Download as pdf or txt
Download as pdf or txt
You are on page 1of 9

Ind. Eng. Chem. Res.

1988, 27, 1775-1783 1775

A. V.; Ireland, H. R.; Callen, R. B.; Simpson, C. A. “Upgrading of B. C..; Petrakis, L. “Capillary Column Gas Chromatography with
Coal Liquids for Use as Power Generation Fuels”. Electric Power Sulphur- and Nitrogen-specific Hall Detectors for Determination
Research Institute PF-444, 1977. of Kinetics of Hydroprocessing Reactions of Individual Com-
Swansiger, J. T.; Best, . T.; Dickson, F. E. “Liquid Coal Compo- pounds in Coal-liquid Fractions”. Fuel 1983, 62, 1376-1378.
sitional Analysis by Mass Spectrometry”. Anal. Chem. 1974, 46,
730-734. Received for review November 9, 1987
Westerman, D. W. B.; Katti, S. S.; Vogelzang, M. W.; Li, C.-L.; Gates, Accepted May 23, 1988

Transport Model with Radiative Heat Transfer for Rapid Cellulose


Pyrolysis
Lee J. Curtis and Dennis J. Miller*
Department of Chemical Engineering, Michigan State University, East Lansing, Michigan 48824-1226

A mathematical model is presented which describes mass and energy transport during rapid pyrolysis
of fibrous cellulose particles. Radiative heat transfer within porous cellulose is modeled by using
the method of zones. The kinetic model for pyrolysis developed by Bradbury et al. is extended to
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

include secondary decomposition of condensible liquids (tars) formed. Solution of the governing
equations shows that both mass- and heat-transfer resistances influence product composition from
Downloaded via NATL TAIWAN UNIV on November 24, 2022 at 08:00:21 (UTC).

pyrolysis even for cellulose particles as small as 0.5 mm in diameter. Heating rate has little influence
on product composition, but increasing the total pressure results in a decreased condensible product
yield. Radiative heat transfer plays a minor role within the solid for the conditions simulated. The
model is useful for identifying critical parameters and conditions in pyrolysis and for predicting trends
in product yields.

I. Introduction even more pronounced as the high porosity char layer is


Since biomass materials are abundant, inexpensive, and formed during pyrolysis. Several models have been de-
renewable resources, their conversion to synthetic fuels and veloped based on the assumption that heat transfer limits
chemical feedstocks appears attractive. For these end uses, pyrolysis rate, and the models consider only heat-transfer
rapid pyrolysis is a potential route since it requires less resistances (Kung, 1972; Kansa et al., 1977). In extreme
energy input than steam gasification, requires no additional cases, a shrinking core mode of pyrolysis is assumed (Chan
reagents, and produces very little char. Pyrolysis does have
et al., 1985; Kanury, 1972).
one important drawback, however: it is difficult to se- Models including both mass- and heat-transport effects
lectively produce high yields of valuable products. have been developed primarily for predicting the rate of
Therefore, the focus of much present pyrolysis research wood pyrolysis and combustion. Fan et al. (1978) have
is on understanding pyrolysis chemistry and physics, both developed a general pyrolysis model which includes gen-
via experimentation and modeling, so that desired product eration, reaction, and diffusion of the gas-phase compo-
yields might be enhanced. The need for this work is clearly nents within the solid. However, convective flow and solid
outlined in the Kona workshop on thermochemical biom- structure changes are not included in their model. Antal
ass conversion (Antal, 1984). (1985) has set forth “general” transport equations for the
Major products from rapid pyrolysis of cellulose are pyrolysis of cellulose, and Kothari and Antal (1985) have
gases and a condensible fraction containing anhydrosugars presented numerical solutions to some simplified forms of
and organics. A number of reaction models have been the equations. In these model equations, however, two
developed from experimental data to describe product important phenomena were not explicitly included: the
formation pathways (Shafizadeh, 1968; Madorsky, 1964; effects of secondary decomposition reactions on overall
Mok and Antal, 1983a); kinetic constants have been de- product yield, and the contribution of radiation to overall
termined for several reaction steps (Bradbury et al., 1979; heat transfer within the cellulose particle. These phe-
Hajaligol et al., 1982; Broido, 1976). nomena may play an important role in determining overall
The influences of heat- and mass-transfer resistances pyrolysis behavior, especially at the high temperatures and
have been experimentally investigated, but unfortunately heat fluxes encountered in flash pyrolysis (Kansa et al.,
most studies only give qualitative information. The pri- 1977). Chan et al. (1985) included both secondary reac-
mary effect of mass-transfer resistance appears to be a
tions and a diffusive radiation term as a correction to the
decrease in condensible tar yield, as secondary decompo- thermal conductivity in their model for slow pyrolysis, but
sition to gases becomes more important with increasing to date, no detailed treatment of radiation has been con-
residence time in the reacting region. Mok and Antal ducted for cellulose.
(1983b) and Shafizadeh and Fu (1973) both report a de- This paper presents a model for flash pyrolysis which
crease in tar yield as pressure is increased, and Hajaligol includes a semirigorous treatment of radiation within a
et al. (1982) report a similar decrease as sample thickness “gray” cellulose solid and the secondary decomposition of
is increased. Others have shown similar trends (Scott and pyrolysis tars. The objectives of developing the model are
Piskorz, 1982; Shafizadeh et al., 1979). to investigate the importance of radiative energy transport
Heat-transfer resistances result both from endother- within the porous solid and the effects of transport re-
micity of the primary reactions and the low thermal con- sistances on overall product yield.
ductivity of cellulose. The thermal resistance becomes II. Model Development
Cellulose is modeled as a one-dimensional porous slab
* To whom all correspondence should be addressed. (half-thickness = L) of randomly oriented fibers located

0888-5885/88/2627-1775$01.50/0 © 1988 American Chemical Society


1776 Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988

Table I. Product Gas Composition Zone]— Zone i


component mole fraction
CO 0.454
co2 0.096
h2 0.094
h2o 0.100
CH4 0.170
C2H4 0.058
c2h6 0.020
total 1.0

between two identical radiating surfaces. During pyrolysis,


the volatile products formed within the solid flow outward
to the surfaces, while residual char remains as the solid
skeleton. The spatial dimension of the solid is taken to
be constant during pyrolysis; this is based on observations Zone i Zone j
in our laboratory (Kim et al., 1987) and by Mok and Antal (Ti) (Tj)
(1983b), in which it is reported that the char has structural µ----------2L----------µ
integrity and retains the shape of the original sample. Si S2
Thus, the solid merely becomes more porous as cellulose (Ts) (Ts)
is consumed. Also, it is assumed that the vapor and solid Figure 1. Radiation to zone i from other zones j and from surfaces
phases are in thermal equilibrium at each point within the Si and S2.
solid matrix.
Volatile pyrolysis products are grouped into two com- The method of zones, which is a discretized version of
the rigorous equations describing radiation within a gray
ponents: (1) noncondensible gases of a fixed composition
of CO, C02, H20, H2, and small amounts of hydrocarbons, gas, is thus used to describe the radiative flux within
cellulose. This technique, developed by Hottel and Sa-
and (2) a condensible “tar” fraction. The gas composition
is given in Table I; the tar fraction is assigned the prop- rofim (1967), treats the absorbing medium as a number
erties of levoglucosan. This binary vapor phase was chosen
of isothermal zones, each of which is interacting with all
to simplify calculations and to allow for secondary reaction
other zones. Although an approximation, the method is
accurate when r < 0.4 for each zone, a condition which
of tars during transport out of the solid.
arises in the present simulation if each zone is designated
II.l. Radiation in Cellulose. II.l.l. Diffuse Radi-
as the region between adjacent difference points.
ation. Radiation within porous solids has traditionally
been treated as an addition to the conduction term, where Figure 1 illustrates the basic idea of the method of zones.
the radiative flux is given in dimensionless form as It is seen that any zone i interacts with other zones j and
surfaces S1 and S2 by both direct and reflected radiation.
d_
16 3 dT The direct view factors, defined as the fraction of one zone
E R,D ”
(1) or surface directly seen by another zone or surface, are
dz 3t dz
functions of absorption coefficient, position, and the overall
This treatment of radiation as a diffusive process is an sample dimension 2L:
accurate approximation (Hottel and Sarofim, 1967) only (1) zone ¿-zone j
when the mean free photon path is much smaller than the
gjij 3( ) 3( ß6) + |3( ß )]
=
2[ 3( ß )
- -

(4a)
particle dimension (i.e., when radiation penetrates only a
short distance in the solid). The optical thickness, r, which (2) zone ¿-surface 1; zone ¿-surface 2
is a measure of the opaqueness of the solid, is defined as
SiEi =
2[£3( 1 )
-

3( 1ß)] (4b)
T12
= K dx (2) g¿S2
=
2[£3(t2S)
-

|3(r2a)] (4c)

where K is the absorption coefficient (inverse centimeters). (3) surface 1-surface 2


In this study, it is assumed that cellulose is a gray material SvS~2
=
2£3(t2L) (4d)
composed of fibers of diameter df and having porosity e. The optical thicknesses used in calculating direct view
If it is assumed that the fibers are opaque and the gas is
factors are given in Figure 2 (Hottel and Sarofim, 1967).
nonabsorbing, then the fraction of radiation which passes The quantity £3(t) is the third exponential integral.
through a thickness of one fiber diameter is equal to the
Knowing the direct view factors, the overall view factors,
porosity, e. The local absorption coefficient for cellulose defined as the fraction of energy emitted by one zone or
is therefore estimated as
surface that is absorbed by another zone or surface both
K (cm-1) = - —

(3) directly and by reflection, are given as


“f (1) zone ¿-zone j
In general, the diffusive approximation of radiative heat
(1 e.)ftS2 + (1
- -

transfer is valid only when r > 3 for any solid region of


GiGj gjgj + giS2 +
-

interest (Hottel and Sarofim, 1967). l-d-^m^)2


II. 1.2. Method of Zones. In the cellulose simulated
in this study, overall sample optical thickness ranges from (1 -

eJftSj + (1 -

e^ftSjSjS,
1.2 to 12. To solve the differential equations describing ftSi (5a)
heat and mass transfer within the solid, 20-40 individual
l-d-e.mjSj)2
difference points must be used, resulting in « 3 for each (2) zone ¿-surface 1
difference region. Therefore, the diffusive approximation
es(§iSl + (1 ea)S1S2giS2)
of radiative heat transfer cannot correctly be used, and a 0
OjG;


_

---
/^1
(5b)
s

more rigorous treatment must therefore be invoked. 1 (1 ee)2( SiS2 )2


- -
Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988 1777

to that obtained by numerous other investigators under


similar conditions (Hajaligol et al., 1982; Agarwal and
McCluskey, 1985; Scott and Piskorz, 1982; Shafizadeh et
al., 1979).
The heats of reaction are estimated from heats of for-
mation of product gases, heat of formation of cellulose
(Bains, 1974), and heat of formation (Bains, 1974), energy
of phase transition (Shafizadeh et al., 1970), and heat of
Zone-Surface Exchange vaporization (Prokorov et al., 1978) of levoglucosan. These
heats of reaction depend on product gas composition and
therefore are specific to these simulations.
*
From reaction models such as those proposed by Mok
¡
WOO j
and Antal (1983a,b), cellulose pyrolysis is seen to be an
—ß— extremely complicated chemical process. It is understood
Zone
Zone
that the kinetic model used in this study does not rigor-
ß ously represent pyrolysis chemistry; however, this model
Pi- 1
is the only one available (1) for which numerical values of
Zone Zone Exchange -

rate constants are given and (2) which is structured to


Figure 2. Definitions of optical thicknesses used in method of zones. explicitly include the secondary reactions that largely
determine product distribution. Most pyrolysis models
Table II. Pyrolysis Reaction Model report kinetic constants for solid cellulose decomposition
only and are thus unsuitable for these simulations. For
AHrxn. £a, these reasons, the model of Bradbury et al. (1979), ex-
reaction cal/g cal/mol tended to include secondary decomposition, is best suited
(1) cellulose -* 0.0 2.83 X 1019 58000 for this work.
active cellulose
II.3. Model Equations. II.3.1. Material Balances.
(2) active cellulose -*
tars +350.0 3.17 X 10u 47300
(3) active cellulose -* -72.0 1.32 X 1010 36600 In the initial model development, material balances were
0.65 gases + 0.35 char written to include both diffusion and pressure-driven flow
(4) tars — gases +8.0 3.18 X 1016 36600 of gas-phase components within the solid particle. Early
simulations, however, showed that terms for diffusion and
The zone ¿-surface 2 view factor, S2G¿, is found by inter- pressure-driven flow in the material balances were several
changing subscripts 1 and 2 in the last equation. orders of magnitude smaller than the convection and re-
If the terms for emission and absorption from all zones action terms; these terms were thus dropped from the
and surfaces are combined, the term in the energy balance model without loss of accuracy. Also, early simulations
describing the net radiative heat-transfer rate out of zone showed that gas-phase residence times within the solid
i is given in dimensionless, discretized form as were small compared to the time necessary for temperature
=
4ATiaecTi4
-

SiG¡ ß » ~ profiles to change significantly, so a pseudo-steady-state


Er,,· _

assumption was made for gas component fluxes. There-


X,GiGjaet.Tji (6) fore, the material balances written for the two gas-phase
-

S2G¿ffe8T84
components, tars and gases (product gases, water, and
Further details of the method of zones are given in Hottel inerts), contain terms only for convective flow and reaction.
and Sarofim (1967). The equations are given in dimensionless form.
II.2. Pyrolysis Reaction Model. Bradbury et al.
tars
(1979) developed a simple model for low-temperature py-
rolysis of cellulose by assuming the volatile products to be ó¿VL/dz =
a2R2
-

R4 (7)
a mixture of tars (levoglucosan) and noncondensible gases.
From experiment they obtained rate constants and acti- gases
vation energies which are assumed to apply at the higher
dÑG/dz =
0.65a3Gfi3 + a4R4 (8)
temperatures of this study. However, the model of
Bradbury et al. does not consider secondary tar decom- The reaction rate terms are given as
position, which is relatively unimportant at low tempera-
tures but becomes important in high-temperature pyro- Rl =
<t>iWc exp[-EA|1(l/T- 1)] (9a)
lysis. Bradbury et al.’s kinetic model is therefore extended Ri =
c exp[-ÉA,,(l/T
-

1)] i =
2, 3 (9b)
for this study to include the secondary decomposition of
tars. This secondary decomposition is assumed to occur R4 -

4>4Xl exp[-£A,4(l/T
-

1)] (9c)
only in the gas phase within the porous solid matrix. The and the concentration (mole fraction in dimensionless
parameters used in the kinetic model are described in
Table II. form) of tars is
Reaction kinetics for secondary vapor-phase tar decom- *L
position (reaction 4) have not been reported in the liter- x L
-

(10)
ature. For the model, therefore, the value of this rate ÑL + ÑG
constant has been adjusted such that the maximum tar Overall yields (grams/gram of cellulose) are calculated by
yield for simulations at 1.01 X 105 Pa pressure and a integrating surface fluxes over time:
heating rate of 250 °C/s is between 55% and 60% by 1 ^
c*
weight. The rate constant was adjusted by assuming the
-

^
-

I ^L,lsurface (11a)
activation energy to be equal to that of reaction 3 and a 2
vo

changing the preexponential to obtain the desired tar yield. 1


This value of tar yield at atmospheric pressure has been
_

dt (lib)
_

=
YG J ÑG|surface
obtained in our laboratory (Kim et al., 1987) and is similar OizGJ 0
1778 Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988

Table III. Dimensionless Groups and Variables Table IV. Properties and Parameters
Groups base case
Wc0 ki0L2PcCpc Wcfi(-AHi) values range
°2 0; =---i
-

i
'—
i =
1,4 ; = ......-

T =-2’3 initial temp, K 298


PoMl «e0 Pc*-'pc1 0
total pressure, Pa 1.01 x 105 1.01 X 103-1.01 X 106
0.02 0.005-0.04
W=,o 0 0(- 4) ]_ half-thickness, cm
250 10-500
“3G -

L 4 =
heating rate, °C/s
PoMG «e0 PcCpCTo surface emissivity (e„) 0.75 0-1.0
cellulose emissivity 0.95
M, PoCph EKi initial porosity 0.70
II :”
<? V'L
-

®A,i -(In t/df) -(0.01 In


is r
Pc^pc RT0 absorption coeff, cm"1 e)/df- (-(In e/df))
vL2pcCpo PoCpG Tr where £Rl· is given by eq 6.
~
-

r Tr =:
All dimensionless groups and variables are identified in
AoT0 Pc*^pc To
Table III.
Variables II.3.3. Boundary Conditions. Four spatial and four
EL2 wnr Wr initial boundary conditions are required for the solution
E =-
WT

77;— of the material and energy balances. Initial conditions for


0 W=,o Weft
pyrolysis are specified as uniform temperature and solid
K Wc nl composition:
k° =
r«e0 wc =
Al
Wc,0 nl + ng at t = 0

NiLpcCpc T f = 1 Wc
= 1 WBC
= 0 Wt = 0 (16)
N¡ =
-r- i =
L,G T = —

P 0Ae0 T0
The spatial conditions at the centerline are specified by
keot symmetry:
i -

'E-i II

PcCPA2 at z = 0

The solid-phase balances give the local concentration df/dz =


0 ÑL = 0 ÑG = 0 (17)
of solid components as a function of time, thus allowing At the solid surface, the boundary condition on the energy
the local solid porosity to be calculated. balance is specified by the pyrolysis environment. For
cellulose these simulations, the surface of the solid is taken to be
in thermal equilibrium with the radiating surfaces, S1 and
dWJdt =
-Rx (12a)
S2, which increase with temperature at a linear rate during
activated cellulose pyrolysis. The boundary condition is therefore
dWBC/dt = -

R2
-

R¡ (12b) at z = 1

char f =
fg = 1 + t (18)
dWt/dt =
-a3tR3 (12c) II.4. Physical Properties and Parameters. The
porosity physical properties and simulation parameters used are
listed in Table IV. The table includes both a “base case”
e = 1 (1 e0)OVc + Wac + Wt) (13) value of the parameter and the range of values of each
- -

II.3.2. Energy Balance. The energy balance in the parameter used in simulations. Several of the parameters
solid includes convective, conductive, and radiative heat are explained in the following sections.
transfer and heats of reaction terms. The equation is given 11.4.1. Effective Thermal Conductivity. Reliable
in dimensionless, continuous form: models of effective thermal conductivity as a function of
void fraction are not generally available in the literature,
and insufficient data are available for cellulose to generate
|[(f-TR)(l-0]
tit
+
¿[(WVL
oz
+
such a model. Therefore, a simple approximation of
-

fR)] + £ =
(14) two-phase thermal conductivity is used to account for
nel changes in this parameter as cellulose is consumed and
The quantity £ represents the dimensionless rate of energy porosity increases. The model used is attributed to
Woodside and Messmer (1961); it is easily applied, has the
removal per unit volume due to conduction and radiation. correct limits at extremes of porosity, and agrees with
If only conduction is included,
experimental values obtained by Woodside and Messmer:

£-£=-!(-'-·!) (i5>) K -

kc{kG/kc)‘
The values of thermal conductivity for cellulose (kc) and
(19)

where kB is the effective thermal conductivity, discussed


in a later section. When radiation is accounted for, £ is gas (kG) phases were taken from Perry and Chilton (1973):
the sum of both radiation and conduction terms. For kc = 5.8 X 10~4 cal/(cnvs-K); kG = 5.9 X -5?’0·94 cal/
diffuse radiation, (cm-s-K).
11.4.2. Specific Heats. Specific heats are necessary to
£ =
£c £r,d (15b) calculate the convective heat-transfer term and the energy
where £R-D is given by eq 1. accumulation term. The specific heat of cellulose is taken
to be constant and equal to that of paper (Perry and
For the method of zones, the equation is transformed
into finite difference form, and Chilton, 1973). The gas-phase specific heat is taken as the
molar-weighted average of all components in the gas phase
£¡ =
£C;; £Ri,· (15c) (Table I).
Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988 1779

gas

Cp,G (cal/(mol-K)) = 5.657 + 0.212T

tars

CPiL (cal/Cmol-K)) =
-8.27 + 44.7f-2.6f2
cellulose

Cp,c (cal/(g*K)) =
0.32

II.4.3. Properties of Cellulose. The following prop-


erties of cellulose are used in the model. The fiber diam-
eter, df, is taken to be 0.0025 cm based on microscopic
examination of several typical papers. The intrinsic den- Figure 3. Cellulose conversion profiles at several surface tempera-
tures during pyrolysis.
sity of cellulose is taken as 1.4 g/cm3 (Perry and Chilton,
1973). The initial porosity is taken as 0.7, based on our
measurements of apparent density of the filter paper. The
emissivity of cellulose/activated cellulose/char is taken to
be 0.95 (Perry and Chilton, 1973).
II.5. Method of Solution. The method of lines was
chosen as the solution method for the simulations. The
method was applied by first establishing a set oiM ( =
20-40 for this application) discrete axial coordinates or
“lines” and then rewriting the four partial differential
equations using centered finite difference approximations
for derivatives in the discretized dimension, while the re-
maining dimension, time, remained continuous. Thus, the
problem was transformed from four partial differential
equations in distance and time to 4M ordinary conditions
in time only. The ordinary differential equations were then Figure 4. Solid temperature profiles at several surface temperatures
solved by using the IMSL subroutine dgear. The ordinary during pyrolysis.
differential equations for gas-phase component fluxes were
integrated in the axial direction by using the improved
Euler method. The surface fluxes of these gas-phase
components were integrated in time by using DGEAR to
evaluate overall product yields.
An interface similar to PDEONE, developed by Sincovec
and Madsen (1975), was written for this problem to sim-
plify the numerical programming task. The boundary
condition approximations used in the interface were de-
rived by using the false boundary method so that finite
difference approximations were used for all discrete ap-
proximations. This removed a slight inconsistency present
in Sincovec and Madsen’s PDEONE.
All simulations were carried out in double precision with
an error tolerance of 10""6. Mass balances checked for Figure 5. Vapor-phase concentration profiles in the solid at several
surface temperatures during pyrolysis.
closure to at least 0.01% by weight above 1% cellulose
conversion. Several simulations were run at twice the As stated, the terms in the material balances for diffu-
number of lines as in the initial simulation; the largest
sion, accumulation, and pressure-driven flow were all
overall difference in temperature was less than 2% for the several orders of magnitude smaller than the convective
two simulations, indicating good convergence of the solu- term and were thus not included. Results for simulations
tion. which ascertain the importance of the term for radiative
heat transfer are reported in this section.
III. Simulation Results 111.1. Detailed Results from One Simulation. Fig-
Simulations have been conducted to fulfill three objec- ures 3-5 give the results obtained for a typical pyrolysis
tives: (1) estimate the magnitude of mass- and heat- simulation. This simulation was conducted for base case
transfer resistances and resulting gradients during pyro- parameters except that L = 0.01 cm. The results show the
lysis, (2) determine the effect of varying process or physical temperature, cellulose conversion, and tar concentration
parameters on overall product distribution from pyrolysis, profiles within the cellulose particle at several times during
and (3) ascertain the relative importance of the various pyrolysis. It is seen that conversion begins at about 700
physical and chemical phenomena occurring in pyrolysis. K surface temperature and is essentially complete by the
The parameters varied in the study of product distri- time surface temperature reaches 900 K. From the cel-
bution are listed in Table IV; both the “base case” value lulose conversion and tar concentration profiles, the overall
and overall range of values of each parameter are given. conversion of cellulose and yields of tar, char, and gases
The base case values represent a likely physical situation. as afunction of time are obtained.
Some simulations include extrapolated values at higher Parametric Studies. Parametric studies have
111.2.
conversions; computation time limits and behavior of the been conducted for three important process variables:
solution prevented completion of these simulations. sample heating rate, total pressure, and sample thickness.
1780 Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988

Figure 6. Overall fraction of cellulose remaining versus time for Figure 10. Cumulative tar yield versus time for pyrolysis at several
pyrolysis at several heating rates. values of absolute pressure.

Figure 7. Cumulative tar yield versus time for pyrolysis at several Figure 11. Cumulative gas yield versus time for pyrolysis at several
heating rates. values of absolute pressure.

Dimensionless Distance From Surface


Figure 8. Cumulative gas yield versus time for pyrolysis at several Figure 12. Cellulose conversion profiles at 357c overall conversion
heating rates. for several sample thicknesses.

CT
C Table V. Overall Product Yields from Parametric Studies0
tar yield,
E char yield, g/g gas yield,
or simulation g/g cellulose cellulose g/g cellulose
o
o pressure
1.01 X 103 Pa 0.023 0.892 0.085
1.01 X 104 Pa 0.025 0.690 0.285
o
1.01 X 106 Pa 0.028 0.330 0.642
1.01 X 106 Pa 0.030 0.070 0.900
heating rate
10 °C/s 0.035 0.275 0.690
25 °C/s 0.032 0.308 0.660
1.5 2.0 2.5 3.0 75 °C/s 0.030 0.320 0.650
Time (seconds) 250 °C/s 0.028 0.330 0.642
500 °C/s 0.027 0.337 0.636
Figure 9. Overall fraction of cellulose remaining versus time for sample thickness
pyrolysis at several values of absolute pressure. gradientless 0.017 0.565 0.418
0.005 cm 0.018 0.552 0.430
The results of these simulations are given in Figures 6-12; 0.01 cm 0.019 0.435 0.546
the final product yields are summarized in Table V. 0.02 cm 0.028 0.330 0.642
0.04 cm 0.034 0.162 0.804
The results of varying sample heating rate are given in
Figures 6-8, in which the fraction of cellulose remaining 0
All parameters at base case values except as specified.
Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988 1781

and product yields are plotted as a function of time. It contribution of radiative heat transfer to overall energy
is seen in Figures 6-8 that changing the sample heating transport was still minor. There results at most a 5%
rate has only a slight effect on overall pyrolysis product change in overall temperature gradient within the solid
distribution; lowering sample heating rate therefore only upon consideration of radiation in the energy balance.
lengthens total pyrolysis time. Second, the value of the absorption coefficient, K, of the
The effects of varying total pyrolysis pressure are given solid was decreased by a factor of 10 and by a factor of 100.
if Figures 9-11. Increasing process pressure, in contrast This decreases the optical thickness of the solid and allows
to heating rate, does not strongly affect the cellulose better penetration of radiation. This also had a minor
conversion rate (Figure 9) but greatly reduces tar yield effect on the simulation results.
(Figure 10) and increases gas yield (Figure 11). This is a Finally, varying the surface emissivity, eB, over its full
direct result of an increased rate of secondary reaction at range of 0-1.0 also resulted in only a slight change (<1%)
higher pressure, which is a consequence of higher tar in simulation results.
concentrations and longer residence times within the
cellulose particle. IV. Discussion
The effects of increasing sample thickness are illustrated The simulation results illustrate that transport resist-
ances have a strong influence on product distribution from
in Figure 12 and Table V. Increasing sample thickness
results in a very significant increase in temperature and pyrolysis; the presence of transport gradients results in
conversion gradients within the solid sample, from rela- enhanced secondary decomposition and thus lower overall
tively mild gradients for L = 0.005 cm to nearly a shell- yields of volatile organic products formed in pyrolysis. The
progressive mode of pyrolysis at L = 0.04 cm. This is organic products formed are potentially valuable as fuel
shown by the cellulose conversion profiles in Figure 12 at and chemical feedstocks; the model therefore provides the
35% overall cellulose conversion. The final product yield important capability of at least predicting trends in
is affected by sample thickness; tar yield is decreased for product yields under actual process conditions. The model
thicker samples. The maximum tar yield is obtained in is able to predict this behavior for rapid pyrolysis via in-
the absence of temperature gradients in the solid; the tar clusion of a secondary decomposition reaction.
The results show that even for cellulose particles as small
yield from the “gradientless” simulation given in Table V
as 0.3 mm in diameter =
6L) there are significant
is the maximum tar yield and was obtained by fixing the (dp
temperature equal everywhere in the solid. This simula- gradients within the sample which affect product yields.
tion was also used to fix the rate constant for secondary In fact, all simulations indicate the presence of significant
tar decomposition. gradients which affect product yields with the exception
of the simulations at very low pressures (P = 1.01 X 103
The surface energy flux has also been calculated for
several sample thicknesses. The energy flux is relatively Pa) in which the tar yield approaches the upper limit
dictated by the reaction kinetics. Even though tempera-
constant once pyrolysis has begun and varies little for
ture gradients exist during pyrolysis that are similar to
different sample thicknesses. The maximum value of those at higher pressures, gas densities and residence times
surface energy flux at a heating rate of 250 °C/s is 2 X 10s
in the high-temperature region are small enough that
W/m2. secondary reactions are negligible.
III.3. Importance of Radiative Heat Transfer in
Secondary tar decomposition becomes more important
Pyrolysis. Three difference models have been used in in the latter stages of pyrolysis, when the surface tem-
simulations to ascertain the role of radiation in pyrolysis.
perature and temperatures of reacted regions are quite
These models differ only in the term describing radiative
high. In Figure 5, it is seen that the tar concentration at
heat transfer in the energy balance equation. In the first
higher surface temperatures decreases rapidly near the
model, radiative heat transfer is neglected (eq 15a); in the surface, indicating almost complete secondary decompo-
second, radiation is approximated as a diffusive process sition of tar. The overall tar yield, which is the integrated
(eq 15b); in the third, radiation is treated rigorously by surface flux, thus increases rapidly in the initial stages of
using the method of zones (eq 15c). pyrolysis but reaches its final value long before pyrolysis
For the base case values of parameters used in the sim- is complete. This is illustrated clearly in the tar yield
ulations, the results from the three models differ very little. curves in Figure 10.
This indicates that at the temperatures for which pyrolysis The results point out the importance of including sec-
occurs radiative heat transfer makes a minor contribution
ondary decomposition reactions as an integral part of any
to overall energy transport. The temperature profiles kinetic reaction model for pyrolysis. These secondary
within the cellulose particle, the overall rate of cellulose reactions occur within the porous solid and cannot be
conversion, and the product distribution differ by no more neglected even for the thinnest samples without grossly
than 2% at any instant for the three cases. Because the misrepresenting the pyrolysis chemistry. The overall
overall pyrolysis is highly endothermic, the temperatures product yields from the rapid pyrolysis show that char
within cellulose remain relatively low, thus rendering the yields via reaction 2 are small (<5%) in all simulations,
radiation term with T4 dependence small. indicating that almost all cellulose decomposition is via
To illustrate the role of radiation in pyrolysis, three formation of tars (reaction 3). If secondary reaction was
parametric studies were carried out in which parameter neglected, tar yields would be over 95% in every case. This
values were changes to enhance the contribution of radi- is not the case in any simulations except those at very low
ation to heat transfer. First, the same physical system with pressures (Figure 10); thus, it must be concluded that
different chemical reaction kinetics was investigated by secondary decomposition always plays a role in rapid py-
increasing the activation energies for the four reactions in rolysis. In this simulation, only vapor-phase decomposition
the kinetic model by 50% to make the reactions occur at of tars has been considered: it has been postulated that
higher temperatures. The intention of using these acti- tars formed as liquids also can decompose before vapori-
vation energies was to magnify the importance of the ra- zation occurs. It may well be that including only vapor-
diative heat-transfer term, which dominates at higher phase decomposition underestimates the importance of
temperatures. Even with the resulting higher pyrolysis secondary reactions; thus, the results presented may be
temperatures (1000-2000 K instead of 700-900 K), the conservative.
1782 Ind. Eng. Chem. Res., Vol. 27, No. 10, 1988

For the physical system modeled, it is clear that radia- product yields to process parameters, but because of
tion within the cellulose sample does not contribute sig- variability in cellulosic materials and experimental con-
nificantly to heat transfer during pyrolysis. This is a result ditions, exact predictions cannot be made. In order to
primarily of the fact that temperatures involved in cellulose better predict pyrolysis behavior, more accurate estimates
pyrolysis are low enough that the radiative T4 term is small of physical properties (thermal conductivity, specific heats,
and thus conductive heat transfer dominates. Because etc.) and pyrolysis chemical kinetics, especially describing
rapid pyrolysis is very endothermic, the solid is essentially secondary reactions, are absolutely essential. In addition,
“cooled” as pyrolysis progresses: the temperature in the the model needs to be expanded to more rigorously de-
reacting region increases only slightly as cellulose is con- scribe the complete physical system in which pyrolysis is
sumed (Figure 4) once pyrolysis has started. conducted. This is necessary because the primary tars
It is possible that for different physical systems radiative formed are thermally fragile and are thus very prone to
heat transfer would make a larger contribution in the subsequent decomposition even after leaving the imme-
energy balance. For instance, if instead of being exposed diate environment of the solid being pyrolyzed.
to a surface heated linearly with time the cellulose was
initially exposed to a surface already at a very high tem- V. Conclusions
perature, then radiative heat transfer would be expected The simulation of rapid pyrolysis of cellulose leads to
to play a larger role. Unfortunately, the extreme gradients several important conclusions:
generated in such a case require a number of difference 1. Diffusional mass transfer and total pressure gradients
points and thus computation times beyond those available. are found to be relatively unimportant in rapid pyrolysis
For such simulations, a finite element approach would be of thin samples.
better suited. 2. Significant temperature, solid conversion, and gas
Quantitative comparison of simulation results with ex- composition gradients exist during rapid pyrolysis even for
perimental data is somewhat difficult in general, as vari- very thin (dp =
0.3 mm) samples.
ations in cellulosic starting materials alter both chemical 3. Secondary reactions are an integral part of pyrolysis
reaction pathways and physical properties pertinent to reaction chemistry and substantially reduce volatile organic
pyrolysis. Moreover, there have not been any detailed product yields from rapid pyrolysis. For the range of
experimental studies of rapid pyrolysis in which intra- conditions studied, the only exception is found at low
particle gradients are directly measured. Therefore, com- pressure, where tar yields approach the upper limit dic-
parisons of model predictions with experimental data are tated by the kinetic model.
limited to those of overall product distribution as a 4. Radiative heat transfer plays a minor role for the
function of pyrolysis parameters. Again, because cellulose physical systems simulated, even for extreme values of
properties such as impurity content so strongly affect physical parameters important in radiative heat transfer.
5. The model predicts the proper trends in product
product distribution and because such a variety of starting
materials are used in pyrolysis studies, quantitative com- yields with important process parameters. However,
parison is difficult. Experiments reported by Hajaligol et quantitative prediction of pyrolysis behavior requires
al. (1982) using a heated screen reactor are closest to the better estimates of physical properties and more rigorous
physical system modeled and thus will be examined. modeling of the particle experimental system or process
The increase in volatile tar yield as pressure is decreased under study.
is reported by several investigators (Agarwal and
Nomenclature
McCluskey, 1985; Shafizadeh et al., 1979; Shafizadeh and
Fu, 1973; Hajaligol et al., 1982; Mok and Antal, 1983a,b). CpL
=
heat capacity of tar, cal/(mol-K)
Agarwal and McCluskey (1985) report tar yields of 40% Cpc
=
heat capacity of cellulose, cal/(g-K)
at 1 atm and 70% at 0.001 atm using newsprint as a CpG
=
average heat capacity of gas components, cal/(mol-K)
starting material; this is in good agreement with our model df cellulose fiber diameter, cm
=

if the influence of impurities on tar yield is accounted for. es


=
emissivity of surfaces S1 and S2
Hajaligol et al. (1982) report that tar yield decreases as ec
=
emissivity of cellulose
E = net energy removal per unit volume from conduction
pressure is increased; the value of tar yield at 5 psig (59%
for 750 °C peak temperature) is in good agreement with and/or radiation, cal/(cm3-s)
the value obtained (55%) in our simulations. £A i = activation energy for reaction i, i = 1, 4, cal/(mol-K)
AHi = heat of reaction for reaction i, i = 2, 4, cal/g
The effect of heating rate on product distribution is very K = absorption coefficient, cm™1
small in our model; heating rate primarily affects the ke
=
effective thermal conductivity of cellulose matrix, cal/
magnitude of temperature gradients and total pyrolysis (cm-s-K)
time. This does not agree well with the results of Hajaligol kc = cellulose thermal conductivity, cal/(cm-s-K)
et al. who report a mild increase in tar yield as heating rate kG
=
gas-phase thermal conductivity, cal/(cnvs-K)
is decreased. This discrepancy is difficult to reconcile; they k¡ = rate constant for reaction i, i = 1, 4, 1/s
suggest, however, that the heated screen enhances tar L = half-thickness of cellulose sample, cm
decomposition at higher heating rates and this leads to the Ml = molecular weight of tar, =162 g/mol
observed trend. Mg = average molecular weight of gas component, =27 g/mol
Increasing sample thickness above 0.01 cm strongly Nl = molar tar flux, mol/(cm2-s)
decreases tar yield according to the model; this effect is Ng molar gas flux, mol/(cm2-s)
=

also reported by Hajaligol et al. except that the critical Ri = dimensionless reaction rates, i = 1, 4
R = ideal gas constant, =1.987 cal/(mol-K)
thickness appears to be around 0.02 cm in their experi-
Ts =
surface temperature, K
ments. Again, taking into account the variation in ex- T = temperature, K
perimental conditions, this agreement (within a factor of TR =
reference temperature, =298 K
2) is quite good. t = time, s
In summary, quantitative prediction of pyrolysis be- VFC
=
apparent solid concentration of cellulose, g/cm3
havior using the model is not yet reliable. The model Wac
=
apparent solid concentration of activated cellulose,
predicts the proper trends and relative sensitivities of g/cm3
Ind. Eng. Chem. Res. 1988, 27, 1783-1788 1783

Wr =
apparent solid concentration of char, g/cm3 Bains, M. S. Carbohydr. Res. 1974, 34, 169.
x = coordinate in solid, cm Bradbury, A. G. W.; Sakai, Y.; Shafizadeh, F. J. Appl. Polym. Sci.
Y¡ =
yield of product i, g/g of cellulose 1979, 23, 3271.
Broido, A. Symp. Therm. Uses and Properties Carbohydrates and
Greek Symbols Lignins; Shafizadeh, F., Sarkanen, K. V., Tillman, M., Eds.; Wiley:
= stoichiometric coefficient of reaction 2 New York, 1976; Vol. 19.
2
=
stoichiometric coefficient for gas formation from reaction Chan, W. R.; Kelbon, M.; Krieger, B. B. Fuel 1985, 64, 1505.
«3G
Fan, L. S.; Fan, L. T.; Tojo, K.; Walawender, W. P. Can. J. Chem.
3
Eng. 1978, 56, 603.
a3r = stoichiometric coefficient for char formation from re- Hajaligol, M. R.; Howard, J. B.; Longwell, J. P.; Peters, W. A. Ind.
action 3, =0.35 Eng. Chem. Process Des. Dev. 1982, 21, 457.
e = void fraction Hottel, H. C.; Sarofim, A. F. Radiative Transfer; McGraw-Hil: New
=
heating rate, °C/s York, 1967.
Po
= initial gas density, mol/cm3 Kansa, E. J.; Perlee, . E.; Chaiken, R. F. Combust. Flame 1977,29,
= cellulose density, =1.4 g/cm3 311.
pc
=
Stefan-Boltzmann constant, =1.36 X 10~12 cal/(s-cm2-K4) Kanury, A. M. Combust. Flame 1972, 18, 75.
Kim, N. W.; Trautz, J. A.; Miller, D. J. IGT Energy from Biomass
=
optical thickness and Wastes XI, Orlando, FL, March 16-20, 1987; Paper #40.
&(*) =
Si“ e-7f3 dt Kothari, V.; Antal, M. J., Jr. Fuel 1985, 64, 1487.
Subscripts Kung, H. C. Combust. Flame 1972, 18, 185.
Madorsky, S. L. Thermal Degradation of Organic Polymers; Inter-
0 = initial condition science: New York, 1964.
ac = activated cellulose Mok, W. S.-L.; Antal, M. J., Jr. Thermochem. Acta 1983a, 68,155.
L =
tars Mok, W. S.-L.; Antal, M. J., Jr. Thermochem. Acta 1983b, 68, 165.
c = cellulose Perry, R. H., Chilton, C. H. Eds. Chemical Engineer’s Handbook,
f = fiber 5th ed.; McGraw-Hill: New York, 1973.
G =
gases Prokorov, A. B.; Cergeva, B. H.; Fainber, E. E.; Kalnins, A. R. Chem.
s = surface Wood Mat. 1978, 3, 18.
r = char Scott, D. S.; Piskorz, J. Can. J. Chem. Eng. 1982, 60, 666.
Shafizadeh, F. J. Appl. Polymer Sci.: Appl. Polymer Symp. 1983,
Registry No. CO, 630-08-0; C02,124-38-9; H2,1333-74-0; CH4, 37, 723; Adv. Carbohydr. Chem. 1968, 23, 419.
74-82-8; C2H4, 74-85-1; C2H6, 74-84-0; cellulose, 9004-34-6. Shafizadeh, F.; Fu, Y. L. Carbohydr. Res. 1973, 29, 133.
Sincovec, R. F.; Madsen, N. K. ACM Trans. Math. Software 1975,
1(3), 232.
Literature Cited Smith, J. M. Chemical Engineering Kinetics, 3rd ed.; McGraw-Hill:
New York, 1981.
Agarwal, R. K.; McCluskey, R. J. Fuel 1985, 64, 1502. Woodside, W.; Messmer, J. H., Jr. J. Appl. Phys. 1961, 32, 1688.
Antal, M. J., Jr. “Biomass Engineering: Thermochemical Conver-
sation Research Needs”. Report of the Kona Workshop, June Received for review November 3, 1987
1984; Hawaii Natural Energy Institute. Revised manuscript received May 24, 1988
Antal, M. J., Jr. Fuel 1985, 64, 1483. Accepted June 27, 1988

Catalytic Hydrotreatment of Vacuum Pyrolysis Oils from Wood


Jean Gagnon and Serge Kaliaguine*
Department of Chemical Engineering, Université Laval, Ste-Foy, Quebec G1K 7P4, Canada

The effects of a mild hydrogenating pretreatment in the presence of a Ru catalyst over the hy-
drodeoxygenation (HDO) of wood-derived vacuum pyrolysis oil have been investigated. The optimal
conditions found for this pretreatment are a temperature as low as 80 °C and a pressure of 600 psig.
The results indicate that the yield of HDO is correlated with the average molecular weight of the
products determined from gel permeation chromatograms. This suggests that polymerization/
hydrogenolysis reactions are occurring in HDO conditions. It is also found that some hydrogenolysis
is produced during the hydrogenating pretreatment.

Upgrading of biomass pyrolytic oils involves essentially baum et al. (1979) and Krishnamurthy et al. (1981).
the removal of oxygen from a variety of organic com- Teman and Brown (1982) performed HDO of coal distillate
pounds, and this can be performed by hydrodeoxygenation liquids, with high oxygen contents. High rates were ob-
(HDO). This type of process has been the object of much served although only the smaller oxygenated compounds
less work than the related reactions of hydrodesulfurization (for example, phenols) were reacted. Bredenberg et al.
(HDS) and hydrodenitrogenation (HDN). Early works on (1982) hydrotreated lignin degradation liquids. They re-
HDO were reviewed by Weisser and Landa (1973). More port that only phenols and cresols are completely con-
recently, Furimsky (1978,1979,1983a-c) studied HDO of verted, whereas higher molecular weight compounds are
various feedstocks and reported kinetics and mechanisms almost not reacting. Chum and her group (Johnson et al.,
for the HDO of furan and tetrahydrofuran. The deoxy- 1986) investigated recently the conversion of lignin into
genation of dibenzofuran was examined by Badilla-Ohl- phenols and hydrocarbons by mild HDO and the hydro-
treatment of lignin model compounds such as 4-propyl-
* To whom correspondence should be addressed. guaiacol (Ratcliff et al., 1987). Train (1986) developed a

0888-5885/88/2627-1783$01.50/0 &copy; 1988 American Chemical Society

You might also like