Steam Deodorization of Edible

Fats and Oils

Theory and Practice
A LTHOUGH a great ~ a r i e t y
of methods have been
proposed for rendering
edible oils odorless and tasteless, A E. BAILEY
obtained from steam refining
tests should be also applicable
t o steam deodorization. This
is important in an investigation
it is now almost universal prac- of the process. for free fatty acid
tice to accomplish this by a The Cudahy Packing Company, South Omaha, Nebr. removal from an oil is easily
process of steam distillation followed by a simple titration
under reduced pressure. Ac- with alkali and is therefore well
cording to a recent historical review of the subject (6), this adapted to quantitative study. Oils are also readily available
process probably originated in France. I n the central and with free acids in any desired concentration.
northern European countries it has been used on a large scale
for many years in the manufacture of margarine fats. Prob- Apparatus for Batch Steam Deodorization
ably its most extensive and useful application has been in the The ordinary type of batch deodorizer is shown in side elevation
United States, however, in converting the large American in Figure 1. It consists essentially of a cylindrical insulated steel
tank, fitted at the bottom with a pipe for the injection of steam
production of cottonseed oil to edible purposes. In the and at the top with a thermocompressor for the maintenance of
United States alone the annual production of deodorized fats high vacuum. In the commercial equipment used in the present
is now not less than 1.5 billion pounds. tests, the injected steam was expanded from the mains (at about
In spite of the relative importance of the deodorization proc- 150 pounds gage pressure) through a throttling valve to 15-75
pounds, and then passed through a metering nozzle to a dis-
ess in oil and fat technology, the published data concerning it tributor at the bottom of the deodorizer where it was divided into
have been meager. Theoretical or semitheoretical treatments thirty separate jets. The injected steam was not superheated.
of the subject have been given by Whiton ( I f ? ) ,Brash ( 1 , g), The diameter of the deodorizer was 8 feet, so that 20,000 pounds
and Singer ( 9 ) , but all of them have neglected factors of im- of oil at 460" F. filled it t o a depth of about 8.5 feet. The oil
was heated by Dowtherni (diphenyl and diphenyl oxide) vapor,
portance in actual operation-for instance, the several sources condensed in roils in the lower part of the deodorizer body. Pres-
of oil loss in deodorizing. sure within the head space of the deodorizer was measured on a
Reports on commercial deodorizer operation have also been closed-end mercury manometer connected to the deodorizer body
few, although those of Thurman (11), Khiton ( I f ? ) , and Dean above the surface of the oil. The injected steam, together wit,h
the thermocompresaor steam, discharged through a Venturi tube
and Chapin (3) may be mentioned. The purpose of this paper to a barometric condenser, so that any substances distilled from
is to consider the theory of the process in some detail and to the oil appeared in the water discharged from the condenser.
report certain experimental data gained from the operation of
commercial equipment. The paper ~villbe confined to batch, Theory- of Distillation by Sleain
as distinguished from continuous, deodorization. In the study of petroleum distillation problems, 137. K.
Lewis and eo-workers have successfully used a differential
Nature of the Process treatment based on the application of Raoult's law. The
same method is applicable to the deodorization process.
The deodorization process is essentially one of steam dis- According to Dalton's law, in the vapors issuing from the
tillation, in which relatively volatile odor-causing substances deodorizer a t any instant, the molar ratio of volatile substance
are stripped from the relatively nonvolatile oil. These sub- to steam will be the same as the corresponding ratio of their
stances are probably of aldehydic and ketonic nature. A few partial vapor pressures. Mathematically,
definite compounds have been identified in coconut oil (4, 5,
8), but in the oils and fats containing no low-molecular-
weight fatty acids, the identity of the odoriferous substances is
still unknown. Hydrogenation, even of a previously odorless where S = moles of steam
oil, imparts a characteristic odor and flavor, and the source of 0 = moles of volatile odoriferous substance
p , = actual partial pressure of steam
this is also still undetermined. In the case of the common p , = actual partial pressure of volatile substance
animal and vegetable fats, such as cottonseed oil, soybean oil,
palm oil, tallow, lard, etc., the vapor pressures and probably But p , is very small in relation to p,, so p , approaches
also the molecular weights of the unkno\\n substances are closely the total pressure. Consequently as a close approxi-
comparable t o those of the CI6and Cl8 fatty acids, since in mation of Equation 1we may write:
practice flavor and odor removal is observed to parallel free dS P
fatty acid removal very closely.
The concentration in which the odoriferous substances exist
a = g
where P = total pressure
in these oils is likewise unknown, although it is ordinarily less
than about 0.20 per cent, since the above-mentioned fats, if From Raoult's law,
0
well refined, can be deodorized with losses not greater than
this.
(mF)
P: = Po . . , (3)
Closely related to the process of steam deodorization is that where p: = equilibrium pressure of odoriferous substance
of steam refining. The two are similar in execution although Po = vapor pressure of pure odoriferous substance
F = moles of fat or oil
different in purpose. I n steam refining the object is not the
removal of odoriferous substances but of free fatty acids. As 0 is very small in relation to F , however, so that (0 F )+
mentioned above, the odoriferous substances and the free approaches closely the value of F . Therefore, as an approxi-
fatty acids of the oil are comparable in volatility; hence data mation of Equatior, 3 we may write:
404
March, 1941 INDUSTRIAL AND ENGINEERING CHEMISTRY 405

p: = P,O/F (4) mitic acid a t 460" r-l

F. is about 33
It is convenient at this point to introduce a new factor, E , mm., of oleic and
the so-called efficiency of vaporization. By definition, other C I S acids,
E = P~/P: (5) about 18 mm.
If a value of 33
Combining Equations 4 and 5,
mm. is taken,
po = EP,O/F Equation 8 gives
0.64 for E. If 18
Finally, from Equatjons 2 and 6, mm. is taken, E
dS PF becomes 1.18. The
a=m (7) mean of these two
valuesis0.91. The
Integrating Equation 7,
s= g ( 1 n0
g)
freefatty acid con-
tent of the oil was
originally 4.6 per
where 0, = initial concentration of odoriferous substance cent, and this had
O2 = final concentration of odoriferous substance been reduced to
2.6 per cent by
Thus, the amount of steam required for deodorization is:

';6
distillation before
(a) directly proportional to the amount of fat or oil, (a) di- the experiment
rectly proportional to the total pressure in the deodorizer, was started. This
(c) inversely proportional to the vapor pressure of the pure
preliminary dis-
volatile or odoriferous substance a t the temperature of tillation, if effec- FIGURE1. DIAGRAM OF RATCH
operation, and (d) inversely proportional to the vaporieation t i v e a t all i n DEODORIZER
efficiency. separating the GO
The vaporization efficiency is, practically speaking, a and Cls fatty acids, could have served only to decrease the
measure of the completeness with which the steam bubble original proportion of the former, so from this standpoint the
becomes saturated with volatile substance during its passage above figure may reasonably be regarded as a minimum one.
through the oil. From the two-film theory of gas absorption However, such experiments as the one outlined above are
(7) we have the condition, that a t any instant the rate of subject to some uncertainties unavoidable in the use of ordi-
transfer of volatile substance from the oil into the steam nary commercial equipment and thus are by no means of a
bubble is equal to the difference between saturation pressure high order of precision. Other experiments with palm oil
in the bubble and actual pressure, times the surface area of have generally indicated vaporization efficiencies of 0.7 to 0.9.
the bubble, times a constant characteristic of the oil and It seems certain that vaporization efficiency in deodorization
steam. Mathematically. is comparatively high, and probable that it is greater than
d p o / d t = IcA(pA - po) (9) about 0.7.
where t = time of contact between steam bubble and oil The progressive removal of free fatty acids with distillation
A = surface area of bubble time in the above experiment is shown graphically in curve 1
k = gas diffusion constant of Figure 2. The rate of fatty acid removal slows appreciably
as the concentration of the acid falls to the lower levels. It is
Integrating Equation 9,

RUN NO. 1 - 19.WO LBS. O I L - 4 0 0 L b I STEAn

(12)
1
= In

Vaporization Efficiency in Practice

A rough idea of the magnitude of E in actual practice has
been gained from observations of the rate of free fatty acid
removal from palm oil under controlled conditions of vacuum,
temperature, and injected steam. The following experiment
is typical:
Oil used, pounds 19,000
Steam, pounds per hour 400
Tem erature O F 460
Absoyute pre;sure; mm. 10
Time. hours 0.50
Average mol wei ht of oil 848
Free fatty aiid re&ction, % 2.6-0.90
0 1 2 3 4
The chief difficulty here is in assigning an average value for
HOURS DCODORIZLD
the vapor pressure of the fatty acids. Palm oil contains
palmitic and oleic (and other 18-carbon fatty acids) in a p FIGURE2. REMOVAL ACIDS
OF FREB FATTY
proximately equal proportions. The vapor pressure of pal- FROM PALMOILAT Two DIFFERENT
PRESSURES
406 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 33, No. 3

believed that this slowing is apparent only, since there is a steamed a t a rate just short of that a t which significant
slight tendency for the steam t o split the fat with the freeing quantities of oil begin to be blown out of the deodorizer. The
of fatty acids even a t the low pressures employed. This entrainment losses (total loss corrected for distillation loss) in
splitting action is constantly in operation, of course, but is deodorizing at two different pressures are shown graphically
apparent only toward the end of the process when its magni- in Figure 4. Appreciable entrainment becomes evident a t a
tude approaches that of the reverse action of fatty acid steam Row of about 400 pounds (181 kg.) per hour at 10 mm.
removal by distillation. The range of free fatty acid con- pressure and at about 600 pounds (272 kg.) per hour a t 25 mm.
centration chosen in the experiment (2.6-0.9 per cent) would pressure. Over the range covered by the experiments, 35 to
appear to be somewhere near the optimum for purposes of cal- 65 per cent more steam is required for a given loss at 25 mm.
culation; it is low enough so that the conditions assumed in than at 10 mm.
Equation 8 should apply, yet high enough so that slight This is in reason-
hydrolyzation of the fat should not vitiate the results. ably good accord
All of the foregoing has been developed on the assumption with theory. En-
that Raoult's law applies to the mixture of oil with fatty acids trainment presum- 5 so
or other volatile material. It is realized that this may not be ably occurs when
entirely justified, and that the factor Po in Equation 8 might oildroplets, thrown E ,,
be more accurately represented by the Henry's law constant. up above the sur- 0
Even should this be true, however, it is certain that the vari- face of the oil, are 2
ation of this constant with temperature, within the range in more powerfully 2 70
which we are interested, cannot be greatly different from the impelled upward $
variation of Po. Any deviation from Raoult's law would by the vapor A,,
affect the above calculations of vaporization efficiency. At stream than they $
present there are apparently no data available on the mix- are downward by
tures in question, and it can only be said that the method out- t h e f o r c e of 0
ZOO 400 6W em 1WO 12w
lined has proved useful in predicting actual deodorizer per- gravity. The rLOW OP INJECTCD STCAM-POUNDS R R HOUR
formance. kinetic energy of
the oil droplets as FIGURE4. ENTRAINMENT LOSSESIN
they are thrown up DEODORIZING AT Two DIFFERENT
Losses from the Deodorizer is known to be a PRESSURES
Although the neutral triglycerides are among the least factor in entrain-
volatile of all organic substances, their vapor pressures are ment in certain classes of equipment. I n view of the great
high enough so that distillation losses become appreciable in distances from the surface of the oil to the top of the deodorizer
deodorizing a t elevated temperatures. Figure 3 shows the and also from the surface of the oil down to the steam jets, it is
results of a typical loss test on a hydrogenated cottonseed oil not believed to be important in the present case, however.
shortening deodorized a t varying temperatures. I n this test Under a given set of conditions, then, there is a critical velocity
deodorization was carried out in stepwise fashion a t tempera- of the steam required for carrying droplets of a given size.
ture levels of from 300' t o 490" F., while the Dressure on the This velocity is given by the expression (IO):
deodorizer was
held a t 10 mm.
and the steam flow
a t 350 pounds per where u =linear velocity of steam
hour throughout
D =diameter of droplet
dl = density of droplet
t h e run. Loss dp = density of steam
figures were ob- K = a constant
tained by deter-
mining the neutral At a given temperature dl is constant and hence may be incor-
fat in the con- porated in a new constant, K'. The magnitude of dz is in-
denser water, the significant with respect to that of dl. As an approximation of
flow of which was Equation 11 we may therefore write:
measured each
time a sample was
taken. The dis-
tillation loss in- It will be remembered that each value of v has corresponding
creases a t a rate to it a definite degree of entrainment and a definite rate of
r50 300 330 400 450 500 approaching ex- loss.
TCMPLRATURE - or.
ponential, dou- A consideration of Equation 12 will reveal that equal
FIQURE3. DISTILLATION Loss FROM blingfor eachincre- losses should theoretically result when the steam flow a t 25
DEODORIZER AT 10 MM. PRESSUFLE ment in tempera- mm. is about 58 per cent greater than a t 10 mm. It would
AND 350 POUNDS PER HOUROF IN-
JECTED STEAM
ture of about 40" seem reasonable to expect the number of small droplets
F. (22" C.). The thrown up into the vapor stream, and hence the entrainment,
data of Figure 3 to bear some relation t o the vapor velocity alone without
also provide the basis for a rough calculation of the vapor regard to its density; but it is evident from the experimental
pressure of the glycerides. The loss a t 10 mm. pressure with data that if this is a factor to be considered, it is a t least not a
350 pounds of injected steam per hour is about 55 pounds per major one.
hour a t 482" F. (250"C,). Assuming an average oil molecular Figure 4 also shows that entrainment losses increase with
weight of 860 and a vaporization efficiency of 0.8, the vapor great rapidity as the steam flow increases t o the higher levels.
pressure of the glycerides so calculated is 0.04 mm. This is also t o be expected. The steam velocity required to
The question of loss by entrainment is also of considerable entrain oil droplets varies with the square root of the diameter
practical importance, since the batch will ordinarily be of the droplet, whereas the weight of the droplets varies with
March, 1941 INDUSTRIAL AND ENGINEERING CHEMISTRY 407
the third power of the diameter. The weight of the largest has been found satisfactory as a guide for operation in the
droplets of oil entrained varies, therefore, with the sixth deodorization of hydrogenated cottonseed oil products. It is
power of the steam velocity or steam flow. No attempt will based upon a steam flow of 300 pounds (136 kg.) per hour at
be made here to correlate the weight of the largest particles of 6 mm. pressure, and batches of 20,000 pounds of oil.
oil entrained with the actual loss in pounds of oil per hour, The remaining temperature to be considered is that of the
but it is obvious that the one will be more or less inclined to injected steam. If the vacuum is high and the requirement
follow the other. of steam correspondingh low, there is no evidence that any
A third source of loss in deodorizing is in the tendency of benefit results from superheating the latter. This is only
steam a t elevated temperatures to split neutral glycerides into reasonable in view of the relative heat capacities of the oil and
their component fatty acids and glycerol, which readily dis- steam. A batch of oil weighing 20,000 pounds gains approxi-
till off. This splitting tendency of the steam seems to bear mately 1,000,000 B. t. u. as it is heated from 350" to 450"E'.
a direct relation to its total pressure and is largely avoided in (177"to 232" C.). The steam required for an hour's blowing
practice by keeping the pressure very low. Determination of of this oil (300 pounds) gains only about 15,000 B. t. u. as i t is
free fatty acids in the total water-insoluble distillate from the superheated from saturation at 350" to 450' F.
deodorizer indicates a loss from this source of about 3 pounds
per hour a t 460" F. (238" C.),and 10 mm. pressure, with a
steam flow of 350 pounds (150 kg.) per hour. At pressures of Influence of Vacuum
25 mm. and higher, with the steaming correspondingly in- One effect of vacuum has been noted-viz., with other fac-
creased, the loss may amount to several times this figure. It tors constant, the steam requirements are directly propor-
is particularly important that the pressure be kept low if a tional to the absolute pressure. With ordinary three-stage
very low free fatty acid content is desired in the finished oil; thermocompressor equipment operating with a barometric
otherwise the building up of acids from fat splitting may condenser, the maximum vacuum easily obtainable is about
nearly approach the rate a t which they are removed by dis- 0.25 inch (6 mm.) of mercury. If this vacuum drops to 0.50
tillation, toward the end of the deodorization period. Figure inch the steam requirements would be expected to be doubled;
2 shows how much more rapidly a low fatty acid in palm oil if it drops to 1.00 inch, they would be expected to be quad-
is attained a t a pressure of 10 mm. than a t 25 mm. rupled. Actual observation seems to c o d i n n the above ex-
pectations, a t least within the somewhat wide limits of error
Influence of Temperature of our methods of measurement, and within the range of 6 to
It was noted above that the amount of steam required for 25 mm. of pressure. Thus in the two experiments of Figure 2
deodorization is in inverse proportion to the vapor pressure of a t equivalent temperatures and with approximately equal
the volatile compounds of the oil. As the temperature range amounts of oil, the total pressures in the two runs stand in the
with which we are concerned is only about 300" to 500" F. inverse ratio of 1to 2.5, whereas the corresponding steam re-
(149" to 260' C.), and within this range there is practically a quirements for equivalent fatty acid distillation are about 1 to
linear relation between the temperature and the logarithm of 8.3. Furthermore, in the series of experiments on the steam
the vapor pressure, the importance of operating at the highest refining of palm oil mentioned in a previous paragraph, there
possible temperature is immediately apparent. was no correlation between the pressures which varied from 6
Just what constitutes the highest possible temperature may to 25 mm. and the vaporization efficiencies which varied from
well depend upon the facilities a t hand for heating the deodor- 0.7 to 0.9. Theoretically the pressure should affect the
izer rather than on the possibility of doing harm to the oil. vaporization efficiency considerably unless the latter is so high
If steam a t the pressures available in most plants is depended as to be very close to unity at all pressures. Eight and one
upon for heating, a top temperature of about 375'F. (191"C.) half feet of oil a t 450"F. exert a hydraulic pressure of some 190
may be reached. Boilers employing a high-boiling-point mm. of mercury. Each bubble of steam during its passage
organic compound instead of water have recently come into through the oil is therefore subjected to a constantly varying
extensive use, and with these, temperatures in excess of 500' F. external pressure. Both its internal pressure and its surface
(260" C.) are easily attainable. High temperatures may area vary continuously, and under these conditions Equations
also be obtained by a flame applied either directly to the oil 9 and 10 are not applicable without modification, since they
or to a liquid heat-transfer medium. comprehend no such variation. A complete mathematical
treatment of this phase of the subject appears to offer some
difficulties and will not be attempted here. It is apparent,
however, that a high vacuum serves not only to make the
TABLE I. TIME REQUIRED FOR DEODORIZATION OF HYDROGEN-
average steam bubb!e comparatively large, thus making its
ATED COTTONSEED OIL
Temp., Vapor Pressure of Time surface area small in proportion to its volume, but also makes
O F. Oleio Acid, Mm. Hour; its average volume during its passage through the oil small in
350 1.4 16.0 proportion to its final volume. Vaporization efficiency is
375 2.5 9.0
400 4.6 4.0 based upon this final volume at the surface of the oil; there-
425 8.3 a.7 fore it should decrease as the pressure falls. That it appar-
450 15 1.5
475 26 0.9 ently does not may be evidence that saturation of the steam
500 43 0.5
bubble a t high temperatures is so rapid as to be substantially
attained at any pressure down to at least 6 mm.
Another highly important consideration with respect to
If due care is exercised in other respects, the temperature vacuum is its effect on the maximum permissible rate of
of deodorization may be surprisingly high without apparent steaming. If steam can be injected at a maximum rate of 350
injury to the oil. The writer has seen cottonseed oil products pounds per hour at a vacuum of 0.25 inch without excessive
successfully deodorized a t temperatures as high as 525" F. entrainment, it can be shown from Equation 12 that the maxi-
(274" C,). Sufficiently rapid deodorization may usually be mum rate of steaming a t a vacuum of 1.00 inch is twice as
obtained a t 425' to 475" F. (218"to 246" C.), however. In great, or 700 pounds per hour. Since the weight of steam re-
practice good results have been obtained by adjusting the quired a t 1.00 inch is four times that a t 0.25 inch, then twice
time of deodorization on the basis of the vapor pressures of the as long a time will be required to deodorize a t the lower
common fatty acids a t the various temperatures. Table I vacuum.
408 INDUSTRIAL AND ENGINEERING CHEMISTRY Vol. 33, No. 3

Design of Deodorizing Vessel ordinary steam distribution, however, the value of elaborate
The optimum size and shape of the deodorizing vessel have distributing or diffusing devices would appear to be question-
always been a matter of some discussion and conjecture. able.
From the preceding results i t would seem possible that these Acknowledgment
are, within wide limits, comparatively unimportant. It is The author is indebted to R. H. McKinney and R. F. Krage
obvious that increasing the depth of the vessel will tend to for preparation of the figures in this article, and to A. R. Diehl
raise the vaporization efficiency, since factor t in Equation 10 and Tales Newby for assistance in conducting the plant tests
is Drouortional to the oil deDth. As mentioned above. how- and experiments.
ever, there is some evidence that good efficiency is attained Literature Cited
even under adverse conditions of pressure. If this is the case, Brash, J. Sac. Chem. I n d . , 45, 73T (1936).
similarly good efficiency should be obtained with shallow I b i d . , 45, 331T (1926).
bodies of oil. The use of shallow vessels will permit a com- Dean and Chapin, Oil & S o a p , 15, 200 (1938).
paratively greater rate of steaming, since the maximum rate Haller and Lassieur, Compt. rend., 151, 697 (1911).
Haller and Lassieur, J. SOC.Chem. I n d . , 28, 719 (1909).
is more or less a function of the surface area of the oil body. Lee and King, Oil & S o a p , 14, 263 (1937).
For the same reason deodorization may be conducted more Lewis and Whitman, IND. ENG.CHEM.,16, 1215 (1924).
rapidly in small vessels than in large ones. Salway, J . Chem. Soc., 111, 407 (1907).
Distribution of the steam in the form of fine bubbles is indi- Singer, Seifensieder-Ztg.,65, 487, 507 (1938).
Souders and Brown, IND. ENG.CHEX.,26, 98 (1934).
cated, since factor A in Equation 10 is thereby increased. Thurman, Ibid., 15, 395 (1923).
I n view of the high vaporization efficiencies obtained with Whiton. Cotton Oil Press, 7. No. 10, 32 (1924).

Calculating Beattie-Bridgeman
Constants from Critical Data
SAMUEL H. MARON AND DAVID TURNBULL
Case School of Applied Science, Cleveland, Ohio

NE of the most accurate and useful equations of state for

0 gases is that of Beattie and Bridgeman (2) :

where A , B , C, D, E , F , and G are given in terms of the

where p, yB, and 6 are defined by: Beattie-Bridgeman constants and the critical constants P, and
T,by:
A =- Bop,
2.303RT,
B = -AJ'o
2.303 (ET,)
c =- -cP,
2.303RTC4
-BobPo'
Ao, Bo,a, b, and c are constants independent of temperature D = 2 X 2.303(RT0)Z
and pressure, which must be evaluated empirically for any AoaPC2
E =
given gas. The purpose of this paper is to show how, on the 2 X 2.303(RTJ8
basis of certain assumptions, these five constants may be -BocPO2
F =
evaluated for any gas from its critical constants and the 2 X 2.303R*TC5
Beattie-Bridgeman and critical constants of some reference BobcPO3
G =
gas, and also to test the calculated constants for each gas 3 X 2.303R3TC6
with known compressibility data.
Newton (11) found that when values of y obtained graphi-
Derivations of Equations cally were plotted against P, a t constant T,, a n average curve
could be drawn through the points from which the activity
In the preceding paper ( I O ) the authors derived a n ex- coefficients of most of the gases studied deviated by no more
pression relating log,, y , where y is the activity coefficient of than 2 per cent. This finding of Newton's may be generalized
a gas, to the Beattie-Bridgeman constants and the reduced to state that a t the same values of T,and P, all gases, with
pressure and temperature, P, and T,. The derived expression possibly a few exceptions, have the same activity coefficient.
was as follows: Applying this condition to Equation 2, i t must follow that