U.S. Environmental P.-otectiort Agency Industrial Environmental Research EPA~600/7~77~073d
Office of Research and Development Laboratory . . -tQ-r-r
Research Triangle Park, North Carolina 27711 JUly tSI f
PROCEEDINGS OF THE SECOND
STATIONARY SOURCE
COMBUSTION SYMPOSIUM
Volume IV. Fundamental
Combustion Research
Interagency
Energy-Environment
Research and Development
Program Report
X.
U.S.
-------
RESEARCH REPORTING SERIES
Research reports of the Office of Research and Development, U.S.
Environmental Protection Agency, have been grouped into seven series.
These seven broad categories were established to facilitate further
development and application of environmental technology. Elimination
of traditional grouping was consciously planned to foster technology
transfer and a maximum interface in related fields. The seven series
are:
1. Environmental Health Effects Research
2. Environmental Protection Technology
3. Ecological Research
4. Environmental Monitoring
5. Socioeconomic Environmental Studies
6. Scientific and Technical Assessment Reports (STAR)
7. Interagency Energy-Environment Research and Development
This report has been assigned to the INTERAGENCY ENERGY-ENVIRONMENT
RESEARCH AND DEVELOPMENT series. Reports in this series result from
the effort funded under the 17-agehcy Federal Energy/Environment
Research and Development Program. These studies relate to EPA's
mission to protect the public health and welfare from adverse effects
of pollutants associated with energy systems. The goal of the Program
is to assure the rapid development of domestic energy supplies in an
'environmentallycompatible manner by providing the necessary
environmental data and control technology. Investigations include
analyses of the transport of energy-related pollutants and their health
and ecological effects; assessments of, and development of, control
technologies for energy systems; and integrated assessments of a wide
range of energy-related environmental issues.
REVIEW NOTICE
This report has been reviewed by the participating Federal
Agencies, and approved for publication. Approval does not
signify that the contents necessarily reflect the views and
policies of the Government, nor does mention of trade names
or commercial products constitute endorsement or recommen-
dation for use.
This document is available to the public through the National Technical
Information Service, Springfield, Virginia 22161.
-------
EPA-600/7-77-073d
July 1977
PROCEEDINGS OF THE SECOND
STATIONARY SOURCE
COMBUSTION SYMPOSIUM
Volume IV. Fundamental
Combustion Research
fc
M
Symposium Chairman Joshua S. Bowen
Vice-Chairman Robert E. Hall
Environmental Protection Agency
Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, North Carolina 27711
Program Element No. EHE624
A
u, s. :...- v . .
Prepared for
U.S. ENVIRONMENTAL PROTECTION AGENCY
Office of Research and Development
Washington, D.C. 20460
-------
PREFACE
These proceedings document the more than 50 presentations and discus-
sions of the Second Symposium on Stationary Source Combustion held August 29
- September 1, 1977 at the Marriott Hotel in New Orleans, Louisiana.
Sponsored by the Combustion Research Branch of the EPA's Industrial
Environmental Research Laboratory-Research Triangle Park, the symposium
presented the results of recent research in the areas of combustion
processes, fuel properties, burner and furnace design, combustion
modification, and emission control technology.
Dr. Joshua' S. Bowen, Chief, Combustion Research Branch, was Symposium
Chairman; Robert E. Hall, Combustion Research Branch, was Symposium Vice-
Chairman and Project Officer. The Welcoming Address was delivered by Dr.
John K. Burchard, Director of IERL-RTP; the Opening Address was delivered by
Robert P. Hangebrauck, Director, Energy Assessment and Control Division,
IERL-RTP; and Dr. Howard B. Mason, Program Manager NOX Environmental
Assessment Program, Acurex Corporation, delivered the Keynote Paper.
The symposium consisted of six sessions:
Session I: Small Industrial, Commercial and Residential Systems
Robert E. Hall, Session Chairman
Session II: Utility and Large Industrial Boilers
David G. Lachapelle, Session Chairman
Session III: Special Topics
David G. Lachapelle, Session Chairman
Session IV: Stationary Engine and Industrial Process Combustion
Systems
John H. Wasser, Session Chairman
Session V: Advanced Processes
G. Blair Martin, Session Chairman
Session VI: Fundamental Combustion Research
W. Steven Lanier, Session Chairman
-------
TABLE OF CONTENTS
- SESSION VI: FUNDAMENTAL COMBUSTION RESEARCH -
Page
"Fundamental Combustion Research Applied to Pollution Control,"
T. J. Tyson, M. P. Heap
"Chemical Reactions in the Conversion of Fuel Nitrogen to N0x:
Fuel Pyrolysis Studies," A. E. Axworthy, V. H. Dayan ....
"Fate of Fuel Nitrogen During Pyrolysis and Oxidation," Y. H. Song,
J. M. Beer, A. F. Sarofim
"Interactions Between Sulfur Oxides and Nitrogen Oxides in Combus-
tion Processes," J. 0. L. Wendt, T. L. Corley, J. T. Morcomb . . .
"Chemical Reactions in the Conversion of Fuel Nitrogen to NOX:
Low-Pressure Flat-Flame Burner Studies," D. R. Kahn, A. E.
Axworthy
"Formation of Soot and Polycyclic Aromatic Hydrocarbons in Combus-
tion Systems Development of a Molecular Beam Mass Spectrometer,"
J. 0. Bittner
"Investigation of NOX, Nitrate and Sulfate Production in Laboratory
Flames," D. 0. Seery, M. F. Zabielski, L. 6. Dodge
"Influence of Aerodynamic Phenomena on Pollutant Formation in Com-
bustion: Phase II Liquid Fuels," L. J. Spadaccini, J. B. McVey,
J. B. Kennedy, F. K. Owen, C. T. Bowman, A. Vranos, A. S.
Kesten
"Two-Dimensional or Axially Symmetric Modeling of Combusting Flow,"
H. McDonald, R. C. Buggeln
"PRemixed One-Dimensional Flame (PROF) Code Development and Appli-
cation," J. T. Kelly, R. M. Kendall
39
79
101
139
183
209
235
283
311
iii
-------
-------
SESSION VI:
FUNDAMENTAL COMBUSTION RESEARCH
W. STEVEN LANIER
CHAIRMAN
-------
-------
FUNDAMENTAL COMBUSTION RESEARCH APPLIED
TO POLLUTION CONTROL
An Overview
By:
T. J. Tyson and M. P. Heap
Energy and Environmental Research Corporation
Santa Ana, California 92705
-------
-------
ABSTRACT
The development of NO control techniques by modification of the
X
combustion process in a timely and cost-effective manner is enhanced if
there is a basic understanding of the phenomena associated with NO forma-
-------
-------
SECTION 1
INTRODUCTION
Tn February 1977 the U.S. linvt ronment.-il Protection Agency n war clod a
70,000 man-hour le-vel-of-ef fort contract to the Energy ;md linvi ronmental
Research Corporation to assume the major responsibility for managing .'ind
conducting its fundamental combustion research efforts. The "Fundamental
Combustion Research Applied to Pollution Control" (FCR) program objectives
are to provide a focused program of basic engineering research designed to:
Provide the EPA/CRB technology development programs with the
necessary understanding of basic combustion behavior required
to achieve the minimization of NO from stationary sources;
Provide guidance and mathematical tools required to effectively
utilize this body of information in the development of NO
X
control techniques; and
Assure that critical information will be generated within a
time frame consistent with the technology development programs.
The words "focused" and "engineering" research are directed at the important
issue of relevance. The program will be focused on well-defined priority
target areas and the research effort will be directed towards engineering
solutions to specific problems. Developing the "necessary understanding"
requires that relevant issues be isolated .so that the program can be dis-
engaged from studies of irrelevant or nonproblem aspects of the physics and
chemistry of combustion.
Sfveral Taotorn ni'cd coiusiclur.'it l.on (f the Information generated by tliu
FCR program Is to he nt.ll.lxml offtsc.t Ivc'ly. Primary amongst these is the
provision of u strong 1 fnk.-ijjc between ihc understand ing of basic, behavior
and tin.' control of NO Jn practical combust f on devices. A unification and
x
genera I.iz.ition of the laboratory observations La required if this linkage
-------
is to be achieved. Mathematical modeling is an important component of the '
FCR program as it provides a formalism whereby this generalization and
linkage can be developed. Modeling allows a systematic exploration of the
implications of the basic understanding with regard to NO control.
X
Effective utilization also requires effective transfer of FCR-generated
technology into the CRB Technology Development (TD) programs. Thus, close
coupling of the FCR and TD programs is an important factor in the develop-
ment of FCR.
The final objective listed above relates to the important question of
when FCR results must be attained. Many judgement factors enter during the
resolution of this question. Obviously inputs to specific CRB Technology
Development programs must occur at a time consistent with the schedules of
these programs. In general this implies that FCR results must be generated
and transferred within a two-co-four year time period. The demands imposed
by such a short time frame strongly influences the nature of the research
to be undertaken.
To date the majority of the effort expended under FCR has centered
around the detailed planning of the program elements and in preparation of
work scopes for upcoming subcontracts. It is the Intent of this paper to
present the results of this initial planning process in the hope that by
describing the constraints around which these planning activities proceeded,
the potential subcontractors will obtain a better feeling for the thrust of the
FCR program, and how they might contribute to fulfilling its objectives.
The motivation for the FCR program stems directly from EPA's stated goal
to establish the technology base necessary for maximum reduction of NO emis-
3v
sions from stationary sources. In the course of generating an overall EPA
strategy to achieve maximum NO reduction it has become evident how difficult
it Is to exploit or transfer a successful control technique developed under
one condition to other situations. The lack of a mechanistic understanding
of the dominant phenomena prevents a generalization of the observations made
during the course of a technology development program. Application of the
!
.results to other scales, fuels, device types, and operating regions often '
becomes impossible.
8
-------
7'he use of detailed measurements obtained from pilot scale combuscors
during the development period to provide the required understanding of con-
trolling mechanisms is a sometimes successful but is more often an unreward-
ing task. The difficulty lies in the complexity of coupling between domi-
nant mechanisms within a practical combustion device. Controlling mechanisms
are not easily identified or characterized under such conditions.
The above factors make the evolution of successful NO control techn.i-
x
ques exclusively through pilot scale development a costly and lengthy trial
and error process. Taken together they provide strong motivation for an F?CK
program which focuses on defining the dominant mechanics and providing process
models which will allow generalization and exploitation of pilot sca.le results.
Further stimulus for a substantial I'CK element within the overall EPA program
is the recognition that through fundamental understanding it is possible to
systematically establish the lower bounds on NO emissions that are theo-
X
retically attainable within the engineering constraints imposed on a parti-
cular device. A knowledge of these lower bounds will provide a much needed
yardstick with which to judge the development of low emissions systems and
to guide their further development.
The organization of the program is shown in Figure 1. As noted in the
middle box planning, review, and redirection functions are carried out by HER
staff and staff consultants in conjunction with the EPA Project Officer. The
organization is shown in throe distinct tiers. The upper tier indicates the
various groups outside of EER will provide Input to the planning function
of the FCR program. Potential subcontractors (their tier) arc encouraged to
provide critical comment on the FCR program as well as to make available
information on their own capabilities and areas in which they are interested
in performing research under FCR. The CRB staff, of course, provides the
necessary insight into the various technology development programs and is
the primary recipient of technology developed under FCR. Coordination will
be sought between KCR activities and rcilatod ocmbustion research activities
-------
within the DOD and DOE, the energy industry,, and university laboratories.
In addition to EER's staff considerable use of consultants will be made
who have recognized expertise in areas of critical importance to the pro-
gram. In certain circumstances these consultants will also be used as
proposal reviewers.
Although not shown in Figure 1, the MIT Energy Laboratory has a special
role in the program that distinguishes it from other subcontractors. The
Energy Laboratory will be under contract during the entire course of the FCR
program providing EER with support in the following areas:
Program Planning and Review
Program Management
Technology Evaluation
Coordination with other CRB Fundamental Research Activities
Engineering Analysis
» Exploratory Laboratory Investigations to Provide Definition
and Guidance of Subcontractor Tasks
* Technology Transfer
The fourth item above is in recognition of the important fact that several
of the CRB Technology Development programs have a fundamental research
component. In addition the CRB sponsors grant type programs at a number of
universities. Hence participation in the coordination of these activities
becomes an important FCR function. Engineering analysis is a significant
part of the EER in-house effort. This will include development and applica-
tion of data analysts techniques, codes designed for numerical experiments,
modeling of specific mechanistic: boliavJor, and modular modeling which links
together relatively simple models In such a manner .IK to allow analysis of
practical combustion systems.
10
-------
SECTION 2
PI..ANNTNC; STUATIW;Y
Having established tin: general objectives of Lhe I-'CR program (L i.s
necessary to make these concrete in terms of specific engineering research
tasks. The flow diagram shown in Figure 2 outlines the strategy through
which this will be achieved. This section discusses the general aspect of
the strategy outlined in Figure 2 and the anticipated products resulting
from the implimentation of such an approach. Presented later in this paper
are specifics regarding the initial planning effort and the definition of an
initial program of research which has resulted from this effort.
Information from a variety of sources including the CRB Technology
Development programs and the NO Environmental Assessment program have been
J\.
used to establish priorities by device type, size, and fuel, and the time
frame in which research results must be generated if they are to be effectively
utilized. These studies clearly show that near and intermediate-term require-
ments will necessarily lead to research focused on retrofit concepts in which
the combustion device type cannot undergo drastic changes in design concept.
Longer time frame objectives, on the other hand, will allow the research
effort to take on a more wide ranging character associated with the possi-
bilities of entirely new combustion system concepts. With these priorities
it is possible to define the general flame characteristics associated with
the combustion devices and fuels which have the highest priority.
At this point the strategy takes cm a more scientific character. Civen
flame types, what course of research should he pursued to yield Lhe necessary
understanding required to aehJevo KI'A's )>na I s for the reduction of NO cmls-
x
sinus? The1 answer to this question rests first, on tin.' identification of Lhe
sign.iric.nnt physical and chemical phunomona which dictate the behavior of the
flame type under conslderat ion. As defined, ".significant" Implies phenomena
whose alteration will yield a 1st order effect on NO or related pollutant
emissions.
11
-------
in order to establish Lhe "significance" of the various phenomena an
attempt must be made to answer a number of related questions.
To what degree are the phenomena controllable through changes in
device design parameters?
Can "cause and effect" history be established which links
significant phenomena and provides insight into control
methods?
When do the phenomena occur in the overall proc.ews (enrly,
late, intermediate, continuously)?
What is time constant of the phenomena in comparison to other
process time constants?
If the process rapidly equilibrates how does the equilibrated
behavior depend upon process boundary conditions?
If the process rapidly goes to slowly varying nonequilibrium
state how does that state depend on process boundary conditions?
To what extent are processes coupled?
Answers to these questions not only provide the necessary insight as to what
phenomena are important but also provide guidance as to the depth of under-
standing required for FCR purposes. For example, in examining processes
with short-time constants in comparison to other significant phenomena the
details of the process path are of little interest. Only the end-point
equilibrated behavior as it is affected by the process environment (boundary
conditions) is of concern. Consider the coal particle heat-up process. If
all other processes in the systems such as rapid devolatilization are slower
than this process then only the equilibrated final temperature of the
particle Is of interest. Under certnin circumstances processes which fall
into thin i-alL'jyjry might be: turbulence In};, particle b;ilJ .1st Irs, prompt
fuel nitrogen chomlsiry, and onrrgy rcluaHr kinetics.
A Hystc-'nmt. U- henrch for the nnsworw to these questions will, leml to
identification oT important gaps In current: understanding and will provide
guidance in the definition of experimental and theoretical tasks to close
these gaps.
12
-------
TECHNOLOGY ASSESSMENTS
In order to identify and characterize significant topics worthy of
FCR support there will be continual assessments of the current state-of-
understanding in the following areas:
Gas phase chemical kinetics and soot formation mechanics
Physics and chemistry of solid nnd liquid fuel combustion
Physics of transport phenomena
Measurement technology
Mathematical modeling
These assessments will form the basis of a series of special reports setting
forth the reasons for supporting particular research programs, and assessing
the value of the results of these programs. The intent is to critically
review past and present experimental and theoretical investigations and to
undertake simple limit-case analyses and numerical experiments which will
aid tho assessments. The primary motivation for the measurement ;jnd modeling
areas is to assure that adequate measurement nnd data analysis ler.hn i<|ues
are available for all experimental programs sponsored under the I''CK project,.
The product of the assessments, in addition to the identification of gaps,
will be to generalize and quantify the information in a form suitable for
use in process models.
EXPERIMENTAL AND THEORETICAL STUDIES
With the gaps in understanding of significant phenomena delineated, the
core of the FCR program becomes the set of experimental and theoretical
studies designed to fill these gaps. The. common thread running through these
programs arc the requirementw to:
ltat.nl> I Ish how ,-md wli;H b.iulc Information will be extracted from
exper Iment.'i ] d;it.;i or theoretical, study.
Show fil.rong linkage hc.'tween plieuoineno log Icn I behavior In experi-
ment.* or theoretic:;! I study and the behavior In prart ic.;i I com-
husl foil systems.
13
-------
Establish of fee 1.1 VIMICMH of iTK-aHiin'rixMil trchn \<\\H-H.
Defliu- how haute- Informal: Ion obtained from InvcHl: I>;,.-jl: ion will
be ust:d lo achieve ends of program.
ESTABLISHMENT OF LOWER BOUNDS ON NO EMISSIONS
A
A key element in the FCR strategy is the search for lower bounds set
on NO emissions under various limiting situations and constraints. For
example, is it possible to establish the lower bounds on NO set by the gas
X
phase chemistry for a particular type of device with its engineering con-
straints on heat transfer, exit temperature, size, etc.? Such a lower
bound study would assume idealized and optimum time-phasing for fuel/air
contacting and heat transfer under the imposed constraints. In a similar
manner Is it possible to establish the lower bounds on NO set by fundamental
constraints associated with solid nml liquid fuel combustion? The value oT
such studios would be substantial. Not only would they provide a yardstick
by which to measure the effectiveness of control schemes but they would pro-
vide strong guidance in the development of control techniques.
The search for lower bounds will encompass both physical and numerical
experiments. These experiments will investigate numerous limiting situations
involving
Controlled heat and mass transfer reactors
Sequential, parallel, and feedback staging
Fuel staging
Natural staging in turbulent diffusion flames,
THE ROLE OF MATHEMATICAL MODELING IN FCR STRATEGY
Mathematical, modeling will play a strong role in the FCR program in the
following wide ranging areas:
Data analysis
Limit-case studies
Modular modeling of complex systems
Relatively simple modeling of complex fluid mechanical behavior
14
-------
Data analysis capability will be required for all experimental programs to
extract the maximum amount of mechanistic information implied by the data.
For example, a pulverized coal-stirred reactor experiment would necessarily
be complimented with a two-phase stirred reactor code. Measurements of gas
and solid phase characteristics could then be used in conjunction with the
code to establish the detailed particle mechanics of devo.latili.zat ion and
fuel nitrogen evolution and speciation. Data analysis techniques will
include sensitivity studies and inverse methods. Sensitivity studies will
allow the determination of phenomenological parameters which have the
greatest influence on the observables. Inverse methods allow the data Co
be used as independent input variabl.es whi.J.e the mechanistic parameters
becomes the calculated dependent variables. The values deduced for the
mechanistic parameters would then be checked for plausibility.
The limit-case studies in which certain phenomena are assumed to dominate
the process while other phenomena become suppressed are particularly useful
in numerical experiments designed to examine the implications of specific
phenomenological behavior. The analysis might be as simple as that of the.
limiting situation of intense backmixing such that the system become per-
fectly stirred. On the other hand the analysis might be as complex as a
two-phase turbulent diffusion flame in the limiting case of no backmixing
or recirculation zones. In any event, a common feature of the limit-cases
under consideration is that they are all amenable to straightforward
numerical, analysis using well-known fast and accurate techniques. In con-
junction with experiments designed to meet the same limiting assumptions,
these codes become data analysis tools to extract basic mechanistic: behavior.
Once the imbedded physics and chemistry has been modified to be consistent
with observations in analogous physical experiments, the models can be
effectively used in nume.r'ic.'il. experiments and us fools in the search for
the Lower hounds on NO emission.
x
Modular modeling implies Llii'. modeling <>l ;i complex system ushij', :\
coll.e.i-l Ion of" lhnit-r.;i.se elements linked together by means of ump i r ic;i I
knowledge of the exchange of heat ;ind m.'tss between these elements. The
combustor is divided into xones whle.li li.'ive a distinct Limit-case character,
e.g., rectrcMi.1 ;it ion /ones, free shenr layers, co-axial turbulent diffusion
15
-------
zones, regions where fuel droplets are transported ballistically, etc. In
the simplest case the model might consist of a collection of stirred and
plug flow reactors connected in series. More complex cases could include
feedback (upstream influence) paths and diffusion flame elements. The zone
sizes and linkage between zones would in general be estimated through the
application of relatively simple fluid mechanical models In conjunction with
empirical observat Ions from Ihe specific device type, under examination.
Stimulus-response experiments would be, part leu larly useful In this regard.
In these technlejues inert and chemlcaJ1y active tracers are Injected at
various points within the system and their presence Is measured at other
points. From such measurements considerabie information on macro- and
and micro-scale mixing patterns, mass and heat transfer rates, and distribu-
tion of residence times can be obtained.
The motivation for semi-empirical modular modeling stems from our
inability to mechanistically describe and couple all of the phenomenon
present in a complex system in a single unified mathematical model. This
is primarily a consequence of two factors our poor understanding of turbu-
lent transport In complex systems and inadequate numerical techniques.
Neither of these conditions Is likely to change during the critical time
period of the FCR program.
Relatively simple fluid mechanical, models will he sought which can
describe in a gross fashion certain important fuel/air contacting charac-
teristics of flames. For example, the near field behavior of swirl
stabilized flames defies description through a complete mathematical solu-
tion of the conservation laws of physics and chemistry. However, a gross
fluid mechanical description may be possible which links influential system
parameters with characteristics such as shear layer scales and strengths,
recirculation zone sizes, ballistic behavior of fuel droplets, etc. If such
models can lend to a rough ability I o .scale observed phenomena then they
wil.l have served an Important fund'Ion. Insight to the scaling behavior of
complex .systems is ilie. single most. Important objective of predictive modeling.
16
-------
FCR PRODUCTS
Referring again to Figure 2 It is appropriate to summarize at this
point tlie products tlt.it will be gen cm Loci by the KCR program.
Identification and characterization of dominant phenomena which
significantly effect NO and related pollutant emissions.
X
Quantification and generalization of results through intuitive
or qualitative models, empirical correlations, mathematical
models for prediction and scaling.
Determination of degree to which critical mechanisms can be
controlled and insight as to how this control might be achieved.
Establishment of lower bounds on NO emissions set by various
x
physical and chemical factors.
Direct input to "technology development" programs through
review of results from fundamental viewpoint, guidance in
design of experiments, and new control concept suggestions.
In addition to the above outputs which related specifically to the objectives
of the EPA programs, the FCR program will generate information and tools of
general use in the development of combustion devices.
17
-------
SECTION 3
INITIAL PROGRAM 01' RESEARCH
This section reviews the initial effort in applying the planning
guidelines outlined in Section 2.
INITIAL PRIMARY OBJECTIVE
A study of stationary combustors indicates that flame types which
contribute significantly to the total NO generated by these devices can
be characterized as follows:
Dominant Flame Characteristic
Transport Is turbulent and
radiative
Fuel/air contacting is by
diffusion and particle
penetration
Flames are large - typically
occurring in industrial/
utility equipment with
characteristic dimension of
the energy release zone on
the order of feet to tens of
foet
Residual oil and coal are Lbi>
principal sources of em.IMS I OIKS
due to lilp.li ii.ltroRun conli-nt
and tho difficulties associ-
ated wllli s InuiLlaiuHHiH NOX and
paniculate control
Significant Exceptions
Flame propagation in spark-
ignited I.C. engines
Flames in domestic and small
commercial furnaces; gas
turbines
Alternate- fuels, natural gas
18
-------
The time-mean motion is in
steady-state
Pressure' is atmospheric
K are confined by cold
walls to which the flame
radiates
Reciprocating I.C. engine
I,C. Engines, supercharged
boilers
Process furnaces
The primary initial objective follows directly from these observations.
The FCR program will focus on the control of NOX and primary related pollu-
tants generated by large confined one-atmosphere turbulent diffusion flames
fueled with pulverized coal or residual oil. and radiating to cold walls.
Such flames account for 60 percent of all stationary NO emissions.
X
Two time frames must be considered for the application of control
technology to this type of equipment which have implications for the FCR
program. In the short-term, control strategy will be associated with
existing technology and will involve simple system changes such as burner
redesign. Thus results must be forthcoming in the two-to-four ytmr time
period if they are to have a significant impact upon technology development
programs. Results from this same period will also be applicable to long-
term NO control strategy since they will help to define the expected
X
limits of control that are achievable through the application of existing
technology and will form a basis for the development of long-term control
technology based upon the design of radical new combustion systems.
SECONDARY OBJECTIVES
A group of secondary objectives have also been identified and are
categorized by device/fuel type as Follows:
Large furnaces fired with co.-i 1-derived fuels
Stoki- r-f 1 red equipment
Comb i.m-d cycle and gas turbine comhustor fired with residual
oil ;mcl coal-derived fuels
C;it:nly t Ic combustors firing high nitrogen fuels
19
-------
Fluid mechanics and kinetics of selective NO reduction devices
x
Reciprocating internal combustion engines
Stoker-fired boilers receive prominence because they represent the most
probable use of coal for systems raising steam at 250,000 Ib/hr or less.
Combined cycles with their potential for 50 percent efficiency necessarily
demand attention. Selective NO reduction may be economically attractive in
X
applications where existing technology is Incapable of reaching the necessary
level of control, and hence, deserves attention especially with regard to
the fluid mechanics of adequately mixing the reducing agent and the NO -
X
laden combustion products.
Although these areas will not immediately receive direct FCR attention
much of the research focused on large turbulent diffusion flames is not
device-specific and is applicable to several of these categories. An under-
standing of pulverized coal combustion provides insight into suspension
phase burning in a spreader stoker. The chemical kinetics of fuel nitrogen
conversion is generally applicable.
PRELIMINARY TECHNOLOGY ASSESSMENTS
Preliminary technology assessments have been undertaken in order to
structure an initial program of research. Some of the conclusions of these
assessments are briefly presented below.
Gas Phase Kinetics
A review is currently in progress of the state-of-understanding of the
chemical kinetics of NO formation and destruction in flames. With regard
X
to fuel nitrogen chemistry, although considerable data and speculation exist
as to the pertinent chemistry, there is a clear need for more reactor experi-
ments. Questions need to be resolved regarding the existence of a common
stable Iiitcrmedlnte nitrogen compound regardless of the nature of the fuel
nitrogen. Does sm:h ;\ compound occur very early In the conversion process
and hence obvinte Liu- need to study the details of "prompt" chemistry? What.
dictates the extend of fuel nitrogen conversion to this stable .intermediate?
Can kinetics studies be concentrated upon the long time constant process by
which the stable intermediate is converted to NO or N .
-------
In examining existing flat flame and stirred reactor data two difficulties
arise. First the uncertainty in the measurements due to probe and molecular
beam skimmer of feels l.n flat flames and unmtxedness and temperature- instru-
mentation in stirred reactors can give rise to a substantial degree of
uncertainty in conclusions drawn from the data. Second, considerable diffi-
culty arises in analyzing the data. Kor example, in flat flames jf interest
extends to the behavior within the flame, front then a knowledge of low
temperature kinetics associated with the upstream portion of the flame may
be required to examine the high temperature kinetics of interest. Further,
if the degree of rapid conversion of fuel nitrogen to a stable intermediate
is process-dependent then the existence of a lean zone in an overall rich
flat flame makes it difficult to draw general conclusions regarding this
process in furnace-type flames. Stirred reactor data analysis requires
careful attention to the sensitivity of species concentrations to reaction
rates in order to extract basic mechanistic: information. Valid sensitivity
analyses are accurate only for .small variations in rate constants and hence,
the process of analysis becomes an iterative one.
There is a clear need for more well-instrumented reactor experiments
designed to cover a wide range of residence times, temperatures, stoichiom-
etries, fuel types and fuel nitrogen dopants. The effect of higher C/H ratio
fuels needs special attention since the speciation of hydrocarbon radicals as
a function of time may influence both the formation of nitrogen intermediates
as well as the long time conversion of these intermediates to NO or N?.
Extending these reactor experiments in a search for the lower bound on
NO emissions set by the chemistry needs to be done and will become an impor-
tant part of the FOR program. Such experiments will involve a variety of
staging conditions both sequential .-ind parallel to give control over the
time-phasing of mass and heat transfer. Additional ga.s phase kinetics areas
which are not well-understood and will receive- attention under !*'CR are:
SO /NO interact ion
x x
Soot format Ion
Soot/NO formation
X
21
-------
PHYSICS AND CHEMISTRY OF TWO-PHASE FLOW
The most apparent gap in experimental evidence concerning pollutant
formation is pulverized coal and heavy oil combustion is in the area of
laboratory reactors which simulate combustion conditions representative of
specific regions within furnace-type flames. Reactor experiments should be
designed such that the fluid mechanics of fuel/air contacting is sufficently
simple ensuring that the results are amenable to analysis. Stirred and plug
flow reactors and co-axial turbulent diffusion flames with well-controlled
ignition regions are representative of such experiments. They are both
analyzable and can be made representative of important furnace-like condi-
tions such as stoichiometry, temperature, particle heat-up times, radiation
intensity, residence time, turbulence level, transport behavior, etc. Reactor
experiments provide the opportunity not only to obtain global process charac-
teristics of the reactor as an entity but also allow the definition of
specific phenomenological behavior relating to oil droplet and coal particle
combustion. Understanding the mechanics of devolatilization and vaporiza-
tion, fuel nitrogen conversion processes, and particulate formulation should
be considerably enhanced as a consequence of these experiments.
Figure 3 is a schematic illustration of spray combustion reactor experi-
ments. It illustrates the type of coupled phenomena which can be explored in
a well-controlled analyzable experiment. The limiting situations shown are
representative of different combustion regimes within a real system and yet
are amenable to relatively straightforward analysis. Liquid fuel combustion
has the unique feature of ballistic penetration of fuel into the combustion
system due to high droplet momentum. This provides the potential for greatly
enhanced fuel/air Contacting rat«8. Key to this behavior aro the relative
vaporization, dynamical eqn I II hralIon, and turbulent I. rannport times. lie fore
dynami«:a 1 e<|u I I $hr;il- Icm or valorization Lake place transport ran be by ballJs-
tics or turbulence or both. rollowlng vaporization or velocity equilibra-
tion transport is, of. course, only by turbulence. Well-controlled reactor
experiments can explore these limits. The sketch at the bottom of Figure 3
simulates the conditions existing in a shear layer surrounding the stabilizing
recirculation zone of a swirl or bluff body burner. The simulation approxi-
mates an injection system designed to place the fuel Into the shear region.
22
-------
The above does not negate the value of single particle or controlled
multi-particle experiments provided a strong linkage can be made between
the results of such investigations and the critical behavior of practical
combustors. For example, the effect of liquid fuel and coal properties
on NO and particulate formation may be further illuminated by continued
X
detailed investigations of pyrolysis and combustion with independent control
of heat-up rates, gas phase composition, and particle spacing.
MATHEMATICAL MODELING
The numerical analysis problems associated with NovLer-Stoken modeling
of complex flame behavior arc:
1. Significant diffusion-type errors are associated with attempts
to gain numerical stability.
2. Near-equilibrium behavior (e.g., turbulence, particle dynamics,
particle heat-up, chemical kinetics) can lead to severe numerical
stability problems.
3. Adequate flow field definition in regions of large gradients
(e.g., shear layers, energy release zones) requires a high
density of grid points.
In general, the resolution of these problems leads to prohibitively long
computation times. Problems associated with near-equilibrium behavior can
be eliminated through reformualtion of the mathematical description such
that when any short-time-constant process closely approaches its equilibrium
state that process is in fact redefined as equilibrated.
New numerical techniques which have the potential to resolve the
problems listed above are the subject of much research In applied mathema-
tics. "Flux-Correcting" methods show considerable promise: for .step-wist:
correction of the numt:r leal ly- induced diffusive flux associated with .stable
integration. "Adaptive" grid o.oiitro.1 im-Lhods must In: developed in which the
grid-point density IK locally adapted as a function of gradient Intons i l y.
Other research rotated to "FiniLt- H lemon I." techniques In which local inte-
gration is performed across finite elements imbedded within the. (Mow field
may eventually prove to be the most satisfactory approach.
23
-------
Adequate Nnvier-Stokcs so IuLion techniques will not be forthcoming in
the time frame of the primary FCR objectives. Long-term (10 year) FCR
objectives, however, may dictate that the program should attempt to influence
the direction of research in numerical methods. To this end, consideration
is being given to supporting a task to define a sensible long-term program
of research in numerical methods suitable to FCR needs. The results of
such an effort would then be used to influence the actions of other govern-
ment sponsors.
With regard to mathematical modeling the FCR program will concentrate
on developing adequate data analysis and modular modeling codes. The primary
thrust of the FCR modeling element will be directed toward the development
of adequate data analysis and module modeling codes. The numerical techni-
ques required for such codes are proven standard methods and will not require
further development.
TRANSPORT PHENOMENA AND FLUID MECHANICS
The principal unresolved issues related to transport phenomena in
furnace-type combustors are:
1. Nature of complex transport phenomena in near-field dominated
flames.
2. Turbulence-kinetics coupling due to segregation.
3. Nature of macroscale transport in long turbulent diffusion flames.
4. Participation of liquid or solid phase in turbulent transport.
5. Turbulence lag phenomena.
6. Radiation transport effect on gas and condensed phase temperatures.
The first issue constitutes perhaps the largest gap in our understanding of
significant transport and flow field behavior. To elaborate on this we note
that large furnace flames can be broadly classified into two groups of very
different character:
Near-field dominated typified by present-day wall-fired furnaces
in which fuel processing, flame holding, energy release, and
quenching are intimately coupled in an intense region near to
and dominated by the burner.
24
-------
Far-field dominated typified by corner-fired boilers in which
stabilization is achieved through flame-flame impingement or
through hot product recirculation on a scale of the furnace size.
These classifications aid in making our primary objective more concrete from
the standpoint of defining the flame characteristics of interest. It is the
fluid mechanics of near-field dominated flames that is the concern of the
first issue raised above, and more work is required to bring our under-
standing of such flow fields to the necessary level. Attempts at gaining
this understanding through solution:; to Lin- full Navier-Stokes equations
have failed due to inadequate knowledge- of physics and mathematics. What
knowledge we do have of turbulent transport Js relegated primarily to simple
classical flow fields such as free shear layers, boundary layers, and coaxial
jets. A well-designed experimental and theoretical effort will be undertaken
to explore the basic fluid mechanical characteristics of burner-dominated
flames. Initially an engineering approach using empiricism and relatively
simple fluid mechanical modeling will be pursued.
The objectives of the investigation will be to link the basic fluid
mechanical behavior and fuel particle mechanics with controllable design
parameters. Flame types which are potentially attractive for NO control
A
must receive emphasis. The study should examine the fluid mechanical conse-
quences of energy release patterns associated with several distinct fuel
distributions and recirculation zone characteristics. Included would be:
(1) injecting most of the fuel at high momentum down the centerline through
a stabilizing "donut" vortex pair; (2) depositing most of the fuel at low
momentum into a central torroidal vortex; and (3) placing a portion of the
fuel into the intense shear layer region surrounding a central vortex.
These fuel distributions are illustrated schematically in Figure 4. Indi-
vidually and in combination such fuel injection and recirculation configura-
tions are representative of the several jiy.pothesis regarding NO control.
JC
For example, placing coal In n hot, rich, long residence time recirculation
zone may allow most of the fuel nitrogen to ho. drlvt-n out of the fuel prior
to chnr burnout and Into a nrar equilibrium gas phase state which would
25
-------
minimize overall conversion to NO . Injecting fue.l down the centerllne,
with just sufficient burner or combust ton chamber-Induced recirculatlon for
flame stability, represents an attempt to shield the fuel region from
oxidative species. Hot combustion products mix with the fuel jet providing
a hot, long residence time environment while the flame prevents the penetra-
tion of oxidizing species into the rich fuel zone. Placing fuel into the
vortex boundary shear zone leads to intense combustion with high NO con-
centrations. A potential control scheme would embody parallel staging
wherein a portion of the fuel is burned in the shear layer yielding high
NO and a portion is burned under conditions yielding high XN compounds.
X
Brought together in the proper manner the possibility exists of the NO
X
being reduced by the XN giving rise to a low final NO level.
X
The main point of this discussion of flame characteristics is to
emphasize the need for sufficient understanding to allow manipulation of
flame behavior in very specific ways. Some questions follow which under-
score the type of understanding required. How is a central or torroidal-
pair vortex generated which has a specific size, residence time, and shear
layer strength? How do these quantities scale? How can fuel be ballis-
tically inserted into specific regions of this flow field? How does the
energy release distribution effect flow patterns? How are burner param-
eters such as divergence geometry, radial profile of swirl and axial
velocities, and fuel injection characteristics related to the significant
flame properties?
The pressure field is an important fluid mechanical variable which can
be used as an intermediate link between burner parameters and velocity field
characteristics. Stagnation conditions and attendent adverse axial pressure
gradients required to induce recirculation can be achieved in a variety of
ways. These include confined flows with sudden area expansion, the presence
of a bluff trailing edge body, coaxial mixing between streams with large
velocity ratios, and swirling flows subjected to sudden area expansion. A
"donut" vortex pair can be generated by a combination of swirl and high
centerline axial velocity. Figure 5 illustrates these basic methods of
inducing recirculatlon. The required oxperi mental and theoretical investi-
gations should provide an under.stand.lng of such flow fields, especially
26
-------
swirl stabilized flows, including the turbulent and ballistic mixing of
fuel and combustion air, and the effect of energy release on the basic
flow configurations. Integral techniques offer one promising theoretical
approach. Stimulus-response methods should be investigated as an aid in
the experimental program.
The remaining issues requiring investigation that are listed i\t the
beginning of this section pertain to both near and f.-ir-field dominated
flames. The significance of turbulent-kinetics coupling rents on the ques-
tion of whether long life Lime coherent structures (eddies) exist in fur-
nace-type flames. If a large eddy remains stable for periods comparable to
molecular diffusion times across the internal lamina of the eddy then
significant segregation effects can be present. Both theoretical and
experimental programs are contemplated to explore these questions. An
adequate description of macroscale transport in long coaxial diffusion
flames rests again on the degree of coherent structure, and hence segrega-
tion, existing in furnace-type flames. Substantial overlap of time-mean
concentrations of fuel and air on the centerline of such flames is the
primary indicator of significant segregation.
Turbulence lag refers to a state of turbulence which is not wholly
dependent on the local kinematics of the flow but rather shows a persistence
of turbulent structure generated at an earlier time. That is the turbulence
"lags" its equilibrium state which is dictated by the local flow kinematics.
The important question again is whether such behavior is characteristics of
furnace-type flows. The answer rests on determining the relative time
constants for turbulence equilibration and the streamwise changes in the
time-mean flow. Plausibility arguments and numerical experiments suggest
that furnace-type flows are equilibrated but the issue needs further
clarification.
INITIAL SUBCONTRACTOR AND EER TASKS
Following the preliminary technology assessments and identification of
gaps in the understanding of significant phenomenon ;m Initial program of
subcontractor-oriented research tasks has been set forth. Initial pro-
curements are contemplated in the following areas.
27
-------
1. Stirred reactor and plug flow investigation of the gas phase
chemistry of fuel nitrogen conversion, soot formation, and
SO /NO interation.
x x
2. Pulverized coal combustion reactor studies stirred, plug flow,
controlled mixing.
3. Residual oil combustion reactor studies stirred, plug flow,
controlled spray combustion.
4. Experimental investigation of the basic fluid mechanical
characteristics of residunl oil burners as they relate to fuel
nitrogen conversion.
5. Evaluation of NO sampling measurement techniques and develop-
X
ment of standard procedures.
6. Theoretical investigation of the basic fluid mechanical behavior
of near-field burner-dominated flames as it pertains to energy
release and fuel nitrogen conversion.
7. Theoretical investigation of the fluid mechanics of coherent
structure turbulent diffusion flames as it pertains to energy
release and fuel ntirogen conversion .
8. Experimental search for the lower bound of fuel nitrogen con-
version in gaseous systems.
9. Application of laser holography to determine behavior of coal
particles during heat up and devolatilization.
In addition, consideration for initial subcontracts is being given to:
1. Liquid fuel pyrolysis studies linkage of fuel properties
to combustion behavior and fuel nitrogen conversion.
2. Residual oil droplet burning experiments linkage between
combustion environment, droplet-droplet interaction and tendency
to produce particulates.
3. Experiments designed to investigate the presence of long life time
coherent structure In furnace-type flames.
28
-------
In-house EER efforts under the FCR program are presently concentrated
on the development of data analysis and modular modeling codes as follows;
1,
Stirred Reactor Sensitivity Analysis predicts influence of
rate constant changes on species concentrations.
Inverse Flat Flame Analysis generates rate constants consistent
with observed species distributions.
Two-Phase Stirred Hunt-tor Co
-------
Figure 6 illustrates the output expected from the initial set of tasks
described in this section. The gas phase kinetic information in conjunction
with the various codes allows a description of NO formation in a variety of
simple reactors and laminar flames for which the transport physics is well-
understood. Armed with this information a systematic search, both experi-
mental and theoretical, can be undertaken for lower bound estimates. Moving
down the figure more information pertaining to transport and two-phase
behavior is generated and used for NO prediction in generally more complex
situations. Although coal flames appear at the bottom it should not .be
implied that they are necessarily the most difficult to describe. For example,
a long turbulent coal flame may act much like a fuel-staged gaseous diffusion
flame followed by an essentially premised two-phase char burnout. In
principle, this situation is less complex than burner-dominated heavy oil
combustion.
30
-------
SUMMARY
A three-year program of fundamental combustion research is underway
which is focused on the control of NO and primary related pollutants from
J\
stationary sources.
The initial emphasis has been placed on the combustion of pulverized
coal and heavy residual oil in large boilers. Initial planning and
technology assessments have been completed. These have led to the defini-
tion of initial subcontractor tasks and EER and MIT Energy Laboratory in-
house tasks. The initial program of research is weighted heavily toward
coal and oil reactor experiments which, in a controlled manner, simulate
different regions and limit situations within practical combustors.
The program is conservative with a high probability of success in
generating useful information for design and development of low emission
combustors within a short time period. This confidence steins in large
part from the program of fundamental research sponsored by the CRB over the
last seven years. The results of these investigations have provided the
insight which allows the delineation of significant phenomena and the
formulation of a directed program of research.
31
-------
ACKNOWLEDGEMENTS
The formulation of the FOR program plan has been a team effort and
includes contributions from amny sources. The authors would like to express
their appreciation to all those organizations who invited us to visit their
laboratories to discuss the program. These discussions were most helpful
and gave us an understanding of many research programs which are planned or
are underway, and which will impact FCR. We would like to thank the staff
of the CRB for their interest and their thoughts on Che support expected
from fundamental research activities in other program areas. We are parti-
cularly indebted to Professors Gerstein of USC, Samuelson of U.C. Irvine and
Wendt of U. of Arizona for their help in defining critical problem areas in
the various scientific disciplines involved in pollutant control. We would
also like to thank Professors Beer, Langwell, Louis and Sarofim of M.l.T.
for their helpful comments in reviewing the initial plan.
32
-------
^~ LO
O f2
LU Q; z
3 S £
2 C =
« £ 5
«"^ 1
UJ O
jr* uj LI_ u.
^-5^^
Q- oo oo
g < a: oe
2 °- i-ti LO
*^ LiJ li I lil
§
O
CO
N
H
«
00
)-l
O
s
2
00
O
00
33
-------
/
«/
to tt
^J H
=J (
Sg
12 i
LU ^
x
^
H
<
(_
t
U
H-
g
>*
I/
S
»
1
/
\v
N
,
i
!
;
5
4
)
:
3
i
*
tf
S
LU
LU
C0
UJ
z
) «
u.
g
y
S,
i-^
1
0
u.
z
»^
cC
U
MH
U
£
O
Q
g
*^
_J
S
1
\n
>-
x
a.
to
<_»
>-H
1 CHARACTERIST
V
>
s
\-
0.
z
*-H
<
i
^j
a.
\
i
>
/
\
o 5
i "
* 3E
S C
JJ l_
s
r
1
CURRENT
c
5 to <-
LU
NTERPRETATION
YNTHESIS OF R
TMVACP TO MAw
>-i -I o t
to h D. a
LU CO 0. C
> LU =sa
z > to o
i i Z -
i-i tO f
_l b^
Mlrf ^^
< -JtOK
h- < >- 3
LU I < 3
SEI-§C
Sfl£<9
Uk ^b _
UJ O < f
a, LU t «s
^ ^ ^^ ^^ I .
2P§i
L
- LU
: o
>-«
> >
LU
s
» »-^
i O
JS
i a.
>
4 Z
J >-.
»
I
S
§
o
. QQ
i on
. LU
. 3B
JO
>IJ
to
s of- i
£UJ
g
: z £
;^x
:
-------
PREMIXED
Stirred Reactor
\
V
Plug Flow Reactor
BALLISTIC
MIXING
-»- Air,
-»- Hot
-»- Products
Particle
Trajectories
TURBULENT
MIXING
A1r
Fuel
/
A
r i r i i i r
Distributed Injection
COMBINED
TURBULENT
AND BALLISTIC
TRANSPORT
Air
Turbulent-Free
Shear Layer
Particle Trajectories
i r
Figure 3. Schematic Spray Combustion Experiments
35
-------
FUEL ON CENTERLINE
FUEL INTO RECIRCULATION REGION
FUEL INTO HIGH SHEAR RECIRCULATION
BOUNDARY
Figure 4. Schematic of Several Fuel Distributions with Divergent-Swirl
Induced Recirculation
36
-------
LU
o:
u_
z>
_l
.00
fiL
K
o
o
<0
H
I
H
O
1
O
m
(0
o
H
U
H
U
tn
0)
n
37
-------
§
LU
o.
o
DE
c
M
a. >
Ul CO
z co
2 -<
O 9C
CO 00 Ul
9C X
O
CO
h- CO CO
co ai=>
1-1 Ul O
ac >ui
Z CO
-t H- h- O M
CO 0 CO CO H
t Z3 LU >-l O «C
OC U. > _| QC 1
O U, Z _J CJ
o. CH o <4 i <
co
U
b
O
00
*
14
00
«
H
Q
I
r-l
fa
-------
CHEMICAL REACTIONS IN THE CONVERSION OF FUEL NITROGEN TO NO,
FUEL PYROLYSIS STUDIES
By:
A. E. Axworthy and V. H. Dayan
Rockwell International/Rocketdyne Division
Canoga Park, California 91304
39
-------
-------
ABSTRACT
The objective of this study Ls to investigate the types of "prof lame" react J
fuel nitrogen compounds can undergo. The inert pyrolysis .study was extended to
include additional coals and alternate- liquid fuels. At 1100 0, the MCN yield
(% fuel nitrogen recovered as HCN) ranged from 23 to 44% for seven No. 6 fuel
oils and from 24 to 30% for six coals. A COED char yielded less than 2% HCN,
while a liquefied (catalytic) coal yielded 51% HCN and two hydrotreated shale
oils produced 60 and 69% HCN, respectively.
A two-stage pyrolysis reactor was developed in which volatile nitrogen compounds
are evolved from fuels in the first stage and subsequently react under conditions
that can be quite different. The two-stage experiments demonstrated that nearly
all of the HCN forms from the secondary reactions of volatile nitrogen compounds
and little forms in the initial pyrolysis process. When the first-stage tempera-
ture is decreased from 1100 to 750 C, while maintaining the second stage at
1100 C, the HCN yield from a fuel oil decreases by one-fourth. For a similar
experiment with three coalK and the first stage at 600 C, the HCN yields are one-
third to three- fourths those obtained with both stages at 1100 C. It is postu-
lated that more nitrogen remains in the char at the lower temperature.
In addition, a study was carried out on the oxidative pyrolysis and oxidation
of the model fuel nitrogen compounds pyridine and benzonitrlle. The major
nitrogen-containing products from both model compounds are mainly HCN and NH^
at lower temperatures (720 C) and are N?, N?0, and NO at higher temperatures
0~-760 C) . With both modal compounds, the maximum N?0 yield exceeds 45% (at
about 770 C).
41
-------
-------
INTRODUCTION
An experimental and analytical program is being conducted to investigate the
chemical processes involved in the conversion of fuel-bound nitrogen to NO, in
ft,
combustion. The results obtained during the first phase of this study, performed
under EPA-Contract 68-02-0635, have been reported previously (Ref. 1 through 'i>.
The study is continuing and the additional results obtained to date from pyrolysis
and oxidation experiments are reported herein (Ref. 4). The objective of this
task of the program was to investigate the types of preflame reactions that
volatile fuel nitrogen species can undergo during pyrolysis and before reaching
the flamefront.
Since the previous symposium, model fuel nitrogen compounds have been pyrolyzed
in an oxidative atmosphere, and oils and coals have been pyrolyzed with emphasis
on inert conditions. A two-stage reactor was developed in which the secondary
reactions of the nitrogen-containing volatiles formed in the initial pyrolysis
process could be studied independently.
BACKGROUND
In the previous pyrolysis studies, that were conducted during the first phase of
this program (Ref. 1 through 3), fossil fuels and model nitrogen compounds were
pyrolyzed in an inert atmosphere (helium) in small quartz flow reactors, HCN
was the major volatile nitrogen compound formed in all cases, indicating that
HCN probably forms from nitrogen-containing fuels during the initial, pre-flame
stages of combustion and may be the principal intermediate in the formation of
fuel nitric oxide.
MODEL COMPOUND PYROLYSIS (TNKRT)
Model compound vapors, diluted to .-i concentration of 3% in helium, were, passed
through a quartz cap 1.1 la ry reactor at r«.:m|>uratures Imtwuun 900 and 1100 C wi.th
a nominal residence time :ln the heated zone- of .1. second. Samples of the model
compound liquid, 0.2 \i'i. In size, were vaporized into the preheated carrier gas
stream before it entered the reactor.
43
-------
Under these inert pyrolysis conditions, HCN and a carbonaceous residue are the
major nitrogen-containing products derived from pyridine and benzonitrile, the
two model compounds that were studied in detail. The fraction of the organic
nitrogen that is converted to HCN rather than being contained in the residue
increases with temperature. The HCN yields from benzonitrile and pyridine at
96,0 C are 49 and 40%, respectively. At 1100 C, however, the HCN yield from
benzonitrile is 82%, and all of the pyridine-N is converted to HCN. No molecular
nitrogen and only minor amounts of ammonia form from the inert pyrolysis of the
model fuel nitrogen compounds. At the lower temperature, pyridine forms a few
percent of benfconltri le, quInoJ.Lne, and acrylonltri le.
Based on the kinetic parameters obtained from the pyrolysiK of the model fuel
nitrogen compounds, it was established that even in the absence of oxygen the
organic nitrogen compounds present in fossil fuels should decompose in fractions
of a millisecond at a temperature around 1700 K. It was postulated, therefore,
that when a diffusion flame surrounds a coal particle or oil droplet the volatile
fuel-nitrogen compounds will be mostly converted to HCN before reaching the flame-
front.
FOSSIL FUEL PYROLYSIS (INERT)
Two bituminous coals and six No. 6 fuel oils were rapidly heated in a quartz
reactor to either 950 or 1100 C. The volatile pyrolysis products formed were
carried through the reactor in a helium stream, and the nitrogen containing
species were measured. The fuel samples, 1 to 2 milligrams in size, were placed
in a minature quartz boat that was moved very quickly into the preheated reactor.
The residence time of the volatile products in the heated zone was about 1 second.
The fraction of organic fuel nitrogen that was converted to HCN under these condi-
tions ranged from 15 to 25% at 950 C and from 23 to 42% at 1100 C (Ref. 1 through
3). In each case, the amount of HCN formed at 950 C was about six-tenths of that
formed at 1100 C. Wilmington crude oil yielded 30% HCN at 950 C and 50% at 1100 C.
The NH_ and Ng formed in these fossil fuel experiments were measured together
(after converting any Nll.^ to N^). The total amount of NH plus N_ formed from
the fuel oils (measured as N,,) was nearly independent of the nitrogen content
44
-------
of the oil and of the temperature, and ranged from 0,5 to 0.8 micrograms of N~
per milligram of oil. Because these values are only slightly greater than the
solubility of N« in oils, it appears that the amounts of NH and N~ formed under
these inert pyrolysis conditions are quite small. One of the coals yielded
slightly more N~ plus NH, (1.1 ±0.1 micrograms per milligram), but this was
equivalent to only 9% of the fuel-N. It was established, therefore, that NH
and N_ are only minor products of the rapid, high-temperature inert pyrolysis
of fosssil fuels. This would be predicted from the model compound results.
It was assumed that the remainder of the fuel nitrogen was contained .in the
char and in the carbonaceous residue that formed on the reactor wall from the
volatiles. Since it has been demonstrated (Ref. 5) that much of the fuel nitro-
gen in coals is nonvolatile at these temperatures and would remain in the char,
the amounts of HCN formed at 1100 C may represent nearly all of the "volatile"
organic nitrogen contained in these fuels. It was not possible to distinguish
in these experiments between HCN formed in the primary pyrolysis process and that
formed from the secondary reactions of the volatile nitrogen species. The model
compound results suggested that most of the measured HCN may have formed down-
stream of the sample boat from the secondary reactions of the volatiles.
EXPERIMENTAL
The experimental procedures employed in this phase of the pyrolysis study are
similar to those developed previously (Ref. 1 through 3). The model compound
experiments were again carried out in a reactor that was simply a heated section
of 2-mm ID quartz capillary, but the carrier gas was changed to a 3:1 0--He
mixture so that the oxidative processes could be investigated.
Shown schematically in Fig. 1 are the single-stage fuel reactor, used in the
previous study (Kef. 1 through 3), and a two-stage reactor developed and used
in the present Investigation. In both reactors, the volatiles formed from the
pyrolysis of the fuel .sample in the quartz boat (that can be moved rapidly into
the preheated reactor) flow through a 2-cc volume downstream microreactor in which
secondary reactions can occur. In the single-stage reactor, the quartz boat
and the microreactor are at the same temperature and the carrier gas is of the
same composition at each of these points. With the two-stage reactor, however,
the temperatures at the stop position of the boat and at the microreactor can
45
-------
differ by as much as 500 C, and oxygen can be introduced into the microreactor
only. Thus the fuel can be pyrolyzed in helium at a given temperature and the
volatiles formed can be reacted downstream at a different temperature in either an
inert or oxidizing atmosphere.
Only the two-temperature-zone feature of this reactor has been employed in the
experiments conducted thus far, i.e., 0,, has not yet been added to the second
stage. A study of the oxidation of the volatile fuel nitrogen species in the
second stage only is planned. It has been demonstrated in the inert pyrolysis
experiments that operation of the two-stage reactor with both stages at the same
temperature gives, as expected, results nearly equivalent to those obtained in
the single-stage reactor. Therefore, for simplicity, such single-temperature
experiments will be referred to as s Lnglt'-stage experiments in this paper.
The remaining experimental details will, be described in the following sections
where the results from each type of pyrolysis experiment are presented and
discussed.
INERT PYROLYSIS OF FOSSIL FUELS
(SINGLE STAGE)
Additional fossil fuels have been pyrolyzed in helium under the single-stage
condition (where both the primary and secondary reactions take place at the
same temperature) and the amount of HCN formed has been determined. The reactor
was again operated at atmospheric pressure, the only change in the procedure
being to increase the helium flowrate from 20 to 40 cc (STP)/tnin so that shorter
residence times could be investigated (M).7 sucond) . It was established that at
1100 C the HCN yield is independent of flowrate over this range. The volatile
nitrogen compounds apparently react completely even at the shorter residence time.
This is in agreement with the results of the previous model compound experiments.
HCN YIELDS FROM OILS AT 1100 C (SINGLE STAGE)
Listed at the top of Table i are the HCN yields obtained at 1100 C from four addi-
tional oils. Also listed for comparison arc .selected results from the previous
study. A No. 6 fuel oil c-.onlnlning 0.71% nitrogen gave an HCN yield of 44.51.
46
-------
This is the largest percentage yield that has been obtained from a petroleum-
based No. 6 fuel oil but is less than the 50% HCN yield obtained from Wilming-
ton Crude Oil (Table I). The HCN yield from this No. 6 fuel oil amounts to
6.1 ygm HCN per mg of oil (or 6.1 gms of HCN per kilogram), which is about
the same as was formed from both a No. 6 fuel oil (obtained mainly from
California crudes) that contained twice; as much fuel n.i trogen and from Wilming-
ton Crude Oil.
A sample of liquefied coal prepared by the Synthoil process was pyrolyzed under
these same inert conditions at 1100 C. This oil, which contained 1.3% nitrogen,
produced twice as much as HCN as any oil that has been tested. The amount of
HCN produced was 12.8 ugm/mg, which represents an HCN yield of 51.0%.
Two hydrotreated shale oils gave large HCN percentage yields of 60 and 70%,
respectively, under these conditions, but the amounts of HCN produced were
only 3.1 and 2.4 ugm/mg because of the low nitrogen contents of these treated
oils. The first of these oils (C in Table I) was a treated full boiling range
shale oil that contained 0.27% N and 0.04 S. The raw oil before treatment con-
tained 2.05% N and 0.57% S. The second oil (D) was a treated middle distillate
that contained 0.18% N and 0.04% S after treatment and 1.60% N and 0.57% S before
These limited results with a liquefied coal and two treated shale oils suggest
that the fuel nitrogen compounds in these oils form HCN more readily during
rapid inert pyrolysis than do those in No. 6 petroleum-based fuel oils. This
also suggests that these alterante fuels may show a greater extent of conver-
sion of NO to fuel nitrogen during combustion with excess air.
X
EFFECT OF TEMPERATURE ON HCN YIELDS (SINGLE STAGE)
The HCN yield from the No. 6 fuel oil that was listed nt the top of Table 1.
was measured at. the three lemperntures 750, 9*50, and 11.00 C. These results
are plotted on linear scales Jn Fig. 2A ;md In the Arrhenius form in Fig. 2B.
Under these single-stage conditions, the HCN yield drops from 46% at 1100 C
to 4% at 750 C. The decrease in the HCN yield of a factor of 2 between 1100
and 950 C is in the range observed for other fuels in the single-stage
experiments.
47
-------
The Arrhenius plot shows a slight curvature, but the average apparent activation
energy for HCN formation is 19 kcal/mole over this temperature range. This
apparent activation energy may result from a combination of the temperature
dependencies of the following effects: (1) the fraction of fuel-N that remains
in the residue, (2) the fraction of the volatile fuel nitrogen species that
decomposes before exiting the reactor, O) the formation of HCN from the fuel-N
species that do react, and (4) the inclusion of nitrogen in the carbonaceous
residue that forms from the volatiles. It will be shown in a later section
that the results of two-stage reactor experiments with this fuel oil indicate
that the amount of nonvolatile nitrogen that remains in the residue, effect
(1) above, does not change appreciably over this temperature range.
HCN YIELDS FROM COALS AT 1100 C (SINGLE STAGE)
The HCN yields obtained from the inert pyrolysis of three coals at 1100 C are
listed at the top of Table II. The results from the two coals studied previously
are also for comparison. All five of these coals gave HCN yields in the range
of 24 to 30%. The amount of HCN produced ranged from 3 pgm/mg to 8.5.
INERT PYROLYSIS OF FOSSIL FUELS
(TWO STAGE)
The production of HCN in the inert pyrolysis of fossil fuels was investigated
using the two-stage reactor described in the Experimental section of this
paper. The fuels were pyrolyzed in helium at one temperature and the volatile
products formed were passed through a 2-cc microreactor held at a different
temperature. The objective of this study was to determine what fraction of the
HCN produced is formed from the secondary reactions of the volatile species
that form in the primary pyrolysis process. The previous model compound results
had indicated that these volatile nitrogen compounds should convert almost com-
pletely to HCN when pyrolyzed in lieliutn at 1.100 C.
It was suggested (Ref. 2) that most of the HCN is formed in secondary reactions
of the volatiles rather than In the initial fuel vaporization/pyrolysis process.
If this is the case, the HCN yield should increase with the temperature at which
the volatiles react and be fairly Independent of the vaporization temperature
particulary for oils where the amount of nitrogen left in the residue is expected
48
-------
to be nearly independent of temperature. In the present two-stage experiments,
the effect on the HCN yield of varying the temperature of each stage independently
was i nves t i. gn tecl.
TWO-STAGE INERT PYROLYSIS OF A NO. 0 FUEL OT.L
The No. 6 fuel oil listed at the top of Table I (Oil A) was employed i.n this
series of experiments. The first stage of the two-stage reactor (Fig. 1) was
constructed of 2-tnm ID quartz capillary tubing and had a volume of only 0.25 cc.
The second stage, which was also quartz capillary tubing, contained the 2-cc
microreactor in which the volatiles had sufficient residence time to react. The
oil samples were nominally 1.8 mg in size; this amount of oil should form about
0.8 cc of hydrogen during pyrolysis (calculated at 950 C). Therefore, the
residence time of the volatiles in the first stage will be even less than 10%
of the second-stage residence time and the microreactor volume is required to
prevent the hydrogen formed from expelling the volatiles from the heated zone
before they can react.
Duplicate experiments were conducted at i:ach of the nine temperature combin;it ions
that can be obtained by setting each stage at 750, 950, or 1100 (.'.. The HCN
yields obtained are listed in Table III.
Experiments With First Stage at 950 or 750 C
Plotted in Fig. 3A and 3B are the HCN yields obtained in experiments in which
the first stage was at 950 or 750 C. The yields are plotted as a function of
second-stage temperature to demonstrate the amount of HCN that was formed in
the second stage. If all of the HCN should form in the first stage, the HCN
yield would be independent of the second-stage temperature. This is definitely
not the case.
The yields shown Ln Fig. 3 are dependent on the first-stage temperature only in
the case where the second stage is at 750 C. Under this condition, increasing
the first-stage temperature from 750 to 950 C increases the HCN yield from 4 to
14%. This demonstrates that in all of the experiments in which the first stage
is at 950 C, from 10 to 14% (absolute) of the HCN yield is formed in the first
49
-------
stage. When the first stage is at 750 C, however, this amount of fuel nitrogen
is vaporized in a form that can be converted to HCN in the second stage if the
second-stage temperature is 950 C or greater. This causes the two curves in
Fig. 3 to coincide at the higher second-stage temperatures. According to this
interpretation of the data in Fig. 3, the amounts of HCN formed in each stage
under these six conditions are:
750 C/750 C 750 C/950 C 750 C/1100 C
0 to 4%/0 to 4% 0 to 4%/0 to 4% 0 to 4%/34 to 38%
950 C/750 C 950 C/950 C 950 C/1100 C
10 to 14%/0 to 4% 10 to U%/10 to 14% .10 to UJE/27 to 31%
With the second stage at 1100 C, therefore, two-thirds or more of the HCN forms
from reaction of volatiles in the second stage. The HCN that forms in the first
stage at 950 C may also from from reaction of volatiles before they exit the
first stage. Thus, most or all of the observed HCN is formed from the reactions
of volatile fuel-nitrogen species in the second stage.
Experiments With First Stage at 1100 C
Plotted in Fig. 4 are the HCN yields fi-om the six experiments in which the first
stage was heated to 1100 C. Under this condition, the yield is much less depen-
dent on the temperature of the second stage and an HCN yield of 35% is obtained
even with the second stage at 750 C. This indicates that 31 to 35% of the HCN is
formed in the first stage when it is at 1100 C. Thus the amounts of HCN formed
in each stage are:
1100 C/750 C 1100 C/950 C 1100 C/1100 C
31 to 35%/0 to 4% 31 to 35%/7 to 11% 31 to 35%/ll to 15%
The most likely explanation for the results obtained with the first stage at
1100 C is that the pyrolysls rate of the volatile nitrogen species becomes so
high that about three-quarters of the volatile nitrogen species react in the
first stage during the short time they are flowing through this portion of the
two-stage reactor. Although the HCN yield has a rather low temperature coeffi-
cient (apparent activation energy) of about 19 kcal/mole (Fig. 2B), the rate of
50
-------
NDH
' (Z "JaH) anpTsaa snoaoeuoqjBO jo srjunome
asnsoaq p^ayA -ta«oi qBu/iaujos B 9At8 04 paqoadxa aq pjnoM
punoduioo -[apoui aqq uj pojrtHeaui qou SBM auo-[t; outr.ou'inb uioaj
oq pappt; u;iq/vv '^
qnq
-sp px^jX fjon aq^ 'jjo lanj 9 -0^ B 03 pappe SBM a
punodiuoo japoui aq^ uj f) QOTT 3^ auo^e 9[taq|:uozu3q wojj XisnofAajd
SB ^^^s 31I3 anoqe ST STIU, -paufB^qo SBW %9*08 J° PI^T^ NDH UB * (TT°
*°N SB PTOS OS^B) lanj xasaTP 2 'ON B 05 pappB SBM afTarj-puozuaq
NDH o:> N-TanJ J° uoTsaaAUoo
squaintjadxa aq^ jo s^insai aqq
NDH UB 3Ae§ -[TO ian.f 9 -ON padopun aqx 'poo§ a
uaas aq UBD 3T PUB sjuaiutaadxa aaaq^ ao owg jo
UT pai&'TI
pauinssB SBM 51: 'pasn SBAI -[TO stq^ qoTt!w UT
uj (j aiqnx) Xpnqs snofAaad aqj uy
SBA
aqrj uo paseq
jo
fapoui aqq moaj pamaojf ST
saioads uaSoaaTU aq^
pasn aaa« spunoduioo -[apoui
auzp-paXa -spunodmoo
s^oajje ^Bqi aauupui auios UT
aqq UT pauuog saToads
os s^uauitaadxa qons
NOH J°
s-[to asoqq jo
AUB JT auTmaaqap 03 SBM asodand aqx '%z 3r*oqB oq squaq.uo3 uaSoaq-pu
03 auT^ou-rnb jo aiTj^TUozuaq qqTn j^adop,, uaaq pBq qeqq SITO
-moj o«q qqxM Q OOTT 3E paqonpuo'o ajaw squauiTaadxa sjsXjoaAd
riaaow
HUM CJSIdOC] S'HO JO SISAlOMAd
S-[BOD asaqq uj sajoads
jo sadA^ aqq UT saouaaajgtp a-[qejiapTsuoo OJB ajaqq qBqq sqsaSSns STUJ,
'D OOTT 3e SB o 009 3B NDH MDnlu SB %g^ qnoqt- saATS (i -ON) TBOO 9 -ON qganq
rui,! aqq 3TTq« 3 Q09 ^B aSeqs qsaTj aqq qqiM NDIJ ipniu SB gfBq-ouo put? p
-auo qnoqB KIUO aonpoad Aaijj 'aaojaaaqx -(A ''[q^l J<' uuinjo:> qsu-j) ;) OOTT
-------
IS
"N3H
0'0£
*°M
aqq
A *°N
paisodxa se '
3noqe ATUO paonpoad 'uiBaSoad HnotAsad aq3 aapun paipms
qotqw * IA *°N PUB A 'ON 'STBOO snoupurujq aqx -/y ajqex uf
3S9qi 'qoE9 uioaj N^H jo p"[3fA sqq auTuiaa^ap 03 y OOTT 3B a3i?^s puooas 3q^ pue
0 009 ^ aSnas asj-pj 3qq I^TM pazXfoa^d oasw aeip aaoD B PUB S~[BOD
jo
puooas 319 u-f pauuoj sj -J;BOO s^qi uioaj painaoj
^qj. '%£ A^uo sj
jo 3som jfjo aAjjp 03
ij aqq q^T*1 'D OS/ o:i
jo 3unouie
3 ' (AI
D OSi ^^
-qaanoj-auo qnoqc Aq p[aj^ KOH sm sascajaap A^uo f) Q09
j aqq Su^aartox ' qoBj uj 'o-jpB^OA X^a^Baapoui aae ^OH
3Bqa T;BOO sfqq ut quasaad saiaads uaSoaqfu ^q^ 3«q3 saqBOTpuj sjqj, *
OOTT 3B ST aS^qs qszjj aqq aaq^aqw %£i qnoqe jo spia-pA NDH "T
SB NOH °3 satoads uaSoaqxu ajj^F^OA aq:j jo qonui SB ^iaAUoo 03 '3 QOTT 3°
uinniTxcni S^T 3B aSB^s puooas aqq SUTPIOR 'M aj^I UT paziaBtmans ajB
ut i *ON IBOO) ^BOO g -ON qgjnqsrnja B q^T* pauTB^qo sa^^sai a8B3s-o«3 aqj,
STVOO 10
29VIS-OMI
aq3 JT
30 adX3 Jo
03
PTTOS B M^Tw 3uB3JtodaiT aaoui sj pauuoj saxoads
uo ajruBaaduia} a3B3S-3SJtj aq3 jo
uj
i.u a[t3B-j;oA jo sife-Aioj/Cd /Ltupuonos oqn uioaj suuoj \\o
3jaut aq3 UT pamaoj M
ua8
stq3 30
asaqi
I TO
'D OOTT
30 aSB3s 33:173 aqq UT anooo p-[noqs saToads uaSoaq-cu
30 qonra 30 uo-pBinjoj aqq 3Bqq 'a^ojaJ
3ou ST 3i -o OOTT PUB 056 uaa«3aq £3 30 ao30B3 aaq30uB Xq puB 3 QS6 pUB
uoTqDBaa B jo a3BJ aqx ' (Z "39>l) uoTqTSoduiooap auTpT^Xd saop SB
aqq uo XSaaua uorqBATq^B UE aAcq p-[noqs saroeds a^T3BXOA aqq 30
-------
The results obtained with the doped oils under these conditions do not indicate
any major interaction between the nitrogen compounds and the oil substrate.
Because of differences between the molecular weights, however, some of the
model compound may pyrolyze before the oil does, thereby masking a part of
any interaction that does occur.
OXIDATIVE PYROLYSIS AND OXIDATION OF
MODEL FUEL NITROGEN COMPOUNDS
A study was conducted of oxidative pyrolysis and oxidation of pyridine and ben-
zonitrile. The major goal was to identify and measure the nitrogen-containing
products of oxidative pyrolysis to determine what nitrogen intermediates may
be important in the conversion of fuel nitrogen to NO in combustion. The
model compound vapor was flowed continuously through the reactor and an excess
of 0_ was employed so that its concentration would remain essentially constant
even at complete reaction.
The experimental conditions and a schematic of the apparatus are presented in
Fig. 5A and 58, respectively. In each experiment a 3:1 0_-He mixture was passed
continuously through a quartz capillary reactor (2-mm ID) at atmospheric pressure
and a flowrate of 40 cc/min (measured at room temperature). The gas entering the
reactor contained 10 torr of model compound vapor in most experiments and 1 torr
in others. The reactor effluent passed through a heated line (~200 C) to a 5-cc
loop in a heated Carle gas sampling valve from which samples were diverted per-
iodically to gas chromatographs, or through an aqueous bubbler system to collect
the HCN or NHL for colorimetric analysis (Kef. 2).
Conditions for gas ehromatographic analyses of the product gas stream were as
follows:
Gas Cons tituent
Unreacted model
compound and
organic products
N and CO
and
(CN),
Column
Chromosorb 103
Molecular
sieve 5A
Molecular
sieve: 5 A
F'oropak CJ
Temperature
Temperature
programmed
(60 to 240 C)
Ambient
temperature
Isothermal
160 C
120 C
63
Detector
Flame ionization
Thermal conductivity
Thermal conductivity
Thermal conductivity
-------
In addition to the gas chromatographic analyses, the amount of NO formed was
A
evaluated by flowing the reactor effluent directly into a TECO chemiluminescence
NO analyzer (with sample-flow-control capillary removed). A needle valve was
X
placed between the reactor and the CA so that atmospheric pressure could be
maintained in the reactor. N0« also was measured by placing Saltzman's reagent
in the bubbler. However, calibration with an N0« in N- standard indicated that
the amounts of N02 measured (by either method) may have been much less than that
actually formed. Apparently, the NO reacts with moisture on the walls of the
system downstream of the reactor.
EFFECT OF TEMPERATURE AND CONCENTRATION ON OKI DATIVE
PYROLYSIS RATES
Shown in Fig. 6 are the amounts of pyridine and benzonitrile remaining unreacted
as a function of temperature. Benzonitrile is much more stable than pyridine
under the conditions of oxidative pyrolysis investigated. Reducing the reactant
concentrations by a factor of 10 had only a small effect on the fraction reacted
at a given temperature. This is in contrast to the inert pyrolysis results
(Fig. 29 of Ref. 2) where a decrease in the pyridine concentration by an order
of magnitude increased the 50% decomposition temperature by 60 C. Measurements
were not made with the higher concentration of benzonitrile below 720 C because
a residue formed at the lower temperatures, which resulted in the eventual plugg-
ing of the system downstream of the reactor. A similar residue formed from pyri-
dine but not to as great an extent.
PRODUCTS FROM PYRIDINE AT HIGHER CONCENTRATION
The yields of nitrogen-containing products from the oxidative pyrolysis of pyridin
are plotted in Fig. 7A as a function of temperature, and the carbon product yield
are shown in Fig. 7B. The unreated pyridine curve from Fig. 6 is reproduced on
each of these plots for reference. The upper curves in Ftg. 7A and 7B demonstrate
that, within experimental error, all the nitrogen and carbon present in the pyridi
is recovered in the-, volatile products above about 750 C. At lower temperatures,
the recoveries decrease with both the. nitrogen and carbon recoveries being about
80% at 720 C. This Indicates that the C/N molar ratio In the nonvolatile pro-
ducts is about 5 as it is in pyridine itself.
54
-------
At temperatures below 720 C, HCN is the major volatile nitrogen-containing pro-
duct from pyridine (Fig. 7A), but NH_ is also formed in appreciable quantities.
This is in contrast to the results obtained in the previous inert pyrolysis
study wherein NH, was formed in only very minor amounts. N~0 was found to
be an appreciable oxidation product of pyridine, starting at 710 C, and the
yield increased to 45% at about 770 C. Molecular nitrogen, which was not
formed in the inert pyrolysis of pyridine, does not form at 710 C, but its
yield increases rapidly at slightly higher temperatures (in the same regi.on
where the yields of HCN and NH decrease rapidly). The yields of NO and N0?
from pyridine, shown in the following table, were quite small under these
conditions (i.e., at the higher pyridine concentration), reaching maximum
values at 900 C:
Temperature ,
C
710
749
900
Yield, %
NO
0.7
0.5
3
N02
0.05
0.1
0.7
At 720 C, equal amounts of CO and CCL are formed from pyridine (Fig. 7B). As
the reactor temperature is increased, the CO yield decreases rapidly while
the COj increases even more rapidly. It is likely that the CO forms at the
higher temperatures via the homogeneous oxidation of CO (i.e., through the
reaction CO + OH = CO + H). Extrapolating the high-temperature rate expres-
sion derived from the post-flame gas results of Sobolev (Ref. 6), gives the
following predicted half-lives under our conditions (assuming that the mole
fraction of H00 is 0.03):
Temperature, C
CO Half-Lf.fc, second
550
0.9
f.OO
0.4
650
0.2
700
0.07
750
0.04
The residence time of the carrlfr gat: In the. heated ssono of the re.-.at:tor under
these conditions (1 atm, 40 cc/min, and 700 C) can be calculated from the
upper curve in Fig. 28 of Ref. 2 for which the residence time was 0.6 second.
Correcting the residence time for the differences in temperature, pressure,
55
-------
and flowrate gives 0.4 second for the present conditions. Therefore, Sobolev's
rate expression predicts that one-half of the CO should oxidize homogeneously
at 600 C, compared with the observed temperature of about 720 C. This is
fair agreement, considering that Sobolev's rate expression was obtained in
the temperature range of 1600 to 2100 C and Dryer et al. (Ref. 7) have demon-
strated that the CO + OH reaction does not have a simple Arrhenius temperature
dependence. Also, the H^O concentration has not been measured in our
experiments.
PRODUCTS FROM PYRIDINE AT U)WER CONCENTRATION
The NO and N0? percentage yields (Fig. 8) were considerably higher when the
pyridine concentration in the reactor feed was lowered by an order of magni-
tude. No other product measurements were carried out under these conditions
because the sample sizes were quite small.
PRODUCTS FROM BENZONITRILE AT HIGHER CONCENTRATIONS
The products from the oxidative pyrolysis of benzonitrile are plotted in Fig.
9A and 9B. As with pyridine, nearly all the carbon is recovered in the volatile
products above 750 C. Nitrogen recovery is approximately quantitative at 750 C,
but drops off to about 80% at higher temperatures. At 722 C, where only 42%
of the benzonitrile reacts, the recoveries of carbon and nitrogen (based on
the reacted benzonitrile only) are 43 and 49%, respectively. This indicates
that the C/N ratio in the nonvolatile products, which apparently form in a
'45% yield at this temperature, is the same as in benzonitrile (i.e., a molar
'ratio of 7), suggesting that some type of "oxidative polymerization" takes
place at the lower reaction temperatures.
Comparing Fig. 7A and 9A, it can be seen that the nitrogen products from ben-
zonitrile and pyridine are quite similar. The major differences are that
with benzonitrile the N.O yield drops off to only 15% at the higher temperatures,
and NO forms in 20% yield while NO was only a minor product from pyridine (at
the higher concentration being discussed In this section). By coincidence.
56
-------
each of the four products (HCN, NH~, N-0, and N ) are produced in about 25%
«J £, Z,
yield from benzonitrile at 740 C. The maximum NH_ yield is about the same (25%)
from each of the two model compounds. However, the maximum HCN yield was
only 25% from benzonitrile, compared with 40% from pyridine.
In both Fig. 7 A and 9A, the N yield increases abruptly from nearly zero at
720 C to about 53% at 750 C. The N yield from pyridine continues to increase
at higher temperatures; for benzonitrile, however, the N» yield decreases to
45%. Although the N_0 yield from both pyridine and benzonitrile reaches the
same maximum (about 47%), the N_0 curves are quite different In shape.
Benzonitrile forms no N?0 at 720 C, whercini? pyridine forms about 15%. The.
N-O yield Erora pyrldent- only decreawes nUghlly from Its maximum value as
the temperature is increased, but the N?0 yield from benzonitrtle drops to
about 15% at temperatures above 800 C. For reasons that are not apparent,
the total nitrogen species balance from benzonitrile also falls off at the
higher temperatures, i.e., the increase in NO yield is not large enough to
account for the decreased yields of N-O and N~.
The N02 yield from benzonitrile was low, reaching about 2% above 800 C. One
possible explanation for the low nitrogen recovery from benzonitrile between
800 and 900 C would be that 20% or more NO. was actually formed, in this
region, but most of it was lost before reaching the TECO CA. Some of the N0?
calibration experiments indicate that one mode of N02 loss, probably reaction
with moisture on the wall, results in the formation of one-half mole of NO
per mole N02 reacted. Thus, if 40% NO formed from benzonitrile at the higher
temperatures and reacted in this manner, this would account for the 20% yield
of NO and the 20% of nitrogen that is missing.
Comparing Fig. 7B and 9B, the C02 and CO yield curves are quite similar. The
CO^ first approaches its maximum at about 760 C in both cases and the CO "bums
out" at only a slightly higher temperature in the pyridine experiments than
in the benzonitrile experiments.
57
-------
PRODUCTS FROM BENZONITRILE AT LOWER CONCENTRATION
The NO and NO yields from the oxidative pyrolysis of benzonitrile at the lower
concentration are plotted in Fig. 10. These results are similar to those
obtained in the lower concentration pyridine experiments. It is noteworthy
that of the four conditions studied (two model compounds at two concentra-
tions) only pyridine at the higher concentration (Fig. 7A) did not produce
about 20% NO at 900 C. An explanation for this is not apparent.
DISCUSSION OF MODEL COMPOUND OXIDATIVE PYROLYSIS RESULTS
It can be seen from Fig. 7A and 9A that HCN, which is a major product of inert
pyrolysis, is also formed in appreciable yields in oxidative pyrolysis.
Ammonia, on the other hand, is a major product of oxidative pyrolysis even
though it was not important under inert pyrolysis conditions. It is not
possible to establish from the available data if the change in product com-
position as the reactor temperature Is increased results from a change in the
primary products of oxidative pyrolysis or from secondary reactions of the
primary products before they oxlt the reactor. Experiments are in progress
to investigate the possible secondary reactions.
If secondary reactions are controlling the product distributions at the higher
temperatures, the differences between the products from pyridine and benzonitrile
must result mainly from the secondary reactions of the low-volatility products
that are formed from the model compounds in the initial pyrolysis process. At
temperatures below about 750 C, much of this material exits the reactor,
leading to incomplete carbon and nitrogen balances and to the formation of
deposits at the reactor exit and in the sampling valve. At higher tempera-
tures, all of this lowvolntility material reacts before exiting the reactor
(at least in the case, of pyridine) and may, therefore, affect the product
dl.stri but Ion.
50
-------
OXIDATIVE PYROLYSIS OF FUEL OILS
(SINGLE STAGE)
The planned iwo-stage fuel oxidative pyrolysis experiments should be the most
informative because only the volatiles that form in the initial (inert) pyro-
lysis process will encounter an oxidative atmosphere. Preliminary oxidative
pyrolysis experiments were carried out, however, with two No. 6 fuel oils under
single-stage conditions where both the fuel sample and the products were in
contact with nn oxidative atmosphere.
The conditions were those used previously in the inert fuel pyrolysis studies,
except that lower temperatures were employed and oxygen was present in the
carrier gas system. It should be noted that under these single-stage conditions
all the sample eventually reacts and, unlike inert pyrolysis, no residue is
left in the quartz boat. The products measured include both those formed
from the oxidation of the volatile portion of the oil and those formed in the
heterogeneous oxidation of the char. In some of the experiments, the boat
was removed from the re-actor after approximately one-half of the sample had
reacted to determine how much of the HCN formed in the later stages of reaction.
Us tod in Table VII are the nominal reaction times and yields of HCN and NH_
obtained in the experiments In which the sample was allowed to react to
completion.
Experiments in addition to those in which the products were measured were
carried out to determine the time for the entire sample to react. This was
done by removing the boat periodically from the heated zone for inspection.
At 700 C with only 0 flowing, the sample flared up and the light reflected
in such n way that the flame was visible outside of the furnace. With 20% 0_
in helium, the t: i me for complete reaction of a l.'J-mg sample of Oil-A increased
from « st-comls nt 700 C. to about 8 minutes at 500 C (Table VTT). Oi L-G reacted
sul>Ktnni i;ill.y slower under the same conditions. The oil samples would Initially
be present a:i ;i <1 rop at one point In th<: hnat. Shortly after the samples were
lif;i!<«.!, they flowed lo cover the interior surface of the boat and then formed
;i Hi.-ir thai' gradiui J.I y reacted.
59
-------
The SIGN yields increased with decreasing temperature (Table VTI). The maximum
yield of HCN was obtained at 500 C where the reaction rate is quite slow and
the reaction is almost certainly heterogeneous. It is possible that at the
higher furance temperatures, self-heating of the sample surface increases the
temperatures to the point where part of the HCN formed undergoes oxidative
pyrolysis. Calibration samples of HCN passed through this fuel reactor under-
went only 15% reaction in 100% 02 at 700 C. Therefore, the surface temperature
of the sample would have to have been well in excess of 700 C for any HCN to
have been lost once it was formed. The maximum yield of HCN from Oil-A under
these oxidative conditions was 23%, while Oil-G produced only about two-thirds
of this HCN yield.
Under conditions wltcro the maximum HCN yield was obtained and the rate was the
slowest, 500 C, It was found that about one-half of the HCN formed from Oil-A
during the first 3-1/2 minutes of reaction and about one-half during the final
3-1/2 minutes. In a similar experiment at 600 C, 72% of the HCN formed during
the first 20 seconds of reaction and 28% formed during the remaining 20 seconds
of reaction.
NH_ measurements were made only with Oil-A'at 650 and 500 C (Table VII). The
ammonia yields were 11.5 and 13.5%, respectively. One-microliter samples of
pyridine also were passed through the fuel reactor under these conditions and
the amount of NH^ formed was measured (Table VIII). Practically no NH formed
below 700 C, indicating that pyrldine-type fuel nitrogen compounds will not
react homogeneously under these conditions. At 700 C, a 12% yield of NH- was
obtained From pyridine, but this decreased to 2.3% when the oxygen concentra-
tion was increased. Tin- decrease in NH may result from oxidation of the NH
Lhat formed before it exited the reactor. Only minor amounts of NO and NO
were formed under these conditions (Table VII).
The measurement of the inorganic nitrogen species formed is more difficult in
the furl oxidation experiments than when a helium atmosphere is employed. If
the fin-! contains sufficient sulfur, the sulfite Ion formed interferes with
measurement- of HCN by the barbituric acid colorimetric method. Tt was found,
60
-------
however, that a reduction in the temperature of the Na2CO trapping solution
I o 0 C ncf(;iifK (his I nto rTcn.-mrf. To measure NO by the Salt/.man procedure,
it imisi be conn-nl: r;-iU!
-------
temperature; at 750 C, for example, increasing the second-stage temperature from
750 to 1.100 C increased the HCN yield from 4 to 38%.
The two-stage inert pyrolysis of three bituminus coals showed a much larger
effect of first-stage temperature than was the case with the No. 6 fuel oil.
Holding the second stage at 1100 C to convert as much of the volatile nitrogen
species to HCN as possible, while reducing the first-stage temperature from
1100 to 600 C, resulted in decreases in the HCN yields from these three coals
of one-fourth, one-half, and two-thirds, respectively. These differences may
result from differences in the amount of nitrogen that remains in the residue,
and suggests that there are considerable differences in the types of fuel
nitrogen species in these coals.
The inert pyrolysis at 1100 C of oils doped with the model compounds benzoni-
trile and quinolino gave about the same HCN yields as did the neat model
compounds. This indicates that there is no major interaction between the nitro-
gen compounds and the oil substrate under these conditions.
The oxidative pyrolysis of the model fuel nitrogen compounds pyridine and
benzonitrile was investigated in a quartz flow reactor. HCN was also a
major product under these conditions and NH , which did not form to any
appreciable extent in inert pyrolysis, was a significant product of oxidative
pyrolysis. Quite surprisingly, N20 formed in a 45% yield from each of these
model compounds at about 760 C. These oxidative pyrolysis results indicate
that, NH3 and N20 in addition to HCN, may be important intermediates in the
formation of fuel NO in combustion.
5C
ACKNOWLEDGEMENT
This HI u
-------
REFERENCES
Axworthy, A. E., G. R. Schneider, and V. H. Dayan, "Chemical Reactions in
the Conversion of Fuel Nitrogen to NO , "Proceedings of the (First)
A
Stationary Source Combustion Symposium, September 1975, Vol. I, Fundamental
Research, EPA-600/2-76-152a, Rockwell International, Rocketdyne Division,
r.nnoga Pnrk, California, .June 1976.
Axworthy, A. E. et al, Chemistry of Fuel Nitrogen Conversion to Nitrogen
Oxides in Combustion, EPA-600/2-76-039, EPA Contract No. 68-02-0635,
February 1976.
Axworthy, A. E., V. H. Dayan, and G. B. Martin, "Reactions of Fuel Nitrogen
Compounds under Conditions of Inert Pyrolysis," FUEL (accepted for
publication).
The results of the flat-flame burner study that is also being carried out
under this program will be reported at this meeting by Dr. Daniel Kahn.
Pohl, J. H., and A. F. Sarofim, 16th Symposium (International) on
Combustion^ 1976.
Sobolev, ("!. F. , Seventh Symposium (International) on Combustion, p. 386,
1958.
Dryer, F., D. Naegeli, and I. Classman, Combustion and Flame, 17, 270,
1971.
63
-------
TABLE 1. HCN YIELDS FROM INERT PYROLYSIS
OF OILS AT 1100 Ca
A.
B.
Oil
No. 6 Fuel Oilb
Liquefied Coal
% N
0.71
1.3
ygm HCN per
% S mg of oi 1
6.1
12.8
HCN yield, %
44.5 ±0.3
51.0 ±0.3
(SYNTHOIL)
Treated Shale OMC
(Full Boiling Range)
Treated Shale Oi1°
(Middle Distillate)
No. 6 Fuel OMC'f
(Mainly California
Crudes)
0.27 0.04
0.13 0.04
1.41 1.6
3.1
2.4
6.31
60.0 ±1.3
69.5 ±2.5
23.2
F.
G.
H.
Wi
No
No
1 m ing ton
. 6 Fuel
. 6 Fuel
Crude Oil"''
Oilb'f
one'f
0
0
0
.63
-38
.22
1.
0.
0.
6
3
9
6
2
1
.06
.58
.56
49
35
37
.8
.2
.0
a.
b.
c.
d.
Single-stage Reactor, 40 cc/min helium carrier gas
Supplied by Ultrasystems Corporation
Supplied by Gulf Research and Development Company
Supplied by U.S. Bureau of Mines, Laramie, Wyoming
e'Supplied by EPA (IERL-RTP)
'Results From Table 17 of Ref. 2 for comparison, 20 cc/min helium
flowrate
G4
-------
TABLE II. HCN YIELDS FROM INERT PYROLYSIS
OF COALS AT 1100 Ca
1.
1 1.
III.
IV.
V.
VI.
Coal
Pittsburg
No. 8C
Montana Liqnite-A
(hvA-b)d
Montana Coal
COED Char9
Bi Luminous
Bi turn! nous
,h
Coal6'9
Coalf'9
% Nb
1.2
0.61
0.93
0-99
1.17
1.8
Ugm
5
2
k
0
6
8
HCN/mg
-5
.8
.8
3
.8
.5
HCN
23
23
26
1
y
.6
.8
.8
.7
30
2k
iel
±1
±1
±0
to
.0
.5
d, I
.5
.6
.3
.1
a.,..
b.
c.
d.
e.
f.
Single-Stage Reactor, kQ cc/min helium carrier gas
As-received basis
Supplted by Aerotherm/Acurex
Supplied by MIT
Supplied by EPA (IERL-RTP)
Supplied by International Flame Research Foundation
^"Results from Ref. 2 for comparison, 20 cc/min
'Supplied by D. Pershing, University of Arizona
65
-------
TAJUJ-; III. HCN YIELDS FROM THE TWO-STAGE PYROLYSIS
OF A NO. 6 FUUL OIL3
(Atmospheric Pressure, Helium at 40 cc/min)
First-Stage
Temperature,
C
Second-Stage Temperature, C
1100 950 750
HCN Yield, %
1100
950
750
46
40
38
.2
.8
.5
±1
±1
±2
-9
.2
.0
41
23
23
.9
-5
-7
±1
±0
±0
.3
.7
.4
34.
13-
4.
8 ±0.
6 ±1.
4 ±0.
8
7
7
aOf1 A in Table I.
66
-------
TABLE IV. RESULTS OF TWO-STAGE INERT PYROLYSIS OF A
PITTSBURGH NO. 8 SEAM COAL (1.2%-N)a
Temperature, C
First Stage Second Stage
1108
745
600
747
1100
1098
1 100
7^9
HCN
23
22
17
7
Yield, %
.6 ±1.5
.6 ±0.3
.6 ±1.3
.2 ±0.7
Coal I in Table I I
TABLE V. TWO-STAGE 1'YROLYSIS OF COALS
(Low-Temperature First Stage)
Coal
Temperature, C HCN Yield, %
First Stage Second Stage Two Stage Single Stage^
1. Pittsburgh
No. 8C (UZfc-N)
V. Bituminous
(1.17S-N)
VI . Biluminous
(I.8Z-N)
IV. COED Char
600
598
600
604
1100
1099
1098
1097
17-6 ±1.3
11 .0 ±1.0
12.7 ±0.7
23-6
30.0
24.5
1.7
See Table I I
From Table II for comparison
From Table IV for comparison
-------
CO 0
u
CO
PM
00
£4 r-t
W 00
O
<0
(0
csj
oo
r-i
-H
oo
-H
-H
CM
CNI
O
OO
CJ
Cvj
CN
O
-------
CO
EH
rJ
s
CO
3
1
CO C
co e
fj U
o u
erf
^ O
pu -d-
f^ «*
^ 0
M »->
H n)
Q -H
l-l
o aT
60
M 4J
O CO
tJ OJ
W rH
f~i hft
E a
H
CO
S.X
M
M
~>
a
5
H
c 2
~ l"~
> r~"~
Q. ^^
f0
O
1
O . *
^
"*""
3 00
MD C3
' '
6
z
TO
O /-^.
z
(U <3^>
U. r~
vO O
6
z
<*'
._
*"
-r
"Z.
*>
*
"O
^
0)
z
CJ
IE
C
.£)
"i3 £
CM LA CM
r-%
CM CM CM
~ CM fA JT
OvOCM CMCT\ O -H- -3-f~.cn
cMrA
CM CM CM
i/l
0)
3
T3 T3 T3 -Q
C C C C E
O O O O
u o o o co
(/» o
~ CO vD O
-a- r~-
O CM CM CM CM
O O O O
o p <>e ,K>
o o o o o
CM CM CM CM
<-> O «-J O O
O O O O O
O O LA O O
r-. l~- vX> vo LA
I/)
c
0
4~f
c
o
u
0)
in
.c
m
>s
5>s
^
(U
>
IM-
w
0
0)
Q.
(U
L.
"
tr\
c
<0
c
ro
«~
4-1
V)
I/I
(U
*
0)
(U
CM
0
Z
o
c
(U
0
Q. Z
e
ro 14
in O
« E -0
.Q cri 0)
ro .
h- >-
(U U (U
OJ o .c
(0 -O O
69
-------
n)
0)
M
,-H
03
U
C/3
Cfi
a.
a-
H >>
(0 W
>>-H
^H N
o m
M rH
M
r-t O
gj u-(
to "O
SI
nj o.
o- w
OD
70
-------
8
o
s!
CO
fl
CO
S
O
M
S
J-J_LLi_LJ J_L
g§o 8 J2 o « H
EC O
oo
-------
*
a
UJ
750 960 1100
SECOND-STAGE TEMPERATURE, °C
A. LINEAR PLOT
Q
1100 960 760
SECOND-STAGE TEMPERATURE, °C
B. ARRHENIUS PLOT
Figure 3. Two-Stage Inert Pyrolysis of a No. 6 Fuel Oil
SO
>
I
40
30
I I I I
I
780 950 1100
SECOND STAGE TEMPERATURE. °C
A. LINEAR PLOT
O
50
40
30
20
15
10
I
IST-1100°C
1100 960 750
SECOND-STAGE TEMPERATURE, °C
B, ARRHENIUS PLOT
Figure 4. Two-Stage Inert Pyrolysis of a No. 6 Fuel Oil
With First Stage at 1100 C
72
-------
to
ZJ
Q.
CL
W
H
CO
cd
"O
"O
I
o
to
o
o
o
a)
M
a
H
73
-------
100
90
80
3?
§ 70
o
I 60
850
5
Q 40
ui
§ 30
cc
z
3 20
10
600
PYRIOINE
REACT ANT AT 1.4 m %
-
-------
0 I
*
to
o
o
UJ
4J
-H O
u at
-
PL, »-'
00
H
75
-------
(0
4-1
C
M
a)
8
00
0°
UJ
fiC
D
QC
UJ
Q.
5
UJ
8
G
O
H
4J
0)
U
e
M
60
H
fe
in
CN
10 O
%'cn3iAionaoad
76
-------
O4o
i
u
o
o
o
ce
o_
C3
z
g
i
O
%'anatAiDnaoad
o
o
01
OC
-
IU
i
o
o
I
C9
O
CC.
01
t-H
H
a
3
a
» O
H N
O C
eu
'a
ai a>
> *J
H U
±J Cfl
(0 CU
o pd
a -o
O 0)
M (0
Fh (0
PQ
CO
4J CO
O T3
3 f-(
O 0)
O i-f
^ !«
Oi *-*
01
M
§»
H
xanaoud
77
-------
e
o
(9
t-i
jj
G
(U
u
a
o
o
3
>.<
-------
FATE OF FUEL NITROGEN DURING PYROLYSIS
AND OXIDATION
By:
Y. H. Song, J. M. Beer and A. F. Sarofim
Massachusetts Institute of Technology
Cambridge, Massachusetts 02139
79
-------
-------
SUMMARY
Regulation of nitric oxide emissions by modification of coal combustion
processes requires a better understanding of the mechanism of fuel nitrogen
conversion to nitrogen oxides. To this end, the fate of coal nitrogen in both
inert and oxidizing atmospheres has been studied in a laboratory furnace at
temperature levels and for time intervals of interest in pulverized coal com-
bustors.
Data on the overall conversion of coal-nitrogen into nitrogen oxides
obtained at several different temperatures are reported. Oxidation experi-
ments were carried out with both coal and char to enable the respective con-
tributions of coal volatiles and char to the emissions of nitrogen oxides to
be determined. The char used in the oxidation studies was prepared by py-
rolysis of coal at the temperature and residence time of the corresponding
oxidation experiments. The conversion to nitric oxide of char-nitrogen was
lower than that of coal-nitrogen. The yields of nitric oxide from both coal-
and char-oxidation were found to decrease monotonically with increasing
fuel/oxygen ratios. Results also show that the conversion of fuel nitrogen to
nitrogen oxides decreases slightly with increasing temperature.
Rates of oxidation of char and char nitrogen were determined at
temperatures of 1250 to 1750 K and oxygen partial pressures of 0.2 and 0.4
atmosphere. The char in these latter studies was produced by pyrolyzing
coal particles at 1750 K during a residence time of one second. The variation
of nitrogen/carbon ratio of the char samples with time suggests that oxidation
and pyrolysis of the char nitrogen occur simultaneously.
81
-------
-------
INTRODUCTION
In staged combustion, the conversion of fuel nitrogen to nitrogen oxides
is reduced by delaying the addition of the oxygen required to complete the
combustion until after the fuel nitrogen has reacted (1,2). In this manner the
fuel nitrogen oxidation occurs under fuel rich conditions which favor the forma-
tion molecular nitrogen. For coal the problem is complicated by the fact that
the release of fuel nitrogen is relatively slow and is a strong function of
temperature (3,4). In order to provide an understanding of the effect of this
slow release of fuel nitrogen on nitric oxide emissions from coal combustors
and to guide the further developments of control strategies, the fate of coal
nitrogen during pyrolysis and oxidation has been studied under well controlled
conditions in a laboratory furnace. The study has included the determination
of the effect of fuel/air equivalence ratio and temperature on nitric oxide
emissions under simulated combustion conditions, the identification of the con-
tributions of the char and volatiles to the total emission, and time-resolved
measurements of coal pyrolysis and char oxidation in order to provide data on
the kinetics of these processes.
Results on the pyrolysis kinetics and some of the oxidation experiments
have been reported previously (4,5,6). In this paper, data on the effect of
temperature on the contribution of volatiles and char to NO emission and on
the rate of oxidation of char nitrogen are presented.
KXPKR TMENTAT.
A schematic of the experimental apparatus is shown in Fig. 1. It consists
of an Astro Model LOOOA electrically heated furnace with an alumina muffle tube,
A preheated inert or oxidizing gas stream is fed to the working section of the
furnace through a honeycomb straightener. The coal or char is injected along
83
-------
the ax in, and the products are withdrawn through a water-cooled probe the
position of which could be suitably adjusted to vary residence time. The
reactions are quenched by injection of argon at the mouth of the probe and
the solid products are collected in a sintered bronze filter. The equipment
can also be operated in a "free-fall" mode where the probe is removed and a
water-cooled section added to the bottom of the furnace. In this mode of
operation, particles are allowed to fall through the cooled section onto a
bronze disk collector at the bottom of the furnace. For the production of
char for use in oxidation experiments, a minor modification is made to the
free-fall mode of operation to permit the collection of the char in a flask.
The maximum residence time in the hot zone of the furnace is about one
second. Details of the apparatus are provided in Refs. 4 and 7.
The properties of the coals and chars used are summarized in Table I.
RESULTS
1. Conversion of Coal Nitrogen to Nitric Oxide
The percentages of the nitrogen in a Montana lignite that are converted
to nitric oxide for a hot zone residence time of one second as a function of
fuel/air ratio are shown in Figs. 2 and 3 for furnace temperatures of 1250 K
and 1750 K, respectively. The coal particles used have a nominal size of
38 to 44 microns. Also shown at the top of the figures are the total solid
weight loss (on a dry-ash-free basis) and, in the middle, the fraction of
the original coal nitrogen retained by any unburned char. Oxidation experi-
ments were also carried out on char in order to identify the separate con-
tributions to the nitric oxide emissions of the volatiles and char. The
char used in the oxidation studies was prepared by pyrolysis of the coal at
the temperature and residence time corresponding to those of the oxidation
experiments. Results of char-nitrogen conversions at 1250 K and 1750 K are
presented in Figs. 4 and 5. As can be ween from Figs. 2 to 5, the conversion
to nitric oxide of char-nitrogen was lower than the corresponding values for
foal-nftrogen wh.Ilo both coal- and rluir-nit rogen conversion to nitric oxide
w«*ro found to decrease monotonically with increasing fuel/oxygen equivalence
ratios. In both cases of coal and char oxidations, the percentages of solid
weight loss decreased and the percentages of the nitrogen retained in the
unburned char increased as the equivalence ratio Increased. The effect of
84
-------
temperature on fuel nitrogen oxidation is small and the effect may be masked
by day-to-day variations in operating conditions. In order to critically
test the effect of temperature on the conversion of coal-nitrogen to nitric
oxide, a series of experiments were conducted in which the fuel/oxygen ratios
were fixed and the furnace temperature was varied. The data in Fig. 6 were
obtained for a Montana sub-bituminous coal in this manner. The nitric oxide
yields were essentially the same at 1250 K and 1500 K while they decreased
by about 20 to 25 percent as the temperature was increased from 1500 to
1750 K. This is consistent with results shown in Figs. 2 and 3. Comparison
of the data in Fig. 6 with those in Figs, 2 and 3 show very little differ-
ences even though differenc coals are involved. This is consistent with
the findings of Pershing (8) and results previously obtained in this labor-
atory.
2. Kinetics of Char Oxidation
Time-resolved measurements of char-nitrogen oxidation were made by the
rapid quenching of samples taken at different distances from the char in-
jector. The char in these latter studies was produced by pyrolyzing lignite
particles at 1750 K for a residence time of one second. The char produced
was then size-graded to select particles in the 38 to 44 micron size range
for further oxidation experiments. Oxidation experiments were carried out
at temperatures of 1250 K, 1500 K, and 1750 K, and oxygen partial pressures
of 0.2 atm and 0.4 atm. The weight loss of char particles as a function of
distance between the feeder and the collector is presented in Fig. 7.
Selected char samples were sent to Galbraith Laboratory, Inc. of Knoxville,
Tennessee for elemental analyses. The variation of nitrogen/carbon ratio,
expressed in terms of the percentage of the original N/C ratio in the char,
is presented as a function of distance from the feeder in Fig. 8. The
nitrogen/carbon ratios were found to decrease with increasing temperatures
nnd/or with decreasing oxygen partial pressures. At high oxygen pressures
nnd tow temperatures, the; nitrogen AM rbon ratio remains constant at its
Initial value.
DISCUSSION
t Nitric. Oxide
A simplified scheme that represents the fate of the fuel nitrogen during
combustion of coal has been suggested previously(4,5,6) , and is summarized
85
-------
below:
1-Y
Unburned
Char
x
N_ + Others
By a first order approximation which assumes that the volatile-nitrogen
conversion to nitric oxide and char-nitrogen conversion are independent, the
overall conversion n* of coal-nitrogen to nitric oxide
overall
conversion
to NO
an H
volatile
contri-
bution
(1 - ct) YH2
char
contribution
where
a « fraction of the coal-N that is released as volatiles
1 - a = fraction of the coal-N that is retained in the char
Y - fraction of the char-N that is consumed
YH0 = fraction of the char-N that is converted to NO
2 x
H1 = fraction of the volatile-N that is converted to NO
.I. X
The values of 1 ~ a have been found to be 0.75 at a temperature of 1250 K
and 0.27 at 1750 K, in good agreement with the values previously reported by
Pohl (4,6). The values of n* are reported as functions of fuel/oxygen equiva-
lence ratio in Figs. 2 and 3 and the values of Yn2 in Figs. 4 and 5. The
values of r^, the volatile-nitrogen conversion efficiency, and anl , the percen-
tage of the volatile-nitrogen contribution to nitric oxide, cannnow be derived
06
-------
from the above in formatJon, The results of this derivation for temperatures
of 1750 K and 1250 K, are shown in Fig. 9. The efficiency of conversion to
nitric oxide of volatile nitrogen and the contribution to the total emissions
of the volatile nitrogen were found to decrease monotonically with increasing
fuel equivalence ratios. As expected, the contribution to nitric oxide by
the volatiles increases with temperature. The conversion to nitric oxide
of the volatile nitrogen, however, decreases with increasing temperature.
These two effects tend to compensate and partially explain the small depen-
dence on temperature of the conversion of coal nitrogen. In this study, a
small net decrease is observed. In a separate study, Pershing (8) has found
practically no effect of temperature on the conversion of fuel nitrogen to
nitric oxide. The present results suggest that it is optimal to operate the
first stage of a combustor at high temperatures since this will favor the
completion of the devolatilization under fuel-rich conditions.
2. Kinetics of Char Oxidation
Figs. 2 and 3 demonstrate that with increasing fuel equivalence ratios,
the percent conversion of coal nitrogen to nitric oxide decreases. Under the
same fuel-rich conditions, however, an increasing amount of the coal nitrogen
may escape the first stage of the combustor with the unburned char. The main
objective of the study of the kinetics of oxidation of char was to establish
the residence time requirements for completing the combustion of the char and
to determine the fate of the char-nitrogen during oxidation. The data in
Figs. 7 and 8 are plotted as a function of distance. For determining kinetic
parameters, residence time measurements are necessary. Such measurements
were made by Kobayashi (7), who used essentially the same system and showed
that the average particle velocity may be approximated by 1.4 times the
average main gas velocity. For the Interpretation of results of the present
study, distances wore converted to residence times, using Kobayashi's approx-
imation. The results with times reported in milliseconds are summarized
l>cl.ow:
Temperature, K
1250
1500
1750
nice1., inch
<"~" -^
3
256
213
183
4
340
283
243
87
5
427
356
305
6
512
426
365
7
599
500
428
-------
It has been previously shown (4) that during coal pyrolysis the carbon
loss appears to become asymptotic at times exceeding 100 milliseonds at
1750 K, the temperature used to produce the char in this study, but that
nitrogn release continues for longer times. The rationalization of the
results was that while carbon is present in relatively stable compounds in
the char, no comparable stabilized nitrogen structures are formed; conse-
quently, char-nitrogen continues to be released until it is completely
eliminated from char. Therefore, when a char particle is oxidized, the
nitrogen loss will be due to both devolatilization and oxidation, but the
carbon loss will'be exclusively due to oxidation.
Assuming the fuel nitrogen is uniformly distributed throughout the
char, the oxidation rate of the nitrogen should therefore equal the product
of the oxidation rate of the carbon and the mole ratio of the fuel nitrogen
to carbon in the char (9). Since the pyrolysis loss and oxidation loss of
char-nitrogen are additive, the consumption rate of char-nitrogen can be
written as
total consumption-rate f
of char-nitrogen
consumption-rate of
char-nitrogen due
to pyrolysis
or
since
dt
ft)
pyrolysin
dC = /dC\
dt \dt)
oxidation
consumption-rate of
char-nitrogen due
to oxidation
oxidation
88
-------
therefore,
d_N
dt
pyrolysis
(E)
oxidation
following pyrolysis
This model indicates that for low oxygen pressure and high temperature char
oxidation, the nitrogen to carbon ratio in the char will decrease as a
function of reaction time. As the oxygen pressure increases and temperature
decreases, the oxidation process tends to compensate the depreciation of
nitrogen/carbon ratio due to pyrolysis and keeps this ratio more or less
constant during the course of the combustion process. The experimental
findings support this conclusion. The data will be used to derive kinetic
parameters for the oxidation and pyrolysis of the char nitrogen. The exper-
iments were carried out with large excess of oxygen so as to maintain a
constant oxygen partial pressure and thus facilitate interpretation of the
results. The basic kinetic parameters derived from the data can then be
used to analyze char burnout under the varying oxygen concentrations that
will be encountered in practical combustors.
CONCLUSIONS
The data on coal and char oxidation show that the devolatllized nitrogen
compounds account for the major fraction of the nitric oxide produced from
coal nitrogen at high temperatures and low fuel/air ratios, but that the char
nitrogen contribution cannot be neglected. Increasing temperature increases
the volatile contribution but decreases the efficiency with which it is con-
verted to NO . For the combustor in this study, there is a net decrease in
X
emissions with increases in temperature above 1500 K, suggesting that NO
omissions can be reduced by increases In both the temperature and the fuel/air
ratio In the first stage of n staged combustor. Oxidation experiments on
chars support the view that there in no selectivity between nitrogen and
carbon Loss during oxidation but that the char nitrogen may undergo pyrolysis
in parallel with the oxidation. The time-resolved measurements may be used
to infer kinetic data needed to size the first stage of a staged combustor.
89
-------
ACKNOWLEDGEMENT
The authors wish to thank Dr. John H. Pohl and Dr. Joel M. Levy
i
for valuable discussions, and Mr. Gerald Mandel and Mr. Anthony J.
Modestino for their technical assistance. The work was supported by
the EPA under grant number R-803242.
REFERENCES
1. Pershing, D.W., Martin, G.B. and Berkau, K.E., Influence of Design
Variables on the Production of Thermal and Fuel NO from Residual
Oil and Coal Combustion, A.I.Ch.E. Symposium Series No. 148, 71 (1975).
2. Armento, W.J. and Sage, W.L., Effect of Design and Operation Variables
on NOX Formation in Coal Fired Furnaces: Status Report, A.I.Ch.E.
Symposium Series No. 148, 71 (1975).
3. Axworthy, A.E., Chemistry and Kinetics of Fuel Nitrogen Conversion to
Nitric Oxide. A.I.Ch.E. Symposium Series No. 148, 71 (1975).
4. Pohl, J.H., Fate of Coal Nitrogen. Sc.D. Thesis, M.I.T. (1976).
5. Pohl, J.H. and Sarofim, A.F., Fate of Coal Nitrogen During Pyrolysis and
Oxidation, paper presented at EPA Symposium on Stationary Source
Combustion, Atlanta, GA (1975).
6. Pohl, J.H. and Sarofim, A.F., Devolatilization and Oxidation of Coal
Nitrogen, paper presented at Sixteenth International Symposium on
Combustion, Cambridge, Mass. (1976).
i
7. Kobayashi, H., Devolatilization of Pulverized Coal, Ph.D. Thesis,
M.I.T. (1976).
8. Pershing, D.W., Nitrogen Oxide Formation in Pulverized Coal flames, Ph.D.
Thesis, University of Arizona (1976).
9. Wendt, J.O.L. and Schulze, O.K., On the Fateof Fuel NitrogenDuring
Coal Char Combustion, A.I.Ch.E. Journal, 22, 102 (1976).
90
-------
CO
u
s
3
^
o
cj
o
z
o
M
^J
£j
H
g
13
^2
3
*TJ
O
M
W
pQ
«<£
£**
H
o >«
X>
cn
O
H
4-1
m
fj
r»
<4H
H
U
01
(X
on co ^o co
-st m CM o o
O vO -* CM rH
*^ CN O> o
oo m o O o
*
r~^ (j\ CM oo O1*
j i j
^^ ^n ^1
O **** O f^J O
\T> CM CS O rH
* * * W
m r-H - > > 60
Cfl O -H (U
C/3 P^ Qfj p^
ft
-l (-1
H C fd cfl CO
&-d x: x! x:
a u c_> u
H 3
KJ 4J cu cu cu
iH 4J 4-1 4J
SCO ,O 'H iH -H
C 1 Ul 6 &i C3 $4 pl
crj CO X) C>0 bO 00
W 4-13 O-rl O'H O-H
c! tico in»_3 in*-J in^
O O CNl ^**~ 1*"^
& '£ rH iH iH
pj
0
H
4-i
CO
^4
0
&
o
X
O)
M
3
0
-------
Main Gos
from Preheater
Alumina Honeycomb
Flow Strolghtener
Power and
Water
Quartz
Observation
Port
Sintered
Bronze Filter
Water Cooled
Flanges
Boll Valve
Char and Carrier Gas
Thermocouple
^
/ .Alumina Muffle Tube
Cooled Flanges
Power and
Water
Graphite
Heating Element
Water Cooled
Collecter
Boll Valve
'- Water
:-- ' Quenching Gas
' Vocuum
Figure 1. Schematic of Experimental Furnace
92
-------
O n
1 r
O
j_
I \ I I I T
o o o
0)
E _
QJ O
I O
o
o
93
-------
]
1
I I \ \ I T
o o o
cp sr CM
O O
00 (0
JDp % "I
sscn *M PMOS N ;o
O
in
5 1
O CO
{- ro
Is
U
c c
i~ CD
1 ] 1
CM
o o o o o o o
h» CD i/) -sr co CM «-
XQN
|DOO
o
o:
u
c
o
cr
X
O
,
O ,0
a
^
a
£
c y
o> a
en -H
^ S5
3
U4
O
en
c
60
^ a
U-l
? I
o
to
I
a
i
a
5
M
4J
4» 95
O
4J
00
41
3
0)
E
94
-------
I 1
i i i i i i r
O O
CO CO
O O O
CD ^ CM
Ul
o
in
eg
55
O
QJ
0)
a ._
»
4)
OJ
Q>
i i i i i i i
CM
O
a
*
X
O
O J=
H U
to
4) 0)
> C
§ y
0 1
*
c . ^
O ,0
H
u C
(0 O
o 1-1
H
V 4-1
ao o
u -^
M
a
a. Q
o
H -a
V D
N
>. -H
y -o
a 1-1
41 X
H O
u
4-1 R)
W
o
a a)
O N
*« >.
(0
3
4) CJ
oo
H
UH i-i 4) 03
O t-> r-l C
H T3 «B
4) 2 TJ *J
U T4 C
flt O £ O
f»< <-> *~f £
60
o o o o o o o
Is* to in ^ en CM *-
N
95
-------
I
I I
O
in
(U (D
a ^
^ CT
go
i. in
o
oc
o
0)
M
g
u
t-l
nJ
o
o ^
X
O
(D
3
U.
ff
H
I
M
H
<0
6
o
-------
00 O CM
* * *
o *-
41 II »
u
O O O
in o tn
co if) f*.
O
o
oo
O
o
(O
o
o
o
o
CM
w
t.
3
*/
O
t_
c»
Q.
6
oq
o
o
c
"o
> ^
ui
c -?
w t;
O) O
>, a
x
O
I?
^ g
o «
e
o
o
u
§
0)
M
§
H
I
U
H
hi
SEi 4
o 3
u
-------
1 I I I
.c
u
OD £
L
O
*-
U
0>
CO
C\J
O
U
0)
TJ
*-)
-------
i i i r
o o o
in o ^
CM if) r->
O
1/1
0.
O
o o o o o o o
^ CM O 00 CD "t CU
00
JC
u
o
*-
u
<1>
5
u
(U
D
0)
-------
c
o
*-
3
JO
c
o
0)
§
O
T3
QJ
0)
O
u
o o
9O
80
70
6O
50
4O
30
20
1O
90
80
7O
60
5O
4O
30
20
10
1 2
Fuel / Oxygen Equivalence Ratio
Figure 9. Contribution of Volatiles to NO Emission:
Total N0x Contributed by Volatiles (Top), and
Conversion of Volatiles to NO (Bottom).
X
100
-------
INTERACTIONS BETWEEN SULFUR OXIDES
AND
NITROGEN OXIDES IN COMBUSTION PROCESSES
By:
J. 0. L. Wendt, T. L. Corley, and J. T. Morcomb
University of Arizona
Tucscon, Arizona 85721
101
-------
-------
ABSTRACT
Turbulent diffusion flame studies on gas and oil show that the
presence of fuel .sulfur could either inhibit, enhance or have no
effect on both thermal and fuel NO , depending on flame aerody-
namics and air preheat. Laboratory flat flames studies showed
that under fuel rich conditions, the presence of fuel sulfur
could have a radical effect on fuel NO formation mechanisms.
i
Both enhancement and apparent inhibition of fuel NO were observed,
depending on the stoichiometric ratio, and the residence time
at which the measurement was made. The results can be interpre-
ted in the light of current theories on fuel NO formation
mechanisms.
103
-------
-------
INTRODUCTION
Chemically bound nitrogen in fossil fuel has been shown to be an
important contributor to nitrogen oxide emissions from combustion
processes. Recent research has focussed on mechanisms governing
the oxidation of fuel nitrogen to fuel NO (Merely, 1976; Haynes,
1977), in the hope that the insight and understanding gained,
thereby, will lead to development of new combustion modifications
for NO control. However, many fuels containing chemically bound
X
nitrogen also contain sulfur, in various amounts. Indeed, in so
far as pulverized coal is concerned, a wide variation in sulfur
content is possible, even from coals with similar nitrogen contents.
The question can, therefore, be posed whether the sulfur content of
a fuel is likely to have a major influence on the resulting NO
emissions and specifically whether the presence of fuel sulfur will
cause major changes in mechanisms of fuel NO formation. Furthermore,
since there exists a wide variety of possible combustion conditions
and combustion modifications, it is to be expected that the poten-
tial importance of SO /NO interactions depends not only on the
X X
fuel quality, but also on the conditions under which the fuel is
burned.
This project was divided into two phases. Phase 1 consisted of
screening experiments on turbulent diffusion flames. The purpose
of this phase was to determine whether fuel sulfur had an important
effect on NO emissions In practical combustion systems. Phase 2
105
-------
continued the investigation of SO /NO interactions, but in a pre-
X X
mixed flat flame environment. The purpose of this phase was to
determine in more detail the conditions under which these inter-
actions occurred. Results in Phase 2 would help generalize the
data obtained in Phase 1 and would help determine how to relate pre-
mixed combustion data to the environment controlling pollutant
formation in turbulent diffusion flames.
Although oxidation of chemically bound sulfur in the fuel can
result in significant sulfur oxide emissions, there is little or
no data concerning the pertinent mechanisms involved. The combustion
of H?S has been studied extensively (Levy and Merryman, 1965;
Merryman and Levy, 1967; Sachyan, Gershezon, and Nalbandyan, 1967).
There is good agreement on the mechanisms involved in forming S0?
from H2S, and on the sulfur species present at different
stages of the combustion process. The limited data available on
combustion of organic sulfur compounds indicate the presence of
sulfur species identified in the combustion of H?S (Cullis and
Mulcahey, 1972). Due to the general lack of data, it would appear
necessary to treat bound sulfur in the fuel as if it were H?S to
analyze the possible interactions with NO formation. Furthermore,
there is evidence (Cullis and Mulcahey, 1972) that sulfur species
are cqul.'Llbrlatcd under fuel rich conditions. Intermediate products
are SO, S and SH. Addition of S02 to a fuel to simulate sulfur, will
produce a distribution of intermediates not very different from those
106
-------
produced using H^S, especially in the post primary reaction zone.
The possibility of interactions between sulfur and nitrogen oxides
at low temperatures is not new (Wendt and Sternling, 1973). This
interaction results in the catalysis, by NO, of the oxidation of
SCL to S0_. At higher temperatures, typical of combustion condi-
tions, free radicals are produced in super-equilibrium concentrations.
Since free radicals are important in NO formation, and since it has
been shown (Durie, Johnson, and Smith, 1971) that sulfur dioxide
is an effective catalyst in reducing super-equilibrium free radical
concentrations, it is reasonable to expect that sulfur dioxide, and
possibly other fuel sulfur compounds, inhibit the formation of NO
in flames. Indeed, Wendt and Ekman (1975) found that the addition
of high levels of S0_ or H2S to premixed, flat, methane-air flames
inhibited thermal NO formation significantly. The inhibition effect
was attributed to the catalysis of super-equilibrium oxygen atom
recombination reactions by S0j.i and was in evidence for both prompt
NO and post flame NO. DeSoete (1975) on the other hand, found that
low levels of sulfur could enhance "prompt NO" formation in rich
hydrocarbon flames. It should be noted that here flame temperatures
were in excess of 2100 K. DeSoete attributed the observed enhance-
ment to mechanisms involving CH, N», HCN, SO^, that is, to interactions
between fuel sulfur and the Fenimore (1971) mechanism for prompt NO.
Clearly, some discrepancies between Wendtfs data and DeSoete's data
still remain to be resolved.
107
-------
DeSoete (1975) also, consistently found that fuel sulfur enhanced
fuel nitrogen (NHL, NO) conversion in argon/hydrocarbon/oxygen flames
under fuel rich conditions. At a stoichiojnetric ratio of 0.6 at
flame temperatures of 2220°K he found that 250ppm H.S in the inlet
mixture led to an increase of 70% in NO emissions. Under fuel lean
conditions, a slight inhibition was observed. Indications from
preliminary but incomplete studies conducted by Wendt and Ekmann
(1975) on fuel sulfur interactions with fuel NO formation were that,
although the effect is not significant under fuel-lean conditions,
it may, under fuel rich conditions, have a marked influence on the
rate at which fuel NO is formed.
The unanswered questions addressed in this paper are: 1) Does fuel
sulfur influence fuel and thermal NO emissions in turbulent diffusion
flames? 2) What is the nature and extent of fuel sulfur/fuel nitro-
gen interactions in premixed flat flames? Furthermore, the applicable
conditions to be examined in the premixed flat flame experiments were
determined by inferences drawn from the turbulent diffusion flame
data. In this sense, the pilot scale experimentation guided the
more fundamental laboratory scale research, which is the converse
of normal procedure. It should be noted that one essential difference
between turbulent diffusion flames and premixed flames is that any
fuel additive does not immediately come into contact with oxygen in
the former, while It does in the Intter.
108
-------
TURBULENT DIFFUSION FLAMES
Experimental Combustor
Swirling turbulent diffusion flames were supported on a small scale
80,000 Btu/hr burner in a 6" ID diameter combustor. Hardware details
can be found elsewhere (Pershing and Wendt, 1977, Pershing, 1976).
It is important to note that by proper adjustment of fuel and air
velocities, of volume fraction of secondary air given tangential
momentum (hereafter denoted by % swirl) , and of injector hole angle,
it was possible to sustain a wide variety of flame types, ranging
from axial Type I flames, with little primary recirculation, to
Type II and Type III flames with increasing intensity of primary
recirculation. The gas injector utilized allowed poorly mixed
turbulent diffusion flames (Type I) to be supported when gas flowed
through an axial port, but also supported more intensely mixed com-
pact turbulent diffusion flames (Type II) when the gas flowed
through radial porta. The total area of the six holes for the
radial case is equivalent to the area of the single hole axial
case. The oil injector used for doped distillate oil was water
cooled and utilized standard pressure atomizing Monarch nozzles
rated at O.Sgph at lOOp.s.i. Most tests were conducted with a
30 R solid cone nozzle. Additives could be bled into the fuel
stream (gas and oil), or Into the n1r stream without changing
total flow rate. Further details can be found elsewhere (Corley
1976). The sampling and analytical system provided continuous
monitoring of NO, N02> CO, C02, 02 and SO- and has been described
109
-------
elsewhere (Pershing and Wendt, 1977). Suffice it to say that S02
was shown not to significantly interfere with the chemiluminescent
NO readings. The diesel oil utilized contained 0.3% wt. sulfur and
147ppra wt. nitrogen, while the natural gas was rated at 1064 Btu/scf
and contained merely 0.07 grains/lOOscf sulfur. Additives used were,
for the natural gas studies, SO and NH-; for the distillate oil
studies, thiophene (C,H,S) and pyridine (C H N).
SO /Thermal NO Interactions
A
Fig. 1 shows results for natural gas and the radial injector at zero
secondary air preheat. This flame could be classified as Type II
and was highly mixed. S0? was injected first into the fuel only,
and then into the air only. In each case, fuel sulfur inhibited
thermal NO , and it made very little difference whether the additive
was injected into the fuel or air. This is consistent with the
premixed flame data of Wendt and Ekmann (1975), and indicates that
a) superequilibrium free radical concentrations are important for
this type flame, and b) mixing is rapid in so far as S0_/thermal HO
^ X
reactions are concerned. At 440 F, secondary air preheat (Fig. 2)
results show essentially the same, with the exception that mixing
does now become important at low excess air values, because there
the inhibition was less with the sulfur in the fuel than in the
air. One might also deduce that the Fenimore prompt NO mechanism
was not of paramount importance for this flame, since DeSoete's
enhancement, thereof, was not observed.
110
-------
Data for the axial injector (Type I flame) show that at low air
preheat values (Fig. 3), SCL added to the air has a larger inhibi-
tory effect then when added to the fuel. Clearly, mixing now
controls the importance of SO /NO interactions,
X X
Local sulfur concentrations will be higher on the fuel lean side
or fuel rich side of the reaction zone, depending on whether SCL
IH injected in the air or fuel. Therefore, one can conclude for
this type flame, that inhibition of thermal NO (probably through
X
reduction of superequilibrium oxygen atoms) occurs on the fuel lean
side of the local reaction zone.
At high air preheat (Fig. 4) the situation is different, in that,
although inhibition is observed when SO- is injected into the air,
an enhancement is observed when S0_ is injected into the fuel.
Since flame temperatures are now high, this would indicate that
the Fenimore prompt NO/SO interactions first observed by DeSoete
X
are of paramount importance in the axial flames. The practical
importance of Figs. 1 and 4 is that fuel sulfur will inhibit
Thermal NO from Type II and Type III flames, but will enhance
X
Thermal NO from Type I flames.
x "
It is of interest to note that only an inhibition of Thermal NO
x
was observed when 3.8% wt. S (as thlophene) was added to distillate
in
-------
oil, both for low and high air preheat values. In these cases,
however, Type II flames were stabilized, allowing rapid mixing.
Fuel NO - Fuel Sulfur Interactions
The addition of thiophene to a pyridine-distillate oil fuel mix-
ture was used to investigate the effect of fuel sulfur on fuel
NO emissions. For the majority of studies conducted, the fuel
sulfur level was 2.05 wt. % S, corresponding to about HOOpptn SO-
in the exhaust for complete conversion; the fuel nitrogen level
was .78 wt. % N, corresponding to approximately 960ppm NO in the
flue for complete conversion. These levels are representative of
fuel nitrogen contents of low-grade crudes or residual oils.
The oil firing rate was about 70,000 Btu/hr. The fuel was
mechanically atomized after passing through a water-cooled injector.
The same injector was used for all studies although the nozzle was
changed before each study. The flames produced were short, bushy,
and luminous (Type II). These flames were attached to the nozzle
about V above the orifice of the nozzle. For the majority of
studies, 30 R solid cone nozzles were used.
To insure that the observed effects on fuel NO were due to the
presence of fuel sulfur, and not to other transients introduced
by the addition of thiophene, efforts were made to ascertain that all
112
-------
other conditions remained constant. First, it was shown that
addition of benzene to a quinolene doped distillate oil did not
affect fuel NO emissions. Since thiophene is itself a fuel
and changes in the stoichiometric air requirements would be
expected when it was introduced into the pyridine-oil fuel
mixture, the following procedure was, therefore, employed in
these studies:
1. Baseline thermal NO data were taken over the excess air
range of interest without the presence of pyridine.
2. Pyridine was added to the fuel and excess air conditions
were varied over the range of interest.
3. Thiophene was added to the pyridine-oil. mixture and NO
measurements were made over a range of excess air conditions.
4. Thiophene addition was stopped and NO measurements were
again taken for the pyridine-oil mixture at selected excess
air conditions.
5. Finally, pyridine addition to the oil was stopped, and
thermal NO readings were taken to compare with the original
baseline data.
The most important problem encountered in the studies conducted with
//2 diesel oil was coke formation on the oil nozzles. The severity
of coke formation increased with air preheat, decreasing excess air
conditions, and when fuel nitrogen was present. A reduction in coking
was observed when thiophene was added to the pyridine-oil mixture.
113
-------
Since the presence of S02 or H?S in premixed and diffusion flames
has been found to reduce carbon formation (Cullis and Mulcahey,
1972), this is not unexpected.
Fig. 5 shows results of 2.05% sulfur addition to a distillate oil
containing 0.78% chemically bound nitrogen. No effect was observed.
Pig. 6, however, shows that at high secondary air preheat values,
fuel sulfur enhanced fuel NO conversion. It is important to note,
however, that not only had the combustion temperature increased,
but the baseline (0% S) fuel NO had decreased, indicating a decrease
in mixing intensity and the persistence of more fuel rich regions.
The observed enhancement of 40% due to sulfur might be due to either
of these effects. Subsequent experiments in which S0» was added
in the air showed that sulfur had to penetrate into fuel rich regions
in which some fuel NO was being formed, before enhancement of fuel
X
nitrogen conversion could be observed.
Indications were, therefore, that enhancement of fuel nitrogen
oxidation by fuel sulfur in turbulent diffusion flames would occur
under poorly mixed (or Type I) flame configurations and at high air
preheats. Experiments were performed with natural gas doped with
ammonia, using the axial injector. This gave a Type I flame,
and results (Fig. 7) clearly show a significant enhancement (40%)
of fuel NO emissions when SO- was added to fuel. Some enhancement
was also observed when S02 was added to the air.
174
-------
PREMIXED FLAT FLAMES
The turbulent diffusion flame results described above indicated
that fuel NO /SO interactions were important primarily in a fuel
X X
rich combustion region. Furthermore, DeSoete's work indicated
that enhancement could occur only under fuel rich conditions.
Therefore, in our premixed flat flame studies, we focussed on
fuel NO/fuel sulfur interactions under fuel rich conditions and
attempted to address the following critical questions:
1) What is the relationship between local stoichiometry and
temperature for the observed enhancement fco occur?
2) To what extent does the presence of fuel sulfur radically
change fuel nitrogen oxidation mechanisms?
3) Does the importance of these interactions under fuel rich
conditions indicate that fuel sulfur will have a major
influence on staged combustion configurations for fuel NO
X.
abatement from pulverized coal.
Experimental Burner
The premixed flat flame burner was similar to that utilized by
Axworthy et. al. (1976). It was nominally 5cm I.D. and allowed a
flat flame to be stabilized over a coarse stainless steel screen
(10 x 10 mesh, 0.028" wire). The burner was enclosed by a 15cm I.D.,
60cm high quartz chimney, the inside surfaces of which were purged
with nitrogen in order to prevent soot deposition. Details can be
found elsewhere (Morcomb, 1977). Temperature mecisurements were made
using the sodium T)-line reversal technique, with experimentally dbtained
115
-------
corrections for the glass chimney surfaces. Flows of major (0,,,
CH,, Ar) and minor species (C^N-, SO-, H S) were controlled and
metered with sonic orifices, rotameters and laminar flow elements
where appropriate.
Sampling was accomplished through both uncooled and cooled quartz
probes. NO analysis was by cherailuminescence and a Molybdenum
J\
converter. Teflon lines were used throughout. We were concerned
about reactions occurring at room temperature in the sampling line.
Special experiments, including some in which S0_ in the sample
line was removed at various points, showed that our NO measure-
ments were not due to an Interference between sulfur and any
potential NO forming species between the probe and the analyzer.
Both the hot and cooled probes withdrew the sample through a
1.5mm orifice, allowing a normal sampling rate of 1.3 liters/min.
Special tests on the hot quartz probe showed that when soot,
previously exposed to S0?, was present, the baseline (no sulfur)
NO was lowered by 7ppm, compared to results from the cooled
probe. The cooled probe did not allow the build up of soot, and
changing sampling rates by a factor of five did not change "exhaust"
ppm NO measured, both in the presence and absence of sulfur. The
cooled probe, however, caused the peak flame temperature of 1953 K
to drop 30°K when it was in the immediate vicinity. This interfe-*
rence became negligible 2cm away from the probe. Most of the
results reported below are with the cooled probe, except as noted.
116
-------
A chart recorder was used to observe transients and baseline
reproducibility. Changing or cleaning the porous disk and screen
on the burner had no effect on results.
Results;
Figs. 8 through 12 show the effects of various fuel sulfur levels
on fuel NO profiles at various stoichlometries. The total dry and
wet molar flow rates of the burned gas were calculated assuming
chemical equilibrium in the fuel rich mixtures. Residence times
and ppm NO corresponding to 100% conversion of fuel nitrogen were
calculated using these molar flow rates. In general, the sulfur
additive ranged from 0.2 through 3.8 mole percent of the fuel; the
nitrogen contents were 0.2 and 1.0 mole percent nitrogen in the
fuel. Specific details for each condition investigated can be
found in the figures and figure captions. The major species were
argon, methane and oxygen. In all cases, there was no evidence
of NOp, either with or without sulfur present, although it should
be cautioned that long term exposure to CO can affect the performance
of the Molybdenum converter.
At a stoichiometric ratio of 0.46 ( = 2.17) sulfur greater than
0.2 mole percent in fuel had a large qualitative and quantitative
effect on the entire* fuel NO profile (F;ig. 8). Open symbols denote
d/ita taken with the coo'U'.d probe, Hhml<;d symbols nre those t«iken
with the hot probe. The bnselliu? fucj.l NO rose quickly to a constant
lOppm. Addition of 3.8 mole percent sulfur dioxide* to the fuel
prompted a rapid Increase of fuel NO with a peak of 130ppm occurring
117
-------
downstream of the luminous zone of the flame. This high value
decayed slowly to 40ppm over a residence time greater than 0.1
seconds. Although sulfur does act as a combustion inhibitor, the
position of the luminous zone of the flame varied only slightly
within a few millimeters of the grid. However, it is important to
note that with sulfur present, NO is destroyed by a slow reaction
at low temperatures (less than 1600°K). This does not occur in
the absence of sulfur. The fact that the baseline NO does not
change over the same distance indicates that along the centerline
dilution is not significant.
Clearly, at very fuel rich conditions, the presence of sulfur
compounds has a marked influence on fuel NO formation mechanisms.
Fig. 9 shows results of a similar experiment but with HLS added
as the fuel sulfur additive. Qualitatively similar profiles were
obtained as with S02, but comparison of this figure with Fig. 8
indicates a less extreme effect. It should be noted that H S
addition does lower the stoichiometric ratio, thus, tending to
lower fuel nitrogen conversions. This is in contrast to DeSoete's
results which showed US to have a larger effect on prompt NO
than did S0_. However, our results indicate that it is sulfur,
rather than the Initial form of sulfur, which is responsible
for the large effect observed.
Fig. 10 shows results at a stoichiometric ratio of 0.61 (4> =
1.64). Again, sulfur (as SO,) added to the fuel radically changed
118
-------
fuel NO mechanisms. The peak now occurred closer to the luminous
zone in the flame, and a steady NO consumption brings the NO pro-
file below the baseline case. In this case, whether fuel sulfur
inhibits or enhances NO depends entirely on where the measurement
is taken. It should be noted that temperatures in Fig. 10 were
significantly hotter than at the richer case. Also, here H_S addi-
tion had a similar effect as did S0».
At higher stoichiometric ratios, sulfur has less of an effect (Fig. 11;
SR - 0.68), although the salient features observed before are present,
i.e. sulfur compounds cause a rise followed by a steady decline in
fuel NO.
Results for a fuel nitrogen content of 1 mole percent and a stoichio-
metric ratio of 0.46, are shown in Fig. 12. These data should be
compared to those in Fig. 8. The results are essentially analogous.
One important difference is that at the higher nitrogen content,
a significant reduction in baseline NO, without sulfur, was observed
close to the flame front. With sulfur, the NO profile again formed
a high peak followed by a slow decay.
Discussion
Previous work (Haynes, 1976), has indicated that the route to NO
from fuel nitrogen occurs through HCN, NCO and NH . Under fuel
X
rich conditions, SO- Is reduced to SO, H S, and SH, and these
species may well interact with NH to change the distribution
119
-------
among NH, NH? species. There exists evidence (Lyon and Longwell,
1976, Wendt et. al., 1973) that NH or its derivatives, will reduce
NO at moderately low temperatures. Note that at very low stoichio-
metric ratios, the equilibrium NHL concentration is greater than
the equilibrium NO. It might be conjectured, therefore, that reduced
sulfur species directly change the NH profiles in a flame, so that
the NH species responsible for NO reduction persists far beyond
A
the flame zone, thus, accounting for the steady NO reduction. This
conjecture might also explain why the early NO peaks are so high.
In the flame zone, the sulfur has delayed the formation of the NO
reducing species (Fenimore, 1976) until the post flame region is reached.
That the presence of sulfur allows NO profiles to decay sometimes
below the baseline case is not surprising, since equilibrium NO
values at SR= 0.61 are well below Ippm. In fact, given that sulfur
promotes this slow decay, it is not surprising that it can allow
measured values to be above or below the baseline values. It is
hard to rationalize that the slow NO reduction observed results
from reactions with CH (Myerson, 1975) or other hydrocarbon frag-
ments, since one would not expect them to persist so far from the
reaction zone.
Our results differ from those of DeSoete (1975) , in that he did
not observe NO reduction prompted by sulfur, although he did observe
significant enhancement of NH_ conversion due to the presence of H_S.
However, measurements made in this work were over much longer resi-
120
-------
dence times, and they involved both higher nitrogen and sulfur
levels, and lower flame temperatures. Furthermore, DeSoete's
experiments on the role of sulfur utilized NH« and NO as nitro-
genous additives, while this work utilized C_N?. The nitrogen
species involved may be important when sulfur is present.
It is interesting to note that the NO profiles measured here
in the presence of sulfur, are qualitatively similar to NO pro-
files in fuel rich pulverized coal combustion (Pershing et. al.,
1977), while the profiles without sulfur are not. The presence
of sulfur in coal may have a significant effect on NO abatement
by staged combustion, since one would expect sulfur levels in
the volatiles to be high.
121
-------
CONCLUSIONS
The presence of fuel sulfur can increase thermal NO emissions from
poorly mixed Type I flames, but decrease thermal NO emissions from
well mixed flames. Furthermore, fuel sulfur has a first order
influence on mechanisms governing the oxidation of fuel nitrogen
under fuel rich conditions. This can manifest itself in a signifi-
cant increase in NO emissions when a fuel containing chemically
X
bound nitrogen is burned with sulfur in an axial diffusion flame.
Fuel oils containing both nitrogen and sulfur are likely to give
higher NO emissions than those containing little sulfur, especially
at high temperatures.
Increased emphasis on NO abatement of dirty fuels through fuel
X
rich and staged combustion would appear to dictate that fuel nitro-
gen oxidation mechanisms in tbe presence of sulfur should be
investigated. The data reported herein lend support to the
hypothesis that the staged combustion of dirty fuels containing
sulfur may be less effective than that of fuels of low sulfur
content, although this may depend on the first stage stoichlometric
ratios and residence times involved. Clearly, fuel sulfur and
fuel nitrogen interactions are of first order Importance for many
practical combustion systems.
122
-------
ACKNOWLEDGEMENTS
The authors would like to express their appreciation to David W.
Pershing and Edward M. Morcomb, without whose help, some of the
data reported herein would not have been obtained. We would also
like to thank Michael P. Heap and W. S. Lanier for their valuable
technical input to this project.
123
-------
LITERATURE CITED
Axworthy, A. E., G. R. Schneider, M. D. Shuman and V. H. Dayan
"Chemistry of Fuel Nitrogen Conversion to Nitrogen Oxides
in Combustion", EPTS Report EPA-600/2-76-039, February,
1976.
Corley, T. L., "Fuel Sulfur Effects on NOX Formation in Turbulent
Diffusion Flames", M.S. Thesis, University of Arizona,
Department of Chemical Engineering, December, 1976.
Cullis, C. F. and M. F. R. Mulcahey, "The Kinetics of Combustion
of Gaseous Sulfur Compounds", Comb. Flame, 18 (1972).
DeSoete, G., "La Formation des Exyde D'Azote Dans La Zone d'Oxydation
des Flammes d'Hydrocarbures", Institut Francais du Petrole,
Final Report No. 23309, June, 1975. Ruiell Malmaison, France.
Durie, R. A., G. M. Johnson, and M. Y. Smith, "The Effect of Sulfur
Dioxide on Hydrogen-Atom Recombination in the Burnt Gas of
Premixed Fuel-Rich Propane-Oxygen-Nitrogen Flames", Comb.
Flame, 17_ (1971).
Fenimore, C. P., "Formation of Nitric Oxide in Premixed Hydrocarbon
Flames", Thirteenth Symposium (International) on Combustion,
The Combustion Institute, Pittsburgh, Pennsylvania (1971).
Fenimore, C. P., "Reactions of Fuel-Nitrogen in Rich Flame Gases",
Comb. Flame ^6_, 249 (1976).
Haynes, B. S., "Reactions of Ammonia and Nitric Oxide in the Burnt
Gases of Fuel-Rich Hydrocarbon-Air Flames", Comb. Flame,
2£, 81 (1977).
Levy, A. and E. L. Merryman, "The Microstructure of Hydrogen Sulfide
Flames", Comb. Flame, J9 (1965).
Lyon, R. K. and J. P. Longwell, "Selective, Non-Catalytic Reduction
of NOX by NH3", The Proceedings of the NOV Technology Seminar,
EPRI Special Report EPRI SR-39, February, 1976.
Merryman, E. L. and A. Levy, "Kinetics of Sulfur-Oxide Formation in
Flames. II Low Pressure ^S Flames", J. Air Poll. Control
Assoc. , 17. (1967).
Morley, C., "The Formation and Destruction of Hydrogen Cyanide from
Atmospheric and Fuel Nitrogen in Rich Atmospheric-Pressure
Flames", Comb. Flame, ^7, 187 (1976).
124
-------
Morcomb, T. J., "Interactions Between Fuel Sulfur and Fuel NOX
Formation Mechanisms", M.S. Thesis, University of Arizona,
Department of Chemical Engineering, December, 1977.
Myerson, A. L., "The Reduction of Nitric Oxide in Simulated
Combustion Effluents by Hydrocarbon-Oxygen Mixtures",
Fifteenth Symposium (International) on Combustion, p. 1085,
The Combustion Institute, Pittsburgh, Pennsylvania (1975).
Pershing, D. W., "Nitrogen Oxide Formation in Pulverized Coal
Flames", PhD. Dissertation, University of Arizona, Depart-
ment of Chemical Engineering, August, 1976.
Pershing, D. W. and J. 0. L. Wendt, "Pulverized Coal Combustion:
The Influence of Flame Temperature and Coal Composition on
Thermal and Fuel NOX", Sixteenth Symposium (International)
on Combustion. The Combustion Institute, Pittsburgh,
Pennsylvania (1977).
Pershing, D. W., J. W. Lee and J. 0. L. Wendt, "Fate of Coal
Nitrogen Under Fuel Rich and Staged Combustion Conditions",
Paper to be presented at 70th Annual AIChE Meeting, New York,
November, 1977.
Sachyan, G. A., Y. M. Gershenzon, and A. B. Nalbandyan, "Multistage
Nature of Combustion in Dilute Hydrogen Sulfide Flames",
Dokl. Aked. Nauk SSSR, 175 (1967)', [English version, 175
(1967)1 L
Wendt, J. 0. L. and C. V. Sternling, "Catalysis of S02 Oxidation
by Nitrogen Oxides", Comb. Flame, Zl (1973).
Wendt, J. 0. L., C. V. Sternling and M. A. Matovich, "Reduction
of Sulfur Trioxide and Nitrogen Oxides by Secondary Fuel
Injection", Fourteenth Symposium (International) on
Combustion, p. 897, The Combustion Institute, Pittsburgh,
Pennsylvania (1973).
Wendt, J. 0. L. and J. M. Ekmann, "Effect of Fuel Sulfur Species
on Nitrogen Oxide Emissions from Premixed Flames", Comb.
Flame 25_ (1975).
Wendt, J. 0. L. and J. M. Ekmann, "Effect of Fuel Sulfur on NOX
Emissions from Premixed Flames", EPTS Report EPA-600/2-
75-075, October, 1975.
125
-------
60
50
40
30
Q.
0.
20
10
O Base: Thermal NOx
n S02 in Gas {6.0 Wt % S;
LJ 3960 ppm S02 in flue)
OSOa in Air (4070 ppm
SOz in flue)
8 16 24 32
PERCENT EXCESS AIR
40
Figure 1: Turbulent Diffusion Flames: Natural Gas/Radial Injector,
Effect of S02 on Thermal NO (38% Swirl; 100°F Air Preheat)
126
-------
120
100
~ 80
x
p
o
z
a.
CL
60
40
20
O Base: Thermal NOx
SOz in Gas (4.86 Wt %S;
Q 3140 ppm S02 in flue)
OS02 in Air (3090 ppm
S02 in flue)
5 10 15 20
PERCENT EXCESS AIR
25
Figure 2: Turbulent Diffusion Flames: Natural Gas/Radial Injector,
Effect of S02 at High Air Preheat (42% Swirl; 440°F Air
Preheat)
127
-------
80
so
(O
o
Q.
CL
40
20
O Base: Thermal NO*
,-, S02 in Gas (6.15 Wt%S;
"-1 4080 ppm S02 in flue)
OS02 In Air (3160 ppm
influe)
5 10 15 20
PERCENT EXCESS AIR
25
Figure 3: Turbulent Diffusion Flames: Natural Gas/Axial Injector,
Effect of S02 on Thermal NO (37% Swirl; 120°F Air Preheat)
128
-------
120
100
g 80
CO
*m*
O
Z 60
a.
a.
40
20
O
D
O
Base: Thermal NOx
S02 in Gas (5.43 Wt%S;
3550 ppm S02 in flue)
S02 in Air (3060 ppm
302 in flue)
10
PERCENT
15
EXCESS
20
25
AIR
Figure 4: Turbulent Diffusion Flames: Natural Gas/Axial Injector,
Effect of SO, on Thermal NO at High Air Preheat (40%
Swirl; 440°F Air Preheat)
129
-------
1000
800
X
O 600
400
O.
£L
200
Pyridine (.776 Wt %N)j
Thiophene (2.05 Wt %S)
Pyridine (.783 Wt % N)
10 20 30 40
PERCENT EXCESS AIR
50
Figure 5: Turbulent Diffusion Flames: Doped Distillate Oil, Thiophene
Does Not Affect Fuel NO at Low Air Preheat (43% Swirl; 100°F
Air Preheat; 30°R Nozzle)
130
-------
600
500
400
x
o
O 300
z
a.
o_
200
100
O
Pyridtne (,783 Wt % N)
A Pyrtdlne (.776 Wt % N);
^ Thlophene (2.05 Wt % S)
8 16 24
PERCENT EXCESS AIR
32
40
Figure 6: Turbulent Diffusion Flames: Doped Distillate Oil, Thiophene
Increases Fuel NO at High Air Preheat (40% Swirl; 310°F Air
Preheat, 30°R Nozzle)
131
-------
500
400
X
o
<2
o
3°°
200
100
O NH3 (1.08 Wt % N)
A NH3 (.968 Wt% N); S02 in
Gas (5.19 Wt% S; 3380
ppm SOe in flue)
O NH3 (1.08 Wt % N); S02
in Air (1900 ppm SOg
In flue)
J_
12 18 24
PERCENT EXCESS AIR
30
Figure 7: Turbulent Diffusion Flames: Doped Natural Gas/Axial Injector
S02 Increases Fuel NO at High Air Preheat (44% Swirl; 430°F
Air Preheat)
132
-------
1900
ID
-------
e 1900
ul
(T
5
QC
UJ
O.
2
UJ
I-
O
UJ
o:
3
CO
<
UJ
2
O
o.
0.
1700
70
50
30
to
\
\
\
\
\
\
l/min ml/mln
CH4 6.08 H2S 0
02 5.62 13
Ar 8.92 110
228
C2N2 6.3
D
Q
O
8 12 16 20
HEIGHT, cm
24
28
Figure 9: Laminar Premixed Flame: Stoichiometrlc Ratio 0.46
(0.21Z fuel N, 100% conversions 553 ppm, lcm%7.67ms
at 1900°K)
134
-------
2100
u
(T
25 '900
OL
yj
300
260
3 22°
UJ
o:
O) 180
<
UJ
2
** 140
CL
0.
100
60
20
\
\
\
\
\
\
CH4
02
Ar
I/ min
6.08
7.40
24.3
ml/min
S02
H2S
C2N2
0
13
114
231
228
6.3
O
O
O
A
6
\
6 12 16 20 24 28
HEIGHT, cm
Figure 10: Laminar Premixed Flame: Stoichiometric Ratio 0.61
(0.21% fuel N, 100% conversion* 357 ppm, lcm«?4.47
at 2100°K) 135
-------
II
QC2000
'D
OS
UJ
0-
Ul
1800
260
I .80
Ul
«* 140
100
60
20
l/min ml/min
CH4 6.08
02 8.25
Ar 32.0
S02
H2S
C2N2
0 0
13 D
114 O
231 A
228 £»
6.3
8 12 16 20
HEIGHT, cm
24
28
Figure 11: Laminar Premlxed Flame: Stoichiometric Ratio 0.68
(0.21% fuel N, 100% conversion*^300 ppm, lcms»4.47
at 1980°K) 136
-------
250
230
220
190
170
O
LJ
(T 150
CO
O
Q.
0.
130
110
90
70
50
10
l/min ml/min
CH4 6.08 S02 0 O
02 5.62
Ar 10.92
8 12 16 20
HEIGHT, cm
24
28
Figure 12: Laminar Premixed Flame: Stolchiometric Ratio 0.46
(1.05% fuel N, 100% conversions 2793 Ppm, 1cm-ss 7
ms at 1970°K) ''
-------
-------
CHEMICAL REACTIONS IN THE CONVERSION OF FUEL NITROGEN TO NO,..
LOW-PRESSURE FLAT-FLAME BURNER STUDIES
By:
0. R. Kahn and A. E. Axworthy
Rockwell International/Rocketdyne Division
Canoga Park, California 91304
139
-------
-------
ABSTRACT
The objective of this study is to use a low-pressure, premixed, flat-flame
burner as a tool for investigating the mechanism and kinetics of the chemical
reactions that are involved in the formation of NO in the combustion of fossil
fuels or alternate fuels derived from fossil fuels. A detailed probing study
of a CH.-Oj-^ flame at 0.1 atm was conducted with the addition of 50 ppm of
MR. or HCN to the feed gas. These additives were used to simulate the expected
products of preflame reactions of the fuel nitrogen species that are contained
in coals and residual fuel oils (mainly heterocyclic aromatic compounds). An
experimental procedure was developed that permitted NO, NCL, NH,, and HCN to be
measured independently below, in, and above a low-pressure flame.
The methane flame experiments were carried out under fuel-rich ($ = 1.5) and
fuel-lean ( = 0.8) conditions at a diluent ratio of 0.7 with and without the
addition of NH or HCN. The formation of "thermal" NO from N (via both the
J X f.
Zeldovich and "prompt" paths) was compared directly with the formation of "fuel"
NO from the added NH. or HCN. In all cases, thermal (prompt) NO and fuel NO
X j (
formed very rapidly at approximately the same location just above the top of
the luminous zone, and the amounts formed were approximately additive. In fuel-
rich flames, both prompt NO and fuel NO appeared to form almost exclusively via
HCN as an intermediate.
In a similar study of NO formation in the fuel-rich flame of a simulated low-
Btu fuel (a mixture of hydrogen and carbon monoxide), NO formed from added NH.
in high yield but more than one-half of this NO was consumed just above the
luminous zone. Only 20% of added NO was consumed in the absence of added NH~.
141
-------
-------
INTRODUCTION
This work is part of an experimental and analytical research program which is
being conducted to provide information on fuel decomposition and flame reac--
tions required for the development of a complete understanding of the chemical
phenomena involved in the conversion of fuel-nitrogen to NO . The program
X
extends the fuel pyrolysis and flame studies previously carried out in this
laboratory (Ref. 1 and 2). Task 1 consists of experimental studies of the
types of chemical reactions fuel nitrogen species can undergo as the fuel
decomposes in the early (preflame) stages of combustion. Low-pressure,
flat-flame burner studies are being conducted under Task 2 to investigate
the mechanism and kinetics involved In the conversion of fuel-nitrogen
Intermediates (such as NH_ and HCN) to NO in the combustion of fossil fuels
J X
or alternate fuels derived from fossil fuels. Under Task 3, the combustion
data for the conversion of fuel nitrogen to NO are being analyzed in terms
X
of the chemical mechanisms and physical processes involved.
The present paper summarizes recent effort completed under Task 2, Combustion
Kinetics. Following a description of the flat-flame burner apparatus and a
brief discussion of the pertinent experimental techniques, the results obtained
in two series of detailed flame probing studies will be presented. The first
study involved the comparison of "fuel" NO and "thermal" NO formation in the
X X
CH.-O^-N^ flame system (0.1 atm) at fuel-rich (equivalence ratio, 4>, of 1.5) .
and fuel-lean (4> « 0.8) conditions using NH» and HCN as additives. The second
study is concerned with fuel NO formation in a simulated low-Btu gas flame
system of !!, CO, 02, and Ar (0.066 atm) at fuel-rich conditions (<)> « 1.4) to
which NH_ or NO has been added.
DESCRIPTION OF THE EXPERIMENTAL APPARATUS
The design of the flat-flame burner employed in the combustion kinetics exper-
iments is depicted in Fig. 1, and a schematic of the overall burner system is
shown in Fig. 2A.
The water-cooled, stainless-steel burner is 5.72 cm in diameter, and can
be positioned vertically with a precision of better than 0.01 mm with a
143
-------
micrometer-drLven translation stage. A stainless-steel sintered porous plate
3.2 mm (1/8 inch) thick cemented In a recess at the burner exit furnishes a
flat velocity profile across the port. An 8-mesh, stainless-steel screen spot-
welded to the top of the burner port serves as a flame holder. The burner
and igniter are enclosed in a Pyrex glass chamber so that a pressure of 0.1 atm
or less can be maintained to spread the flame and permit more detailed probing.
An uncooled quartz microprobe is used to remove gas samples from the flame
reaction zone along the centerline of the burner. The reacting gases are
f
quenched aerodynamically by expanding them from the reaction chamber pressure
(50 to 76 torr) to about 1 torr. The microprobe was constructed from 9 mm
quartz tubing drawn to a very small-diameter tip containing the sonic sampling
orifice. The diameter of the sonic orifice was between 75 and 100 microns,
while a taper of approximately 30 degrees was used to ensure minimum flow
disturbance occurring upstream of the probe as a result of gas sampling
(Ref. 3).
Temperature measurements were performed along the centerline of the flame with
a 3-mil Pt/Pt-10% Rh thermocouple coated with a 1- to 3-micron thick layer of
A120, using a high-vacuum sputtering technique. The thermocouple probe Itself
consisted of 10-mil thermocouple wire support arms of Pt and Pt-10% Rh strung
through a two-hole alumina tube held in place with Sauereiaen cement.
The inlet feed gases CH. or H2/CO, 0-, and N2 or Ar are metered with critical
flow orifices. The flow system is similar to that reported by Anderson and
Friedman (Ref. A). The fuel-nitrogen additives (NH-, HCN, and NO), and the
purge gas flowing along the annular region of the burner chamber, are metered
with Brooks flowmeters,
GAS ANALYSIS PROCEDURES
Gas samples withdrawn from the flame are analyzed for the mole fraction of
stable species using two different techniquesmass spectrometry and chemi-
luminescent analysis. The major stable species (CH,, CO, CO,,, 00, H0, N ,
*r &£.£*£
or Ar) are measured by mass spectrometry using a batch gas-sampling procedure.
144
-------
An on-line sampling technique using a Thermo Electron Corp. (TECO) Model 10A
chemiluminescent analyzer (CA) enables the nitrogen species NO, N0?, NH,,, and
HCN to be determined by converting them to NO (as described later). The CA was
modified to operate with on-line sampling by bypassing its capillary flow
metering system and using the sonic sampling orifice at the tip of the quartz
microprobe to regulate the gas sample flow to the CA reaction chamber. The
pressure in the CA reactor with no sample feed was reduced to 0.8 torr by low-
ering the ozone feed rate. With sample feed, the CA reactor pressure is main-
tained between 0.9 to 1.1 torr, depending on the diluent chosen and other flow
conditions. Under the conditions employed, the photon emission rate is directly
proportional to the product of the NO concentration and the sample flowrate.
Over the range of sample feed rates employed, the CA reaction chamber pressure
remains nearly constant. Because the sample flowrate is small, the composi-
tion of the sample does not have an appreciable effect on the sensitivity of
the CA.
Contained within the TECO CA is a stainless-steel catalytic converter for
converting N02, NH_, and HCN to NO. A small flow of 0. is added to the sample
just before it enters the TECO CA. This added Q is used in all experiments
in which fuel-rich conditions are employed. The purpose of the 0^ is to ensure
that the gas sample is overall oxidizing in composition to prevent catalytic
removal of NO in the stainless-steel converter. A modular molybdenum catalytic
NO-NO converter unit (and a temperature regulator) also manufactured by TECO
X
is used to convert only N02 to NO. At 800 C under overall oxidizing condi-
tions, the conversion of NH. -* NO in the stainless-steel converter is 100%
over the full range of concentrations employed in these studies (0 to 2800
ppm NH-). The conversion efficiency for HCN * NO at these same conditions is
dependent on the HCN concentration, decreasing from 100 to approximately 80%
for the range 80 to 2500 ppm.
Since the system of catalytic converters only permitted the sum of NH- and HCN
to be determined, a selective chemical trapping agent (Cosorb, manufactured
by Malllnckrodt) can be introduced into the gas sampling line upstream of the
catalytic converters to remove HCN quantitatively while allowing the NH_ to
145
-------
pass through the absorbent bed. The Cosorb solid reagent also removes N09
quantitatively and removes a minor amount of NO.
The procedure used to uniquely determine the concentrations of NO, N02, NH_,
and HCN in the sample that is flowing continuously from the probe is shown sche-
matically in Fig. 2B. The sample (at a pressure of 1 torr) is passed through
each of the four paths shown, and the concentration of NO in the gas reaching
the reaction chamber of the CA is measured in each case. Individual species are
obtained from the appropriate differences in the amount of NO measured. The
amount of NO in the sample is determined first without the use of either con-
verter or the chemical trap, path (1). In path (2), the moly converter converts
only N0_ to NO, and the amount of NO measured is equal to (NO) + (N0?) in the
sample. The stainless-steel converter converts all of these nitrogen species
to NO, so the (NO) from path (4) is equal to (NO) + (N02) + (NH_) + E(HCN) In
the sample where E is the efficiency of HCN conversion in this converter. In
path (3), the HCN and N0? are removed in the chemical trap before the sample
passes through this converter, giving a measure of (NO) + (NH,) in the sample.
The species concentrations are essentially calculated from these four measure-
ments by solving four equations with four unknowns. It should be noted that
any nitrogen-containing free radicals present in the flame gases as they enter
the sampling probe will be measured also if they form NH. or HCN in the probe.
If they form N_ they will not be measured.
CH -02-N FLAME STUDIES:
FUEL-RICH FLAMES
The purpose of this series of experiments with a CH,-0,,-N? flame was to investi-
gate the mechanism and rate of thermal NO formation under fuel-rich conditions
X
and to compare directly thermal and fuel NO formation under identical condi-
tions. Three detailed probing experiments (No.'s 2, 3, and 6) were carried out
at an equivalence ratio
of 1.5, a diluent ratio (DR)* of 0.7, a total pres-
sure (PT) of 76 torr, and an inlet gnu feud rate (FT) of 7520 cm /min (STP).
The reactants in experiment 2 contained no additive and, therefore, all of the
*The diluent ratio is defined as
Diluent/0,.
"
-------
NO formed from N or from upecles formed from N_. In experiments 3 and 6,
50 ppm NH» and 50 ppm HCN, respectively, werc> added to Investigate the interac-
tions, if any, between the reactions that form thermal NO and those that form
fuel NO.
The temperature profile measured along the axis of the burner (corrected for
radiation) from experiment No. 2 is shown in Fig. 3A, and is typical of all the
temperature profiles obtained under these fuel-rich conditions. The tempera-
ture rises rapidly below and through the luminous zone region, reaches a maxi-
mum just above the luminous zone, and thereafter decreases gradually which axial
distance. This almost linear decrease beyond the maximum temperature point is
attributed mainly to flame radiation losses*, although some gradual mixing of
the outer, cooler purge gas with the inner core of hot combustion gas may occur
far above the flame front. In these three fuel-rich experiments, the luminous
zone extended between about 3.5 and 5.5 mm above the top of the burner, with
the maximum temperature of about 2000 K occurring at 7.6 mm.
In experiment No. 3, the mole fractions of the major reactant and product spe-
cies (CH,, 0?, H2, CO, CO., N_, and H_0) were determined as a function of dis-
tance above the burner using mass spectrometry. The mass spectrometric analy-
sis determines the concentrations of the species CH,, 02, H2> and CO relative
to the sum of CO + N-. The CO and N_ cannot be measured separately in a low-
resolution mass spectrometer because they have the same molecular weight, and
masses of the lower-molecular-weight fragments formed are not unique in these
two species. The partial pressure of water vapor also cannot be measured
accurately because of its affinity for surfaces. The CO concentrations were
set equal to the amount required to balance the C/N ratio of each sample to
the C/N ratio in the reactants, while the concentration of H?0 in each sample
was adjusted to balance the H/N ratio to that of the reactants. The major
species mole fraction profiles for the fuel-rich flame, calculated using this
procedure, are plotted in Fig. 3B.
The calculated H20 concentration are not shown in Fig. 3B; the H-0 begins to
increase jus't .above 3 mm and levels off at 19 mole percent at about 5.5 mm.
The theoretical final concentration of HO is 20.5%, indicating that the H2
Calculations are being carried out to verify this assumption.
147
-------
balance is quite good. The predicted equilibrium concentrations of CO and CCL
at this temperature are 10.5 and 4.9%, respectively. This is in fair agreement
with the measured values at 8 mm of 8.5 and 5.7% (Fig. 3B).
The concentrations of 0_ and CH, measured in an unburned sample of reactants
are plotted at the left side of Fig. 3B (X = 0). These are in good agreement
with the metered concentrations denoted as Oj a*id CH, .
Because H,, diffuses so rapidly, its concentration profile is quite flat. It
can be seen that H» formed in the flame front diffuses upstream toward and
possibly through the screen flame holder. The H_ concentration is nearly 4%
at the first data point which is only 0.6 mm above the top of the screen (the
screen extends from 0 to 1.1 mm on the assigned distance scale). This is in
agreement with the predictions of preliminary kinetic-diffusion flame model
calculations that have been carried out.
A plot of the species flux provides more information since it takes into con-
sideration the diffusion of each species and gives a more precise indication
of the exact location at which species are forming and disappearing. Following
the one-dimensional flow model presented in Ref, 1 and 2, measuring the slopes
of the mole fraction profiles at selected points and knowing the gas tempera-
'ture profile enabled the molar flux profiles to be calculated. It can be seen
from Fig. 3C that the CH, and Q react in the luminous zone, as expected, and
'that the CO and H^ also form in this region. The slowest reaction involving
the major species is the oxidation of CO to CO,,.
THERMAL NO FORMATION IN FUEL-RICH CH4-02~N2 FLAME
The NO mole fraction profile measured in experiment 2 is plotted in Fig. 4A.
This fuel-rich methane flame contained no additive. Little NO is present in
the luminous zone, and the maximum NO mole fraction of about 50 ppm is not
reached until beyond 30 mm above the burner where the temperature has dropped
more thnn 100 degrees from its maximum value. Also plotted in Fig. 4A are the
calculated equilibrium NO values for these conditions. Fenlmore (Ref. 5) had
obtained a correlation in ethylene flames that indicated that the amount of
prompt NO formed in very rich flames was approximately equal to the amount of
148
-------
NO that would form at chemical equilibrium. Since our NO profile does not have
the shape obtained by Fenimore, it is not possible to separate prompt NO, but
the total NO is greater than the equilibrium NO. The calculated equilibrium
NO is fairly temperature sensitive under these conditions, doubling every 68
degrees. Thus, if the measured temperatures were low by 70 degrees for some
reason, the measured NO would not exceed the equilibrium values. Even if the
measured temperatures were in error by this amount (which is unlikely), the
fact that the NO profile continues to increase in the region where the tempera-
ture is decreasing establishes that our NO yield is not limited by chemical
equilibrium.
Plotted in Fig. AB on an expanded distance scale are the measured mole fractions
of HCN, NO, and NO . The HCN, formed in this flame from the reaction of N2,
reaches a maximum mole fraction of 35 ppm just above the top of the luminous
zone.
Using the temperature profile and the mole fraction plots for NO, N02, and HCN,
the molar flux profile for each of the fuel nitrogen species was calculated.
The NO and HCN flux profiles are plotted in Fig. 4C and on an expanded distance
scale in Fig. 5D where they are compared with the results of experiment 3 that
will be discussed later. As shown in Fig. 4C and 5D, most of the NO forms well
above the luminous zone where the HCN is disappearing. Less than 20% of the NO
has formed at the point of maximum temperature where the HCN flux has reached
its maximum. The maximum HCN flux is 75% of the maximum NO flux. More than
one-half of the HCN appears to form in the luminous zone.
The fact that the maximum measured HCN flux is only 75% of the maximum NO flux
does not preclude the possibility that all of the NO forms via an HCN interme-
diate under these conditions. The reason for this is that HCN is probably being
both formed and consumed in some regions of the flame. If the 10 to 20% of the
NO that forms below the point of maximum HCN flux forms from HCN, this amount
of HCN will not contribute to the maximum HCN flux. Likewise, any HCN that is
formed after this maximum is reached will not contribute to the maximum value.
At the point of maximum HCN flux, the rates of HCN formation and consumption
are equal. It is unlikely, therefore, that HCN only forms below the point where
this maximum occurs and is only consumed above this point. The decrease in the
149
-------
HCN flux between 7 and 19 mm is about 90% of the increase in the NO flux in
that region, and some HCN undoubtedly forms in that region also. The rate of
NO formation at 12.5 mm is 83% of the rate of HCN decay, suggesting that HCN is
the principal intermediate in that region.
) i
The N0? flux profile is not plotted but, as expected from the N07 mole fractions
^ -Q
in Fig. 4B, the flux increases from zero at 4.5 mm to a maximum of 1 x 10 at
5.6 mm and then decays rapidly.
Discussion of Results: "Prompt" NO Formation
Fenimore (Ref. 5) observed that in fuel-rich hydrocarbon-air flames at 1 atm a
rapid transient formation of NO occurs in the "primary reaction zone" followed
by the slower formation of additional NO in the post-flame gases. He termed the
initial rapidly formed NO "prompt NO" and, since the slow rate observed in the
post-flame gases agreed with that predicted by the well-known Zeldovich mech-
anism (0 + N = NO + N, etc.), concluded that prompt NO forms by some other
mechanism. He proposed reactions such as:
CH + N2 = HCN + N (1)
C2 + N2 » CN + CN (2)
Bauer (in the discussion at the end of Ref. 5) cited data that indicates that
attack on N? by C.. and C_ are more likely reaction paths.
Sarofim and Pohl (Ref. 6) studied NO formation in premixed methane-air flames
at equivalence ratios ( = 1.32,
and that he observed HCN at the ppm level under these conditions.
Iverach et al. (Ref. 7) showed that in ethylene-air flames, for example, the
Zeldovich mechanism could not reasonably account for the rapid early rate of NO
formation above an equivalence ratio of about 1.15 (cf. Fig. 8 of Ref. 7).
150
-------
Their results at = 1-4 would require the oxygen atom concentration to be
about 3000 times its equilibrium value.
Perhaps the strongest evidence that much of the NO forms via a non-Zeldovich
mechanism in fuel-rich premixed flames are the recent observations that con-
siderable HCN forms as a measurable intermediate under these conditions.
Bachmaier et al. (Ref. 8) found that in their propane-air flame at = 1.1
HCN reaches a concentration of 12 ppm about 5 mm above the burner, and then
reacts completely. The yield of prompt (end total) NO is 60 ppm. At (j> = 1.5,
however, the measured HCN reaches 23 ppm at 15 mm and remains unchanged to the
last measurement point at 45 mm, while the amount of NO formed is less than
10 ppm under this condition. Haynes et al. (Ref. 9) measured up to 250 ppm
HCN in acetylene-air flames at 4> = 1.95, but obtained less than 3 ppm HCN
from their methane flames. Morley (Ref. 10) found that HCN formed from N2 in
the reaction zone of rich hydrocarbon flames, the amount being roughly propor-
tional to the N- concentration but not very dependent on the type of fuel
(CH^, C2H^, or C2H2) or on the temperature (between 2000 and 2560 K). The
amount of HCN formed was strongly dependent upon equivalence ratio and in-
creased as the flame became richer. The thermal HCN subsequently reacted
completely within the flame, in part to NO. At an equivalence ratio of 1.58
in an atmospheric-pressure CH.-air flame at 2560 K, Morley obtained a peak
HCN mole fraction of 71 ppm with a resultant prompt NO formation of 102 ppm,
i.e., [HCN] /[NO] t = 0.70. This ratio is identical to that obtained
max prompt ...
in experiment 2 of this study: [HCNJ /[NO] = 35/50 = 0.70.
t&cLX D13.X
The thermal NO flux profile obtained in experiment 2 (Fig. 4C) can be used to
determine accurately the maximum rate of NO formation in this fuel-rich flame,
and a comparison can be made to the rate predicted by the Zeldovich mechanism
at these same flame conditions. Ignoring the slight surge of NO at the top of
the luminous zone (Fig. 5D), the maximum rate of NO formation occurs between 6
_Q O
and 9 mm on the assigned distance scale and has a value of 1.5 x 10 mol/cm -
sec. The temperature is 1991 K in this region. The mole fraction of N2 is
0.54, while the calculated equilibrium oxygen atom mole fraction is 4.0 x 10 .
Using Baulch's rate expression (Ref, 11) of
k - 7.6 x 1013 exp(-38,000/T), cm3/mol-sec
151
-------
for the reaction 0 + N -* NO + N, then
d(NO)/dt = 2k (0)(N2) = 6.2 x 10~13 mol/cm3-sec
Although the NO flux curve in Fig. AC does not have the typical prompt NO shape,
it is indeed "prompt" NO since the maximum rate of formation exceeds the pre-
dicted Zeldovich rate (assuming equilibrium 0 atom) by a factor of 2.4 x 10 .
The apparent NO decay beyond 35 mm is a result of the gas temperature dropping
rapidly beyond the point of maximum temperature, and some mixing of purge gas
into the central flame region beyond 30 mm above the burner.
NH3 ADDITION TO A FUEL-RICH CH4~02-N2 FLAME
"-<,
Shown in Fig. 5A through 5D are the mole fraction and molar flux profiles for
experiment 3 with 50 ppm NH,, added to the conditions employed in experiment 2.
Only a relatively small difference existed between the temperature profiles
measured in experiments 2 and 3.
The thermal NO mole fraction curve from experiment 2 is plotted in Fig. 5A for
comparison. The addition of 50 ppm NH- (0.026 wt % N in fuel) increased the
maximum NO mole fraction from 49.2 to 90.5 ppm, indicating an 83% yield of NO
from the added NH, (assuming that the yields of thermal NO and fuel NO are
-9
purely additive). The maximum NO flux (Fig. 5C) increased from 11.0 x 10 to
-9 -9
20.0 x 10 for an added NH» flux of 10.0 x 10 . The conversion of NH» to NO
i's, therefore, 90% based on maximum fluxes.
In Fig. 5D the HCN and NO fluxes from experiments 2 and 3 are compared. The
HCN flux peaks at the same distance above the burner in each case, and the HCN
begins to form rapidly just below the top of the luminous zone in each. It can
_Q
be seen that the addition of NH,, at a flux of 10.0 x 10 increases the maximum
_q *
HCN flux by 11 x 10 , indicating that most of the NH3 is converted to HCN under
these conditions and the HCN then forms NO.
HCN ADDITION TO A FUEL-RICH CH.-O^-N, FLAME
In experiment 6, 50 ppm (nominal) HCN (0.026 wt % N in fuel) was added to the
conditions used in experiment 2. Thus, the conditions in experiments 6 and
3 were identical except for the substitution of HCN for NH- as the fuel nit-
rogen additive. The temperature profile obtained in thiB experiment agreed
152
-------
fairly well with that measured in experiment 2, with the position of the lumi-
nous zone and the maximum temperature point occurring at virtually the same
location. The maximum temperature measured in experiment 6 was 2040 K compared
to 1992 K in experiment 2. The two profiles differ by approximately 40 to 50 K
over the entire range of the measurement (2.0 to 77.0 mm).
The mole fraction profiles for the fuel-nitrogen species are plotted in Fig.
6A and 6B together with the thermal NO and thermal HCN profile obtained in
experiment 2. The addition of 49.8 ppm HCN increased the maximum NO mole frac-
tion from 49.2 to 82.9 ppm indicating a 68% yield of NO from the added HCN.
This assumes that the yields of thermal and fuel NO are purely additive. The
NO formed well above the luminous zone where the HCN is disappearing.
Note that the measured HCN mole fraction in experiment 6 initially decreases
and reaches a minimum of 42.5 ppm at a point (2.6 mm) where the thermal HCN
is also very low, then increases to nearly 60 ppm, Indicating the formation
of "thermal" HCN. Both the fuel and thermal HCN mole fraction profiles have
maximum points above the top of the luminous zone and just below the point
where the maximum flame temperature is reached. Using the Cosorb solid trap-
ping agent in conjunction with the stainless-steel converter, NH_ levels of
0.6 to 1.8 ppm were measured below and in the flame front between 1.8 and
5.0 mm. However, It should be pointed out that these values were obtained by
taking small differences between two relatively large numbers and, therefore,
must be considered only as approximate.
The molar flux profile for NO obtained in experiment 6 is compared in Fig.
6C with the NO profiles obtained in experiments 2 and 3. Similarly, the HCN
molar flux profiles are compared in Fig. 6D for the same three experiments.
~9
It can be seen that the maximum NO flux increased from 10.9 x 10 to
9 2 9 2
18.4 x 10 mol/cm -sec for an added HCN flux of 9.9 x 10 mol/cm -sec.
The conversion of HCN to NO is thus 76% based on maximum fluxes. For com-
parison, the conversion of NH- to NO was determined to be 92% based on max-
imum fluxes in experiment 3. The NO flux profiles obtained in experiments
2, 3, and 6 are all very similar in shape. The maximum rate of NO formation
occurs at a point approximately midway between the top of the luminous zone
153
-------
and the location of the maximum gas temperature. The maximum NO flux occurs
at about 35 mm downstream, then decreases slightly.
9 2
The HCN flux profile in experiment 6 reaches a maximum at 15.5 x 10 mol/cm -
sec at 7.0 mm, just below the maximum temperature point, and thereafter de-
creases, approaching zero at 20.0 mm downstream. It is likely that more than
this amount of HCN forms, since undoubtedly some HCN is reacting before the
maximum is reached and some forms beyond the maximum. It can be seen in Fig.
6D that, regardless of the form from which the HCN is obtained, the maximum
HCN flux occurs at the same location. The thermal and fuel HCN flux curves
decay in a similar manner, as might be expected. Note that the maximum rate
of HCN disappearance occurs at approximately the same point as the maximum KG
formation rate. In experiment 6 at 7.0 mm, [d(NO)/dt] was found to be
3 max -ft
3.1 x 10 mol/cm -sec, while [-d(HCN)/dt] was determined to be 2.7 x 10
., max
mol/cm -sec.
Discussion of Results; HCN Oxidation
In a recent publication, Morley (Ref. 10) observed that when various nitrogen
compounds (NO, NH~, CH-CN, and pyridine) are added to fuel-rich atmospheric-
pressure hydrocarbon flames, they are converted quantitatively to HCN in the
reaction zone regardless of the type of N-additive. He subsequently studied
HCN oxidation within flames between 2300 and 2560 K, and found the rate to
be consistent with the following reaction being the rate-limiting step:
CN + OH -*- NCO + H (Mechanism A)
A rate expression was derived for this reaction:
d£n(HCN)/dt = k1Q (H00)/(H0) 6 x 1025X2 exp(-135,000/RT>
J.O Z Z
where X is a factor which takes into account any disequilibrium In the radi-
-10 3
cal concentrations in the flame, and k,g = 1 x 10 cm /mol-sec. Morley's
rate data are plotted in the Arrhenius form in Fig. 7 for equivalence ratios
of 2.05, 1.8, and 1.56.
154
-------
Using the rate data obtained in the present experiment 6 at 7.5 mm downstream
(T = 2040 K), letting X = 1 and
(H20)/(H2)=
> at 2040
2.11,
[d£n(HCN)/dt](H2)/(H20) - 425
at = 4.90 x 10
4 '
If a straight line is drawn approximately through
Morley's data points in Fig. 7 for <|> = 1.56 between 2000 and 2400 K, the data
point calculated above for experiment 6 of this work appears to lie very close
to this line. It is suggested that the data points from the lower temperature
flames lie above the expected trend because of a superequillbrium of radicals
(A > 1). Morley proposed an alternate mechanism for HCN oxidation, namely
0 + HCN + NCO + H (Mechanism B)
but the predicted rate for this reaction is too slow to explain the experi-
mental results (Fig. 7).
CH4-0 -N2 FLAME STUDIES: FUEL-LEAN FLAMES
The purpose of these experiments (No. 4 and 5) was to investigate the mechanism
and rate of thermal NO formation at fuel-lean conditions and to compare directly
thermal and fuel NO formation under identical conditions. Both experiments
were carried out with a CH.-O -N2 flame at $ = 0.8, DR = 0.7, P = 76 torr,
and F = 7520 cc/min at STP. The reactants in experiment No. 4 contained no
additive and, hence, all of the measured NO formed from N?. In experiment
No. 5, 54 ppm NH on a total volume basis (0.044 wt % N in CH.) was added to
investigate the interactions, if any, between the reactions that form "thermal"
and "fuel" NO.
The flame temperature profile obtained in experiment No. 4 is given in Fig.
8A. The luminous zone extended from 1.8 to 3.0 mm above the burner top, while
the maximum temperature of 1953 K occurred at 5.2 mm. The temperature profile
measured in experiment No. 5 agreed closely with that in experiment No. 4,
155
-------
the maximum temperature being 1940 K also at 5.2 mm. The temperature profiles
were strikingly similar in the vicinity of the luminous zone, and differed by
only about 30 K at 80 mm downstream.
In these CH.-O^-N- fuel-lean flames, the mole' fractions of the major reactant
and product species were also determined as a function of axial distance using
mass spectrometry. Following the same computational technique described in
the previous section, the mole fraction and molar flux profiles were calculated
and are shown in Fig. 8B and 8C for CH,, 0 , H-, CO, and (XK.
The concentrations of 0. and CH, in an unburned sample of reactants are plotted
at the left side of Fig. 8B, while the metered concentrations are indicated at
X * 0. The CH, values agree fairly well, while the measured 0- concentration
is about 2% higher than the metered value.
The predicted equilibrium concentration of CO and CO at 1935 K are 0.12 and
9.8%, respectively, which is in good agreement with the measured values at
7.0 mm of 0.25 and 9.5% (Fig. 8B). The measured 02 concentration at this loca-
tion is 6.2% compared to the predicted value of 4.8%, but this is not unreason-
able considering the measured inlet 0_ concentration. The H mole fraction
profile again is seen to be relatively flat, but the measured concentration of
0.70% at 7.0 mm is considerably higher than the calculated equilibrium value
of 0.055%.
It can be seen from Fig. 8C that most of the CH, consumption and about one-half
of the 0» consumption occurs in the luminous zone. EL forms near the top of the
luminous zone and then reacts just above this region where the CO is forming.
The CO oxidation step again appears to be the slowest reaction involving the
major species.
THERMAL NO FORMATION IN A FUEL-LEAN CH.~0.-N,. FLAME
422
The mole fraction and molar flux profiles for NO and NO^ in experiment 4 are
shown in Fig. 9A through 9C and IOC. The maximum NO mole fraction was deter-
mined to be 9.3 ppm (at 47 mm) compared to the maximum NO- mole fraction of only
2.7 ppm (at .1.0 mm). It can be Hhown from Fig. 9C that the maximum rate of NO
156
-------
formation occurs just above the top of the luminous zone, and the NO formed in
that region diffuses into the luminous zone and is rapidly consumed. The rate
of formation becomes nearly constant just beyond the location of the maximum
gas temperature, and eventually becomes zero at about 40 to 50 mm downstream.
N0~ appears to be forming in the luminous zone and the maximum rate of N0« dis-
appearance coincides with the maximum NO production rate at a point just above
the top of the luminous zone. No HCN or NH^ formed in these experiments.
Discussion of Results: Thermal NO Formation
Table I compares the rate of NO formation (slope of the NO flux profile)
measured at three points along the NO flux curve of Fig. 9C with that pre-
dicted by the Zeldovlch mechanism, assuming equilibrium oxygen atom concen-
tration. The latter is calculated In the manner dencribed earlier using
Baulch's rate expression for the reaction 0 + N_ * NO -t- N. Aa can be seen
from Table I, the measured rate exceeds the predicted Zeldovlch rate (assum-
3
ing equilibrium oxygen atom concentration) by a factor of 4.6 x 10 at the
top of the luminous zone, but decreases to only a factor of 27 at 5.15 mm
downstream (i.e., 2 mm above top of luminous zone).
Sarofim and Pohl (Ref. 6) have studied NO formation in a premixed methane-
air flame (1 atm) at an equivalence ratio of 0.89 (T = 2000 K)_, and ob-
served that the NO mole fraction peaked at approximately 10.5 ppm. This is in
good agreement with the maximum NO mole fraction of 9.3 ppm obtained in this
study at 0.8 (DR - 0.7, T = 1953 K). They determined that the peak
rates of NO formation, which they observed in the "flame zone" are in agree-
ment with Zeldovich kinetics when the free radicals are calculated from the
concentration of stable species using the partial-equilibrium assumptions
for the fast-flame reactions. Thus, the experimental data and partial equi-
librium calculations of Sarofim and Pohl provide a plausible explanation for
the differences in measured versus calculated NO formation rates shown in
Table I at fuel-lean conditions, i.e., the presence of excess [o] radicals.
157
-------
NH ADDITION TO A FUEL-LEAN Ct^-O.-N FLAME
Kor comparison, the mole fraction and molar flux profiles for experiment 5 are
shown In Fig. 10A through 10D together with the thermal NO profiles obtained in
experiment 4. The addition of 54.5 ppm NH, to the conditions of experiment 4
increased the maximum NO mole fraction from 9.3 to 60.9 ppm, indicating a 95%
yield of NO from the added NH_ assuming that the yields of thermal and fuel NO
are purely additive. The maximum NO,, mole fraction measured in experiment 5
with NH, addition is 6.6 ppm at 1.7 mm (Fig. 10B). However, the trend of the
data suggests that more than this amount of N0_ may have formed in the region
near the top of the burner.
i
9
From Fig. IOC it can be seen that the maximum NO flux increased from 1.86 x 10
_q o 9 2
to 12.14 x 10 mol/cm -sec for an added NH flux of 10.87 x 10 mol/cm -sec. .
The conversion of NH, to NO is thus also 95% based on maximum fluxes. As in
experiment 4, the maximum rate of NO. formation (Fig. 10D) occurs just below
the top of the luminous zone, and the N02 reacts rapidly just above the top of
the luminous zone.
Figure 10D reveals that the maximum rate of NO formation in experiments 4 and
5 occurred at exactly the same location, namely, 3.2 mm above the top of the
burner and just above the top of the luminous zone (3.0 mm). This is in the
region where both the NH_ and N02 flux profiles are decreasing rapidly. The
NH~ flux begins to decrease abruptly at the bottom of the luminous zone and
finally disappears just beyond the point where the NO flux has reached its
maximum value (3.5 to 4.0 mm).
H2-CO-02-Ar FLAME STUDIES: FUEL-RICH FLAMES
There is considerable interest in burning low-Btu gas, which has been obtained
from coal gasification processes, in commercial power plants. However, these
synthetic gases may contain trace quantities of fuel nitrogen species, such as
NH.,, which can be converted to NO during combustion. Preliminary calculations
reported in the literature (Ref. 12 through 14) have shown that proper combus-
tion design can contribute significantly to reduction of NO emissions from
158
-------
low-Btu gas syBtcmB and that staging of the combustion process has the poten-
tial for reducing NO emissions by a factor of 3 or 4. Fundamental kinetic
data on the conversion of fuel-nitrogen species in a low-Btu gas system, par-
ticularly under fuel-rich conditions, would be very useful in developing com-
bustion models for determining optimum conditions for reducing NO emissions frota
A
low Btu gas power generation systems. To employ a low-pressure, flat-flam*
burner system to collect such information, the data should be obtained at
flame temperatures which correspond approximately to those measured in real
combustor configurations when using low-Btu gas derived from a typical coal
conversion process.
The basis for the experimental burner conditions subsequently chosen for the
present study came from the recent work by Tyson (Ref. 13), who developed
a kinetic model for NO emissions in a low-Btu gas combustor, validated:
the model by comparison with experimental data, and applied this Information
to idealized combustor configurations for two representative combined cycle
systems. In one case, for an advanced-technology, high-temperature gas tur-
bine with a waste-heat boiler, the optimum design for an adiabatic gas gener-
ator was found to be a rich (1.33 < $ < 1.45) primary reactor section (200-msec
residence time) followed by a gradual mixing of the primary products into the
dilution air stream. Using the staged system as opposed to an unstaged system
resulted in calculated NO emissions reduced by a factor of 4.
The fuel composition employed by Tyson in his gas turbine combustor calcula-
tions was that of a typical air-blown Lurgi gas and contained 10.1% H~0, 19.6%
H2, 13.3% CO, 13.3 C02, 5.5% CH^, 37.6% N2> and 0.4% NH.J. This value for the
NHL mole fraction was arrived at by assuming a hot-gas removal process for
H_S (1088 K), and was based on the high-temperature cleanup data reported by
Robson (Ref., 15). The fuel gas inlet temperature and the air inlet tempera-
ture were 1500 and 1000 F, respectively, while the total pressure was main-
tained at 10 a tin.
The H2/CO ratio chosen in the present work was set at 1.4 to simulate the H2/
CO ratio found in the low-Btu gas derived from the typical air-blown Lurgi
process. Argon was selected initially as the diluent to preclude the
159
-------
possibility of thermal NO formation and to permit the CO to be measured mass
X
spectrometrically. From Tyson's conclusions, a fuel-rich condition appeared
to be the most interesting for these studies, and an equivalence ratio ($) of
1.4 was chosen. From adiabatic flame temperature calculations at 4> = 1.4,
it was found that a diluent ratio of 1.9 assigned to the simulated low-Btu gas
(H_-CO-0_-Ar-NH3) at 0.1 attn would nearly match the adiabatic flame tempera-
ture (2150 K) of the Lurgi-air system at 10 atm.
Preliminary screening experiments with the H -CO-02~Ar flame system revealed
that this fuel burns much differently in a low-pressure burner than does ]
methane. To position the luminous zone above the burner so that suitable
detailed probing measurements could be made, the total system pressure was
reduced to 50 torr, while the total inlet gas feed rate was increased 17% over
that employed in the previous CH, flame studies to 8750 cc/ram (STP) .
Two experiments were conducted to determine the rate and extent of the con-
version of fuel nitrogen species added to a simulated low-Btu fuel flame of
H , CO, 02, and Ar. NH, was added to the flame in experiment 8 at 2740 ppm
on a total feed basis, which is equivalent to 4200 ppm on a Lurgi fuel basis.
the conditions utilized in experiment 9 were identical to those employed in
. o
experiment 8 ( = 1.4, DR = 1.9, 50 torr, and 8750 cm /min) , with the excep-
tion that 2230 ppm NO was added as the fuel-N species instead of NH- to deter-
mine how much of the NO, if any, is consumed as it passes through the flame
front.
The temperature profile along with the centerline of the burner (corrected for
radiation) obtained in experiment 8 is given in Fig. 11A. The luminous zone
extended between 3.3 and 8.2 mm above the top of the burner. The maximum
flame temperature of 1857 K (Tadlabat:lc = 2144 K) was at 12.9 mm. The detailed
probing was performed with a 3-mil Al 0--coated Pt/Pt-10% Rh thermocouple.
There was evidence that the thermocouple coating was deteriorating and that
some catalytic reactivity of the thermocouple was occurring in the region
below and through the luminous zone. If the measured temperatures are somewhat
160
-------
high, the net effect would be to place an upper limit on the N-species mole
fraction and molar flux profiles. However, the conclusions drawn from the
experiment do not appear to be critically dependent on the accuracy of the
temperature measurement.
As in the CH, flame experiments, the mole fractions of the major reactant and
product species were measured as a function of axial distance using mass spec-
trometry. The data were reduced to yield the mole fraction and molar flux
profiles shown in Fig. 11B and 11C for H , CO, 0 , and CO . In this analysis,
CO was determined directly, since no N was fed initially to the burner, and
H-0 was calculated as before by balancing the H/Ar ratio to that of the burner
feed.
As expected, H2 reacts at a much faster initial rate than does CO. Above the
top of the luminous zone, however, some H appears to reform. If this small
increase in the concentration is real and results from the recombination of
H atoms, the H atoms present at 9 mm must have reacted with 0_ in the probe
rather than recombining to H .
Equilibrium calculations indicated that virtually no 02 should remain in this
fuel-rich system at complete reaction. It was not expected that any 0. would
remain unreacted above the luminous zone in the presence of excess H_ . The
kinetic-diffusion calculations that are being carried out will establish what
the 0- reaction rate should be in this region.
NH ADDITION TO A FUEL-RICH H2-CO-02-Ar FLAME
The molar flux profiles for NH_ and NO in experiment 8 are shown in Fig. 12A
and 12B. There was no measurable N02 or HCN within the detection limits of
the analytical system in either experiment. The added NH_ (2740 ppm total
basis) in experiment 8 appears to have completely reacted by .Just above the
bottom of the luminous zone. The location of the maximum Nil- decay rate cor-
responds approximately to the point of Initial NO formation. The maximum
rate of NO formation is 1.58 x 10 mol/cm -sec and occurs at approximately
the snmn point where the NH., Flux becomes zero. The maximum NO flux of
161
-------
-7 2
4.84 x 10 mol/cm -sec (76% yield) occurs at a point roughly three-fourths
of the distance into the luminous zone. Thereafter the NO is consumed, reach-
ing a minimum near the maximum flame temperature location, and then gradually
increases and levels off at 70 mm downstream. The flux at this latter location
f\ 0
(3.10 x 10 mol/cm -sec) corresponds to a NO yield of 49%.
The most interesting feature of the N-species profiles can be found in Fig.
12B where the sum of the NH and NO fluxes are plotted versus distance. Note
that at 4.0 mm, only 26% of the initial fuel nitrogen is present in a form
which can be measured by the chemlluminescent analyzer/converter system (KH«, ,
NO, N02, or HCN). Obviously there in an undertermined radical species which
forms in the flame as an intermediate, and which subsequently can react further
to yield either NO or N? (but presumably forms mostly N9 in the probe).
SITRIC OXIDE ADDITION TO A FUEL-RICH H.-CO-O -Ar FLAME
Because much of the NO that forms from the added NH_ is unexpectedly consumed
above the luminous zone and then partially reforms, NO was added in the absence
of NH» to see how much, if any, is consumed above the luminous zone. With
2230 ppm NO (total basis) added instead of the NHL, the NO flux profile in
experiment 9 (Fig. 12A and 12B) remains relatively constant at its initial value
Lup to about 4.5 mm. The initial decrease in NO flux in experiment 9 corresponds
to the point in the luminous zone where the NO formation rate in experiment 8 "
begins to decrease from its maximum value. In both experiments, the NO flux
decreases rapidly near the top of the luminous zone. However, the NO flux in
experiment 8 decreases significantly more than in 9. Both curves reach a mini-
mum at the same point (approximately 18 mm) just above the maximum temperature
location. It would appear that there are additional free radical species
present in the experiment with NH« addition which can also promote NO depletion.
At least one very long-lived nitrogen intermediate must be involved in experi-
ment 8 because the NO flux continues to Increase more than 40 mm above the
burner (and the NH additive changes the visible color of the flame at even
greater distances), whereas in experiment 9 the NO flux begins to decrease above
40 mm.
162
-------
SUMMARY AND CONCLUSIONS
Low-pressure flat-flame burner experiments provide an excellent tool for study-
ing the types of reactions that are involved in the formation of NO during the
A
combustion of fossil fuels or alternate fuels derived from fossil fuels. In
CH,-0 -N flame studies at fuel-rich conditions ( = 1.4), adding NH- to the flame (2740 ppm) results in a maximum conversion
to NO of 76% near the top of the flame front. The NO is rapidly consumed above
this point, reaches a minimum just beyond the point of maximum flame tempera-
ture, and thereafter begins to increase gradually. Near the bottom of the lum-
inous zone only one-quarter of the initial fuel nitrogen is in a form that can
be measured by the chemiluminescent analyzer/catalytic converter system (NH^,
HCN, N0?, or NO). There is obviously an undertermined radical species that
forms in the flame as an intermediate and subsequently reacts to yield either
NO or N2. 163
-------
Adding NO to a similar flame under identical conditions to determine whether it
too would be consumed in the region just above the luminous zone did not result
in the same rapid NO decline. Apparently, there are additional free radical
species present when NH~ is added that can promote NO depletion in the region
of the flame just above the luminous zone and can also cause the NO flux to
increase far above the burner.
ACKNOWLEDGEMENT
This study was sponsored by the U.S. Environmental Protection Agency under
Contract No. 68-02-1886 with G. Blair Martin as the Project Officer. The fol-
lowing individuals have contributed to the technical aspects of this program:
V. Dayan, L. Grant, M. Heap, G. B. Martin, W. Nurick, D. Seery, G. R. Schneider,
B. Tuffly, and T. Tyson.
164
-------
REFERENCES
1. Axworthy, A.E., G. R. Schneider and V. H. Dayan, "Investigation of the
Chemistry of Fuel Nitrogen Conversion to Nitrogen Oxides in Flames - Final
Report", EPA-600/2-76-:039, February 1976.
2. Axworthy, A. E., G. R. Schneider and V. H. Dayan, "Chemical Reactions in
the Conversion of Fuel Nitrogen to NO }', Proceedings of the Stationary
Source Combustion Symposium, Vol. I, pp 185-216, EPA-600/2-76-152a,
June 1976.
3. Fristrom, R. M. and A. A. Westenberg, Flame Structure, McGraw-Hill, New
York, 1965.
4. Andersen, Jn W. and R. Friedman, "An Accurate Gas Metering System for
Laminar Flow Studies," Rev. Sci. Instr., 20, 61 (1949).
5. Fenimore, C. P., "Formation of Nitric Oxide in Premixed Hydrocarbon Flames,"
Thirteenth Symposium (International) on Combustion, The Combustion Institute,
Pittsburgh, pp 373-380, 1971.
6. Sarofim, A. F. and J. Pohl, "Kinetics of Nitric Oxide Formation in Premixed
Laminar Flames," Fourteenth Symposium (International)on Combustion, The
Combustion Institute, Pittsburgh, pp 739-754, 1973.
7. Iverach, D., K. S. Basden and N. Y. Kirov, "Formation of Nitric Oxide in
Fuel-Lean and Fuel-Rich Flames," Ibid., p. 767, 1973.
8. Bachmaier, F. et. al., "The Formation of Nitric Oxide and the Detection
of HCN in Premixed Hydrocarbon-Air Flames," Comb. Sci. and Tech. _7_» PP-
77-84 (1973).
9. Haynes, B. S., D. Iverach and N. Y. Kirov, "The Behavior of Nitrogen Species
in Fuel-Rich Hydrocarbon Flames," Fifteenth Symposium (International} _on
Combustion. The Combustion Institute, Pittsburgh, p. 1103, 1975.
10. Morley, C., "The Formation and Destruction of Hydrogen Cyanide from Atmos-
pheric and Fuel Nitrogen in Rich Atmospheric-Pressure Flames," Combustion
and Flame. ^7, pp. 113-121 (1977).
165
-------
11. Baulch, D. L., D. D. Drysdale and D. G. Home, Evaluated Kinetic Data for
High Temperature Reactions. Vol. 2; Homogeneous Gas Phase Reactions of the
-2"-2~N2 Sv3tem» Butterworths, London, (1973).
12. Martin, G. B., "NO Considerations in Alternate Fuel Combustion," presented
at EPA Symposium; Environmental Aspects of Fuel Conversion Processes,
Hollywood, Florida, December 14-18, 1975.
13. Tyson, T. J., "Chemical and Physical Limitations on Low BTU Gas Combustor
NO Emissions," presented at EPA Combustion Research Branch Contractors
X
Meeting, MIT, Cambridge, Mass., August 1976.
14. Heap, M. P., T. J. Tyson, J. E. Cichanowicz, R. Gershman, C. J. Kau, G. B.
Martin, and W. S. Lanier, "Environmental Aspects of Low BTU Gas Combustion,"
presented at the 16th Symposium (International) on Combustion, MIT, Cam-
bridge, Mass., August 1976.
15. Robsen, F. L., and A. J. Giramonti, "The Environmental Impact of Coal-
Based Advanced Power Systems" Symposium Proceedings; Environmental Aspects
of Fuel Conveys ion Technology, EPA-650/2-74-118, October 1974, pp. 237-257.
166
-------
TABLE I. COMPARISON OF EXPERIMENTALLY MEASURED (EXPERIMENT NO. 4)
VS CALCULATED NO PRODUCTION RATES
AX = 3-0 to 3-2 mm
At Top of Luminous
Zone
AX - 5-15 mm
Location of Maximum
Gas Temperature
AX 8.00 mm
Constant NO
Formation Rate
Measured* Gas
Temperature,
K
1900
1953
1935
Measured**
NO
Production
Rate,
mo 1 /cnr -sec
1.16 x 10"7
1.59 x 10"9
3.70 x 10"10
Calculated***
NO
Production
Rate,
mol/cm -sec
2.55 x 10* !1
S.Bk x 10"11
3.66 x 10~H
R
jneas
Rcalc
*»550
27
10
* Corrected for thermocouple radiation loss
**See Fig. 9
***Assumes equilibrium concentration of [0]
167
-------
(U
e
0)
*i
<44
O
I
n
ft
K
1 _ *
111
168
-------
TO
VACUUM PUMT
AMOIXHAIffT
TOMOOO
THERMOCOUPLE
Ij
I
PROBE
BURNER
INLET REACTANT
OASES FROM GAS
METERING SYSTEM
NO INN,
STANDARD OA5
IN ALUMINUM CYLINDER
(FOR CALIBRATION OF
TECO ANALYZER)
Figure 2A. Schematic of the Flat-Flame Burner System
INLET
MIXTURE
(=1 torr)
NO
NOj
HCN
(1)
NO
N02
HCN
NH3
1 FROM SAMPLE 1
| FROM N0>
(2)
MOLYBDENUM
CATALYTIC
CONVERTER
AT3SO°C
HCN
NH3
j
... , ^
FROM SAMPLE ]
(4)
STAINLESS STC EL
CATALYTIC
CONVERTER
AT = BOO°C
NO
FROM HCN
FROM NH3 I
-i
<8HJ
TO
CHEMILUMINESCENT
ANALYZER
-»
REACTOR
CHAMBER
Figure 2B. Experimental Scheme for Determination of
NO, N02, NH3 and HCN
169
-------
03*aiOM
to «
LU o
O h-
o: a.
a. X
o
«/>
I- UJ
o
< o
K UJ
U. a.
to
LJ
_l ftl
'.I
oa
=> O
a.
x
UJ
u
<
oo
4J
ID
C
o
u
o
oa 'oo tm xmj MV-IOH
i i i I i I i i
u
H
OS
^H
ac
170
-------
-------
UK» 'Noaovud aiow saioads
5(5
_ ae
U. M
UJ (/)
o
LU Z
O.
CM
CM --»
?-d fl
3 -H
-------
24
20
LUMINOUS ZONE
TMAX-200BOK
AT 7.40mm
ZO 30 40 50 80 70
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
80 90
C. N-SPECIES MOLAR FLUX PROFILES
HCN
("THERMAL''* "FUEL
EXPT. #3)
60 PPM NH3 ADDED
("THERMAL"* "FUEL"
EXPT. #3)
60 PPM NHa ADDED
NH0 NH3 (EXPT
S, NO
^ ("THERMAL" ONLY
EXPT. #2)
4 6 8 10 12 14
AXIAL DISTANCE ABOVE TOP OF BURNER1, mm
16 18
D. COMPARISON OF MOLAR FLUX PROFILES FOR
HCN AND NO IN LUMINOUS ZONE REGION
Figure 5. (Concluded)
-------
8
8 S 8 8
OHM 'NOUWHd 310WS3I03<«
§
8
o
I
Ul
1
s
8
u.
Q
G
Ul
i
ta
N
UJ
uj s>
o
Lkl Z
^ «i^
I
o.
5?
SH
"3
d
OS
ow
go O
w tN
r-l 4), H
1 ' S
« -O u
H § <«
1?7
ffl ftl
vO
I
31OW 83l33dS
174
-------
20
16
1"
i
i.a
8 4
a
LUMINOUS
ZONE
NO
("THERMAL"* FUEL, EXPT#3
WITH BO PPM NH3, TMAX - 2006OK)
("THERMAL" + FUEL, EXPT #6
WITH GO PPM HCN, TMAX ' 2040°K>
("THERMAL" ONLY, EXPT #2,
10 20 30 40 GO 60
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
70
80
C.
COMPARISON OF NO MOLAR FLUX PROFILES
FROM EXPERIMENTS NO. 2, 3, and 6
"THERMAL"*FOEL\
EXPT. 13
BO ppm NH3 ADDED /
ADDITIVE
FLUX:
EXPT IT'S
NO / "THERMAL" + FUEL\
EXPTI6
\60ppmHCNADOED /
NO /"THERMAL"ONLVN
EXPTiK )
4 6 8 10 12 14
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
IB
D. COMPARISON OF MOLAR FLUX PROFILES FOR HCN
AND NO FROM EXPERIMENTS NO. 2, 3, AND 6
IN LUMINOUS ZONE REGION
Figure 6. (Concluded)
175
-------
11
Figure 7.
Arrhenius Plot for HCN Oxidiation Given by Morley (Hef . 10)
Solid Line Corresponds to Mechanism A (136 kcal/mole) for
Equivalence Ratios :Q> 2.05;Q, 1.8; A, 1.56. Broken Line
( --- ) Corresponds to the Predicted Rate for Mechanism B.
21001
TANCE ABOVE TOP OF BURNER, mm
A. FLAME TEMPERATURE VS DISTANCE (EXPT NO. k)
Figure 8. Fuel-Lean CH4-02-N2 Flame ($ - 0.8, DR - 0.7, P - 76 torr,
T5 mm "fCOrt ^ ^ I t |_ ft fWVl \ *
176
FT - 7520 cc/mln at STP)
-------
12 - 24
11
8"
5*
if
UJ
J e
^ 28
- 22 -
M V> \
18 - V>\ \
8
* 1
i
10 -
8
8
4
2
0
r
LUMINOUS
TMAX i9«^
02'
0 \\
^
H i r°H i 1
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
B. MOLE FRACTION PROFILES OF MAJOR
SPECIES (EXPT NO. 5)
4 ?
8"
x
o
snsssnsi
V*. ALUMINOUS
ZONE TMAX
t 2 3 4 6 6
AXIAL DISTANCE ABOVE TOP OF BURNER! mm
C. MOLAR FLUX PROFILES OF MAJOR
SPECIES (EXPT NO. 5)
Figure 8. (Concluded)
177
-------
« to
AXIAL DISTANCE ABOVE TOP OF BURNER, i
A. NO MOLE FRACTION
PROFILE
TMAX-1«W°K
AT fi. 15 mm
NO,
3 « 8 8 10
AXIAL DltTANCt ABOVE TOP OF BURNER, mi
0 2 4 6 « 10 1Z
AXIAL DISTANCE ABOVE TOP OF BUHNER, mm
B. MOLE FRACTION PROFILES FOR NO
AND N02 IN LUMINOUS ZONE REGION
C. MOLAR FLUX PROFILES FOR NO AND N02
IN LUMINOUS ZONE REGION
Figure 9. Thermal NO and N02 From a Fuel-Lean CH4-02-N2 FlameExperiment No. 4
( - 0.8, DR - 0.7, PT » 76 torr, FT = 7520 cc/min)
178
-------
i
§
TMAX
AT 546i
LUMINOUS
ZONE
FUEL » THERMAL NO (EXPT «»
.-THERMAL NO ONLY (EXPT |M|
*ATT-Hr
10 20 30 40
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
A. COMPARISON OF NO MOLE
FRACTION PROFILES
01 t » * »
AXIAL DISTANCE ABOVE TO? OP BURMA, mm
B, N-SPECIES MOLE FRACTION PROFILES
IN LUMINOUS ZONE REGION
14
,12
> 10
<
' *
TH6BMAL _
f
-INLiT NN3 ADDITIVE FLUX LEVEL
NO (THERMAL ONLY - EXPT M)
10 W M 40 10 M 70 60
AXIAL DISTANCE ABOVE TOF OF BUANEfl. mm
C. COMPARISON OF NO MOLAR
FLUX PROFILES
0 1 Z 3 466
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
D. N-SPECIES MOLAR FLUX PROFILES IN
LUMINOUS ZONE REGION
Figure 10, Thermal and Fuel NO Formation From Fuel-Lean
Flames With 54 ppm NH3 AddedExperiments No. 4 and No. 5
(4> - 0.8, DR - 0.7, PT *° 76 torr, FX a 7520 cc/min)
179
-------
1WM
1800
f? 1600
| 1600
1400
1300
1200
irssi
..LUMINOUS ZONE
10 20 30 40 M 60 70 SO
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
A. FLAME TEMPERATURE VS DISTANCE
I
B
10
LUMINOUS ZONE
g. I 10 16
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
4.0
0.0
. 1887«>K
0 5 10 IS
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
B. MOLE FRACTION PROFILES OF
MAJOR SPECIES
C. MOLAR FLUX PROFILES OF
MAJOR SPECIES
Figure 11. Fuel-Rich H2-CO-02-Ar Flame (H2/CO - 1.43, <|> 1.4, DR - 1.9,
P - 50 torr, T - 8750 cc/min at STP) - Experiment No. 8
180
-------
I P77>*-LUMINOUS ZONE
10 20 30 40 BO 60
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
A. NO MOLAR FLUX PROFILES
"e
x
3
§
7.0
8.0
5.0
4.0
3.0
2.0
1.0
LUMINOUS ZONE
NO (EXPT. #9)
0 24 6 8 10 12
AXIAL DISTANCE ABOVE TOP OF BURNER, mm
B. N-SPECIES MOLAR FLUX PROFILES
IN LUMINOUS ZONE REGIONS
Figure 12. Fuel NO From and NO Stability in Fuel-Rich H2~CO-02-Ar
Flames - Experiments No. 8 and No. 9 (H2-CO - 1.43, $ - 1.4,
DR - 1.9, PT = 50 torr, T - 8750 cc/min)
181
-------
-------
FORMATION OF SOOT AND POLYCYCLIC AROMATIC HYDROCARBONS
IN COMBUSTION SYSTEMS
Development of a Molecular Beam Mass Spectrometer
By:
J. D. Bittner
Massachusetts Institute of Technology
Cambridge, Massachusetts 02139
183
-------
-------
ABSTRACT
The overall objective of this research program has been to study
the production of soot and polycyclic aromatic hydrocarbons (PCAH) in com-
bustion systems. The program has two phases. In one phase the production
of soot and PCAH in a turbulent diffusion flame was studied. The results
have been reported elsewhere. In the phase to be discussed in this paper,
a low pressure flat laminar premixed flame is being used to study the
kinetic relationships between soot, PCAH and other hydrocarbon species that
may be important as soot nuclei or surface growth species. The molecular
beam-mass spectrometer system developed to study the gas phase species will
be described. Preliminary mass spectral data on an acetylene-oxygen flame
will be reported.
185
-------
-------
INTRODUCTION
The formation of soot In combustion systems has been studied for many years
with the objective of attaining a working knowledge of the process to aid in
predicting flame emissivities for radiative heat transfer problems. Recently,
concern has been expressed that soot and some of the organic compounds
associated with incomplete combustion are pollution problems that may have
to be controlled. It is known that some polycyclic aromatic hydrocarbons
(PCAH) produced in flames and adsorbed on soot particles are carcinogenic.
Soot and other organic particulate matter may not be the most abundant particu-
lates emitted, but because the small particles (<0.2ym) are easily injected
deep into the respiratory system, they may be one of the types of particulates
most hazardous to human health. The effects of two expected changes in the
operation of combustion systems threaten to increase the emissions of particulate
organic matter. First, the staged combustion strategy for controlling NO
requires primary combustion in a fuel rich atmosphere that leads to increased
soot and PCAH formation. Second, the planned increased use of coal and coal
derived liquids will also lead to Increased soot and PCAH formation since
1 2
aromatic fuels are known to be troublesome in this respect. * The necessity
for learning how to burn low H/C fuels (aromatics) without considerable soot
and PCAH formation is especially important for small systems such as gas
turbines and internal combustion engines where residence times available for
burnout are limited.
The mechanism of soot formation in combustion systems Is not well under-
stood. Some of the mechanisms proposed over the years and the confusion
surrounding the role of PCAH in the process are Illustrated in Figure 1.
Since soot particles contain mnny more carbon atoms and a much lower H/C
ratio than the fuel molecules, soot formation must involve processes of
187
-------
aggregation (top to bottom) and dehydrogenation (left to right). The extreme
routes, the C2 route and the saturated polymer route, are unlikely to occur
under typical combustion conditions. The C~ route may be important at
temperatures around 3000°K. At lower temperatures, around 700°K, and long
reaction times the saturated polymer route may occur. But, in the range
1200-2200°K most mechanisms can be represented by the central portion of
Figure 1. From this figure it is seen that PCAH species have been postulated
to serve as nuclei for soot growth by surface decomposition of any number of
hydrocarbon species (including themselves). They have been postulated to
have sufficiently low vapor pressures to physically condense to liquid drop-
lets which then form soot. They have also been postulated to be stable by-
products of the aoot formation process.
In light of the concern over particulate organic matter as a pollutant
and the confusion surrounding soot formation mechanisms and the role of PCAH,
the specific objective of this research is to study the chemistry of soot and
PCAH formation. Emphasis is on the identification and measurement of relative
concentration profiles of the gas phase hydrocarbon species and the measurement
of soot particle concentrations, size distributions and total mass of soot
at different stages of combustion in a flat premixed flame.
1 EXPERIMENTAL APPARATUS
DESCRIPTION
The molecular beam mass spectrometer system that has been developed
during this project to study premixed low-pressure flat flames is schematically
represented in Figure 2. The flame is stabilized on a flat water-cooled
i
drilled copper plate burner that is 7.1 cm in diameter. The burner chamber
is pumped by a 60 cfm Stokes mechanical pump. The flame is sampled along
i
the centerline of the burner by moving the burner relative to the rest of
the apparatus. The sample is withdrawn supersonically through a quartz
4
nozzle. The nozzle is a hybrid type, as discussed by Biordi et al. The
tip has about a 40° outside angle that expands to a 90° angle about 1.5 cm
from the tip. The central core of the expanding flame gases passes through
the orifice in the tip of the cone-shaped (60° total outside angle) aluminum
188
-------
skimmer. Formation of the molecular beam is completed by collimation with
a circular orifice placed between the second and third (mass spectrometer)
stages. The beam is modulated by an American Time Products tuning fork
chopper placed near the entrance to the ionizer of the Extranuclear quadru-
pole mass spectrometer. Orifice diameters, separation distances and typical
operating pressures are shown in the following table. The partition
between the second and third stage is a hollow annulus designed to be filled
Orifice diameter,
mm
Distance from Pressure downstream,
nozzle, cm flame at 20 torr,torr
nozzle
skimmer
collimator
ionizer
0.
0.
3.
3.
7
9
0
0
"
1.8
v 36
* 40
1.
1.
1.
1
2
4
-4
x 10
x 10~6
x 10~7
(stage
(stage
(stage
1)
2)
3)
with liquid nitrogen to provide cryogenic pumping near the ionizer to reduce the
amounts of water and other condensible species, especially hydrocarbons,in the
background. This recent improvement has been fabricated but not yet tested
and the pressures reported in stages two and three are without the benefit of
this cryogenic pumping. The system is equipped with an effusive source that
can be moved into place to calibrate for mass spectrometer sensitivities and
study mass discrimination effects in the beam.
For studying the particulate phase the mass spectrometer is removed and
the top flange is replaced by a flange that has a mechanism for mounting
electron microscope grids and moving them in and out of the beam. The grids are
exposed to the beam for a measured time interval with the use of a mechanical
shutter. The particles collected on the grid are analyzed by making electron
micrographs using either shadowing or transmission procedures. Number densities
and size distributions are obtained with the aid of a Zelss semi-automatic
particle counter. Computation techniques developed by Wersborg will be used
to calculate profiles of average soot particle size, soot particle number
concentration, rates of nucleatlon, surface growth and coagulation.
A schematic of the mass spectrometer instrumentation is shown in Figure 3.
189
-------
The molecular beam is chopped by the fixed-frequency tuning fork operating
at 220 Hz. The signal from the electron multiplier at any particular mass has
a DC component corresponding to ions originating from the background gas
molecules and an AC component at 220 Hz corresponding to ions originating from
the beam molecules. The signal is amplified with an Extranuclear fast
electrometer preamplifier. The amplified signal is DC coupled to an Extra-
nuclear electrometer to obtain a signal corresponding to the background
component. A reference signal from the chopper and the AC component of the
signal from the preamplifier are introduced into a Princeton Applied Research
HR-8 lock-in amplifier to obtain a signal proportional to the beam component.
Signals corresponding to beam and background components are then displayed
either on a dual-beam storage oscilloscope or a chart recorder.
CALIBRATION
Cons iderat ions
! Calibration of supersonic molecular beam systems for flame sampling is not
a straight forward task due to the many effects that can influence the compo-
sition of the beam prior to reaching the mass spectrometer ionizer. Shock
formation in front of the skimmer orifice, species condensations, pressure
diffusion in the free jet, Mach number focussing downstream of the skimmer, scattering
of the beam by the background gas in-any.-or all of the chambers, and effusion of the
background gas into the beam are effects that can distort the beam composition
and are discussed extensively in a review article by Knuth . Skimmer interference
problems can be avoided by proper design of external and internal angles of
th'e skimmer and the selection of the proper distance between the tip and the
supporting wall. Species condensations in free jet expansion are not
believed to be important at flame temperatures and sub-atmospheric pressures
for most flame constituents , however little is known about the behavior of the
high molecular weight hydrocarbons and care must be taken to look for condensation
effects in the interpretation of the data. Pressure diffusion, Mach number
focussing and background scattering and effusion are more difficult effects
to avoid altogether and all can be functions of source conditions. A study
of their Importance in this system and techniques for calibration of these
effects is now underway. The limited data indicate that background scattering and
effusion are negligible under the flame conditions presented here. Pressure
190
-------
diffusion and Mach number focussing are lumped together as "mass discrimination"
and can be studied in this system with the aid of the effusive source placed
in the second stage upstream of the collimator.
The system for studying mass discrimination is shown schematically in
Figure 4. Mixtures of the major stable species are made by metering through
critical orifice flow meters and mixing dynamically. These mixtures may be
introduced into the mass spectrometer ionizer either as an effusive beam in
which no mass discrimination effects occur or as a supersonic beam (through
the burner and quartz nozzle) in which high molecular weight species are
usually preferentially concentrated relative to low molecular weight species.
To introduce the mixture effusively without mass discrimination, a
stainless steel sintered disc with a nominal pore size of 0.5 urn is used to
leak into the effusive source in the second stage about 1% of the total gas
flow by the disc. The pressure on the high pressure side of the sintered disc
is maintained in the range 0.1 to 1 torr. Tubing sizes and pressures are
designed to give viscous flow (no mass discrimination) everywhere upstream of
the sintered disc and effusive flow through the disc and downstream from it.
Under these conditions" the flow rate of a component through the disc is in-
versely proportional to the square root of its molecular weight but its
density anywhere downstream of the disc is proportional to its partial pressure
upstream. Since the electron impact ionizer is a density detector, ratios of
ion signals are proportional to the ratios of partial pressures in the gas
mixture upstream of the disc. The effusive source inlet system has been tested
with commercially prepared standard gas mixtures and reproduces the suppliers
analysis to within 2% on any component.
By comparing signal ratios in the effusive beam to signal ratios in the sonic
beam,the mass discrimination between two species in a mixture can be character-
ized by an enrichment factor, a
AB'
"
where
AB
B
AB
S
E
enrichment factor of species A relative to species B, dimensionless
beam signal intensity of species A, amperes
beam signal intensity of species B, amperes
sonic introduction
effusive introduction
191
-------
XA = mole fraction of species A
&
X - mole fraction of species B
D
The effusive source system is also used to obtain relative mass spectrometer
sensitivity factors for stable species, S. :
S kB
*»*!* o, -Jl
B k
KB
where k = sensitivity for species A, amperes/torr
A
k_ = sensitivity for species B, amperes/torr.
Mass discrimination effects have been studied most extensively by
Sharma et al. The enrichment factor may or may not be a function
of' the source conditions (flame conditions at the sampling point in
this case) depending upon how the system geometry and source conditions dic-
tate the point of transition from continuum to free molecular flow. Although
the study of these effects in this system is not complete, preliminary experi-
ments at room temperature with a mixture approximating the burned gas compo-
sition of the acetylene-oxygen flames studied here Indicate that enrichment
factors for the major species except that for hydrogen are relatively insen-
sitive to source conditions. An increase in source density by a factor of
five resulted in a decrease in a CQ of 23% and changes in all other enrich-
ment factors of less than 12%. Further experimental work to assess the effect
of gas composition and source density on the enrichment factors is planned.
The calibration procedures for mass discrimination in the sonic beam
and for mass spectrometer sensitivities are carried out at room temperature.
The effect of temperature on these two calibration steps must be considered.
The mass spectrometer relative sensitivities in some cases might be functions
of temperature. In conventional (residual gas analyzer) mass spectrometer
ionizers variation with temperature of absolute ionizer sensitivities for to-
tal ionizatlon have been found to be due only to the effects of temperature
on gas density and the speed of the neutral molecule (which effects ion
collection efficiency)« implying that actual total ionizatlon cross-sections
g
are not temperature sensitive. In the type of molecular beam ionizer used
here the neutral beam molecules are directed along the axis and toward the
quadrupoles, therefore the ion collection efficiency might be expected to be
high and less sensitive to translatlonal temperatures perpendicular to the
beam axis. Although total ionization cross-sections are temperature indepen-
dent, fragmentation patterns and therefore mass spectrometer sensitivity
192
-------
(when u specific ion is used to follow concentration changes) is highly
9
dependent upon the vibrational energies of the molecule-ion . As temperature
increases more fragmentation occurs. With current knowledge of vibrational
relaxation processes in supersonic expansions* vibrational energies are
impossible to predict and one must if possible avoid the use of high electron
energies where fragmentation occurs. The appearance potentials of fragment
10
ions, like heats of reaction, seem to vary but little with temperature.
So, if fragmentation does not occur in room temperature calibrations it is
not likely to occur in the case of the flame. However, in spite of these
arguments, care must be taken in the interpretation of the data and simple
experiments must be done to check temperature effects on mass spectrometer
sensitivities due to fragmentation effects wherever possible.
The effects of temperature on the mass discrimination enrichment factors
must be assessed using the models designed to handle the prediction of enrich-
ment factors for pressure diffusion and Mach number focussing. * Analysis
of these models suggests that these mass discrimination effects depend upon
temperature in only the way in which it effects the source density. Therefore
enrichment factors from room temperature calibrations carried out at a reduced
pressure to reproduce the source density conditions should apply to the
higher pressure and temperature conditions in the flame.
Ma j or Stable Species
In accordance with the considerations discussed above, the calibration for
most major stable species involves the measurement of enrichment factors and
relative mass spectrometer sensitivities. With this calibration information
ratios of mole fractions in the flame are calculated from ratios of signal
intensities. In the flames studied here 5 mole% argon has been added to the
unburned gases as an aid in following density changes in the flame front.
Since calibration of this system for water is very uncertain the following
relationships are used to calculate mole fractions from ratios of mole
fractions in the tail of the flame, where diffusion effects are negligible:
Ar
CO
V + V
an oxygen balance
193
-------
x.
CO
an argon balance
n.
Ar
and from experimental measurements
X X X
"O ^*O*^n ^1
CO
CO,
X. ' X.
Ar Ar
X. ' X 'X.
Ar Ar Ar
where n /n Is the ratio of initial number of moles to final number of moles
and (X,)_ refers to the mole fraction of species A in the unburned gas. Carbon
A I
and hydrogen balances can then be used to check the consistency of the data
j^
and calibrations. From the calculated ratio H.O in the tall of the flame
*IT
and experimental HLO ratio it is possible to obtain a calibration factor for
Ar
water:
where I = signal intensity of species i, amperes
X. = mole fraction of species i
c,, n . = calibration factor for water relative to argon.
H_u,Ar
Now with this additional piece of experimental information the oxygen and argon
balances are not needed in the reaction zone where steep gradients require the
use of diffusional terms. The assumption is made that C
is constant
,- ,
throughout the flame (i.e. the enrichment factor Is not a function of gas
density or composition and mass spectrometer relative sensitivity factors are
not functions of flame position.)
194
-------
Higher Molecu1ar We igh t Hydrocarbons
Due to the difficulty of introducing compounds that are liquids of low
vapor pressure and solids at room temperature into the system to obtain mass
spectrometer sensitivities and mass discrimination enrichment factors, all
higher molecular weight hydrocarbon species measurements are made on a
relative basis. To account for any changes in mass spectrometer relative
sensitivities that might occur,the fragmentation pattern at 20 eV electron
energy of perfluorotributylamine is recorded before and after each experiment.
EXPERIMENTAL PROCEDURE
To check out the experimental technique and procedures several fuel-rich
acetylene-oxygen flames have been investigated. All flames had a cold gas
velocity of 50 cm/sec and burned at a pressure of 20 torr. Five mole percent
argon was added to the unburned gas. Four fuel equivalence ratios were
investigated, - 1.5, $ = 2.0, = 2.4 (sooting limit) and $ = 3.0. Complete
profiles were made for 4> - 1.5, 2.0 and 2.4. Two points near the end of the
oxidation zone were sampled in the = 3.0 flame.
The mass spectrometer resolution controls were set to give unit resolution
from 18 to 502 a.m.u. The electron energy was set at a nominal 20eV which
corresponded to an actual electron energy of 17.1eV according to argon ioniza-
tion potential measurements. At these conditions, room temperature
fragmentation of 0?, CO., CO and H20 does not occur. For acetylene, peak 25 is less
than 1% of peak 26. Mass spectrometer sensitivity and beam enrichment factor
calibrations were made before and after each flame run. Two sweeps across
the mass range 0 - 230 a.m.u. were made at each flame position at a speed of
^ 1 a.m.u./sec.
RESULTS AND DISCUSSION
Profiles of signal intensities relative to argon for the $ = 2.4 flame
are shown in Figures 5 and 6. The mole fractions of the major stable species
in the burned gas, as calculated from the procedure described above, are included
in Figure 5.
195
-------
The experimental points with the most scatter are the hydrogen and water
profiles. The characteristics of quadrupole mass spectrometers make it
difficult to have a high sensitivity for hydrogen and high molecular weight
hydrocarbon species simultaneously. To improve upon this measurement
different tuning conditions can be used for the low molecular weight gases
and the high molecular weight hydrocarbons. The noise in the water signal
is due to a large background peak at 18 a.m.u. The additional cryogenic
pumping in the ionizer region should improve this measurement.
The general features of the profiles are in agreement with the measurements
11 12
on a similar flame by Homann and colleagues. ' About 5.5 mole % of the
acetylene is present in the burned gases. The other hydrocarbons in the
burned gases are mostly polyacetylenes. They Increase through the early
part of the oxidation zone and maximize at about 1 cm. Detectable amounts
of polyacetylenes up to CgH are still present in the burned gas. Polyacety-
lene formation is preceded in the oxidation zone by vinylacetylene (C^H,) and
a C,H.. compound that could be benzene or a straight chain hydrocarbon. These
o 6
compounds have concentration maxima that are at least an order of magnitude
lower than the polyactylene of the same carbon number.
The interaction between the major stable species and the hydrocarbon
species throughout the flame is complex and interesting. Oxygen is completely
consumed by about 1.4 cm. At this point CO and H2 have nearly reached their
final values. C0? and H_0 molefractions peak near 1 cm and decrease slightly
i
into the burned gas. Acetylene continues to decrease until about 2.5 cm above
the burner. These basic features can be explained qualitatively by the
following scheme of competing reactions early in the oxidation zone:
196
-------
Reactions leading to polyacetylenes and C0? and H?0 are faster than reactions
involving the same radicals leading to CO and H_. This ties up oxygen in
the form of CCL and HO and produces super equilibrium amounts of polyacety-
lenes. Near the end of the oxidation zone where the temperature reaches
its maximum, reactions producing OH, CO and H» from CO- and H.,0:
£* £ £,
H20 + H -> OH + H
C02 + H -» CO + OH
become important and the oxidation of polyacetylenes becomes faster than their
13
production. The oxidation rate decreases quickly because of the drop in OH
14
radical concentration leaving some polyacetylenes in the burned gas.
The features of the flame with 4> = 2.0 were essentially the same but
with lower CO, H?, acetylene and polyacetylene concentrations as might be
expected. At = 1.5 no acetylene or polyacetylenes were detectable in the
burned gas.
Signals were observed at masses other than those shown in Figures 5 and
6. Table I lists those masses at which positive identifications were not made
or contributions from several species were not sorted out. Masses 15, 16 and
17 peak in the middle of the oxidation zone. The most probable contributor to
peaks 15 and 16 is methane in this fuel rich flame. Masses 29 and 30 are
maximum at the lowest sampling point, 0.17 cm and the contribution of H2CO to
peak 29 has not been determined at these mass spectrometer conditions. Mass
34 appears only at the lowest sampling point. Mass 39 maximizes at 0.62 cm
from the burner. Mass 42, probably corresponding to propylene,maximizes at
0.35 cm. Appearance potential measurements along with measurements at lower
electron energies to prevent contributions from fragmentation of hydrocarbons
must be made to make positive identifications.
Mass spectra were observed at two points near the end of the oxidation
zone of a sooting flame of fuel equivalence ratio, = 3.0. In addition to
the species observed in the = 2.4 flame several higher molecular weight
aromatic hydrocarbons were observed. Their masses, molecular formula and
possible structures are listed in Table II. The sensitivity for higher mole-
cular weight hydrocarbons (100 a.m.u. and above) should be increased by an
order of magnitude when the cryogenic pumping in the ionizer region is added
197
-------
since the present sensitivity is limited by noise in the beam spectrum caused
by high levels of hydrocarbons in the background.
SUMMARY
A molecular beam mass spectrometer system has been developed for studying
sooting flames. Preliminary measurements suggest that molecular beam mass
discrimination effects may be relatively insensitive to source conditions
and that absolute concentration measurements of major stable species profiles
will be possible.
Profiles of relative signal intensities of several non-sooting and barely
sooting acetylene-oxygen flames have been made. The results are in agreement
with those of previous investigators. Polyacetylenic hydrocarbons up to mass
146 were detected and are the major hydrocarbons present other than the
unburned fuel. Maxima in the carbon dioxide and water profiles are supportive
of the suggestion that reactions that store oxygen in these species are rapid
compared to reactions that lead to CO formation from the fuel in the early part
of the flame. At the higher temperatures present at the end of the oxidation
zone this oxygen is released in the form of OH that attacks polyacetylenes and
unburned fuel.
Under strongly sooting conditions (4 = 3.0) aromatic hydrocarbons have been
detected. The addition of cryogenic pumping in the ionizer region is expected
to increase the sensitivity for these species by an order of magnitude.
ACKNOWLEDGEMENTS
This work has been done under EPA Grant No. R 803242.
198
-------
REFERENCES
1. Butze, H.F. and R.C. Ehlers NASA Tech. Mem. NASA JMX-71789. Paper presented
at Western States Section of the Combustion Institute, Palo Alto, CA.
Oct. 20, 1975.
2. Schirmer, R.M. in "Emissions from Continuous Combustion Systems", W.
Cornelius and W.G. Ahnew, ed., p. 189, Plenum Press, N.Y. (1972).
3. Longwell, J.P., "Synthetic Fuels and Combustion" Plenary lecture presented
to Sixteenth Symposium (International) on Combustion, Cambridge,
Massachusetts, August, 1976.
4. Biordi, J.C., C.P. Lazzara and J.F. Papp, Combust. Flame 2_3, 73 (1974).
5. Wersborg, B.L., J.B. Howard, and G.C. Williams, Fourteenth International
Symposium on Combustion, p. 929, The Combustion Institute (1973).
6. Knuth, E.L., "Direct Sampling Studies of Combustion Processes" in "Engine
Emissions, Pollutant Formation and Measurement", G.S. Springer and
D.J. Patterson, ed., p. 319, Plenum Press, N.Y. (1973).
7. Sharma, P.K., E.L. Knuth and W.S. Young, J. Chem. Phys. 64, 4345 (1976).
8. Field, F.H. and J.L. Franklin, "Electron Impact Phenomena and the Properties
of Gaseous Ions," pp. 202-203, Academic Press, N.Y. (1970).
9. Ibid. pp. 203-204.
10. Ibid., p. 81.
11. Homann, K.H. and H.Gg. Wagner, Berichte der Bunsengesellschaft 69, 20 (1965).
12. Bonne, U., K.H. Homann, and H. Gg. Wagner, Tenth Symposium (International)
on Combustion, p. 503, The Combustion Institute (1965).
13. Homann, K.H. Combust, and Flame 11, 265 (1967).
14. Bonne, U., H. Gg. Wagner, Berichte der Bunsengesellschaft 69. 35 (1965).
199
-------
Table I. MASSES AT WHICH SIGNALS WERE OBSERVED BUT POSITIVE IDENTIFI-
CATIONS WERE NOT HADE.
2.4, P = 20 Torr, V = 50 cre/s
MASS
POSSIBLE CONTRIBUTING SPECIES
14
15
16
17
25
29
30
34
38
39
42
CH2
CH3,CH4
CH4 (MOST PROBABLE), 0
OH
r u r* LJ
i/p n | up np
HCO , H2CO, CI30
H2CO, C2H6 (UNLIKELY)
H202
C3H2
C3H3 ' C3H3
C3H6
200
-------
Table II. ADDITIONAL SPECIES DETECTED NEAR THE END OF THE REACTION
ZONE IN A C2H2/02 FLAME, * = 3.0, P = 20 Torr, VQ = 50 cm/s
MASS MOLECULAR POSSIBLE STRUCTURE
FORMULA SPECIES
CH,
92
102
104
118
128
130
C7HQ
CsH6
CB^S
C9HIO
CIOH8
^inH,n
TOLUENE lOJ
PHENYLACETYLENE (6)
LJ /"* " f* LJ
nu~ wno
STYRENE (g)
^"^^ HCHCHo
METHYL STYRENE (Qj
^^CH3
NAPHTHALENE ©©
PHENYL BUTADIENE fOTJu
DIHYDRONAPHTHALENE
142 C|,H,0 METHYL NAPHTHALENE
146 C(2H2 HEXACETYLENE
CH,
201
-------
SUMMARY OF PROPOSED MECHANISMS OF SOOT FORMATION
o
UJ
QC
O
O
<£
DEHYDROGENATION
PARAFFINIC
FUEL
\
MOLECULES
'2H2-Z
HYDROCARBON
IONS
POLYMERIZATION
AND
DEHYDROGENATION
CONDENSATION
SOOT
NUCLEUS
STABLE
PC AH
BY-PRODUCTS
PCAHCPOLYACETYJ^NEg
/
SURFACE
GROWTH BY
DECOMPOSITION
PHYSICAL
CONDENSATION
I
DEHYDROGENATION
AND
POLYMERIZATION 1
POLYMERIZATION
!£MI lUt
SATURATED OR
UNSATURATED
POLYMERS
LIQUID DROPLETS CARBON PARTICLES
CONTAINING * WITH GRAPHITE-
POLYCYCLIC COMPOUNDS LIKE STRUCTURE
Figure 1. Summary of proposed mechanisms of soot formation
202
-------
ELECTRON
MULTIPLIER
QUAORUPOLE MASS
FILTER
IONIZER
LN2-COOLED
WALLS ~
TUNING FORK
CHOPPER
CALIBRATION GAS
EFFUSIVE
SOURCE f
1
6 INCH
DIFFUSION
PUMP
4 INCH
DIFFUSION
PUMP
COLLIMATOR
6 INCH
DIFFUSION
PUMP
SKIMMER
QUARTZ NOZZLE
BURNER
MECHANICAL
VACUUM PUMP
b=» PREMIXED GASES
Figure 2. Molecular beam mass spectrometer system
-------
o
4J
<0
M
c
O)
O
0>
Q.
Vt
O
U
*«~
M
cn
o>
«^
U-
204
-------
MECHANICAL
VACUUM
PUMP
1
1
^M
i
c
ALUMINA
TRAP
OTA i hi i roc __, ' n"r
STAINLESS \
STEEL |
SINTERED t
DISC \ L
1
1
DIFFUSION /
3UMP /
MKS BARATRON
PRESSURE
TRANSDUCER
k EFFUSIVE
SOURCE-^
J°i°
1* " t
L PTA
L is^xi
MIXING
VOLUME
M.S.
STAGE
^r°
2ND
STAGE
1ST
STAGE
BURNER
CHAMBER
> >
llll CRITICAL
- -- E-E + ^E E^ ORIFICE
T T T T T FLOW METERS
:0 COo H0 Co Ho Oo Ar
Figure 4. Schematic of gas introduction system for studying mass
discrimination
205
-------
o
o
cr
LU
LU
or
CO
Z
HI
h-
z
CO
C2H2/02/Ar P = 20TORR
0.469/0.485/O.O5 VQ= 50CM /S
30xH2
3xC02 Xco =0.061
I 2 3
DISTANCE ABOVE BURNER, CM
Figure 5. Profiles of signal intensities realtive to argon for major
stable species in an acetylene-oxygen flame near the sooting
limit.
-------
2.0
o
o
a:
.5
>
t*
LJ
* 1.0
>-
z
LU
h-
0.5
Z
O
en
C2H2/02/Ar P= 20 TORR
0.469/0.485/0.05 V0 = 50 CM/S
1234
DISTANCE ABOVE BURNER , CM
Figure 6. Profiles of signal intensities relative to argon for minor
species in an acetylene-oxygen flame near the sooting limit.
$ = 2.4, P = 20 torr, V = 50 cm/s, 5m% argon
207
-------
-------
INVESTIGATION OF NOX, NITRATE AND SULFATE PRODUCTION
IN LABORATORY FLAMES
By:
D. J. Seery, M. F. Zabielski, and L. G. Dodge
United Technologies Research Center
East Hartford, Connecticut 06108
209
-------
-------
SECTION 1
INTRODUCTION
This research program is directed toward an understanding of the
chemical kinetics of the formation of NOX, nitrates and sulfates in lab-
oratory flames. In the first year, efforts will be directed toward
elucidating the mechanisms of "prompt" KO formation and NO-WOg conversion.
As part of the KO-N02 conversion study an investigation will be made of
the possible perturbation of the JUK^ concentration by chemical reactions
in the sampling probes.
The experimental procedure for this program consists of measurements
of temperature and composition profiles obtained on a low pressure, flat
flame burner and subsequent analysis of these data in terms of chemical
mechanisms. Temperature profiles are obtained using miniature thermo-
couples and gas samples are extracted from the flame for concentration
analysis using both a molecular beam sampler and quartz microprobes.
The samples extracted by the microprobes will be analyzed by mass spectro-
meter and chemilumine scent analysis (for NO and NOX). The molecular beam
sampler is combined with a TOF mass spectrometer for analysis of both
stable and unstable species. Molecular beam sampling has been selected
as a standard for comparison with the microprobe since this technique
seemed the one most likely to provide an unperturbed sample of flame gases.
In addition to probe sampling, measurements of HO and OH are being
carried out using optical spectroscopy in order to provide an independent
check on the accuracy of the concentration measurements of these species.
211
-------
SECTION 2
EXPERIMENTAL
BUKHERS AND ACCESSORIES
Two identical flat flame burners are used in the UTRC Flame Laboratory.
One burner pictured in Fig. 1 is housed in a 6-inch Pyrex cross and is
used for measurements involving optical spectrometry and microprobe sam-
pling. The second burner is housed in a high vacuum chamber which is
attached to the molecular beam sampler. A picture of this apparatus
is shown in Pig. 2. Both burners have 3-inch diameter, water-cooled,
sintered copper surfaces which provide flat flames with negligible radial
gradients. The burners can be translated vertically by low-geared motors
and the location is measured with a cathetometer to ± .OOOV. Cooling
water to the burners is measured on entering and exiting and the exit
temperatures are maintained at 50 ± 5°C. Cooling water to other parts
of the burner housing is controlled so as to avoid water condensation.
Gas flow rates are measured by critical flow jewelled orifices. The
upstream temperature and pressure are monitored continuously and the mass
flow rates can be reset to ±1$. Both chambers are evacuated through 1.5"
lines by a 30 CFM Kinney KD-30 pump. This pump is isolated by a dry ice-
acetone trap and provides a two hour run time without water contamination
of the pump oil.
Temperatures are measured by means of fine thermocouple probes. For
temperatures below 1900°K, 0.003" Pt/Pt-10 percent Rh thermocouples, coated
with a mixture of BeO/YgO^ to prevent catalytic reactions (Eef. l) are used.
212
-------
Above 1900°K, Ir/Ir-ljO percent Rh thermocouples are used. The thermocouple
probe is mounted in a flexible bellows housing so that it con be translated
both vertically and horizontally. Thermocouple data are rapidly obtained
with a Fluke 2100A Digital Thermometer. Both types of thermocouples are
flame-welded so that the bead diameter is approximately O.OOU". The
large radiation correction caused in part by the high emissivity of the
coating, poses a serious problem, since the theoretically calculated
corrections (Ref. 2) have large uncertainties. Because of this uncertainty
the radiation correction factor for these thermocouples is obtained by a
detailed comparison with sodium line reversal measurements at several
temperatures and extrapolated to other temperatures using the Stephan-
Boltzmann relationships. The emissivity of the BeO/YgOo coating has been
measured to be 0.6U ±.02. Using this value radiation corrections were
calculated which deviated from the measured correction by up to a factor
of two depending on the assumptions used.
An example of the thermocouple measurements ia presented in Fig. 3«
The data are for a stoichiometric methane air flame burning at a pressure
of 76 torr. At two locations in the flame., radial temperature profiles are
shown which indicate that the flame does indeed have a flat profile. The
dashed line above the temperature data is from a thermocouple with a
cracked coating. The increase in temperature is caused by catalytic reac-
tions at the nobel metal surface and is a convenient indication of a coat-
ing failure.
Concentration profiles for OH are being measured using the A^n-X^jXOjO)
band at 306U A. A water-cooled hollow cathode lamp operating with flowing
Ar saturated with water vapor at a total pressure of 2 torr and a current
of 100 mA is used as the source of narrow-line OH radiation. The vertical
spectrometer slit is rotated by a quartz dove prism to a horizontal image
in the flame with a resulting vertical resolution in the flame of better
than 1.5 ram. A 0.5 meter spectrometer with a 1200 g/mm grating operated
213
-------
in second order is used to achieve a resolution of better than 0.15 A.
The optical system is shown schematically in Fig. ^. Included in the dia-
gram is a multipass system which is required for measurements of KD at
low concentrations. Rotational temperature measurements are made to assure
rotational equilibrium and that concentration measurements of OH may be
made from single spectral line measurements.
A concentration profile for OH as shown in Pig. 5 was determined for
the region from 2 mm to Itf) mm above the base of the burner by averaging
values for the three lines RgCS)* ^C1*)* and ^(lU). Data for both temper-
ature and concentration were reduced assuming uniform temperature and con-
centration profiles across the burner.
MCCROFROBE SAJPLIKG
Micrbprobe sampling of low pressure flames can be discussed in terms
of two problems; (l) perturbation of the flame due to the sampling probe
and (2) perturbation of the gas sample by chemical reactions in the probe.
The first problem, perturbation of the flame, can be minimized by
working with very fine, uncooled-probes to decrease both the aerodynamic
and the thermal disturbance. The design and operation of such probes has
been discussed in detail by Pristrom and Westenberg (Ref. 3). It should
be pointed out however that even for very smallrnmicrciprobes there is a
measurable thermal sink which can have significant effects on the tempera-
ture history of the gas sample. Figure 6 presents measurements in a CHl|.-Air
flame of the gas upstream of a microprobe. The temperature decrease with
a water-cooled probe is, of course, larger but in any case this temperature
change can and should be measured.
The perturbation of a gas sample by chemical reaction in the probe is
a poorly understood phenomena (Ref. U). Although the initial quench rate
in a probe can be readily calculated, because of shock waves and boundary
layer effects in the probe the actual aerodynamic conditions are uncertain.
214
-------
Calculations using the UTRC (Ref. 5) probe deck indicate that for many cases
the gas temperature in uncooled probes rapidly returns to the stagnation
conditions. Chemical reactions in probes are most important for species
present in low concentrations. For example, the concentration of a major
product like C0g is not likely to be affected by 0.1 % of 0 atom, whereas
the NO or HO concentrations which can be comparable to the 0 atom can be
greatly altered.
The micro-probes used in this study are similar to those described in
Ref.3 and by Merryman and Levy (Ref.io), that is, quartz with 75 to 100
micron orifices both cooled and uncooled. These probes appear to provide
reliable gas samples for major constituents based on atom mass balances.
The calculated temperature profile shown in Pig. 7 however indicates that
the rate of quenching may not be rapid enough to preserve the original
concentrations of minor species. The microprobes are attached to a 0.5"
diameter, teflon-coated, aluminum line which leads to the specific gas
analyzers or a mass spectrometer. Because of the facility's design, sample
line lengths can be kept to six feet or less. The sample line is maintained
at 125°C to minimize adsorption and condensation of water.
For microprobe measurements, HO and NO are determined with a Thermo-
«rv
Electron Corporation (TECO) Series 10 Chemiluminescent Analyzer. This
instrument has a stated minimum detectable concentration of 50 ppb and a
linearity of ±1 percent. Maximum NO concentration is 10,000 ppm. For NOX
measurements, a TECO Jfodel 300 Ifolybdenum NOX Converter is used.
MDLECULAR BEAM SAM>IJBG
A sampling technique that does not have the major disadvantages of
microprobes is the molecular beam sampler. Figure 2 is a picture of the
molecular beam sampler and Fig, 8 schematically shows the flat flame burner,
molecular beam sampler and mass spectrometer ion source. The pressure in
the reaction chamber is maintained at 76 torr for alT flames. The gases
215
-------
sampled from the flame are supersonically expanded through the nozzle ori-
fice into the nozzle chamber. The mean pressure in this chamber is typically
10~3 _ icf1* torr. The nozzle is a 120°(Ho°) "hybrid" (Ref. 6) made of quartz
with a .05 cm orifice. Because of the large pressure gradient across the
orifice, the hot sampled gases experience a dramatic decrease in temperature
that "freezes" most chemical reactions. The central core of this super-
sonic expansion is then isolated by a skimmer with about a 0.10 cm orifice.
This skimmed core constitutes the molecular beam that traverses the chamber.
The pressure in the skimmer chamber is maintained at approximately 1 x 10~"
torr. The beam is then collimated and enters the ion source of the mass
spectrometer. Typically, the pressure within the mass spectrometer chatriber
is 5 x 10"^ torr.
The photograph in Fig. 2 shows all of the significant features of the
apparatus as presently constituted. Also shown in this photograph are an
optical access port (rear port not visible), cooling water thermometers,
six-inch diffusion pumps, and associated pluiribing. The nozzle and skimmer
chambers are separated by a diagonal wall. This is done to insure maximum
pumping speed for the nozzle chamber, which handles the bulk of the gas
flow. The vacuum line connecting the nozzle chamber to the diffusion puap
has a six-inch diameter while the skimmer chamber line has a three-inch
diameter line. Nozzle Knudsen numbers of 10"^ > K^, 2 2 x 10"3 are desired
to minimize the boundary layer at the sampling orifice. In the present
study the molecular beam has been designed to have KQ = 2 x 10"2.
A photograph of a methane-air flame in the beam sampling apparatus
is shown in Fig. 9-
MASS SPECTROMETER
The primary analytical instrument used in the low pressure flame in-
vestigations is a 1 meter, bipolar 1 KeV, time-of-flight (TOF) mass speetro-
216
-------
meter designed and built at UTRC. The instrument has a normal resolving
power of 200, and a maximum value of yOO. All of the ion sources for this
instrument employ electrostatic focusing of the ionizing beam which provides
superior low mass sensitivity in comparison with the magnetic focusing found
on commercial instruments. This is particularly useful in H and Kg measure-
ments. Also, the ionizing electron energy can be automatically scanned to
further facilitate appearance potential measurements.
The vacuum system is sufficient to limit the residual gas background
pressure to about 5-10 x 10~9 torr. Hence, with a source pressure of
1 x 10~3 torr, the interference at a given mass is usually of the order of
or better.
217
-------
SECTION 3
DATA PROCESSING AMD ANALYSIS
In order to facilitate the data reduction and to improve the precision
of the data, the mass spectrometer has been coupled with a Northern Scienti-
fic NS-575 signal averager. In the measurements that have been made, typi-
cally 32 spectra per sampling location have been averaged. This corresponds
to an improvement in the signal-to-noise ratio by a factor of 5«7« Using
this method, the Ar3° ana Ar3° isotopes in air can be detected. This corre-
sponds to approximately 30 and h ppm, respectively, without resorting to
ion counting.
The data from the signal averager is recorded on 7-track magnetic tape
and then transferred to a Digital Equipment Corporation PDP-6 time-share
computing system. A schematic diagram of the burner, sampler, mass spectro-
meter and data handling equipment is shown in Fig. 10.
Once the data are. in the computer, two separate programs are used in
the actual reduction. The first program consists of the following sections:
(l) a fast Fourier transform (FFT) routine to analyze the noise content of
the spectra; (2) Martin-Graham digital filter to smooth the data; (3) a
routine for baseline subtraction; (1*) a mass marker routine; (5) a peak area
and height reading routine, and (6) a deconvolution routine for unfolding
overlapping peaks. The deconvolution routine is based on both a Fourier
and a modified Gauss-Seidel iterative technique. In addition, the iterative
technique has been shown to work in situations Where the Fourier techniques
fail (Ref. 7); thus, even when small peaks reside on the shoulders of large
peaks, quantitative measurements can be made. The second program is used
218
-------
to invert the data matrix generated by the first program. This data matrix
can be fairly complex, depending on the range of masses observed and the
constituents of the gas being analyzed. Frequently, more than one species
contributes to a given mass peak; fortunately, however, the observed peak
is the result of the linear superposition of all the ion currents at the
particular mass. The overlap of ion currents at a particular mass normally
arises from two or more species having the same nominal mass like ethylene,
nitrogen and CO at mass 28 and from species that fragment under electron
impact such as the ethylene fragment from ethane.
The value of mass spectrometry data obtained from flames is highly
dependent on calibration of the experiment and control of minor variables.
In the present experiments checks have been made on the effects of varia-
tion in chamber pressure, changes in the temperature drop in the water-
cooled burner and temperature drop in the flame gases because of the heat
sink of the sampling probe. The concentration data are checked for mass
balance consistency and, of course, the mass spectrometer sensitivity and
the sampling line-efficiency are determined for each species to be measured.
The calibration of the microprobe and analytical instrumentation is
accomplished by using the critical orifice gas metering system for major
species and premixed gas bottles for minor species. Several concentrations
representative of the expected flame concentrations are normally chosen.
This is required because of the limited dynamic range of the specific gas
analyzers and also as a check on the linearity of both the analyzers and
the mass spectrometer.
The calibration of the molecular beam is considerably more complicated.
First, for the specific range of flame conditions (pressures, temperatures),
the optimum nozzle orifice-skimmer orifice distance is determined. If this
distance is too small, skimmer interaction with the central core of the
free jet expansion occurs. If this distance is too large, then skimming
is performed beyond the Mach disk which results in low beam intensity. The
optimum location for low pressure flames is usually in the region of transla-
219
-------
lational freezing point. Once the optimum beam intensity is determined,
the burner is removed from the chamber and a quartz chamber and furnace
are installed for heating stable species. This high temperature calibra-
tion is required because molecular species can be "frozen" in excited
states which, consequently, change the ionization potential of the molecule,
the appearance potential for certain reactions, and the fragmentation
pattern even at 70 eV (Ref.'S). Measurements of the temperature dependence
of the methane fragmentation pattern are presented in Pig. 11. For the
conditions of the present study the 16/15 ratio for methane does not appear
to be dependent on temperature. The net result of this calibration is a
set of sensitivity factors and fragmentation patterns that are incorporated
into the data reduction program.
A difficulty in analyzing mass spectrometer data is illustrated in
Fig. 12, which shows the signal averager data for a methane-air flame in
the mass range from 27-31- In (a) the large single peak is at mass 28 and
is predominantly attributable to N^ with only a small contribution from CO.
The contribution from COg is negligible for this case. As the sensitivity
of the scale is increased, (b) and (c), the peak at mass 29, which is mostly
lf^5jf" becomes apparent. Also in (c) the mass 30 peak can be seen,and with
increasing sensitivity (d) it can be conveniently measured. The mass 30
peak is composed of contributions from NO, Cr^O-"* and IT^lr'-'and accurate
data is required on the latter two species in order to calculate the UO
concentration.
In analyzing concentration data from flames, the diffusion velocity
is a major correction and cannot be neglected. A major advantage of low
pressure flames is that the concentration gradients are more gradual than
in atmospheric flames and thus the diffusion velocity can be calculated
with greater accuracy.
The diffusion corrected mass flux data are used to calculateathe rates
of formation and disappearance of the individual species. Reaction mechan-
220
-------
isms and rate constant data can be obtained from this data by the kinds
of analysis exemplified in the works of Fristrom and Westenberg (Ref. 3)
and Fenimore (Ref. 9)? or by matching the data with model calculations.
221
-------
SECTION U
REFERENCES
1. Kent, J. H. : Combust. Flame 1*4-, 279 (1970).
2. Kaskan, W. E. : Sixth Symposium (international) on Combustion.
Rheinhold, New York, 1?U (1975).
3. Fristrom, R. M. and A. A. Westenberg: Flame Structure. McGraw-Hill,
New York (1965).
U. Lengelle, G. and C. Verdier: Gas Sampling and Analysis in Combustion
Phenomena. AGAKDograph No. 168, AGAKD, Paris (1973K
5. Cohen, L. S. and R. N. Guile: AIAA Journal 8, 1053 (1970 K
6. Biordi, J. C., C. P. Lazzara and J. F. Papp: Combust. Flame 23 t 73
(197^).
7. Mine, T. A., J. E. Beachey and F. T. Greene: J. Chem. Phys. 56,
3007 (1972).
8. Zabielski, M. F. and T. M. McHugh: Proceedings of 21st Annual Confer-
ence on Mass Spectrometry and Allied Topics, 198 (1973).
9. Fenimore, C. P.: Chemistry in Premixed Flames. MacMillan, New
York (196U).
10. Merryman, E. L. and A. Levy: Fifteenth Symposium (international) on
Combustion. The Combustion Institute, 1073 (1975)
222
-------
FLAT FLAME BURNER
Fig. 1
74-37
7C 04 151 ?
-------
MOLECULAR BEAM SAMPLER
FIG. 2
MASS SPECTROMETER
SKIMMER CHAMBER ^^Kf> /
224
76-04-151-1
-------
FIG. 3
LL
EC
UJ X
X 7
h- a
225
77 0V 51 :i
-------
UJ
g
X
o
o
DC
K 2
Z DC
W D
S CD
^ W
5 s
« ^
< i
LU "
O u.
CO
CO
o
a.
O
FIG. 4
77 07 61 A
226
-------
FIG. 5
DC
h-
Z ^
Z <="
UJ
QC
UJ
>
u
CN *
of
UJ
CM Z
CN £
oo
i
00 <
«- yj
O
(A
5
00
Wcfd eOf< NOI1VU1N30NOO HO
227
77-07-28-1
-------
FIG. 6
UJ
00
O
DC
a.
.
i
a
a
LU
8
o
D
GO
oc
DC
UJ
a.
LU
CC
QC
LU
i
LU
Xo '3UniVH3dlAI31
228
77-07-28-3
-------
FIG. 7
UJ
00
O
oc
a.
O
cc
O
5
cc
O
LL
O
UJ
CC
CC
UJ
a.
S
UJ
77-07-21'. 4
-------
FIG. 8
SCHEMATIC REPRESENTATION OF MOLECULAR BEAM MASS SPECTROMETRIC SYSTEM
TIME-OF-FLIGHT
MASS SPECTROMETER ( j
ION SOURCE ! 1
MOLECULAR BEAM CO
COLLIMATOR S
RE
LIT >-
TUNING FORK CHOPPER
tKIMMFR
FREEJET
IMOZZLI: "
\
\
i r
x
I
1
h
1
fi
vAsi
Sv
\\
I1'
1
1
4
1
y
IONIZATION VOLUME
i- TO ION DETECTOR
MASS
SPECTROMETER
CHAMBER
P3
^ SKIMMER
CHAMBER
P2
1
V
P1
po
TO MECHANICAL
VACUUM PUMP
i
/
'-BURNER
NOZZLE
"* CHAMBER
^ , R POTION
CHAMBER
230
N12-181-1
-------
Fig. 9
231
-------
FIG. 10
K
UJ
*? >-
^2
U
X
2£ /-
o 3 2
uj 5
C
UJ
8
-
< in
tr M
uJ 2
ETER
cc
t-
o
UJ
Q.
CO
u.
O
UJ
77_07-61-1
-------
TEMPERATURE DEPENDENCE OF METHANE FRAGMENTATION
FIG. 11
LO
to
O
1.30
1.28
1.26
1.24
1.22
t.20
AVERAGE OF 9 PTS
IONIZATION ENERGY * 856V
I
200 400
600
GAS TEMP, °K
800 1000
233
77-07-28-2
-------
FIG. 12
Q
£ -
S s
O iu
OC ">
rr «
H <
O 5
S1 *
55 £
I 1
Oil. U O 3 Z t- «0
z
e
i I
i 8
1
OIL O O
+r
i
T
*
O to. U O 9
oik w o a z «- w
234
77 07-61 -2
-------
INFLUENCE OF AERODYNAMIC PHENOMENA ON POLLUTANT
FORMATION IN COMBUSTION
Phase II Liquid Fuels
By:
L. J. Spadaccini, J. B. McVey, 0. B. Kennedy, F. K. Owen,
C. T. Bowman, A. Vranos, A. S. Kesten
United Technologies Research Center
East Hartford, Connecticut 06108
235
-------
-------
ABSTRACT
An experimental investigation of the effects of the interaction between
physical and chemical processes on pollutant formation and destruction in a
liquid fuel turbulent diffusion flame burner has been carried out. In this
investigation, the effects of fuel type, inlet air swirl, inlet air tewpera-
ture and combustor pressure on the spray characteristics and the time-mean
and fluctuating flow field structure have been determined using probing and
optical techniques. Changes in the spray and flow field structure have been
correlated with changes in pollutant emissions from the burner. The results
of this investigation show that variation of these operating parameters pro-
duce major changes in spray dynamics and vaporization rates and in the time-
averaged fuel/air distribution within the burner which significantly influence
energy release rates and pollutant formation and destruction. In addition,
it was found that there are significant differences between the mean velocities
of the gas and fuel droplets which likely influence droplet vaporization rates
and mixing of the vaporized fuel and air.
237
-------
-------
ACKWOWLEDGMEMTS
A number of individuals at UTRC made significant contributions to
the experimental investigation. Dr. M. F. Zabielski and Mr. G. L. Dodge
designed the gas sampling system used in the investigation and developed
the calibration procedures employed in the gas sampling portion of the
experiments. Mr. T. A. Murrin assisted throughout the experimental pro-
gram and was responsible for operation of the combustor and for reduction"
of much of the experimental data.
-A special debt of gratitude is owed to Mr. W. S. Lanier, the EPA
Project Officer for this contract effort, for asking the critical questions
and questioning the critical answers.
239
-------
-------
SECTION 1
IWTROIXJCTION
A large number of continuous cotribustion devices, including furnaces and
gas turbines, operate on liquid fuels. Investigations of pollutant emissions
from these devices indicate that changes in injector design which change the
droplet size distribution in the spray and variations in air inlet conditions
which alter the interaction between the fuel spray and the surrounding gas
stream (l-6) can have a significant effect on pollutant formation and destruc-
tion. Norster and Lefebvre (l) found that atomization techniques can affect
pollutant emissions, particularly exhaust smoke, in a gas turbine combustor.
Grobman (2) reported that improving fuel atomization reduces hydrocarbon and
carbon monoxide emissions during idle in a wide range of conventional and
experimental gas turbine combustors. Mellor and his co-workers (3>5) and
Pompei and Heywood (I*) have attributed changes in carbon monoxide and nitric
oxide emissions from gas turbine combustion with fuel injection pressures to
changes in fuel atomization and vaporization rates. Inlet air temperature
and swirl, combustor pressure and combustor reference velocity* significantly
influence hydrocarbon, carbon monoxide and nitrogen oxide emissions (1-3,6).
Hence, it appears that appropriate modifications of fuel atomization techniques
and coiribustor inlet conditions can result in significant reductions in the
emissions of most pollutant species from continuous combustion devices.
*The combustor reference velocity, Vref , is a measure of combustor residence
time and is defined by Vref s ^ir/PairNiax T*61*6 *kir = a*r fl°w rate,
p . = inlet air density and
- maximum combustor cross- sectional area.
240
-------
The combustion of liquid fuel sprays is a complex process involving
simultaneous heat, mass and momentum transfer and chemical reaction which are
influenced by the fuel characteristics, the droplet size distribution and
number density, the relative velocity between the droplets and surrounding gas
and the ambient gas temperature and composition. Although qualitative models
and empirical correlations of pollutant emissions from liquid-fueled combustore
have been developed (see, for example, 7 and 8), our present understanding of
spray burning is insufficient to permit quantitative predictions of the effects
of changes in fuel injection techniques and operating conditions on pollutant
emissions. Investigations of burning sprays have been hampered by difficulties
associated with measuring the characteristics of the spray and with determining
the interaction of the spray with the surrounding gas stream. However,
recently-developed optical and probing techniques appear to be promising
diagnostic tools for measurements on burning sprays.
This paper outlines the results of an experimental investigation, sponsored
by EPA under Contract 68-02-1873* of the effects of several operating parameters
on the spray characteristics and flow field structure in a liquid-fuel turbulent
diffusion flame burner and the subsequent effects on pollutant formation and
destruction. This work is documented in more detail in Ref. 9«
241
-------
SECTION 2
EXPERIMENTAL APPARATUS AND INSTRUMENTATION
The experimental apparatus and approach used in this study are similar
to those employed previously to study pollutant formation and energy release
in gaseous-fuel turbulent diffusion flames (10). The only significant change
in the coiribustor configuration is the fuel injector modification required for
liquid fuel operation. A schematic diagram of the water-cooled combustion
system is presented in Figure 1. It consists of an electric resistance-type
air heater, a 12.2 cm-dia water-cooled cylindrical coiribustor section having
a centrally located pressure-atomizing fuel injector and an extension section
which contained an exhaust probe rake and a water-cooled orifice plate which
can be installed to increase the combustor pressure. As in the gaseous fuel
stady, flame stabilization in the high velocity flows investigated was
achieved by producing a recirculation zone in the initial region of the
combustor by imparting a swirl component to the airflow. Swirl was imparted
by inserting replaceable sets of straight swirl vanes into the annular passage
which surrounds the fuel injector. The trailing edges of the swirl vanes are
located upstream of the injector exit plane to permit measurement of the air-
flow characteristics entering the combustor. These measurements permit deter-
mination of the inlet conditions which fire needed for the analytical model'-
ing effort. The oombuetor was designed to permit independent variation of
each of three operating parameters inlet air swirl, combustor pressure
and air preheat which are known to significantly influence emissions from
liquid-fuel combustors.
In the present study, three fuels were investigated -- liquid propane,
n.so-octane, and No. 2 distillate fuel. These fuels provide an orderly
242
-------
progression in complexity of molecular structure and distillation character-
istics and permit an evaluation of the effects of chemical and physical
properties of fuels on pollutant formation and energy release. Typical
liquid fuel properties and the results of limited quantitative fuel analyses
are given in Ref. 9« The injector assembly was water-cooled so that the fuel
was not heated by the high-temperature inlet air, and fuel injector design
and injection pressure were chosen to ensure liquid injection. A conventional
pressure-atomizing swirl-type nozzle which produced a nominal 60 deg hollow-
cone spray with a nominal droplet Sauter mean diameter of 100 tarn in quiescent
air at atmospheric pressure was used for iso-octane and No. 2 distillate fuel.
This type of nozzle proved unsuitable for propane since vaporization occurred
internally due to expansion in the nozzle swirl chamber. Therefore, a
tangential-feed, pressure-atomizing nozzle in which the full pressure drop
occurred across the exit orifice was used to maintain the propane liquid to
the point of injection. The direction of rotation imparted by the swirlers
to both the fuel and the air streams were identical for each of the configura-
tions tested. The fuel injectors, air swirl vane designs, and the fuel
injector assembly are shown in Ref. 9.
Measurements in the combusting flow were made through 6.^-cni dia window
ports in the combustor sections (Figure l). A pair of window ports iBO^deg
apart are present at each location and permit the use of optical measurement
techniques (e.g., laser velocimetry and laser holography). The location of
a port directly downstream of the injector exit plane allowed an unhindered
view of the flame in the vicinity of the fuel injector and permitted acquisi-
tion of flow field data close to the injection plane. The combustor probing
devices are compatabile with all window ports and may replace a quartz window
or water-cooled plug in any given port. In addition, the entry section was
designed to permit axial relocation of the fuel injector between tests, thereby
increasing the number of axial locations at which radial traverse can be made.
A porous-metal disc installed in the air entry section serves to provide a
243
-------
uniform inlet flow, which was verified by laser velocimeter measurements,
and the combustor wall temperature was maintained at a constant value (~500°K)
along the entire 100 cm length by regulating the coolant flow rates.
The concentration of/nitrogen oxides (NO, HOg), oxygen (Og), carbon
monoxide (CO), carbon dioxide (COg)? and unburned hydrocarbons (THC) within
the coiribustor were measured using cooled traversing sampling probes coupled
to on-line analytical instrumentation. Nondispersive infrared analyzers were
used to measure the CO and COg concentrations in the gas sample and a para-
magnetic analyzer was used to measure ©2 concentrations. The MO and Kt>2
concentrations were measured using a chemiluminescence analyzer. An exhaust
probe rake was used to aspirate gas samples from equal annuli for determina-
tion of the average concentrations of pollutant species in the exhaust flow.
The inlet flow into the gas sampling probes was maintained choked, resulting
in aerodynamic cooling of the sample by means of a rapid internal expansion.
The combined effects of expansion and wall-cooling served to quench further
chemical reactions. A liquid-vapor phase-discriminating sampling probe and
a heated flame ionization detector were used to measure the total hydrocarbon
(i.e., liquid plus vapor) and gaseous-hydrocarbon concentrations. The flow
was sampled isokinetically and phase separation was achieved within the probe
by aspirating a portion of the flow into a perpendicularly-oriented vapor-only
sampling tube. Temperature distributions at the exhaust plane and within the
coiribustor were obtained using a traversing calibrated-heat-loss thermocouple
probe. Although conventional thermocouple materials are limited to tempera-
tures below 2000°K, cooling the exposed junction by conduction extends the
range of thermocouple utilization to gas temperatures above the melting point
of the material to the 2000-2500<>K range. In order to obtain the local stream
temperature, the measured stream thermocouple temperature must be corrected
for conduction and radiation heat losses; therefore, calibration data were
acquired simultaneously with the required temperature measurement. Mean and
rms gas and droplet velocities were measured using a dual-beam, frequency-
offset laser velocimeter and single-particle time-domain signal processing.
244
-------
This velocity measurement technique removes directional ambiguity errors which
arise in recirculating flows. The laser velocimeter also provided a qualitative
measure of droplet number densities in the burning spray which could be used
to establish the spray trajectory. Selective seeding of the airstream with
micron-sized phenolic resin particles was used to obtain a sufficient signal-
to-noise ratio in regions of low fuel droplet number density to permit measure-
ments of the local gas velocity. The number density, trajectory and mean
diameter of droplets in nonburning and burning liquid fuel sprays were deter-
mined using an off-axis, transmitted-light type laser holography system. Fringe
patterns were produced on a holographic plate by the interaction of an object
beam, directed through the combustor section, and a reference beam, directed
around the combustor section. Reconstruction of the holograms was accomplished
with a second optical system and the data were reduced manually using a 12-
power loupe. High-speed color motion pictures (500 frames per second) of the
flame in the vicinity of the injector were obtained to assist in the inter-
pretation of the test results. Detailed descriptions of the sampling probes
and associated instrumentation are given in Ref. 9«
245
-------
SECTION 3
EXPERIMENTAL RESULTS
DESCRIPTION OF EXPERIMEOTS
The principal objective of the experimental program was to investigate
the interaction of physical and chemical processes in heterogeneous combustion
on pollutant formation and destruction. This was accomplished by (l) deter-
mining the effects of combustor operating conditions on pollutant emissions,
(2) obtaining detailed maps of the combustor flow field, and information on
fuel spray characteristics and liquid-vapor concentration distributions for a
range of operating conditions and (3) correlating changes in flow field
structure with changes in pollutant formation and energy release. The experi-
mental results will be used to evaluate the combustor flow analysis (CHRISTY
code) being developed in the theoretical portion of this program 'Snd to .
assist in assessing the validity of various models for turbulent transport
and droplet burning.
The combustor was designed so that it would be amenable to analytical
modeling and yet would exhibit many of the essential features of practical
burners. Ultimately, it is intended that the information obtained from the
experimental and theoretical studies will be utilized for evaluating potential
emission control strategies.
The experimental program comprised two different types of tests: (l)
input-output tests to establish the relationship of liquid fuel and air input
conditions to average exhaust species concentrations, and (2) flow-field napping
tests to obtain radial and axial distributions of temperature, species concen-
tration and mean and rms gas and droplet velocities within the combustor and
to evaluate fuel spray characteristics. The matrix of combustor operating
246
-------
conditions for tests conducted using iso-octane, No. 2 fuel oil and propane
is presented in Table 1. These conditions were selected to encompass varia-
tions in operating parameters which, based on the results of Refs. 10 and 12,
are believed to have the greatest influence on pollutant emissions. Tests
were conducted at nominal combustor pressures of 1 atm and 3.3 atm, inlet
air temperatures of 533K, 6UUK, and 755K, and for fuel-air equivalence ratios
in the range 0.9 to 0.5 (10 to 100 percent excess air). The inlet airflow
rate was held constant at a nominal value of 0.137 kg/sec and the swirl
number was varied from low (s = 0.3) to moderate (S = 0.6) by interchanging
swirl vanes. (Swirl number is defined in Table l). In the input-output
tests, measurements were made at the exit of the combustor extender section
(see Fig. 1), while in the mapping experiments detailed measurements were
made within the combustor at a minimum of four axial locations. Variations
in the average exhaust concentrations with overall fuel-air equivalence ratio
and detailed flow field maps describing the effects of inlet air swirl,
combustor pressure and air preheat are summarized below.
INPUT-OUTPUT TESTS RESULTS
Emissions data showing the effects of fuel type, inlet air swirl, com-
bustor pressure, and air preheat at equivalence ratio 0.65 are summarized in
Table 2. A complete tabulation of all (input-output and mapping) species
concentration data is given in Ref. 9. Because neither the Total Efy-drocarbon
Analyzer nor the sample transfer line were heated to prevent condensation of
high molecular weight hydrocarbon species, exhaust THC concentration measure-
ments were obtained only for tests in which propane fuel was used. However,
use of the phase-discriminating sampling probe and the discrete-sampling
heated hydrocarbon analyzer for the flow-field mapping tests permitted deter-
minatione of the total hydrocarbon (i.e., liquid plus vapor) and gaseous
hydrocarbon concentrations within the combustor. These measurements are
discussed in subsequent sections of the report. The high concentrations of
oxygen (compared to equilibrium) measured with the exhaust rake and the low
247
-------
exhaust CC^ concentrations, particularly for propane and iso-octane, are not
indicative of incomplete combustion but rather of a sampling problem. It is
likely that the problem was caused in part by the limited number of ports
sampling a flow stream with sharp concentration gradients and the potential
for blocking of some of these ports by particulates generated during combustion.
Significant particulate loading was noted for propane (because of the narrow
angle fuel injector employed) and for iso-octane. W), K>2 and CO concentra-
tions are presented in Table 2 as measured and also corrected by the ratio
of measured oxygen used to equilibrium oxygen used.
Tests conducted at fixed inlet conditions using each of the three fuels
resulted in similar trends in exhaust emissions with increasing overall
fuel-air equivalence ratio WO, CO and C0p exhaust concentrations increased
while the exhaust concentration of 02 decreased. These general trends are
similar to what would be predicted for gas-phase premixed combustion. Cal-
culated equilibrium exhaust concentrations for premixed, adiabatic combustion
are presented in Table 2 for comparison. As would be expected for finite
residence times, measured CO levels exceed equilibrium levels and measured
J3O levels are far below equilibrium levels.
Jt
Comparison of the emissions data obtained for each of the fuels tested
indicates a dependence of exhaust emission on fuel type. These trends reflect
in part the different carbon/hydrogen ratios of the fuel but also suggest
that differences in the atomization, vaporization and mixing characteristics
of the three fuels can affect pollutant formation and destruction. Detailed
data describing the influence of fuel spray characteristics on the combustor
flow field are discussed in subsequent sections of this report.
The influence of inlet air swirl on exhaust species concentration levels
was evaluated using each of the three fuels at a combustion pressure of 1 atm.
The emissions data, summarized in Table 2, indicate significant changes in
concentration levels as a result of increasing swirl. In addition, variations
in the trends of the exhaust composition data were observed for different
fuels suggesting that there are differences in the structure of the flow
248
-------
fields which affect the pollutant formation and destruction. These differences
were explored in greater detail in the mapping tests discussed in the follow-
ing sections. Increasing the inlet air swirl from S=0.3toS=0.6 resulted
in increased exhaust concentration of UOX in the liquid propane tests but in
decreased NOX concentrations in the iso-octane and No. 2 fuel oil tests.
The trends observed for liquid propane are in agreement with those previously
reported for gaseous propane and natural gas and are characteristic of a
flow field having high fuel concentrations near the centerline (10 and 12).
Visual observations of the combustion process in the vicinity of the injector
revealed that liquid propane was being injected in a concentrated stream
having a very narrow spreading angle. The opposite trends observed in tests
conducted with iso-octane and with Wo. 2 fuel oil probably are due to differ-
ences in the fuel/air distribution in the combustor. The effects of swirl
on the combustor flow field are discussed later on in the report.
Input-output tests were also conducted to determine the effect of
combustor pressure on exhaust species concentrations. Increasing the combustor
pressure while maintaining constant inlet mass flow and temperature results
in longer residence times and generally higher temperatures and reaction rates
which produce increased NO emissions and decreased hydrocarbon emissions.
These trends were found for liquid propane (cf., Tests 11 and 1^) and No. 2
fuel oil (cf., Tests 6 and 9). However, the iso-octane tests show opposite
trends with combustor pressure (cf., Tests 1 and ^). The measured exhaust
MO concentration level decreased significantly and the exhauat 00.
concentration increased for combustion of iso-octane at 3*3 a-tm, suggesting
a change in the fuel spray characteristics or fuel/air distributions at
elevated pressure for this fuel. For the 3.3 atm iso-octane test, the exhaust
gas temperature levels decreased and high concentrations of particulate carbon
were observed, indicating a reduced combustion efficiency. The effects of
pressure on mixing and vaporization are discussed later in the report.
A final series of input-output tests was conducted to determine the
effect of inlet air temperature on pollutant emissions. Increased inlet air
249
-------
temperature should result in higher combustion temperatures and more rapid
oxidation of fuel and CO and more rapid W formation. The combustor tempera-
ture distributions (discussed in the following section) and the exhaust emis-
sions data presented in Table 2 indicate that an increase in inlet air tempera-
ture from 533K to 750K resulted in a significant increase in the exhaust ISO
concentration and reduced UHC and CO concentrations. Other investigators (13)
have reported similar emissions trends with inlet air temperature, and similar
increases in NO emissions levelfc were measured in previous natural gas combus-
tion tests (12).
FLOW FIELD MAPPING RESULTS
Examination of the results of the input-output tests indicates that
pollutant emissions levels are particularly sensitive to inlet air swirl,
combustor pressure and inlet air temperature. Variations in these parameters
produced some experimental trends which cannot be predicted on the basis of
thermochemistry alone, suggesting that there is significant coupling between
the fluid dynamic and chemical processes in the combustor. Such a coupling
was observed in the previous gaseous fuel tests (10). Therefore, detailed
maps of the flow field and holograms of the fuel spray were obtained for the
six test conditions listed in Table 1. These conditions were selected to
encompass variations in combustor operating conditions which have the greatest
influence on pollutant emissions, as determined from the input-output tests.
As in the gaseous fuel tests referenced above, the combustor mapping
data were reduced to isopleth form to permit visualization of the radial
and axial variation of individual flow field parameters and to facilitate
comparisons between these parameters for each of the flow configurations
investigated. However, since the radial distributions of mean flow properties
were determined at a discrete number of axial locations within the combustor,
some interpolation between stations was required. Typically a radial traverse
consisted of 9 to 15 measurements cpaced approximately uniformly across the
combustor diameter. A complete tabulation of the experimental data is presented
in Ref. 9.
250
-------
THE EFFECT OF SWIRL ON FLOW FIELD STRUCTURE AHD NO FORMATION
When a gaseous fuel is injected axially into a swirling air flow the
primary mode of dispersion of the gaseous fuel is turbulent transport and the
primary effects of increasing swirl are to increase local mixing rates and to
increase radial and axial pressure gradients. Previous tests with natural
gas (10) have confirmed that energy release rates increase with increased
swirl. However, liquid fuels of low volatility injected with a radial compo-
nent of velocity may penetrate the airstream primarily as a result of droplet
inertia. In this case, swirl would influence relative velocities between fuel
droplets and air and would affect droplet vaporization and burning rates.
The input-output tests indicate that the effects of swirl are very
different for the three fuels tested. For propane, increasing swirl results
in a decrease in CO levels and an increase in TSO^. Here, propane behaves like
a gaseous fuel and combustion rates are enhanced by increased turbulent trans-
port rates. However it is found that with iso-octane, CO levels increase and
10 levels decrease with increased swirl. This result suggests that insufficient
X.
vaporization occurs close to the injector to achieve the rapid air/vapor fuel
mixing allowed at higher swirl number with an entirely vaporized fuel. In
addition the effect of increased swirl is to shift air flow radially outward
and increase the radial pressure gradient; this would tend to retard the pene-
tration of small, partially vaporized fuel droplets. Since shear levels decay
rapidly with axial distance, vaporized fuel introduced further downstream
into the annular airstream does not burn rapidly.
For No. 2 fuel oil, increasing swirl results in a decrease in both CO
and WOX levels. Temperature and composition profiles presented later on in
this report indicate that radial droplet penetration was the dominant mode of
dispersion of this fuel. Once the fuel is dispersed across the airstream, Hslie
tlroplet combustion rates are probably increased with increasing swirl by
higher relative velocities and turbulence levels. Ifore rapid combustion and
smaller flame standoff distances in burning droplets or droplet arrays would
result in lower levels of both CO and NOX.
251
-------
Flow field mapping tests conducted with iso-octane confirm the results
of the input-output tests and allow some tentative conclusions to be made about
the interactions of swirl with fuels of moderate volatility. The time-mean
temperature distributions obtained for iso-octane/air combustion at one atmos-
phere pressure and for inlet air swirl levels of 0.3 and 0.6, respectively,
are presented in Figure 2. An initial examination of the data reveals the
similarity of the flow field structure obtained for each of these liquid-fuel
test configurations and a general correspondence with the flow fields obtained
previously using gaseous fuel (10), i.e., the characteristic shape usually
associated with axisymmetric, turbulent diffusion flames. The contours are
characterized by peak temperatures occurring off the centerline in an annular
region. Variations in liquid fuel and air inlet conditions altered the rela-
tive rates at which heat was released within the combustor, and specific trends
resulting from these variations are evident with more detailed analysis of
the data.
The temperature contours are not symmetric about the combustor centerline,
but instead are displaced slightly toward negative values of R/RO. Since
the uniformity of the inlet flow was verified by laser velocimeter measurements
and by temperature measurements in the inlet section, the apparatus was eliminated
as the source of this asymmetry. Furthermore, the species concentration dis-
tributions appear symmetric about the combustor centerline. Therefore, the
asymmetry must be the result of blockage introduced into the flow by traversing
the comparatively large thermocouple probe (9) from the positive to negative
radial direction.
Specific trends resulting from systematic variation of the inlet swirl
are evident from the temperature data. For example, increasing the swirl
level from 0.3 to 0.6 results in an initial Increase in the axial rate of
heat release (X/D<2), followed by a gradual decrease in the axial rate of
heat release (X/D>2). This initial increase of heat release rate is attributed
to more rapid mixing of the vapor fraction. However, insufficient initial
penetration of fuel droplets leads to an extended flame as indicated by the
252
-------
-.4
radial temperature gradients Which remain steeper for a greater axial distance.
For the case of 0.3 awirl, the greater part of the available chemical energy
was released within an axial distance of approximately six combustor diameters
(X/D = 6) and the downstream temperature distributions are relatively flat.
Isotherms corresponding to S = 0.6, on the other hand, indicate a larger flame
length, as evidenced by peak temperatures extending the full length of the
combustor (X/D = lU),
Mean axial gas- and droplet-velocity contours obtained for iso-octane/air
combustion at one atmopshere pressure are presented in Figures 3 and k. Differ-
ences between the local gas and droplet velocities are apparent near the in-
jector, as are areas of flow recirculation. Also, some unsteadiness of the
flow was indicated by fluctuations in the droplet velocity measurements.
Farther downstream (X/D>1.0), droplet sizes and concentrations are reduced
by evaporation and corribustion and droplets are convected at the local gas
velocity. At S = 0.3, it was not possible to distinguish between local fuel
droplet and gas velocities in the vicinity of the spray near the centerline
because of the high droplet concentration. Consequently no gas flow recircula-
tion is shown; however, the existence of a recirculation zone may be inferred
from the droplet velocity data, Figure UA. Gas velocity measurements were
possible at S = 0.6 and a torroidal-shaped recirculation zone was identified
(Figure 3). A primary effect of increasing the inlet air swirl from 0.3 to
0.6 was to shift the regions of droplet recirculation closer to the injector,
thereby influencing flame stabilization and energy release in the initial
region of the combustor.
The local time-mean axial velocities are somewhat higher at S = 0.6 and
the diffusion-flame-like flow field structure persists for a greater axial
distance. The insensitivity of the fuel droplet axial velocities to the
level of inlet air swirl is also apparent from the similar appearance of the
droplet velocity distributions in the initial mixing regions; however, at
S = 0.6, high droplet velocities persist farther downstream. Spray trajectories,
as determined from laser velocimetry and laser holography measurements are also
253
-------
shown in Figure k and are in good agreement with the nominal spray angle of
60 degrees. The persistence of droplet velocity and the higher gas velocities
at S = 0.6 are associated with the increase in recirculation zone size and
consequent increase in mass flux density outside the recirculation?'zone.
Gas composition contours for low (S = 0.3), and moderately (S = 0.6)
swirling flows, discussed above, are shown in Figures 5, 6 and 7» These data
indicate the tendency of increased swirl to suppress mixing beyond a zone of
rapid initial mixing near the injector. Radial concentration gradients are
sharper with increased swirl and CO burnout is slower. Further insight into
the effect of swirl on iso-octane vaporization and combustion is obtained from
comparisons of total hydrocarbon concentration distributions (Figure 8) together
with profiles of the percentage of hydrocarbons vaporized (Figure 9)- In the
upstream section of the combustor, total hydrocarbon concentrations are higher
for the lower swirl number (Figure 8) while the concentrations of unvaporized
fuel (computed from the product of total hydrocarbon concentration from Fig. 8
and 1 minus the fraction vaporized from Fig. 9) are about the same. In this
upstream region increased swirl promotes mixing of vaporized fuel. However,
for the higher swirl number, unvaporized fuel persists further downstream
and the total hydrocarbon concentrations are greater in the downstream sections
even after the fuel has vaporized. It is likely that this is a result of
reduced droplet penetration into the airstream with increased swirl and the
#
rapid decay of swirl (shear) induced mixing with axial distance.
ITO concentration distributions are shown in Figure 10. The regions of
high NO concentration within the combustor are coincident with the regions of
locally high temperature. At low swirl, higher concentrations of NO were
measurer! close to the combustor centerline and in the vicinity of the injector.
In contrast, at moderate swirl the reaction zone is moved radially outward
and closer to the injector, and NO formation occurs in a narrow annular region.
The combined effect of low oxygen concentration and low temperature result in
a reduced rate of NO formation and, therefore, lower NO exhaust emissions levels
at S => 0.6.
254
-------
The mapping data indicate that, in the present combustor configuration,
increasing swirl from 0.3 to 0.6 increases mixing of partially vaporized fuel
with air in the initial region of the flow, resulting in increased energy
release rates. Hence, increased swirl tends to move the region of flame
stabilization closer to the fuel injector. Beyond this initial region, in-
creasing swirl appears to suppress vaporization of the liquid fuel and subsequent
mixing of the vaporized fuel with air. At sufficiently high swirl numbers, the
radial pressure gradients reduce penetration of partially vaporized droplets
into the airstream, resulting in a relatively cold fuel-rich region on the
combustor centerline. The reduced vaporization rates result in an extended
mixing region. In addition, axial decay of swirl-induced shear levels tends
to reduce mixing rates downstream from the injector. Reduced mixing rates
result in generally slower fuel oxidation and CO burnout rates and in lower
HO formation rates.
EFFECT OF PRESSURE ON FLOW FIELD STRUCTURE AND NO FORMATION
Previous tests (10) with natural gas have demonstrated several effects of
pressure on exhaust emissions at constant mass flow rate. Increasing pressure
from 1 to 3 atmospheres decreased CO levels and increased exhaust BO levels
principally because combustor residence time increases with pressure. However,
the local rate of energy release decreased indicating that mixing rate was
suppressed at higher pressure. This would be expected since shear levels
decrease with decreasing velocity. This conclusion is supported by the fact
that at even higher pressure, 7 atm, exhaust hydrocarbon levels increased
despite the increased residence time.
For two of the liquid fuels investigated, No. 2 distillate oil and propane,
increasing pressure from 1 to 3«3 atmospheres decreased CO levels and increased
NO levels. However, when iso-octane was used as a fuel, CO levels increased
and NO emission decreased as the pressure was Increased from 1 to 3.3 atmos-
pheres. The flow field mapping tests conducted using iso-octane confirm the
results of the input-output tests and indicate the effect of pressure on the
flow field structure and pollutant formation.
255
-------
The time-mean temperature distributions, Figure 11, indicate that longer
flames are obtained at higher pressure. Furthermore, peak temperatures are
lower at higher pressure suggesting lower energy release rates. Examination
of the species concentration distributions obtained for iso-octane/air com-
bustion at elevated pressure are consistent with the temperature data. Examina-
tion of Figure 12 reveals that at 3«3 atm pressure, the ©2 concentrations near
the combustor centerline are lower, indicating a reduced mixing rate. Similarly,
initial breakdown of the fuel to CO and oxidation of CO to COg is slower at
3.3 atm. Figure 13 shows that NO is formed in an annular region close to the
injector at approximately the same radial location as the peak temperature.
There are steep radial gradients and low NO concentration levels at the aom-
bustor centerline. Peak NO concentrations at the elevated pressure are much
lower than were observed at atmospheric pressure. An increase in pressure from
1.0 atm to 3-3 atm results in a significant decrease in NO emissions which
may be attributed in part to lower temperatures.
One possible explanation for the different effect of pressure on flow
field structure for iso-octane in comparison with natural gas, propane and
No. 2 distillate oil is as follows: The propane rapidly vaporizes on injection
into the combustor. Hence, both natural gas and propane may be considered
gaseous fuels. In spite of reduced mixing rates resulting from the reduced
shear levels associated with the lower air velocities, combustion is enhanced
and NO emissions increase due to increased residence time and increased reaction
rates. In contrast, No. 2 distillate oil burns largely inhomogeneously since
vaporization rates are relatively low due to higher boiling points. Increased
droplet penetration at higher pressure partially offsets the effect of reduced
mixing due to shear and combustion goes to completion because of increased
residence time. Iso-octane is more volatile than No. 2 distillate oil, and
droplet vaporization tends to limit droplet penetration. With relatively
little penetration of iso-octane liquid, combustion efficiency would be governed
largely by droplet vaporization rates. But the droplet vaporization rate is
a function of droplet boundary layer thickness, which in turn is a function of
256
-------
the product of gas density and relative velocity between droplets and air.
For air moving at a velocity higher than the droplet velocity, as pressure
increases the relative velocity might well be reduced far more than the density
is increased. This could reduce vaporization rate with increasing pressure.
Increasing boiling point with higher pressure would also tend to reduce the
heat transfer rate with vaporizing fuel which is proportional to the differ-
ence between the ambient temperature and the boiling point temperature. In-
creased reaction rates due to increased pressure and increased residence time
do not compensate for reduced shear levels, poorer penetration and lower
vaporization rates and a significant amount of fuel vaporized prior to burning.
EFFECT OF FUEL TYPE ON FLOW FIELD STRUCTURE AND NO FORMATION
Liquid propane, iso-octane and Wo. 2 fuel oil differ widely in the
physical properties which influence the atomization and vaporization (velocity,
surface tension, heat capacity, latent heat of vaporization, vapor pressure).
The energy added to the airstream by combustion at a given equivalence ratio, is
of similar magnitude for each of the three fuels; thus little difference in
flow field structure or emission levels can be expected on the basis of equili-
brium thermodynamic considerations. Also, the amount of fuel-bound nitrogen
found in all of the fuels is quite small, and thus this factor is not believed
to contribute significantly to the overall level of nitric oxide production.
Typical properties of the liquid fuels and the results of limited quantitative
fuel analyses are given in Ref. 9«
Significant differences existed between the temperature patterns observed
in the burner when using liquid propane as compared to patterns produced when
using iso-octane or fuel oil see Figure lU. This difference in pattern is
due largely to the difference in the fuel distributions achieved when injecting
propane. These fuel pattern differences are illustrated in Figure 15 which
presents levels of total unburned hydrocarbons within the combustor as deter-
mined by use of the phase-discriminating probe. Most of the propane was found
to be concentrated near the centerline of the combustor; this fuel distribution
is believed to have resulted from flashing of the liquid propane within the
257
-------
injector with the result that a conical spray was not achieved. Because of
the initial fuel distribution, combustion was slow and peak temperatures were
not achieved in the initial regions of the conibustor (Figure 14A). In the
case of the iso-octane and fuel oil, spray patterns were similar and fuel
penetrated to the outer coiribustor radii within two test section diameters f?).
Temperature patterns produced were also qualitatively similar (Figures lUB
and lUc), the most significant difference being the higher temperatures at
the outer radii of the combustor in the case of the fuel oil. Combustion
appears to be more intense in the case of the iso-octane spray resulting in
slightly higher peak temperatures and steeper temperature gradients. The
lower volumetric heat release rates in the case of the fuel oil are probably
associated with the fuel oil droplet characteristics. The fuel oil droplets
were somewhat larger than the iso-octane droplets initially and the fuel oil
vaporizes less rapidly than iso-octane. Thus, although the distribution of
unburned fuel in the initial region of the combustor is qualitatively similar
for the fuel oil and the iso-octane (Figures 15B and 15C), the fuel oil droplets
were larger and required greater time, and hence, distance to burn completely.
The fact that a greater amount of reaction took place in the outer radii in the
case of the fuel oil is confirmed by measurement of the oxygen concentration
which shows that lower oxygen concentrations were found in this region,(Figure
16).
EFFECT OF AIR PREHEAT ON FLOW FIELD STRUCTURE AMD NO FORMATION
An increase in the inlet air temperature will influence flow field
characteristics by affecting flow velocities, chemical reaction rates, and
heat transfer rates. With all other conditions held constant, an increase
in temperature will result in correspondingly higher temperatures throughout
the combustor and will create higher flow velocities. These higher flow
velocities will have the primary effect of decreasing the residence time of
the combustor gases. The diminished time available for completion of the
chemical kinetic processes is in most cases more than offset by the strong
temperature dependence of individual reaction rates. Decreased time available
258
-------
for the droplet vaporization will be compensated for by the increase heat
transfer rate associated with the greater temperature difference between the
gas and the droplets.
Examination of the temperature patterns (Figure 17) obtained for the
case where the entrance temperature was increased by 220K (a *K3 percent increase)
shows that the combined effect is primarily to increase the temperature levels
qualitatively, the temperature pattern did not change significantly. Correspond-
ingly, the inlet temperature change resulted in only snail changes in the un-
burned fuel pattern (Figure l8A and 18B). As would be expected, the fraction
of the unburned fuel existing in the vapor state was greater for the increased
temperature level case (Figure l8c and 18D).
The rate of formation of nitric oxide is very sensitive to local tempera-
ture and accordingly, the increased temperature levels resulted in an approxi-
mate doubling of the local NO concentration ratios (Figure 19). This dramatic
increase occurred over the complete equivalence ratio range tested intthe input-
output experiments. Emissions of CO would be expected to decrease with in-
creased preheat level because of the increased rate of CO oxidation and higher
temperature levels, and this, indeed, was found to be the case.
In conclusion, the effect of the increased preheat level was primarily
to increase the temperature levels throughout the combustor and thereby to
increase the production of nitric oxide; temperature patterns and composition
patterns remain relatively unchanged.
259
-------
SECTION U
REFERENCES
1. Norster, E. R. and A. H. Lefebvre. Effects of Fuel Injection Method
on Gas Turbine Combustor Emissions. Emissions from Continuous Combustion
Systems, Cornelius, W. and W. G. Agnew (eds.), New York, Plenum Press,
pp. 255-278, 1972.
2. Grobman, J. S.. Effect of Operating Variables of Pollutant Emissions
from Aircraft Turbine Engine Combustors. Emissions from Continuous
Combustion Systems, Cornelius, W. and W. G. Agnew (eds.), New York,
Plenum Press, pp. 279-303, 1972.
3. Tattle, J. H., R. A. Altenkirch and A. M. Mellor. Emissions from and
Within an Allison J-33 Combustor. II. The Effect of Inlet Air Tempera-
ture. Comb. Sci. Technol. J_: 125-13U, 1973.
k, Pompei, F. and J. B. Heywood. The Role of Mixing in Burner-Generated
Carbon Monoxide and Nitric Oxide. Comb. Flame 19; U07-U18, 1972.
5. Mellor, A. M.. Simplified Physical Model of Spray Combustion in a Gas
Turbine Engine. Comb. Sci. Technol. 8: 101-109, 1973.
6. Bowman, C. T. and L. S. Cohen. Influence of Aerodynamic Phenomena on
Pollutant Formation in Combustion. Environmental Protection Agency,
Research Triangle Park, N. C., Publication Number 650/2-75-06la, p. 159,
July 1975.
7. Tuttle, J. H., M. B. Colket, R. W. Bilger and A. M. Mellor. Characteristic
Times for Combustion and Pollutant Formation in Spray Combustion. Paper
presented at the l6th Symposium (international) on Combustion. Cambridge,
Mass., August 1976.
260
-------
8. Appleton, J. P. and J. B. Heywood. The Effects of Imperfect Fuel-Air
Mixing in a Burner on NO Formation from Nitrogen in the Air and the Fuel.
Fourteenth Symposium (international) on Combustion. Pittsburgh, PA.
The Combustion Institute, pp. 777-786, 1973.
9. Spadaccini, L. J.,et.al.. Influence of Aerodynamic Phenomena on Pollutant
Formation in Combustion (Phase II - Liquid Fuels). Final Report under
Contract 68-02-1873, Environmental Protection Agency, Research Triangle
Park, NC, in press.
10. Spadaccini, L. J., F. K. Owen and C. T. Bowman. Influence of Aerodynamic
Phenomena on Pollutant Formation in Combustion of Gaseour Fuels. Environ-
mental Protection Agency, Research Triangle Park, NC, Publication Number
6oO/2-76-2»47a, September 1976.
11. Kerr, N. M. and D. Fraser. Swirl. Part I: Effect on Axisymmetrical
Turbulent Jets. J. Inst. Fuel 3J3: 519-538, 1965.
12. Bowman, C. T. and L. S. Cohen. Influence of Aerodynamic Phenomena on
Pollutant Formation in Combustion. Environmental Protection Agency,
Research Triangle Park, HC, Publication Number EPA 650/2-75-06la, July
1975.
13. Tuttle, J. H., R. A. Altenkirch and A. M. Mellor. Emissions from and
Within an Allison J-33 Combustor II: The Effect of:Inlet Air Temperature.
Comb. Sci. Technol. J_: 125-13^, 1973-
361
-------
TABLE 1. NOMINAL TEST CONDITIONS
Fuel Swirl Press. TAIR
No.*
(atm) (K)
Input- Mapping
Output
Iso-Octane
No. 2 Oil
Propane
0.3
0.3
0.3
0.3
0.3
0.6
0.3
0.3
0.3
0.3
0.3
0.6
0.3
0.3
0.3
0.3
0.3
0.6
1.0
1.0
1.0
1.0
3.3
1.0
1.0
1.0
1.0
1.0
3.3
1.0
1.0
1.0
1.0
1.0
3.3
1.0
533
533
61*
750
533
533
533
533
61*
750
533
533
533
533
6M+
750
533
533
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0.137
0,137
0.5-0.9
0.65
0.65
0.65
0..65
0.65
0.5-0.9
0.65
0.65
0.65
0.65
0.65
0.5-0.9
0.65
0.65
0.65
0.65
0.65
x
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
*As defined in Ref. 11:
s. 111.- Z3)
3 (1 -
X
X
X
tan
where Z = hub-to-tip ratio
Ti - angle of vanes
262
-------
XI
1
El*
-H
1
D
ft
3
H
*
|
33
rH
X)
^
x|
*
U
1
i
-p-p
0)0 W
O
,wflj
<*«
o
i i
p<
fc
o
t
* K S
o p^
^^ O^
* fc" a
o" ^
te
o s
S
*6v
>\ i i
0° S
o S
es.
o «>
O«^i
n
S
(^ 5
v^ 1 1
0 A
<§" H"
s
»
> o
H -H
g- tj
t-J bW
w «
M
HS-
W -^~s
03 S
flj Lj
U' ^^
J? *
HM *
H
*<
t_t
1
4-'
02 O
t^
E-l
VD
CO
1
g
OJ
00
-3-
VO
00
LA
-*
OJ
o
CO
aT1
f a
H O
LA
H
00
VD
OO
LA
VO
O
c7
LA
O
H
CO
o
H
VO VO
O rH
H H
1 1
1 1
1 t
OO f-
oo H
H OJ
OO LA
d S
££
OJ CO
ON 3F
H
s a
0 0
IA O
H H
O 0
O O
CO CO
-* J-
ON ON
C^
-=t IA
VO vo
O O
fc ^
VO ,1
O 0
H r-t
CO 00
O O
OJ CO
VD
LA
H
1
l
1
ON
IA
£
$
ON
00
00
o
LA
OJ
o
LA
t-
OJ
S
OJ
LA
VD
O
OJ
CO
LA
00
00
00
O
-sf
p
-d-
OJ
i
« P
P
t~
OJ
LA
OJ
VO
S S
J-
J-
s
0
H
OJ O
S, S o,
0 0 ^
^ OJ
VO ON
ON LA
3 ^
H
LA LA
VD VO
0 0
P 00
00 00
ir\ LA
0 O
H H
H
VD H
0 S
IA
-d- oo
* I *
S ! *-
i i i
ON vo o
CO <*0 t-
H H
& $ S
CO LA ON
oo oo vg
vo p vo
CO OJ VO
1^ I"1
P ON 5*
OJ P O
0 P 0
.
P CO Jt
OJ O P
P P 0
CO
VD ON
CO CO
C- CO t-
t- 00 f-
LA xt
LA LA 3-
VO VO VO
O O O
OJ ^ CO
oo 3- 3-
LA VO IV
O O O
H H H
CO CO OO
O P P
vo tv co
263
VO
IA
1
1
1
1
^
S
3
3
H
«
O
«
O
-*
*
3
oo
VO
LA
IA
VO
O
CO
CO
LA
00
00
OO
O
ON
ON
ON
i °
LA
IV
£
ON
VO
vo oo
Co co
CO
vo
CO
1
p
LA
t + co
& ^ {3T
S °. cT
0 0 °
ON Iv
4
00 ON
tv oJ
b- t-
IA
LA LA
VD VO
O O
O IA
CO IA
IA t-
O 0
H H
r-i
^O -Q
0 s1
s
O t- CO
OJ ON t-
rH rH
H VO H
3 £ fS
* 3 Si
OJ ON 00
VO CO rj
co cy vo
^1 f"i
OJ ON O
VD CO H
VD -* OJ
00 CJ H
0 P P
vO OO CO
CJ H P
p p d
IA
OO VO H
t- fr- t-
CO ON LA
a 3 a
IA LA
IA IA LA
VD VO VO
0 P 0
8 3 R
LA VO C
POO
H H H
CO OO OO
P P O
H OJ oo
H H H
o- oo
LA p
H CJ
19 i
§ a
-* OJ
ON OO
8 g
H H
J* OJ
ON OO
§" 8
p P
LA
8 »
P 0
OO 00
t- VD
H ON
oo g
rH
LA
LA VO
vo 05
0 P
00 ON
oo oo
LA IA
OO O
OO H
OO VD
O P
J; LA
H H
P
SI
fn
*' t
rfi
*j^j
8
p
\R
CO
tt
t-
IA M
^o 'd
0 «
>3
fH
oo -o
oo _,
LA "
d
o
o
p £
d
. 1
_|
3 s
a w
$ -H
w
it)
«v
*<
**<
*
-------
FIG. 1
SCHEMATIC DIAGRAM OF AXISYMMETRIC COMBUSTION FACILITY
WATER-COOLED FUEL INJECTOR
WINDOW PORTS /-EXHAUST
SAMPLING
PROBE
REPLACEABLE
SWIRL VANES
COMBUSTOR SECTION*
ORIFICE
264
76 05 -108-1
-------
TIME-AVERAGED TEMPERATURE DISTRIBUTIONS
ISO-OCTANE/AIfl, 1 ATM, TA)R - 533 °K,
-------
FIG. 3
MEAN AND RMS GAS VELOCITY DISTRIBUTIONS
ISO-OCTANE/AIR, 1 ATM, TA(R = 533°K, 0- 0.65
RECIRCULATION ZONE
M URMS>10M/SEC
(A) SWIRL=0.3
-1.0
O
IX
g
<
u
O
Q
<
oc
.25 -50
AXIAL DISTANCE, X/D
.75
(B)SWIRL=0.6
-1.0
O
I
O
8
_l
5
<
tc
-0.5
0.5
1.0
SPRAY TRAJECTORY
1 ^ , _< 20 M/S u '. 2o M/SEC
,25 .60
AXIAL DISTANCE, X/D
.75
266
-------
FIG. 4
MEAN AND RMSDROPLE T VELOCITY DISTRIBUTIONS
ISO OCIANMAiM . 1 AIM, -iA)H W3"K. ',, fl
tt:-si£ Kt CIRCULATION 2
WJ$M URMS>1°M/S
(A)SWIRL--0.3
-1.0
O
cc
z
O
O
O
<
Q
<
QC
1.0
AXIAL DISTANCE, X/D
O
DC
O
O
O
<
oz
(B)SWIRL=0.6
-1 0
-0.5
AXIAL DISTANCE, X/D
267
-------
(Ai SWIRL-0.3
TIME-AVERAGED Q2 DISTRIBUTIONS
ISO <><:IAI\H,AIH,IATM, i. i,:i.-!"K , ..'. ()«',
FIG. 5
i.o r
0.5
o
a:
in
O
a.
Q
<
DC
O.'j
1.0L
(B)SWIRL=0.6
I.Oi-
0.5
O
tr
rr
z
o
W)
O
a.
Q
<
EC
0.5
t.0«-
L.
10
II)
1
1 ' 1 1 1 L_ ___ . 1
I! H 10
AXIAI DISI ANCt . X I)
14
6
II
AXIAI DISIANCI X/D
268
1?
M
16
-------
O
IT
O
V)
O
0.
-J
<
Q
TIME-AVERAGED CO DISTRIBUTIONS
ISO OCTANE/AIR, 1 ATM, TA|R = 533°K , 0 0.66
FIG. 6
(A)SWIRL=0.3
1.0 r-
0.5
-0.5
-1.0
6 8 10
AXIAL DISTANCE, X/D
12
14
16
(B) SWIRL=0.6
I.Or-
O.S
O
oc
Q
<
C
-0.5
-1.0
6 8 10
AXIAL DISTANCE, X/D
269
12
14
16
-------
FIG. 7
O
OC
2"
O
oo
O
a.
_!
<
Q
<
£T
TIME-AVERAGED CO2 DISTRIBUTIONS
ISO-OCTANE/AIR, 1 ATM, TA|R = 533°K, 0=0.65
(A) SWIRL-0.3
LOT
0.5
-0.5
-1.0
_L
6 8 10
AXIAL DISTANCE, X/0
12
14
16
(B) SWIRL=0.6
1.0r-
O
IE
2
O
I-
8
Q
<
OC
0.5
-0.5
-1.0
-7
5
-3
_L
6 8 10
AXIAL DISTANCE, X/D
270
12
14
16
-------
FIG. 8
TIME-AVERAGED DISTRIBUTIONS OF UNBURNED HYDROCARBONS
ISO-OCTANE/AIR,! ATM,TA)F,»533°K,0 -0.6S
(A)SWIRL=0.3
-1.0
0
2345
AXIAL DISTANCE,X/D
(B)SWIRL=0.6
O
DC
1.0
0.5
<
-------
FIG. 9
TIME-AVERAGED PERCENTAGE OF HYDROCARBONS VAPORIZED
ISO-OCTAN£/AIR,1ATM,TA|R=533°K,<£=0.65
(A) SWIRL=0.3
3456
AXIAL DISTANCE, X/D
_J
8
(B)SWIRL=0.6
23456
AXIAL DISTANCE, X/D
272
-------
(A) SWIRL=0.3
1.0 r
O
tr
V)
O
CL
_j
<
O
0.5
-0.5
1.0
TIME-AVERAGED NO DISTRIBUTIONS
ISO OCTANE/AIR, 1 ATM, TA)R 533"K ,0»0.65
FIG, 10
G 8 10
AXIAL DISTANCE, X/D
12
14 16
(B) SWIRL=0.6
1.0 r~
O
cc
0.5
0-
<
s
-0,5
10
100
50
10
J_
6 8 10
AXIAL DISTANCE. X/D
273
14
f6
-------
FIG, 11
TIME-AVERAGED TEMPERATURE DISTRIBUTIONS
ISO-OCTANE/AIR, TA|R " 533 °K,0- 0.6S
(A) SWIRL * 0.3, 1 ATM
1.0
O
DC
0.5
g
t o
V)
2
< -0.5
Q
-1.0
600-
4 6 8 10 12 14
AXIAL DISTANCE, X/D
16
(8) SWIRL = 0.3, 3.3 ATM
O
DC
1.0
0.5
O
E 0
5-0.5
oc
-1.0
46 8 10 12
AXIAL DISTANCE, X/D
14 16
274
-------
TIME-AVERAGED O2 DISTRIBUTIONS
(A) SWIRL 0.3. 1 ATM
1.0 i
0.5
O
tr
z
o
Q
<
ac
-0.5
-,.oL
10
10
6 8 10
AXIAL DISTANCE, X/D
12
FIG. 12
16
SWIRL=0.3. 3.3 ATM
0.51
-------
{A> SWIRL=0.3,1ATM
1.0 r ' ~
O
IT
O
0.5
§ °
Q
<
OC
-0.5
-1.0 L
TIME-AVERAGED NO DISTRIBUTIONS
'/MR
ISO OCTANE/AIR, T«,H 533°K , fy 0.65
10
1
L
6 8 10
AXIAL DISTANCE, X/D
12
FIG. 13
14
16
O
DC
z"
8
_i
<
o
EE
{B> SWIRL=0.3,3.3 ATM
t.or
0.5
0.5
1.0
__l 1
324
i
f> J
1 j
3 10
12
14
16
AXIAL DISTANCE, X/D
276
-------
TIME-AVERAGED TEMPERATURE DISTRIBUTIONS
FIG. 14
5=0.3,1 ATM.TA|R=533°K, £=0.65
(A) PROPANE
-1.01
(B) ISO-OCTANE
1.0,
o
tr
0.5
8
S -4s
o
-1.0
(C) N0.2 FUEL OIL
1.01
O
I 0.5
I
5 -0.5
I
-1.0
1600
2 4 6 8 10 ""iTl?
AXIAL DISTANCE, X/D
16
4 6 8 10 12 14 16
AXIAL DISTANCE, X/D
4 6 8 10 12 ,4Te
AXIAL DISTANCE X/D
277
-------
FIG. 15
TIME-AVERAGED DISTRIBUTIONS OF TOTAL UNBURNED HYDROCARBONS
5 = 0.3,1 ATM, TA,R=533°K,0<=0.65
(A) PROPANE
1.0r
O
cc
z
O
O
a
_i
<
a
. L. __ 1. . L. L .. 1
-0.5 -
-1.0" L
(B) ISO-OCTANE
1.01
0.5
z
O
55 0
0.5
-1.0
(C> NO.2 FUEL OIL
1.0
AXIAL DISTANCE.X/D
278
-------
(A) ISO-OCTANE
1.0 r
0,5
O
cc
<
O
oc
-0.5
-1.0
TIME-AVERAGED C>2 DISTRIBUTIONS
S= 0.3, 1 ATM,TA|R - 533°K, #
16
10
3
> 5
10
J_
_L
JL
6 8 10
AXIAL DISTANCE, X/D
12
FIG. 16
14
16
(B) NO.2 FUEL OIL
1.0
0.5
O
oc
0
I
oc
-0.5
-1.0 *-
8 10
AXIAL DISTANCE, X/D
279
12
14
16
-------
TIME-AVERAGED TEMPERATURE DISTRIBUTIONS
NO.2 FUEL OIL/AIR, s -0.3, 1 ATM,
-------
FIG. 18
.09
1
cc
I
QC
0
a
I ?
09 o
§ f
s -
W c
§ I
§ -
m W
CO D
Q
ui
§
CC
LU
ui
in
o
in
q
T
°d/U 'NOIllSOdlVIQVU
281 §
-------
(A) TA|R=533K
1.0f-
0.5
O
cc
ec
O
Q
<
CC
-0.5
-1.01
TIME-AVERAGED NO DISTRIBUTIONS
NO.2 FUEL OIL/AIR, S=0.3. 1 ATM, 0 = 0.65
20 ,50
100
J_
I
I
6 8 10
AXIAL DISTANCE, X/D
12
FIG. 10
14
(B) TA|R=750K
0.5
O
C
Q
<
CC
-0.5
-1.0
6 8 10
AXIAL DISTANCE, X/D
282 *
12
14
16
-------
-------
TWO-DIMENSIONAL OR AXIALLY SYMMETRIC MODELING
OF COMBUSTING FLOW
By:
H. McDonald
Scientific Research Associates
Glastonbury, Connecticut 06033
and
R. C. Buggeln
United Technologies Research Center
East Hartford, Connecticut 06108
283
-------
-------
ABSTRACT
A numerical technique has been developed to solve the time-averaged turbulent
reacting flow equations in general two-dimensional orthogonal coordinate systems,
including the effects of flow recirculation. The numerical procedure utilizes
aii efficient robust relaxation scheme which solves a finite difference representa-
tion of the coupled set of elliptic partial differential equations of conservation
of mass, momentum and energy. Physical models representing the effects of
turbulent transport, liquid droplets and radiative heat transfer are incorporated
into the scheme.
To date the main emphasis of this work has been on obtaining numerical
results as free as possible from numerical inaccuracies and truncation error then
comparing the experimental results from a methane burner with theoretical predic-
tions. The predictions were made using three versions of two-equation models of
turbulence, Results of comparison show good qualitative agreement between the
theory and experimental results in regards to the basic flow patterns. For the
swirl number equals 0.3 an experimentally observed toroidally shaped recirculation
zone was predicted. When the flow passed round the recirculation zone the swirl
velocity rapidly accelerated and downstream the swirl velocity profiles filled out
and flattened appreciably as was observed experimentally. In addition the
theoretical axial velocity profiles exhibited the observed local maximum at the
centerline. Quantitatively the predicted axial velocity profiles compare
favorably with the experimental data; however, although in qualitative agreement, .
the comparison of the swirl velocity profiles does not fare quite as well
quantitatively. The principle discrepancies in the flow predictions axe thought
to be mainly due to the difference in the axial location of the theoretically
predicted and the measured location of the toroidal recirculation zone. This
research was conducted in fulfillment on Contract No. 68-02-1873 by United
Technologies Research Center under the sponsorship of the U.S. Environmental
Protection Agency. This paper covers the period of April 1975 up to the present.
285
-------
286
-------
INTHOLUCTIOHf
During the last two decades, numerical techniques for the prediction
of viscous flows have made significant advances in their ability to
represent the flow fields resulting from a wide variety of practical
problems. Initially work concentrated on the simplier parabolic version
of the viscous flow equations, the so-called boundary layer equations.
Considerable success was achieved in making predictions in the boundary
layer case where the diffusional effects in the primary flow direction can
be neglected and the pressure field determined independently. Because of the
parabolic nature of these equations, that is the problem possesses a time-
like quality of dependence on initial and boundary conditions, the boundary
layer equations lend themself to relatively simple spatial inarching integra-
tion techniques. However, there is still a considerable practical interest
in being able to predict the location of and the detailed features of
recirculatlng flows, which evidently do not possess this time-like quality
of the parabolic problem, since in a recirculatlon zone flow is being
convected back against the time-like marching direction. Thus it becomes
necessary to consider a more complete set of viscous flow equations which
at the very least would not neglect the diffusion in the primary flow
direction. For these cases the differential equations are elliptic in nature
which is to say that at any point the flow can, In principle, be influenced
by all the boundaries and, of course, the spatial marching techniques are no
longer applicable except perhaps in some Iterative sense. In order to
make predictions for these elliptic cases, two basic techniques have become
popular for steady state calculations, the time inarching and residual
relaxation techniques. Both techniques consist of replacing the partial
differential equations which describe the elliptic viscous flow phenomena
by some form of algebraic analogy. Many investigators have chosen to use a
287
-------
finite difference representation to approximate derivatives; however, other
algebraic approximations can and have "been used. The time or pseudo time
marching technique consists of making an initial guess of the solution at
time zero and then integrating a time or pseudo time dependent form of the
governing equations out In 'time* until the desired steady state solution
is achieved. The residual relaxation technique also relies on making an
initial ^UGES of the solution. From this initial guess a measure of the
deviation of this guess from the solution is made by inserting the approxi-
mate solution in the governing equation arid computing the error or residual
as it is often called. The error or residual is then used in some manner
to make a correction to the guess in an iterative sense until a converged
solution is obtained. The time marching technique is more general than the
residual technique as it also has the capability of yielding the transient
solution as well as the steady state solution, should a steady state indeed
exist. In many cases, however, only the steady state solution is of interest,
and for these cases the residual relaxation technique can present a
competitive and in some senses a superior method for calculating elliptic
viscous flow situations since the transient solution does not necessarily
have a rapid approach to the steady state. Direct i.e., noniterative
techniques have In the past not been competitive with iterative techniques
for the desired dense meshes. In recent years there has however been a
revived interest in direct techniques but none have reached the point of
being able to be applied in the problem of current interest. The numerical
basis of the residual relaxation technique used in this investigation, the
Field Relaxation Elliptic Procedure (FREP) technique, will not be discussed
in this paper as it has been previously described in several other documents
(Refs. 1, 2, and 3)- Instead this paper will concentrate on four general
topics. The first of these general topics will be to discuss the overall
objectives and general philosophy followed in applying the FRKP numerical
technique to combustor problems. Secondly a brief discussion will be made
of tho history of the development of the numerical scheme and problems
encountered during its development. The third topic will discuss the
rocmit and current activities performed under EPA contract 68-02-1873.
Finally, some general observations will be given and some recommendations
made for what future work should be performed in this particular area of
modeling.
288
-------
Taking the first topic, the overall objective of this combustor modeling
effort has always been to develop a numerical procedure that is capable of
being able to economically represent the combusting flow and related
phenomena associated with a wide range of combustor problems. Consideration
has been given to developing a computer code which has the ability of
analyzing a large number of geometrical configurations and different fuels.
The general approach has been to use "state of the art" physical models of
transport and chemistry and insofar as possible the general structure of the
code has been developed in such a manner that if the "state of the art"
proved inadequate, new technology could relatively easily be incorporated
Into the code as it became available. Numerical techniques were used which
were felt to have the highest probability of success as a general prediction
method for the solution of the governing partial differential equations.
Consistent with the generality desired, the full Navier-Stokes equations were
treated and as few additional approximations as possible were made. These
additional approximations occur mainly in the areas of the physical models
and in the application of "wall function" boundary conditions, which are
approximations designed to make major savings in computer storage and run
time with the minimum impact on the overall plausibility.
Developing a combustor analysis comparison of experimental data with
computer predictions of combustor flows is a valuable tool for determining
the predictive accuracy of the combination of the numerical technique used
to solve the differential equations and the physical models used in the computer
code. Results of the comparisons are valuable not only in guiding the
investigator in his choice of alternate physical models or numerical schemes,
but also enables the numerisist to aid the experimenter in his choice of
future experiments. For example, in the present study the experimental
inlet conditions were essentially unknown. A simple theoretical study using
a range of reasonable inlet conditions indicated that the solution downstream
of the inlets could be critically affected by these inlet conditions, thus
strengthening the motivation of the experimenter to perform the often
difficult measurements to map the inlets conditions. Another example from
the present study is the observation that the axial velocity profiles are in
reasonably good agreement with the experimental data, but the temperature
profiles show poor agreement. In view of the close coupling between
289
-------
temperature density and velocity it does not seem likely that the velocities
would be predicted correctly unless the temperatures were 'reasonable',
and this thought does cast doubt on the temperature measurements since the
velocities were measured by a nonintrusive laser velocimeter. To date the
indications are that only a limited amount of the theoretical predictions
compares reasonably well on a quantitative basis with the experimental
combustor data taken under this particular contract. The relatively poor
quantitative results (qualitatively the predictions shown remarkable
similarities to the observed flow features) may be due to errors arising from
the numerical techniques or the physical models used or in part is due to
the lack of well defined experimental inlet conditions, Identification of
the role of these potential problem areas is in itself a significant
conclusion.
290
-------
BACKGROUND
Insofar as the numerical aspects are concerned the general evolution of
the KREP computer code has been that of taking a rather general numerical
procedure and gradually adding to it numerous physical models, boundary
conditions, additional partial differential equations to consider new
effects, and geometric capabilities to consider more complex configurations.
At various points in this development process it has also been necessary to
modify the numerical procedure in order to both economize on computer time
and storage and to generalize the technique to adequately consider the
effects of the new physical models and additional differential equations.
The first version of the PREP computer code solved the stream function,
vorticity, stagnation enthalpy and mass fraction equations. The turbulence
model assumed that some global mixing length model was available and the
chemistry was an equilibrium model. The above equations were solved in
an uncoupled manner} however, it soon became evident that because of the
strong coupling of the vorticity to the stream function in many cases of
interest it was necessary to solve the stream function and vorticity in a
coupled manner. This was done by generalizing the numerical procedure to
allow for systems of equations which are solved simultaneously as opposed to
the sequential but individual solution of each equation in the system. The
formal introduction of the coupling into the system allowed computations
at a much higher Reynolds number than had previously been possible as well
as of course an increased rate of convergence relative to the uncoupled
system. Next came the addition of the swirl equation and elementary
radiation equations so that the user could consider the effects of these
phenomena. In order to calculate the nitric oxide emission a non-equilibrium
(rate) NO equation was added. The assumption was made that the formation
of the NO had little effect on the fluid dynamics (little energy release)
291
-------
and hence the NO levels could be calculated after a converged fluid dynamics
solution had been obtained. The rate NO equation presented the first
exposure to an equation which had a strong nonlinear source term. What could
have been a potentially difficult problem turned out to be easily handled by
a Newton-flaphson iteration scheme embodied in the FREL' procedure.
Work on the second itfA contract (EPA 68-02-1092) falls into four
basic areasJ (l) basic numerical work, (2) addition of new physical models,
(3) added geometrical capabilities and (4) comparison of computer predictions
and experimental data for the EPA-UTRC burner. The original version of the
FRE1' computer code used a straightforward upwind differencing on the
convection terms. It was found that in some cases as a result of numerical
truncation error, mass was not conserved (Ref. 4) to a sufficiently close
tolerance. In order to correct this deficiency, central differencing and
donor cell differencing of the convective terms, two additional techniques,
were incorporated into the computer code. The central differencing and
donor cell differencing of the convective terms both conserve mass (and
any other term), Central differencing which is second order accurate in
space does have the so-called cell fieynolds number problem of wiggles
developing in the solution when the cell Reynolds number exceeds two},
however, this problem can be minimized by the proper choice of grid spacing
and the application of artificial viscosity. Donor cell differencing does
not have any stability problems, but it is formerly only first order
accurate. At present the technique Is to use central differencing whenever
possible and to use donor cell differencing for all other cases.
Initially the FHEP computer code used a mixing length model to calculate
the eddy viscosity and an equilibrium model for the chemistry. It was felt
that these models were inadequate for the preliminary evaluation of the code
predictions. When this initial evaluation was completed the mixing length
model was replaced by a two-equation model of turbulence, the turbulence
kinetic energy-dissipation of turbulence kinetic energy model (Ref. 5).
The two-equation model of turbulence requires the solution of two addition
partial differential equation with nonlinear source terms, but its strength
lies in the generality of the flow situation that it can consider. The
chemistry model was also Improved to account for some of the rate phenomena.
Dryer's (ttef. 6) global kinetic methane model was incorporated into the code
292
-------
and it should, be mentioned that this implementation required some further
approximation. The rate chemistry also required the additional solution of
a partial differential equation. The source term of this equation represents
the rate of burning of the methane and by assuming that the products are
in equilibrium a new chemistry model is evolved. This new model is
especially useful for the premixed flows where the equilibrium models
predict much too rapid "burning which as well as being unrealistic could
often lead to numerical instability.
During the initial runs for the comparison of the theoretical results
with the experimental (see Figure 1 for a schematic of the experiment), it
became evident that effects in the combustor were being felt up in the
fuel injector, but of course these effects were prevented from doing so
in the calculation by the imposed boundary conditions. In order to account
for the phenomenon the computational domain was increased to include the
region inside the fuel injector. When the computation was performed with the
new domain the results showed that the influence of the main combustion
region were indeed felt upstream in the fuel inlet.
Subsequently in performing comparison of computed predictions with the
experimental data, it was assumed that the radial velocity was zero at the
air inlet plane. It soon became evident that immediately downstream of the
air inlet the predicted and measured radial velocity profiles were not in
reasonable accord, hence several air inlet profiles of radial velocity
were assumed and the resulting downstream characteristics calculated. It
was found that the assumed radial velocity profile had a very significant
effect on the resulting downstream flow field. The location of the
recirculating and in fact the existence of a reclrculating zone could be
determined by changing the assumed radial velocity profile. Since the
air inlet radial velocity profile at the dump plane is rarely measured or
otherwise known it then becomes necessary to extend the computational zone
up into the air injector where the assumption of zero radial velocity
profile is more likely to be a valid approximation. Incorporation of this
capability and the numerical problems associated with it will be discussed
later.
One of the motivations for going to the more complex Dryer's global
kinetic methane scheme was the lack of agreement between the computational
293
-------
and measured temperature profiles. The computational temperature profiles
in general had much larger gradients and peak temperatures than the experi-
mental data. It was thought that this was due to the equilibrium assumption
and that Dryer's kinetic model as implemented here would tend to diffuse
the computational temperature profiles and that the agreement with the data
would improve. Unfortunately the present application of Dryer's model was
unable to improve the poor predictions. The computational profiles did
become slightly more flat, but the agreement with the data was in general
not significantly Improved.
The computed axial velocity profiles were in general in relatively
good agreement with the laser velocimeter measurements of velocity. In
the region immediately downstream of the dump plane on the axis of symmetry
the computed axial velocity did not exhibit a local maximum, as the experi-
mental data did; however, the profiles further away from the centerline
were in good agreement. Farther downstream agreement between theory and
experiment was good across the entire combustor. Predictions of swirl
velocity were consistently lower than the measured values. Among the
possible explanations for this deficiency is that the values of turbulent
viscosity used in the swirl equation could bs too high.
294
-------
THE PRESENT EFFORT
Work performed under the present EPA contract has concentrated in
four main areas (l) incorporation and evaluation of alternate two-equation
turbulence models into the FREP computer code, (2) improve the geometrical
capability, (3) improve the generality of the wall function portion of the code
and (4) perform further comparisons between theory and experiment for the EPA-
UTRC burner. Two alternate two-equation turbulence models were incorporated
into the FREP computer code. Both models utilized the turbulence kinetic
energy as the first variable. The first model had as the second variable the
product of the turbulence kinetic energy and a 'mixing length1 (Ref . ?) while
the second model had as its second variable a quantity related to the fluctua-
tion of vorticity (Ref. 8). Computations performed with these two alternate
two-equation models produced results that differed only slightly from the re-
sults produced by the turbulence kinetic energy-dissipation of turbulence
kinetic energy equations. It is felt that this is due mainly to the fact that
the theoretical basis of the source terms of the second equation of each model
are related and hence if the equations are source dominated, similar results
should occur. This phenomenon has also been observed by other investigators
(Ref. 9)i and in part stems from the originator of the particular model
ensuring that certain limiting forms of the turbulence model are recovered.
As mentioned in the previous section, there are cases when it is
necessary to have as the computational domain regions inside both fuel and
air inlets. During the present contract the computer code was modified to
allow the computational domain to include the regions inside an arbitrary
number of inlets. This required a modification of the basic alternating
direction implicit (AUI) scheme to allow for the integration through solid
surfaces. This was successfully done by modifying the elements within the
ADI matrices in a manner such that the wall boundary conditions were satisfied
295
-------
at each surface. Results performed with this new version of the PREP computer
code substantiated the belief that the radial velocity profile on the dump
plane of the air injector of the EPA-UTRC burner could not be assumed to be
of zero magnitude. In addition a considerable effort was made to generalize
the wall function concepts Introduced in the previous contract. Wall
functions can be viewed as an economical means of approximating the turbulent
flow phenomena in the regions of walls. In the regions of walls the
derivatives normal to the wall of many variables become very large and
hence a large number of grid points is needed in the wall regions to
adequately resolve these derivatives. Wall function concepts are based on
simplified closed form approximations to the boundary layer profiles of the
flow variables in these regions to calculate their derivatives, Under
the previous contract it was assumed that the first grid point off the wall
was in the well known boundary layer logarithmic velocity profile region.
This assumption proved to be invalid in some regions of the flow field
where the first grid point actually lay within the viscous sublayer and
hence necessitated the need for a more general wall function formulation. The
assumption of being in the logarithmic region was replaced by a formulation
that allowed for a continuous movement away from the near wall viscous
sublayer region through the transition region into the logarithm region. The
generalized wall function scheme affects not only the velocity profile near
walls, but also the stagnation enthalpy profile for cooled walls, the
derivatives of viscosity near walls and the boundary conditions for both
equations of the two-equation models of turbulence. Once these improvements
were incorporated into the FREP computer code, attempts were made to compare
the predictions with the experimental data for the EPA-UTRC burner. However,
problems emerged in two areas (l) with the two-equation model of turbulence in
the form of the physically Impossible negative turbulence kinetic energy and
(2) with a lack of overall outer convergence of the system of equations. The
predicted, negative turbulence kinetic energy occurred in regions of the
flow field where the system desired to have a very low level of (positive)
turbulence kinetic energy and during the course of the iterative development
minor excursions into the domain of negative energy occurred. Once these
physically implausible negative values occurred within the solution domain
the solutions diverged immediately. Upon close examination of the two
turbulence model equations, it was seen that the linearized version of
296
-------
the particular form of the equations In widespread use were In fact singular
as the turbulence kinetic energy approached zero. Hence "based upon an
equivalent but reformulated, version of the model equations, a new linearized
form which circumvented the singularity problem was developed and incorporated
into the FREP computer code. This revision of the form of the turbulence
model equations not only helped with the predictions from the turbulence
equations, but also speeded the rate of overall outer convergence. However,
even at this point the outer convergence of the system of equations could
not be obtained to the expected degree, although the general qualitative
features of the predicted flow were consistent with the experimentally
observed results. For both the swirl number equals 0.3 and 0.6 cases the
approximate location of the measured and calculated recirculation zones and.
the qualitative form of the measured and calculated axial velocity profiles
were similar. The predicted axial velocity profiles also exhibited a
local maximum on the centerline as was experimentally observed.
In order to improve the degree of outer convergence the form of the
swirl equation was reformulated such that the dependent variable was
changed from the swirl velocity times the radius to the swirl velocity
alone. Some improvement in the swirl velocity predictions were observed
but the problem of relatively poor outer convergence still remained.
Another problem, previously mentioned and felt to be a possible source for
the lack of outer convergence was the very large gradients of temperature,
not observed experimentally, that appeared in the predictions and the fact
that small physical movements of the flame front could cause locally large
residuals. Suspecting that an inadequately assumed value for turbulent
effective Prandtl number could be responsible for allowing the large
gradients to exist, the turbulent transport of enthalpy was increased by
varying the assumed turbulent Prandtl number, but the results indicated that
a much larger effect was required to reduce the predicted temperature
gradients. With this in mind a simplified radiation effect was included In
the stagnation enthalpy equation and with very reasonable emlasivity
assumptions the large temperature gradients were substantially reduced.
Unfortunately this also did not cause the hoped for improvement in the
degree of outer convergence.
297
-------
At this point in the development of the FREP computer code, It Is
desirable to evaluate the codes performance in the two cases run under the
present contract. It appears that the main problem area at the present time
is the inability of the code to obtain a satisfactory level of overall outer
convergence for these two cases. Further preliminary investigation has
indicated that at least some of the difficulty is associated with the
slow rate of outer convergence of the two equations associated with the two
equation model of turbulence. The "stiffness" of these equations can
perhaps be test overcome by still further linearization of these equations
or by Iterating on the solution of these equations during each outer
1 te ration.
It should be pointed out that tho combustor cases generated by the
EPA-UTRC combustor are not typical of normal furnaces. The ratio of the
velocities in the air and fuel injectors is on the order of 20il thus
causing the existence of extremely large velocity gradients in the vicinity
of the splitter plate. Further adding to the numerical difficulties Is
the existence of large temperature gradients which are of course typical
of diffusion burners. Large gradients with a limited number of grid points
to resolve them are always a potential source of numerical difficulties.
In addition, further numerical problems can possibly be caused by the thin
splitter plate between the fuel and air inlets. Since it was desired
to include in the computational domain the regions inside both inlets,
it was necessary that the splitter plate be taken to be at least one grid
spacing thick. The use of such a small grid spacing in the vicinity of the
splitter plate and hence a much larger grid spacing elsewhere is another
potential problem area that could only be resolved by the addition of more
grid points. Simplier cases (i.e., cases which have smaller gradients of
velocity and temperature) involving a premlxed fuel and air have been
successfully run with the FREP computer code by the present authors (Ref. 10)
lending credence to the validity of the overall numerical scheme.
Other Investigators have seemingly obtained converged solutions to the
various flow equations for combusting flow and in some cases have achieved
reasonable agreement between theory and experimental data. In most cases
the problems considered have been much less severe and have predominately
been premlxed or for diffusion burners have been a low ratio of air to fuel
298
-------
axial ratio cases. In view of the fact that in the present study it was
found that the assumed inlet conditions had a great effect on the downstream
flow and that the other investigators did riot have inlet information avail-
able, it is doubtful if the good agreement between theory and data was
obtained without some "experimenting" with the inlet conditions. In some
cases it is doubtful if convergence is actually achieved as many of the
dependent as well as the derived variables are heavily under-relaxed.
This c:aa result in the slowing down of numerical evaluation of the solution
and falsely ,<;ivc the impression that thfc solution has converged. In addition
several cases have been reported where extremely coarse grid spacing has
Leon used. This results in the introduction of large amounts of truncation
error which in turn causes the large gradients to be smeared out thus
resulting in a much less numerical severe case where a converged solution
can be obtained. If a much finer mesh had been used, the solution may not
have converged. Certainly much work needs to be done on mesh refinement
and its effect on both the achievement of a converged solution and on the
quality of the converged solution, if it exists.
In spite of what is considered relatively poor outer convergence the
following comparisons are presented to indicate the qualitative features
of the predictions. It is not expected that the overall flow features will
change by an appreciable amount when satisfactory outer convergence is
obtained. The overall qualitative features of the predicted flow patterns are
in fairly good agreement with the experimentally observed results. The
existence of the toroldally shaped central recirculation zone is in
approximately the same region ae was measured. Figures 2, 3> an
-------
the isotherms predicted by the FREP computer code. These Isotherms when
compared with the measured values of temperature have a poor agreement. The
predicted temperature profiles in general have much larger radial gradients
and exhibit much larger peak temperatures. This lack of agreement, of course,
produces a corresponding lack of agreement in NO^ profiles as the predicted NO
general follows the temperature.
300
-------
SUGGESTION FOR FUTURE DEVEJjQPMENT
It is suggested that for the near future that work concentrate both
on examining the outer convergence problems and on attempting to compare
theoretical predictions with experimental data for cases not as severe as
those considered under the present contract, since indeed the cases
considered to date are not at all representative of industrial furnaces.
Next some basic numerical work should be performed on some simple cases
such as the flow in a laminar driven cavity and simple turbulent prefixed
flows to establish the effect of the grid spacing and mesh refinement on
the quality of the solution. In addition further effort should be undertaken
to improve the numerical scheme in regards to computational speed and
numerical accuracy and robustness, and also to develop a means of eliminating
the need to use artificial viscosity when the cell Reynolds number becomes
too large. Furthermore, it would be advisable to consider some cases of a
diffusion flame where the splitter plate is not of essentially infinitesmal
thickness. The consideration of cases which have thicker splitter plates
will enable the use of a more approximately equally spaced radial grid spacing
T* >
and thus eliminate the problems associated with rapid variation of the grid
spacing. Since it is felt that the assumed inlet conditions can greatly
effect the resultant downstream flow field, it would be desirable to
consider experimental cases for which valid inlet data exists or cases where
it is reasonable to assume conditions up in the injectors. As a final
suggestion, some numerical experimentation could be performed in the area of
temperature predictions in combusting flow. Experimental measurement of
temperature is difficult in combusting turbulent flow due to the high
fluctuation level in both temperature and velocity and because of the
angular sensitivity of the measuring apparatus. Some simple hot versus cold
flow field comparative studies could be made to determine the effect that
301
-------
combustion has on the development of the flow field. Another potentially
valuable study could be performed by setting the temperature field to its
experimentally measured values and then letting the velocity field and other
variables develop as the numerical solution converged. If the resulting
velocity field converged to the experimentally measured one, this would
lend credence to the measured temperature field and thus demonstrate the
need for further work in the means by which the theoretically calculated
temperature fields are predicted.
302
-------
REFERENCES
1, Anasoulis, R. F, and H. McDonald! A Litudy of Corabustor Flow Computations
'rfith Experiment. EPA-65012-73-045, Dec. 1973.
2. Anasoulis, R. F. , H. McDonald and R. C. Buggelni Development of a
Combustor Flow Analysis, Part Ii Theoretical Studies. AFAPL-TB- 73-98 ,
Jan.
3. Buggeln, R. C. and H. McDonald J Influence of Aerodynamic Phenomena on
Pollutant Formation in Combustion. Environmental Protection Agency,
Research Triangle Park, to be published.
4. Roache, P. J. * Computational Fluid Dynamics. Hermosa Publishers,
Albuquerque, New Mexico, 1972.
5. Jones, W. P. and B. E. Launder* The Prediction of Laminarization with
a Two-Equation Model of Turbulence. Int. J. Heat Mass Transfer, Vol. 15,
1972, PP. 301-314.
6. Dryer, F. L. ; High Temperature Oxidation of Carbon Monoxide and Kethane
in a Turbulent Flow Reactor. AFCSR Report No. AFOSR-TB-?2-1109,
March 1972.
7. flotta, J.s Recent Attempts to Develop a Generally Applicable Calculation
Method for Turbulent Shear Flow Layers. Proceedings of AGARD Conference
on Turbulent Shear Flows, 1971.
n. Spalding, D. B. » The Prediction of Two-Dimensional Steady Turbulent
Flows. Imperial College, Heat Transfer Section flep. EF/TN/A/16, 19*9.
9. Laundex-, B. K. and D. B. Spaldingi Mathematical Models of Turbulence.
Academic Prese, New Jfork, 1972.
10. Gibeling, H. J., H. McDonald and W. R. Brileyi Development of a Three-
Dimensional Combustor Flow Analysis. AFAPL-TR-75-59, Vol. II, Oct. 19?6.
303
-------
304
-------
o
-------
o
oo
o
VO
o
Jt
o
CVI
o
o
a
o
CO
_ o
o
OJ
o
O
i!
J-4
o
Q>
W
CO
OJ
*
O
II
-p
03
-p
CS
4?
H
a
5
(U
o
306
-------
8
8
w
09
I
4-1
ON
0)
«5
O
H
00
O
O
O
307
-------
o
a
C5
II
M
4)
*<
I
e
CM
O
11
VO «
to
i
O
-------
8
o
0)
CO
CTs
eg
I
O
00
O
(M
309
-------
ON
CO
OJ
CM
CM
CM
03 H
310
-------
PREMIXED ONE-DIMENSIONAL FLAME (PROF)
CODE DEVELOPMENT AND APPLICATION
By:
J. T. Kelly, R. M. Kendall
Acurex Corporation/Aerotherm Division
Mountain View, California 94042
311
-------
-------
ABSTRACT
The PRemixed One-Dimensional Flame (PROF) code numerically models complex
chemistry and diffusion processes in premixed flames. Since the code includes
diffusion, realistic solutions of coupled combustion and pollutant formation
processes can be obtained in the flame zone, as well as downstream in the post-
flame region. Previous experience (Reference 1) has shown that the code can be
a valuable aid when analyzing experimentally-found coupled combustion and pol-
lutant formation phenomena such as "PROMPT NO."
The code's accuracy and reliability has now been increased. Also, several
models for heat loss to bounding walls have been added to the code to model con-
fined flames, such as those occurring in surface and catalytic combustors. Also,
plug and well-stirred reactor options have been added to model combustion and
pollutant formation processes in a wide variety of conventional and experimen-
tal combustion systems.
Sample calculations are presented to demonstrate the capabilities of the
revised PROF code. Also, to illustrate the value of the code in analyzing com-
plex combustion and pollutant formation processes, predictions of NO formation
in premixed CH4/air flames over a wide range of equivalence ratios are presented.
These calculations show that the well-known extended Zeldovich mechanism is ade-
quate for stoichiometric and lean conditions, but inadequate under fuel-rich
conditions. A possible fuel-rich NO mechanism is then postulated to account for
NO formation at equivalence ratios greater than one.
313
-------
-------
INTRODUCTION
Detailed experimental sampling of premixed flat flames have shown
that some pollutant formation processes, such as NO formation i.n fuel-rich
hydrocarbon flames, are strongly coupled to combustion processes within the
flame zone. The chemical interpretation of data from these flames is usually
not clear, since the data is affected by the experimental probe, chemical
events are obscured by diffusive processes, and insufficient species concen-
tration data are taken. To help interpret this data, elementary chemical
kinetic and diffusion data can be used in comprehensive theoretical flame
models to predict species concentrations through flames. These predictions,
combined with the experimental data, can provide valuable insights into the
complex chemical events taking place within the flame.
The PRemixed One-Dimensional Flame (PROF) code treats combustion and
pollutant formation events occuring in premixed flames. Comparing H2 and CO
flame predictions and data (Reference 1) showed that the PROF code is a
valuable aid when interpreting and extending flame data. Also, these predic-
tions showed that the PROF code is reliable, quick, and accurate. The need
to apply PROF to more complex flames such as CH4/air with pollutant formation
and combustion systems such as catalytic combustors prompted the present study
to upgrade and expand the PROF code.
One of the primary objectives in developing the original PROF code was
to demonstrate how efficient matrix procedures can be applied to solve flame
equations. For this demonstration, the solution procedure was simplified to
consider only first order diffusive fluxes, spatially-uniform heat losses, and
a mean Lewis number of unity for gas mixtures. Although these simplifying
assumptions are acceptable for many flames, in some cases the assumptions give
errors. To remedy this situation the PROF code has been upgraded to include:
315
-------
Higher order species diffusive fluxes
Flameholder heat losses
Radiation heat losses
Nonunity Lewis number effects
Besides the additions, the code has been expanded to model flames with
heat losses to reactive or nonreactive bounding walls. Important confined
flame problems, such as those in catalytic and surface combustors and tubular
reactors can now be treated. In addition, well-stirred and plug flow reactor
options have been included in the code so that furnace, gas turbine and other
combustion systems can be modeled. With these additions a wide variety of
experimental and practical combustion and pollutant formation problems can now
be treated with the PROF code.
316
-------
FORMULATION
The PROF code models the confined flame problem shown schematically below.
The code treats both axial and radial transport of heat and mass. The boundary
conditions assigned are the state of the gas at the tube wall, the initial bulk
species fluxes, and temperature. Given these conditions, the code predicts bulk
gas properties along the tube axis.
Post-flame
zone
Luminous
zone
Preflame
zone
^M
I
t tf t
i
Well-stirred and plug flow reactor problems can also be modeled by the
PROF code. These options use the generalized chemistry solution procedure
applied in the flame option, but do not consider diffusive species or heat
transport between reactor stations.
As described in the following sections, the PROF code solution pro-
cedure has been optimized to predict premixed one-dimensional flames. The
assumptions and equations used in the development of the PROF code are dis-
cussed below. The well stirred and plug flow reactor equations are subsets
317
-------
of the flame equations and will not be discussed herein. The predictor-
linearized corrector solution procedure used to solve the flame equations
remains essentially unchanged from the original formulation detailed in
Reference 1.
FLAME CONSERVATION EQUATIONS
The equations for the flame option are developed by integrating the steady
two-dimensional species, mass, and energy equations across a plane perpendicular
to the axis. This produces a set of quasi-one-dimensional flame equations in
terms of bulk gas properties. These properties vary along the flame axis as a
result of fluxes of heat and mass at the radial edge of the flame and chemical
reaction and axial diffusive fluxes within the bulk gases. For a large free
flame with an initially uniform flow, the edge fluxes are small and the bulk
properties are equivalent to local conditions across the flame. For a flame
confined in a tube such as in a catalytic combustor, the edge fluxes can be
significant. The code treats these fluxes using a transfer coefficient approach
in which the fluxes are assumed to be directly proportional to bulk and wall gas
states. Using transfer coefficients reduces an essentially two-dimensional
problem to one-dimensional form.
The quasi-one-dimensional flame conservation equations can be written as:
Specves
dY d(AJ.)
'_ - fl U J
ds~ " A Wi ds
(1)
Energy
ft.*)-"
(2)
where global continuity has been incorporated into the above equations and the
momentum equation has been replaced by assigning a fixed pressure. As shown in
Equation (1) the species mass fraction along the flame axis, Y-j, is altered by
species bulk gas phase chemical production, W-j; axial diffusive flux, J^; and
wall diffusive flux, Jw-j. In the energy equation, the enthalpy, h, along the
flame axis is altered by the bulk gas heat loss rate, Q; axial diffusive heat
flux, XJ.h. + k dT/ds; and wall heat loss, q . Expressions for W-, J^, Jwi»
ill W 1 1
q, and q , and boundary conditions complete the definition of the flame problem.
W
Species AxialDiffusional Flux
In the PROF code, the simple yet adequate bifuracation approximation
for the binary diffusion coefficients is used with the Stefan-Maxwell relations
to develop an explicit expression for species flux, J.. This approximation,
318
-------
developed in Reference 2, assumes that the contributions of species i and j to
the diffusion coefficients, P., can be separated in the following manner:
ij
(3)
where D is a reference self-diffusion coefficient and F-J and Fj are diffusion
factors. The explicit formulation for species flux, J^, developed using this
approximation is:
J - * /"'I + "i
Ji - - iTTTldr + T
(4)
where
and M2 "MZVF1
In Equation (4), D is usually taken as the self-diffusion coefficient of a
reference species, e.g., 02. The Fj then are found to be constants independent
of temperature and pressure. Computations using this formulation are more ef-
ficient than those that use arbitrary expressions for the P,-j and either apply
the Stefan-Maxwell implicit equations or develop the concentration-dependent
multicomponent diffusion coefficients.
Species Production Terms
Each Wi in the species conservation equation is a summation of contribu-
tions from all reactions which include that species. For reactions of the type
"I"'
where jj are the stoichiometric coefficients on species B., W. can be written
as
- M.
(5)
m
319
-------
where
m
m
In P, SyJ In p.-ln K
' m
A
(6)
and kf has the Arrhenius form
m
=aTbe-(E/RT)
(7)
m
and v. denotes third body efficiency of species i.
m
As discussed in Reference 1, solving the species equations, including the
above terms, is difficult. In a typical flame problem, reaction rates will
range from nearly equilibrated to inactive. A nearly equilibrated reaction
will dominate all of the species equations in which it appears, making them
very difficult to solve since they are all nearly identical. To solve this set
of equations, several species equations must be combined so that the dominating
reaction appears in only one equation. This equation is effectively an equi-
librium relationship. By rearranging equations, a solution is more easily
obtained.
However, even if the equations are rearranged, the iterative Newton-
Raphson procedure used to solve them may not work, if the character of the equa-
tions has changed significantly. In the PROF kinetics package, a combination
of gradual recharacterization of equations and damping has proved highly suc-
cessful in finding a solution. This is of great importance if an efficient
grid-type coupled diffusion/kinetics boundary value solution is to be achieved.
Species Flux at the Wall
The PROF formulation uses an efficient transfer coefficient approach to
establish the species flux at the wall. In this approach, the species flux is
assumed to be directly proportional to the difference between the bulk and wall
gas composition. Since detailed radial distribution of the species concentra-
tion need not be known, this formulation remains one-dimensional. For consis-
tency, the wall flux proportionality factor is developed similar to the axial
species diffusion, and is written as:
wi
-f-4
(8)
w
320
-------
The scale factor f is included in Equation (8) to bring the effective diffusion
coefficient into line with detailed theoretical results for flow in a tube. The
value of f depends on the geometry and the state of boundary layer development.
Bulk Gas Volumetric Heat Loss
The term Q in the energy equation is a bulk gas phase volumetric heat
source or sink term. If desired, Q can be assigned as a constant or as a func-
tion of distance along the flame axis. However, most often Q is the volumetric
heat loss due to radiation. The PROF code uses the emission-dominated or opti-
cally thin limit approximation to model radiation heat loss. For nonsooting
flames of interest, this approximation is reasonable, and is written as:
Q = -4K
(9)
where K is the Planck mean absorption coefficient defined as
K
P
f
J
(10)
The individual Planck mean absorption coefficients, used to construct Kp, are
developed from the gas emissivity wide band correlation parameters of Edwards
and Balakrishnan"{Reference 3). Gas emissivities of the major radiating species,
COg, H20 and CO, obtained using this method differ from experimental data by less
than 10 percent (Reference 4).
Wall Heat Flux
For nonreactive tube walls, the PROF code formulation treats the wall heat
flux by assuming that the heat flux is proportional to the difference between
bulk and wall gas temperatures. For the reactive wall, an additional term is in-
cluded to account for the heat released by chemical reactions which occur be-
tween the bulk gas and wall. The general expression for wall heat flux is then:
_f_£D
(T -
r£
(11)
where f is the same factor as was introduced into the wall species flux expres-
sion and F, Is a
"mean"
Lewis number given by
= Pp
(12)
321
-------
Sped al 1 zed Boundary Condi t i ons
For flames In free space, the upstream boundary conditions are the temper-
ature and composition of the unburnt gas and the downstream boundary conditions
are negligible diffusion of species and heat. In the laboratory, flames are
anchored to their burners with flameholder grids. These grids stabilize the
flame by providing a sink for the flame's "excess" heat or reactive species,
but they also alter the flame properties. To accurately predict laboratory
burner flames, the effects of flameholder processes on flame properties must
be included in the code.
One characteristic of flameholders is the absence of heat or reactive
species diffusion upstream of the grid. The PROF code now can model this
effect by setting the upstream diffusion equal to zero. In addition, the tem-
perature of the gas at the flameholder can be set equal to the measured flame-
holder temperature. This specification defines the heat loss from the gases to
the flameholder surface, and is useful for cases where nonnegligible amounts of
heat are lost to a flameholder of known temperature,
Solution of the Finite Difference Form of the Conservation Equations
The species equation (1) is reduced to algebraic form by introducing a
normalized distance coordinate, dividing the species concentration equation
through by molecular weight, M-j, to obtain concentration per unit mass, a-j,
and applying linear finite differencing. The lengthy expression produced is
given in Reference 5. The algebraic form of the energy equation can be deter-
mined similarly, and is also given in Reference 5. These algebraic equations
are solved by a predictor-linearized corrector solution procedure which consists
of the following steps:
1. Initial values are selected for a-j, T, h, Cp at all grid points.
These may be output from a prior run or may be generated by a linear
interpolation between initial and guessed final values.
2. By applying known upstream conditions and the initial guessed values,
grid point values for a-j, h, T, etc., are found through matrix solution
of the equation set.
3. When the downstream boundary is reached, the no-diffusion boundary
condition is applied. {Guessed solution values are not required for
this grid point.)
4. Using the derivatives of o^ obtained from chemistry solutions at all
grid points, the rate of change of all a-j's with respect to initially
guessed ot^'s at each grid point are constructed.
5. Assuming the system is linear, corrections to all ot^'s are made by
applying the derivatives from Step 4.
6. Using the corrected a-j's as new guesses, Steps 2 through 5 are repeated
until the guessed a-j equals the corrected a-j.
322
-------
Although solutions have been found without using Step 5, the corrector
step can reduce the number of iterations by factors of 2 to 10. Since Step 5
takes only 10 to 15 percent of the total computation time, it is a very effec-
tive method for reducing total computer time. The corrector step also allows
the flame speed to 6e predicted by applying constraints on the grid variables.
The flame speed, or mass flow rate, m, is obtained by assigning a concentration
between the inlet and final equilibrium value to a species at a grid point
away from the inlet. This value may be assigned to any stable species which
monotonically increases or decreases through the flame.
323
-------
APPLICATIONS
The PROF code can predict confined flames, as shown in the sample calcula-
tions of CH^/air flames in nonreactive and reactive tubes. In addition, the
PROF code can be applied to complex coupled combustion and pollutant formation
problems, as shown below in predictions of NO pollutant formation in CH/^/air
free flames over a wide range of equivalence ratios.
CONFINED FLAME PREDICTIONS
To demonstrate the PROF confined flame predictive capability, calculations
of flame quench in nonreactive tubes and flame phenomena in reactive wall cata-
lytic combustors are presented.
Nonreactive Wall Heat Loss
To demonstrate the PROF nonreactive wall heat loss model, stationary Clfy/
air flames in small diameter tubes are predicted and compared with experimental
data. The flow velocity necessary to maintain stationary CHa flames in small
diameter tubes is a function of tube diameter. It is commonly believed that the
primary process governing flame velocity in tubes is conductive heat loss from
the flame to the tube walls. Wall chemical kinetics, radiant heat exchange and
buoyancy induced convection effects are secondary to this heat loss for methane
flames in small diameter tubes. Therefore, the PROF option for nonreactive wall
heat loss is applied to model stationary flames in small diameter tubes. PROF
flowrate predictions and methane flame quench and propagation data and natural
gas incipient flame flashback data are compared in Figure 1. Predictions and
quench data were for a 9.5 percent methane air mixture at 1 atmosphere pressure.
To model CH4 combustion 15 species (CH/i, 02, No. CHo, CH20, CO, C0?, H?0,
Ho, CHO, H, HO, 0, H02, CH) and 37 reactions were specified in the code.
(The reactions and their associated rates are given in Table I.) When applied
to freely propagating flames this mechanism gives reasonable flame speeds.
Figure 1 shows the good agreement between the predicted and measured flame
quench diameter (i.e., the zero flowrate tube diameter). Some agreement is also
found between the predicted and measured natural gas and methane flame flowrate
versus tube diameter values. However, these latter comparisons are not strictly
valid since the natural gas mixture is not equivalent to methane in composition.
In addition, propagating methane flames are not equivalent (in terms of tube
heat loss) to stationary methane flames. However, the predictions and data show
the same trends, and the agreement at the quench point shows that the PROF non-
reactive wall heat loss model is adequate.
324
-------
Reactive Wall Heat Loss
Predictions of flame phenomena in catalytic combustors are presented below
to demonstrate the PROF reactive wall heat loss or gain model. These predic-
tions also illustrate the Importance of "flame" type phenomena in these com-
bustors.
The catalytic combustor used for this prediction consists of a 3.66-inch
diameter cylindrical block of ceramic material. Within this cylinder, 2300
circular channels 0.06 inch in diameter have been constructed. The interior
surface of these channels are coated with a catalyst material which helps
convert fuel and air to product molecules. The rate of this conversion de-
pends on the flowrate of gases through the channels, the surface area of the
channels, the reactivity of the catalyst, and the length of the catalytic com-
bustor.
Catalytic combustor experiments have shown that under certain conditions,
fuel concentration is rapidly reduced beyond that expected on the basis of wall
reactions. These rapid reductions abruptly disappear once the bed flow veloc-
ity is increased above a limiting value. The abruptness of the reductions,
coupled with the short length over which they occur indicates that a "flame"-
type phenomena is responsible for reducing the fuel concentrations.
To demonstrate the "flame" phenomena in catalytic combustors, several PROF
calculations were carried out. A 200 percent theoretical air Cfy/air mixture
at 540°K was assigned as the initial gas condition. The wall condition was
assumed to be the equilibrium composition of the inlet mixture at the adiabatic
flame temperature. This condition is reasonable for the low flowrate (3.25 gm/
sec) specified. Predictions were made for flows without axial diffusion, but
with gas phase chemical reaction; with axial diffusion, but without gas phase
chemical reaction; and with both axial diffusion and gas phase chemical reac-
tion.
As the predicted fuel concentration decay rates in Figure 2 show, both
axial diffusion and gas phase chemical reaction must be included in the modeling
to predict a rapid "flame" like reduction in fuel concentration. If either of
these effects 1s not Included, the decay 1s solely due to wall reactions and a
much longer bed length is required to reduce fuel concentrations to very low
levels.
The "flame" type phenomena 1s also shown in Figure 3, where including both
axial diffusion and gas phase chemical reaction causes a rapid rise In gas
temperature at roughly the 1.5 cm station. Examining the detailed output for
the "flame" case shows that surface reactions, prior to the "flame" region,
preheat the gases causing a "light off" of the gas phase reactions. This phe-
nomenon is flame-Uke since it 1s characterized by a rapid decay in the fuel
concentration, a rise in radical (0, H, OH) concentrations, peaks in H£ and CO
concentrations, and decay of CO and Hg to product molecules, COg and H20.
Since the surface reactions are important in preheating the gas to sustain
the flame, varying the tube diameter (which alters the wall heat and mass trans-
fer effect) may change the "light off" length. A series of PROF calculations
325
-------
has shown that the "light off" length does change with tube diameter for all
other parameters fixed. These results indicate that the tube diameter can be
varied to reduce the overall length of the catalytic combustor. Besides
determining how "flame" characteristics vary with tube diameter, the PROF code
also can be used to assess the impact of initial mixture composition, preheat,
and tube geometry on "flame" phenomena.
METHANE FLAME COMBUSTION AND NO POLLUTANT FORMATION CHARACTERISTICS
PROF predictions of low pressure CH4/02 flame species concentrations are
compared with detailed measurements to demonstrate that the formulation and
the CH4 chemical combustion mechanism and transport data used in the calcula-
tions are adequate. This combustion mechanism and the well-known extended
Zeldovich NO mechanism is then applied to atmospheric pressure CH^air flames
in an attempt to predict NO pollutant formation over a wide range of equivalence
ratios. Comparing predictions and data, the Zeldovich mechanism is found to
be adequate for lean and stoichiometric flames, but inadequate for fuel-rich
flames. An alternate chemical mechanism is then suggested which can account
for NO production in fuel-rich CH4/air flames.
Verification of CH4 Chemical Combustion Mechanism
Reliable predictions of coupled combustion and pollutant formation pro-
cesses in CH4 flames requires specification of adequate chemical combustion
mechanisms. For this study, 13 species (CH4, Og, CHg, CHO, CH20, CO, 0, OH, H,
H20, C02, H02, H2) and 28 reactions were used in PROF to model CH4/02 combustion.
The reactions ana their associated rates are given in table II. To verify this
mechanism and the transport data input into the code, predictions of low pres-
sure CH4/02 flames employing this mechanism are compared with the extensive data
of Peeters and Mahnen (Reference 8).
The data of Peeters and Mahnen is an important basis of comparison, since
in this data 12 stable and unstable species concentrations and temperature are
available for comparison with predictions as a function of position through the
flame. Also, comparison of predictions and this data are straightforward, since
the experimental conditions under which the data was taken gives accurate spatial
resolution of flame properties and minimal impact of flameholder on flame prop-
erties.
PROF predictions are favorably compared with the detailed species and
temperature data of Peeters and Mahnen in Figures 4 through 7. The pressure for
this flame was 0.053 atmosphere and the initial temperature was 298°K. Initial
concentrations of CH4 and 02 were 0.095 and 0.905, respectively. The comparisons
in Figures 4 through 7 show that temperature and CH4, CO?, H20, 0£, CO, CHaO,
H2, OH, 0 and H concentrations are adequately predicted throughout the flame,
but CH3 and H02 predictions are higher than the experimental data. This poor
agreement indicates that some uncertainty still remains in the CH4 combustion
mechanism. Nevertheless, considering experimental errors and the uncertainty
of rate data, the combustion mechanism is adequate.
The PROF code was then used to predict atmospheric pressure CH4/air flames
as a function of equivalence ratio, and the predicted flame speeds were com-
pared with data (Reference 9) in Figure 8. Agreement of flame speed using this
326
-------
chemical mechanism is good for lean and stoichiometric flames. However, for
fuel-rich flames ($ > 1.15), predicted flame speeds are higher than measured
values.
As noted in Reference 10, predicted flame speed varies with the maximum
level of radical ([OH], [H]) concentrations in the flame for flames at various
pressures and fixed composition, as well as those at fixed pressure and vari-
able composition. Using the predicted flame speeds and maximum radical con-
centrations ([OH], [H]}, from Reference 10, it has been shown in Reference 5
that flame speed is roughly proportional to maximum radical concentrations
([OH], [H]) in the flame. Furthermore, the proportionality of flame speed to
maximum H atom concentration has been experimentally observed (Reference 11)
and theoretically demonstrated (Reference 1) for fuel-rich H2/air flames.
These results indicate that the CH4 chemical combustion mechanism given in
Table II probably generates excess of H radicals under fuel-rich ($ > 1.15)
conditions.
To check this conjecture, a reaction which depleted H atoms by recombining
them with CH3 was input into the code along with the reaction set given in
Table II. Predicted maximum H atom concentrations and flame speed were reduced
considerably. Figure 8 compares the flame speeds achieved by the revised mech-
anism with data at equivalence ratios of 1.05, 1.15 and 1.3. These results
demonstrate the relationship between radical concentrations and flame speeds.
At an equivalence ratio of 1.3, the baseline chemical mechanism given in
Table II will overpredict H atom concentrations by as much as 15 percent. This
conclusion will be important in subsequent discussions of NO pollutant forma-
tion in fuel-rich flames. At lower equivalence ratios, the mechanism in Table II
should be accurate.
NO Pollutant Formation in CH4 Flames
The formation of NO in combustion systems is typically assumed to occur at
a much slower rate than the rate of combustion. Predictions of NO concentra-
tions can be made by applying the Zeldovich mechanism:
NO + N
N + 0,
NO + 0
where 0 and 0? concentrations are determined by assuming all of the combustion
products are in equilibrium. For fuel-rich conditions, this NO mechanism is
usually augmented by the reaction:
N + OH
NO + H
with OH determined through the assumption of equilibrium combustion products.
NO predictions using this approach fall short of measured values for many pre-
mixed flame cases, particularly for fuel-rich hydrocarbon/air combustion.
327
-------
Experimentally, it has been found that within and near the flame zone, the
NO formation rate is much greater than downstream measurements and predictions.
This rapid NO formation near the flame zone has been termed "PROMPT NO" by
Fenimore (Reference 12). This "PROMPT NO" accounts for most of the differences
between measurements and predictions. Recently, several studies (References 1,
13, 14, 15, 16} have shown that the extended Zeldovich mechanism can treat flame
zone or "PROMPT NO" in many flame systems if actual nonequilibrium concentra-
tions of 0, OH and H radicals are used instead of post-flame equilibrium values.
NO produced by the Zeldovich mechanism is increased significantly by this "non-
equilibrium radical" approach, since radical concentrations are typically far in
excess of calculated equilibrium values in the flame zone. These results show
that accurate predictions of premixed flame NO concentrations require a coupled
approach, which includes both a combustion mechanism to model the radical con-
centrations and a pollutant formation mechanism to determine NO concentrations.
Several approaches can be used for coupling NO formation and combustion.
In Reference 15 a partial equilibrium approach was used to determine the radical
concentrations in premixed CH/j/air flames. These radical concentrations were
then applied in the Zeldovich NO mechanism to predict NO concentrations in
1 atmosphere, CH/j/air flames at 0.89, 1.05 and 1.15 equivalence ratios.
Good agreement was found between the NO levels predicted and data at equi-
valence ratios of 0.89 and 1.05. However, at the highest equivalence ratio,
1.15, the measured rates of NO production did not.match those predicted. For
this case, the data indicated a sharp peak and a rapid fall off in NO production
rates through the flame, whereas predictions showed a lower peak and a sustained
high rate through the flame. The authors suggested a number of reasons for the
lack of agreement. However, it was clear that under fuel-rich conditions, the
Zeldovich mechanism may not account for all of the NO formation.
In Reference 16, a straightforward coupled approach, which includes both
combustion and pollutant formation chemical kinetic mechanisms, was used to
determine NO formation in atmospheric pressure CH4 combustion.* The detailed
CH4 chemical combustion scheme of Bowman (Reference 13) was combined with the
extended Zeldovich mechanism to determine NO formation over a range of equiv-
alence ratios. The predictions qualitatively had the same NO formation behav-
ior as was observed in Reference 15. However, the important flame initiating
processes of diffusive heat and mass transport were not treated in these cal-
culations. This necessitated the assignment of arbitrarily and artificially
high initial temperatures in the calculations to initiate combustion. Therefore,
this approach can only provide qualitative information on flame zone processes
where diffusion is significant and most of the "PROMPT NO" is formed.
The uncertainties in partial equilibrium and nondiffusive approaches pre-
vent one from concluding whether the Zeldovich NO formation mechanism 1s ade-
quate or inadequate for premixed CH4/air flames. However, using the PROF code,
which can treat coupled chemistry as well as diffusion, removes these uncer-
tainties and clearly demonstrates the limitations and range of applicability
of the Zeldovich mechanism.
The use of the term flame was avoided in Reference 16 since the important
flame processes of diffusion were not included in the calculations.
328
-------
PROF predictions of atmospheric pressure CH4/ai,r premixed flames were
carried out and compared with data over a wi;de range of equivalence ratios.
The extended Zeldovtch NO formation reactions used in the calculations, and
thefr associated rates are;
OH + N -» H + NO kf = 4.00 * 1013 T°'s
0 + N2 * N + NO
-6,300
N + 02 * NO + 0 kf 6.50 x 109 T e
kf » 1.4. x If)1" e
-75.200
RT
The combustion mechanism and transport data applied in the NO predictions are
identical to those used to calculate low pressure CH4/02 flames. The reactions
and their rates are given in Table II. An initial temperature of 298°K was as-
sumed for the reactants and predictions were made for equivalence ratios of
0.7, 0.89, 0.952, 1.05, 1.15 and 1.3. Figure 9 presents NO concentrations
for the various equivalence ratios as a function of time through the flame.
These results show that, as the equivalence ratio increases from 0.7 to
0.952, the NO concentration rises from a negligible value to a maximum. Further
increases of equivalence ratio decrease NO until at an equivalence ratio of 1.3
very low levels of NO are predicted. For lean and near stoichiometric equi-
valence ratios, this behavior is in qualitative agreement with the data in
Reference 15.
At equivalence ratios of 0.89 and 1.05, peak NO production rates obtained
from the detailed computer printout are in good quantitative agreement with peak
rates derived from the data in Reference 15. At an equivalence ratio of 1.15,
the predicted ppst-f1ame zone formation of NO appears to be qualitatively correct.
However, the rapid increase of NO within the flame zone, observed in Reference 15
is not found in the predictions.
Comparison of the maximum predicted and measured peak NO formation rates
indicates that the Zeldovich mechanism is significantly underpredicting NO
formation within the flame zone under fuel-rich conditions. This is clearly
demonstrated in Figure 10 where predicted flame zone NO concentrations are com-
pared to "PROMPT NO" values taken from Reference 10. The experimental levels of
"PROMPT NO" (equivalent to flame zone NO) were determined by extrapolating post-
flame NO concentrations back to the burner face and using the nonzero intercept
value. In the predictions, the flame zone was defined as the post-City depletion
and radical peak region. This roughly corresponds with the 2-msec time in
Figure 9 for all of the equivalence ratios investigated.
The comparison of predictions and data in Figure 10 shows that "PROMPT NO"
is adequately treated by the coupled nonequilibrium radical and Zeldovich NO
329
-------
mechanism for equivalence ratios jp to 1.05. Above 1.05, the predictions fall
rapidly to zero (* - 1,3) whereas measurements indicate a rise up to an equi-
valence ratio of 1.4, followed by a decrease to very low levels at an equivalence
ratio of about 1.6,
As discussed previously, considerable care was taken in developing a com-
bustion mechanism which generates accurate levels of radical concentrations
through the flame. Comparisons of predictions and flame speed data indicated
that radical concentrations were adequately predicted for equivalence ratios up
to 1.15. At an equivalence ratio of 1.3, it was estimated that the radical con-
centration was probably high by 15 percent. Therefore, at equivalence ratios
greater than 1.05, the Zeldovich mechanism cannot account for measured "PROMPT
NO," even when the correct or even enhanced radical concentrations are used in
the NO reactions. For equivalence ratios lower than 1.05, the Zeldovich mecha-
nism can account for "PROMPT NO" if radical concentrations are properly evaluated.
Having shown that the Zeldovich NO mechanism is inadequate in fuel-rich
flames, an alternate mechanism was sought to account for "PROMPT NO" in these
flames. Fenimore (Reference 12) and Iverach, et al. (Reference 17), experimen-
tally measured hydrocarbon flames, and concluded that the "PROMPT NO" formed
under fuel-rich conditions results from energetic hydrocarbon fuel fragment
molecules reacting with molecular nitrogen. Reactions such as:
CH + N0 + HCN + N
C + N0 * CN + N
were thought to produce N atoms that would react with OH via the extended Zeldo-
vich mechanism to form NO. The unique presence of significant fuel- rich "PROMPT
NO" in hydrocarbon flames and the excessively high 0 atom radical concentration
required to produce this NO by the Zeldovich mechanism reinforced the above
author's hydrocarbon fragment hypothesis.
Recently, several experimental and theoretical studies have added more evi-
dence to support the hydrocarbon fragment theory. In References 18 through 21,
HCN was found 1n significant quantities (>10 ppm) near the reaction zone of fuel-
rich hydrocarbon flames. Also, References 19 through 21 showed that NO rapidly
forms within the flame zone as HCN decays. In Reference 19, data indicated that
about 90 percent of "PROMPT NO" can be formed by way of HCN. Theoretical studies
of CH4 combustion (Reference 22) in jet-stirred reactors demonstrated that fuel-
rich NO can be accounted for 1f HCN is included as an intermediate in the NO
formation mechanism.
The fuel-rich NO mechanism applied in this study consists of:
CH, CH2 hydrocarbon fuel fragment attack on molecular nitrogen
Formation of atomic nitrogen or nitrogen bearing intermediates (HCN,
CN, NCO, NH)
330
-------
o Oxidation of atomic nitrogen or nitrogen-bearing intermediates to
NO
Alternate reaction paths for NO back to N2 or HCN
This mechanism may be schematicized as:
HCN * CN + NCO * NHj + N
The reactions assumed to take place between these species are listed in
Table III. Although additional nitrogen-bearing species and many other re-
actions can be included in this mechanism, the mechanism is plausible and In-
cludes enough reaction steps to account for the observed phenomena. However,
considering the limited amount of data and the uncertainty of key reaction
rates, other mechanisms, which were not examined in this study, also might be
able to account for the data.
Flame speed and "PROMPT NO" levels predicted by the fuel-rich mechanism at
1 atmosphere pressure are compared with experimental data in Figures 8 and 10.
As shown in Figure 8, predicted flame speeds and data agree at equivalence ratios
greater than or equal to 1.3. Below 1.3 the predictions fall below data. These
comparisons show that the chemical combustion mechanism, which includes NO pol-
lutant formation, is reasonable under fuel-rich conditions.
In Figure 10, prompt NO predictions, over a wide range of equivalence ratios,
show the same trend as experimental data. Predictions fall somewhat higher than
data. However, this lack of agreement is not critical, since there are consider-
able uncertainties inherent in defining experimental "PROMPT NO." At an equi-
valence ratio of 1.15, the calculated maximum rate of NO production closely
agrees with the experimental value found in Reference 15.
Examining the calculations in detail shows that the fuel-rich mechanism
reduces to the Zeldovich mechanism at equivalence ratios less than 1.05. At
equivalence ratios above or equal to 1.3:
t Total NO production rate decreases due to the reaction steps
N + NO -* N2 + 0
NO + CH -» HCN + 0 or CHO + N
331
-------
t The species sets HCN and CN, and NH?, NH and N become equilibrated
through reactions such as
HCN + H £ CN + H,
NH -f H £ N + H,
NH2 + H * NH + H2 etc.
Significant NH^ is produced within the flame and remains downstream
of the flame
It is also noted that HCN decays rapidly at equivalence ratios of 1.15 and
1.3, but much slower at equivalence ratios greater than or equal to 1.5. At
these higher ratios, significant concentrations (-20 ppm) of HCN remain down-
stream of the flames.
In summary, the hypothesized fuel-rich mechanism used in this study can
account for "PROMPT NO" over a wide range of equivalence ratios, whereas the
Zeldovich mechanism by itself is inadequate at ratios equal to or greater than
1.15. However, firm conclusions on the applicability of this mechanism to
flames require further comparison between predictions and detailed data at a
variety of conditions.
332
-------
CONCLUSIONS
Predictions of confined flame and plug-flow reactor properties have shown
that the upgraded and expanded PROF code is versatile, reliable, quick and
accurate. By comparing extensive CH4/02 flame data (Reference 8) and PROF pre-
dictions, an adequate chemical combustion model for CH4 has been developed.
When coupled to a chemical kinetic combustion mechanism, the Zeldovlch NO
formation mechanism is adequate for predicting CH4/air flame zone "NO" forma-
tion at near-stoichiometric and lean conditions. However, under fuel-rich con-
ditions {$ >1.05), the Zeldovich mechanism significantly under-predlcts
CH4/air flame zone, or "PROMPT NO." formation.
A "hydrocarbon fuel fragment" chemical mechanism can account for "PROMPT
NO" in fuel rich CH4/air flames, as demonstrated by comparing predictions and
data. The mechanism considered HCN as an intermediate, with N and NH molecules
also being produced. Most of these nitrogen-containing species were then
oxidized to NO or converted back into N-
333
-------
LIST OF SYMBOLS
B
w
D
E
f
F1
h
1
J
«P
cross sectional area
monochromatic radiation intensity
specific heat
circumference of bounding tube
binary diffusion coefficient
diffusion constant defined by Equation (3)
activation energy for kinetic reaction
scale factor on radial heat and mass transport terms
diffusion factor of species i
enthalpy of bulk gas
denotes species when used as subscript
flux of species 1 in axial direction
flux of species 1 at bounding tube wall
thermal conductivity of bulk gas
spectral absorption coefficient
Planck mean absorption coefficient
334
-------
m
M
P
q
q,
w
w
R
S
T
Wi
Xi
m
LIST OF SYMBOLS Continued
equilibrium constant for reaction m
mass rate of gas
molecular weight
pressure
axial heat transport
heat transport at the bounding tube wall
volumetric heat loss
radius of bounding tube
gas constant
distance along flame axis
temperature
chemical production rate of species i
mole fraction of species i
mass fraction of species i
species concentrations in moles per unit mass
third body efficiency of species i in reaction m
density
Stefan-BoItzmann constant
335
-------
LIST OF SYMBOLS Concluded
Superscripts
P reaction products
R reaction reactants
336
-------
REFERENCES
1. Kendall, R. M. and Kelly, J. T., "Premixed One-Dimensional Flame Code
(PROF) - Its Formulation, Manipulation, and Evaluation," Aerotherm
Report TR-75-158, July 1975.
2. Bartlett, E. P., Kendall, R. M., and Rindal, R. A., "A Unified Approxi-
mation for Mixture Transport Properties for Multicomponent Boundary
Layer Applications," Aerotherm Final Report 66-7, Part IV (also NASA
CR-1063), March 14, 1967.
3. Edwards, D. K. and Balakrishnan, A., "Thermal Radiation by Combustion
Gases," Int. J. Heat and Mass Transfer, Vol. 16, 1973, pp. 221-230.
4. Hottel, H. C. and Sarofim, A. F., Radiative Transfer, McGraw-Hill Book
Company, New York, 1967, pp. 229-236.
5. Kelly, J. T. and Kendall, R. M., "Further Development of the Premixed
One-Dimensional Flame (PROF) Code," Aerotherm Final Report 76-229,
August 1976.
6. Singer, J. M. and von Elbe, G., "Flame Propagation in Cylindrical Tubes
Near the Quenching Limit," Sixth Symposium (International) on Combustion,
Reinhold Publishing Corporation, New York, 1957, pp. 127-130,
7. Lewis, B. and von Elbe, G,, Combustion, Flames and Explosions of Gases,
Academic Press, Inc., New York, 1961.
8. Peeters, J. and Mahnen, G., "Reaction Mechanisms and Rate Constants of
Elementary Steps in Methane-Oxygen Flames," Fourteenth Symposiurn (Inter-
national) on Combustion, The Combustion Institute, Pittsburgh, Pennsyl-
vania, 1973, p. 133.
9. Andrews, G. E. and Bradley, D., "The Burning Velocity of Methane Air
Mixtures," Combustion and Flame, Vol. 19, 1972, pp. 275-288.
10. Smoot, L. D., Hecker, W. C., and Williams, G. A., "Prediction of Propa-
gating Methane Air Flames," Combustion and Flame, Vol. 26, No. 3, June
1976, pp. 323-342.
11. Padley, P. J. and Sugden, T. M., "Chemiluminescence and Radical Recombi-
nation in Hydrogen Flames," Seventh Symposium (International) on Combus-
tion, Butterworths, London, 1959, p. 235.
12. Fenimore, C. P., "Formation of Nitric Oxide in Premixed Hydrocarbon
Flames," Thirteenth Symposium (International) on Combustion, The Combus-
tion Institute, 1971, p. 373.
13. Bowman, C. T., "Kinetics of Nitric Oxide Formation in Combustion Pro-
cesses," Fourteenth Symposium (International) on Combustion, The Com-
bustion Institute, 1973, p. 729.
337
-------
14, Homer, J. 8. and Button, M. W., "Nitric Oxide Formation and Radical Over-
shoot in Premixed Hydrogen Flames," Combustion and Flame, Vol. 20, 1973,
pp. 71-76.
15. Sarofim, A. F. and Pohl, J. H,, "Kinetics of Nitric Oxide Formation in
Premixed Laminar Flames," Fourteenth Symposium (International) on Com-
bustion, The Combustion Institute, 1973, p. 739.
16. Ay, J. M. and Sichel, M., "Theoretical Analysis of NO Formation Near the
Primary Reaction Zone in Methane Combustion," Combustion and Flame, Vol.
26, No. 1, February 1976, pp. 1-17.
17. Iverach, D., Kirov, M. Y., and Haynes, B. S., "The Formation of Nitric
Oxide in Fuel-Rich Flames," Combustion, Science and Technology, Vol. 8,
1973, p. 159.
18. Bachmaier, F., Eberius, K. H., and Just, T., "The Formation of Nitric
Oxide and the Detection of HCN in Premixed Hydrocarbon-Air Flames at One
Atmosphere," Combustion. Science and Technology, Vol. 7, 1973, p. 77.
19. DeSoete, 6. G., "Overall Reaction Rates of NO and Ng Formation From Fuel
Nitrogen," Fifteenth Symposium (International) on Combustion, The Com-
bustion Institute, Pittsburgh, Pennsylvania, 1975, p. 1103.
20. Kahn, D., "Low Pressure Flat-Flame Burner Studies," presentation made at
U.S. Environmental Protection Agency Contractor's Meeting on Fundamental
Combustion Research, Boston, Massachusetts, August 11-13, 1976
21. Haynes, B. S., Iverach, D., and Kirov, N. Y., "The Behavior of Nitrogen
Species in Fuel-Rich Hydrocarbon Flames," Fifteenth Symposium (Interna-
tional) on Combustion, The Combustion Institute, 1975, p. 1103.
22. Waldman, C. H., Wilson, R. P., and Maloney, K. L., "Kinetic Mechanism of
Methane/Air Combustion with Pollutant Formation," Environmental Protection
Technology Series, EPA-650-/2-74-045, June 1974.
338
-------
CD
o o o
\c m M
o o o o
» « * *
o o 3-
o o o
o »o
« in
K> o m o CM *
o o
o in
o o
o
o
o o
in
n
o
t-« » O
in o a a o
in o in
o o o o
ol O O O O
* o
r-t O
CO
o
o
oooooooooooinooooaaaoaoaoooaoooooaoocio
oooooooooooMoootnoooknoooooinominoaomointno
cc
o c\r *
a. J-
CM O IM <->
X O X X O X
-------
* e e o e e o o e o o o o e e e o e o o e o o e e o e e o
u ooooeoooeeoooeoooooooooooooo
n oeeeocceeoeeeeeoeoceaoeeocoo
U.
u
M ooooeeeoeoeoooooooooeeoooeoo
x e e e e o e e o e o o o o e e o e o e o e o e a o 9 a o
K eeeeoeooooeeoeeoooeoooeooooo
o
C/)
eeoeooeeooeoeoeooooooeoooeeo
oooooeoOBoeooeooeeoooeoooceo
OOOOOOOOOOOOOBOOOOONOmotfOOOO
in ir> »
««
-------
t> e» o o «a c
o o a a oooo>oc»«»c»&oc»oooo
o o o o
0000
UJ
oeoaoobooooooaooc
oa a a a a o r» a xC o a a o a <
^ O i-*
,* ni P- »-t
CJ
LU
T T T t £ O 3 J C,
i J. O fi 1 t X T ,.-
no a I ttO<>'vo'M(
LU
CO
M> i*) *) ") OJ f\* O O (J
JC X 3'ZX I O -L J. X O O O O
C» I I O irw
O II II II |l It LI II II It II II IJ II II II II II II It M M II II a ll II II H M II ii II (I |l it II (I II 41 II It M II II It U U U II U M II tl H II U M «
" t I I I I I I I I I I I I I llll*lllllllllll!l|lltlllll|||lttllttttll«t*
»- Illttlllllllltlllllllltllllllllllllllllllllllllltlllllllt*
ftt CJ CM O
O O 'V CJ O
(M T\| f\t fy tv f
O Q Q Q <* W O
:ooo xoxoxci 2.u^iXisDxj.Xi.£*oxioooxo-i:so.z«r;:ri*_ix£otoi, ^ -r z o X *
XX3.XZ1XX
I U M LJ l_l O .
: i x o 3. oouiUL
., ^ ? t-l i I
341
-------
<1)
0)
(V
M
01
«3
in
$
to
-------
o c c
i- O O
m -i- -i-
3 -»->«/>
M- O 3
l»_ <0 H-
i- OJ <4-
* ^'-5
o o> i
c c
C O.
o
i l/l
S C 0)
s_ o tf
3 Q-
i/l
O)
l/>
fO
D- »"
13 CD
t/\
as O O
13 Z Z
0)
a;
343
-------
-g
i-
o
V)
-O
O
U
(O
u
o
r
Id
M
(O
M
m
01
o.
03
C3
(X-,)
344
-------
0.12
0.10
0.08
c
o
cu
o
0.06
0.04
0.02
Experimental data (Reference 8)
PROF predictions
CH4/0?:0.095/0.905
p = 0.053 atmosphere
T = 298K
0.2 0.4 0.6
Distance above burner (cm)
0.8
fboo
1600
1200
800
400
1.0
Figure 4. Comparison of predicted and measured low pressure
CH4/02 flame species concentrations and temperature.
345
-------
0.88
0.84
CM
o
0.80
o
0.76
0.68
Experimental data (Reference 8)
PROF predictions
CH4/02:0.095/0.905
p = 0.053 atmosphere
T = 298K
).20
0.16
o
o
o
CM
X
0.12
-------
10"
(J
re
o
ae
in"
10
-5
Experimental data
(Reference 8)
PROF predictions
CH4/02:0.095/0.905
p = 0.053 atmosphere
T = 298K
0.2 0.4 0.6
Distance above burner (cm)
0.8
1.0
Figure 6. Comparison of predicted and measured low pressure
CH/0 flame species concentrations.
347
-------
lo"
g io~3
o
a
HI
"o
10
CHa/02:0.095/0.905
[i - 0.053 atmosphere
T - 298K
Exnerimental data (Reference 8)
PROF predictions
0.4 0.6
Distance above burner (cm)
1.0
Figure 7. Comparison of predicted and measured low pressure
CH4/02 flame species concentrations.
348
-------
1 1 1 1 1 1 1 1 1
at <«-
a>
t */
s~ fo oo ,/ X"
r CTl J
£=\
o> 1/1 -t- i/) \
0 C C .
C O (0 C \
QJ - x: io \
i. »-> o x:
QJ O QJ D \^
o: QJ o» o ^^P
t C -0 -r- ^i
O- -i- at s_ v
^ u. "a; £ r- ^.
+* o w) > aj x
c a: ID ai ri N
Oj Q- CO QJ LI
'^-
| X
UJ
I 1 1 1 1 1 1 1 1
-
-
-
-
-
-
_
-
i-~
"~~
^
in
CO
*" -e-
CM .?
'
i.
V)
O)
a.
u
t.
ai
o
B
(O
3
V}
o
o
Q.-P-
5
O 3
I/I O
(O
O -M
u re
oo
a>
(oas/iuo) ^LDOi.aA
349
-------
c
o
^
M
U
r"
o
O)
Q.
en
a>
en
350
-------
-V
00
eg csj
or a:
a> c
en a;
a>
*
c
O)
QJ
CD
(U
a;
en
o -c x:
0)
i
LU
a>
O) O)
0)
rxj u.
oo
o
UD
00
vo
CM
(mdd) ON
351
-------
TECHNICAL REPORT DATA fl
(Please reed /niirucribni on the reverse before completing) g
1. REPORT NO. 2.
EPA~600/7-77-073d
4. TITLE AND SUBTITLE PROCEEDINGS OF THE SECOND
STATIONARY SOURCE COMBUSTION SYMPOSIUM
Volume W. Fundamental Combustion Research
7 AUTHORtS) Symposium Chairman J.S. Bowen, Vice-
Chairman R. E. Hall
9. PERFORMING ORGANIZATION NAME AND ADDRESS
NA
12. SPONSORING AGENCY NAME AND ADDRESS
EPA, Office of Research and Development
Industrial Environmental Research Laboratory
Research Triangle Park, NC 27711
3. RECIPIENT'S ACCESSION NO. 1
5. REPORT DATE 1
July 1977 1
6. PERFORMING ORGANIZATION CODE 1
8. PERFORMING ORGANIZATION REPORT Nojj
1O. PROGRAM ELEMENT NO. fl
EHE624 I
1 1 . CONTRACT/GRANT NO. 1
H
NA (Inhouse)
13. TYPE OF REPORT AND PERIOD COVERED
Proceedings; 8/29-10/1/77
14. SPONSORING AGENCY CODE
EPA/600/13
15. SUPPLEMENTARY NOTES ffiRL-RTP project officer for these proceedings is R.E. Hall,
Mail Drop 65, 919/541-2477.
6. ABSTRACT The proceeciings document the 50 presentations made during the Second
Stationary Source Combustion Symposium held in New Orleans, LA, August 29-
September 1, 1977. Sponsored by the Combustion Research Branch of EPA's Indus-
trial Environmental Research Laboratory--RTP, the symposium dealt with subjects
relating both to developing improved combustion technology for the reduction of air
pollutant emissions from stationary sources, and to improving equipment efficiency.
The symposium was divided into six parts, and the proceedings were issued in five
volumes: Volume I--Small Industrial, Commercial, and Residential Systems; Volume
II--Utility and Large Industrial Boilers; Volume fflStationary Engine, Industrial
Process Combustion Systems, and Advanced Processes; Volume IV--Fundamental
Combustion Research; and Volume V--Addendum. The symposium was intended to
provide contractor, industrial, and Government representatives with the latest infor-
mation on EPA inhouse and contract combustion research projects related to
pollution control, with emphasis on reducing nitrogen oxides while controlling other
emissions and improving efficiency.
17. ' KEY WORDS AND DOCUMENT ANALYSIS
a. DESCRIPTORS
Air Pollution, Combustion, Field Tests
Combustion Control, Coal, Oils
Natural Gas , Nitrogen Oxides , Carbon
Carbon Monoxide , Hydrocarbons , Boilers
Pulverized Fuels , Fossil Fuels , Utilities
Gas Turbines , Efficiency
is. DISTRIBUTION STATEMENT
Unlimited
t>. IDENTIFIERS/OPEN ENDED TERMS
Air Pollution Control
Stationary Sources
Combustion Modification
Unburned Hydrocarbons
Fundamental Research
Fuel Nitrogen
Burner Tests
19. SECURITY CLASS (This Report )
Unclassified
30. SECURITY CLASS (Thiipage)
Unclassified
C. COSATt Field/Group
13B 21B 14B
21D UH
07B
07C 13A
13G 14A
21. NO. OF PAGES
355
22. PRICE
KPA Perm 2220-1 («-73)
352
-------
IBP 600/7 EPA
I Ind. Env. Res. Lab.
AUTHOR
Proc. of the second stationar
TITLE source combustion symposium.
: Fundamental combustion
DATE cWI5** arCft BORROWER'S NAME
OAVLORD *9
-------
DATE DUE
BORROWER'S NAME
i*y*
-------
DATE DUE
-------
------- |