FIRE PERFORMANCE OF FLAME RETARDED

POLYMERS USED IN CONSUMER ELECTRONICS

Matthew Bundy* & Thomas Ohlemiller

National Institute of Standards and Technology, Gaithersburg, MD, USA

ABSTRACT

An experimental study was performed to compare the bench-scale and full-scale fire
performance of commercial polymeric materials used in electronic equipment. The ignition resistance,
self-extinguishing behavior, heat release rate (HRR), and combustion product yields of 18 different
materials at two thicknesses were characterized using three standard bench-scale fire tests. Five of the
18 materials were molded into 19” computer monitors for full-scale fire testing (using real and
simulated internal components). The results of this study were used to assess the predictive value of the
bench-scale tests in determining full-scale fire performance and to describe the fire hazard of the full-
scale specimens when exposed to three different ignition scenarios. A UL94 HB rated monitor
enclosure was easily ignited using a (38 ± 2) W needle flame and resulted in a peak HRR of (200 ± 25)
kW. The ignition threat distance (determined using the measured radiant heat flux distribution) for this
fire was found to be (58 ± 15) cm for piloted ignition of a stack of paper and (112 ± 28) cm for ignition
of insulated cotton fabric. The (23 ± 3) kW fire resulting from ignition of a keyboard was used as a
more severe ignition source for the monitor housings. Tests were also performed using a radiant heat
panel to simulate an existing burning item. All of the monitor specimens achieved ignition and at least
partial burn-up from the larger ignition sources. The full-scale test results were examined to determine
the degree of correlation with the bench-scale results. The UL94 vertical burn test showed good
agreement with the needle flame ignition results and the bench-scale peak HRR showed some
qualitative agreement with the keyboard fire and radiant panel full-scale results. All of the resins
exhibited complex physical behavior when burning (i.e. melting and charring) which made comparison
with small scale tests more difficult.

INTRODUCTION

Although fires originating from consumer electronics are rare, the hazard presented when
exposed to a small external ignition source (such as a candle) is not well known. Even when the
equipment is not the first item involved in a fire, its contribution to the total fire load and impact on
flashover of a room can be significant. Both of these issues are important due to an increasing number
of both candles and electronics in the home. It should also be noted that in recent years the number of
television fires has increased in many European countries following a reduction in the use of some
flame retardant compounds due to environmental concerns1. It is anticipated that this trend could
follow in the United States. The objective of this work is to relate the full-scale flammability and fire
hazard of consumer electronics assemblies having enclosures made from different resin formulations to
bench-scale fire performance of these resins. A research consortium was established between NIST,
UL, Dow, PolyOne, Albemarle and Samsung Cheil to conduct this research. Eighteen commercial
resins were evaluated using three different standard bench-scale flammability tests. Based on the
bench-scale test results, five of these resins were molded into 19” computer monitor housings and
examined in full-scale fire tests that measured the heat release rate (HRR) and the radiative ignition
threat to surrounding objects. The results are compared and contrasted to bench-scale results to infer
useful guidelines.

BACKGROUND

A number of previous studies have examined the fire performance of electronic equipment. A
comprehensive study was sponsored by the Society of the Plastic Industry (SPI) in 1981 2 that rated the
relative performance of 5 flame retarded plastic materials (UL94 V-0 and V-1) using 10 different bench-
scale fire tests and compared the ranked performance to the fire performance of model electronic
enclosures exposed to a 3 kW propane sand burner. The results showed reasonable qualitative
agreement between the overall bench-scale and full-scale performance. The Flame Retardant Chemical
Association (FRCA) sponsored a study by the National Bureau of Standards (NBS) in 1988 3 that
showed a significant reduction in the fire hazard of TV cabinets using flame retardant materials.
Several reports4-7 by the Swedish National Testing and Research Institute (SP) and the National
Association of State Fire Marshals (NASFM) compared the fire growth of off-the-shelf printers,
computer monitors and CPU’s to the UL94 ranking of the enclosure material and concluded that
enclosures using HB rated plastics are vulnerable to ignition by a small flame and can lead to flashover
of a room.

BENCH-SCALE FIRE TESTING

Three standardized bench-scale flammability tests were used to characterize a set of

commercially available resins. The bench scale flammability tests included the Cone Calorimeter test
(ASTM E 1354), the UL94 vertical burn test, and the Glow Wire Ignitability Temperature test (GWIT)
(IEC 695-2-1/3). A detailed description of the test methods can be found in a previous report8.

Materials

The formulations used in this study were chosen based on industry use and flame retardant (FR)
approach. Industry experts were consulted in choosing a set of 18 resins which included a variety of
resin types, FR levels and FR approaches. Commercial resins were chosen instead of model
formulations so that the effects of processing aids and other additives are included in the fire
performance results. The compounded formulations were provided by four different resin
manufacturers. The 18 different material identification labels used in this study are listed in the first
column of Table 1. The format of the label is: number - resin type – FR type. The resin types include
Acrylonitrile Butadiene Styrene (ABS), High Impact Polystyrene (HIPS), Polycarbonate (PC),
Polypropylene (PP), Polyvinyl Chloride (PVC) and a PC/ABS blend. The flame retardant types
include Bromine/Antimony (BFR), Phosphate (PFR), non-halogenated (NH) and no flame retardant
(NFR). Specimen thicknesses of 1.6 mm and 3.2 mm were chosen to represent typical electronic
housings. The resins were injection molded into 10 cm diameter round plaques for the Cone
Calorimeter and Glow Wire tests, and standard 125 mm by 13 mm bars for the UL94 tests.

Bench Scale Test Results

The results of the bench scale testing are listed in Table 1. The UL94 vertical burn test mean
afterflame times (time to self extinguishment) listed in the fourth column of Table 1 represent the
average of ten replicate measurements. The standard uncertainty for the afterflame time measurement
ranged from ±1 s to ±10 s. The Cone Calorimeter results listed in Table 1 are the average of 3 replicate
measurements with an external heat flux of 50 kW/m2. HRRpeak is the peak heat release rate, HOC eff is
the effective heat of combustion, tign is the time to sustained ignition and tpeak,HRR is the time to the peak
heat release rate. The relative standard uncertainty for the HRR peak measurement ranged from 2% to
20% with an average value of 8%. The results from the Glow Wire Ignition/Flammability
Temperature (GWIT/GWFT) test are shown in the last 2 columns of Table 1. A more detailed
description of the uncertainty of the measurements in Table 1 can be found in a previous report 8.
Several noteworthy observations are evident from examination of the results in Table 1. The 3.2 mm
thick specimens generally showed an improved performance in the UL94 test compared to the 1.6 mm
specimens. However, for 13 of the 18 materials, the opposite effect was observed for the peak heat
release rate (thinner specimens had a lower HRR peak). The effect of flame retardant additives showed
both increased UL94 performance and a reduction in the HRR peak, although a decrease in the time to
ignition was observed for most of the FR specimens. There was no obvious effect of thickness or FR
additive on the Glow Wire Ignition Temperature results. Some effort was given to obtaining a
qualitative correlation between the results of the different bench scale tests in Table 1 (can we predict
UL94 performance from HRR peak or some other measurement from the Cone?). Although correlations
could be found within a particular set of resins (such as HIPS), a more general relationship was not
found. This is not unexpected since the upward flame spread and melting/dripping mechanisms
inherent to the UL94 test are not captured by the geometry of the cone.
Table 1 Summary of Bench Scale Test Results at 50 kW/m2 heat flux.

Cone Calorimeter measurements were also

performed at 3 different heat flux levels using the 1800
3-HIPS-NFR - 30kW/m
2

3.2 mm thick specimens. The heat release rate 1600 3-HIPS-NFR - 50kW/m
2
2
3-HIPS-NFR - 90kW/m
curves for the non-FR HIPS sample at incident 1400
fluxes of 30 kW/m2, 50 kW/m2, and 90 kW/m2 are 1200
shown in Figure 1. The results show a greater
HRR (kW/m )
2

1000
HRRpeak and shorter time to ignition with increasing
800
heat flux and the general shape of the HRR curve
was unchanged. This trend was observed for most 600

of the materials examined; however some materials 400

(such as 6-PC/ABS-NFR) showed a decrease in 200

peak HRR as the heat flux was increased. 0
0 50 100 150 200 250 300 350
Time (s)
It has been shown that a steady state energy
Figure 1. HRR curves for 3-HIP-NFR (UL94 HB)
balance can be used to predict the functional 3.2 mm sample thickness.
relationship between the steady HRR and external
heat flux9. HRRss=HRRo+HRP(qext). In this
expression the intrinsic heat release rate, HRR o, represents the heat flux at zero external flux and the
heat release parameter, HRP, represents the material sensitivity to external flux and has been used to
predict fire propagation. It was observed that for a wide range of polymers the HRR o value was a good
predictor of UL94 performance, where self-extinguishing materials generally had a HRR o of less than
100 kW/m2. 10
A summary of the Cone results at 3 different
external heat flux levels is shown in Table 2.
Because many of the materials in Table 2 are
thermally thin and charring, a steady HRR
was never reached. The HRR o and HRP
values were determined using a linear
regression of the initial peak HRR value. In
general, larger values of HRR o and HRP
relate to increased fire hazard. A comparison
of the HRR o and UL94 rankings is shown in
Figure 2. From this limited data set it appears
that the criteria for self extinguishing
materials (V-0 and V-1) is that the HRR o is Figure 2. Comparison of UL94 ranking and
less than 500 kW/m2. Although these results intrinsic HRR for 3.2 mm specimens.
are qualitatively meaningful, their quantitative
values are questionable due to the large uncertainly in choosing an appropriate steady HRR. Because
the UL94 ranking can depend on sample thickness (ie. material 12 from Table 1), it cannot be
determined solely from an intrinsic property such as HRR o.

Table 2. Summary of Cone results at three heat flux levels, specimen thickness = 3.2 mm.
Sample Peak Heat Release Time To Sustained HRR0 HRP
Identification Rate (kW/m2) Ignition (s) (kW/m2) (kJ/kJ)
Irradiation (kW/m2) 30 50 90 30 50 90
1-PC-NH 457 482 532 193 67 17 420 1.3
2-HIPS-BFR 304 461 566 87 31 9 211 4.1
126
3-HIPS-NFR 1108 5 1623 151 50 16 843 8.6
4-PC-NFR 734 703 984 500 129 40 548 4.6
5-PC-BFR 321 343 437 260 78 24 254 2.0
6-PC/ABS-NFR 850 790 762 137 56 21 877 -1.4
7-ABS-BFR 243 370 515 126 50 17 126 4.4
8-PC/ABS-PFR 428 567 611 154 53 23 378 2.8
9-HIPS-BFR 930 760 827 136 44 17 909 -1.2
10-PC-BFR 191 167 258 461 72 16 133 1.3
165 209
11-PP-BFR 0 0 2391 221 62 20 1383 11.7
12-PP-NH 265 337 392 50 25 10 218 2.0
168 220
13-PP-BFR 9 6 2529 124 46 12 1396 13.2
167 220
14-PP-BFR 7 0 2581 110 35 16 1343 14.3
15-PP-NH 315 487 530 140 48 19 261 3.2
17-PVC-NFR 179 243 305 103 23 11 128 2.0
18-HIPS-PFR 391 445 639 98 30 9 251 4.2
19-ABS-PFR 262 293 454 99 36 11 148 3.3

FULL-SCALE FIRE TESTING

The full scale tests were performed under a 3 m square exhaust hood designed to accommodate
sustained fires with a net heat release rate (HRR) of up to 1 MW. The exhaust mass flow rate was set
to 2 kg/s (3600 SCFM) for these tests to provide optimal resolution for fires less than 400 kW in size.
Heat release rate measurements were based on the well-established oxygen consumption principle 11, 12.
A complete description of the hardware and test setup can be found in the final report 13.
The 5 materials used in the full scale tests reported here were selected to represent a wide range of fire
performance in the bench scale tests. A list of materials used in these tests and a summary of results
from 3 bench-scale tests (3.2 mm thick samples) are given in Table 3. The full scale specimen mass
listed in Table 3 is the combined mass of the front and rear pieces of the pre-assembled monitor
enclosure.

Table 3. Summary of bench-scale performance for materials used in full-

scale tests, and initial mass of combustible material on full scale specimens.
Cone peak Full-Scale
Resin HRR (kW/m2) UL94 GWIT (oC) / Enclosure
Identification @50 kW/m2 Classification GWFT (oC) Mass (g)
3-HIPS-NFR 1307 HB 750 / 725 2335
13-PP-BFR 1916 V-2 800 / 960 1999
18-HIPS-PFR 398 V-1 NT 2453
1-PC-NH 586 V-0 825 / 960 2839
7-ABS-BFR 409 V-0 750 / 960 2683

Full-scale specimens

Two piece 19” CRT computer monitor housings (Figure 3) were molded from the materials listed in
Table 3. The average specimen wall thickness was (3.0 ± 0.25) mm. The fire characterization was
performed using both real and simulated internal components. The simulated internal frame consisted
of 1.6 mm steel sheet metal formed roughly into the shape
of the real components. Draped over the frame was a
sheet of aluminum foil that served to increase the lateral
cross section of the sheet metal frame. The frame also
served to partially support the plastic enclosure during the
fire test and shield interior surfaces from radiation. The
frame was thus intended to achieve the same qualitative
effects that real monitor components have on a fire. The
real internal components consisted of a cathode ray tube
(CRT) and other various electronic components, some of
which were combustible. The 22 cm x 29 cm opening at
the base of the enclosure was covered with 1.6 mm sheet
metal. The specimen was placed on a brick such that the
mid-length side lower edge was (8 ± 1) cm from the table
surface. Figure 3. 19” Computer monitor specimen
used in full-scale testing.
Ignition Sources

Small local ignition source: A small local ignition source representing a typical candle sized flame was
used for these tests. A 0.5 mm I.D. needle flame burner was used to produce an n-butane flame with a
height of (20 ± 1) mm. The fuel mass flow rate was measured as (0.84 ± 0.02) mg/s and the net heat
release rate of this flame was calculated as (37.5 ± 2) W. The flame was applied mid-length along the
side of the specimen, (3 ± 1) cm above the lower edge. This emphasized both the potential for upward
flame spread and for the development of an interactive melt pool fire on the table surface. The initial
flame application was for a period of 20 s. If the burning specimen extinguished within 60 s, the flame
was immediately re-applied for 60 s. The 60 s application was repeated 3 times for a total flame
application time of 200 s. In the case that a hole was formed in the specimen, the test flame was moved
laterally to remain in contact with the enclosure (chasing the receding material). This ignition method
was designed to provide information on the fire performance of the equipment when exposed to a
localized short duration ignition source, and also the possibility of a much longer duration ignition
source such as an unattended candle.

Large ignition source (radiant heat panel): The response of the specimens to a larger ignition source
was simulated using a 48 cm x 33 cm natural gas radiant heat panel. This was intended to represent a
situation where the monitor was not the first item involved in a fire. The centerline heat flux was (21 ±
1) kW/m2 at a location 15 cm from the front surface of the panel. A removable copper plate shutter
was used to protect the specimen from the heat flux panel prior to the start of the test. The shutter was
water cooled and painted black to minimize the tendency of the gas-fired panel to increase in
temperature when shielded. A 1.6 mm I.D. open tube burner was used to produce a 10 cm n-butane
pilot flame with a net heat release rate of (178 ± 5) W. This pilot was applied in a location similar to
that of the needle flame; the entire side of the monitor was irradiated. In some tests it was held in
contact with melted material that had fallen to the table top.

Large ignition source (polystyrene keyboard): A generic, non-FR polystyrene keyboard was used as
an ignition source for the monitor specimens in some tests. The total weight of the keyboard was
(580 ± 5) g. The keyboard was placed under the front bezel of the monitor and ignited using the 20 mm
needle flame described previously. The needle flame was applied for 20 s to the side of the F9 key.
Tests were also conducted using only the keyboard to determine its contribution to the heat release rate.

Heat Flux Measurements

An array of four total heat flux gauges (Schmidt-Boelter type, 6 mm diameter sensor face, 13 mm
diameter body) was placed in a position to view one side of the burning object (right side of Figure 4).
The goal of the array was to obtain data on the distribution of radiative heat flux versus distance along
a line perpendicular to the object surface being viewed. This information is used below to infer the
maximum distance at which different materials could be ignited as a result of radiative heating from the
object fire. The four gauges were arranged as follows. The front and rear gauge were on the same
horizontal axis, both facing along this axis, and separated by 25 cm to 30 cm. The remaining two
gauges were on a single vertical axis, both facing in the same direction as the first two, i.e., toward the
side of the burning object. That vertical axis was displaced 5 cm from the horizontal axis of the first
two gauges. The two gauges on this vertical axis were separated vertically by a distance of 15.3 cm,
symmetrically above and below the horizontal axis of the first two gauges. This arrangement ultimately
supplies three measures of the heat flux versus perpendicular distance away from the viewed surface of
the burning object. In addition, since this distribution depends also on the height at which it is
measured, the two vertical gauges provide a first order correction for this effect. Since the gauges (and
their physical supports) had a finite size and could be within the field of view of those behind them,
corrections had to be made to their readings for this shadowing.

Needle Flame Ignition Results

None of the 4 flame retarded materials produced a measurable fire when exposed to the 20 mm needle
flame for a total application time of 200 s. The monitor enclosure molded using the non flame retardant
material, 3-HIPS-NFR, was easily ignited during the initial 20 s application of the flame and produced
a fire that consumed the entire monitor housing. The HRR curves for these specimens are shown in
Figure 4. The initial test flame was applied at 4 min into the data file for all tests described here. The
fire growth and peak HRR on the monitors with the simulated internal frame (test 1 and 4 in Figure 4)
were very reproducible. For each of the specimens the fire grew slowly during the first 3 min then
rapidly accelerated to its peak value during the next 3 min. The presence of the CRT (test 26 in Figure
4) decreased the fire growth rate and lowered the peak HRR by approximately 30 %.

The total mass loss measurement in this series of tests was compromised by several factors. The
calcium silica sheet below the specimen contained roughly 3 % water weight (150 g) that partially
vaporized during the test. In addition, some of the glass from the CRT was ejected from the monitor
during the test. For these reasons it was not possible to distinguish the mass loss due to the burning
plastic from the overall mass loss. The initial weight of the monitor enclosures for tests 1, 4 and 26
was (2335 ± 5) g. The combined expanded relative uncertainty (95 % confidence level) of the peak
heat release rate measurement was ±12 %, based on propagation of measurement uncertainty.
 250
monitor 1
monitor 4
1 monitor 26
200

4
150
HRR(kW)

26
100

50

tign = 4 min
0
0 5 10 15 20 25 30 35
Time (min)

Figure 4. Heat release rate curves for needle flame ignition of 19” monitors. Ignition flame applied to
specimen at t = 4 min. Image of monitor test, 3-HIPS-NFR, near peak heat release rate (right). Array
of heat flux gauges is shown on right side of image
Keyboard Fire Ignition Results

The keyboard ignition tests were performed

250
following the needle flame ignition method Stand alone keyboard
for the four flame retarded specimens that did 1-PC-NH (V-0)
7-ABS-BFR (V-0)
not ignite and had only local fire damage to 200
13-PP-BFR (V-2)
18-HIPS-PFR (V-1)
the enclosure. Tests conducted to 13
characterize the stand-alone keyboard as an
150
ignition source showed an average peak HRR
HRR (kW)

of 22.7 kW, approximately 10 min after

ignition. When the keyboard burned, its resin 100
did not flow outward more than 1 cm to 2 1
cm. Thus, when used as an ignition source, it 50
was essentially stationary. Portions of the tign = 4 min 18 7
monitor (the front bezel) immediately above k
it were partially immersed in its flames. 0
0 5 10 15 20 25 30 35 40
More remote portions of the monitor saw Time (min)

only limited radiation from the keyboard fire Figure 5. HRR curves for keyboard ignition of 19”
plume. computer monitors.

The HRR curves for the keyboard fire ignition of the monitors are shown in Figure 5. The HRR data
for the stand alone keyboard fire test is also shown in this figure for reference. With the exception of
material 7-ABS-BFR, the monitor specimens were ignited by the burning keyboard and the enclosures
were completely consumed by the resulting fire. During the test of monitor 7-ABS-BFR, only the
enclosure material directly in contact with the fire from the burning keyboard was ignited. The
contribution of the enclosure to the peak HRR was between 10 kW and 20 kW and the rear half of the
enclosure was not involved in the fire.

Of the FR specimens, material 13-PP-BFR exhibited the greatest hazard when exposed to the keyboard
fire. Once ignited the fire quickly grew into a large pool fire that covered the entire test surface. The
edge of the test surface was protected with aluminum foil to contain the melt pool, however a small
amount of burning plastic spilled over the edge. It should be noted that polypropylene is not typically
used for electronic enclosure housings.

The monitors using materials 1-PC-NH and 18-HIPS-NH had similar performances in the keyboard
fire ignition configuration. Although the peak HRR was lower than specimen 13-PP-BFR, the fire
spread to the rear part of the enclosure and consumed most of the mass of the enclosure. This result
illustrates that the fire hazard is not the same for all V-0 rated materials. As with all of the specimens,
the presence of the real CRT delayed the fire growth and lowered the peak HRR. This is likely due
to the considerable heat sink of the massive CRT. A summary of the peak HRR
results is given in Table 4.
Table 4. Summary of peak heat release rates for all full-scale fire tests.

UL94 Needle Flame Keyboard Radiant Panel

Rating Ignition Ignition Ignition
Specimen ID @3.2 mm CRT HRRpeak (kW) HRRpeak (kW) HRRpeak (kW)
PS-keyboard ---- 22.5 ---- ----
PS-keyboard HB ---- 22.6 ---- ----
PS-keyboard ---- 23 ---- ----
1-PC-NH No no-ign 46 124
1-PC-NH No no-ign 120 117
V-0
1-PC-NH Yes no-ign 55 ----
1-PC-NH No no-ign 55 ----
3-HIPS-NFR No 208 ---- 240
3-HIPS-NFR HB No 200 ---- 190
3-HIPS-NFR Yes 144 ---- ----
7-ABS-BFR No no-ign 44 no-ign
7-ABS-BFR V-0 No no-ign 31 25
7-ABS-BFR Yes no-ign 35 ----
13-PP-BFR No no-ign 205 193
13-PP-BFR V-2 No no-ign 199 167
13-PP-BFR Yes no-ign 180 ----
18-HIPS-PFR No no-ign 115 88
18-HIPS- PFR V-1 No no-ign 89 94
18-HIPS- PFR Yes no-ign 73 ----

Radiant Heat Panel Ignition Results

250
1-PC-NH (V-0)
The results of the radiant heat panel ignition of 3-HIPS-NFR (HB)
the monitor specimens are summarized in 200
7-ABS-BFR (V-0)
13-PP-BFR (V-2)
Table 4. All of these tests were performed 18-HIPS-PFR (V-1)
using the specimens with simulated internal 3
150
components. The radiant panel was positioned
HRR (kW)

13
2
so that a 21 kW/m total heat flux was imposed 1

at the point on the side of the specimen where 100

the local ignition source was applied. As in the
previous tests a barrier of foil was applied to 50
18

the edge of the support surface to prevent tign = 4 min

material from dripping onto the floor and 7
damaging the load cell and instrument wires. 0
0 5 10 15 20 25 30
Two 3.2 mm thick steel bars were positioned in Time (min)

front of the heat panel to prevent the enclosure Figure 6. HRR curves for radiant heat panel
from tipping over and contacting the face of the “forced ignition” of 19” monitors.
panel. 4 min after the start of the test the water-
cooled radiation shield was removed and the pilot flame was immediately applied to the side of the
monitor.

Specimen 1-PC-NH was easily ignited with this method and the fire quickly consumed the entire
specimen and grew to exceed 100 kW in size. The time from ignition to peak HRR was about 4 min.
As expected, the non-FR HIPS specimen, shown in Figure 6, ignited and quickly developed into a large
pool fire in less than 5 min. The 7-ABS-BFR specimen did not ignite during the first two attempts. All
of the material on the side of the enclosure melted away from the 10 cm pilot flame before it could
ignite. During the third attempt, shown in Figure 6, the pilot flame was lowered to remain in contact
with the developing melt pool on the surface of the calcium silicate board. The enclosure ignited and
slowly spread to the far side of the specimen. The HRR had a peak of 25 kW nearly 15 min after the
initial flame application. Similar behavior was observed for the 13-PP-BFR specimens, except that the
result was a larger fire. Only after applying the pilot flame to the melt pool for approximately 5 min
did the fire begin to propagate to the other sides of the enclosure. Once ignited however the fire grew
very quickly and had a similar growth rate and peak HRR to the keyboard fire ignition method of the
same material. The heat release rate curve of the 18-HIPS-NH (UL94 V-1) specimen ignited using the
radiant panel and pilot flame is also shown in Figure 6. The average peak heat release rate, 91 kW,
was substantially lower than the HIPS resin with no flame retardant.

Threat of Ignition of Other Objects

Any fire poses an ignition threat to objects in its surroundings. Such ignition could occur by one of
several modes including: direct flame contact with the surface of another object, movement of flaming
material, and remote ignition by radiation. For the purpose of comparing relative fire hazard we will
consider only radiative ignition here. An important feature of radiative ignition is that below some flux
level (corresponding to some distance away from the radiating fire plume), the surface of the heated
object will not get hot enough to be ignitable even if a pilot flame is present. The distance beyond
which this is true defines the “threat radius” of the fire.
Different target objects have differing ignitability as a result of their specific chemical and physical
properties. Since it is not possible to make predictions for all objects which may plausibly be near an
electronic equipment fire, surrogate materials are used. Here we consider two materials which are
surrogates for common objects of interest in the vicinity of a desktop computer. The first is a stack of
paper. Specifically, the surrogate material is a 2.54 cm thick unbound pile of copier paper. This is a
surrogate for books, magazines, manuals or printer paper that could be on a desk top. The second is a
medium weight cotton fabric which is a surrogate for a drape or for the seatback of an upholstered desk
chair. The fabric is 100 % cotton and weighs 0.41 kg/m2 (12 oz/yd2). The piloted ignition behavior of
these materials was measured in the Cone Calorimeter over a flux range from 70 kW/m2 down to the
minimum flux for ignition. In the monitor fires, the heat flux that would impinge on a target object
rises and falls as the monitor fire builds and recedes, therefore the Cone data cannot be used directly.
Instead, it forms the basis for inferring the effective ignition properties of the surrogate material. A
simple thermal ignition model is used to find
effective property values which closely reproduce
the measured ignition behavior from the Cone.
These properties are the ignition temperature and the
apparent thermal inertia (product of density (), heat
capacity (c) and thermal conductivity (k)). A
comparison of the cone ignition data and the ignition
model for the cotton fabric material is shown in
Figure 7. Given these effective ignitability
properties and the measured heat flux versus time
from the monitor fire tests, this information can be
used in the model(s) to predict the farthest distance
from the fire at which ignition of the surrogate
materials can just occur.
Figure 7. Radiative ignition of cotton fabric in the
Since the number of heat flux gauges was quite Cone Calorimeter.
limited, only some portion of the flux versus
distance profile was measured in each test and it
was necessary to extrapolate/interpolate the
measured values. Digital images of the fire,
taken from behind the flux gauges, were used to
determine an approximate area of the fire which
was then used to account for partial shadowing
of the rear gauges and to extrapolate the flux
distribution based on the radiative view factor.
Figure 8 shows this view factor based
extrapolation curve and the point measurements
for one of the monitor specimens. Because
the flux data have substantial
variability (due to the turbulent nature Figure 8. View-factor based extrapolation curve and
flux measurements from keyboard ignition of 1-PC-
of the fire plume), it is simpler to NH monitor.
approximate the transient nature of
the flux by using a Gaussian time
dependence. The Gaussian is specified to match the peak flux and its time width
at 50 % of the peak. A detailed description of the flux gauge geometry, ignition
model , and extrapolation methods can be found in the full report 13.
Table 5 shows the computed maximum piloted ignition reach values for the cotton fabric and the paper.
This was done only for a select set of fire tests since it was quite labor intensive. The cases reported
here are for the monitor enclosures containing the pseudo-CRT interior; the main goal is a comparison
of the differing resins, plus some guidance regarding “acceptable” fire sizes. The results for the cotton
fabric were computed for the two extremes; adiabatic back surface or equal back and front surface heat
losses (re-radiation plus convection). The ignition behavior of the paper surrogate was computed with a
thermally thick model using a 2.5 cm depth. This is thick enough that the back surface condition is
irrelevant on the time scale required for ignition. The relative error in the maximum ignition threat
distance was estimated to be less than 25%.
Table 5. Results of maximum ignition threat distance analysis.
(a)=needle flame ignition, (b)= keyboard ignition, (c)= radiant heat panel ignition.
Max.
Ignition Max.
Pea Distance Ignition Max.
k (m) Distance Ignition
HR Non- (m) Distance
Resin R insulated Insulated (m)
(configurati (kW Cotton Cotton Stack of
on) ) Fabric Fabric Paper
3- HIPS-NFR 200 0.71 1.12 0.58
(a)
13-PP-BFR (b) 198 0.65 0.97 0.56
18-HIPS-PFR 89 0.36 0.56 0.29
(b)
7-ABS-BFR (b) 31 N. A. N. A. N. A.
1-PC-NH (b) 120 0.36 0.58 0.31
1-PC-NH (c) 124 0.36 0.51 0.26
Keyboard (a) 22.5 0.12 0.27 0.08

The best-behaved case was for 7-ABS-BFR for which ignition reach values could not be directly
calculated (indicated by N.A. in Table 5). The reason for this was that the fire stopped before it
propagated as far along the side as the flux gauge array and the peak flux values recorded even for the
front flux gauge, were less than 4 kW/m2. Even the most ignitable surrogate case (cotton fabric with
adiabatic rear surface) required a minimum peak flux of about 8 kW/m 2 to ignite. Extrapolation of the
available flux data imply that this would not be reached even at the monitor surface for 7-ABS-BFR, at
least in the vertical plane of the flux gauges. The worst fires here threaten to ignite such objects from
as much as 1 m away, the more moderate fires from more than half a meter. It is not possible to make a
statement about how many secondary fires these situations would induce in the real world. One can
only infer that the probability of a secondary fire is roughly proportional to the area encompassed by
the ignition reach and thus to the square of the ignition reach value. Even a heat release rate
peak of about 23 kW (here seen for the keyboard alone) can potentially ignite
the back of an upholstered desk chair or a drape nearly 0.3 m away.

Note that the preceding results show that the peak HRR from an electronic
enclosure depends on the size and intensity of the ignition source. Thus it is
necessary to put the object in context, decide on plausible ignition sources, and
test the object in full-scale to find its peak HRR. Only then can the ignition reach
results presented here pertain to a specific monitor. If the only plausible ignition
source is match-sized, and the object is not used in proximity with non-FR
peripherals, then all FR resins here would be adequate since none led to a
significant fire.

Full-Scale/Bench-Scale Comparison
The results from the bench-scale tests and the full-scale monitor tests are shown in Table 6. The bench-
scale test results are for 3.2 mm thick samples. The full-scale peak HRR results in Table 6 represent
the largest of the replicate measurements. Qualitatively, the UL94 vertical burn test gave the best
indication of the full-scale monitor fire performance when exposed to the needle flame ignition source.
All of the materials that self-extinguished in the UL94 test resisted sustained ignition in the full-scale
monitor tests. The time to ignition in the Cone and the Glow Wire Ignition Temperature were the
poorest indicators of full-scale fire performance in these tests. Although the presence of a flame
retardant additive can cause a material to ignite faster and at a lower temperature, these factors did not
help predict whether or not the flame would propagate once ignited. The peak HRR from the Cone was
not a good predictor of the full-scale response to the local ignition source in this study. A notable
example of this is the polypropylene specimen (13-PP-BFR) that has a very high peak HRR in the Cone
but did not ignite when exposed to the needle flame. The Cone results compared more favorably with
the full-scale monitor tests having a larger ignition source. The two specimens with the highest peak
HRR in the Cone tests also produced the largest fires when exposed to the keyboard fire and radiant
panel ignition sources. More research is needed to develop and interpret bench-scale tests capable of
predicting full-scale performance. The most reliable existing measure is full-scale testing, assuming the
appropriate ignition scenario can be identified.
Table 6. Comparison of Bench-Scale and Full-Scale test results.

CONCLUSIONS

Progressively larger and more intense ignition sources caused the burning of an increasing
number of tested resins. The use of flame retardant materials (including non-halogenated) provided
adequate protection against the needle flame that represented a “candle size” ignition source. The fire
hazard from needle flame ignition of the enclosure having a non-flame-retarded material (3-HIPS-NFR)
was significant and resulted in the threat of fire spread to nearby objects. The keyboard fire ignition
source produced a significant fire hazard for all but one (7-ABS-BFR) of the monitor enclosures. The
radiant heat panel used to simulate an existing fire produced significant burning for all of the monitor
specimens.

Several bench-scale flammability measures were assessed for ability to predict full-scale monitor
behavior with limited success. The UL94 vertical burn test was a good indicator of the likelihood of the
full-scale specimens to resist sustained ignition by a “candle size” flame. The Cone Calorimeter test
was a reasonable indictor of the response of the monitor specimens to ignition by a nearby burning
object. The Glow Wire Ignition Test was a poor indicator of the full-scale response to an open flame
ignition source.

The radiant ignition of remote objects was analyzed for several of the monitor fires. A 200 kW (peak
HRR) fire produced a threat distance of 1.1 m for upholstery fabric. A (20 to 25) kW (peak HRR) fire
produced a threat distance of 0.3 m for upholstery fabric. A peak HRR of less than 10 kW would likely
produce a minimal radiant ignition threat to its surroundings, comparable to the threat of direct flame
contact. Peripheral items (such as the keyboards in this study) are often made from non-FR materials
and can serve as an ignition source for other items if ignited.

ACKNOWLEDGEMENTS
 The authors would like to thank Scott Dillon and Ken Steckler from the ATF Fire Research
Laboratory for their assistance in performing the full-scale experiments. The following individuals also
made contributions to this work: Sanghyun Hong, Alex Morgan, Tony Kingsbury, Susan Landry,
Douglas Wetzig, Robert Backstrom, David Edenburn, Robert Grifffin, Gordon Nelson, Edward Watt,
Michael Smith, Richard Harris, Takashi Kashiwagi, Jeffrey Gilman and Richard Lyon.

REFERENCES
1
De Poortere, M., Schonbach, C., Simonson, M., The Fire Safety of TV Set Enclosure Materials,
a Survey of European Statistics, Fire and Materials, 24 (1), 2000, p.53 - p.60.
2
Bieniarz, R., Fire Experiments on Structural Foam Plastic Equipment Enclosures, Underwriters
Laboratories, 1981.
3
Babrauskas, V., Gann, R.G., Harris, R.H., Levin, B.C., Fire Hazard Comparison of Fire-
Retarded and Non-Fire-Retarded Products, NBS-SP749, 1988, Gaithersburg, p.1-86.
4
Simonson, M., Report for the Fire Testing of Computer Monitors, R31151, SP 1999.
5
Simonson, M., Report for the Fire Testing of Deskjet Printers, P005848, SP 2000.
6
Simonson, M., Report for the Fire Testing of One Printer and Two CPU's, P008664, SP 2000.
7
Bliss, D., Simonson, M., Fire Performance of IT Equipment Studied in the Furniture
Calorimeter, Interflam 2001, p.171 - p.179.
8
Bundy, M., Ohlemiller, T.J., Bench-Scale Flammability Measures for Electronic Equipment,
NIST-IR7031 2003, Gaithersburg, MD, p.1-18.
9
Tewarson, A., Generation of Heat and Chemical Compounds in Fires, Soc. of Fire Protection
Engineers 1995, Boston, MA, p.53-124.
10
Lyon, R.E., Fire & Flammability, Fire and Materials 2003, San Fransico, CA.
11
Huggett, C., Estimation of Rate of Heat Release by Means of Oxygen-Consumption
Measurements, Fire and Materials, 4 (2), 1980, p.61 - p.65.
12
Parker, W.J., Calculations of the Heat Release Rate by Oxygen-Consumption for Various
Applications, Journal of Fire Sciences, 2 (5), 1984, p.380 - p.395.
13
Bundy, M., Ohlemiller, T., Full-Scale Flammability Measures for Electronic Equipment, NIST-
TN1461, 2004, Gaithersburg, MD, p.1-46.