Salt Lake City, UT 84103

475 E 6th Ave. #13

Salt Lake City, UT 84103

December 8, 2005

Professor Terry A. Ring

Department of Chemical Engineering

University of Utah

Salt Lake City, UT 84112

Dear Dr. Ring:

During the period of November 15 to December 6, 2005, the members of Group B, which consisted of Kara Stowers, Nathan Brown, and myself, analyzed the vapor-phase catalytic decomposition of cumene to benzene and propylene. In the analysis, there were four main objectives: (1) to determine the reaction order with respect to cumene and rate constant as a function of temperature, (2) to determine deactivation kinetics, (3) to determine the time how long it takes for the catalyst to achieve line-out conditions, and (4) to determine the reaction order with respect to cumene and rate constant as a function of temperature at line-out conditions. To accomplish these, a small fixed-bed, catalytic reactor was loaded with a small amount of SiO2-Al2O3 catalyst. The reactor temperature was varied from 335°C to 450°C and helium gas was used as the carrier to simulate the inert, isooctane, in the customer’s system. It was hypothesized that the reaction was first order with respect to cumene. When analyzing the data, it was found that there was a large amount of deviation in the analytical measurements, mainly from the gas chromatograph (GC). Because of this, the objectives were not met in the allotted time to perform the experiments. Even when maximum conversion was achieved, only is 40% ± 10% of the original cumene would have reacted. From the analysis, it was concluded that other methods of cumene removal should be investigated. Several recommendations were also made to minimize error to find reasonable results for future experiments with this reactor.

The fixed-bed catalytic reactor is located in the Senior Lab, Merrill Engineering Building (MEB) 3520C. The four main parts of the reactor consist of: (1) the helium mass flow controller, (2) the cumene bubble vessel, that sits in a temperature controlled water bath, (3) the reactor tube, and (4) the furnace, which is used to control the temperature of the reactor. Helium flows from a cylinder to the mass flow controller, which was manufactured by Sierra, which operates by measuring thermal conductivity of the flow stream against a small reference stream to control a variable position solenoid valve. After the mass flow controller, helium can flow into the cumene bubble vessel then to the reactor, or bypass it to purge pure nitrogen through the reactor. It was assumed that the helium would be saturated with cumene coming out of the bubble vessel, and if not, then at least at a steady state value. The temperature of the water bath controlled the vessel temperature. The water bath was manufactured by Forma Scientific, model 2095. The reactor tube was simply a glass tube with an inner diameter measured to be 8.0 ± 0.1 mm, which had a glass wool plug midway down the tube holding a plug of SiO2-Al2O3 catalyst, which weighed 1.55 grams. The catalyst was manufactured by Davison Specialty Company, grade 980-25%, sieved to a size range of 351 to 850 μm using Tyler Screens 42 and 20. The reactor was held in the heater using a ring stand and a clamp above and below the furnace. The reactor was heated by a Lindberg furnace, which was controlled using an external Honeywell PID controller. The reactor temperature was the controlled variable and the heater power input was the manipulated variable. There were sampling septa before and after the reactor, where samples could be removed using a syringe. The samples were then analyzed using gas chromatography (GC). The gas chromatography instrument used is located in the Instrumental Analysis Lab, MEB 3215. The instrument was a Hewlett-Packard 5890A, which was equipped with a flame ionization detector (FID). The column was an HP-5 Crosslinked 5% Ph ME Silicone 25 m (lenth) ´ 0.32 mm (ID) ´ .53 mm (film thickness). This instrument was connected to an interface PC running with ChromPerfect software to log the chromatograph generated from the instrument.

In order to analyze reaction kinetics, samples were generated using the packed-bed catalytic reactor. The water bath was set to a constant bath temperature of 60°C, which would produce an initial concentration of cumene at 0.0125 ± .0001 mol/L. Operation of the reactor was as follows:

1. The flow rate of the carrier gas, helium, was varied between 2 to 4 SLPM to achieve different residence times in the reactor.
2. The reactor temperature was varied from 335°C to 450°C using the external PID temperature controller to find reaction rate as a function of temperature.
3. Change the inlet composition by changing the water bath temperature to 30°C to find the reaction order.
4. To achieve line-out conditions, reactor temperature was held constant at 450°C and the flow rate of carrier was held at a constant 4 SLPM. This procedure was run until steady state conversions were measured to attempt to achieve line-out conditions.

The samples were analyzed using the GC, which was operated according to the standard operating procedure provided. The GC settings were set such that the elution times on the chromatograph for benzene and cumene were at approximately 2 minutes and 4.5 minutes, respectively. The sample size for gas samples was 0.2 mL and 0.1-0.4 μL for liquid calibration samples. The GC was calibrated by mixing a 50:50 by volume mixture of pure benzene and cumene, then diluting with a solvent. This calibration was analyzed to find how the peak areas of benzene and cumene related to each other. The samples could then be analyzed using the differential method[1] to find rate constant and order of reaction. Assuming the reaction and decay are both first order, the deactivation kinetics may be found by fitting the data to the equation:

where kd is the rate constant at line out, W is the catalyst weight, k is the rate constant, CAO is the initial concentration of cumene, and CA is the concentration of cumene at a point, or in the product stream[2].

In the past, the samples could be analyzed using a GC that was located in the same room as the reactor. This instrument was not working, so it was required to transport the sample from the reactor to a separate room down the hall to a separate room where the GC that was used was located. It was also found that analyzing the data using the GC did not generate very reproducible data. One principle of gas chromatography says for equal amounts of two different components, the burn ratio, the ratio between the areas of the two respective peaks, will be constant. When calibrating the instrument, it was found that for the 50:50 benzene and cumene mixture, the burn ratio was 1.0 ± 0.3 after three measurements (Table 1). This had a very good injection and sample preparation. This shows that there is a significant error in the instrument itself. In the reaction from cumene to benzene and propylene, there is a 1:1 production rate of benzene and propylene. There should be a constant burn ratio of benzene and propylene. It was found that for all of these measurements the average was 0.8 ± 0.6. It was also shown that for several samples at the same conditions, the measurements were inconsistent (Table 2). This proves that the measurements from the GC were too far off to make any conclusions concerning reaction kinetics. These are off due to errors in sampling, transportation and injection. Because of the time it took to remove a sample until the time of injection into the GC, some of the components, mainly benzene and cumene, could have condensed inside the syringe or some may have diffused out of the needle. There are also some residual components from former samples in the needle, which might not have been completely purged out of the syringe and needle. Also, one of the major constraints on this data was the time. If there were more time allotted to do the experiment, more samples could have been taken to eliminate some of the uncertainty. There needs to be at least five measurements at each condition to obtain the deviation of the measurement. In order to achieve all of the objectives, it was required that hundreds of samples be analyzed. The GC gave a time restraint, because each sample required about 8 minutes to run and bring back to initial conditions. There simply was not enough time allotted to make any reasonable conclusions.

It was attempted to achieve line-out conditions, by running the reactor at high temperature and the same feed concentration for a long period of time. It may not be determined whether line-out conditions were met; however, at long periods of time several measurements seemed to be around the same value (Figure 1). The maximum conversion at this point was found to be 0.4 ± 0.1. This means that there was only about 40% of the original cumene fed to the reactor that was converted. This means that there is still a high amount of cumene in the stream. If cumene affects the downstream process, there may have to be other reactions or separations to attempt to remove the cumene from the stream, such as a flash vessel or distillation column.

There are several recommendations that can be made to this project more fruitful in the future. The first is that an inline GC be implemented in the product stream of the reactor. There was a lot of our error that can be attributed to the sampling method. An inline GC would eliminate a lot of the human error and also the errors associated with condensation of components. If an inline GC may not be obtained, it would be helpful if the GC were at least located in the same room as the reactor. This would allow the operators to take a sample when they needed to, instead of taking a sample, transporting it, and waiting until the last run was over. It would also help if more time were allotted to analyze at least five samples at each reactor condition. If there were less error, there could be more analysis on reaction rate and order.

There are also some recommendations to be made to the customer. Cumene should be removed from this isooctane stream using another method. Implementing a full-scale packed-bed catalytic reactor would require a large investment and would take a while to install. This seems to be impractical, because the maximum conversion that can be achieved is about 40%. If a lower concentration of cumene is specified, it may be easier to purify the isooctane by means of separation. It would also be recommended to investigate the separation capability of a flash vessel or distillation column to remove the cumene. If these recommendations are implemented, there will be more conclusive data and the customer could be sure whether or not to implement this reactor into their isooctane stream.

Sincerely,

Michael Siddoway

Figure 1 - Plot of conversion versus time obtained when investigating line-out conditions. At large times, the data shoould be reaching a steady state, but it is very unclear what is happening. This plot illustrates the difficulty of the uncertainty of the measurements.

Table 1 – Table of GC peak areas and peak ratios for the calibration samples. Peak ratio for benzene and cumene should be consistent when a sample of equal concentrations in injected into the instrument several times. This illustrates the deviance of the instruments.

Cumene Peak Area / Benzene Peak Area / Peak Ratio / Ratio Average / Ratio Deviation
447538 / 619426 / 0.722504 / 1.0 / .3
370375 / 349760 / 1.05894
527309 / 501891 / 1.050644
1185373 / 882940 / 1.34253

Table 2 - Table of GC peak areas and conversion for running the reactor at constant condiditons. This was shown for the two different flow rates. The peak areas and conversions should, be the same, but they are not. This also illustrates the deviation of the instrument.

Bath Temperature (°C) / Reactor Temp (°C) / Helium Flow Rate (SLPM) / Benzene Peak Area / Conversion
50 / 400 / 2 / 93644 / 0.5
50 / 400 / 2 / 5410 / 0.229053
50 / 400 / 4 / 2817 / 0.36171
50 / 400 / 4 / 154453 / 0.631754

[1] Fogler, Scott. Elements of Chemical Reaction Engineering. 3rd Ed., Prentice-Hall, New Jersey. 2002 (p.225)

[2] Fogler, Scott. Elements of Chemical Reaction Engineering. 3rd Ed., Prentice-Hall, New Jersey. 2002 (p.661)