xvEPA
United States
Environmental Protection
Agency
Office of
Toxic Substances
Washington DC 20460
EPA-560/5-84-009
December, 1984
Toxic Substances
Thermal Degradation
Products from
Dielectric Fluids
-------
THERMAL DEGRADATION PRODUCTS FROM DIELECTRIC FLUIDS
by
Mitchell D. Erickson, Christopher J. Cole, Jairus D. Flora, Jr.,
Paul G. Gorman, Clarence L. Haile, Gary D. Hinshaw,
Fred C. Hopkins, and Stephen E. Swanson
WORK ASSIGNMENT NO. 23
INTERIM REPORT NO. I
EPA Contract No. 68-02-3938
MRI Project No. 8201-A(23)
November 19, 1984
For
U.S. Environmental Protection Agency
Office of Toxic Substances
Field Studies Branch, TS 798
401 M Street, S.W.
Washington, DC 20460
Attn: Frederick W. Kutz, Project Officer
Daniel T. Heggem, Work Assignment Manager
-------
DISCLAIMER
This document has been reviewed and approved for publication by
the Office of Toxic Substances, Office of Pesticides and Toxic Substances,
U.S. Environmental Protection Agency. The use of trade names or commercial
products does not constitute Agency endorsement or recommendation for use.
-------
PREFACE
This report presents the results of Work Assignment No. 23 on U.S.
Environmental Protection Agency Contract No. 68-02-3938, "Incineration Test-
ing of PCBs." The work was done at Midwest Research Institute (MRI) during
the period May 22, 1984 to November 19, 1984. Mitchell D. Erickson was the
MRI Work Assignment Leader. This report was prepared by Dr. Erickson,
Jairus D. Flora, Jr., Clarence L. Haile, Gary D. Hinshaw, and Stephen E.
Swanson. The thermal destruction system was operated by Mr. Hinshaw,
Christopher J. Cole, Paul G. Gorman, and Fred C. Hopkins. Laboratory work
was done by Mr. Swanson, with assistance from Alice Cheng, Michael McGrath,
and Edward Olsen. The GC/MS data were acquired by John Gamble, Jon Onstot,
Gil Radolovich, and Margaret Wickham. Mass spectral data were interpreted by
Dr. Erickson, Mr. Swanson, and Leslie Moody. Additional support was provided
by Audrey Sanford.
The EPA Work Assignment Manager, Daniel T. Heggem of Field Studies
Branch, provided helpful guidance and advice.
MIDWEST RESEARCH INSTITUTE
Clarence L. Haile
De/puty Program Manager
fohn E. Going
Program Manager
Approved:
ames L. Spigarelli, Director
Chemical and Biological Sciences
Department
m
-------
TABLE OF CONTENTS
Page
I. Introduction 1
II. Summary 1
III. Recommendations 2
IV. Background 3
A. PCDF Occurrence in PCB Fires 3
B. Reaction Mechanisms 3
C. Relationship of Conditions of PCDF Formation. . . 5
V. Experimental Plan 8
A. Thermal Destruction System 8
B. Phase 1 - System Refitting and Preliminary Plans. 8
C. Phase 2 - Experimental Optimization 9
D. Phase 3 - Aroclor 1254 Test Runs 12
VI. Experimental Methods 13
A. Reagents and Supplies 13
B. Destruction Facility Operation 15
C. Chemical Analysis 22
D. Statistical Analysis 30
VII. Results and Discussion 33
A. Phase 1 33
B. Phase 2 33
C. Phase 3 50
VIII. References 90
-------
LIST OF TABLES
Table Page
1 Thermochemical Conversion of PCBs to PCDFs 6
2 Concentrations of PCB Congeners in Mineral Oil for Phase 2 . 13
3 Concentrations of Aroclor 1254 in Feed Samples Used in
Phase 3 14
4 Phase 3 Surrogate Spiking Solution 24
5 Types of Analyses Used for Samples from Phase 3 27
6 Operating Parameters for Gas Chromatography/Mass Spec-
trometers Used to Analyze Phase 3 Samples 28
7 Scan Ranges and Selected Ions Monitored for Individual
Analytes 29
8 PCDD/PCDF Standard Used in Phase 3 31
9 Type of Quantisation Used During Phase 3 HRGC/EIMS Analysis. 32
10 Phase 1 Non-PCB Combustion Test Conditions . 34
11 Operating Conditions for Phase 2 Tests 36
12 Nominal and Actual Values for Operating Conditions During
Phase 2 Tests 37
13 Weights of PCBs Used During Phase 2 Tests 42
14 Weights of PCDFs in Combined XAD-2/Rinse Samples from
Phase 2 Tests 44
15 Conversion Efficiencies (PCBs to PCDFs) for Phase 2 Tests. . 45
16 Full Model Analysis of Variance 49
17 Reduced Analysis of Variance Model Using Only Temperature
and Oxygen 51
18 Means for Total PCDF Conversion Efficiency (%) Grouped
by Variable 52
19 Analysis of Variance for TetraCDF 53
20 Means for TetraCDF Conversion Efficiency (%) Grouped by
Variable 54
-------
LIST OF TABLES (continued)
Table Page
21 Operating Conditions for Phase 3 Tests 56
22 PCB Feed Characteristics in Phase 3. 57
23 Amounts of PCDFs Formed in Phase 3 60
24 PCDF Formation in Phase 3 61
25 Conversion Efficiencies (PCBs to PCDFs) for Phase 3 62
26 Amounts of PCDDs Formed in Phase 3 63
27 PCDD Formation in Phase 3 64
28 Conversion Efficiencies (PCBs to PCDDs) for Phase 3 65
29 Results of Analysis of PCBs in Phase 3 Samples ....... 83
30 PCB Destruction Efficiencies in Phase 3 Runs 84
31 Means of PCDF Formed in Phase 3, Grouped by Matrix and
Concentration 86
-------
LIST OF FIGURES
Figure Page
1 Laboratory-scale combustion system 16
2 Temperature profiles in combustion furnace 18
3 Sample analysis scheme 23
4 Two-part cleanup column 25
5 Continues gas monitoring results for Run 6-20-13-MMH 39
6 Continuous gas monitoring results for Run 6-19-11-MHM .... 40
7 Combustion efficiency versus temperature 41
8 Total PCDFs formed as a function of oxygen 47
9 Tetra CDFs formed as a function of oxygen concentration ... 48
10 Mono and di CDFs and CDDs in sample 8-22-52-S500 66
11 Tri and tetra CDFs and CDDs in sample 8-22-52-S500 67
12 Penta and hexa CDFs and CDDs in sample 8-22-52-S500 68
13 Hepta and octa CDFs and CDDs in sample 8-22-52-S500 69
14 Mono and di CDFs and CDDs in sample 8-30-62-ASKL 70
15 Tri and tetra CDFs and CDDs in sample 8-30-62-ASKL 71
16 Penta and hexa CDDFs and CDDs in sample 8-30-62-ASKL 72
17 Hepta and octa CDFs and CDDs in sample 8-30-62-ASKL 73
18 Tri and tetra CDFs and CDDs in sample 8-15-43-M5 74
19 Tri and tetra CDFs and CDDs in sample 8-17-47-S5 75
20 Tri and tetra CDFs and CDDs in sample 8-28-57-CLBZ 76
21 Average PCDF formation versus PCB concentration for Phase 3 . 78
22 PCDF formation in PCB-spiked mineral oil by homolog 79
23 PCDF formation in PCB-spiked silicone oil by homolog 80
24 PCDF formation from PCB askarel fluid 81
25 PCDF and PCDD formation from trichlorobenzene transformer
fluid 82
26 Comparison of PCDFs formed with PCB feed composition
(mineral oil) . . . 87
27 Comparison of PCDFs formed with PCB feed composition
(silicone oil) 88
28 Comparison of PCDFs formed with PCB feed composition
(askarel) 89
-------
I. INTRODUCTION
The Environmental Protection Agency (EPA) issued a final rule on
August 25, 1982, authorizing indefinite use of certain electrical transformers
containing polychlorinated biphenyls (PCBs). This rule is known as the PCB
Electrical Use Rule. At that time, information available to the EPA indicated
that fires involving electrical transformers were rare, isolated incidents.
However, several recent transformer fires in buildings have brought into ques-
tion EPA's earlier assumption. The Agency has therefore issued an Advance
Notice of Proposed Rulemaking (ANPR) (USEPA 1984a) and proposed rule (USEPA
1984b) to gather data on the specific risks posed by fires involving electri-
cal transformers that contain PCBs and also on mechanisms for mitigating or
eliminating these risks. Depending upon the results of EPA's analysis of
these data, the Agency may propose further control measures on the use of this
equipment.
This report describes the methods and results of a study, conducted
in support of EPA's data gathering activities under the ANPR, of the potential
for formation of polychlorinated dibenzofurans (PCDFs) and polychlorinated
dibenzo-p_-dioxins (PCDDs) from uncontrolled fires involving PCB-containing
dielectric fluids. The following two sections present a summary of the study
and recommendations for further study, respectively. Section IV provides a
brief literature review of PCDF and PCDD formation from PCBs. Sections V and
VI describe the experimental plan and methods. The results of the study are
presented and discussed in Section VII.
II. SUMMARY
At high temperatures, such as those in transformer fires, polychlo-
rinated biphenyls (PCBs) react to form polychlorinated dibenzofurans (PCDFs)
and other toxic by-products. The purpose of this study was to optimize condi-
tions for PCDF formation in order to examine the potential for formation of
PCDFs and polychlorinated dibenzodioxins (PCDDs) from combustion of selected
PCB-containing dielectric fluids.
The study was conducted in three phases. In Phase I, a bench-scale
thermal destruction system, developed by MRI, was refitted with specific com-
ponents installed to accommodate this study. Then, a few test runs were made
under preliminary temperature and oxygen conditions to ensure an acceptable
system blank. The concentrations of CO, C02, and 02 in the effluent were mon-
itored continuously. The entire effluent from the thermal destruction system
was passed through an XAD-2 trap to collect PCDFs and other semi volatile organ-
ics. The XAD-2 trap and a rinse of connective tubing were Soxhlet-extracted
and cleaned using column chromatography to isolate the PCDFs and PCDDs. All
samples were analyzed for PCDFs (and PCDDs in Phase 3), using high resolution
(capillary) gas chromatography/electron impact mass spectrometry (HRGC/EIMS)
in the selected ion monitoring (SIM) mode. No PCBs or PCDFs were detected in
the Phase 1 samples.
The Phase 2 experiments were conducted to determine the optimum tem-
perature, oxygen, and residence time conditions for PCDF formation. The feed
-------
into the system was mineral oil spiked with three individual PCB congeners
which form PCDFs by the four known reaction mechanisms. A statistical analy-
sis of the results for 33 runs indicates that both temperature and oxygen have
significant effects on the PCDF yield and that the interaction between temper-
ature and oxygen is synergistic. The results indicate that the optimum values
are a temperature of 675°C and an excess oxygen concentration of 8%. The res-
idence time did not significantly affect the yield in the range of 0.3 to 1.5
s, although lower times appeared to yield less PCDFs. Since residence time
does not appear to significantly affect the PCDF formation, 0.8 s was chosen
as convenient for the Phase 3 runs.
In Phase 3, duplicate test runs were conducted with mineral oil and
silicone oil dielectric fluids containing PCBs (Aroclor 1254) at concentrations
of 0, 5, 50 and 500 ppm. A PCB askarel containing 70% Aroclor 1260 and a non-
PCB askarel, containing mostly trichlorobenzene with some tetrachlorobe_nz
were also tested in duplicate. PCDFs were found in all samples,
PCDDs were found in the samples from the trichlorobenzene runs a
at low levels in some of the other samples. Up to 5,700 ng total PCDFs/mL of
spiked feed oil or 4% conversion efficiency (PCBs to PCDFs) was observed for
the mineral oil and silicone oil runs. The PCB destruction efficiencies cal-
culated for the 5, 50, and 500 ppm runs ranged from 79 to > 99%. Up to
19,000,000 ng total PCDFs/mL feed oil (19 mg/mL) or 3% conversion efficiency
was observed for the askarel fluid. Statistical analysis showed a linear
relationship for PCDFs formed versus the amount of PCBs. Although not statis-
tically different, about twice the quantity of PCDFs was formed from PCBs in
silicone oil than from the corresponding mineral oil samples. All eight PCDF
homologs were detected in the askarel runs with a maximum at the pentaCDF.
With a few exceptions, only tri- through hexaCDFs were observed in the lower
level runs, with a maximum generally at the triCDF homolog.
PCDFs and, to a lesser extent, PCDDs are formed from the trichloro-
benzene dielectric fluid under the optimum PCB-to-PCDF conversion conditions.
Up to > 110,000 ng total PCDFs/mL feed oil (> 0.004% yield) and 1,900 ng total
PCDD/mL feed oil (0.0001% yield) were observed for the trichlorobenzene runs.
The homolog distribution of PCDFs is similar to that found from feeding PCBs.
The amount of PCDFs formed is one to two orders of magnitude lower than for
the askarel, but substantially higher than that for dielectric fluid contain-
ing 500 ppm or less PCBs.
The results of this work indicate that the optimum conditions for
PCDF formation from PCBs are near 675°C for 0.8 s or longer, with 8% excess
oxygen. Under these conditions, PCDFs are formed from mineral oil or silicone
oil contaminated with PCBs at 5 ppm or greater. PCDFs are also formed from a
trichlorobenzene dielectric fluid which contained no detectable PCBs.
III. RECOMMENDATIONS
Further work on the Phase 3 sample extracts can yield important ad-
ditional information. These samples should be analyzed by full scan HRGC/EIMS
to (a) identify any other products of the thermal destruction, (b) confirm the
identity and amount of the PCDFs and PCDDs detectable by this technique, and
(c) attempt to identify and quantitate the polychlorinated biphenylenes (PCBPs).
-------
Additional statistical treatment of the data may be useful, especially if the
full scan HRGC/EIMS analysis yields additional quantitative information.
Additional runs of the thermal destruction system should (a) repli-
cate some of the runs reported here for QC purposes, (b) examine the PCDF for-
mation from other PCB mixtures, specifically Aroclors 1242 and 1260, and (c)
examine the formation of toxic products from other dielectric fluids, such as
tetrachloroethylene and HTH. Since the formation of PCDFs, PCDDs, and other
toxic products from the non-PCB fluids likely occurs via different reaction
mechanisms, the optimum conditions established for PCBs in Phase 2 may not be
appropriate for these fluids. Additional optimization runs should be con-
ducted for these fluids. This optimization should also be conducted on the
chlorobenzene fluid for which there are preliminary results in this report.
If additional information becomes available on the actual conditions
in transformer fires, the thermal destruction system could be operated at
these conditions to mimic the formation of toxic products during transformer
fires.
IV. BACKGROUND
A. PCDF Occurrence in PCB Fires
A number of fires involving electrical equipment containing PCBs
have been reported in Europe and the United States. Following a capacitor
fire in Sweden in 1978, Jansson and Sundstrom (1982) determined that the
amount of PCDF in the PCB oil had increased from about 1 pg/g before the fire
to an average concentration of 81 ng/g after the fire. Jansson and Sundstrom
also reported their analysis of the soot from a transformer fire in Toronto
(Canada) in 1979. The soot from this fire contained PCDF at 5 (jg/g soot. As
an aftermath of an electrical fire in Binghamton (New York) in 1981, soot was
spread throughout an office building. Soot samples were analyzed by two
groups of researchers (Smith et al. 1982) and 2,3,7,8-TCDF was determined in
the samples in the range of 3.7 to 2j160 ppm. Rappe et al. (1983) discuss
the results of analyses of PCDFs from various combustion sources. They re-
ported the results of the analysis of wipes from a metal treatment factory in
Skb'vde (Sweden) taken after an electrical fire in which 12 capacitors were
damaged. The results showed total tetrachlorodibenzofurans at 1 to 600 ng/m2
and pentachlorodibenzofuran at < 1 to 100 ng/m2. No other PCDFs were detected
above 100 ng/m2. Previously unpublished data from a number of other elec-
trical fires were presented in a state-of-the-art review by Vuceta et al.
(1983). The PCDF concentrations detected in transformer oils and in soot from
electrical fires indicate that chlorinated dibenzofurans are formed in elec-
trical fires and may pose a significant threat of exposure (USEPA 1984).
B. Reaction Mechanisms
The formation of polychlorinated dibenzofurans by air oxidation of
PCBs has been studied by only a few researchers. Morita (1977) reported that
heating Aroclor 1248 to 300°C in a sealed glass ampule for 2 weeks produced
approximately a four-fold increase in the amount of PCDF in the Aroclor.
-------
Morita again reported on the formation of PCDFs in Aroclor 1248 in 1978
(Morita et al. 1978). In the presence of oxygen, PCDF formation began when
samples were heated for 1 week above 270°C, reaching a maximum of approxi-
mately 0.2% conversion at 300°C. Dichloro- and trichlorodibenzofurans were
also formed from the tetrachlorobiphenyls in Aroclor 1248. Reaction mechan-
isms involving the loss of C12 or HC1 were postulated from the observations.
Buser et al. (1978) reported the formation of PCDFs from three in-
dividual PCB isomers. In these experiments, the PCB isomers were heated from
room temperature to 550 or 850°C in about 55 s and held at these temperatures
for 5 s. In these experiments, more than half of each of the chlorinated bi-
phenyl congeners (2,6,2',6'-tetrachloro-, 2,4,5,2',4',5'-hexachloro-, and
2,4,6,2',4',6'-hexachloro-) had decomposed at 550°C. Decomposition was es-
sentially complete above 650°C. The hexachlorobiphenyls formed tetrachloro-
and pentachlorodibenzofurans at 550-650°C at yields of 0.1% to 1.6%. The
tetrachlorobiphenyl formed dichloro- and trichlorodibenzofurans at 550° at
yields of 1.6 to 2.5%, depending on the temperature. At temperatures above
700°C, apparent complete destruction of the PCDFs was observed. It was pro-
posed from this work that a third reaction mechanism involving rearrangement
was responsible for formation of some isomers.
In the work of Buser and Rappe (1979), 18 individual PCB isomers
were pyrolyzed in the presence of air at 600°C in sealed ampules. The indi-
vidual PCDF isomers formed were identified. From this study four thermochem-
ical reaction mechanisms were proposed for formation of PCDFs from PCBs. The
four reaction mechanisms and the observed reactions are shown below.
1. Mechanism 1: Loss of Ortho-C12
Example:
Cl Cl
Cl Cl
Cl Cl
2,4,6,2',4',6'-Hexachlorobiphenyl 1,3,7,9-Tetrachlorodibenzofuran
2. Mechanism 2: Loss of HC1 Involving 2,3-Chlorine Shift at
the Benzene Nucleus
Example:
Cl Cl
Cl Cl
Cl Cl
2,4,6,2',4',6'-Hexachlorobiphenyl 1,3,4,7,9-Pentachlorodibenzofuran
4
-------
3. Mechanism 3: Loss of Ortho-HCl
Example:
Cl Cl
Cl Cl
Cl 'Cl
2,3,5,6-Tetrachlorobiphenyl 1,2,4-Trichlorodibenzofuran
4. Mechanism 4: Loss of Ortho-H?
Example:
Cl Cl
Cl
Cl Cl Cl u Cl
3,4,5,3',4',5'-Hexachlorobiphenyl 2,3,4,6,7,8-Hexachlorodibenzofuran
The results observed by Buser and Rappe are summarized in Table 1, which lists
each chlorinated biphenyl isomer, its chlorinated dibenzofuran reaction prod-
ucts, and the associated thermochemical reaction mechanism.
The studies of PCDF formation from PCBs are continuing. In his re-
view of PCDD and PCDF analyses, Rappe (1984) cites recent results giving fur-
ther support to thermochemical generation of PCDFs from PCBs. Rappe1s review
also discusses the work of Buser (1979) in which PCDF and PCDD formation was
observed during the sealed ampul pyrolysis of chlorobenzenes. Buser reported
a complex mixture of PCDD and PCDF isomers with a thermochemical conversion
of approximately 1%.
C. Relationship of Conditions to PCDF Formation
Morita et al. (1978) and Buser et al. (1978) each reported on the
conditions which produced maximum formation of chlorinated furans and the con-
ditions at which these furans were subsequently destroyed. Morita studied
PCDF formation from PCBs heated to 250 to 330°C for 3 to 28 days in air, oxy-
gen, and nitrogen atmospheres. The optimum furan formation conditions were
found by heating Aroclor 1248 in an oxygen atmosphere for 1 week at 300°C.
Under these conditions, a conversion rate of 0.2% was achieved. However, the
yield was quite temperature-dependent. At 270 and 330° the furan yields were
100 times less than at the 300° optimum. When the Aroclor 1248 was heated to
300° in an air atmosphere, the optimum furan formation was not achieved until
after 14 days of heating. Only 0.04% of the PCBs were converted to PCDFs.
-------
Table I. Thermochemical Conversion of PCBs to PCDFs
(From Buser and Rappe, 1979)
Reaction
mechanism
4
3
3
2
1
3
1
2
3
4
3
4
1
2
3
1
2
3
1
2
4
1
2
3
1
2
PCB congener
studied
2,3,4,5-
2,3,5,6-
2,6, 2' ,6'-
2
2
2
2
2
2
2
2
2
2
2
2
3
2
2
,3
,4
,4
,4
,4
,4
,4
,4
,4
,4
,3
,3
,4
,4
,4
,4
,5
,5
,5
,5
,5
,5
,6
,6
,6
,6
,6
,5
,5
,6
,5,
,2'
,2'
,2'
,2'
,3'
,3'
,2'
,2'
,2'
,2'
,2'
,3'
,2'
,2'
6-
,5'-
,5'-
,5'-
,5'-
,4'-
,4'-
,4'-
,4'-
,4'-
,5'-
,4',6'-
,4',5'-
,4',5'-
,4',6'-
PCDF reaction products
1,2,3,4-
2,3,4-*'
1.2.4-3
l,449-a>
1
1
2
2
2
1
2
1
1
1
1
1
1
1
1
1
1
1
2
2
2
1
1
1
a
'b
b
,9"
,2
,3
,3
,3
,3
,3
,3
,3
,3
,3
,4
,4
,2
,2
,4
,2
,2
•}
,3
,3
,3
,3
,3
,3,
4-
a
,8-b
,6,
,6,
,4,
,7,
,4,
,7-
,6,
,7,
,8-
,6,
,4,
,6,
,6,
,8,
,4,
4
,7,
,4,
,4,
,7,
,4,
8-
9-
6,
8-
7,
b
7-
9-
a,
8-
8-
9-
9-
9-
8,
6
8-
and 2,3,4,8d .
a
and l,3,4,8-d
9a
b
8-
and 2,3,6,7-a
and l,3,4,6,7-a
and l,3,4,7-d
b
a,
a
a
a
-* >
a
9b
7
b
and l,2,8-a'b
D, l,2,6,8-a, and
and l,4,6,9-a
l,2,6,9-a, and
and l,2,4,6,9-a'b
8-
7,8-
7,8-
9-
7,9-
-------
Table 1 (continued)
Reaction
mechanism
1
2
3
1
3
1
3
1
3
4
PCB congener
studied
2, 4, 5, 2', 4', 6'- 1
2
1
2, 3, 4, 2', 3', 4'- 3
1
2, 3, 5, 21, 3', 5- 2
1
2, 3,4, 5, 2' ,3', 4'- 2
1
1
,3
,3
,3
,4
,2
,4
,2
,3
,2
,2
,7
,4
,4
,6
,3
,6
,4
,4
,3
,3
,8-
,7,
,7,
,7-
,6,
,8-
,6,
,6,
,4,
,4,
PCDF reaction products
a,b
9-a and 1,3,4,7,8-
9-
7-
8-
7" a b
6,7-a and 1,2,3,6,7,8°
7,8,9-a
1,2,3,4,6,7,8- and
l,2,3,4,6,7,9-a
1
2
3
4
1
3
4
2,3,4, 5, 2' ,4' ,5'- 2, 3, 4, 7,8- u
2,3,4,6,7,8-
1,2,3,4,7,8- and l,3,4,6,7,8-a
1,2,3,4,6,7,9-
2, 3, 4, 5,2', 3', 4', 5'- 2,3,4,6,7,8- and (other hexa)
1,2,3,4,6,7,8-°
Octachlorodibenzofuran
,b
^Tentative isomer identification.
Major isomer among reaction products.
-------
The results reported by Buser et al. (1978) indicate that the opti-
mum conditions for furan formation may be different for each reaction mechan-
ism. When the reaction temperature was raised from 550 to 650°, the amount
of chlorinated furan formed by loss of ortho-C12 (Mechanism No. 1) decreased.
However, the amount of furan formed by loss of ortho-HCl (Mechanism No. 3)
increased. At a reaction temperature of 650 to 700°, no chlorinated furans
were detected. The reaction time for all of Buser1s experiments was stated
to be 5 s. However, a 55-s heating period and an undefined cooling period
may imply that the PCBs were hot enough for some PCBF formation for periods
longer than 5 s (i.e., a minute or more).
The incidence of transformer fires, their effects, and the forma-
tion of PCDFs from PCBs under fire conditions were recently reviewed by USEPA
(1984a) in the ANPR on use of PCBs in electrical transformers.
V. EXPERIMENTAL PLAN
A. Thermal Destruction System
MRI has developed a bench-scale thermal destruction test system
which can be used to examine various combustion processes. It has also been
used to provide data on the incinerability (i.e., destruction efficiency) of
hazardous compounds in solid or liquid waste material, and to provide data on
products of incomplete combustion that may be formed. This system can pro-
vide destruction data on gram-quantity samples of materials in either solid
or liquid form or even semisolid materials, such as tars. In contrast, other
related systems often can handle only very small quantities of pure compounds.
The system is described in detail in Section VI. Briefly, the sys-
tem consists of a volatilizing/pyrolysis heater for the sample, an air pre-
heater furnace, and the main combustion furnace (all electrically heated).
Separate volatilization/pyrolysis furnaces are used for sol id/semi sol id feed
and for liquid feed. Gas flow through the combustion furnace is laminar and
can be varied to provide different gaseous residence times. The combustion
temperature can also be varied, up to a maximum of 1200°C (2192°F).
Primary operating conditions that can be varied and controlled in
this system are temperature, oxygen concentration, and residence time. By-
products that can be determined include 02, CO^, CO, and total hydrocarbons
(by continuous monitoring) and volatile or semivolatile organic compounds (by
adsorption and concentration in a sampling trap, followed by extraction or
desorption and analysis). The system has been operated using continuous in-
jection (for liquid feed) and by batch feeding (for solid feed). For this
program, continuous liquid injection of feed was used.
B. Phase 1 - System Refitting and Preliminary Runs
Prior to testing with PCBs, the system had to be refitted to accom-
modate the special needs of this project and then tested without feeding PCBs
to ensure proper operation of the system and to assess the system blank for
interfering compounds and background levels of PCBs, PCDFs, and other analytes
of interest.
8
-------
C. Phase 2 - Experimental Optimization
1. Operational Conditions
Experiments in this phase were to investigate the influence of com-
bustion conditions on PCDF formation. The conditions which were to be varied
were oxygen concentration, incineration temperature, and residence time. The
other factors, the PCB concentration and matrix effects, were to be investi-
gated in Phase 3. Each parameter was initially evaluated at three levels,
high, medium, and low.
2. Values of Evaluation Parameters
The incineration temperature was to be evaluated at three levels.
High - 750°C
Medium - 600°C
Low - 450°C
These selected values were based on the PCDF formation conditions reported in
the literature.
The excess oxygen content, as measured after the combustion furnace,
was to be evaluated at three levels.
High - as large an excess as is feasible (e.g., 16%)
Medium - about 3%
Low - essentially 0%
These levels span the range of possible conditions in a fire, from oxygen-rich
to oxygen-starved.
The residence time was to be evaluated at three values.
High - 1.5 s
Medium - 0.8 s
Low - 0.3 s
These times span the operating range of the bench-scale reaction system.
These values will either identify an optimum residence time, or as a minimum,
show a trend line. It is not known how closely they approximate the residence
time of a PCB in a specific temperature region of a fire.
3. Experimental Design and Analysis of Results
The primary goal is to determine the values of temperature, oxygen
concentration, and residence time that will produce a maximum PCDF yield.
Under the assumption that the runs can be made sequentially and that the GC/MS
results can be obtained in time to plan the subsequent runs, a sequential ap-
proach to the experimental design is the most efficient means of identifying
the optimum conditions.
-------
The range of each factor was initially divided into three levels
with L, M, and H denoting the low, medium, and high levels of a given factor.
A complete factorial design of the three factors at three levels would entail
33 = 27 experiments. The levels selected represent points on a continuum.
The optimum is probably not precisely at one of the three values selected, so
an efficient method for converging on the optimum parameters was sought. If
only two levels (say L and M for purposes of demonstration) are selected, the
trend of the PCDF formation can be assessed with a complete 23 factorial de-
sign (8 runs). Even more efficiently, a half-fractional design (4 runs) can
still show the trends. An example half-fractional design is:
Run Temperature
1 L
2 L
3 M
4 M
Oxygen
L
M
L
M
Residence Time
L
M
M
L
Although this example used L and M values for all three parameters, other com-
binations could be used (e.g., L, M; M, H; L, H).
The results from the four runs listed above can yield one of four
conclusions:
(1) The response (PCDF yield) increases as the level increases for
all three factors;
(2) The response increases as the level increases for two of the
three factors, and decreases with increasing level of the third factor;
(3) The response increases as the level increases for one factor
and decreases with increasing level of the other two factors; or
factors.
(4) The response decreases as the level increases for all three
If situation (1) arises, the data imply that the optimum for all three param-
eters is between M and H. Therefore, the logical choice of a level for the
next iteration is the midpoint between M and H: (M + H)/2. The second set
of four runs would then be:
Run
5
6
7
8
Temperature
(M + H)/2
(M + H)/2
M
M
Oxygen
(M + H)/2
M .
(M + H)/2
M
Residence Time
(M + H)/2
M
M
(M + H)/2
The results of runs 5 through 8 will again yield one of the four situations
above. This process of iteration and narrowing the interval of the values
should result in a rapid convergence to the optimum values.
10
-------
If situation (2) arises, the second iteration of the half-fraction
of the 23 design would entail changing the low level of the two factors show-
ing increasing response with increasing levels to (M + H)/2, and changing the
low level of the third factor to (M + L)/2.
If situation (3) arises, one can proceed as in (2) with one substi-
tution of L to (M + H)/2 and two of L to (M + L)/2.
If situation (4) arises, the second iteration would be designed by
substituting all L levels with (M + L)/2.
This procedure can theoretically be iterated until the optimum com-
bination of levels of the three factors which yields the highest response is
found. After the first set of four runs, the maximum is likely to be within
R/2 for each factor, where R denotes the range for a given factor. After a
second set of four runs, it is likely to be within R/4. After a fourth set
of runs, it is likely to be within R/16 on each variable. Thus, the best a
priori knowledge about the ranges of the different factors will accelerate
the convergence process to locate the solution.
The procedure described above can allow for readjusting the upper
or lower limit of the factor ranges, if necessary. This might be necessary
if the ranges were chosen without sufficient data. For example, if the opti-
mum residence time is higher than 1.5 s, then one could raise the upper limit
in subsequent iterations, subject to limitations imposed by the reaction
system.
In the absence of problems, two runs can be made per day, and the
results from the GC/MS can be obtained about 3 days later. This would then
require a week for the results from four runs and thus require at least a
3-day shut-down of the thermal destruction facility between each set of runs.
In view of the project deadlines, these delays would have been intolerable.
Therefore, after the first set of four runs, a second set was to be conducted
immediately thereafter. This second set of four runs would begin with the
other half-fraction of the 23 factorial design, using the medium-high combina-
tion if the first sequence used low-medium. As the experiments progress, the
results of both sequences were to be considered and the levels of the factors
adjusted appropriately. Since at each stage the next set of runs was to be
determined on only that part of the data that was available from the GC/MS
analysis, convergence may not have been as efficient, but was more rapid.
4. Selection of Compounds for Evaluation
The Phase 2 incineration tests were performed using individual PCB
congeners rather than an Aroclor mixture. This approach simplified the chem-
ical analysis and data evaluation associated with selecting the optimum condi-
tions for use in Phase 3. The use of specific congeners also allowed assess-
ment of the PCDF formation via all four proposed reaction mechanisms (Buser
and Rappe 1979). Optimization using a commercial mixture such as Aroclor 1254
would likely have generated a confusing array of products.
11
-------
Three PCB congeners were selected for the Phase 2 experiments:
(a) 2,3,5,6-tetrachlorobiphenyl,
(b) 3,4,,5,3',,4',5'-hexachlorobiphenyl, and
(c) 2,4,6,2',4',6'-hexachlorobiphenyl.
The first compound, 2,3,5,6-tetrachlorobiphenyl (Ballschmiter No. 65), forms
1,2,4-trichlorodibenzofuran by loss of ortho-HCI (Mechanism No. 3). The com-
pound 3,3',4,4',5,5'-hexachlorobiphenyl (No. 169) forms 2,3,4,6,7,8-hexachlo-
rodibenzofuran by loss of ortho-H2 (Mechanism No. 4). The final compound,
2,2',4,4',6,6'-hexachlorobiphenyl (No. 155), forms two PCDF reaction products
in nearly equal amounts (Buser et al., 1978). The reaction product 1,3,7,9-
tetrachlorodibenzofuran is formed by loss of ortho-C!2 (Mechanism No. 1) and
1,3,4,7,9-pentachlorodibenzofuran is formed by loss of HC1 with a 2,3-chlorine
shift at the benzene nucleus (Mechanism No. 2).
Using these three PCB congeners, only four chlorinated dibenzofurans
should be formed and each thermochemical reaction mechanism could be studied.
The identification of thermochemical reaction products was also simplified
since only one each of the tri-, tetra-, penta-, and hexachlorodibenzofurans
are produced. In the one reaction mechanism where two products are formed
(2,2',4,4',6,6'-hexachlorobiphenyl to 1,3,7,9-tetrachlorodibenzofuran and
1,3,4,7,9-pentachlorodibenozfuran), these two products are produced in nearly
equal amounts (Buser and Rappe 1979). Although the PCB congeners chosen for
Phase 2 are not found in Aroclors at significant levels, these compounds are
commercially available and produce a single isomer of each PCDF homolog as
reaction products using all four reaction mechanisms. Hence, they were appro-
priate for the proposed range-finding experiments.
D. Phase 3 - Aroclor 1254 Test Runs
During Phase 3, PCBs in dielectric fluids were to be subjected to
the conditions determined to be optimal for PCDF formation in Phase 2. The
PCB mixture chosen was Aroclor 1254, a commercial mixture commonly used in
transformer askarels. The concentrations were to range from 5 ppm to about
70% PCB. The samples used were:
(a) 50 ppm PCB in mineral oil,
(b) 50 ppm PCB in silicone oil,
(c) 500 ppm PCB in mineral oil,
(d) 500 ppm PCB in silicone oil,
(e) Chlorobenzene dielectric fluid,
(f) 5 ppm PCB in mineral oil,
(g) 5 ppm PCB in silicone oil, and
(h) Used askarel-type dielectric fluid (e.g., 70% Aroclor 1254/30%
trichlorobenzene, Type D).
12
-------
VI. EXPERIMENTAL METHODS
A. Reagents and Supplies
1. PCBs
In Phase 2, three PCB isomers--2,3)5,6-tetrachlorobiphenyl,
3,3',4,4',5,5'-hexachlorobiphenyl, and 2,2',4,4',6,6'-hexachlorobiphenyl--
were added to the mineral oil (Exxon HPLX 355077) feed. These compounds were
purchased from Ultra Scientific. The concentrations of the feed solutions
are presented in Table 2.
Table 2. Concentrations of PCB Congeners in Mineral Oil for Phase 2
Concentrations (mg/mL)
2,3,5,6- 3,3',4,4',5,5'- 2,2',4,4' ,6,6'-
Run no. Tetrachlorobiphenyl Hexachlorobiphenyl Hexachlorobiphenyl
5-7
8-14
15-27
28-38
0.54
0.51
0.50
0.59
0.37
0.48
0.46
0.59
1.08
1.2
1.0
1.2
Four types of dielectric fluids were fed during Phase 3: mineral
oil, silicone oil, a chlorobenzene fluid, and an askarel. The mineral oil
(Exxon type HPLX 355077) and silicone oil (Union Carbide type L-305) were
spiked with Aroclor 1254 at 5 to 500 ppm. The actual concentrations spiked
are shown in Table 3. The chlorobenzene fluid (mostly trichlorobenzene iso-
mers with some tetrachlorobenzene, Electro-Chem FR-15, Standard Chlorine
Chemical Company, Kearney, New Jersey) and askarel (PPM, Inc., Kansas City,
Missouri) were run without modification. The askarel was obtained from a
transformer draining operation and was characterized (by MRI) as containing
70% Aroclor 1260 and the balance trichlorobenzenes.
13
-------
Table 3. Concentrations of Aroclor 1254 in Feed
Samples Used in Phase 3
Concentration of Aroclor 1254
Run type Run no. in feed (ug/g)
5 ppm Mineral oil
5 ppm Si li cone oil
50 ppm Mineral oil
50 ppm Si li cone oil
500 ppm Mineral oil
500 ppm Si li cone oil
43,
47,
45,
49,
39,
51,
44
48
46
50
40, 41
52
5
5
50
50
500
500
2. Calibration Gases
A Scott Specialty Gases Acublend mixture of 15.05% 02, 12.34% C02,
and 2,069 ppm CO was used during Phase 2 and part of Phase 3. A second
Acublend mixture with 15.11% 02, a lower C02 concentration of 4.10%, and a
higher CO concentration of 3,987 ppm was used for the remaining Phase 3 runs.
3. Surrogates, Standards, Reagents, and Adsorbents
Surrogate spiking compounds, (13C12)-2,3,7,8-tetrachlorodibenzofuran
and (13C12)-2,3,7,8-tetrach1orodibenzo-p_-dioxin, were purchased from Cambridge
Isotopes. During Phase 3 a column recovery surrogate solution was also used.
This column recovery mixture was purchased from KOR Isotopes and contained a
mixture of 37Cl-labeled tetrachlorodibenzofurans and pentachlorodibenzofurans.
Tetrachlorodibenzofurans were quantitated against a standard pur-
chased from Cambridge Isotopes containing unlabeled 2,3,7,8-tetrachlorodiben-
zofuran. Tetrachlorodibenzodioxins were quantitated against a 2,3,7,8-tetra-
chlorodibenzodioxin standard reference solution (Lot No. 20603-01/83) supplied
by the EPA Quality Assurance Materials Bank, EMSL-LV.
All solvents used for extraction, probe rinses, and sample cleanup
were Burdick and Jackson "Distilled-in-Glass" grade. The 100-200 mesh acid
alumina used for sample cleanup was Part No. AG-4 purchased from Bio-Rad
Laboratories. The silica gel used for sample cleanup was Kieselgel 60 (70-230
mesh) purchased from L. M. Reagents.
Each 200-g batch of XAD-2 (Supelco) resin was precleaned by succes-
sive washing in a continuous extraction column with:
14
-------
• One liter distilled water
• One liter 0.1 N sodium hydroxide
• One liter distilled water
• One liter 0.1 N hydrochloric acid
• One liter distilled water
• One liter methanol (2 times)
The resin was then Soxhlet-extracted 24 h with methanol, followed
by 24-h Soxhlet extractions with acetonitrile, and methylene chloride. It
was then placed into a container with a screw~cap top and dried in a vacuum
oven at 110°C for several hours. The container was capped immediately upon
removal from the oven. The resin was stored under methanol in a tightly
capped jar until it was loaded into the sampling cartridges.
B. Destruction Facility Operation
1. System Refitting
A schematic diagram of the MRI laboratory-scale system is shown in
Figure 1. The system was previously used in a batch fed mode of operation.
For the current program, the system was first refitted to the continuous liq-
uid injection mode. This operation involved disassembly at the tee located
near the incineration furnace inlet, removal of the vaporization/ pyrolysis
furnace chamber (tube section) used for batch fed operation, replacement of
the vaporization/pyrolysis furnace chamber (tube section) used for continuous
liquid injection, installation of the pumping system (syringe pump, polytetra-
fluoroethylene lines, and flowmeter), and replacement of all heating tape and
insulation. While disassembled, all tubes and chambers were cleaned with
solvent and allowed to dry. All connections and fittings were carefully in-
spected and replaced if defective. The system was then reassembled and leak
checked.
Modifications were also performed at the outlet side of the incin-
eration furnace. The condensation tube was modified so that it was heated to
~ 300°C above a compression nut fitting. A short length (6-12 in.) of 3/4
in. stainless steel tubing was attached at the compression nut fitting. The
other end of the tube was connected to a ball joint fitting that attaches to
the XAD-2 trap. This section allows cooling of the heated gases prior to
entry of the XAD-2 trap and forms part of an interchangeable sampling train
assembly. Thus any PCDFs that condense in the cool-down tube are recovered
during the rinse of the train.
After reassembly of the system, the furnaces were operated at high
temperature for several hours, with air, to bake out any organic residues.
2. Feed Oil
The composition and nature of the various dielectric fluids that
were subjected to thermal decomposition during this program are discussed in
Section V.
15
-------
Interchangeable
Samplin9
Met°r Assembly
Exliuust
0.2-2.0//min
Figure I. Laboratory-scale combustion system.
16
-------
The fluid feed rate was maintained at a constant 13.5 uL/min for
all of the runs. The constant displacement pumping system consisted of a
syringe pump (Sage Instrument Model 355) and 30-mL glass syringe with a PTFE-
tipped plunger (Manostat P5178-30-L Varipet). Narrow-bore PTFE tubing (1/16
in. ID) connected the pumping system to the thermal destruction system, pass-
ing the feed through a flow meter (Manostat 36-541-03) that verified constant
flow. The pump/syringe combination was calibrated by pumping a liquid of
known density (mineral oil) through the system into a vial which was placed
on a top-loading analytical balance. The weight of the pumped liquid was
periodically recorded to verify both the amount of oil pumped and the consis-
tency of the flow rate.
3. Thermal Destruction System
a. System Description
As discussed above, the system was operated in a continuous injec-
tion mode, using a motorized syringe pump to slowly feed a liquid into a
vaporization/pyrolysis furnace (see Figure 1). The vaporization/pyrolysis
furnace consisted of a section of 3/4-in. stainless steel pipe heated by an
electric tube furnace to a nominal operating temperature of 300°C. The feed
oil solution entered the heated pipe section in the center of the furnace via
a 1/16-in. diameter section of stainless steel tubing connected to the pipe
with appropriate reducers. A small flow of inert carrier gas (prepurified
nitrogen) entered the vaporization/pyrolysis furnace as shown and continuously
purged the chamber of vaporized waste. The flow rate used for this carrier
gas was a constant 0.05 L/min. These gases exited the vaporization/pyrolysis
furnace through a 1/4-in. diameter Inconel® tube.
A mixture of prepurified nitrogen and room air was first passed
through a charcoal trap and then preheated (nominally to 900°C) before mixing
with the waste vapors at a tee (maintained at 300°C). The resulting mixture
was passed through a short section of 1/4-in. diameter Inconel® tubing into
the combustion furnace where exposure of the waste/air mixture occurred at a
constant elevated temperature for a predetermined residence time. The com-
bustion furnace consisted of a section of 3/4-in. Inconel® pipe, heated by a
larger electric tube furnace. The effective volume of the incineration fur-
nace was ^ 50 cm3. A rupture disk was provided at a point just outside the
combustion furnace as a safety measure, should any explosive condition arise
in the furnace.
The temperature profiles of both the pyrolysis furnace and the
combustion furnace were determined by measuring gas temperatures at incre-
mental distances using a thermocouple probe. This characterization of the
furnaces was performed prior to running any destruction tests. The permanent
locations for thermocouples inside these furnaces was selected from the re-
sults of those studies. Also, the effective volume of the incineration fur-
nace was determined by multiplying the cross-sectional area enclosed by the
Inconel® pipe by the axial distance through the electric tube furnace in
which at least ± 25°C of the desired temperature was maintained. The mean
gaseous residence time was determined by dividing this effective volume by
the temperature-corrected total gas flow rate through the furnace. Figure 2
17
-------
500
00
24
26
28
30
32
34
36 44
Electric Tube
i i
46
Prof Me at 450° C
Nominal
Furnace Temp.
Profile at 700° C
Nominal
Furnace Temp.
Temp. Monitor-
ing Point
(Thermocouple)
Used for Tests
J 1-
-t75Q
740
730
720
710 £
O
700 |
690 |
f.
680 J
670
660
650
640
48
50
Distance (cm \ from Inlet End of Combustion Furnace
Figure 2. Temperature profiles in combustion furnace.
-------
shows the profiles of the combustion for two furnace temperature set points,
using a typical gas flow rate. The method of obtaining the effective distance
(length) of the retention volume is illustrated by the shaded regions. A
length of 15 cm was used to calculate the effective furnace volume for all
runs in Phases 1, 2, and 3.
After exiting the combustion furnace, the effluent gases passed
through a vertical section of 3/4-in. stainless steel pipe where cooling of
the hot gases occurred. After cooling in this condensation tube, the gas
stream entered an XAD-2 adsorbent resin sampling trap that collected semivol-
atile organic compounds. The sampling system is described in detail below.
A vacuum pump pulled the gases through the sampling train.
The gases then passed sequentially through continuous gas mon-
itors for oxygen (02), carbon dioxide (C02), and carbon monoxide (CO) deter-
mination. The operation of these instruments is described below. The gases
were pumped through a dry gas meter (Singer Model 817) before venting into a
Class A laboratory fume hood.
Water manometers were used to monitor the pressure of the com-
bustion furnace relative to atmospheric pressure and also the pressure drop
across the XAD-2 sampling trap (to indicate plugging of the trap). Locations
of pressure sensing points are shown in Figure 1 above.
Thermocouples were used to monitor both refractory and internal
gas temperatures in the vaporization/pyrolysis furnace, the makeup air heater,
and the combustion furnace. Gas temperatures were also monitored in the
vaporization/pyrolysis furnace outlet, the tee (combustion furnace inlet),
and the combustion furnace outlet, and the condensation tube inlet. Electric-
ally powered heating tape maintained the gas temperature on either side of
the incineration furnace near 300°C. Locations of all thermocouples in the
system are also shown in Figure 1.
b. System Operation
The generalized operating procedures for the destruction system
are as follows. At the beginning of each working day, the entire system was
first leak-checked. All furnaces and heaters were then turned on at desired
settings and allowed to warm up.
The nitrogen carrier gas, used to purge the vaporization/
pyrolysis furnace, was started and the flow rate checked for stability. The
pump was turned on and the total flow rate adjusted to pull air through the
system for the desired temperature-corrected residence time in the combustion
furnace.
The combustion system was fitted with a special assembly sim-
ilar to the XAD-2 sampling train assembly, which contained activated charcoal
and XAD-2 resin as adsorbents for organic compounds. This served as a safety
measure for operating periods when a test was not actually being conducted.
This provided for capture of any toxic compounds in the effluent gases while
the facility was in operation, and no test was in progress. The activated
charcoal was replaced routinely and disposed of as hazardous waste.
19
-------
The continuous gas monitors were calibrated with standards and
zeroed as necessary. The ratio of air and nitrogen in the makeup gas heater
was then adjusted so that the effluent 02 concentration was near the desired
operating value. Final adjustment for effluent 02 occurred after flow of the
feed oil was started and the continuous monitors had stabilized. A detailed
discussion of the continuous gas monitor operation is provided below.
The syringe pump was loaded with the desired sample and the
pump was started when the proper operating conditions were achieved. When
all operating conditions were stabilized after adding feed oil flow, a sam-
pling train assembly (described below) was installed, replacing the assembly
containing activated charcoal. This step marked the beginning of a test run.
Once the sampling train was installed, the test start time and initial gas
meter volume were recorded in a laboratory notebook. Operating data were re-
corded every 10-15 min, including feed oil flow rate, total gas flow rate,
nitrogen carrier gas flow rate, makeup airflow and nitrogen rates, continuous
gas monitor readings, and appropriate pressures and temperatures.
After sampling for the desired time (typically 60 min), the
test was concluded by removing the sampling train assembly and replacing it
with the activated charcoal assembly. The ending time and final gas meter
volume were then recorded. Appropriate changes in operating conditions were
made and the system allowed to stabilize before another run was initiated.
Quality control measures were integrated into the preparation
of the system for this program and also into daily operations. During the
refitting activities, temperature profiles were checked, flow meters and dry
gas meters calibrated, and the system checked for leaks to ensure proper per-
formance. During each test run, the operation was checked by monitoring the
pressure and temperature sensors. Temperature and flow rate were monitored
and recorded at least every 15 min. All of the continuous monitors were cali-
brated and checked according to either manufacturer's instructions or in-house
standard operating procedures (SOPs). In addition, some of the test runs were
blank (no feed) runs to check carryover of analytes in the system.
Before any organic samples were collected, the combustion sys-
tem was operated under a variety of conditions during preliminary testing.
Only after the system was demonstrated to be capable of performing under most
of the operating conditions desired for this program did testing commence.
4. Sample Collection
An interchangeable sampling system was designed, and several units
were fabricated and tested before use. This sampling system consists of a
stainless steel condensation tube, XAD-2 trap, cyclone, condensate collection
flask, and appropriate fittings and transitions. The sample collection sys-
tem for the thermal destruction tests is outlined by dashed lines in Figure 1,
showing its relationship to the rest of the thermal destruction system. All
effluent gases were drawn through this sampling train.
20
-------
The sampling system was designed to collect all organic constitu-
ents of interest. Condensation of these organic compounds prior to entering
the sampling system was prevented by maintaining the incinerator exit above
300°C. A thermocouple was used to monitor the gas temperature immediately
ahead of the sampling assembly. The condensation tube provided a temperature
gradient from > 300° to ambient, before the effluent gases entered the sorbent
trap.
Each effluent sample consistgd-tsf two parts: the XAD-2 resin car-
tridge and a solvent rinse of the^s«ifiple collection apparatus. Following each
experimental run, the XAD-2£papwas removed from the sampling assembly, the
ends capped with precl^afrea aluminum foil, and the trap labeled. The conden-
sation tube was^tireticl amped to the sampling cyclone, and
The solvent rinse was collected in a
abel ed bottle. Both the XAD-2 trap and rinse sample were stored
"at 4°(T~until the samples were prepared for analysis.
5. Continuous Monitoring of 02, C02, and CO
The 02, C02, and CO concentrations in the effluent gases were con-
tinuously monitored during each run. The 02 concentration is an evaluation
parameter selected for Phase II experiments. The C02 and CO concentrations
were measured in the effluent gas so that the combustion efficiency could be
calculated. These monitors and their operation are described below.
Oxygen was measured using a Beckman 7003 polarographic analyzer.
In this type of analyzer, oxygen diffuses across a membrane to a cathode where
it is electrochemical ly reduced. This results in a current flow proportional
to the partial pressure of oxygen in the sample. The instrument can operate
in four ranges, 0-1%, 0-5%, 0-10%, and 0-25% 02. Calibration was performed
by spanning with a specific calibration gas and zeroing with nitrogen gas.
Instrument precision of 0.3% absolute concentration or ± 6% of a measured
concentration, whichever is greater, at 95% confidence intervals, can be ob-
tained, according to the manufacturer's specifications.
The Horiba Model PIR-2000S carbon dioxide analyzer uses a nondisper-
sive infrared (NDIR) method of detection. The instrument can operate in three
ranges, 0-5%, 0-15%, and 0-25% C02. Concentrations can be read directly on a
meter and also displayed on a recorder'. Calibration is performed by spanning
with a calibration gas at the desired concentration and zeroing with nitrogen
gas. Instrument precision of < 0.1% absolute concentration or ± 0.6% of a
measured concentration, whichever is greater, at 95% confidence intervals,
can be obtained, according to the manufacturer's specifications.
A Horiba PIR-2000L carbon monoxide analyzer also uses an NDIR de-
tector. A silica gel/Ascarite cartridge was used to remove interferences such
as C02 and water.
Multi component standard gas mixtures were used to concurrently cali-
brate all three instruments. Two tees in the sample line allowed the sample
stream from the reaction system to be vented and the calibration gas to be
directed to the monitors. This allowed a calibration of the instruments to
be done immediately before and after each run.
21
-------
A data-logging system, consisting of an Epson HX-20 portable com-
puter with a Wintec MCS data interface, was used to collect and reduce the
data for each run. Concurrently, a strip chart recorder for each monitor gave
a real-time, visual indication of the state of the run. An advantage of the
data-logging system was that it printed out a minute-to-minute quantitation
of the three gases, automatically correcting for instrument nonlinearity and
for C02 removal in the CO analyzer. In addition, the computer was able to
estimate a corrected CO value up to 2-1/2 times the calibration span. Three
records of the monitoring data are retained: the strip chart recording, the
minute-to-minute run data from the computer, and the reduced data from the
computer.
C. Chemical Analysis
Each effluent sample consisted of two parts, the XAD-2 resin car-
tridge and a solvent rinse of the sample collection apparatus. As shown in
the analysis scheme in Figure 3, the XAD-2 samples were Soxhlet-extracted.
This extract combined with the associated solvent rinse to make a combined
effluent extract. The combined extract was evaporatively concentrated to 2
ml and one-half of each extract was cleaned by chromatography on acidified
silica and acidified alumina. These cleaned extracts were analyzed for chlo-
rinated biphenylenes, chlorinated dibenzodioxins, and chlorinated dibenzo-
furans. The fraction of each extract which had not been cleaned up was an-
alyzed for PCBs. Some of these extracts were screened for other chlorinated
organics. The specific analysis procedures are given below.
1. Sample Extraction and Concentration
The contents of the XAD-2 resin trap were transferred from the sam-
pling cartridge to a Soxhlet extractor. The resin was then spiked with surro-
gate compounds. During Phase 2, each XAD-2 resin was only spiked with 25 ng
of (13C12)-2,3,7,8-tetrachlorodibenzofuran in isooctane. For Phase 3, each
XAD-2 was spiked with the surrogate solution listed in Table 4. The resin
samples were then extracted for approximately 16 h with benzene. This extract
was combined with the rinse of the sampling apparatus and concentrated to < 5
ml using a Kuderna-Danish concentrator. Each extract was further concentrated
to 2 mL under a gentle stream of dry nitrogen.
22
-------
XAD Resin
-Add 13C Surrogate(s)
Soxhlet Extract
Apparatus Rinse-
Resin
Extract
K-D
Split Extract.
1/2 of Extract
Column Clean-up
Acidified Silica
Acidified Alumina
Concentrate
GC/EIMS-SIM
Identify PCDF/PCDD
Quantitate PCDF/PCDD
1/2 of Extract
GC/EIMS
Identify PCB's, Other
Chlorinated Organics
Quantitate PCB's, Estimate
Concentration of Other
Chlorinated Organics
Figure 3. Sample analysis scheme.
23
-------
Table 4. Phase 3 Surrogate Spiking Solution
Surrogate compound Spiking level, ng
(13C12)-2,3,7,8-TCDF 250
(13C12)-2,3,7,8-TCDD 250
(13C12)-OCDD 250
(I1 ,2' ,3' ,4' ,5' ,6'-13C6)-4-Chlorobiphenyl 104
(13C12)-3,3' ,4,4'-Tetrachlorobiphenyl 257
(13C12)-2,2' ,3,3' ,5,5' ,6,6'-Octachlorobiphenyl 407
(13C12)-Decachlorobiphenyl
2. Cleanup
One-half of each concentrated extract was cleaned by elution from a
multi-phase column, shown in Figure 4, using the following elution procedure:
a. Add sample extract to top of column.
b. Elute with 45 ml of hexane.
c. Remove top (silica) column.
d. Elute lower column with 20 mL of hexane.
e. Archive eluate.
f. Elute with 20 mL of 50% dichloromethane in hexane.
g. Collect eluant, concentrate just to dryness, under a gentle
stream of dry nitrogen. Redissolve residue in 0.5 ml isooctane.
h. Elute with 20 ml of MeCl2 and archive eluate.
This cleanup procedure was monitored by observing the recovery of
the two surrogate compounds (13C12)-2,3,7,8-TCDD and (13C12)-2,3,7,8-TCDF.
The recoveries observed for the extraction and cleanup procedures were between
50 and 70%.
24
-------
Reservair
• 4g of 40% (w/w) Sulfuric Acid/Silica Gel
>• Ig Activated Silica Gel '.Type 60, EM Reogent 100-200 Mesh)
1 g Anhydrous
• 6g Acid Alumina (AG4 Bio-Rad Labs)
Figure 4. Two-part cleanup column.
25
-------
3. GC/MS Analysis
During Phase 2, each cleaned extract was analyzed for trichloro-,
tetrachloro-, pentachloro-, and hexachlorodibenzofurans, using high resolution
gas chromatography and electron impact mass spectrometry detection with se-
lected ion monitoring (HRGC/MS-SIM). Analytical conditions were as follows:
Column: 30 m x 0.25 mm fused silica column, wall -coated with DB-5
Column temperature: 100°C (2 min hold) to 320°C at 10°C/min
Injector: Grob-type, 45 s splitless, 280°C
Electron energy: 70 eV
During Phase 3, each cleaned extract was analyzed for polychlorin-
ated dibenzodioxins (PCDDs), and polychlorinated dibenzofurans (PCDFs). The
second half of each extract, which had not been cleaned, was analyzed for PCBs
and some of these extracts were screened for other chlorinated organics.
Table 5 lists the samples collected during Phase 3, the analyses which were
performed on these samples, and the instrument which was used. Among the
types of analyses performed on the Phase 3 samples were full scan HRGC/MS,
HRGC/MS-SIM, and HRGC/MS- limited mass scan. Three different instruments were
used to perform these analyses. The operating parameters for the instruments,
Finnigan 4023, Finnigan MAT 311A, and Kratos MS-50, are listed in Table 6.
The specific ion masses and mass ranges monitored for each analyte are shown
in Table 7.
4. Analyte Quantisation
The PCDFs in the sample extracts from Phase 2 were quantitated by
comparing the responses for the sample extract to the response of 2,3,7,8-
tetrachlorodibenzofuran (2,3,7,8-TCDF) for an authentic standard solution.
These concentrations were calculated using the internal standard method. The
internal standard used was (13C12)"2,3,7,8-TCDF. First, a response factor
(RF) for the standard (2,3,7,8-TCDF) was calculated according to the equation:
(AS)(C )
-
" (Ais)(Cs)
where A = Area of the primary characteristic ion of the 2,3,7,8-TCDF
s (m/z 306)
C. = Concentration or amount of the internal standard
A. = Area of the primary characteristic ion of the internal standard
(13C12)-2,3,7,8-TCDF (m/z 318)
C = Concentration or amount of 2,3,7,8-TCDF in the calibration
standard
The response factor calculated for 2,3,7,8-TCDF was used to quanti
tate all chlorinated dibenzofurans identified in the effluent samples in
Phase 2. The quantitation was performed using the equation:
(AS)(C )
Concentration =
(Ais)(RF)
where A , A. , C. , and RF are described above.
26
-------
Table 5. Types of Analyses Used for Samples From Phase 3
ro
—i
^v. Type of
^\analysis
Sample noX.
8-07-39-M500
8-13-40-M500
8-14-41-M500
8-14-42-BLK
8-15-43-M5
8-15-44-M5
8-16-45-M50
8-16-46-M50
8-17-47-S5
8-20-48-S5
8-21-49^550
8-21-50-S50
8-22-51-S500
8-22-52-S500
8-22-53-BLK
8-23-54-CLBZ
8-23-55-CLBZ
8-23-56-BLK
8-28-57-CLBZ
8-28-58-CLBZ
8-28-59-BLK
8-29-60-ASKL
8-30-61-ASKL
8-30-62-ASKL
8-30-63-BLK
Full scaq
analysis
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Limited mass
scan analysis
for PCBsa
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Analysis of
dioxins, furans
Clt + C12 C13 + C14
Xb
* \
\
x3
xc
xc
xcr
xc
xcc
*b
\
xb
K XK
n n
XK x
D C
V y
C
h Xr
V ¥
f
b xc
xb xc
xb xc
Xh Xr
xb xc
chlorinated
and biphenyls
n + n
^ 1 5 * L '6
xc
X
X
x
Xh
x
xc
xk
x
xc
xc
xc
x^
p
xb
X
xc
xb
xb
u
xb
xb
x
C17 + C18
xb
xb
xb
x
b
xb
x
xb
X
b
x
b
xb
xb
x
.Analyzed using Finnigan 4023.
Analyzed using Finnigan MAT 311A.
Analyzed using Kratos MS-50.
-------
Table 6. Operating Parameters for Gas Chromatograph/Mass
Spectrometers Used to Analyze Phase 3 Samples
ro
CO
Instrument model
Mass spectrometer type
Mode of operation
Finnigan 4023
Quadropole
Full scan Limited mass scan
Finnigan MAT 311A
Magnetic sector;
Selected ion
monitoring
Kratos MS-50
Magnetic sector;
Selected ion
monitoring
Operating parameters
(GC)
Column
Column temperature
Injector
Injector temperature
(mass spectrometer)
Scan range (m/z)
30 m x 0.25 mm DB-5 30 m x 0.25 mm DB-5 30 m x 0.25 mm DB-5 15 m x 0.25 mm DB-5
30° (1 min hold) to
325° at 10°/min
Grob 45 s splitless
280°
95-550
80° (1 min hold) to
325° at 6°/min
Grob 45 s splitless
280°
Variable, relative
to analyte
100° (2 min hold)
to 320° at 10°/min
Grob 45 s splitless
280°
Variable, relative
to analyte
100° (2 min hold)
to 320° at 10°/min
Grob 45 s splitless
280°
Variable, relative
to analyte
Scan time
Resolution
Ion source temperature
Electron energy
Data system
1 s
Unit
270°
70 eV
Incos 2300
1 s
Unit
270°
70 eV
Incos 2300
0.9-1 s
800
250-270°C
70 eV
Incos 2400
0.9-1 s
1200
250-270°C
70 eV
Incos 2400
Scan ranges and selected ion listed for individual analytes in Table 7.
-------
Table 7. Scan Ranges and Selected Ions Monitored for Individual Analytes
1X3
Analyte
Homolog
Cli
C12
C13
C14
C15
C16
C17
C18
C19
Clio
Polychlorinated biphenyls
Chlorinated
biphenylenes
Primary
ion
Limited mass scan range (m/z) (m/z)
186-196
222-226
256-262
290-308
324-330
358-365
392-399
426-447
460-467
493-511
(±o.
(±0.
(±o.
(±o.
(±o.
(±o.
(±o.
(±0.
(±o.
(±o.
5)
5)
5)
5)
5)
5)
5)
5)
5)
5)
186
220
253
289
323
357
391
427
.04a
.00
.96
.92
.88
.84
.80
.76
-
-
Secondary
ion
(m/z)
190. 04a
221.99
255.96
287.92
321.88
359.84
393.80
427.76
-
-
Chlorinated
dibenzofurans
Primary
ion
(m/z)
202. 02a
235.98
269.14
305.90
339.86
373.82
407.78
443.74
-
-
Secondary
ion
(m/z)
204. 02a
237.98
271.94
303.90
337.86
375.82
409.78
441.74
-
-
Chlorinated
dibenzodioxins
Primary
ion
(m/z)
218
251
285
321
355
389
423
459
.Ola
.97
.94
.88
.85
.82
.78
.74
-
-
Secondary
ion
(m/z)
220. Ola
253.97
287.93
319.88
353.86
391.81
425.77
457.74
-
-
Scan range for selected ion monitoring was that of the reported m/z ±0.18.
-------
During Phase 3, a mixed PCDD/PCDF standard was used. This standard
contained a number of PCDD isomers and three PCDF isomers. The composition
of the PCDD/PCDF standard used during Phase 3 is shown in Table 8. The re-
sponse factors for this standard were calculated using the equation shown
above. As listed in Table 9, any mono- or diCDFs were quantified using the
response factor for 2,8-diCDF. Any tri-, tetra-, penta-, or hexaCDFs were
quantitated using the response factor for 2,3,7,8-tetraCDF. Any hepta- or
octaCDF was quantitated using the response factor for octaCDF. Any PCDDs were
quantitated using the corresponding response factors. Two internal standards
were used for these calculations. All PCDDs were quantitated using (13Ci2)~
2,3,7,8-TCDD as the internal standard, and all PCDFs were quantitated using
(13C12)-2,3,7,8-TCDF as the internal standard.
D. Statistical Analysis
1. Phase 2
Intermediate statistical analyses were performed as each set of data
became available. Each set of data consisted of a half-replicate of a 23 fac-
torial design. These were analyzed separately to estimate the main effects.
The data then were combined with previous data and the entire available data
set analyzed. The purpose of these interim statistical analyses was to indi-
cate the levels that should be tested at the next set of runs. Analysis of
variance was used to analyze the half-replicate data. These half-replicates
are saturated designs. That is, there is one parameter to be estimated for
each data point. Consequently, the analysis could only estimate the main ef-
fects and partition the total sum of squares into components for each main
effect. No estimate of error or interactions is available.
When the data were pooled with other data, more detailed analysis
became possible. This analysis varied with the amount of data and the struc-
ture. At one point the usual analysis of variance for a complete 23 factorial
design was used, but generally the data available were an unbalanced, incom-
plete design. The analysis for these data sets was that of a general linear
model (GLM). Computations were performed using the GLM program in the SAS
package. The results and the data structure that finally resulted are dis-
cussed in Section VII, where the analysis and conclusions are presented.
2. Phase 3
The data for the PCDFs (in nanograms) were statistically analyzed
using regression and ANOVA programs in the SAS package.
30
-------
Table 8. PCDD/PCDF Standard Used in Phase 3
Compound Concentration (ng/mL)
2-Chlorodibenzodioxin 4
2,7-Dichlorodibenzodioxin 4
2,8-Dichlorodibenzofuran 4
1,2,4-Tri chlorodi benzodi oxi n 4
2,3,7,8-Tetrachlorodibenzodi oxi n 20
2,3,7,8-Tetrachlorodibenzofuran 20
1,2,3,7,8-Pentachlorodi benzodi oxi n 20
1,2,3,4,7,8-Hexachlorodibenzodioxin 10
1,2,3,4,6,7,8-Heptachlorodi benzodi oxi n 10
Octachlorodibenzodioxin 200
Octachlorodibenzofuran 200
(13C12)2,3,7,8-Tetrachlorodibenzofuran 200
(13C12)2,3,7,8-Tetrachlorodibenzodioxin 200
(37Cl4)l,2,3,4,6,7,8-Heptachlorodibenzodioxin 2,000
(13C12)0ctachlorodi benzodi oxi n 2,000
31
-------
Table 9. Type of Quantitation Used During Phase 3 HRGC/EIMS Analysis
CO
ro
Chlorinated dibenzofurans
No. of
chlorines
1
2
3
4
5
6
7
8
Type of
quanti-
tation
ES
ES
IS
IS
ES
ES
IS
IS
Internal
std used
-
-
13C-2,3,
7,8-TCDF
13C-2,3,
7,8-TCDF
-
13C_
octaCDD
13C.
octaCDD
Quantitation
standard
2,8-diCDF
2,8-diCDF
2,3,7,8-
tetraCDF
2,3,7,8-
tetraCDF
1,2,3,7,8-
pentaCDD
1,2,3,4,7,8-
hexaCDD
OCDF
OCDF
Chlori
Type of
quanti-
tation
ES
ES
IS
IS
ES
ES
IS
IS
nated dibenzodioxins
Internal
standard
-
-
13C-2,3
7,8-TCDD
13C-2,3
7,8-TCDD
-
13C_
octaCDD
13C.
octaCDD
Quantitation
standard
2-monoCDD
2,7-diCDD
1,2,4-tri-
CDD
2,3,7,8-
tetraCDD
1,2,3,7,8-
pentaCDD
1,2,3,4,7,8-
hexaCDD
1,2,3,4,6,7,8-
heptaCDD
OCDD
ES - external standard; IS = internal standard.
-------
VII. RESULTS AND DISCUSSION
A. Phase 1
The thermal destruction system was operated at two temperatures and
two gas flow rates to determine the effects on effluent levels of 02, C02, CO,
and on combustion efficiency (CE). Test conditions included two different
flow rates and two different types of mineral oil, as noted in Table 10. Sys-
tem blank runs were performed at 600° and 700°C at high 02 levels (Runs 5-22-01
and 6-05-03, respectively), and at 700°C at low 02 levels (Run 6-06-04). The
rinse samples from the lower temperature run and the low 02 run were a light
yellow color, indicating a high level of organics due to incomplete combustion
of the mineral oil. Also, a soot-like material appeared at the entrance to
the XAD-2 trap during the low 02 run. In the 700°C, high 02 run, both the
rinse sample and the XAD-2 sample appeared clean.
The Phase 1 samples were extracted and analyzed for PCBs by full
scan HRGC/EIMS. No PCBs were detected. The extracts were analyzed by HRGC/
EIMS in the selected ion monitoring mode for the tri- through hexaCDFs. Ini-
tially, the samples contained too much background to detect low levels of
PCDFs; however, after column chromatographic cleanup, no PCDFs were detected.
Based on this information, all Phase 2 samples were cleaned prior to the
analysis.
It is apparent from the test results summarized in Table 10 that
the CE was affected by incineration temperature, generally increasing with
increasing temperature. The data suggested that near ideal stoichiometry re-
sults at 700°C or higher temperatures. When the gas flow rate was maintained
at a nominal I L/min, C02 increased with increasing temperature and CO re-
mained fairly constant. However, for tests conducted at a nominal flow rate
of 0.5 L/min, the C02 still increased with temperature, while CO maximized at
an intermediate temperature (600°C). CE was lowest at the 600°C test.
B. Phase 2
1. Test Conditions
Thirty-four test runs were completed during Phase 2. Two of these
were considered invalid for reasons noted below. A run was defined as the
test period in which an XAD-2 sample was collected. Each run was assigned a
unique number, consisting of three fields: first, the date of the run (in-
cluding the month, a dash, and the date); second, a two-digit sequential run
number, beginning with "01" for the first run; and last, a three-letter code
for the operating conditions. Low is designated by "L," medium is designated
by "M," high is designated by "H," and intermediate between medium and high
is designated by "M+." The first letter of this code represents combustion
temperature, the second letter represents effluent oxygen concentration, and
the final letter represents residence time. System blanks, indicated by the
suffix "(B)," were conducted under the same operating conditions as the pre-
ceding run, except without feed oil flow.
33
-------
Table 10. Phase 1 Non-PCB Combustion Test Conditions
CO
Run no.
NAa
NA
NA
NA
NA
NA
5-22-01
5-31-026
6-05-03
6-06-04
Combustion
temperature
450b
600b
700b
450b
700b
•<• 730
610d
728d
705d
Vaporization/ Effluent Residence "Waste" Effluent 02 Effluent C02 Effluent CO Combustion Comments
pyrolysis gas flow time feed rate concentration concentration concentration efficiency
temperature rate (s) (ul/min) (X) (X) (ppm) (X)
(°C) (L/min)
300b •<• 0.5 i. 2.3 ~ 5C 16.2 0.63 2,060 75
300b •<• 0.5 * 1.9 •>• 5C 16.3 0.81 5,050 62
300b •* 0.5 •«. 1.7 •<• 5C 16.0 1.24 1,250 90.9
300b -v 1.0 -v 1.2 -v- 10C 19.1 0.27 1,630 62
300b ~ 1.0 -v 0.9 •v. 10C 17.5 1.28 2,050 86
300b ~ 1.0 -v 0.9 -v. 10C 16.6 1.82 270 98.5
362d 1.51 0.7 •>• 10C 16.5 ~0.37 1,380 73 Rinse was light yellow.
416d 0.86 1.0 -v 10f 15.1 1.6 1,880 89 XAD and rinse appeared
clean and colorless.
Incineration tempera-
ture varied over •<• 50°C
range.
425d 1.02 0.9 •>• 10f 0.51 0.33 3,210 51 Soot-like material at
entrance to XAD. Both
XAD and rinse were
light yellow.
.Not assigned. Testing performed for combustion efficiency determination. XAD-2 sample not taken.
Nominal furnace (refractory) temperature.
."Waste" was pharmaceutical grade mineral oil.
eAverage of gas temperatures read at 10-min intervals.
^Operating conditions were unstable due to inadequate temperature control in vaporization/pyrolysis furnace.
"Waste" was technical grade mineral oil.
9Gas temperature variations apparently due to fluctuations in flow rate.
XAD-2 sample held but not analyzed.
-------
The actual operating conditions for each run are provided in Table
11. The runs are listed in a hierarchical order: first, by temperature;
second, by oxygen level; and third, by residence time. The average C02 and
CO continuous monitor readings taken during each run and combustion efficien-
cies calculated from those readings are also presented. Combustion efficiency
is calculated from the equation:
re = _ ECO?] _
_ _
[C02] + [CO]
where both C02 and CO are in the same concentration units.
The selection of operating conditions used for Phase 2 tests was
based upon an interactive combination of a statistical experimental design
and the results of the chemical analyses of effluent samples. That is, the
experimental design was updated when analytical results became available with
the objective of maximizing the formation of PCDFs. The evolution of the ex-
perimental design is discussed in Section V.C.2. The influence of the various
operating conditions upon PCDF formation is also discussed at length in Sec-
tion VII. Table 12 summarizes the nominal values, means, standard deviations,
and ranges of the actual values for the various operating conditions. In
general, actual combustion temperatures are within 15° of the nominal values.
[Effluent oxygen levels are more difficult to set to prespecified values since
the oxygen consumption by the waste itself is a variable.] Actual residence
time values are generally within 10-20% of the nominal values.
2. Continuous Monitor Results and Combustion Efficiency
As discussed in Section VLB above, the desired effluent oxygen con-
centration was obtained by varying the nitrogen: air ratio in the makeup gas
furnace, and thus was an independent variable. However, both carbon dioxide
and carbon monoxide concentrations were dependent variables, since they are
affected by feed oil flow rate, residence time and total gas flow rate, com-
bustion temperature, and possibly adjusted oxygen concentration. Aside from
system blanks, which had very low C02 and CO levels, the effluent C02 concen-
trations ranged from 0.08% (in both Run 6-14-06-LLL and Run 7-05-17-LML) to
3.23% (in Run 6-18-09-HHH) , and the effluent CO concentration ranged from
0.0015% or 15 ppm (in Run 6-18-09-HHH) to 0.587%, or 5,870 ppm (in Run
7-13-21-HMH).
35
-------
Table 11. Operating Conditions for Phase 2 Tests
Run no.
6-14-06-LLL
7-10-18-LLM
7-10-19-LLM(B)a
7-05-17-LML
6-14-07-LMM
6-22-14-LMM
7-03-16-MLL
6-13-05-MLM
6-15-08-MML
7-03-15-MMM
7-12-20-MMM
7-20-27-MMM
7-26-32-MMM.
6-20-12-MMH
6-20-13-MMH
7-26-31-MMH
7-31-37-MMH
7-31-38-MMH(B)a-
7-24-28-MMHC i
7-25-29-MMHC i
7-19-25-MM+M
6-19-11-MHM
7-16-23-MHH
7-30-35-MHH
7-30-36-MHH
7-17-24-M+MM
7-25-30-M+MM+
7-27-33-M+MH
7-27-34-M+MH(B)a
7-19-26-M+M+M
6-19-10-HMM
7-13-21-HMH
7-16-22-HHM
6-18-09-HHH
Temp.
(°C)
459
464
464
454
458
452
597
605
581
629
605
606
612
614
615
615
611
607
nvalid
nvalid
608
604
616
610
611
680
689
677
677
683
759
754
771
750
Oxygen
(%)
0.79
0.48
0.86
3.66
3.93
3.67
1.18
0.61
3.49
3.65
3.33
3.93
3.07
3.72
3.64
3.45
3.45
3.79
invalid
invalid
8.15
18.45
10.49
12.61
12.65
2.67
3.22
3.04
3.52
7.71
3.27
3.32
11.55
12.93
Res. time
(s)
0.31
0.78
0.79
0.29
0.74
0.76
0.30
0.81
0.29
0.74
0.80
0.80
1.21
1.62
1.65
1.55
1.66
1.59
invalid
invalid
0.86
0.79
1.68
1.49
1.52
0.79
1.22
1.60
1.56
0.80
0.82
1.67
0.82
1.52
C02
(%)
0.08
0.10
0.05
0.08
0.13
0.11
0.10
0.30
0.15
0.26
0.29
0.29
1.07
1.46
0.53
1.79
1.72
0.07
invalid
invalid
0.37
0.36
1.50
1.56
1.62
0.35
1.49
2.22
0.05
0.37
1.39
1.07
1.55
3.23
CO
(%)
0.015
0.020
0.0005
0.011
0.030
0.031
0.033
0.098
0.041
0.111
0.078
0.111
0.233
0.531
0.149
0.347
0.336
0.0017
invalid
invalid
0.1394
0.161
0.572
0.536
0.502
0.1670
0.362
0.393
0.0000
0.298
0.133
0.588
0.0077
0.0015
CE
(%)
85
84
98.9
88
81
78
75
75
79
70
79
72
82
73
78
84
84
98
invalid
invalid
73
69
72
74
76
67
80
85
100
55
91
65
99.5
99.95
.No waste flow (system blank).
Only XAD-2 sample analyzed, not rinse.
Operational problems with the reaction system prevented completion of this
test run.
36
-------
Table 12. Nominal and Actual Values for Operating Conditions During Phase 2 Tests
Operation condition
Test codes for various levels of operating conditions
L M M+ H
GO
Combustion temperature (°C)
Nominal
Mean ± std. dev.
Range (no. tests)
Effluent oxygen concentration (%)
Nominal
Mean ± std. dev.
Range (no. tests)
Residence time (sec)
Nominal
Mean ± std. dev.
Range (no. tests)
450
458 ± 5
451 - 464(6)
< 1
0.78 ± 0.27
0.48 - 1.18(5)
0.3
0.30 ± 0.01
0.29 - 0.31(4)
600
609 ± 10
581 - 629(17)
3-4
3.56 ± 0.25
3.07 - 3.93(15)
0.8
0.79 ± 0.03
0.74 - 0.86(14)
675
681 ± 5
677 - 689(5)
8
7.93 ± 0.31
7.71 - 8.15(2)
1.15
1.22 ± 0.01
1.21 - 1.22(2)
750
758 ± 9
750 - 771(4)
> 12
13.11 ± 2.77
10.49 ± 18.45(6)
1.5
1.59 ± 0.06
1.52 - 1.68(12)
-------
The observed values for 02, C02, and CO are averages of readings
taken during each test. Strip chart recordings of effluent levels of 02, C02,
and CO are shown in Figures 5 and 6 and for two representative tests (Run
6-19-11-MHM and Run 6-20-13-MMH, respectively). In each figure, background
levels are shown prior to turning on the waste flow, followed by a stabiliza-
tion period prior to beginning a test. The start and end times for sampling
are also shown. Some perturbation of concentration of the monitored gases
occurs following switchover of sampling systems. This causes the spikes at
the beginning and end of each run. Under some test conditions, wide varia-
tions (swings) occur in some or all of the monitored gases, as shown in Fig-
ure 6. Other test conditions produce relatively stable patterns, as shown in
Figure 5. There is no clear explanation for these differences in apparent
stability. The average of readings taken at 1-min intervals was reported to
dampen out the fluctuations. Absolute values for C02 and CO should not be
directly compared between runs, since the gaseous flow rates varied while the
feed oil flow rate remained constant (13.5 uL/min). That is, the source of
the CO and C02 remained constant, but different dilution factors resulted from
varying the total gas flow rate to establish the desired residence time. The
calculated combustion efficiencies compensate for variations in flow rate and
can be compared between runs.
In Phase 2 tests, the CE ranged from 55 to 99.95%, i.e., less than
one "nine" to greater than three "nines." It is not certain whether CE is
related to PCDF formation, but it may be a useful parameter in relating the
conditions in the combustion unit to those occurring in a transformer fire.
Figure 7 shows a plot of CE as a function of combustion temperature for the
Phase 2 tests, excluding system blanks.
The target sampling period for each run was 60 min. Since the
spiked mineral oil was fed at a constant rate of 0.0135 mL/min, each run had
a nominal feed oil volume of 0.81 ml. For the Phase 2 tests, a solution of
three specified PCB congeners in Exxon mineral oil comprised the feed oil
solution, at the following nominal concentrations: 0.5 mg/mL 2,3,5,6-tetra-
chlorobiphenyl, 0.5 mg/mL 3,3',4,4',5,5'-hexachlorobiphenyl, and 1.0 mg/mL
2,2',4,4',6,6'-hexachlorobiphenyl, or a total of 2.0 mg/mL PCBs. The amounts
of each PCB congener and the total input during each run are given in
Table 13.
3. PCBs Input During Tests
The actual sampling periods for the Phase 2 tests ranged from
34 min (during Run 6-14-07-LMM) to 65 min (during Run 6-20-12-MMH). The short
run (6-14-07-LMM) was terminated early because the syringe pump was nearly
emptied of feed oil. It was repeated as Run 6-22-14-LMM, with fairly consis-
tent results between runs. The total input feed oil volume varied from
0.46 mL to 0.88 mL, depending on the length of the sampling period. In addi-
tion, the concentrations of the three PCB congeners also varied slightly from
one batch of feed oil to the next (as shown in Table 2).
38
-------
02
8
7
6
5
4
3
2
1
0
2.5
2.0
1.5
CO
(ppm)
0.5
0
6000
5000
4000
3000
2000
looo;
1530 1540 1550 1600 1610 1620 1630 1640
Time |
Waste Sample Sample
On Start End
1 t
Figure 5. Continuous gas monitoring results for Run 6-20-13-MMH.
39
-------
02
20
19
18
17
i r
\ I I I r
J I I I I I I
1.0
0.8
C02 °'6
(%) 0.4
0.2
0
I I
CO
(ppm)
I800p—i r
1600 -
1400 -
1200 -
1000 -
800 -
600
j i
1320 1330 1340 1350 1400 1410 1420 1430 1440
| t Time 4
Waste Sample Sample
On Start End
Figure 6. Continuous gas monitoring results for Run 6-19-11-MHM.
40
-------
1 \-/\J> —
9O -
8
il 8O ~
LU
O
§ 70 -
DQ
O
O
60 -
o\_/
D u a
D
a
a
a a
a CD
n a ,
D
a p Dn
sfi
a a
a
a
a
i i i i i i i
4OO
5OO 6OO 7OO
COMBUSTION TEMP (DEC C)
Figure 7. Combustion efficiency versus temperature.
8OO
-------
Table 13. Weights of PCBs Used During Phase 2 Tests
Chlorobiphenyl congener
Run no.
6-14-06-LLL
7-10-18-LLM
7-10-19-LLM(B) ^
7-05-17-LML
6-14-07- LMM
6-22-14-LMM
7-03-16-MLL
6-13-05-MLM
6-15-08-MML
7-03-15-MMM
7-12-20-MMM
7-20-27-MMM
7-26-32-MMM+
6-20-12-MMH
6-20-13-MMH
7-26-31-MMH
7-31-37-MMH
7-31-38-MMH(B)
7-19-25-MM+M
6-19-11-MHM
7-16-23-MHH
7-30-35-MHH
7-30-36-MHH
7-17-24-M+MM
7-25-30-M+MM+
7-27-33-M+MH
7-27-34-M+MH(B)
7-19-26-M+M+M
6-19-10-HMM
7-13-21-HMH
7-16-22-HHM
6-18-09-HHH
Feed oil
vol. (ml)
0.78
0.81
0.82
0.81
0.46
0.81
0.76
0.81
0.81
0.81
0.81
0.81
0.81
0.88
0.81
0.81
0.81
0.80
0.84
0.81
0.82
0.81
0.81
0.84
0.81
0.81
0.81
0.81
0.68
0.86
0.84
0.81
2,3,5,6-
Tetra
(mg)
0.42
0.41
0.41
0.41
0.25
0.41
0.38
0.44
0.41
0.41
0.41
0.41
0.41
0.44
0.41
0.41
0.41
0.40
0.42
0.41
0.41
0.41
0.41
0.42
0.41
0.41
0.41
0.41
0.34
0.43
0.42
0.41
2, 2', 4,
4', 6, 6'-
Hexa
(mg)
0.84
0.83
0.84
0.83
0.49
0.81
0.77
0.87
0.81
0.83
0.83
0.83
0.81
0.88
0.81
0.81
0.81
0.80
0.85
0.81
0.84
0.81
0.81
0.85
0.81
0.81
0.81
0.83
0.68
0.88
0.85
0.81
3, 3', 4,
4', 5, 5'-
Hexa
(mg)
0.29
0.38
0.38
0.38
0.17
0.39
0.35
0.30
0.39
0.38
0.38
0.38
0.41
0.42
0.39
0.41
0.41
0.40
0.39
0.39
0.38
0.41
0.41
0.39
0.41
0.41
0.41
0.38
0.32
0.40
0.39
0.39
Total
PCB
(mg)
1.55
1.61
1.64
1.61
0.91
1.61
1.50
1.61
1.61
1.61
1.61
1.61
1.62
1.75
1.61
1.62
1.62
1.59
1.66
1.61
1.64
1.62
1.62
1.66
1.62
1.62
1.62
1.61
1.34
1.72
1.66
1.61
42
-------
4. PCDF Results
The results of analysis of the Phase 2 samples for trichlorodibehzo-
furan (triCDF), tetrachlorodibenzofuran (tetraCDF), pentachlorodibenzofuran
(pentaCDF), and hexachlorodibenzofuran (hexaCDF) are shown in Table 14. When
the PCDF homolog was not detected, "< 5 ng" is noted. The total PCDFs result-
ing from the four homologs of interest are given in the final column. In the
final column, the summation of four "< 5 ng" values yields a total PCDF value
of "< 20" where no PCDFs were detected. As footnoted in the table, only the
XAD-2 sample from Run 6-20-12-MMH was analyzed, as the rinse portion of the
total sample was lost and not analyzed. Run 6-20-13-MMH was a replicate of
that run, and the XAD-2 extract and rinse were analyzed individually to deter-
mine the relative abundance of the PCDFs in the two fractions. The amounts
were summed for the reported value. Only a minor amount of triCDF was de-
tected in the rinse sample, with none of the tetra-, penta-, or hexachloro-
dibenzofurans detected. Therefore, the PCDF values for Run 6-20-12-MMH were
judged valid.
The conversion efficiencies of the specific PCB congeners to the
anticipated PCDF, i.e., percent yield, are shown in Table 15. According to
results previously reported in the literature, it was presumed during this
study that triCDF is formed from 2,3,5,6-tetrachlorobiphenyl, that both tetra-
and pentaCDFs are formed from 2,2',4,4',6,6'-hexachlorobiphenyl and that
hexaCDF is formed from 3,3',4,4',5,5'-hexachlorobiphenyl, as noted previously
in Section IV. A total conversion efficiency for each run is also given in
Table 15. In this case, the total nanograms of the four PCDF homologs are
divided by the total of the three PCB congeners fed in the feed oil. This is
not an average of the four individual conversion efficiencies.
The primary PCDFs formed from the three individual PCB congeners
were triCDF and tetraCDF. Maximum values detected were 8,870 ng and 3,520 ng,
respectively. In contrast, relatively low levels of the pentaCDF and hexaCDF
were found. Maximum values detected were 244 ng and 125 ng, respectively.
Apparently, the reactions for producing these PCDFs are not as efficient as
those for producing tri- and tetrachlorodibenzofuran. The maximum conversion
efficiency observed for producing triCDF was 2.2%; for producing tetraCDF it
was 0.43%; for producing pentaCDF it was 0.03%; and for producing hexaCDF it
was 0.03%. The maximum total conversion efficiency of PCBs to PCDFs during
Phase 2 tests was 0.78% (Run 7-19-26-M+M+M). Conversion efficiency is used
to evaluate the PCDF formation, rather than the absolute amount of PCDFs
formed, because the conversion efficiency normalizes the results with respect
to the different amounts of PCBs fed during the runs.
In most Phase 2 samples a single isomer triCDF and a single tetraCDF
were observed. In those samples which had the highest concentration of these
isomers (7-27-33, 7-30-35, 7-30-36, 7-19-26), six to ten additional isomers
were also observed at lower concentrations for each homolog. These isomers
were not included in the results listed in Tables 14 and 15. The presence of
these other isomers suggests that reactions in addition to those described in
Section IV.B are occurring during the combustion process. These reactions
may involve dechlorination of penta- or hexaCDFs or may involve rearrangements
of the tri- or tetraCDFs.
43
-------
Table 14. Weights of PCDFs in Combined XAD-2/Rinse Samples
from Phase 2 Tests
Run no.
6-14-06-LLL
7-10-18- LLM
7-10-19-LLM(B)a
7-05-17-LML
6-14-07-LMM
6-22-14-LMM
7-03-16-MLL
6-13-05-MLM
6-15-08-MML
7-03-15-MMM
7-12-20-MMM
7-20-27-MMM
7-26-32-MMM:!;
6-20-12-MMH0
6-20-13-MMH
7-26-31-MMH
7-31-37-MMH
7-31-38-MMH(B)a
7-19-25-MM+M
6-19-11-MHM
7-16-23-MHH
7-30-35-MHH
7-30-36-MHH
7-17-24-M+MM
7-25-30-M+MM+
7-27-33-M+MH
7-27-34-M+MH(B)a
7-19-26-M+M+M
6-19-10-HMM
7-13-21-HMH
7-16-22-HHM
6-18-09-HHH
TriCDF
(ng)
< 5
< 5
< 5
< 5
< 5
< 5
< 5
< 5
690
< 5
< 5
1,800
605
990
1,316
599
508
< 5
2,280
905
1,160
1,788
2,111
1,190
2,066
1,992
< 5
8,870
178
92
35
192
TetraCDF
(ng)
303
< 5
< 5
. — -*
946
95
256
324
169
240
985
174
740
446,
517
966
263
498
< 5
800
629
240
1,159
1,241
337
942
1,186
< 5
— — — — '
3,520
224
59
12
23
PentaCDF
(ng)
5
< 5
< R
< 5
21
< 5
< 5
15
31
< 5
< 5
6
125
49
100
54
170
< 5
19
19
45
213
208
38
194
244
' 5
96
19
5
< 5
0.4
HexaCDF
(ng)
< 5
< 5
< 5
< 5
< 5
< 5
< 5
< 5
2
< 5
< 5
< 5
21
4
< 5
7
36
< 5
< 5
< 5
< 5
125
115
trace
93
81
< 5
Jk-2
< 5
< 5
< 5
< 5
< 5
PCDFs
(ng)
308
< 20
< 20
=.
946
116
256
324
184
963
985
174
2,546
1,197
1,560
2,382
923
1,212
< 20
i
3,099
1,553
1,445
3,285
3,675
1,565
3,295
3,503
< 20
__
— . — ,
12,486
421
156
47
215
.System blank; no feed oil flow.
Only XAD-2 sample analyzed; rinse sample was not analyzed.
44
-------
Table 15. Conversion Efficiencies (PCBs to PCDFs) for Phase 2 Tests
Conversion efficiency (%)a
Run no.
6-14-06-LLL
7-10-18- LLM .
7-10-19-LLM(B)a
.-T-flSrrFLMT
6-14-07-LMM
6-22-14-LMM
7-03-16-MLL
6-13-05-MLM
6-15-08-MML
7-03-15-MMM
7-12-20-MMM
7-20-27-MMM
7-26-32-MMM+
6-20-12-MMH6
6-20-13-MMH
7-26-31-MMH
7-31-37-MMH .
n
7-31-38-MMH(B)a
1 7 — TQ— -9-f^.MM-t-M
6-19-11-MHM
7-16-23-MHH
7-30-35-MHH
7-30-36-MHH
7-17-24-M+MM
7-25-30-M+MM+
7-27-33-M+MH
7-27-34-M+MH(B)
7-19-26-M+M+M
6-19-10-HMM
7-13-21-HMH
7-16-22-HHM
6-18-09-HHH
TriCDF
0C
0
0
. — •
0
0
0
0
0
0.17
0
0
0.44
0.15
0.22
0.32
0.15
0.13
0
_—ft— ^A— —
0.22
0.28
0.44
0.52
0.28
0.51
0.50
0
2.2
0.052
0.021
0.0084
0.047
TetraCDF
0.036
0
0
0.11
0.020
0.032
0.042
0.019
0.03
0.12
0.02
0.090
0.055
0.059
0.12
0.033
0.061
0
«_ nnfl ""
0.077
0.029
0.14
0.15
0.040
0.12
0.15
n
0.43
0.033
0.0067
0.0014
0.0028
PentaCDF
0.0006
0
0
• — .
0
0.0042
0
0
0.0017
0.0038
0
0
0.0007
0.015
0.0056
0.012
0.0067
0.0210
0
— •
n no??
U . UU££
0.0023
0.0054
0.026
0.026
0.0045
0.024
0.030
Q_
0.012
0.0028
0.0006
0
0
HexaCDF
0
0
0
0
0
0
0
0
0.0005
0
0
0
0.052
0.0009
0
0.0017
0.0089
0
. .
n
U
0
0
0.0309
0.0284
0
0.0230
0.0200
0
0
0
0
0
PCDFsb
0.020
0
0
0"7059
0.013
0.016
0.022
0.012
0.060
0.061
0.011
0.16
0.074
0.089
0.15
0.057
0.075
0
~~~"~~— — — "^
n 'i Q
U • -L.7
0.096
0.088
0.20
0.23
0.094
0.20
0.22
0
— • — --•
0.78
0.031
0.0091
0.0028
0.013
.Conversion efficiency = ng PCDF formed/ng PCB fed x 100%.
The total nanograms of the four PCDF homologs are divided by the total of the
three PCB congeners fed in the feed oil. This is not an average of the four
individual conversion efficiencies.
All "not detected" values from Table 10 are expressed as "0" is this table
.for statistical calculation purposes.
System blank; no feed oil flow.
Only XAD-2 sample analyzed, not rinse.
-------
The total conversion efficiencies for PCDF and tetraCDF formation
are plotted versus the effluent oxygen concentration in Figures 8 and 9. It
can be seen that the highest conversion efficiencies occurred at the medium
and medium-high temperatures. Extreme variability occurred in the medium
(3-4%) oxygen range, but conversion efficiencies were more consistently high
for oxygen levels of ^ 8%. The Y-scale is shown broken in both figures, since
the highest values for conversion efficiency, from one test, are significantly
higher than all other values. It is possible that this test was an anomaly.
Differences in residence time are ignored in these figures, since statistical
analysis indicated residence time to be the least significant operating con-
dition.
5. Statistical Evaluation
Statistical analysis of the measured efficiencies of conversion from
PCBs to PCDFs relative to temperature, oxygen, and residence time conditions
was employed to identify optimum conversion conditions for Phase 3 runs. The
total PCB to PCDF conversion efficiencies were given highest priority. How-
ever, conversion efficiencies to tetraCDFs were also evaluated in view of
their typically higher toxicity and general environmental concern.
Data from 29 runs, excluding blanks and invalid runs, were obtained
under various combinations of the levels of temperature, residence time, and
oxygen concentration, as shown in Table 15. The table also includes a number
of blank runs that were done for quality control to ensure that no contamina-
tion was present from one run to a subsequent run. All of the blank determi-
nations gave zero on all components of the PCDFs and were not included in the
statistical analysis.
The data were analyzed using the general linear model approach to
the analysis of variance from an unbalanced and incomplete design. The total
conversion efficiency was used as the dependent variable for analysis. Table
16 gives the analysis of variance for the full model that includes main effects
and all two-way and three-way interactions. As can be seen from the low values
in the "P" column, the main effects of temperature (T) and oxygen (0) are sta-
tistically significant (P < .05), as is the two-way interaction of these two
variables (T*0). The interpretation of these results is that the mean conver-
sion efficiency differed among the levels of temperature and oxygen, but not
by residence time. The interaction was that mean conversion efficiency at
the optimum level for temperature and oxygen was higher than would be expected
from the addition of the temperature and oxygen effects. This is indicative
of a synergistic relationship between temperature and oxygen. None of the
other effects was significant.
46
-------
u
c
0
o
U
0.78
0.77
0.24
0.22
0.20
0.18
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
O
00
I I I I
I
J I
o
I A I
A
I
LEGEND
Nominal Combustion Temperature:
• 450°C
O 600°C
* 675°C
A 750°C
I I I I
J I
8 10 12
Effluent Oxygen Concentration (%)
14
16
18
20
Figure 8. Total PCDFs formed as a function of oxygen.
47
-------
X
g
.2
'o
£
o
0.43
0.42
*
O.lo'
0.15
0.14
0.13
0.12
0.11
0.10
0.09
0.08
0.07
0.06
0.05
0.04
0.03
0.02
0.01
0
- O
O
I
O
o
LEGEND
Nominal Combustion Temperature:
• 450°C
O 600"C
4 675°C
A 750°C
J I
J I
8 10 12
Effluent Oxygen Concentration (%)
14
16
18
20
Figure 9. Tetra CDFs formed as a function of oxygen concentration.
48
-------
Table 16. Full Model Analysis of Variance
Dependent variable: Total
Source
Model
Error
Corrected total
Source
Th
(T
RC
T*R
T*0
0*R
T*0*R
DF
20
8
28
DF
3
3
3
4
3
2
2
Sum of squares
0.5852
0.0268
0.6120
Type III SS
0.1617
0.1652
0.0084
0.0084
0.1000
0.0013
0.0005
Mean square F value
0.0293 8.72
0.0034
F value
16.06
16.42
0.84
0.62
9.94
0.20
0.08
P R-square
0.0019 0.9561
Root MSE
0.0579
P
0.0010
0.0009
0.5110
0.6594
0.0045
0.8254
0.9282
C.V.
55.6118
Total mean
0.1042
.Temperature.
C0xygen.
Residence time.
-------
The analysis of variance model was refit using only the main effects
and the oxygen by temperature interaction. The resulting analysis of variance
is shown in Table 17. Again, the main effects of temperature and oxygen and
their interaction are significant. The analysis of variance was also run on
each PCDF separately. The tri and tetra components gave the same results as
the total PCDF. The pattern for the pentaCDF was slightly different. For
pentaCDF, the main effects of temperature and residence time were significant
(P = 0.04 and P = 0.02, respectively), while the effect of oxygen was nonsig-
nificant. None of the effects was significant when hexaCDF was used as the
dependent variable in the analysis, probably due to the high number of "not
detected" values.
The mean total PCDF conversion efficiencies, grouped by variable,
are presented in Table 18. These are the mean values of all of the conver-
sion efficiencies in Table 15 which have the same level of a given variable.
For example, five runs at 450°C (T = L) have a mean conversion efficiency of
0.0214%. Table 18 also shows that conversion was highest at a temperature of
675°C (M+). Likewise, conversion was highest at 8% oxygen (M+). Conversion
was about the same for all levels of residence time. Although the mean was
lower at the lowest residence time, this effect did not reach significance.
The final part of the table gives the mean by each combination of temperature
and oxygen, together with the number of observations in that combination.
Because of the particular toxicological importance of the tetra com-
ponent, the analysis of variance and associated table of means are presented
in Tables 19 and 20. The conclusions are the same as for the variable total
percent conversion.
The results of the Phase 2 statistical analysis indicate that,
within the ranges studied, the variables of temperature, oxygen, and residence
time appear to have maximum conversion near a temperature of 675°C, oxygen
concentration of 8%, and for a residence of about 0.8 s. Conversion is sig-
nificantly higher near the middle of the ranges for temperature and oxygen,
but does not vary significantly by residence time. Nevertheless, the sugges-
tive lower value at the short residence time of 0.3 s indicates that such a
short residence time should be avoided in the next phase. Consequently, the
recommended conditions for the next phase would be temperature 675°C, oxygen
8%, and residence time between 0.8 and 1.6 s. A convenient residence time
could be chosen within that range as it does not appear to affect conversion.
C. Phase 3
1. Test Conditions
A total of 24 tests were performed in Phase 3, employing four types
of dielectric fluids and spanning a range of PCB concentrations from 0 to 70%.
All runs used the same nominal operating conditions obtained from statistical
evaluation of Phase 2 results. These nominal conditions are a 675°C combus-
tion temperature, an 8% effluent oxygen concentration, and a 0.8-s. residence
time in the combustion zone. The gas temperature in the pyrolysis furnace
was about 395°C in all tests except the askarel tests, in which a 461°C tem-
perature was used, for reasons discussed below.
50
-------
Table 17. Reduced Analysis of Variance Model Using Only Temperature and Oxygen
Dependent
Source
Model
Error
Corrected
Source
Th
Rb
S<
T*0
variable: Total
DF
12
16
total 28
DF
3
3
3
3
Sum of squares Mean square
0.5754 0.0480
0.0366 0.0023
0.6120
Type II SS
0.1175
0.0066
0.1939
0.1088
F value P R-square
20.96 0.0001 0.9402
Root MSB
0.0478
F value P
17.13 0.0001
0.96 0.4377
28.26 0.0001
15.85 0.0001
C.V.
45.92
Total mean
0.1042
.Temperature.
C0xygen.
Residence time.
-------
Table 18. Means for Total PCDF Conversion Efficiency (%)
Grouped by Variable
Variable
T
L
M
M+
H
0
L
M
M+
H
R
L
M+
M
H
T
L
L
M
M
M
M
<
M
H
H
0
L
M
L
M
M+
H
M
M
M
H
Sample size (N)
5
16
4
4
4
17
2
6
4
13
2
10
2
3
2
9
1
4
3
1
2
2
Mean
0.0214
0.0979
0.3225
0.0142
0.0132
0.0808
0.4813
0.1051
0.0400
0.1121
0.1387
0.1126
0.0099
0.0291
0.0166
0.0814
0.1864
0.1536
0.1712
0.7762
0.0202
0.0081
52
-------
Table 19. Analysis of Variance for TetraCDF
en
CO
Dependent
Source
Model
Error .
Corrected
Source
T
R
0
T*0
variable:
DF
12
16
total 28
DF
3
3
3
3
Tetra
Sum of squares Mean square
0.1600 0.0133
0.0311 0.0019
0.1911
Type II SS
0.0356
0.0027
0.0494
0.0382
F value
6.85
F value
6.10
0.46
8.46
6.54
P R-square
0.0003 0.8371
Root MSE
0.0441
P
0.0057
0.7108
0.0013
0.0043
C.V.
60.42
Tetra mean
0.0730
-------
Table 20. Means for TetraCDF Conversion Efficiency (%)
Grouped by Variable
Variable Sample size (N) Mean
T
L
M
M+
H
0
L
M
M+
H
R
L
M+
M
H
I P_
L L
L M
M L
M M.
M M
M. H
MI M+
M M
H M
H H
5
16
4
4
4
17
2
6
4
13
2
10
2
3
2
9
1
4
3
1
2
2
0.0402
0.0714
0.1820
0.0110
0.0243
0.0643
0.2598
0.0677
0.0555
0.0747
0.0857
0.0752
0.0180
0.0551
0.0307
0.0651
0.0937
0.1006
0.1007
0.4260
0.0199
0.0021
54
-------
The list of runs, operating conditions, gas concentrations and cal-
culated combustion efficiencies is provided in Table 21. The run numbering
system is similar to that used for Phase 2. The first three segments are the
month, date, and sequential run number, all separated by dashes. In the suf-
fixes "M" represents mineral oil, "S" represents silicone oil, "CLBZ" repre-
sents chlorobenzene, and "ASKL" represents askarel. The 5, 50, and 500 repre-
sent total PCB concentrations in ppm (w/w, i.e., ug/g). Systemjalank runs
are noted by a "(B)."
The combustion efficiencies (CEs) are generally low, as would be
expected for the combustion temperature selected for maximum PCDF generation.
However, CE for blank runs should be near 100%, as CO levels should be near
zero. This was not the case in Runs 8-22-53-S500(B) and 8-23-56-CLBZ(B) as
unexpectedly high CO levels were observed. This is presumably due to slow
decomposition of carbonaceous material remaining in the pyrolysis furnace, as
discussed below.
For the tests involving mineral oil and silicone oil, three concen-
trations of PCBs (Aroclor 1254) were used. The chlorobenzene dielectric fluid
was analyzed by GC/ECD and contained mostly trichlo'robenzene isomers, some
tetrachlorobenzene, and no detectable PCBs, PCDFs or PCDDs. The askarel-type
dielectric fluid contained a high level of PCBs, 70% (w/v) Aroclor 1260 by
GC/ECD analysis, and no PCDFs or PCDDs. For each fluid except askarel, the
density was determined to allow conversion of the concentrations units from
weight/weight to weight/volume. This step was necessary because the syringe
pump used for injecting the feed oil solution operates at a constant volume
rate. The sample period for each test and resulting dielectric fluid volume
pumped are shown in Table 22. Also shown are the fluid densities and PCB con-
centrations (in both ug/g and ug/mL). Finally, the total milligrams of PCBs
subjected to thermal degradation are indicated for each test.
2. Operational Problems
No serious difficulties were encountered during the Phase 3 mineral
oil tests. As noted in Table 21, a longer residence time was used (1.23 s)
for the first run (8-07-39-M500). This was a preliminary test performed be-
fore final test conditions had been identified.
Two major problems developed during the silicone oil tests. The
first problem became readily apparent when the first run (8-17-47-S5) com-
menced. Large quantitites of a fine, white silicate powder were generated.
Although the bulk of this material was deposited on the condensation tube
walls and on the glass wool plug of the XAD-2 trap, significant quantities of
the particulate passed through the resin trap and into the tubing leading to
the continuous gas monitors. This particulate condensed on the inlet filter
in the C02 monitor and stopped the gas flow to both the C02 and the CO moni-
tors. C02 and CO data were not obtained from this run. In addition, the
deposition of large quantities of particulate clogged the glass wool at the
front of the XAD-2 trap, which presented flow control problems. Although the
placement of in-line filters at different locations helped to control this
problem, frequent flow rate corrections and shutdowns were necessary during
the silicone oil tests.
55
-------
Table 21. Operating Conditions for Phase 3 Tests
en
CTi
Run no. Combustion temperature Oxygen
(°C) (%)
8-15-43-M5
8-15-44-H5
8-17-47-S5
8-20-48-S5
8-16-45-H50
8-16-46-H50
8-21-49-S50
8-21-50-S50
8-07-39-M500
8-13-40-H500
8-14-41-M500
8-14-42-H500(B)
8-22-51-S500
8-22-52-S500
8-22-53-S500fB)
8-29-60-ASKL0
8-30-61-ASKL
8-30-62-ASKL
8-30-63-ASKL(B)
8-23-54-CLBZ
8-23-55-CLBZ
8-23-56-CLBZ(B)
8-28-57-CLBZ
8-28-58-CLBZ
8-28-59-CLBZ(B)
679
679
677
675
679
679
678
679
685
678
678
675
680
680
679
680
680
679
679
680
680
679
679
679
680
8.0
8.1
8.5
7.6
8.0
7.8
8.2
7.8
8.3
7.7
7.9
8.2
8.2
8.3
8.2
ND
8.8
8.5
7.9
8.3
8.3
8.2
8.1
7.7
8.2
Residence time
(s)
0.84
0.83
0.85
0.78
0.83
0.81
0.83
0.82
1.23
0.81
0.82
0.81
0.79
0.82
0.82
ND
0.87
0.81
0.75
0.81
0.81
0.82
0.82
0.83
0.80
C02
(%)
0.56
0.58
ND
0.06
0.55
0.54
0.08
0.06
1.53
0.53
0.55
0.06
0.05
0.06
0.05 ~~
NO
0.07
0.06
0.05
0.07
0.07
0.06
0.07
0.07
0.07
CO Combustion efficiency
(%) (*)
0.28
0.29
ND
0.13
0.29
0.28
0.32
0.32
0.55
0.24
0.24
0.001
0.35
0.36
0.14
ND
0.23
0.18
0.001
0.41
0.33
0.12
0.05
0.10
0.002
67
67
ND
32
65
66
20
16
73
69
70
99
12
14
26
ND
23
25
99
15
17
33
56
41
98
Pyrolysis temperature
(°C)
396
395
395
394
395
395
395
394
396
395
395
395
394
394
395
461
462
461
461
395
395
395
394
395
395
*ND = no data.
Conditions unstable; continuous monitoring data was unreliable.
-------
Table 22. PCB Feed Characteristics in Phase 3
Run no.
8-15-43-M5
8-15-44-M5
8-17-47-S5
8-20-48-S5
8-16-45-M50
8-16-46-M50
8-21-49-S50
8-21-50-S50
8-07-39-M500
8-13-40-M500
8-14-41-M500
8 ~ 1 4 ~ 4 2 ~ M5J3£LCB-X—_
"^B^ -51^500
8-22-52-S500
8~22~E)3~St}fln(R'\
^8-29-60-ASKb-
8-30-61-ASKL
8-30-62-ASKL
_8j:30-63-ASKL(B)
8-23-54-CLBZ
8-23-55-CLBZ
8- 23- 56-CLBZI81—
\J C*. *J *J\J *^ I \ PI \**J ~~
rt""rtQ^QT^^| Q7
O kO 3/ L. L n /
8-28-58-CLBZ
O™"tO""Oy~*wl— LJ^\ D 7
— — -—• ~ "*
Sample
period
(min)
67
73
58
67
61
63
59
53
60
61
61
60
56
61
-^-—58——
120
35
35
___^15-
64
58
CO
60
59
^— — &u
Feed
volume
(ml)
0.90
0.99
0.78
0.90
0.82
0.85
0.80
0.72
0.81
0.82
0.82
______Q__ --
0.76
0.82
.Q
1.62
0.47
0.47
n
0.86
0.78
__ Q
0.80
———_____
PCB
cone.
(pg/g)
5
5
5
5
50
50
50
50
500
500
500
500
500
NA
UNK5""
UNK
UNK
NA
— g— -
0
NA
tin
- — — -Q
0
M A
NA
•"•— -
PCB
cone.
(|jg/mL)
4.28
4.28
4.66
4.66
42.8
42.8
46.6
46.6
428
428
428
--—NA
466 — — '
466
-J1A
700,000
700,000
700,000
-NA^
0
0
NA
nn .^.
-^ 6
0
kl A
NA
PCBs in
feed
(mg)
0.004
0.004
0.004
0.004
0.035
0.036
0.037
0.033
0.35
0.35
0.35
^_^^ 0
(OB
0.38
0 3
1,130
330
330
0
0
n
\j
~D "^
0
.NA = not applicable.
UNK = unknown.
57
-------
The second problem that developed during the silicone oil tests was
less obvious initially. High CO levels in the effluent gases remained even
when the waste feed was shut off. This situation continued even during the
first three chlorobenzene runs (Runs 54, 55, and 56). Also, both the system
blanks run during this period showed very high CO levels. At this point, the
pyrolysis furnace tube was opened and several milliliters of an oily liquid
were visible in the tube. It was apparent that the operating temperature for
the pyrolysis furnace was too low to completely vaporize the silicone oil.
Slow decomposition of this residue was likely caused by the high CO levels
observed up to several hours after the flow of silicone oil was stopped. This
residual pool of oil was rinsed out before repeating the now suspect chloro-
benzene tests. Fortunately, subsequent analysis of the rinse sample did not
detect any PCBs. This indicated that either (1) all PCBs passed through the
system, or (2) if any PCBs remained, they were eventually decomposed in the
pyrolysis furnace. As discussed below, thermal analysis was later performed
on the silicone oil, verifying that the pyrolysis furnace temperature was too
low for complete vaporization. The second group of three chlorobenzene runs
was conducted without incident.
Thermal degradation of the askarel fluid resulted in the formation
of black, sooty materials that passed through the XAD-2 trap and resulted in
minor clogging and flow control problems. The pyrolysis furnace temperature
was elevated during these runs to enhance vaporization, although later thermal
analysis of the askarel fluid indicated that this measure was probably unnec-
essary.
3. Thermal Analysis of Dielectric Fluids
Because of operational problems encountered during the Phase 3 deg-
radation tests, as described above, attempts were made to better characterize
the various dielectric fluids. Two types of thermal analysis, differential
scanning calorimetry (DSC) and thermogravimetric analysis (TGA), were con-
ducted. DSC indicates chemical or physical changes (either exothermic or
endothermic) occurring in a substance as it is heated. TGA indicates weight
loss (or gain, if appropriate) of a substance as a function of temperature.
Both DSC and TGA were performed using a heating rate of 20°C/min and an atmo-
sphere of prepurified nitrogen, flowing at 20 or 30 mL/min. Sample sizes for
the analyses ranged from 3 to 6 mg. All four dielectric fluids used in the
Phase 3 thermal degradation tests were studied by DSC. Only the silicone oil
and askarel were analyzed by TGA.
Several interesting results were obtained from the thermal analyses.
Mineral oil undergoes an exothermic reaction at 232°C (probably a decomposi-
tion or rearrangement) and boils at less than 348°C at atmospheric pressure.
This value was obtained from a closed-pan DSC analysis, which may have had an
elevated pressure and thus represents an upper limit. A weight loss of 97%
resulted from heating to 400°C. The chlorobenzene dielectric fluid had a mea-
sured boiling point at or below 264°C. Heating to 300°C resulted in a 97%
weight loss. The askarel fluid exhibited a boiling point of approximately
230°C. Essentially complete weight loss occurred from heating to 300°C.
58
-------
The polydimethylsiloxane (silicone oil) exhibited a strongly exo-
thermic reaction at approximately 285°C. This is probably a polymerization
reaction. Heating to 350°C caused very little weight loss, leaving a sticky,
viscous liquid. The TGA indicated a boiling point of approximately 450°C with
about 90% weight loss occurring between 300 and 500°C. Thus, only about 30%
of the silicone fluid should have vaporized at the pyrolysis furnace operating
temperature (395°C) used during the Phase 3 tests. This coincided with the
visual observation of residual oil remaining in the pyrolysis furnace follow-
ing the silicone oil tests.
4. PCDF and PCDD Analysis Results
The results of the analysis of the Phase 3 samples are
Tables 23 through 28. PCDFs were found in all samples
.blanks,. PCDDs were found in the effluent from the c
occasionally at low levels in other samples (Tables 26 through 28). ^
are not reported, since they were not distinguishable from PCBs under^the
analysis conditions.
As described in Section VII.C.2, runs 39, 54, 55, and 56 were con-
ducted under different conditions than the other runs. Also the first askarel
run (No. 60) produced CO, C02, and 02 levels which varied from those for the
other runs. Since these five runs could not be compared with the other runs,
these runs were not included in Tables 23 through 28.
Since the sampling periods varied from 35 to 120 min and the amount
of feed oil fed varied commensurately, the amount of PCDFs (Tables 23 and 26)
found is also expressed in nanograms (ng) PCDF per milliliter (mL) feed oil
in Tables 24 and 27. This permits direct comparison between runs. The con-
version efficiencies shown in Tables 25 and 28 were calculated in a different
manner than for the Phase 2 samples. The conversion efficiencies shown in
Table 15 (Phase 2) for individual PCDF homologs were calculated by dividing
the concentration of the PCDF by the concentration of the corresponding indi-
vidual PCB isomer in the feed. In Tables 25 and 28 (Phase 3), the conversion
efficiencies for the individual PCDF and PCDD homologs were calculated by
dividing the concentration of the PCDF or PCDD homolog by the total PCB con-
centration in the feed. Consequently, a direct comparison of Tables 15 and
25 cannot be made for the individual PCDF homologs. The total (PCDF) conver-
sion efficiencies in these two tables are, however, comparable, since each
was calculated by dividing the total amount of PCDF formed by the total amount
of PCB feed and then multiplying by 100.
Although a single value is reported for each homolog, characteristic
clusters of isomers were generally observed. The responses for all identified
peaks were simply summed to give the reported values. Representative isomeric
distributions are shown in Figures 10 through 20.
59
-------
Table 23. Amounts of PCDFs Formed in Phase 3
CTl
o
Run no.
8-15-43-M5
8-15-44-M5
8-17-47-S5
8-20-48-S5
8-16-45-M50
8-16-46-M50
8-21-49-S50
8-21-50-S50
8-13-40-M500
8-14-41-M500
8-14-42-M500(B)
8-22-51-S500
8-22-52-S500
8-22-53-5500(8)
S-SO^eP'ASKL
8- 3th 62- /|s KL
8-30M3^«SKL(B)
8-28/S7-CLBZ
8-28\58-jCLBZ
8O O ^C Q ^f* \ D 7 f D *\
L- O 3*5*^ Lr l_ D Z_ I D I
Lab Blank
MonoCDF
(ng)
_a
-
-
-
-
-
-
1,700
-
-
-
50
0.4
810
1,900
28
-
2,000
-
0
DiCDF
(ng)
_
-
-
-
-
-
-
-
0
-
-
-
1,300
0
5,100
7,000
190
-
29,000
-
0
TriCDF
(ng)
130
43
26
31
200
140
290
530
2,200
1,300
0
2,000
5,000
0
440,000
220,000
310
2,400 "
>13,000
81
0
TetraCDF
(ng)
49
23
0
9
110
82
73
640
690
620
13
740
2,100
0
1,400,000
1,100,000
1,200
2,600
>19,000
25
0
PentaCDF
(ng)
NQb
0
90
150
39
21
62
83
43
170
0
340
170
0
6,400,000
4,700,000
17,000
5,000"
>22,000
5
0
HexaCDF
(ng)
Oc
0
0
0
8.5
2.2
0
0
7
13
0
45
12
0
910,000
660,000
3,000
0
5,200
0
0
HeptaCDF
(ng)
_
-
-
-
-
-
0
0
0
0
0
-
0
0
29,000
19,000
-
-
0
-
™
OctaCDF
(ng)
_
-
-
-
-
-
0
0
0
0
0
-
0
o
3,400
1,300
-
-
0
-
~
PCDFs
(ng)
180
66
116
190
350
250
420
1,300
4,700
2,100
13.
3,100
8,600
0,4 .
9,200,000
6,700,000
22,000 ^
9,900 7S
>90,000
110
0
. - = not analyzed.
NQ = not quantitated.
0 = not detected.
-------
Table 24. PCDF Formation in Phase 3
CTi
Run no.
8-15-43-H5
8-15-44-H5
8-17-47-S5
8-20-48-S5
8-16-45-H50
8-16-46-M50
8-21-49-S50
8-21-50-S50
8-13-40-M500
8-14-41-H500
.8-14-42dM5.0.01BJ
8-22-51-S500
8-22-52-S500
8-22-53-5500(8)
8-30-61-ASKL
8-30-62-ASKL
8-30-63- ASKL(B)
8-28-57-CLBZ
8-28-58-CLBZ
8-28-59-CLBZ(B)
HonoCOF
formation
(ng/mL)
_a
2,100
NA"
61
NA
1,700
4,000
NA
2,500
NA
OiCOF
formation
(ng/mL)
0
NA
1,600
NA
11,000
15,000
NA
36,000
NA
TriCDF
formation
(ng/ml)
150
44
33
34
240
170
360
740
2,700
1,600
NA
2,600
6,100
NA
940,000
470,000
NA
2,900
> 16,000
NA
TetraCDF
formation
(ng/mL)
54
23
0
10
130
96
92
900
830
760
NA
980
2,500
NA
3,000,000
2,400,000
NA
3,200
> 24,000
NA
PentaCOF
formation
(ng/mL)
NQb
0
120
170
48
25
78
120
52
210
NA
450
210
NA
14,000,000
9,900,000
NA
6,100
> 28,000
_ _,NA
HexaCOF
formation
(ng/mL)
Oc
0
0
0
10
3
0
0
9
16
NA
60
15
NA
1,900,000
1,400,000
NA
0
6,500
NA
HeptaCDF
formation
(ng/mL)
0
0
0
0
0
NA
0
0
NA
61,000
39,000
NA
0
NA
OctaCOF
formation
(ng/mL)
0
0
0
0
0
NA
0
0
NA
7,200
2,700
NA
0
NA.._
PCDFs
formation
(ng/mL)
200
67
150
210
430
290
530
1,800
5,700
2,600
NA
4,100
11,000
NA
19,000,000
14,000,000
NA
12,000
> 110,000
NA
. - = not analyzed.
°NQ = not quantified.
JjO = not detected.
NA = not applicable.
-------
Table 25. Conversion Efficiencies (PCBs to PCOFs) for Phase 3
Run no.
8-15-43-H5
8-15-44-H5
8-17-47-S5
8-20-48-S5
8-16-45-H50
8-16-46-H50
8-21-49-S50
8-21-50-S50
8-13-40-M500
8-14-41-M500
8-22-51-S500
8-22-52-S500
1 ->^ 8-22-53-5500(8)
8-30-61-ASKL
8-30-62-ASKL
-r- ^ 8-30-63-A5KL(B)
8-28-57-CLBZ
8-28-58-CLBZ
— ,8-28-59-CLBZ(B)
rvi
HonoCDF
conversion
efficiency
(%)
_a
0.49
NAd
0.013
NA
0.0002
0.006
NA
NA
NA
NA
01CDF TrlCOF
conversion conversion
efficiency efficiency
(%) (X)
0
NA
0.35
NA
0.0015
0.0021
NA
NA
NA
NA
3.4
1.0
0.71
0.74
0.55
0.40
0.78
1.6
0.64
0.38
NA
0.56
1.3
NA
0.13
0.067
NA
NA
NA
NA
TetraCDF
conversion
efficiency
(X)
1.3
0.55
0
0.21
0.31
0.22
0.20
1.9
0.19
0.18
NA
0.21
0.54
NA
0.43
0.34
NA
NA
NA
NA
PentaCDF HexaCOF
conversion conversion
efficiency efficiency
(X) (X)
NQb
0
2.5
3.6
0.11
0.058
0.17
0.25
0.012
0.048
NA
0.096
0.044
NA
1.9
1.4
NA
NA
NA
NA
0C
0
0
0
0.024
0.0060
0
0
0.0020
0.0037
NA
0.013
0.0031
NA
0.28
0.20
NA
NA
NA
NA
HeptaCDF
conversion
efficiency
(X)
0
0
0
0
NA
0
NA
0. 0087
0.0056
NA
NA
NA
NA
OctaCOF
conversion
efficiency
(X)
0
0
0
0
NA
0
NA
0.0010
0.0004
NA
NA
NA
NA
PCOFs
conversion
efficiency
(X)
4.7
1.6
3.2
4.5
1.0
0.67
1.1
3.8
1.3
0.61
NA
0.88
2.3
NA
2.8
2.0
NA
NA
NA
NA
.- = not analyzed.
NQ = not quantitated.
•JO = not detected.
°NA = not applicable.
High levels saturated the detector signal.
-------
Table 26. Amounts of PCDDs Formed in Phase 3
01
OJ
Run no.
8-15-43-M5
8-15-44-M5
8-17-47-S5
8-20-48-S5
8-16-45-M50
8-16-46-M50
8-21-49-S50
8-21-50-S50
8-13-40-M500
8-14-41-M500
8-14-42-M500CB")
8-22-51-S500
8-22-52-S500
8-22-53-5500(6)
8-30-61-ASKL
8-30-62-ASKL
8-30-63-ASKLCB1
8-28-57-CLBZ
8-28-58-CLBZ
8-28-59-CLBZCB)
Lab Blank
MonoCDD
(ng)
_a
-
-
-
-
-
-
-
0
-
-
-
0
0
0
0
0
0
-
0
DiCDD
(ng)
.
-
-
-
-
-
-
-
0
-
-
-
0
0
0
0
0—
0
-
0
TriCDD
(ng)
Ob
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
1,100
630
0
0
TetraCDD
(ng)
0
0
0
0
0
0
0
0
0
0
0
0
0 .
0
0
0
0
440
520
0
0
PentaCDD
(ng)
0
0
0
0
0
0
0
0
0
0
0
7.7
1.7
0
0
0
0
0
0
0
0
HexaCDD
(ng)
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
0
72
0
0
HeptaCDD
(ng)
_
-
-
-
-
-
0
0
0
0
0
-
0
0
330
230
_
-
0
-
—
OctaCDD
(ng)
_
-
-
-
-
-
0
0
0
0
0
0
0
37
51
_
-
0
-
—
PCDDs
(ng)
0
0
0
0
0
0
0
0
0
0
0
7.7
1.7
..._.._ Q
360
280
0
1,500
1,200
0
0
,- = not analyzed.
0 = not detected.
-------
Table 27. PCDO Formation in Phase 3
en
Run no.
8-15-43-H5
8-15-44-H5
8-17-47-S5
8-20-48-S5
8-16-45-M50
8-16-46-H50
8-21-49-S50
8-21-50-S50
8-13-40-M500
8-14-41-M500
8-14-42-H500(B)
8-22-51-S500
8-22-52-S500
8-22-53-5500(8)
8-30-61-ASKL
8-30-62-ASKL
8-30-63-ASKL(B)
8-28-57-CLBZ
8-28-58-CLBZ
8-28-59-CLBZ(B)
HonoCDD
formation
(ng/mL)
_a
-
-
-
-
-
-
-
0
-
NAC
-
-
NA
0
0
NA
-
0
NA
DiCOD
formation
(ng/mL)
_
-
-
-
-
-
-
-
0
-
NA
-
-
NA
0
0
NA
-
0
NA
TriCDO
formation
(ng/mL)
Ob
0
0
0
0
0
0
0
0
0
NA
0
0
NA
0
0
NA
1,300
790
NA
TetraCDO
formation
(ng/mL)
0
0
0
0
0
0
0
0
0
0
NA
0
0
NA
0
0
NA
540
650
NA
PentaCDD
formation
(ng/mL)
0
0
0
0
0
0
0
0
0
0
NA
10
2
NA
0
0
NA
0
0
NA
HexaCOD
formation
(ng/mL)
0
0
0
0
0
0
0
0
0
0
NA
0
0
NA
0
0
NA
0 .
90
NA
HeptaCDD
formation
(ng/mL)
_
-
-
-
.
-
0
0
0
0
NA
-
-
NA
690
490
NA
-
0
NA
OctaCDO
formation
(ng/mL)
_
-
-
-
-
-
0
0
0
0
NA
-
-
NA
78
110
NA
-
0
NA
PCDOs
formation
(ng/mL)
0
0
0
0
0
0
0
0
0
0
NA
10
2
NA
770
600
NA
1.900
1,500
NA
h- = not analyzed.
°0 = not detected.
NA = not applicable.
-------
Table 28. Conversion Efficiencies (PCBs to PCDDs) for Phase 3
CTi
cn
Run no.
8-15-43-M5
8-15-44-M5
8-17-47-S5
8-20-48-S5
8-16-45-H50
8-16-46-M50
8-21-49-S50
8-21-50-S50
8-13-40-M500
8-14-41-H500
8-14-42-H500(B)
8-22-51-S500
8-22-52-S500
8-22-53-S500(B)
8-30-61-ASKL
8-30-62-ASKL
8-30-63-ASKL(B)
8-28-57-CLBZ
8-28-58-CLBZ
8-28-59-CLBZ(B)
HonoCOD
conversion
efficiency
(%)
_a
-
-
-
-
-
-
-
0
-
NAC
-
0
NA
0
0
NA
NA
NA
NA
D1CDD
conversion
efficiency
(X)
_
-
-
-
-
-
-
-
0
-
NA
-
0
NA
0
0
NA
NA
NA
NA
TrICDD
conversion
efficiency
(%)
Ob
0
0
0
0
0
0
0
0
0
NA
0
0
NA
0
0
NA
NA
NA
NA
TetraCOD
conversion
efficiency
(%)
0
0
0
0
0
0
0
0
0
0
NA
0
0
NA
0
0
NA
NA
NA
NA
PentaCOO
conversion
efficiency
(X)
0
0
0
0
0
0
0
0
0
0
NA
0.0022
0.0004
NA
0
0
NA
NA
NA
NA
HexaCDD
conversion
efficiency
(X)
0
0
0
0
0
0
0
0
0
0
NA
0
0
NA
0
0
NA
NA
NA
NA
HeptaCDO
conversion
efficiency
(X)
_
-
-
-
-
-
0
0
0
0
NA
-
0
NA
0.000099
0.000070
NA
NA
NA
NA
OctaCDD
conversion
efficiency
(X)
-
-
-
-
-
-
0
0
0
0
NA
-
0
NA
0.000011
0.000015
NA
NA
NA
NA
PCODs
conversion
efficiency
(X)
0
0
0
0
0
0
0
0
0
0
NA
0.0022
0.0004
NA
0.00011
0.00009
NA
NA
NA
NA
.- = not analyzed.
0 = not detected.
•JNA = not applicable.
High levels saturated the detector signal.
-------
CTl
CTl
MID MASS CHROI1ATOGRA1S DATA: 8201I12R3 «1
03/12/84 10:56:00 CALI: MI0180I12 #2
SAMPLE: ?-22-52-S508 FCDD/F/BFH 1UL CL1-2
CONDS.: -2098EW 70E1.1 IMA DB5-30M-100-2H-320-10/
RANGE: G_ 1, 309 LABEL: N 0, 4.0 QUAH: A 0, 1.0 J 0 BASE: U 23, 3
626
Mono CDFs
SCANS 6f?9 TQ 950
232 _
13.6-
218 _
188.8-1
236 _
7?9
663
698
C38
790
830
No mono CDDs identified
652 672 . 632.702 725 746
"i 1 1 1 pi 1 1"_ c _ li" * [J. l i n—i—i—l \z £ .[ rSr
752
di COFs
.201.939
844 ± 0.580
113280.
217.935
± 0.500
577536.
235.929
± 0.500
8 J
0.3-
252 _
No di CDDs identified
1 1 1 1 | 1 1 F 1 | 1 1
600 S50 700
11:41 12:33 13:38
737
1 ^T | 1
750
14:36
788
777 f, 805
ji H j jl )
889
15:34
jfv
919 '
r ^ i \ '
A
&
16
5136.
251.924
850 SCAN
16:33 TIME
Figure 10. Mono and di CDFs and CDDs in sample 8-22-52-S500.
-------
en
109.0-1
279 _
0.6-
285 _
20.8-
396 _
0.5-
322 _
MID MASS CHROMATOGRAH5 DATA: 8201H27R6 #1
08/27/84 11:27:00 . CALI: MID250H21 *2
SAMPLE: 8201A23-RUN 8-22-52-S500 1UL INJ
CONDS.: -2090EMU 70EY IMA DB5-30n-80-2H-320-10/
RANGE: G 1/1200 LABEL: N 0, 4.0 QUAN: A 0, 1.0 J 0 BASE: U 23, 3
953
Tr1 CDFs
SCANS 900 TO 1100
989
18:18
983
1037 1053
1074
-ar
1060
No tri CDDs identified
953 354
1016
"P!\ f i*»
i
1 A 1086
1 1 /\ ,A ys.
1061
1069
Tetra CDFs
1036
1013
1060
No tetra CDDs identified
1110010.
269.319
± 0.500
6840.
285.314
± 0.500
230312.
305.998
± 0.500
5624.
321.903
± 0.53?
350
19:10
1000
20:11
1058
21:11
1100 SCAN
22:12 TIME
Figure 11. Tri and tetra CDFs and CDDs in sample 8-22-52-S500.
-------
CTl
OO
MID MASS CH90MATOGRAHS DATA: 8201112X4 #1
03/12/84 17:16:00 • CALI: MID315I12X1 #3
SAMPLE: 3-22-52-S500 1UL PCOD/F/BPN CL5-6
CONDS.: MG=3.5 70EU BC=3 DB5-15M 190-1H-325-10/ 45 SEC SPLTL.
RANGE: G 1,1574 LABEL: N 0, 4.0 QUAN: A 0, 1.0 J 0 BASE: U 20< 3
1110
Penta CDFs
SCAHS 1650 TO 1300
1278 1291
1132 1151 1165 1177
No hexa CDDs identified
1130
374
390 _
1650
16:10
1100
16:56
1200
18:29
1250
19:15
234312.
340.102
± 0.500
7696.
356.105
± 0.509
31520.
374.112
± 0.569
2380.
390.117
± 0.500
1300 SCAH
20:01 TIME
Fiqure 12. Penta and hexa CDFs and CDDs in sample 8-22-52-S500.
-------
MID MASS CHROHATOGRAfIS DATA: 8291I11R2 #1 SCANS 650 TO 850
09/11/84 11:19:60 ' CALI: MID390I11 #2
SAMPLE: 8201-A23 RUN 8-22-52-S500 111 !NJ. CL7-8 PCDD/^/BPH
COND3. : -2000EMU 79EU IMA DB5-30I1-288-2H-328-10/
RANGE: G 1,1000 LABEL: N 0, 4.0 QUAN: A 0, 1.8 J 0 BASE: U 22, 3
408 _
8.0-1
424 _
0.0n
444 _
0.0-1
450 _
//v
No hepta CDDs identified
No octa CDFs identified
No octa CDO identified
12836.
487.803
654* ^fo+J' wwv" ~ * 0-530
No hepta CDFs identified
1.
423.800
± 0.500
1.
443.708
± 0.500
1.
459.700
± 8.583
—i 1 i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
658 780 750 800 850 SCAM
13:11 14:12 15:12 16:13 17:14 TIME
Figure 13. Hepta and octa CDFs and CDDs in sample 8-22-52-S500.
-------
MID MASS CHRQMATOGRWS DATA: 8201I12R9 $1
09/12/84 17:01:00 . CALI: MID180I12 #2
SAMPLE: 8-39-62-ASKL PCDD/F/BPN 1UL CL1-2
CONDS.: -2000EM'J 70E1,1 IMA DB5-39M-108-2H-320-10/
RANGE: G 1, 900 LABEL: N 0, 4.0 QUAN: A 0, 1.8 J 0 BASE: U 28- 3
SCANS 550 TO 839
100.0-1
202 _
651 673 685 701 713 734
No mono CDDs identified
1.4-1
252 _
805
No di CDDs identified 739
550
19:42
600
11:41
650
12:39
700
13:38
750
14:36
15:34
4338SS0.
201.939
± 0.580
248576.
217.935
± 0.580
1533958.
235.929
± 0.500
58432.
251.924
9.569
SCAN
TIME
Figure 14. Mono and di CDFs and CDDs in sample 8-30-62-ASKL.
-------
59.8-1
270 _
MID MASS CHROMATOGRAMS DATA: 8201I15RS #1
09/15/84 16:34:00 • CALI: MID250H21 #2
SAMPLE: 8201A23 RUN 8-30-62 ASKL 1/100 OIL 1UL IHJ
CONDS.: -2200EMY 78EV IMA DB5-3PM 188-1H-320-1P/
RANGE: G 1,1380 LABEL: N 0, 4.0 QUAN: A 3, 1.0 J 0 BASE: U 20, 3
Tr1 CDFs
327 338 953_967980
SCANS 850 TO 1840
0.6-1
286
100.8-,
306
322
1815
1001
No tri CDDs identified
920
1914
0-
No tetra CDDs identified
859 980
15:05 15:59
953 967
950
16:52
1023 h
1801 1 A /'
985 ^A^_ / v y \ r^./
1000
17:45
1380350.
269.919
± 0.580
14368.
285.914
± 0.508
2310149.
305.908
± 0.500
69376.
321.903
± 0.500
SCAN
TIME
Figure 15. Tri and tetra CDFs and CDDs in sample 8-30-62-ASKL.
-------
—I
ro
100.0-1
340 _
MID MASS CHROMATOGRAMS DATA: 8201I15R1 #1
69/15/84 8:35:00 CALI: MID315I14R1 »3
SAMPLE: 8-30-62-ASKL 1/108DIL PCDD/F/BFN 1UL CL5-6
COHDS.: -2200EMU 79EV IMA DB5-30M 100-1H-320-10/
RANGE: G 1,1608 LABEL: N 0, 4.0 QUAN: A 0, 1.0 J 0 BASE: U 20, 3
1250
1298
Penta CDFs
SCANS 1200 TO 1599
No penta CDDs identified
No hexa CDDs identified
390 _
1289
18:41
1250
19:28
29:15
1350
21:01
1400
21:48
1450
22:35
7020548.
339.898
± 0.580
76032.
355.893
± 0.500
448512.
373.888
± 0.580
4540.
389.883
1500 SCAN
23:21 TIME
Figure 16. Penta and hexa CDDFs and CDDs in sample 8-30-62-ASKL.
-------
GO
MID MASS CHRWATOGRAMS DATA: 8201I11R12 #1
09/11/84 16:50:00 CALI: MID390I11 »2
SAMPLE: 8-30-62-ASKL (D/F) 1UL FCDO/F/BPN CL7-8
CONDS.: -2000EMU 70EV IMA QB5-30M-200-2H-320-10/
RANGE: G 1, 900 LABEL: N 0, 4.0 QUftN: A 0, 1.0 J 0 BASE: U 20, 3
696
SCANS 689 TO 850
100.8-1
403 _
2.2-1
424
10.1-1
444
0.6-, 6I8 695
460 _
:735 Hepta CDFs
A
,/ V. 747 758 786
817
723
Mepta CDDs
825
729 742 751 753 773 _782
689 709 720
13:47 14:12 14:36
740 768 780 800 828 840
15:00 15:25 15:49 16:13 16:38 17:02
5455870.
407.609
± 0.503
117760.
423.800
± 0.500
551935.
443.780
± 0.500
35072.
459.700
± 0.500
SCAN
TIME
Figure 17. Hepta and octa CDFs and CDDs in sample 8-30-62-ASKL.
-------
100.0-1
270 _
20.5-1
236 _
39.3-1
306 .
56.4-1
322 _
MID MASS CHROMATOGRAMS DATA: 8201111X5 #1
09/11/84 11:22:00 ' CALI: HID250I10X1 #3
SAMPLE: 8201A23 RUN 8-15-43-M5 1UL INJ
CONDS.: MG=3.5 70EU BC=3 DB5-15M 100-1H-325-10/
RANGE: G 1,1280 LABEL: H 0, 4.9 QUAN: A 0, 1.0 J 0 BASE: U 20, 3
847
SCANS 789 TO 1020
No tri CDDs identified
969
see
13:68
No
-. L -J , X.
350
13:57
tetra CDDs Identified
878 911
L 1 1 C .
300
14:47
934 946 *2!L>
\~^ J®3- __J5i2~
— i 5 T"^ ' " ' ' 1 ' '
950 1000
15:36 16:25
2121728.
270.831
± 8.533
434S8S.
28S.085
± 0.590
833536.
306.092
± 0.500
1196030.
322.096
± 0.560
SCAM
TIME
Figure 18. Tri and tetra CDFs and CDDs in sample 8-15-43-M5.
-------
err
48.9-1
270 _
13.7-
285 _
84.1-
306 _
MID MASS CHROMATQGRAMS DATA: 8291111X3 #1
09/11/84 10:18:00 , CALI: MID250I10X1 #3
SAMPLE: 8201A23 RUN 8-17-47 S5 XAD 1UL INJ
CONDS.: MG=3.5 70EU BC=3 DB5-15M 100-1H-325-10/
RANGE: G 1,1200 LABEL: N 0, 4.0 QUAN: A @, 1.0 J 0 BASE: U 20, 3
846
SCANS 750 TO 1059
1027
J.../W,
449536.
270.081
± 0.509
126238.
286.085
± 0.5S0
772036.
306.092
± 0.500
1008
A... . 1*2,
100.0-
322 _
7
12
No tetra CDDs identified
50 800 850 900
:19 13:08 13:57 14:47
3fU
1807
950 1000
15:36 16:25
1042
10
17
918528.
322.096
± 0.500
1050 SCAN
17:14 TIME
Figure 19. Tri and tetra CDFs and CDDs in sample 8-17-47-S5.
-------
MID MASS CHROMATOGRAHS . DATA: 8201116X2 #1
03/10/84 15:45:00 CALI: I1ID250I10X1 #3
SAMPLE: 8201A23 RUN 8-28-57 1UL INJ
CONDS.: MG=3.5 70EU BC=3 DB5-15M 109-1H-325-19/
RANGE: G 1,1258 LABEL: N 0, 4.0 QUAH: A 0, 1.0 J 0 BASE: U 28. 3
SCANS 809 TO 1050
100.0-1
270 _
43.2-1
236 .
33.1-1
306 .
846
333
931
879 J[ 904 A
-f\—^J w\L / V
878
tri CDDs
816 838 A
A A/A AA.
17.6-1
322 .
3
13
90 859
:03 13:52
tetra CDDs
934
372 832 ^v 921 ^A^
900
14:41
951
A
— i —
958
15:30
1010
1000
16:19
3203070.
278.081
4 0.509
1576350.
286.085
± 0.500
3141630.
306.092
± 0.500
563200.
322.036
t 0.508
1050 SCAN
17:07 TIME
Figure 20. Tri and tetra CDFs and CDDs in sample 8-28-57-CLBZ.
-------
s">kown in Tables"23 and 24 and Figure 21, the amount of PCDF vv •//
formed Wnecal-TyHnepeased with amount of PCBs fed. It should also be noted J^jr^
in Table 23 that two bfa^:=rw4^;8-28-59-CLBZ(B) and 8-30-63-ASKL(B)] con- fi V\
tained measurable amounts of PCDPs>xEach of these blank runs was made imme- ' \
di ateTy^arier runs which produced P£DFs in the microgram to milligram range.
This possTfrH4t-yJjad ^ been^anti-e-fpated , and the run order was designed to min-
imize the influence ~6T~anaTyte carryover on consecutive runs.
In addition to investigating the effects of PCB concentration on
the rate of PCDF formation, it is also instructive to compare the relative
distribution by homolog. This is presented in Figures 22 through 24. The
askarel tests yielded a normal distribution, peaking at pentaCDF. The M500
and S500 tests yielded somewhat less regular patterns, although the maximum
formation was in the tri-, tetra-, and pentaCDF homologs. This irregular pat-
tern is particularly apparent in Figure 22, where a large amount of monoCDF
was found in one M500 sample, but no diCDF was found in that sample.
PCDFs and, to a lesser extent, PCDDs are formed from chlorobenzene
dielectric fluid under the optimum PCB-to-PCDF conversion conditions. The
conversion efficiency of the trichlorobenzene feed was > 0.004% to PCDFs and
0.0001% to PCDDs. The homolog distribution of PCDFs is similar to that found
from feeding PCBs as shown in Figure 25. The rate of formation is one to two
orders of magnitude lower than for askarel, but substantially higher than that
for dielectric fluid containing 500 ppm or less PCBs.
5;. PCB Analysis Results
The Phase 3 samples from the spiked mineral oil and spiked si li cone
oil runs (Runs 40 through 53) were analyzed for PCBs. Table 29 lists the re-
sults of these analyses. No PCBs were detected in the blank samples above
the limit of detection. However, these results were not included in this
table. The results of analysis of the 5-ppm and 50-ppm spiked mineral oil
samples are not included because the high hydrocarbon background in these
samples prevented the quantisation of PCBs. Table 29 also lists the percent
composition of Aroclor 1254 by homolog as presented by Brinkman and De Kok
(1980).
From these PCB concentrations, the PCB destructon efficiency (D.E.)
was calculated for each run using the following equation:
D.E. = x 100
Win
where W. = PCB feed (Table 22)
Wout = PCBs 1n Combust1on effluent (Table 29)
77
-------
10,000
-J
00
100
500
PCB ConcenfraHon (ppm)
Figure 21. Averaqe PCDF formation versus PCB concentration for Phase 3.
-------
10,000
U5
1.000
l 100
Q
k!
10
ND
Mono Di
Tri
Tehra Penta
PCDF
* 500 ppm
S—— 50 ppm
9 5 ppm
Hexa Hepta Octa
Figure 22. PCDF formation in PCB-spiked mineral oil by homolog. Closed symbols are averages
of two values; open symbols are single determinations; missing points are no data.
-------
00
o
10.000
1,000
TJ
E 100
_i
£
Q
10
ND
Mono
Di
Tri
Telra Penfa
PCDF
Hexa
+. 500 ppm
• 50 ppm
•—— 5 ppm
I
Hep»a
Octa
Figure 23. PCDF formation in PCB-spiked silicone oil by homolog. Closed symbols are averages
of two values; open symbols are single determinations; missing points are no data.
-------
00
Q
u
Ou
O)
c
100,000,000
10,000,000
1,000,000
100,000
10,000
1,000
100
10
Not'
Detected
1 % Conversion
Efficiency
Mono
Di
I
1
I
Tri
Tetra Penta
PCDF
Hexa
Hepta Octa
Figure 24. PCDF formation from PCB askarel fluid. Points are averages of two values.
-------
00
2
'D
U_
E
u_
Q
O
Q.
c5
IUU,UUU,UUU
10,000,000
1,000,000
100,000
10,000
1,000
100
10
0
Not^
Detected^
-
-
_
Q^
- _/' ~~~. — -— — -*~£5f.
' / X
• \Q) • \
.* \ ^/^ ^* V »
/ \J S ^- x
; / \ /' \\ J
- o O \S \b._ Q
i i i i I ii I
Mono Di Tri Tetra Penta Hexa Hepta Octa
Homolog
Figure 25. PCDF and PCDD formation from trichlorobenzene transformer fluid. Closed
symbols are averages of two values; open symbols are single determinations.
-------
Table 29. Results of Analysis of PCBs in Phase 3 Samples (ng/sample)
^^\^^ Sample
^\^ no.
Analyte^~\^^
Monochlorobiphenyl
Oichlorobiphenyl
Trichlorobiphenyl
Tetrachlorobiphenyl
Pentachlorobiphenyl
Hexachlorobiphenyl
Heptachlorobiphenyl
0° Octachlorobiphenyl
Nonachl orobi pheny 1
Oecachlorobiphenyl
Total PCBsb
8-17-47-S5
< 20
< 20
< 20
< 50
< 75
< 75
< 75
< 100
< 100
< 100
8-22-48-S5
< 20
< 20
< 20
360
160
< 75
< 75
< 100
< 100
< 100
520
8-21-49-S50
< 20
< 20
< 20
600
6,000
1,100
< 75 .
< 100
< 100
< 100
7,700
8-21-50-S50
< 20
< 20
150
1,800
2,600
1,100
< 75
< 100
< 100
< 100
5,600
8-22-51-S500
< 50
< 50
96
5,600
7,400
2,800
< 150
< 200
< 200
< 200
13,000
8-22-52-S500
< 50
< 50
640
17,000
19,000
10,000
540
< 200
< 200
< 200
47,000
8-13-40-M500
< 50
< 50
< 50
20,000
11,000
3,000
< 150
< 200
< 200
< 200
34,000
8-14-41-M500
< 50
< 50
< 50
13,000
16,000
5,800
< 150
< 200
< 200
< 200
35,000
Aroclor 1254
(% composition)
0
0
1
15
53
26
4
0
0
0
^Literature value, taken from Brinkman and De Kok (1980).
In calculating total PCB concentration, less than values were considered zero.
-------
The calculated destruction efficiencies are shown in Table 30.
Table 30. PCB Destruction Efficiencies in Phase 3 Runs
Ruh no.
Destruction efficiency
8-13-40-M500
8-14-41-M500
8-17-47-S5
8-20-48-S5
8-21-49-S50
8-21-50-S50
8-22-51-S500
8-22-52-S500
90
90
> 99
87
79
83
96
86
As described above, destruction efficiency calculation for the 5-ppm
and 50-ppm spiked mineral oil runs was not possible. Destruction efficiencies
were not measured for the askarel runs (Runs 61 and 62).
6. Statistical Analysis
The data for the PCDFs from the three concentrations, 5, 50, and
500 ppm, in both mineral oil and silicone oil were statistically analyzed.
Separate analyses were performed on the total PCDFs and the four homologs,
tri-, tetra-, penta-, and hexaCDF, which exhibited sufficient response for
analysis.
Because of slightly different burn times, the amounts of PCDFs from
different runs would not be directly comparable. Consequently, a multiple
regression approach to the analysis was preferred over the two-way analysis
of variance. This allowed for consideration of the amount of PCBs in the feed
oil rather than just the three nominal levels. The results of both methods
of analysis (regression and ANOVA) agreed closely.
All five of the variables analyzed were consistent with a zero in-
tercept. That is, the test of the intercept equal to zero was nonsignificant
at the 5% level in all cases. A quadratic in the concentrations was also con-
sidered. This was also nonsignificant at the 5% level, so the linear compo-
nent in the means is sufficient.
The mean PCDF values in the silicone oil were higher than in mineral
oil. However, the differences were not statistically significant for the
total nor for any homolog except pentaCDF. This is interpreted to mean that
production of PCDFs under these conditions may be higher for silicone oil.
The differences were substantial, but were not statistically significant,
probably because of large variability and relatively small sample size. No
significant interaction between concentration and oil type was found.
84
-------
The levels of triCDF, tetraCDF, pentaCDF, hexaCDF, and total PCDF
all showed a significant relationship to the levels of PCB in the feed oil.
In each case, the mean amount of the isomer of furans produced in the thermal
combustion system increased monotonically with the concentration of PCB in
the waste. Table 31 presents the means by isomer, by matrix, by concentra-
tion, and by matrix-concentration combination.
In conclusion, the Phase 3 results indicate that PCDFs are readily
formed from PCBs and trichlorobenzenes under the conditions used in this study
(675°C for 0.8 s, with 8% excess oxygen). The statistical analysis indicates
that the PCDF formation is approximately linear with the amount of PCBs in
the feed, with a zero intercept. The amount of PCDF produced may be about a
factor of two higher for a silicone oil matrix than for a mineral oil matrix.
7. Comparison of Feed and Product Compositions
As can be seen from the results (Tables 23 through 25 and Figures 22
through 24), the PCDFs formed from the feed oils containing PCBs have a homo-
log distribution which maximizes at triCDF for the Aroclor 1254 feeds and at
pentaCDF for the Aroclor 1260 feeds. Figures 26 through 28 present the data
in Figures 22 through 24 with an overlay of the Aroclor feed. In both cases,
the PCDF curve is about two orders of magnitude lower, reflecting about 1%
conversion efficiency (as rioted in Table 25). In addition, the PCDF profiles
peak at a lower chlorination number than for the corresponding PCB feed, indi-
cating a loss of chlorine in the thermochemical reactions. The PCDFs formed
from the M500 and S500 oils contained 2.7 and 3.0 chlorines per molecule,
respectively. The Aroclor 1254 has an average of 5 chlorines, indicating that
the average reaction has a loss of about 2 chlorines. This would be consis-
tent with Mechanism 1 in Section IV.B. For Runs 61 and 62, Aroclor 1260 with
an average of 6.25 chlorines per molecule was fed. The PCDF composition of
the products had an average of 4.8 chlorines per molecule. Thus, for these
runs, the average reaction involved a loss of 1.5 chlorines, indicating that
other mechanisms, in addition to Mechanism 1, must be involved.
85
-------
Table 31. Means of PCDF Formed in Phase 3, Grouped by Matrix and Concentration
oo
Matrix
M
S
M
M
M
S
S
S
Concentration
5
50
500
5
50
500
5
50
500
N
6
6
4
4
4
2
2
2
2
2
2
TriCDF
682
1,310
58
290
2,641
88
170
1,789
29
409
3,494
TetraCDF
262
591
20
227
1,033
36
95
656
5
359
1,410
PentaCDF
46
149
60
51
181
0
30
107
120
73
255
HexaCDF
5
10
0
3
19
0
5
10
0
0
29
PCDFs
995
2,060
139
570
3,874
124
301
2,561
153
840
5,187
-------
00
Q
u
Q-
O)
c
,000,000
100,000
10,000
1,000
100
10
Not
Detected
53%
/
/
/
PCB Feed
500 ppm Aroclor 1254
26%
\
\
\
X«4%
1%
9
Mono Di
Tri Tetra Penta
PCDF
Hexa Hepta Octa
Figure 26. Comparison of PCDFs formed with PCB feed composition (mineral oil)
-------
00
00
Q
U
Q_
O>
c
1,000,000
100,000
10,000
1,000
100
10
Not
Detected
53%
PCB Feed
500 ppm Aroclor 1254
Mono Di Tri Tetra Penta
PCDF
N»4%
Hexa Hepta Octa
Figure 27. Comparison of PCDFs formed with PCB feed composition (si "I i cone oil)
-------
PCB Feed
42%
38%
CO
'5
u_
PCDF/mL
O)
100,000,000
10,000,000
1,000,000
100,000
10,000
1,000
100
10
Not'
Detected,
1 2 % 9r '•••x^
* 1 % Conversion ^^^^"'* ^^^^ *""£••
_ Eff,c,ency ^CP^* .
- / \
-
-
1 1 1 1 1 1 1 1 1
Mono Di Tri Tetra Penta Hexa Hepta Octa Nona
PCDF
Fioure 28. Comparison of PCDFs formed with PCB feed composition (askarel).
-------
VIII. REFERENCES
Brinkman UATh, De Kok A. 1980. In: Halogenated biphenyls, terphenyls,
naphthalenes, dibenzodioxins and related products. Production, properties
and usage. Kimbrough RD, ed. New York: Elsevier/North-Holland Biomedical
Press, pp. 1-40.
Buser HR, Bosshardt H-P, Rappe C. 1978. Formation of polychlorinated di-
benzofurans (PCDFs) from the pyrolysis of PCBs. Chemosphere 7(1):109-119.
Buser HR, Rappe C. 1979. Formation of polychlorinated dibenzofurns (PCDFs)
from the pyrolysis of individual PCB isomers. Chemosphere 8(3):157-174.
Buser HR. 1979. Chemosphere 8:415.
Jansson B, Sundstrom G. 1982. Formation of polychlorinated dibenzofurans
(PCDF) during a fire accident in capacitors containing polychlorinated bi-
phenyls. In: Chlorinated dioxins and related compounds, impact on the en-
vironment. Hutzinger 0, et al., eds. Elmsford, NY: Pergamon Press.
Morita M, Nakagawa J, Akiyama N, Minura S, Isono N. 1977. Detailed examina-
tion of polychlorinated dibenzofurans in PCB preparations and Kanemi Yusho
Oil. Bull Environ Contam Toxicol 18(1):67-73.
Morita M, Nakagawa J, Rappe C. 1978. Polychlorinated dibenzofuran (PCDF)
formation from PCB mixture by heat and oxygen. Bull Environ Contam Toxicol
19:665-670.
Rappe C, Marklund S, Bergquist P-A, Hansson M. 1983. Polychlorinated di-
benzo-£-dioxins, dibenzofurans, and other polynuclear aromatics formed during
incineration and polychlorinated biphenyl fires. In: Chlorinated dioxins
and dibenzofurans in the total environment. Choudhary G, Keith LH, Rappe C,
eds. Butterworth Publishers, pp. 99-124.
Rappe C. 1984. Analysis of polychlorinated dioxins and furans. Environ-
mental Sci and Tech 18:78A.
Smith RM. 1982. Analysis for 2,3,7,8-tetrachlorodibenzofuran and 2,3,7,8-
tetrachlorodibenzodioxin in a soot sample from a transformer explosion in
Binghamton, New York. Chemosphere 11:715-720.
USEPA. 1984a. Polychlorinated biphenyls (PCBs); manufacture, processing,
distribution in commerce and use prohibitions; use in electrical transformers.
Advanced notice of proposed rulemaking. (49 FR 11070-11083).
USEPA. 1984b. Polychlorinated biphenyls (PCBs); manufacture, processing,
distribution in commerce and use prohibitions; use in electrical transformers.
Proposed rule. (49 FR 39966-39989).
Vuceta J, Marsh JR, Kennedy S, Heldeman L, Wiley S. 1983. State-of-the-art
review: PCDDs and PCDFs in utility PCBF fluid. CS-3308. Electrical Power
Research Institute. Palo Alto, California.
90
-------
TECHNICAL REPORT DATA
(Please read Instructions on the reverse before completing)
1. REPORT NO.
EPA 560/5-84-009
3. RECIPIENT'S ACCESSION NO.
4. TITLE AND SUBTITLE
Thermal Degradation Products from Dielectric Fluids
5. REPORT DATE
December 1984
6. PERFORMING ORGANIZATION CODE
8201-A(23)
?.AUTHORtsrMitchell D. Erickson, Christopher J. Cole,
Jairus D. Flora Jr., Paul G. Gorman, Clarence L. Haile
Gary D. Hinshaw, Fred C. Hopkins, Stephen E, Swanson
8. PERFORMING ORGANIZATION REPORT NO.
Interim Report No. 1
9. PERFORMING ORGANIZATION NAME AND ADDRESS
Midwest Research Institute
425 Volker Boulevard
Kansas City, Missouri 64110.
10. PROGRAM ELEMENT NO.
Work Assignment 23
11. CONTRACT/GRANT NO.
EPA Contract No. 68-02-3938
12. SPONSORING AGENCY NAME AND ADDRESS
U.S. Environmental Protection Agency
Office of Toxic Substances
Field Studies Branch, TS 798
401 M Street, S.W., Washington, DC 20460
13. TYPE OF REPORT AND PERIOD COVERED
Interim (Mav-November 1984^
14. SPONSORING AGENCY CODE
15. SUPPLEMENTARY NOTES
The EPA Work Assignment Manager is Daniel T. Heggem, (202) 382-3990.
The EPA Project Officer is Frederick W. Kutz, (202) 382-3569.
16. ABSTRACT
Electrical transformer fires can cause extensive smoke damage, especially when poly-
chlorinated biphenyls (PCBs) are involved since they can form polychlorinated dibenzo-
furans (PCDFs) and other toxic by-products. To characterize the potential for by-
product formation, this study was undertaken to optimize conditions for PCDF formation
from PCBs and to study the potential for formation of PCDFs and polychlorinated diben-
zodioxins (PCDDs) from combustion of selected dielectric fluids, including those con-
taminated with PCBs. A bench-scale thermal destruction system was used to combust the
samples. The dielectric fluid was fed continuously using a syringe pump. The concen-
trations of CO, C02, and 02 in the effluent were monitored continuously. The entire
effluent from the thermal destruction system was passed through an XAD-2 trap to col-
lect PCDFs and other semivolatile organics. The XAD-2 trap and a rinse of connective
tubing were Soxhlet extracted. Extracts were cleaned using column chromatography to
isolate the PCDFs and PCDDs. All samples were analyzed for PCDFs using HRGC/EIMS in
the selected ion monitoring mode. The results of this work indicate that the optimum
conditions for PCDF formation from PCBs are near 675°C for 0.8 s or longer, with 8% ex-
cess oxygen. Under these conditions, percent levels of PCDFs are formed from mineral
oil or silicone oil contaminated with PCBs at 5 ppm or greater. PCDFs and PCDDs are
also formed from a trichlorobenzene dielectric fluid which contained no detectable
PCBs.
17.
KEY WORDS AND DOCUMENT ANALYSIS
DESCRIPTORS
' b.IDENTIFIERS/OPEN ENDED TERMS
c. COSATI Field/Group
PCBs Polychlorinated biphenyl
PCDF Polychlorinated dibenzofuran
PCDD Polychlorinated dibenzo-p_-dioxin
TCDD Tetrachloro dibenzo-p_-dioxin
TCDF Tetrachloro dibenzofuran
Combustion Transformer
P.yrolysis PCB fires
18. DISTRIBUTION STATEMENT
UNLIMITED
19. SECURITY CLASS (This Report)
UNCLASSIFIED
21. NO. OF PAGES
98
20. SECURITY CLASS (Thtspage)
UNCLASSIFIED
22. PRICE
EPA Form 2220-1 (Rev. 4-77) PREVIOUS EDITION is OBSOLETE
------- |