The General Applicability of in Situ Transesterification for the Production of Fatty Acid Esters from a Variety of Feedstocks.

Abstract

We previously described a method for fatty acid methyl ester (FAME) production wherein acylglycerol transesterification was achieved by reacting flaked full fat soybeans with alkaline methanol to create a product that met ASTM specifications for biodiesel. In the present work we explore the general applicability of this approach, termed in situ transesterification, to feedstocks other than soybeans. Materials investigated were distillers dried grains with solubles (DDGS), which is a co-product of the production of ethanol from corn, and meat and bone meal (MBM), a product of animal rendering. For both feedstocks, reaction conditions giving maximum lipid transesterification were predicted by statistical experimental design and response surface regression analysis, and then verified experimentally. Successful transesterification was achieved at ambient pressure and 35 °C. For DDGS, partial drying markedly reduced the methanol requirement to achieve a high degree (91.1% of maximum theoretical) of transesterification. Elevated reaction temperatures (to 55 °C was explored) caused little or no shortening of the time to completion. Protein was not removed from the DDGS during this treatment. For MBM, drying was not required to achieve a high degree (93.3%) of transesterification. The remaining meal retained approximately 90% of the protein originally present. Coupled with the previous work with soybeans, the data presented here indicate that in situ transesterification is generally applicable to lipid-bearing materials, which could substantially increase the supply of biodiesel.

Full text

Page 1

ORIGINAL PAPER
The General Applicability of in Situ Transesterification
for the Production of Fatty Acid Esters from a Variety
of Feedstocks
Michael J. Haas Æ Æ Karen M. Scott Æ Æ
Thomas A. Foglia Æ Æ William N. Marmer
Received: 29 January 2007/Revised: 11 July 2007/Accepted: 24 July 2007/Published online: 23 August 2007
? AOCS 2007
Abstract
acid methyl ester (FAME) production wherein acylglycerol
transesterification was achieved by reacting flaked full fat
soybeans with alkaline methanol to create a product that
met ASTM specifications for biodiesel. In the present work
we explore the general applicability of this approach,
termed in situ transesterification, to feedstocks other than
soybeans. Materials investigated were distillers dried
grains with solubles (DDGS), which is a co-product of the
production of ethanol from corn, and meat and bone meal
(MBM), a product of animal rendering. For both feed-
stocks,reactionconditions
transesterification were predicted by statistical experi-
mental design and response surface regression analysis,
and then verified experimentally. Successful transesterifi-
cation was achieved at ambient pressure and 35 ?C. For
DDGS, partial drying markedly reduced the methanol
requirement to achieve a high degree (91.1% of maximum
theoretical) of transesterification. Elevated reaction tem-
peratures (to 55 ?C was explored) caused little or no
shortening of the time to completion. Protein was not
removed from the DDGS during this treatment. For MBM,
drying was not required to achieve a high degree (93.3%)
oftransesterification.The
approximately 90% of the protein originally present.
Coupled with the previous work with soybeans, the data
We previously described a method for fatty
givingmaximumlipid
remainingmealretained
presented here indicate that in situ transesterification is
generally applicable to lipid-bearing materials, which
could substantially increase the supply of biodiesel.
Keywords
Fats and oils utilization ? Fatty acid ester ?
In situ transesterification ? Meat and bone meal ?
Transesterification
Biodiesel ? Distillers dried grains ?
Introduction
The production and use of fatty acid esters, particularly
methyl esters (FAME), as a fuel for compression ignition
(diesel) engines is a rapidly growing technology that has
penetrated the transportation fuels sector in much of Eur-
ope and is in the process of doing so in the United States
and other countries. Compared to petroleum-derived diesel
fuel, the use of biodiesel offers advantages of reduced
engine emissions of pollutants, domestic sourcing, and
renewability.
It is estimated that the United States produced and
consumed approximately 30 million gal of biodiesel in
2004 [1]. European consumption exceeded this value by at
least tenfold. Biodiesel production is growing rapidly with
2006 US production estimated to be 245 million gal and
expected to exceed 700 million gal in 2015 [2]. A majority
of biodiesel is produced from edible vegetable oils, soy-
bean oil in the United States, and presently accounts for
approximately 10% of annual US soybean oil production if
all was made from that feedstock. There is concern that at
anticipated future production levels the use of edible oils
for fuel will compete significantly with food uses. This
would result in undesirable increases in both food and
biodiesel costs, a particularly damaging occurrence in the
Mention of brand or firm names does not constitute an endorsement
by the US Department of Agriculture over others of a similar nature
not mentioned.
M. J. Haas (&) ? K. M. Scott ? T. A. Foglia ? W. N. Marmer
US Department of Agriculture, Agricultural Research Service,
Eastern Regional Research Center, 600 East Mermaid Lane,
Wyndmoor, PA 19038, USA
e-mail: michael.haas@ars.usda.gov
123
J Am Oil Chem Soc (2007) 84:963–970
DOI 10.1007/s11746-007-1119-4

Page 2

case of biodiesel, which even at current prices, has diffi-
culty competing with petroleum fuel on an economic basis.
Feedstock expense is the major contributor to the cost of
biodiesel production [3].
Alkali-catalyzed transesterification of fats and oils in the
presence of a simple monohydroxy alcohol, usually
methanol, is the predominant technology used for FAME
synthesis. For vegetable oils this requires the prior recovery
of the lipid by pressing or solvent extraction, and the partial
purification of this oil. We have previously described an
alternate method of FAME synthesis from soybean oil,
termed in situ transesterification, which does not require
the extraction and purification of the oil prior to FAME
synthesis. Rather, thinly flaked soybeans are themselves the
feedstock for this process, and alkali-catalyzed transeste-
rification takes place directly within the flakes during
incubation in alkaline alcohol [4]. This process operates at
high efficiency, and under optimum reaction conditions
achieves maximum theoretical conversion of the acylgly-
cerols of the feedstock into FAME. Furthermore, the
resulting ester preparation meets the ASTM specifications
for biodiesel [5].
The desirability of ensuring future supplies of biodiesel
feedstocks, while at the same time minimizing economic
pressures on high quality lipids, has led to the develop-
ment of methods to use low quality lipids, such as waste
greases and soapstock for biodiesel production. Other low-
value lipid sources exist, but have received little or no
attention in this regard. One of these is Distillers Dried
Grains with Solubles (DDGS), a co-product of the fer-
mentation of corn to produce ethanol via the dry grind
process. Another is Meat and Bone Meal (MBM), a
product of the animal rendering industry that is composed
of the inedible portions of animals slaughtered in meat
production. Both are produced in substantial quantities,
have low economic value, and contain 5–10% lipid, which
could be converted to FAME. These materials are not
suitable for the recovery of their lipid by hexane extraction
or yield preparations of poor quality. In the present work
we report the results of an investigation of the ability of
in situ transesterification to convert the lipids in DDGS
and MBM to FAME.
Experimental Procedures
Chemicals
DDGS was the product of an industrial corn-to-ethanol
fermentation plant, and was provided by Dr. Vijay Singh,
Department of Agricultural and Biological Engineering,
University of Illinois, Urbana, IL, USA. It consisted of
solid particles with a size range such that 88 wt% passed
through a 20-mesh screen and 24% passed through a 50-
mesh screen. Oven drying (77 ?C) the DDGS to constant
weight indicated 8.7 wt% moisture. Extraction with hexane
for 4.5 h in a Soxhlet apparatus followed by HPLC analysis
(below) of the resulting oil indicated a 8.8 wt% triacyl-
glycerol (TAG) content.
MBM, obtained from a commercial rendering facility,
was provided by Dr. Rafael Garcia of our facility. It con-
sisted of particles with a size range such that 74% of the
mass passed through a 20-mesh screen, and only 2% passed
through a 50-mesh screen. The moisture content, deter-
mined by oven drying as described above, was 2.0 wt%.
Soxhlet extraction with hexane followed by HPLC of the
resulting oil indicated 9.1 wt% triacylglycerol content.
DDGS and MBM were stored at 4 ?C.
As required, feedstocks were incubated in a convection
oven (77 ?C) until the desired moisture levels were
attained.
Soygold brand biodiesel, produced from refined soybean
oil, and used as a chromatography standard was the product
of Ag Environmental Products, Lenexa, KS, USA. Soybean
oil TAG and FFA for use as chromatography standards
were obtained from Sigma-Aldrich. Palmitic, stearic, oleic,
linoleic, and linolenic acids, mixed in amounts proportional
to their mass abundance in soybean oil [6], served as the
FFA standard. A mixture of FAME, whose composition
reflected the fatty acid content of soy oil (RM-1), was the
product of Matreya, Inc. (Pleasant Gap, PA, USA). Organic
solvents were B&J Brand
Jackson, Inc., Muskegon, MI, USA). Other chemicals were
Analytical Reagent grade quality or better.
TMHigh Purity Grade (Burdick &
Conduct and Optimization of in situ Transesterification
The lipid-bearing material (5.00 g for MBM, and an
amount equal to 5.00 g of dry matter in the case of DDGS,
for which multiple moisture levels of substrate were
examined) was mixed with methanol in which sodium
hydroxide had been previously dissolved (‘‘alkaline alco-
hol’’) in screw-capped bottles of a capacity at least five
times the reaction volume. These were mixed by orbital
shaking at a speed sufficient to keep the solids well sus-
pended. Reactions were conducted in heated chambers at
the desired reaction temperatures.
Central Composite Response Surface design methods
[7] were employed to investigate coordinately the effects
and interactions of the amount of alkaline methanol, its
NaOH concentration, and reaction time on the yields of
FAME, FFA and unreacted TAGs in the liquid phase.
Preliminary studies (data not shown) were conducted to
focus the statistically designed work in the region of var-
iable space giving greatest FAME production.
964J Am Oil Chem Soc (2007) 84:963–970
123

Page 3

For the final optimization of the reaction conditions for
5-g reactions of DDGS, the amounts of alkaline methanol
tested were 12, 14, 18, 22, and 24 mL; NaOH concentra-
tions were 0.20, 0.25, 0.35, 0.45, and 0.50 N; and reaction
times were 0.5, 1.5, 2.5, 3.5, and 4.5 h. This series of
variables was tested at DDGS moisture levels of 2.62, 6.86,
and 8.7 wt%. Each series involved 20 reactions at various
combinations of these levels.
For optimization of the reaction conditions for transe-
sterification of 5 g of MBM, the amounts of alkaline
methanol tested were 5.0, 6.67, 10.0, 13.3, and 15.0 mL,
NaOH concentrations were 0.1, 0.133, 0.20, 0.27, and
0.30 N; and reaction times were 0.5, 1.0, 2.0, 3.0, and
3.5 h. Each series involved 20 reactions.
Following reaction, bottles were allowed to sit for
15 min at room temperature to allow the solids to settle.
The liquid phase was removed, the solids were washed
twice by resuspending in 10 mL methanol, the washes
were pooled with the reaction liquid, and the mixture was
analyzed by HPLC to determine the contents of FAME,
TAG, diacylglycerols (DAG), monoacylglycerols (MAG)
and free fatty acids (FFA). (Preliminary experiments
demonstrated that two washes were sufficient to remove all
FAME from the solids.)
For DDGS and MBM, best-fit equations correlating the
amounts of FAME, FFA and TAG produced with the
reactant compositions were constructed using SAS/STAT
software [8]. Numerical analysis of these equations and
examination of the corresponding three-dimensional (3D)
surfaces allowed identification of the conditions predicted
to give maximum FAME yield with minimum contami-
nation by FFA and TAG.
To conduct time-course studies, multiple replicate reac-
tions were conducted, each consisting of 5 g substrate and
the amounts of alcohol and NaOH predicted to give maxi-
mum reaction. At selected times, a pair of tubes was
processed as described in ‘‘Analytical Methods’’, to deter-
mine the FAME content. The results of duplicate
determinations are presented;individualvalues differedfrom
the average by between 2 and 15% in these determinations.
Analytical Methods
Acylglycerols (AG), FAME, and FFA in the products of
in situ transesterification were determined by HPLC on a
diol-modified silica column (LiChrosorb 5 Diol, Varian,
Walnut Creek, CA, USA), which was developed isocrati-
callywitha mixtureof
(99.9:0.1 vol%) at a flow rate of 0.5 mL/min. Peaks were
detected by evaporative light scattering and quantitated by
reference to standard curves constructed with known pure
compounds.
hexaneand aceticacid
Identification of the fatty acids in the FAME product
was achieved by GC, with flame ionization detection. The
column was a Hewlett–Packard Innowax fused-silica cap-
illary, 30 · 0.53 mm (internal diameter) run as previously
described [9]. FAME assignments were made by compar-
ison with known FAME.
Nitrogen contents (wt%) were determined using a
LECO FP2000 Nitrogen Analyzer (LECO Corpn., St
Joseph, MI, USA). To estimate protein content, nitrogen
values were multiplied by 6.25 for DDGS, the value
applicable to corn [10], and 6.45 for MBM (R. Garcia,
personal communication).
Results and Discussion
In Situ Transesterification of DDGS
Reactions were conducted at 35 ?C since reaction rates
were unacceptably low at lower temperatures. Higher
temperatures were not explored in detail since complete
transesterification occurred in less than two hours at 35 ?C.
Previous work with flaked soybeans demonstrated that
in situ transesterification occurred, and that the methanol
required for the reaction was greatly reduced, when
moisture was completely removed from the flakes prior to
reaction [5]. In contrast, with fully dry DDGS the degree of
transesterification never exceeded 25% of the theoretical
maximum. In an effort to achieve higher FAME yields, the
effect of substrate moisture level on the reaction was
investigated.
Table 1 lists the reaction conditions, predicted by Cen-
tral Composite Design experiments conducted at each
selected feedstock moisture level, for maximum transe-
sterification of the endogenous acylglycerols in DDGS into
FAME. These experiments also identified conditions that
would yield only partial conversion of the AG in DDGS
into FAME. The latter data are not discussed here. How-
ever, if one wished to convert only a portion of the lipids in
Table 1 The effect of moisture content on the reaction conditions
predicted to yield optimum FAME production during the in situ
transesterification of 5 g (dry weight) of DDGS at 35 ?C
Moisture
(wt%)
Methanol
(mL)
NaOH
(N)
Time
(h)
FAME yield
(% max. theoretical)
2.614.00.40 1.2100
2.618.00.30 0.8398.7
6.9 22.00.351.599.5
6.924.0 0.201.8 100
8.718.00.454.299.5
8.7 20.00.250.599.7
J Am Oil Chem Soc (2007) 84:963–970 965
123

Page 4

a feedstock to FAME, leaving some of them unreacted,
e.g., to enhance the nutritional energy value of the resid-
uum for use as an animal feed, it is clear that conditions to
achieve this could be identified.
We sought reaction conditions giving optimum FAME
production using substrate with as near its natural moisture
level as was possible. This would reduce feedstock drying
costs. However, in the case of full moisture (8.7%) DDGS
a relatively large amount of methanol, 20 mL, was required
to achieve near-quantitative FAME yield from 5 g of
DDGS (Table 1). This is undesirable, due to the high cost
that would be subsequently expended in recovering unre-
acted methanol by distillation. Increasing the NaOH
concentration in the methanol phase from 0.25 to 0.4 N
slightly reduced the methanol requirement, to 18 mL
(Table 1). However, the cost of the additional alkali would
probably negate any cost reduction due to this small
decrease in methanol use. In addition, with the latter con-
ditions the reaction time to achieve high levels of
transesterification increased approximately eightfold, to
4.2 h.
The removal of 20% of the moisture from DDGS (to
6.9%) did not reduce the methanol requirements of the
reaction (Table 1). In fact, at this moisture level optimal
transesterification required at least 22 mL of methanol per
5 g of substrate.
With more complete drying, however, a reduction in
methanol requirement for high level transesterification was
observed (Table 1). With DDGS at 2.6% moisture, which
corresponds to removal of 70% of the substrate moisture,
the methanol requirement was reduced to 18 mL of 0.3 N
NaOH per 5 g DDGS at a reaction time of 0.83 h (50 min).
By increasing the NaOH concentration to 0.4 N the amount
of methanol required for high degrees of transesterification
was further reduced to 14 mL (Table 1) with 70 min
reaction time.
Thus, high degrees of transesterification with minimal
methanol usage were predicted to occur at moisture levels
approximately 30% of that typically found in DDGS. At
that moisture level (2.62 wt%) the relationship between
FAME yield in 35 ?C reactions containing a 5-g dry wt
equivalent of DDGS and other reaction variables, as pre-
dicted from the Central Composite Design data, is
described by Eq. 1:
FAME ¼ 1110 ? 2:33V þ 1180N ? 0:974T
? 63:6VN ? 0:0158VT þ 1:19NT
þ 0:963V2þ 131N2þ 0:00158T2
where FAME is the predicted FAME yield (mg), V meth-
anol volume (mL), N normality of NaOH, and T reaction
time (h).
ð1Þ
The R2value for the fit of this equation to the experi-
mental data was 0.84. This indicates that the reliability of
the model is reasonably high. We have typically observed
R2values in excess of 0.92 for similar equations describing
the in situ transesterification of flaked soybeans [4]. With
DDGS, however, we have repeatedly observed poorer fit
between predicted and observed degrees of esterification.
Duplicate reactions often show greater variability in results
than we have experienced with other feedstocks. For
DDGS, the standard deviation of the mean of six replicate
reactions conducted at the center-point conditions of the
statistically designed experiment was 16% of the value of
the mean. The cause of this variation is not yet known.
Figure 1 presents the 3D surface of FAME yields pre-
dicted by Eq. 1 at 1.2 h (70 min) of reaction as a function
of the amount of methanol and the concentration of NaOH
in that methanol. Maximum theoretical FAME yield in this
system is 397 mg. The surface in Fig. 1 generally exceeds
this value. This suggested that shorter reaction times might
achieve full transesterification. We found that this was not
the case for 30 min reactions. We did not explore reaction
times between 30 and 70 min, since there would be little
value in exercising such control over reaction length on an
industrial scale. The surface also suggested high FAME
production with volumes of methanol below 12 mL.
However, this area was outside the region explored
experimentally, and thus Eq. 1 lacked predictive reliability
at low volumes. Experiments subsequently conducted
using small methanol volumes (6–10 mL) gave FAME
yields below 35% of theoretical maximum, probably
Fig. 1 Predicted FAME production (from Eq. 1) at 70 min of
reaction time as a function of methanol volume and NaOH
concentration during the in situ transesterification of 5 g of DDGS,
2.62 wt% moisture, at 35 ?C. Maximum theoretical FAME produc-
tion: 397 mg
966J Am Oil Chem Soc (2007) 84:963–970
123

Page 5

because the DDGS substrate was not thoroughly wetted by
the liquid phase.
To validate the predictions of Eq. 1 and Fig. 1, in situ
transesterification reactions were conducted under the
optimal conditions for the transesterification of 2.62 wt%
moisture DDGS with 14 and 18 mL of methanol. FAME
production was 91.1 ± 1.7% and 90.1 ± 0.1% of maximum
theoretical for these conditions, respectively. This is
acceptably close to the full transesterification predicted by
the experimental model. Reactions of slightly longer
duration have consistently yielded theoretical maximum
transesterification. Under optimum conditions for FAME
production, the production of FFA and AG was negligible,
with these species constituting only 0.3–0.4 wt% of the
lipid-based mass in the product.
Since preliminary experiments indicated that a high
degree of transesterification was achieved in a relatively
short time at a low temperature (35 ?C), the effects of
higher temperatures on the reaction were not explored in
the statistically designed experimentation. It was reasoned
that if this method were applied in industry, the lowest
temperature giving an acceptable degree of completion in a
relatively short reaction would be used to save energy. To
briefly explore the effect of temperature, replicate time-
course reactions were run at 35, 45, and 55 ?C. The results
(Fig. 2) suggest that an elevation in reaction temperature
caused, at best, only a modest acceleration of reaction rate.
The fatty acid composition of the FAME fraction pro-
duced from DDGS by in situ transesterification at optimal
conditions (14 mL methanol/0.4 N NaOH) was compara-
ble to that typically seen for corn oil (Table 2). Thus, as
reported for soybeans [5], in situ transesterification does
not exhibit fatty acid selectivity in the transesterification
event.
As received, the DDGS contained 28.0 ± 0.2 wt% pro-
tein. Following in situ transesterification under optimal
reaction conditions (14 mL liquid phase reaction) the
protein content was 32.4 + 0.2 wt%. An increase in protein
content would be expected if little or no protein was sol-
ubilized during transesterification, since lipid removal by
the reaction increases the relative proportion of protein.
Given an AG content of 8.8% in DDGS, and a transeste-
rification efficiency of 91%, one calculates that the loss of
lipid should increase the protein content to 30.4% if no
protein were lost. The actual measured protein content was
greater than this following transesterification suggesting
that all or nearly all of the protein was retained while a
small amount (approximately 6% of meal mass) of non-
proteinaceous, non-lipid material was removed from the
meal during transesterification.
Conditions for the optimal transesterification of dry
flaked soybeans involved amounts of methanol comparable
to those identified here optimal for DDGS, but alkali
concentrations of only 0.1 N NaOH, not 0.3–0.4 N as for
DDGS [5]. The pH of a mixture of 5 g of DDGS in 15 mL
water was 4.2, whereas the pH of a comparable mixture of
flaked soybeans was 6.4. Some, or all, of the greater alkali
requirement for effective transesterification of DDGS was
thus probably due to the inherent acidity of this substrate.
Modifications in the technology of preparing DDGS that
reduced its acidity could lower the alkali requirement for
effective in situ transesterification. Of course, an analysis
Time (min)
) e
v i t a l e
r
%
(
d l e i
Y
r
e t s
E
0
20
40
60
80
100
120
0 20
40
6080
100
120 140
160 180
Fig. 2 Time course of FAME production over time as a function of
reaction temperature in reactions containing 5 g of DDGS (2.3 wt%
moisture), and 14 mL of methanol containing 0.4 N NaOH. Reaction
temperatures: filled circles: 35 ?C, filled squares: 45 ?C, filled
triangles: 55 ?C. Data presented are the averages of duplicate
determinations. Most individual determinations differed from the
averages by 10% or less
Table 2 Fatty acid composition (wt%), determined by GC, of methyl esters produced by in situ transesterification of DDGS and MBM
Source material16:018:0 18:118:2 18:3
DDGS 12.91.6 28.5 55.51.4
Corn oil [14]9.2–11.8 1.1–1.719.5–30.453.0–65.31.2–2.1
MBM 25.219.735.61.9 0.3
Beef tallow [15]20.9–28.9 7.0–26.5 30.4–48.0 0.6–1.80.3–0.7
Typical fatty acid compositions of the triacylglycerols of corn oil and beef tallow are provided for reference
J Am Oil Chem Soc (2007) 84:963–970 967
123