409 S University St.
SLC, Utah 84102
December 8, 2005
Professor Terry A. Ring
Department of Chemical Engineering
University of Utah
SLC, Utah 84112
Dear Dr. Ring:
The members of Group B (N.D. Brown, M.R. Siddoway, and myself) were asked
to explore the possibility of removing a 10-25% impurity of cumene from an isooctane
stream by vapor-phase catalytic decomposition to benzene and propylene during the
period of November 15 to December 6, 2005. In order to determine the practicability of
removal our objectives were to characterize the order of the reaction and the rate constant
as a function of temperature. Furthermore, we were asked to look at the catalyst
deactivation kinetics, specifically the amount of time needed to reach line out as well as
the reaction order and rate constant at line out conditions. The model carrier gas flowed
through a small-scale fixed bed reactor with low-grade SiO2-Al2O3 catalyst to decompose
a <5% impurity. Due to the lack of reproducible data the results were inconclusive and
most of the objectives could not be determined. However, we conclude that the removal
of cumene from an isooctane stream by decomposition to propylene and benzene is not
possible below 60% of the initial cumene present since our data showed that the extent of
the reaction was at a maximum of 0.4±0.1 for line out conditions. A précis of the
experiment and recommendations follow:
Located in room 3520 C of the Merrill Engineering Building (MEB), a glass tube
fitted within a heating jacket modeled the small-scale fixed bed reactor. The jacket was
manufactured by Lindberg, and a Honeywell controller modified the reactor temperature.
The catalyst was ground with a mortar and pestle and sifted with trays to obtain a 351800 um particle size. The prepared SiO2-Al2O3 catalyst was added to the tube and
positioned in the center by a glass-wool plug supported by a metal spring. The helium
used to model the isooctane was controlled using a Sierra mass flow controller. The
amount of cumene present in the stream was obtained by bubbling helium through liquid
cumene kept at a set temperature controlled by a Forma Scientific model # 2095 water
bath. The reactor was a continuously flowing system. The samples were taken from a
septum located after the cumene exited the reactor. Composition analysis of the product
stream was completed using a HP 5890A gas chromatography instrument (GC). An HP5 Cross linked 5% Ph ME Silicone 25 m long column was installed on the GC with a
flame ionization detector and analysis was completed using ChromPerfect Software. The
GC was located in MEB room 3215. Gas tight syringes (2 mL) were used to sample the
product streams. Both the reactor and the GC were shut down after use and restarted
each day.
The assumption was made that the helium stream becomes saturated with cumene
at the vapor pressure correlating to the bath temperature. This being the case the
maximum concentration of cumene was obtained at a set point of 60oC and using the
Antoine constants of cumene a concentration of 0.0015±0.0001 mol/L was calculated.
The settings on the GC eluted the benzene and cumene at 2 minutes and 4.95 minutes.
The procedure was as follows:
a)
The flow rate was varied from a minimum flow rate of 2 SLPM to a
maximum flow rate of 4 SLPM. This allowed a varied range of
residence time in the reactor.
b)
The reactor temperature was varied from 335oC to 450oC. This range
allowed the rate constant to be found as a function of temperature.
c)
Line out conditions were achieved at a reactor temperature of 450oC
and a maximum flow rate of 4 SLPM. These conditions were held
constant while samples were taken.
d)
Samples were analyzed on the GC using a 0.2 mL sample size.
Calibration of cumene and benzene was completed with equivalent
volumes in sample sizes of 0.1-0.4 uL diluted in acetone.
Because the carrier gas of the experiment and the GC carrier gas were the same,
we were unable to compare the initial amount of cumene in the feed stream to the amount
reacted using GC analysis. This required analysis of the conversion in the product stream
to rely solely on the respective areas of the products. A fundamental principle of gas
chromatography is that the burn ratio of chemicals should be the same. The burn ratio is
the ratio of the areas of equal amounts of chemicals on a GC experiment. Calibration of
the GC and an average of four measurements showed that the burn ratio of cumene to
benzene was 1:1 with a relative error of 30% as shown in Table 1. From this we were
able to conclude that the area of benzene present in the product stream sample would be
equivalent to the amount of cumene reacted to form the detected area. In the reaction,
cumene is assumed to decompose to propylene and benzene in equal amounts, therefore
the areas of the products should have a consistent burn ratio. In taking an average of the
burn ratio of benzene to propylene for the thirty four experiments the burn ratio was
found to be 0.8:1 with an error of 80%. From this high deviation in burn ratio we
suspected that something was erring in the amounts of propylene and benzene collected.
Further evidence for this suspicion is shown in repeated experiments that show an
error of greater than 20% in conversion as shown in Table 2. This error was consistent
among all repeated experiments averaging to >50% relative error. Assuming that the
reactor is converting the same amount of cumene to benzene and propylene the error
must be attributed to the sampling. Variation in obtaining the sample is attributed to the
small measurement samples taken. At 0.20±0.05 mL the accuracy in the samples due to
small measurement is very low. Taking a larger sample size would attenuate the error
due to its relativity. Further error in sampling the product stream could be caused by the
residual amounts of components left in the needle. Flushing the needle with the stream
before obtaining the new sample may not have purged the needle of the residual
components, which might have condensed. These remaining concentrations of
propylene, benzene or cumene left in the needle may have subsequently been injected
into the GC thereby causing the areas to be non-representative of the sample taken. The
cause in the condensation could be due to the time-delay from taking the sample to
injecting the sample. At the point where samples were taken, the stream was cooled to
ambient temperature due to the surroundings. Because of this, I do not expect
condensation to be large. However, because liquid and gas samples are detected with
different sensitivity it can be logically concluded that the conversion would be detected
as widely different for two similar runs dependent on the amount of condensation or
residual components in the needle. The enormous error and irreproducible results are
attributed to this suspicion. Consequently, most of the results are inconclusive in the
determination of the objectives.
The principle findings of our experiment are such that the maximum conversion
of cumene is estimated to be no greater than 0.4±0.1 (taken at close to line out
conditions). This value is found by averaging the line out conditions. As shown in
Figure 1, the line out conditions can be approached at 450oC and a flow rate of 4 SPLM
for more than two hours. The measurements show a slight downward trend, yet the noise
in measurement implies grouping of conversion amounts instead of a trend. This analysis
leads to the reasoning that the removal of cumene from an impure stream is unpractical.
At that conversion, the stream will retain at least 60% of the original cumene impurity
after reaction with the catalyst. This conclusion would be supported by characterizing the
reaction kinetics. Investigation of the data showed that to get a reasonable standard
deviation with the amount of inaccuracy as mentioned, at least five samples of the
product stream would be needed at each variation of residence times and reactor
temperatures. The maximum number of runs completed in three hours is approximately
20 samples, therefore in order to obtain reproducible results within a 10% error the time
allotted for the four objectives would need to be increased significantly. If the
decomposition of cumene is assumed a first order reaction the data should show a linear
trend. Determination of the deactivation kinetics of the catalyst requires the order of the
reaction to be known and an initial guess of the order of the deactivation kinetics.
According to Fogler (2002)1, the first order decay kinetics and the first order reaction are
can be used to find the decay order of the reaction as shown in Equation 1.
 k d t  n( 0 / Wk )  n(n(Cao / Ca )) .
(1)
Plotting the residence time versus natural logarithm of the natural logarithm of the ratio
of initial concentration to final concentration should show a linear trend if the decay and
reaction are, in fact, found to be first order. However, in order to use the equations, a
reasonable standard deviation would need to be used to have any confidence in the
reaction order and kinetic rate constants obtained.
The catalytic reactor is assumed to be performing adequately and consistently. It
is also reasoned that the GC is running with a 30% relative error. The following
problems in the method of sampling the product stream should be addressed before this
lab can provide conclusive results:
a)
The time delay between taking the samples of the data and injection
into the GC would need to be decreased. This would ensure that
1
Fogler, H. Scott. Elements of Chemical Reaction Engineering. Prentice-Hall, Inc., New Jersey (2002)
b)
c)
condensation does not occur and that the composition of the product
stream injected is similar to what it was when sampled. Decreasing the
time delay could be achieved by moving the analysis instrument closer
to the reaction process, or decreasing elution time on the instrument.
The method of sampling would need to be calibrated to account for
error. Inline injection would ensure that the sample is injected
uniformly for all experiments. Another recommendation would be to
provide a chemical or inert air stream to be used in completely purging
the needle of any residual components before taking subsequent
samples.
Adequate time is given to complete all objectives. With the necessity
of taking five samples due to the deviation, only two of the four
objectives could have been met. A lower deviation in measurement
would require at least three samples due to the standard 30% error in
the instrument.
If these recommendations can be implemented the cumene reaction kinetics can
be adequately characterized and the objectives met. This would allow supporting
evidence to be obtained for whether or not the removal of cumene impurity from an
isooctane stream is possible.
Sincerely,
Kara Stowers
Table 1. Calibrated burn ratio of Cumene:Benzene
As shown the Cumene to Benzene has a 1:1 burn ratio. From this data we see
that for equal amounts of cumene and benzene, equal areas will be read by the
GC. This allows a basis for conversion.
Cumene
Benzene
Ratio
Calibration Peak Area Peak Area
Burn Ratio
Average
Deviation
1
447538
619426
0.723
2
370375
349760
1.059
1.0
0.3
3
527309
501891
1.051
4
1185373
882940
1.343
Table 2. Repeated Experiments with very high error and deviation
Experiments were completed at a reactor temperature of 400oC and a water bath
temperature of 40oC. Similar experiments show a relative error of 80% for a
flow rate of 2 SPLM and a relative error of 20% for a flow rate of 4 SPLM.
Individual Peak deviation was calculated assuming a 20% error in GC
determination of the peak area.
Flow
Cumene
Benzene
Peak
Deviation
Rate(SPLM) Peak Area Peak Area Conversion Deviation between runs
2
44543
12563
0.22
0.05
0.38
2
39550
128340
0.76
0.05
4
192156
322388
0.63
0.07
0.12
4
69497
56878
0.45
0.07
Figure 1. Line out conditions at a temperature of 450oC and flow rate of 4 SPLM
Line out was approached as shown in the figure. The general trend was
decreasing but as shown the deviation in measurement showed huge variance
over time.
conversion (cumene)
Approaching Line Out
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
20
40
60
time (min)
80
100
120