834
Energy & Fuels 1988,2, 834-842
Correlations between Physical and Chemical Properties of Pyrolysis Liquids Derived from Coal, Oil Shale, and Tar Sand M. Rashid Khan Morgantown Energy Technology Center, U S . Department of Energy, Morgantown, West Virginia 26507-0880 Received February 2, 1988. Revised Manuscript Received August 15, 1988
Characterization techniques based upon the liquid’s refractive index are used with petroleum distillates to predict fuel-related properties; however, essentially nothing has been reported on the application of this technique to pyrolysis liquids. Measurements of the refractive indices of the pyrolysis liquids derived from various feedstocks (coal, oil shale, and tar sand) were made and appear to correlate well with the liquids’ physical and chemical properties. The refractive indices of the pyrolysis liquids show good correlations with liquid density (correlation coefficient of 0.98), carbon and proton aromaticities (correlation coefficients of 0.88 and 0.91, respectively), and liquid carbon residue (i.e., correlation coefficient of 0.88 with the Conradson carbon residue). The above and other correlations were developed using data from a t least 7 to as many as 32 discrete samples. These correlations have been used to develop empirical models. These findings demonstrate for the first time the potential of using the liquid’s refractive index as a rapid technique to characterize the fuel-related properties of fossil fuel liquids generated by pyrolysis (before they are hydrogenated).
Introduction and Objectives The Morgantown Energy Technology Center (METC) results demonstrated that relatively high-quality liquid fuels (low sulfur, high H/C) can be produced by low-temperature devolatilization of coal.’t2 Physical and chemical properties characterization of these liquids has been the focus of several studies during the last few y e a r ~ . l - ~ Correlations based on the data obtained from relatively simple characterization techniques for predicting fuel-related physical and chemical properties of pyrolysis liquids would facilitate the utilization of these liquids in various processing schemes. Properties such as molecular weight and aromaticity of liquids influence their utilization behavior/properties (e.g., viscosity, smoke point). Measurements of molecular weight or aromaticity can be difficult, time consuming, and expensive; can require skilled operators; and are often beyond the resources of most small laboratories. The correlations between properties of petroleum liquids and their refractive indices are available in the literature.Sturm et a1.8afractionated the pyrolysis liquid generated (1) Khan, M. R. Fuel Sci. Technol. Znt. 1987,5, 105-231. (2) Khan, M. R. Proceedings-Znternationa~ Conference on Coal Science; Elsevier: Amsterdam, 1987; pp 647-651. (3) Khan, M. R. “Characterization and Mechanism8 of Mild Gasification Processes: Low-Temperature Devolatilization Studies.”Proceedings of the Fifth Annual Gasification Contractors Meeting; DOE/ METC-85/6024; NTIS/DE85008618; NTIS: Springfield, VA, 1985. (4) Khan, M. R.; Kurata, T. M. The Feasibility of Mild Gasification of Coal: Research Needs; DOE/METC-85/4019; NTIS/DE85013625; NTIS Springfield, VA, 1985. ( 5 ) Khan, M. R. Production of a High-Quality Liquid Fuel From Coal by Milk Pyrolysis of Coal-Lime Mixtures; DOE/METC-86/4060 (DE86006603);NTIS: Springfield, VA, 1986. (6) (a) White, C.; Perry, M. B.; Schmidt, C. E.; Douglas, L. J. Energy Fuels 1987, 1, 99-105. (b) Majumdar, B. K. Energy Fuels 1988, 2, 230-233. (7) Riazi, M. R.; Daubert, T. E. Znd. Eng. Chem. Process Des. Deu. 1980, 19, 289-294. (8) (a) Sturm, G. P., Jr.; Woodward, P. W.; Vogh, S. A.; Holmes, S. A.; Dooley, J. E. BERC/RI-75/12, November 1975. (b) Holmes, S., personal communication, 1988. (c) Katzer, J.; Gates, B. “Catalytic Processing in Fossil Fuels Conversion.” in AZChE Tody Series; American Institute of Chemical Engineering: AIChE New York, 1975; I-B-35-I-B-42.
in a rapid-rate reactor (Coal-Oil-Energy Development, COED) into various distillate cuts. The COED liquids were severely hydrogenatedsb to remove the heteroatom contents (e.g., sulfur, nitrogen) and to improve their fuel value. Following the leading work of Riazi and Daubert’ on petroleum distillates, White et al.@compared the refractive indices of coal liquefaction distillate cuts with their chemical properties (e.g., molecular weight) and developed three sets of exponential equations.sb These authors showed that the rate of change of a given property (e.g., density, molecular weight) in various distillate cuts when compared with the refractive indices of the corresponding cuts are essentially uniform for a given liquefaction distillate. MajumdarGbreevaluated the data presented by White et a1.6a and concluded that linear models (opposed to nonlinear models used by White et al.&) gave much simpler and statistically more reliable and accurate equations. White et a1.6aincorrectly described the samples used by Sturm et al. as “pyrolysis liquids”. Although the liquids were generated by the Coal-Oil-Energy Development (COED),a pyrolysis-basedprocess, the COED liquids were extensively hydrogenated before they were studied by Sturm et al. Catalytic processing of these COED liquids was also investigated by Katzer and Gates.& Furthermore, the entire study of White et al. was based on the hydroliquefaction products of only two coals. In previous studies, it was demonstrated by KhanQthat the pyrolysis liquids generated from weathered fossil fuel samples (including coal and oil shale) had higher liquid refractive indices than those obtained from unweathered samples, suggesting that heteroatoms present in various pyrolysis liquids would significantly influence the measured refractive indices. The objective of this paper is to illustrate the relationship between the refractive indices of pyrolysis liquids and the fuel-related physical (density, Conradson carbon res(9) Khan, M. R. Energy Fuels, 1987, 1, 366-376.
This article not subject to US. Copyright Published 1988 by the American Chemical Society
Properties of Pyrolysis Liquids
idue) and chemical properties (aromaticity, molecular weight, hydrocarbon type, etc.) of liquids derived from a range of feedstocks (e.g., coal, oil shale, and tar sand). This work is, therefore, considerably different from that by White et al. who measured the "rate of change" of properties of various distillate cuts for a given liquefaction distillate. Experimental Section A fixed-bed reactor was used to generate pyrolysis liquids at 500 "C. More details on this reactor system as well as the experimental procedures used and the reproducibility of data are available.'~2~6 A range of feedstocks (primarily coal but also oil shale and tar sand) were devolatilized in this reactor. Coal samples were supplied by the Penn State/DOE coal data bank.l0 The origin of shale samples (eastern and western) and the Pittsburgh No. 8 coal has been discussed elsewhere? Tar sand samples were procured from the Western Research Institute (WRI). Sample preservation and avoidance of air oxidation of samples were key considerations in this investigation as reported in previous s t ~ d i e s . ~ J ~ Availability -'~ of fresh (well preserved, not weathered) samples was the criterion used for sample selection. Some coal samples utilized by Given et al.15 for an investigation on direct liquefaction were selected in this study for comparison. Primarily bituminous coals were used in this study as these are known to yield the highest liquid product during pyroly~is.'~~ All sample preparation and handling were performed in an inert atmosphere. The refractive index of liquid samples were measured at 20 "C by Huffmann Laboratories, Inc., applying ASTM D-1218 methods. The refractive index of 1-bromonaphthalene was measured as a standard before the refractive index of each pyrolysis liquid. The refractive index measurement of 1-bromonaphthalene was within *0.0005 units of the reference value (reported by Hoffmann). Repeatability of the refractive index measurements was the same as that reported by White et al.,& as the same laboratory procedures were used for measurements in both studies. Conradson carbon residue was determined at Saybolt Lab (Pasadena, TX) by using ASTM D-189. The density measurements (reproducible within *O.OOOl) were also performed at Saybolt at 15.5 "C by ASTM D-70. The aromaticity of the liquids was measured at the University of North Dakota Energy and Minerals Research Center (UNDEMRC) by proton NMR, using the procedure by Clutter et al.'& with a Varian XL200 NMR spectrometer. 'H NMR was used for determination of aromaticities. Molecular weights were calculated on the basis of the NMR data by applying the equation used by Clutter et al.16a The aromaticity of selected pyrolysis liquids generated by entirely separate pyrolysis experiments was also characterized a t Virginia Polytechnical Institute and the Morgantown Energy Technology Center. A comparison between the data generated at these labs showed excellent agreement (well within 0.03, the estimated accuracy of measured aromaticity, as determined by NMR).lGbThe liquid chromatography (LC) separation of the mild pyrolysis liquids was performed by using sequential elution solvent chromatography (SESC). The procedure is described by Farcasiu, Seshadri, and Cr~nauer."~'~In mild pyrolysis, the liquids produced are light, and nearly 85 wt% of all tars was recovered in the first six fractions. Thus, after the column was eluted with methanol, the remaining tar was ducarded (10)Penn State/DOE Coal Data Bank, Office of Coal Research, The Pennsylvania State University, University Park, PA 16802. (11)Khan, M. R. Preprint, 193rd National Meeting of the American Chemical Society,Petroleum Chemistry Division, Denver, CO, April 1987. (12)Khan, M. R.19th Oil Shale Symposium Proceedings; Colorado School of Mines: Golden, CO, 1986;pp 139-148. (13)Khan, M. R.;Jenkins, R. G. Fuel 1985,64,189-192. (14)Khan, M. R.; Jenkins, R. 0. Fuel 1985,64, 1618-1622. (15)Given, P. H.;et al. Fuel 1982,61,971. (16)(a) Clutter, D. R.; et al. Anal. Chem. 1972, 44,1395. (b) Retcofsky, H. L.; Schweighardt, F. H.; Hough, M. Anal. Chem. 1977,49,
---
5R5-5RR ---.
(17)Farcasiu, M. Fuel 1977,56,9. (18)Seshadri, K.S.;Cronauer, D. C. Fuel 1983,62,1436.
Energy & Fuels, Vol. 2, No. 6, 1988 835 along with the column packing. The first fraction containing saturates, alkenes, and aromatics was studied with high-performance liquid chromatography (HPLC). The HPLC chromatographic instrumentation consisted of a Perkin-Elmer Series 4 quarternary solvent delivery system, an IS-100 autosampler, an LC-85 UV detector, and a Model 3600 data station. Analytical chromatography was performed on a dual-column system (25 cm X 4.6 mm) with 5 pm silica gel bonded with cyano group (Supelco LC-CN) in one column and 5 pm plain silica (Supelco LC-Si) in the other. SESC fraction 1was dissolved in hexane (2 mg/mL), and 20 pL of the solution was injected onto the column. Then the column was eluted with hexane at a flow rate of 1 mL/min and eluant from the column was monitored at 254 nm, which provided fairly good resolution for aromatics. After the chromatogram of aromatics in the sample was recorded, the column was cleaned with methylene chloride and equilibrated with hexane prior to the analysis of the next sample. In addition, selected samples were characterized by field ionization mass spectroscopy (FIMS) at the Stanford Research Institute (SRI).lg In order to identify individual components, the SESC fraction-1 (saturate/olefin) portion was analyzed and the data were reduced. In separate experiments for GC/MS measurements, the neutral fractions of pyrolysis liquids were generated chromatographically on activated alumina. A 20-in. X 3/8-in. column was loaded with 25 g of activated alumina (Biorad). A sample of 0.2 g was charged to the column and the neutral fraction was eluted with 50 mL of benzene. The benzene was removed from the sample by rotary evaporation. To ensure that the lower boiling point components of the sample were not lost during benzene removal, the benzene was not completely removed from the sample. The concentration of the residual benzene was then determined by combined gas chromatography/mass spectrometry (GC/MS) and the final weights of the neutral fractions were adjusted to account for the residual benzene. The polar material was not eluted from the column, but was determined by difference. The neutral fractions were analyzed by using WRI's GC/MS hydrocarbon group type analysis. A Hewlett-Packard Model 5985B GC/MS system was used. The GC column was a 50-m quartz capillary, coated with methylsilicone and operated in programmed-temperature modes, which optimized the resolution of components. The method determines the composition of the hydrocarbon group by using selected ions that are representative of each hydrocarbon group type and relies on the gas chromatographic separation to minimize interferences from fragment ions of other hydrocarbon groups. The results from the analyses are reported as weight percentages of the neutral fractions. The results are provided as weight percentages of the totalsample that was charged to the column. The results for the neutral fractions reflect the adjusted weights of the neutral fractions after accounting for the residual benzene. Detailed analysis and distribution of the components identified in various fractions will be presented in a future communication. The distillate cut between 300 and 600 OF is of interest because its properties are expected based on literature studies to match the target properties of the high-density fuels.20p21 Liquids were combusted in a thermogravimetric analysis system (TGA) at 100 OC/min in air (flowrate of 120 cm3/min). In addition TGA distillation of the liquids was performed (heating rate, 5 "C/min, He flowrate 50 cm3/min). TGA is a readily available and convenient technique for rapid determination of the evaporation profiles of liquids. The evaporation profiles obtained by TGA are dependent on the experimental conditions used, as discussed by Mondragon and O ~ c h i .ASTM-D86 ~ was also applied to determine the distillation profiles for selected liquids. ASTM-D86 was performed at Saybolt (19)John, G. A.; Buttrill, S. E., Jr.; Anbar, M. "Field Ionization and Field Desorption Mass Spectroscopy Applied to Coal Research." In Organic Chemistry of Coal; Larsen, J., Ed.; ACS Symposium Series 71; American Chemical Society: Washington, DC, 1978; p. 223. (20)Korosi, A.; Rubin, J. N. "Hydroprocessing of Light Pyrolysis Fuel Oil for Kerosene Type Jet Fuel"; Technical Report AFWAL TR-83-2048; Wright-Patteron Air Force Base: Dayton, OH, 1980. (21)High-Density Candidate Hydrocarbon Fuels from Refinery ByProduct Streams; Technical Report (Interim), Contract No. F33615-83C-2360/585842-83-C-0383, Project No. 745,Task 001, 002; Universal Energy Systems: Dayton, OH, 1985. (22) Mondragon, F.; Ouchi, K. Fuel 1984,63,61.
Khan
836 Energy & Fuels, Vol. 2, No. 6,1988 Table I. Origin and Apparent Rank of the Samples PSOC geological location" apparent no. (Dr0vince:reeion:seam name:citv. rank - . state) hvAb 123 East:Appal:Lower Kittanning:Bickmore, WV Sub A 181 InkEaskUDDer BlockPatricksbureh. IN hvAb 267 East:Appai:klintwoodAirport, V i ' hvAb 275 East:AppakOhio No. GACadiz, OH hvAb 296 East:Appal:Ohio No. 5:North Lima, OH hvAb 306 East:AppakOhio No. 12:Yorkville, OH hvAb 355 East:AppakLower Kittanning:Darlington, PA hvAb 375 EaskAppakHazard No. QHarlan, KT hvBb 1109 Rock Mt:SW Utah:King CannekMt Caramel Jct, UT mvb 1313 East:Appal:Lower Kittanning:Mt Zion, PA hvBb 1323 1nterior:East:Ill No. 6:Danville, IL hvAb 1448 North Great P1ns:York Canyon, NM hvAb 1469 East: AppakMary Lee:Goodsprings, AL hvAb 1470 East: AppakPratkOakman, AL hvAb 1471 East:AppakPee Wee:Oliver Springs, TN hvAb 1472 EaskAppakLower Banner:Dante, VA hvAb 1473 East: AppakUpper Banner:Dante, VA hvAb 1475 East:AppakElkhorn No. 3Wayland, KY hvAb 1481 East:Appal:Upper C1arion:Blairs Corners, PA hvCb 1492 InkEaskI11 No. 6:Cutler, IL hvCb 1502 Rock Mt:Uinta:Hiawatha:Emery, UT hvAb 1504 Rock MkUinta:Upper Sunnyside:Sunnyside, UT hvAb 1517 East:Appal:Ohio No. 5: Petersburg, OH hvAb 1523 East:Appal:WV No. 5 Block:Ward, WV 1524 East:Appal:Upper Kittanning:Philippi, WV hvAb apparent sample rank Pittsburgh No. 8:Arkwright Seam:WV hvAb hvCb Illinois No. 6:Burning Star Mine, IL North Dakota lignite lignite Colorado oil shale (Mahogany Zone, Green River) Sunbury oil shale (Kentucky Shale, obtained from KCERLb) Asphalt Ridge Basin tar sand (Utah tar sand, obtained from WRI') resinite sample separated from Utah coal (procured from the University of Utah)
-
"Location: East = Eastern; Int = Interior; Appal = Appalachian; Rock Mt = Rocky Mountain; North Great Plns = Northern Great Plains. KCERL = Kentucky Center for Energy Research Laboratory. WRI = Western Research Institute. Laboratory. The Statistical Analyses System (SAS)package developed by the SAS Institute was used for data analyses. The regression programs available in this package were applied to develop empirical model^.^^^^^
Results Table I summarizes the origin and rank of the samples used in this study. A detailed study on the influence of feedstock type on pyrolysis liquid product yield and composition will be presented separately. As one would expect, the physical and chemical properties of the liquids derived from various feedstocks vary widely. A number of characterization techniques were applied to investigate quantitatively the compositional differences between various feedstock liquids. Some typical examples of these variations will be briefly discussed in this presentation. A comprehensive treatment on the characterization data will be presented separately. Tables I1 and I11 show some of the differences in these liquids. Table I1 shows that the first two SESC fractions constitute the major portion (from 40 to over 90 wt%) of (23) Box,G. E. P. Statistics for Experiments; Wdey: New York, 1978. (24) S A S User's Guide: Basics, Version 5 Edition; SAS Institute:
Cary, NC, 1985. ( 2 5 ) Guffey, F., personal communications, 1987.
Table I1
(A) Sequential Elution Solvent Chromatographic (SESC) Fractions of Mild Pyrolysis Liquids" wt % of liquid fraction Pittsburgh resin Colorado tar sand Sunbury 1 37.0 82.6 74.6 71.1 48.5 2 2.5 8.6 3.2 3.8 5.0 3 20.7 6.4 7.9 6.4 17.8 4 9.7 0.4 0.5 2.2 3.9 5 11.2 2.5 3.9 3.4 6.3 6 2.5 1.7 3.2 0.7 2.2 residue 16.4 1.3 13.0 17.5 1.4 (B) Class Types of Predominant Components of Fractions Separated in SESC fraction eluant major components 1 hexane alkanes/alkenes, neutral aromatics nonbasic N-, 0-, 2 hexane/l5% toluene S-heterocyclics, neutral aromatics monophenols 3 chloroform polyphenols, carbonyls, 4 chloroform/4% diethyl ether 5 diethyl ether/3% ethanol polyphenols, basic nitrogen heterocyclics 6 methanol highly functional (polar) molecules (large heteroatom contents) a Key: Pittsburgh, Pittsburgh No. 8 coat tar;resin, concentrated resin tar separated from Utah coal, Colorado, Colorado shale oil; tar sand, Asphalt Ridge tar sand liquid; Sunbury, Kentucky Sunbury shale oil.
Table 111. Separation Results for SESC Fraction 1 of Mild Pyrolysis Liquids wt % in SESC fraction 1 saturates and % tar samole olefins aromatics recoverv Pittsburgh No. 8 coal 15.1 81.5 96.6 concentrated resin 9.5 87.6 97.1 Colorado oil shale 50.7 45.2 95.9 Kentucky Sunbury oil shale 16.3 78.8 95.1 Asphalt Ridge tar sand liquid 56.7 33.6 97.1
the totalliquids. These fractions contain saturate/alkene and neutral aromatics (e.g., furans, diary1 ether, benzofurans, and benzothiophenes). The HPLC procedure was also used to separate the saturate/olefins from aromatics (Table 11). The results demonstrate that while the tar sand oil contains large amounts of saturates, the liquids from the Pittsburgh No. 8 coal and concentrated resin samples are highly aromatic. The chromatograms showed that Sunbury shale oil contains a higher concentration of multiring aromatic compounds than the Colorado shale oil. Additional discussion on the SESC results is available?J* The cyclic alkanes present in the saturate/alkene fractions of various liquids derived from the Pittsburgh No. 8 coal, Colorado oil shale, and Asphalt Ridge tar sand have been identified based on the Z-series type of analyses of FIMS data. The results demonstrate that while the coal liquids contain relatively more alkanes and dicyclic alkanes, the Asphalt Ridge tar sand oil is richer in tetra- and pentacyclic alkanes (Figure 1). These latter components make suitable feedstocks for the production of high-energy-density aviation fuels. Table V presents a summary of equations used to predict the properties of pyrolysis liquids. The hydrogento-carbon ratio (H/C) of the pyrolysis liquids is an im-
Energy & Fuels, Vol. 2, No. 6, 1988 837
Properties of Pyrolysis Liquids
Table IV. Characterization Data Used for the Predictive Equationso obsb 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32
sample PSOC 123 PSOC 181 PSOC 267 PSOC 275 PSOC 296 PSOC 306 PSOC 355 PSOC 375 PSOC 1109 PSOC 1313 PSOC 1323 PSOC 1448 PSOC 1469 PSOC 1470 PSOC 1471 PSOC 1472 PSOC 1473 PSOC 1475 PSOC 1481 PSOC 1492 PSOC 1502 PSOC 1504 PSOC 1517 PSOC 1523 PSOC 1524 Pitt No. 8 I11 No. 6 ND lignite Col Shale Sun Shale ASP tar sand conc. resin
~~
n 1.5518 1.5586 1.5900 1.5939 1.5792 1.5812 1.5686 1.5854 1.5102 1.5827 1.5887 1.5385 1.5654 1.5650 1.5623 1.5620 1.5750 1.5928 1.5897 1.5862 1.5609 1.5362 1.5816 1.5679 1.5768 1.5803 1.5747 1.5584 1.5100 1.5422 1.5087 1.5477
C, 0.46 0.50 0.57 0.57 0.51
H, 0.16 0.19 0.24 0.24 0.19
0.52 0.56 0.28 0.57 0.58 0.43 0.55 0.55 0.49 0.57 0.54 0.53 0.55 0.57 0.48 0.46 0.56 0.51 0.51 0.55 0.59 0.51 0.28 0.38 0.22 0.33
0.20 0.22 0.08 0.25 0.25 0.15 0.22 0.18 0.18 0.22 0.20 0.20 0.22 0.24 0.16 0.15 0.23 0.19 0.18 0.21 0.27 0.22 0.06 0.12 0.05 0.07.
H/C 1.44 1.35 1.26 1.40 1.34 1.29 1.30 1.31 1.78 1.20 1.41 1.47 1.30 1.23 1.31 1.31 1.33 1.43 1.43 1.38 1.48 1.46 1.34 1.33 1.35 1.35 1.42 1.46 1.61 1.44 1.63 1.53
H 10.19 9.36 8.90 9.58 9.38 8.66 8.81 8.66 11.66 8.59 8.59 10.41 9.07 8.22 8.69 9.37 9.52 10.08 8.91 8.60 10.01 10.33 8.88 9.25 9.17 9.10 9.05 9.14 11.34 9.87 11.71 11.15
MW 239.7 209.8 194.2 185.2 203.3 195.6 186.7 401.3 205.6 169.2 245;6 186.9 218.1 192.1 219.3 245.7 194.1 184.0 168.5 219.3 222.2 185.9 207.1 204.6 206.1 171.7 186.0 355.1 283.9 531.9 311.0
T,,OF
234
185 199 290 257
TM,OF
198 172 163 158 177 172 177 163 180 178 180 176 189 178 158 192 185 165 163 164 164 185 183 175 165 157 153 175 187 202 233
d,g/cmS
c,
0.9019
2.16
1.0306 1.0444
4.62 3.94
0.9024 0.9548 0.9059 0.9930
1.97 3.83 0.94 .2.28
Key: n = refractive index; H/C = hydrogen-to-carbon ratio (atomic);.C, = carbon aromaticity; H, = proton aromaticity; MW = molecular weight; T, = temperature of middistillation point (ASTM D-86);TM = temperature of middistillation point (TGA); d = density; C, = Conradson carbon residue. Observation number. TarSand
0011Shale Coal Monocycllc AlkaneJlOlefins
2
0
Pentacyclic Tetracycllc Alkanes Alkanes
-2
-4
-6
-8
-10
L
Figure 1. Comparison of liquid composition derived from coal (Pittsburgh No.8),Colorado oil shale, and Asphalt Ridge tar sand.
Z-series type of analyses were performed by FIMS on the saturate/olefin HPLC fraction.
portant fundamental property of liquid fuels that influences various physical (e.g., viscosity, density) and chemical (e.g., molecular weight, aromaticity, etc.) properties. The linear relationship between the liquid H/C ratios and the refractive indices of the liquid is presented in Figure 2A (correlation coefficient of -0.74). The following linear regression equation is used for predicting the liquid H/C ratios: [H/C] = 6.876 - 3.50n (1) The comparison between the predicted and experimental results is presented in Figure 2B. The refractive indices (n) of the liquids correlated with the hydrogen (H) contents of the liquids (correlation coefficient = -0.83; Figure 2C). Equation 2 was used to predict the hydrogen contents H = 57.264 - 30.5n (2)
(Figure 2D). The R2 for the model is 0.70. R2 measures the proportion of total variations explained by the regression. It is calculated by dividing the s u m of squares due to regression by the total sum of squares. In simple linear regression, R2is related to correlation coefficient, r, as follows: r = (R2)lI2,and r has the same sign as the slope of the computed regression line. Figure 3A shows the relation between the carbon aromaticities (C,) and the refractive indices (n) for the pyrolysis liquids derived from various fossil fuels. The correlation coefficient between the variables was 0.91. The carbon aromaticity ( C A could be predicted by the linear regression equation C, = 3.657n - 5.228 (3) The R2for the model is 0.82. The data were tested for linearity (by Natrella‘s model%) for eq 3. The test showed that the linear model defined by eq 3 is not very appropriate. Thus,a quadratic fit was attempted, which resulted in C,
-
= 102.280n - 31.759n2 81.770
(44
with R2= 0.86 and the mean residual percent deviation equals 5.80. The mean residual percent deviations for various fits are displayed in Table VI. Data presented in Table VI show that the quadratic model has the smallest mean residual deviation and largest R2. The individual residual percent deviations were also calculated for the quadratic fit. Out of 31 points, in only 5 points do the residual percent deviations exceed 10%. Figure 3B plots ~~
(26) Natrella,M.“Experimental Statistics”;NBS H a d b . (U.S.)1963, 91.
838 Energy & Fuels, Vol. 2, No. 6,1988
Khan
2
1.8
-
1.7
-
1.6
-
1.5
-
0
! 9
Y
1.0
A
l4 1.3
-
1.2
-
I . I 1.0
1.5D
1.52
1.54
1.56
1.58
B
1.60
1.0
1.2
1.4
1.6
1.8
PREDICTED VALUE
REFRACTIVE INDEX
13 i
c
8 I,
I
1.508
1.518
1.527
1.537
1.548
1558
1.585
1575
1584
8
I
I
I
I
I
1.594
9 10 11 12 13 D 8 REFRACTIVE INDEX PREDICTED VALUE Figure 2. (A) Correlation between the liquid hydrogen-to-carbonratio and the refractive index (n)of the liquids. (B) Predicted and measured hydrogen-to-carbonratio based on a single-variablemodel. (C) Correlation between the liquid hydrogen content (dry basis) and the refractive index of the liquids (correlationcoefficient-0.83). (D) Predicted and measured hydrogen content based on refractive index of the liquids.
the predicted data with the actual values. The predicted results are based on the quadratic model. The carbon qromaticity of the same data set could be predicted with a two-variable model that includes the liquid H/C atomic ratio: C,, = 2.887n - 0.227(H/C) - 3.71
(4b)
Figure 3C compares the predicted values with the measured carbon aromaticity data. Figure 4A compares the proton aromaticity with the refractive indices of the liquids (correlation coefficient of 0.88). The actual aromaticity data can be predicted (Figure 4B) on the basis of a single-variable model, defined by H,, = 2.102n - 3.103 (5) The predicted equation, based on a two-variable model, can be defined by H,,
= 1.739n - 0.105(H/C)
- 2.39
(6)
The test of linearity by the use of Natrella's was applied to validate that the above linear models were ap, , . Nevertheless, a nonlinear model was propriate for H attempted. The predicted and actual data based on a two-variable model are shown in Figure 4C. Proton aromaticity could also be predicted on the basis of the following nonlinear model. (See also Table V.) H,,
= n-0.56168 (H/ C)4.4"
The molecular weights (MW) of the pyrolysis liquids (determined by using NMR data) could be predicted from
the refractive indices. It is well-known that the molecular weight measurement of the pyrolysis liquids is a strong function of the technique used. In this study, NMR data were utilized to calculate MW of the hydrocarbons. In addition, selected samples were analyzed by FIMS to obtain molecular weight versus ion intensity profiles. A comparison of FIMS number average molecular weight with the MW determined by NMR showed good qualitative agreement, although the absolute values showed significant differences. We considered measuring the molecular weight of our samples using vapor pressure osmometry (VPO). However, because the pyrolysis liquids studied in our samples contained a wide variation in polar functional groups (Tables I1 and 111), the VPO molecular weight resulta were questionable, especially for the heavier polyfunctional materials. A single-variable model based on the refractive indices of the liquid could be used for this prediction. This model is given by the equation MW = 4383.2- 2655n (7) The predicted and measured molecular weight based on this model is shown in Figure 5A. A two-variable model based on the liquid density (d) and the middistillation point (T,) could also be used for predicting the MW of the pyrolysis liquids (Figure 5B): MW = 1.348Tm- 1249.24d 1234.6 (8)
+
The distillation data were obtained by using the ASTM D-86 procedure. A different model, based on the same two-variable model (similar to eq 8) but with the distilla-
Properties of Pyrolysis Liquids
Energy & Fuels, Vol. 2, No. 6,1988 839
Table V. Summary of Regression Equations for Pyrolysis Liquids eq set
reeression ea (H) = 57.264 - 30.50n (H/C) = 6.876 - 3.50n n = refractive index H/C = hydrogen-to-carbon ratio of liquids (atomic) C, = 3.657n - 5.228 C, = 2.887n - 0.222(H/C) - 3.71 C, = carbon aromaticity (see also Table VI) , H 2.102n - 3.103 H, 1.739n - 0.105(H C) 2.39 H, = n*.""(H/C)-'.4$ H, = proton aromaticity MW = 4.383.2 - 2655n MW = 1 . 3 3 2 T ~- 1573.28d + 1589.1 MW = 1.348Tm- 1249.24d + 1234.6 MW = molecular weight Tw = temperature of middistillation point (TGA) T, = temperature of middistillation point (ASTM D-86) d = 1.977n - 2.08 d = density, g/cm3 C, = 37.75n - 55.29 C, = Conradson carbon residue
R2" 0.70 0.55
F' 65.3 35.4
0.59-
Pd ~
0.51-
~~
o.Ooo1 o.Ooo1
>
55 0.43-
3 5 0.82 0.86
131.9 82.5
o.Ooo1 o.Ooo1
0.77 96.8 50.5 0.79 0.97* 531.5
o.Ooo1
5 0.35P
s
OZ7iF
o.Ooo1 o.Ooo1
0.70 0.82 0.94
1.b
1.64
1.58
l.iO
REFRACTIVE INDEX
65.3 o.Ooo1 9.03 0.033 23.0 0.015
-
C~ro=-81.77+102.28n 31.759nZ
lRZ = 0.86) 0.50
0,40i 0.45
0.97
155.2
o.Ooo1
0.77
16.5
0.0097
0.35
0.25
I, R2measures how much variation in the dependent variable can
be accounted for by the model (i.e., independent variable). *R2is redefined for the no-intercept case by SAS.= Note that this is a nonlinear equation in n but the fitting can be handled by linear regression models. 'The F ratio is the ratio produced by dividing the mean square for the model by the mean square of error. It tests how well the model as a whole (after adjusting for the mean) accounts for the behavior of the dependent variable.agpu d P defines the "observed level of significance". In statistical terms, the level of significance, a, of a test is defined as the probability of rejecting the null hypothesis (Le., no linear relatiohship between the dependent and independent variables) given the null hypothesis is true. P gives us the largest value of a that would lead to the acceptance of the null hypothesis. Table VI. Prediction of Carbon Aromaticity Based on Linear and Nonlinear Mbdels" mean residual % prediction R2 dev y = -81.770 + 102.280n - 31.759 n2 0.86 5.80 y = -5.227 + 3.657nb 0.82 6.97 = e-1.028n14.018 0.80 11.43 "Key: y = carbon aromaticity (C,). Residual percent deviation is defined as (Iresiduall/lyl) X 100. *Note: the linearity test of NatrellaZ6is significant.
tion data obtained from TGA evaporation, can be defined by MW = 1.3322'50 - 1573.28d + 1589.1 (9) In the above equation, TN stands for the middistillation temperature as determined by TGA. A comparison between the predicted and measured MW is provided in Figure 5C. The correlation coefficient between the refractive index and the density (Figure 6A) of the liquids was 0.98. A linear regression fit provided the following model: d = 1.977n - 2.08
1.kz
1.bo
A
(10)
The measured and predicted density based on the above model is presented in Figure 6B. The correlation coeffi-
1 /
0.20/,/,
B
O ,
2 OBS Hidden
,
,
,
,
,
,
,
,
,
,
,
0.20 0.23 0.26 0.29 0.32 0.35 0.38 0.41 0.44 0.47 0.50 0.53 0.56 0.59
0.51
0.53
E
5
0.45
8
0.37
PREDICTED
-,
-I
L
3
U
0.29
0 21 021
0 31
0 41 PREDICTED VALUE
0 51
0 51
Figure 3. (A) Correlation between the carbon aromaticity and the refractive index (correlation coefficient 0.91). (B)Predicted and measured carbon aromaticity, with the prediction based on refractive index (n) (quadratic model). (C) Predicted and measured carbon aromaticity with predicted values based on a twovariable model that considers both the refractive index (n) and the liquid hydrogen-to-carbon ratio.
cient between the refractive indices of various distillation cuts and the corresponding density for the COED syncrudes (reported by Sturm et al.&) was 0.99, calculated in this study. The correlation coefficient between the Conradson carboh (C,,,) residue of the mild pyrolysis liquids derived in this study and the refractive indices of the liquids was 0.88 (Figure 7A). A single-variable model was defined by C,,, = 37.75n - 55.29 (11) The measured and predicted C,,, based on the above model is compared in Figure 7B. A liquid combustion
840 Energy & Fuels, Vol. 2, No. 6, 1988
Khan A
0.23 0.25
/ 0
6ool 500
0
I c
b
400-
5
8
22 300'O01
A
1.50
I
I
I
I
1.52
1.54
1.56
1.58
I
A
160
2b0
360
4b0
5d0
SbO
PREDICTED VALUE
160
REFRACTIVE INDEX
6ool
i_l 5001
>
c
g 0.19
3
0.17 0.15
& 0.13
300
E 0.11 0.09 100
0.07
B
100
I
200
B
I
300
460
I
500
Bob
560
600
PREDICTED VALUE
0.05 0.05
0.10
0.15
0.20
0.25
6ool
0.30
PREDICTED VALUE
500
H
0.1.j
t 0.13
c
/",
0.051I
c
I
0.05
0.10
I
I
0.15
0.20
0.15
1
0.30
PREDICTED VALUE
Figure 4. (A) Correlation between proton aromaticity and refractive index (correlation coefficient 0.88). (B) Predicted and actual proton aromaticity data based on a single-variable model. (C) Predicted and measured proton aromaticity data based on a two-variable model (including refractive index and liquid H/C ratio).
parameter (second maximum peak of combustion) measured in a thermogravimetric analyses system (TGA) could be positively correlated with the refractive indices of the liquids (data not shown). The significance of this TGA second maximum peak is not known. It may be worthwhile to attempt correlations of liquid combustion properties with their refractive indices, keeping previous relationships observed in this study in mind. Additional correlations between the refractive indices and the pyrolysis liquid characterization data me presented in Figures 8 and 9. Figure 8 shows a relationship between
I
100
260
3b0 4b0 PREDICTED VALUE
Figure 5. (A) Predicted and measured molecular weight based on a single-variablemodeL (E!) predictad and m e a s d molecular weight based on a two-variable model (including density and middistillation point (T,)with distillation data obtained by the ASTM D-86 procedure. (C) Predicted and measured molecular weight based on a two-variable model (including density and middistillation point (Tm)as determined by a TGA procedure).
the refractive indices of the whole liquid and the tricyclic saturated compounds (so-called "high-density fuels") in the 300-600 O F cut of the neutral fraction of the pyrolysis liquids as determined by a GC/MS technique (the correlation coefficient of the data set was 0.80). The liquid distilled between 300 and 600 O F was of particular interest because this boiling range approximates the range of JP-8, a high-density aviation fue1.8b*21 The total naphthalenes content in cut 2 (boiling between 300 and 600 O F ) of the pyrolysis liquids present in the neutral fraction could be directly correlated with the refractive indices of the whole liquids (Figure 9; correlation coefficient 0.98).
Energy & Fuels, Vol. 2, No. 6,1988 84 1
Properties of Pyrolysis Liquids
1
1.05
A
1.508
I
I
I
I
I
I
I
I
1.518
1.527
1.537
1.546
1.556
1.565
1.575
1.584
I 1.594
REFRACTIVE INDEX
A
I
I
1.50
1.52
I
I
I
I
1.54
1.56
1.56
1.60
51
REFRACTIVE INDEX
7
/
/
1.01
B
I
o
1
2
3
4
5
PREDICTED VALUE
B 0.90
I
0.95
I
100
I
1.06
I
110
PREDICTED VALUE
Figure 6. (A) Relationship between the density and the refractive index of the pyrolysis liquids (correlationcoefficient 0.98). (B) Predicted and measured density of pyrolysis liquids. The predicted density was based on the refractive index of the pyrolysis liquids.
Figure 7. (A) Relationship between the refractive index and the Conradson carbon residue, in percent (correlation coefficient 0.88). (B) Predicted and measured Conradson carbon residue. The predicted values were based on refractive indices of the liquids.
Figure 10 compares the cetane index of the hydrogenatedsb COED liquids distilled a t various temperature ranges (data from Sturm et a1.8a)with the corresponding refractive index. A summary of correlations developed in this study is presented in Table V.
t
t
t
Discussion The results of this study demonstrate that the refractive index, used in petroleum literature to correlate fuel properties, is a relevant property to evaluate pyrolysis liquids derived from coal, oil shale, and tar sand. It is demonstrated that the liquids' H/C, density, hydrogen content, molecular weight, and aromaticity correlate with refractive indices of the pyrolysis liquids. The relationship between the refractive index (n)and density (d) and other properties (e.g., Conradson carbon residue) of the liquids is consistent with our physical understanding of the structural properties of hydrocarbons. Paraffins have the highest hydrogen content and tend to have "fluffy" chain structures with low densities. For pure compounds, the lower the density, the lower the refractive indices. On the other hand, relatively compact cyclic (dense) aromatics are deficient in hydrogen. Cycloparaffins, which are both cyclic and hydrogen saturated, represent the best compromise between density and hydrogen content.20i21 ,The refractive indices of pyrolysis liquids correlate with the liquid H/C ratio (correlation coefficient -0.74) or the hydrogen contents of the liquid (correlation coefficient
1
O-I,', 1.508
, , , , , , , , , , , 1.518 1.530 1.542 1.554 1.566 1578
I
,
1.590
REFRACTIVE INDEX
Figure 8. Correlation between refractive index and the total tricyclic saturated compounds of the liquids (in cut 2, i.e., the 300-600 O F fraction of the neutral portion of pyrolysis liquids) (correlation coefficient of 0.8).
-0.83). In this study, models have been developed to predict the H/C ratio or hydrogen contents of the liquids based on their refractive indices. It is not surprising that refractive index correlates well with the liquid aromaticity. An increase in hydrogen content of fuels often leads to a decrease in aromaticity. For example, benzene with H/C = 1has fa = 1,while C6H12 (cyclohexane) with H/C = 2 has fa = 0. Here, fa is the proton or carbon aromaticity. The correlation coefficients between refractive index and proton and refractive index and carbon aromaticity were 0.88 and 0.91, respectively. One- and two-variable models were applied to predict the carbon and proton aromaticities of the pyrolysis liquids, based on their refractive indices. For a given carbon
Khan
842 Energy & Fuels, Vol. 2, No. 6, 1988 50
tion as determined in a TGA (Tho) or by an ASTM technique (ASTM D-86, T,,,)]. Conradson carbon residue (C,) could be correlated with the hydrogen content (correlation coefficient of -0.92). C, also showed correlations with the carbon and proton aromaticities (correlation coefficients of 0.91 and 0.84, respectively). Thus, the observed correlation between the refractive index and C, is not surprising. The observed relation may reflect the well-known fact that aromatic compounds are more susceptible to coke formation during severe heat treatment. We could also correlate Conradson carbon residue with the monopolar contents of the liquids (data not shown). The higher the polar fraction in the liquid, the greater was the residue. This suggests the significance of polar molecules in residue formation.
/ P
1I
1.50
1.52
1.54 1.56 REFRACTIVE INDEX
1.56
1.60
Figure 9. Relationship between the total naphthalene contents of the pyrolysis liquids (in cut 2) and their refractive indices (correlation coefficient 0.98).
--66 -
* *
74 72 70 68
*
8:;:
t
*
9 Bf 1 2 54 8 52 50 48 46 44 42
---
* *
-
*
I
1.40
1.41
1.42
1.43
1.44 1.45 1.46 1.47 1.48 REFRACTIVE INDEX
1.49
1.50
1.51
1.52
Figure 10. Relationship between the cetane index and the refractive indices for the COED syncrude.
aromaticity, refractive indices correlated with the middistillation points of the liquids. Molecular weights are fundamental properties of liquids. For various classes of pure compounds, refractive indices tend to change rapidly a t low molecular weight but less rapidly a t higher molecular weight. I t appeared that refractive indices of the pure liquids correlated with the negative half-power of molecular weight (data not shown). In this study, molecular weight could be predicted based on a single (e.g., refractive index, n) or a two-variable model [including density ( d ) and temperatures of 50% evapora-
Summary and Conclusions I t is demonstrated in this study for the first time that the refractive indices of pyrolysis liquids derived from a range of coal, oil shale, and tar sand serve as a useful property to evaluate the fuel-related physical and chemical properties of these fuels. The refractive indices of the liquids correlate well with the liquid hydrogen content (weight percent) and the H/C ratio, aromaticity (carbon and proton), molecular weight, density, and Conradson carbon residue. Because a number of physical and chemical properties for the pyrolysis liquids derived from a diverse range of feedstocks correlate, it is suggested that the changes in the liquid properties follow a certain order that can be treated mathematically. A number of empirical equations have been developed on the basis of these correlations. Results of this study also identify some of the compositional differences in the liquids derived from coal, oil shale, and tar sand. These differences may serve as guidelines for determining appropriate upgrading procedures for the pyrolysis liquid fuels derived from various feedstocks. Acknowledgment. Funding for this work was provided by the U.S. Department of Energy, Assistant Secretary for Fossil Energy, Office of Coal Utilization, Advanced Conversion and Gasification. I thank Dr. K. Seshadri of EG&G for the LC data, J. Wescott and T. Kowalski, undergraduate student trainees through the Oak Ridge Associated Universities (ORAU), for performing the fiied-bed pyrolysis runs, and Dr. E. Gunel for his input in the statistical aspect on a portion of this study. Registry No. C, 7440-44-0.
