Fuel Processing Technology 124 (2014) 7077

Contents lists available at ScienceDirect

Fuel Processing Technology

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

Esterication of free fatty acids to fatty acid alkyl esters in a bubble

column reactor for use as biodiesel
Colin J. Stacy, Cory A. Melick, Richard A. Cairncross
Chemical and Biological Engineering Department, Drexel University, Philadelphia, PA 19104, USA

a r t i c l e

i n f o

Article history:
Received 17 May 2013
Received in revised form 5 February 2014
Accepted 7 February 2014
Available online 12 March 2014
Keywords:
Biodiesel
Fatty acid methyl ester
Free fatty acid
Trap grease
Esterication

a b s t r a c t
Biodiesel is readily produced from rened vegetable oils; however, many low-value degraded and waste oils contain a high concentration of free fatty acids, which are difcult to convert to fatty acid alkyl esters for use as biodiesel. This paper evaluates the performance of an acid-catalyzed bubble column reactor that is highly robust for
the esterication of free fatty acids to fatty acid alkyl esters. The bubble column reactor typically operated at
120 C and ambient pressure; methanol bubbling through the reactor reacts with free fatty acids and strips byproduct water, which enables high conversions and makes the reactor more robust to water than other reactor
designs. This paper shows the effects alcohol, feedstock type and quality, alcohol ow rate, and oil feedstock
on the reactor performance. Most reaction conditions produce greater than 98% conversion in less than 2 h,
including reactions with alcohols containing 10% water by volume and lipids extracted from trap grease.
2014 Elsevier B.V. All rights reserved.

1. Introduction
Rising prices of crude oil and concerns about carbon dioxide
emissions have led to increased research into alternative fuels
and renewable energy. There are several other pressures for increased research into renewable fuels including: worsening air
quality by emissions of sulfur oxides, particle matter, and other
gases; security of domestic energy supply; and limited long-term
supply of petroleum. This paper analyzes performance of a bubble
column reactor for producing biodiesel from low-value lipid feedstocks containing high concentrations of free fatty acids (FFA);
these low-value lipids can be obtained from non-food crops
grown on marginal land or from waste fats, greases, and oils.
Biodiesel is a promising renewable fuel. Biodiesel is a name given to
fatty acid alkyl esters that are suitable diesel substitutes. Biodiesel is typically produced chemically by reacting plant or animal derived lipids
with an alcohol. The majority of biodiesels are produced by reacting
lipids with methanol to produce fatty acid methyl esters (FAME).
Throughout this paper, the term FAME is used to refer to the reaction
product. Before selling FAME as biodiesel, it is necessary to demonstrate
that the product satises the full suite of ASTM specications for biodiesel as a diesel substitute. The majority of current biodiesel is produced
from rened lipids with low FFA concentrations, such as soybean oil
(in USA), rapeseed oil (in Europe), and palm oil (in Asia), which are

Corresponding author at: 3141 Chestnut Street, Philadelphia, PA 19104. Tel.: +1 215
895 2230.
E-mail address: cairncross@drexel.edu (R.A. Cairncross).

http://dx.doi.org/10.1016/j.fuproc.2014.02.003
0378-3820/ 2014 Elsevier B.V. All rights reserved.

agricultural crops that are relatively high-cost, require signicant fertilizer and chemical inputs, and compete with food crops for land. However, biodiesel is not a complete solution for the United States energy
demand. Van Gerpen [1] noted that only about 14% of current diesel
demand can be replaced by biodiesel produced from crop-based
lipids. Other waste lipid feedstocks can add signicant production
capacity [2].
Published techno-economic analyses of biodiesel production processes have predicted attractive conversion economics, but the results
are highly sensitive to the feedstock used. Apostolakou et al. [3] and
Haas [4] showed that the cost of raw materials can be 75%90% of the
cost of manufacturing. Marchetti et al. [5] reported three scenarios for
producing biodiesel from lipids containing 5% FFA, and all three processes were protable with raw materials being more than 80% of the
manufacturing costs. Zhang et al. [6,7] compared four biodiesel processes; the production of biodiesel from rened lipids had the lowest capital
expense and highest cost of raw materials. Processes converting waste
lipids to biodiesel required more methanol and larger distillation columns to recover methanol for recycle. Because feedstock is a major fraction of the manufacturing costs, processes that use low-value feedstocks
(such as waste oils) are likely to show much greater economic protability [2,4].
The reactor described in this paper has several potential advantages
related to sustainability including: exibility to varying FFA content,
exibility for alcohol feed, robustness to moisture, and reduced energy
requirements. The results section of this paper presents parametric
studies of how the bubble column reactor performance varies with
some key process variables. A robust reactor that can handle a variety
of feedstocks without additional pretreatment will reduce economic

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

71

hurdles to constructing and operating biodiesel production facilities and

will result in lower consumer prices of biodiesel products.

feedstock costs and is less prone controversies associated with creating

fuels from food-grade lipids [21,22].

1.1. Chemistry for production of fatty acid methyl esters

1.3. Process options for the esterication of high-FFA lipids

The most common reaction pathway for producing FAME is the

transesterication of triacyl glycerides (TAG), or triglycerides. During
transesterication, a TAG molecule reacts with three alcohol molecules
to form three FAME molecules with by-product glycerin:

There are several technologies available for converting high-FFA lipids

to FAME; acid-catalyzed esterication is effective over a large range of FFA
concentration and is often used for pretreatment prior to base-catalyzed
transesterication in a two-step process. A signicant disadvantage of
acid catalysts is slower reactions. There are several ways to increase
acid-catalyzed esterication reaction rates, including the following: increasing temperature, increasing catalyst concentration, and removing
by-product water. For feedstocks containing 1%10% FFA, a two-step process (low-temperature acid-catalyzed esterication followed by basecatalyzed transesterication) is the most common method. For feedstocks
containing more than 50% FFA, multiple moderate-pressure reactors with
intermediate removal of water are used effectively [23]. Multiple, identical reactors with intermediate water removal not only increases the reaction time and conversion, but it also increases the capital and operating
costs signicantly. To achieve acceptable reaction rates, temperatures
above the boiling point of methanol are often used, which requires elevated pressure to maintain methanol in the liquid phase; for example, Van
Gerpen reports using 240 C and 90 bar [1] and Berry and Ratigan report
115 C and 5.4 bar [23].
However, Kocsisova et al. [13] showed that high conversion can be
achieved with acid-catalyzed reactions by bubbling methanol vapor
through FFA at ambient pressure and elevated temperature. Bubbling
methanol vapor through hot lipid provides several benets: use of
high temperatures corresponds to faster reaction kinetics, simultaneous
removal of water reduces equilibrium limitations, and intensive agitation of the reaction mixture enhances mass transfer. Kocsisova et al.
[13] demonstrated that esterication in a bubble reactor at temperatures 50%60 C higher than the boiling point is effective and requires
lower amounts of methanol (methanol:FFA molar ratio of 3:1 to
4:1 is sufcient) for high conversion. Kocsisova et al. [13] claims
that feeding the methanol as a liquid into the reactor and local excess
of methanol near the feed were necessary for high conversion. The
reactor described in this paper extends on the work of Kocsisova
et al. [13] by using a column reactor conguration and by providing
a more detailed study of how the rate of methanol fed to the reactor
affects reaction conversion.

1
In this reaction equation, the alcohol is methanol, and R represents a long aliphatic fatty acid chain, which typically contains 8
22 carbon atoms [8]. The reaction in Eq. (1) is a simplication of a series of three reactions in which TAG sequentially reacts with methanol molecules to form a diacyl glyceride molecule (DAG), a monoacyl
glyceride (MAG), and nally glycerin with one FAME molecule produced at each step. Industrial biodiesel processes predominantly
use transesterication reactions and base catalysts because this reaction requires low operating temperature and achieves high conversions within a couple of hours. A major drawback of the basecatalyzed transesterication reactions is that the lipid feedstock
must be high purity in TAG; if the FFA content exceeds 1%, the
soaps that form from the reaction between the base catalyst and
the FFA hinder transesterication and downstream purication
[1,9,10]. The alternative acid-catalyzed reactions do not form soaps
with FFA, but acid-catalyzed transesterication reactions are much
slower than base catalyzed transesterication reactions [1,11].
For high-FFA lipids, acid-catalyzed esterication is effective for producing FAME from FFA:

2
The esterication of FFA by Eq. (2) catalyzed with acids like sulfuric
acid have been the subject of a number of studies [9,1218]. Esterication reactions are reversible and are equilibrium-limited by the accumulation of the by-product water. The presence of water generally
limits the conversion to FAME that can be achieved in an acidcatalyzed reactions, and its continuous removal has been shown to dramatically increase yields [14]. The bubble column reactor described in
this paper is effective at continuously removing water from the system,
which results in high yields of FAME when using FFA feedstocks.
1.2. Biodiesel process feedstocks
The quality and the price of potential lipid feedstocks are related to
their FFA content. Edible lipids have low FFA content and high prices. Inedible lipids tend to be high in FFA and have low prices. The high-FFA
lipids are mostly waste products and have limited commercial value,
while low-FFA lipids tend to be viable food sources. For example, soybean
oil currently sells for about $3.52 per gallon, and yellow grease (ltered
and dewatered waste cooking oil with FFA content b15%) sells for $2.19
[19]. Trap grease is a potential source of high-FFA lipids; wastewater utilities charge $0.06 or more per gallon to dispose of trap grease [20]. Trap
grease contains 2%10% lipids, and lipids separated from trap grease can
be over 95% FFA; more details about trap grease are provided in
Section 3.4 [2]. Producing biodiesel from high-FFA lipids entails low

2. Materials and methods

2.1. Chemicals and materials
Oleic acid was used to model an FFA feedstock for the experiments in
this paper unless otherwise noted. Oleic acid at technical grade purity
(N 90%) as well as toluene and methanol with purities above 99% were
purchased from Sigma Aldrich and used without further purication.
Ethanol was anhydrous and denatured with 5% isopropyl alcohol and
was also purchased from Sigma Aldrich. Isopropyl alcohol was purchased from Azer Scientic and was 99.99% purity. Sulfuric acid at 93%
weight (66 Baume) was purchased from Fischer Scientic. Paratoluenesulfonic acid (PTSA) was purchased from Sigma-Aldrich and dissolved in methanol for use. Triglyceride samples were rened soybean
oil purchased from local supermarkets. Trap grease was donated by
Russell Reid waste management, and more information about the
experiments with trap grease is included in Section 3.4.
Fig. 1 displays a schematic of the reactor system. The primary reaction chamber was a jacketed glass column with several pumps used
for alcohol feeds and recirculation. Throughout the system, all parts
wetted by the reactants and products were glass, stainless steel, or
PTFE. The reactor and the alcohol-vaporizer/lipid-reheater were heated
by circulating hot silicone oil from a heating bath.

72

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

Fig. 1. Schematic of reactor showing bubbling action of reactor and water and methanol transfer between liquid and vapor phase.

2.2. Analysis of conversion by titration and NMR

The conversion of fatty acid chains to FAME was quantied using
two techniques: (1) base titration and (2) and nuclear magnetic resonance (NMR). The titrant used was a 0.1 molar solution of sodium hydroxide in methanol, and samples from the reactor were dissolved in
a titration solution containing equal parts toluene and isopropyl alcohol
with trace phenolphthalein as an indicator; this is a titration procedure
similar to AOCS Cd 3d-63 and ASTM D-664 [24,25]. The base titrant was
prepared from a standard base concentrate (Fixanal purchased from
Sigma Aldrich) and tested against an acid standard prior to experimentation. This titration determined the acid number of the sample (mg
KOH/g sample). For samples where the average molecular weight of
the fatty acid is known (for example, oleic acid), the mole fraction of
FFA was readily determined.
For experiments with partial TAG feedstocks, titration cannot
completely determine fatty acid conversion to FAME. So proton
NMR (H-NMR) spectroscopy was used to measure the FAME content
of samples over time. The machine used for analysis of FAME samples
was Drexel University's 500 MHz Inova Varian NMR. H-NMR peaks
were assigned to their appropriate functional groups as shown in
the supplementary documentation. A combination of NMR and titration enables determining the content of FFA, FAME, and acyl glyceride fatty acids.

2.3. Bubble column reactor design

Although the bubble column reactor experiments were motivated
by the work of Kocsisova et al. [13], the reactor was redesigned several
times before constructing the current bubble column reactor design
shown in Fig. 1. The bubble column reactor was a jacketed glass column
18-in. tall with a 1-in. internal diameter. Typically, the reactor operated
at temperatures of about 120 C with about 180 mL of feedstock lipids.
The top of the reactor was open to the atmosphere, which allows the
alcohol and water vapor bubbling through the reactor to exit into a
fume hood in which the experiments were performed. The interior of
the reactor had stainless steel tubing to transport chemicals into and
out of the reactor and stainless steel-cladded thermocouples to monitor

the reactor temperature. The column was topped with a glass funnel
that prevented overow of reactor contents during bubbling. As
shown in Fig. 1, methanol entered the bottom of the reactor as a
vapor. Although the solubility of methanol in lipids is small, some methanol from the vapor bubbles was believed to dissolve into the liquid
lipids where it reacted with the FFA. The by-product water formed by
the reaction evaporated and was removed by the bubbles. Some
methanol reacted to form FAME, while some methanol remained
unreacted and exited the reactor in the bubbles along with water
vapor. The reactor design shown here was much more reproducible
and reliable than previous designs (such as a round-bottomed ask)
tested by the authors.
The reactor contents were circulated from the bottom of the reactor to the top of the reactor through an external sampling loop. Before returning to the top of the reactor, the circulating lipids passed
through heat exchanger tubes immersed in hot silicone oil. This
same heat exchanger was also used to heat and vaporize the methanol feed, which was fed to the heat exchanger as a liquid by syringe
pumps and vaporized in separate tubes in the exchanger before
reaching the reactor. Two syringe pumps were used sequentially
during reaction experiments; these pumps were congured in a
way that allowed one syringe pump to be lled with alcohol while
the other was discharging alcohol into the reactor through the heat
exchanger. Kocsisova et al. [13] explained that methanol should be
fed to the reactor as a liquid; however, feeding liquid methanol to
the reactor caused a drop in reactor temperature due to the latent
heat of vaporization of methanol. For all of the experiments in this
paper, methanol was fed to the reactor as a vapor, which led to
more stable reactor temperatures.
Liquid catalyst, typically sulfuric acid, was mixed with a small volume of alcohol to form a catalyst solution. The quantity of catalyst
added to the solution was 0.1% of the weight the lipid feedstock unless
otherwise specied. This solution was gradually added to the reactor
at the top by a peristaltic pump during the rst 5 min of the reaction.
3. Results and discussion
Several parametric studies were performed to test the bubble column reactor performance for a variety of conditions. The typical

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

experimental conditions for results in this paper were as follows: a reactor temperature of 120 C, ambient pressure, lipid volume of 180 mL of
pure oleic acid, alcohol feed at 0.75 mL of liquid methanol per minute,
and 0.1% (wt. catalyst/wt. lipids) of sulfuric acid catalyst added during
rst 5 min of reaction. Deviations from these conditions are indicated
in the gure captions. Oleic acid was used as the FFA for these experiments as an appropriate surrogate for naturally occurring FFA.
Before conducting the following studies, a study of varying catalyst
concentration was conducted. Concentrations between 2.0% (w/w) and
0.01% (w/w) were evaluated. The authors evaluated 0.1% (w/w) to be
the lowest concentration within the kinetically limited regime and
chose this concentration of sulfuric acid for future experiments. More details of this study can be found in the supplementary documentation.
3.1. Effect of methanol ow rate on the esterication of oleic acid
Methanol was continuously fed to the bubble column reactor by syringe pumps. The effect of methanol feed rate to the reactor on conversion

FFA Content (%)

73

to FAME is shown in Fig. 2. The syringe pumps fed liquid methanol, which
was vaporized before reaching the reactor; so methanol ow rates are
often displayed in liquid milliliters per minute at which the syringe
pumps operated, and alternative expressions for feed rates are shown in
Table 1. Fig. 2(a) shows that as the reaction proceeded, the amount of
FFA in the reactor decreased rapidly at rst and later asymptotically
approached zero. The rate at which FFA decreased, or the slope of the
conversion prole, varied with the rate at which methanol owed into
the reactor with higher initial slopes corresponding to higher methanol
ow rates. At liquid methanol ow rates higher than 1.16 mL/min (for
180 mL of FFA in reactor), conversion of greater than 95% of FFA to
FAME was achieved in less than 60 min. As the ow rate of methanol
decreased, so did the reaction rate, and at ow rates of less than
0.43 mL/min, 95% conversion of FFA to FAME took more than 100 min.
In Fig. 2(a), the conversion proles for the lowest ow rates are
nearly linear because the supply of methanol limited the reaction
rate. In the low ow reactions, the measured decrease in concentration of FFA was close to the moles of methanol fed to the reactor

100

3.5 mL/min
2.57 mL/min
1.75 mL/min
1.16 mL/min
0.75 mL/min
0.43 mL/min
0.20 mL/min

80
60
40
20
0
0

20

40

60

80

100

120

Time (min)

Unreacted Methanol Ratio

8
6.2 min

8.7 min

12.3 min

19.1 min

29.1 min

52.5 min

95% conversion

113.5 min

0
0

20

40

60

80

100

120

Time (minutes)
7

180
160

Time to 95% conversion (left axis)

Excess MeOH at 95% conversion (right axis)

140
120
100

80

60

40
1

20
0

Unreacted Methanol Ratio

Time to 95% conversion of FFA

(minutes)

0
0

0.05

0.1

0.15

0.2

Normalized MeOH Feed Rate (1/min)

Fig. 2. Effect of methanol ow rate on the conversion of FFA to FAME in a bubble column reactor. (a) Conversion versus time. (b) Amount of methanol fed to the reactor that passes through
the reactor without reacting as a function of time with black dashed line indicating the time at which 95% conversion to FAME is achieved. (c) Effect of methanol ow rate on the time to
reach 95% conversion (blue triangles, left axis) and excess methanol at 95% conversion (red square, right axis).

74

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

Table 1
Representations for alcohol ow rates for experiments shown in Fig. 2.
Volumetric ow rate of liquid MeOH (mL/min)

Molar ow rate of MeOH (mol/min)

MeOH ow normalized by FFA0 (1/min)

Time to stoichiometric delivery of MeOH (min)

3.5
2.57
1.75
1.16
0.75
0.43
0.2

0.0859
0.0631
0.0430
0.0285
0.0184
0.0106
0.0049

0.1617
0.1146
0.0816
0.0523
0.0343
0.0190
0.0088

6.2
8.7
12.3
19.1
29.1
52.5
113.5

Unreacted methanol ratio

:
N

 t NFFA t
1
NFFA;0

MeOH

where NMeOH is the molar ow rate of methanol, NFFA(t) is the moles of

FFA in the reactor at time t measured by titration, and NFFA,0 is the initial
number of moles of FFA. Fig. 2(b) plots the data of Fig. 2(a) as unreacted
methanol verses time with methanol ow rate indicated as time to stoichiometric delivery of methanol (last column of Table 1). At low ow
rates, the unreacted methanol was close to zero, which means that
most of the methanol fed to the reactor reacted with FFA to produce
FAME, i.e., the methanol used is close to the stoichiometric ratio of
methanol to FAME. As the methanol ow rate increased, unreacted

methanol increased to over eight times the stoichiometric ratio of methanol required. In all cases, the unreacted methanol was initially lower
and increases approximately linearly at long times. At long times, the reaction rate approached zero, so nearly all of the methanol that entered
the reactor passed through unreacted, leading to a linear increase in
unreacted methanol. The dashed curve of Fig. 2(b) indicates the time
at which each experiment achieves 95% conversion of FFA to FAME;
lower methanol ow rates correspond to both lower fractions of
unreacted methanol and longer time to reach 95% conversion.
Fig. 2(a) shows that high ow rates lead to faster reaction rates, and
Fig. 2(b) shows that high ow rates also lead to a high excess of methanol fed through the reactor. As ow rate decreased, both of these quantities decrease. Consequently, there is an important trade-off in a bubble
column reactor: between short reaction time and high methanol excess.
At high methanol excess, the amount of methanol used and the cost of
equipment required to collect, separate, and recycle excess methanol
increases. At low reaction rates, the rate of FAME production per unit
volume of reactor is small, so larger reactors are required. It is desirable
to nd a balance between low methanol excess and fast reaction rate.
Fig. 2(c) plots an alternative representation of the reaction conversion
data as the time required to reach 95% conversion of FFA to FAME
(t95) and the fractional unreacted methanol at 95% conversion as functions of ow rate. At low ow rates, t95 increases hyperbolically as
ow rate decreases. At high ow rates, t95 asymptotically approaches
a constant value of about 40 min. So for a pure methanol feed at
120 C in this reactor conguration, 40 min is the shortest time to
reach 95% conversion of oleic acid to FAME.
The unreacted methanol ratio at 95% conversion is a function that is
nearly linear with respect to methanol ow rate. While an economic
analysis has not yet been performed for the ow rate of the bubble

100
Pure MeOH
90:10 MeOH:H2O
80:20 MeOH:H2O
Pure EtOH
90:10 EtOH:H2O
80:20 EtOH:H2O

90
80

FFA Content (%)

divided by the volume of lipids in the reactor; consequently, nearly

all the methanol fed to the reactor reacted with FFA to produce
FAME. Because the cumulative amount of methanol fed to the reactor
is linear with time, at low ow rates, the measured FFA content decreases nearly linearly with time due to a limited supply of methanol.
As the ow rate increased, the reaction rate increased until the reaction approached an apparently kinetically limited regime. Faster
methanol ow rates created more vigorous bubbling, which leads
to a reduction in mass transfer resistance and contributed to faster
reaction rates and faster decrease in FFA content as shown in Fig. 2.
For methanol ow rates higher than 2.57 mL/min, the reaction rate
was independent of the ow rate, which indicates that for these reaction conditions, the reaction kinetics had a dominant effect on
the conversion prole.
The curves in Fig. 2(a) are labeled by the ow rates of liquid
methanol used in the experiment in mL/min. Table 1 contains alternative expressions of those ow rates that may be useful for
interpreting the results. The rst column in Table 1 is the liquid
methanol ow rates in milliliters per minute. The second column
shows the ow rate of methanol in moles of methanol per minute.
The third column shows the molar ow rate of methanol that is
scaled by the initial moles of oleic acid in the reactor; this is a normalized ow rate expressed in minutes. This normalized ow rate
is useful in representing how much methanol is available to react
with the oleic acid in the reactor; the normalized ow rate is also
useful for scaling between reactors with different volumes of FFA.
The fourth column is the reciprocal of the third column and is the
time required for the moles of methanol fed to the reactor to be
equal to the initial moles of oleic acid in the reactor. This is the theoretical minimum time for complete conversion if all methanol entering the reactor reacted with oleic acid to produce FAME.
At all rates, some methanol passed through the reactor without
reacting with FFA to form FAME. The ratio of the unreacted methanol
to the initial amount of FFA in the reactor is the unreacted methanol
ratio, and the unreacted methanol was vented to the fume hood, in
which the experiments were conducted. This unreacted methanol
could be collected and recycled, which is especially important for fullscale production processes. The amount of unreacted methanol was calculated from the known methanol ow rate and the measured conversion of FFA to FAME:

70
60
50
40
30
20
10
0
0

20

40

60

80

100

120

Time (minutes)
Fig. 3. Measured FFA content during reactions in bubble column reactor during experiment measuring the effect of varying alcohol quality at constant catalyst concentration
and a constant methanol ow rate of 0.75 mL/min. Methanol (MeOH) and ethanol
(EtOH) were used as the alcohols with varying water content (on a volume basis).

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

column reactor, the far-left and far-right regions are regions where the
reactions have either low conversion rate or high excess of methanol.
The middle region of Fig. 2(c) may be the region with the most desirable
ow rates that balance the trade-off between faster reactions (smaller
reactor at high ow rates) and lower unreacted methanol (less excess
methanol that has to be recovered and recycled).
3.2. Robustness for varying alcohol feed quality on the esterication of
oleic acid
Fig. 3 displays FFA conversion proles for a series of reactions that
were performed with different alcohol feedstocks. Both ethanol
(EtOH) and methanol (MeOH) were used as primary alcohols. Each alcohol was also mixed with water in concentrations of 90:10 and 80:20
alcohol:water by volume as a liquid. In typical lipid to fatty acid alkyl
ester (FAAE) reactions, water hinders the esterication reactions of
FFA because it enhances the reverse reaction. However, in the bubble
column, rising methanol vapor bubbles strip water from the lipids and
draw the water out of the reactor. This allows high conversions to be obtained with alcohol feedstocks containing in excess of 10% water by
volume.
The reaction proles in Fig. 3 show that reaction rates were
lower in experiments with water-diluted alcohols. The decrease
in rate is likely because as alcohol feed rates were maintained at
the same volumetric rate but lower purity, the molar feed rate of
alcohol is reduced. For example, Fig. 3 shows that an increase

from 0% to 10% water in the alcohol feed lengthens the time to

90% conversion from 60 to 80 min for methanol and from 80 to
120 min for ethanol. Also, the conversion proles in Fig. 3 show
that all of the reactions with methanol proceed faster than reactions with ethanol at the same volumetric water concentration.
Surprisingly, the conversion prole for pure ethanol feed was
nearly the same as the conversion for methanol feed with 10%
water; likewise, the prole for ethanol with 10% water overlaps
the prole for methanol for 20% water. Hence, under these conditions, switching from methanol to ethanol had roughly the same
effect on conversion time as adding 10% water by volume to the
methanol feed.
The experiments exploring different types of alcohol feedstock and
alcohol moisture content (displayed in Fig. 3) demonstrated that the
fastest esterication reactions used pure methanol as the alcohol feedstock. However, the bubble column reactor is robust and achieved a
high conversion of FFA to FAAE for varying alcohol feeds and with
lower quality alcohol feedstocks containing water. If the reactor can utilize lower-quality alcohol feedstocks, then the cost of raw materials for
making FAAE can be reduced. Additionally, the cost of recycling the alcohol/water exhaust exiting from the reactor can be reduced because
it is not necessary that the recycled alcohol be completely free of
water, possibly eliminating the need to break any alcohol/water azeotropes. In addition to being robust for impure feeds, these studies
show that larger alcohols can react with fats, greases, and oils for
FAAE production in the bubble column. Using ethanol as the alcohol

c
200
180
160

100/0

140

90/10

120

50/50

100

10/90

80
60
40
20
0
0

15

30

45

60

75

90

FAME Content (% of fatty acid chains)

Acid Number (mg KOH/g sample)

100

100/0
90/10
50/50
10/90

90
80
70
60
50
40
30
20
10
0
0

15

Time (minutes)

30

45

60

75

90

Time (minutes)

d
100

Acyl glyceride fatty acid Content

(% Fatty acid chains)

Acid Number/Initial Acid Number (%)

75

90
80

100/0

70

90/10

60
50

50/50

40

10/90

30
20
10
0
0

15

30

45

60

Time (minutes)

75

90

100

10:90

90

50:50

80

90:10

70
60
50
40
30
20
10
0
0

15

30

45

60

75

90

Time (minutes)

Fig. 4. Performance of bubble column reactor for lipids containing a mixture of FFA/TAG (%v/v) (a) Change in acid number from titration versus time, (b) Percent of initial acid number
versus time, (c) FAME content from NMR versus time, and (d) acyl glyceride content versus time.

76

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

feedstock could have several advantages, including the potential to be

produced from renewable feedstocks and having lower toxicity than
methanol.
3.3. Esterication of mixed FFA/TAG feedstocks
A series of experiments were conducted to examine the performance of the bubble column reactor for creating FAME using feedstocks containing a mixture of FFA and TAG. The FAME, FFA, and
TAG content were analyzed using titration and NMR as described
in Section 2.2. Fig. 4(a) plots changes in acid number versus time
for several experiments with different initial FFA content, which
follows the pattern of an exponential decay, similar to that of
pure FFA experiments. The data showed that after about 70 min,
more than 95% of the initial FFA is converted to FAME. Fig. 4(b)
shows the percent change of acid number normalized to the initial
acid number of the reactions in Fig. 4(a). The curves in Fig. 4(b)
show that conversion proles for the esterication of FFA collapse
to a master curve when normalized to initial acid number.
Fig. 4(c) shows the FAME content of the reactor during mixed feedstock reactions as measured by NMR. The FAME content begins at zero
and increases in all reactions until it reaches a plateau value, which
varies depending on initial FFA content. The FAME content for reactions
with high FFA contents plateaued above 90%, while reactions with equal
parts FFA and TAG plateaued near 60% FAME and reactions with an initial FFA concentration of 10% plateaued at only about 40% FAME. Although the FFA content (acid number from Fig. 4(a)) fell to zero, not
all fatty acids were converted to FAME. These results indicate that the
transesterication reaction (Eq. (1)) is incomplete, and some fatty
acid chains in the reactor remain as acylglycerides: TAG, DAG, and
MAG. This indicates that under the conditions of this reaction with an
acid catalyst, the rate of transesterication is signicantly slower than
esterication. Additional experiments that utilized an alcohol feed diluted with up to 20% water (not shown here) produced FAME formation
rates similar to those seen in Fig. 4(c). This indicates that hydrolysis of
acylglycerides to FFA is also slow under these experimental conditions.

a
Acid number (mg KOH/g oil)

The traditional route to produce FAME is a base-catalyzed

transesterication, which is readily accomplished with lipids low in
FFA and can be carried out at temperatures below 40 C. If the FFA content of lipids can be lowered to less than 1%, those lipids can be used in a
transesterication process [9]. For lipid feedstocks containing more than
1% FFA content, advanced reaction technologies or 2-step methods can
be utilized to convert FFA to FAME. Data from the study of varying FFA
and TAG contents of lipids (Fig. 4) show that the FFA content of any tested oil can be reacted by esterication to FAME using the bubble column
reactor of this paper. This acid catalyzed bubble column reactor can then
be used as the rst step of a two-step reaction process such as one proposed by Van Gerpen. This two-step technology can utilize any renewable lipid resource that contains FFA and TAG in any ratio, and because
the bubble column reactor does not require temperatures or pressures
as high as other esterication methods, it may prove to be a costeffective esterication step in a 2-step conversion to FAME. This robustness of the reactor with respect to lipid feedstocks allows biodiesel
manufacturers employing this technology to greatly diminish their
feedstock costs. Because feedstock costs are a dominant expense in producing biodiesel, it could lower the price of biodiesel to a level that is
more competitive with petroleum diesel.

3.4. Esterication of trap grease lipids

Trap grease is a waste by-product of the food service industry that
contains water, lipids, detergents, food particles and other waste. Trap
grease is collected in grease interceptors, which are storage tanks ranging from a few gallons to several thousand gallons. Grease interceptors
remove grease and sediments from kitchen efuent that could otherwise enter the sewage systems and cause blockages. Grease interceptors
are required by law to be emptied at specic intervals, usually by commercial grease handlers, who are paid to pump and remove trap grease
and then must pay to process and dispose of the grease properly. If the
lipids could be separated from trap grease and converted to FAME, then
the feedstock cost of the resulting biodiesel process can be a fraction of

200
180

Trap Grease 1% PTSA

0.75 mL/min
Oleic acid 1% PTSA,
0.75 mL/min

160
140
120
100
80
60
40
20
0
0

20

40

60

80

100

120

Time (minutes)
Fig. 5. Esterication of FFA from trap grease in a bubble column reactor. (ad) Photos of various stages of trap grease processing: (a) raw trap grease; (b) layers after heating and settling
trap grease from top to bottom: lipid, oating solids, water, and sediment; (c) trap grease lipids (FOG) in solid state at room temperature; (d) crude trap grease FAME in liquid state at room
temperature. (e) Changes in acid number versus time of lipid layer of trap grease in bubble column reactor with oleic acid reaction results for comparison. In these experiments, the acid
catalyst was p-toluenesulfonic acid (PTSA).

C.J. Stacy et al. / Fuel Processing Technology 124 (2014) 7077

that for rened soybean lipids and other conventional sources of TAG
[2,4].
Trap grease samples (Fig. 5(a)) from Russell Reid, a commercial
waste management company, were heated and gently stirred in the
lab. The heating allowed the lipid, water, and sediment layers to separate (Fig. 5(b)) so that the lipid layer could be collected and ltered
(Fig. 5(c)). Once ltered, the trap grease lipids were used as a lipid feedstock in the bubble column reactor and converted into FAME (Fig. 5(d))
using sulfuric acid and occasionally para-toluenesulfonic acid (PTSA) as
catalysts; the latter was used in the reactor in the same manner as the
sulfuric acid catalyst. The FAME produced had a visibly lower viscosity
than the lipid layer of the trap grease. Also, although the trap grease
lipids solidied at room temperature, the crude FAME produced from
those lipids remained a liquid at temperatures below 10 C.
Fig. 5(e) displays the conversion versus time for FFA from trap
grease with the oleic acid conversion data shown for comparison. The
reaction of trap grease lipids to FAME was roughly 50% slower than
pure oleic acid with approximately 95% conversion achieved in about
110 min. Additionally, it was observed during experiments that the typical 0.1% (w/w) acid catalyst was not sufcient for the esterication of
trap grease lipids; instead, 1% (w/w) catalyst was used for the experiments in Fig. 5(e). The need for higher catalyst concentration could indicate that an unknown contaminant in the trap grease is consuming
the acid catalyst before a reaction proceeds.

4. Conclusions
A bubble column reactor that converts free fatty acids to fatty acid
alkyl esters (for use as biodiesel) was tested and developed. The bubble
column reactor achieves a conversion of over 95% of free fatty acid to
fatty acid methyl esters in less than 2 h for a variety of reaction conditions.
For example, with a sulfuric acid catalyst concentration of 0.1 wt%, a reaction temperature of about 120 C and ambient pressure, a methanol to
fatty acid molar ratio of less than 3:1, and bubbling alcohol vapor through
the reactor at a ow rate of 0.034 moles of alcohol per mole of FFA per
minute into liquid, the time to 95% conversion is about 70 min. At higher
methanol feed ow rates, the time to 95% conversion decreases to about
40 min but requires a methanol to fatty acid molar ratio of ve or higher.
The reactor is effective with low-quality alcohol feedstocks containing
water and is robust for different types of alcohol feed. Because of these attributes, the reactor is potentially useful for the conversion of low-value
alcohols and low-value lipids into biodiesel. As a result of potentially lowering the feedstock costs, the cost of producing biodiesel from renewable
resources could be lowered substantially.

Acknowledgments
The authors would like to thank Brittany Preston for her work with
NMR analysis, Minhazuddin Mohammed for his research on biofuels
and a model for the bubble column reactor, Drexel University's Chemistry Department for access and use their 500 MHz NMR, Jacklyn Briguglio
for her work with ltering of trap grease, Megan Hums for consultation
on process impacts, and Gary Weiner Russell Reid for supplying trap
grease samples. Work was funded by EPA P3 Design Award (SU83352401).

77

Appendix A. Supplementary data

The supplemental documentation submitted with this paper includes NMR spectra used to determine FAME content in mixed feedstock studies and conversion data documenting the effect of catalyst
concentration on conversion. Supplementary data to this article can be
found online at http://dx.doi.org/10.1016/j.fuproc.2014.02.003.
References
[1] J. Van Gerpen, Biodiesel processing and production, Fuel Processing Technology 86
(2005) 10971107.
[2] K.S. Tyson, J. Bozell, R. Wallace, E. Petersen, L. Moens, Biomass Oil Analysis: Research
Needs and Recommendations, NREL, Golden Colorado, 2004.
[3] A.A. Apostolakou, I.K. Kookos, C. Marazioti, K.C. Angelopoulos, Techno-economic
analysis of a biodiesel production process from vegetable oils, Fuel Processing
Technology 90 (2009) 10231031.
[4] M. Haas, Improving the economics of biodiesel production through the use of low
value lipids as feedstocks: vegetable oil soapstock, Fuel Processing Technology 86
(2005) 10871096.
[5] J.M. Marchetti, V.U. Miguel, A.F. Errazu, Techno-economic study of different alternatives for biodiesel production, Fuel Processing Technology 89 (2008) 740748.
[6] Y. Zhang, Biodiesel production from waste cooking oil: 1. Process design and
technological assessment, Bioresource Technology 89 (2003) 116.
[7] Y. Zhang, M.A. Dub, D.D. McLean, M. Kates, Biodiesel production from waste
cooking oil: 2. Economic assessment and sensitivity analysis, Bioresource Technology 90 (2003) 229240.
[8] D. Samios, F. Pedrotti, A. Nicolau, Q.B. Reiznautt, D.D. Martini, F.M. Dalcin, A
transesterication double step processTDSP for biodiesel preparation from fatty
acids triglycerides, Fuel Processing Technology 90 (2009) 599605.
[9] S.V. Ghadge, H. Raheman, Biodiesel production from mahua (Madhuca indica) oil
having high free fatty acids, Biomass and Bioenergy 28 (2005) 601605.
[10] A.N.A. Aryee, F.R. van de Voort, B.K. Simpson, FTIR determination of free fatty acids
in sh oils intended for biodiesel production, Process Biochemistry 44 (2009)
401405.
[11] M.G. Varanda, G. Pinto, F. Martins, Life cycle analysis of biodiesel production, Fuel
Processing Technology 92 (2011) 10871094.
[12] M. Berrios, J. Siles, M. Martin, A. Martin, A kinetic study of the esterication of free
fatty acids (FFA) in sunower oil, Fuel 86 (2007) 23832388.
[13] T. Kocsisova, J. Cvengro, J. Lutisan, High-temperature esterication of fatty acids
with methanol at ambient pressure, European Journal of Lipid Science and Technology 107 (2005) 8792.
[14] I.L. Lucena, R.M.A. Saboya, J.F.G. Oliveira, M.L. Rodrigues, A.E.B. Torres, C.L.
Cavalcante, E.J.S. Parente, G.F. Silva, F.A.N. Fernandes, Oleic acid esterication with
ethanol under continuous water removal conditions, Fuel 90 (2011) 902904.
[15] E. Santacesaria, R. Tesser, M. Di Serio, M. Guida, D. Gaetano, A. Garcia Agreda, F.
Cammarato, Comparison of different reactor congurations for the reduction of
free acidity in raw materials for biodiesel production, Industrial & Engineering
Chemistry Research 46 (2007) 83558362.
[16] F.A.P. Voll, C. da Silva, C.C.R.S. Rossi, R. Guirardello, F. de Castilhos, J.V. Oliveira, L.
Cardozo-Filho, Thermodynamic analysis of fatty acid esterication for fatty acid
alkyl esters production, Biomass and Bioenergy 35 (2011) 781788.
[17] Z.-M. Wang, J.-S. Lee, J.-Y. Park, C.-Z. Wu, Z.-H. Yuan, Optimization of biodiesel production from trap grease via acid catalysis, Korean Journal of Chemical Engineering
25 (2008) 670674.
[18] B.D. Wahlen, B.M. Barney, L.C. Seefeldt, Synthesis of biodiesel from mixed feedstocks
and longer chain alcohols using an acid-catalyzed method, Energy & Fuels 22 (2008)
42234228.
[19] USDA, National Weekly Ag Energy Round-Up, USDA Livestock, Poultry & Grain
Market News, United States Department of Agriculture, Des Moines, Iowa, 2014.
[20] DELCORA, Resolution NO. 2013-21, Hauled Waste Rates, Delaware County Regional
Water Control Authority, Chester, PA, 2013.
[21] M. Canakci, The potential of restaurant waste lipids as biodiesel feedstocks,
Bioresource Technology 98 (2007) 183190.
[22] K.S. Tyson, DOE analysis of fuels and coproducts from lipids, Fuel Processing
Technology 86 (2005) 11271136.
[23] W.W. Berry, B.J. Ratigan, Process of making alkyl esters of free fatty acids, Philadelphia
Fry-o-Diesel Inc., US, 2010.
[24] ASTM, Standard Test Method for Acid Number of Petroleum Products by Potentiometric Titration, in, ASTM International, West Conshohocken, PA.
[25] AOCS, Acid Value, in: AOCS Methods for Biodiesel Feedstock Quality, American Oil
Chemists Society, Urbana, IL.