very high as well. It seems that the formation of the lower
CBs are controlled by different parameters (mechanisms)
than the higher CBs and/ or that they can act as precursors
for the formation of higher CBs. As can be seen in the loading
plot, the CPh formation, particularly the di- and triCPh
homologues, is distinctly influenced by the amount of air in
the system. The lowest formation of CPhs is found in
experiments 3, 5, 7, and 9 which all were performed at
relatively inefficient combustion. Experiments 4 and 8,
performed with the highest combustion efficiency, have
significantly higher levels of CPhs formed than the other
experiments. Consequently, the two variables di- and triCPh
are placed in the lower right corner of the loading plot, very
close to the O2 variables and opposite the CO variables. The
levels of tetra- to octaCDD/ Fs, tetra- to pentaCPhs, and penta-
to hexaCBzs are generally higher in samples taken from
experiments 7-11. All these experiments were performed at
different total amounts of air but all had high preheated
secondary airflow, i.e. performed at efficient secondary
oxidization. These homologues are placed in the upper right
corner of the loading plot. This shows that the total amount
of air is not the only important parameter for the formation
of these higher chlorinated homologues. How the air is
distributed between the primary and secondary is also
important, i.e. at efficient secondary oxidization the formation
increase. The importance of efficient secondary oxidization
in reducing the amount of formed OMPs has been shown in
other studies (3, 31). A difference between these studies and
our study is that their samples are taken after the ESP or after
the boiler at a temperature around 200-300 °C, i.e. after
both primary and secondary formation, while our study only
is discussing the primary formation at high temperature.
Oxygen is essential in the formation process of chlorinated
OMPs in three ways: the chlorination reactions through
Deacon mechanism, the oxidative breakdown of carbon on
the surface of the fly ash (de Novo synthesis), and in the
chlorination reaction of aromatics by CuCl2 (32-37). During
incomplete combustion at high temperatures, most of the
reactions involve H, OH, O, and O2H radicals. Cui et al. (38)
suggested that chlorination reactions are unfavored when H
atoms are present in the combustion gas, and Hagenmaier
et al. (14) suggest that fly ash catalyze dechlorination/ hydro-
genation reactions if the reactions occur during starved air
conditions. Their results also indicate that the CDD/ F forma-
tion increased during oxygen surplus conditions, a correlation
that first was described by Vogg and Stieglitz (15). The present
of H-atoms and the low levels of oxygen in the flue gas can
be two of the reasons for the low formation of higher
chlorinated homologues during the experiments performed
at inefficient combustion conditions. An explanation of the
high formation at high secondary oxidization conditions could
be that the Deacon reaction is enhanced by the high oxygen
level in the flue gas. The only conditions in which the higher
chlorinated OMPs (Cl > 4) can be formed or chlorinated is
during efficient secondary oxidization conditions due to the
oxygen-demanding chlorination reactions. These results
show that there is a complex correlation between the
oxidization reactions and the formation of chlorinated OMPs.
TABLE 4. Dominating CDD/F Isomers in Each Homologue
Presented in Percentage of the Homologuea
homologue
CDDs
%
CDFs
%
m ono
di
tri
2-
40 2-
no particular
20 [234-,238-,237-] 30
60
30
[23-,28-]
138-
tetra
penta
hexa
hepta
1368-
[12468-,12479-]
[124679-,123468-,124689-] 20 123467-
50/50 of the two HpCDDs 40 1234678-
15 [1234-,2349-]
15 23467-
40
50
70
a
[-]: coeluating isom ers. The m ono-, di-, and triCDD are only form ed
in experim ent 8.
of almost all congeners show that the dibenzofuran is formed
through non- or low-chlorinated precursors, followed by
further chlorination to form the CDFs. A more selective
congener formation would have been expected if the CDFs
were formed from chlorinated precursors. Chlorination
experiments of DF and/ or DD have been performed by Luijk
et al. (18) and Addink et al. (39) in microscale quartz reactor
systems. These experiments show that majority of the CDDs/
CDFs were present as octaCDD/ octaCDF and that the lateral
positions (2, 3, 7, 8) were preferred for chlorination, i.e. the
law of electrophilic aromatic substitution was followed. Such
a CDD/ F pattern is normally not found in flue gas samples
from normal MSW combustion processes and so is also the
case in our study. A reason for the disagreement in the
congener pattern between microscale laboratory results, and
real flue gas samples can be the distinction between the
conditions in the systems. Many reactions other than the
specific chlorination reactions ruled by electrophilic aromatic
substitution are very likely to occur in complex combustion
systems. The congener pattern and homologue profile in
flue gas samples taken from different MSW incinerators is
quite constant, independent of variations in combustion
conditions. The stable pattern implies that the formation
could be controlled by thermodynamic properties. The
2,3,7,8-substituted CDD congeners have been shown to be
more thermodynamically stable than the 1,4,6,9-substituted
(40-42). However, the thermodynamically controlled model
is not able to entirely explain the stable extensive congener
pattern in the flue gas from MSW incineration. Therefore,
contributions of other mechanisms had to be included in
the formation. The relative activity, i.e. the differences in
frontier orbital (HOMO and LUMO) for all CDD congeners,
was calculated by Wehrmeier et al. (43). The 2,3,7,8-
substituted congeners were shown to be more sensitive for
oxidative/ reductive breakdown than the other congeners,
and the most stable congeners were the 1,4,6,9-substituted.
Considering this unstability of the most thermodynamic
stable congeners, the calculated most probable congener
pattern according to the law of thermodynamic will shift. A
pattern where the congeners with chlorine substituted in
position 1,4,6,9 has a larger importance. A combination of
chlorination/ dechlorination (formation/ destruction) reac-
tions of the CDF congeners may explain the extensive
congener distribution found in our experiments.
3.3. Congener Distribution of the CDD and CDF in the
Flue Gas. The congener patterns and homologue profiles
were closely evaluated to enhance the understanding of any
similarities or dissimilarities in the CDD and CDF formations.
The levels of each homologue group changed due to
differences in combustion conditions, while the congener
distribution patterns were more or less the same, independent
of variations in the combustion conditions. Table 4 lists the
dominating isomers and their percentage part in each
homologue group. Almost all of the 135 congeners of the
mono- to octaCDF are formed during primary formation,
except a few di- and triCDF isomers. The extensive formation
The similarity in substitution pattern of the tetra- to
heptaisomers in our data shows that the formation mech-
anism can be the same, formed by the same precursor or by
the same chlorination reactions of the lower chlorinated CDF
homologues. Only half of the 75 CDD congeners are formed
during primary CDD formation. The extensive congener
formation found for the CDFs is not noticed. The 1,3,6,8-
and 1,3,7,9-TCDD isomers constitute for 50 and 20%,
respectively, of the TCDD pattern. These two TCDD isomers
are well-known as products from condensation of two 2,4,6-
triCPh. The same can be noticed within the PeCDDs,
approximately 50% origin from the two coeluating isomers
9
VOL. 33, NO. 23, 1999 / ENVIRONMENTAL SCIENCE & TECHNOLOGY 4 2 6 7