United States
Environmental Protection
Agency
Atmospheric Research and
Exposure Assessment Laboratory
Research Triangle Park NC 27711
Research and Development
EPA7600/S3-90/052 Aug. 1990
&EPA Project Summary
A Chamber and Modeling
Study to Assess the
Photochemistry of Formaldehyde
H.E. Jeffries, K.G. Sexton, J.R. Arnold, Y. Bai, J.L. Li, and R. Grouse
A new analytical method for
formaldehyde (HCHO) was implemen-
ted for use in the UNC outdoor smog
chamber, and HCHO measurements
obtained with this method were
compared with those obtained using
other analytical techniques. Six dif-
ferent calibration standards for HCHO
were found to agree within ± 2%, and
the different HCHO analytical
methods had precisions of ±10%.
New experiments in which HCHO was
an initial reactant and in which HCHO
was produced chemically were
performed. The older and newer
HCHO analytical methods agreed to
within 10% in these experiments.
A very explicit photochemical
reaction mechanism for ethene and
propene was formulated to explain
the chamber observations. The
ethene mechanism showed excellent
agreement with chamber observa-
tions; the propene mechanism, how-
ever, did not perform as satisfactory.
A comparison of these explicit
mechanisms with the Carbon Bond IV
(CB4) mechanism, which is used in
several EPA air quality simulation
models, showed excellent agreement
for ozone (<10% error), nitrogen ox-
ides, and hydrocarbon oxidation
rates; the CB4, however, consistently
underpredicted the HCHO maximum
by about 13%.
An analysis of a simulation of an
urban scenario showed that chemical
production of HCHO was the dom-
inant factor governing afternoon
HCHO concentrations. Ethene and
other olefins were the source of 58-
62% of the HCHO produced and
aromatics were responsible for 10-
12%.
It was concluded that the CB4
mechanism can be used to predict
ambient HCHO levels with an error of
about 20%.
This Project Summary was
developed by EPA's Atmospheric
Research and Exposure Assessment
Laboratory, Research Triangle Park,
NC, to announce key findings of the
research project that is fully
documented in a separate report of
the same title (see Project Report
ordering information at back).
Introduction
Because of a growing concern about
the health risks of exposure to
formaldehyde, the U.S. EPA is
considering regulating HCHO ambient
levels. An initial exposure assessment
conducted by EPA concluded that 80-
90% of ambient HCHO is produced by
the photochemical oxidation of
hydrocarbon (HC) species. Thus, photo-
chemical reaction models will have to be
used to relate emissions and emissions
controls to ambient levels of HCHO.
Current photochemical models accepted
by EPA for use in oxidant models were
not specifically developed to predict
HCHO. In fact, the model developers had
reservations about the HCHO chamber
data and, therefore, only made limited
comparisons between predicted and
observed HCHO. In particular, it had
been suggested by model developers
that the University of North Carolina
(UNC) HCHO chamber measurements
may have suffered from positive
interferences and thus could be too high
-------
by almost a factor of two. HCHO data
obtained in a chamber facility operated
by the Statewide Air Pollution Research
Center (SAPRC) were lower (and thus
more consistent with model predictions)
than that from UNC, but the SAPRC data
were more variable. Thus the accuracy of
HCHO predictions obtained with existing
photochemical models was uncertain
mainly because of a lack of believable
measurements of HCHO in the chamber
data used to develop and test models.
The work described in this summary
was carried out to determine if existing
EPA photochemical reaction mechanisms
can be used to predict accurate ambient
levels of HCHO. The approach taken was
to a) implement and test newer and more
reliable HCHO analytical methods and
calibration sources in the UNC chamber,
b) conduct new chamber experiments
with the newer and older HCHO analytical
methodology, c) compare new results
with those in the existing 12-year data
base used to develop the EPA models, d)
perform modeling exercises to diagnose
problems in data and model formulations,
and e) examine HCHO predictions in light
of the new information.
Formaldehyde Measurements
A new diffusion-scrubber, fluorescence
detector HCHO monitor (Dasgupta
Monitor) was implemented and its
performance was compared to seven
other techniques for measuring HCHO.
Several conclusions were drawn from this
work:
• All formaldehyde sources used as
calibration standards agreed to within
±2%.
• The paraformaldehyde permeation
tube was shown to be a very accurate
and stable calibration standard for
HCHO.
• Significant interferences did not
appear in any test of either the CEA
555 HCHO monitor (in use for 11
years at UNC) or the Dasgupta HCHO
instrument (in use for 3 years at UNC).
• There is weak evidence that the CEA
response, when not zero-checked for
twenty minutes every three hours,
may exhibit a positive drift; data
collected during such periods may not
be correctable. When zero-checked
regularly, the CEA response does not
show any significant bias.
• There is no evidence that the
Dasgupta HCHO instrument exhibits
more than a 5% bias at any time.
Several aspects of this particular wet-
chemical fluorescence technique, how-
ever, can make some individual data
points more uncertain than others.
When the Dasgupta HCHO instrument
is operated in a well-maintained mode,
it exhibits a precision of ~6 ppb at a
level of -155 ppb (3.8%).
• Despite concerted and on-going
efforts at QA and calibration,
uncertainty persists regarding some of
the HCHO data taken over the past 12
years in the UNC chambers. Errors in
recording and processing the data are
always possible, and calibration
standards have not always been as
available and reliable as they are
today. Thus, there are some HCHO
data in the UNC chamber database
that cannot be shown to be wrong, but
which are contrary to expectations
given our current understanding of
HCHO chemistry.
• Based upon the comparison of the
Dasgupta monitor with the
chromotropic acid bubbler method and
the FTIR method used at SAPRC, it
appears that the HCHO standards
used at SAPRC are accurate and
consistent with the UNC and EPA
standards to ±2%.
• The SAPRC chromotropic acid
bubbler method appears to be erratic
at high, reasonably steady levels of
HCHO. It also can have a low bias for
all measurements taken on a given
day, even if it had no bias the day
before. The bubbler method also
never measured higher than the
Dasgupta method in seven
experiments.
• The newer and older instruments
showed excellent agreement for a
wide range of chamber HCHO levels.
Thus we have strong evidence that the
UNC HCHO chamber data are
accurate. Therefore, any disagreement
between model and chamber
observations is probably due to
inaccurate representations in the
models.
Explicit Photochemical
Mechanism
An explicit photochemical reaction
mechanism was formulated using existing
kinetics data reviews and the literature.
The mechanism includes explicit chem-
istry for methane, formaldehyde, ethane,
acetaldehyde, organic nitrates and
peroxyacetylmtrate, ethene, glycoalde-
hyde, glyoxal, propene, and detailed
radical representations. Particular atten-
tion was paid to the state of knowledge
for the fate of the Criegee biradicals as
review suggested that this was the most
ill-defined aspect of olefin chemistry in
the existing EPA mechanisms. The
explicit mechanism was used to simulate
a) ozone-ethene kinetics experiments
conducted by several investigators, b)
nighttime ozone-ethene and ozone-
propene experiments conducted in the
UNC chamber, and c) sunlight irradiated
experiments containing oxides of nitrogen
(NOX) and CO, HCHO, acetaldehyde.
ethene, or propene that were also
conducted in the UNC chamber. In
almost all cases except propene, the
agreement between the simulations and
the observations was excellent for NOX,
03, HCs, and HCHO. For propene, the
theory and the sparse kinetics
information for the two-carbon Criegee
biradical were in substantial
disagreement with the chamber results
for NOx-propene experiments. Some
modelers made ad hoc adjustments to
the propene-ozone mechanism to allow
better simulation of chamber data and in
doing so may have compromised HCHO
predictions. Several conclusions were
reached based on the simulation tests:
• Our chamber simulations support the
higher absorption cross-sections of
Moortgat compared to Bass.
• We present indirect and nonconclusive
evidence that hydroxy-ethyl nitrate is
formed through reaction of NO with
the ethene-OH adduct to an extent of
4%.
• The yields given in the literature for
stabilized Criegee biradicals for O3 +
olefin reactions are based on the loss
of either 03 or the olefin, both of which
have other consumption pathways in
the experimental system. Therefore,
when these yields are used in a
mechanism, they should be corrected
to reflect these other losses.
• The reactions of the CT Criegee
biradical appear to be reasonably well
understood, with the exception of its
reaction with water. We presented
indirect and inconclusive evidence that
a large fraction of the reaction of this
biradical with water could lead to
HCHO and H202- If our assumed
pathway is correct, there exists a large
potential for forming H202 by dark
reactions.
• The chemistry of glycoaldehyde may
be sufficiently different from that of
acetaldehyde that the CB4 approx-
imation of representing glycoaldehyde
as acetaldehyde is questionable.
• New chamber experiments using
glycoaldehyde as the primary reactant
are needed for continued model
testing. Chamber experiments emp-
loying glyoxal as the primary reactant
would also be beneficial.
-------
• A technique for the reliable, real-time
measurement of formic acid in ethene-
NOX chamber experiments should be
developed.
• More high concentration ethene-NOx
chamber experiments should be
conducted on cool days to investigate
the anomalously high HCHO formation
observed in the cool-weather exper-
iments carried out in this study.
• Although the model fits to the very
wide range of experiments simulated
here were excellent, we believe that
further work is needed on the UNC
chamber wall model.
• The excellent "goodness-of-fit" of the
explicit ethene mechanism to a large
range of conditions in the UNC
chamber over a 12-year period is
strong evidence that the mechanism
does explain the observed chemistry
in a manner consistent with known
kinetics data. Although there remain
some doubtful elements in the ethene
mechanism, it is at present the best
explanation we can construct, and we
believe that it can be taken as a
standard against which to measure the
performance of the ethene portion of
air quality simulation mechanisms in
common use by EPA.
Performance of the CB4
Mechanism
The Carbon Bond Four mechanism
was used to simulate several ethene and
propene chamber experiments and the
results were compared to those obtained
with the explicit mechanism. These tests
of CB4 used a) HCHO photolysis rates
derived from the cross-sections reported
by Moortgat, b) the newest UNC chamber
characterization model, and c) the latest
information on in-chamber actinic fluxes.
The conclusions reached from analyzing
these simulations were:
• With regard to predicting 03, the CB4
mechanism is excellent for ethene-
NOx experiments. It is slightly worse
than the explicit mechanism for NO
and NOX, but not significantly so.
• At all levels of ethene tested, the CB4
consistently underpredicted the HCHO
maxima observed in ethene-NOx ex-
periments by about 13%. The cause
of this underprediction is complex and
related to a number of approximations
used in the CB4:
a) There are too few radical-radical
termination reactions in the CB4
for it to simulate the correct
efficiency of HO2 production
from the ethene + OH reaction;
it therefore predicts too much
ethene consumption and too
much production of HO2 and
N02.
b) Acetaldehyde is not a good
substitute for glycoaldehyde in
these high concentration exper-
iments.
c) Certain important pathways in
the Criegee biradical reactions
have been omitted in the CB4.
• Like the explicit propene mechanism
described in this work, the CB4
mechanism does poorly in simulation
propene-NOx experiments with initial
propene concentrations above ~- 1.5
ppmC.
• Previous testing of the CB4 with
toluene and xylene experiments from
the UNC chamber showed that the
CB4 overpredicts HCHO. Confirming
toluene and xylene experiments were
conducted in this study and these
experiments show levels of HCHO
similar to those obtained in the
previous experiments. Thus, we con-
clude that the CB4 mechanism over-
predicts HCHO yields from toluene
and xylene by about a factor of two.
The observed yields of HCHO are low,
however, amounting to only about 3%
of the aromatic carbon.
Urban Simulations
A time-resolved mass balance method
was used to analyze predictions of HCHO
obtained in a simulation of a typical urban
scenario. This analysis showed that:
• Chemical production of HCHO ac-
counts for 70-90% of the total HCHO
observed. A "pseudo-photostationary
state" is reached at approximately the
same level regardless of whether or
not initial HCHO and HCHO emissions
are included in the simulation.
• Large early morning dilution rapidly
reduces any HCHO initially present
and most of the initial HCHO is gone
by 1300 LOT (maximum 03 occurred
at 1900 LOT).
• Over the day, the olefins and
acetaldehyde were responsible for ap-
proximately 60% of the HCHO that
was produced. The non-ethene olefins
made their contribution in the morning,
ethylene contributed all day, and
acetaldehyde increased its contribu-
tion to HCHO production near the end
of the day.
• An analysis of the CB4 mechanism
and the UNC chamber data suggests
that the CB4 is able to predict HCHO
formation in urban HC mixtures fairly
accurately. It appears, however, that
the CB4 underpredicts HCHO forma-
tion from olefinic HCs and over-
predicts HCHO formation from
aromatic HCs. Thus, the good
agreement appears to be due to
compensating errors.
Conclusions
The purpose of this study was to
determine if present photochemical
reaction mechanisms can be used to
predict accurate ambient levels of HCHO
for various scenarios of HC and NOX. We
believe that this is probably the case for
the most commonly used EPA mech-
anism, the Carbon Bond Four mechanism
(CB4). Some parts of the CB4 (and other
similar mechanism), however, are not
fully accurate descriptions of the impor-
tant HCHO producing chemistry. The
errors are probably small, only resulting
in approximately a 20% error in max-
imum HCHO predictions. The magnitude
of the error depends upon the
composition of the HC mixture used in
the simulation because there are
compensating errors in different HC
classes. A second weakness that CB4
has for this application is that it has a
very compressed representation for the
most important species producing HCHO,
the various olefin classes. For example,
acetaldehyde is used in the CB4 to
represent internal olefins. The accuracy
of the CB4 for predicting HCHO formation
could be improved by slightly expanding
the representation of olefins in the
mechanism.
.S. GOVERNMENT PRINTING OFFICE: 1990/748-012/20063
-------
H.E. Jeffries, K.G. Sexton, J.R. Arnold, Y. Bai, J.L Li, and R. Grouse are with the
University of North Carolina, Chapel Hill, NC 27514
Marc/a C. Dodge is the EPA Project Officer (see below).
The complete report, entitled "A Chamber and Modeling Study to -Assess the
Photochemistry of Formaldehyde," (Order No. PB 90-240 581/AS; Cost:
$39.00, subject to change) will be available only from:
National Technical Information Service
5285 Port Royal Road
Springfield, VA22161
Telephone: 703-487-4650
The EPA Project Officer can be contacted at:
Atmospheric Research and Exposure Assessment Laboratory
U.S. Environmental Protection Agency
Research Triangle Park, NC 27711
United States Center for Environmental Research
Environmental Protection Information
Agency Cincinnati OH 45268
Official Business
Penalty for Private Use $300
EPA/600/S3-90-052
------- |