FIBROUS BED COALESCENCE OF W A T E R Role of a Sulfonate Surfactant in the Coalescence Process ROBERT
N. H A Z L E T T
Naval Research Laboratory, Washington, D.C. 10390
The effect of sodium sulfonate surfactant on coalescence in a fibrous bed was examined under a variety of conditions. The flow velocity of the continuous phase exerted a slight control. The fiber size and order in the filter bed were important, but the surface chemical properties of the fiber were not critical. The bed depth was not important above a critical amount. The surfactant is not active in the approach or attachment steps of coalescence, Rather the sulfonate interferes with droplet release from the downstream face of the filter. It reduces the interfacial tension between the two liquids and thus limits droplet growth prior to detachment.
HE passage of an emulsion of water in an immiscible liquid Tthrough a fibrous bed is widely used for coalescing small water droplets. The larger water droplets formed by this means are separated more readily from the continuous phase. The coalescence process is easily altered by additives or contaminants in either phase. Some surface active materials interfere with coalescence at a concentration of 1 p p.m. or less. Hazlett (1969) examined the processes occurring in coalescence and suggested probable mechanisms which control the drop size of the released drops. He included possible modes of action by surfactants. This paper presents experimental results for coalescence studies involving one type of surfactant, sodium alkyl benzene sulfonate. In addition to surfactant concentration, a number of parameters were examined: fiber size and material, bed depth, packing density, water content, and flow velocity of the continuous phase. Experimental Details
The size of the water drops in an emulsion is not readily controlled. A11 methods of emulsion preparation give a range of particle sizes. Segregation into a narrow size fraction might be feasible experimentally but the routine determination of sizes in the micron range is extremely cumbersome. Therefore an emulsion formed by feeding water into the inlet side of a pump during circulation of fuel was utilized in the experimental work described here. The size distribution of a pump-formed emulsion, which is typical of emulsions in the field, has been reported by Bitten (1967). Ninety per cent of the water droplets were less than 4.8 microns and 100% were less than 8.6 microns in diameter. The fuel properties chosen for these experiments were those which make coalescence more difficult. JP-5. the primary jet fuel used by the Navy, or closely related fuels were used, since these materials have higher viscosities and higher densities. The range in these properties (Table I ) was limited, since other factors were of greater interest. The basic tool used was the Water Separometer, designed by Krynitsky and Garrett (1961) and further developed in cooperative studies in the Coordinating Research Council (1962, 1964). It was developed to rate jet fuels with respect to their coalescing properties and is used in the water separation specification test for purchase of jet fuels. In this use, the Separometer is standardized with respect to configuration
and operating procedures. The instrument prepares a 0.1% water in fuel emulsion, passes the emulsion through a selected coalescer bed of bonded glass fiber and a settling chamber where coalesced water drops fall to the bottom, and determines the turbidity of the effluent water-fuel emulsion leaving the settling chamber by routing it through a turbidity analyzer, The light transmission in the turbidity analyzer a t a prescribed time is taken as an indication of the water-separating performance of the fuel being tested. The light transmission is not directly related to the amount of water in the effluent fuel. since the amount of light scattered by an emulsion is a function of the particle size as well as the number of particles. A quantity of water scatters light much more efficiently dispersed as small droplets than dispersed as large droplets. The relationship between scattering and particle size is not amenable to a formulation useful for the current work. Even so, the turbidity is a useful indicator of coalescing action, although one cannot differentiate between two interpretations of a turbidity decrease: All of the water drops have increased in size but the free water content has not changed; or some of the water drops have not changed size but the free water content has decreased. Positions in between these two extremes are also possible. A concurrent determination of water content is helpful in resolving this uncertainty. The Water Separometer is an excellent tool for coalescence studies. In addition to the versatility which permits adjustment of the many variables referred to above, limited quantities of fuel and coalescer material are required and all components of the apparatus are contained in a single unit which can be placed on a laboratory bench. The second Coordinating Research Council (1964) study on the Water Separometer showed that equilibrium conditions are reached 6 minutes after beginning of emulsion flow with a 1/16-inch coalescer bed depth. Accordingly, the %minute reading has been adopted for this bed depth. Deeper beds sometimes require longer run periods to reach equilibrium conditions because of greater capacity for water holdup. For the purposes of this paper, fuel samples are classified as good if the per cent T reading is 85 or higher, while samples yielding readings less than this value are classified as poor. The flow velocity of the fuel through the coalescer was varied in two ways. The volumetric flow rate through the coalescer can be varied and this was adjusted from 25 to VOL.
0
NO. 4
NOVEMBER
1 9 6 9
633
Table I.
Table II.
Properties of Fuels Used in Coalescence Studies Density, Fuel
Viscosity,
2 5 O C . , G./MI. 25' C., Cp.
0.777 0.789 0,810 0.802
BayoP
RF- '1 JP-5(434) JP-5 (435)
2.63 1.60 1.72 1.47
Coalescer Cell Studies
Cell Opening Cell
Diam., cm.
Flow Velocity, Cm./Sec.
-- --
Area,
150
37
sq. cm.
ml./min.
ml./min.
Highly reflned kerosine available from Humble Oil and Reflning
co.
* Reference fluid composed of 15% toluene in Bayol.
150 ml. per minute. The 150-ml. per minute rate for specification tests was used in a majority of the tests. The flow velocity was also altered by varying the area of the coalescer pad exposed to fuel flow, by fabricating coalescer cells with different cell openings (dimension D, Figure 1). The four cells listed in Table I1 cover an area ratio of 30 and this allows an even greater velocity ratio when combined with the different volumetric flow rates. When the cell opening is altered to change velocity through the coalescer cell. the velocity in the settling chamber remains the same, but it is affected when the coalescer cell velocity is adjusted by changing the volumetric flow rate. The flow velocity in the settling chamber varies linearly with the flow rate and a t 150 ml. per minute is 0.50 em. per second. The velocity in the settling chamber determines the size of particle which will collect in the sump, A fourfold velocity increase would decrease the maximum size of particles retained in the sump by a factor of 2. The coalescer bed depth was varied by adjusting the effective length of dimension E of the coalescer cell (Figure 1). One or more l/l&inch spacers were used to reduce this dimension from the standard dimension of 5/8 inch for the insert, increasing the fiber bed depth from the standard 1/16 inch. Coalescer Material. Glass fibers with average diameters from 0.4 to 11 microns were used throughout these studies. The fibers down to 0.75-micron diameter were available with a phenolic bonding. Unbonded fibers were available in the 0.4- to 11-micron range or could be prepared by removing the bonding agent with a 1-hour heat treatment at 450' C. I n view of the importance of fiber dimensions, it was felt
r-1 2"
I
Figure 1. 634
l&EC
Coalescer cell
FUNDAMENTALS
Standard cell for modifled Water Separometer. Standard cell for original Water Separometer.
that valid comparisons of different surfaces must take this into consideration. By modifying the surface of a coalescer pad containing the desired fiber diameter, pads which differed only in a surface coating could be compared. The bonded glass fiber afforded a medium energy surface and the unbonded material gave a high energy surface. A low energy surface was attained by treating the unbonded glass so that a hydrocarbon moiety controlled the properties of the surface. Two chemical and one physical method were utilized. The hydrophilic hydroxyl sites were esterified by refluxing with anhydrous n-octyl alcohol (Iler, 1955) or complexed by treatment with Quilon, a dilute chromium stearate solution which is a fiber-treating reagent produced by E. I. du Pont de Nemours & Co., Inc. The physical method involved dipping the glass fiber in a boiling xylene solution of polyethylene, followed by evaporation of the xylene a t 100' C. A concentration of 0.3 to 0.5y0 was suitable for this last procedure. The effectiveness of these three treatments mas determined by observing the behavior of a coalescer pad in a two-phase system, Bayol over distilled water. Unbonded glass when placed in the Bayol layer readily penetrates the interface and water displaces the Bayol. An organic coating alters this behavior and the coalescer pad will not move through the interface without additional force. Even then the Bayol is displaced only with vigorous agitation and the water adsorbed can be easily removed by re-exposure to Bayol. This behavior is typical of the phenolic bonded glass fiber as well as the coatings produced for a low energy surface. The coating prepared by treatment with octanol appeared to be the most resistant to displacement of Bayol by water. Surfactants. Sodium petroleum sulfonates frequently contaminate fuels as a result of certain refinery processes. Two samples from such refinery processes were used in these studies. Sample A had a molecular weight of 400 and sample B had a molecular weight of 420. Stock solutions of the surfactants in toluene were added in the desired amount to a fuel shortly before a coalescence experiment. One-gallon glass bottles were used to mix samples. Analysis of Water. The water content of the effluent fuel stream was determined in some experiments. Analysis of samples collected downstream of the turbidity chamber was by one of two methods. One method, a determination with Karl Fischer reagent, must be corrected for water dissolved in the fuel. This method is not very accurate for water contents less than 25 p.p.m. (wt./vol.) because of a relatively large correction. The other, which might be called the dyecentrifuge method, determines only free water and was developed for the present study. In the latter method, a water-soluble dye (Acid Fuchsin Red has been used successfully) is added to the water used in the preparation of the water-in-fuel emulsion. The water is normally adjusted to pH 5.5 by the addition of sodium acetate, since the dye is a strong acid. The effluent fuel sample collected after passing through the Water Separometer is treated with a known amount of water buffered a t p H 4, the p H of maximum dye intensity, then alternately shaken and centrifuged t o separate the colored water. An aliquot of this centrifuged water is diluted 1 to 1 with isopropyl
alcohol and the intensity of the dye determined in a spectrophotometer a t 535 mp. The amount of free water in the fuel sample can then be related to the dye concentration in the diluted sample. The light transmission of water-infuel emulsion is not significantly altered by the presence of the dye in the water. R a t e r contents determined by the two methods-Karl Fischer titration and dye-centrifuge technique-were comparable. The dye is insoluble in fuel and did not alter the coalescing performance. Coalescence in a Transparent Cell. A square transparent cell constructed from poly (methyl methacrylate) was used for visual and photographic studies. The coalescer section was 1 em. square, thus being comparable to cell C in area and flow velocities (Table 11). The bed depth could be varied with gaskets. Experimental Results
Effect of Flow Velocity. The flow velocity through the coalescer material was varied over a wide range. The per cent transmission of the turbidity cell was taken a t comparable total fuel flows for each flow velocity. The effect of flow velocity on per cent T shown for three different fuels in Figure 2 is rather slight. Additional data (not plotted), indicate only a slight velocity effect for other conditions also. A 30-fold decrease in the flow velocity raises the per cent T of RF-1 with 1.0 p.p.m. of additive less than 10 points, although it causes a larger response with 0.5 p.p.m. of additive. The flow velocity for RF-1 plus 0.5 p.p.m. of sodium petroleum sulfonate was altered both by using the four coalescer cells described in Table I1 and by adjusting the flow rate. The data obtained were interconsistent and are plotted together in Figure 2. Figure 2 also illustrates the remarkable effect of surfactant on the per cent transmission in coalescence experiments. As can be seen, doubling the surfactant level influences this coalescence phenomena more than a 30-fold change in velocity. A comparison of Figure 2 with the interception efficiency for droplet approach to a fiber (Hazlett, 1969) indicates a flat response for both single fiber interception efficiency and 100 -
I
-
* I
I
-
,
I
I
-,
I
*
EAYOL
i
90 -
1
30
I
I
I
I
I
1
I
Table 111.
Effect of Flow Velocity on Coalescence Using BayoC Bonded Fibers
Flow Velocity, Cm./Sec.
% T
0.83
99
2.5
1 0
8
99
25 a
Hz0 content, p.p.m.
100 100
8.3
Unbonded Fibers
34
R T 100
Hg0 content, p.p.m.
10
1 1
100 99
...
...
Cell gap 1/16 inch. Standard coalescer bed.
per cent T with respect to flow velocity. Although the slopes are opposite for the two criteria, the effect of velocity in each case is slight. The apparent upturn in per cent T a t lower velocities might be explained by the increasing efficiency of the diffusion mechanism a t lower velocities. The slope is less pronounced than that expected for a diffusion-controlled process, however (Hazlett, 1969). Other phenomena downstream from the point of initial droplet approach to a fiber are influenced by velocity and such effects would be superimposed on the approach mechanisms. An indication of the complicated nature of coalescence is obtained when a water analysis is made in addition to the per cent transmission. About one half of the water from a O.lyoemulsion is trapped in the apparatus and only 500 p.p.m. of water reaches the fibrous bed. For Bayol fuel, the water content of the effluent from the Water Separometer increased with flow velocity, although the data in Table I11 indicate that per cent T does not change with velocity. When an additive is present in the fuel, the opposite behavior is observed. The water content in the effluent for flow velocities used is about 500 p.p.m. for a sulfonate-containing fuel, while the per cent T changes as shown in Figure 2 . Physical Characteristics of Coalescer Bed. The amount of fibrous material, the fiber sizes, and the arrangement of different components of a bed have a significant effect on coalescence. The amount of filter needed for efficient coalescence is shown in Figure 3. The data were obtained with bonded glass fiber beds in which a constant amount of 4.6micron-diameter fibers and a variable amount of 1.0-microndiameter fibers were used. The smaller fiber preceded the coarse fiber in all the beds. The amount of filter material is indicated by a pressure differential for air flow (8 liters per minute) through the beds as measured in cell C. The data indicate that efficiency is not raised by increasing the quantity of glass fiber beyond a certain amount, for a good fuel which gives a high reading of per cent T or for a fuel with additive. The influence of fiber diameter as well as amount of coalescer material is shown in Figure 4. These results were obtained with a constant amount of 4.6-micron bonded glass fiber preceded by varying amounts of smaller diameter unbonded glass fiber. As before, the flat response of per cent transmission to quantity of glass fiber is noteworthy, as is the greater efficiency of small diameter fibers. The data for 1.0-micron-diameter fibers shown for this same fuel as the lower curve in Figure 3 would fall between the 0.35- and 2.1-micron fibers. The 1.0-micron pad data are for bonded fibers, but the data still fall in the same order. Attempts were made to obtain comparable data for the 4.6-microndiameter bonded material alone, but an air flow resistance beyond 10 em. of water could not be attained without tearing the pads a t the high packing densities required. The per cent T values were in the low ~ O ’ S ,but no general trend could VOL.
8
NO.
4
NOVEMBER
1969
635
I
90 -
I.
FUEL--J-435
50
FUEL--J-435
+ 0.5 PPM
I
w
-
SODIUM SULFONATE ( A )
20
I
I
8
12
Figure 3.
I
I
I
I
16 20 24 28 A I R F L O W RESISTANCE (CM O F WATER)
I
I
32
36
3
Effect of amount of coalescer material on per cent T
Coalercer. Bonded glass fiber, varying quantity o f 1 -micron-diameter fibers, and a constant quantity o f 4.6-micron-diameter fibers Cell gap. 1/16 inch Fuel velocity. 2 5 cm./sec.
be established. The data generally support the pattern established for the smaller fibers. The efficiency with the larger fibers may still be increasing as bed resistance increases and might approach the level attained by the small fibers if sufficient material was added. The efficiency with respect to fiber size is in the order expected from the calculations presented by Hazlett (1969). The per cent transmissions for coalescer beds composed of two or more pads with different fiber diameters are depicted as bar graphs in Figure 5. The necessity for including more than one fiber size is clearly evident. A small fiber size (0.75 and 1.0micron) and a coarser fiber (2.8 and 4.6 microns) are both inefficient by themselves. Combinations of a small and large size greatly improve the efficiency. Thus a 0.75or 1.0-micron pad with a 2.8-micron pad is efficient and a pad made from either of the small fibers is efficient when combined with a 4.6-micron fiber pad. A gradation of fiber sizes does not further enhance the efficiency. A bed consisting of a 0.75-micron pad plus a 4.6-micron pad is slightly better than a graded bed containing pads of each of the fiber sizes used in Figure 5. 70
I
I
I
I
I
-
I
I
0.35-p FIBERS
a rr 50 ap
I
i-
30
20
4
I
I
8
12
1
I
I
I
I
16
20
24
28
32
AIRFLOW RESISTANCE (CM OF WATER)
Figure 4.
Effect of glass fiber diameter on per cent T
+
0.5 p.p.m. sodium sulfonate (A) Fuel. J-435 Other conditions as in Figure 3
636
l&EC
FUNDAMENTALS
I
The order of the fiber sizes is important. The performance of a bed made from 1.0- and 4.6-micron fibers deteriorates if a small fiber pad is added to the downstream side of the bed (1.0 to 4.6 combination us. 1.0 to 4.6 to 1.0 combination), but there is no change if the same pad is added to the upstream side (1.0 to 4.6 us. 1.0 to 1.0 to 4.6). Addition of a coarse pad on the upstream side is not detrimental (4.6 to 1.0 to 4.6 combination). It thus appears that the downstream fiber size has a marked control on coalescence. The interception, attachment, and growth of droplets can occur in efficient processes but may be nullified if droplet release is inadequate. Surface Chemical Considerations. The behavior of different surfaces with respect to coalescence was examined. The important physical characteristics were kept constant by modifying the surfaces of glass fiber samples in several different ways. Surface effects for a particular fiber size with a variety of surface treatments could then be compared. I n Table IV, comparison is made between 1.0-micron bonded glass fiber pads heated for 1 hour a t various temperatures. In each case, the 1.0-micron pad was followed in the coalescer bed by a bonded 4.6-micron fiber pad. Most of the bonding material is removed a t 300' C. and all is burned off at 400' C. The behavior in a two-phase fuel-water system is consistent with this fact. A fiber pad heated a t 200' C. as well as an unheated pad prefers the fuel layer, whereas fuel is readily displaced by water with pads heated a t 300' C. and above. The only significant difference with the observed per cent transmissions in coalescence experiments was found for the pad heated to 600' C. Since this pad fused slightly at this temperature, the changes in fiber physical characteristics rather than surface properties are probably responsible for the poorer performance. Comparisons were next made with modifications of the 1.0-micron fiber surface, the 4.6-micron fiber surface, or both. In each case, the larger fiber was used with and preceded by the smaller fiber pad in the coalescer bed. Three surfaces were compared: glass fiber with phenolic bonding; unbonded glass fiber prepared by heating bonded glass fiber a t 450' C. for 1 hour; and Quilon-treated surface prepared by treating unbonded glass fiber with a chromium stearate complexing
R
u Figure 5.
u
u
T
R
s
T U
s u
s u s
u s u
OF FUEL FLOW
Per cent T for coalescence with combinations of bonded glass fiber diameters
+
Fuel. RF-1 0.5 p.p.m. sodium sulfonate (6) Cell gap. 1/8 inch Fuel velocity. 25 cm./sec.
reagent. The comparisons are listed in Table V for a high quality fuel and the same fuel to which a surfactant was added. All combinations performed well with RF-1. With the additive fuel, the unbonded samples gave per cent T readings slightly higher than the bonded glass fiber. The hydrophobic fibers were similar to the bonded samples. The comparable behavior of bonded and unbonded fibers with a good fuel is also indicated when the water content of the effluent is determined at different flow velocities (Table 111). The 1.0-micron fiber pads were treated in other ways and then combined with the bonded 4.6-micron material in coalescence experiments. Two additional treatments gave hydrophobic surfaces similar to the Quilon reagent: esterification of the hydroxyl sites on the glass and coating the fibers with polyethylene by dipping in a dilute solution of this polymer in boiling xylene. The per cent T readings with pads treated in either way were very close to those for bonded fibers. Partial degradation of the phenolic bonding material by treatment with acids, bases, or oxidizing agents also failed to alter the coalescing behavior of 1.0-micron material. The absorption of sulfonate additive by glass fiber was examined with respect to the effect of this material on fiber surface characteristics, An emulsion made with a fuel containing 2.0 p.p.m. of additive A was passed through bonded glass fiber coalescers. After 5.5 liters of this poor fuel had passed through, a good fuel was used with the same coalescers. The per cent T rose from 17 for the additive sample to 89 for the follow-up sample. This latter value is Table IV.
Effect of Heat Treatment on Coalescencea
Temp., C.
Wettability
No heating
Hydrophobic Hydrophobic Hydrophilic Hydrophilic Hydrophilic Hydrophilic
200
300 400
500 600
+
% T 42 42
46 40 42
31
Fuel, J-434 0.5 p.p.m. sodium sulfonate (A). Velocity, 25 cm./sec. Cell gap, 1/16 inch. Standard coalescer bed. a
similar to those of 88,90, and 93 obtained for the good fuel in three separate experiments with unused filter pads. A poor fuel (2 p.p.m. of sulfonate A) was also passed through coalescer pads without the addition of emulsifying water. The follow-up fuel, which was the same as that above, gave readings of 87, 85, and 95 in three different experiments. Absorption studies were also undertaken with unbonded glass fiber and additive B. Experiments similar to those above were carried out, both with and without water addition during the pretreatment step. Unbonded glass fiber pads were also soaked up to 48 hours as a pretreatment step on 0.25% solutions of additive B in toluene or Bayol. None of these procedures degraded the coalescing performance of a follow-up fuel sample. The results given above on fiber surfaces and surfactant absorption studies indicate that the surface chemistry of the fiber is not critical. Rather the effects of surfactants on coalescence must be assigned to the water-fuel interface as Lindenhofen and Shertzer (1967) have proposed. Surfactant Concentration. The phenomenon of coalescence is extremely sensitive to traces of sodium petroleum sulfonate. Figures 6 and 7 show that 0.1 p.p.m. of this additive alters the behavior of Rayol fuel, which passes no
Table V.
Variation of Coalescer Surface
1 .o-
4.6-
Micron Fibers
Micron Fibers
B U B U
B B U U B
2Q
% T RF-1
99 100 100 99,98 99, 98, 97 98 95
Q
a
RF-1-B
63, 59, 55
67,66 66. 61 67; 74, 70 56, 55, 56
64
51
B. Glass fiberbonded. U. Glass Bberunbonded (hydrophilic). Q. Glass flber-unbonded and treated with Quilon (hydrophobic). 0.5 P.p.m. sodium petroleum sulfonate ( B ) . RF-1-B. RF-1 Cell gap, 1/16 inch. Velocity, 25 cm./sec.
+
VOL.
8
NO.
4
NOVEMBER
1969
637
A* 70
\.-
‘O.
2
6
Figure 6.
TIME10I M I N )
8
12
14
18 0.5 PPM
16
26
4
Effect of sulfonate (B) concentration in Bayol on per cent T
Coolercer bed. Three 1 .O-micron-diameter pads followed by one 4.6-micron-diameter pad Cell gap. 1 / 8 inch Fuel velocity. 2.5 cm./rec.
I
I
I
- 500 a I -a
W
/ / / / /P’
-
s
P
L
/ A AI J’
300
I
I
-
+-W
0
I
I
0 5 P
5 400 ~
I
I
/
/’
/
..-e-
-
,D-----
0 2 P r M . ...A
..-._.--.
,e.. .-’....--A ,I
&..”’ *...-.
03 P P M a
g--I
-
I -
-
water a t this flow velocity in the absence of the sulfonate. Increasing amounts of additive increase the turbidity in a regular way (Figure 6), but the water concentration as measured by the “dye-centrifuge” technique does not follow a regular pattern (Figure 7 ) but rather shows a sharp step between 0.1 and 0.2 p.p.m. of additive and a gradual rise above 0.2 p.p.m. Although the emulsions for these experiments were made with 1000 p.p.m. of water, about one half of the water is held up in the apparatus. Most of the water being released from the coalescer a t the higher additive concentrations is not retained in the settling chamber. The difference in per cent T for the 0.3- and 0.5-p.p.m. samples is ascribed to the difference in water droplet size and number but, not to the total amount of water. Water Content in Emulsion. The amount of water used to prepare a fuel emulsion containing sulfonate also influences the free water content of the fuel after passage through a 638
l&EC
FUNDAMENTALS
coalescer. The relationship between input water and unsettled water is shown in Figure 8. Greater amounts of water were found in the effluent at low water inputs. A minimum amount of effluent water was found a t an intermediate input water level. The surfactant samples used in these studies are much more soluble in water than in fuel. Consequently, the concentrations of the surfactant in the water phase is reduced as the water content is increased, with a resulting concentration decrease a t the water-fuel interface and improved coalescence. Transparent Cell Studies. In his movies of coalescence phenomena, Brown (1966) noted that the water drops on the downstream face of the glass fiber bed were normally a few millimeters in diameter at the time of release. Drops grew into balloon-shaped drops and detached from the same sites by rupture of the neck. This study with good quality fuel established the normal mode of droplet release.
I
I
I
I
I
I
I
I
I
I
I
H 2 0 CONTENT OF EMULSION
I
2
3 3 0 1
TIME (MINI
Figure 8.
Effect of input water on water content of effluent fuel
Fuel. Bayol with 0.1 p.p.m. sodium sulfonate (B) Same conditions as Figure 6
Additives in the fuel alter this behavior (Hazlett, 1969). The drop size on the bed face is limited by some additives. Also, release may occur from an internal bed site, in which case growth cannot be observed. Sodium petroleum sulfonate has a considerable effect on droplet growth and release at the bed face. As with good fuel, activity occurs repeatedly from the same sites. The drops do not grow into balloons, however, but are extended by the force of the flowing fuel into thin fingers, which oscillate normal to the fuel flow and release drops from the tip. These observations, made with a fuel containing 0.2 p.p.m. of additive, are a variation of the jet rupture mechanism for injection of one liquid into another immiscible flowing liquid (Merrington and Richardson, 1947). These workers observed rupture of a liquid jet by Rayleigh instability (Rayleigh, 1879), in which case a series of nearly uniform droplets was formed. In the case where water is being released from a fibrous bed, the rate of water feed is low and extension of the water thread to the point that a series of uniform droplets forms is not observed. Release of drops from the tip of the water thread is probably controlled by the Rayleigh phenomenon, however. Discussion
The experimental work presented in this report focuses our attention on four salient features of this coalescence study: the importance of fiber sizes in the coalescer bed; the strong effect of surfactants on coalescence; the minor influence of the flow velocity on the per cent transmission readings; and the flat response of the per cent transmission to the quantity of coalescer material once a critical amount has been exceeded. The fact that the per cent T for a poor fuel reached a plateau similar in shape to that for a good fuel as the amount of coalescer material increased shows that the efficiency of the approach step can reach unity for both types of fuel with proper bed design. Further, since the only effect of surfactant on theapproachmechanism is through a change of particle size of the water drops, the droplet size in the emulsion must not be greatly affected by sodium petroleum sulfonate in the
fuel. Measurements by Bitten (1967) are in agreement with this conclusion. The slight effect of flow velocity on per cent T supports the viewpoint that the interception process is the primary approach mechanism for water in fuel emulsions. Theoretical calculations agree with this finding (Hazlett, 1969). In addition, the minor effect of velocity indicates that the drop size formed a t release may be insensitive to flow velocity. An additional piece of evidence which bears on the approach problem is the relationship of the water content to the per cent T reading for fuel samples containing sodium sulfonate. The per cent T of such a fuel leaving a good coalescer is much higher than that of the entering fuel and may even be above 90% if the additive concentration is low. The water content of the effluent fuel does not change accordingly, and all of the incoming water may be retained in the effluent from the settling chamber. This is interpreted to mean that approach has been accomplished successfully; coalescence of all small drops has occurred but drop size is controlled by the release process. Visually it has been observed that effluent samples which have a high per cent T and a high water content have little or no haze and any haze that, is present quickly settles to the bottom of a container. The loss in light transmission for such samples is thus due to scattering by relatively large coalesced drops which are not massive enough to sett,le in the settling chamber, rather than scattering by small droplets from the entering emulsion which have passed through the coalescer bed without any interaction. Sareen et al. (1966) have observed coalescence in a fibrous bed in which the enlarged droplets were too small to allow settling. It has been reasoned that the interception mechanism is the significant approach step for coalescence and that an efficiency of unity for approach can be attained even for an additivecontaining fuel. The site of additive action must therefore be assigned to the attachment or release processes. The former process is a time-dependent one, since the drainage of a fuel film from between an approaching water drop and a fiber requires a finite time. If this were the crucial step, the flow velocity should have a significant effect on over-all VOL.
8
NO. 4
NOVEMBER
1969
639
coalescence efficiency, that efficiency increasing as the velocity is reduced. The experimental results indicated that this was not the case. The other arguments presented in favor of an efficient approach mechanism-namely, the flat response of per cent T to the amount of coalescer material and the observance of high per cent T readings in conjunction with high water contents-also support an efficient attachment mechanism. Lawson (1967) indicates that film drainage time is very short for small emulsion drops and attachment of such droplets is not a critical process. Valid reasons have been presented which suggest that the sodium petroleum sulfonate in a fuel does not alter the approach or attachment steps of coalescence such that either of these steps becomes controlling. Let us look a t the final step, the release of the water droplet. In his observations of droplet detachment, Beatty (1966) found that the droplets oscillated prior to release. A small residual water drop remained attached to the filament after release and detachment occurred much more readily in fuel containing powerful sulfonate surfactants. Events subsequent to release occurred rapidly and were not observed. Merrington and Richardson (1947) examined the phenomena associated with the flow of a liquid from a capillary into a moving liquid. They found a t velocities higher than those characteristic of coalescers that the droplet size was controlled by interfacial tension and capillary diameter. The velocity did not influence the droplet size. Although the experimental conditions and details are far different for the work by Merrington and Richardson (1947) and this study, there is a similarity in control for the two flow regimes. The coalescence process for a sulfonatecontaining fuel is controlled by fiber size-Le., diameter of a water channel a t the point of release-and interfacial tension but influenced only slightly by velocity. Measurements of the interfacial tension of water-soluble surfactants by the pendant drop method have been reported by the CRC (1964). Strong sulfonate surfactants lower the tension of fuel-water interfaces below 10 dynes per cm. and degrade coalescer performance significantly. Observations in the transparent coalescer cell reinforce the argument that the sulfonate additive interferes with the release of water droplets from the downstream fiber face. No haze was observed which would indicate the passage of minute droplets through the fiber bed without attachment. Small droplets were emitted from the tips of water threads projecting from the downstream face of the coalescer. The following picture of coalescence in a fiber bed emerges from this study. Water droplets in a fuel emulsion approach the fibers and reach a fiber primarily by the interception mechanism. Hydrodynamic factors-fiber size, particle size, differential density, and fuel viscosity-control the efficiency of this mechanism, but beds can be and have been designed which afford an efficiency of unity. The drops
640
l&EC
FUNDAMENTALS
attach to a fiber and grow by further droplet interception to a larger size. The extent of growth is determined by the balance between the water-fuel interfacial forces and the flow forces of the fuel. The fiber-fiber distance is small compared to the coalesced droplet size and the droplet is attached simultaneously to several fibers as it threads through them. At the downstream side of the fiber bed the droplets are released as discrete drops. The concentration of sulfonate surfactant and the fiber size at the point of release are important influences on the size of the released water drops, and largely determine the amount of water left in the fuel after passage through a coalescer. This proposed picture needs to be tested in a study which relates the drop size distribution of coalesced water to other experimental parameters. Goren (1968) carried out this type of work on oil-in-water emulsions. Acknowledgment
Thanks are extended to Naval Air Systems Command for Rupport of this work under project A33-536/652/ 69F32.543.301. The experimental work included in this report was performed by Cora A. McLean. Glass fiber samples were furnished by Owens Corning Fiberglas, . Toledo, Ohio, and Johns Manville, New York, N. Y. Surfactant samples were made available by Chevron Research Co. literature Cited
Beatty, H. A., Ethyl Corp., Ferndale, Mich., Interim Tech. Rept. GR-66-80 (July 1966). Bitten, J. F., IIT Research Institute, Chicago, Ill., Annual Rept., Project C6088-4 (May 1967). Brown, R., Engineer Research and Development Laboratories, Fort Belvoir, Va., private communication, 1966. Coordinating Research Council, New York, Rept. 368 (February 1962). Coordinating Research Council, New York, Rept. 376 (February 1964). Goren, S.L., University of California, Berkeley, Calif., private communication, 1968. IND.ENG.CHEM.FUNDAMENTALS 8, 625 (1969). Hazlett, R. Iler, R. K., Colloid Chemistry of Silica and Silicates,” p. 170, Cornell Universitv Press. Ithaca. N. Y.. 1955. KryniGky, J . A., Gkrett, W. D., Naval Research Laboratory, Washington, D. C., Rept. 6686 (Aug. 22, 1961). Lawson, G. B., Chem. ProC. Eng. 48,45 (1967). Lindenhofen, H., Shertzer, R. H., Aeronautical Engine Laboratory, Philadelphia, Pa., NAEC-AEL-1862 (April 17, 1967); NAEC-AEL-1866 (July 7, 1967). Merrington, A. C., Richardson, E. G., Proc. Phys. SOC.69, 1
y.,
(1 \ - Q47\ _-.,.
Rayleigh, Lord, Proc. Roy. SOC.27, 71 (1879). Sareen, S. S.,Rose, P. M., Gudesen, R. C., Kintner, R. C., A.Z.Ch.E.J. 12, 1045 (1966). RECEIVED for review August 14, 1968 ACCEPTEDApril 16, 1969
