Wat. Res. Vol. 20, No. 8, pp. 977-986, 1986 Printed in Great Britain
0043-1354/86 $3.00 + 0.00 Pergamon Journals Ltd
PHOSPHORUS REMOVAL BY GRANULAR ACTIVATED ALUMINA H . BRATTEBO 1 and H. ODEGAARD 2
~A. R. Reinertsen Consulting Engineers, Erling Skakkesgt. 25, N-7000 Trondheim and ='The Norwegian Institute of Technology, Div. Hydr. San. Eng., N-7034 Trondheim-NTH, Norway (Received January 1985)
Abstract--Phosphorus removal from wastewater may be carried out by fixed-bed adsorption using activated alumina. In order to prevent unacceptable head-losses coarse-grained alumina must be used. Such systems have been referred to by several authors in literature. The mass transport characteristics of the system has so far, however, not been given a thorough investigation. This study uses the homogeneous surface diffusion model (HSDM) to describe the process as influenced by the system parameters. A sensitivity analysis is presented to optimize the process design for given conditions. The pH, the alumina particle size and the column length are found to be very important parameters determining the column performance. The process is very well suited for designing a beds-in-series system. Key words--wastewater, phosphorus, removal, adsorption, activated alumina, modelling
NOMENCLATURE A= A0 = A~ = Bi = Co = CFop =
cross section area of bed (L 2) constant (dimensionless) constant (dimensionless) Biot number (dimensionless) inlet phosphate concentration (M L -3) operating cost factor = water production/alumina consumption (L L i) Dg = solute distribution parameter (dimensionless) Ds = intraparticle diffusivity (L 2 t-~) DT = constant (dimensionless) F = flow of water (L 3 t i) kj= liquid film transfer coefficient (L t -I) Kr= Freundlich equilibrium constant (L 3 M-I)l'n L = bed length (L) n = Freundlich exponent (dimensionless) Q = adsorption capacity (M/M) R = alumina particle radius (L) u = hydraulic load (L t i) V = bed volume (L 3) t = bed porosity (dimensionless) z = contact time within the bed (t) zmln = the necessary contact time for the mass transfer zone (MTZ) to achieve a constant pattern (t) ZMrz = the necessary contact time for the MTZ to be totally included in the bed length (t) 0 = detention time of the alumina within the bed (t) p = apparent density of alumina particles (M L-3).
ered attractive (Shiao and Akashi, 1977; Vinyard and Bates, 1979), however, such products may not always be sufficiently available. Furthermore, product handling and process operation may favour the use of high quality adsorbents. The use of activated alumina has been suggested by several investigators in literature during the past 15 years. Recent research in the Federal Republic of Germany (Donnert, et al., 1982) describes a very promising and compact process based on an in-series system of stirred reactors containing fine-grained alumina. All the other papers presented consider systems of fixed-bed reactors. Even though no fullscale plants exist today, there is a lot of information on how such a system is influenced by changes in the different physical, chemical and operating parameters. The subject may so far, however, be characterized by inadequate process knowledge, since no thorough study has been devoted to the determination of the adsorption kinetics in the light of accepted mass transfer theory. The scope of this paper is to contribute to such an understanding in interpreting the results from isotherm and column experiments with the help of a mathematical model based on general accepted mass transfer theory.
INTRODUCTION PREVIOUS WORK
There has been an increasing attention towards alternative methods for phosphorus removal in wastewater treatment during the last decade. For several reasons, the use of inorganic adsorbents seems to be a promising alternative. One type of commercial available adsorbent which is very well suited for this application is activated alumina, due to its high capacity and selectivity for phosphates. Low-cost adsorbents like red mad and fly ash are also consid977
The type of activated alumina which is studied in literature is mostly 7-A1203, for which the following affinity series for anion adsorption is found (Schwab and Dattler, 1937): O H - > P O 3- > F - > SO42 > C1- > N O r . The series shows that, besides hydroxyl, sulphate is the anion that is most likely to compete with phosphate for adsorption from wastewater. However, concentrations of 4 8 0 0 m g l t SOl are
H. BRATTEBOand H. ODEGAARD
found not to have an adverse effect on phosphate adsorption from a synthetic secondary wastewater (Ames and Dean, 1970). HCOj-, which is not included in the series above, is known to be preferred only negligibly by alumina. Organic matter present in wastewater are removed by activated alumina (Eberle et aL, 1975; Kummert, 1979; Kummert and Stumm, 1980; Rohmann and Sontheimer, 1978; Davis, 1982; F6rster and Solbach, 1979), but organic matter present in secondary treated wastewater are found only to influence the phosphate adsorption to a negligible extent (Eberle et al., 1980). The phosphate adsorption mechanisms are related to several types of ion exchange and complexation reactions (Rohmann and Sontheimer, 1978; Stumm et al., 1976; Davis and Leckie, 1980; Anderson and Malotky, 1979). The complexity of the adsorption mechanisms is expected to be due to the complexity of the alumina surface, which contains several types of possible adsorption sites (Peri, 1965). In general, phosphate adsorption gives a release of hydroxyl ions to the water, and hence an increase in pH. The adsorption capacity is strongly pH-dependant, i.e. capacity increases for decreasing pH. Below a pH value of about 4 ~ . 5 the capacity decreases due to a higher solubility of alumina. The pH-dependency is both related to the amphoteric properties of the alumina surface and to the polyprotic nature of phosphate, however, several explanations to these phenomena are proposed in literature (Eberle et al., 1980; Hingston et aL, 1972). Above p H = 8 the capacity increases when Ca 2+ and Mg 2+ are present in the wastewater due to precipitation on the alumina surface (Narkis and Meiri, 1981). The adsorption capacity at a given pH may be modelled by the classical Freundlich or Langmuir isotherms. Activated alumina is in fact a rather heterogeneous solid. Therefore, since adsorption at the more active sites is favoured, heats of both monolayer physical, exchange and chemical adsorption might be expected to become significantly less exothermal as surface coverage increases. In this way the assumption of the Langmuir isotherm concerning a constant heat of adsorption independent of surface coverage may not hold for this type of adsorbent. A two-site Langmuir isotherm has been reported to be valid often to microporous adsorbents (Koresh and Softer, 1983), and has given good agreement to a phosphatealumina system (Eberle et al., 1978). For practical engineering purposes, however, a simple Freundlich or Langmuir isotherm may model the capacity fairly well. In evaluating the performance of activated alumina in fixed-bed operations, several important factors have been investigated. Ames and Dean (1970) found that the time until phosphate breakthrough was inversely proportional to the inlet concentration. However, data was not available to determine how the apparent linearity of
this relationship was affected by factors such as flow rate, bed volume, and the considered range of phosphate concentration. Purushothaman and Yue (1970) reported that a decrease in flow rate gave an increase in the volume of water treated until breakthrough. Narkis and Meiri (1981) treated a secondary wastewater with acidic and basic 28-48 mesh activated alumina at a hydraulic load of 3.2 m h - ~. They found no significant difference between the two types of alumina. Breakthrough occurred at about 1000 bed volumes at an inlet concentration of 9.8 mg P 1 ~, giving a relative adsorption of 9.7 mg P g J A1203. Effects of changing loading conditions and changing alumina particle sizes were not considered. For onsite treatment applications, Detweiler (1978) found a relative adsorption of 10.9 mg P g ~ A1203 at a low hydraulic load. Donnert et al. (1978) concluded that hydraulic loads greater than 4.7 m h ~were not to be recommended when using a coarse-grained (1-3 mm) alumina. Relative adsorption was reported to be 10mg P g ~ A1203. Eberle et aL (1980) reported that an 85 day experiment on 0.15-0.3 mm alumina gave a relative adsorption of 7.1 mg P g ~ Al203 at a hydraulic load in the range of 10.5-18.8 m h -~ and an inlet concentration of 1.0 mg P 1-~. Pollutech Pollution Advisory Services Ltd (1976) found that activated alumina was an economically attractive process when effluent concentrations less than 0.5 mg P 1were required. Contact time and inlet concentration were found to be significant factors in the performance of the process. All these results state that there is a significant relationship between the alumina particle size, inlet concentration, pH, contact time and the effluent quality vs time of operation. This corresponds very well with general adsorption theory, as one should expect. MATHEMATICAL MODELLING OF THE ADSORPTION SYSTEM
The results presented in literature indicate that the alumina system behaves very similar to a granular activated carbon (GAC) system. In addition special attention obviously has to be taken concerning the pH, since the adsorption capacity is strongly pHdependent and since the adsorption itself affects pH significantly. The fact that the alumina particle size is a very important parameter indicates that the reaction rate is controlled by intraparticle diffusion. For these reasons the homogeneous surface diffusion model (HSDM) is considered a proper choice in process modelling. This model has lately been used with great success in G A C applications for the removal of organic compounds and humic acids (Thacker et al., 1981, 1983; van Vliet and Weber, 1981; Lee et al., 1983), both for single solute and for multisolute systems. The HSDM model describes the process by a set of nonlinear differential equations. The model is based on the fact that physical liquid film and intraparticle diffusion control the overall
Phosphorus adsorption from wastewater mass transfer. The two diffusion p h e n o m e n a are coupled by the a p p r o p r i a t e a d s o r p t i o n isotherm. The model is presented in detail elsewhere (Crittenden a n d Weber, 1978; Crittenden et al., 1980). In this study the single solute version ( o r t h o - p h o s p h a t e ions) with o r t h o g o n a l collocation routines for numerical solution of the P D E system was used. Even t h o u g h the model is mathematically complex and requires a numerical solution by the use of a computer, only very few parameters have to be determined by l a b o r a t o r y techniques. These are the isotherm c o n s t a n t s and the intraparticle diffusivity, Ds, which describe the a d s o r p t i o n capacity and the intraparticle diffusion properties o f phosphate, respectively. The model also requires a value for the liquid film transfer coefficient, k t, which describes the diffusion properties t h r o u g h a s t a g n a n t liquid film s u r r o u n d i n g the particle surface. This last p a r a m e t e r may, however, be calculated from empirical correlations in literature (Williamson et al., 1963; Gnielinski, 1978). The model is based on the following assumptions: (1) the intraparticle diffusivity does not depend on the p h o s p h a t e c o n c e n t r a t i o n n o r on the particle size, (2) the a l u m i n a particle is a homogeneous ideal sphere a n d (3) there is a local equilibrium at the particle surface represented by the Freundlich isotherm. The isotherm constants may be determined by use of the bottle point technique which is given in several general t e x t b o o k s o n adsorption. In order to app r o a c h equilibrium quickly, it is necessary to g r o u n d a g r a n u l a r a l u m i n a thoroughly. C o n t a c t time during the experiment should be greater t h a n 5 days since reaction is slow (Brattebo, 1983). Due to pH-increase during adsorption, it is necessary to add the required a m o u n t s o f dilute hydrochloric acid several times d u r i n g the experiment (Brattebo, 1983). The intraparticle diffusivity is conventionally experimentally o b t a i n e d from batch slurry agitated reactors ( F u r s a w a a n d Smith, 1973). W h e n studying g r a n u l a r adsorbents this technique is, however, detrimental to the particle size due to erosion a n d particle break-up. New techniques are developed to eliminate this effect (H61zel and Sontheimer, 1981; L a d e n d o r f et al., 1972). Here the a d s o r b e n t is kept in a differential bed within a batch reactor. Alternatively, a m i c r o - c o l u m n technique with a c o n t i n u o u s load of p h o s p h a t e solution may be used (Weber and Liu, 1980; Brattebo, 1983). A last way to determine the intraparticle diffusivity is to predict large scale column breakt h r o u g h numerically using the model with different Ds input. The actual Ds here will be the value giving best fit to a control experiment b r e a k t h r o u g h curve. MATERIALS AND METHODS In order to standardize the adsorption conditions for the different experiments, a synthetic inorganic secondary wastewater was constructed for use in all the experiments: l l 0 m g 1 i Na2CO3 ' 60mgl i KHCO3, 47.2mgl-t K2504, 15 mg I KNH4C 1and 20 mg 1- I HNO3" H3PO 4 and HC1 were
added separately to the water already containing the standard concentrations of inorganic compounds, to give the desired phosphate concentrations and pH values of the liquid solutions. In the isotherm experiments the chemicals were added to distilled water, otherwise tap water was used. The isotherm experiments were conducted in a Julabo SW1 shaking water bath. The granular alumina was ground in a mortar and sieved through a 70#m standard sieve. Prior to use the alumina powder was pH-adjusted in a slurry placed on a magnetic stirrer, until a stable pH was obtained, equal to the one during the following experiment. Afterwards the alumina was settled, dried for 24 h at 105C and stored in closed bottles within a desiccator. Erlenmeyer flasks of 50ml volume were filled with the phosphate solution and the chosen amount of alumina, and placed in the water bath at 20°C for 7 days. The 7 days contact time was selected after an initial experiment to verify the adsorption as a function of time. During the experiment, pH-control was obtained by manual addition of 0.1 M HC1 once every day. After shaking, the suspension was filtered through a 0.45/~m Schleicher and Schull membrane filter prior to photometric determination of the remaining orthophosphate in the filtrate. One selected type of granular activated alumina, 1-3 mm Compalox AN/V 813 (Martinswerk GmbH, F.R.G.) was tested in 4.4 cm dia, 1.5 m long upflow loaded columns. The breakthrough curves were examined as a function of hydraulic loads, influent phosphate concentrations and pH, at different sampling points along the column. The sampling points were located at 0.0, 0.20, 0.55, 0.85, 1.15 and 1.50m column length. The synthetic phosphate concentrated solution was added to tap water batch-wise in a 1.1 m 3 storage tank. From this tank water was pumped to a 12.01. stirred reactor for automatic fine pH-adjustment, using a CFG Prominent Dulcometer PHW 212 F2. Then the water was pumped into head-providing columns connected to the bottom inlet of the alumina columns. All pumps in the system were of the type CFG Prominent membrane pumps. The packing of the columns with granular alumina was carried out by following a standard procedure. After the experiment a constant part of the alumina volume within the bed was dried and weighed. The phosphate analyses were performed by the use of a Hitachi model 10~20 spectrophotometer, following the ascorbic acid method (APHA, 1980). Prior to addition of chemicals, the samples were diluted by use of a Hamilton microlab M diluter. The samples from the isotherm experiments were analysed in triplicate, while the samples from the column experiments were analysed in duplicate. Initial experiments showed that the adsorption rate was strongly dependent on the adsorbent particle size. In the column experiments the alumina was used as received from the manufacturer without any further sieving (1-3 mm dia). To get a more precise measure of the particle size, a statistically adequate number of random particles were photographed. The average particle diameter was determined to be 1.90 mm with a standard deviation of 0.45 mm. Thus the variation in particle size is in fact considerable for such an alumina grade. We would here like to comment that one in fact should pay attention to the variation in particle size when planning experimental procedures. The object should be to carry out the experiments under conditions one might expect in practice. When conducting experiments of a short duration, a large variation in the particle size may cause significant errors in the determination of model parameters. This is due to the fact that smaller particles indeed will carry out most of the adsorption in experiments of a short duration. The dilemma is that the model uses the mean particle size of the lot. However, the experiments referred here were carried out at a duration of several days, and therefore at a time scale close to reality. Consequently, the standard deviation in
H. BRATTEBOand H. ODEOAARD 3'
g -4 log
Fig. 1. Adsorption isotherms for Compalox AN/V at 20°C. • p H =5.0, KF=736.2, n =0.127; O p H = 6 . 0 , KF=714.5, n=0.207; C) pH=7.0, KF=531.7, n=0.208; A p H = 8 . 5 , K~=317.1, n =0.253. Q =KF.C", Q(mmol P kg t A1203)' C(mmot P I l).
particle diameter of 24 percents was regarded to be acceptable. Nevertheless, it is very important to be aware of this problem and use a particle size as correct as possible in modelling. RESULTS AND D I S C U S S I O N
The results from the isotherm experiments are given in Fig. 1, at four different pH values. The adsorption capacity of this adsorbent may very well be expressed by a simple Freundlich's isotherm. The isotherm parameter values listed in Fig. 1 were found by linear regression. One may see that the Freundlich exponent n is far below 1.0, i.e. the adsorption is termed favourable. This implies that the mass transfer zone (MTZ) in a column will move at a constant pattern after it has been established. The capacity is highly pH-dependant, however, the Freundlich ex-
ponent does not seem to be much influenced by the pH. In a given column experiment, the mean pH of the column may be anywhere between the pH-values for which the isotherm parameters have been determined. Consequently, in order to determine the actual operating capacity of the alumina within the bed, interpolation must be carried out. For this purpose the curves of Fig. 2 are presented, which are calculated from the isotherm parameters given in Fig. 1. In Fig. 2 each curve represents one specific liquid equilibrium phosphate concentration, C. The adsorption isotherm at any pH may now be calculated by computing the linear regression line for the logarithmic values of the corresponding five Q-C data sets from the figure. In this way the interpolation technique is simply carried out in the same way as
700 ;..,.. I~m
E •= t2 t~ Q.
Fig. 2. Adsorption capacity at intermediate pH-values. V C = 0.32mmol P 1 ~; C) C = 0.16mmol P 1 t: • C = 0.032 mmol P l-I; • C = 0.016mol P 1-';  C = 0.0032 mmol P I ~.
~, 2.50 E - 1 0 E v
~ 2 . 0 0 E -10
- 1.50 E - l O " @
"~ 1.00 E - I O "
~ .~ 0
phosphate concentration 6"0 (rng P 1-11
Fig. 3. O b s e r v e d i n t r a p a r t i c l e d i f f u s i v i t y vs t h e inlet p h o s phate concentration.
constructing an isotherm based on experimental data. In addition, Fig. 2 provides a good visual picture of the changes in capacity with changing pH. The column experiments are listed in Table 1, together with the Ds of best fit to the breakthrough data o f L = 1.50 m for each run. One may see that the mean Ds, Ds, varies considerably with the inlet phosphate concentration. This fact is also represented in Fig. 3. The phenomena may be due to the heterogeneity of the activated alumina. The alumina surface contains several types of adsorption site, with the ability to adsorb phosphate at different energetic levels. The observed decrease in Ds with decreasing concentration may be due to occupation of the more energetic sites. At higher concentrations also the less energetic sites have to be occupied. It is a fact that as the adsorbed ion diffuses from site to site along the adsorbent surface, the ion is more mobile when less energetic sites are involved. Hence the adsorbed phosphate will be more mobile at higher liquid concentrations, and a higher Ds will be observed. The Ds of best fit is, however, not changed with changing pH in a systematic manner. Thus one may conclude
that pH influences the diffusivity only negligibly, however, capacity is influenced considerably. Some examples of the fit between computed and experimentally observed breakthrough curves are given in Fig. 4. One may see that the model is able to describe the column performance very well. An exception is the deviation at early times of the runs. At these times the pH of the system was considerably higher than the inlet pH, giving a less actual capacity than the one used in the model. This higher pH was due to the original basic nature of the alumina, and to a high hydroxyl release through exchange adsorption. Since the lack of fit at early times of operation is mainly a result of the original state of the alumina, one must conclude that the mathematical model is very good in predicting column breakthrough when the model parameters are known. Thus the model may be used to carry out a sensitivity analysis of the column performance. The object here is to determine to what extent a change in the different free variables responds to the column performance. Such an analysis is carried out based on the following procedure: (1) Consider a continuous countercurrent system. (2) Apply a total column length that gives a contact time within the bed equal to the necessary contact time to incorporate the whole mass transfer zone (MTZ). 17 _~_ "CMT z
L = u "~MTZ/E.
(3) Assume steady-state conditions, i.e. no movement of the MTZ. Since the adsorption is favourable, this should be satisfied by establishing a constant countercurrent flow of alumina. For such a system Fig. 5 illustrates the conditions. The detention time of alumina within the bed is called 0. A somewhat indirect measure of the operating costs may be 1/CFop where the cost factor CFop is defined according to Water production
Co (mg P I ~) u (m h ~)
CF°p = Alumina c o n s u m p t i o n - L (mm 1) (2)
Table 1. A s u m m a r y of the column experiments Run No.
Ds (cm 2 s ~)
I 2 3 4 5 6 7 8
~ 10.0 ~10.0 ~ 10.0 ~ 10.0 ~ 10.0 ~ 10.0 ~ 10.0 ~ 10.0
3.9 6.0 6.0 6.0 9.0 12.0 12.0 12.0
~7.0 ~7.0 ~6.5 ~4.3 ~6.8 ~7.0 ~6.5 ~4.3
1.30E-IO 2.50E-I0 2.40E-10 2.10E-10 3.40E-10 2.40E-10 2.35E-10 2.40E-10
(cm 2 s i)
11 12 13
~2.0 ~2.0 ~2.0
6.0 12.0 18.0
~7.0 ~7.0 ~6.5
9.50E-I1 9.50E-11 9.50E-11
H. BRATTEBOand H. ODEGAARD
( h )
Time ( h )
Time ( h )
Fig. 4. Observed and computed breakthrough curves for Compalox AN/V 813. Computation is based on the respective Ds-values in Table 1. (a) Co=10.0mgP1 ~; u = 6 . 0 m h t; O L = 0 . 2 0 m , Ds=l.4OE-lOcm :s-~; V L = 0 . 5 5 m , Ds=l.8OE-lOcm 2s-t; [ ] L = 0 . 8 5 m , Ds=2.5OE-lOcm 2s-I; A L = I . 1 5 m , Ds=2.5OE-lOcm2s-I; C) L = 1.50m, Ds=2.5OE-lOcm2s -~. (b) Co= 10.0mgPl-t; L = l . 5 m ; V R u n 2 , u = 6 . 0 m h - l ; • R u n 5, u = 9 . 0 m h ~; • R u n 6, u = 1 2 . 0 m h - ' . (c) L = l . 5 m ; u = 1 2 m h - L ; • R u n 7, Co=10.0mgPl t; © R u n 10, C o = 5 . 0 m g P l - I ; V R u n 12, Co=2.0mgPl l; A R u n 15, Co=l.0mgP1 -I.
Phosphorus adsorption from wastewater
F l o w of w a t e r
Flow of a l u m i n a
Fig. 5. Schematic illustration of the steady-state fixed-bed countercurrent system with r = TMTZ.
and calculated for a period of time equal to 0. CFop should of course be as high as possible. The term alumina consumption may express both the a m o u n t of fresh alumina and the a m o u n t of all regeneration chemicals that are consumed during the period of the time equal to 0. In addition, however, other effects like head loss development, pH-adjustment and fouling will contribute to the total operating costs. The following equations are valid for such a system (Hand et al., 1983). rMT Z =
R "(Ao'Bi + AI) Tmin - -
k#(l - E ) k # R ' ( 1 - E)
Ds" Dg" E Dg
p ' Q . ( 1 --E)
Q = KF.C ~.
The parameters A0, A~ and D T are constants depending on the Blot n u m b e r and the Freundlich exponent. For Bi > 10 is A l = 0 (Hand et al., 1983). Consequently, A~ may be set to zero for the type of column conditions involved in this study• By substituting equations 1 and (3)-(7) into equation 2, one obtains the following expression:
The equation is split into three separate terms. The last term includes parameters which are fairly constant for a given type of alumina and a given inlet concentration. The second term includes nonconstant parameters related to the alumina, and these parameters will influence strongly on CFop. One may see that the pH and especially the alumina particle size are important parameters. The type of alumina to be used at a given pH, particle size, and inlet concentration, should be selected by optimizing the calculated product of the last two terms of equation 8. The first term, however, has nothing to do with the alumina characteristics. It includes only
the 0 value which describes the countercurrent flow of alumina. Thus the operating costs are inversely proportional to the chosen value of 0. It is very interesting to observe that the operating costs do not depend on the applied hydraulic load, u. Since capital costs are expected to increase with increasing column volume, equation 2 is rearranged for a further study:
u.O L - CFop.
F.O L .A = V = CFop
Equation l0 shows that the column volume, V, is proportional to the water flow, F. The volume is, however, independent of 0 since this parameter also is included in CFop. For a continuous countercurrent steady-state system with a complete removal of phosphorus, the required column length may be calculated by equation 9. For a Compalox A N / V 813 system where the parameters are: C o = 0.161 m m o l P 1-1 pH = 6.0 u = 6 . 0 m h -l 0 = 25 days Ds = 1.8 x 10-1°cm2s -I k r = 3.0 × 1 0 - 3 c m s -l R = 0.095 cm E = 0.29 p = 1.49 g c m - 3 /(.i=714.5 n = 0.207 DT = 1.08 A0 = 0.28 the required column length is calculated to be L = 27.1 m. For these conditions CFop = 133 m 3 m -3. The required column length is of course unacceptably high. The reason is that the mass transfer is too slow to establish a constant M T Z for a shorter column length, hence it is impossible to obtain steady-state conditions in practice. In order to keep a constant effluent quality, the 0 value must be reduced proportional to the time of operation. This results in a need
H. BRATTEBO and H. ODEGAARD
(b) .>, "0 cO
5°i ~o] 4O
~8 E t-
c) 5O [
Fig. 6. The time of operation of a fixed-bed Compalox AN/V 813 system until the effluent concentration reaches a value equal to 0.5 mg P1 i.(a) C 0 = 5 . 0 m g P l - t ; p H = 7 . 0 ; •u=6mh-I; O u=9mh-~; Vu=12m h ~. (b) C 0 = 5 . 0 m g P l-]; U = 9 . 0 m h ~; A p H = 5 . 0 ; • p H = 6 . 0 ; OpH=7.0; • pH = 8.5. (c) u = 9 . 0 m h - ' ; pH = 7.0; ~ 1 ppm; • 2ppm; O 5ppm; • 10ppm.
Phosphorus adsorption from wastewater for more frequent regenerations as time passes. The regeneration efficiency itself will probably not be 100%. The frequency of regeneration therefore requires to be further increased as time passes. The considerations above were based on a system with complete phosphorus removal and steady-state conditions. For practical applications it is, however, in most cases not necessary to satisfy a requirement of effluent concentration much less than 0.5 mg P 1 ~. In order to satisfy such a requirement, the mathematical model has been used to compute the time of operation for a fixed-bed system without countercurrent flow at different operating conditions. This time of operation will be equal to the 0 value of a countercurrent system in the start-up phase. The results from these calculations are presented in Fig. 6. The results in Fig. 6 show that the time of operation (or the 0 of a countercurrent system in the start-up phase), at an effluent concentration requirement of 0.5mg P1 -~, is acceptable for a column length of about 5.0 m at pH = 7.0, Co = 5.0 mg P Iand u = 9.0 m h -1. The figure also shows how 0 is influenced by changes in pH, Co and u. If such a system was to be operated countercurrently at steadystate, the 0 value had to be reduced from 25 days at run start-up to less values as the time of operation passed. CONCLUSIONS This study clearly shows that the pH, the alumina particle size and the column length are very important parameters in minimizing the operating costs of a fixed-bed system. The pH should be reduced to levels of about 6.0 by adding hydrochloric acid or CO2 gas to the inlet water. The alumina particle size is indeed a very important parameter. Thus the water should be given a careful pretreatment in order to produce an inlet water low in suspended solids. This is considered to be decisive in preventing head-loss build-up when using fine-grained alumina. The column length should be as high as possible in order to get a fair utilization of the available adsorption capacity of the alumina within the bed. The adsorption process is well suited for designing a countercurrent system, for which the detention time of the alumina should be as high as possible to minimize the operating costs. However, there has to be found an optimum choice between this detention time and the hydraulic load in order to minimize the total annual costs of the process. A proper process design seems to be a beds-inseries system. For such a system the above mentioned factors will be favourable. A countercurrent alumina flow may be obtained by regularly regenerating the head column of the series, and coupling this column back into the process as the last one after regeneration. In this way a quasi-steady-state condition will be obtained, with an effluent concentration which
always may be kept below the requirements. The frequency of regeneration must, however, be increased as time passes. With the use of the activated alumina Compalox AN/V 813, a coarse-grained type y-alumina, a very good effluent quality may be obtained at hydraulic loads less than 10 m h-~ for a total column length of about 5.0m. When using a more fine-grained alumina, careful pretreatment of the water is more critical, however, a higher hydraulic load may be used. A hydraulic load of 10 m h -1 is several times the one of a conventional flocculation/sedimentation process. Thus the adsorption process seems to be a very attractive alternative when a high effluent quality has to be achieved, as a polishing step in the overall treatment. This is indeed a fact for countries with a cold climate where most of the plants are covered by buildings, and the total annual costs depend strongly on the area requirement of the unit processes. This paper has been devoted to the more mechanistic phenomena of the adsorption process for phosphorus removal. Although we feel to have succeeded in giving an interpretation of the process in the light of accepted mass transfer theory, several important factors still have to be examined. The most urgent ones may be efficiency and chemical consumption during regeneration of exhausted alumina, long term fouling effects, head loss development and pretreatment requirements, and adsorbability of more complex phosphorus compounds. Acknowledgements--This study was made possible with
funds provided by the Royal Norwegian Council for Scientificand Technical Research (NTNF). The writers also wish to thank Dr Crittenden at Michigan Technological University, U.S.A. for providing the computer programmes used to solve the HSDM equations.
Ames L. L. and Dean R. B. (1970) Phosphorus removal from effluentsin alumina columns. J. Wat. Pollut. Control Fed. 42, RI61-R172. Anderson M. A. and Malotky D. T. (1979) The adsorption of protolyzable anions on hydrous oxides at the isoelectric pH. 3. Colloid Interface Sci, 72, 413-427. APHA (1980) Standard Methods for the Determination o[ Water and Wastewater, 15th edition. American Public Health Association, New York. Brattebo H. (1983) Phosphorus removal from wastewater by fixed-bed adsorption on granular activated alumina. Thesis submitted to the Department of Civil Engineering, the Norwegian Institute of Technology, in partial fulfillment of the requirements for the degree of Doctor of Engineering, University of Trondheim. Crittenden J. C. and Weber W. J. Jr (1978) Predictive model for design of fixed-bed adsorbers: parameter estimation and model development. Am. Soc. cir. Engrs J. envir. Engng 104, 185 197. Crittenden J. C. and Weber W. J. Jr (1978) Model for design of multicomponent adsorption systems. Am. Soc. cir. Engrs. J. envir. Engng 104, 1175-1195. Crittenden J. C., Wong B. W. C., Thacker W. E., Snoeyink V. L. and Hinrichs R. L. (1980) Mathematical model of
H. BRATTEBOand H. ODEGAARD
sequential loading in fixed-bed adsorbers. J. Wat. Pollut. Control Fed. 52, 2780-2795. Davis J. A. (1982) Adsorption of natural dissolved organic matter at the oxide/water interface. Geochim. cosmochim. Acta 46, 2381-2393. Davis J. A. and Leckie J. O. (1980) Surface ionization and complexation at the oxide/water interface. 3. Adsorption of anions. J. Colloid Interface Sci. 74, 32~,3. Detweiler J. C. (1978) Phosphorus removal by adsorption on alumina as applied to small scale waste treatment. M.S. Thesis, University of Wisconsin, Madison. Donnert D., Eberle S. H., Gruber P., Sapandowski H. P. and Sch/ifer J. (1982) Application of activated alumina for phosphorus removal and recovery from waste waters. Japanese-German Workshop on Waste-water Treatment, Tokyo, Japan. Donnert D., Eberle S. H., Metzger H., Endres G. and Sapandowski H. P. (1978) Untersuchungen fiber die Rfickgewinnung von Phosphaten aus Abw/issern mittels Aluminiumoxid. Vom Wass. 50, 125 136. Eberle S. H., Donnert D. and Sapandowski H. P. (1980) Abtrennung und Rfickgewinnung von Phosphat aus biologische gereinigtem Abwasser mittels Aktivtonerde. Gewiiss.- Wass.-Abwiiss. 42, 835-859. Eberle S. H., Klopp R. and Anwar S. (1978) Quantitative Beschreibung von Adsorptionsgleichgewichten mit Aluminium-oxid auf Basis einer iiberlagerung yon zwei Langmuir-Adsorptionsisothermen. Kernfoschungszentrum Karlsuhe, Kfk, 2588 UF. Eberle S. H., StSber H. and Donnert D. (1975) Versucbe fiber die Adsorptionseigenschaften von Aluminiumoxid und seine Anwendung zur Reinigung eines Humistoffhaltigen Grundwassers. Engler-Bunte-lnstitut der Univ. Karlsruhe, VerSff. der Bereichs und der Lehrustuhls ffir Wasserchemie, H.9, Karlsruhe. FSrster H. and Solbach H. (1979) Entwicklung und Erprobung eines Verfahrens zur Reinigung von Zellstoffabw~issern mit Aluminiumoxid. Bundesministerium ffir Forschung und Technologie, BMFTWA 072. Fursawa T. and Smith J. M. (1973) Fluid-particle and intraparticle mass transport rates in slurries. IEC Fundament. 12, 197-203. Gnielinski V. (1978) Gleichungen zur Berechnung des W/irme- und Stoffaustausches in durchstr6mten rubenden Kugelschfittungen bei mittleren und grossen Pecletzahlen. vt. Verfahrenstechnik 12, 363-366. Hand D. W., Crittenden J. C. and Thacker W. E. (1983) Simplified models for design of fixed-bed adsorption systems. Am. Soc. cir. Engrs. J. envir. Engng. In press. Hingston F. J., Posner A. M. and Quirk J. P. (1972) Anion adsorption by goethite and gibbsite. 1. The role of the proton in determining adsorption envelopes. J. Soil Sci. 23, 177-192. H61zel G. and Sontheimer H. (1981) Eine Apparatur zur Messung der Adsorptionskinetik k6rniger Adsorbentien. Vom. Wass. 56, 165-170. Koresh J. and Softer A. (1983) Application of the two-site
langmuir isotherm to microporous adsorbents. J. Colloid Interface Sci. 92, 517-524. Kummert R. (1979) Die Oberfl/i.chenkompleksbildung von organischen S/iuren mit ~,-Aluminiumoxid und ihre Bedeutung ffir natfirliche Gew/isser. Dissertation ETH, Zfirich. Kummert R. and Stumm W. (1980) The surface complexation of organic acids on hydrous y-AI2 03. J. Colloid Interface Sci. 75, 373 385. Ladendorf K.-F., Schlfinder E. U. and Sontheimer H. (1972) ~ber die Austauschkinetik organischer Anionen an Macro-porSsen Anionenaustauschharzen. ChemieIngr.- Tech. 44, 337-341. Lee M. C., Crittenden J. C., Snoeyink V. L. and Ari M. (1983) Design of carbon beds to remove humic substances. Am. Soc. cir. Engrs J. envir. Engng 109, 631-645. Narkis N. and Meiri M. (1981) Phosphorus removal by activated alumina. Envir. Pollut. (Series B) 2, 327-343. Peri J. B. (1965) Infrared and gravimetric study of the surface hydration of y-A1203, J. phys. Chem. 69, 2211-2219. Pollutech Pollution Advisory Services Ltd (1976) Effluent Polishing by Filtration Through Activated Alumina, Vols. I and II. PPSD Ltd, Oakuiue, Ontario. Project no. 73-5-4. Purushothaman K. and Yue C. M. (1970) Removal of phosphates by sorption on activated alumina. Missouri Water Resources Research Center, University of Missouri, Rolla, Missouri. Rohmann U. and Sontheimer H. (1978) Untersuchungen zum Sorptions-mechanismus von organischen S/iuren an Aluminiumoxid. Vom Wass. 50, 301 318. Schwab G. M. and Dattler G. (1937) Angew. Chem. 50, 691. Shiao S. J. and Akashi K. (1977) Phosphate removal from aqueous solutions on activated red mud. J. Wat, Pollut. Control Fed. 49, 280-285. Stumm W., Hohl H. and Dalang F. (1976) Interaction of metal ions with hydrous oxide surfaces. Croat. chem. Acta 48, 491-504. Thacker W. E., Crittenden J. C. and Snoeyink V. L. (1981) Modelling of adsorber performance: variable influent concentration and comparison of ad~,orbents. Presented at the 54th Annual Meeting of the Water Pollution Control Federation, Detroit, Mich. Thacker W. E., Snoeyink V. L. and Crittenden J. C. (1983) Desorption of compounds during operation of GAC adsorption systems. J. Am. Wat. Wks Ass. 75, 144-149. Vinyard D. L. and Bates M. H. (1979) High-calcium fly ash for tertiary phosphorus removal. Wat. Sewage Wks 126, 62 64. Vliet B. M. van and Weber W. J. Jr (1981) Comparative performance of synthetic adsorbents and activated carbon for specific compound removal from wastewater. J. Wat. Pollut. Control Fed. 53, 1585-1598. Weber W. J. Jr and Liu K. T. (1980) Determination of mass transport parameters for fixed-bed adsorbers. Chem. Engng Comm. 6, 49-60. Williamson J. E., Bazaire K. E. and Geankoplis C. J. (1963) Liquid-phase mass transfer at low Reynolds numbers. IEC Fundament. 2, 126-129.