BT and DBT. Both regression analysis andstatistical significance were carried out in predicting responses forall experimental regions. From analysis of response surfaces,ultrasonication time, OP:AP ratio and Fe(VI) concentration werefound to have significant effects on the reduction of BT whileultrasonication time, PTA and OP:AP ratio were considered to besignificant on sulfur reduction of DBT. The optimal conditions of16.4 min ultrasonication time, 122.1 mg PTA, an OP:AP ratioof 29.7 mL:10.3 mL and 204.8 ppm Fe(VI) for BT and 29.5 minof ultrasonication time, 111.6 mg PTA, an OP:AP ratio of16.2 mL:23.8 mL and 245.3 ppm Fe(VI) for DBT were determined.Under the optimal conditions, total sulfur content of diesel oil wasreduced from 1428.6 ppm to 204.5 and 128.2 ppm for BT and DBT,respectively. Hence, production of low sulfur diesel fuels could beachieved using Fe(VI) and PTA in a UAOD system.

Acknowledgments

The authors would like to acknowledge the Taiwan NationalScience Council for financial support in this research (NSC 100-2221-E-041-005, NSC 101-2221-E-041-010-MY3).

References

[1] Wan MW, Yen TF. Enhance efficiency of tetraoctylammonium fluoride appliedto ultrasound-assisted oxidative desulfurization. Appl Catal A-Gen 2007;319:237–45.

[2] Etemadi O, Yen TF. Surface characterization of adsorbents in ultrasound-assisted oxidative desulfurization process of fossil fuels. J Colloid Interf Sci2007;313:18–25.

[3] Mello PA, Duarte FA, Nunes MAG, Alencar MS, Moreira EM, Korn M, et al.Ultrasound-assisted oxidative process for sulfur removal from petroleumproduct feedstock. Ultrason Sonochem 2009;16:732–6.

[4] Chen TC, Shen YH, Lee WJ, Lin CC, Wan MW. An economic analysis of thecontinuous ultrasound-assisted oxidative desulfurization process applied tooil recovered from waste tires. J Clean Prod 2013;39:129–36.

[5] Breysse M, Mariadassou D, Pessayre S, Geantet C, Vrinat M, Perot G. Deepdesulfurization: reactions, catalysts and technological challenges. Catal Today2003;84:129–38.

[6] Chen TC, Shen YH, Lee WJ, Lin CC, Wan MW. The study of ultrasound-assistedoxidative desulfurization process applied to the utilization of pyrolysis oilfrom waste tires. J Clean Prod 2010;18:1850–8.

[7] Song C, Ma X. Ultra-deep desulfurization of liquid hydrocarbon fuels: chemis-try and process. Int J Green Energy 2004;1:167–91.

[8] Babich IV, Moulijn JA. Science and technology of novel processes for deepdesulfurization of oil refinery streams: a review. Fuel 2003;82:607–31.

[9] Mei H, Mei BW, Yen TF. A new method for obtaining ultra-low sulfur diesel fuelvia ultrasound assisted oxidative desulfurization. Fuel 2003;82:405–14.

[10] Ali MF, Al-Malki A, El-Ali B, Martinie G, Siddiqui MN. Deep desulphurization ofgasoline and diesel fuels using non-hydrogen consuming techniques. Fuel2006;85:1354–63.

[11] Schmitz C, Datsevitch L, Jess A. Deep desulfurization with diesel oil: kineticstudies and process-improvement by the use of a two-phase reactor with pre-saturator. Chem Eng Sci 2004;59:2821–9.

[12] Tailleur RG, Ravigli J, Quenza S, Valencia N. Catalyst for ultra-low sulfur andaromatic diesel. Appl Catal A: Gen 2005;282:227–35.

[13] Dai Y, Qi Y, Zhao D, Zhang H. An oxidative desulfurization method usingultrasound/Fenton’s reagent for obtaining low and/or ultra-low sulfur dieselfuel. Fuel Process Technol 2008;89:927–32.

[14] Liu S, Wang B, Cui B, Sun L. Deep desulfurization of diesel oil oxidized by Fe (VI)systems. Fuel 2008;87:422–8.

[15] Tam PS, Kittrell JR, Eldridge JW. Desulfurization of fuel oil by oxidation andextraction 1. Enhancement of extraction oil yield. Ind Eng Chem Res1990;29:321–4.

[16] Zaykinaa RF, Zaykina YA, Yagudinb SG, Fahruddinov IM. Specific approaches toradiation processing of high-sulfuric oil. Radiat Phys Chem 2004;71:465–8.

[17] Matsuzawa S, Tanaka J, Sato S, Ibusuki T. Photocatalytic oxidation of diben-zothiophenes in acetonitrile using TiO2: effect of hydrogen peroxide andultrasound irradiation. J Photoch Photobiol A 2002;149:183–9.

[18] Zannikos F, Lois E, Stournas S. Desulfurization of petroleum fractions byoxidation and solvent extraction. Fuel Process Technol 1995;42:35–45.

[19] Te M, Fairbridge C, Ring Z. Oxidation reactivities of dibenzothiophenes inpolyoxmetalate/H2O2 systems. Appl Catal A-Gen 2001;219:267–80.

[20] Al-Shahrani F, Xiao T, Llewellyn SA, Barri S, Jiang Z, Shi H, et al. Desulfurizationof diesel via the H2O2 oxidation of aromatic sulfides to sulfones using atungstate catalyst. Appl Catal B-Environ 2007;73:311–6.

[21] Ramı́rez-Verduzco LF, Murrieta-Guevara F, Garcia-Gutierrez JL, Martin-Cas-tanon RS, Martinez-Guerrero MC, Montiel-Pacheco MC, et al. Desulfurizationof middle distillates by oxidation and extraction process. Petrol Sci Technol2004;22:129–39.

[22] Hagenson LC, Doraiswamy LK. Comparison of the effects of ultrasound andmechanical agitation on a reacting solid-liquid system. Chem Eng Sci1998;53:131–48.

[23] Kuppa R, Moholkar VS. Physical features of ultrasound-enhanced heteroge-neous permanganate oxidation. Ultrason Sonochem 2010;17:123–31.

[24] Chakma S, Moholkar VS. Physical mechanism of sono-Fenton process. AIChE J2013;59:4303–13.

[25] Bolla MK, Choudhury HA, Moholkar VS. Mechanistic features of ultrasound-assisted oxidative desulfurization of liquid fuels. Ind Eng Chem Res2012;51:9705–12.

[26] Bhasarkar JB, Chakma S, Moholkar VS. Mechanistic features of oxidativedesulfurization using sono-fenton � peracetic acid (ultrasound/Fe2+ � CH3-

CH3COOH � H2O2). Ind Eng Chem Res 2013;52:9038–47.[27] Wan MW, Yen TF. Portable continuous ultrasound-assisted oxidative desul-

furization unit for marine gas oil. Energy Fuels 2008;22:1130–5.[28] Chan NY, Lin TY, Yen TF. Superoxides: alternative oxidants for the oxidative

desulfurization process. Energy Fuels 2008;22:3326–8.[29] Li Y, Zhao D, Lin J, Yuan Q. Preliminary study on oxidative desulfurization of

diesel via power ultrasound. Energy Sources 2009;31:191–8.[30] Gonzalez LA, Kracke P, Green WH, Tester JW, Shafer LM, Timko MT. Oxidative

desulfurization of middle-distillate fuels using activated carbon and powerultrasound. Energy Fuels 2012;26:5164–76.

[31] Zhang G, Yu F, Wang R. Research advances in oxidative desulfurizationtechnologies for the production of low sulfur fuel oils. Petrol Coal2008;51:196–207.

[32] Collins FM, Lucy AR, Sharp C. Oxidative desulphurization of oils via hydrogenperoxide and heteropolyanion catalysis. J Mol Catal A: Chem 1997;117:397–403.

[33] Bortolini O, Furia FD, Modena G, Seraglia R. Metal catalysis in oxidation byperoxides, Sulfide oxidation and olefin epoxidation by dilute hydrogen per-oxide catalyzed by molybdenum and tungsten derivatives under phase trans-fer conditions. J Org Chem 1985;50:2688–90.

[34] Sharma VK. Potassium ferrate(VI): an environmentally friendly oxidant. AdvEnviron Res 2002;6:143–56.

[35] Sharma VK, Luther III GW, Millero FJ. Mechanisms of oxidation of organosulfurcompounds by ferrate(VI). Chemosphere 2011;82:1083–9.

[36] Liu Y, Hua S, Wenchao Z. Oxidative desulfurization of model sulfur compoundby potassium ferrate in the presence of phosphomolybdic acid catalyst. ChinaPet Process PE 2013;15:61–5.

[37] Derkyi NSA, Adu-Amankwa B, Sekyere D, Darkwa NA. Optimization of processparameters using response surface methodlogy for the extraction of formal-dehyde-condensable phenolics from pine bark. J Emerg Trends Eng Appl Sci2011;2:64–9.

[38] Sachdeva TO, Pant KK. Deep desulfurization of diesel via peroxide oxidationusing phosphotungstic acid as phase transfer catalyst. Fuel Process Technol2010;15:1133–8.

[39] Baskan MB, Pala A. A statistical design approach for arsenic removal bycoagulation process using aluminum sulfate. Desalin 2010;254:42–8.

[40] Jiang Z, Lu H, Zhang Y, Li C. Oxidative desulfurization of fuel oils. Chin J Catal2011;32:707–15.