Energy & Fuels 1994,8, 117-123
117
Development of Standard Direct Coal Liquefaction Activity Tests for Fine-Particle Size, Iron-Based Catalysts Frances V. Stohl’ and Kathleen V. Diegertt Process Research Department 6212 and Statistics and Human Factors Department 323, Sandia National Laboratories, P.O. Box 5800, Albuquerque, New Mexico 87185 Received July 12, 1993. Revised Manuscript Received November 2, 199P
The use of low amounts (10.5 wt %) of unsupported, fine-particle size (C40 nm), iron-based catalysts in direct coal liquefaction has the potential to enhance desired reactions and minimize retrogressive reactions because of the very high surface areas and good dispersions attainable with these catalysts. The objectives of this project are to establish standard activity test procedures for these types of catalysts and to use these procedures a t Sandia National Laboratories for evaluating and comparing the activities of the many catalyst formulations being developed ,in the Advanced Research Coal Liquefaction Program that is funded by the United States Department of Energy’s Pittsburgh Energy Technology Center (PETC). The standard testing procedure that has been developed uses a factorial experimental design with three variables (temperature, time, and catalyst loading), Blind Canyon coal, and phenanthrene as the reaction solvent. Reactions are performed in batch microautoclaves and results are reported as tetrahydrofuran conversions,heptane conversions, amounts of hydrogen donor in the recovered reaction solvent, and product gas yields. The strategy behind the development of this test and the results of using this test with a commercially available pyrite are presented. The results show that the experimental procedures can be implemented consistently by different operators, and that use of phenanthrene as the reaction solvent allows significant eatalytic effects to be observed.
Introduction There are several potential advantages of using cheap, unsupported, fine-particle size (C40nm) catalysts in direct coal liquefaction; among these are improved coal/catalyst contact due to good dispersion’ of the catalyst, and the potential for using low quantities of catalyst (10.5% based on the weight of coal) because of their very high surface areas. These catalysts could be combined with the coal as either active catalysts or catalyst precursors that would be activated in situ. Research efforts that have been performed to develop fine-particle size, unsupported catalysts for direct coal liquefaction2J indicate that the use of these catalysts could result in significant process improvements, such as enhanced yields of desired products, less usage of supported catalyst, and possibly lower reaction severities. These improvements would result in decreased costs for coal liquefaction products. The Advanced Research (AR) Coal Liquefaction Program, which is managed by the United States Department of Energy’sPittsburgh Energy TechnologyCenter (PETC), is funding numerous research efforts aimed at developing these types of catalysts for direct liquefaction. PETC’s program covers a wide variety of iron-based catalysts, such as iron oxides, iron oxyhydroxides, sulfated iron oxides, sulfated iron oxyhydroxides, and iron sulfides. These iron materials are not truly catalysts because they change Statistics and Human Factors Department 323. e Abstract published in Advance ACS Abstracts, December 1, 1993. (1) Huffman, G. P.; Ganguly, B.; Zhao, J.; Rao,K. R. P. M.; Shah,N.; Feng, 2.;Huggins, F. E.;Taghiei, M. M.; Lu, F.; Wender, I.; Pradhan, V. R.; Tierney, J. W.; Seehra, M. S.; Ibrahim, M. M.; Shabtai, J.; Eyring,E. M. Energy Fuels 1993, 7,285-296. (2) Pradhan, V. R.; Tierney, J. W.; Wender, I. Energy Fuels 1991,5, 497-507. (3) See papers in this issue. t
composition during reaction and are recovered as pyrr h ~ t i t e .However, ~ because it is common usage, we will refer to these materials as catalysts. Although most catalyst developers have the capability of testing the performances of the catalysts they develop, it is difficult if not impossible to compare results among researchers because of the different testing procedures used. Some of the differences include reactors, reaction temperatures, reaction times, pressures, hydrogen donor solvents,solventto-coal ratios, and workup procedures. Therefore, to guide the research and development efforts for these fine-particle size, unsupported catalysts and to identify the best catalysts for further evaluation in larger-scale, continuous reactors, it is necessary to evaluate each catalyst’s performance under standard test conditions so that the effects of catalyst formulations from different laboratories can be compared. The objectives of this project are to develop standard coal liquefaction test procedures and to use these procedures at Sandia National Laboratories to evaluate and compare the novel fine-particle size liquefaction catalysts being developed in the PETC AR Coal Liquefaction Program. This paper describes the strategy used in the development of this test and the results of evaluating this test with a commercially available pyrite.
Experimental Section Materials. The coal being used in this project is the DECS17 Blind Canyon Coal obtained from the Penn State Coal Sample Bank. The coal, which is packaged under an inert atmosphere in sealed foil bags with a plastic liner, is stored in a refrigerator prior to use. Each bag contains 20 g. The coal is a HVA bituminouscoalwith0.36%iron,0.02%pyriticsulfur,and7.34% (4) Lambert,Jr., J. M.; Simkovich, G.;Walker,Jr., P. L. Fuel 1980,59, 687-690.
0887-0624/94/2508-0117$04.50/00 1994 American Chemical Society
118 Energy
Stohl and Diegert
&Fuels, Vol. 8, No, 1, 1994
mineral matter (on a dry basis). The particle size is -60 mesh. 1,2,3,6,7,8-hexahydropyrene ( H a y ) ,9,lO-dihydrophenanthrene (DHP), and phenanthrene were evaluated as reaction solvents (98% purity) and phenanthrene for use in the standardtest. H y& ' (98% purity) were obtained from Aldrich Chemical Co. Inc., and DHP was obtained from either Aldrich (94% purity) or Janssen Chimica (97% purity). Pyrite (99.9% pure on a metals basis) with a -100-mesh particle size was obtained from JohnsonMatthey. Histologicalgrade, stabilized tetrahydrofuran (THF) and reagent grade heptane from Fischer Scientific are used in product workups. Microautoclave Reactors. The testing is performed using batch microautoclaves made of type 316 stainless steel components including a 0.75in. 0.d. Swagelok union tee with the branch connection attached with a reducer to 0.375in. 0.d. high-pressure tubing and the two run connections cappedwith plugs. A Whitey plug valve at the top of the tubing is used to insert a thermocouple into the reactants in the tee, to attach a pressure transducer, and to pressurize and depressurize the reactors. The total volume of a reactor is 43 cm* with a liquid capacity of up to 8 cms. Four reactions can be run simultaneously. Reaction Procedures. The coal is riffled four times with the splits being recombined after the first three times and the splits taken after the fourth. Each split contains a little more than needed for two reactions. The splits that are not immediately used are put in glass scintillation vials with an argon cover and stored in a refrigerator. The union tees with the plugs in place are loaded with 1.67 g of coal and 3.34 g of reaction solvent. If the reaction is catalytic, the catalyst loading will be either 0.5wt % or 1.0wt % on an as-receivedcoal basis. The reactors are then assembled and charged to 800 psig of Hz (cold charge). Each reactor is suspended over a fluidized-sand bath at a height so that the reactor temperature is maintained between 35 and 50 "C.The reactor is shaken at 200 cycles/sfor 10min with a Burrell wrist-action shaker. While still being shaken, the reactors are immersed into the sand bath and rapidly heated to reaction temperature. Temperatures, pressures, and times are recorded with a digital data acquisition system every 30s during the course of the experiments. Following the heating period, the reactors are rapidly quenched to ambient temperature in a water bath and a gas sample is collected. The reaction data is analyzed to determine the actual reaction time at temperature and the averages and standard deviations for reaction temperature and pressure. Heat-up times and quench times are also determined. Product Workup Procedures. Upon completion of the reactions, the union tee is separated from the tubing, and the run connection plugs are removed. The tee and plugs (with liquid and solid products) are placed in a beaker with 250 mL of THF alongwith THF washingsfrom rinsing out the tubing. The beaker is sonicated for 30 min, left to soak overnight, and sonicated again for 30 min the next morning. The reaction products are then rinsed out of the reactor tee and plugs with THF. Solvent solubilities are measured using a Millipore 142 mm diameter pressure filtration device with air pressurization and Duropore (0.45rm) filter paper. The total time the liquid and solid products are in THF prior to filtration is about 17 h. The filter cake is rinsed twice with THF prior to dismantling the pressure filtration device. After the filtration is complete, the filter paper is dried under vacuum at 70 O C , cooled to room temperature, and weighed to determine the insoluble portion. The THF solublesare rotoevaporated under vacuum to reduce the volume to about 50-60 mL, quantitatively transferred to a 100-mL volumetric flask, cooled to room temperature, and brought to 1WmLvolumeusing THF. A 1-mLaliquot is removed and used for gas chromatographic (GC) determination of the recovery and composition of the reaction solvent. These quantitative GC analyses use methylnaphthalene as an internal standard and phenanthrene and DHP as external standards. The remaining 99 mL of THF with THF solubles is rotoevaporated under vacuum for 40 min after THF stops dripping into the receiver flask. After 200 mL of heptane is added to the flask containing the THF solubles, the flask is sonicated for 10 min. This heptane/product mixture is then pressure filtered. The
CATALYST
Figure 1. Factorial experimental design. Temperature in O C , time in minutes, catalyst loading in w t % on an as-received coal basis; letters are order in which reactions performed.
filter cake is washed twice with heptane prior to dismantling the pressure filtration device. The filter paper is dried under vacuum at 70 O C , cooled to room temperature, and weighed to obtain the weight of heptane insolubles. Conversions are calculated on a dmmf coal basis. The quantity of gases (excluding HzS) produced in a reaction is calculated using the postreaction vessel temperature and pressure with the ideal gas law and the mole percents of the gases in the gas sample as determined using a Carle GC and a standard gas mixture. Factorial Experimental Design and Analysis. The factorial experimental design (Figure 1) that was chosen for this project evaluates the effects of three variables at two levels: time (20 and 60 min), temperature (350and 400 "C),and catalyst loading (0 and 1wt % based on as-received coal). With this full factorial experimental design, the experimental results are evaluated for all combinations of levels of the three variables so that 23 evaluations are required. Reactions at the center point in this cubic design are also performed. In this project, an analysisof variance (ANOVA)was performed to estimate the effects of the experimental variables, and to statistically test their significance. Using the Statistical Analysis System (SAS version 5.18 from the SAS Institute, Cary, NC), full factorial models with all two-way and three-way interaction terms were originally fit to the measured experimental results. Tests of statistical significance were performed for the effects of all variables; those variables with significance levels greater than 0.05 were dropped from the model. However, if an interaction term was significant, the main effects of the corresponding variables were always included in the model. The totalvariability in a measurement was partitioned into amounts attributable to each of the variables in the experiment and their interactions and to measurement error. Replication of the experiments was used to estimate measurement error and to reduce its impact on the estimated effects of the variables. Statistical tests were performed to determine if the amount of variability attributed to an experimental variable was too large (relative to the measurement error) to be due to chance. If so, the effect of the variable at different levels was estimated. Models were constructed using the estimates of the effects of the variables to calculate the expected experimental results for specified sets of reaction conditions.6 The experimental variables used in the ANOVA were the measured average reaction temperature, measured reaction time, and the actual weight of catalyst used. The operator effect was treated as a random effect, and its contribution to variability was estimated.
Results and Discussion
Two important aspects of this catalyst testing project are t h e development of standard experimental procedures ( 5 ) John, P. W. M. Statistical Design and Analyses of Experiments; MacMillan Co.: New York, 1971.
Coal Liquefaction Activity Tests and the development of a factorial experimental design that would allow evaluation of the catalyst over a range of process variables. Development of the StandardTest Procedures. The details of the standard experimental procedures that have been developed in this project are described in the Experimental Section. The Blind Canyon coal was chosen by Malvina Farcasiu (PETC) and Alan Davis (the Pennsylvania State University) because it has low inherent catalytic activity due to its low iron and low pyritic sulfur contents. A large coal sample was specifically collected and prepared by the Pennsylvania State University for use in PETC's Fine-Particle Catalyst Development Program. The reaction solvent for the standard test should yield low enough thermal conversions so that catalytic effects can be detected when using a high solvent-to-coal ratio that would help ensure good mixing of the catalyst and coal. Since current direct coal liquefaction processing configurations use approximately a 2:l solvent-to-coal ratio, this appeared to be a reasonable ratio for the standard test. DHP was evaluated as a possible reaction solvent because it is a very good hydrogen donor with the potential for donating up to 1.1 wt 5% hydrogen, which is slightly higher than the 1 wt % hydrogen contained in some of the better recycle solvents from large-scale, coal liquefaction processes.6 Experiments performed with a 2:1 DHP:asreceived coal ratio a t 400 "C for 60 min and 350 "C for 20 min gave THF conversions of about 94% and 43 % , respectively. Several similar experiments with pyrite addition showed increases of about 3 %I in THF conversion as compared to the thermal reactions. Experiments were also performed to evaluate the use of H6Py as a reaction solvent. H6Py is an excellent hydrogen donor solvent with the potential for donating up to 3 wt % hydrogen. Results from thermal liquefaction experiments performed at 400 "C for 30 min and at 350 "C for 30 min with a very low H6Py:coal ratio of 0.41 yielded 90% and 44% THF conversions, respectively. Increasing the solvent-to-coal ratio to help give good mixing of the catalyst and coal would increase these conversions, thus making separation of thermal and catalytic effects difficult. Therefore, H6Py would not be a good reaction solvent for this test. Thus, DHP with a 2:l reaction solventxoal ratio was chosen for the first evaluation of the factorial experimental design. Because of DHP's low melting point (32-35 "C), the 10min shaking of the reactor was incorporated into the test procedure prior to the start of the reaction to help ensure good catalyst-coal mixing. Pyrite was chosen as the standard catalyst because it is a known catalyst in direct coal liquefaction,7i8is commercially available, and is easy to use since it does not require complicated pretreatments or special handling procedures. Thus, pyrite will provide a good baseline comparison for the novel fine-particle size, iron-based catalysts that will be tested in this program. The surface area of the pyrite was 0.7 m2/g as determined using BET techniques. An X-ray diffraction pattern showed no evidence of any phases other than pyrite. Definition of Conditions for the Factorial Experimental Design. A factorial experimental design was used (6) Whiphuret,D. D.;Mitchell,T. O.;Farcaaiu,M. Coal Liquefaction: The Chemistry and Technology of Thermal Processes; Academic Press: New York, 1980; p 295. (7) Stephens,H. P.;Stohl,F. V.; Padrick,T. R o c . Znt. Conf.Coal Sci., Duseldorf, Germany, Sept. 7-9,1981 1981,368-373. (8)Stohl, F. V. Fuel 1983,62, 122-126.
Energy Q Fuels, Vol. 8,No. 1, 1994 119
in this project because it gives an efficient means, using ANOVA, for estimating the individual and joint effects of a number of variables on one or more experimentalresults. In an ANOVA, the effect of each variable is estimated using all data collected. Thus, factorial designs that are properly analyzed are more efficient in the use of data than one variable at a time studies, where the effect of a particular variable can be computed only from runs in which that variable is changed and all other variables are fixed. Thus, ANOVA of factorial experimental designs permits reliable comparisons of the performances of the novel catalysts. The reason for choosing a factorial experimental design (Figure 1)with three variables at two levels (times of 20 and 60 min, temperatures of 350 and 400 "C, and catalyst loadings of 0 and 1 w t % based on as-received coal) and a center point (375 "C, 40 min, 0.5 wt % catalyst) was to obtain process information for the variables that were considered most important while keeping the number of experiments to a reasonable number. The higher severity reaction conditions are consistent with process conditions used in coal liquefaction, and the lower severity conditions will give information on catalyst activity during heat up of the coal and solvent to reaction conditions. The maximum amount of a disposable fiie-particle size catalyst that will be used in a commercial process will probably be less than 0.5 wt % . However, since the purposes of this testing project are not only to compare different catalysts, but also to guide research efforts, it was felt that a larger range of catalyst loading should be used. Time and temperature were chosen as variables because they were believed to have the biggest impacts on coal conversion. Evaluation of the Factorial Experimental Design with DHP. The first evaluation of the factorial experimental design was performed using the experimental procedures developed in this project, pyrite as the catalyst, and a 2:l DHP:as-received coal ratio. Two operators independently performed the reactions associated with the experimental design. Replicates were performed by one or both operators at the center point, at 400 "C for 60 min both with and without catalyst, and at 350 "C for 20 min with catalyst. Four experimental results were measured and statistically analyzed THF conversion,heptane conversion, product gas yields, and the amount of DHP in the recovered reaction solvent. The order in which the reactions were performed by each operator is shown by the letters in Figure 1. Table 1 shows the measured experimental results obtained by each operator for the nine sets of experimental conditions. Reactor heat up and quench times are about 3 min each. Standard deviations for reaction temperatures measured over the course of each reaction are routinely
