Professional Documents
Culture Documents
C. Dierent
liquid feeds are dosed with a peristaltic pump. The feed consisting of water,
TBA and sulfuric acid is prepared at low temperatures because otherwise
the heat of mixing leads to the formation of oligomers.
2
Mass ow rates into and out of the CSTR are measured. At the end of
cooler 1 (gas exit) and cooler 2 (liquid exit), gas, respectively liquid samples
are taken. Inside cooler 2 a siphon prevents gas from leaving the reactor. So
gas leaves at end of cooler 1 and liquid at end of cooler 2. For residence time
computation a colored tracer is used which is detected by a phototransistor
at cooler 2.
At steady-state condition the liquid samples are cooled down immedi-
ately by means of liquid nitrogen. Each gas sample is collected over a xed
time interval, so that the ow rate of isobutene is determined by the sam-
ples weight. It can be achieved that the liquid compounds are completely
condensed by cooler 1, hence gas samples contain isobutene exclusively.
The composition of feed and outlet samples are determined from infrared
spectroscopy measurements. For this purpose the ATR-technique (attenu-
ated total reection) is employed. The ATR-crystal consists of ZrO
2
and
is resistant against strong mineralic acids. Calibration is done by means of
the chemometric PLS-method (partial least squares), where weight percents
instead of molar concentrations is used to avoid temperature dependence.
The actual concentrations are then calculated using density functions, which
are computed from density measurements at 20
C, 25
C, 30
C. It is shown
in Steinkemper [4] that these functions are also valid for the applied temper-
atures of 63
C, 68
C, 73
C.
3 Reaction Mechanism
Motivated by studies of Xu [5], whose research describes the dehydration of
t-butanol in compressed liquid water at 250
6.41
c
T
c
W
, (12)
=2.718 + 2.755 exp
8.69
c
T
c
W
. (13)
Due to the fact that most of the intermediate species are inaccessible to
measurements, the individual rate constants k
i
, k
z
i
in (1)-(8) cannot be ob-
tained with reasonable eort. On the other hand, the intermediate species
c
2
, c
3
, c
4
and c
E
are expected to be short-lived, hence application of quasi-
steady-state-approximation (QSSA) for these chemical species is allowed.
After application of QSSA, performed by means of a lengthy but elemen-
tary computation, it turns out that dehydration of TBA can be modeled by a
single reaction rate which formally corresponds to the acid catalysed overall
reaction
T
PSfrag replacements
k
F
k
B
I + W. (14)
The rate function of this overall reversible reaction is given by
r = k
F
a
P
c
T
+ k
B
a
P
c
I
c
W
. (15)
The structure of r is the same as if the reaction rate would follow from
mass action kinetics, but the rate constants that result from the QSSA-
procedure turn out to be rather complex functions of c
T
, c
W
and a
P
. These
5
read as
k
F
= p
1
1 + p
2
c
T
1 + p
3
c
W
+ p
4
a
P
1 + p
5
c
W
+ p
2
1 + p
3
c
W
1 + p
3
c
W
+ p
4
a
P
c
T
(16)
and
k
B
= p
6
p
5
+ p
2
p
3
c
T
1 + p
3
c
W
+ p
4
a
P
1 + p
5
c
W
+ p
2
1 + p
3
c
W
1 + p
3
c
W
+ p
4
a
P
c
T
. (17)
Here the parameters p
i
are related to the rate constants k
i
as follows.
p
1
=
k
1
k
2
k
z
1
+ k
2
, p
2
=
k
4
k
5
k
3
(k
z
4
+ k
5
)
, (18)
p
3
=
k
7
k
6
, p
4
=
k
z
4
k
z
5
k
6
(k
z
4
+ k
5
)
, (19)
p
5
=
k
z
1
k
z
2
k
3
(k
z
1
+ k
2
)
, p
6
= k
z
3
. (20)
5 Validation
Experimental validation of the kinetic model is done by means of CSTR-
measurements. In contrast to batch experiments, this allows to keep the
reaction volume constant during each measurement. Mass balances lead to
V
dc
T
dt
=
V
f
c
f
T
V c
T
V k
F
a
P
c
T
+ V k
B
a
P
c
I
c
W
, (21)
V
dc
W
dt
=
V
f
c
f
W
V c
W
+ V k
F
a
P
c
T
V k
B
a
P
c
I
c
W
, (22)
V
dc
I
dt
= V k
F
a
P
c
T
V k
B
a
P
c
I
c
W
k
L
A(c
I
c
g
I
/H
I
) , (23)
V
g
dc
g
I
dt
= k
L
A(c
I
c
g
I
/H
I
) n
out
. (24)
Under steady-state conditions, the two balances for isobutene, i.e. (23), (24)
lead to
n
out
= k
L
A(c
I
c
g
I
/H
I
) = V (k
F
a
P
c
T
k
B
a
P
c
I
c
W
) = V r. (25)
6
This can be simplied to
n
out
= V r
F
= V k
F
a
P
c
T
, (26)
since the concentration of isobutene in liquid phase is very small due to
spontaneous phase change of isobutene into a gaseous phase. Note that the
isobutene balance is most appropriate for validation, since the mass ux of
isobutene can be measured with high accuracy.
Finally, use of V =
V with residence time together with
V c
T
= n
T
=
m
T
M
T
(27)
leads to the following formulation of the isobutene balance, which is suitable
for measurement analysis and parameter estimation:
m
g
I
M
I
= n
out
=
m
T
M
T
a
P
k
F
(28)
with k
F
from (16).
We performed experiments under isothermal steady state conditions at
three dierent temperatures for various feed compositions. Evaluation of
the measurements is done by means of nonlinear t of model parameters
p
1
, . . . , p
5
in equation (16). For this purpose we utilized only experimental
data corresponding to those 13 feed compositions where reliable measure-
ments were possible at all temperatures. The feed compositions in these
measurements are shown in the following Table 1. Parameter tting shows
that not both parameters p
2
and p
5
are signicant, but only their ratio is.
This leads to arbitrarily large values for p
2
and p
5
, hence the tting algo-
rithm is not convergent. Note that large values for p
2
and p
5
are reasonable,
since both parameters contain the factor 1/k
3
, and the reaction step with
rate constant k
3
characterizes the nonpreferred reaction path, i.e. k
3
takes
small values [10]. To overcome this defect, we x p
2
at a suciently large
value and use the ratio p
5
= p
5
/p
2
as a new parameter. This corresponds to
the following simplied relation for k
F
.
k
F
=
p
1
c
T
1 + p
3
c
W
+ p
4
a
P
p
5
c
W
+
1 + p
3
c
W
1 + p
3
c
W
+ p
4
a
P
c
T
. (29)
7
no. c
f
A
[mol/L] c
f
W
[mol/L] c
f
T
[mol/L]
1 31.3 15.6 53.1
2 30.9 15.2 53.9
3 29.8 15.6 54.5
4 29.0 29.0 41.9
5 30.2 34.2 35.6
6 30.7 20.4 48.9
7 35.4 23.6 41.0
8 25.5 12.2 62.3
9 26.6 7.9 65.5
10 24.8 7.9 67.3
11 33.8 28.1 38.1
12 30.4 4.6 65.0
13 19.4 2.7 77.9
Table 1: Feedcompositions
The results of the parameter estimation is shown in Table 2, while the qual-
ity of tting the model predictions to the measurements is illustrated by
Figures 3, 4 and 5. Unfortunately, due to the restricted number of these
quite elaborate experiments, a specication of statistical quantities is not
reasonable here. Nevertheless, the gures show a sucient agreement.
63
C 68
73
p
1
[ - ] 0.0071 0.0074 0.0345
p
3
[L/mol] 0.7821 0.2684 0.2182
p
4
[ - ] 0.6633 2.1985 10.6578
p
5
[ - ] 7.5169 2.5244 1.7657
Table 2: Parameter
The values of the obtained model parameters are dicult to interpret on
the basis of individual reaction constants, since each term is a combination
of several rate constants. For this reason the activation energies cannot
be calculated. Still, the model parameters show a monotonic dependence
on temperature. The model shows that formation of isobutene due to the
decomposition process of the ether is relevant and its contribution increases
8
with temperature.
6 Simulation of the dehydration of t-butanol
in a batch reactor
At a rst glance, the kinetic model developed above seems unapplicable for
the real process since the formation of oligomers has been omitted. But
production of oligomers has to be avoided anyhow, hence a combination
of our model with the production rates of oligomers, obtained empirically
by Allenbach [2] can be used for process optimization in the regime where
formation of oligomers is negligible.
We illustrate this approach in case of batch simulations. Since the batch
volume V changes signicantly during the reaction, we have to formulate the
balance equations in terms of the total amounts of molar masses instead of
concentrations. Together with (15) this leads to the system
dn
T
dt
=k
F
(a
P
,
n
T
V
,
n
W
V
) a
P
n
T
, n
T
(0) = n
0
T
dn
W
dt
= k
F
(a
P
,
n
T
V
,
n
W
V
) a
P
n
T
, n
W
(0) = n
0
W
dn
I
dt
= k
F
(a
P
,
n
T
V
,
n
W
V
) a
P
n
T
, n
I
(0) = n
0
I
(30)
with k
F
(a
P
, c
T
, c
W
) from (16), where a
P
is computed according to (10) - (13).
Simultaneously, the amount n
Ol
of oligomers is calculated by solving the
initial value problem
dn
Ol
dt
= V r
Ol
(
n
T
V
,
n
W
V
; T, c
0
T
, c
0
A
), n
Ol
(0) = 0, (31)
where r
Ol
(c
T
, c
A
; T, c
0
T
, c
0
A
) is the rate of formation of oligomers at tempera-
ture T and for initial composition c
0
T
, c
0
A
due to Allenbach [2]. To keep the
paper self-contained, this correlation is reproduced here. It reads as follows,
where temperature, time and molar concentrations are made dimensionless
using K, min and
mol
L
as reference dimensions, for T, t and c, respectively.
r
Ol
= k(T) (c
0
T
, c
0
A
)(c
T
, c
A
) (32)
with
log k(T) = 25.81
6539
T
, (33)
9
log (c
0
T
, c
0
A
) = 12.6(0.414c
0
T
1) log c
0
A
38.5 log c
0
T
, (34)
(c
T
, c
A
) = c
2
T
c
7
A
. (35)
Let us note in passing that this experimental correlation is based on the
molar acid concentration c
A
instead of an acidity function. This makes the
large exponent in the last equation plausible.
Since the batch volume V is not constant, we use conservation of sulfuric
acid, i.e.
V
0
c
0
A
= V (t)c
A
(t), (36)
for solving (30) and (31). Here c
A
(t) is determined from the relation
c
A
=
1000(c
T
, c
W
, T) 74c
T
18c
W
98
(37)
with the experimentally obtained correlation (see [4])
(c
T
, c
W
, T) = 2.55 + a
1
(T 293.15)
+(0.163 a
2
(T 293.15)) c
T
(38)
+(0.028 a
3
(T 293.15)) c
W
.
In (38), all concentrations are dimensionless with 1 mol/L as reference quan-
tity and the parameter values are
a
1
= 1.84 10
4
, a
2
= 3.02 10
4
, a
3
= 4.45 10
5
.
Altogether, given n
T
, n
W
as well as T and c
0
A
, the batch volume V is implic-
itly determined by the relation
V
0
c
0
A
V
=
1000(
n
T
V
,
n
W
V
, T) 74
n
T
V
18
n
W
V
98
. (39)
Knowing the molar masses hence allows for computing the corresponding con-
centration. The ratio between production of oligomers and that of isobutene
is then given by
n
Ol
(t)
n
I
(t)
=
V
0
c
0
A
n
I
(t)
t
0
r
Ol
(c
T
(s), c
A
(s); T, c
0
T
, c
0
A
)
c
A
(s)
ds. (40)
10
Now we proceed as follows. The numerical calculation is continued until
either
n
Ol
(t)
n
I
(t)
= q
crit
at a rst t = t
crit
or (41)
1
c
T
(t)
c
0
T
> U
at a rst t = t
, (42)
i.e. until a critical fraction of oligomers (say q
crit
= 0.01) or the desired
conversion, say U
} (43)
and the associated conversion
U(c
0
T
, c
0
A
) = 1
V (t) c
T
(t
proc
)
V (0) c
0
T
. (44)
Hence U(c
0
T
, c
0
A
) denotes the conversion which is reachable with suciently
small production of oligomers. This is shown in Figure 6, where the plateau
region corresponds to those initial conditions for which our model is valid
because formation of oligomers is negligible. Recall that the rate function
(15) is based on experiments without formation of oligomers.
For optimization of the dehydration process we dene a performance by
means of
P(c
0
T
, c
0
A
) =
0 if t
proc
= t
crit
U(c
0
T
, c
0
A
)
t
proc
if t
proc
< t
crit
, (45)
where the performance is dened to be zero if production of oligomers is not
negligible. The process performance is shown in Figure 7, which illustrates
the existence of an optimal initial composition at approximately (c
0
A
= 3.8
mol/L, c
0
T
= 8.0 mol/L, c
0
W
= 3.32 mol/L) within realizable range of initial
concentrations.
The optimal composition corresponds to about 5% of water and is only
achievable by mixing almost concentrated sulfuric acid with TBA. In practice
one has to be careful concerning release of mixing enthalpies, since this can
lead to oligomerization. The peaks that can be noticed in Figure 7 are due
to numerical instabilities and hence articial.
11
The dynamical behavior during the dehydration process is illustrated by
Figures 8-11, where the initial composition is chosen as
c
0
T
= 8 mol/L, c
0
A
= 3.8 mol/L, c
0
W
= 3.32 mol/L (c
0
I
= c
0
Ol
= 0 mol/L).
For this feed composition and for a batch volume of 1L, say, the total amount
of oligomers, produced until a conversion of 95% is reached, is about 0.04
mol. Hence the ratio n
Ol
/n
I
is below 0.5% (cf. Figures 8,9). Observe also that
the acidity attains a gentle maximum at approx. 200 min (see Figure 10).
This agrees with the changes in volume and in water concentration. Indeed,
the fast decrease of the batch volume (cf. Figure 11) leads to an increase of
the acid concentration in the initial phase, followed by a decrease in the long
run due to dilution by increasing amounts of water.
7 Conclusions
The dehydration of t-butanol (TBA) in sulfuric acid is a common way for the
production of very pure isobutene. The demands concerning the quality are
high, in particular for the synthesis of polymers with high molecular weight.
Depending on composition of the acid phase, cleavage of TBA is problematic
since it leads to dierent yields of isobutene. Especially at temperatures
above 70
C, 68
C and 73
C.
Under quasi-steady-state-assumptions concerning intermediates, a sim-
plied kinetic model is obtained via QSSA. Coupling of the resulting kinetic
equations to CSTR balance equations leads to the complete mathematical
model. Fitting of the remaining model parameters to experimental data
yields parameter values which allow for better understanding of the reaction
12
mechanism. In particular, the formation of isobutene via decomposition of
DTBE is conrmed by the values of the tted model parameters.
Combining this model with the results of Allenbach, numerical simula-
tions are done to compute reaction conditions, for which oligomerization of
isobutene is avoided and yield of isobutene is maximized.
13
Nomenclature
A interfacial area [m
2
]
H
0
Hammett acidity function
H
I
Henry coecient
K equilibrium constant
M molar mass [g]
P performance
T temperature [K]
U conversion
V reaction volume [L]
V volumetric ux [L min
1
]
a activity
c concentration [mol L
1
]
c
2
concentration of t-butyloxonium ion [mol L
1
]
c
3
concentration of t-butylcation [mol L
1
]
c
4
concentration of protonated di-t-butylether [mol L
1
]
k reaction rate constant [(L mol
1
)
m1
min
1
] for m-th order
k
L
mass transfer coecient
m mass [g]
m mass ow [g s
1
]
n molar amount of species [mol L
1
]
n mol ow [mol s
1
]
p, p
parameter
q mol ratio
r reaction rate
s integration variable
t time [min]
Greek symbols:
tting parameter of Hammett function [L mol
1
]
tting parameter of Hammett function [ - ]
, correlation function
density [kg L
1
]
residence time [min]
14
Subscripts:
0 initial value
A sulfuric acid
B reverse reaction
E di-t-butylether (DTBE)
F forward reaction
I isobutene
T t-butanol
W water
Ol oligomers
crit critical
out outlet
proc process
innity
Superscripts:
0 initial value
f feed
g gas phase
z reverse reaction
Abbreviations:
CSTR continuous stirred tank reactor
DIB diisobuten (2,4,4-trimethyl-penten-1 and 2,4,4-trimethyl-penten-2)
DTBE di-t-butylether
TBA t-butylalcohol
ATR attenuated total reection
References
[1] H.-J. Warnecke, J. Pr uss, B. Bienek, R. G. Presenti, Modeling isobutene
extraction from mixed C
4
-streams, Chem. Eng. Sci. 47, 533-541 (1992).
[2] U. Allenbach, Untersuchung der schwefelsauer katalysierten Bildung von
Oligomeren des Isobutens, PhD-Dissertation, Berlin (1978).
[3] U. Allenbach, B. Wichtendahl, W.-D. Deckwer, Makrokinetik und Se-
lektivitat der Bildung von Oligomeren des Isobutens, Erd ol+Kohle 31,
353-355 (1978).
15
[4] A. Steinkemper Modellierung der Dehydratisierung von t-Butanol
unter Vermeidung der Oligomerenbildung, PhD-Dissertation, Paderborn
(2001).
[5] X. Xu and M. J. Antal, Kinetics and Mechanism of Isobutene Forma-
tion form t-Butanol in Hot Liquid Water, AIChE Journal 40, 1524-1534
(1994).
[6] E. J. Smutny and A. Bondi, Di-t-Butyl Ether: Strain Energy and Phys-
ical Properties , J. Phys. Chem. 65, 546-550 (1961).
[7] J. O. Fenwick, Thermodynamic Properties of Organic Oxygen Com-
pounds: Enthalpies of Formation of Eight Ethers, J. Chem. Thermo-
dynamics 7, 1943-1947 (1975).
[8] Hammett, L. P.: Physical organic chemistry : reaction rates, equilibria,
and mechanisms. (2
nd
ed.), McGraw-Hill 1970.
[9] A. Friedrich, H.-J. Warnecke, H. Langemann Die Hammettsche Acidi-
t atsfunktion von Schwefels aure in Wasser-t-Butanol L osungen, Z. Phys.
Chemie 265, 11-16 (1984).
[10] E. Velo, L. Puigjaner Inhibition by Product in the Liquid-Phase Hy-
dration of Isobutene to tert.-Butyl Alcohol: Kinetics and Equilibrium
Studies, Ind. Eng. Chem. Res. 27, 2224-2231 (1988).
16
0.00g 0.00g
PSfrag replacements
balances
measure unit
PC
pump
cooler 2
cool traps
cooler 1
T
exible connection
reactor
magnetic stirrer
Figure 1: Experimental set-up
+
+
+
+
PSfrag replacements
k
1
+H
+
H
+
k
z
1
+H
+
k
z
3
k
3
H
+
k
2
H
2
O
+H
2
O
k
z
2
k
7
+H
2
O
+H
+
k
z
5
k
5
H
+
O
O
O
O
H
OH
OH
OH
OH
OH
2
+
+
c
T
c
2
c
3
c
I
c
4
c
E
k
4
k
z
4
k
6
2
Figure 2: Postulated reaction mechanism of the dehydration of t-butanol
17
PSfrag replacements
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.0
1.0
2.0
3.0
4.0
5.0
n
I
[
1
0
4
m
o
l
/
s
]
No. of Experiments
Model
Samples
Figure 3: Fit for 63
C
PSfrag replacements
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
n
I
[
1
0
4
m
o
l
/
s
]
No. of Experiments
Modell
Samples
Figure 4: Fit for 68
C
18
PSfrag replacements
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
0.0
1.0
2.0
3.0
4.0
5.0
6.0
7.0
8.0
9.0
10.0
n
I
[
1
0
4
m
o
l
/
s
]
No. of Experiments
Modell
Samples
Figure 5: Fit for 73
C
19
3.5
4
4.5
5
c
A
@molLD
4
5
6
7
8
c
T
@molLD
0.4
0.6
0.8
1
Conversion
3.5
4
4.5
5
c
A
@molLD
Figure 6: Simulation of conversion depending on c
0
T
und c
0
A
3.5
4
4.5
5
c
A
@molLD
4
5
6
7
8
c
T
@molLD
0
0.02
0.04
0.06
0.08
Performance
3.5
4
4.5
5
c
A
@molLD
Figure 7: Simulation of process performance depending on c
0
T
und c
0
A
20
0 200 400 600 800 1000
t @minD
0
2
4
6
8
n
I
@
m
o
l
D
Figure 8: Simulated molar mass of isobutene
0 200 400 600 800 1000
t @minD
0
0.005
0.01
0.015
0.02
0.025
0.03
0.035
n
O
l
@
m
o
l
D
Figure 9: Simulated molar mass of oligomers
21
0 200 400 600 800 1000
t@minD
0
20000
40000
60000
80000
100000
a
H
H
2
S
O
4
L
@
m
o
l
L
D
Figure 10: Evolution of acidity
0 200 400 600 800 1000
t @minD
0.2
0.4
0.6
0.8
1
V
s
V
0
@
D
Figure 11: Simulated change in relative acid volume
22