Вы находитесь на странице: 1из 35

This article was downloaded by: [Stanford University Libraries]

On: 11 October 2012, At: 03:02


Publisher: Taylor & Francis
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer House,
37-41 Mortimer Street, London W1T 3JH, UK

Combustion Science and Technology


Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/gcst20

A Review of NOx Formation Under Gas-Turbine


Combustion Conditions
SANJAY M. CORREA
a

General Electric Corporate Research Center, Schenectady, New York, 12301, U.S.A.

Version of record first published: 27 Apr 2007.

To cite this article: SANJAY M. CORREA (1993): A Review of NOx Formation Under Gas-Turbine Combustion Conditions,
Combustion Science and Technology, 87:1-6, 329-362
To link to this article: http://dx.doi.org/10.1080/00102209208947221

PLEASE SCROLL DOWN FOR ARTICLE


Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions
This article may be used for research, teaching, and private study purposes. Any substantial or systematic
reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution in any form to
anyone is expressly forbidden.
The publisher does not give any warranty express or implied or make any representation that the contents
will be complete or accurate or up to date. The accuracy of any instructions, formulae, and drug doses should
be independently verified with primary sources. The publisher shall not be liable for any loss, actions, claims,
proceedings, demand, or costs or damages whatsoever or howsoever caused arising directly or indirectly in
connection with or arising out of the use of this material.

Combust. Sri. and Tech., 1992, Vol. 87, pp. 329-362

Gordon and Breach Science Publishers S.A.


Printed in United Kingdom

Photocopying permitted by license only

A Review of NO x Formation Under Gas-Turbine Combustion


Conditions

co R REA General Electric Corporate Research Center,


Schenectady, New York 72301, USA.

SANJAY M.

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

(Received January 3, /99/; in finalform October 28, /99/)


Abstract-Gas turbines offer very high cycle efficiency, exceeding 50% in modern 250 MW-c1ass combinedcycle units for power generation, as well as very low NO, in the lean premixed combustion mode with
natural gas fuel. They also account for virtually all commercial aeropropulsion systems, in which case
kerosene-based fuel is used, To meet future NO, and CO regulations, a higher level of understanding of
turbulence, chemical kinetics and their interactions is required. NO" in particular has become a pacing
consideration, although other constraints are present. Selected NO., data obtained at laboratory and
machine conditions with gaseous fuel are reviewed here. Although the important chemical reactions cover
a wide range in effective Damkohler number, the measure or turbulence-chemistry interactions, it appears
that NO. is formed in a distributed zone manner. Equilibrium and superequilibrium effects can broaden
the NO, -Iorming zones beyond the fine scales or turbulence, even in non-premixed flames. Pressure and
the structure or the turbulent flowfield have a diminishing effect on NO,. emissions in progressively leaner
premixed combustion. Generalizations such as these cannot be made for the related problems of CO
emissions and combustion-driven pressure oscillations.

NOMENCLATURE
A
D

o;
E"
k

In

M}
p

R
Re

Sc
T
II
\I'

X,
Y;
e
I)

8;
)_;
v
~

p
ai'

:E(

pre-exponential constant in reaction rate


diffusivity
Damkohler number, 8)!}
activation energy
reaction rate
exponent in pressure scaling
molecular weight of species "l"
pressure
gas constant
Reynolds number
Schmidt number, vjD
temperature
velocity
production rate
spatial coordinate
mole fraction of species "i "
mass fraction of species "i"
mechanical dissipation rate
reaction progress variable
time scale of mixing process "i"
length scale of mixing process" i"
kinematic viscosity
mixture fraction
density
sensitivi ty coefficient for species "l' to reaction" i", Eq. (13)
scalar scale
329

330

S. M. CORREA

time scale of reaction "1'''


equivalence ratio
subscripts
a
activation energy
f
flame
k
coordinate direction"k"
I
integral scale
K Kolmogorov scale
t
turbulence
superscripts
equilibrium
e
u
unburned
r

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

INTRODUCTION
Concerns regarding acid rain and depletion of stratospheric ozone have made NO,
emissions a pacing item in the design of gas-turbine powerplants and propulsion
systems. Stationary gas-turbines are regulated to 25 ppm or less of NO, in many
geographical regions, e.g., 9 ppm NO, at 15% 0, in parts of Southern California. NO,
emissions are corrected to constant (e.g., 15%) oxygen levels for ease of comparison.
Correction to constant oxygen implies correction to a standard reference state by
actual or conceptual post-flame addition of air, and is similar to the use of "emission
index" (grams of NO,/kg of fuel) in that the reported NO, level is not affected by the
degree of post-flame dilution. Emissions from stationary devices are regulated in
terms of so-corrected ppm's, while emissions from aeroengines are regulated in terms
of emission index. Low NO, regulations and the current worldwide availability of
natural gas have made lean premixed combustion the principal technology of interest
in stationary gas-turbines.
Since equilibrium NO, levels are on the order of 10' ppm in combustion, NO,
formation is a kinetically-limited process in the sub-I 00 ppm regime of interest.
"NO/' includes NO and NO" but in most flames it is predominantly NO that is
formed, with oxidation to NO, occurring in a post-flame process (Sano, 1985) that
does not affect total NO,. N,O is not a significant emission and is not defined as a
constituent of NO" although it can sometimes play the role of an intermediate in NO,
formation.
NO, emissions targets of 100ppm (at 15% 0,) can be met by management of
thermal NO, processes in non-premixed combustion (in the absence of fuel-bound
nitrogen), but a goal of 10ppm requires lean premixed combustion and a more
detailed level of understanding. Additional kinetic mechanisms and interactions with
turbulence that occur in lean premixed combustion are the subject of renewed
research. There is also the possibility that emissions standards will drop continuously,
according to "best available technology", and thereby present an open-ended challenge.
This paper is organized as follows. First, relevant chemistry and turbulence issues
are introduced. Although understanding of the chemistry of NO, has progressed a
great deal in the last two decades (Miller and Bowman, 1989), interactions between
turbulence and chemistry determine the yield of NO, and can appear in a variety of
ways. Theory can account for certain idealized cases only, limiting the guidance that
can be provided to the combustor designer. Second, performance characteristics
required of stationary and propulsion gas-turbines are presented. It will be seen that
NO, emissions, although of great importance, are not the only issue. In some cases,

331

NO, FORMATION
104
<Il

.
E

10'

0.
0.

a:
c:
0

.~

10

LL

z 10-'

<0

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

(;;

s:
l-

E 10-4

'"~

0
LL

10-
0.4

0_6

0_8

1.0

1.2

1.4

Equivalence Ratio

FIGURE 1 Forward thermal NO formation rate with inlet conditions typical of laboratory (I atm,
300K), utility gas-turbine (lOatm, 600K) and aeropropulsion gas-turbine (30atm, 900K) combustion.

the other issues rule against certain low-NO, combustion technologies. Finally, NO,
data from non-premixed and premixed combustion of gaseous fuels are reviewed.
Since NO, is produced in the flame and is not significantly affected by post-flame
processes under conditions of interest here, it may be used to diagnose the interaction
between mixing and flame chemistry. Such a viewpoint is important to consider in the
development of a comprehensive theory of flame structure, besides contributing to
low-NO, combustion technology.
Chemistry of NO, Formation
NO, is formed in methane flames by three mechanisms (Miller and Bowman, 1989):
the thermal mechanism, the so-called "prompt" mechanism, and the nitrous oxide
mechanism. Thermal NO, is formed by the reactions which occur between atmospheric
nitrogen and oxygen atoms. The formation rate IV is given by
w =

A[Nz][O] exp (- EaIRT),

(I)

where A is a constant and "[I]" represents the concentration of species "i ", Equation
(I), convenient for the purposes of discussion, may be written in this Arrhenius-like
form if the concentration of NO is limited to being small compared with the equilibrium
level. This assumption is justified at conditions of interest. The thermal mechansim is
significant only at temperatures above 1800K or so, because the activation energy E,
is large (-76 kcal/rnol). Figure I shows the forward rate of formation of thermal NO,
under laboratory (I atm, 300 K inlet), and reasonably typical utility-class power
generation (10 atm, 600 K inlet) and aeropropulsion conditions (30 atm, 900 K inlet).
The fuel is methane. This rate is based on equilibrium N z, O-atoms and temperature;
non-equilibrium effects are discussed in Section 3 and are less significant to thermal

332

S. M. CORREA

NO,. at elevated conditions. It is clear that very little time can be spent in stoichiometric gas under machine conditions, if NO x goals are to be met. A significant
difference between formation rates under laboratory and machine conditions is also
apparent.
NO, formation is very sensitive to temperature perturbations if in the thermal
regime, as may be deduced from Eq. (I). The explicit temperature dependence is

w o: exp (- Ta/T),

(2)

where the activation temperature T; == Ea / R is about 38000 K. Writing the flame


temperature T as

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

= Tf + tlT,

(3)

where 1fis the original flame temperature and .1. Tis the perturbation, and since it may
be assumed that tlT ~ Tf , the exponential term can be expanded to yield
w

o: exp (- Ta/1f) exp (Ta.1.T/TJ)

(4)

With 1f typically 2200 K and T; as above, it follows that a perturbation of 90 K will


change the rate w by a factor of - 2. The effects of temperature on radical levels would
contribute further. Temperature decrements caused by non-adiabaticity can therefore
be very consequential.
Prompt NO x was identified by Fenimore (1970), in a study of laminar premixed
ethylene-air flames in the range 1-3atm. When the axial profile of NO,. along the
burner centerline was extrapolated back to the surface of the burner, the non-zero
intercept obtained was called "prompt" NO,. This apparently instantaneous formation
of NO., observed in hydrocarbon flames has two contributions: part is due to the
thermal mechanism, enhanced by rapidly-formed superequilibrium radicals, and the
rest is due to a hydrocarbon-plus-nitrogen mechanism (Fenimore, 1970; Iverach et al.,
1972). Both contribute to production of NO, too close to the burner to be resolved
spatially. The hydrocarbon route has come to be identified with the term "prompt"
NO,. Fenimore's (1970) original paper estimated that prompt NO, accounted for
30% of the total NO, in a non-premixed gas-turbine combustor of the time.
The - 14 kcal/rnol activation energy of the key step (CH + N, = HCN + N)
implies that the rate of the prompt NO x sub-mechanism does not fall as rapidly with
temperature as does the thermal rate. A recent upward revision of the activation
energy from - 14 to - 22 kcal/rnol (Dean et al., 1990) is unlikely to affect this
conclusion in a significant way.
The nitrous oxide mechanism is analogous to the thermal mechanism but is
important at low temperatures (Malte and Pratt, 1974). It involves the production of
N,O as an intermediate
N,

N,O

M,

and its subsequent conversion to NO. Another possible outcome is reduction back to
N,. N,O itself does not survive as a significant emission; Muzio et al. (1990) have
found that measurements of N,O in combustion, reported previously, were artifacts
of the sampling technique. Sensitivity analysis (Correa and Srnooke, 1990; Drake et al.,
1990) indicates that the nitrous oxide sub-mechanism is significant in lean (e.g.,
</J = 0.6) premixed laminar flames.

333

NO, FORMATION
Reaction Zone

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

Fuel

Air

(a)

FIGURE 2

Air

(b)

Limiting cases of turbulence-chemistry interaction. (a) Flamelet. (b) Distributed reaction

zone.

Turbulence-Chemistry Interactions
A hypothesis regarding the nature of turbulent combustion is needed not only for
modeling, but also to assist in the interpretation of data. For example, are turbulent
flames better viewed as ensembles of strained laminar "flarnelets" (Peters, 1986)
or as broader "distributed" zones of reacting species (Bilger, 1988)? These extreme
possibilities are illustrated in Figure 2. In the flamelet model (Peters, )986), all
chemistry would occur in a region of well-characterized strain since by assumption all
the reactions are fast enough to occur close enough to some interface. Here "close
enough" means within a Kolmogorov scale, or the corresponding scalar scale. It is
assumed here that the Kolmogorov and Batchelor scales are approximately equal,
confining the discussion to gas-phase systems (Schmidt number Sc - I); additional
possibilities arise in the case that Sc is not of order unity, as is the case in water
(Sc ~ 600; Koochesfahani et al., 1985). By contrast, reactions may be slow enough
to occur over a broader "distributed" zone: the term "distributed zone" implies a zone
which has sufficient spatial extent to permit internal turbulence structure, without a
unique value or perhaps even a monotonic variation of dissipation rate (Bilger, 1988).
A chemical reaction may be characterized in terms of a length scale-the width of
the reactive zone in mixture fraction space being transformed appropriately to physical space-or a time scale. The multiple chemical reactions involved in combustion
introduce a wide range of Darnkohler numbers (Da., == OJr) into the flow and pose
a challenge that has not yet been met by theory; here 0; is a time scale associated with
mixing process "i" and r j is a time scale associated with chemical reaction "j". To
quantify the Da., for a practical combustion system, selected characteristic kinetic
times were obtained from a stirred reactor calculation for 10atm. methane-air
combustion (Correa, 1989). The conditions-I 0 ms residence time, methane at 300 K,
air at 610 K, and equivalence ratio t/J = 0.7-are representative ofaveraged "head-end"
combustor conditions in a 10w-NOx premixed methane-fueled gas turbine. The kinetic
scheme was taken from Glarborg et al. (1986), excluding Cy-chemistry and fuel-bound
nitrogen but including the NOx mechansims given above. Concentrations, density and
temperature from the stirred reactor solution were used to estimate characteristic
times of selected reactions. Characteristic chemical time scales were estimated as

334

S. M. CORREA
TABLE I
Selected chemical time scales in 10atm combustor

Reaction

fl"

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

I. H + 0, = OH +
3. H + OH + M = H,O + M
5. CH, + M = CH, + H + M
7. CH) + OH = CH, + H,O
9. CH + N, = HCN + N

2
7
5
2
3

Reaction

10- 1

x 10-'
x 10-'
X 10-'
X 10-'

2.
4.
6.
8.
10.

r,

0+ H, = OH + H
N, +
= NO + N
CH, + OH = CH, + H,O
CO + OH = CO, + H
CH, + N, = HCN + NH

2 x
4 x
2 X
2 x
0.20

10-'
10-'
10- 7
10-'

r, = 1./(k[X]) where "k" is the reaction rate and [Xl is the concentration of the more
abundant reactant, with an additional concentration factor for three-body reactions.
Time scales for selected reactions are shown in Table r.
Given the wide disparity in reaction time scales, it is clear that the same turbulence
field could appear as fast mixing (Da., --> 0; stirred reactor limit) or as slow mixing
(Da'j ~ I; equilibrium flamesheet limit) depending on the reaction in question. Thus
approaches which classify turbulent flames based on a single chemical time scale are
not comprehensive enough. Existing theories cannot predict NO, from first principles
except in certai n cases.
To improve predictive capabilities in the regime of interest, more data are needed
on reactions that occur at intermediate Damkohler numbers. Since different NO,
sub-mechanisms can be dominant depending on the flame temperature and pressure,
NO, data can be used to fill in the Damkohler number regime addressed by traditional
laboratory experiments. The following list presents typical experiments on turbulent
mixing of passive and/or reactive scalars:
I. Constant-density non-reacting mixing flows: Images of conserved scalars in
non-reacting turbulent flow, resolved down to the mixing scales, are available
for the high Schmidt number case of water (Planar laser-induced fluorescence
(PLIF); Dahm and Buch, 1989), and also for gaseous jets where Sc is of order
unity (Rayleigh scattering; Bilger et al., 1990). Both show that mixing consists
of strained diffusive layers embedded in broader regions of lower strain rate.
2. Constant-density fast-chemistry flows: Spatial resolution becomes an important
and potentially limiting issue in direct imaging techniques when the finest
turbulence scales are of interest. Another technique is to use the product of a
known chemical reaction as an indicator of the net effect of the turbulencechemistry process; it is suggested below that NO, data can play this role in
real-world systems where other diagnostics are not available. The technique is
particularly useful in the limit of fast chemistry, i.e., "mixed is reacted". For
example, Koochesfahani et al. (1985) studied the interaction between turbulence
and a very fast acid-base reaction in a turbulent mixing layer in water. The
profound influence of the mixing field on net reaction rate has been demonstrated
by Bilger (in Libby and Williams, Eds., 1980), in the limit of fast chemistry
between non-premixed reactants. Under fast chemistry, the species mass
fractions are given uniquely by the mixture fraction ~ (normalized elemental
mass fractions and enthalpy),

y,

Y,'(~);

i =

(5)

I, NS.

Substituting this relation into the fundamental transport equation for y"
=

o (pD OY)
OX~ +

oX

Wi'

(6)

335

NO, FORMATION

and with the homogeneous transport equation for


instantaneous reaction rate is obtained as

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

W;

iJ~ )2 d' y,'


-pD ( iJx
de'
k

and the chain rule, the

(7)

Under fast chemistry, d 2 Y,'Ide is a delta function at the stoichiometric mixture


fraction ~" which selects the scalar dissipation at the stoichiometric isopleth as
being critical. This elegant theory decouples the turbulence-chemistry interaction into an a priori determination of the laminar flame structure over a range
of strain rates, and a determination of the joint pdf of mixture fraction and
dissipation rate in the turbulent flow (Peters, 1986).
3. Atmospheric pressure flames: The effects of non-constant density, order-ofunity Schmidt number, and finite-rate chemistry are among the phenomena
introduced by gas-phase combustion. Finite-rate chemistry invalidates the
hypothesis Eq. (5) since the composition is not determined by stoichiometry
alone. Reactions will occur at a finite rate and in a region not necessarily
confined to the vicinity of the stoichiometric isopleth. Pulsed laser Raman
spectroscopy provides time-resolved but usually not quite space-resolved information in laboratory combustion experiments. Bimodality in the temperaturemixture fraction scattergrams is commonly used to deduce the mode of
turbulence-chemistry interaction; however, lack of probe resolution (scattering
volume) down to the Kolmogorov/Batchelor scale, the development of chemical
metastable states on the rich side in hydrocarbon "flamelets" (Bilger, 1988),
and the possibility of re-ignition following relaxation of the strain (which
occurs on a time scale 'K) make it difficult to draw conclusions.
Laboratory research on NO, in turbulent flames is often conducted in atmospheric turbulent jet "diffusion" (non-premixed) flames. NO, production in
these flames is so dominated by the thermal mechanism that the range of
turbulence-chemistry interactions revealed by a NO, measurement becomes
limited. Moreover, diffusion flames yield so much NO,_ that they will see limited
use in the future; turbulence-chemistry interactions in lean premixed combustion
are of increasing interest.
Planar imaging has contributed information such as planar laser-induced
fluorescence images of OH in laminar, transitional and turbulent H 2 flames.
The OH existed in discrete "islands", which were not connected in the plane of
the image and extended spatially over the order of the integral scales of
turbulence (Kychakoff et al., 1984). Such data apparently refute flamelet ideas,
but are not available at a scale which resolves the turbulence.
Needless to say, none of these techniques is used in a routine way in the
machine environment, where the mode of turbulence-chemistry interaction can
be quite different from that in the laboratory.
4. High pressure flames: Pressure has a substantial effect on reaction rates, and so
on the relevant Damkohler numbers. The available database on NO, formation
covers pressures from I to 40 atm., inlet temperatures from 300 K to 900 K, and
both premixed and non-premixed flames. As will be shown, NO, formation
sweeps across the range of possible modes of turbulence-chemistry interaction.
These flames are not influenced by buoyancy, so the effects of Froude number
cannot be deduced; work in non-premixed turbulent jet flames (Peters and
Donnerhack, 1980; Turns and Lovett, 1989) has addressed scaling with
buoyancy.

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

336

FIGURE 3

S. M. CORREA

150MW class stationary gas-turbine. Combustion hardware has been removed; the combustor

cans would lie between the last compressor stage and the three turbine stages at the back.

ISSUES IN GAS-TURBINE COMBUSTION

Gas turbines may be grouped in two categories, one used for baseload power generation
and the other for aeropropulsion. This section first reviews the principal performance
characteristics required of each category, and then the common issues.
Stationary Gas-Turbines
Stationary gas-turbines range in size from under 10MW to more than 150MW, the
latter being designed for baseload power generation (Figure 3). Combustors may be
of annular, can (the cans are generally in an annular plenum, hence the name
"cannular" is often used) or off-board ("silo") design (Figure 4). There are fourteen
to eighteen combustor cans in a 150 MW-class cannular machine; each has an airflow
of about 25 kg/s and is on the order of I m in length and 30em in diameter (Figure 5).
The can residence time is typically about 20 ms. The exhaust from the gas turbine is
often coupled to a heat-recovery steam-generator and thence to a steam-turbine,
yielding about 250 MW in total and capable of continuous operation for one year or
more at over 50% thermal efficiency. The high efficiency may be viewed as a savings
in both fuel and in CO 2 emissions, although the latter are not regulated as yet.
(Regulation of NO x in terms of NO x per unit of energy produced, rather than ppm
in the exhaust gas, would automatically introduce a CO 2 standard.) Gas-turbine
powerplants do not require much on-site construction and so can be brought on line
far more rapidly than competing powerplants, which contributes to their popularity.
Gas-turbines fired on natural gas are in particularly great demand at present.
Stationary gas-turbine combustors traditionally operated in the non-premixed

337

NO, FORMATION

ANNULAR
SILO

c
T

CANNULAR

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

T
C = Compressor

c
FIGURE 4

T = Turbine

Illustration of annular, cannular and silo combustors. Flowpaths are indicated conceptually.

mode. The resulting stoichiometric interfaces in the turbulent flame lead to high NO,
(about 200 ppm NO, at 15% 0,), but such combustors have wide stability limits. To
reduce peak temperatures and thermal NO x formation rates, steam or water is often
injected into the combustor. Figure 6 shows typical NO, levels as a function of the
rate of water injection; the data were obtained in an aeroderivative machine at full
power, but are typical of other machines as well. The approach requires a source of
high-pressure steam or purified water, constituting a significant capital item and a
significant loss of cycle efficiency. The large steam/water flow rate required for
adequate NO, control can also have the undesirable side effects of quenching CO
burnout (Figure 7), possibly necessitating a CO catalyst in the exhaust, ofcompromising
flame stability and of leading to mechanical corrosion within the combustor. Despite
these difficulties, steam or water injection is used to suppress NO, in some applications.
Optimization generally focuses on minimizing the steam or water flux for a given
reduction in NO" and requires an understanding of mixing processes in the flame.
To further reduce NO, to the levels required in some regions, ammonia-based deNO,
is implemented in the exhaust. This process has the disadvantages of adding significant
capital and operating expense, as well as possibly emitting ammonia. Yet another
approach is rich-lean combustion in which a fuel-rich primary zone is followed by a
lean zone to produce the correct turbine inlet temperature while, in principle, avoiding
stoichiometric temperatures. Staged combustion fails the ultra low-NO, application
because it is limited by the rate at which hot rich gas can be mixed with the rest of
the air to lean conditions. Even a 10-100 ps residence time in a stoichiometric
turbulent eddy formed during mixing could produce unacceptably high levels of NO,.
Rich-lean staged combustion has been of interest for fuels with high fuel-boundnitrogen content in which case the NO x goals are not as ambitious; the rich zone
converts much of the bound nitrogen to N, rather than NO,.
Designers have chosen lean premixed combustion to meet NO x objectives (Davis
and Washam, 1988). Leaning the flame does not reduce the thermodynamic efficiency
of present gas-turbine cycles. This is because the original (near stoichiometric) flame
would have been diluted by addition of air in the latter part of the combustor to
produce the appropriate combustor exit equivalence ratio (~0.4) and temperature, a
prime determinant of the efficiency. (A potential source of confusion is the fact that

S. M. CORREA

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

338

FIGURE 5 Combustor cans from 150MW-class stationary gas-turbine; dilution holes are visible.

combustor exit temperatures exceed turbine rotor inlet temperatures by an amount


corresponding to the enthalpy absorbed by the air used for cooling the first stage
turbine nozzle; this difference can be as much as 100 K. It is sufficient here, however,
to equate combustor exit temperature with cycle efficiency.). Thus the flame temperature may be lowered, .by using the air for premixing rather than dilution, for NO x

339

NOx FORMATION

200,-------,------,--------.---------,

150
ON
;f.
~

1ii 100
E
0.
0.

0"

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

50

O'-----------'------L--------'-

0.5

1.5

Water/Fuel Mass Flow Ratio

FIGURE 6

No.-.: as a function of water addition in a gas-turbine combustor running on natural gas at

a pressure ratio of 30.

control without affecting the combustor exit temperature. At some point in the future,
if turbine materials improve such that combustor exit temperatures themselves rise
into the regime of significant NO x formation, further improvements in NO x might
perhaps be gained only at the expense of thermodynamic efficiency.
Aeropropulsion Gas-Turbines
Aeropropulsion gas-turbines have some unique constraints and requirements. A
cutaway of a modern turbofan engine used for widebody transports (Figure 8a) shows
the annular combustor. Such an engine has a fan airflow of about 800 kgfs and a
bypass ratio of about five, which means that about 150kgfs of air pass through the
200

150
ON
;f.

'"
1ii
E

100

0.
0.

0"

50 -

0
0

100

200
CO, ppm at 15% 0,

300

FIGURE 7 Concomitant increase in CO emissions corresponding to Figure 6.

400

S. M. CORREA

340

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

FAN

HOT SIDE
Panel Film Cooling Air

FIGURE 8 Large modern turbofan aeroengine. (a) Cutaway showing location of annular combustor
within engine; multiple fuel injection swirl cups are visible. (b) Details of combustor wall showing dilution
holes and film-cooling slots.

combustor. The ~ I galls kerosene burning rate (at takeoff power) and ~ 15cm
combustor length indicate the intensity of the combustion process. A few of the
twenty or so fuel injection "swirl cups" are visible. At high power (takeoff and climb,
at least) spray atomization and vaporization occur very rapidly. Combustor residence
times are typically 2-3 ms, about an order of magnitude less than in the baseload
power generation machines. The combustor in Figure 8a resides within a fully annular
plenum which is fed by high-pressure high-temperature compressor discharge air.
Figure 8b provides details of a typical combustor "liner", i.e., the wall between the
combustor and the plenum, showing film-cooling slots and dilution holes. A typical
average liner temperature is about 1100 K, but the overall combustion process is
essentially adiabatic since all the air used for cooling enters the combustor as shown
in Figure 8b.
(i) Altitude Relight: A key non-emissions requirement of aeropropulsion combustors
is relight at high altitude in the event of combustor flameout. Spray atomization and
vaporization, and ignition chemistry, are compromised by the low (absolute) pressure

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO, FORMATION

341

and temperature of compressor-discharge air at stratospheric altitudes. All these


complex phenomena become important, which is why relight cannot be predicted
reliably. Relight capability is improved in practice by providing a region oflower axial
velocity within the combustor, leading to double-annular designs. In these, one
annulus furnishes a pilot flame for the other.
(ii) Emissions: It is difficult to attain 50-100 ppm NO x levels in operational aeropropulsion combustors (non-premixed), since they operate at higher pressures
(20-40 atm VS. 10-15 atm) with attendant higher combustor inlet temperatures than
their stationary counterparts. Non-premixed flames with preheated air produce NO x
copiously, because of the contributions from stoichiometric and near-stoichiometric
zones (Figure I). Radiation and its coupling with turbulence may also playa significant
role in thermal NO, under these conditions.
There have been many efforts to develop lean premixed combustion for aeroengines
(e.g., Ekstedt and Fear, 1987). The fuel for aeroengines is likely to be kerosene-based
for the foreseeable future. Semerjian and Vranos (1976) showed that Jet A fuel
(kerosene) could be prevaporized and burned in a premixed manner over a perforated
plate, yielding low NO, as expected. The inlet temperature was 750 K, but the pressure
was atmospheric. Autoignition is of great concern given that the fuel is in liquid form;
autoignition delay times are on the order of 1-10 ms at the compressor discharge
pressures and temperatures of interest. This concern can limit the time available for
premixing.
Because space and weight limitations inhibit premixing, and because of the hazards
of autoignition and flashback, premixed combustors are not operational in aeroengines
at present. Lean premixed combustion is, however, being considered for future
machines such as the High Speed Civil Transport (HSCT). With conventional technology the HSCT would produce on the order of 1000 ppm NO" at altitudes injurious
to the ozone layer. The target for NO, emissions is at present 3-8 g NO,/kg of fuel
(about 50 ppm, depending on the engine cycle). Another option is rich-lean staged
combustion. Although rejected for the ultra low-NO, application, staged combustion
might meet HSCT goals. Development of this technology will require, among other
items, much improvement in mixing control between stages and in the turbulence
models embodied in design codes. Suppression of soot in the rich zone will also be
critical.
(iii) Derivatives: Some aeroengines are re-packaged with a power turbine to provide
shaft power for intermittent-duty industrial or marine use. Most of these machines
employ non-premixed combustors and so typically inject steam or water for NO,
suppression (Figures 6 and 7). The areas of these technologies which need improvement
have been discussed. Lean premixed combustor designs are becoming available for
these aero-derivatives.
Common Problems
The following issues are among those common for both types of gas-turbine combustor.
(i) Pattern Factor: The combustor exit temperature is determined by the overall
fuel-air ratio and, as mentioned above, is a key determinant of the thermodynamic
efficiency of the machine. Combustor exit temperatures are lower in baseload stationary
gas-turbines-less than 1700 K at present-for reasons of longevity; operation for one
year or more is usually required. Aeropropulsion combustors must generally produce
hotter gas.
The above addresses only the bulk-averaged combustor exit temperature. The
exit-plane radial profile must be tailored by appropriate placement of dilution holes

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

342

S. M. CORREA

FIGURE 9 Combustor exit temperature distribution over the full annulus in a typical aeroengine.
Temperatures are in of; these data are from a low-power test which accounts for the low temperatures.

and management of in-combustor aerodynamics to ensure that the turbine blade tips
(which experience lower centripetal stress) run hotter than the hubs, for longevity. The
"pattern factor" is a measure of this profile. Residence times are adequate-given the
high pressures of modern combustors-for the combustor exit temperature profiles
to correspond to chemical equilibrium. Thus the problem becomes one of understanding the post-flame mixing of hot products with dilution jets and film-cooling air.
Figure 9 shows combustor exit temperature contours over the full annulus in a typical
aeroengine.
Limitations of turbulence models and 3D numerics make it a somewhat cut-and-try
endeavor to optimize pattern factor, although 3D fast chemistry/assumed-shape pdf
modeling is making an impact (Correa and Shyy, 1987). Heat transfer to the combustor liner (wall) is also a critical issue, since it dictates the amount of cooling flow
needed to maintain the liner within allowed temperatures ( ~ I J00 K). The complicated
plenum-liner cooling flow distributions are often computed by simple one-dimensional

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO, FORMATION

343

models. Radiation plays a non-negligible role in heat transfer from the flame to the
liner.
(ii) Higher temperature-rise combustors will be required as turbine materials
improve. Future aeropropulsion combustors may operate at an overall stoichiometry
increasing towards unity, favoring performance over emissions for non-civilian
purposes. The higher exit temperatures will mean that there is less air available
downstream of the flame to complete combustion of rich pockets which may include
soot, to tailor the combustor exit temperature profiles, and to cool the combustor
liner. Techniques for efficient mixing in the flame zone, soot control, and advanced
cooling with limited air will be needed.
The impact of higher turbine inlet temperatures on NO, has already been discussed.
(iii) Turndown and dynamics: Turndown describes the ability to reduce fuel flow
(power output) without compromising flame stability, and is very influential in
combustor design. Stationary machines used for power generation have the special
constraint that the turbomachinery must run at constant speed, being connected to
a generator whose frequency must not vary with load. The airflow through the system
is therefore constant as load varies, and so a reduction in fuel flow is tantamount to
a reduction in equivalence ratio. Throttling at the inlet to the machine offers only a
partial solution to this lean-limit problem. In contrast, the shaft speed (air flow) may
vary with load in an aeroengine, which alleviates the problem.
In a non-premixed combustor, turning down the fuel flow for part-load operation
will result in a smaller but stable flame. A lean premixed combustor, however, may
already be operating near its lean limit. Reduced fuel flow can be accommodated by
techniques such as (a) designing the combustor to have a number of independent
burners, and then shutting off the fuel to some of these ("fuel-staging"), (b) redirecting
some of the premixing air to the latter part of the combustor to maintain equivalence
ratio in the flame zone above the blowoff limit ("air-staging"), (c) shifting into a
non-premixed mode, at the expense of emissions. Techniques which require moving
parts, particularly in the hot sections, are less attractive in practice.
Many interesting problems are provided by the unsteady aspects of bluff-body
stabilized premixed flames. Different dynamic regimes appear as the mixture is leaned
to blowoff. Consider first the near-stoichiometric case, which is also of interest in
afterburners for military jet engines; in the laboratory the frequency of the fluctuations
often corresponds to a quarter longitudinal wave upstream of the flame-holder.
One-dimensional models which relate the unsteady heat release to the unsteady
velocity fluctuation through an empirical chemical induction time are quite successful
in predicting the longitudinal modes. Strouhal shedding may be excited and may
amplify the pressure fluctuations (Ghoniem, 1986). In practice, much higher frequencies
associated with radial or circumferential modes may be dominant. Near the lean
blowofflimit, lower frequencies associated with bulk (Helmholtz) modes may dominate.
These characteristics have been found to depend heavily on the flameholder and
combustor geometries. Dynamic over-pressures produced during power transients or
unsteady interactions between flow and heat release may significantly exceed the
pressure drop engineered across flameholders, causing flashback. Such effects can be
minimized by prevention of large-scale organized motion in the flow, for example, by
having multiple fuel injectors and by adding three-dimensionality to otherwise
axisymmetric flameholders. Dynamics in cannular systems can also lead to unsteady
can-to-can coupling and so modulate the air flow (to each can), the equivalence ratios,
and therefore the emissions.
NO, formation in different modes of continuous combustion is discussed next.

344

S. M. CORREA
10'

."'
E

00-

oi

10

1ii

a:
c

.g

'"

E
(;

u,

10'

z
<ii

'"
f-

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

.c

"E

lO'

'~"

u,

10.6 I
0.020

0.030

0.040

0.050

0.060

0.070

0.080

MixtureFraction

FIGURE 10 Forward thermal NO formation rate in mixture fraction space, based on equilibrium
temperature and O-atoms; CH 4 -air flame at pressure of I aim and initial temperature of 300 K.

NON-PREMIXED TURBULENT COMBUSTION

Since non-premixed flames involve stoichiometric interfaces and attendant high


temperatures, the thermal mechanism dominates NO x formation. Much more NO x is
formed in the vicinity of this interface than on the rich and lean sides where the other
sub-mechanisms contribute, limiting the utility of NO, as a diagnostic of turbulencechemistry interactions. Moreover, emission levels are too high for low-NO, applications
such as power generation. On the other hand, the very extensive database of turbulent
jet diffusion flames makes them worth examining. These flames also exhibit equilibrium
broadening, superequilibrium broadening and radiation effects on NO" as discussed
next.
The high activation energy of thermal NO, suggests that NO, would be formed in
thin flamelet-Iike regions of stoichiometric gas. Studies of turbulent jet diffusion
flames, however, indicate otherwise. First, equilibrium dissociation itself permits
significant radical levels and NO formation rates at non-unity stoichiometry. The
width in mixture fraction space of significant thermal NO, formation depends on the
temperature and pressure. Figure I0 repeats, in mixture fraction space, the thermal
NO formation rate for a 1atm methane-air flame at 300 K inlet temperature. The zone
of significant NO, formation rate (greater than I ppm/ms) is approximately 0.03 units
wide in mixture fraction space.
Can zones of this width be characterized as "flamelets", or do they contain
significant turbulence structure? Bilger et al. (1990) answer such questions by comparing
the width of the reaction zone with the scalar scale L~, the scale at which inhomogeneities in the scalar field are mixed away by diffusion. This scale is associated with the
regions of greatest gradient in the scalar field. The numerical value of stoichiometric
mixture fraction (~,) is also important. For example, ~, is approximately 0.056 for
CH, but is in the range 0.3-0.5 for the CO/H 2/N2 blends widely studied (Drake et al.,

345

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO, FORMATION

1987;Correa and Gulati, 1988),and is also large (0.2) in the CH./N, flames of Barlow
et al. (1990). The undiluted hydrocarbon flames have narrow reaction zones in
mixture fraction space.
Returning to the relevance of flamelets to NO" estimates following Bilger et al.
(1990) are made: assuming a turbulence intensity of 2m/s and an integral scale of
5 mm, the scalar scale follows E, "" 0.3~';~' is the intensity of scalar fluctuations and
is in the range 0.01 to 0.1 at xld = 25 in the 40% CO/30% H,/30% N, jet flame
(~, = 0.3). The scalar scale is therefore in the range 0.03 to 0.003, at most comparable
to the width of the NOx-forming zone. One would not expect a flamelet theory to be
applicable.
Non-equilibrium effects also broaden the NOx-forming zone in physical space, as
shown by extensive studies of the above turbulent 40% CO/30% H,/30% N, jet
diffusion flame in co-flowing air (Drake et al., 1987). This effort resulted in a database
on mean NO x and the pdf's of major species and temperature (by Raman scattering),
OH (by laser-induced fluorescence), and axial velocity and turbulence (LOY) at
various locations in the flame. A partial equilibrium/assumed shape pdf model
(Correa et al., 1984) was used to predict temperature, major and minor species, and
thermal NO x in the above flame. Partial equilibrium in the oxyhydrogen radical pool
leads to relationships such as [0) = K[H,)[O,)/[H,O), where K is the product of the
equilibrium constants associated with the elementary steps, and allows for superequilibrium levels of radicals because of the finite rate of radical recombination (mole
reducing) reactions. The degree of superequilibrium increases when the fluid mechanical
time scales are decreased (e.g., in more intense turbulence) and when the recombination
time scales are increased (e.g., when the pressure is decreased). Superequilibrium
radicals had to be accounted for to predict temperature, OH and NO, in agreement
with the data. The greatest discrepancies with data were in the cool core of the jet
where partial equilibrium in the radical pool is not expected to hold. Errors due to the
k-e model and assumed shape pdf/moment equation closure were fairly insignificant,
as shown by an independent joint pdf transport/Monte Carlo calculation of the same
flame; the two calculations agreed to within a factor of two on sensitive quantities
such as mean and rms of the major and minor species (Pope and Correa, 1986;Correa
et al., 1988).
Superequilibrium broadens the NO x forming zone, as may be shown in terms of a
reaction progress variable for the oxyhydrogen radical pool. A linear combination of
species mass fractions Yit, is first defined
Yit, =

I M H2

YH ,

+ "2 M O H

YO H

M H,
M Yo
o

3 M H,

+ "2

YH

M H,
M
Yeo
eo

(8)

The choice of coefficients ensures that Yit, is affected only by recombination reactions,
and not by the two-body shuffle reactions which are taken to be in partial equilibrium.
Equation (8) leads to the definition of the reaction progress variable rt
_ Yit, -

Yit,u

rt = y*e _ y*u
H2

(9)

Hz

The progress variable n varies between frozen (rt = 0) and equilibrium (rt = I)
values. Although superequilibrium radical levels (rt < I) lowered the mean flame
temperature by as much as 200 K, the degree of superequilibrium is so large (factors
of 100-1000 above equilibrium levels in off-stoichiometric gas) that the thermal NO x
rate is greater than if the radical pool was in equilibrium at that stoichiometry

346

S. M. CORREA

T(K)
2.0

2250

2000

3
1.0

::.

o
?;z

1750

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

>-

1500

0.0

0
0.7

0.8

1.0

0.9

Reaction Progress Variable '1


FIGURE II Effect of superequilibrium on a-atoms, temperature and thermal NO formation rate at
stoichiometric conditions for a 40% CO, 30% H" 30% N, fuel in air (from Drake et al., (987).

(Figures II and 12). The partial equilibrium model predicted superequilibrium OH


levels in good agreement with the laser-induced fluorescence data; O-atom levels were
predicted but not measured. The model also predicted changes in NO, effected by
changing the diluent in the CO/H 2/N2 fuel from N 2 to CO 2 or Argon; besides the effect
on adiabatic flame temperature, these species have different third-body efficiencies for
radical recombination and so affect the degree of superequilibrium. Substitution of
CO 2 for N 2 caused a factor offourdecrease in measured NO" in close agreement with
the prediction. With Argon, the model predicted a small increase in NO" while the
2.0 r - - - - r - - - , - - - r - - - - . - - r - - - - ,

e-

Ol

,JI
(; 1.0

io

~ = 0.35

s:

....

.................

~=0.25

-."

.....
"

,, ".
,,

0.0 L-_----'_ _---.J_ _----'_ _---'-_ _--'--_ _- '


0.7

0.8
0.9
Reaction Progress Variable ~

1.0

FIGURE 12 Composite effect of superequilibrium on thermal NO formation rate in lean, stoichiometric


and rich gas; the fuel is the same as in Figure II (from Drake et al., 1987).

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO, FORMATION

347

data indicated a substantial decrease. Myhr and Turns (1991), however, have repeated
the experiments and find a small increase agreeing with the predictions for Argon.
Thus there is broad support for the concept of radical superequilibrium and its effect
on thermal NO,. Superequilibrium broadening further upsets the notion of a flameletlike topology even for thermal NO, in a diffusion flame.
At progressively higher (superatmospheric) pressures the three-body recombination
reactions speed up and drive the system towards equilibrium within the time scales of
consequence, reducing the influence of superequilibrium radicals.
Superequilibrium was introduced in the context of the partial equilibrium model
only for convenience in modeling. Drake and Blint's (1989) calculations of a laminar
counterflow diffusion flame indicate that partial equilibrium is a simplistic assumption;
however, superequilibrium and the effect of broadening the NO, forming zone are
supported by detailed kinetics, and partial equilibrium is attained at high temperatures. (It may also be noted that the partial equilibrium sub-model is tractable in
multi-scalar pdf computations whereas full kinetics would require the assumption of
a f1amelet topology.)
Barlow et al. (1990) have reported joint Raman-OH data in atmospheric piloted
turbulent CH.-N, and CH. -air jet diffusion flames. Broad zones of superequilibrium
OH were found to exist and the concentrations were not in accordance with predictions
of flamelet theory. Partial equilibrium in the oxyhydrogen radical pool suggests a
similar broad structure for 0 atoms, and that superequilibrium broadening of thermal
NO, formation zones is significant in hydrocarbon flames as well.
Earlier studies of NO, formation in turbulent jet diffusion flames of H, have
provided two widely-debated apparent consequences of turbulence-chemistry interactions. Measurements indicated a shift in the peak NO, production rates towards the
rich zone and a dependence on the jet exit Reynolds number, Re- 1/2 (Kent and Bilger,
1972; Bilger and Beck, 1974). These results are important because of the implications
regarding how and where NO, is formed; however, it should be noted that these early
data have not been reproduced in subsequent studies.
Drake et al. (1987) could not reproduce the "rich shift". They found that NO, was
maximized in flame zones which were, in the mean, closest to stoichiometric with the
highest temperature and the largest radical concentration. They attributed the earlier
data to errors in the sampling technique. A rationale for the "rich shift" not being
physically realistic is that although superequilibrium radicals increase the thermal
NO, formation rate in rich gas, this increase is not enough to equal the rate in the
stoichiometric gas. As mentioned above, these superequilibrium effects are important
in atmospheric flames. Peters and Donnerhack (1980) tried to explain the rich shift
by convoluting the equilibrium thermal NO x formation rate with given {J-function
pdf's for the mixture fraction. Pdf's with successively increasing variance (fluctuations)
were used and were found to preferentially increase the NO, formation rate on the
rich side. Elements of this argument are plausible, but again it is handicapped by the
assumption of equilibrium radicals. In any event, the mechanism of enhancement due
to fluctuations is intrinsically contained in the partial equilibrium/assumed shape
(jJ-function) pdf model.
The apparent dependence on jet exit Reynolds number is also being re-examined.
Although viscous effects in high Re laminar flow are confined to thin regions (boundary
layers or internal layers) whose thicknesses scale as v-II', one must be sure that the
NOx dependence is on Re, and not on the jet exit velocity. (It is the jet exit velocity
which is actually varied in experiments, changing Reynolds numbers, Damkohler
numbers and radiation characteristics simultaneously.) A Reynolds number dependence implies an interaction between thermal NO, chemistry and the turbulent strain

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

348

S. M. CORREA

field. Since the time-averaged/z-s/assumed shape pdf model does not identify NO,
formation with flamelet structure, and neglects molecular transport in comparison
with turbulent transport, it does not predict a Reynolds number dependence (Drake
et al., 1987). Bilger (1976) developed a theory of superequilibrium O-atoms caused by
microscale turbulence, resulting in the rich shift and the Re- 1/' scaling. The model of
Broadwell and Mungal (1988) for mixing layers also produces a dependence on
Reynolds number. The model postulates that turbulent mixing and reaction commence with entrainment into large-scale structures, allowing formation of thin
stretched interfacial regions or "flarnelets". Within the core of the large-scale structure and once the dimensions have been reduced to near the Kolmogorov scale,
molecular diffusion rapidly homogenizes the fluid within the structure. Reaction
then is presumed to occur in the manner of a well-stirred reactor. The empiricallydetermined split between the strained flamelet and stirred reactor modes introduces
the dependence on the Reynolds number.
More recently, it has been suggested that radiation losses may explain the apparent
Re-scaling of NO, (Turns and Lovett, 1989). Radiant fractions in propylene flames
were found to decrease with increasing velocity (Delichatsios et al., 1988), so that
higher velocities led to more nearly adiabatic flames. As shown above, the effects of
even small perturbations in temperature are significant in the thermal NO, regime.
Turns and Lovett (1989) found that simple radiation estimates applied to their
propane jet flame led to an appropriate scaling with jet exit velocity, which could be
perceived as a dependence on Reynolds number. Subsequent experiments produced
a Re- O.41 dependence of NO" and calculations of temperature decrements due to
radiation from stoichiometric gas endorsed the idea that radiation could account for
the observed scaling (Turns and Myhr, 1990). Peters and Donnerhack (1980) found
a similar NO, scaling (Re- 0 66 instead of Re- 1/' ) in turbulent jet diffusion flames of H,
CH 4 and C 3 H s. Soot is, of course, absent from the H, flames but radiation may still
playa role. Further experiments would be helpful in assessing a radiation-based
explanation of the Re-scaling in the general case. A calculation formulated in terms
of the joint pdf for mixture fraction (elemental composition only), enthalpy and
radical pool reaction progress variable would also be instructive; unknowns in soot
formation and transport could be eliminated by studying non-sooting flames, as long
as radiation remained significant to enthalpy transport.
It remains unclear whether there is a scaling of NO, with Reynolds number (as
opposed to velocity) and, if so, what the origin is. It is also not clear whether
observations in turbulent jet flames are relevant to practical recirculation-stabilized
combustion. In turbulent jet diffusion flames, combustion occurs in the outer layer
where strain rates are maximum. This structure is not universal. Bluff-body stabilized
flames provide a different and more complex turbulence field, and should receive more
attention. For example, in practical recirculation-stabilized combustors, NO, reduction
by steam injection depends more heavily on the relative flow rates of steam and fuel
than on the relative locations of the injection points. Intuitive attempts to optimize
injection systems-for example, by shrouding the fuel jets in an annular steam flow
-are often unsuccessful. These results can be rationalized if the thermal NO, formation
zones are broadened and/or the turbulent mixing is so rapid as to mix reactants and
products to a significant extent. Scaling laws based on turbulent jet flows are not
applicable, even though fuel injectors appear to be jet-like.

349

NO, FORMATION

TABLE II
Characteristic turbulence time scales
u'

Laboratory flame
Turbine combustor

E(m'ls 3 )

(m/s)
5
10

I
5.0

0,005
0.005

X
X

10- 3
10-'

2.5

10'

2.0 x 10'

OK (s)

Re,

1.4 X 10-'
1.8 x 10-'

-50
-800

Note that the laboratory flame is assumed to have a very high turbulence intensity (20%), attainable

for LDI purposes. A corresponding turbulence intensity for the turbine combustor would require an
unreasonably high pressure drop, leading to a perhaps unacceptable reduction in cycle efficiency.

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

LEAN DIRECT INJECTION

Is it possible to introduce fuel and air separately into a combustion chamber in overall
lean proportions and mix them rapidly enough to achieve the NO, levels typical of
turbulent premixed combustion? Ifproduct formation in this process of simultaneous
mixing and reaction cannot be distinguished from that in a process of premixing
followed by reaction, it must be that reaction is slow relative to the mixing process
(Da., ~ I). Again, this provides a clue to flame structure. The question also addresses
the viability of lean direct injection (LOI) combustion, a potential low-NO, alternative
to lean premixed combustion without the problems of flashback and autoignition.
The turbulence intensities required for LOI may be estimated from the thermal
NO, rate and simple turbulence relationships, The forward thermal NO, formation
rate is taken from Figure I for inlet conditions of (i) I atm, 300 K and (ii) 10atm,
600 K. The equilibrium flame temperature goes up with pressure, since radical
recombination is favored at higher pressure and is exothermic, and with inlet temperature (albeit by an amount less than the inlet temperature increase, due to enhanced
dissociation at the higher temperature). As shown in Figure I, there is a speed up of
a factor of about 40 between the maximum NOx-forming rates at laboratory and at
typical utility-class gas-turbine combustor conditions. More intense turbulence would
be required to make LOI work at elevated conditions than at atmospheric conditions,
The viability of LOI is assessed in each case by computing the limits on Damkohler
numbers, Turbulence scales are estimated as follows. The integral time scale (8,) is
(10)
where u' is the Lm.S. velocity fluctuation and A, is the integral length scale. From the
kinematic viscosity v and the mechanical dissipation rate e,
e

cu"/)",

(\ 1)

where c is a constant of order unity, the Kolmogorov time scale 8K follows

8K

(v/e)I/2.

(12)

With c == I, these relations also imply that the Reynolds number of the turbulence
(U'A,/V) is the square of the ratio (8,(8d.
Estimates are made for the two cases assuming the following data (i) laboratory
flame: mean velocity (u) = 25 ta], u' = 5 mis, A, = 0,005 m, and (ii) turbine
combustor: mean velocity (u) = 100mis, u' = 10m/s, A, = 0.005 m. The numerical
value of kinematic viscosity is appropriate for the burned gases, yielding a lower
turbulence Reynolds number (Rei) than that computed for the unburned gases,
Turbulence time scales are shown in Table II.

350

S. M. CORREA

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

TABLE 1II
Turbulent Damkohler numbers
Laboratory flame
I atm, 300 K inlet

Turbine Combustor
10atm, 600 K inlet

0.05
0.007

1.05
0.038

The characteristic chemical time scale r is taken as the time to form 100ppm of
NO" using the maximum rates shown in Figure I. The turbulent Damkohler numbers
Da, == ed". and Da, == eK !,,, the latter also being an inverse Karlovitz number,
quantitatively compare the turbulence and chemical scales. Da ~ I indicates a wellmixed mode for the particular reaction and mixing process used to define Da, while
Da ~ I indicates an unmixed mode (which is to be avoided for thermal NO x reactions).
Damkohler numbers computed for the two cases (Table HI) indicate that NOx
formation in the laboratory flame lies in the well-mixed regime, while that in the
turbine combustor lies between the well-mixed regime (at the Kolmogorov scale) and
the intermediate Da ~ I regime (at the integral scale).
The relevance of the Kolmogorov scale in these estimates is worth discussion.
Planar imaging of mixture fraction in non-reacting gaseous jets has shown that the
zones of intense scalar dissipation form what Bilger et al. (1990) have called "patches",
surfaces which are on the order of the integral length scale in two dimensions and on
the order of the Kolmogorov scale in the third direction (perpendicular to the
"patch"). The high wave number part of the mixing process is experienced by only a
small part of the flow at any given instant. Bilger (1988) has claimed that the effective
mixing length scale is about ten times larger than the Kolmogorov length scale.
Assuming that this conclusion applies to turbulence in the presence of strongly
exothermic reactions, the estimates in Tables H and III should be considered to be
bounding values. If both bounding values of Damkohler number are either small or
large compared with unity, however, the exercise is still instructive. On this basis it
may be concluded that, in a machine and regardless of the (non-premixed) injection
scheme, most of the time NO, formation in LDI would not lie in the well-stirred
regime. (At an overall lean equivalence ratio, however, the temperature is much lower,
"is much larger and therefore the formation of NO, occurs in a well-mixed manner.)
The conclusions of Table III find support in experiment. Stattelmayer et al. (1990)
discuss a partially-premixed injector, operated in a near-adiabatic mode at atmospheric
pressure. Data on LDI combustion of propane (Figure 15 in Stattelmayer et al., 1990)
indicate that the NO x levels are comparable to those in a premixed flame, in agreement
with column I of Table HI above. On the other hand, the large degree of turndown
achieved is evidence that heat release reactions lay in the partially-premixed or
non-premixed regime; a diffusion flame has no lower limit on equivalence ratio in the
overall sense. Aigner et al. (1990) focus on high pressure tests of the above partiallypremixed injector, at pressures up to 16atm and inlet temperatures up to 700 K. Of
particular interest are their observations that:
i. NO, emissions scaled strongly with pressure, approximately as pO.". It will be
shown below that this strong dependence on pressure is a symptom of NO x production
in near-stoichiometric gas.
ii. NO, emissions were higher than those in a premixed flame at the same overall
equivalence ratio.

NO, FORMATION

351

Ill.
The combustor could be operated well below the lean limit of the equivalent
premixed flame, e.g., at rf> = 0.2 overall.
These observations indicate that the combustor was not operating in a well-stirred
mode, with respect to NO, formation, in accordance with column 2 of Table III.

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

5 PREMIXED COMBUSTION
Unlike the situation in non-premixed flames, stoichiometric interfaces cannot form in
lean premixed flames and so the associated high temperatures do not occur. Since the
rate of formation of thermal NO, is very sensitive to temperature, this component of
NO, is minimized and can be virtually eliminated from playing a role. Laminar and
turbulent flames have been studied over the pressure range of interest.
Correa and Smooke (1990) used the Miller-Bowman (1989) scheme in a NavierStokes simulation of one-dimensional freely propagating flames, described by the
steady-state form of Eqs. (6) plus the energy equation. The range 0.6 '" rf> '" 1.0,
I '" P (atm) '" 10 was covered. 43 species and 154 reactions including Cy-chemistry
were employed, and the numerical solution was processed to obtain coefficients
describing the sensitivity of any species to any reaction. The numerical Jacobian
matrix assembled in course of the Newton method provides coefficients for the firstorder sensitivity of species "j" to reaction "i", After normalization by the largest
value of mole fraction of "j",
(13)
these coefficients (Jji may be read as the fractional response in mole fraction Aj, to unit
increase in the forward rate of reaction "r", kf,i. Note that the reverse rate would also
be increased since it is constrained by the thermodynamic equilibrium constant, and
that species "l" may not occur in the reaction "i", A positive (negative) sensitivity
implies that the reaction contributes to production (consumption) of species ''j''.
Only the sensitivity coefficients with the largest absolute maxima are presented. The
actual numerical value of the sensitivity coefficient depends on the normalization in
Eq. (13).
Typical species mass fractions and NO-sensitivity profiles are shown in Figures 13
and 14 for atmospheric stoichiometric and lean flames, respectively. Figure 15 shows
axial profiles of the pressure sensitivity coefficient. The sensitivity profiles show that
(i) thermal NO, is the major contributor in stoichiometric atmospheric pressure
flames, and there is a dependence of NO, on pressure, (it) N 2 + 0 + M = N 20 + M
is the major nitrogen fixing mechanism in lean flames; Drake et al. (1990) report a
similar result in lean (sub-1800 K) high-pressure laminar premixed ethane flames; and
(iiI) there is no dependence on pressure in lean flames; Heberling (1976) measured a
reduced dependence of prompt NO, on pressure in lean laminar premixed flames, over
the pressure range 1-18 atm.
Sensitivity to the prompt NO, mechanism was never among the top five sensitivities
for NO. N02 was found to exist in the initial part of the flame due to reaction between
NO and H0 2 present at low temperature, and consequently is short-lived. This N0 2
was sensitive to the prompt mechanism.
The dwell time in a lean high-pressure laminar flamefront (flame velocity 3-5 cm/s)
is 10-30 rns, much greater than the residence time in the stabilization zone of a
gas-turbine combustor. Since radical levels and NO, formation chemistry would be

352

S. M. CORREA

,1"---

,
,,,
,
,,

--7---------0

,,
I

: ,.' 1',

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO,':

f\

,! !' \"
:"

Iid':, i
f'

:
," \:I

10"

\.

CH

\.

.
\
0.05
0.10

0.00

0.15

0.20

0.25

0.30

Axial Distance, x (em)

o
z

a:

oLL

r-

[/)

i=

Is

i=

5>

(jj
Z

:lClao,

l.U
[/)

l.U

>--

<l:

a:

z
o

l.U

a:

::;:
::>

63-_ _ _ 63 _ _ _ 6 3 - - - .3~

~ ~O;-W---~07---I07
- - - 9---29
\ 6 - - - 1 6 - - - 1 6 - - - I.

~CII--II_-II

~<l:

(b)

N + NO = N2 +
H + 02 = OH +0
CO + OH = CO 2 + H
HCO + M = H + CO + M
29. CH 2 + 02 = CO 2 + H + H

l.U

107.
63.
16.
11.

-,

::;:
X
<l:
::;:
-2

0.0

0.1

0.2

0.3

0.4

0.5

AXIAL DISTANCE, x (em)


FIGURE 13 Numerical results for atmospheric pressure q, = I laminar premixed methane-air flame.

(a) Selected species mass fractions. (b) Sensitivity coefficients defined by Eq. (13).

different, the relevance of lean premixed laminar flames to their turbulent counterparts in not clear. At stoichiometric conditions, and if heat release reactions are of
primary importance, this conclusion could change.
Leonard and Correa (1990) studied NO, formation in premixed turbulent highpressure methane-air flames stabilized above an uncooled perforated plate made of

353

NO, FORMATION
10"

/---1---------- _

10.3
I

,"

10"

NO

1O, s

u,

10'

'"
'"
::;;

10.7

I
I
I

I
I
'
I "

<Il

e:

.~

..

,.'

,:

//

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

10'
10.9

/ I"

r,
I

0.05

/\

r-.

1 . . . __ - - - - -

-.-

.~

'--CH

j \.
\: \
~!\ i\

i \.\

0.10

NO,

\.

._.-.-.-.-.L.-.-.-.-
0.15

0.20

0.25

Axial Distance. x (em)


1.0 ,---~-,_-~-,_-~-,_-~-,__-~-__,
I
I

o
z

a:

u,

(/)

0.5 l-

t-

s
;::

117.
125.
103.
65.
118.

N,O+M = N,+O+M
HCN + OH = HOCN + H
NH+NO = N,O+H
H + 0, + M = HO, + M
N,O+H = N,+OH

(fj
Z

00

}---IIl~Z5.5- .5--- ..'-:,....

--125-_-1257"
Wi
,\f.l!

<C

,~~

a:

II'):

;::
U
...:
~ -0.5
::;;
::::l
::;;

-----117

117

...:

::;;

FIGURE 14

AXIAL DISTANCE. x (em)


Numerical results for atmospheric pressure t/> = 0.6 laminar premixed methane-air flame.

(a) Selected species mass fractions. (b) Sensitivity coefficients defined by Eq. (13).

alumina (Figure 16). Conditions varied from I to 10atm, inlet air temperatures from
300 K to 610 K, and equivalence ratios from 0.5 to 0.9. The data were compared with
a model consisting of a stirred reactor for the back-stirred region followed by a plug
flow reactor, using the Glarborg et al. (1986) scheme without Cj-species. Figure 17
compares measured and predicted NOx levels at 10.4atm. for three different initial

354

S. M. CORREA
1.0

_ _- ,_ _~,_,

,-~--,--_----,-_~

....

z
w
l3

u:
U.

0 5

....>-

i=

0.0

= 0.6

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

Ci5
z
w

en
w
a:
::>
en

f3 -0.5

a:
a.

-1. 0

O:--~-_L-_---'----:':--~-_...L_-----'-_...J

AXIAL DISTANCE, x (em)

FIGURE 15 Pressure sensitivity coefficients in the stoichiometric and lean atmospheric pressure laminar

premixed methane-air flames.

temperatures. NO, increases from ~ I ppm levels as the equivalence ratio increases.
Several features were evident from the data and calculations. The low temperatures
characteristic of lean flames preclude significant formation of NO" by eliminating the
thermal mechanism. The prompt and nitrous oxide mechanisms, however, continue
to function at lower temperatures and so are dominant below about 1800K. As
mentioned above, the key prompt NO, reaction CH + N 2 = HCN + N has a
relatively low activation energy compared with N 2 + 0 = NO + N, the key thermal
NO, reaction. CH levels are, however, very low. The NO, levels are much lower than
in the thermal regime, and there is little or no growth of total NO, with distance as
long as thermal NO x remains insignificant (flame is lean).
It is not clear why predictions of the stirred-plug flow reactor model agree with the
above data as well as shown. As indicated in Table I, not all the reactions would be
in the "distributed reaction zone" regime. In particular, CH + N 2 = HCN + N has
a very short time scale and so is flamelet-like. Corr (1990) shows that pyrolysis might
occur in a non-well-stirred manner even in a jet-stirred reactor. Not well-stirred,
however, does not automatically indicate fiamclet-like behavior. Schefer et al. (\ 990)
directly imaged CH, OH and CH. in a lifted turbulent jet diffusion flame of methane
and found that OH and CH zones were many times greater than the expected laminar
flame thickness. One may conclude heuristically that the overall chemical process of
NO x formation in premixed flames occurs in a well-enough distributed manner for a
stirred-plug flow reactor description to work. It is unfortunate that better models are
not available.
The results also indicate a diminishing influence of pressure of NO x in lean premixed

355

NO, FORMATION

(al

Combustion Air
Orifice Meter

Choked) (
Orifice

,....--11---

Sampling

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

Probe

Cooling
Air

Burner

" ' - - Pressure


Vessel

(b)

Quartz Liner

Alumina Flame Holder


Stainless Steel
Flame Arrestor

Premixed Fuel
and Air

FIGURE 16 Apparatus used for high-pressure lean premixed combustion studies. (a) Layout of facility.
(b) Perforated-plate burner.

356

S. M. CORREA

40

"if.
';2

310K

530K

615K

30

1ij

a.
a.
O

20

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

10

0.5

0.7

0.6

0.8

Equivalence Ratio

FIGURE 17 Measured and predicted NO, levels at 10.4atm for three different initial temperatures.

combustion. The behavior may be expressed as


(NOx){(NO, at I atm.)

= p",

(14)

where "p" is the pressure (in atm.) and "m" is an unknown exponent. It should be
noted that since NO, levels are low in lean flames, "m" is sensitive to errors in the
measurements. Figure 18 shows NO x data from the perforated plate burner at three
different pressures; the mass flow is proportional to the pressure, so that the bulk
residence time remains constant (to within the effects of pressure on temperature)
between the runs. The dependence on pressure diminishes at the lower equivalence
ratios (m = 0 within experimental error), as in the laminar flame (Heberling, 1976;
Correa and Smooke, 1990).
The exponent "m" is larger than zero in the richer cases in Figure 18. In a related
issue, NO, in turbulent diffusion flames (e.g., computations in Drake et al., 1987) and
practical combustors (engineering data) exhibits an approximate scaling as
(m = 1{2). At high enough equivalence ratios (T > 1900K) the NO, formation rate
becomes predominantly thermal in origin and is limited by the abundance of'O atoms.
atoms scale approximately as
The absolute concentrations of equilibrium
which would explain the observed behavior if all other parameters were constant. In
richer combustion systems, however, pressure will affect temperature. For example,
a I to 10atm change in a stoichiometric methane-air system results in a te')!Perature
increase of about 40 K. One may therefore have expected a greater than ...;p scaling.
The fact that this is not so in a wide range of combustion experience, to the point that
is accepted "design practice" in non-premixed combustion, may be because the
effect of higher temperature at higher pressure is offset by the effect of superequilibrium atoms at lower pressure. Further experiments and calculations are needed

JP

JP

JP,

357

NO x FORMATION

,.'
""' ...

3.1Atm

... 6.0Alm

....

~
1ii
E
a.
a.
O

'

10.3Atm

" I,'

,~"" """
, '~

20

~l,

,,I.
, "

.....y '/

~' ..,~'

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

30

,.

,,
," ,
,, ,, ...

10

......
0.65

..

.... ,

II
II
1/
//

LY

--.:.-

-_:.:.::.::-0.70

0.75

0.80

0.85

Equivalence Ratio

FIGURE 18 Measured and predicted NO x levels at an initial temperature of 322 K for three different
pressures.

to clarify these speculations on the role of pressure, and implications regarding the
mode of NO, formation in different flames.
Altemark and Knauber (1987) measured NO x emissions from a 300 kW turbulent
lean premixed gas flame, in the pressure range 4-10 atm. Their results confirm that
sub- 10ppm levels of NO, are possible, and that the NO, emissions are independent
of pressure in premixed flames at equivalence ratios of 0.6 or less.
Decoupling of NO Production from the Turbulent Flowfield
There is direct evidence that NO formation in turbulent premixed combustion can be
rendered slow enough, by leaning the premixture, to decouple from the turbulence.
Figure 19 compares NO measurements made in turbulent lean premixed methane-air
flames (Gulati, 1990), with flameholders of three different types: a V-gutter, a
perforated plate, and a 60 swirler. The velocity and temperature of the approach flow
were held constant at 10 mls and 560 K, respectively. Based on fuel flow, each
combustion process was on the order of 100kW in power; the walls were not adiabatic
but conditions were the same for each flameholder. One-dimensional plug-flow
residence times to the probe were 25-30ms, and dilution air was not added downstream of the flame. The recirculation zones responsible for flame stabilization were
quite dissimilar in each case, but the data are remarkably consistent. NO formation
is evidently not dependent on the fine structure of turbulence in this lean premixed
regime.
6 CONCLUSIONS
Gas turbines are principally used for aeropropulsion-with some spin-off to landbased power generation and marine propulsion-and for continuous utility-class

358

S. M. CORREA

15

f0-

<E.-

v-qutter. 70% blockage

000
000
000
000

Perforated plate.77% blockage

Single cupswirler

ow

a:
a:
Q

::J

10

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

6
z
E

c.
c.

0'-'--'--.J....L..:.:l-L...L...L-L...L-l-.L...L-l-.L...L-l-.L...L-L.l-l.-L.l-l......L..L.l-LJ
0.40

0.45

0.50

0.55

0.60

0.65

0.70

EQUIVALENCE RATIO

FIGURE 19 Measured NO emissions with various flameholders, The methane-air premixture approaches
the flameholder at I atm, 560 K and 10 m/s.

baseload power generation, where advantages include short delivery times and high
efficiency in the combined cycle mode. Aeropropulsion combustors have special
constraints such as size and altitude relight. Parameters such as pattern factor,
turndown, liner heat transfer and life, and so on, are important in both types
of combustors. These traditional concerns will increasingly be complicated by
requirements on emissions. NO" in particular, has led to a significant re-examination
of combustion technology.
Lean premixed combustion is probably the only combustion technology which will
provide sub IO-ppm NO., levels, at least in the immediate future. Catalytic combustion
is not yet a viable technology for baseload power generation, in part because of the
short life of catalyst materials under machine conditions and duty cycles. Pulsed
combustion for NO, has not been attempted in systems of this size.
Substantial progress has been made over the last decade in identifying and quantifying the elementary chemical kinetic mechanisms by which NO x is produced in
methane-air flames. Complex kinetic schemes have been advanced, accounting for the
details of methane oxidation including NO x formation by the thermal, prompt and
nitrous oxide mechanisms. These sub-mechanisms have well-defined reaction rates,
which differ enough to allow interaction with turbulence at a variety of scales.
Temperature and stoichiometry can be selected to make different ones of the submechanisms dominate, e.g., thermal NO, is generally dominant in diffusion flames.
The data reviewed here contain the following observations and scaling laws:
I. NO, production via the thermal mechanism dominates turbulent diffusion
flames. The fact that the thermal contribution diminishes with strain in laminar
diffusion flames may not be relevant given the lack of a flarnelet-like topology

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO, FORMATION

359

in turbulent flames. Zones that form NO x by the thermal mechanism at a rate


greater than I ppm/ms, even with equilibrium radicals, are not thin compared
with turbulence scales. Superequilibrium broadening of NO, forming zones
cannot be ignored at pressures below 10atm. The rich shift of NO x injet flames
appears to have been an artifact of the sampling technique. The origin of the
Re- I /2 scaling is still under study, but is due in part at least to radiation effects.
In any case NO, in the intensely turbulent recirculation-stabilized combustors
of practical interest does not appear to follow fuel jet-based scaling laws and
is not strongly dependent on the injection geometry. NO, scales approximately
as JP in non-premixed combustors in which the overall residence time is held
constant (mass flow must vary linearly with pressure).
2. Imaging of reaction zones down to the dissipative scales would help to assess
models for turbulent combustion. Current data are of insufficient resolution.
NO, formation can be a diagnostic of the mode of turbulence-chemistry
interaction. Data on NO, in a wide range of relevant Damkohler numbers are
provided by CO/H 2 , methane and natural gas combustion under laboratory
and elevated conditions. It is apparent that while some reactions fall in the
flamelet regime and some in the distributed regime, many important reactions
fall in between. The simultaneous prediction of kinetically sensitive phenomena
such as pollutants, flame stability and combustion efficiency, will require new
theories.
3. Turbulent mixing of fuel and air to overall lean conditions within the combustion
chamber itself, so as to produce NO, levels approaching those of premixed
combustion, is possible under non-preheated conditions. This shows that the
interaction of thermal NO, chemistry with turbulence does not necessarily
occur in an interfacial manner even in non-premixed systems, if the turbulence
is intense enough. The accelerated NO, formation rates that result from
combustion at high inlet temperatures suggest that lean direct injection for NO,
control may not be realizable in machines.
4. In lean (enough) laminar premixed flames, NO, is independent of pressure and
is formed predominantly by the nitrous oxide pathway. Comparisons with
stirred reactor data and calculations-particularly the identification of differences if any between laminar flames and stirred reactors-e-could provide further
clues to the mode of NO, formation in turbulent premixed flames.
5. In lean (enough) turbulent premixed combustion, NO, is once again independent
of pressure. Stirred reactor simulations indicate that again the nitrous oxide
pathway plays a role. NO, is independent of the fine structure of the turbulent
flowfield, suggesting that turbulence-NO, chemistry interactions in these
lean systems occur in the distributed reaction zone mode. Combinations of
stoichiometry and inlet temperature (related to the pressure ratio of the
machine) which result in flame temperatures below 1800K will meet a sub10ppm NO, (at 15% 0,) goal with some margin to spare. Predictions based on
stirred-plug flow reactors are quite useful for NO, in this lean premixed regime.
This review has indicated that lean premixed combustion is the most likely means
to sub-10ppm NO, in practical machines. Secondary problems must, however, be
pointed out. Large amounts of CO are produced in the stabilization zone of a lean
turbulent premixed flame, and must be oxidized in the turbulent post-flame gas. CO
oxidation occurs via CO + OH = CO, + H, with time scales of the right order to
couple strongly with (be quenched by) post-flame turbulence. Unlike NO" CO can
therefore depend on the flameholder shape and the turbulent post-flame flowfield.
Fluid mechanical simplifications such as that of stirred reactor theory are not useful;

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

360

S. M. CORREA

theoretical models that have been found to be useful are generally based on simplified
chemistry and a pdf description of the scalar turbulence (Chleboun et al., 1987).
Theories will have to be improved, however, before CO can be predicted with confidence in three-dimensional premixed turbulent flames. Rapid-quenching probes to
sample non-equilibrium CO in high-temperature gas, or their non-intrusive equivalents robust enough for the machine environment, have also been found to be a
critical item in the development of lean premixed gas-turbine combustors. Emissions
of UHC are more complicated kinetically, but generally track those of CO. Dynamics
and flame stability are also strongly dependent on geometry, making them difficult to
discuss in general terms and often translating into expensive cut-and-try development.
Modeling of turbulent combustion in realistic three-dimensional geometries is not
yet advanced enough to be a reliable tool for the range of issues presented above. The
prevalent k-e/assumed shape pdf closure model is used for pattern factor (Correa and
Shyy, 1987) and for thermal NO, in non-premixed systems (Priddin and Coupland,
1986), but must be improved upon or replaced before the other quantities mentioned
in Section 2 can be usefully predicted. An alternative is the Monte-Carlo/pdf
approach (Pope, 1990); although well proven for fully-developed shear flows, this
method needs to be adapted to pressure-dominated flow in complex geometries.
Suitable algorithms for mixing are also under active development. Furthermore,
accounting for emissions will require computationally tractable simplifications of the
chemistry (e.g., the three- or four-step mechanisms under development, Paczko et al.,
1986) and suitable turbulence-chemistry interaction models. Unsteady models
(Ghoniem, 1986) are needed to account for dynamics and stability in premixed
combustion. It is unlikely that a single comprehensive code will be developed but,
even with restricted regimes of applicability, models can make an impact.
ACKNOWLEDGEMENTS
This work was supported in part by the Air Force Officeof ScientificResearch, Contract F49620-88-C-0066,
Dr. Julian Tishkoff, Program Manager, and the Gas Research Institute, Contract 5089-260-1912, Dr.
Robert Gemmer, Program Manager. The author has also benefited greatly from discussions with colleagues
inside and outside General Electric Company.

REFERENCES
Aigner, M., Mayer, A., Schissel, P., and Strittmatter, W. (1990). Second-Generation Low-Emission
Combustors for ABB Gas Turbines: Tests Under Full-Engine Conditions. Paper 90-GT-308, ASME
Gas Turbine Conference, Brussels, Belgium, June 11-14.
Altemark, D. and Knauber, R. (1987). Ergebnisse von Untersuchungen an einem Vormischbrenner unter
Druck mil extrem niedriger NO,-Emission. VDI Berichte Nr. 645, pp. 299-311.
Barlow, R.S., Dibble, R.W., Starner, S.H., Bilger, R.W., Fourgette, D.C., and Long, M.B. (1990). Reaction
Zone Structure in Dilute Methane Jet Flames Near Extinction. AIAA Paper 90-0732, 28'" Aerospace
Sciences Meeting, January 8-11, Reno, NV.
Bilger, R. W. (1976) Reaction Zone Thickness and Formation of Nitric Oxide in Turbulent Diffusion
Flames. Comb. and Flame 26, 115-123.
Bilger, R. W. (1980). Turbulent Flows with Non-premixed Reactants. In Turbulent Reacting Flows, (Libby,
P. A. and Williams, F. A., Eds.) Springer-Verlag, New York.
Bilger, R. W. (1988). The Structure of Turbulent Nonpremixed Flames. Twenty-Second Symposium
(International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp. 475-488.
Bilger, R. W. and Beck, R. E. (1974). Further experiments on Turbulent Jet Diffusion Flames. Fifteenth
Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp. 541-522.
Bilger, R. W., Yip, B., Long, M.B., and Masri, A.R. (1990). An Atlas ofQEDR Flame Structures. Comb.
Sci. and Tech. 72, 137-155.
Broadwell, J. E. and Mungal, M. G. (1988). Molecular Mixing and Chemical Reactions in Turbulent Shear

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

NO, FORMATION

361

Layers. Twenty-Second Symposium (International) on Combustion, The Combustion Institute,


Pittsburgh, PA, pp. 579-587.
Chleboun, P. V., Hubbert, K. P., and Sheppard, C G. W. (1987). Modelling of CO Oxidation in Dilution
Jet Flows. Paper 38, AGARD Conference Proceedings No. 422, Chania, Crete, October 19-23.
Corr, R. A. (1990). NO, Formation in Lean Premixed Hydrocarbon Combustion, M.S. Thesis, Mechanical
Engineering, University of Washington.
Correa, S. M. Flame Structure: A Discussion, Proceedings of NASA Langley/ICASE Combustion
Workshop, Oct. 2-4, 1989, Springer-Verlag, in press.
Correa, S. M., Drake, M. C, Pitz, R. W., and Shyy, W. (1984). Prediction and Measurement of a Nonequilibrium Turbulent Diffusion Flame. Twentieth Symposium (International) on Combustion, The
Combustion Institute, Pittsburgh, PA, pp. 337-343.
Correa, S. M. and Gulati, A. (1988). Non-premixed Turbulent CO/H, Flames at Local Extinction
Conditions. Twenty-Second Symposium (In/ernational) on Combustion, The Combustion Institute,
Pittsburgh, PA, pp. 599-606.
Correa, S. M., Gulati, A., and Pope, S. B. (1988). Assessment ofa Partial-Equilibrium Monte-Carlo Model
for Turbulent Syngas Flames. Comb. and Flame. 72,159-173.
Correa, S. M. and Shyy, W. (1987). Computational Models and Methods for Continuous Gaseous
Turbulent Combustion, Prog. Energy Comb. Sci., 13, 249-292.
Correa, S. M. and Smooke, M. D. (1990). NO, in Parametrically Varied Methane Flames. Twenty-Third
Symposium (International) on Combustion. The Combustion Institute, Pittsburgh, PA, pp. 289-295.
Davis, L. B. and Washam, R. M. (1988). Development ofa Dry Low NO, Combustor. ASME 89-GT-255,
Gas Turbine and Aeroengine Congress and Exposition, June 4-8, Toronto, Ontario, Canada.
Dahm, W. J. A. and Buch, K. A. (1989). High Resolution Three-Dimensional (256') Spatio-Temporal
Measurements of the Conserved Scalar Field in Turbulent Shears Flows, Seventh Symposium on
Turbulent Shear Flows, August 21-23, Stanford, CA.
Dean. A. J., Hanson, R. K., and Bowman, C T. (1990). High-Temperature Shock Tube Study of Reactions
of CH and C-atoms with N,. Twenty-Third Symposium (International) on Combustion. The
Combustion Institute, Pittsburgh, PA, pp. 259-265.
Delichatsios, M. A., Markstein, G. H., Orloff, L., and deRis, J. (1988). Turbulent Flow Characteristics and
Radiation from Gaseous Fuel Jets, GRI Report 88/0100.
Drake, M. C and Blint, R. J. (1989). Prompt Nitric Oxide Formation in Stretched Laminar Opposed Flow
Diffusion Flames. Comb. and Flame, 76, 151-167.
Drake, M. C, Correa, S. M., Pitz, R. W., Shyy. W., and Fenimore, C P. (1987). Superequilibrium and
Thermal Nitric Oxide Formation in Turbulent Diffusion Flames. Comb. and Flame, 69, 347-365.
Drake, M. C, Ratcliffe, J. W., Blint, R. J., Carter, C D., and Laurendeau, N. M. (1990). Measurements
and Modeling of Flamefront NO Formation and Superequilibrium Radical Concentrations in
Laminar High-Pressure Premixed Flames. Twenty-Third Symposium (International) on Combustion,
The Combustion Institute, Pittsburgh, PA, pp. 387-395.
Ekstedt, E. E. and Fear, J. S. (1987). Advanced Low Emissions Combustor Program, AIAA 23" Joint
Propulsion Conference, San Diego, California, June 29-July 2.
Fenimore, C P. (1970). Formation of Nitric Oxide in Premixed Hydrocarbon Flames. Thirteenth
Symposium (International) on Combustion, The Combustion Institute, Pittsburgh, PA, pp. 373-380.
Ghoniem, A. F. (1986). Effect of Large-Scale Structures on Turbulent Flame Propagation. Combust.
Flame. 64, 321-336.
Glarborg. P., Miller, J. A., and Kee, R. J. (1986). Kinetic Modeling and Sensitivity Analysis of Nitrogen
Oxide Formation in Well-Stirred Reactors, Comb. and Flame, 65, 177-202.
Gulati, A. (1990). unpublished work.
Heberling, P. V. (1976). Prompt NO Measurements at High Pressures. Sixteenth Symposium (International)
on Combustion. The Combustion Institute, Pittsburgh, PA, pp. 159-168.
Iverach, D., Basden, K. S., and Kirov, N. Y. (1972). Formation of nitric-oxide in fuel-lean and fuel-rich
flames, Fourteenth Symposium (International) on Combustion, The Combustion Institute, Pittsburgh,
PA, pp. 767-775.
Kent, J. H. and Bilger, R. W. (1972). Turbulent Diffusion Flames. Fourteenth Symposium (International)
on Combustion, The Combustion Institute, Pittsburgh, PA, pp. 615-625.
Koochesfahani, M. M., Dimotakis, P. E., and Broadwell, J. E. (1985). A "Flip" Experiment in a
Chemically Reacting Turbulent Mixing Layer. AIAAJ. 23, 8,1191-1194.
Kychakoff, G., Howe, R. D., Hanson, R. K., Drake, M. C, Pitz, R. W., Lapp, M., and Penney, C. M.
(1984). Visualization of Turbulent Flame Fronts with Laser-Induced Fluorescence, Science, 224,
382-384.
Leonard, G. L. and Correa, S. M. (1990). NO, Formation in Lean Premixed High-Pressure Methane
Flames, 2"' ASME Fossil Fuel Combustion Symposium, New Orleans, Louisiana, January 14-18,
1990; ASME/PD Vol. 30, Singh, S.N. Ed., pp 69-74.

362

S. M. CORREA

Multe. P. C. and Pratt, D. T. (1974). Measurement of Atomic Oxygen and Nitrogen Oxides in Jet-Stirred
Combustion. Fifteenth Symposium (lnternationalv on Combustion. The Combustion Institute,
Pittsburgh, PA, pp. 1061-1070.
Miller, J. A. and Bowman, C. T. (1989). Mechanism and Modeling of Nitrogen Chemistry in Combustion.
Prog. Energy Comb. Sci, 15, 287-388.
Muzio, L. J., Montgomery, T. A., Samuelsen, G. S., Kramlich, J. C., Lyon, R. K., and Kokkinos, A.
(1990). Formation and Measurement of N,O in Combustion Systems. Twenty-Third Symposium
(International) on Combustion. The Combustion Institute, Pittsburgh. PA, pp. 245-250.
Myhr, F. H. and Turns, S. R. (1991). Oxides of Nitrogen Emissions from Turbulenl Jet Flames: Effects
of Fuel Dilution and Partial Premixing, in preparation.

Paczko, G., Lefdal, P. M., and Peters, N. (1986). Reduced Reaction Schemes for Methane, Methanol and
Propane Flames. Twenty-First Symposium (International) on Combustion, The Combustion Institute,

Pittsburgh, PA, pp. 739-748.

Downloaded by [Stanford University Libraries] at 03:02 11 October 2012

Peters, N. (1986). Laminar Flamelet Concepts in Turbulent Combustion. Twenty-First Symposium


(Internouonah on Combustion, The Combustion Institute, Pittsburgh, PA, pp. 1231-1250.

Peters, N. and Donnerhack, S. (1980). Structure and Similarity of Nitric Oxide Production in Turbulent
Diffusion Flames, Eighteenth Symposium (International) on Combustion, The Combustion Institute,
Pittsburgh, PA, pp. 33-42.
Pope, S. B. (1990). Computations of Turbulent Combustion: Progress and Challenges. Twenty-Third
Symposium (Iruernationali on Combustion. The Combustion Institute, Pittsburgh. PA, pp. 591-612.
Pope, S. B. and Correa, S. M. (1986). Joint PDF Calculations of a Non-Equilibrium Turbulent Diffusion
Flame. Twenty-First Symposium (International) on Combustion. The Combustion Institute, Pittsburgh,
PA, pp, 1341-1348.
Priddin, C. H. and Coupland, J. (1986). Impact of Numerical Methods on Gas Turbine Combustor Design
and Development. ASME Winter Annual Meeting, Anaheim, CA, December 7-12.
Sano, T. (1985). NO, Formation in the Mixing Region of Hot Burned Gas with Cool Air. Comb. Sci. Tech.
38, pp. 129-144.
Schefer, R. W., Namazian, M., and Kelly, J. (1990). CH, OH and CH, Concentration Measurements in
a Lifted Turbulent-Jet Flame. Report SAND89-8806, SANDIA Livermore, CA, 1990; also, TwentyThird Symposium (International), on Combustion, The Combustion Institute, Pittsburgh, PA,
pp. 669-676.
Semerjian, H. and Vranos, A. (1976). NOx Formation in Premixed Turbulent Flames, Sixteenth Symposium
(International) on Combustion. The Combustion Institute, Pittsburgh, PA, pp. 169-179.
Stattelmayer, Th., Felchlin, M. P., Haurnann, J., and Styner, D. (1990). Second-Generation Low-Emission
Combustors for ABB Gas Turbines: Burner Development and Tests At Atmospheric Pressure, Paper
ASME 90-GT-162, ASME Gas Turbine Conference, Brussels, Belgium, June 11-14.
Turns, S. R. and Lovett, J. A. (1989). Measurements of Oxides of Nitrogen Emissions from Turbulent
Propane Jet Diffusion Flames. Comb. Sci. and Tech.. 66, 233-249.
Turns, S. R. and Myhr, F. H. (1990). Oxides of Nitrogen Emissions from Hydrocarbon Jet Flames: Fuel
Effects and Flame Radiation, Fall Technical Meeting, Eastern States Section of The Combustion
Institute. Orlando, FL. Dec. 3-5.

Вам также может понравиться