Flammability of Polymers

C. P, F~mMORE and F. J. M~TXN

General Electric Research Laboratory, Schenectady, New York
(Received Julr 1965)

The/tammability of polymers has been measured by determining the oxyge~ content of the,
atmospheres just capalde o/ burning them. The effects of the ]ollowing three commonly used
al,,ents for reducing l~ammability have been compared in polyethylene. (I) A moderate degre,~
o] inhibition is con/erred by antimony trioxide plus su~cient chlorine: the maximum effect
requires about 0.01 antimony atoms per Ce group in the polyethylene, and develops just as
well when chlorine~antimony is six as ¢~hen it is twenty. The lull effect o] the antimony does
not develop when chlorine~antimony is two. (2) Various phosphorus compounds are all
o/ similar effectiveness per atom o] phosphorus added to polyethyh;ne. (3) With su~cient
substitution o/ hydrogen by hatogen in polyethylene, a greater inhibition is achieved than
~s possible by (t) or (2). We have no evidence how, (1) and (2) inhibit burning, bat present
evidence is that in (3) a large substitution of chlorine works mostly by affecting the
degradation of the polymer rather than by inter/ering with gas phase flame reactions.

Introduction flow orifices, mixing the gases, and feeding

H. G. WOLFHARD and co-workers1,2 compared
the flammabilities of some light fuel~ l:y finding
the limiting mixture of air and nitrogen in which
each could just hum. They described the mix-
tares by their oxygen indices,
Oxygen index = n = [Oz] 7 ([O~] + IN2])
For air, n=0'21.
The smaller the limiting n, the more flammable
the fuel since the more its flame survived cool-
iiJg by nitrogen. Some of their limiting indices
for diffusion flames at atmospheric pressure
were:
n = 0.054 for hydrogen
= 0.056 for an equimolecular mixture of hydro-
gen plus carbon monoxide
---=0.071 for monomeric formaldehyde
=0.183 to 0"134 for benzene and normal
paraffins containing 6 to 10 carbon atoms.
We have correlated the effects of substituents
and additives in polymers, particularly in poly-
ethylene, by their oxygen indices. So-called
'~lf-extinguishing' materials, which do not
continue burning in air after being ignited,
evidently have n greater than 0,21; and the air
must be enriched with oxygen to investigate
them.
Experimental
Known atmospheres were prepared by metering
air and either nitrogen or oxygen through critical
185
them to the bottom of a vertical glass tube 60
cm long by 8.4 cm diameter. The gases passed
through a bed of glass beads at the bottom of
the tube to smooth their flow, and then up the
tube and out of the open top. A sample of
polymer, usually a strip 0-6 cm x 0.3 crn x 8 cm,
was clamped with its long dimension vertical
in the centre of the tube; it was ignited at the
upper end with a hydrogen flame; the igniter
flame was withdrawn and the atmosphere found
in which the sample just burned for its entire
length.
Polyethylene, polymethyl methaerylate,
Teflon, or other substances which burnt without
any residue, had steady flames which gri~w
smaller until they extinguished in atmospheres
containing not quite enough oxygen for con-
tinuous combustion. In contrast, silica-filled sill
cones had little visible flame; a zone of glowing
silica worked through the specimen, or ceased to
propagate and cooled at extinction. Other
materials went out less smoothly. Polycarbon-
ates and polyvinylehloride charred as they
burnt and seemed to smother their flames in
char. Polyethylene containing antimony oxide
plus chlorine formed a mobile liquid running
over the surface as it burnt; the flame flickered
badly in the neighbourhood of the limiting oxy-
gen index and ortega remained large, if unsteady,
until it extinguished abruptly. Polyethylene
136 £. P. Fenimore and F. J. blsrtln Vol. 10

containing phosphorus compounds had a The burner maintained a sphere of liquid fuel
similar flickering flame and the ap~arance of 3.8 cm in diameter through which kerosine
being blown out at extinction, thouga it was not could be pumped independently of the amount
really blown out by the flow of oxidant because vaporized and burnt. When more fuel was
the n value was independent of thL, flow. pumped through the sphere than was burnt in
A few indices are reported at re]uced pres- the surrounding flame, excessive heat was
sures, They were obtained by burn ng samples drained from the flame which became small and
in a bell jar in which constant p:'essure was blue instead of yellow and smoky.
maintained. 0.6 .... ,--'
Most sample strips were cut from materials
milled together and pressed by Ralph Beers of
this laboratory. Our interest was in £ ammability ¢I
and the other properties of the samples were ir
iI
ignored.
Indices were computed on the assumption that
air contains 0.21 oxygen and 0.7~ nitrogen. I
The n values were usually reproduci~le to about ,~ s for Yeflon
one per cent, independent of the ~;as velocity .~ CILsubstituted ,/
in the range 3 to 12 cm sec-~, and independent in potyethytene /
of strip thickness in the range 0.1 to 0.6 cm.
The only noteworthy limitation to the experi-
mental method was that some mate::ials melted F substituted
¢z
too easily and abstracted more he~-.t from the / / ~Burnt at 14 cm ,/
surrounding diffusion flame than was required / /6 of mercury pressure//
to pyrolyse the polymer. The n obtr~ined when
this happened was too large, as c~.n be illus-
0'3 ,/ //
trated by the following example. / / ~ / / 2 added to /'
/ polyethylene ,s/
A strip of low molecular weight polypropy-
lene (Epolene NIS) burnt with a light blue
flame which rode down the specimen on a wave
of molten polymer and left the ce:atre of the 0"2 ~t , . - o - ~ p ~, containing
e.~ HCt or el=
strip unconsumed. The n--_-0.23 r;quired for
such burning was high because heat was lost to C2H4,C~H3C( mixtures
give the molten wave. The same ~ ~,terial cast
about a ceramic wick and burnt ~s a candle O . t L ~ s .._
gave n =0.177+0"005. That the lower index was 0 1 2
characteristic for polypropylene was shown by Halogen or sulphur atoms per two carbon atoms
burning a strip of higher molecular v.'eight poly- Figure 1. Limiting oxygen indices ]or materials con.
mer (Hercules Profax 6505) which did not taining halogens (or sulphur). All samples burnt at
undergc excessive melting and gave n=0.174_+ atmospheric pressure except/or the three points indi-
cated at 14 cm of mercury. "./'he F substituted samples
0.002. We avoided excessive melting with other ~,Jere: polyvinyl [luoride, DuPont Tedlar 200 SG 40
materials by using only high molecular weight TR; polyvinylidene fluoride, Kynar," Teflon /tom
polyethylene, polymethyl methacrs'late, etc., stocl¢. The CI substituted samples w~re: chlorinated
Madex-50, or mixtures of Marlex-5002 with polyvinyl
and by excluding polymers from consideration chloride; polyvinyl chloride, Geon-IOI or -404; poly-
which would not burn down properly from the vinylidene chloride, Dow Saran 281S905, The, two
ignited end. For example, we did not find a sulphur points were obtained by milling elementary
sulphur into Marlex-5002 and pressing. The two lower
proper end-burning specimen of poly-cx-methyl- curves are for gaseous mixtures o/ ethylene and
styrene. i:hlorine compounds burnt as di//usion /lames
A related instance of excessive he~t loss from
a diffusion flame was described by l?J. B. SPA~,- The measurements are given in the tables and
OING"~ who burnt kerosine on a special burner. graphs. Figure 1 also shows indices for a few
June 1966 Flammability of polymers 137

gaseous fuels. These were determined by feed- Table !. Limiting oxygen indices o/vaviou, materials
(at atmospheric pressure unless o~,huwise stated)
ing the fuels through an 0.7 era diameter tube
countercurrent to the ascending flow of oxidant. Polyoxymethylene n
The flames of the gaseous fuels surrounded the Delrin 0.150 4-0.003
Celcon, O.1 cm thick 0.148 4-0.001
tip of the fuel delivery tube a t a distance which 0"3 c m thick 0.149
depended on the fuel flow, usually about I cm. Celcon at 14 cm Hg pressure 0'190
Polyethylene oxide,
Polyox WSR-35, Union Carbide 0.150
Discussion Kitchen candle 0.16
When a polymer burns, the oxygen may be Polymethyl methacrylate,
Rohm and ltaas Plexiglas 0.173 + 0.001
mostly consumed in a nearby combustion region Polypropylene, Hercules Pro[ax 6505 0.174 + 0"002
which gives heat to pyrolyse the polymer and Polyethylene
is fed by the pyrolysis products. The nearby Allied Chem-1220 0.174+0.001
Phillips Marlex-5002 0.175 + 0.001
combustion region may be a diffusion flame sur- Marlex at 14 cm Hg 0"210
rounding the polymer, as when polyethylene Polystyrene
burns, or a region of surface combustion, as 0.1 cm thick ]rom stock 0.151
0"6 cm thick, Westlatce high
when silica-filled silicones burn and a zone of temperature sheet 0"183
glowing silica works through the sample. Polybutadiene, erosslinked with
For such a model, the oxygen index might dicumyl peroxide . 0"183
Polyvinyl alcohol, DuPont Elvanol
reflect the ease or diffÉculty of pyrolysis and also 70-0.5 0-225 4-0.004
the ease or difficulty of burning the pyrolysis Chlorinated polyether,
products in the nearby combustion region. The Penton, Hercules
[.CH,,---C(CH,CI),,---CH.,---C~-]- 0'232
n =0'95 indicated in Figure I for Teflon may be Polycarbonates, various clear Le~:an
large both because Teflon is difficult to pyro- resins, General Electric
lyse and because the monomer, which is pre- [C, H.t--C(CHz).f.--C Ha---OC(O)O] 0.26 to 0.28
sumably the main pyrolysis product, is difficult Polyphenylene oxide, General Electric
[C6H.,(CH3).,--O] 0"28 to 0"29
to burn. The n =0.60 indicated in Figure i for Graphite
polyvinylidene chloride is about the iudex re- Porous carbon, PC-25, National
quired to burn carbon, listed in Table I; and Carbon 0.559 4-0.003
Carbon electrode, L 8109, National
since m u c h transient char is formed from poly- Carbon 0"635
vinylidene chloride and the material goes out Carbon electrode at 50 cm Hg 0.80
unless the char is consumed, it may be that Carbon electrode at 24 cm Hg 1"00
polyvinylidene chloride burning can be repre- Asbestos if[led polypropyletce, "slow
sented thus: burning" ]MDC-4400, Union Carbide 0.205+_0.005
"Sell-extinguishing' polypropylene,
(C2Ho.CI.~)~---+ char + HC1 (pyrolysis) ]MDA-9490, Union Carbide, probably
C h a r + O..,--->- carbon oxides (combustion) • contains chlorine and antimony oxide 0.282+0.003
Silica-filled silicone rubber, General
Electric
and that the limiting n mostly reflects the diffi- SE 9029 0'3C~3+0"005
culty of burning the pyrolysis product. SE 9036 0"326+0"005
Polyvinylidene chloride is exceptional because SE 9036at 14 cm fig 0'391
it does not give much flammahle gas on pyro-
lysis. Most polymers give off gases which are In the series [CH.,CHX-], it can be seen from
more combustible than the original material. Table I and Figure I that n increases as X is
Thus, the n=--0.15 required to burn polyoxy- changed from CH 3 or H, to CcHa, t.3 OH or F,
methylene is larger than the n quoted in the to CI. This might be the order of increasing
introductibn as necessary to burn the probable stability against pyrolysis as it occurs in burn-
pyrolysis products, monomeric formaldehyde or ing, although it is not the order according to
hydrogen and carbon monoxide, when the lat- S. L. MADORSKY4 for thermal decomposition at
ter are supplied independently to a diffusion rates thousands of times slower than the decom-
flame. It is reasonable to attribute the larger n position rates of burning polymers. Madorsky
for polyoxymethylene to the need to decompose gives C,~H~, to CH a or F, to H.
the polymer in order to get the flammable gases. We investigated how chlorine inhibits the
138 C . P . Fenimore and F. J. Martin
burning of .[CH,CHX~ as X is cltanged from
H to el; whether mostly by affecting the pyro-
lysis, or mostly by making the gas phase flame
reactions more difficult. If the lat:er were the
case, n should "t,~ affected about as strongly by
mixing equivaie~t chlorine into the atmosphere
as by substituting chlorine in the polymer. The
equivalence of chlorine in the attrosphere can
be calculated b.~i, supposing that the ratio of
chlorine to oxygt n entering the ram ~ is the same
as the ratio mixed in the atmosphere, and by
assuming some overall stoichiometry, say
[CHoCH2-]-+ 30= ~ 2CO~. + 2H~O. On these
assumptions, the n values were de:ermined for
the curve in Figure I labelled 'polyethylene burnt
in atmospheres containing HC1 or CI='. By com-
parison with the curve labelled 'C. substituted
in polyethylene', it is clear that ch orine added
to the atmosphere is ineffective.
The two lowez curves in Figure I represent
experiments in which gaseous mixtures of ethy-
lene and chlorine compounds we:'e burnt as
diffusion flames. They show that chlorine does
not inhibit gas-phase flame reactions of hydro-
carbons as strongly as it inhibits polyethylene
burning when it is substituted in the polymer.
Therefore, substituted chlorine probably in-
creases n by affecting the pyrolysis.
It is often suggested that hydrogen chloride is
split out and additional carbon-c~,rbon bonds
are formed when polyvinyl chloride begins to
decompose. An analogous splitting rmt of water
and formation of ~dditional ca:bon-carbon
bonds is possible in polyvinyl alcohd, n--=0.225,
but not in the isomeric polyethylene oxide,
n = 0.15. If the additional bonding decreased the
generation of flammable gases, the dammability
of the polymer would be decreased.
All the polymers which cont~;ined much
chlorine charred as they burnt; as did polyvinyl
alcohol, though to a s, ~tller extew than poly-
vinyl chloride, It was stated in the cxperimental
section that polymers which charted heavily
seemed to extinguish by smothering their own
burning; and one might wonder it the forma-
tion of much char is a sufficient exp'~anation
for the decrease in flam~aability in such cases.
We think not. Similar heavy chars ,to not make
all polyme~ equally no:n-flammabh;. Polyvinyl
chloride has n=0.45 according to Figure !;
Penton [-CH.,--C(CH:CI).,--CH~--O-] whid~
Vol. 10
contains almost as much chlorine and also chars
heavily has n=0.23 according to Tab!e 1. The
easier break-up of the polyether chain still tells
even in similarly charring materials; and the
decomposing polyvinyl chloride more than the
decomposing chlorinated polyether must go
through a relatively inert, perhaps crosslinked,
stage which does not readily generate flammable
gases.
Other known ways of making polymers less
flammable are illustrated by the remaining
figures and tables. Figure 2 shows that anti-
0"30
Solid circles, SbzO3 in C2H3.I Br0 0 e l
(Polyethylene plus tetrabromo benzene)
0-25 'fftg ?op . . . . ircies, SbzO3 in CzH3tC|o.2o~
/f . _ \Open triangles, in C2H394Cto.ot4 )
//'7- "~ . . . .
Squares, Sbl O3 in "~,,
CzH3n CI0 020 ~.~.
0"20
, ~ . o In C2H4 (p01yethyLene)
' I
0"150 0"5 1"0
Moles Sb203 per 100C2 units
Figure 2. Inhibition by antimony trioxide in poly-
ethylenes containing chlorine or bromine
mony trioxide (Sb.,Os) is not an inhibitor in
polyethylene in the absence of halogen, but it
is in the presence of chlorine or bromine. The
amount of chlorine present is not critical as
long as it is enough; a chlorinated polyetilylene
of composition C2H~.o.tClo.o,4has about the same
oxygen index as one of composition C~H~.~CIo.=
when 0.5 mole of Sb20 ~ is added per 100 C2
units. The composition C2H~.98Clo.o2 contains
too little chlorine to develop the inhibitory
action of antimony fully. It is commonly be-
lieved that the inhibition involves some anti-
mony-chlorine compound, and such a view is
consistent with Figure 2 although we do not
know why the inhibition should saturate at
around n=0"26. Table 2 records the results of
adding some other compounds to chlorinated
June 1966 Flammability of polymers 139

Polyethylene, none of them being as effective as Table 3. Limiting oxygen indices for mixtures ol
polyethylene (100 CsH, units) plus two moles ot
antimony trioxide.
additive (except where noted otherwise)
Table 2. Limiting oxygen indices/or mixtures o] tO0 Additives n
C~H,.sCIo.a units (chlorinated polyethylene) plus 0"5
or 1 mole of various additives None 0.175
Triphenyl phosphate 0.194 t: 0.001
0.5 mote oI n Triphenyl phosphite 0.195
nolle 0.211 Triphenyl phospMne 0.193
Bi20 a 0.216
Triphenyl methane 0.177
As.~Os 0.239 Diphenyl sulphone 0.177
Sb~,Os 0.257 Diphenyl carbonate o.177
! mole of
Dicumyl peroxide* 0.172
tricresyl phosphate 0'235 Lead dioxide O.178
Bisphenol A O"177
Sb20s 0"257
SnO~ O.227
Tetrabromo bisphenol A t 0.203
PbO 0.208
Sodium chloride* 0.179
Boric oxide, BoO3 0"175
0'25 Borax, NaaBdOr • 10H20§ 0" i80
*1 mole dtc-mylperoxidetmr 100 C,2Ht units. Two sampleswere
prepared; one ~ncsscd at 125~C (uncured), the other cured 0'5 h
Of~n circles; phosphonitrillc ehloride,PNCl2~ at 1550C.Bothgavethe ~ntc n.
l CV~wd circles;tricresyt phoc~pi'mte ]~.o *Comparewith Flsure 1. Br added as tetrabromobisphenolA does
not increase n much re..re *hallthat added as tetrabromobenzeae.
/;TwentymolesNaCIIxr tOO C2H4 units.N~21has liUl¢effectcom-
pared to equivalentC1substitutedin polyethylene.
§1 moleNa2BdO7• 10H20per 1 0 0 C2Hs units,
in the gas phase is present in about equal
amounts as unsaturated hydrocarbons, as car-
bon monoxide, and as carbon dioxide; the
0"2[ / rriph,,~yt phospl'~,te
hydrogen in the gas phase is about a third
present in unsaturated hydrocarbons, a sixth in
molecular hydrogen, and a half in stemla.
Af AS203
From these experiments, the oxygen index
// $b2 03
seems a useful measure of polymer flammability.
hJ Bi2 O3 I It is easy, reproducible, and numerical. Flame-
2'5 5!0 7"5
Moles additive per 100 C2H4units resistant plastics are usually called 'slow burn-
Figure 3. Limiting oxygen indices for polyethylene, ing', 'self-extinguishing', etc., but these quali-
Marlex-5002, containing various additives tative terms could be recast numerically. For
example, Table ! shows that 'self-extinguishing'
Figure 3 illustrates the action of phosphorus plastics can have n as small as 0.282 ('self-
compounds in Polyethylene, and additional data extinguishing' polypropylene) or 4).26 to 0.28
in Table 3 show that triphenyl phosphite and (polycarbone.tes), while 'slow burning' plastics
triphenyl phosphine are as effective as ~ne phos- can have n as large as 0.225 (polyvinylalcohol).
phate. The similar effects of different phosphorus Perhaps any substance for which n is larger
compounds prompt the question whether they than about 0.27 can be considered 'self-extin-
might all form the same species at the surface guishing'.
of the burning Polymer. For example, could References
reduced phosphorus species be oxidized at the t St,~tMc,~s, R. F. and WOLFHARD, H. G. Combus-
surface? This is not impossible; some prelimin- tion ~ Flame, 1957, 1, 155
ary analyses of gases collected within the diffu- 2 thLL, A. R., McCoUBREY, J. c. and WOLFHARD,
H. G. Combustion "6, Flame, 1957; !, 53
sion flame, 0.1 cm from the suface of burning 3 Si'ALDtXG, D. B. Some Fundamentals of Combus-
polyethylene, show them to he pretty well oxi- tion, p 126. Batterworths: London; Academic
dized although the concentration of molecular Press: New York, 1955
•t MADOJ:SKV,S. L. Thermal Degradation o[ Organic
oxygen is very small at this point. The carbon Poly;,,ers, p 293, Interscienee: New Yorh, 1964