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, US . Department of Energy,
Morgantown, West Virginia 26507-0880
Received February 2, 1988. Revised Manuscript Received August 15, 1988
Characterization techniques based upon the liquids 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 at 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 liquids 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-tem-
perature devolatilization of coal.t2 Physical and chemical
properties characterization of these liquids has been the
focus of several studies during the last few year~.l -~
Correlations based on the data obtained from relatively
simple characterization techniques for predicting fuel-re-
lated 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 be-
havior/properties (e.g., viscosity, smoke point). Mea-
surements of molecular weight or aromaticity can be dif-
ficult, 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 Gasifi-
cation Processes: Low-Temperature Devolatilization Studies. Proceed-
ings 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;
NTI S 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., J r.; 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 dis-
tillate.
MajumdarGb reevaluated the data presented by White
et a1.6aand 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 a1.6a incorrectly 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-based process, 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 hydro-
liquefaction products of only two coals.
In previous studies, it was demonstrated by KhanQ that
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 mea-
sured refractive indices.
The objective of this paper is to illustrate the relation-
ship 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 prop-
erties of various distillate cuts for a given liquefaction
distillate.
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-per-
formance liquid chromatography (HPLC). The HPLC chroma-
tographic 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 pmsilica gel bonded with cyano group (Supelco
LC-CN) in one column and 5 pm plain silica (Supelco LC-Si) in
the other. SESC fraction 1 was 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 chro-
matogram 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 I n order to
identify individual components, the SESC fraction-1 (satu-
rate/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 chroma-
tographic 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 total sample that
was charged to the column. The results for the neutral fractions
reflect the adjusted weights of the neutral fractions after ac-
counting for the residual benzene. Detailed analysis and dis-
tribution of the components identified in various fractions will
be presented in a future communication. The distillate cut be-
tween 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~chi . ~ ASTM-D86 was also applied to determine the distillation
profiles for selected liquids. ASTM-D86 was performed at Saybolt
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 ex-
perimental procedures used and the reproducibility of data are
available.'~2~6 A range of feedstocks (primarily coal but also oil
shale and tar sand) weredevolatilized 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 pre-
vious st~di es. ~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 pro-
cedures 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
The aromaticity of the liquids was measured at the University
of North Dakota Energy and Minerals Research Center (UN-
DEMRC) 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 at Virginia Polytechnical Institute and the Mor-
gantown 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).lGb The liquid chromatography (LC) sep-
aration of the mild pyrolysis liquids was performed by using
sequential elution solvent chromatography (SESC). The proce-
dure 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
ASTM D-70.
(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, PetroleumChemistry Division, Denver, CO, April 1987.
(12) Khan, M. R. 19th Oil Shale SymposiumProceedings; 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) Ret-
cofsky, 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.
(19) John, G. A.; Buttrill, S. E., J r.; Anbar, M. "Field Ionization and
Field Desorption Mass Spectroscopy Applied to Coal Research." In Or-
ganic Chemistry of Coal; Larsen, J., Ed.; ACS SymposiumSeries 71;
American Chemical Society: Washington, DC, 1978; p. 223.
(20) Korosi, A.; Rubin, J. N. "Hydroprocessing of Light Pyrolysis Fuel
Oil for KeroseneType J et Fuel"; Technical Report AFWAL TR-83-2048;
Wright-Patteron Air ForceBase: Dayton, OH, 1980.
(21) High-Density Candidate Hydrocarbon Fuels from Refinery By-
Product Streams; Technical Report (Interim), Contract No. F33615-83-
C-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.
836 Energy & Fuels, Vol. 2, No. 6, 1988
Table I. Origin and Apparent Rank of the Samples
PSOC geological location" apparent
no. (Dr0vince:reeion:seamname:citv. state) rank
181
267
275
296
306
355
375
1109
1313
1323
1448
1469
1470
1471
1472
1473
1475
1481
1492
1502
1504
1517
1523
1524
- -.
123 East:Appal:Lower Kittanning:Bickmore, WV
InkEaskUDDer BlockPatricksbureh. IN
East:Appai:klintwoodAirport, V i '
East:AppakOhio No. GACadiz, OH
East:Appal:Ohio No. 5:North Lima, OH
East:AppakOhio No. 12:Yorkville, OH
East:AppakLower Kittanning:Darlington, PA
EaskAppakHazard No. QHarlan, KT
Rock Mt:SW Utah:King CannekMt Caramel
East:Appal:Lower Kittanning:Mt Zion, PA
1nterior:East:Ill No. 6:Danville, IL
North Great P1ns:York Canyon, NM
East: AppakMary Lee:Goodsprings, AL
East: AppakPratkOakman, AL
East:AppakPee Wee:Oliver Springs, TN
EaskAppakLower Banner:Dante, VA
East: AppakUpper Banner:Dante, VA
East:AppakElkhorn No. 3Wayland, KY
East:Appal:Upper C1arion:Blairs Corners, PA
InkEaskI11 No. 6:Cutler, IL
Rock Mt:Uinta:Hiawatha:Emery, UT
Rock MkUinta:Upper Sunnyside:Sunnyside,
East:Appal:Ohio No. 5: Petersburg, OH
East:Appal:WV No. 5 Block:Ward, WV
J ct, UT
UT
hvAb
Sub A
hvAb
hvAb
hvAb
hvAb
hvAb
hvAb
hvBb
mvb
hvBb
hvAb
hvAb
hvAb
hvAb
hvAb
hvAb
hvAb
hvAb
hvCb
hvCb
hvAb
hvAb
hvAb
East:Appal:Upper Kittanning:Philippi, WV hvAb
sample rank
Pittsburgh No. 8:Arkwright Seam:WV hvAb
Illinois No. 6:Burning Star Mine, IL hvCb
North Dakota lignite lignite
Colorado oil shale (Mahogany Zone, Green River)
Sunbury oil shale (Kentucky Shale, obtained from
Asphalt Ridge Basin tar sand (Utah tar sand,
resinite sample separated fromUtah coal (procured
apparent
KCERLb)
obtained fromWRI')
fromthe University of Utah)
"Location: East =Eastern; Int =Interior; Appal =Appalachi-
an; 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 com-
position will bepresented separately. As one would expect,
the physical and chemical properties of the liquids derived
from various feedstocks vary widely. A number of char-
acterization techniques were applied to investigate quan-
titatively the compositional differences between various
feedstock liquids. Some typical examples of these varia-
tions 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 I1shows 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: NewYork, 1978.
(24) SAS User's Guide: Basics, Version 5 Edition; SAS Institute:
( 25) Guffey, F., personal communications, 1987.
Cary, NC, 1985.
Khan
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 5.0 8.6 3.2 3.8
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 1.4 1.3 13.0 17.5
(B) Class Types of Predominant Components of Fractions
Separated in SESC
fraction eluant
1 hexane
2 hexane/l5% toluene
3 chloroform
4 chloroform/4%
5 diethyl ether/3% ethanol
6 methanol
major components
alkanes/alkenes, neutral
nonbasic N-, 0-,
aromatics
S-heterocyclics, neutral
aromatics
monophenols
polyphenols, carbonyls,
diethyl ether
polyphenols, basic nitrogen
heterocyclics
highly functional (polar)
molecules (large
heteroatomcontents)
a Key: Pittsburgh, Pittsburgh No. 8 coat tar; resin, concentrated
resin tar separated fromUtah coal, Colorado, Colorado shale oil;
tar sand, Asphalt Ridge tar sand liquid; Sunbury, Kentucky Sun-
bury shale oil.
Table 111. Separation Results for SESC Fraction 1 of Mild
Pyrolysis Liquids
wt % in SESC
fraction 1
saturates
tar samole olefins aromatics recoverv
and %
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 total liquids. These fractions contain saturate/alkene
and neutral aromatics (e.g., furans, diary1 ether, benzo-
furans, 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 frac-
tions 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 al-
kanes, 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-en-
ergy-density aviation fuels.
Table V presents a summary of equations used to pre-
dict the properties of pyrolysis liquids. The hydrogen-
to-carbon ratio (H/C) of the pyrolysis liquids is an im-
Properties of Pyrolysis Liquids Energy & Fuels, Vol. 2, No. 6, 1988 837
Table IV. Characterization Data Used for the Predictive Equationso
obsb sample n C, H, H/C H MW T,,OF TM,OF d,g/cmS c,
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
~~
PSOC 123 1.5518
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
I11No. 6
ND lignite
Col Shale
Sun Shale
ASP tar sand
conc. resin
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
0.46
0.50
0.57
0.57
0.51
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
0.16
0.19
0.24
0.24
0.19
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.
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
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
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
198
172
163
158
177
172
177
163
178
180
176
189
178
158
192
185
165
163
164
234 180 0.9019
164
185
183
175
185 165 1.0306
157 1.0444
153
199 175 0.9024
187 0.9548
290 202 0.9059
0.9930 257 233
2.16
4.62
3.94
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
0 011Shale
Coal
Monocycllc
AlkaneJlOlefins Pentacyclic
Tetracycllc Alkanes
Alkanes
2 0 -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 satu-
rate/olefin HPLC fraction.
portant fundamental property of liquid fuels that influ-
ences 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:
(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)
[H/C] =6.876 - 3.50n
(Figure 2D). The R2 for the model is 0.70. R2 measures
the proportion of total variations explained by the re-
gression. It is calculated by dividing the sum of squares
due to regression by the total sum of squares. In simple
linear regression, R2 is 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 aro-
maticities (C,) and the refractive indices (n) for the py-
rolysis liquids derived from various fossil fuels. The cor-
relation coefficient between the variables was 0.91. The
carbon aromaticity (CA could be predicted by the linear
regression equation
(3)
The R2 for the model is 0.82.
The data were tested for linearity (by Natrellas 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
( 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
C, =3.657n - 5.228
C, =102.280n - 31.759n2 - 81.770
~~
(26) Natrella, M. Experimental Statistics; NBS Ha d b . (U.S.) 1963,
91.
838 Energy & Fuels, Vol. 2, No. 6, 1988
1.8 -
1.7 -
1.6 -
2 1.5 -
0
! l4 -
9 1.3 -
1.2 -
Y
Khan
1.0 .I,
A 1.5D 1.52 1.54 1.56 1.58 1.60
REFRACTIVE INDEX
1.0 I
B 1.0 1.2 1.4 1.6 1.8
PREDICTED VALUE
13 i
8 I , I
c 1.508 1.518 1.527 1.537 1.548 1558 1.585 1575 1584 1.594
REFRACTIVE INDEX
8 I I I I I
9 10 11 12 13
PREDICTED VALUE
D 8
Figure 2. (A) Correlation between the liquid hydrogen-to-carbon ratio and the refractive index (n) of the liquids. (B) Predicted and
measured hydrogen-to-carbon ratio based on a single-variable model. (C) Correlation between the liquid hydrogen content (dry basis)
and the refractive index of the liquids (correlation coefficient -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 mea-
sured 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 bepredicted (Figure
4B) on the basis of a single-variable model, defined by
(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-
propriate for H,,. Nevertheless, a nonlinear model was
attempted. The predicted and actual data based on a
two-variable model are shown in Figure 4C. Proton aro-
maticity could also be predicted on the basis of the fol-
lowing 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
H,, =2.102n - 3.103
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 ob-
tain molecular weight versus ion intensity profiles. A
comparison of FIMS number average molecular weight
with the MW determined by NMR showed good qualita-
tive agreement, although the absolute values showed sig-
nificant differences. We considered measuring the mo-
lecular weight of our samples using vapor pressure osmo-
metry (VPO). However, because the pyrolysis liquids
studied in our samples contained a wide variation in polar
functional groups (Tables I1and 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
0.59-
0.51-
>
5
5 0.43-
3
5
P
s
5 0.35-
Energy & Fuels, Vol. 2, No. 6, 1988 839
Table V. Summary of Regression Equations
for Pyrolysis Liquids
eq
set reeression ea R2" F' Pd
(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 =1.332T~ - 1573.28d +1589.1
MW =1.348Tm- 1249.24d +1234.6
MW =molecular weight
Tw =temperature of
T, =temperature of mid-
middistillation point (TGA)
distillation point (ASTM D-86)
d =1.977n - 2.08
d =density, g/cm3
C,, =37.75n - 55.29
C, =Conradson carbon residue
0.70
0.55
0.82
0.86
0.77
0.79
0.97*
0.70
0.82
0.94
0.97
0.77
65.3
35.4
131.9
82.5
96.8
50.5
531.5
65.3
23.0
9.03
155.2
16.5
~ ~~
o.Ooo1
o.Ooo1
o.Ooo1
o.Ooo1
o.Ooo1
o.Ooo1
o.Ooo1
o.Ooo1
0.033
0.015
o.Ooo1
0.0097
, I R2 measures how much variation in the dependent variable can
be accounted for by the model (i.e., independent variable). *R2 is
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. I t
tests how well the model as a whole (after adjusting for the mean)
accounts for the behavior of the dependent variable.agpu d P de-
fines 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 hypothe-
sis 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
NatrellaZ6 is significant.
tion data obtained from TGA evaporation, can be defined
by
(9)
In the above equation, TN stands for the middistillation
temperature as determined by TGA. A comparison be-
tween 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 (10)
The measured and predicted density based on the above
model is presented in Figure 6B. The correlation coeffi-
MW =1.3322'50 - 1573.28d +1589.1
OZ7i F
1.bo 1.kz 1.64 1.b 1.58 l .i O
REFRACTIVE INDEX
A
C~r o=- 81. 77+102. 28n - 31.759nZ
lRZ =0.86)
0.50
0.45
0.35 0,40i
0.25 1 /
2 OBS Hidden
0. 20/ , / , O , , , , , , , , , , , ,
B
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
PREDICTED
0.51 -,
-I
0.53
E
5 0.45
L
8 0.37
3
U
0.29
0 21
021 0 31 0 41 0 51 0 51
PREDICTED VALUE
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 mea-
sured carbon aromaticity with predicted values based on a two-
variable 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 syn-
crudes (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
(11)
The measured and predicted C,,, based on the above
model is compared in Figure 7B. A liquid combustion
C,,, =37.75n - 55.29
840 Energy & Fuels, Vol. 2, No. 6, 1988
0.25 -
0.23 -
Khan
A
/
0 0
b
I I I I I
A 1.50 1.52 1.54 1.56 1.58 160
REFRACTIVE INDEX
>
c
g 0.19
0.17
0.15
& 0.13
0.11
E
0.09
0.07
0.05
B 0.05 0.10 0.15 0.20 0.25 0.30
PREDICTED VALUE
0.1.j
t 0.13
0.051 I /",
I I I 1
c 0.05 0.10 0.15 0.20 0.15 0.30
PREDICTED VALUE
Figure 4. (A) Correlation between proton aromaticity and re-
fractive index (correlation coefficient 0.88). (B) Predicted and
actual proton aromaticity data based ona single-variablemodel.
(C) Predicted and measured proton aromaticity data based on
a two-variablemodel (including refractive index and liquid H/C
ratio).
parameter (second maximum peak of combustion) mea-
sured 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
500 6ool
c I
400-
5
8
2 300-
2
' O01
A 160 2b0 360 4b0 5d0 SbO
PREDICTED VALUE
6ool
5001
3 i_l 300
100
I I I
B 100 200 300 460 500 B o b
PREDICTED VALUE
500 6ool
H
I
c 100 260 3b0 4b0 560 600
PREDICTED VALUE
Figure 5. (A) Predicted and measured molecular weight based
ona single-variablemodeL (E!) predictad and measd 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 corre-
lation 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).
Properties of Pyrolysis Liquids Energy & Fuels, Vol. 2, No. 6, 1988 84 1
1.05 1
I I I I I I
A 1.50 1.52 1.54 1.56 1.56 1.60
REFRACTIVE INDEX
1.01
/
I I I I
0.90 0.95 100 1.06 110
PREDICTED VALUE
B
Figure 6. (A) Relationship between the density and the refractive
index of the pyrolysis liquids (correlation coefficient 0.98). (B)
Predicted and measured density of pyrolysis liquids. The pre-
dicted density wasbased onthe refractive index of the pyrolysis
liquids.
Figure 10 compares the cetane index of the hydrogen-
atedsb COED liquids distilled at 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.
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. I t 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 un-
derstanding 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. Cyclo-
paraffins, which are both cyclic and hydrogen saturated,
represent the best compromise between density and hy-
drogen 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
I I I I I I I I I
A 1.508 1.518 1.527 1.537 1.546 1.556 1.565 1.575 1.584 1.594
REFRACTIVE INDEX
51 7
/
I
B o 1 2 3 4 5
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 werebased onrefractive indices of the liquids.
PREDICTED VALUE
t
t
t
1
O - I , ' , , , , , , , , , , , , I ,
1.508 1.518 1.530 1.542 1.554 1.566 1578 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.
I t 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 =1 has f a =1, while C6H12
(cyclohexane) with H/C =2 has f a =0. Here, f a 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
842 Energy & Fuels, Vol. 2, No. 6, 1988
74 -
72 -
70 -
68 -
66 -
8: ; :
9 Bf 1
2 54 -
8 52 -
50 -
48 -
46 -
44 -
42 -
Khan
50 1 /
I P
1.50 1.52 1.54 1.56 1.56 1.60
REFRACTIVE INDEX
Figure 9. Relationship between the total naphthalene contents
of the pyrolysis liquids (in cut 2) and their refractive indices
(correlation coefficient 0.98).
*
* *
* *
t
*
*
I
1.40 1.41 1.42 1.43 1.44 1.45 1.46 1.47 1.48 1.49 1.50 1.51 1.52
REFRACTIVE INDEX
Figure 10. Relationship between the cetane index and the re-
fractive indices for the COED syncrude.
aromaticity, refractive indices correlated with the mid-
distillation points of the liquids.
Molecular weights are fundamental properties of liquids.
For various classes of pure compounds, refractive indices
tend to change rapidly at low molecular weight but less
rapidly at higher molecular weight. I t appeared that re-
fractive 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-
tion as determined in a TGA (Tho) or by an ASTM tech-
nique (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 aro-
maticities (correlation coefficients of 0.91 and 0.84, re-
spectively). 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. Wecould 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.
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 chem-
ical 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 cor-
relations. 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 proce-
dures 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 Con-
version and Gasification. I thank Dr. K. Seshadri of
EG&G for the LC data, J . Wescott and T. Kowalski, un-
dergraduate student trainees through the Oak Ridge As-
sociated Universities (ORAU), for performing the fiied-bed
pyrolysis runs, and Dr. E. Gunel for his input in the sta-
tistical aspect on a portion of this study.
Registry No. C, 7440-44-0.