Volatile Compounds Produced During Deodorization

of Soybean Oil and Their Flavor Significance

Jian-Wen Kao, Earl G. Hammond, and Pamela J. White*
Department of Food Science and Human Nutrition, Iowa State University, Ames, Iowa 50011

ABSTRACT: Freshly deodorized soybean oil has a character- perature treatments; therefore, little work has been done to
istic nutty flavor but often yields no detectable headspace examine the flavor compounds in freshly deodorized veg-
volatiles. The cause of this flavor was investigated by deodoriz- etable oil.
ing soybean oil in an apparatus with a double cold trap that al- In this study, we wished to see if we could account for the
lowed the volatile compounds formed from the initial decom- flavor of freshly deodorized oil by examining the volatile fla-
position of hydroperoxides to be collected separately from those
vor compounds generated during deodorization and by ex-
produced during the normal deodorization process. The chief
volatile components from the normal deodorization process
ploring the possible role of nonvolatile flavor compounds.
were hydrocarbons, which contributed little to no odor to the
oil. The compounds with the greatest odor were carbonyls, es- EXPERIMENTAL PROCEDURES
pecially heptanal and cis-4-heptenal. Although these compo-
nents should accumulate at some steady-state concentration in Collection of volatiles. Newly purchased Wesson brand soy-
an oil during its deodorization, none seemed to account for the bean oil (800 g) was deodorized with steam at 235°C under
flavor of the deodorized oil. By using a particle detector, it was vacuum (0.005 mm Hg) as described by Stone and Hammond
shown that small particles could be generated in the human (2) and modified by Moulton (3). The volatiles that formed
mouth that could provide a mechanism to bring oil with non- were collected in one cold trap (Fig. 1) for 2 h, and then the
volatile flavor components into contact with the olfactory apparatus was switched to a second cold trap for 4 h.
organ. Attempts to separate possible nonvolatile flavors in de-
Separation of volatiles from trap 2 into carbonyls and hy-
odorized oil from triacylglycerides by chromatography on alu-
mina or reaction with 2,4-dinitrophenylhydrazine were unsuc-
drocarbons. The volatiles in trap 2 were extracted three times
cessful. Possibly, the flavor is caused by the glycerol esters with 10 mL pentane, and the pentane extract was concentrated
themselves. to 1.5 mL. To retain the low-boiling volatile compounds, the
JAOCS 75, 1103–1107 (1998). pentane extract was placed in a 10-mL tube sitting in a water
bath, which was set at 2°C above the boiling point of pentane.
KEY WORDS: Carbonyl, deodorization, flavor, GC–MS iden- The solvent was slowly evaporated while keeping the level of
tification, hydrocarbon, sensory evaluation, soybean oil, pentane higher than the surrounding water. Florisil (magne-
volatile compounds. sium silicate, 60–100 mesh; Sigma Chemical Company, St.
Louis, MO) was activated at 250°C overnight, and 10% of its
weight of distilled water was added and allowed to equilibrate
Freshly deodorized soybean oils have a nutty flavor that is for 1 d. Next, 700 µL of the pentane concentrate was applied
considered to be acceptable by consumers. Sensory panel
members, looking for flavors caused by oxidation in soybean
oils, learn to discount this nutty flavor, but when the flavor in-
tensity of freshly deodorized oils is compared with a standard,
such as an odorless mineral oil, the freshly deodorized oil re-
ceives a flavor intensity that cannot be accounted for by rec-
ognized flavor volatiles that arise from oxidation (1). Nonfat
or reduced-fat foods often lack desirable flavors associated
with fresh vegetable oils, such as soybean oils, and this ob-
servation may be partly related to the nutty flavor. It is likely
that volatile compounds do not exist in large amounts in
freshly deodorized oil because of the high-vacuum and tem-
*To whom correspondence should be addressed at Department of Food Sci-
ence and Human Nutrition, 2312 Food Sciences Building, Iowa State Uni-
versity, Ames, IA 50011. E-mail: pjwhite@[Link] FIG. 1. Diagram of deodorization apparatus.

Copyright © 1998 by AOCS Press 1103 JAOCS, Vol. 75, no. 9 (1998)
1104 J.-W. KAO ET AL.

to a 10-mm i.d. × 26-cm column filled with 10 g Florisil in an acetone dry-ice bath, and ozone was passed through it at
pentane/ether (99:1). Hydrocarbons were eluted with 15 mL 20 mL/min. The presence of an excess of ozone was detected
pentane/ether (99:1), and the carbonyls were eluted with 15 by passing the gas leaving the reaction vessel into a solution
mL ether (4). of 5% potassium iodide and a starch indicator in 5% sulfuric
Instrumental analysis and gas chromatography (GC) odor acid, which turned blue in the presence of ozone. When ozone
evaluation. The volatiles were analyzed by GC (HP 5890 se- absorption was complete, an excess of triphenylphosphine
ries II; Hewlett-Packard, Palo Alto, CA) on a Supelco SPB-1 (100 mg) was added, and the reaction mixture was allowed to
fused-silica capillary column (30 m, 0.25 mm i.d., 0.25 µm warm to room temperature. The methyl 9-oxononanoate in
film thickness; Bellefonte, PA) at a helium flow rate of 1.7 the ending reaction mixture (1 mL) was purified by chroma-
mL/min. The temperature program was held for 4 min at tography on a 10-mm i.d. × 26-cm column filled with Florisil
30°C, raised from 30 to 210°C at 6°C/min, and finally held at in hexane. Hexane/ether (9:1, 100 mL) was used to remove
210°C for 10 min. Peaks were detected by a flame-ionization triphenylphosphine, and then ether was used to elute methyl
detector (FID) held at 230°C. Volatiles were identified by 9-oxononanoate. The first 15 mL of elution was discarded,
mass spectrometry (MS) (HP 5970 series Mass Selective De- and the next 10 mL was collected. The ether fraction contain-
tector), and when possible, by comparison of their retention ing methyl 9-oxononanoate was concentrated and rechro-
times with those of known standards. Odor evaluations were matographed on Florisil. The refractionated methyl 9-ox-
accomplished by three trained, experienced sensory panelists, ononanoate ether solution was concentrated. The odor of
who smelled individual volatiles as they issued from the GC methyl 9-oxononanoate was evaluated by two of the three
detector port. For peaks with detectable odor, the intensity trained panelists noted earlier, by smelling the GC detector
and sensory descriptors were recorded. The GC temperature port. Also, after evaporating the ether, mineral oil was added
program for odor evaluation was the same as that used for GC to make up 1 and 10 ppm methyl 9-oxononanoate in mineral
analysis. The sniffing port was adapted at the elution end of oil, and the flavor intensity and character (presence of typical
the GC column by blowing out the flame of the FID. No ad- nutty flavor) were compared with a blank mineral oil by the
ditional attachments were added. same two panelists.
Measurement of aerosol particles generated in human Preparation of carbonyl-free soybean oil. A 2,4-dinitro-
mouths. Particles in the 0.5 to 5 µm diameter range were mea- phenylhydrazine (DNPH) column was prepared according to
sured with a Met One particle detector (Met One Inc., Grants the procedure of Schwartz and Parks (9). The DNPH (Aldrich
Pass, OR). The counter was placed in a biological safety cab- Chemical Company Inc., Milwaukee, WI) (0.5 g) was dis-
inet (Nuaire, Plymouth, MN), and panelists breathed particle- solved in 6 mL 85% phosphoric acid (Fisher Scientific, Pitts-
free air from inside the cabinet until their breath revealed no burgh, PA) by grinding in a mortar, and 4 mL distilled water
particles (~1 min). Then, the panelists generated particles by was added to form a yellow solution. Celite (10 g) was added
smacking their lips or pulling their tongues away from the to the solution and ground to make a homogeneous mixture.
roofs of their mouths. The number of particles so generated Then the DNPH-impregnated Celite was tamped in a 16-mm
were counted by inserting the inlet hose of the counter near i.d. × 61-cm column filled with hexane.
or between the lips and sampling at a flow of 2.8 L/min. Commercial soybean oil (Wesson, 50 g) was dissolved in
Sensory evaluation of soybean oil before and after alumina 60 mL hexane and applied to a DNPH column. A yellowish
fractionation. Wesson brand soybean oil (200 g) was deodor- oil was eluted with hexane (100 mL) and then passed through
ized at 220°C for 6 h under vacuum (0.005 mm Hg), and 100 a column (35-mm i.d. × 57-cm) filled with 100 g alumina in
g was immediately fractionated by passage through a 35-mm hexane. The eluate (100 mL) obtained from the alumina col-
i.d. × 57-cm column filled with 200 g alumina in hexane (5). umn was passed through a new 100-g alumina column again
Triacylglycerides (TAG) and the compounds less polar than to obtain a carbonyl- and DNPH-free, colorless soybean oil
TAG were eluted with 2 L of hexane/ether (9:1). A yield of in hexane. The hexane was removed by rotary evaporation at
62 g of TAG resulted after rotary evaporation of the solvent 45°C and steam deodorization at room temperature under
at 45°C and steam distillation at 120°C under vacuum (0.005 vacuum, and the flavor intensity and sensory description of
mm Hg) for 1 h. The TAG, with and without alumina frac- the carbonyl-free soybean oil were compared with freshly de-
tionation, were diluted 100-fold with water to make emul- odorized soybean oil twice by the three trained panelists, who
sions (6), and their flavors were evaluated by 10 panelists in a tasted the oil directly.
triangle test on two separate occasions. The samples (60 mL) Synthesis of glyceroltricaprate (GTC). Capric acid (178 g,
were presented to the panelists in 266-mL plastic cups, cov- Sigma) and glycerol (30 g, Sigma) were heated and refluxed
ered with aluminum foil, and tasted at room temperature. Sig- with 66 mL high-performance liquid chromatography-grade
nificance was accepted at P ≤ 0.05 (7). benzene (Fisher Scientific) and 3.13 g p-toluenesulfonic acid
Synthesis of methyl 9-oxononanoate. Methyl 9-oxono- (Sigma) in a flask equipped with a Dean-Stark trap and con-
nanoate was prepared by the procedure of Hammond et al. (8) denser (10). The reaction was refluxed for 20 h, and 17.3 mL
for identifying positional isomers of unsaturated fatty acids, of water was formed and trapped in the Dean-Stark trap. Dis-
except the carbon disulfide was replaced by pentane. A solu- tilled water (300 mL) was added to the reaction mixture, and
tion of 25 mg methyl oleate in 3 mL pentane was cooled in the organic phase (upper layer) was separated with a separa-

JAOCS, Vol. 75, no. 9 (1998)

 VOLATILE COMPOUNDS PRODUCED DURING DEODORIZATION 1105

tory funnel. The organic phase (GTC and some benzene) was hydrocarbons that varied in length from 8 to 18 carbons
washed three times with 100 mL 5% sodium carbonate and (Fig. 2). Identification was based on the molecular weight of
three times with 100 mL distilled water. The organic phase the largest mass peak. Compounds of less than eight carbons
was then dried with 23 g sodium sulfate. may have been generated but lost during evaporation of the
Ninety-two grams of synthesized GTC (with a trace of solvent. Probably, the chains were generated by heat degra-
benzene) was dissolved in 100 mL hexane and passed through dation of the carbon chains as follows:
a 35-mm i.d. × 57-cm column filled with 200 g alumina in O O
hexane. The GTC was eluted with 600 mL of hexane/ether R–O–C–CH2CH2CH2R′ R–O–C–CH=CH2 + CH3–R′
(9:1). After rotary evaporation at 80°C and steam distillation
at 80°C under vacuum (0.005 mm Hg) for 1 h to remove Because 15-carbon chains were the most prevalent, seem-
hexane, ether, and benzene, 50 g of GTC was obtained. The ingly scission between carbon 3 and 4 of the C18 chains is fa-
GTC was then deodorized at 200°C under vacuum (0.005 mm vored. Scission between carbon 6 and 7 or 5 and 6 also seems
Hg) for 1 h and purified again by passing through the 100-g to be favored.
alumina column. Twenty grams of GTC was obtained after Limonene also was identified in the nonpolar fraction by
rotary evaporation of solvent at 80°C and steam deodoriza- its MS and retention time. It was the only compound of the
tion at 80°C under vacuum for 30 min. The GTC was evalu- nonpolar fraction to exhibit a noticeable odor when smelled
ated for flavor intensity and description by the three trained at the exit port of the GC. Probably, the branched structure of
panelists on two separate occasions. limonene is generated from heat degradation of sterols or
squalene rather than from fatty acyl groups.
Polar fraction. The polar fraction from the Florisil column
RESULTS AND DISCUSSION
consisted primarily of aldehydes (Fig. 3, Table 1), including
Trap contents. The double cold trap on the deodorizer was de- heptanal, cis-4-heptenal, a heptanal dimer, and heptanoic
signed to capture volatiles arising from previous oxidation of acid. Cis-4-heptenal has been shown to arise from 2,6-nona-
the oil in trap 1 and to capture volatiles produced during de- dienal by a reversal of the aldol condensation (11), and a sim-
odorization in trap 2. Examination of the contents of trap 2 ilar breakdown of 2-nonenal or its precursor could lead to
by GC indicated that the major components were hydrocar- heptanal. The carbonyls likely come from the slow decompo-
bons, but that carbonyl compounds accounted for most of the sition of the stable dialkylperoxides formed in free-radical
flavors. For better identification, the trap contents were sepa- termination reactions. The heptanal dimer is presumed to be
rated into polar and nonpolar fractions by chromatography on formed by aldol condensation of the heptanal after its release
a Florisil column. from its precursor.
Nonpolar fraction. The nonpolar fraction from the Florisil Many of the other compounds of trap 2 are well-known ox-
column was shown by GC–MS to be primarily unsaturated idation products (1). The menadione, like limonene, may

FIG. 2. Gas chromatography (GC)–mass spectrometry (MS) chromatogram of hydrocarbons

formed during deodorization of soybean oil. The silicone peaks are attributed to degradation
of the GC stationary phase.

JAOCS, Vol. 75, no. 9 (1998)

1106 J.-W. KAO ET AL.

FIG. 3. GC–MS chromatogram of carbonyls formed during deodorization of soybean oil. Peak
numbers are identified in Table 1. See Figure 2 for abbreviations.

come from the rearrangement of sterols or squalene or their in Figure 3 were not further identified by MS; however, none
oxidation products. 3-Methyl-2,4-nonanedione has been re- of these had odor significance. Thus, although several volatile
ported as a photoxidation product of furanoid fatty esters in odor compounds were found in the deodorized oil, none
soybean oil (12), but a trace seemed to be generated during seemed to be present in sufficient amounts or to have suffi-
deodorization. Phenylpropanone, a relatively strong fruity- cient odor intensity to account for the flavor of freshly de-
rose flavor as determined by the three trained, experienced odorized soybean oil. Previous unpublished work from our
sensory panelists, may come from the degradation of linole- laboratory also showed that contact of freshly deodorized
nate. The odors of other carbonyls found in trap 2, and de- soybean oil with human saliva did not release volatile flavor
scribed by the three sensory panelists, are listed in Table 1. compounds. The flavor perception of freshly deodorized oils
The odors of the heptanal and cis-4-heptenal were the most seems to be mediated by the olfactory organ, because the
potent in the polar fraction, but because they are fairly nutty flavor cannot be detected if the oil is tasted with the ob-
volatile, not much accumulated in the oil being deodorized, server’s nose blocked.
and no compounds with low volatility were present in signifi- Flavor from aerosol particles. An alternate mechanism that
cant amounts or had an unusually intense odor. Several peaks might account for the flavor of freshly deodorized oil is sug-

TABLE 1
Carbonyl Compounds Found in Trap 2
Identification
Retention time
Number Compound (min) MSa RTb Odor descriptorc

1 Hexanal 6.15 + + Green

2 cis-4-Heptenal 9.54 + + Fish oil
3 Heptanal 9.72 + + Heptanal
4 Benzaldehyde .21 + + Cherry
5 2,4-Heptadienal 12.96 + + Fruity
6 Nonanal 15.93 + + Slight fruity
7 Heptanoic acid 16.3 + + Sweaty
8 Phenylpropanone 17.27 + + Fruity-rose
9 2-Nonenal 17.32 + + Aldehyde
10 3-Methyl-2,4-nonadione 19.3 + + Licorice
11 2,4-Decadienald 20.73 21.22 + + Beany
12 Decano-γ-lactone 24.3 + + Buttery lactone
13 Menadione 25.28 + + Spicy
14 Heptanal dimer 27.01 + — —
a
MS, mass spectrometry.
b
RT, retention time.
c
Odor identified, within the retention time of each peak, by three trained, experienced sensory
panelists.
d
cis-cis and cis-trans isomers elute at different times.

JAOCS, Vol. 75, no. 9 (1998)

 VOLATILE COMPOUNDS PRODUCED DURING DEODORIZATION 1107

gested by the work of Hammond and Smith (13), who reported Thus, if there is a nonvolatile flavor component in freshly
that the odor of swine houses was carried on particulates and deodorized vegetable oils, it seems not to be very polar nor to
that particulates enhanced odor intensity because they were be a carbonyl. The nutty flavor of GTC suggests that the nutty
scrubbed from air streams in the nasal passage much more ef- flavor may be caused by the glycerol esters themselves.
ficiently than molecularly dispersed odorants. This argument
is based on the observation that the olfactory organ is placed ACKNOWLEDGMENTS
at a sharp bend in the air stream and particles in the 0.5 to 5
This is Journal Paper J. 17471 of the Iowa Agriculture and Home
µM range are removed and deposited on the olfactory organ Economics Experiment Station, Ames, Iowa, Project No. 3128. We
by a centrifugal effect. Thus, the load of odorants on particles gratefully acknowledge the Iowa Soybean Promotion Board for par-
of this range is released directly on the olfactory organ. tial financial support.
Using the three trained panelists, we found that several
hundred particles in the 0.5 to 5 µM range could be generated REFERENCES
by smacking one’s lips and by pulling the tongue from the
1. Lee, I., S.H. Fatemi, E.G. Hammond, and P.J. White, Quantita-
roof of one’s mouth. Interestingly, these are precisely the con- tion of Flavor Volatiles in Oxidized Soybean Oil by Dynamic
ditions in which one can taste freshly deodorized oil. If one Headspace Analysis, J. Am. Oil Chem. Soc. 72:539–546 (1995).
puts freshly deodorized oil in one’s mouth and breathes in and 2. Stone, R.R., and E.G. Hammond, An Emulsion Method for the
out while keeping the lips slightly parted and tongue still, Sensory Evaluation of Edible Oils, Ibid. 60:1277–1281 (1983).
there is no flavor perception, but if one opens and closes the 3. Moulton, K.J., Laboratory Deodorization of Vegetable Oil, Ibid.
66:302–308 (1989).
mouth or pulls the tongue from the roof of one’s mouth while 4. White, P.J., and E.G. Hammond, Quantification of Carbonyl
breathing gently through the mouth, the flavor is perceived. Compounds in Oxidized Fats as Trichlorophenylhydrazones,
Flavor from nonvolatile compounds. These observations Ibid. 60:1769–1773 (1983).
encouraged us to look for nonvolatile flavor compounds in 5. Jensen, R.G., T.A. Marks, J. Sampucna, J.G. Quinn, and D.L.
the oil. It seemed likely that the flavors would be more polar Carpenter, Purification of Triglycerides with an Alumina Col-
umn, Lipids 1:451–452 (1966).
than TAG, considering the nature of the flavors found in oxi- 6. Dixon, M.D., and E.G. Hammond, The Flavor Intensity of Some
dized oil. Thus, we tried an experiment that might concen- Carbonyl Compounds Important in Oxidized Fats, J. Am. Oil
trate the flavors from freshly deodorized soybean oil by pass- Chem. Soc. 61:1452–1456 (1984).
ing the oil through alumina columns. No difference in flavor 7. Roessler, E.B., R.M. Pangborn, J.L. Sidel, and H. Stone, Ex-
intensity before and after treatment was perceived in the TAG panded Statistical Tables for Estimating Significance in Paired-
Preference, Paired-Difference, Duo-Trio and Triangle Tests, J.
fractions eluted from such columns. This treatment would Food Sci. 43:940–943 (1978).
eliminate mono- and diacylglycerides or other polar com- 8. Hammond, E.G., W.C. Ault, E. Heftmann, H.K. Mangold, N.
pounds that might contribute to the flavor. Pelick, H. Schlenk, and R.J. Vander Wal, Lipid and Related
Aldehydes and ketones are only slightly more polar than Compounds, in Specifications and Criteria for Biochemical
TAG, so it seemed possible that azelaic semialdehyde esteri- Compounds, edited by Donald L. MacDonald, National Acad-
emy of Sciences, Washington, D.C., 1972, pp. 119–147.
fied with glycerol, or ketoacyl groups generated by decompo- 9. Schwartz, D.P., and O.W. Parks, Preparation of Carbonyl-Free
sition of acyl hydroperoxides esterified with glycerol could Solvents, Anal. Chem. 33:1396–1398 (1961).
cause the flavor of freshly deodorized oils. To this end, 10. Quinn, J.G., J. Sampugna, and R.G. Jensen, Synthesis of 100-
methyl azelate semialdehyde (methyl 9-oxo-nonanoate) was Gram Quantities of Highly Purified Mixed Acid Triglycerides,
synthesized by ozonolysis of methyl oleate, and its odor and J. Am. Oil Chem. Soc. 44:439–442 (1967).
11. Josephson, D.B., and R.C. Lindsay, Retro-Aldol Degradations
flavor were examined by smelling it as it exited from the GC of Unsaturated Aldehydes: Role in the Formation of c-4-Hepte-
and tasting the compound in mineral oil solutions. Methyl 9- nal from t-2, c-6-Nonadienal in Fish, Oyster and Other Flavors,
oxo-nonanoate had no detectable flavor or odor. Commercial Ibid. 64:132–138 (1987).
soybean oil also was passed through a DNPH reaction col- 12. Guth, H., and W. Grosch, Detection of Furanoid Fatty Acids in
umn to convert any volatile or nonvolatile carbonyls it con- Soya-bean Oil—Cause for the Light-Induced Off-Flavor, Fat
Sci. Technol. 93:249–255 (1991).
tained to DNPH derivatives. Then, the DNPH derivatives 13. Hammond, E.G., and R.J. Smith, Survey of Some Molecularly
were removed by chromatography on alumina, and the oil Dispersed Odorous Constituents in Swine-House Air, Iowa
was tasted after removal of the solvent at low temperature. State J. Res. 55:393–399 (1981).
The oil had a nutty flavor typical of freshly deodorized oils.
Finally, GTC was synthesized from glycerol and capric acid.
After purification on an alumina column and steam deodor-
ization, the flavor of GTC was perceived by two panelists as
nutty, whereas the other panelist identified no odor. [Received June 19, 1997; accepted April 4, 1998]

JAOCS, Vol. 75, no. 9 (1998)