You are on page 1of 17

COMBUSTION A N D F L A M E 6 8 : 2 3 1 - 2 4 7 (1987)

231

A Simplified Model for the Pyrolysis of Charring Materials


I N D R E K S. W I C H M A N and A R V I N D A T R E Y A
Department of Mechanical Engineering, Michigan State University, East Lansing, MI 48824-1226

A simplified model of the pyrolysis of charring materials is analyzed. The effects of moisture are neglected, and the
heat of pyrolysis is assumed equal to zero. Four stages of pyrolysis are obtained: (i) inert heating, (ii) initial pyrolysis,
(iii) thin char, and (iv) thick char. Formulas for the volatile mass efflux, m, are obtained in stages (ii), (iii), and (iv);
m = 0 in the first stage. The calculations indicate that the surface temperature controls the volatile production rate in
the initial pyrolysis stages (the kinetically controlled regime), while the temperature gradient controls the volatile
production rate in the thick char stage (the diffusion-controlled regime). Comparisons of the calculated results with
numerical computations are made for the volatile mass efflux, the surface temperature, and the density.

I. I N T R O D U C T I O N
Wood and other cellulosic materials constitute a
substantial fraction of the fuel load in many
building fires, and are increasingly being utilized
as renewable sources of energy. Thus, it is
important to understand their behavior under hightemperature burning conditions. When heated
sufficiently these materials eventually undergo
thermal decomposition, producing combustible
gases (volatiles), water vapor, and char. Quantities that are of particular importance in the field of
fire research are the evolution rate of the volatile
gases and the surface temperature of the pyrolyzing sample. Knowledge of these quantities is
required in both the ignition and growth stages of
the fire.
A substantial amount of experimental and theoretical work on the pyrolysis of thermally thick
samples of wood and cellulosic materials has
already been done. The recent experimental work
is reviewed in [1], and will not be considered
further here. The previous theoretical studies can
generally be divided into two groups: (1) detailed
numerical studies that attempt to provide a comprehensive description of wood pyrolysis by ineluding as many physical processes as possible and
Copyright 1987 by The Combustion Institute
Published by Elsevier Science Publishing Co., Inc.
52 Vanderbilt Avenue, New York, NY 10017

(2) simplified analytical studies that attempt to


develop practical and useful formulas for quantities such as the volatile mass efflux. The initial
numerical studies [2] included volatile heat convection, Arrhenius decomposition kinetics, net
decomposition exothermicity, and variable thermophysical properties. Subsequent studies have
added the effects of volatile gas momentum [3],
moisture, and a simulation of char cracking [4].
The numerical study of Ref. [5] develops an
overall energy balance by computing individual
energy balances across N thin slices of the
pyrolyzing material. The individual slices retain
their identities as the material shrinks and cracks
during pyrolysis and char formation. The problem with the numerical models is that they are
neither simple to use nor comprehensive enough to
include all of the relevant physics and chemistry.
Notable omissions from the models are the effects
of ambient oxygen concentration, variable char
yield, and moisture content. In addition, it is
difficult to describe the relationships between the
various parameters of the problem. On the other
hand, the existing analytical studies [6] ignore all
physical and chemical processes except the transient propagation of a constant temperature pyrolysis front into the solid, which advances into the

0010-2180/87/$03.50

INDREK S. WICHMAN and ARVIND ATREYA

232

reaching the surface. The sample is assumed to be

virgin solid while leaving behind a char layer.


Thus, the analytical models ignore the transient
chemical decomposition process, which limits
their range of applicability and makes them unsuitable for analyzing certain physical processes, such
as ignition.
The purpose of this study is to analyze a
simplified model of wood pyrolysis that includes
the effects of global decomposition chemistry and
variable thermophysical properties. The analysis
produces formulas for the volatile mass flux that
are applicable during all stages of the pyrolysis
process, and exhibit its dependence on the most
important thermal, chemical, and thermophysical
parameters.
The physical configuration studied here is illustrated in Fig. 1. The thermally thick (semiinfinite)
sample, initially at equilibrium with the ambient
surroundings, is subjected at/" = 0 to a constant
incident radiative heat flux F. It is assumed that
the gas phase is inert (no oxidative effects) and that
the influences of solid-phase cracking, shrinkage,
surface regression, and grain direction are negligible. The interaction between / and the gaseous
products of pyrolysis is also neglected; thus, ~ is
assumed to be the time-average radiative heat flux
A .'4

~-

nondiathermic and opaque to the incident radiative


flux.
The initial heating phase, herein called the inert
heating stage, is characterized by the constant
density heat up of the surface to 7~p, the pyrolysis
temperature, defined as the value of 7~ at the
inception of pyrolysis. No chemical reactions
occur in this stage, and since the material is
considered dry (i.e., only water of constitution),
there are no effects of moisture. In the next heating
phase (the initial pyrolysis stage) the initial release
of volatiles from the surface coincides with the
formation of a pyrolysis front. In this phase a rapid
rise in the volatile mass flux is observed, and heat
losses from the surface by convection and reradiation become important. The surface temperature
and the volatile mass efflux continue to increase
until a thin char layer forms there. In this heating
stage (thin char) the mass flux attains its maximum
value. The final heating stage (thick char) is
characterized by a gradual decline in th and a
correspondingly gradual increase in the char layer
thickness. The expected time span of each of these
four stages is illustrated qualitatively in the
versus f plot of Fig. 2.

~4.

o ~Ts -T= .'

o.

H(Ts-T=)

....
/

-xa#_

I////////I

F-H(T-T.)
,.,

~dT

-E'/RT
,4

.A

p--p,., T=T.
Fig. 1. The physical configuration. The dimensional governing equations and boundary
conditions are also shown.

233

PYROLYSIS MODEL OF CHARRING MATERIALS

0 _ Inert _ l T r a n s i t i o n
r h e a t i n g l - regime

_ _i Thin
- I - - ] char

Thick
char

?
"

Fig. 2. A qualitative plot of the volatile mass flux, rh, versus time, /', indicating the separate

pyrolysis regimes to be examined.


II. F O R M U L A T I O N
Before stating the nondimensional governing
equations and boundary conditions, three important assumptions made in this study must be
discussed. The first assumption is that the interactions between the volatile gases and the reacting
material can be neglected. It is known that the
generation of volatile gases inside the solid produces high pressures (up to 0.3 atm [7], depending
on the wood porosity), which force the volatiles
toward both the hot char layer and the interior of
the solid. The gases traveling into the interior of
the solid condense, only to be subsequently regasifled [8]. The net heat transfer between the volatiles
and the hot char, whose magnitude is measured by
the quantity Cpg(Ts - Tc), is ignored for two
reasons. First, for small char layer thickness,
(~pgT"c)(7~s/T'c - 1) is negligible because (Ts/7"c
- 1) ,~ O(1). Second, when the char-layer
thickness is large, and Ts/Tc >> O(1), most of the
volatiles are issued through cracks that form in the
char layer, which makes the net heat transfer
between volatiles and char negligible. Thus, the
heat transfer between the volatiles and the solid
depends on the extent of char cracking, which has
been disregarded from the outset in this model.
The effects of condensation and regasification are
ignored because the volatile gas physical proper-

ties, such as condensation temperatures as functions of pressure, are unknown; therefore, the
calculation of their influence is impossible.
The second assumption is that the chemical
processes occurring in the decomposition of
" d r y " wood to char can be modeled by a single,
one-step rate equation containing three fixed parameters, the char density, the preexponential
factor, and the activation energy. This assumption
is an idealization of the actual process since wood
decomposes in a complex manner, producing
hundreds of compounds. However, there is presently much controversy concerning its precise
nature (even for pure cellulose, see, e.g., [9, 10]),
and no generally accepted reaction pathway, analogous to those used to model gas-phase reactions,
has yet been developed. Thus, since the physical
parameters appearing in the one-step rate equation
have direct physical meaning, and since much of
the recent experimental work has been devoted to
their measurement [11], it is felt that use of the
one-step scheme is appropriate.
The third assumption is that the heat of pyrolysis term, Q, which appears in the equation for
conservation of energy, is negligible. As discussed
in Ref. [4], there is much confusion concerning Q.
Reported values vary between 180 cal/g (endothermic) and - 4 5 0 0 cal/g (exothermic). The
exothermicity apparently arises from secondary

234

INDREK S. WICHMAN and ARVIND ATREYA

reactions between the oxygen and the hot char, or


the pyrolysis gases. However, the net exothermicity of the process must be relatively small, since
thermal runaway (i.e., explosion) has never occurred during wood pyrolysis. An estimate for the
upper bound of Q is c p ( T p - Ta,) ~ O(100 cal/
g), which is the sensible heat required to raise lg
of wood to its pyrolysis temperature. Thus, Q _>
100 cal/g could lead to thermal runaway. The
detailed numerical study of Ref. [12], using
previous experimental data, gives only slightly
endothermic Q. In the absence of an oxidizing
atmosphere, this seems to be the more reasonable
conclusion, since wood pyrolysis is a degradative
chemical process. Therefore, since [Q[ - O(10
cal/g), whether exothermic or endothermic, is
generally much smaller than the thermal diffusion
term in the equation for conservation of energy [~
OET/O.~ 2 - Cp(7"p- Tea) - O(100 cal/g)], Q = 0 is
assumed here. This assumption has also been
employed by previous authors [4, 5].
With these assumptions the model equations and
boundary conditions describing the pyrolysis of
wood reduce to those shown in Fig. 1. If it is
additionally assumed that the thermal conductivity
is proportional to the density, i.e., ~ = ~, where
~ = constant = ~oo/~=, then the nondimensional
equations become

sional activation energy and preexponential factor


for the overall decomposition reaction, t5 = ~c/~.,
the charyield, and F = / - I T ~ / P + ( ~ C 4 / / F ) ( f s
+ 1)(Ts2 + 1), the linearized heat loss term,
containing the effects of convective and linearized
radiative heat losses from the surface. Here, ~ is
an average nondimensional surface temperature,
chosen to match the heat losses to their actual
values over the appropriate temperature range.

Energy:

P at

cgx P
Op

Decomposition:

--= -A(p-6)e
Ot

Mass:

am

ap

ax

at '

(1)

'

-E/r,

(2)

(3)

Boundary and initial:


T ( x , 0 ) = p ( x , 0)= T ( ~ , t ) = 1
aT

- P ~xx (0, t ) = 1 - r ( T s -

1).

III. ANALYSIS
There are four distinct stages of pyrolysis that
will be considered in the limit of high activation
energy, E ~ ~ . The important features of each
are illustrated qualitatively in Figs. 3a-3d. In the
inert heating stage the density is constant, because
Ts < Tp is always satisfied (Fig. 3a). The initial
pyrolysis stage contains the preliminary effects of
chemical reaction. A narrow reaction zone develops near the surface, in which 6 < p < 1 and
Tp < Ts < Tc (Fig. 3b). When Ts passes through
To, a thin char layer forms at the surface (Fig. 3c).
In this stage the conditions satisfied by p and Tare
p = ~, Ts -> Tc in the char layer, and t5 < p < 1,
Tp < T < Tc in the pyrolysis zone preceding it.
The temporal and spatial matching of these three
regimes is easily achieved because they fall in the
initial, or kinetically controlled, pyrolysis regime.
The thick char stage, however, represents the long
time behavior of the system (Fig. 3d). Here, the
surface temperature approaches its maximum
value, Tmax = 1 + 1 / r , obtained from the second
of Eqs. (4) by putting a T s / S x = 0. Temporal
matching of the solution in this stage to that in the
kinetically controlled regime is very difficult,
because the time t appears only as a parameter in
the equations.

III.1. The Inert Heating Stage


(4)

Here, T = T/~**,p = ~ / ~ , x = / [ / S =
05=/~)7~**/P], and t = /'/? [? = 2/&=] were
used for nondimensionalization. The four nondimensional parameters appearing in Eqs. (1)-(4)
are E = /~//~7~ and A = A?, the nondimen-

Here, one defines D = A e x p ( - E / T p ) as the


Damk6hler number, thereby allowing Eq. (2) to
be rewritten as Op/at = - D ( p - ~) e x p [ ( - E /
T p ) ( T p / T - 1)]. Thus, for large values of E, a p /
at will be very small until T is very close to the
pyrolysis temperature, Tp, giving p = 1 as the
approximate solution for the density. Therefore,

235

PYROLYSIS MODEL OF CHARRING MATERIALS


Reaction
zol~m

Ts<TO

p= T=

pT
{a)

ps=pf - -klye
. ---x
\

"Is:,.Tp

(b)

Ts>Tc

Ts'Tmax

' t~~ 1 ~ ~-Tc

Tc
layerzone
Tc 4._.-----Reaction

"z'& I
p= T
(c)

p=,T=
(d)

Fig. 3. The four pyrolysis regimes: (a) the inert-heating stage, in which no reaction occurs
and the density is constant, (b) the transition from inert heating to charring conditions, (c) the
thin-char regime, in which Ts > To, and (d) the thick-char regime.

the solution for the temperature, from Eqs. (1) and


(4), is

where

TI = 1 +

a=(-OTi/OX)x=o=er2tp

1[

erfc

X erfc I"v~+

- e rx+r2t

(5)

Consequently, the surface temperature, TI(0, t),


reaches Tp when t = tp, the pyrolysis time. The
definition of Tp (or tp) is presently arbitrary, but
from Eq. (5) the specification of one immediately
determines the other. The procedure for calculating Tp is now described.
Consider tp as a known parameter. Then the
expansion of Eq. (5) for x < O(1) and t - tp <
O(1) gives
T -

Tp-ax+b(t-tp)+...,

(6)

and
OT
-- -

ax

erfc(r4~p),

(8)

t=tp
1
ra,

(9)

To= TI(0, tp)= 1 + ~ (1 - a ) .

(10)

b = (OTx/at)x=o -

t=tp N~tp

and
1

An approximation for the value of the density at


which pyrolysis begins, Pp, is obtained by evaluating Eq. (2) at the surface, and using P = 1 and the
temperature field given by Eq. (5) on the righthand side, viz.,
Op

A (1 - 6 ) e - E / T 1 ('t).

Ot
- a + b [ x + I'( t - tp)] + " " ,

(7)

Since most of the changes in P occur for T~ near


Tp, the expansion given by Eq. (6) is used.

INDREK S. WlCHMAN and ARVIND ATREYA

236
Asymptotic evaluation of the resulting integral
over t in the limit E ~ oo leads to

Op = 1

A(1 - 5 )

- e-E/Tp.
bE/Tp 2

(I 1)

Thus, fixing 0p allows Tp and tp to be determined


iteratively from Eqs. (8)-(11), since A, E, 6, and
F are known input parameters. Although the
specification of pp is also arbitrary, it seems much
more reasonable to specify the extent of decomposition beyond which one is unwilling to continue
the inert solution f o r the temperature, than to
specify T 0 arbitrarily and to terminate the inert
solution at a different level of decomposition for
each case.
By usingA = 1011, E = 40, 5 = 0.3, F =
0.4, and assuming P0 = 0.99, one finds T 0 = 1.5,
tp = 0.27, a = 0.8, and b = 0.76. These values
are used in Section IV to compare the theoretical
calculations with the numerical computations. The
available experimental data indicate that reasonable values for the pyrolysis temperature lie in the
range 400K _< Tp < 600K [4], i.e., 3/2 _< Tp _<
2. The calculated result for T o falls in the lower
end of this range.

111.2. The Initial Pyrolysis Stage


In this stage it is useful to rewrite Eqs. (1)-(4) in
terms of a new temperature, defined as the
difference between the actual temperature and the
inert temperature field given by Eq. (5), viz., T =
TI + O. Thus, Eqs. (1)-(4) become

(i)

0t

0x ~

~\~-x

(ii)

a__pp= _ D(p - 5) exp


Ot

Ox

'

ro

(iii)

O0
(l-o)
-O-~(O't)=
p

(iv)

O(x, 0 ) = 1 ,

( OT,'~ FO
\-~X/---'O

O(x, 0)=0(o% t)=O.


(12)

Since E/Tp is large, the departures from inert


heating (O = 0) will first occur near the surface
and for t close to tp. Therefore, into Eq. (12i)
introduce the expansion for TI given by Eq. (6),
and expand for small O, retaining only the lowest
order terms;

019= _D(p_5)e_(E/Tp2)[ax_b(t_tp)_Ol
Ot

(13)

In this initial pyrolysis zone it is necessary to


include both the space and time variations of the
inert temperature field. Therefore, stretched variables are introduced such that spatial and temporal
variations are both retained in the exponent of Eq.
(13), viz.,
(i)

~'= (aE/Tp2)x,

(ii)

r=(bE/Tp2)(t-tp)+lnDo;
Do =D/(bE/Tp2),

(14)

and the following expansion is introduced for O:

0 = Oo/(E/Tp 2) + o [(E/To2)-I].

(15)

Note from Eq. (13) that since O(1) changes in p


result from small changes in ~', r, and Oo, the
quantity 1 - 19 is not stretched.
With these scalings of x, t, and O the transient
term in Eq. (12i) vanishes; however, transient
effects are retained in Eq. (12ii), which is coupled
to Eq. (12i) through the diffusion term. Consequently the effects of diffusion, reaction, and
transients are all present in the governing equations. This result contrasts with those for solidphase ignition problems, in which only the energy
equation containing transient, diffusion, and reaction terms is analyzed [13]. The uniform scaling of
the spatial and temporal variables in the initial
ignition stage results in the loss of the transient
term from the energy equation and hence in the
formation of a purely reactive-diffusive inner
zone, which must subsequently be matched to an
outer transient-diffusive zone. In the present study
no such matching of spatially inner and outer
solutions is necessary, because all of the initial and
boundary conditions can be satisfied by the inner
solution alone.

237

PYROLYSIS M O D E L OF C H A R R I N G M A T E R I A L S
By substituting Eqs. (14) and (15) into Eqs.
(12), (6), and (7), and noting that the initial
conditions (12iv) apply at t = tp (i.e., at z = In
Do), one obtains the following lowest order
equations;
(i)

p -~/=~,

(ii)

0p
--=
Or

- ( p - 6)eo +T-r,

(iii)

OOo
1-O(0, r)
- (0, ~-)=

(iv)

p(~', In Do) = 1,

~"

(16)

p(0, r)

Note from the definition o f s that as Do ~ 0, So


- oo. Putting so = - oo directly into Eqs. (18)
gives p = 1, q = 0 as the trivial solution.
However, a linearized critical point analysis near
(p, q) = (1, 0) gives q = qo exp(s - So), and p =
1 + qo exp(s-So). Use of the initial condition qo
= - ( 1 - 6) exp(so) indicates that the solutions
for p and q become independent of so as So ~ oo. Therefore the line q = 0 in the p - q plane is a
line of unstable nodes. The only remaining parameter in Eqs. (19) in the limit E ~ oo is the char
yield, dt.
An expression for the volatile mass flux is found
by integrating the equation O m / a x = Op/at, viz.,

Oo(~', In Do)

i o -OoOt- d x .

m= -

=00(0% r)=0.
By introducing the similarity variable

s=r-~=(bE/Tp2)(t-tp)+lnDo-~,

C o ( s ) = is_~

p(u----3--du,

(18)

which satisfies eo(OO, v) = 0 and the condition


(16iii) at the surface. For the initial condition, t --,
tp in Eq. (17) gives s -* In Do - ~'; as the
activation energy increases and Do becomes vanishingly small, s -} - o o and Oo -* 0.
In the similarity coordinate, s, and with the
solution for Oo(S) given by Eq. (18), the decomposition equation, (16ii), can be reduced to the
following pair of first order ordinary differential
equations;
(i)

(ii)

do

--=q,
ds

p(So) = 1,

q
- - ~--,
ds ( p - ~) p

dq

q2

q(so) = - (1 - a)eSo,

b iv do d s = b [ l - p ( 0 ,
a -=ds
a

(17)

and assuming that Oo and p are functions only o f s


(i.e., that the pyrolysis process is wavelike), the
solution for Oo is obtained as

. b. i. Oo ds
a o Or
r)].

(20)

Therefore the mass flux attains its m a x i m u m


value when the surface f i r s t chars. An analytical
approximation for m is obtained by integrating Eq.
(16ii) directly,

m= --

o -~T

d~ - - e "
a

(o-6)eo-~

b
=- (1- 6)e',

d~

(21)

where O = 1, Oo = 0 were used in the final step.


Substitution for z from Eq. (14ii) gives
A(1 - 6 )
m

-"

e - el Tp +(bE/Tp2)(t -

tp),

(22)

( a E / T p 2)

which indicates that m increases exponentially


with increasing t - tp. By using Eqs. (6) and (7),
evaluated at the surface, Eq. (22) becomes

m -

A(1-6)

T~2

e-E/rs,

(23)

( - OT/Ox)s E
(19)

where So = In Do - ~'. The initial condition for q


was obtained by evaluating Eq. (16ii) at t 0.

which is a more general


The solution of Eqs.
tained numerically, by
Runge-Kutta numerical

result than Eq. (22).


(18)-(20) was also obusing a fourth order
integration scheme. As

I N D R E K S. W I C H M A N and A R V I N D A T R E Y A

238

e#

2.0

LO-

1.5 l

p,

O,

-1.0

.5

..5

=0.6

~:o

2'0

/b

'

'/o
-

JlO

2.'o

Fig. 4. Plot o f p, rh = 1 - p, a n d Oo as functions o f the similarity v a r i a b l e s = r -

~'in the

transition s t a g e o f h e a t i n g , for the c a s e 6 = 0 . 3 , so = - 13. C h a r r i n g conditions are attained at


s = s* = 0 . 6 .

Do --" 0 (i.e., as E --* oo) the numerical solutions


were found to depend only on ~, as expected.
Shown in Fig. 4 is a plot o f p, Oo, and r~ = m / ( b /
a) versus s, for the case (5, So) = (0.3, - 1 3 ) .
Note from Eq. (18) that as p approaches ~, Oo
approaches [(1 - 5)/~]s; thus, the gradient o f Oo
approaches its m a x i m u m value. Note also from
Fig. 4 that the attainment of/~max " = 1 - ~ occurs
when s = s* = 0.6; consequently, the approximate time to charring conditions can now be
calculated from Eq. (14ii), viz.,
tc = tp + ( b E / T p 2 ) - 1 [s*(6) - In Do].

t l.O5"

.5"

(24)

The corresponding approximation to the maximum


value o f the volatile mass flux is
mmax = ( b / a ) [ 1 - 6].

1.0

.5

Fig. 5. Plot o f s * ( ~ ) v e r s u s (5, indicating the b r e a k d o w n o f the


a s y m p t o t i c a n a l y s i s a s / i --* 1.

(25)

A plot o f s*(5) versus a is shown in Fig. 5. The


decrease in s* for a >_ 0.75 indicates that chemical
reaction is becoming increasingly difficult. In the
limit a --* 1, no reactions can occur, and the
concepts o f mm~x and to, as well as the asymptotic
analysis, break down.
An estimate for the pyrolysis zone thickness can
be obtained from Fig. 4. By defining s.95 as the

location o f p = 0.95 one finds (for fixed r) from


the formula s = r - ~" that the pyrolysis zone
thickness is given by A~" = ~" - ~ = s* - s.95 =
3.1. Therefore,
3.1
AX -

aE/Tp2

(26)

which becomes vanishingly small

as

E/Tp 2 -'* oo.

PYROLYSIS M O D E L OF CHARRING MATERIALS

239

111.3. The Thin-Char Stage

Eq. (28) reduces to [compare Eq. (16ii)]

Following the procedure of the previous section


the temperature is written as T = TCHAR + ~ ,
where TCHARis the solution for T until t = t~, and
if' is the correction thereafter. Thus, by definition,
q/(x, to) = 0. The lowest approximation to Tcrt~ is
obtained by combining Eqs. (5), (15), and (18):

00
--=
0re

Oo
TC.AR = TI +E/Tp--------5= TI

I t(l-p(u))/o(u)]
-~

du
(27)

E / Tp2

By substituting T = T c ~ a + if' into the decomposition equation, and expanding for small x and
small t - t~, one obtains the lowest order equation
[compare Eq. (13)],

-(O-6)e

*o+,c-l~.

(31)

These variables can be used to rewrite the energy


equation, and the boundary and initial conditions,
whose numerical integration then gives m(t) for t
near to.
However, instead of pursuing such an involved
computation, a simple formula for the volatile
massflux maximum in terms of the char parameters is derived here. Thus, by putting t = t~ and
~I'o = 0 into Eq. (31) and then integrating over the
spatial coordinate, one finds

mmax

dx = - - -

--

A ( 1 - - 6 ) e-E/rc I ' ( 0 - 6 )

-- a~E/Tc
------2

O._O=
p - A e - E/rc(p _ 6 )e_(E/Tc2)[acX_bc(t_tc)_ ~ l,
Ot
(28)
where [compare Eqs. (8) and (9)]
(i)

a c = ( -- OTCHAR/OX)x=

0
I=t c

= er2t~ erfc ( r x/~c) + ( ~ - ~ )

(ii)

bc =

(33)

satisfies the conditions Plx=0 = 6, Op/OXlx=o =


O, Op/Ox[x=~. = 0, andplx=ax = 1 - (1 - 6)e(1

(29)
By defining the stretched variables

+ k)e -+k), where Ax is the pyrolysis zone


thickness. For the case 6 = 0.3 the choice Plx=a~
= 0.95 gives k = 4.3. By recalling Eq. (26), and
noting from Eq. (30i) that x = ~cTO/acE, Eq.
(33) can be rewritten as
p-6

~ = (a~E/T~2)x,
F Ae-E/Tc ]

(ii)

r~ = ( bcE/ Tc2)( t - t~) + In L ~

(iii)

~I, =

(E/Tc 2)

The integral is difficult to evaluate because P is


obtained numerically from Eqs. (19). Therefore,
an approximation that allows evaluation of the
integral is made for the function (O - 6)/(1 - 6)
at the first instant of char formation at the surface
Lo(0, tc) = 6]. The function

P-6
11-6= - ( l + k - ~ x ) e-~X/aX

- rer2'~ erfc(r x/~)

~Io

(32)

--

I=fC

(i)

e - t~ d~c.

a~

1
(OTcHAR/Ot)x=O

d~c

+ o [(E/T~z) - ~],

_l '

(30)

1-6

= 1 - [ 1 + kc ~'~]e-kcr,

(34)

where
4.3 a

k c - 3 . 1 ac

Tp/

"

(35)

240

INDREK S. WICHMAN and ARVIND ATREYA

Thus, Eq. (32) gives

OT/Ox, which are used as boundary conditions for

~2
A(1-~)
1 + kc,/ ( - a T / O x ) s
kc

mmax - -

TcEe-e/rc'
t=t c

(36)
which should be compared with Eq. (23) for
functional dependence of m on T.
111.4. T h e T h i c k - C h a r

T in the unburnt-wood and hot-char layers on


either side of the reaction zone. Matching of the
solutions in the three zones gives oR(t) in Eq. (38).
Since the advance of a constant-temperature front
into a solid is proportional to ~ , its rate of
advance varies as t - 1/2 ; thus, m ~ (1 - tS)t- 1/2
from Eq. (38). The asymptotic analysis of the
following two subsections produces an approximation for the constant of proportionality.

Stage
III.4.1.

In this stage the reaction zone is assumed infinitesimally thin with respect to the length scale L,
while the surface temperature is assumed close to
its maximum value, given by Eqs. (4) when a T/Ox
= 0:

Ts,max= 1 + 1/F.

(37)

The solution for rn in this heating stage cannot be


matched in time to the solution in the thin-char
stage because the spatial ordering in the inner
reaction zone reduces t to a parameter in the
governing equations.
A simple mass balance across the reaction layer
gives (see Fig. 6)

The Inner (Reaction)

Zone

Based on the calculations of the preceding two


subsections, the temperature difference T - Tc
and the immediate vicinity of the reaction layer
x - x~(t) are stretched with the factor E / T J ~,
O(I); no stretching is employed for t. Thus,
(39)

0 = ( E / T c 2 ) ( T - To)

and
S = - ( E / T c 2) [ x - Xc(t)],

(40)

so that in the lowest order Eqs. (1) and (2) become


Energy:

~-~ p ~

--0,

(41)

dxR

m = (1 - ~) - - ~ = (1 - 5) OR,

(38)

Decomposition:
where OR = dxR/dt is the velocity of the reaction
layer. To solve for m, the reaction zone must be
studied in detail. This leads to jump conditions on

ap
- - = - (p - 6 ) K ( t ) e - , (42)
aS

where
Ae-E/Tc

(43)

K(t) = (E/Tc)2(dxc/dt) ,

while the boundary conditions are now given by


(see Fig. 6)

%hot
' "

,~-7 . . . . . . . . .t

2" "

"////////v'/I/////////////f/////////4
" ......
.....
"T
dt

"

R~tion Layer

(i) char layer:


S--*~:
(ii) unburnt wood layer: S ~ o o :

O--' - 0%

p ~ I.

p-*~
(44)

The solution of Eq. (41) gives

VJ~in Wood

O = g (t)
Fig. 6. Mass balance across the reaction layer, giving m = (l
- it)oR. Also shown are the boundary conditions on Eqs. (41)
and (42) in the stretched S-coordinate system.

O--+oo,

iso o(z, t) ,

(45)

where g(t) is an as-yet-unspecified function of t.

241

PYROLYSIS MODEL OF CHARRING MATERIALS


Note that g(t) > 0 since - OT/Ox = O0/OS > O.
Note also from Eq. (41) that the heat flux across
the reaction layer, O O0/OS, is constant, since Q
= 0 (see the discussion of Section II). Equation
(45) satisfies the conditions
lim O(S, t ) = [g(t)/6] lim S = oo
and
lim O(S, t ) = [g(t)/6]

lim S = - oo.

Thus, the jump condition for the temperature


gradient across the reaction zone is given by

111.4.2. T h e Outer H o t - C h a r and U n b u r n t Wood Zones

In the two outer zones the density is constant (see


Fig. 6); thus, the temperature fields obey Eq. (1),
with p = 6 in the hot-char layer and p = 1 in the
unburnt-wood layer. The boundary conditions are
TH(0, t ) = Tmax = 1 + 1" - I , Tn(xc(t), t ) = Tc
(hot-char layer) and Tw(xc(t), t) = To Tw( ao, t)
= 1 (unburnt-wood layer). Integration of the two
energy equations gives TH = A + B erf(x/2x/~),
Tw = C + D erf(x/2x/~, where A, B, C, and D
are constants. Application of the boundary conditions and the jump condition [Eq. (46)] requires
x~(t) = 2Fx/t

(48)

and

= ( ~ 6 6 ) g( t),

(46)

which will be used in the following subsection to


match with the solutions in the two outer zones.
The solution for p is obtained by substituting
Eq. (45) into Eq. (42), which can then be written
as the following first-order system of equations:

do

(i)

~-~=q,

p(S--, - oo) = 1,

(ii)

dq - - - qZ ~ g ( t ) - q,
dS 0 - 6
o
q ( S ~ - oo) - (1 - 6 ) K ( t ) e gins.
(47)

This system is identical to the one derived in


Section III.2, except for the factors g(t) and K(t).
The analysis of the following subsection shows
that g(t) - t -1/2 and K(t) ~ tl/2; thus, Eqs. (47)
were integrated numerically with a fourth-order
Runge-Kutta scheme, starting with K(t) = g(t)
= 1, and then increasing t. Figure 7 shows a plot
of AS, the reaction-zone thickness, and K(t) and
g(t), versus t x/z. A nearly linear increase of AS
with t 1/2 is observed, indicating that the assumption of an infinitesimally thin reaction layer
becomes progressively worse as t increases.

g( t) = J/'ftt,

(49)

where F and J are constants. Since there are six


constants (A, B, C, D, F, J ) and only five
boundary conditions (including the jump condition), an additional condition is necessary. This is
obtained from Eq. (41) by noting that the heat flux
across the pyrolysis zone is constant, whereby
( -- 6 0 T H / O X ) x R ( t )
= ( - - aTw/aX)xR(t), which, with
the above solutions for TH and Tw leads to the
requirement, ,,/'6-B exp( - F2/6) = D exp( - F2).
Thus,
fs-_
Tmax- Zc ~ / ~ e -F2/~,

(50)

J = erf(F/x/-6)
and F is determined from the transcendental
equation

e_F2[(l/6)_l I

erfc F
1 To- 1
e r f ( F / x / 6 ) - x / 6 Tm~x- Tc "

(51)

For small F, Eq. (51) reduces to

F=

T-I/"

(52)

The formula for m in this heating regime is


obtained by substituting Eq. (48) into Eq. (38),

242

INDREK S. WICHMAN and ARVIND ATREYA

K(t)

t.O

A5

K(t
q(t)),

5.5

AS(t)
I

1.5

Fig. 7. The reaction-zone thickness, AS, as a function t wz, indicating that AS -- t v2. The
functions K(t) and g(t), used in the calculation of AS, are also plotted against t j/2.

viz.,
m=

(1-tS) F

(53)

IV. N U M E R I C A L COMPUTATIONS:
COMPARISON AND DISCUSSION
Equations (1)-(4) were also integrated numerically
with an iterated Crank-Nicolson scheme [2].
Shown in Fig. 8, for the case (,4, E, ~, r ) =
(10 n, 40, 0.3, 0.4) is the resulting m versus t plot
and its comparison to the predictions of Eqs. (20),
(22), and (23). The mass flux rises until p = ~ at
the surface, and decreases monotonically thereafter. By using pp = 0.99 the parameters shown in
Table 1 can be calculated. The m versus t profiles
of Eqs. (22) and (23) approximate the exponential
rise of m well for small t - tp. However, they rise
much faster as t - tp increases, since reactant
depletion has been neglected. The estimate for m
given by Eq. (20) is also an overprediction;
although reactant depletion has been included, the
asymptotic formulation uses the linearized Arrhenius exponent, ( - E / T p ) + ( E / T p 2 ) ( T Tp),
which for T > Tp is always less than its actual

value, ( - E / T ) . Thus, the reaction rate in the


asymptotic formula is always higher and the mass
flux peak must therefore occur sooner. The
asymptotically predicted values of mr~x [Eq. (25)]
and tc therefore underestimate the actual values
[see Table 1].
A plot of the surface temperature and density
profiles for the same case is shown in Fig. 9, along
with the m versus t profile of Fig. 8. The
asymptotic estimate of ps(t) is obtained from the
integration of Eqs. (19), with ps(tp) = 1. The rate
of decrease of the asymptotic estimate is faster
than the numerical predictions of Eqs. (1)-(4), as
previously discussed for m. For Ts(t), the inert
profile of Eq. (5) follows the numerical solution
until t = tp, when Ts rises rapidly. For t _> tp,
Tis(t) is a poor estimate of Ts. The asymptotic
estimate of Ts for t > tp is obtained by using Eqs.
(5), (15), and (18) in T = TI + O. The agreement
with numerical predictions is good until t = 0.7.
The char formation temperature, from Fig. 9, is
Tc = 1.97 (i.e., 7~c = 600K), at tc ---- (~.7. Then,
from Eqs. (35) and (36), kc = 0.74 and mmax --0.73. This point is also plotted in Fig. 8.
For the thick char stage, the substitution of T
= 1.97, {5 = 0.3, and Tmax = 1 + 1 / r = 3.5

PYROLYSIS

MODEL

OF CHARRING

243

MATERIALS

A(/-,~

bE

/.6

.5.

~m= ~[/_,,~o,r)l

.5

LO

L5

Fig. 8. The mass flux m versus t for the case A = 10 n , E = 4 0 , 6 = 0.3, a n d F = 0.4.
Shown are the numerical evaluation of Eqs. (1)-(4), the asymptotic predictions of Eqs. (22)
and (23), and the numerically evaluated mass-flux profile of Eq. (20). The numerical
computations give tp = 0.27, tc = 0.69, and rnmax = 0.825. The best-fit curve, m = 0 . 5 2 / ~ ,
is plotted in the thick-char regime. The point rn~x, evaluated from Eq. (36), is also shown.

TABLE

Calculation of Parameter Values for Each Heating Stage, for the Case A = 10 u, E = 40, 6 = 0.3, l" = 0.4 (Where Possible,
Comparisons Are Made between Calculated and Computed Values, and Percent Difference Is Shown)

Heating
Stage
Inert

Initial pyrolysis

Thin char

Quantity
Tp
tp
a

Calculated
Value

Source

1.5
0.27
0.80
0.76

F__.qs. (8)-(1 l)
Eqs. (8)-(11)
F.q. (8)
Eq. (9)

s*
tc
rn,~,
Ax

0.6
0.61
0.66
0.22

Fig. 4
Eq. (24)
Eq. (25)
Eq. (26)

a~

2.58

b~

2.16

1.76

Eqs.
used
Eqs.
used
Eqs.

(8), (29i);
tc = 0.7
(9), (29ii);
tc = 0.7
(5), (15),

Numerical
Value

Percent Difference
[(~NUM -- t~CALC)/~NUM)] X 100

-----

--

---

-0.7
0.825
--

-13
20
--

--

--

--

--

1.97

i0

--

--

0.825

12

-0.52/~-

-62

--

( 1 8 ) , at t = 0 . 6 1

Thick char

kc

0.74

mmax

0.73

F
m

0.27
0.2/x/t

Eq. (35); used


T = 1.97
Eq. (36); used
Tc = 1.97
Eq. (51)
Eq. (53)

244

I N D R E K S. W I C H M A N and A R V I N D A T R E Y A

/////

m,

*" .5.

t,

. . . . .

.27

.5

tO

L5

Fig. 9. Surface temperature and density profiles for the case of Fig. 8. The asymptotic
estimates of p~ and Ts, obtained in Sections III.1 and II1.2 are also shown (---). The inert
temperature profile is shown as Tt( . . . . . ). Note that Tmax = 1 + r - ~ = 3.5.

into Eq. (51) gives F = 0.27, which by Eq. (53)


leads to m = 0.2/x/~. This is an underestimate, by
nearly a factor of three, of the actual profile,
shown in Fig. 8 as m = 0.52/x/~.
Numerical profiles of p and T in the pyrolyzing
sample are shown in Figs. 10a and 10b, as
functions o f x and t, for the case (A, E, tS, F) =
(10 I1, 40, 0.3, 0.5). The wavelike nature of the
decomposition process is evident from Fig. 10a.
Note that ps = ~ occurs at t = 0.8. Thus, in 0 _ t
< 0.8 the volatile mass flux is increasing, while
for t > 0.8 m decreases monotonically. Figure
10c shows the dependence of the local mass flux,
mL = A(O - ~) e x p ( - E / T ) ,
on x and t; note
that m = f~*mLdX- When t -< 0.4, m L i s t o o
small to appear on the plot. The maximum value of
mL, for t = 0.8 and t = 1.0, occurs when p =
0.55, which is considerably displaced from the
char front. The temperatures at the maxima of
mL(X, t) are T = 1.83 (t = 0.8) and T -=- 1.73 (t
= 1.0). Since p = 0.55 for b o t h maxima, p
cannot be a function only of T. Furthermore, if p
= p ( T ) , then mL = A ( p - ~) e x p ( - E / T )
=
mL(T) and m = I~' mL d x w o u l d a l w a y s be
constant. However, m ( t ) decreases as t - 1/2 in the
thick char stage, indicating strong functional
dependence on quantities other than T.

Examination of Fig. 10b shows that the temperature profile across the reaction zone (denoted as
the region within hatchmarks) becomes less steep
as the reaction zone propagates into the sample.
This and the above discussion suggest that the m
versus t behavior is controlled primarily by aT~
Ox. From Section 111.4, the substitution of Eq. (49)
for g ( t ) into Eq. (46) shows that the jump in 07"/
@x across the reaction zone decreases as t -~/2.
From Eq. (53), m oc t-1/2, which follows the
same time dependence. Therefore, the volatile
mass flux in the thick char stage depends on the
diffusion of thermal energy into and out of the
reaction zone, which is controlled by the local
temperature gradient.
By contrast, the mass flux in the initial pyrolysis
stage is controlled primarily by the instantaneous
surface temperature, Ts(t) [see Eq. (23)]. The
functional dependence of m on Ts is much
stronger than its dependence on ( - @ T / O x ) s .
Consequently the initial pyrolysis stage is referred
to as the kinetically controlled pyrolysis stage [4].
V. S U M M A R Y A N D C O N C L U S I O N S
The model equations for the pyrolysis of a
charring material were analyzed in four distinct

PYROLYSIS

MODEL

OF CHARRING

MATERIALS

245
t= . 2

t.O,

2.C

1.5

.01

.02

- x - .03=

.04

.05

Fig. lOa

0 ......

o.b5

0 '/

0 '2
~

03

"

Fig. 10b

n-.54

7/.83
]~

mL= A (n - 8) e -E/ T

L 3.

2.

o.1

o.'5

'

,.b
X

,,

"~

Fig. 10c
Fig. 10. Numerically calculated [from Eqs. (1)-(4)] profiles of p, T, and mL = (p -- ~i) e x p ( - E / T ) as
functions o f x and t, for the case A = l 0 n, E = 40, ~ = 0.3, 1" = 0.5. Note that the m a x i m u m of rn
occurs when t = 0.8 (i.e., when p~ = ~5 = 0.3 is first attained). The wavelike character of pyrolysis in the
transition stage is evident from (a). The m a x i m a of mL for t = 0.8 and t = 1.0 occur when p = 0.55.
Therefore, mL and p are n o t functions of T alone.

246

INDREK S. WlCHMAN and ARVIND ATREYA

regimes. Formulas for the volatile mass efflux,


re(t), were derived in the kinetically controlled
stage [Eqs. (20), (22), (23)] and in the diffusion
controlled stage [Eq. (53)]. An estimate for the
char formation time is given by Eq. (24). The
influences of the activation energy, preexponential
factor, char yield, and charring temperature on
mmax are given by Eq. (36), which has been
derived in the thin char stage. The wavelike nature
of the initial pyrolysis process is evident fromthe
asymptotic analysis, where the similarity variable
s = r - ~"reduces the mathematical description of
pyrolysis to ordinary differential equations. The
t - 1/2 dependence of the mass flux in the thick char
stage has been discussed by previous authors [6].
The analysis of Section 111.4 shows that the
assumption of an infinitesimally thin reaction zone
becomes poorer as t increases, since the reaction
zone thickness increases as t ~/2.
The present model can be made more realistic
by including the full nonlinear reradiation term in
the boundary conditions (recall that the reradiation
term was linearized), and by adding the effects of
moisture and char oxidation. The linearized reradiation term and the assumption )x = pk are
believed responsible for the anomalous temperature rise in the initial pyrolysis stage (see Fig. 9,
for 0.4 < t < 0.7), which is not observed
experimentally [14]. The available surface temperature profiles show a monotonic and much steeper
rise in Ts to its maximum value. The influences of
atmospheric oxygen and the oxygen of molecular
decomposition on the rate of pyrolysis are believed
to be significant [14]. The influences of moisture,
which is driven from the wood in the initial heating
phase, can be described by a kinetic equation for
the moisture density analogous to Eq. (2) [4]. A
theory based on the assumption of local thermodynamic equilibrium between the rates of adsorption
and desorption of moisture in the porous wood
sample can also be developed.
NOMENCLATURE
a

ac

nondimensional surface temperature gradient, Eq. (8)


see Eq. (29i)
nondimensional preexponential factor, A

b
bc
Cp
D
Do
E

g(t)
K(t)
kc
L
m
r~
q

nondimensional time rate of change of Txs,


Eq. (9)
see Eq. (29ii)
specific heat
Damk6hler number, D = A e x p ( - E~ Tp)
Damk6hler number, Do = D/(bE/Tp 2)
nondimensional activation energy,
incident heat flux
see Eqs. (46) and (49)
convective heat-loss coefficient
see Eq. (43)
numerical constant in the thin-char stage
characteristic length,/S = Xf'oo/F
nondimensional mass efflux

r~ = m / ( b / a )
q = da/ds, see Eqs. (19)
heat source term; Q = 0 is used here
Qs
nondimensional net heat flux into surface
/~
ideal gas constant
s
similarity variable, s = r - ~"
so
s evaluated at t = tp, So = In Do - ~"
S
stretched spatial coordinate in thick-char
regime, see Eq. (40)
AS
reaction-zone thickness
s*
value o f s for which m = mmax;S* depends
on 8 [see Fig. 5]
t
nondimensional time, t = f/
T
nondimensional temperature, T = 7~/f'~
TCnAR nondimensional solution for T up to charring time
OR
nondimensional reaction-zone propagation
velocity, OR = dXR/dt
XR
nondimensional reaction-zone location
Greek
&
r
8
~x
0

thermal diffusivity, & = ~,/~dp


nondimensional linearized heat-loss parameter
nondimensional char yield, 8 = p c / ~
pyrolysis zone thickness
nondimensional temperature correction in
the initial pyrolysis stage
thermal conductivity
stretched nondimensional spatial coordinate

PYROLYSIS M O D E L O F C H A R R I N G M A T E R I A L S
P
?
xI,

nondimensional density, p = ~ / ~
Stefan-Boltzmann constant
characteristic time, ? = /S2/&~,
nondimensional temperature correction in
the thin-char stage

Subscripts
c
H
I
max
o
p
s
w
oo

char
hot-char
inert
maximum
lowest-order approximation
pyrolysis
surface
unburnt-wood
ambient

Superscripts
dimensional quantity
average

This research was performed while the first


author was an N R C postdoctoral research staff
member at the National Bureau o f Standards
(Fire Research program). The opportunity provided by this program to study problems of
one's own choosing is greatly appreciated.

247

Partial support for the second author was


provided by the National Science Foundation
under Grant Number CBT-8415423.

REFERENCES
1.

Atreya, A., and Emmons, H., submitted to Combust.


Flame (1986).
2. Kung, H., Combust. Flame 18:185 (1972).
3. Kansa, E. J., Perlee, H. E., and Chaiken, R., Combust.
Flame 29:311 (1977).
4. Atreya, A., NBS-GCR-83-449, 1984.
5. Parker, W. J., NBSIR 85-3163, 1985.
6. Delichatsios, M. A., and deRis, J., Factory Mutual
Research, OKOJ1.BU, 1984.
7. Lee, C., Chaiken, R., and Singer, J. M., Sixteenth
Symposium (International) on Combustion, The
Combustion Institute, 1976, pp. 1459-1470.
8. Min, K., and Emmons, H., Proc. Heat Transfer and
Fluid Mech. Inst., 1972.
9. Shafizadeh, F., in The Chemistry o f Solid Wood (R.
Rowell, Ed.), ACS, Washington D.C., 1984, pp. 489529.
10. Hirata, T., NBSIR 85-3218, 1985.
11. Roberts, A. F., Combust. Flame 14:261 (1970).
12. Kung, H., and Kalekar, A. S., Combust. Flame 20:91
(1973).
13. Lififin,A., and Williams, F. A., Comb. Sci. Tech. 3:91
(1971).
14. Kashiwagi, T., Ohlemiller, T. J., and Atreya, A., 8th
U.S.-Japan National Resources Meeting on Panel of Fire
Research, 1985.
Received 14 April 1986; revised 22 January 1987

You might also like