Anda di halaman 1dari 10

Ind. Eng. Chem. Process Des. Dev.

1901, 20, 349-358 949

Kinetics of Donor-Solvent Liquefaction of Bituminous Coals in


Nonisothermal Experiments
Govlndan Mohan and Harry Sllla
Department of Chemistry and Chemical Englmrhg, Stevens Institute of Technobgy, Hoboken, New Jersey 07030

An Illinois No. 6 coal was liquefied in a batch reactor under nonisothermal conditions from 330 to 450 O C using
tetralin as a hydrogendonor solvent at total pressures up to 70 atm and reaction times from 5 to 60 min. A
separation scheme was developed using liquid-solid chromatography on alumina to separate the Ilquefled products
into classes of compounds consisting of mixtures of aromatics, ethers, nltrogens, hydroxyls, and multifunctionals.
Using these classes of compounds several kinetic models were tested by flttlng experimental nonlsothermal
concentration-time profiies using a nonlinear parameter estimation technique. The multifunctional compounds were
considered as reaction intermediates which then decomposed reversibly Into aromatics, ethers, nitrogens, and
hydroxyls. It was found that this kinetic model predicts concentration-time profiles satisfactorily. This work has
also established the use of preparative liquld-solkl chromatography as a reliable and useful method for separating
coal liquids for kinetic studles.

Introduction solvents with the definition for the various solubility classea
In view of the need to develop commercially viable is given in Table I. These experiments were carried out
processes for deriving liquid fuels from coal a description with and without catalysts and donor solventa using mostly
of coal liquefaction kinetics is necessary for reactor design bituminous coals. The proposed kinetic models for li-
and scaleup. Most coal liquefaction processes involve quefaction with or without donor solvents are in many
dissolution of the coal by using a hydrogen-donor solvent cases similar. In many of the studies a hydrogen-donor
with added gaseous hydrogen, but among the various solvent with an added catalyst was used.
solvents employed in laboratory studies tetralin has been In the earlier proposed kinetic models the reaction
found to be the most common hydrogen donor. Fur- products, oils and asphaltenes, were combined or treated
thermore, synthetic solvents designed to simulate a typical separately to determine reaction rate constants. Pelipetz
process solvent invariably contain significant amounts of et al. (1955) and Hill et al. (1966) assumed that only a
tetralin. The addition of gaseous hydrogen in donor-sol- single thermal decomposition process occurs. Falkum and
vent liquefaction in several laboratory investigations has Glenn (1952) found that the hydrogenolysis of Spitsbergen
been shown to yield higher conversions (Han et al., 1978). coal occurred in two distinct stages and suggested that two
Recently, Maddocks and Huffman (1978) proposed that constituents are present in coal, one of which degraded
no hydrogen gas be added in the first dissolution stage of much faster than the other. Such a model, involving two
a two-stage coal liquefaction process. Also, in recent lab- reactive constituents, was investigated by Curran et al.
oratory studies Mitchell and Whitehurst (1978) and Lon- (1967) and Struck et al. (1969) and was found to fit their
ganbach et al. (1979) liquefied coal at short contact times experimental conversion data on coal extraction and
in the absence of gaseous hydrogen. In our studies tetralin coal-extract hydrocracking processes. Weller et al. (1951)
was used as the donor solvent without added gaseous hy- proposed that the conversion of coal to oils involves two
drogen. consecutive first-order reactions with asphaltenes as an
The kinetics of coal liquefaction are very complex intermediate product. They determined the rate constants
leading to the formation of numerous compounds. The for both steps as shown in Table I. On the other hand,
approach in kinetic studies has been to attempt to separate Liebenberg and Potgieter (1973) proposed two simulta-
these compounds into kinetically similar classes based neous reactions for the formation of oils and asphaltenes
on some separation technique. Most studies used methods in order to explain their experimental data. Yoshida et
of separation that depend on solubility in various solvents al. (1976) expanded this model to allow for oils formation
to separate coal liquids. These studies are summarized in by two paths. Recently, Chiba and Sanada (1978) assessed
Table I. Recently, Brunson (1979) used distillation to the previous models consisting of oils and asphaltenes and
measure the kinetics of formation of gaseous and liquid suggested using the formation of oils from two active coal
products using a coal-derived hydrogen-donor solvent. components to explain the experiments in the literature.
Prather et al. (1977) indicated the feasibility of using Recently, a third class of compounds, the preasphalt-
analytical liquid-liquid chromatography (LLC) to char- enes, was isolated from coal liquefaction products and was
acterize coal liquefaction solvents for studying kinetics, but shown to be a key reaction intermediate by Sternberg et
they did not attempt to use this approach to obtain al. (1975). Thus a more complex model was proposed by
product compositions for estimating kinetic parameters. Schweighardt and Sharkey (1976) using preasphaltenes in
Whitehurst et al. (1977) separated coal liquid products by addition to oils and asphaltenes. Subsequently, simplified
liquid-solid chromatography (LSC) on silica using various versions of this model have been satisfactorily tested in
eluting solvents in sequence. They concluded that the coal a batch reactor (Shalabi et al., 1979) as well as in a con-
liquefaction products contained a large number of com- tinuous reactor (Cronauer et al., 1978; Traeger, 1979).
pounds with very similar hydrocarbon skeletons differing Squires (1978) proposed a more comprehensive working
only by the various functional groups, and they also did model for donor-solvent coal liquefaction, but the model
not propose any kinetic model based on the separated has not been tested to date because the isolation procedure
products. for the product oil and semi-coke has not been es-
A summary of kinetic models and rate constants based tablished.
on product compositions obtained by solubility in various In this study LSC was employed to separate the coal
0196-4305/81/1 120-0349$01.25/0 0 1981 American Chemical Society
350 Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981

Table I. Summary of Coal Liquefaction Mechanisms Baeed on Solubility Separations


hydrogen donor rate constant at
investigators kinetic schemea coal solvent added catalyst 400 OC, min-I
Pelipetz et al. (1955) Wyoming none none k , = 0.0030
Hill et al. (1966) Utah tetralin none k , = 0.0372
Falkum and Glenn Spitsbergen none Ca-Cu-Cr k , = 0.0068
(1952)

Curran et al. (1967) Pittsburgh tetralin none k , = 3.0200


k , = 0.1160
Struck et al. (1969) Pittsburgh none ZnO/ZnCI, k , = 0.2400
extract k , = 0.0062
Weller et al. (1951) Pittsburgh none SnS, NH,CI k , = 0.0270
Anthraxylon k , = 0.0011

Liebenberg and bituminous tetralin none k , t k , = 0.0039


Potgieter (1973)
Yoshida et al. (1976) Japan decrystallized red-mud k , = 0.0037
anthracene oil and sulfur k , = 0.0049
k , = 0.0170

Cronauer et al. (1978) Belle Ayr hydrogenated none k , = 0.1492


phenanthrene k , = 0.1979
k , = 0.0855
k , = 0.0240
k , = 0.1546
k , = 0.0735

Shalabi et al. (1979) Kentucky tetralin none k , = 0.00043


k , = 0.00007
k , = 0.00075
k , = 0.00005
k , = 0.00120
k , = 0.00005
G, gases; 0, oils (hexane solubles); A, asphaltenes (hexane insolubles, benzene solubles); P, preasphaltenes (benzene in-
solubles; tetrahydrofuran solubles).

liquids into classes of compounds for kinetic analysis. reactor to minimize reactant heat-up times. This system
Chromatographic separation promises to be a more effi- included a double-pipe heat exchanger to prevent over-
cient and perhaps more fundamental method. heating of the Teflon seat of the ball valve caused by heat
conduction from the reactor. Nevertheless, the Teflon
Experimental Section seats still had to be cleaned or replaced after a few in-
System Design. Since Storch et al. (1940) reported the jections because of coal deposits or erosion. The holdup
use of batch reactors for kinetic studies, many investigators of the slurry in the injector, which was washed and mixed
have employed these massive reactors of various capacities. with the reactor wash at the end of a run, was found to
These reactors require long heat-up and cool-down times, be less than 0.5% of the slurry injected. This method of
during which significant reactions can occur. The ap- slurry injection caused a temperature drop of 30 "C of the
proaches used to minimize these nonisothermal heat-up reaction mixture upon injection, thus requiring a method
times have been to use small reactors constructed from to correct for the nonisothermal reaction conditions.
tubing (Neavel, 1976) or to first heat up the solvent to the The furnace temperature was controlled by a propor-
reaction temperature and inject the coal slurry with a tional action temperature controller using a chromel-alu-
pump (Shalabi et al., 1979). If isothermal conditions me1 thermocouple and solid-state switching (Barber-Col-
cannot be obtained experimentally, then a calculation man, Series 530). The reaction temperature was measured
procedure is required to correct for the nonisothermal manually (*l OC) with another chromel-alumel thermo-
conditions, which is the approach adopted here. Earlier, couple inserted into the thermowell of the reactor using
Yergey et al. (1974) applied the nonisothermal analysis to a cold-junction compensator and a potentiometer. The
estimate the kinetic rate parameters for the hydro- temperature was also recorded with a Leeds & Northup
desulfurization of coal. Recently, Morita et al. (1979) used Speedomax X/L 680 recorder with a full scale response
a nonisothermal analysis for estimating the reaction rate of 1 8.
parameters in studying the effect of hydrogen pressure on All the components of the Magnedrive impeller were
the rate of direct liquefaction of some Japanese coals. disassembled and cleaned thoroughly after each run.
The experiments discussed here were conducted in 1-L Neither the graphite bearings of the Magnedrive impeller
stirred reactor (Autoclave Engineers, Model AFP-1005, nor the stainless steel gasket needed replacement during
Magnedrive), shown in Figure 1, where all materials in the course of this work, because of the care and cleaning
contact with the coal slurry are 316 stainless steel. A coal procedure adopted. The reaction mixture was stirred at
slurry injection system was designed and added to the 1500 rpm since the mass transfer resistance has been re-
Ind. Eng. Chem. hocess Des. Dev., Vol. 20, No. 2, 1981 351

NITROGEN
REACTION
CHECK VALVE
SOLVENT PREHEATING WENCH

530

400
0
STATOR FOR
TACHOMETER u'
BALL VALVE
28 300
ir DOUBLEPIPE
HEAT EXCHANGER
200

IO0

I I I I

0 I 2 3 4
TIME, h

Figure 2. Typical temperature-time profile.


- W E N C H WATER
INLET Experimental Procedure. In a typical experiment, the
reactor is assembled and leak tested with nitrogen. Then
264 g of tetralin is introduced into the reactor at ambient
pressure, flushed with nitrogen, and heated to a steady-
COOLING COIL-
-FURNACE state temperature while stirring at 1500 rpm. A slurry
consisting of 30 g of coal in 36 g of tetralin is injected from
THERMOWELL- INJECTION TUBE the feed reservoir using nitrogen at a pressure of 30 atm
IMPELLER as shown in Figure 1. The coal will settle from more
concentrated slurries in the feed reservoir whereas less
concentrated slurries cool the reaction mixture appreciably.
The final solvent/coal ratio in the reactor was 10 in all
Figure 1. Coal liquefaction reactor system. experiments to keep the tetralin concentration relatively
constant during reaction (Hill et al., 1966; Shalabi et al.,
Table 11. Ultimate and Proximate Analyses of 1979). After injection of the cold slurry, the reaction
Illinois No. 6 Coalo mixture is cooled by approximately 30 O C after which the
ultimate proximate temperature rises steadily during the experiment as can
be seen in a typical temperature-time profile shown in
mass % mass % Figure 2. The pressure rise during reaction was about 25
(moisture (moisture
component free basis) component free basis) atm. After the desired reaction time, the reaction is rapidly
quenched by passing cold water through the cooling coils
hydrogen 4.8 volatile matter 42.3 and by blowing cold air over the reactor after dropping the
carbon 68.9 fixed carbon 45.2 furance with the aid of laboratory jacks.
nitrogen 1.2 ash 12.5
sulfur 3.1
The experiments were carried out at temperatures from
(total) 330 to 450 OC and pressures under 70 atm. Temperatures
oxygen 8.9 below 330 O C are ineffective for coal dissolution, resulting
ash 12.5 in low conversions (Guin et al., 1979) and temperatures
* Coal source, River King Mine, Ill. Donated by U.S. greater than 450 "C lead to excessive coke formation
Department of Energy, Pittsburgh, PA. Dried prior to (Storch et al., 1940). Reaction times of 5, 10, 15, 20, 30,
use. 40, and 60 min were selected for all the experiments.
Because injection of the coal slurry required 1 min and
ported to be insignificant at this speed (Hill et al., 1966). quenching 2 min, reaction times less than 5 min were not
Also, when the impeller speed was reduced to 1300 rpm considered. After 60 min a limiting conversion was
in this investigation, there were no significant changes in achieved.
the product concentrations. Product Separation Procedure. The separation
Materials. The coal used in all experiments was a procedure for the reaction products is shown in Figure 3.
high-sulfur and a high-ash Illinois No. 6 coal from the River After the reactor is cooled down to room temperature, the
King mine. The analyses of the coal are given in Table gaseous products are bubbled through 0.4 M lead acetate
11. The coal was dried, ball-milled and sieved through a solution to remove the H a and thus determine the amount
200 U.S.standard mesh screen and stored in a desiccator of sulfur in the gas phase. The black precipitate of PbS
under nitrogen atmosphere prior to its use. formed is filtered, washed with 0.1 M HN03 and then with
Technical grade tetralin was used as received from City distilled water before drying at 110 "C to constant weight.
Chemicals Corp., New York, and was analyzed by LC and Overall mass balances made for each experiment showed
found to contain 25 mol % of decalin and less than 1% better than 96% recovery of the total charge to the reactor.
of naphthalene. The reagent grade solvents used for The unwashed material sticking to the walls of the reactor,
washing the reactor and for the chromatographic separa- estimated to be 1%of the total charge, and the product
tions, and the alumina packing (Fisher No. A-540), used gaws constituted the losses. In early experiments methane,
to separate the liquid reactions products, were obtained ethane, propane, butane, and hydrogen were found from
from Fisher Scientific Co. a GC-MS analysis of the product gases, but no attempt was
352 Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981

0.25 r
THFWIB*OFREACTDR
4-4
REACTOR
GASES
t L E A 0 ACETATE
SCRUBBER

p-/-'T' SLURP"

'
i ' I
A 800 1200 I IO
VOLUME ELUTED, mi

Figure 4. A typical preparative liquid-solid chromatogram for


TETRALIN I "*C""U I solvent refined coal.

PACKING: BONDAPAK CI8/PORASIL B PEAKS 1. DECALIN


SOLVENT: METHANOLIWATER (77:23) 2. TOLUENE (Internal Standard1
FLOWRATE. 2.5 mllmin 3. NAPHTHALENE
4. DIHYDRONAPHTHALENE
5. TETRALIN

NiTRDGlNS

HIOROXYLS
*SH UULTlFUNCTlONilLI

Figure 3. Separation scheme for coal liquefaction reaction products.

made to account for the gaseous products in subsequent


runs.
The slurry products were collected in two stainless steel
flasks and centrifuged to separate the sediment (heavy
liquids plus solids) from the liquids. After the reactor
surfaces were washed with tetrahydrofuran (THF), the
wash was combined with the sediment. The total wash
was then extracted with THF in a Soxhlet extractor to DISTILLATE SATURATES
recover THF solubles. Subsequently, the THF in the Figure 5. Comparison of analytical liquid chromatograms of the
extract was evaporated in a Rotavap and the undissolved distillate and the saturates fractions.
material dried to constant weight and its ash content de-
termined. Any water produced during the liquefaction with the sequence of solvents used by Schiller and Mat-
would be evaporated in this step. hiason (1977). Each eluted sample was concentrated in
The Soxhlet extract was combined with the centrifuged a Rotavap and the solutions transferred to tared vials
liquids and the mixture was then distilled at 125 "C and where the solvents were evaporated at 25 "C and 2 mmHg
8 cmHg to remove most of the unreacted tetralin and its and the residues weighed. The uneluted material, which
decomposition products, naphthalene and dihydro- was obtained by difference, is the irreversibly adsorbed,
naphthalene. Volatile coal liquids could also be removed multifunctional and highly polar components of the SRC.
at this point but these losses are small because the total Six fractions were produced, five eluted and one adsorbed,
loss for the whole experiment is only 2%. The bottoms as shown in Figure 4. Reproducibility of the separations
product, referred to as Solvent Refined Coal (SRC), was in duplicate determinations ranges from 2% for the small
stratified. In order to obtain a uniform sample for liq- fractions and up to 5% for the large fractions.
uid-solid chromatography (LSC) the SRC was dissolved The coal products are divided into six classes of com-
in T H F to produce a homogeneous solution. Approxi- pounds, namely, aromatics, ethers, nitrogens, hydroxyls,
mately 5 mL of this solution was evaporated in a Rotavap multifundionals, and unreaded coal. The volatile portions
at 25 OC and 2 mmHg to eliminate THF in order to obtain of similar chromatographic fractions were analyzed by
an accurate sample (1.5 f 0.001 g) for LSC separations. Schiller and Mathiason (1977) using high resolution mass
Then the sample for the chromatographic column was spectrometry and it was found that the aromatics fraction
redissolved in THF and mixed with 10 g of alumina in a contained polyaromatic hydrocarbons and some aromatic
gas washing bottle to produce a material resembling wet ethers; the ethers fraction, benzofurans and naphthofurans;
sand. The alumina was then dried at 25 "C and 2 mmHg the nitrogens fraction, carbazoles, quinolines, and aza-
in the wash bottle where the evaporating THF vapors pyrenes; and the hydroxyls fraction, phenols, hydroxy
smoothly fluidized the alumina with no visible solids en- aromatics, and hydroxy ethers.
trainment. The dry alumina with the adsorbed sample was The distillate and the saturates fraction from the SRC
then loaded onto the top of a 25 mm i.d. X 600 mm were analyzed by analytical liquid-liquid chromatography
preparative chromatographic column, which had been (LLC) using a methanol-water mobile phase, and the
slurry-packed with 60 g of alumina in hexanes and chromatograms are compared in Figure 5. It is clear that
equilibrated overnight. The adsorbed SRC was then eluted both these fractions contain only decalin, naphthalene,
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981 353

s
1 0.16 i-

3800I
I / I IO 20 30 40 50 60
TIME, mln
REACTION TIME, min
Figure 6. Chebyshev fit to the measured reaction temperature-time Figure 7. Desulfurization of Illinois No. 6 coal.
profile.
dihydronaphthalene, and tetralin. Furthermore, it was The starting values for the rate parameters, Aj and Ej,
found that in the blank runs, using tetralin without coal were obtained by executing the isothermal version of the
under similar reaction conditions, the same product con- program using values for the rate constants that are con-
centrations were produced within experimental error. sistent with those in Table I. The optimal parameter
Hence no significant amounts of naphthalene and di- estimates of this program were the maximum likelihood
hydronaphthalene are derived from the coal under the estimates and they are neither unique nor global. Ob-
reaction conditions used, and the saturates fraction con- taining a satisfactory fit of a kinetic model to the exper-
tains the unrecovered tetralin and its reaction products imental concentration-time profiles by this method does
from the distillation step. Therefore, the saturates fraction not verify the correctness of a kinetic scheme because good
from the SRC was combined with the distillate and was fits can always be obtained if the kinetic scheme is made
not considered in the kinetic schemes. The aromatics sufficiently complex so as to introduce additional param-
fraction was partially soluble in the mobile phase and when eters. The method, however, does allow one to reduce the
analyzed showed the absence of decalin, naphthalene, number of possibilities because some kinetic schemes may
dihydronaphthalene, and tetralin. The other fractions be eliminated because of poor fits. Once a satisfactory fit
obtained from the preparative chromatographic separa- has been obtained additional experimental information is
tions are insoluble in the mobile phase. required before the kinetic scheme can be accepted. Also,
parameter estimation is useful in planning experiments.
Kinetic Modelling and Parameter Estimation Desulfurization. The coal used in this work is a very
The kinetic models are obtained by performing a dif- reactive Illinois coal with 3.7 mass % sulfur and 12.5% ash
ferential mass balance for each class of compounds present content. In order to estimate the extent of desulfurization,
in reactor a t any instant. Thus, the general set of dif- the mass fraction of sulfur removed as H2Swas calculated
ferential equations representing a proposed reaction on the basis of total sulfur in the coal. With such high ash
scheme and the initial conditions are written as content any iron oxides in the coal might be converted to
sulfides, and the pyrites into pyrrhotite and hydrogen
sulfide. Some sulfur may be present as organic compounds
in the THF solubles, but this was not determined. It was
Xi(0) = XIJi found that up to 18% of the total sulfur is removed as H a
The rate, Rj, depends on the component concentrations, during the experiment with large desulfurization occurring
X i , and temperature. The kinetic models developed here at higher temperatures and extended times as shown in
are restricted to systems of fmt-order reactions, where the Figure 7. These results are consistent with those reported
temperature dependence for the rate constant is given by by Longanbach et al. (1979) for the liquefaction of a West
the classical Arrhenius expression Kentucky 9/14 coal using Wilsonville recycle solvent under
similar conditions of temperatures and pressures except
k j = Aj exp(-Ej/(RT)) (2) that the reaction times were less than 5 min and sol-
The objective then is to estimate the rate parameters, vent/coal ratios less than 3.
Aj and E,, for proposed kinetic models by using nonlinear Coal Conversion Kinetics. Coal conversion is based
regression techniques (Bard, 1974). The estimation pro- on the fraction of liquefied coal that is insoluble in tet-
cedure consists of fitting the solutions of eq 1to the con- rahydrofuran. The coal converted on the dry and ash-free
centration measurements made at different reaction times basis, defined as
and temperatures since the experiments are nonisothermal.
The temperature variation during an experiment (Figure WC - WR
x = (3)
6) is fitted by a 10-term series expansion in Chebyshev WC
polynomials (Esch, 1974) with the Chebyshev coefficients
determined by unweighted least-squares fitting to the was reproducible within 4% absolute error. A first-order
reaction temperature-time profile. Hammings modified irreversible reaction scheme for coal dissolution is assumed
predictor-corrector method is used to numerically integrate in accordance with earlier investigations (Han et al., 1978)
the set of simultaneous linear differential equations given and the kinetic model is represented as
by eq 1. The nonlinear estimation algorithm (Meeter, dx
1965) is based on Marquardts method (1963), which - = -k(X, - X ) (4)
dt
combines the best features of the Gauss-Newton and
steepest descent methods. x(0) = 0
354 Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981

hydroxyls ( H )
nitrogens (N)
ti." y
coal (c) 2 multifunctionals ( M ) (5)
1.. A ethers (E 1
aromatics (A )

-
d
dt

I/
REACTION TIME, min

Figure 8. Conversion of Illinois No. 6 coal during liquefaction.

Table 111. Kinetic Parameters for Conversion of


Illinois No. 6 Coal
~~

limiting activation frequency


temp conversion, energy, factor,
range,"C Xe kcal/g-mol min-*
330-390 0.75 18.2 1.86 x 105
390-450 0.88 19.1 3.53 x 105

where the limiting conversion of a given coal, xe, obtained


at long reaction times, is a function of temperature as well
as the donor-solvent type (Cronauer et al., 1978). The
numerical solution of the above model is compared with
the experimental data in Figure 8. The best-fit rate pa-
rameters, determined by the estimation program, are given
in Table 111. The value of activation energy for coal
'MAlI'AM

aromatics (A)
e ethers (E)
dissolution agrees very well with that of 18.7 kcal/g-mol
for bituminous coals liquefied in a well-mixed batch reactor all the reactions shown in eq 5 are made reversible with
(Hanet al., 1978). the exception of the formation of the multifunctional
Coal Liquefaction Kinetics. To obtain kinetic models compounds from coal. The kinetic model corresponding
for coal liquefaction, the liquefied coal products were to the kinetic scheme in eq 10 is given by the set of
equations in matrix form in eq 11 and eq 12, 13, and 14.
separated into classes of compounds according to their
chemical functionalities, namely, aromatics (A), ethers (E), 0 0 0
nitrogen compounds (N), hydroxyl compounds (HI, large
multifunctional compounds (M), and unreacted coal (C). 0
Because significant qualitative differences in concentra- kMN 0 -kNM 0 0
tion-time profiles for the multifunctional compounds were ME 0 0 EM 0
observed at high and low temperatures, the entire exper-
imental range was divided into a high temperature range MA 0 0 0
(390-450 "C) and a low-temperature range (330-390 "C).
Several models were tested in order to obtain the best-fit
to the experimental concentration-time profiles.
A. High-Temperature Liquefaction. The experi-
mental concentration-time profile for the multifunctional
compounds in the high-temperature range (390-450 " C )
suggests that they me reaction intermediates. Perhaps this
is due to the initial thermal breakup of the "coal [XC, XM, XH, XN, XE, XAlO = [I,0, 0, 0, 0, 01 (12)
molecules'' to produce large free radicals, which are then P2 = ( ~ M H+ ~ M +N ME + MA) (13)
stabilized by abstraction of hydrogen from tetralin. Sub-
sequent bond breaking of these stabilized-multifunction- x , = 0.88 (Table 111) (14)
al-large molecules, which are highly polar, produce several The concentration-time profiles for the proposed mod-
smaller and less polar compounds. Accordingly, kinetic els, calculated according to the procedure outlined earlier,
models 1 and 2 were tested. are compared with the experimental concentration-time
Kinetic Model 1. Model 1 consists of a series-parallel profiles in the high temperature range in Figures 9 and 10.
reaction scheme of first-order irreversible reactions with Undoubtedly, model 2 with all but one reaction reversible
the large multifunctional compounds as reaction inter- fits the data best, which is further confirmed by the lower
mediates (eq 5). The kinetic model corresponding to the value of the minimum sum of squares of the residuals
kinetic scheme in eq 5 is given by the set of equations in shown in Table IV. Numerical values of the kinetic pa-
matrix form in eq 6 and eq 7, 8, and 9. rameters for model 2 are given in Table V. Reversible type
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981 355

I.",

1I Q 6 1 hMULTIFUNCTIONALS

NREACTEDCOAL

REACTION TIME, min


REACTION TIME, min
Figure 9. Comparison of kinetic model 1with experimental product Figure 10. Comparison of kinetic model 2 with experimental
concentrations in the high-temperature range. product concentrations in the high-temperature range.
Table IV. Kinetic Model Discrimination Using
Regression Analysis
sum of squares of residuals
kinetic model high temp low temp
Based on Chromatography
H f 0.61 \
1. C -M 4N 0.3378 0.1817
7;

a 0.0346

REACTION TIME, min


Based on Solubility
Figure 11. Comparison of kinetic model 1 with experimental
/O product concentrations-in the low-temperature range.
4. c - 0 0.0696 0.0256
\ P 1.01 r

5. c - A 0.0359 b
\)

0.0336 b i 0.61 \

Model does not apply. Estimation converges to neg-


ative rate constants.

Table V. Rate Parameters for Kinetic Model 2


high temperature low temperature REACTION TIME, mm

E, E, Figure 12. Comparison of kinetic model 2 with experimental


rate para- kcal/ kcall product concentrations in the low-temperature range.
meter A, min-' g-mol A, min-' gmol
CM 8.21 x 105 19.1 2.11 x lo6 19.5 will not fit the experimental data because it will not pro-
k MH 2.21 X lo8 29.1 7.34 x 105 19.4 duce the peak shown in Figure 10, which is characteristic
~ H M 2.57 X lo5 19.3 2.31 X lo6 19.5 of consecutive reactions. The activation energies for all
~ M N 6.41 X 10' 29.2 1.87 x 105 19.1 the reaction steps range from 17 to 29 kcal/g-mol. I t is
~ N M 1.54 X lo8 28.8 2.33 X lo6 19.2 found that all the reverse reactions are kinetically faster
ME 1.52 X lo8 28.1 1.16 X lo6 19.6 than the corresponding forward reactions, which agrees
EM 4.32 X lo5 17.2 1.25 x 107 20.1
with the high concentration for the large multifunctional
MA 1.18 X lo8 29.3 3.92 x 105 19.6
AM 4.74 x 105 20.0 1.94 x lo6 19.0 compounds obtained experimentally. It is also found that
the hydroxyl compounds dominate the other functional
reactions to describe coal liquefaction kinetics have been compounds, which is consistent with the abundance of
suggested by Squires (1978). Model 3 shown in Table IV hydroxyl groups found in the bituminous coal structure
356 Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981

Table VI. Combination of Coal-Derived Products Obtained from Chromatographic Separations t o Oils,
Asphaltenes, and Preasphaltenes
chromatographic fraction elution solvent combination rule solubility description classical definition
saturates hexanes
aromatics ( A ) toluene A + 0.5E oils (0) hexane solubles
ethers (E) chloroform
nitrogens (N) chloroform 0.5E + N + 0.5H asphaltenes (, A,) hexane insolubles
benzene solubles
hydroxyls (H) 90% THF/10% ethanol 0.5H + M preasphaltenes (P) benzene insolubles
THF solubles
large multifunctionals (M) uneluted

(Whitehurst el al., 1977; Wiser, 1978). 1.01


B. Low-Temperature Liquefaction. In the low-tem-
perature range (330-390 "C)it is again found that not
model 1 but model 2 (Table IV) fits the experimental
concentration-time profiles as shown in Figures 11 and 12.
It can also be seen in Figure 12 that the multifunctional 0.6-
compounds do not peak as was found in the high-tem-
perature range because in the low-temperature range the E
i
net rate of formation of the multifunctional compounds E 0.4-
is slower. Thus model 3, where the reaction products were 2
L
assumed to form directly from the coal, was tested. Y
Kinetic Model 3. In this kinetic scheme all reaction 8 0.2-
HYDROXYLS-,
steps are assumed to be first order and irreversible. The
multifunctional (MI
hvdroxvls ( H) I
10 20 30 40 50 60 70
REACTION TIME, mm

Figure 13. Comparison of kinetic model 3 with experimental


product concentrations in the low-temperature range.

1.01
aromatics (A) ethers (E)

kinetic model corresponding to the above kinetic scheme Cl.1 -


is given by the set of equations in matrix form in eq 16 and QUI
0 0 0 0 0
PREAWHALTENES 7
0 0
0 0

E].[ti.]
kCN 0 0 0 0 0
kCE 0 0 0 0 0
UNREACTEDCOAL
kCA 0 0 0 0 0

-kCN (l-xe) (16)


OA Ib io 3b 40 510 60 7
REACTION TIME, mm
-k C E
Figure 14. Comparison of kinetic model 4 with experimental
XA -kCA
product concentration in the high-temperature range.
eq 17, 18, and 19.
of SRC fractions based on their solubilities in various
[XC, X M , X H , X N , XE, X A l O = (17)
0, 0, 0, 0, 01 solvents are preasphaltenes, asphaltenes, and oils. Since
6 3 = ( k C M + k C H -t- k C N + ~ C +
E kCA) (18) LSC separations produced five fractions, some way of
combining these fractions is required to obtain the frac-
x, = 0.75 (Table 111) (19) tions based on solubility. Whitehurst el al. (1977) sug-
In Figure 13 model 3 is fitted to the experimental con- gested a formula for combining the nine chromatographic
centration-time data. A slightly better fit of the data, fractions obtained in their work into oils, asphaltenes, and
however, is obtained with model 2 as can be seen by the asphaltols (pyridine solubles, benzene insolubles) on the
lower value of the s u m of squares of the residuals, possibly basis of chemical functionality, C-H skeletal structures,
due to the greater number of parameters in model 2 as and molecular weight rangea that might be common among
compared to model 3. The fact that model 2 fits both the the two different product classifications. A similar com-
experimental concentration-time data for both the high bination rule, as given in Table VI, is used in this inves-
and low temperature range, whereas none of the other tigation.
kinetic models tested do, may be significant. The acti- Three simplified versions of the kinetic model recently
vation energy for the various steps in model 2 has been proposed by Cronauer et al. (1978) shown in Table I were
found to be approximately 19 kcal/g-mol. fitted to the concentration-time profdes obtained by these
Comparison of Kinetic Models Based on Solubility. rules of combining the LSC fractions. The tested models
As it has been discussed earlier, the classical definitions are given in Table IV. Model 4 fits the data reasonably
Ind. Eng. Chem. Process Des. Dev., Vol. 20, No. 2, 1981 357

Table VII. Comparison of Kinetic Parameters for the Formation of Oils, Asphaltenes, and Preasphaltenes
reactor batch CSTRa batch batchb
technique nonisothermal isothermal nonisothermal isothermal
(390-450"C) (400-470"C) (330-390 "C) (350-400"C)
coal Illinois No. 6 Belle Ayr Illinois No. 6 Kentucky No. 9
(bituminous) (subbituminous) (bituminous) (bituminous)
solvent tetralin hydrogenated hydrogenated tetralin tetralin
anthracene oil phenanthrene
reaction E A E A E A E A E A
coal + oils 19.8 1.71 X lo5 14.1 3.11 X lo3 28.9 2.10 X l o 8 19.1 1.58 X lo5 40.0 1.21 X 1 O ' O
coal ---f asphaltenes 18.9 1.44 X lo5 15.6 1.12 X 10" 8.6 9.63 X 10 19.4 3.61 X lo5 30.0 8.34 X lo6
coal ---f preasphaltenes 29.3 1.49 X l o 9 13.8 2.81 X lo3 4.3 4.94 19.2 1.20 X lo6 29.0 2.25 X lo6
preasphaltenes + as- 29.2 2.05 X 10' 12.8 9.66 X 10' 33.9 2.48 X lo9
phaltenes
asphaltenes + oils 29.0 1.14 X lo7 16.0 1.42 X l o 3 25.6 1.53 X lo7
a Cronauer et al. (1978). Shalabi et al. (1979). E , activation energy, kcal/g-mol;A, frequency factor, min-I.

1
0
PREMPHALTENES 7 0

0.2 L:ASQH~EST

'0 IO
0

20
UNREACTED COAL

OILS

30
7
40
,

50
9
60 70
UNREACTEDCOAL

REACTION TIME, mm REACTION TIME, mm

Figure 15. Comparison of kinetic model 4 with experimental Figure 16. Comparison of kinetic model 7 with experimental
product concentrations in the low-temperature range. product concentrations in the high-temperature range.

well in the low-temperature range but not as well in the


<
R
high temperature range as shown in Figures 14 and 15. /
cr1.- el01
Models 5 and 6 were found to fit the data in the high- 1
temperature range but yielded negative values for some P
of the rate constants. By analogy to model 2, model 7 with
reversible steps is proposed and is found to fit the data
: I\
very well in both high- and low-temperature ranges as
shown in Figures 16 and 17.
Unfortunately, no kinetic data exist in the literature for
Illinois No. 6 coal based on solubility measurements to
make a strictly valid comparison. It is of interest, however,
to make the best comparison that can be made at this time.
Thus, the kinetic parameters obtained by Cronauer et al.
(1978) and Shalabi et al. (1979) are compared in Table VI1
using other bituminous coals. Model 6 (Table IV) is es- I
sentially identical with the kinetic model proposed by REACTION TIME. mh
Shalabi et al. (1979) shown in Table I with the exception
of the reaction where oils are formed from the preas- Figure 17. Comparison of kinetic model 7 with experimental
phaltenes. Most of the kinetic parameters differ sub- product concentrations in the low-temperature range.
stantially from the values calculated here. This dis-
agreement may be caused by the differences in the coal- compounds, fall into two temperature regions is of par-
solvent systems or the method of combining the LSC ticular interest, but the reason for this effect is not known.
fractions to obtain the oils, asphaltenes and preasphaltenes. Model 3, with fewer parameters, fits the concentration-
time profiles in the low-temperature range almost as well
Conclusions as model 2, but model 2 has been found to fit both tem-
It has been found that nonisothermal experiments with perature ranges. A tentative method for combining the
Illinois No. 6 coal liquefied in tetralin as the donor solvent five chromatographic fractions is proposed to obtain the
will yield reliable kinetic data for evaluation of kinetic classical grouping of the compounds, preasphaltenes, as-
models. Liquid-solid chromatography has also been es- phaltenes, and oils, based on solubilities. In the high-
tablished as an adequate separation method to model temperature region model 7 (Table IV) was found to be
donor-solvent coal liquefaction kinetics. The fact that the the best, but all models gave reasonably good fits. Model
rate data, and in particular, that of the multifunctional 7 also fitted the data well in the low-temperature region.
358 Ind. Eng. Chem. Process Des. Dev. 1981, 20,358-360

Although the fit for model 7 is good, the calculated kinetic Falkum, E.; Glenn, R. A. F d I 9 5 2 , 31, 133.
parameters differ considerably from literature values. This Guh, J.; Tarrer, A.; Taybr, L., Jr.; Rather, J.; Green, S., Jr. Ind. ng. Chem.
Rocess Des. Dev. 1976, 75, 490.
may be due to the difference in coal-solvent systems or the Han, K. W.; Dixit, V. B.; Wen, C. Y. Ind. Eng. Chem. Process Des. Dev.
method of combining chromatographic fractions to obtain 1978, 17, 16.
Hili, 0. R.; Harlri, H.; Redd, R. 1.; Anderson, L. L. A&. Chem. Ser. 1966, No.
the solubility fractions. Whether kinetic models using LSC 55, 427.
separations are better than kinetic models using separa- Llebenberg, B. J.; Potgieter, H. G. J. Fuel1973, 52(4), 130.
tions based on solvent solubility remains to be answered, Longanbach, J. R.; Droege, J. W.; Chauhan, S. P. Repr. Div. Fuel Chem.,
Am. Chem. Soc. 1979, 24(2). 52.
but LSC is probably more suitable for on-line analysis in Maddocks, R. R.; Huffman, E. L. Prepr. Div. Pet. Chem., Am. # e m . Soc.
coal liquefaction plants. 1978, 23(4), 1296.
Marquardt, D.W. J . Soc. Ind. Appl. Meth. 1963, 17(2), 431.
Acknowledgment Meeter, D. A. Nonlinear Least Squares (&ushaus), Program No. COO17-00/
SOO17-00, Unhrerslty of Wisconsin Computing Center, Madison, WI, 1965.
The computational assistance of Ms. Joann Gherardi, Mitchell, T. 0.; Whltehurst, D. D. Paper No. 116, 71st Annual Meeting of the
AIChE, American InstlMe of Chemical Engineers, New York, Nov 1978.
Mr. David German, and Mr. Jay Fleishman, supported by Morlta, M.; Sato, S.; Hashlmoto, T. Repr. Dlv. Fuel Chem., Am. Chem.
Stevens Work Study Program, is appreciated. Soc. 1979, 24(2). 270.
Neavel, R. C. Fuel 1978, 55(7), 237.
Nomenclature Pelipetz, M. G.; Salmon, J. R.; Bayer, J.; Clark, E. L. Ind. Eng. Chem. 1955,
47(10), 2101.
A = frequency factor, min- Rather, J. W.; Tarrer, A. R.; Guh, J. A.; Johnson, D. R.; Neely, W. C. Llquid
Aii = stoichiometriccoefficients (positive if reaction j generates Fuels from Coal, Elllngton, R. T., Ed.; Academic Press: New York, 1977;
p 245.
class i, negative if it consumes class i, and zero if it does Schiller, J. E.;Mathiason, D. R. Anal. Chem. 1977, 49, 1225.
not involve class i) Schwelghardt, F. K.; Sharkey. A. 0. Prepr. 1976 Coal Chemistry Workshop,
E = activation energy, kcal/g-mol pp 72-85, Stanford Research Instltute, Menio Park, CA, Aug 1976.
kj = first-order rate constant for reaction j
km = first-order constant for reaction step; A B
R = gas constant, 1.987 cal/g-mol K
- Shalabi, M. A.; Baldwln, R. M.; Baln, R. L.; Gary, J. H.; Golden, J. 0. Ind.
Eng. Chem. Process Des. Dev. 1979, 18, 474.
Squires, A. M. Appl. Energy 1978. 4. 161.
Sternberg. H.; Raymond, R.; Schwelghardt, F. K. Pfepr. Div. Fuel Chem.,
Am. Chem. Soc. 1975, 20(4), 198.
R, = rate of reaction step j , min- Storch, H. H.; Fisher, C. H.; Eisner, A.; Clarke, L. Ind. Eng. Chem. 1940,
t = time, min 32(3), 346.
T = absolute temperature, K Struck, R. T.; Clark, W. E.; Dudt, P. J.; Rosenhoover, W. A.; Zlelke, C. W.;
Wc = mass of dry, ash-free feed coal &in, E. Ind. Eng. Chem. Process Des. Dev. 1989, 8 , 546.
W , = mass of dry, ash-free residue as THF insolubles Traeger, R. K. Paper No. 19% 86th National Meeting of the AIChE, American
Institute of Chemical Engineers, New York, Apr 1979,
x = coal conversion defined by eq 3 Weller. S.; Pelipetz, M. G.; Friedman, S. Ind. Eng. Chem. 1951, 43(7), 1575.
x , = limiting coal conversion Whitehurst, D. D.; Farcaslu, M.; Mitchell, T. 0.; Dickert, J. J., Jr. The Nature
Xi = mass fraction of class i, based on dry, ash-free feed coal and Orlgin of Asphakenes in Processed Coals. EPRI Final Report, AF-
460, Electric Power Research Institute, Palo Alto, CA, July 1977.
L i t e r a t u r e Cited Wiser, W. H. Chemisw of Coal Liquefaction: Status and Requirements.
Scientific Problems of Coal Utlllzation, Cooper, B. R., Ed.; p 232, DOE
Bard, Y. Nonllnear Parameter Estimation, Academic Press, New York, Symposium Series 46, CONF-770509, National Technlcal Information
1974. Services, Springfield. VA. 1978.
Brunson, R. J. Fuel 1979, 58(3), 203. Yergey. A. L.; Lampe. F. W.; Vestal, M. L.; Day, A. G.; Ferguseon, G. J.;
Chiba, T.; Sanada. Y. Nenryo Kyokai Sbi 1978, 57(612), 259. Johnson, W. H.; Snyderman, J. S.; Essenhigh, R. H.; Hudson, J. E. Id.
Cronauer. D. C.; Shah, Y. T.; Ruberto, R. G. Ind. Eng. Chem. Process Des. Eng. Chem. Process Des. Dev. 1974, 13, 233.
Dev. 1978, 17, 281. Yoshida, R.; Maekawa, Y.; Ishii, T.; Takeya, G. Fuel 1976, 55(10), 337.
Curran, G. P.; Struck, R. T.; Gorln, E. Ind. Eng. &em. Process Des. Dev.
1967, 6, 166.
Esch, R. Handbook of Applied Mathematics, Pearson, C. E., Ed.; Van Receioed for reuiew December 31, 1979
Norstrand-Reinhold: New York, 1974; p 964. Accepted November 12, 1980

Use of Solvent Extraction in the Production of the Ammonium Salt


of 3,5-Dinitro-l,2,4-triazole
Klen Y. Lee, Donald G. Ott, and Mary M. Stlneclpher
Los Alamos Scientific Laboratory, Los Alams, New Mexico, 87545

A process for the production of the ammonium salt of 3,5dinitro-l,2,4-triazole (ADNT) has been developed. The
process includes solvent extraction of 3,5dinitro-l,2,4-triazole (DNT) with water-insoluble, high-molecular-weight
amines and the preparation of the ammonium salt with anhydrous ammonia. The extraction of DNT is based on
the reaction of the free-base form of the amine with the organic acid to form the water-insoluble DNT-amine salt.
The production of ADNT is achiived by regeneration of the amine to its free-base form with a basic stripping agent
in a suitable diluent. The complete process requires 6 to 7 h on a laboratory bench scale and appears to be feasible
for large-scale production. The process could be applied to the production of any ammonium salt with properties
similar to those of ADNT.

Introduction Type 12/12B dropweight impact sensitivitiesof 59/80 cm,


The ammonium salt of 3,5-dinitro-1,2,4-triazoie(ADNT) and has a critical temperature of 222 C in 0.8 mm thick
is an explosive with several interesting properties. It is slab geometry.
soluble in water (28 g/100 mL), has a calculated PCJof 262 The parent compound, 3,5-dinitro-l,2,4-triazole (DNT),
kbar at its crystal density of 1.632 g/cm3, melts with loss was first prepared by U.S. Navy researchers (Burchfield
of ammonia at 170 O C , has ERL (Fedoroff et al., 1975) et al., 1963; Wiley et al., 1964), but a better method was
0196-4305/81/1120-0358$01.25/00 1981 American Chemical Society

Anda mungkin juga menyukai