(95) Trotman-Dickenson, A. F.. Steacie. E. W. R., J. Chem. Phys., 19, 169 (1951). (96) Trotman-Dickenson, A. F., Milne, G. S.. Nat. Bur. Stand., Ref. Data Ser. No. 9,108 (1967). (97) Truby, F . K.. Rice, J. K., lnt. J. Chem. Kinet.. V, 721 (1973). (98) Van Damme, P., et al., AlChEJ., 21, (1975). (99) Voevodsky, V. V., Trans. Faraday Soc., 55, 65 (1959). (100) Voevodsky, V. V., Kondratiev, V. N., "Progress in Reaction Kinetics", Vol. ill, G. Porter, Ed., Pergamon Press, London (1961). (101) Wang, Y. L., etal., lnd. Eng. Chem., Fundam., 2, 161 (1963).
(102) Yang, K. A., J. Am. Chem. SOC.,84, 719 (1962). (103) Zalotai, L., et al., lnt. Cbem. Eng., 10, 133 (1970).
Received f o r review September 30,1976 Accepted June 29,1977
Supplementary Material Available. Pyrolysis mechanisms and rate data (Tables I-I11,21 pages). Ordering i n f o r m a t i o n is given o n any current masthead page.
The Removal of Emulsified Oil Particles from Water by Flotation Christina Angelidou, Elaheh Keshavarz, M. J. Richardson, and G. J. Jameson" Department of Chemical Engineering, lmperial College, London, S. W. 7, England
The flotation of emulsified oil particles suspended in low concentrations in water has been studied. Two oils were used: a spontaneously emulsifying cutting oil or machining lubricant, and white spirit, a petroleum based turpentine substitute which was emulsified by intense agitation. The oil concentrations were up to 200 mg/L. To effect the separation, various cationic surfactants were used in the flotation cell which was operated batchwise with an external total recycle. It was found that the rate of flotation in water increased with addition of surfactant up to a limit. The presence of sea salt reduced the flotation rate. Simple mathematical models of the flotation cell are developed for predicting the flotation rate from first principles. The agreement between predicted and measured flotation rates is ,quite good.
Introduction Oil-water separation techniques have gained increasing attention in recent years because of the need for treating oilpolluted waste waters. Oil-water emulsions are found in waste water effluent streams from many sources, including petroleum refineries (Steck, 1966; Quigley, 1966), the discharge of bilge and ballast water from ships (Hernandez et al., 1966), washrack and hanger waste waters, rolling mills, and chemical processing and manufacturing plants. Many commercial techniques and devices are being developed and marketed for the removal of emulsified oil from water, but a single, economical, and efficient method is not yet available (Wang et al., 1975). Flotation, one of the adsorptive bubble separation techniques (Karger et al., 1967; Lemlich, 1972), being basically a large-scale process offers a great deal of potential for the treatment of sewage and industrial wastes (Grieves, 1970; Boyd and Shell, 1972; Jenkins et al., 1972). Air flotation applied to the separation of emulsified oil from water involves the injection of fine air bubbles on which oil agglomerates or oil drops adhere in the presence of a surface-active agent (or surfactant) and are removed by rising to the water surface and being trapped into the resulting foam which is subsequently removed ( E f f l u e n t W a t e r Treat. J., 1969; Rohlich, 1954; Barry, 1951). In dispersed-air flotation, air bubbles are generated either by electrolysis or by forcing air through a spinnerette, porous glass frit, single orifice, or other suitable sparger. Small bubbles of diameters less than 0.1 mm can be produced electrolytically but the bubble size distributions obtained in this way are not reproducible (Reay and Ratcliff, 1975). The size distributions of bubbles produced by spargers are governed mainly by the size distribution of the sparger pores and the air flow rate (Shah and Lemlich, 1970). Small bubbles with reproducible size distribution can be generated on a porous 436
Ind. Eng. Chem., Process Des. Dev., Vol. 16, No. 4, 1977
glass frit by adding up to 0.5 vol% of a frother (a short chain aliphatic alcohol) to the solution (Zieminski et al., 1967; Reay and Ratcliff, 1975), although for economic reasons such alcohol addition would not be practicable on the large scale. Effluent treatment flotation plants tend to use small bubbles often of diameter less than 0.1 mm (Reay and Ratcliff, 1973) which distinguishes this process from the froth flotation process used for many years in mineral dressing. A theoretical analysis exploring the effect of bubble size on flotation rate in dispersed-air operation, supported by experimental confirmation (Reay and Ratcliff, 1973, 1975), has shown that for small bubbles of diameter up to 0.1 mm around which Stokes flow can be applied, the flotation rate should, to a first approximation, be proportional to bubble frequency and be independent of bubble diameter over a range of particle sizes (up to 20 Fm in diameter) often encountered in effluent streams. Therefore for a given air rate the bubbles should be made as small as possible in order to maximize their frequency. Of course, this must be balanced against the extra cost of generating smaller bubbles in the case of spargers and porous plates, the method of electrolysis being left to date as the main alternative for industrial applications (Kuhn, 1974; W a t e r W a s t e Treat. J., 1968).Because of the dependence of the flotation rate on the bubble size, it is clear that a need exists to produce small gas bubbles of controllable and predetermined size, by an economical process. Most of the work on effluent flotation has dealt with the chemical aspects of the process and very little work has been done on the physics of bubble-particle attachment. Collins and Jameson (1976,1977) recently conducted an experimental investigation on the physical variables that control the kinetics of the flotation of fine particles by small bubbles of diameters less than 0.1 mm, stimulated in part by a suggestion made by Reay and Ratcliff (1975),and showed that the rate of flotation depends strongly on the double layer repulsion between par-
ticle and bubble, the flotation rate falling rapidly as the double layer repulsion increased. The motivation of the present study was to find conditions under which it is possible to float particles of emulsified oil in dilute suspensions, at a reasonable removal rate. In existing processes, flotation is generally preceded by a flocculation step in which particles are first coagulated to form larger aggregates, which are then removed by sedimentation. However, flocculation is a slow procedure and relatively long holding times are necessary. Also, the cheapest and often most effective coagulants are salts of aluminum and iron, and the bulky sludges they produce are often difficult to dispose of. Our aim is to produce a flotation system which can remove emulsified oil particles without the need for a flocculation stage, and using minimum retention times. In the present work, the kinetics of flotation of emulsified oil particles in water has been studied in a batch flotation cell with air bubbles being generated by a dispersion mechanism. The effects of surfactant concentration and of the addition of dissolved salts have been measured. Simple theoretical models are developed to predict the flotation rate constant. Materials Two oils were used. (1)The first was a “soluble” cutting oil from a commercial source which emulsifies spontaneously in water to form a white fluid of milky appearance. It contains 80% solvent-refined mineral oil, the balance being composed of sodium naphthenate emulsifier, polyethylene glycol monooleate, oleyl alcohol, bactericides, etc. (2) The second oil was white spirit, a turpentine substitute wholly a petroleum product. It complies with British Standard BS 245, which specifies the following distillation figures: not more than lo?? by weight to boil a t less than 155 “C; not less than 90% below 195 “C; final boiling point not to exceed 210 “C. The product corresponds to a heavy naphtha. The soluble oil emulsion was prepared by injection of the required volume of oil into a specified volume of London tap water with gentle agitation. In the case of the white spirit, the measured quantity was fed to the high shear region of a high-intensity laboratory disperser containing the tap water to be contaminated, and the whole was left to mix for 1 h. During mixing a considerable quantity of heat was generated so the vessel containing the oil and water was kept in a water bath at constant temperature (23 “C). The white-spirit particle size distribution was determined photographically, using a Vickers projection microscope with a X600 objective. After projection of the photograph on a screen a total magnification of XlOOO was obtained and the particle diameters would be measured. The counting procedure gave a histogram of particle frequency f ( d p ) against particle diameter d, shown in Figure 1,giving a mean particle diameter of 3.2 fim and a standard deviation of 0.85 pm. Hexadecylbenzyldimethylammonium chloride (HBDA-C1) and cetyltrimethylammonium bromide (CTAB), shpplied by B. D. H. Chemicals Ltd. as general laboratory reagents, employed in the experiments for the removal of cutting oil and turpentine substitute oil respectively, each added to function as collector and frother. Octylphenol ethylene oxide condensate (Nonidet P.42, Shell Chemicals Ltd.), a nonionic surfactant, was also added as a frother in some of the flotation experiments for the separation of emulsified turpentine substitute oil from sea salt and pure salt solutions. The surfactants were freshly prepared for each experiment by dissolving in warm water. Sea salt solutions containing 4% salt were made up by dissolving the required quantity of dried sea salt crystals (Tidman’s Sea Salt (from Spain); Harrods Ltd., London) in tap water. These sea salt crystals were not in any way purified and
100
80
-
a V
-?a 6 0 -
TJ
L
LO
-
20 -
dp
(Am)
Figure 1. Size distribution of oil particles. Mean particle diameter = 3.2 pm.
c Figure 2. Flow scheme of the flotation system. The liquid in the flotation column (F) overflows a t a weir (A) and is pumped through a rotameter (C) and bubble generator (E).Foam is collected in the vessel G. The pressure in the pump delivery line is measured a t D.
thus contained all impurities such as calcium salts, soluble organic matter, etc. Pure salt solutions containing 4% chloride salt were made up by dissolving pure 99.9% Analar sodium chloride in tap water. Experiments A batchwise dispersed-air flotation system was employed with external, total recycle and cocurrent flow of air and oilwater emulsion. I t consisted essentially of a vertical glass column of height 151 cm and diameter 10.8 cm with provision a t the bottom for air-liquid entry and a t the top for liquid recycle and foam removal; a container to receive the recycled liquid and to smooth out any surges in recycle flow rate caused by excess foaming at the head of the column; and a mixing device equipped with two capillaries to introduce the air in a dispersed form. By pumping the recycled liquid from the container through the mixing device to the bottom of the column, air was also sucked through the capillaries and mixed with the flowing liquid in the form of small bubbles. The recycled liquid flowed out of the top of the column through an adjustable lute which was used to control the position of the foam-liquid interface so as not to allow recirculation of oilenriched foam. The flow scheme of the system is shown in Figure 2. The total volume of the system was 16 L. Air and recycled liquid flow rates were 12.5 and 30 cm3/s, respectively. The pH of the prepared emulsion was measured and was in the range 7.5 to 8.0 for all cases. Three flotation runs were performed with cutting oil emulsions, at concentrations between 50 and 100 mg/L, using HBDA-C1 as the cationic surfactant a t a concentration of 200 mg/L. In the course of a run, the surfactant is partially removed by the air bubbles, and in a batch process, one would expect the concentration of surfactant to fall with time. T o Ind. Eng. Chem., Process Des. Dev., Vol. 16,No. 4, 1977
437
1600-
. T 0
D
z
1200-
0 0
01
02
03
04
05
06
db l r n m l
Figure 3. Bubble size distribution. Mean bubble diameter
= 0.147
mm.
Figure 5. Effect of surfactant concentration on flotation rate of white spirit. HTA-Br added half prior to flotation and half continuously during the run: 0 , 1 2 . 5 mg/L CTAB; 0 , 2 5 mg/L CTAB; A , 50 mg/L CTAB; m, 100 mg/L CTAB.
0
5
10
15
20
25
time ( r n i n )
Figure 4. Effect of method of addition of surfactant on flotation: A, 200 mg/L HBDA-CI added prior to flotation; K = 0.006 min-I; 0 , 2 0 0 mg/L HBDA-CI added continuously; K = 0.03 min-'; 0 , 1 0 0 mg/L HBDA-CI added prior to flotation; 100 mg/L HBDA-C1 added continuously; K = 0.013 min-'.
test if this was important in the time scale of these experiments, three different methods of adding surfactant were used: (a) HBDA-Cl solution was added continuously during the first 20 min of the run; (b) HBDA-Cl solution was added, in a single dose into the emulsion prior to the start of flotation (Le., prior to the transfer of emulsion to the column); ( c ) HBDA-C1 solution was added half into the emulsion prior to the start of flotation and half continuously during the first 20 min of the run. The rest of the flotation runs were performed to separate emulsified white spirit from water, sea salt, and pure salt solutions, using CTAB as the cationic surfactant. In the runs with sea salt and pure salt emulsions, Nonidet P.42 was also added as a frother apart from CTAB. The initial oil concentrations for the runs ranged from 40 to 87 mg/L and the CTAB and Nonidet P.42 concentrations employed in the system ranged from 25 to 100 mg/L and from 25 to 50 mg/L, respectively. The prescribed quantities for any particular run of the CTAB solution and Nonidet P.42 condensate were added half into the emulsion prior to the start of flotation and half continuously over the duration of the run. For each run samples were taken at the start of flotation (at reference time = 0) and a t specified time intervals thereafter, from the recycle exit at the top of the column. Samples were analyzed for oil concentration by measuring their optical densities in a Hilger Spekker absorptiometer which had previously been calibrated using reference emulsions of known oil concentrations. The diameter of the bubbles generated in the system was obtained by taking photographs of the foam just after the exit 438
Ind. Eng. Chem., Process Des. Dev., Vol. 16, No. 4, 1977
from the mixing device. To avoid distortions due to the curved wall of the glass tubing, the foam was played against a flat glass plate and photographs were taken from the opposite side. The foam was fairly dense, approximately two-thirds water by volume, and there was insufficient time for segregation of the bubbles to occur. By measuring bubble diameters within a sector of a photograph taken with a surfactant concentration of 50 ppm CTAB, the histogram shown in Figure 3 of bubble frequency f ( d b ) against bubble diameter d b was constructed. The calculated arithmetic mean bubble diameter was 0.147 mm with a standard deviation of 0.122 mm. The low value obtained for the standard deviation implies small dispersion of the diameter values, Le., good bubble size uniformity.
Experimental Results The results of the flotation runs with the cutting oil emulsions are shown in Figure 4. The points lie on reasonably straight lines indicating that the removal follows first-order kinetics of the form
where Cp is the concentration of oil particles at time t and K is the rate constant. First-order kinetics with very small particles have also been found recently by Reay and Ratcliff (1975) and Collins and Jameson (1976). It is evident that the highest rate constant, and therefore the fastest removal rate, was obtaiped when the surfactant was added continuously during a run. In the experiments with white spirit, first-order kinetics were again observed as in Figure 5 , and the rate constants obtained by least-squares regression are shown in Figure 6. A t higher concentrations of surfactant the rate constants appear to approach a limiting value. The results of the runs performed to separate emulsified white spirit from sea salt solutions are shown in Figure 7. In analyzing for the oil concentrations of the samples taken during the runs, the true values of concentrations of oil could not be obtained due to the presence of solid or colloidal impurities from the sea salt crystals. However, plots of oil concentration against time are given in Figure 7 to show the trend in the operation. During the runs it was observed that the presence of sea salt in the emulsion decreased the foaming
0
0
0
20 CTAB
1
I
I
LO
60
eo
concentration
I
IW (mgili
Figure 6. Rate constant of flotation of white spirit a t various surfactant concentrations.
power of the surfactant, and the impurities seemed to be separated before the oil. The use of a nonionic surfactant (Nonidet P.42) instead of the cationic CTAB did not improve the process; indeed it had a markedly worsening effect on it as it can be seen from Figure 7. This underlines the importance of the surfactant charge in the flotation mechanism. A 1:l mixture by weight of the cationic and nonionic surfactants gave better results since the charge effect of the former and the foaming power of the latter were brought together in the process. In the runs employing emulsions of white spirit in pure salt solutions, a true representation of concentrations of oil against time could be obtained since impurities in the form of suspended solids were no longer present. The results are plotted in Figure 8 as the log of oil concentrations against time and show that the addition of a mixture of 1:l by weight of CTAB/Nonidet P.42 gave little removal (34% in 16 min) although much foaming. When the amount of Nonidet was decreased to a ratio of 3:l CTABINonidet P.42 the removal of oil was slightly worse (31.3%in 16 min) and the foaming still excessive. The addition of pure CTAB resulted in a slightly better removal (35.7% oil removal in 16 min). The straight lines in Figure 8, obtained by a least squares regression on the data points, indicate that the flotation kinetics of oil removal from salt water is also first order. The results show several interesting aspects. In Figure 4 it is seen that the removal rate of cutting oil was largest when surfactant was added continuously into the recycle stream, which also carried the injected bubbles, during a run. Thus at the beginning of flotation the concentration of surfactant in the vertical cell was essentially zero. Part of the added surfactant would already be adsorbed on the fresh bubbles in the injected stream, and the excess would be available to neutralize the charge on the particles. The results suggest that the surface condition of the bubbles is a t least as important as the condition of the particles, and that the surface concentration of adsorbed ions, the charge per unit area, and the orientation of adsorbed species could all have a bearing on the flotation rate. In the presence of large concentrations of NaC1, the flotation of the white spirit was depressed, when qualitative judgment would have suggested that it ought to be enhanced. Thus one would have expected that with a large concentration of ions, the double layers around the bubble and the oil particles would be “collapsed”, and the effective charge of both would be zero. The oil is naturally hydrophobic, and in the absence of charge it could be expected to float more readily than when bubbles and particles possessed the same sign. The fact that the rate of removal was actually slower suggests that charge effects remained important. It is worth noting in this context that Collins and Jameson (1977) have measured the
-
0
A lot 01 0
A
IO
20
30
time I m i n I
Figure 7. Flotation of white spirit in sea water a t various surfactant concentrations: .,50 mg/L CTAB; 0 , 5 0 mg/L Nonidet P.42; A. 25 mg/L CTAB; 25 mg/L Nonidet P.42.
0 c
0 L
0 0
34
_I
32
-
\\
30 0
10
20 time ( m i n l
Figure 8. Flotation of white spirit in pure salt solution (4% NaCl by weight) a t various surfactant concentrations: O , 5 0 mg/L of CTAB;
K = 0.023 min-I; A ,25 mg/L of CTAB, 25 mg/L of Nonidet P.42; K = 0.023 min-I; 0,37.5 mg/L of CTAB, 12.5 mg/L of Nonidet P.42; K = 0.023 min-’.
mobility of small bubbles in surfactant solutions similar to those used here, and found that the charge corresponded roughly to that of the oil droplets we have studied, in the range 3.9 to 5.0 pmlslVlcm.
Theoretical It is customary in the field to treat flotation as a first-order rate process by analogy with chemical kinetics. Thus the process is assumed to follow an equation like (l),defining a rate constant K which is easily derived from experimental data. However, this view of flotation rates is really quite misleading, since the so-called “rate constant” is not a constant at all, but depends on a number of variables not the least important being the gas flow rate. In fact, the flotation of very small particles has much more in common with mass transfer operations in which the rate of transfer of particles is proInd. Eng. Chem., Process Des. Dev., Vol. 16, No. 4, 1977
439
portional to a mass transfer coefficient h, defined by the equation Np
= hp(Cp - Cps)Ab
(2)
where N, is the mass flow rate (kg/s) of particles to the surface of the bubble, whose area is A b ; C, is the concentration of particles (kg/m3) in the bulk liquid and C,, is the limiting concentration of particles in the liquid layer adjacent to the surface of the bubble. We assume that in our work with dilute systems, the effective value of C, is zero, Le., that as soon as a particle nears the surface it breaks through and is retained in it, and the number accumulating on the surface of the bubble is so low that further particle captures are unhindered. We have no way of characterizing the degree of backmixing in our flotation cell, so for simplicity we shall assume that it is well mixed so that the particle concentration C, is uniform throughout the cell. This is not unreasonable, because liquid from the base of the cell will be entrained in the wakes of rising bubbles, with corresponding replacement by fluid from the top. Further, our recycle system took liquid from the top of the cell and replaced it a t the base with added air bubbles. With these assumptions, the total rate of removal of particles by bubbles is the product of the rate of removal per bubble, the residence time of a single bubble, and the number of bubbles generated per unit time, thus
where nb is the number of bubbles per unit volume. Across any cross section in the cell, continuity of gas flow requires that n b f d b 3 A ( U f + ub) = Q 6 Thus from (7) and (8) we obtain
or
From our data we have calculated the rate constant K using the two different approaches. The data relevant to the equipment are: d, = mean particle diameter = 3.2 X 10-6 m; m; A = column area db = mean bubble diameter = 147 X m2; Ub = terminal velocity of bubble = 2.5 X = 91.56 X m/s; Uf = upward superficial velocity of recycle fluid = 4.64 X m/s; and Q = gas flow rate = 12.5 X m3/s. The bubble size distribution given in Figure 3 was measured for a CTAB concentration of 50 mg/L, for which the measured s-l. We shall first rate constant is, from Figure 6,2.7 X use the diffusional mass transfer coefficient approach and calculate k , from standard relations. Although the particles are rather too large to be truly colloidal, we try the singlesphere correlation of the form (Bird et al., 1962)
(3) where A is the cross-sectional area of the column, h is the depth of liquid, Q is the volumetric gas flow rate, Vb is the bubble volume, U b is the terminal velocity of the bubble, and U f is the upward superficial velocity of the fluid (liquid gas) in the column because of the recycle stream. On substitution in (2) and integration we find
+
(4) where CO , is the initial concentration of oil emulsion a t t = 0. In comparing (3) and (1)we see that the rate constant K is
Sh = 2.0
G
= (Pp
- Pf)dp2g/18c(Ub
(6)
In addition, k , will be dependent on the charge on the particles and on the bubbles and also on the relative orders of magnitude of the intermolecular forces between water and oil, and water and itself (Prieve and Ruckenstein, 1974). An alternative to the “mass-transfer” approach is t o treat the removal of the particles by the bubbles through the concept of “collection”. Thus the bubble is assumed to rise through the flotation column, and some of the particles which initially lie in the volume swept by the bubble will collide with and adhere to it. The collection efficiency E, may be defined (Reay and Ratcliff, 1973) as the fraction of the particles in the path of the bubble which collide and remain with it. The rate of removal of particles in the liquid in the cell is then -hA 440
dC = C p i db2U&,nbhA 4 dt
Ind. Eng. Chem., Process Des. Dev., Vol. 16, No. 4, 1977
(7)
+ 0.6Re1/2Sc1I3
(11)
The particle diffusivity is found by the Stokes-Einstein equation to be
a=--kT
- 1.36 X m2/s 3 ~ d p giving a Schmidt number Sc = 7.34 X lo6. The bubble Reynolds number is 1.62, and the mass transfer coefficient is found to be 1.39 X 10-7 m/s. From (5), the calculated rate constant is therefore K = 1.1 X s-l. We now use the collection model to estimate K . Reay and Ratcliff (1973) showed by a simple hydrodynamic model that the collection efficiency E, is E , = (1 + :)2
The effective bubble mass transfer coefficient h, accounts of the capture of particles by the bubble, by a variety of mechanisms. For particles of the size considered here (