Oil Slick Removal Using Matrices of Polypropylene Filaments Mohamed A. Zahid, James E. Halligan,l and Robert F. Johnson Texas Tech University, Lubbock, Tex. 79409
Use of matrices of polypropylene filaments in the removal of oil from the surface of seawater was investigated. The data indicated that the distance between the filaments and the cross-sectional area of the filaments was an important factor in the observed removal capacity of the matrices. An increase in the turbulence level of the slick surface markedly decreased the oil retention capabilities of the matrices. At similar interfilament distances, changes in the filament geometry from circular to rectangular led to much higher oil retention capacities while changes in the surface properties of the filaments appeared to have a minimal effect on the observed capacities.
o i l spills are almost inevitable due to the rapidly growing demand for petroleum products. The world oil production is about 1.8 X 1Ol5 gal./year, 60y0 of which is transported across water before final usage (Blumer, 1969). Spills of these petroleum products during transport damage recreational areas, contribute a n undesirable taste and odor to drinking water, and endanger marine life. Therefore, techniques for containment and removal of oil slicks have been under active development. Several devices t h a t employ some type of rotating drum or endless belt are either currently available or are being developed. Continuous sorbent belts constructed of polymers hold great promise since the oil is removed from the water surface by the oleophilic material from which the belt is made. I n many of these systems, the device involves moving a rotating belt through an oil slick prior to removing the oil from the belt using squeeze rolls mounted on a ship (Walkup et al., 1969). Polyurethane, polyethylene, polystyrene, and polypropylene, in foam form as well as polypropylene in fiber form, are among the materials which have been considered for use in these belts (Walkup, 1971). Johnson et al. (1971) reported that polypropylene fiber has one of the highest sorption capacities in a series of synthetic and natural polymer staple fibers. They also observed that a decrease in the capacity, measured in terms of grams of oil recovered per gram of fiber immersed, was experienced with a n increase in the fiber denier, a decrease of crude oil viscosity, or a n increase in the temperature. Schatzberg and Nagy (1971) also reported sorption capacities of a n unstructured length or wad of polypropylene fibers from fibrillated film were tested with different types of crude oil. The data of Schatzberg and Johnson differ considerably in terms of the capacity of polypropylene fibers to remove oil from seawater. Although it is not entirely clear, it is suspected that the variation in their results was due to differences in fiber geometry and spacing. KO study has appeared in the literature concerning the optimization of the distance of separation of fibers in any of the proposed recovery systems. Therefore a study was undertaken to optimize the distance of separation of polypropylene 1
To whom correspondence should be addressed.
550 Ind. Eng. Chern. Process Der. Develop., Vol. 1 1 , No. 4, 1972
filaments in structured matrices. Factors such as filament geometry, cross-sectional area, and surface roughness t h a t could have a n influence on the recovery capacities of the matrices were also investigated. Data were obtained in the presence and absence of wave action. Experimental Method
The apparatus used in this study was simple in concept and design. It consisted of a variable speed mechanism which created waves in a Plexiglas chamber by a reciprocating horizontal motion. The chamber was 1 ft long and 6 in. wide. The bottom was covered by about 1500 ml of a 3.5% aqueous sodium chloride solution on which 150 ml of a light crude oil formed an oil layer approximately 0.20 in. thick. The filament matrix was securely held on the liquid surface by means of a ring stand. At a level of 35 reciprocations per minute, each revolution of the variable speed mechanism gave rise to three rather gentle waves that passed over the matrix on the surface. The average amplitude of the waves was about 1 in. An oil-water emulsion, which contained approximately 6 0 4 5 % water, was formed on the surface of the chamber a t this condition. Since this is similar to emulsions reported in the field, this reciprocation level was used to obtain most of the turbulent data in this study. Before the crude oil was placed in the chamber, it was pretreated by placing trays containing thin layers of the oil in a n autoclave under a suction pressure of 22 in. of mercury for 24 hr. This was done to simulate the process of “weathering” that the oil encounters in the open sea. This greatly reduced the rate of evaporation of the more volatile components when the matrices were tested in the Plexiglas chamber. ,411 of the capacity data were experimentally obtained using a one-day pretreated oil obtained from the Friendswood Field near Houston. Some of the physical properties as \%-ell as the infrared absorbancy ratios for characterization (Kawahara and Ballinger, 1970) are listed in Table I. I n the construction of the matrices used in this study, it was essential that the rows of polypropylene filaments be evenly spaced and parallel. This meant that matrix construction was rather tedious and difficult because polypropylene has a high
Table 1. Characterization of Friendswood Crude O i l at Different Conditions SP 9'
API
Kinematic viscosity,
O i l condition
60160
gravity
cs
7201 1 3 7 5 cm-I
As received 1-Day pretreatment 1-Week pretreatment
0.885 0.587 0.895
28.4 28.0 26.6
21.93 30.58 44.76
0.2251 0.2348 0.2427
bending modulus. T h e filaments were laced into the form of a matrix by pulling them tightly around a perforated sheet metal support. Since these frames were only about 0.125 in. wide and 0.02 in. thick, it took considerable amount of experience and caution to avoid bending the frames or having loose filaments when t'he construction process was complet,ed. When successfully completed, each matrix resembled a screen made of polypropylene. X sketch of the filament lacing pattern is shown in Figure 1. Each frame was made of four thin strips of perforated steel about 3 in. long soldered together. The four sides of the support were designed to be a t the same level so that when t'he filaments were laced into a matrix they would touch each other. Since the cost of having specially prepared perforated sheet metal with closely spaced holes was prohibitive, the study was restricted to spacings which could be achieved using readily available metal screens. The holes in the screens were about 0.76 mm in diameter while t,he spacings from center to center ranged from 1.12 to 1.41 mm. d spacing of twice this distance was also attained by using alternate holes in the metal. However, some very desirable intermediate spacings could not be obtained owing to a lack of screens with the proper hole spacings. By carefully cutting the metal screens, partial holes a t the frame edge were made to avoid pulling 15-30 ft of filament through twice the number of required holes. The filaments were interlaced into alternating openings of two parallel and opposite rows of holes, instead of adjacent holes in a single row. Experience indicated that this pattern produced tighter filaments, and relatively constant spacing in a matrix. After the filaments were interlaced] triangular shaped wires were diagonally attached to the frames so the matrix could be clamped and easily handled. Each of the test matrices was held by means of a clamp on the surface of the oil-water pool a t 0 and 35 rpm reciprocat'ion conditions. h l-min holding period was employed, after which the matrices were raised, and the excess fluid was allowed to drip for 1 min. The saturat'ed matrices were then placed in light aluminum dishes and weighed. Saturated matrices, where the voids were filled with a n oil-water emulsion, were allowed to stand for about 1 hr in the aluminum dishes to allow the emulsion to break. By means of a syringe, the bulk of the water was then carefully removed. However, small droplets of water normally remained immersed within the oil. The aluminum pans containing the matrices rvere then placed in a n autoclave a t a suct'ion pressure of 20-24 in. of mercury for 4 h r to eraporate t h e remaining mater. Since this was adequate to remove the last' traces of water, t,he pans were then weighed again, and the amount of oil picked u p by the frames and the matrices was determined. For t h e zero turbulence immersions] no lvater was trapped within the fluid retained b y the capillary bridges between the filaments. The amount of oil picked u p by the frames and the
Infrared absorbency ratio classification 801 / 81 01 16001 1 3 7 5 cm-I 7 2 0 cm-' 1 3 7 5 cm-'
0.2418 0.2499 0.2445
1.0743 1,0642 1,0075
0.2337 0.2348 0.2445
16001 7 2 0 cm-I
1.0382 1 . 0000 1.0075
matrices was then determined directly without the use of t h e autoclave. The frames were subjected to a t'reatment similar t o t h a t described for the matrices (on frames) to determine the quantity of fluid retained on immersion by the metal frames. Knowing the weight of the filament used in lacing the matrix and the amount of oil retained b y the bare frames permitted the amount of oil retained after immersion to be calculated. The matrix was then carefully cleaned with trichloroethylene and dried prior to a subsequent immersion. The amount of fluid retained by the matrices varied more for the runs when the surface was turbulent. Therefore, a t zero turbulence, each of the data points represents a n average of four immersions, while those reported a t 35 rpm represent averages of eight duplicate immersions. Discussion of Results
Some of the previous investigators who have taken data on unstructured lengths or wads of staple fibers, have used the term sorption to refer to the process by which oil is ret,airied by the fibers. Since the term sorption is rather vague, one of the first objectives was to determine 11-hether the controlling mechanism was absorption, adsorption, or capillary action or a combination of any of these. The high crystallinity of polypropylene filaments gives rise to a compact molecular structure making this polymer essentially impenetrable by liquids. lIoisture absorption in the amorphous regions is dependent upon the presence of
Figure 1 . Metal support and filament lacing pattern Ind. Eng. Chem. Process Des. Develop., Vol. 1 1 , No, 4, 1 9 7 2
551
0 14
148 Denier 246 Denier
0 380 Denier
.^ (bl
% 413 Denier RPM = 35
T=
78 i 2°F
I SDdClng
Figure 2. Qualitative model for the dependence of capacity on spacing
hydrophilic or polar groups not available in the polypropylene structure. Because little swelling of the filaments was observed, it was concluded t h a t absorption probably is not a significant effect. T h e amount of fluid adsorbed on a n isolated filament was small compared to the quantity retained by the capillary bridges in the voids between the filaments. Inasmuch as this ratio suggests t h a t capillary bridging is the controlling phenomenon, the stability of these bridges becomes very important. This stability is related to the physical properties of the materials involved and the length of the bridge and reinforces t'he idea that' the ability of these matrices t o ret,ain fluid should be closely related to'the interfilament distance. T o determine if the experimentally observed relationship between the capacity of the bridges and the distance between adjacent filaments was reasonable, it was desirable to have a t least a qualitative model to predict the form of this relationship. Figure 2 shows a pair of fixed and equal diameter filaments a t a series of different spacings ranging from two adjacent filaments touching each ot,her to a separation where the surface tension force is no longer able to support the bridge and rupture occurs. The geometry of the bridges suggests that there should be a certain spacing at which the amount of fluid held between the two filaments is a maximum. Beyond this spacing a decrease in capacity would be anticipated. and is attributed to the continually decreasing magnitude of the angle a and therefore the thinning of the liquid bridge as the interfilament distance increases. Because the variation of a with spacing is not known, a n expression by which t'he exact capacities can be computed could riot be developed. X camera and a microscope were used to estimate the contact angle, 8, to be about 5' for the air-oil-natural polypropylene system a t 28°C. Physically, this implies that polypropylene filaments are well wetted by the oil-expected, because they are both hydrocarbons. This qualitative model would strictly apply only to infinitely long parallel filaments. Because the matrices used 552 Ind. Eng. Chem.
Process Des. Develop., Vol. 1 1 , No. 4, 1972
Figure 3. Matrix fluid retention capacity on a weight of filament basis
in this study involved not only horizontal but also perpendicular rows of filaments, the model should be used only to assure that the data are in general agreement with this view. The experimental data concerning the influence of spacing on the capacity of a matrix to pick up a n oil-water niixt,ure exhibited a considerable variation with changes in spacing when the reciprocator was set a t 35 rpm. I n the uppermost curve in Figure 3, where the diameter was held constant a t 148 denier, the data show that the capacity is consistent with the qualitative model in that it increases steadily to a maximum, after which any increase in the spacing decreases the overall capacity. Optimum spacings were observed for only the two smaller filaments, namely, 148 and 246 denier due to limitations in the available hole spacings. Denier is a measure of the linear density of filaments, being defined as the weight in grams of 9000 meters. Since the density of all the fibers was the same, the diameter mas directly related to the denier. These data also suggest t h a t the smaller filaments achieved a maximum at a smaller spacing than the larger filaments. This may be due to the higher ratio of the distance between the filaments to the radius, which results in a higher fraction of voids for the matrices constructed using small filaments. This was somewhat substantiated by the fact that Figure 3 shows that the capacity data obtained for the tw-o largest filaments were increasing after the capacity had started t o decrease for the smaller filaments. Unfortunately, the larger filaments will probably reach their maximum capacities between 1.2 and 1.7 mm of spacing. Between 1.2 and 1.7 mni no data have been taken because no metal supports which had the desired hole spacings could be obtained. At higher values for filament separations (1.7-2.5 mm), all the curves suggest the existence of a local maximum relative to their optimum capacities. KO good explanation of this phenomenon was developed. At lower spacings (0.6-1.3 mm), where essentially all the pores in a filament matrix were normally observed to be
10.0
+
0 148
-
9.0
-
8.0
-
0
Denier
148 Denier 246 Denier
246 Denier
c
Y\
380 D e n i e r 413 Denier
a
Y
% 7.0 I u'
-
'4
:
\
0 L
6.0 -
P
E 3
:m:
5.0
-
u
0.6
0.9
I 1.2
I
I
1.5
1.9
2.2
2.4
Spacing, m
I
3.0 0.6
,
I
0.9
I ,2
1
I .I
1 1.8
I
2.1
2.4
Figure 5. Matrix oil retention capacity on a unit area basis
Spacing, m
Figure 4. Matrix oil retention capacity on a weight of filament basis
filled with an oil-water mixture, the average standard deviation of the capacity data for all t h e immersions a t 35 rpm was calculated to be 1Sy0 of the mean. At higher spacings (1.7-2.5 mm) where only about 20-40y0 of the voids were bridged, the average standard deviation was 30y0 of the mean. The higher deviation a t larger spacings was attributed mainly to the fact t h a t only a fraction of the voids were bridged due to the action of waves. At higher interfilament distances, different quantities of fluid were retained by a filament matrix depending upon the position of wave relative to the matrix when it was removed from the fluid surface. If, for inst'ance, a saturated matrix was exposed to a wave just as it was being removed from the fluid surface, some of the voids would lose their contents. However, a t lo\ver spacings when the fluid contained in a void was forced down b y a wave, the void was immediately refilled because of greater bridge stability. Probably partial bridging was due not only to the wave action but also to the nature of the water-in-oil emulsion containing 60-85Qj, water, formed under the 3 5 r p m reciprocation conditions. The resulting mixture may well change the solid-liquid-air contact angle, making bridge formation more difficult. Figure 4 shows the effect of spacing on the capacity at zero turbulence. Under these conditions essentially all of the voids were filled with bridges for all the different spacings used. Therefore, changes in capacity were due merely to spacing with no effect due to the probability of bridging. The capacity curves have approximately the same general shapes as those in Figure 3. However, the top curve seems t o have passed through a local maximum and is rising again, perhaps toward another peak. The average standard deviation for all the data points at zero turbulence was approximately 11% of the mean. At lower spacings, where all the voids were bridged, the data in Figures 3 and 4 show t h a t matrices with comparable spacings had higher capacities at higher levels of turbulence owing to the density difference between oil and water, even
though approximately the same volume of fluid was retained in both the zero rpm and turbulent conditions. Figures 3 and 4 also show the effect of diameter size on the capacity, defined in this case as grams of fluid per gram of filament, a t different spacings. The curves that represent the data were ordered such that the filament diameter was inversely related to capacity. The two lower curves are relatively close to each other because their diameters differ by only 0.01 mm. The two upper curves, however, are farther apart,. The size difference of their diameters, 0.05 mm, is greater than that of the largest two filaments by a factor of five. S o comparison could be made between the 380 and 246 denier curves for the latter had achieved a maximum and data for the former were obtained only during the preliminary and terminal stages of its capacity curve. It can be reasonably speculated that the 380 denier curve will have a n optimum capacity lower than optimum capacities measured for the 148 and 246 denier filaments, but higher than the opt'imum capacity of the 413 denier curve. Because a t the same spacing a gram of the smaller filament permits a larger void space for bridging, a larger capacity in berms of grams of oil per gram of filament would be anticipated. The capacity of a matrix may also be computed on the basis of grams of fluid retained by the matrix per unit area. This procedure would be realistic inasmuch as a n endless belt would probably be designed on a n area rather than a weight basis. The matrix area in which the rows of filaments crossed each other was chosen as the area to use in this computation. Experimentally, this area contained bridges when the filament ends immediately adjacent to the met'al supports did not support bridges. Figure 5 shows t h a t on this basis, the data indicate that larger capacjties were definitely observed for the larger diameter filaments. The curves are labeled according to their diameter size, and it was observed that the 413 denier filament had the highest capacities and the 148 denier the lowest. It is anticipated t h a t the 413 denier filament, however., would have the highest of all the capacities a t its maximum, which incidentally should be to the right of all other peaks because of its smaller interfilament distance to radius ratio a t a fixed spacing. Ind. Eng. Chem. Process Der. Develop., Vol. 1 1 , No.
4, 1972
553
d < a) Side view o f a f l u i d b r i d g e
>
d
b ) Top v i e w o f t h e f l u i d i n a void
Figure 6. Approximate model of an oil bridge
The change in the observed order of the capacit,y curves with a change in basis can be rationalized using a n approximate rectangular model showing the relationship bet'ween the amount of fluid retained by the voids and the filament diameter as shown in Figure 6. If we refer to this figure, we see t'hat on a weight of fiber basis, the lveight, of oil in a void will be proportional to the product of the diameter of the fiber, D,, t,he square of the interfilament distance, d, and the density of the oil, po. T h a t is: wt of oil in a void mt of fiber per void
(Dj)(d2)p0
0:
(:)
(Dj2)(4d)P,
Therefore, when everything is held constant except the diameter of the fiber, D,, grams of oil (Dj ) (d2)( P O ) a grams of fiber ( T ) (Of2)( d ) p ~~
a -
1
D
However on a n area basis wt of oil in a void
a
area of a void
( D j )(d2)( P O ) 0:
d2
diameter of the 246 denier fiber but its width was 3.73 times that diameter. The rectangular filaments were shown to have a larger capacity, defined in this case as grams of oil per unit matrix area, than the circular ones for a n equivalent height and spacing. This was true in spite of the fact that the fraction of voids per unit area of matrix was larger for the round filaments. The porosity of the round filament matrices was about four times that of the rectangular ones. The higher capacity shown by the rectangular than the round filaments was most probably due to the formation of a larger fluid layer. Due to close packing, the rectangular filaments enhance the formation of a much thicker layer of fluid than the round, observed by examining the lower portions of the matrices. Figure 7 indicates that a t smaller spacings, the rectangular shaped filaments have higher capacities than the circular ones even a t the optimum spacings of the latter. However, the capacity of the rectangular filaments was observed to decrease a t a much higher rate with increasing spacing. For instance, a t a spacing of about 1.3 mm, where essentially all of the voids mere still bridges, the data show that the rectangular filaments have capacities t h a t are approximately equal in magnitude to those exhibited by the circular filaments.
Therefore grams of oil unit area
(D,) (d2)( P O ) a
(az)
3.0
D,
The data given in Figures 3-5 are consistent with this very approximate model. The importance of proper spacing in these systems can be seen in Figure 5, for example, where on the 148 denier curve the optimum capacit'y is 30 times that of the lowest obt'ained, both being at' similar conditions of turbulence. Because the larger diameter filaments are expected to have higher capacities at their opbimum spacings in matrices and all of the curves are more or less grouped toget,her a t higher spacings, it is anticipated that the difference between the capacity extremes for a larger diameter filament would be greater than 30 t'imes. It mas expected that a change in the geometry of the filament from circular to rectangular would also have a n influence on the capacity of the matrices. To test this, rectangular filaments were obtained which had the same height as some of the circular filaments examined previously. Matrices were then carefully prepared by orienting the rectangular filaments such that the height of the total array would be the same as those constructed using round filaments. Figure 7 shows the observed effect of filament geometry on the capacity a t different spacings under the 3 5 r p m reciprocation condition. The 507 denier filament' had a height equal to the 148 denier filament but its width was 3.17 times its diameter. The 940 denier filament had a height equal to the 554
Ind. Eng. Chem. Process Des. Develop., Vol. 11, No. 4, 1972
0 148
Deniep (round)
246 3 e n i e t ( r o u n d )
0 507
Denier ( r e c t a n g u l a r ) a r )
953 3 e n i e r ( w c t a n j u l a r ) a r ~
RPY = 35 T = 7 8 : 2.F 2.0
z
-5 0 L
0
*
\
i
1 .o
.
I
"
,
0.0
0.6
0.9
I
1
1.2
1 .5
1 .e
2.1
2.4
S p a c i n g , mm
Figure 7. Effect of filament geometry on oil retention capacity
y+ 413 Denier (smooth) 4 1 3 Denier (rough)
A
41 3 Denier (rougher) RPM = 0 T = 78 t 2°F
3.0
1
0.9
I
1.2
I
I
1.5 Spacing
1.8
I
2.1
2.4
, mn
Figure 8. Influence of surface roughness on oil retention capacity
Due to a lack of curvature of the supporting walls, bridge rupture was anticipated a t a lower spacing for the rectangular filaments than for t h e circular ones which had equivalent filament heights. However, with the spacing range 1.8 to 2.0 mm, the 507 denier filament has a capacity slightly higher than t h e 148 denier filament, while the capacity difference between the 950 and 246 denier filaments is greater. Since most of the voids were not bridged at this spacing, adsorption was increasingly significant, and for the wide flat filaments the capacity difference was more or less expected. Figure 7 is useful because a t a fixed spacing, t h e 950 denier rectangular filaments produced the highest experimentally observed capacity for oil removal based on the effective area of a matrix. This data point should be used to estimate t h e capacity of a continuous belt. The influence of surface roughness on the capacity of polypropylene filaments a t different spacings could not easily be measured with the available apparatus. However, to develop a qualitative estimate of its importance, samples with similar diameters but varying degrees of roughness were solicited from industrial suppliers. The variation in roughness was due t o differences in the titanium dioxide concentration included during filament manufacture. Figure 8, which contains data measured at zero turbulence, shows that for three filaments which had different magnitudes of roughness but the same diameters, the observed capacities are grouped together. Kevertheless, the capacity of the smoother filaments was observed to be slightly higher. The effect of the surface roughness on the capacity, however, is not nearly so pronounced as the effect of the filament diameter. Polypropylene is attacked by atmospheric oxygen, and the reaction is stimulated by heat and ultraviolet light. Therefore, the decrease in capacity of matrices constructed using t h e rougher filaments would probably be compensated
b y the reduction of the sensitivity to sunlight of the filaments owing to t h e incorporation in the polymer melt of titanium dioxide-also responsible for roughening the surface. The primary usefulness of the capacity data would be in the design of a continuous belt. Laboratory experience with these matrices indicated that almost all of the fluid can be easily removed from the voids using a simple air blower and t h a t the matrices can be recycled many times without any significant reduction in capacity. However, changes in the angle of inclination of the matrix during removal from the oil-water surface can cause a serious reduction in the capacity. All of this suggests a relatively large but partially submersible rotating belt of these matrices. By varying the degree of submersion, the angle of inclination of the matrices as they are removed from the surface could be changed. This would also permit some adjustment for different sea states. T o develop an estimate of the capacity of a continuous belt of these matrices, a belt 10 ft wide was assumed t o be rotating a t 5 mph while attached to a boat moving a t 5 mph. If we use the maximum experimentally observed oil removal capacity under turbulent conditions of 2.36 X g oil/cm2, t h e amount of oil removed by such a belt would be: capacity
2.36 X
=
g oil)( 5-rnles)( 5280 ft) _ _ X mile
(y)(%)(--)
(10 ft wide)
0.88 g oil
bbl 159016 em3
=
41 bbl hr
The above calculation assumes the presence of a continuous layer of a n oil-water emulsion with a thickness a t least twice that of the flat filaments used (0.408 mm). If additional emulsion were present, the speed of the belt may be varied to remove additional material. I n conclusion, the data support the contention t h a t proper structuring of polypropylene filament matrices can result in significant increases in their capacity to remove oil from seawater. Acknowledgment
Acknowledgment is made to the Enjay Chemical Co. which supplied the filaments used in this research. literature Cited Blumer, AI., in “Oil on the Sea,” D . P. Hoult, Ed., p 6, Plenum
Press, New York, N.Y., 1969. Johnson, R. F., Manjrekar, T. G., Halligan, J. E., “Textile Materials Solutions to Water Pollution Problems.” Dresented at the Amer. Soc. Eng. Ed. Ann. Meeting, Annapolis, Md., - ti..in^ ... 21-4. 1971. ~~~
I
~
Kawahara, F. K., Ballinger, D. G., Ind. Eng. Chem. Prod. Res. Develop., 9, 553-8 (1970). Schatzberg, P., Nagy, K. V., ‘Sorbents for Oil Spill Removal,” oresented at the Joint Conf. on Prevention and Control of Oil spills, Washington, D.C., June 15-17, 1971. Walkuo. P. C.. Water Pollut. Contr. Fed.. 43. 1069-50 11971). Waikubi P. C.: et al., Dept. of Navy, Naval Facilities Engineering Command, Rept. CR 70.0001, 1969. RECEIVED for review February 11, 1972 ACCEPTEDJune 16, 1972
Ind. Eng. Chem. Process Des. Develop., Vol. 1 1 , No. 4, 1972
555