Reactor Model for Production of Aluminum Fluoride

John Karlström
Department of Chemical Engineering II, Lund University, P.O. Box 124, SE- 221 00 Lund, Sweden

A reactor model was developed to simulate the production of aluminum fluoride from
fluosilicic acid and aluminum hydroxide in a semibatch reactor. The reaction
mechanism was determined and a kinetic model was combined with the material and
energy balances to create a reactor model. The reactor model was used to study the
effects of the preheating temperature of fluosilicic acid and aluminum hydroxide, the
concentration of fluosilicic acid as well as the grain size of aluminum hydroxide.

Introduction The production of aluminum fluoride is carried

out preferably between 70°C and 100°C. The
Aluminum fluoride is used in many industrial
concentration of fluosilicic acid can be as high as 35
processes. It is one of the minor constituents added
wt-% in a water solution. The acid concentration as
to the electrolytic cells during the production of
well as the particle size of the aluminum hydroxide
metallic aluminum. It is also used in the preparation
affect the rate of the reaction. The resulting
of white enamels, as an anti-reflection coating in
aluminum fluoride solution is metastable, and the
complex optical systems, as a constituent in welding
trihydrate begins to crystallize out quickly at
fluxes, and in the preparation of fluorine containing
temperatures around 90°C. Precipitated solid silica
glasses, (Gernes, 1962).
must therefore be removed as quickly as possible.
The present study was carried out with the aim to
The main problems in the process are firstly to
investigate the reaction between fluosilicic acid and
carry out the reaction in such a way that the filter
aluminum hydroxide under the production of
cake obtained by separating out the silica exhibits
aluminum fluoride and silica. The process for
good filtering properties, and secondly to avoid the
production of aluminum fluoride and silica is
aluminum fluoride being contaminated with silica.
schematically depicted in Figure 1.
This leads to the following requirement on the
The total process of fluosilicic acid interaction
conditions for carrying out the reaction (Arankathu,
with aluminum hydroxide can be described by the
1980):
following overall reaction:

H2SiF6 + 2 Al(OH)3 → 2 AlF3 + SiO2 + 4 H2O (I) 1 . The silica precipitates out in different forms,
depending on the pH of the solution. At low pH
The reaction is exothermal and proceeds in values (<3) the readily filtered form of the silica
several steps. It can be described by the following predominates; with increasing pH the amount of
three reactions (Skyler, 1966; Dmitrevskij and fine and colloidal silica increases. The fine
Semenova, 1970): particulate form makes filtration difficult,
therefore the pH should not exceed a certain
3 H2SiF6 + 2 Al(OH)3 → Al2(SiF6)3 + 6 H2O (II) value (<3-4) which is determined by the method
of filtration used.
Al2(SiF6)3+ 6 H2O → 2 AlF3 + 3 SiO2 + 12 HF (III) 2. The reaction time plus the filtering time should
not be extended to the point that aluminum
12 HF + 4 Al(OH)3 → 4 AlF3 + 12 H2O (IV) fluoride begins to crystallize.
3. On the other hand, to obtain a good yield and
also a practically silica-free aluminum fluoride
solution, it is important that the fluosilicic acid
is converted as completely as possible.
Al(OH)3 Reac- Filter Cryst-
tor alli-
zation Reaction mechanism
AlF3
H2SiF6 To determine the kinetics of the reactions, the
experiments and the results of Skyler (1966), and
Dmitrevskij and Semenova (1970) have been used.
SiO2 When evaluating the results in the latter article, the
Figure 1. Process for production of aluminum fluoride
rate equation of the first reaction in the mechanism
above can be described as:
r1 = k1 * CA* CB2/3 * CB01/3 (1)
In this relation r1 is the rate of the first reaction
and k1 is the rate constant. CA is the concentration of
fluosilicic acid, CB is the concentration of aluminum
hydroxide and CB0 is the start concentration of
aluminum hydroxide. The first reaction is the rate
determining step of the overall reaction. The rate
constant of the first reaction is inversely proportional
to the particle size of aluminum hydroxide. Both the
second and third reactions have been calculated to be
second order reactions. The second and third
reaction can be described as:
r2 = k2 * CC * CG (2)
r3 = k3 * CB * CD (3)
In these equations r2 and r3 are the rates of the
reactions and k2 and k3 are rate constants. In Eq. [2]
CC and CG are the concentrations of aluminum silica
fluoride and water, respectively. In Eq. [3], CB and
C D are the concentrations of aluminum hydroxide
and hydrogen fluoride, respectively. The third
reaction should be substantially fast because of the
acidic conditions.
The activation energy for the overall reaction has
been calculated to 21,3 kcal/(K*mol) (Dmitrevskij
and Semenova, 1970). An approximation in this
model is that the activation energy for the overall
reaction is the same as the energy for each
elementary reaction step. In other words the
activation energies for the three reactions E1 = E2 =
E 3 = 21,3 kcal/(K*mol) = 89121 J/(K*mol). To
calculate the frequency factors k01, k02 and k03 for
each reaction the Arrhenius equation has been used.
The rate constants for a specific temperature were
calculated according to Eqs. [1], [2] and [3]. The
Arrhenius equations can be described as:
k1 = k01 * exp(-E1/(R*T)) (4)
k2 = k02 * exp(-E2/(R*T)) (5)
k3 = k03 * exp(-E3/(R*T)) (6)
In these equations T is the temperature in Kelvin
and the gas constant R = 8.31 J/(K*mol). The
calculated frequency factors were used for
calculation of the rate constants when the
temperature of the reactor was varied. The results of
several simulations were compared with a standard
case. The standard case corresponded to an acid
concentration of 28.3 wt-%, a particle diameter of
100 µm and stoichiometrical amounts of fluosilicic
acid and aluminum hydroxide. Furthermore, the
preheating temperature of fluosilicic acid and
aluminum hydroxide were 65° C and 35°C,
respectively. In comparison with the standard case
the temperature profile of the simulation was not
correct. In other words the frequency factors did not
resemble the real operation. The reason of the
problem was that the specific surface area of the
aluminum hydroxide used in the experiments by
Dmitrevskij and Semenova (1970) was about 100
times greater than that of the hydrates used in
commercial processes. To correct the frequency
factors they were adjusted so that the simulation
corresponded to the standard case. The result of the
simulation is depicted in Figure 2, where the
symbols represent the measured temperatures of the
standard case and the line represents the simulation.
To get values of the factors that described the real
operation k01 had to be decreased and k03 had to be
increased, while the frequency factor of the second
reaction did not have to be corrected.
The perfect match corresponded to the following
frequency factors:
k01 = 8,1 * 109 s-1*(mol/l)-1
k02 = 7,6 * 108 s-1*(mol/l)-1
k03 = 2,0 * 1011 s-1*(mol/l)-1
Kinetic model
A kinetic model has to have different kinds of
elements that easily can be separated. At first it is
important to give the components in the reactions
different symbols, due to the fact that this makes the
model easier to follow and, if necessary, to correct.
Secondly, the different elements that the model shall
include have to be determined. In this case the
model will be partly based upon the reaction
mechanism and partly upon the size of the aluminum
hydroxide particles. Finally, for the purpose of
making the model complete, material and energy
balances are used together with the kinetics.
105
100
)
C simulation
( 95
e
r
u 90
t
a
r 85
e
p 80
m
e
t 75
70
65
60
0 100 200 300 400 500 600 700 800 900 1000
reaction time(s)
Figure 2. Simulation of the standard case
 As mentioned earlier the overall reaction is For material and energy balances general
divided into three steps. With symbols the schedule equations have been used. The equations for the
of the reaction becomes a bit easier to follow. With rates of the reactions can be transformed from terms
the symbols A=H2SiF6, B=Al(OH)3, C=Al2 (SiF6)3, of reactions into terms of species.
D=HF, E=SiO2, F=AlF3 and G=H2O the reaction can
be described in another way: rA = –3 * r1 (8)
rB = – 2 * r1 – 4 * r3 (9)
3 A + 2 B→ C + 6 G (II) rC = r1 – r2 (10)
rD = 12 * r2 – 12*r3 (11)
C + 6 G → 2 F + 3 E + 12 D (III) rE = 3 * r2 (12)
rF = 2 * r2 + 4*r3 (13)
12 D + 4 B → 4 F + 12 G (IV)
rG = 6 * r1 – 6 * r2 + 12 * r3 (14)
As mentioned earlier the rate equations of the
reaction can be described as: In these equations rA through rG are the rates of
change of the elements in mole/(l*s) according to
r1 = k1 * CA * CB2/3 * CB01/3 (1) the symbols presented earlier.
Finally, material balances can be presented.
r2 = k2 * CC * CG (2) These balances will calculate the change in the
amount of moles related to the reactions.
r3 = k3 * CB * CD (3)
dnA/dt = rA*V + FA (15)
The rate constant of the first reaction has to be dnB/dt = rB*V + FB (16)
adjusted for different particle sizes. The standard dnC/dt = rC*V (17)
case corresponds to a size of 100µm, which in this dnD/dt = rD*V (18)
model is called dp0N. This size gives the value k1N for dnE/dt = rE*V (19)
the rate constant of the first reaction. To calculate dnF/dt = rF*V (20)
the rate constant k1 for a different grain size dp0 the dnG/dt = rG*V + FG (21)
following equation can be used:
In these equations V is the volume in liters in the
k1=(dp0N/dp0)*k1N (7)
reactor. FA , FB and FG are the feed flows in moles
per second of fluosilicic acid (100%), aluminum
hydroxide and water in the acid solution,
Material balances respectively. The volume will increase as long as
The reaction cycle of the model has been the either fluosilicic acid or aluminum hydroxide still is
basis of the whole investigation. In this model pumped into the reactor. Therefore, a balance of the
fluosilicic acid is pumped into the reactor initially, at change of volume in the reactor has to be
the time t0, with a constant flow. At the time t1 the implemented in the model:
start of pumping aluminum hydroxide into the
reactor with a certain flow takes place. The flow of dV/dt = (m1/DA) + (m2/DB) (22)
aluminum hydroxide is decreased at t2, the so called
fine feeding is started. At the times t3 and t4 the feed In this equation m1 is the mass flow of acid and
of aluminum hydroxide and fluosilicic acid, water and m2 is the mass flow of aluminum
respectively, are stopped. At the latter time the hydroxide. Both in kg/s. D A is the density of
double amount of aluminum hydroxide in relation to fluosilicic acid in kg/l and it varies proportionally
fluosilicic acid has been pumped into the reactor. In with the concentration of the acid in the solution.
other words the reaction is performed This according to the next equation:
stoichiometrically. After the time t4 the reaction
proceeds until emptying of the reactor takes place at DA = 0.878*(acid conc. (%/100)) + 1.00 kg/l (23)
t5. In Table 1 all the time points of the different
moments are presented. DB is the density of the aluminum hydroxide and
it has a value of 2.4 kg/l.
Table 1. Different operations during one reaction cycle
Operation Time Thermodynamic data
Start of acid filling, t0 0s In the model ∆H f-values of the components are
Start of hydrate filling, t1 37 s used to calculate the heat of reaction of the three
Start of fine feeding of hydrate, t2 78 s reactions. Although the temperature increases during
Stop of fine feeding of hydrate, t3 90 s the reaction the ∆H f-values at 25°C will be used.
The ∆Hf- and Cp-values of the components are
Stop of acid filling, t4 128 s
presented in Table 2.
Emptying of reactor 700-800 s
Table 2. ∆Hf- and Cp-values at 25°C (SI Chemical Data) The first thing to do is to take the derivative of
Eq. [29]. Then the temperature T is solved from the
Component ∆Hf (J/mol) Cp (J/(K*mol)) equation. The result is an equation that describes the
H2SiF6 (aq) -2331300, HA 94∗, Cp1 temperature derivative in the reactor:
Al(OH)3 (s) -1276000, HB 93, Cp2
Al2(SiF6)3 (aq) -8066300∗, HC - dT/dt = (dH/dt – (T – 273.15)*Cpmix*∑dni/dt) /
HF(aq) -333000, HD -107 (Cpmix * (nA+nB+nC+nD+nE+nF+nG)) (30)
SiO2 (s) -847300, HE 44 Equation [30] is used to calculate the
AlF3 (aq) -1510000, HF 75 temperature in the reactor during the reaction.
H2O (aq) -286000, HG 75 Before the reaction takes place the temperature is
∗
estimated value
assumed to be constant, and then the temperature of
the preheated fluosilicic acid gives the temperature
The only two Cp-values that are used in the
in the reactor. When the reaction starts the
model are that of aluminum hydroxide and the
temperature increases adiabatically until 102°C,
assumed one of fluosilicic acid. A mean value for
which is the boiling point of fluosilicic acid. The
the Cp-value of the mixture in the reactor during the
concentration of the acid affects the boiling point.
reaction has been assumed through adjustment
Into this effect the model does not take any
between the temperature in the simulations with the
consideration. When the boiling point has been
temperature of the standard case. The Cp-value of
reached the temperature in the reactor in stays
the mixture, Cpm i x , has been estimated to 94
constant. In other words the temperature derivative
J/(K*mol).
is held constantly at zero both before aluminum
hydroxide is pumped into the reactor and after that
Energy balance time when the boiling point of fluosilicic acid has
been reached.
The heat of the reactions are represented as ∆H1
for the first reaction, ∆H2 for the second reaction and
∆H3 for the third reaction in the following equations: Results
The investigation was based upon how a few
∆H1 = 6*HG + HC – 2*HB – 3*HA (24) different parameters affect the reaction. These
parameters were the preheating temperature of both
∆H2 = 12*HD + 3*HE + 2*HF – 6*HG – HC (25) fluosilicic acid and aluminum hydroxide, the particle
size of aluminum hydroxide, the acid concentration
∆H3 = 12*HG + 4*HF – 4*HB – 12*HD (26) and a possible excess or deficit of aluminum
hydroxide. The results were evaluated in terms of
increased or decreased conversions, temperature
Finally, the energy balance can be derived. The profiles and enthalpy profiles in comparison with the
following equation describes the enthalpy change of base case.
the system during the reaction: Simulation of the base case was, as mentioned
earlier, performed with the following values of the
dH/dt = (–r1*∆H1 – r2*∆H2 – r3*∆H3)*V – (FA + FG)* parameters. The acid concentration was 28.3 wt-%,
*Cp1*(273.15 – T10) – FB*Cp2*(273.15 – T20) (27) the particle diameter 100µm and there were
stoichiometrical amounts of fluosilicic acid and
In Eq. [27] T10 and T20 are the temperatures of aluminum hydroxide at a mole ratio of 1:2.
the feed of fluosilicic acid and aluminum hydroxide, Furthermore, the preheating temperature of
respectively. Before aluminum hydroxide is pumped fluosilicic acid and aluminum hydroxide was 65°C
into the reactor, which in this model is before 37 and 35°C, respectively. According to commercial
seconds, no reaction takes place. Then the energy processes the emptying of the reactor is carried out
balance has to be described in another way: after approximately 700-800 seconds. This fact can
be of great interest when the parameters vary. The
dH/dt = – (FA + FG)*Cp1*(273.15 – T10) (28) conversions of the reactants can be compared with
the conversions of the base case. When the degree of
The enthalpy in the reactor varies with the the conversion increases, emptying of the reactor can
temperature according to: be performed earlier, while higher concentrations of
G
the products are received earlier. This could lead to a

H = Cpmix* ∑ ni * T (29)
more efficient production. The concentration profiles
for the base case are shown in Figure 3.
A
) Concentrations of the components
l 6
/
l H2SiF6
Al(OH)3
Conversion of fluosilicic acid
o
m 5 Al2(SiF6)3
( HF 2200
n SiO2 2000
o 1800
i AlF3
t 4 1600

Time (s)
a 1400
r 1200 50% conversion
t 1000 95% conversion
n 3 800
e 600
c 400
n
o 200
2 0
c
45 50 55 60 65 70 75 80 85 90
Preheating temperature of H2SiF6 (C)
1

0 Figure 4. Variation of conversion of fluosilicic acid with

0 100 200 300 400 500 600 700 800 900 1000 the preheating temperature of fluosilicic acid
reaction time(s)

Figure 3. Concentration profiles for the base case

Conversion of fluosilicic acid

The preheating temperature of fluosilicic acid 1600

and the acid concentration have the largest effects on 1400
the reaction. An increased temperature of the 1200
Time (s)

fluosilicic acid leads to a faster reaction and greater 1000

50% conversion
heat production. Even if measures like this could 800
95% conversion
600
lead to a more efficient process the heat production
400
causes some problems. An excess of enthalpy is
200
supplied to the reactor which could cause the reactor 0
content to violently boil and foam. Accordingly, a 15 20 25 30 35 40
decrease in the preheating temperature gives a Acid concentration (%)
slower reaction and a temperature profile that
finishes below 102° C. An increased acid
concentration affects the reactor in principle the Figure 5. Variation of conversion of fluosilicic acid with
same way as an increased temperature of fluosilicic the acid concentration
acid. This fact gives reason to lower the preheating
temperature of fluosilicic acid when the acid
) Concentrations of the components
concentration is higher than the normal case. l 6
Accordingly a higher preheating temperature is / H2SiF6
l Al(OH)3
necessary when the acid concentration is well below o
m
Al2(SiF6)3
5
28.3 wt-%. The reaction times, when 50% and 95% ( HF
n SiO2
conversion of fluosilicic acid are reached, variation o AlF3
i
with the preheating temperature of fluosilicic acid t 4
and the acid concentration are shown in Figures 4 a
r
and 5, respectively. In these diagrams the other t
n 3
parameters have the same values as the base case. e
c
The preheating temperature of aluminum n
o 2
hydroxide does not affect the reaction much. An c
increase in this temperature makes the reaction a bit
faster, but not much. The particle size of aluminum 1
hydroxide affects the reaction much more. With a
smaller particle diameter the rate of the first reaction
0
increases, which affects the total reaction. Smaller 0 100 200 300 400 500 600 700 800 900 1000
particles however give only a small excess of reaction time(s)
enthalpy. In other words a more fine material could
give a more efficient production. Bigger particles
lead to the opposite effect. The concentration Figure 6. Concentration profiles for a 50% decrease of the
profiles of a decreased particle diameter with 50% to particle diameter
50 µm are shown in Figure 6.
 An excess of aluminum hydroxide could in the
future be of interest due to the fact that it leads to a
higher pH-value without giving a filter cake that is
more difficult to handle. This measure could give a
better production because the aluminum fluoride
might be produced in a more pure form, free from
phosphorus. In other words a more phosphorus
containing fluosilic acid into the reactor would not,
with a pH-value at about 4 when the reaction is
completed, give a less pure aluminum fluoride. The
results of an excess of hydrate show that this would
lead to some what of a faster reaction without any
direct problems. A deficit of aluminum hydroxide on
the other hand leads to unwanted effects on the
process. The production of both aluminum fluoride
and silica decreases and large amounts of hydrogen
fluoride are produced.
Literature cited
1. Donald C. Gernes, Producing aluminum fluoride, patent
US 3,057,681, Oakland, California, 1962.
2. Arankathu Skaria, Felix Hartmann, Process for
producing aluminum fluoride, patent GB 2,049,647,
London, England, 1980.
3. L. D. Skrylev, Kinetics and mechanism of interaction of
fluosilicic acid with aluminum hydroxide, Zh. Prikl.
Khim, 1966 , 39(1), 58-64. Journal written in Russian.
4. G. E. Dmitrevskij, E. B. Semenova, Kinetics of
fluosilicic acid interaction with aluminum hydroxide,
Odess. Gos. Univ. im. Mechnikova, Odessa, USSR.
Izv. Vyssh. Ucheb. Zaved., Khim. Khim Tekhnol,
1970, 13(7), 960-962. Journal written in Russian.
Received for review April 22 2002