COMBLtSTIONAND FLAME 18, 85-97 (1972) 85

Radiant Ignition of a Reactive Solid with In-Depth Absorption

A. LINAN
Escuela de IngenieroxAeronduticos, &mdad UniverJitw~a,Madrid, and
Instituto National de T&cnicaAerorpecial "'Esteban Terradas'; Madrid, Spain
and
F. A. V ~ L , L ~ S
Department o f the A erospaceand Mecha~:icalEngineering Sciences, University o f California,
San Diego, La Jolla, California

An asymptotic analy~s of the limit of large activation energy is presented for radiant ignition of a
sold that experien:es a one-step Ar~enins reaction in the condensed phase. Both constant and
time-dependent radiant-etm~ fluxes are considered, and the complete ra%~e of values is coveled for
the absorption coe~cient u. It is shown that as p inczeases, the structure of the transition stage, which
foilog~ the inert heat-conduction stage, passes from a thermal explosion without heat conduction, to a
~ l e transient heat-conduction zone with distn'buted chemical heat release, to a two-zone structure
composed of a reactive-diffusive-absorptive zone near the surface and a tr~sient-diffusive zone in the
interior. For very high values o f ~, the reactive-diffusive-absorptive zone further splits into a surface
absorption zone and an interior reactive.diffusive zone, thereby reproducing results obtained
previously for ignition by a surface-applied ener&v flux. The analyfis shows that contrary to earlier
expectation, the nondimen~onal absorption coefficient must be at least as large as the nondimensional
activation enexgy for in-depth absorption to affect the ignition time negligibly.

1. Introduction absorption bands, ~ may depend appreciably on

In a previous paper [1] we have employed •
asymptotic methods to investigate, in the limb of
large activation energy, the ignition of a reactive
solid by a constant energy flux appfied to its
surface. It has generally been assumed that during
radiant ignition of solid propellants, the absorp-
tion layer is thin enough for the approximation of
a surface-applied energy flux to be valid. However,
a theoretical determination of conditions neces-
saw for the neglect of in-depth absorption has not
be~.n reported. Our previous finding that a narrow
layer at the surface is of critical importance during
the Wan~/tion stage preceding ignition, suggests
that the traditional approximation may be ques-
tionable. Therefore we have performed the further
analyses reported he.,ein.
Difficulties in estimating the effect of in-depth
absorption s t ~ from two problems that arise in
defming the absorption coefficient./c First, par-
ticularly for pokymeric materials in the vicinity of
the frequency of the incident radiation. Analyses
neglecting this wavelength dependence will be
most nearly applicable either for monochromatic
laser radiation or for arc-image radiation apt~lied to
materials whose absorpfivities vary but slightly over
the energy-containing wavelength region of the
arc. The principal part of the following develop-
ment ignores any wavelength distribution. How-
ever, in the final section we shall show that
concepts of superposition enable some conclusions
to be drawn concerning ignition of materials with
wavelength-dependent absorbtivities.
A second problem that is especially severe for
heterogeneous materials, concerns the conse-
quences of scattering of radiation. In a crude
approximation, one may pze'eume that for hetero-
geneities of sufficiently fine scale, the overall
effect of scattering is to provide an additional
diffuse contn~oution to the reflection coefficient
and a modification to the effective value ofF. The

Copyright © 1972 by The Combustion Institute

Published by American Elsevier Publi-~ing Company, Inc.

constant U enqdoyed herein will be presumed to
represent an- effective overall absorption coeffici-
ent that includes, in an averaged sense, both
scat~rin~ and ~, f ~ u e n c y disU~ution. The ac-
curacy of this approximation has not been investi-
gated.
comld~tiom s u g a r that only careful
measurements can produce useable values of
Measmements h r , e been made for polymers such
as polystyrene and epoxy by O h l e m ~ r and
~~eld 121, who have ~
.'he negl~.~tof in-depth absorption D]-
T h ~ abso~tivities range from 74 a n -t for pure
polystyrene to 1200 cm-: for material containing
20~ carbon black. On the other hand, Evans,
Beyer, and McCu.rley [4] report effective values
/J from 103-los cm -s for p~essed ammonium
perctdorate powder, and they state that the values
are higher if small amounts o f catalyst are added.
Certainly trmstmrencies of dilferent materials
differ greatly. We estimate that in radiant ignition
experiments effective values of # may range from
10 cm-' (~.g., for nonopacified double-base for-
mulations) to 10s cm"~ , although these bounds in
fact may be too n ~ r o w .
In the present work we employ, as closely as
pnss~le, the methods and notation of Ref. I.
Therefore the ~ermal conductivi~ X and thc
thermal diffnsivity g of the solid are a~-umed to be
constant, and the initial temperature To of the
solid is taken to be uniform. The ratio of kTo to
the nonre.~Aed portion ~ o f the incident radiant
energy flux at ~nition will be employed as the
characteristic length in forming the nondimen-
sional sl~tial variable ~ .The quantity (~kT.~)IO}2/K
is the characteristic heat-conduction time that will
be used to form the nondimensional time variable
7. Dependent variables are 0(~, r) =_ T/To, where
T is the local instantaneous temperature, and
~(~:, z), the ratio of the product ~acenL~ation to
the final product. ~ncentration. Nondimensional
W~ae*.ers that appeared in the previous paper are
a Damko-hler numberA, defined as the ratio of the
characteristic heat-conduction time to a character-
i ~ c chemical time, the lat',er being based solely on
the. p~-exponential rate factor; the nondimen-
sion~ ~fivation energy Z " - Ta/To, where T=
is the activation tempe~.*ure, the order a of the
A. LIHAN and F. A. WIr_I]AMS
chemical reaction, and the heat-re!ease parameter
B - ( T / - To)/To,wheTe TI isthe adiaba~name
temperature of the ignition react/on. The addi-
tional nondimensional pel~meter needed herein is
the nondimensional absorption coefficient
a -- ~ T o I ~ .
T h e range of values of a that is of practical
interest can be inferred by ruing the previously
discu~-~ed range of values o f / ~ Since sofids with
high absorbti~ities often also have high thermal
conduc~ivities (e.g., materials containing high per-
centages of metalic constituents) and throe with
low values of/~ tend to have low v~hJes of X, the
range of a exceeds that o f / L Employ~mg To -
300°K, X from 3 x 10-4-3 x 10 -7 cal/cm-sec°K
and t~ from 10-100 cal]cmZ-sec we fred the:
l 0 -2 ~ a ~< lif t- Since generany l << ~ ' ~ l0 2 ,
it is clear that them is reason to study a range of
values of a that includes the extremes a << I and
a >> E ' . This complete range w~l be covered
herein.
In the absence of an analysis inch as that
performed below, sufficient conditions for in-
depth absorption to be negJ~ble cannot be stated.
For example, Wise, Inami, and McCuGey [5] us'd
the condition a2r> lO as a justification for
neglecting the influence of in-depth absorption on
the ignition ~t~e. We shall show that a more
precise condition is a > E'. We shaft also derive
formulas that predict the extent to which the
ignition time is ~ for a < E ' .
Section 2 briefly formulates the "~mition prob-
lem. In -~ecfion 3, properties of the inert stage for
constant flux with in-depth absorption are dis-
cussed. Next, the various regimes of a t~at arise
will be indicated. The two major regimes are
analysed separately in Sections 5 and 6. The
merging of these two _ _ , e ~ and the limit of
thermal explosions are discussed in Section 7.
Effects of time-yawing flux, wavelength dism'bu-
tions and multidimensionality are descn'bed in
Section 8.
Readers who are interested p d m a r ~ in final
results for ignition times under conditions of
constant radiant flux may turn directly to Eq. 36
for a > E', to Eq. 51 for a -,- 1, to Eq. 53 for
1<< a << E ' o r to Eq. 56 for a << 1; alternatively,
they may consult Fig. 3, which summarizes these
RADIANTIGNITIONOF A REACI'IVESOLID
results. Readers who are concerned with variable
flex will f'md "L~nition-tim¢formulas in Eqs. 57 and
58 and an appropriate plot in Fig. 4.
2. Fmmabtioa
Appropriate nondimensionai parameters and vari-
ables have been defined in the introduction. In
addition to these quantities, we m~.~oduce the
function l(r), which is def'med as the ratio of the
instantaneous incident radiant flux to the incident
radiant flux at ignition. For cla~ty of presenta-
tion, we shall assmne that l(r) = 1, except in the
f'm21 section, where the effect of an arbitrary flux
history is ascertained. Gasification and surfa~
n~ession are n ~ e c t e d throughout.
The mathematical problem that must be solved
ca~ be written a s .
Of" =
(1)
and
er = ( ~ . / ( l _ d - exp l_ ~ / ' (2)
with the initial conditions
e(~',0) = 0 , 0(~,0) = 1 (3)"
and the boundary conditions
¢(..,r) = 0 , O(=,r) = 1 (4)
and
oe(o,,') = o . (5)
In Eq. 1 the factor e , wldch is equivalent to
Beer's law, should provide an acceptable descrip-
tion of in.depth absorption of radiation, although
in no sense is it an exact deduction from the
relevant equation for transport of radiation. Ac-
cording to Eq. 5, heat loss from the surface is
neglected; this should be a good approximation for
solids igniting in gaseous atmospheres of moderate
or low density.
87
3. The l n ~ t St2ge
We parallel our earlier work in seeking the value of
A that produces thermal runaway at a spe~zed
time rc, which is selected consist±nt with the
requirement that the corresponding increase in
surface temperature under inert heating, 0c - 1, is
of order uniW. This requirement is imp.Jsed to
assure that there exists an inert heating stage prior
to the events that lead to ignition and that ignition
occurs before temperatures are reacbed that are
too high for a well-defined "~nition event to exist.
It can be shown that unless Oc - 1 >> 1 / g ' , a
distinct inert stage disappears, causing ignition to
merge with the phenomenon of adiabatic thermal
explosion, and unless Oc - ! << E ; chemical
energy will be released at too low a rate to
produce signh~cant departures from inert heating.
In view of these bounds, the fact that E" seldom
exceeds lO 2 demonstrates that there is little
physical interest in considering the order of Oc - 1
to be otl~r than unity.
In th~ inert stage that precedes transition to
ignition| to lowest order, ~ = 0, and 0 = 0t(~, r),
which fA~constant flux obeys the equation
Olr = 0t¢¢ + a exp(-a~) , (0
with the initial and boundary conditions
oz( ,o) = ot( , ,) = 1 , = o. (7)
The solution to the problem defined by Eqs. 6 £nd
7 has been given by Boehringer and Spindler [El ;
it is
0j = 1-+ tx-lf(ct~,a2r), (8)
where
f(u, v) -
2 v exp -uerfc u _e-=
+g
k
+ e-" effe (~/~ - 2 " ~ v ) ] "
88
F~luatiom 8 and 9 predict that the peak tent-
pentture always occurs at the surface of the solid.
The expamic~ about the surface is
0 i ~-- 1 + 2 ~ / : / = - ( l - g ~ ( a -I +a~'2/2),
where we have introduced the abbreviation
g - exp(a2r)erfc(~v%i
s/nee this particular function of a2r
repeatedly. The expansion of the preceding expres-
s/erafor Or about .% is
Or "" Oc +~c(r-rc)-a(1-gc)~2/2,
where
Oc = l + 2 ~ / r ¢ / = - a - l ( l - g c ) ,
and where & denotes the function g evaluated at
a2r = a ' r e . For small values of a2rc, Eq. 12
n~b_~,ces~_to
O. -~ 1 + arc,
and for large values of a2rc it becomes
Oc ~- 1 + 2 X / r J : : .
Equation 12 shows that if a is of order unity..
then 0c - 1 becomes of order unity for rc of
order unity. Equation 13 implies that for a << 1,
0c - 1 wil become o f order unity for r~ large, of
onter a - l ; E q . 13 is valid in this case becanse a2r~
is small of order a. Equation 14 imp:lies that for
a >> 1, 0~ - I becomes o f order unity for r~ of
order nnity; in ~ ~ a2rc is large, of order a 2,
and s o E q . 14 is valid. Thus, we shall assume that
r~ is of order unity for a > 1 and of order a-a
for a << 1. From this observation, we can estab-
l~h the bounds
I<<A<<~,[(I+a)E'] -l exp(Eg,
which me ~ ~:. ~atisfied for the present analysis to
be ~ The first condition assures that the
chemical heat-release rate can become l~rge
A. LINAH and F. A. WILIJAMS
enough compared with the radhi~nt heating rate to
c-~_,_~ tl~nnal runaway (see Ref. 1), and the
second that adiabatic thermal explosion with
neElk,ifbte p r e h e a t i ~ which requires a time
rr = (AE~ - l e x p ( E g , does not precede ignition.
For small vahles of a 2 r , unla~ zz~ is sinai]
compared with unity, Eqs. 8 and 9 reduce to
(10) 0r ~ 1 + aTe-*~', (16)
which simply represents absorption with negligible
heat conduction; there is a layer at the surface in
which the order ~ does not exceed the order of
, where transient heat conduction is of impor-
(11) tance. For large values of a2r, unless a~ is large
compared with unity, Eq$. 8 and 9 reduce to
0i -~ 1 + 2 q r / = - ~ - a - I e - ' ¢ , (17)
(12)
which d e s c ~ e s absorption and heat conduction
with a negli~'ok ~ t term in Eq. (6); for
large compared with the thickness a -1 of the
absorption h.~r, e.g., f ~ ~ of the order of ~-r,
t ~ t heat conduction becomes important and
(13) absorption negligible. Only in thi~ 1~nit of large
a2r does one retrievethe inert ~ l e m of Ref. I;
specifically, an expansion in the artifici~.,small
parameter a -x produ~-s an inner problem whose
solution is
(14)
0! = O/(O,r)-~-a-l(e -'~- D
and an outer problem d e s c n ~ l by the transient
heat-conduction equation with initial and
boundary conditions 0t (~, 0) = 0t(-, r) = 1,
0t~(0, r) = - 1, the last condition being ob-
tained from matching with the first term of the
itmer expansion.
4. Establisimtem of a Regimes
it is -dear from the preceding disc~t~clon that under
the conditions of tim problem analyzed in the
previous paper, the thickness o f the absorption
(15) layer is of order a - 1 . Since the thickness of the
reaction layer was found in this case to be of ordei
1 / Z ' , R fonows that the l~evlous analysis is
applicable only if a >> E ' .
When a becomes comparable with g ' , during the
RADIANTIGNITIONOFA REACTIVESOLID 89
Um~fion ~ a surface h ~ w ~ ~ e l ~ in ~ term of an e x t r a , on
~ r e a ~ m and heat c o n ~ o n occur m a
~ o n whcee t e m ~ r a ~ field is m f l ~ n ~ d by b = b~ + 8b~ + - - . (19)
~th ~so~n, and a new ~ h w ~ be
n e e d . S ~ h an ~ n ~ is ~ e n in ~ f ~ o w ~ of the quantity

T ~ k ~ ~ u e ~ of E" are s u ~ ~ t when a

becomes small compared with E', the next inter- b-In
f wr.)~'4(kE.
/~2\1/2
) ,2o,
esting case occurs for ct of order unity. In this case \o /j
rc is of order unity, and therefore ~ r c is not
large. It follows that the inert problem of the Since a is large, in these expressions 0 c w~l be
pre~ous paper is not applicable here and Eqs. ~ e n by Eq. 14. Thermal nmaway occurs at
11 and 12 must be considered in their entirety. An a = b~ to ~n ~ b ~ c orders in $ , and the higher
ana.~is o f ' h e ~ f i o n stage for this ~ s e will be c n e m ~ e n ~ b ~ , . . . , in Eq. 19 must be ~hosen to
in Section 6, wl~re it will be found that a minimize the severity of the singularity at o = b~.
~k zone of transient heat conductkm and Such a prescription is quite remonable on physical
dh;x~buted chemical .--ea~_~ondevel.el~, instead o~ grotmds, since its effect is to extend the validity of
~ o - m n e structme obtained in the previous the expansic~ procedme, to as late a time as
~ and m ~ c f i o n 5. pothole.
It w ~ be seen in ~ f i o n 7 that fl~ n ~ d ~ of We anticipate a two-zone structure, with the
~ f i o n 6 for ~ a me~ ~ ~ of~fion 5 thickness of the inner zone, in which absorption,
for ~ n ~ ~ / ~ a n d ~ t ~ ~ off,on 6 for heat conduction and chemical heat release occur,
e describe a therm~ ex~omon ~ t h o u t he~t of order 1/~.T, and with the thickness of the outer
c o n ~ f i o n whkh occurs in a ~ a ~ d h~er of zcme of t r s ~ , n t "heat ~ n & ~ ¢ ~ , -wb,~,¢ the
thickne~ a-S/4/V~. This e x ~ o n ~ an chemically h~erated heat is d~Ln~uted, of order
~ f i o n ~ - ~ that ~ u3mew~t ~ o ~ y w~h 1/~/'~. As in our previous paper, both zon~ axe
a ~dt ob~ned'qm~ ~ by u~ng Eq. 1 ~ thin compared with th~ width of the inert-~age
• e heat-con~cfion term deleted entirely, even tr~n~ent heat-conduction region, whose thickness
• o u O the ~tter ~ w o ~ d p r ~ u c e Eq. 16 is of order umW in ~ . Thus, we introduce
and t h ~ b y i ~ o r e the con~cfion ~yer flint e x ~ tuner spatial ~ l e
at ~ e mfface for sm~ll a ~ d n g the ine~ ~ .
From fl~ present ~ x ~ m n it is ~ r e n t ~ t x = (~)~ (21)
t ~ n ~ d ~ w ~ w ~ be ~ d in ~ o n s 5, 6,
7 w ~ cover the compkte ~ of~ of-,
~ t h of ~ mt~ and the outer spatial variable

5. T i n S ~ for a m ~
T/
~ of the Same O n l ~
When the o n k r of a is fl~ same as that of E', we
w~te a = a s / ~ 2 , w ~ r e ~ = ~ , and as- (22)
s a m e ~ t at is of o n ~ r u m ~ . As in our ~ o u s
~r, we e x a c t to Vmd a t r ~ f i o n ~ a ~ in Expansions for 0 - Ot will be sought in the
w ~ the a ~ m p ~ fin~ ~ u ~ is forn~

0 - ~ = 82~0(x,o) + 83~(x,o) +... (23)

o. = + (1~)
koU and

where b~,whichwea.~ume t o b e oforder unity, is O - O r = B2@0(q,o) + 8s~l(~,o) + . . . , (24)

 90 A. LINAN and F. A. WILLIAMS

for the inner and outer zones, respectively. In the It follows that
inner zone the expansion for • is
¢~,o~ = - (=rc)-JJ4Gexp(~0 + o), (32)
• = & l ( , ~ O) + 82e2(x.,
~) + . - - , (25")
where G ~sa function of ea, defined by
and in the outer zone • will be found to be
exponentially small. When expanded in powers of
8, Eqs. 8 and 9 give
G -~exp[-x-a'~lexp(-axx)ldx.
(33)

= 0. + B:z[o• - / ~ - x - a ~ exp(-a:~z)] It is easy to show from Eqs. 1, 2, 18-20, 22, 24,

+01~1 (26)
and 27 that the first term of the outer expansion
of 0 obeys the equation of transient heat conduc-
in the inner variable and tion. The initial condition obtained from Eq. 3 is
00(7/,-~0) = O, and the boundary condition im-
plied by Eq. 4 is q~0(~,o) =O.Matching to the
inner expansim mquims(bo(0,o) = ~o and
in tl:~ outer variable.
Substitution of Eqs. 18-21, 23, 25 and 26 into 4~o~(0,o) = - Gexp(~o + o),
E ~ . 1 and 2 produces the inner equations
where use b ~ been made of Eq. (32). The small
time translation o* = o + In G then produces the
~o= + ~,,,,, = - 8(=r~)-U4ecp[~o + o - x
problem
-- e l I exp(_,l:r) ] (28)
4)oa.= 4)07. • ~ o ( r / , - - ) = Oo(~,a9 0,~
/-
up to terms of order ~2 and 0o~ (0, o') = - exp [~0(0, o') + o'].)

q ~ = 8"qm'~) u 4 exo[.~o + o - x (34)

- a~ l exP(-,,lxH (29) which has been sobed l~eviously in Ref. 1.

Thermal runaway was found to occur at
o" = - 0.431 - b0- It follows that in the present
up to terms o f order 8 . From the term of order problem
tmity in Eq. 28 and the boundary condition for
#'o implied by £q. 5, we conclude that .~0 = be - Inn. 05)
depends only ou o . Equation 29 and the initial
condition for e implied by Eq. 3 dmn yield Equations 19-20 then show that, to lowest order
inS,
el = B-1"~rc)V4exp[-x

-- a~: exp(--alX)]C e...xp(~o4- o ) d o . (30)

(36)
The terms of order 8 in Eq. 28, along with the
boundary conC~ion ~t=(0,o) = 0, implied by This expression for A differs from that obtained
~ . 5, then y~e,~ previously only in the presence of the factor G. It
can be shown tim*.G can be expressed in terms of
~I~ = - ( ~ c ) - u 4 exp(~o the incomplete Gamma function as
+ o)Fexp[-x - aF t exp(-alX)] dx. (31) G = a~Ir{a[1)y*{a~1 , a l l ) , (37)
ao
RADIANT IGNITION OF A REACTIVE SOLID 91

where y* is a single-valued analytic function of 10

l
both of its arguments. Tables and graphs of r
and of y* exist [7]. However, to obtain accurate
values of G, it is better to perform numerically
the integration in Eq. 33 than to use the pub-
lished results. We have calculated G as a func-
I • I
tion o f az by numerical integration and show
the results in F ' ~ I. Also shown in Fig. I is
A(m'c)I/4a-I/2e3q~(-E'/Oc) as a f u n c t i o n o f
according to Eq. 36. By u ng last .5
curve a~d Eq. 14 for 0c, the nondimensional
rate constant A can be obtained if a, E" and G
rc ~e imow~ 2
As at -. ~, G approachesunity, and we recover
the result of our previous paper. An ~n~lytical
exl~aSlon for large values o f al, obtained by twny.
Eq. 33, is shown in Fig. I, from which it can be
seen that the expansion is qmte accurate for .05
al -> 1 . For ax of order un/ty, G is of order
unity, and Eqs. 1~ and 26 show that approxi-
r t ly O, lOU = - o Ig'U, in which an .02
2 5 t 2 5 l0
o,
factors are of order unity except E', which is
Fig. 1. G and the n o n d i w n a l rote c o ~ t for
large. Therefore for al of order unity, the "~nition
ignition,as functionof aO~2/E; for= of ord~ E'.
time d..~ends only weakly on G, and very little
difference will be found between the i~ition time
obtained from Eq. 36 and that derived in the
previous paper. On the other hand, for low values
of a b we obtain from Eq. 33

lnG ~ - a [ "]' + ln[Vra'-~I'~'~+ ~ + O( aV:~-l)], (39)

(3S)
where co, which is of order unity, is the first term
which 5ecomes large in magnitude and thereby
of the expan~on
produces a result which differs appreciably from
that o f our previous work. It can be seen from Fig.
C = CO + ~2C1 +''', (40)
1 that for at <- 1, the numerical results agree
quite well ~ h the asymptotic expmssi~ ~ by
of the quantity
Eq. 38. As ~,~ becomes small, a tends to become
o f order unity, and the analysis given in the
following section becomes appropriate. c -- In[a'Ig'~IAexp(-E'/Oc)], (41)

6. Tmmition Stage f o r a of Order Unity

To analyze the transition stage for a of order
unity, we follow a procedure analogous to that of
the preceding section, except in that we introduce
only one new spatial variable, since a s/ngle zone
occm3. The appropriate time variable for the
transition stage becomes
in which 0~ is Oven by Eq. 12. Thermal nmaway
occurs at t = co to all algebraic orders in 8, and
c't, c2 . . . . are to be chosen to min/mize the
order of the singularity at t = co. The thickness
of the zone in which thermal zunaway occun is
def'med b y y of order unity, where
92 A. LINANand F. A. WII.LI.~[S

tion, derived from Eqs. 3-5, are

Y= k~/ (ctgc)ll2~" (42)
Xo (00, t) = X0 (Y. - oo) = Xoy(0, t) = 0 . (50)

From. this equation and Eq. 10 it is seen that the
thickness o f the nmaw.-v zone in ~ is of order
l/,v~: w ~ n a is o f onter unity or later, and
larger wben a is small
An expansion for 0 - 01 ~ be s o ; ~ ' , L". *..he
form
0-6 i = 82Xo(y,t)+84Xl(y,t)+---,
and for • in the f,-mn
The problems for Xo, given by Eqs. 47 and 50, is
well set, aoC- there is only one parameter, F, but
the :olution must be obtained numerically. By
ana!ogy with Eq. 34, it is dear that the solution
w ~ exhibit thermal runaway, at time t = co,
which will depend on F and therefore on a2rc. If
a2% is of order unity, then F and c0 w111be of
(43) order unity. From F.qs. 40 and 41, the formula for
A, to lowest order in B2, becomes

A exp ( - ~ ) = ag~ exp(co), (5D

• = 8 2 , , ) d y . t) + 8(o.)2(y, t) + . . . . (44)

The relevant expam/'on of 0:, obtained from Eq.

ll,is in which gc is related to a2rc according to Eq. 10.
Equation 47 has been integrated nung'rical~- by
a Crank-NichoLvon lint/Licit ~ employing
0s = 0. + B ~ t - c o - F y ~ + O i ~ I, (45)
q-,m'lhwarizatic~, of the source term. For large
negative value e f t , the solution is
where

(1 - g~)
V~ e ~c 4~e:FeY erfc 1 x/-fy)
~" - (46)
2g~
+ e-Y erfe ( 2 ~ - V~y)]- (52)
Sul~d-tution of Eqs. 39-45 into Eqs. 1 and 2
yields, up to terms of order B2,
In performing the numerical integration for other
Xot - X o ~ = exp (X'0 + t - F y 2) (47) values of t, tle quantities e t and err(y) were
employed as independent variables, to ~oduce
and convenient step sizes and finite boundaw loca-
tions. By completing the numerical integration for
~u = W "~ e a P ( X o + t - F y ~ . (48) various values of F, the dependence of c o on F
was obtained. Use of this result in cm0anction
The sohnion to Eq. 48, consistent with the initial with Eqs. 10 and 46 provides c o as a function of
conc~ion i m ~ ~ Eq. 3, is a2rc. From this result and Eq. 10, the quantity
(A/a) exp(-g'/Oc) can be calculated as a func-
wl = B -t ezp(- tion o! a2rc by means of Eq. 51. Equation 12
e z P [ X o (y, t) ÷ t ] d t ,
shows that a(Oc - D is also a function of a2rc.
(49) various functions are plotted in Fig. 2.
A graph of re, as a function of A, for various
which can be evaluated only a f a r *o(Y, O is values of a and g', is shown in Fig. 3. Tbe curves
found. Equation 47 fm X0 is the heat conduction in Fig. 3 labeled log a = - have been given earlier
equation with a distributed nonlinear heat source; b y n ~ [8] and are in agrcement w i t h E q . 36,
the initml and boundary conditions for this ¢qua- while those '.abeled log a = - 0o are limiting lines
RADIANT IGNITION OF A REACTIVE SOLID
//-
.i
Fig. 2. co, F, gc, n o ~ n m l ignition temperatu~.
and nondimendonal rote constant, as function of non-
dimensional ignition time, foz • of o~lex unity.
that correspond to the adiabatic thermal-explosion
time defined after Eq. 15. Figure 2 was used for
calculating all other curves shown in Fig. 3, except
those with log a = - 2 and -4, for which the
~ p l e r limiting formula given in Eq. 56 was em-
ployed. From Ftg. 3 it can be inferred that for
given values of A and g ' , the ignition time ob-
tained from Eq. 51 or from lrtg. 2 is appreciably
longer than that cor~sponding to Eq. 36. This is
primarily a consequence of the fact that the time
required to reach any given value of 0c according
to Eq. 12 is longer than that required according to
Eq. 14, when a is of order unity or smaller. Since
the expansion parameter here is 8 2 instead of 8
as in the previous work, for typical values of E" we
expect the error in the ~alue of A given by Eq. 51
to he less than 5% instead of the ~ error
expected from Eq. 36.
7. Limit Cases
Further iusigh-', into the character of the depen-
deuce o f i~-nition time upon a can be gleaned by
investi~ting limiting cases of the previous anal.
yse,.
Co,.'..'..'..'..'..'m~fast the form taken by Eq. 47 for large
values of a. From Eqs. I0 and 46 it follows that
F ~, aV~e/'2 Ls a -~ o,, and therefore the coefS-
dent of _y2 in the extxmential in Eq. 47 becomes
large. It follows that for y of order unity the
source term in Eq. 47 is n~li~'ble; an inner layer
develops, where y is of order F -t/2, in which the
source term must be considered. In this inner
zone, the time derivative becomes small, and a
93
0 .. ~ "w
-4o -3o -zo -Io o
10~ott/AI
Fig. 3. Hondimen~ml ignition time as a function of
nond~en.~nal rate const;mt for various values of the
nondimensional activation enex1~ and absotbtivity. The
numbers that label the c u r v e s axe lo~s oa.
problem emerges that resembles Eq. 28, with
the exception that the x dependence inside the
exponential becomes proportional to - x 2 instead
of - x - a~ I ex~(-alx). H,~wever, in the
limit al -, 0 , this last quantity becomes
- a i I - alx2/2, which, with aii absorbed into a
time translation, is of the same general form as
- x 2. The respective length scales for the inner
zones become the same, both being ~ j a E ' ; in
the ~ coordinate. In addition, from Eq. 42 it can
be seen that the length scale for the outer transient
heat-conduction zone obtained from Eq. 47 is
(nrc)b'4 ( 0 2 / E .) 1/2, which is in agreement with
the length scale for the problem de'f'med by Eq.
34, as can be seen from the defmition of ~ t#ven
in Eq. 22. Finally, we note that Eqs. 36 and 5i
become identical in the limits al -, 0, a -, oo, both
reducing to
= (wrc)-l/=(2a rcV~-~V2exp(bo ), (53)
since the solution to the two-zone problem
derived from Eq. 47 reveals that Co --.b0 +
ln(2a~ I/2 in the limit. In Eq. 53,
94
Eq_. 14 ~ m be. u...*~_for fl~. From these observe_
tions, it becomes clear that the results of Section 5
for small values o f at agree with those o f Section
6 for large values of~, and that therefore solutions
are now avmlable which extend from a of order
unity to a large compared with E'.
Consider next the form taken by Eq. 47 for
small values o f a . Since Eqs. l0 and 46 show that
F -. aVrclb as axl'~r¢-.. O, and since w e h a ~
seen that aV~'~c-.O as a -.0, it follows that the
coefficient of _y2 in the exponential in Eq. 47
approaches zero as a - - 0 . Therefore the source
strength chm~,~ negl~'bly across a layer whose
t h i c ~ ; (~/~-"a)I/2 iS such that y is of order
unity. To obtain a region in which the source
c~m~es~ appreciably, we ~ust employ a spatial
variable with the larger characteristic t h i c ~
(zr/¢c) t/ 4q82/E') l/ 2/a. In this variable, the space
derivative in Eq. 47 is neglip~ole in compadsm
with the time derivative, and we obtain a simple
thermal explosion with n~ii~ifble heat conduction,
viz., X0t = exP(X0 + t - F y ~ , whose solution
No = - ln(1--et- F~'~)" (54)
fL~Stdi~rge~ at t : 0 for y = 0.
Equation 54 satisfies all of hhe initial and
boundary conditiom given in Eq. 50, as well as ELI.
47, provided that heat conduction is negligible. It
im~ that in the limit of very low atr~rbtivity
Co = 0. A morn formal calculation reveals that
Co ~ 2 F - 2 + 4 ~,2F2 + O(F3 ) ($5)
3
as F approaches zero. Equation 55 can be used in
conjunction with Eqs. 10, 46, and 51 to provide
~m ~ forA that is valid for small values of
a. When use is made of Eq. 13 for 0o it is found
that the limiting_ form of this expresdon for very
small values o f e can be reduced to
= -l]/a. (s6),
This same remit can be derived by using only Eqs.
I, 3, and 4, and by neglecting heat conduction
~hroughout; such an amdysis produces Eq. 16 for
the inert stage, in physi:al variables, Eq. 56 c~re-
A. LINAN and F. A, WILLIAMS
sponds approx~nately to a.-- ~ t i o n time that is
inversely proportional to the first power, instead
of the square, of the incident radiant flux. These
results extend the range of a, for which ignition-
time formulas are available, to values that are small
compared with uni~y.
8. C,enetaliz,tiom ~mi Disctmion
In principle there is no difficulW in extending the
results to arbitrary histories of radiant flux as a
function of time, the only restriction being that
the time rate of change of flux not be exceedingly
large at the time of ignition. Since the tranfition
stage is short compared with the inert stage,
time-varying flux does not affect the structure of
the transition stage, provided that in the time
variable of the transition stage the logarithmic
derivative of the flux is small In the appendix, we
presume that this condition it ~ f i ~ d and indi-
cate the changes that must be made in the
preceding ~m~lyses to account for arbitraw flux-
time histories. The results demonstrate that the
only new problem requiring soluficm is the ~ . r t -
heating problem whose solution is expressed by
Eq. 8. From the new function [(~,v) that
corresponds to the modified inert-heating prolv
lem, the quantities 0¢ ~nd ~c - OtflO, re)
both be exp _re~_~le in tenm of re. It can then be
shown (see Appendix) that the appropriate gener-
alizatiom o f Eqs. 36 ~ d 51, respectively, to
conditions of arbitrary flux-time history, are
A exp ~-¢ ~--~-¢2
/ G-lexp(b°) (57)
and
Aeo ',8,
with Co in the last expre~on related to the n e w F
def'med in Eq. (A4) by Fig. 2. If gc is interpreted
as ~ c / a and the ~ n t a l scale is interpreted as
an arbitra W parmnet~ that is not related to a2rc,
then all of tic cur~es in Fig. 2, except for
a ( O ~ - D, remdn valid fm arbitrary flux-time
histofie~
Because e f its more general applicability, a graph
R A D I A N T I G N I T I O N O F A REACTIVE S O L I D 95

ignition and consider instead the total amount of

| ~
-I- l
I" I t" "1
'I
'
l-J heat absorbed by the solid up until the time of
ignition, as some authors have attempted to do
(see some studies reviewed in Ref. 9).
If there is a distribution of the absorption
• 20 /
coefficient over the frequency v of the incident
radiation, then in a fairly obw.'ous notation the
radiant heating term in Eq. 1 becomes

o lvav e x p ( - a v ~ ) dv .
.05 a~2!E0
~1~ smal~
This modification will require new solutions to be
.02 obtained for the inert-heating problem, although
these solutions can be constructed by superposi-

.00~
.01 , I [ i
.002 ] I ~ , I ,
0 2 , .4 ,6 .8 tO
(~el( 0£1al''~ ~ (-£'/Ocl
F'g. 4. C¢~pb of ~nifion conditions for in-depth absorp-
tion with an arbitrary flux-time history.
of~/a, as a function of
(Ala) (~ /a)- ~ 2 e x p (- E'IO~);
for various values of a0e2/E" , as obtained from
Eqs. 57 and 58, is shown in Fig. 4. To use Fig. 4,
one mint first solve his particular inert heat-
conduction problem to obtain 0e and 0~ as a
functio~ of r c.
It is remerkabl¢ that only the heat flux at the
time o f ignition (through the nondimemionaliza-
tion), the inert surface temperature at the time of
ignition, and the inert time rate of increase of the
surface temperature at the time of ignition affect
the "n,nition conditions of any given solid material.
M2ny of the detafls of the history of inert heating
-,av unimportant. However, in general., it is not
p o t ~ l e to eliminate any of the tluee essential
~ambles stated here in terms of other quantifies.
For example, one cannot ignore the inert time rate
of increase of surface temperature at the firae of
I?
tion, since with O! - 1 + . Ov d r , each Ov sepa-
rately satisfies Fxl. 6 with
0~(~,0) = 0,,(~,r) = 0~¢ (0;-r) = 0
I The analysis of the transition stage will depend on
whether the incident radiation corresponds princi-
pally to av = O(E') or to a v = O ( D . The
transition stage that develops will be the one that
produces the shortest ignition time. Since the
ignition time for a v = O(E') is appreciably less
than that for a v = O (D, ignition will occur in the
regime of Section 5, unless only a small fraction of
the incident energy falls within that regime. A
generalization of our analysis of Section 5 shows
that in this case Eq. 57 is again obtained, provided
that the more general definitiun
(sg)
is used for G. In the lowest ~proximation, Eq. 57
then reveals that the facto~ G can be ignored,
provided that only the portion of the hucident
energy flux corresponding to the r ~ ,
at, = O ( E ' ) is employed in the ~ that is used to
form the nondimensional time which appears in A
and in 0~.
96
If almost al~ of the incident flux corresponds to
" v = O ~D, the,a :6~ition occurs ia ~ e regime of
Sectio~ 6, and it can be shown that Eq. 58 remains
applicable, with co related th,ough Fig. 2 to the
gener~liTed quantity
F-\_(!: lv'cav~--0i , (6o)
where the int~-al must exclude those values of z,
for which a j is of order ~ ' o r larger. After solving
the inert-heaticg problem, one should calculate
two ignition times, corresponding respectively to
Eqs. 57 and 58 with the generalized definitions of
F and G, and select the shorter time, to complete
the computation in the presence of a frequency-
dependent absorbtivity.
It is of interest to note from the analyses leading
to Eqs. 57 and 58 that for sofids of practical
dimensions, the results given here are independent
of the assumption of one-dimensional heat flow.
only requiretr~,nt is that the narrow surface
layer in wh/ch reaction ~-st occurs must be
approximated well as being one-dim=ntionaL For
the inert stage, s three - 2 h ~ i o n a l tra,'ment heat-
cmduct~n problem must be solved. From this
solution, for each point on the so]kl mrface, the
remits given in hhit section can be applied to
calculate ~ conditions. The thota.est ignition
time so obtained will represent the ignition time
for the ",hree-dimmsioml mild. The degree of
generality that can be achieved with ~ s e through
asymptotic methods is imp~e~ve.
The:e results can be used to suggest an improve~
experimental technique for radiant ignit/o~
studies. The heat flux at the time o f ignition and
the histo W of surface temperature at the point of
ignition are both amenable to ~ t . From
.~'dl measurements, Oc and 6~c can be obtained,
and F~g. 4 can then be used to calculate chemical
rate constants from a few data points, provided, of
creme, that ignition occurs through a condemed-
#ase reactioo.
Appetdtx: Variable Heat Flux
To m . d i f y the obserratiom o f Section 8 in more
explkJ! terms, we fnst note that one can write
A. LINA~ aria F. A. ggLIJA]~.~
l(r) = J(a2r), and that this ~ctor will then
appear in the last term of Eq. 6. It then becomes
clear that we can write the solution for 0z in the
form given by Eq. 8, but the function [(u, v) wRl
depend on the function dfv) and will no longer be
given by Eq. 9. It follows that Eqs. ll-14wfll no
longer be valid. However, it will still be possible to
expand Oi(O,r) about re, obtaining
O,(O,r) = 0c + 0 ~ . ( r - ~ ) , (AI)
where 0~ - 0tr(0, to). We shall see that the
dependence of the exl~mion of 01 about ~ = 0,
needed in the transition-stage analysis, can be
expressed in terms of Oc and 0~. Therefore
formulas for ignition conditions can be obtained in
terms of the two parameters Oc and 0~, which
then must be expressedin terms of rc from an
inert-heating analysis to obtain the ignition time.
Consider first the case analyzed in Section 5. In
Eq. !8, the factor (=re)- 1/2 must be replaced by
0"_.:which is assumed to be of order unity. In Eqs.
20 and 22, (re'c) i/4 is replaced by ~ - t / 2 . W i t h i n
the context of the asymptotic analysis, ignition
cannot occur with negaiive values of 0~; if, for
example, the flux is dis.~ntinued before ignition
occurs, so that the surface temperature hegira to
decrease, then ignition never will occur. ~ t h x
and o as independent vm-iables, the ~ne-derivat~-e
terms in Eq. 1 and in Eq. 6 are of higher order
than the space-derivati~ terms, and to lowest
order Eq. 6 becomes
Olz.~B - 2 = - a i J e x p ( - a s x ) , (A2)
whose solutiou is seen to be gi~n by Eq. 26, when
use /~ made of Eq. (AI) and the definition
d(a2rc) = L Thus, Eqs. 28-32 still follow, with
(~rc) v4 replaced by 0~ -1/2. The problem ex~
pressed by Eq. 34 is ~ obtained, and the
solution, corresponding to Eq. 36, is given in Eq.
57.
Consider next the case analyzed in Section 6.
When the modified form of Eq. 6 is solved for
01~£(0, %), and use is made of Eq. AI, then i.'. is
found that Eq. i I becomes
01 = 0~ + 0~(r - r~ + 2 • (A3)
RADIANTIGNITIONOF A REACTIVESOLID
In Eqs. 39, 41, and 42, we replace age by 0~, and
from Eq. A3 we then obtain Eq. 45, w~-th
F - (a- (A4)
From Eq. A4 it might appear that a new condition
for the generalized analysis to be valid is a > ~ c ,
but in fact this condition is assured by the
inert-heating problem. The remainder of the anal-
ysis of Section 6 proceeds unchanged, and the
generalization of Eq. 51 becomes Eq. 58.
The work o f the first author was partia~y
supported by the Air Force Office o f ScientiF~
Research, through the European Of]'~e o f the
Office o f Aerospace Research, United States Air
Force, under Contract No. F61052-69-C-0036.
The secor~t author wishes to acknowledge fellow-
ship support from the John Simon Guggenheim
Memorial Foundation and addi~'onal supFort from
the Air Force O ~ o f Scient~ffc Research, Office
o f Aerospace Research, United States A ~ Force
under Project THEMIS Contract
F44620-68-C~010. ICe v,.~h to thank Mr. V.
Torroglosa f o r carrying out the computer calcula-
tion of co.
97
References
1. lJ~n~ A., and ~ F. A., Combustion Science
and Technology 3, 91 (1971).
2. OMemfller, 1". J., and Summed]eid, M., "Radiative
Ignition of Polymeric Fuels in an Oxidizing Gas,"
AFOSR Scienti~c Rept. No. /~FOSR 69-2147TR
(1969); Oldemfller, 1". J., Ph.D. Thesis, Princeton
University (1969, unpublished).
3. Oldemiiler, T..~, and Summm4iekl,M, AL4A Jr. 6,
878 (1968).
4. Evan~M.W., Beya, lLB, andMcOmey, L, ~ O:en~
Phys. 40, 2431 (1964).
5. Wise, H, Imuni, S. H., and McOdley, L., Combustion
andF~me 11,483 (1967).
6. Boehringer, J. C., and Spm~ler, IL w AIAA J. 1, 84
(1963).
7. Abrmnovitz, ~ and Stegan, I. A., Handb~,k of
Mathematical Function~, Dover, New York (1964), p.
261.
8. Bradley, H. H, Jr., Combustion Science and Tech-
nology 2, 11 (1970).
9. Merzhanov, A. G., and Ave~son, A. E., Combustion
and F/ame 16, 89 (1971).
(Received May 1971; revised t~r~on
received August 19 71 )