Water Res. Vol. 18, No. 11, pp. 1411-1419, 1984 0043-1354/84$3.00+0.

00
Printed in Great Britain. All rights reserved Copyright © 1984Pergamon Press Ltd

EQUALIZATION/NEUTRALIZATION MODELING
AN APPLICATION TO FLUORIDE REMOVAL

DESMOND F. L A W L E R ~ and D O N A L D H. WILLIAMS 2

~Dcpartment of Civil Engineering, University of Texas at Austin, Austin, T X 78712
and 2U.S. Environmental Protection Agency, Dallas, TX, U.S.A.

(Received March 1984)

Abstract--The chemistry of a fluoride removal system (gas scrubbing to capture hydrogen fluoride and
subsequent precipitation of calcium fluoride with lime) was studied, with particular emphasis on the effects
of and on pH. Carbonate, when present, was found to interfere with the precipitation in the pH range
7-11 but little carbonate was in the system studied because of the low pH and high gas/liquid exchange
in the scrubber. A general approach to modeling equalization/neutralization problems, using mass
balances on conservative substances and a charge balance to determine pH, was presented and illustrated
via application to the fluoride removal system.

Key words--equalization, neutraliTation, fluoride removal, pH control, fluoride precipitation, fluoride

INTRODUCTION duction facility led to this study. Hydrogen fluoride

The control of pH is of major importance in a wide
variety of industrial wastewater applications. Legal
requirements for effluent discharges include con-
straints on the pH, and the success of several waste-
water treatment processes depends on attaining and
maintaining the wastewater within a specific pH
range. In those instances in which pH control or
neutralization are necessary, it is often accomplished
in an open tank with some mixing; a mathematical
model to predict effluent pH in such tanks based on
flow patterns and influent chemical characteristics is
developed in this paper. The modeling approach
given herein could be useful in design (sizing of the
tank) or operation of many equalization/neutral-
ization facilities. However, the development is shown
in the context of a specific application: pH control in
an existing fluoride removal system. Before
presentation of the model, an analysis of the need for
pH control in fluoride removal is presented; this is
done in part because previous investigators have
misunderstood certain aspects of fluoride chemistry,
particularly its precipitation with lime. In addition,
gas HF(g), is produced in the reduction of aluminum
ore through the use of cryolite, Na3AIF~. The hydro-
gen fluoride gas is often removed from the waste gas
stream by wet scrubbing, thereby transferring the
fluoride into water as hydrofluoric acid, HF(aq). The
scrubber wastewater is then treated with lime,
Ca(OH)2, resulting in the precipitation of calcium
fluoride, CaF2. An existing treatment system of this
general description at an aluminum production fa-
cility was the basis of this study; a schematic diagram
of the system is shown in Fig. 1. After precipitation
and sedimentation, the water is recycled to the scrub-
hers.
For optimal use of this system (i.e. both capture of
the fluoride in the scrubbers and removal via pre-
cipitation and sedimentation), plant operators at-
tempted to maintain the pH in the water going to the

~rubbers
an understanding of the fluoride chemistry is critical

'O',n'n'o"on
I I
to understanding the application of the mathematical
model.

FLUORIDE REMOVAL

Fluoride is present in significant concentrations in

Sludge"-I
the wastewaters of several industrial processes; its
control is important for health, ecological and legal
Lime~,~
reasons (Rabosky and Miller, 1974). The generation Fig. 1. Schematic diagram of the fluoride removal system at
and removal of fluoride from an aluminum pro- the Alcoa plant in Rockdale, Texas.
1411
1412 DEsuoso F. L^WLERand DONALD H. WILLIAMS
30
25
20
L.,
t~
E 15
n and means that the solubility of fluoride rises
10
0 I I I I I
4 6 8 10 12 14
pH
Fig. 2. Experimental results for residual soluble fluoride
after lime precipitation reported by Parker and Fong 0975).
scrubbers in the narrow range 8.0 < pH < 8.3. The
principal reason for the choice of this range was the
published results of Parker and Fong (1975) which
showed that residual fluoride after lime precipitation
attained a minimum in that pH range, as shown in
Fig. 2, which is re-drawn from their results. It was
also thought that fluoride capture in the scrubber
declined at lower pH values. Maintenance of the pH
in the desired range had proven to be quite difficult;
it was not uncommon for the pH (measured in the
clear well) to drop to values less than 5.0.
With this emphasis on pH control, it was decided
to study (a) the effect of pH on each of the treatment
processes (scrubbing; lime addition, mixing and pre-
cipitation; and sedimentation) and (b) the effect of
each treatment process on pH.
The study of the role of pH was based on
measurements of the characteristics of the wastewater
of various points in the system, experiments on both
real and synthetic wastewater and an analysis of the
relevant chemistry from both equilibrium and kinetic
viewpoints. The description of this work in this paper
is divided according to the treatment processes. Sub-
sequenfly, the development of the mathematical
model is presented and discussed.
Scrubber chemistry
The principal chemical equations relevant to all the
treatment processes at this plant are presented in
Table 1. The equilibrium between the gaseous and
aqueous forms of H F [equation (1)] is described by
Henry's Law. The equilibrium constant is not known
well but is extremely high, i.e. hydrogen fluoride is
very soluble in water. The dissociation of
hydrofluoric acid in water is expressed by equation
(2); the value of the equilibrium constant indicates
that at pH values higher than 3.17, most of the total
fluoride is present as fluoride ion, F-. The combina-
tion of these two equilibria is shown conceptually in
Fig. 3 in which the total soluble fluoride concen-
tration [defined in equation (3)] is shown to depend
strongly on pH. Since neither the Henry's Law
constant, KH.NF, nor the partial pressure of hydrogen
fluoride gas, PHF, in the system studied, is known, the
correct vertical placement of the line on Fig. 3 is
unknown; however the shape of the curve is correct
significantly with pH. Since, as shown in equation (4),
the driving force for mass transfer of fluoride at any
point in the scrubber is the difference between the
saturation value of HF(aq) and the actual value in
solution, the capture of fluoride should also rise
significantly with pH. Therefore, the idea that the
efficiency of the scrubber in capturing hydrogen
fluoride from the gas stream' declines as the influent
pH declines is well-founded. The meaning of ~0.F in
equation (4) follows standard terminology (Stumm
and Morgan, 1982; Snoeyink and Jenkins, 1980) and
is explained in equation (5); ~.F declines with rising
pH. The pH of the scrubber water declines
significantly through the scrubber, since HF(aq) is a
moderately strong acid [equation (2)]. The capture of
hydrogen fluoride is sufficiently high in the plant
studied that the pH of the wastewater is always below
4.0.
The effect on the wastewater of carbon dioxide,
CO2, in the air being scrubbed was considered, since
CO 2 is a constituent of air. The chemistry of the
carbonate system is assumed to be familiar to the
reader. It is quite similar to that of fluoride, except
that carbonic acid, H2CO3, is a diprotic acid; the
equations for the carbonate system analogous to
those presented above for fluoride are shown as
equations (7)-(16) in Table I. The equilibrium be-
tween atmospheric levels of carbon dioxide
(Pco: = 3.42 x 10-4atm) and water is shown as a
function of pH in Fig. 4. For the system under
consideration, it is reasonable to assume that (1) the
air passing through the scrubber has a CO2 concen-
pH
0 2 4 6 8 10 12 14
I I I
-2
-4
°~ -6
-10
Fig. 3. Gas/liquid equilibrium of fluoride in water.
 Equalization/neutralization modeling 1413

Table 1. Chemical information relevant to fluoride removal system

Equation Equifibrium
number Equation constant
(1) HF(g) = HF(aq) ~H,HF
(2) HF + H20 = F- + H3O÷ Ka, HF = 10 -3.17

(3) FT = [n~ + [F-]

(4) r r = k ~ {[HF], - [HI=']} = k ~ {KH,HF -- %,F FT }
[HI=] [H30 +]
(5) %.r FT K~.,r+ [H30÷]
[F-] K~HF
(6) ~.F FT g~..r + [H30 +]
(7) COs(s) = COs(aq)
(8) CO~(aq) + H20 = H2CO3
(9) [H2CO*]= [CO2(aq)]+ IH2COgl
(10) H2CO~'+ H20 = HCO; + H~O+ Ki = 10-s'3s
(11) HCO; + H20 = CO~3-+ H30 + K, = 10-t°33
(12) CT= [H2CO~]+ [HCO;I + [CO]-I
(13) rCT= kta {[HsCO~]s -- [HsCO*]}
= k~{K..oo/~o~- %.c~}
[H2CO~I [H~O+]~
(14) %,c = C.r -- [H30+]2+ KI[H30+] + KIK2
[HCO~] K~[H30+]
(15) ~t.c Ca. [H30+]2+ Ki[H30+] + KiKs
[c~-] .,¢,K~
(16) c~2"c CT [H30+]s + KI[H30+] + KIK2
(17) Ca(OH)2= Ca2+ + 2OH- K~,- 10-9.9
(18) CaF2(s) = Cas+ + 2F- K~ = 10-I°'s
(19) CaSO~(s)= Ca2+ + 80,2- g~, = 10-('~
(20) CaCO~(s)= Cas+ + COl- K,e = 10-I'~

tration not significantly different from the atmos- [equation (18)], the possibility of precipitation with
pheric level and (2) equilibrium is nearly attained. sulfate [equation (19)] and/or carbonate (equation
The latter assumption is justifiable since mass transfer (20)] was investigated.
is high in scrubbers and since the water is recycled. Sulfate concentrations, measured according to
With those assumptions, and the fact that the s c r u b Standard Methods ( A P H A , 1980) indicated that no
ber effluent wastewater always has a p H less than 5.0, significant precipitation o f calcium sulfate should
the total carbonate in the wastewater should be less occur. Sulfate is added to the system only in the
than 10 -4 M (log CT = --4), as shown in Fig. 4. That make-up water which is put into the clear well to
concentration is equivalent to a total inorganic car-
bon concentration o f 1.2 mg 1-~; i.e. one should ex-
pect very low levels of carbonate in the water. This
pH
expectation is important in the analysis o f the sub-
0 2 4 6 8 40 12 14
sequent processes. I I I I
T o summarize, the capture o f hydrogen fluoride is
improved by a high influent p H to the scrubber, the 2
efficient capture o f the hydrogen fluoride leads to a
low effluent pH, and the low effluent p H means that 4
the total inorganic carbon concentration (CT) should
be very low.
o
Chemistry of precipitation with lime
8
The addition o f hydrated lime, Ca(OH)2, leads to
a rise in the p H due to the dissolution and dis-
10 --
sociation described in equation (17), as is well-
known. The resulting calcium ions, Ca 2+, are then
available to precipitate with other ions in solution; in Fig. 4. Gas/liquid equilibrium of carbonate in water
addition to the desired precipitation with fluoride (Pco~ = 3.42 x 1 0 - ( a t m ) .
1414 DESMONDF. LAWU~ and DONALDH. WILLIAMS

replenish water lost by evaporation and removal of 60

sludge. Because of evaporation and the recycle sys-

tem, sulfate levels can build-up in the system to the
50
point of precipitation, but only a negligibly small
amount of sulfate precipitation occurs at any time in
comparison to the fluoride precipitation. 40
The possibility of calcium carbonate precipitation
was investigated in significant detail, since it was
30
hypothesized that the results obtained by Parker and
Fong (1975), shown earlier in Fig. 2, reflected a
competition for the available calcium between ZO
fluoride and carbonate. Parker and Fong's results Initiot

were cited by Paulson (1977) in recommending that

a pH near 12 be used for fluoride precipitation with 10 o 106 atrnos
lime in order to avoid the poor removal at lower pH o 106 0
values. Other investigators (Rabosky and Miller, A 240 0
t ~ I I }
1974; Zabban and Helwick, 1975; Zabban and Jew- 4 6 8 ~I0 12 14
ett, 1967) have also cited the need for this high pH pH
for sufficient fluoride removal. Fig. 5. Experimental results: residual fluoride after precipi-
Parker and Fong (1975) added varying doses of tation with lime.
lime to 250 ml aliquots of a hydrofluoric acid solution
in distilled water with a concentration of 87 mg FI- ~.
After 30 rain of stirring and 30 rain of settling, soluble the curve of soluble fluoride as a function of pH is
fluoride was measured. The present authors hypoth- due to carbonate interference. Under the conditions
esized that the results of Parker and Fong were of the "open system", equilibrium with the atmos-
influenced by contamination of the distilled water phere was considered to occur at all pH values, so
with carbonate, which could easily occur by dis- that Cr varied as shown above in Fig. 4. Most
solution of CO2 from air. To check this, three sets of wastewater treatment systems would more closely
experiments were performed similar to those of approximate the closed systems at some fixed total
Parker and Fong. In one set, distilled water was carbonate concentration, since most basins are
aerated for 2 h prior to the addition of hydrofluoric reasonably deep and have a small amount of surface
acid to ensure that carbonate would be in the water. area in contact with the atmosphere per unit volume
In the second and third sets, nitrogen gas was bub- of solution. If that fixed level is nearly zero, as
bled through the distilled water for 2 h prior to the expected for the system under consideration in this
hydrofluoric acid addition; the bubbling of nitrogen study, the residual fluoride achieved should be essen-
was continued throughout the experiment to ensure tially independent of pH in the range
that no carbonate was allowed into the system. The 7.5 < pH < 10.5. At measureable levels of total car-
first and second sets were performed at a fluoride bonate, the residual fluoride is highly dependent on
concentration of 106 mg F l-', similar to that used by pH.
Parker and Fong; the third set was performed at a Total carbonate concentrations were not measured
higher concentration. The results of these experi-
ments, shown in Fig. 5, clearly exhibited the follow-
ing trends: first, results very similar to those of Parker
and Fong were found in the first experiment, with a
drastic "hump" in the residual fluoride near pH 11, • ' Open system

and a minimum in the pH range 8-9; second, no such -I -- ...... C T • 10 -3 M

hump in the residual fluoride curve was found when .... CT. 0
carbonate was excluded; and third, the residual
-2
fluoride was independent of the initial fluoride con-
centration in the carbonate-free experiments.
_o
To confirm the hypothesis that the hump in the -3
residual fluoride curve was caused by competition /~..
with carbonate, equilibrium calculations were made
using the REDEQL2 program (McDuff and Morel,
1974) for the soluble fluoride concentration after
precipitation with calcium for various carbonate con- -5 I I I I I
centrations. The results are depicted on a logarithmic 4 6 8 'tO t2 14
graph in Fig. 6. While there is not exact numerical pH

agreement with the experimental results (10-3M is Fig. 6. Theoretical results: residual fluoride after precipi-
19 mg F l-t), these results confirm that the hump in tation with lime.
 Equalization/neutralization modeling 1415

'12-- sufficient lime be added to meet the stoichiometric

oOOO0 O0
requirements for the fluoride precipitation. This
would automatically maintain the pH at a sufficiently
10--
O
high level not only to achieve the precipitation but
0 also to ensure the efficient capture of the hydrogen
8-- O fluoride gas. With the recycle system employed at the
0 plant studied, the primary effect of a temporary
condition of low pH would not be the higher fluoride
~6 0
concentrations in the water but the less efficient
O

@ capture of the hydrogen fluoride by the scrubber.

4 0
D more than one opportunity to precipitate fluoride
, Q
I I I I
0 4 8 '12 '16
NaOH added (mequtv l -I)
Fig. 7. Titration of scrubber effluent wastewater.
directly in this work. However, samples of the scrub-
ber effluent and the sedimentation basin influent were
titrated with a strong base and strong add, re-
spectively; some results are shown in Figs 7 and 8.
Both show a negligible buffering intensity in the pH
range near pH 6.3, an indication that very little
carbonate was in the wastewater. The negligible
carbonate concentration in the scrubber effluent was
expected, as noted above, but these results also
indicate that little carbonate entered the wastewater
as CO2 from the atmosphere in the mixing chambers
prior to the sedimentation basins. Hence, fluoride
removal should be independent of pH over a wide
range above pH 7.5 so that there was no real need to
maintain the pH in the narrow range of 8.0-8.3.
Process overview
The major condition for the lime addition was that
12-
4C
e
~6
4--
This is so because, with the recycle system, there is
from the wastewater but only one opportunity to
capture the gaseous pollutant. A possible effect of
high pH in the clear well is that the pH would not be
depressed sufficiently in the scrubber to ensure low
I
20 carbonate concentrations in the scrubber effluent; in
such a case, the fluoride precipitation would be highly
dependent on the pH because of the competition with
calcium carbonate precipitation, as explained above.
This potential negative effect of high pH is unlikely
in the case studied because the scrubbers capture a
large mass of fluoride in a relatively small volume of
water, but could be important in other applications.
These potential effects of high or low pH emphasize
again the importance of the pH in the clear well, a
point in the system far removed from the lime
addition which ultimately controls the pH.
At the plant studied, the pH was monitored con-
tinuously with the results shown on a strip-chart;
periodically an operator checked the strip chart and
used the information to adjust the lime feed rate, if
considered necessary. However, it was unclear how
often the strip chart should be checked or to what
extent the lime feed rate should be changed. Know-
ledge of the sensitivity of the pH in the system to
changes in either the lime feed rate or the fluoride
concentration just before the lime addition could
yield better guidelines for operating decisions. Hence,
it was considered valuable to develop a mathematical
model to predict changes in the pH at the clear well
due to either type of change.
MATHEMATICAL M O D E L OF
EQUALIZATION/NEUTRALIZATION
The sedimentation tanks, besides achieving excel-
lent solids removal, served as an equalization/neutral-
ization tank to dampen fluctuations in the concen-
trations of soluble fluoride and soluble calcium and
in the pH. The clear well also added to this available

! equalization volume. To be able to model the pH

changes with time, a three-step approach was taken.
First, the inputs to the equalization volume were
defined; one critical assumption was made in this step
2I I I I I I
and some simplifying conditions were chosen to aid
0 04 0.8 1.2 '1.6 20 the analysis and presentation. Second, mass balances
H 2S0 4 added (rneQuiv L-1) were used to determine the concentration of conser-
vative substances in the clear well as a function of
Fig. 8. Titration of sedimentation basin influcm. time. Third, since acid-base (proton transfer) reac-
1416 DESMONDF. LAWLERand DONALDH. WILLIAMS

tions occur very rapidly in water (Stumm and Mor-

gan, 1982) equilibrium was assumed to exist at any
instant for acid/base reactions and a charge balance
was applied to determine the pH at any time. This ~v, v~
approach is similar to that used by DiToro et al.
C~ C2
(1981).
Sedimentation Cleor
wel I
Model inputs
Fig. 9. Schematic representation of sedimentation basin
The critical assumption in the definition of the and clear well for equalization/neutralization model (C is
inputs to the equalization volume was that all precipi- concentration, V is volume, C~' is concentration of constit-
tation reactions are completed in the mixing basin. uent i in the influent and in all tanks prior to time zero, kiC~
This means that, beyond the mixing tanks, the total is constant input concentration of constituent i after time
zero).
soluble concentration of each constituent is a con-
servative quantity in the equalization volume (the
9. The choice of the two tanks in series for the
sedimentation tanks and clear well). The assumption
sedimentation tank was made since it was the sim-
that the precipitation is completed in the mixing basin
plest non-ideal flow which could be represented ana-
is reasonably valid for the plant studied: for the dates
lytically. The flow rate was considered to be constant,
on which observations were made, the concentrations
since water was pumped to the scrubbers at a con-
of soluble fluoride were reduced an average of 74%
stunt rate at all times. Under the conditions assumed
between the point before the lime addition and the
at time zero, the concentration (G) of some con-
sedimentation basin influent, but only an additional
servative substance (i) was the same throughout the
1.4% through the sedimentation basin. Not all pre-
system and is designated C~*. At time zero, the input
cipitation reactions are completed so rapidly, so
was changed to k,C~* where k~ is a positive constant
similar assumptions must be made with caution in
(greater or less than l). The response of each of the
other applications.
three completely mixed tanks in series (the two halves
The conditions chosen for analysis included the
of the sedimentation tank and the clear well) to this
following. First, the soluble fluoride input to the lime
changed input at time zero was determined using
addition/mixing process is constant. As noted above,
sequential mass balances on each tank and LaPlace
pH changes could occur due to changes in the lime
transforms to solve the differential equations.
feed rate or in the soluble fluoride in the scrubber
The mass balance for a conservative substance for
wastewater. The two cases are identical conceptually
the nth tank in a series of completely mixed con-
so that only the first is considered in the analysis.
tinuous flow reactors is written as:
Second, the initial conditions had to be specified, i.e.
concentrations of each soluble species must be known v, -de.
~ = QC,_I - QC,. (21)
at each point in the equalization volume. For each
species, these were chosen to be the same throughout In equation (21), v, is the volume of the n th tank, Q
the volume considered. Third, at time zero, a sudden is the flow rate, t is time, C, is the concentration in
change occurs in the lime feed rate (L) from the the nth tank of the substance under consideration
previous value (L*) to another value (expressed as and C,_1 is the effluent concentration from the pre-
the product of a constant kf~ and the previous value, vious tank (equivalent to the influent of the n th tank).
i.e. L =kf~dL*), which remains constant for a For the first tank (n = 1), the input concentration,
significant time period. These conditions are reflective C,_~, is the input of the entire system, defined above
of the system studied since changes in either lime feed
as kC*. Dividing each term in equation (21) by V.
rate or the fluoride captured in the scrubbers did not
and recognizing that V,/Q = i',, the hydraulic de-
occur frequently, i.e. at intervals significantly less
tention time of the nth reactor, yields:
than a few detention times. Finally, the change in the
lime feed rate causes corresponding changes in the dG G-1 G
= _ ---~. (22)
soluble calcium and total soluble fluoride concen- dt tn tn
trations after the precipitation reactions.
The solution proceeds
)eeeds from the first tank to the
Flow characteristics and mass balances others in turn, since the effluent from one tank is the
influent to the next. The time-varying solutions for
Mass balance equations for conservative sub-
each of the three tanks were as follows:
stances flowing through the equalization volume were
then developed. The equalization volume provided in For tank 1:
the system under consideration was the sedimen-
tation tank and clear well. It was assumed that the C,(t) = C*{k - k exp(-t/~) +exp(-t/~)}. (23)
sedimentation tank exhibited a non-ideal flow pattern For tank 2:
equivalent to two ideal completely mixed tanks in
series (each with half the total volume), and that the
C2(t) = k C * + (I - k ) C * e x p ( - t/h)
clear well was completely mixed, as depicted in Fig. +(1 - - k ) C * t/h e x p ( - t / h ) . (24)
 Equalization/neutralization modeling 1417

And for tank 3: Table 2, Conditionsused in example for model calculations

C3(t) = k C * +
C*(l -k)[l 1 +=t}
_1 l Total soluble fluoride: 1.00x 10-3 M
t3(~__ ~ ) l t.(~3 __ ~l) t|
x exp(-- t/~) + 1 1 - h
F3 F~
In equations (23)-(25), t is time, ~ is the hydraulic
detention time of one-half the sedimentation tank, F3
is the hydraulic detention time of the clear well and
all other terms have been previously defined. The
solutions account for the fact that the sedimentation
tank was conceptually divided into two equal tanks,
so that ~ = F2. C2(t) represents the effluent of the
sedimentation tank, and Ca(t) represents the effluent
(and contents) of the clear well. The results in the
clear well were of principal interest in this work.
At very long times, the exponential terms in equa-
tion (25) become negligible and a new steady-state
condition with Ca, =k~C* is approached, as ex-
pected. The solution is general for the flow pattern
depicted, i.e. it is not specific for any one substance.
Hence, a reduction in the lime feed (k,~ < 1) would
result in a decrease in the soluble calcium in the
system (kc~2+< 1) and an increase in the soluble
fluoride, (kFr > 1). Knowledge of the appropriate k~
values for calcium and fluoride for a specific change
in the lime feed allows calculation, via equation (25),
of the gradual changes with time in the calcium and
fluoride concentrations in the clear well. Such know-
ledge of the appropriate k values can be gained
experimentally, by assuming equilibrium is achieved,
or by some other reasonable estimate.
Charge balance
With the knowledge of the concentrations of all
conservative substances in the water as a function of
time, a charge balance equation can be solved to
determine the pH. The charge balance is based on the
principle of electroneutrality, i.e. the number concen-
tration of positive charges must equal the number
concentration of negative charges. For the system
considered here (assuming carbonate is not negli-
gible), the charge balance equation can be written as
(A) Prior to time zero
(Conditions in intluent and throughoutsystem)
Total soluble calcium: 3.65 x 10-3 M
(B) After time zero
kf~ kc~2+ kFT
0.85 0.82 2.21
0.75 0.79 3.16
(25) 0.50 0.55 6.10
0 0 15.84
sociation constant for water, Kw:
[H~O +] + 2[Ca 2+]
+ ( ~ all other c a t i o n s - ~ all other anions)
K.
-- [i..I30+
] - -F ~I.FFT + (0~1.C + 2~2.c)C T. (27)
With Ca z+, FT and CT known from the mass balances
and a value for the combined term for the unspecified
cations and anions, the only unknown in equation
(27) is the pH (i.e. [H~O+]), which can be determined
by trial and error solution.
RESULTS
An example for the fluoride removal system was
solved under several conditions of reduced lime feed
below that necessary to maintain the pH at 8.3. The
conditions chosen for reduced lime feed and the
resultant changes in the influent calcium and total
fluoride (determined experimentally) are shown in
Table 2. The solutions to the mass balances for
calcium, shown in Fig. 10, reflect the decline in the
expected calcium concentration in the clear well.
Even if the lime feed were shut-offcompletely at some
instant (kca÷ = 0), there should be a significant delay
in the reduction of calcium in the clear well. The
fluoride concentrations would be expected to rise, as
shown in Fig. I l, and are a mirror-image of the
calcium results, since there is a stoichiometric require-
ment for calcium by the fluoride. The changes in
calcium and fluoride are independent of any con-
4

follows:
3 ~ - - 085
[HaO +] + 2[Ca 2+] k
075
\ ~ 050
+ ~ all other cations = [OH-] + [F-] 2 \

+ [HCO~-] + 2[CO~-] + ~ all other anions. (26) , \

..o
In equation (26), the brackets [] refer to molar
concentrations, the units of the terms under the I [ t 1 I I I
1 2 3 4 5 6 7
summations are equivalents 1-' and the substances Time (h)
included under the summations cannot participate in
Fig. 10. Predicted reduction in calcium concentration in the
acid/base reactions within the pH range of interest. clear well with various changes in the lime feed rate (num-
Equation (26) can be rewritten as follows, using bers associated with each line are values of kf,,d, i.e.
several substitutions from Table 1 and the dis- fractions of the lime feed rate required to maintain pH 8.3).
 1418 DESMOND F. LAWLERand DoNxu:) H. WILLIAMS

J 20

Inltl0( CT
(mof L-~)
?~; 't6
0 \ _ _ ,10-3

/
\
/ \
/
/
~ 8 /
/ 05 . . . . . . .
:3 \
~ 4 / ~ ~ ~ ~ ~ ~ 075
•1 -
¢3.
6 \
~__

I I I [
q 2 3 4
Time (h)
Fig. 11. Predicted increases in soluble fluoride concen-
tration in the clear well with various changes in the lime feed
rate (numbers associated with each line are values of kr~,
i.e. fractions of the lime feed rate required to maintain
pH 8.3).
siderations of the carbonate concentration, CT, as-
suming the fluoride solubility is not influenced by the
interference from carbonate below pH 8.3. Two sets
of solutions for p H were obtained using different
assumptions for the carbonate concentrations in the
water. In the first, it was assumed that equilibrium
with the atmosphere was achieved, so that a
significant amount of carbonate was in the equal-
ization volume prior to the reduction of the lime feed.
In the second case, it was assumed that the carbonate
concentration was zero.
The results for the pH variation in the first case are
shown in Fig. 12. Under the condition that the lime
feed is 85% o f that required to maintain the pH at 8.3,
the reduction in pH is predicted to be gradual and a
significant amount of time passes before the pH is
predicted to fall below 5.0. Even if the lime feed were
shut off completely, more than 30 min would pass
before the pH dropped to below 6.0, according to the
model calculations. Under the second assumption
(Cr = 0), the results are significantly different, as
indicated by the results shown in Fig. 13. The pre-
dicted variations of pH with time are shown for the
condition in which 75~o of the stoichiometric require-
ment of lime is fed to the system under both assump-
10
8 and knowing the concentrations for each species at
6 to a series of short term disturbances could be
\ \
\ "-..075
4 opment that cannot be altered without significantly
05
2 are complete prior to the influent to the equalization
I I I I
0 "1 2 3 4 5
Time (h)
Fi B. 12. M o d e l predictions for changes in p H when carbon-
ate is originally present (numbers associated with each line
are values of kr=d, i.e. fractions of the lime feed rate required
to maintain pH 8.3).
I
085
I I
\
5 6 7 5 \
\
3 I I I I I
0 30 6o 90 120 '150
Time (rain)
Fig. 13. Model predictions for changes in pH: effect of
carbonate (both curves with kfnxl = 0.75).
dons. With no carbonate present, there is little
buffering and the pH decreases rapidly after a brief
lag period. The presence of carbonate provides a
significant delay in the loss of pH to below pH 5.0.
The real situation at the plant studied is closer to the
case with no carbonate present, as noted earlier. Plant
operators once shut the lime feed off completely and
found the p H to drop to below pH 5.0 in less than
10rain; the model predicted this precipitous drop
under the conditions that CT was zero.
DISCUSSION
The model predictions shown were made under
very limited conditions, but most of the assumptions
made could be relaxed without severely complicating
the problem. The assumption of the same chemical
conditions throughout the sedimentation tank and
clear well provides a simple starting point, but the
only real requirement is that the chemical conditions
in each part of the system be known. The assumption
that the perturbation which occurs at time zero
remains for a significant time period could also be
relaxed. By redefining time zero at each perturbation,
that time from the previous predictions, the response
calculated. The only critical assumption in the devel-
changing the model is that the precipitation reactions
volume. Stated more generally, the assumption is that
I I I
6 7 all reactions other than the acid/base reactions of
soluble species are completed at the influent. It is this
assumption that allows the mass balance/charge bal-
ance approach to be used in the model. The mass
balances must be made on quantities that are con-
servative in the soluble portion of the wastewater, e.g.
 Equalization/neutralization modeling
total soluble fluoride, total soluble calcium and total
soluble carbonate. Were these quantities to change
beyond the influent because of reactions, the mass
balances would have to include reaction kinetics and
thereby complicate the problem considerably.
The development of the model has been described
above in conjunction with its application to the
fluoride removal system at a specific industrial plant.
However, the general framework of the neutral-
ization model presented has much wider applica-
bility. While some neutralization problems are more
complicated than the one used for illustration here,
many others are simpler. Neutralization often does
not include the complication of (potential) precipi-
tation and usually occurs in a single tank. Such
applications of the model would allow simpler chem-
istry and simpler flow patterns.
The model requires chemical inputs and a descrip-
tion of the flow characteristics of the tank. With such
inputs, mass balances on all conservative (non-
reactive) quantities can be performed. A charge bal-
ance equation can then be applied to determine the
pH at any time. The charge balance equation must
contain explicitly all chemical species which enter into
acid/base reactions in the pH range to be encoun-
tered, and other species which change with time
(those for which mass balances are performed). Other
species which neither enter into acid/base reactions
nor change significantly with time can be lumped into
a single term expressing the sum of residual cations
minus the sum of residual anions in terms of equiv-
alents 1-1.
In many cases of interest, pH fluctuation is caused
by only a small number of constituents. In addition,
the number of weak acid/base pairs is also small.
Hence the required data to be able to develop a model
for other applications are few. If the fluctuation in the
key parameters which drive the pH change is known
or can be estimated reasonably well, the model could
be used in the design of neutralization facilities to aid
in the sizing (detention time) of the tank required (td).
In operating systems, the model could be used to aid
in the development of operating rules, as it was in the
case illustrated.
SUMMARY AND CONCLUSIONS
The need for pH control in a fluoride removal
system was analyzed, and a model to predict pH
changes in the system developed. The following con-
clusions were reached from this work:
(1) Based on Henry's Law and the dissociation
chemistry of hydrofluoric acid, the capture of hydro-
gen fluoride gas in the scrubber water can be expected
to increase directly with increases in the pH of the
influent to the scrubber, for pH greater than 3.0;
(2) The pH of the effluent of the scrubber is
inevitably low due to the capture of the hydrogen
1419
fluoride; the low pH leads to the virtual absence of
carbonate in the water, which in turn leads to the
difficulty of controlling the pH in the range 6-10;
(3) The precipitation of calcium fluoride in the pH
range 6-11 is strongly dependent on pH only when
carbonate is present; this conclusion is supported
both by experimental results and by chemical equi-
librium calculations;
(4) The mathematical model to describe the pH
fluctuation could follow the pattern of pH decrease
observed in plant records. The results of the model
were useful in understanding the sensitivity of the pH
to changes in influent conditions and in developing
operating rules for the system;
(5) The approach used in the mathematical model,
i.e. to perform a mass balance to define the changes
in conservative quantities over time and then a charge
balance to determine the pH fluctuations with time,
is applicable to a wide variety of equalization/
neutralization problems.
Acknowledgements--The authors appreciate the financial
assistance of the Aluminum Corporation of America (Al-
coa) to support one of us (Donald H. Williams) during the
course of this research; the views expressed herein are those
of the authors alone and not those of Alcoa. Mr Glade
Lantz and Mr Flavio Oiivarez of Alcoa were particularly
helpful in defining the project and making plant records
accessible. We are also grateful to our colleagues at the
University of Texas: Mr Frank Hulsey, who assisted in the
solution of inevitable laboratory problems and Dr Howard
Liljestrand, for several enlightening discussions with respect
to this work. The excellent suggestions of the reviewers for
improvements to the manuscript are also appreciated.
REFERENCES
APHA 0980) Standard Methods for the Examination of
Water and Wastewater, 15th Edition. American Public
Health Association, Washington, DC.
DiToro D., LaPadula J. E. and Watkins J. P. (1981)
Equalization-neutralization basin design. J. envir. Engng
Div. Am. Soc. civ. Engrs 107, EEl.
McDuff R. E. and Morel F. M. M. (1974) Description and
use of the chemical equilibrium program REDEQL2.
Technical Report EQ-73-02, California Institute of Tech-
nology.
Parker C. L. and Fong C. C. (1975) Fluoride removal:
technology and cost estimate. Ind. Wastes 21, No. 6, 23.
Paulson E. G. (1977) Reducing fluoride in industrial waste-
water. Chem. Engng Deskbook Iss. 89, 17 October.
Rabosky J. G. and Miller J. P. Jr (1974) Fluoride removal
by lime precipitation and alum and polyelectrolytecoagu-
lation. Proc. 29th Ind. Waste Conf. Purdue Univ., p. 669.
Snocyink V. L. and Jenkins D. (1980) Water Chemistry.
Wiley, New York.
Stumm W. and Morgan J. J. (1982) Aquatic Chemistry,
Second Edition. Wiley Interscience, New York.
Zabban W. and Helwick R. (1975) Dcfluoridation of waste-
water. Proc. 30th Purdue Ind. Waste Conf. Purdue Univ.,
p. 479.
Zabban W. and Jewett H. W. (1967) The treatment of
fluoride wastes. Proc. 22nd Purdue Ind. Waste Conf.
Purdue Univ., p. 706.