Effect of co-solvents on production of biodiesel via transesterification insupercritical methanol{

Ruengwit Sawangkeaw, Kunchana Bunyakiat and Somkiat Ngamprasertsith*

Received 29th September 2006, Accepted 22nd March 2007

First published as an Advance Article on the web 18th April 2007

DOI: 10.1039/b614064e

Previous studies on the transesterification of vegetable oil in supercritical methanol in a batch

reactor resulted in a non-saponified product with high methyl esters content and high glycerol

purity. For the continuous reactor, the high viscosity of vegetable oil might result in problems in

the flow system. This study selected THF and hexane as co-solvents to reduce the viscosity of the

vegetable oil. The effect of co-solvents was investigated in both 250 mL and 5.5 mL batch reactors

by 2-replicate 23 factorial design at temperatures from 290–350 uC, a molar ratio of methanol to

vegetable oil from 12–42 and a molar ratio of co-solvent to vegetable oil from 0–5. The reaction

time was fixed at 10 min. The products from the employed and unemployed co-solvent process

were analyzed by GC-MS to confirm that the reaction among the vegetable oil, methanol and co-

solvent was non-existent. However, some thermal cracking was observed in a 250 mL reactor at

350 uC and 30 min reaction time. The amount of co-solvents had no significant effect on methyl

esters content and also did not allow the reaction to be completed under milder conditions. Thus,

it was concluded that both THF and hexane were appropriate co-solvents to reduce the viscosity

of vegetable oil for the continuous production of biodiesel in supercritical methanol.

Introduction

Biodiesel, an alternative diesel fuel, refers to the lower alkyl

esters of the long chain fatty acids which are synthesized by

transesterification (also called alcoholysis) of vegetable oils

with lower alcohols. The conventional methods for biodiesel

production use a basic or acidic catalyst. With an acid catalyst,

a reaction time of 1–45 hours is necessary for the formation of

the respective esters, and with a basic catalyst it is somewhat

faster, depending on the temperature and molar ratio of

methanol to vegetable oil. Unfortunately it still poses some

separation and purification problems.1 A more recent method

involves the non-catalytic transesterification of vegetable oil

in supercritical methanol.2–9 Saka et al.3,4 found that the

reactions of rapeseed oil were complete within 240 s at 350 uC,

19 MPa and a molar ratio of methanol to vegetable oil of 42 in

a 5 mL batch reactor. While Demirbas5 found that the reaction

of cotton seed oil, hazelnut kernel oil, safflower seed and

sunflower seed oil were completed at a molar ratio of methanol

to vegetable oil of 41, 250 uC, and a 300 s reaction time in a

100 mL batch reactor. Madras et al.6 found that the reaction of

sunflower seed oil was completed at a molar ratio of methanol

to vegetable oil of 40, 400 uC, and an 1800 s reaction time in a

250 mL batch reactor.

Recently, the reactivity of various alcohols in a transesteri-

fication reaction under supercritical conditions was studied at

300 uC reaction temperature.7 It was found that the alcohol

reactivity relates to their critical properties, and methanol

has the highest reactivity. Furthermore, the effects of water

and free fatty acid on biodiesel production in supercritical

methanol, compared with those of biodiesel prepared by acidic

and basic catalyzed methods, were investigated.8 The presence

of water and free fatty acid did not have a significant effect on

the yield in the supercritical process. Moreover, transesteri-

fication in supercritical methanol employing propane and CO2

as co-solvents was developed.9,10 Their studies indicate that the

addition of co-solvent allows the reaction to be completed

under milder conditions due to the lower critical properties of

the reaction mixture. Recently, biodiesel production in a lab-

scale plug flow reactor was developed by us.11 Vegetable oils

(palm kernel and coconut oils) and methanol were fed into a

247 mL reactor by HPLC pumps at approximately 10 g min21

maximum flow rate. The optimal conditions of this flow

reactor were: molar ratio of methanol to vegetable oil, 42;

350 uC; and, 400 s reaction time. The results of the lab-scale

reactor were thus used for the reactor scale-up design.

In running the scale-up reactor at higher flow rates, the

viscosity of vegetable oil posed some problems, such as

fluctuation of system pressure and lower flow rate output.

The first viscosity reduction method was attempted by heating

the vegetable oil to around 70–80 uC. By doing this, the

viscosity of the vegetable oil and the fluctuation of system

pressure were reduced, but flow rate output was still low. In

the second method, the vegetable oil was mixed with a liquid

co-solvent. For simplification, this study selected tetrahydro-

furan (THF) and hexane as co-solvents.

THF improves the solubility of methanol in vegetable oil

and forms a single-phase mixture. A single high-pressure pump

could be used to feed the vegetable oil–THF–methanol mixture

into the reactor. Hexane, on the other hand, which is the

Fuels Research Center, Department of Chemical Technology, Faculty ofScience, Chulalongkorn University, Bangkok, Thailand.E-mail: somkiat.n@chula.ac.th; Fax: +66 2255 5831; Tel: +66 2218 7678{ This paper was published as part of the special issue from the ‘‘GreenChemistry for Fuel Synthesis and Processing’’ symposium at the 232ndACS National Meeting.

PAPER www.rsc.org/greenchem | Green Chemistry

This journal is � The Royal Society of Chemistry 2007 Green Chem., 2007, 9, 679–685 | 679

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

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

conventional solvent in the vegetable oil extraction process,

could be another good solvent for vegetable oil for the

continuous production of biodiesel in supercritical methanol.

Moreover, as their boiling points are close to that of methanol

(65 uC, 66 uC and 69 uC for methanol, THF and n-hexane,

respectively) both co-solvent and methanol could be recovered

simultaneously at the end of reaction and recycled.

Experimental

Materials

Crude palm kernel oil (PKO) was supplied by Chumporn Palm

Oil Industry, PCL. Its analysis, together with diesel analysis,

is presented in Tables 1 and 2. The sample was warmed and

filtered prior to use. Commercial grade methanol, hexane

(mixture of C6 isomers containing more than 65% n-hexane)

and tetrahydrofuran (THF) were used with no further purifica-

tion. All standard methyl esters for gas chromatograph

calibration and methyl undecanoate (internal standard) were

supplied by Fluka. Analytical grade carbon disulfide (CS2),

which was used as a dilution solvent for the gas chromato-

graphy, was supplied by Merck.

Experimental design

The 2-replicated 23 factorial design with 4 central runs was

employed in this study, requiring 20 experiments for each

reactor and each co-solvent.12 The factors were temperature

(290–350 uC), molar ratio of methanol to vegetable oil (12–42)

and molar ratio of co-solvent to vegetable oil (0–5) and the

resulting response was methyl esters content. Tables 3 and 4

show the factors, levels and experimental result employing the

THF and hexane process in a 250 mL reactor, respectively.

The result in Table 4, combined with zero moles of co-solvent

in oil (unemployed co-solvent process, run order 13–20) from

Table 3, was used to minimize the number of experiments.

To eliminate the effect of pressure, the amount of reactants

and co-solvent was adjusted to a specified pressure (19.0 MPa)

by using the Redlich–Kwong equation of state and the Lorentz–

Berthelot-type mixing rule.13 Unfortunately, the calculated

pressure was not exactly equal to the observed pressure. In

some experiments, where there was a large difference between

the calculated and observed pressure, the amounts of reactant

and co-solvent were readjusted by trial and error. The observed

pressures for each experiment are shown in Tables 3 and 4.

As for the 5.5 mL reactor, the experimental design was

similar to that for the 250 mL reactor. The amounts of

reactants were also adjusted for a pressure of 19.0 MPa.

Table 1 Physical properties of PKO sample and diesel fuels

Properties

Highspeeddiesel

Lowspeeddiesel

Palmkerneloil

Kinematic viscosity at 40 uC/cSt 4.1 6.3 30.8Gravity/uAPI 33.2 26.1 24.6Cetane index 54 46 35Flash point/uC 75 78 218Heating value/MJ kg21 43.1 42.1 34.6Acid value/mg KOH g21 oil — — 11.3Saponification value/mg KOH g21 oil — — 198.6Iodine value/g I2 per 100 g oil — — 12.8

Table 2 Fatty acid profile in PKO sample

NameComposition

Molecular weight

(%wt) Fatty acid Pseudo-triglyceride

Caprylic acid 2.99 144 470Capric acid 3.06 172 554Lauric acid 49.51 200 638Myristic acid 17.91 228 722Palmitic acid 8.64 256 806Stearic acid 4.19 284 890Oleic acid 11.64 282 884Linoleic acid 2.07 280 878Average molecular weight 704

Table 3 Experimental data from employed THF process in 250 mLreactor for 10 min with crude PKO as reactant

Runorder

Temperature/uC

Molar ratioof THFto oil

Molar ratioof MeOHto oil

Methyl esterscontent (%wt)

Pressure/MPa

1 350 0.0 12.1 79.7 17.92 290 5.6 41.3 72.6 17.53 350 4.8 39.9 86.5 19.34 290 5.0 41.9 73.2 17.45 350 0.0 41.3 84.9 19.06 290 0.0 42.2 63.7 16.67 320 2.4 24.0 79.6 19.08 350 5.0 12.2 79.6 19.89 320 2.5 23.9 80.4 17.8

10 350 4.9 41.9 79.3 19.411 350 0.0 12.1 80.9 19.812 320 2.6 24.1 79.7 19.113 350 0.0 42.1 85.1 19.614 290 5.1 11.9 45.0 16.315 350 5.1 12.1 82.5 21.616 290 0.0 12.1 43.7 15.617 290 0.0 11.2 47.6 16.218 290 5.1 12.3 47.3 15.919 320 2.5 24.1 78.6 18.020 290 0.0 42.1 62.3 16.9

Table 4 Experimental data from employed hexane process in 250 mLreactor for 10 min with crude PKO as reactant

Runorder

Temperature/uC

Molar ratioof hexaneto oil

Molar ratioof MeOHto oil

Methyl esterscontent (%wt)

Pressure/MPa

1 290 4.7 41.3 62.4 19.62 350 4.6 40.9 87.6 18.73 290 5.1 12.4 46.2 19.24 290 4.8 42.6 65.2 20.05 350 5.0 12.4 79.8 18.66 320 2.5 24.2 77.8 19.27 290 4.9 12.2 48.5 20.08 320 2.6 24.3 78.2 18.69 350 5.0 43.4 88.1 17.7

10 320 2.5 24.2 76.5 18.211 320 2.4 24.1 76.6 19.612 350 5.2 12.4 85.9 19.013 290 0.0 12.1 43.7 15.614 290 0.0 11.1 47.6 16.215 350 0.0 12.1 80.9 19.816 350 0.0 12.1 79.7 17.917 290 0.0 42.2 63.7 16.618 290 0.0 42.1 62.3 16.919 350 0.0 42.1 85.1 19.620 350 0.0 41.3 84.9 19.0

680 | Green Chem., 2007, 9, 679–685 This journal is � The Royal Society of Chemistry 2007

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

View Article Online

Tables 5 and 6 show the factors, levels and experimental result

for the employed THF and hexane process in 5.5 mL reactor,

respectively.

Experimental setup

First, the effect of co-solvents was studied in a 250 mL batch

reactor (Parr Instrument Company, Model 4842), equipped

with a mechanical stirrer and internal cooling. The reactor

was heated with an external electrical heater. The maximum

pressure and temperature values of the equipment were 50 MPa

and 500 uC, respectively. The temperature of the reactor was

measured with a j-type thermocouple and controlled at ¡5 uCfor a set time. The pressure of the reactor was measured with a

pressure gauge and transducer.

As the highest methyl esters content obtained from the

250 mL reactor was slightly lower than expected, the effect of

co-solvent was studied in the 5.5 mL batch tube reactor, the

dimensions of which were 0.375 mm od, 0.124 mm thickness

and 200 mm length. The reactor was heated by immersion in a

fluidized sand bath. The temperature was measured by a

k-type thermocouple.

Experimental procedure

(a) 250 mL reactor. The reaction vessel was charged with a

given amount of vegetable oil, methanol and co-solvent and

was then heated to the desired temperature. The reaction

time and stirring speed were fixed at 10 min and 500 rpm,

respectively, for every experiment. At the end of the reaction,

the reactor was quenched in an ice–water bath to approxi-

mately room temperature and pressure. The content in the

reaction vessel was weighed and put in a rotary evaporator to

remove the solvent phase (co-solvent and methanol). The oil

phase was left to settle for at least 8 hours in a separating

funnel to ensure complete separation. Two liquid phases were

obtained: ester (top layer), and crude glycerol (bottom layer).

The ester layer was then analyzed for methyl esters content.

(b) 5.5 mL reactor. The reaction vessel was charged with a

given amount of reactant and co-solvent, immersed in a

fluidized sand bath at the designed temperature and was

shaken manually from time to time to ensure uniform mixing.

The reaction time was held constant at 10 min. At the end of

the reaction, the reactor was then quenched in an ice–water

bath to stop the reaction. The solvent phase was then

evaporated by warming in a water bath at 80 uC for 2 hours.

Glycerol was separated by centrifuging at 4000 rpm for 10 min.

The ester (top layer) phase was then analyzed for methyl

esters content.

Methyl esters analyses

For the methyl esters content measurement, a gas chromato-

graph (Varian Model CP-3800), equipped with a capillary

column coated with polydimethylsiloxane (30 m 6 0.25 mm 60.25 mm, DB-1, J&W Scientific) and an FID detector, was

used with helium as the carrier gas. The ester product and

the known amount of internal standard was diluted with

CS2 before injection and standardized by the internal

standard method. The temperature of the injection port and

detector were 250 uC and 280 uC, respectively. The column

oven was held at 110 uC for 2 min and then raised to 260 uCat 15 uC min21. The final temperature was held constant

for 10 min.

The methyl esters content was calculated from their content

in the biodiesel product as analyzed by GC. The content (or

purity) was defined as a ratio of the weight of methyl esters, as

obtained from GC, to the total weight of the biodiesel product.

For the GC-MS analysis, a Shimadzu Model GCMS-

QP2010 gas chromatograph coupled with a mass spectrometer

and equipped with a capillary column coated with poly-

dimethylsiloxane (30 m 6 0.25 mm 6 0.25 mm, DB-1ms, J&W

Scientific) was used with helium as the carrier gas. The sample

was diluted in CS2 before injection. The injection port, ion

Table 5 Experimental data from employed THF process in 5.5 mLreactor for 10 min with refined PKO as reactant

Runorder

Temperature/uC

Molar ratioof THFto oil

Molar ratioof MeOHto oil

Methyl esterscontent (%wt)

1 350 0.0 44.5 99.32 320 3.7 24.3 83.23 350 5.2 12.0 83.34 320 2.1 20.2 81.35 290 0.0 12.0 53.66 350 5.0 43.0 97.57 350 0.0 12.0 81.18 350 0.0 40.8 99.49 290 8.3 20.1 56.5

10 290 4.8 40.7 78.311 290 5.0 12.0 54.212 320 2.3 21.7 82.513 350 5.1 41.9 98.714 350 5.3 12.2 84.615 290 0.0 41.9 78.216 320 2.5 22.9 84.617 350 0.0 12.0 82.618 290 0.0 41.8 76.819 290 5.2 42.4 77.520 290 0.0 12.0 57.0

Table 6 Experimental data from employed hexane process in 5.5 mLreactor for 10 min with refined PKO as reactant

Runorder

Temperature/uC

Molar ratio ofhexane to oil

Molar ratio ofMeOH to oil

Methyl esterscontent (%wt)

1 320 2.6 23.9 80.52 290 5.4 42.0 79.53 320 2.5 23.9 81.24 290 5.4 12.0 52.25 350 5.1 11.9 76.06 290 5.0 11.9 54.47 350 6.8 43.5 97.18 320 2.6 23.7 80.29 290 5.0 41.1 80.1

10 320 2.5 23.0 80.211 350 4.9 37.9 86.712 350 5.0 12.2 78.113 290 0.0 12.0 57.014 290 0.0 12.0 53.615 290 0.0 41.8 76.816 290 0.0 41.9 78.217 350 0.0 12.0 81.118 350 0.0 12.0 82.619 350 0.0 44.5 99.420 350 0.0 40.8 99.3

This journal is � The Royal Society of Chemistry 2007 Green Chem., 2007, 9, 679–685 | 681

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

View Article Online

source and interface temperature were 250, 200 and 230 uC,

respectively. The molecular weight scan range was 50–800 m/z

and 3 min of solvent cut time. The column was held at 90 uCfor 5 min and then raised to 260 uC at 20 uC min21. The final

temperature was held constant for 10 min.

Results and discussion

Physical characteristics of PKO sample versus diesel fuel

From Table 1, it is clear that the PKO had a lower heating

value and much higher viscosity than those of diesel fuels.

This indicates troublesome atomization and poor engine

performance if these oils are to be used as neat fuels.

Furthermore, this PKO had higher acid value, which implies

that biodiesel with lower methyl esters content would result

if this PKO was used as the reactant in alkaline catalytic

transesterification.8

From Table 2, the PKO sample contained lauric acid as the

major fatty acid, which is consistent with the literature.14 The

chemical formula of PKO was represented by a pseudo-

triglyceride, which consists of 3 molecules of fatty acid and a

molecule of glycerol. The average molecular weight was

calculated by the pseudo-triglyceride molecular weight and

corresponding fatty acid composition and was found to be 704.

The critical properties of the sample, as estimated by the GC

method,15,16 are given in Table 7. Additional details of these

calculations can be found in our previous work.11

Reaction between vegetable oil, methanol and co-solvents

To ensure that the reaction between co-solvent and other

reactants did not occur, the GC-MS chromatograms of mixed

methyl esters standard and biodiesel products, obtained from

the THF used at 350 uC, a molar ratio of methanol to

vegetable oil of 42 and 5 mole of THF in vegetable oil at 10 min

reaction time, were obtained as illustrated in Figs. 1 and 2.

Comparing Fig. 1 with Fig. 2, it can be seen that the

biodiesel composition from the employed THF process was

basically the same as the mixed fatty acid methyl esters

standard. On the other hand, from the THF and hexane

chromatograms in Figs. 3 and 4, one can deduce that THF and

hexane peaks did not show up in Fig. 2. Therefore, it is

concluded that there was no co-solvent interference in the

transesterification reaction.

The critical properties of the PKO sample, methanol,

n-hexane and THF were the parameters in the Redlich–

Kwong equation of state and the Lorentz–Berthelot-type

mixing rules employed to calculate the amount of methanol,

co-solvent and vegetable oil for a specified pressure.

Effect of co-solvent on methyl esters content in biodiesel

production in a 250 mL reactor

The experimental order (Table 3 and run order 1–12 in Table 4)

was done randomly. For run order 13–20 in Table 4, i.e. the

unemployed co-solvent process, the experimental data were

obtained from Table 3. All experimental data were analyzed

Table 7 Critical properties of methanol, THF, n-hexane and PKOsample

Criticalproperties Methanol THF n-Hexane PKO Unit

Tc 512.6 540.2 507.6 926.1 KPc 7.91 5.12 3.06 0.59 MPaVc 0.118 0.224 0.373 2.476 L mol21

Fig. 1 GC-MS chromatogram of mixed methyl esters standard.

Fig. 2 GC-MS chromatogram of biodiesel from employed THF

process in 250 mL reactor for 10 min with crude PKO as reactant.

Fig. 3 GC-MS chromatogram of THF phase in 250 mL reactor for

10 min with crude PKO as reactant.

Fig. 4 GC-MS chromatogram of hexane phase in 250 mL reactor for

10 min with crude PKO as reactant.

682 | Green Chem., 2007, 9, 679–685 This journal is � The Royal Society of Chemistry 2007

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

View Article Online

by the factorial design procedure13 to obtain the analysis of

variance tables (ANOVA), which are shown in Tables 8 and 9.

From Tables 8 and 9, the molar ratio of co-solvent to

vegetable oil (factor B) and its interaction (factors AB, BC and

ABC) had no significant effect on methyl esters content, as

noticed from the probability of F value less than 0.05, at the

confidence level of 95%. Thus, it was concluded that the

addition of a co-solvent in this process did not show either

negative or positive effect on methyl esters content.

The regression models in terms of coded units of employed

THF and hexane process then can be correlated as shown in

equations 1 and 2, respectively. Temperature (A) and the

molar ratio of methanol to vegetable oil (C) has positive effects

on methyl esters content, and the temperature effect has a

higher magnitude than the molar ratio of methanol to veget-

able oil effect by approximately two orders. The interaction

term (AC) indicated the complete regression model might be

second order, which is consistent with our previous finding.11

%ME = 71.88 + 12.67A + 5.90C 2 4.67AC (1)

%ME = 71.21 + 14.53A + 5.12C 2 2.99AC (2)

where %ME is methyl esters content in biodiesel product

(%wt), and A is temperature in terms of coded unit, derived by

the equation:

A~Temperature 0Cð Þ{320

30(3)

C is the molar ratio of methanol to vegetable oil in terms of

coded unit, derived by the equation:

C~MeOH moleð Þ{27

15(4)

and AC is the product of A and C.

Effect of reaction time on methyl esters content

From Tables 3 and 4, the highest methyl esters content

obtained was found to be not over 88%, which is lower

than the 95% reported in the latest literature of experiments

employing co-solvents.9,10 According to relevant literature, the

conditions were: reaction temperature 350–400 uC; molar

ratio of methanol to vegetable oil 40–42; reaction time

2.4–40 min; and, reactor volume 5–250 mL.3–6,9,10 To confirm

that the reaction reached equilibrium in this work, we ran

another set of experiments for 5–60 min and the result is

illustrated in Fig. 5.

From Fig. 5, maximum methyl esters content was reached

after 30 min reaction time and a methyl esters content of

92.0 ¡ 1% was observed. Because of the slightly lower methyl

esters content, we further established two possible hypotheses

and verified them as follows.

Effect of temperature gradient between reactor wall and bulk

fluid

In the 250 mL reactor, the reaction vessel employed was heated

externally and the contents in the vessel were mixed by a

Table 8 Analysis of variance from employed THF process in 250 mL reactor for 10 min with crude PKO as reactant

Source Sum of squares Degree of freedom Mean square F value Probability . F

A (Temperature) 2583.77 1 2583.77 69.47 ,0.0001B (THF to oil) 18.28 1 18.28 0.49 0.4966C (MeOH to oil) 538.21 1 538.21 14.47 0.0025AB 31.45 1 31.45 0.85 0.3759AC 338.75 1 338.75 9.11 0.0107BC 10.41 1 10.41 0.28 0.6065ABC 39.51 1 39.51 1.06 0.3230Residual 446.29 12 37.19Total 4043.78 19

Table 9 Analysis of variance from employed hexane process in 250 mL reactor for 10 min with crude PKO as reactant

Source Sum of squares Degree of freedom Mean square F value Probability . F

A (Temperature) 3374.01 1 3374.01 142.32 ,0.0001B (Hexane to oil) 15.18 1 15.18 0.64 0.4391C (MeOH to oil) 421.18 1 421.18 17.77 0.0012AB 2.16 1 2.16 0.09 0.7679AC 141.41 1 141.41 5.96 0.0310BC 4.31E-03 1 4.32 E-03 1.8 E-04 0.9895ABC 0.08 1 0.08 3.4 E-03 0.9539Residual 284.49 12 23.71Total 4249.48 19

Fig. 5 Changes in methyl esters content of unemployed co-solvent

process as a function of reaction time in 250 mL reactor with crude

PKO at 350 uC and 42 molar ratio of methanol to vegetable oil.

This journal is � The Royal Society of Chemistry 2007 Green Chem., 2007, 9, 679–685 | 683

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

View Article Online

stirrer. As the temperature near the wall of the vessel is

somewhat higher than in the center, it may cause an in situ

thermal cracking reaction, resulting in slightly lower methyl

esters content. To verify this hypothesis, the experiments were

performed at high temperature, i.e. 350 and 400 uC, in both 5.5

and 250 mL reactors at 30 min reaction time, and the biodiesel

samples were analyzed. The chromatograms obtained were

then compared with that obtained from the conventional

method at 60 uC, 30 min reaction time, 1% NaOH, and 6 moles

of methanol in vegetable oil.

From Fig. 6, sample (a), it was assumed that no thermal

cracking occurred. In the chromatogram of this sample, no

noise was detected, as in sample (b). Therefore, it was con-

cluded that thermal cracking did not take place in the 5.5 mL

reactor at 350 uC. Sample (c) had higher noise than samples (a)

and (b). The retention times of the noises in both sample (c)

and (d) were nearly the same, so these noises were probably the

same compounds derived from the thermal cracking reaction

at temperature over 350 uC. However, the results from GC-MS

(Fig. 2) did not show this noise, probably because that sample

was run at 10 min reaction time.

According to the methyl esters content obtained from the

5.5 mL reactor, it is assumed that there was no temperature

gradient between the reactor wall and the bulk fluid, at 350 uC,

a molar ratio of methanol to vegetable oil of 42, a molar ratio

of co-solvent to vegetable oil of 0–5, and 10 min reaction time.

The experimental data are illustrated in Table 10. The methyl

esters content of the biodiesel products was slightly higher

than those obtained from the 250 mL reactor under the same

conditions (see Tables 3 and 4). Therefore, the assumption that

thermal cracking occurred in the 250 mL reactor might be

plausible. However, the methyl esters content was still slightly

lower than the literature value, especially for the co-solvent

process employed. Therefore, we established a second hypo-

thesis, which is discussed in next section.

Effect of contaminants in crude palm oil

The crude palm oil contained 95–98% of triglyceride and 2–5%

complex minor compounds such as wax ester, hydrocarbons,

pigments and alcoholic compounds.17 Thus, the slightly lower

methyl esters content obtained from crude PKO compared

with that from refined PKO was probably due to the

percentage of triglyceride. This hypothesis was verified in the

5.5 mL reactor, to ensure thermal cracking did not take

place, by using refined PKO (cooking oil shelf product) as the

reactant. The effect of co-solvent by using the experimental

design as mentioned earlier was also investigated. The

experimental data, as shown in Tables 5 and 6, were treated

by the factorial design procedure and the ANOVA tables, as

shown in Tables 11 and 12, respectively.

From Tables 5 and 6, the maximum methyl esters content

was 99.4%. It was clear that the maximum methyl esters

content from refined PKO was higher than that from crude

PKO, which was 96.3% (Table 10).

From Tables 11 and 12, it is clear that the co-solvents did

not affect the methyl esters content. Also, the interaction

term between temperature and molar ratio of methanol to

vegetable oil (Factor AC) had no significant effect. The

regression model in terms of coded units for a 5.5 mL reactor

for the employed THF and hexane process are given in eqns 5

and 6, respectively;

%ME = 79.79 + 12.32A + 9.08C (5)

%ME = 78.24 + 10.51A + 10.11C (6)

where %ME is methyl esters content in biodiesel product

(%wt), A is temperature in terms of coded unit, derived by

eqn (3) and C is molar ratio of methanol to vegetable oil in

terms of coded unit, derived by eqn (4).

For the 5.5 mL reactor, the regression model indicated that

the effect of temperature and molar ratio of methanol to

Fig. 6 Comparison of GC chromatogram of biodiesel product from

crude PKO at various temperatures, indicating some noises observed

in (c) and (d). (a) Conventional method (60 uC), (b) 350 uC in 5.5 mL

reactor, (c) 400 uC in 5.5 mL reactor and (d) 350 uC in 250 mL reactor.

Table 10 Experimental data from 5.5 mL reactor for 10 min withcrude PKO as reactant, temperature controlled by fludized sand bath

ProcessTemperature/uC

Molar ratioof co-solventto oil

Molar ratioof methanolto oil

Methyl esterscontent (%wt)

Unemployedco-solvent

350 0.0 42.0 95.5350 0.0 42.1 96.3

THF 350 5.1 41.1 94.7350 6.8 45.1 93.8

Hexane 350 5.2 40.3 93.9350 5.8 42.3 94.5

684 | Green Chem., 2007, 9, 679–685 This journal is � The Royal Society of Chemistry 2007

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

View Article Online

vegetable oil had a similar magnitude and also had a positive

effect. Furthermore, the overall mean (the first term on the

right hand side) was higher than the regression model of the

250 mL reactor. This indicates that the methyl esters content

in the biodiesel product, which is obtained from the same

temperature and molar ratio of methanol to vegetable oil in

the 5.5 mL reactor, is always higher than that obtained from

the 250 mL reactor.

Conclusions

The effects of co-solvents (THF and hexane) on the production

of biodiesel from PKO in supercritical methanol were

successfully investigated. This system achieves equilibrium

after 30 min reaction time in a 250 mL reactor and in less than

10 min in a 5.5 mL reactor. The reaction of co-solvent with

other reactants (vegetable oil and methanol) did not occur at

supercritical conditions. The addition of co-solvents in this

process did not show either negative or positive effects on the

methyl esters content. The methyl esters content from crude

PKO was slightly lower than that of refined PKO, plausibly

because of the lower triglyceride content.

Acknowledgements

We would like to express our sincere appreciation to

Chumporn Palm Oil Industry, PCL, for supplying the palm

kernel oil samples. We are also grateful for the financial

support from Clean and Green Fuel Research Unit, and

Graduate School, Chulalongkorn University.

References

1 S. Marinkovic and A. Tomasevic, Fuel, 1998, 77, 1389–1391.2 M. Diasakou, A. Louloudi and N. Papayannakoa, Fuel, 1998, 77,

1297–1302.3 S. Saka and D. Kusdiana, Fuel, 2001, 80, 225–231.4 D. Kusdiana and S. Saka, Fuel, 2001, 80, 693–698.5 A. Demirbas, Energy Convers. Manage., 2002, 43, 2349–2356.6 G. Madras, C. Kolluru and R. Kumar, Fuel, 2004, 83, 2028–2033.7 T. Warabi, K. Kusdian and S. Saka, Bioresour. Technol., 2004, 91,

283–287.8 D. Kusdiana and S. Saka, Bioresour. Technol., 2004, 91, 289–295.9 W. Cao, H. Han and J. Zhang, Fuel, 2005, 84, 347–351.

10 H. Han, W. Cao and J. Zhang, Process Biochem., 2005, 40,3148–3151.

1 1 K . B u n y a k i a t , S . M a k m e e , R . S a w a n g k e a w a n dS. Ngamprasertsith, Energy Fuels, 2006, 20, 812–817.

12 D. Montgomery, Design and Analysis of Experiments, John Wiley& Sons, Inc., New York, 2001, pp. 218–276.

13 S. Espinosa, T. Fornari, S. Bottini and E. Brignole, J. Supercrit.Fluids, 2002, 23, 91–102.

14 A. E. Bailey, D. Swern, M. W. Formo and T. H. Applewhite,Bailey Industrial Oil and Fat Products Volume 1, ed. M. W.FormoJohn Wiley & Sons, New York, 1979, p. 298.

15 N. P. Chopey, Handbook of Chemical Engineering Calculation,McGraw-Hill, Inc., New York, 1994, pp. 1–8.

16 Perry’s Chemical Engineers’ Handbook, R. H. Perry, D. W. Greenand J. O. Maloney, McGraw-Hill, Inc., New York, 1997,pp. 227–246.

17 A. Cert, W. Moreda and M. C. Perez-Camino, J. Chromatogr., A,2000, 881, 131–148.

Table 11 Analysis of variance from employed THF process in 5.5 mL reactor for 10 min with refined PKO as reactant

Source Sum of squares Degree of freedom Mean square F value Probability . F

A (Temperature) 2400.84 1 2400.84 138.81 ,0.0001B (THF to oil) 2.00 1 2.00 0.12 0.7399C (MeOH to oil) 1340.93 1 1340.93 77.53 ,0.0001AB 6.11 1 6.11 0.35 0.5634AC 58.91 1 58.91 3.41 0.0898BC 0.02 1 0.02 0.00 0.9756ABC 8.39 1 8.39 0.49 0.4994Residual 207.55 12 17.30Total 3950.57 19

Table 12 Analysis of variance from employed hexane process in 5.5 mL reactor for 10 min with refined PKO as reactant

Source Sum of squares Degree of freedom Mean square F value Probability . F

A (Temperature) 1783.85 1 1783.85 136.94 ,0.0001B (Hexane to oil) 22.61 1 22.61 1.74 0.2123C (MeOH to oil) 1640.57 1 1640.57 125.94 ,0.0001AB 27.62 1 27.62 2.12 0.171AC 53.23 1 53.23 4.09 0.0661BC 9.30 1 9.30 0.71 0.4146ABC 2.88 1 2.88 0.22 0.6466Residual 156.31 12 13.03Total 3673.21 19

This journal is � The Royal Society of Chemistry 2007 Green Chem., 2007, 9, 679–685 | 685

Publ

ishe

d on

18

Apr

il 20

07. D

ownl

oade

d by

Ren

ssel

aer

Poly

tech

nic

Inst

itute

on

19/1

0/20

14 0

6:04

:22.

View Article Online