Removing particles in water and wastewater First of a six-part series on wastewater treatment processes
1 Desmond E Lawler University of Texas at Austin Austin, Ta. 78712 In conventional water and wastewater treatment plants, most pollutants are removed either as particles or are attached to particles before they are removed. The objectives of this article are to present the current understanding of some fundamental phenomena that influence particle removal processes in water and wastewater treatment and to show how these phenomena are, or, in some cases, are not accounted for in current engineering of treatment systems. The normal methcds for treatment of surface waters for drinking or for use in industrial processes are coagulation (chemical addition, rapid mixing, and flocculation), sedimentation, and fiitration, as shown in Figure 1. In addition to the particles naturally present, particles often are created during precipitation to remove soluble constituents, such as calcium and magnesium, or to improve flocculation by adding alumnum or iron hydroxides. Other processes, such as chemical oxidation, that have nothing to do with the particles
856 Environ. Sci. Technol., vol. 20,No. 9, 1986
design is dictated primarily by their ability to remove particles. In wastewater treatment plants, the removal of pollutants as particles is equally important, as Figure 2 shows. Primary sedimentation, secondary sedimentation, and dewatering are direct solid-liquid separation processes. In addition, the objective of biological treatment can be viewed as the conversion of primarily soluble organic pollutants into microorganisms (particles) for subsequent removal. Op tional tertiary treatment processes can include the additional removal of particles by Filtration, the precipitation of phosphate and subsequent solid-liquid separation, and the nitrification and denitrification processes with subsequent sedimentation of sludge. Physics and chemistry are important
experimental results from Wirojanagud (1). The effect of initial concentration and surface charge on the removal of suspended solids by continuous-flow sedimentation at.steady state is illustrated in Figure 3. In separate experiments by Wirojanagud, four suspensions of the same material were put into a compartmentalized, laboratory scale, continuous-flow reactor at the same flow rate. At steadystate conditions, the removal of particle mass was greater for higher concentration suspensions and for suspensions at pH 2.5 (the zero point of charge for these particles) than it was for those at pH 3.5. The differences resulted from flocculation that occurred simultaneously
-7
co159!36WW09200856$01.~/0 0 1986 American Chemicfd Society
flocculation (and therefore of sedimentation) was influenced by the particle concentration and by the surface charge, which were controlled in these experiments by pH. The concentration effect is due to physics, the result of a greater opportunity for particle-particle collisions at higher concentration; the pH effect is the result of surface chemistry. OMelia has characterized the pbysical and chemical aspects of particle interactions as particle transport and attachment, respectively (2). Particles are transported and brought into contact with one another by physical processes and can become attached to one another as a result of the surface chemistry of the particles and the chemistry of the surrounding liquid. Although both the physical and the chemical aspects of particle interactions are important, the focus of this article is on the physical characteristics of particles and their effect on removal. Chemistry is important in all solidliquid separation processes. Organic polymers are added just prior to fdtration to improve particle removal. Habibian and O’Melia (3) and Glaser and Edzwald (4) have presented dramatic evidence of chemical effects in fdtration. Similar chemical additions are made before dewatering.
Flocculation Flocculation can be used to illustrate some of the physical issues to be considered in particle interactions. It is perhaps the simplest process used in particle removal because it focuses only on growth in particle size. In the analysis of flocculation, settling is commonly ignored and precipitation is usually considered complete prior to flocculation. Also, with flocculation there are no complications of history, such as
the effect of previously retained particles on additional retention and pressure development in filtration. In a classic work, Smoluchowski described the mathematics of flocculation: how the number concentration of each size of a particle size distribution changes over time in a batch system in which particle collisions result in attachment (5).In discrete form, the Smoluchowski equation can be written as follows:
C
an,
, Z P(i,k)ni
1 In Equation 1, i, j, and k are subscripts denoting particle size, and c is the maximum allowable value of i, j, or k; n is the particle number concentration; f is time; B(i,j) is the collision frequency function taken as the sum of separate functions for different modes of interparticle contact; and a is the collision efficiency factor (the fraction of predicted collisions that result in attachment). The first term on the right side of Equation 1 describes the rate of formation of size k particles by the flocculation of (two) smaller particles. The second term describes the loss of size k particles by their flocculation with particles of any sue to form larger ones. Particles in water are subject to several forces that influence their transport and cause collisions. The thermal activity of molecules in solution leads to Brownian motion of the particles, fluid motion brought about by external forces causes particle motion, and gravity results in a downward flux of particles. Separate collision frequency functions for interparticle contacts I=
brought about by Brownian motion, fluid motion, and differential sedimentation have been developed and are expressed, respectively, as follows:
p(‘. bJ) =
(2)
B(i,j) = 1 (di 6
+ 4)3G
(3
(di + 4)3Idi - 41
In Equations 2-4, di and 4 are the diameters of pFticles of sizes i and j , reSpeCtively; k is Boltzmann’s constant; T i s absolute temperature; p is absolute fluid viscosity; G is the rootmean-square velocity gradient; g is the gravitational constant; and pp and pi are the densities of the particles and liquid, respectively. For turbulent conditions, the velocity gradient is often replaced by (e X p,lL)‘”, where e is the energy dissipation per unit mass. It is worthwhile to consider the level of complexity that Equations 1-4 represent. Designers of flocculation tanks apparently believe that these equations are too complex; no textbook commonly used by environmental engineers describes a design approach based on them. The usual approach is to consider only the ramifications of a simple subset of the above equations, one obtained by considering all particles to be the same size. In that case, the two terms on the right-hand side of Equation 1 collapse into one term, and the collision frequency function for differential sedimentation goes to zero. The particles often are considered large enough that contact of particles by Brownian mo-
Environ. Sci. Tmhnol., Mi. 20,No. 9, 1986 857
tion is negligible in comparison with The projected surface area is the area of the contact of particles by fluid motion. a circle with radius ai + aj: Unfortunately, the simplifications asA = T(aj a# (6) sumed by designers are not well founded in reality. Lawler and Wilkes measured particle size distributions The collision frequency function for from an operating flocculationtank and differential sedimentation is therefore showed significant particle number derived as follows: concentrations and particle volume (7) M j ) =2r@,-dg concentrations spanning the diameter 9P range from 1 pm (the lower l i t of detection in their work) to more than 40 pm (6). From the trend of increasExpressing Equation I in terms of ing particle number concentration with decreasing sue just above the detection diameter instead of radius, and factorlimit, it is reawnable to assume that ing the last term yields Equation 4. One many more particles were smaller than important liitation of thisderivation is 1 pm and that the particle sues spanned that it ignores the influence of the momore than two orders of magnitude. tion of one particle on that of the other. Because the equations for the collision It assumes that if the two particles are frequency functions for fluid shear and heading toward one another at some diff&nth sedimentation are strong point. they will continue on those trafunctions of the differences in the m i - lectones until they collide. This decle diameters (asymptotic to the-third icription is a p i r representation of and fourth power, respectively), this truth. A large particle moving downtwo-order-of-magnitude difference in ward toward a small particle pushes size can lead to a much larger differ- water out of its path, and the water pushes the small particle as the distance ence in collision frequency. Equations 1 4 are apparently too between them becomes small. Collicomplex from a design standpoint; they sions between large particles and small are overly simplified from a fundamen- particles are less likely than those that tal or microscopic view of particle in- can be predicted by geometric considteractions. A conceptual derivation of erations alone. The hydrodynamic considerations the collision frequency function for differential sedimentation makes this omitted from the derivation above are clear. Consider a particle of radius a, likewise omitted in the collision fre and the origin of a coordinate system at quency functions for Brownian motion the center of that particle. Further- and fluid shear. Their derivations are more, consider a particle of (larger) ra- shown by Friedlander (3.Because the dius a,. If the two particles touch, their hydrodynamic interactions tend to reapart. tard collisions, Equations 2 4 r e p a n t centers are a distance a, Hence, a sphere with a radius o a, + a, overestimates of the actual collision defines the volume of interaction be- rate. The 01 term in Equation 1 can actween the two particles. The collision count for this overestimation in part but frequency function is the product of the must also account for other forces, such relative velocity of particles of size j as van der Waals attraction and electrowith respect to particles of size i and static repulsion, which are ignored in the horizontal area defined by the pro- the derivation. The hydrodynamics of jected surface area of the top half of the collisions between qua-sized particles sphere. The relative velocity is the dif- in fluid shear have been studied both ference in the Stokes's settling veloci- mathematically and experimentally for ties of the particles: and co-workmany years by Mason (8) ers. Subsequent work by Adler extended the research to include a variety of particle sues (9,la). The hydrody-
+
.
+2
1158 Envimn. Sci. Technol., Vol. 20. No. 9, 1986
namic interactions of particles in Brownian motion were investigated by Spielman (11). The goal of treatment process modeling using a particle approach is to predict the performance of the process on the basis of fundamental characteristics of the particles and on the basis of the engineering involved. As that goal is approached, modeling for the design and operation of real systems will become not merely useful but indispensable. There is now a gap between the modeling of flocculation and its utility in real systems. Existing models are too simple to reflect the reality from a microscopic viewpoint and too complex (or perhaps not yet accurate enough) to use routinely in design. Continued progress is needed to narrow that gap. In the following sections, the current state of modeling and design for several other treatment processes is reviewed, although in somewhat less detail than for flocculation.
Sedimentation The approach to modeling the sedimentation of dilute suspensions is similar to that for flocculation. It involves the same terms as those shown in Equations 1-4, and it incorporates additional terms to express sedimentation, V i ally all sedimentation in environmental systems is flocculent; particles flocculate as they settle, as demonstrated in Figure 3. Vpically, the concentration, size, and density of the particles are sufficiently low that St?kes's law applies (Equation 5). In modeling, the reactor can be segmented concephdly into several layers. An equation similar to Equation 1 is written for every particle size in each layer, with two additional terms; one expresses the settling into the layer under consideration from the layer above, and the second expresses how the particle moves into the layer below. For batch or idealized plug flow conditions, a numerical integration of the described set of differential equations is sufficient. For more complicated bydraulic configurations, a ffite-element approach has been taken by Valioulis
and List (12). Such modeliig is reasonably well advanced and shows that influent particle size disaibution, concentration, and density, as well as design considerations such as overtlow rate, detention time, and the depth of the tank can have a significant effect on particle removal. For example, the modeling of sedimentation with this approach shows that removal efficiency decreases with decreasing influent concentration; this result is now reflected in modern practice, in which no sedimentation tank is included between the flocculation and fiitration steps in the treatment of lowturbidity waters. Sedimentation is not only unnecessary but ineffective. wically, removal efficiency in sedimentation tanks is measured in terms of mass removal. For constant density with particle size, the particle volume distribution is also a measure of the distribution of mass with size. Wilkes obtained measurements of particle size distributions in an operating sedimentation tank at a water-softening (CaC03 precipitation by lime) treatment plant (13). For example, the volume distributions from sedimentation tanks at the Davis Water lkeatment Plant in Austin, Ex., are shown in Figure 4. The area under each curve represents the total particle volume concentration; from the reduction in that area, it is obvious that sedimentation occurred through the tank (from Location 1 to Location 3). The increase in particle volume in the smallest sizes shown is the result of flocculation from smaller particles. Improvements to this sedimentation tank would be difficult: Most of the particle mass near the f l u e n t (Location 1) is in the size range of 20 pm < 4 < 60 pm (1.3 < log 4 < 1A), but most of the particle mass near the effluent (Location 3) is in the size range of 7.9 pm < dp < 15.8 pm (0.90< log < 1.20). Nocculeit himentition selectively removes the large particles by direct sedimentation. The small particles are removed by flocculation; intermediatesized particles are lefi behind. Demands for improved removal efficiency become difficult to meet because the remaining particles are so small that their sedimentation velocities are slow. Design of sedimentation tanks is primarily based on experience. The size, density, and concenmtion of particles affect mass removal and hence are reflected in that experience. However, these particle characteristics are not directly recognized or accounted for in typical design practice.
a
Thickening At sufficiently high concentrations, the sedimentation of batch suspensions
exhibits a bewildering phenomenon: A well-defined solid-liquid interface forms, and the particles appear to settle en masse. The minimum concentration necessary for this phenomenon to occur is a function (as yet unknown) of the particle size distribution. It is possible that this minimum might range from as little as 1 glL for activated sludge to several hundred grams per liter for minerals with small panicles. Conceptually, all of the phenomena expressed by the Smoluchowski and Stokes equations (Equations 1 and 5) are important in thickening, but there are two complications. First, Stokes's
law assumes that each particle settles in an infinite medium. This is obviously untrue in such highancentration snspensions; the hydrodynamics are significantly more complicated. There are models that account for complications such as these in the case of nonflocculent, monodisperse suspensions (1416). These models result in a predicted settliig velocity that is the product of the Stokes velocity and a factor (less than unity) that decreases as the suspension concentration increases. Even in this relatively simple case, there is still no agreement about which model is best.
FIGURE4
I
Volume distdbutitm during sedimentathnFb
? b
5
m.0
3--near effluent
E,
-0
21.0
5
I
2
140
d
$z
70
0.0 050
075
1.00
1.25
1.50
175
2.W
Lcg of particle diameter (dpin rm) *\Mume dl61ribUliOns Imm a lank a! the Davis Water Tmatmant Plant Auslm Tex oSedimentatmntank 88 mw&!dsr 53 m x 24 m
Envimn. Sci. Technol.. MI. 20. No. 9, 1986 859
Second, at sufficiently high concentrations, the weight of each particle is partially sup ported by the particles below. In compression, as this p h e nomenon is called, the d i g velocity is hindered further because the driving force is lowered by the support of the other particles. Kos provided considerable insight into this prohlem (17). The design of continuousflow thickeners is based on the limiting solids flux (the maximum mass of solids that can pass through a unit area per unit time). used to increase the driving force for The relationship between the intedace solid-liquid separation in various settling velocity and the concentration equipment designs. Dewatering of sludges involves flow in batch tests must be defmed. The procedure is explained by Dick (18) and through porous media. The Kozeny Vesiliid (19). K e i t h extended the use equation can be used to derive an exof the limiting-flux concept to aid oper- pression for the specific resistance, a common measure of dewaterability, as ation of thickeners (20). For design and operation of treat- shown by Gale (21). The result is ment plants, thickening appears to be = W ( 1 -4 reasonably well understood where there 7. (8) is a hown relationship between settling e3 P p velocity and concentration. The challenge in modeling thickening from a In Quation 8, r. is the specific resistfundamental or particle point of view is ance ( d k g ) , E is the p o d t y , K is the to predict settling velocity as a function Kozeny constant, and S, is the surface of the particle size distribution, thecon- area per unit vohune of particles @mz/ centration, the particle (wet) density, pm3). The specific resistance is related the viscosity of the fluid, and the chem- to the particle density and, through S,, istry of the system. Such a model to particle size and shape. would include the particle interactions Based on their particle size measureof Smoluchowski, the hydrodynamics ments, Knwke and cc-workers have of high-concentration suspensions, a shown that this equation is valid for quantitative explanation for the fact that various water treatment plant sludges most particles a p p t r to seale at the (22.23). Lawler et al. confirmed it for same rate despite their different sizes, wastewater slndges (24).In the creation and some weighted average of the d e and treatment of sludges, the design sired individual settling velocities of and operation of a system that will indifferent-sized particles to predict the crease particle density or reduce sursettling velocity. Although a difficult face area per unit volume of particles task,progress appears possible. (increase the average particle size) will improve the dewaterabiiity by redwing lkW&hg the s p i f i c resistance. Dewatering is the last in the series of Organic polymers often are added to solid-liquid separation processes with increase the rate of solid-liqnid separaincreasing concentration. The object is tion during dewatering. The chemistry to increase the solid content of a sludge of the suspension is changed to increase so that it no longer has fluid properties. the fraction of interparticle contacts This is achieved by the application of that are successful in forming flocs and external forces greater than gravity. thereby to increase the average particle Conceptually, it is possible to substitute size. At high concentrations of sludges, another term for the gravitational con- the freque.ncy of collision is extremely stant in Stokes’s law. For example, in high even without any specific engicentrifugation, the substitution is the neering design to promote contacts; term dR w is the rotational speed, and flocculation is a mnd-order reaction R is the distance from the center of ro- with respect to particle concentration, tation. Vacuum and pressure also are as made clear by Equation 1. WO Envimn. ai.Technol., Mi. 20, No. 9, lS86
The dramatic improvements in dewatering devices over the past 10 years have come primarily from reducing the thickness of the sludge layer to whicb the external force is applied and from incming that force. Because of compression effects (which are identical to those discussed with respect to thickening), a pressure gradient exists in the sludge layer; the smaller the distance over which this gradient exists, the better. For treatment Dlant designers, the questions of whkh kind ana what size of equipment to use are answered empirically during testing with nearly fullscale equipment in conjunction with polymers. This design procedure should be improved to Bccount for the fnndamental particle characteristics and equations that describe flow through
porousmedia. Fillration Filtration is the last process used for solid-liquid separation in drinking-water treatment plants and tertiary wastewater treatment plants (Figure 1) and hence represents the last defense for particle removal. Fdtration is the most complicated process; there are many influential variables and, because of the semicontinuws and semibatcb nature of fdtration, steady-state conditions are never achieved. In addition to time,independent variables include characteristics of the following: the particles (density, surface charge or potential, shape, and size distribution). the water (temperature-and thereby viscosity and density-and chemical characterization), the medium (depth, size, density, and surface characteristics), and the operation (flow rate per unit area, or superficial velocity). The dependent variables are particle removal and pressure development; either of these can determine the length of the fdtex run before backwasbing is
necessary. The modem approach to filtration modeling is to consider the removal efficiency of a single grain of the fdter medium as particles in the flowing water collide with it hecause of their Brownian motion, the fluid motion,or their gravitational settling. The removal in the entire fdter bed is found
by integration from the removal of a single grain. A model by O'Melia and Ali predicts the increase in removal over time as previously retained particles help capture particles from the water, a process called filter ripening (25). Refinement of the model is necessary, especially to account for surface chemistry (and therefore attachment), hydrodynamic retardation, and the heterodisperse nature of the particles in the influent to the filter. Spielman and Fitzpatrick provide the basis for incorporating the hydrodynamic interactions in filtration (26). Filter design has been influenced by the mathematical modeling of filtration, as demonstrated by J. M. Montgomery Engineers (27). Mathematical models are reasonably successful in predicting the sensitivity of filter performance to the large number of variables and therefore are useful aids in design and operation. For example, the current trends of increasing flow rates through filters and increasing the size of the filtration media have been shown by modeling not only to be possible but to allow for greater use of the depth of the filter bed. Nevertheless, the models are not yet accurate enough to be used alone in design, and pilot studies are usually performed.
lkeatment proeess systems Individual treatment processes are combined in series to form overall treatment schemes. A major challenge is to consider systems as a whole to determine and predict the effects of one treatment process on those that follow. For particle removal in water treatment plants, Ramaley et al. modeled a system that included flocculation, sedimentation, and filtration (28). In wastewater treatment, the interactions between the secondary settling tank and the activated sludge process have long been recognized because of the recycling of sludge. The effect of operational variables of the activated sludge process on the microorganisms produced, and therefore on sludge-settling characteristics, has been addressed by Jenkins and co-workers (29). Many questions regarding the interaction of treatment processes remain largely unanswered. What are the effects of flow equalization, now often installed prior to primary sedimentation, on subsequent treatment in wastewater treatment plants? How does digestion affect dewatering? How do the return flows from sludge treatment processes affect the main processes? How does particle removal in primary treatment affect subsequent treatment? The answers to these and similar questions could well have drastic ef-
fects on the design of treatment plants. For example, an answer to the last question might mean that the 65% removal of suspended solids typically achieved in primary sedimentation tanks in municipal wastewater treatment plants is unacceptable; perhaps filtration should be considered at that point in a treatment scheme.
Summary Most pollutants are removed from water and wastewater as particles or attached to particles. Treatment plants normally work well with respect to particle removal. Problems of design (overdesign or insufficient capability) and operation (poor or uneconomical performance) could be alleviated by a p plication of a fundamental understanding of particle treatment. Physics (Brownian motion, fluid m e tion, and gravity) and surface chemistry influence particle transport and attachment and therefore influence particle removal. Improved quantitative understanding of these two aspects of particle interactions is necessary to predict removal in treatment. Particles in water span two to three orders of magnitude in diameter. Traditional methods of design and analysis of treatment processes ignore this variation. For many suspensions most of the particle mass is accounted for by the relatively small number of particles that are larger than 20 pm, even though most of the particles are smaller than 5 pm. Increasing the removal efficiency on a mass basis is difficult b e cause of the size distribution. Concentrations of particles in treatment processes span four orders of magnitude, from effluents of filters ( < 1 mglL) to dewatered sludges (> 10,ooO mg/L). Some phenomena (for example, transport mechanisms and surface chemistry) are important over the entire concentration range; other phenomena (for example, compression) are relevant only at one extreme. Treatment plants are engineered systems. Progress in the design and operation of particle removal processes will come first from the incorporation of the fundamental characteristics of the particles and the science of particle interactions into the engineering of individual processes and second from recognition of the effects of one treatment process on another. References (.I ). Wiroianaeud. Wirojanagud, W.. Ph.D. Dissertation. Univerhy of Teras, Austin. Tex., 1983. University if (2) O'Melia. C. R. In Physicochemical Proce s s e s / ~Wovr ~ Quality Conrrol: Weber. Walter J.. J., Jr.. Ed.: Wile": Wiley: New York. York, 1972: 2, pp. 6l-lO9: 61-109. Chapter 2,'pp.
(3) Habibian. M. T.: O'Melia. C. R. J. Envi,on. Eng. Div. Am. Soc. Civ. Eng. 1975, 101 5. h .7 -.. S. . .., .
(4) Glaser, H. T.: Edzwald, J. K. E n v i r m Sci. Rchnol. 1979, 13. 299-305. ( 5 ) Smoluchawski. M. 2. Phys. Chem. 1917, 92, 129-68. (6) Lawler. D. F.;Wilkes. D. R. 3. Am. Wnru Work.?Arroc. 1984, 76. 90-97. (7) Friedlander, S . K . Smokp. Durr and HUE: Wiley: New York. 1977. (8) Mason, S. G. J. Colloid lnrcrfarr Sci. 1977.58, 275-85. (9) Adler, P M. 1. ColloidSri. 1981, 83, 106I5
(IO) Adler, I? M. J. Colloid lnrrr/orr Sci. 1981,84,461-74. (1 I ) Spielman. L. A. J. Colloid Inrrr/ace Sci. 1970,33, 562-71. (12) Valioulis, 1. A,: List. E. 1. Environ. Sci. Techno/. 1984, 18. 242-47. (13) Wilkes. D. R., Master's Thesis. University ofTeexaa. Austin. Tex., 1983. (14) Brinkman, H. D. Appl. Sci. Res. 1947,
- -
.A..I , 71-?4 . ..
(15) Happel. 1. Am. Insr. Chrm. Eng. J . 1958, 1
In7
,n,
7 , Ill-'.",.
(16) Richardson. 1. F.: Zaki. W. N. Trans. Insr. Chem. Eng. 1954,32, 35-53. (17) Kas. P; Adrian. D. D. J. Environ. Eng. Div. Am. Soc. Civ. E m . 1975. 101. 947-65. for WaferQualirv C"nrro1: Weber, Walter I . ,
Jr.. Ed.: Wiley: New York. 1972; Chapter 12. pp. 533-96. (19) Vesilind, P A. Trearmenr and Disposol of Wasreworer Slud,vs; Ann Arbor Science: Ann Arbor, Mich.. 1979. (20) Keinath. T. M. 1. Warrr Pollur. Conrrol F d 1985.57, 770-76. (21) Gale. R . S. Wnrer Pollur. Comrol 1967. 66, 622-32. (22) Knocke. W. R.: Ghosh. M. M.: Novak, J. T. 1. Environ. E m . Div. Am. Soc. Civ. Eng. 19R0, 106, 363:76. (23) Knocke. W. R.: Wakeland, D. L. J . Am. Worer Works Assoc. 1983, 75, 516-23. (24) Lawlcr, D. F.el al.. accepted for publication in J. W o w Pollur. Conrrol Fed (25) O'Melia, C. R.; Ah, W. P m g . W r e r Rchnol. 1978. IO: 167-82. (26) Spielman, L. A,: Fitzpatrick. 1. A. J . Colloid lnrerface Sci. 1973. 42. 607-23. (27) 1. M. Mohgornery. Inc.'W& Treurmmr Principles and Design: Wiley: New York. ~~
19x5
Desmond E Lawler I nn associare professor in the environmmral and wafer resources engineerin8 Rroup of the Deparrmen1 of Civil Engineering at rhe Universiry of Texas at Ausrin. He has been there since receiving his Ph. D. from rhe Universiry of Norrh Carolina ar Chapel Hill in 1980. Lawler is a member of ACS and has published several research papers concernrng parricle behavior in rreormenr processes. He is a Norional Science Foundarion Presidenrial Young Invesrigaro,: Environ. Sci. Technol..Vol. 20. NO. 9, 1986 861