Fine Particulate Matter Emissions from Candles
Zhishi Guo and Ronald Mosley
U.S. Environmental Protection Agency, Office of Research and Development, National Risk
Management Research Laboratory, MD-54, Research Triangle Park, NC 27711
Jenia McBrian and Roy Fortmann
ARCADIS Geraghty & Miller Inc., P.O. Box 13109, Research Triangle Park, NC 27709
ABSTRACT
Five types of candles purchased from local stores were tested for fine particulate matter (PM)
emissions under close-to-realistic conditions in a research house. The test method allows for
determination of both the emission rate and deposition rate. Most tests revealed low PM
emission rate except two, in which excessive sooting occurred and the PM concentration
approached 1000 |ig/m3 with six and nine burning wicks, respectively. Wax breakthrough
significantly increased the PM emission rate. Smoldering generated more fine PM than several
hours of normal burning, causing very high concentrations in a short period of time, which raises
concern over potentially acute health effects, especially for children and the elderly. A simple
source model is proposed to represent both stable PM emissions during normal combustion
conditions and the sudden concentration surge following flame extinction.
INTRODUCTION
Candles have been associated with human living conditions for at least 1000 years. Although no
longer a major means of lighting in modem society, they are still widely used in homes, mostly
for creating unique, warm, and tranquil atmospheres. It was believed from observation that the
candle flame produces carbon particles and that, under perfect combustion conditions, the carbon
is totally consumed by the flame. In his famous book The Chemical History of a Candle,
originally published in 1861,1 Michael Faraday — one of the greatest experimental scientists of
1
-------
all time -- uses simple, yet very ingenious, experimental methods to prove that carbon particles
exist in the candle flame and that it is these solid particles that create "the very beauty and life of
the flame." He further points out that, under imperfect combustion conditions, the carbon
particles cannot be consumed entirely by the flame, resulting in emissions of soot.
In recent years, concerns over the impact of candle burning, especially the property damage they
may cause due to soot deposition, have been on the rise.23'4'5 Analysis of the potential impact on
human health due to inhalation of particulate matter (PM) has also been reported,5 According to
the study by Fine et al.,2 a sooting flame and a smoldering wick produce much higher fine
particle mass emission rates than a quiet normal burning candle and are responsible for the vast
majority of fine particle emissions from this source. Another human-health-related issue
associated with candle burning is the inhalation of particlebome lead (Pb) generated by certain
types of candles, whose wicks have a lead core.6'7
This paper investigates the PM emissions from candles under close-to-realistic conditions and
their contribution to indoor PM levels with emphasis on the fine fraction of the PM — particles
having aerodynamic diameters of less than 2.5 jam (PM2.5). The goals were to: identify the
emission patterns, measure the emissions under certain real-life scenarios, and determine the key
parameters needed for estimating inhalation exposure.
EXPERIMENTAL
Test Facility
Emissions tests were performed in a research house located in Gary, NC.8 One bedroom, used as
a test chamber, was isolated from the rest of the house by blocking the air supply registers and
closing the interior door. The room has dimensions of 3.78 (length) x 3.28 (width) x 2.44
-------
(height) m. It has vinyl flooring, painted gypsum board walls, and a textured gypsum board
ceiling. The particle-free air supply was generated by an in-line fan (FanTeck Model FR250, 230
W), which passed outdoor air through a high efficiency particle air (HEPA) filter into the test
room, keeping the room under slightly positive pressure (~2 Pa) to prevent the infiltration of
particles from the outdoors and adjacent rooms. Prior to a test, a stand-alone HEPA filter air
cleaner (Bionaire, Model CH-3580 or Honeywell Enviracaire, Model 13520) operated inside the
test room for 30 minutes to reduce the background fine PM concentration to less than 2 fig/m3,
A ceiling fan was used to keep the room air well mixed. The test candles were placed on a table
away from the direct air draft created by the ceiling fan. Under the standard test conditions, the
ceiling fan was set at low speed and normal wind direction (i.e., downward), which gave an
average air speed of 11 cm/s near the top of the candle ~ close to the air speed commonly found
in indoor environments. Other speed/direction combinations of the ceiling fan provided an air
speed range from 14 to 27 cm/s near the candle. When the fan was turned off, the air speed was
reduced to less than 5 cm/s. Additional discussions of this test facility can be found in these
proceedings.9
Test Specimens
All candles tested were purchased from local stores and are described in Table 1.
-------
Table 1. Description of test candles!
Sample ID
PI
P2
P3
BW1
BD1
Type
paraffin/aroma
paraffin/aroma
paraffin/aroma
beeswax
birthday candle b
Shape of
Cross Section
square
square
round
round
round
No, of Wicks
9
9
3
1
1
Color
orange
red
mauve
yellow
white
a PI and P2 are made by the same manufacturer,
b Material type was not mentioned on the label.
PM Sampling
Particles with aerodynamic diameters of less than 10 \im (PM10) and less than 2.5 jim were
sampled simultaneously onto Teflon filters, using cyclones (University Research Glass) with
corresponding size cut-points. Each cyclone has a two-stage filter pack (47-mm, R2PJ047). The
mass concentrations of PM25 and PM 10 were determined gravimetrically. The PM size
distribution and real-time concentration were determined with an electric low pressure impactor,
or ELPI (Dekati Ltd.)10, which measures an aerodynamic diameter range from 0.03 to 10 urn
with 12 stages, and has a response time of 5 sec while experiment data were recorded every 60
sec.
Test Method
The ELPI was used to monitor the indoor PM concentrations throughout the experiment. After
the room was pressurized and the background PM concentration reduced to less than 2 [o.g/m3, the
test candle was lit with a butane lighter. Matches were used to light the birthday candles. The
-------
burning duration was 8 minutes for the birthday candles and 4 to 6 hours for the other candles.
In the latter case, the candles were allowed to burn for 1 hour before filter sampling. After the
test candles were extinguished, the EPLI continued to operate for at least 4 more hours. Sulfur
hexafluoride (SF6) tracer gas was injected after the extinction of the flame to determine the
ventilation rate and the PM deposition rate. No SF6 was injected prior to or during the burning
period because of possible decomposition of SF6 when in contact with a flame, resulting in
interferences."
RESULTS
Correlation Between the Gravimetric Method and ELPI
Comparison of paired fine PM concentration data showed that, in the low concentration range (<
5 M-g/m3), the two methods agree with each other reasonably well and that the ELPI gave higher
readings as the concentration increased (Figure 1). A good correlation exists between the two
methods, however. In this paper, all the ELPI data were corrected based on Equation 1 :
Equation 1. Correlation between the gravimetric method and ELPI for fine PM concentrations
In C = 0.829 In CELPI - 0.475 (r2 = 0.967, n = 14)
where Cg^ = fine PM concentration from the gravimetric method (p.g/m3), and
CELPI = fine PM concentration from the ELPI
-------
Figure 1. Correlation between the gravimetric method and ELPI for fine PM
1E4
J: 1E3-
O)
ZL
•c 1E2
-i—>
6
1EO-
1EO
1E1 1E2 1E3
ELPI (MQ/m3)
1E4
PM Concentrations and Emission Rates by the Gravimetric Method
The average concentration and emission rate calculated from the filter samples are presented in
Table 2. In most cases, the fine PM concentration was low except in two tests for candle P2, in
which excessive sooting occurred and the fine PM concentration reached 955 and 1137 ng/m3,
respectively. These unusually high concentration results were supported by the ELPI data. A
higher burn rate in test C121499 is another indication of possible imperfect combustion.
However, no visual observations were made during these tests and we were unable to determine
the exact causes of the high emission rates.
-------
Table 2. Average fine PM concentrations and emission rates based on filter samples
Test
Candle
PI
P2
P3
BW1
Test
ID
C121399
G121699
C121499
C121599
C121799
C041100
C041200
COS 1600
COS 1700
C041700
Wicks
Lighted
9
9
9
6
1
1
1
9c
9C
1
Burn Rate
(g/h/wick)
2.77
3.71
4.67
NAb
2.29
3.33
2.50
3.89
4.00
7.86
Air Speed ACH Concentration
(cm/s) (h-1) (lJ.g/m3)
11
<5
11
11
14
15
27
11
11
14
0.99
0.95
0.98
0.93
1.1
1.56
1.51
1.54
1.5
.1.45
99.6
13.0 a
955
1137
15.8
8.79
8.13
14.9
17.3
4.32
Emission Rate
(^g/h/wick)
329
41
3120
5287
521
411
368
76
87
188
a This value is based on the ELPI data. The result from filter samples was discarded because the
relative standard deviation for duplicate samples was too large.
b Burn rate was not measured in this test.
c Three 3-wick candles.
PM Size Distribution
Most particles emitted from candles were in the fine particle size range. This is evidenced by the
fact that the filter samples for PM2.5 and PM10 taken in parallel are almost the same in most tests
(Table 3). Typical size distribution for fine PM is shown in Figure 2. Like some other
combustion sources, two peaks appear in the size distribution. Although smoldering creates
more larger particles, the shape of the distribution curve did not change significantly.
-------
Table 3. Comparison of filter samples for PM25 and PMi0
Concentration (|ig/m3) a
PM2J
2.7 ±0.5
8.1 c
10.3 ±0.4
14.9 ±4.6
15.8 ±3.0
17.3 ±1.1
29.2 ± 2.5
49.8 ±1,7
99.6 ± 2.5
955 ± 3.2
1137 ±248
PM10
3.4 ±0.2
8.6 c
11.3 ±0.9
15.2 ±5.1
16.7 ±1.5
18.5 ±1.4
29.1 ±0.1
50.8 c
99.6 ± 2.5
951 ±20.2 :'
1128 ±86.7
Percent
Difference b
23.7
5.5
9.8
1.9
5.1
7.1
-0.5
2.0
0.0
-0.5
-0.8
2 Mean ± standard deviation for duplicate samples.
b Percent = 2 x (PM10 - PM2.5) / (PM10 + PM2.5) x 100.
c Single filter sample.
Figure 2. Fine PM size distribution observed in test COS 1600
0.1 1
Aerodynamic Diameter (jjm)
10
Normal Burn-e- Smoldering
8
-------
Emission Patterns
General Emission Patterns
The real-time concentration data from the ELPI were used to determine the PM emission
patterns. In most cases, the fine PM emission rate was fairly steady during the normal burning
period. However, the emission rate was higher immediately after the candle was lit. Two
slightly different emission patterns are shown in Figures 3 and 4. In both cases, a concentration
surge occurred when the flame was extinguished. In most tests, the smoldering period generated
more particles in a few seconds than during the whole period of normal burning. Emissions due
to smoldering are discussed further in the following section. Also note that the emissions data
from filter samples presented in Table 2 do not include the smoldering period because the
sampling pumps were turned off before the candles were extinguished.
-------
Figure 3. Fine PM concentration profiles for candle P3 in duplicate tests
0
8
Elapsed Time (h)
10
-------
Figure 4. Fine PM concentration profile for candle P2 in test C121699
100
x~, 8°
1
"a* eo
0- 40
-------
where Wx = amount of fine PM emitted during smoldering period (|J.g), Q = ventilation flow rate
(m3/h), C = fine PM concentration (p.g/m3), t = time (h), V = room volume (m3), tx = time
at which the flame is extinguished (h), and t,, = time for the last data point (h).
The time-concentration curve was integrated using the trapezoidal rale, and the calculated
emissions from smoldering are presented in Table 4. It appears that candle P3 emits more fine
PM during the smoldering period than the other two paraffin candles, although it emits less
during the normal burning period (see Table 2). Results in Table 4 also suggest that using a
snuffer to extinguish the flame produced less PM than blowing out the flame. Blowing out 30
birthday candles produced the highest fine PM concentration in the room (about 500 [ig/m3 in the
mixed air), and the concentration remained above 100 [ig/m3 for more than an hour (Figure 5).
Table 4. Amounts of fine PM emitted due to smoldering
Test ID
C121399
C121699
COS 1900
COS 1600
COS 1700
C052300
COS 1900
C041700
C050800
Candle
PI
P2
P3
BW1
BD1
Number of
wicks
9
9
9
9
9
9
9
1
30
Extinguishing
Method
blowout
blowout
blowout
blowout
blowout
blowout
snuffer
blowout
blowout
Peak Cone.
(|ig/m3)
144
88.3
153
257
250
213
82
32.5
483
Emissions
(Hg/wick)
273
261
262
609
650
554
115
816
151
12
-------
Figure 5. Fine PM concentration profile for lighting and blowing out 30 birthday candles
500
0
0
50 100 150
Elapsed Time (min)
200
Emissions after Wax Breakthrough
Wax breakthrough occurs when the rim of the solid wax surrounding the liquid pool softens from
the heat of the flame and slumps off to one side of the candles, causing the release of the liquid
wax from the pool that has concentrated around the wick. It is one way to cause imperfect
combustion. Wax breakthrough occurred in one test with candle P3. As shown in Figure 6, a
higher emission rate resulted.
13
-------
Figure 6, Effect of wax breakthrough on fine PM emissions (test COS 1800)
160
JT120
Q_
CD
C
LJ.40
Blowout
Breakthrough
468
Elapsed Time (h)
10
12
PM Deposition Rates
PM deposition rate is an important parameter to estimate the PM concentrations from indoor air
quality simulation. In this work, the decay part of the ELPI data (i.e., after the candle was
extinguished) was used to estimate this parameter. The first-order deposition rate constant was
calculated by comparing the decay rate for PM with that for the tracer gas. As shown in Figure
7, a U-shaped curve was obtained. Note that the emission rate data presented in Table 2 do not
consider PM deposition and thus should be considered as the lower bound of the actual emission
rate.9
14
-------
Figure 7. First-order deposition rate constants for test C121699
0
0.01
0.1 1
Aerodynamic Diameter (|jm)
10
Data
Best Fit
MODELING CONSIDERATIONS
Source Model
Considering the unique emission pattern for PM emissions from candles, an average emissions
rate may not be adequate for exposure estimation, especially when acute health effects are of
concern. A simple model, represented by Equations 3 and 4, is proposed to account for
emissions for both normal burning and smoldering:
15
-------
Equations 3 and 4. Model for PM emissions from candle burning
D _ D
-K- — JV.
for to < t < tx
AC =
W.
att =
where R = PM emission rate (jJ-g/h), R,, = constant emission rate during the normal combustion
period from to (burning start time) to t,; (M-g/h). and AC = an instant increase of PM
concentration in room air at I* (jag/m3).
This model can be easily implemented in a spreadsheet or an indoor air quality simulation
program. An example application of this model is shown in Figure 8.
Figure 8. Modeling of test C051700 using Equations 3 and 4 as a source model
300
Q_
LL
0
24 6 8
Elapsed Time (h)
10
16
-------
PM Accumulation on Interior Surfaces
With knowledge of both the emission rate and deposition rate, the amount of PM deposited on
interior surfaces can be estimated from Equations 5 and 6. An example simulation shown in
Figure 9 represents a simple case, in which all interior surfaces are treated as the same type and
an average deposition rate constant of 0.4 h"1, which is equivalent to a deposition velocity of 0.2
m/h in the test room, was used for fine PM.
Equations 5 and 6. Estimation of PM deposition on interior surfaces
dC
V-r=R
ai
dM-
'_ r>
— JLJ,
dt
where n = number of interior surface types, S; = area of surface i (m2), Df = PM deposition
velocity for surface i (m/h), and M; = amount of PM accumulated on surface i
17
-------
Figure 9. An example simulation of PM accumulation on interior surfaces in the test room
250
0
0
4 6
Elapsed Time (h)
8
10
CONCLUSION
Under normal combustion conditions, the candles tested do not produce significant amounts of
particles — the average PM2,5 emission rate ranges from 41 to 521 jig/h/wick. Excessive sooting
occurred in two tests with average PM2,5 concentration approaching 1000 (J,g/m3 and an emission
rate in the 3000 to 5000 (ig/h/wick range. However, the exact cause of sooting is not clear.
Smoldering often generates more particles than several hours of normal burning. The amount of
PM2-5 emitted from extinguishing the flame ranges from 115 to 569 |ig/wick. In a similar test
18
-------
with 30 birthday candles, the peak fine PM concentration was near 500 |J,g/m3 after extinguishing
the flames. Given the dramatic concentration surge due to smoldering, using an average
emission rate to represent the fine PM emissions from candles may not be adequate for exposure
estimation in some cases, especially when acute health effects are of concern. As a first
approximation, a combination of constant and instant source models is recommended.
REFERENCES
1. Faraday, M. The Chemical History of a Candle; originally published in 1861; reprinted by
Cherokee Publishing Company, Atlanta, GA, 1993.
2. Fine, P. M.; Cass, G. R.; Simoneit, B. R. T. Environ, Sci. & Techn.1999, 33, pp 2352-2362.
3. Krause, D. In Indoor Environment: the State of the Industry, Presentations from the 7th Annual
Indoor Environment Conference, IAQ Publications, Inc.; Bethesda, MD, 1999; pp 157-159,
4. Al-Ahmady, K. In Indoor Environment: the State of the Industry, Presentations from the 7th
Annual Indoor Environment Conference, IAQ Publications, Inc.: Bethesda, MD, 1999; pp 159-
164.
5, Krause, J. D. Characterization of scented candle emissions and associated public health risks,
Master's thesis, University of South Florida, Tampa, FL, 1999.
6. Alphen, M. The Sci. of the Total Environ. 1999, No. 243/244, pp 53-65.
7. Nriagu, J. O.; Kim, M. The Sci. of the Total Environ., 2000, No. 250, pp 37-41.
8. Jackson, M. D.; Clayton, R. K.; Stephenson, E. E.; Guyton, W. T.; Bunch, J. E. EPA's Indoor
Air Quality Test House, I, Baseline Studies, in Proceedings of the 1987 EPA/APCA Symposium,
Research Triangle Park, NC, May 3-6,1987, Environmental Monitoring Systems Laboratory,
Research Triangle Park, NC, EPA-600/9-87-010 (NITS PB88-113402), 1987.
19
-------
9. MeBrian, J.; Fortmann, R,; Guo, Z.; Mosley R. Test methods to characterize particulate matter
emissions and deposition rate in a research house, in these proceedings.
10. Marjamaki, M.; Keskinen, J.; Chen, D. R.; Pui, D. Y. H. J. Aerosol ScL, 2000,31, pp 249-
261.
11. Fisk, W. J.; Wallman, P. H.; Prill, R. J.; Mowris, R. J.; Grimsrud, D. T. Lawrence Berkeley
Laboratory Report, LBL-24216, Berkeley, CA, 1988, pp 6-25.
KEY WORDS
indoor air, candles, soot, emissions, particulate matter
20
-------
NRMRL-RTF-P-527
TECHNICAL REPORT DATA
(Please read /xstmctiora on the reverse before completing}
i. REPORT NO.
EPA/600/A-00/056
2.
4. TITLE AND SUBTITLE
Fine Partieulate Matter Emissions from Candles
3. REC
S. REPORT DATE
6. PERFORMING ORGANIZATION CODE
7.AUTH0R(s) z
------- |