Formation of chlorinated dioxins and furans in a hazardous-waste-firing industrial boiler

Article abstract of DOI:10.1021/es991196+

This research examined the potential for emissions of polychlorinated dibenzodioxin and dibenzofuran (PCDD/F) from industrial boilers that cofire hazardous waste. PCDD/F emissions were sampled from a 732 kW (2.5 x 10⁶ Btu/h), 3-pass, firetube boiler using 2 fuel oil cofired with 2,4- dichlorophenol or 1,2-dichlorobenzene and a copper naphthenate mixture. PCDD/F levels were significantly elevated when improved combustion conditions (reduced carbon monoxide, increased carbon dioxide) followed periods of flame wall-impingement and soot formation/deposition on the boiler tubes. Boiler tube deposits became a sink and source for PCDD/F reactants (copper and chlorine) and PCDD/F, resulting in continued formation and emissions long after waste cofiring ceased. The role of deposits in PCDD/F formation makes emissions dependent on current as well as previous firing conditions, resulting-in uncertainty regarding prediction of emissions based solely on the type and rate of cofired hazardous waste. This research examined the potential for emissions of polychlorinated dibenzodioxin and dibenzofuran (PCDD/F) from industrial boilers that cofire hazardous waste. PCDD/F emission were sampled from a 732 kW (2.5 × 10⁶ Btu/ h), 3-pass, firetube boiler using #2 fuel oil cofired with 2,4-dichlorophenol or 1,2-dichlorobenzene and a copper naphthenate mixture. PCDD/F levels were significantly elevated when improved combustion conditions (reduced carbon monoxide, increased carbon dioxide) followed periods of flame wall-impingement and soot formation/deposition on the boiler tubes. Boiler tube deposits became a sink and source for PCDD/F reactants (copper and chlorine) and PCDD/F, resulting in continued formation and emissions long after waste cofiring ceased. The role of deposits in PCDD/F formation makes emissions dependent on current as well as previous firing conditions, resulting in uncertainty regarding prediction of emissions based solely on the type and rate of cofired hazardous waste.

Full text of DOI:10.1021/es991196+

Research
Some research has examined PCDD/ F emisssions from
systems that have limited applicability to industrial boilers.
DeFre and Rymen (3) examined PCDD/ F emissions from a
domestic oil burner and gas burner, varying doped hydrogen
chloride (HCl) concentration and combustion conditions via
Formation of Chlorinated Dioxins
and Furans in a
Hazardous-Waste-Firing Industrial
Boiler
2
oxygen (O ) and carbon monoxide (CO) emission concentra-
tions. Emissions of PCDD/ F were always present, up to levels
exceeding 50 000 ng/ m3 (oil, 11% O
typically in the range of 660 ng/ m [methane (CH
, 3.7% HCl) but more
2
), 9% O ,
2
3
B R I A N K . G U L L E T T , * , †
4
A B D E R R A H M A N E T O U A T I , ‡ A N D
3.7% HCl]. The quality of combustion conditions (measured
by CO) was more influential in affecting yields than changes
in HCl concentration although, at fixed combustion condi-
tions, HCl had an exponential effect on PCDD/ F emisions.
†
C H U N W A I L E E
U.S. Environmental Protection Agency, Office of Research and
Development, National Risk Management Research
Laboratory (MD-65), Research Triangle Park,
4
Tests with CH combustion found higher yields than with
propane and ethylene. When radiative flames were produced
with fuel-rich combustion, yields increased, suggesting that
sooting combustion can affect PCDD/ F formation.
Halonen et al. (4) examined the effect of metal chlorides,
variously with tetrachloroethylene, cofired with an aliphatic
(oil) fuel into a 32 kW furnace. They found that small amounts
of chlorine were sufficient to result in formation of PCDD/ F,
which was greatly enhanced in the presence of a copper (Cu)
or iron (Fe) catalyst source. Moreover, the short residence
times (<3 s) led them to conclude that formation of PCDD/ F
was a fast process. Formation was found to be rapid, and
poor combustion conditions resulted in gas-phase homo-
logue profiles dominated by the less highly chlorinated
PCDD/ F.
Other tests for the effect of metal and waste cofiring on
PCDD/ F formation used a slip stream from a natural gas
combustor that cofired 1,6-dichlorohexane, passing through
a presooted, Cu-laden quartz tube (5). Extremely high levels
of PCDD/ F were found, and these levels persisted even when
the 1,6-dichlorohexane was ceased. These results suggest
that Cu-containing soot deposits have a residual, or
North Carolina 27711 and ARCADIS Geraghty & Miller,
Research Triangle Park, North Carolina 27709
This research examined the potential for emissions of
polychlorinated dibenzodioxin and dibenzofuran (PCDD/F)
from industrial boilers that cofire hazardous waste. PCDD/F
6
emissions were sampled from a 732 kW (2.5 × 10 Btu/
h), 3-pass, firetube boiler using #2 fuel oil cofired with 2,4-
dichlorophenol or 1,2-dichlorobenzene and a copper
naphthenate mixture. PCDD/F levels were significantly
elevated when improved combustion conditions (reduced
carbon monoxide, increased carbon dioxide) followed periods
of flame wall-impingement and soot formation/deposition
on the boiler tubes. Boiler tube deposits became a sink and
source for PCDD/F reactants (copper and chlorine) and
PCDD/F, resulting in continued formation and emissions long
after waste cofiring ceased. The role of deposits in
PCDD/F formation makes emissions dependent on current
as well as previous firing conditions, resulting in
“
memory”, effect that may play a significant role in field
mechanisms of PCDD/ F formation.
uncertainty regarding prediction of emissions based
solely on the type and rate of cofired hazardous waste.
To begin to characterize the potential for emissions of
PCDD/ Fs, an emission sampling program under varied
combustion conditions was undertaken using an industrial
boiler. Since the program could not practically reflect all of
the variety of waste types and systems in practice, efforts
were focused on understanding the general conditions under
which PCDD/ F formation could occur. This knowledge can
be applied to the broad boiler types, operating conditions,
and waste types present in this population to focus the effort
required to characterize industrial boiler emissions. Emis-
sions were sampled under varying boiler operating condi-
tions, chlorinated organic waste types, and in the presence
of a Cu catalyst. Soot deposits from three different locations
inside the boiler were sampled and analyzed at various times
throughout the program in an effort to understand the
relationship between PCDD/ F emissions and PCDD/ F in the
boiler deposits. The effect of soot deposits and their ability
to act as sites for further reactivity was addressed to
understand how general boiler operations can affect the
PCDD/ F formation potential.
Introduction
Emissions of polychlorinated dibenzodioxin and dibenzo-
furan (PCDD/ F) from industrial boilers that cofire hazardous
waste are scantly characterized. This potential source of
PCDD/ F emissions will be receiving attention from the U.S.
Environmental Protection Agency (EPA) as part of an effort
to consider revising emission standards for hazardous waste
combustors. It is common practice for combustible hazard-
ous waste, typically liquid waste, to be cofired in industrial
boilers. These boilers have a variety of designs (i.e., watertube,
firetube, and stoker), types of primary fuels (i.e., coal, oil,
natural gas), and amount and type of hazardous waste
cofiring. Data from two of the 136 hazardous-waste-firing
industrial boilers/ furnaces had PCDD/ F emissions averaging
<
2 ng/ dscm, considerably below existing regulatory limits
for other facilities (1). The average toxic equivalency (TEQ)
for these facilities is 0.64 ng/ kg of waste feed, although the
limited sample size provides low confidence that this value
can be defensibly applied to the whole population (2).
Experimental Methods
Emissions were sampled on a boiler facility in which cofiring
of hazardous waste was simulated by injection of chlorinated
and copper-laden wastes during varied combustion condi-
tions. The North American Package Boiler (NAPB) is a 732
*
Corresponding author phone:
(919)541-1534; e-mail:
gullett.brian@epa.gov.
†
‡
U.S. Environmental Protection Agency.
ARCADIS Geraghty & Miller.
6
kW (2.5 × 10 Btu/ h), 3-pass, firetube, marine utility boiler
1
0.1021/es991196+ CCC: $19.00

2000 Am erican Chem ical Society
VOL. 34, NO. 11, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2 0 6 9
Published on Web 04/14/2000

FIGURE 1. North American Package Boiler. Arrows indicate gas flow direction.
capable of firing natural gas or a variety of fuel oils (Figure
). The burner is a ring-type natural gas burner with an
atomizing air, center nozzle, oil burner capable of firing #2
through #6 oils. The boiler has 27.9 m (300 ft ) of heating
surface and generates up to 11 ton/ h (24 000 lb/ h) of saturated
steam at pressures up to 1.02 atm (15 psig). The steam
generated by the boiler is extracted by a condensing heat
exchanger tied to an industrial water chilling system to
simulate actual boiler load. Fuel flows are measured with a
liquid volume meter, and stoichiometric ratios are verified
in municipal waste combustors. Similarly, the chlorinated
aromatic feed rate was set to yield an HCl concentration
sufficiently high (100-500 ppm, 7% O ) so as to not limit the
potential for observable formation. The choice of 1,2-
dichlorobenzene and 2,4-dichlorophenol was based on the
desire to cofire a prechlorinated aromatic and their tested
ability to produce PCDD/ F (7).
1
2
2
2
Initial testing (0-38 h) on the NAPB established baseline
emissions measurement variability for PCDD/ F during firing
of #2 fuel oil with the new, clean boiler tubes. Subsequent
combustion conditions were intended to promote formation
of PCDD/ F. The NAPB was run under relatively “poor”
combustion conditions, believed to be conducive to PCDD/ F
formation. A high firing rate and slight burner nozzle
misalignment caused flame impingement on the Pass 1
combustion chamber wall, resulting in flame quenching
which produced sooting conditions, elevated CO levels, and
through O
tions.
2 2
and carbon dioxide (CO ) emission concentra-
Continuous emission monitors (CEMs) are coupled with
a data acquisition system to record CEM measurements as
well as steam and flue gas temperatures. Samples for CEM
analysis were taken through a probe inserted into the exhaust
stack, 2 m above the boiler. The sampled gases from the
stack or the boiler pass through a 20 µm filter and are
connected to two different CEM benches through heated
and insulated Teflon tubing lines. A Rosemount Analytical
2
relative “low” CO values indicative of incomplete combus-
tion. This period of the testing was retrospectively named
the “Sooting” phase. Test conditions after 107 h of burner
operation were altered to minimize flame impingement on
the walls of the primary combustion tube and to improve
the combustion efficiency. This improvement in combustion
conditions reduced the visible soot formation, increased the
bench measures O
while a Perkin-Elmer (MCS 100 Model) uses hot, wet gas
analyzers for HCl, sulfur dioxide (SO ), and water (H O). The
2 2
, CO , and CO via a hot, dry method,
2
2
polycyclic aromatic hydrocarbons (PAHs) were measured
by a Photoelectric Aerosol Sensor (PAS 2000) which photo-
ionizes particle-bound PAHs for a real-time, qualitative
indicator of combustion conditions.
Flue gas sampling for PCDD/ F followed U.S. EPA Test
Method 0023A (6), and samples were analyzed by high-
resolution gas chromatography/ high-resolution mass spec-
trometry (HRGC/ HRMS). Two VG 70 HRMS instruments were
used, on with a DB-5 column and one with a DB-225. The
DB-225 column was used to obtain a more accurate quan-
2
CO levels, and was subsequently observed to result in
significant PCDD/ F formation. Therefore, this period of the
test program was termed the “Formation” phase. The next
stage of the sampling program determined the residual or
“memory” effect of the NAPB on emissions of PCDD/ F.
During this stage, which lasted for about 400 h of burner
operation, only #2 fuel oil was fired, without any additional
nonfuel carbonaceous precursors, Cl, or Cu source. This stage
was termed the “Decontamination” phase. The last stage of
the work intended to repeat the conditions under which
formation was observed by sequential addition of the nonfuel
sources of Cl, C, and Cu(II). This last stage is called the
“Verification” phase. Variously throughout the testing, soot
deposits were sampled from each of the three boiler passes.
Solids were scraped off the inside of the tubes at the end of
each pass and subject to PCDD/ F analysis using a Hewlett-
Packard 5890/ 5971 [HRGC/ low (L) RMS system (J & W DB-
Dioxin column) and methods that are adaptations of EPA
Method 23 and SW-846 Method 8290 (8, 9). The method has
titation of the 2,3,7,8-TeCDF. Sample volume ranged from
3
4
to 5 dry, standard m (dscm). Test samples are denoted by
the cumulative hours of burner operation in which sampling
was initiated. Twenty-eight emission samples were gathered
during 600 h of burner operation over a 1 y period. Prior to
testing, all of the boiler tubes were replaced to provide a
clean boiler tuber surface. The NAPB was dormant between
the project’s last three of five testing phases, allowing time
for analysis of the samples. Since analytical results lagged
testing by at least 4 weeks, test conditions during each phase
were preplanned and could not be adjusted based on recent
prior results.
been expanded to include mono- to trichlorinated PCDD/ F
13C standards (10). Isotopically labeled internal standards
To simulate cofiring of hazardous waste, chlorine (Cl)
and carbon (C) precursor sources, either 1,2-dichlorobenzene
for each congener class were incorporated prior to extraction.
Recovery of spiked labeled internal standards was determined
by comparison with other labeled dioxin standards added
just before analysis. The spiked internal standard acceptable
recovery limits are 40-120% recovery with the lower limit
extended to 25% for mono-tri-PCDD/ F and hepta-octa-
PCDD/ F. Soot deposits sampled at the end of both the
formation phase and decontamination phase were subjected
to X-ray fluorescence (XRF), X-ray diffraction (XRD), and
inductively coupled plasma (ICP) spectrophotometry. Test
conditions are summarized in Table 1.
(
diClBz) or 2,4-dichlorophenol (diClPh), were injected with
the fuel oil. The Cu catalyst source was copper(II) naphth-
enate (a mixture of acyclic carboxylic acids, 77% mineral
spirit, 8% Cu) doped into the #2 fuel oil. The halogenated
aromatic, copper naphthenate, and #2 fuel oil were mixed
in a 38 L pressurized steel tank and metered into the #2 fuel
oil line for injection. Copper was injected at a nominally
constant rate (20 mg/ kg #2 fuel oil) anticipated to be sufficient
for PCDD/ F formation, based roughly on Cu/ fuel ratios found
2
0 7 0
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 11, 2000

TABLE 1. Test Conditions and Emission Resultsa
PCDD/Fc
(ng/dscm)
test hourb
organic/Cl source
metal source
[HCl] (ppm)
[CO] (ppm)
[O2] (%)
[CO2] (%)
[SO2] (ppm)
Baseline Emissions
1
3
4, 18, 31
8
none
none
none
none
0
0
0-49
>1%
2.9d
1.7
12.3, 12.8, 13.4
13.1
30
30
all ND
ND
Sooting Phase
0, 240
5
6
7
8
0, 57
7
5, 81
7, 93
1,2-diClBz
1,2-diClBz
2,4-diClPh
none (natural gas
firing)
Cu Naph
Cu Naph
Cu Naph
520
540
160
0
2.0d
2.4
12.2, 14.2
12.3
13.7, 13.7
8.0, 7.8
0
0
0
0
4, 5
ND
12, 10
4, 2
>1%
370, 750
30, 120
d
1.7
d
4.8
Formation Phase
1
1
1
1
1
07, 114
20, 126
32
38
47
1,2-diClBz
2,4-diClPh
2,4-diClPh
2,4-diClPh
2,4-diClPh
Cu Naph
Cu Naph
Cu Naph
Cu Naph
Cu Naph
535
240, 400
320, 480
165
1.5d
1.5
1.3
2.9
2.9
15.4, 15.2
15.3, 15.4
15.4
13.8
12.5
0
10
0
0
0
300, 307
257, 139
287
1953
588
d
95
165
416
451
19
12
Decontamination Phase
1
5
53, 158
53
none
none
none
none
0
0
6, 7
8
3.1d
2.8
13.1, 12.4
14.0
29d
34
111, 134
1
Verification Phase
5
5
5
5
5
5
5
58
64
77
81
86
92
98
HCl
none
none
218
8
5
4
140
5
2.7
2.7
3.1
1.4
3.1
2.9
3
14.5
14.3
14.2
15.3
14.5
14.1
14.0
0
0
0
0
0
4
6
2,4-diClPh
2,4-diClPh
2,4-diClPh
1,2-diClBz
1,2-diClBz
1,2-diClBz
466
411
375
538
97
Cu Naph
Cu Naph
Cu Naph
Cu Naph
Cu Naph
51
98
73
11
20
7
7
0
99
408
a
. b Italicized test hours are LRMS analyses. c Mono- to octa-CDD/F. d Average value for tests in row.
2
All concentrations reported dry at 7% O
reduced CO, PAHs, and visible soot formation. Subsequent
emission results (Figure 2) from 107 to 147 h showed
Results and Discussion
The four Baseline runs had nondetectable PCDD/ F formation,
even for the fourth run (38 h) when combustion was poor
and [CO] >1%. This suggests that trace levels of Cu and Cl
in #2 fuel oil are insufficient to result in detectable formation.
Tests during the Sooting phase aimed to determine whether
combustion conditions or type of simulated waste promoted
high PCDD/ F yields. Runs 50, 57, and 67 h, in which 1,2-
dichlorobenzene with copper naphthenate was cofired with
significantly higher levels of PCDD/ F formation. All but one
run had PCDD/ F emissions >250 ng/ dscm. The high PCDD/ F
formation rates are similar to those observed from previously
mentioned work with a chlorinated waste fired through a
Cu- and soot-laden tube (5). The highest formation, at 138
h, was 1953 ng/ dscm, which exceeded 49 ng I-TEQ/ dscm.
This value is over 2 orders of magnitude higher than proposed
regulations for hazardous waste incinerators in the U.S. The
next test, at 147 h of firing time, was run under virtually
identical conditions, although it was chronologically 10 weeks
later. Its emissions dropped 3-fold, to 588 ng/ dscm (≈15 ng
I-TEQ/ dscm). The deposits also showed extremely high levels
of PCDD/ F: samples from the Pass 2 deposits (138 h)
exceeded 16 000 ng/ g of PCDD/ F (see Table 2).
#
2 fuel oil, were conducted under relatively poor combustion
conditions. These conditions resulted in highly sooting, broad
flames (that impinged on the walls of the Pass one firing
tube), relatively low CO levels, and high levels of PAHs.
2
The first seven emission samples collected during the
initial 67 h of NAPB run time (both Baseline and Sooting
phases) did not result in formation of any significantly
quantifiable (via LRMS) levels of PCDD/ F. Extension of the
analytical methods to HRMS for the next two runs (75 and
These test phases indicated that either (1) poor combus-
tion conditions, resulting in highly sooting conditions, are
not necessarily sufficient to produce PCDD/ F levels of
concern or (2) sufficient time of operation may be necessary
to establish a surface deposit environment that is amenable
to PCDD/ F formation. Hypothesis #1 is consistent with the
8
1 h) likewise did not indicate appreciable formation. This
suggests that waste firing under poor combustion conditions
leading to high CO levels, visibly sooting flames, and low
CO
at 87 and 93 h were conducted without waste cofiring and
with #2 fuel oil replaced by methane (CH ) to determine if
2
does not necessarily lead to high PCDD/ F yields. Runs
observation that sufficient O
formation (11-13). It is possible that (a) in an O
2
is necessary for PCDD/ F
-limited
4
2
potential emissions from the sooting phase continued and,
if so, how long these emissions would persist (the reader is
reminded that samples were only analyzed between the last
three test phases, so previous runs’ results were unknown).
The low emissions from earlier runs at 75 and 81 h declined
slightly, but did not cease, to 4 and 2 ng/ dscm.
Prior to sampling at run time 107 h, the fuel injection
nozzle was realigned to be coaxial with the first pass, the
atomization air was adjusted to result in lower CO, and the
firing rate was reduced up to 20% to minimize flame
impingement on the near-burner, inner wall of the first pass.
This had the effect of improving the combustion conditions
environment, high concentrations of higher molecular weight
hydrocarbons are formed which compete for Cl with PCDD/ F
precursors, (b) in a fuel-rich, O -starved environment, the Cl
2
contained in the dopant feed is converted mostly to HCl
which is less likely to chlorinate organics, or (c) production
of molecular chlorine (Cl
reaction (14) is limited by O
2
) or Cl radical via the Deacon
2
availability. In fuel-lean flames,
although less hydrocarbon molecular growth and less
products of incomplete combustion (PICs) are produced,
the excess O increases the Cl (15), which in turn might have
2 2
a positive effect on PCDD/ F formation. Hypothesis #2
suggests that the presence of sufficient surface deposits acts
as a sink and mediating surface for all reactants. Pass 2, as
2
of the boiler, reflected by higher CO levels and (typically)
VOL. 34, NO. 11, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2 0 7 1

FIGURE 2. PCDD/F concentrations in the emissions and Pass 2 deposits (five data points only) during the Sooting (75-93 h), Formation
107-147 h), and Decontamination (153-553 h) phases (results prior to 75 h showed little discernible formation and are omitted).
(
profiles of the deposits reflect those of the emissions, despite
large vapor pressure differences between the homologues.
Even after stopping the chlorinated waste feed at 147 h, the
153 and 158 h emission homologue profiles show little change
TABLE 2. Deposit Sampling Time and Results
soot deposit PCDD/F
soot deposit PCDD/F
a
a
concn (ng/g)
concn (ng/g)
(
5
not shown) and, in fact, are quite similar to the profile at
53 h. This suggests that there are no significant mechanistic
test hour Pass 1 Pass 2 Pass 3 test hour Pass 1 Pass 2 Pass 3
8
7
393
0
2
548
558
564
577
598
190
227
241
113
97
1464
2778 1595
238
196
165
435
changes in the absence of waste feed. These emission and
deposit profiles persist throughout the remainder of the
testing, despite large variations in PCDD/ F formation.
1
1
2
3
38
58
98
78
253 16203 1512
524 11549 3936
1669 18409 3660
36
304
105
Subsequent runs were conducted without hazardous
waste cofiring (no diClPh, diClBz, or Cu ) so that only #2 fuel
oil was fired (the Decontamination phase). The reduced but
significant levels of PCDD/ F emissions indicated both the
extent of the residual, or memory, effect on the PCDD/ F
emissions and the persistent reactivity of the deposits. After
210
8211 4509
a
LRMS analyses.
shown later, is the region where the highest concentrations
of PCDD/ F, Cl, and Cu are found in the soot deposits. Thus,
the PCDD/ F emissions are shown to be dependent not only
on a temperature window favorable to their formation but
also on the soot deposit layer and copper/ chlorine concen-
tration. The extremely high emission values obtained during
run 138 h (0.2000 ng/ dscm) suggest that very high emission
levels are possible from these boiler sources during moderate
cofiring combustion conditions when a sufficient reactive
surface is formed after a certain boiler operation time. These
ideas are supported by similar findings by Lee et al. (5). While
the potential for these emission levels are clearly specific to
the operating conditions used, these results do suggest that
boiler designs which have high potential for sooting will also
have high potential for PCDD/ F formation.
1
1 h of run time, emissions had dropped considerably but
persisted at >100 ng/ dscm (tests 153 and 158 h in Figure 2).
Indeed, approximate calculations suggest that the rate of
post-waste-firing emissions (the average emissions from runs
153 and 158 h) should reduce the boiler deposit concentra-
tions (taken from the 138 h run, Table 2, Pass 2) to zero
within 100-200 h for PCDD and 60-70 h for PCDF, given no
further PCDD/ F formation reactions. However, the concen-
tration of PCDD/ F in the tube deposits, measured by
sampling tube scrapings, showed continued high concentra-
tions (samples at 158, 298, and 378 h) even though the
chlorinated organic waste feed was ceased after 147 h. Over
150 h after cessation of chlorinated organic feed, the 298 h
deposits indicate a PCDD/ F concentration similar to that
observed after the highest emitting test at 138 h (order of
magnitude comparisons only are appropriate given the
uncertainties in the soot deposit sampling procedures and
analysis). Only when post-Cl-feed run time reached 300 h
The molar PCDD/ F emission rate from the 138 h run is
-
6
roughly equivalent to 1 × 10 times its molar feed rate of
2
,4-diClPh. Despite fairly consistent run and feed conditions,
the changes in PCDD/ F emission levels during the Formation
phase (up to 10×) suggest that formation conditions are
actually fairly dynamic. The waste feed conversion rate varied
up to 2 orders of magnitude throughout the test program,
suggesting that observed emission levels are not a function
solely of waste or Cl feed rate but, rather, are affected by
other conditions such as combustion efficiency and/ or
surface effects.
(
2
run 548 h), did the Pass 2 solids concentration reduce below
000 ng/ g. These yield and homologue results from the gas
and solid samples show conclusively that the solid deposits
act as a sink and source for continued PCDD/ F formation
and that gas-phase emissions result from PCDD/ F vaporiza-
tion and/ or surface mediated reactions and subsequent
vaporization.
During the Sooting and Formation phases, the tetra- to
octa-CDD/ F homologue profiles shift with the onset of
significantly large emissions. Observation of representative
homologue emission profiles from 87 to 147 h (Figure 3)
shows that the tetra-CDD-dominated profiles (87 h) gradually
give way to profiles dominated by higher chlorinated DD
homologues. Similarly, the PCDFs evolve to a profile domi-
nated by the HxCDF and HpCDF homologues. The shift to
higher chlorinated PCDD/ F homologues with time was also
observed in the related fixed bed tests of Lee et al. (5). The
The soot deposits may serve as the source of PCDD/ F
emissions. A recent study (16) suggested that PAHs, which
are fuel byproducts often attached on soot particle surfaces,
are a carbon source for formation of polychlorinated
aromatics such as PCDF and polychlorinated naphthalenes.
Alternatively, or concurrently, the soot deposits may serve
as an adsorption substrate for any undestroyed diClPh or
diClBZ which, in turn, may undergo condensation reactions
to form PCDD/ F (see ref 7).
2
0 7 2
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 11, 2000

FIGURE 3. Homologue profiles for PCDD/F emissions (lines) from the early and late stages of the Formation phase compared with solid
deposits (bars, Pass 2).
FIGURE 4. PCDD/F concentrations in the emissions and Pass 2 deposits (four data points only) during the Verification phase (run 553 h
is the last emission sample from the Decontamination phase).
A final “Verification” phase was run to determine the
reproducibility of the PCDD/ F formation and to begin to
isolate causal emission factors. Results from sequential
introduction of Cl by source (HCl then diClPh), metal
the availability of gas-phase Cl or Cu. Return of the 1,2-
diClBz feed rate to the 586 h level plus addition of SO to 100
ppm (598 h) resulted in a 73% lower yield than the 586 h test,
showing that SO -induced prevention mechanisms (5, 17-
20) have at least a prompt (1-4 h) effect on formation.
Verification of the deposit sink/ source effect would suggest
that emissions can be substantially affected by the influence
2
2
catalysts, alternate organic waste source (diClBz), and SO
2
suppressant are shown in Figure 4. Introduction of HCl (558
h) and then 2,4-diClPh (564 h) both resulted in very minor
increases in the PCDD/ F yield, without yielding a distinction
between the Cl sources. These results suggest that the
reactant-depleted deposits did not immediately return to
their previous reactivity with only the introduction of a
chlorine source, HCl. However, addition of Cu naphthenate
to the 2,4-diClPh (577, 581 h) resulted in significant increases
in PCDD/ F levels. This suggests that the PCDD/ F formation
is Cu-limited at this point, either by the ability of the Cu to
act as a catalyst and/ or its ability to sequester then serve as
a transfer agent for Cl. Changing the organochlorine source
to 1,2-diClBz (586 h) resulted in a likely insignificant change
in PCDD/ F formation. A 4-fold reduction in the 1,2-diClBz
feed rate and Cu feed rate (592 h) resulted in a 4-fold drop
in PCDD/ F levels, suggesting that changes in waste feed rates
can have a substantive, first order effect on PCDD/ F emissions
even in the presence of potential concurrent memory effects.
This sensitivity to waste feed rate was not observed in the
Formation phase, suggesting that reactivity conditions then,
and the high PCDD/ F levels observed, were more likely
mediated by surface effects and less sensitive to changes in
2
of firing conditions (such as periodic SO increases) on deposit
properties for some time prior to the current sampling test,
as has been found in field scale municipal waste combustion
(20). Overall, the emissions during this Verification phase
did not reach levels observed throughout the Formation
phase, suggesting that the presence of a Sooting phase and
the subsequent build up of reactive deposits are important
to subsequent PCDD/ F formation rates.
Prior to firing changes at 107 h, only one deposit sample
was analyzed (at 87 h); it showed that PCDD/ F formation
(about 400 ng/ g) occurred only in Pass 1 (see Table 2). With
the change in firing conditions at 107 h of operation, the
highest PCDD/ F deposit concentrations were typically
observed in Pass 2. Conditions for formation and/ or deposi-
tion/ adsorption appear to be optimal in this region perhaps
due to the build up of the deposits. Pass 2 concentrations
(up to 18 000 ng/ g) were approximately 2-10 times the
concentrations observed in Pass 3 and 20-60 times the
concentrations observed in Pass 1. The flue gas temperature
values in the three passes were greater than 600 °C for Pass
VOL. 34, NO. 11, 2000 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
2 0 7 3

1
, 600-300 °C for Pass 2, and 300-130 °C for Pass 3. Pass 2
(5) Lee, C. W.; Kilgroe, J. D.; Raghunathan, K. Environ. Eng. Sci.
998, 15(1), 71-84.
1
appears to be in the oft-cited optimal temperature region for
PCDD/ F formation (21). The large differences in deposit
PCDD/ F concentrations in each pass are likely related to
effects of the gas temperature on gas phase species and
concentrations as well as the effect of surface temperature
on adsorption and reaction of precursors. XRF analyses of
soot deposit samples, taken after the emissions sample with
the highest PCDD/ F formation (138 h), showed that the
primary metals were iron (Fe), likely from oxidation of the
tube surfaces, and Cu, from the waste feed. ICP analyses
indicated that the Pass 2 Cu concentrations was 83.70 mg/ g,
while those in Pass 1 and Pass 3 were 5.81 mg/ g and 19.7
mg/ g, respectively. XRD analyses identified these to be
primarily CuCl, with some Cu(OH)Cl. XRF analysis verified
that Cl followed the Cu trends. After an additional 400 h of
undoped #2 fuel oil firing (test 548 h), the sampled Cu
concentrations in the soot deposits in Passes 1, 2, and 3 were,
respectively, unchanged, 6.13 mg/ g, and 1.45 mg/ g, reflecting
significant reduction. Similar reduction in the Pass 2 Cl
concentration from 54.2 to 6.6 kilocounts/ s (kcps) was noted,
confirming the correlation of Cu and Cl concentrations in
the deposits with high PCDD/ F concentrations.
(
6) U.S. EPA. Test Method 0023A, “Sampling Method for Poly-
chlorinated Dibenzo-p-Dioxins and Polychlorinated Dibenzo-
furan Emissions From Stationary Sources, in Test Methods for
Evaluating Solid Waste; SW-846 (NTIS PB88-239223); Environ-
mental Protection Agency, Office of Solid Waste and Emergency
Response: Washington, DC, December 1996.
(7) Ghorishi, S. B.; Altwicker, E. A. Haz. Waste Haz. Materials 1996,
3, 11-22.
1
(
8) Method 23. In Title 40 Code of Federal Regulations Part 60,
Appendix A; U.S. Government Printing Office: Washington, DC,
1
991.
(
9) Method 8290. In Test Methods for Evaluating Solid Waste, Volume
1B: Laboratory Manual Physical/Chemical Methods, 3rd ed.;
EPA-SW-846 (NTIS PB88-239223); U.S. Environmental Protec-
tion Agency: Washington, DC, 1986.
(
10) Gullett, B. K.; Ryan, J. V.; Tabor, D. Organohalogen Compds.
999, 40, 121-124.
1
(
(
11) Vogg, H.; Stieglitz, L. Chemosphere 1986, 15, 1373.
12) Hagenmaier, H.; Kraft, M.; Brunner, H.; Haag, R. Environ. Sci.
Technol. 1987, 21, 1080-1084.
(
13) Addink, R.; Olie, K. Environ. Sci. Technol. 1995, 29(6), 1586-
1590.
(14) Tilly, J. J. Chem. 1981, 55, 2069-2075.
(15) Procaccini, C. Ph.D. Thesis, Massachusetts Institute of Technol-
ogy, Cambridge, MA, 1999.
(
16) Iino, F.; Imagawa, T.; Takeuchi, M.; Sadakata, M. Environ. Sci.
Technol. 1991, 33, 1038-1043.
Acknowledgments
The authors wish to thank Shirley Wasson (U.S. EPA/ NRMRL-
RTP) for the XRD and XRF analyses and Dennis Tabor and
Bill Preston (Arcadis G&M) for the HRGC/ LRMS analyses.
(
17) Lindbauer, R. L.; Wurst, F.; Prey, T. Chemosphere 1992, 25, 1409-
1414.
(18) Raghunathan, K.; Gullett, B. Environ. Sci. Technol. 1996, 30(6),
827-1834.
19) Gullett, B. K.; Raghunathan, K.; Dunn, J. E. Environ. Eng. Sci.
998, 15(1), 59-70.
20) Gullett, B. K.; Dunn, J. E.; Raghunathan, K. Environ. Sci. Technol.
000, 34, 282-290.
21) Stieglitz, L.; Zwick, G.; Beck, J.; Roth, W.; Vogg, H. Chemosphere
989, 18(1-6), 1219-1226.
1
(
(
(
Literature Cited
1
(
1) U.S. EPA. Combustion Emissions Technical Resource Document,
EPA/ 530-R-94-014; Office of Solid Waste and Emergency
Response: Washington, DC, May 1994.
2
(
2) U.S. EPA. The Inventory of Sources of Dioxin in the United
States, Review Draft; EPA/ 600/ P-98/ 002Aa (NTIS PB98-137037);
Office of Research and Development: Washington, DC, April
1
1
998.
Received for review October 19, 1999. Revised manuscript
received February 10, 2000. Accepted February 29, 2000.
(
(
3) De Fre, R.; Rymen, T. Chemosphere 1989, 19, 331-336.
4) Halonen, I.; Tuppurainen, K.; Ruuskanen, J. Chemosphere 1997,
3
4(12), 2649-2662.
ES991196+
2
0 7 4
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 34, NO. 11, 2000