NIST GCR 09-927

Fire Safety Risks Associated With

Leaks in Hydrogen Systems
 NIST GCR 09-927

Fire Safety Risks Associated With

Leaks in Hydrogen Systems
Prepared for
National Institute of Standards and Technology
Building and Fire Research Laboratory
Gaithersburg, MD 20899-8202

By
Peter B. Sunderland
Department of Fire Protection Engineering
University of Maryland
College Park, MD 20742

Grant 60NANB5D1209

December 2009

U.S. DEPARTMENT OF COMMERCE

Gary Locke, Secretary

NATIONAL INSTITUTE OF STANDARDS AND TECHNOLOGY

Patrick D. Gallagher, Director
 Notice

This report was prepared for the Building and Fire Research Laboratory of
the National Institute of Standards and Technology under grant number
60NANB5D1209. The statements and conclusions contained in this report
are those of the authors and do not necessarily reflect the views of the
National Institute of Standards and Technology or the Building and Fire
Research Laboratory.
 Final Report

Fire Safety Risks Associated with Leaks in Hydrogen Systems

Grant number: 60NANB5D1209

Grant PI: P.B. Sunderland

Awarded by: National Institute of Standards and Technology

Building and Fire Research Laboratory
Gaithersburg, Maryland

Submitted by: P.B. Sunderland

Assistant Professor
Department of Fire Protection Engineering
University of Maryland
[email protected]

Submission date: November 18, 2009

Grant period: September 1, 2005 – August 31, 2009

Investigators: Peter B. Sunderland, University of Maryland (PI)

Richard L. Axelbaum, Washington University (Co-I)
Beei-Huan Chao, University of Hawaii (Co-I)

NIST Contact: Jiann Yang

1
1. Motivation
Hydrogen presents several unusual fire hazards, including high leak propensity, ease of ignition,
and invisible flames. This research concerns experiments, analysis, and computations to identify
the hazards of leaks in hydrogen systems that could result in combustion. The work seeks to
identify the types of hydrogen leaks that can support flames. A small leak in a hydrogen system
could ignite easily, support a flame that is difficult to detect, and lead to a catastrophic failure.

2. Objectives
This research seeks an improved understanding of hydrogen fire safety that will lead to hazard
reduction in hydrogen systems. Specific objectives include:
1. Measure limits of flaming (at ignition, quenching and blowoff) for hydrogen issuing from
circular and slot burners of various sizes.
2. Measure flame quenching limits of hydrogen leaks in plumbing components.
3. Examine material degradation arising from an impinging hydrogen diffusion flame.
4. Prepare analytical and models of spontaneous ignition.
5. Perform CFD analyses of the flames to complement the experiments, yield additional
physical insight, and allow the consideration of untested conditions.

3. Summary of Progress
• The investigator team has published in Combustion Theory and Modeling a manuscript
entitled A Theoretical Study of Spontaneous Ignition of Fuel Jets in an Oxidizing Ambient
with Emphasis on Hydrogen Jets. This paper is attached here in Appendix A.
• The investigator team has published in the International Journal of Hydrogen Energy a
manuscript entitled Limits for Hydrogen Leaks that Can Support Stable Flames. This paper is
attached here in Appendix A.
• To date 15 papers and posters on this work have been published at conferences.
• To date 17 oral presentations on this work have been presented. These have been in the U.S.,
Canada, U.K., and Germany. Six of these presentations have been by graduate students.
• To date one Ph.D. Dissertation and four M.S. Theses have been published on this work.
• An analysis was performed for the spontaneous ignition of a hydrogen (or other gaseous fuel)
jet emanating from a slot into an oxidizing ambient (e.g., air). A similarity solution of the
flow field was obtained. This was combined with the species and energy conservation
equations, which were solved using activation energy asymptotics. Limits of spontaneous
ignition were identified as functions of slot width, flowrate, and temperatures of the
hydrogen jet and ambient gas. Two scenarios are examined: a cool jet flowing into a hot
ambient and a hot jet flowing into a cool ambient. For both scenarios, ignition is favored with
an increase of either the ambient temperature or the hydrogen supply temperature. Moreover,
for the hot ambient scenario, a decrease in fuel Lewis number also promotes ignition. The
Lewis number of the oxidizer only has a weak effect on ignition. Because spontaneous
ignition is very sensitive to temperature, ignition is expected to occur near the edge of the jet
if the hydrogen is cooler than the ambient gas and near the centerline if the hydrogen is
hotter than the ambient gas.
• Quenching and blowoff limits of hydrogen diffusion flames on small burners were observed.
Four burner types, with diameters as small as 8 μm, were considered: pinhole burners,
curved-wall burners, tube burners, and leaky fittings. In terms of mass flow rate, hydrogen

2
 had a lower quenching limit and a higher blowoff limit than either methane or propane.
Hydrogen flames at their quenching limits were the weakest flames recorded to date, with
mass flow rates and heat release rates as low as 3.9 mg/s and 0.46 W. The quenching limit
for a hydrogen flame at a 6 mm leaky compression fitting was found to be 28 mg/s. This
limit was independent of supply pressure (up to 131 bar) and about an order of magnitude
lower than the corresponding limits for methane and propane.
• The quenching limit measurements from this project were incorporated into the new SAE
J2579 standard for hydrogen vehicles. Hydrogen vehicles for U.S. use will now be required
to demonstrate that localized leaks are smaller than our measured quenching limits. We have
requested that a similar provision be added to the new NFPA 2, Hydrogen Technologies
Code.
• Flames with heat release rates as low as 0.25 W have been observed and photographed.
These are believed to be the weakest flames ever observed. The work has applications to fire
safety and microcombustors.
• Materials degradation upon exposure to hydrogen diffusion flames has been observed.
Hydrogen flames were observed to be more corrosive to metals and silicon carbide fibers
than methane flames were.

4. Peer-Reviewed Journal Papers (see full text in Appendix A)

1. K.B. Lim, B.H. Chao, P.B. Sunderland, R.L. Axelbaum, A Theoretical Study of Spontaneous
Ignition of Fuel Jets in an Oxidizing Ambient with Emphasis on Hydrogen Jets, Combustion
Theory and Modeling, 12 (2008) 1179-1196.
2. M.S. Butler, C.W. Moran, P.B. Sunderland, R.L. Axelbaum, Limits for Hydrogen Leaks that
Can Support Stable Flames, International Journal of Hydrogen Energy 34 (2009) 5174-
5182.
3. V.R. Lecoustre, C.W. Moran, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Experimental
and Numerical Investigation of Extremely Weak Hydrogen Diffusion Flames, Combustion
Symposium, in preparation.

5. Conference Proceedings and Posters

1. N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Fire Hazards of Small Hydrogen
Leaks, poster, Bridging the Transition…Hydrogen Internal Combustion Engines Symposium,
weststart.org, San Diego (2006).
2. N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Fire Safety Risks Associated with
Leaks in Hydrogen Systems, poster, Annual Fire Conference, NIST, Gaithersburg (2006).
3. M.S. Butler, C.W. Moran, P.B. Sunderland, R.L. Axelbaum, Quenching Limits of Hydrogen
Diffusion Flames, Eastern States Section of the Combustion Institute, Charlottesville (2007)
8 pp.
4. K.B. Lim, B.H. Chao, P.B. Sunderland, R.L. Axelbaum An Asymptotic Analysis of
Spontaneous Ignition of Hydrogen Jets, 5th U.S. Combustion Meeting, San Diego, 11 pp.
(2007).
5. N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Fire Hazards of Small Hydrogen
Leaks, SAE World Congress, Detroit, SAE Paper 2007-01-0429, 4 pp. (2007).
6. N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Fire Hazards of Small Hydrogen
Leaks, The National Hydrogen Association Annual Conference, San Antonio (2007) 8 pp.

3
7. N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Quenching Limits and Materials
Degradation of Hydrogen Diffusion Flames, 5th U.S. Combustion Meeting, San Diego, 8 pp.
(2007).
8. M.S. Butler, R.L. Axelbaum, C.W. Moran, P.B. Sunderland, Flame Quenching Limits of
Hydrogen Leaks, SAE World Congress, Detroit, SAE Paper 2008-01-0726 (2008) 8 pp.
9. M.S. Butler, C.W. Moran, P.B. Sunderland, R.L. Axelbaum, Fire Safety of Hydrogen Leaks,
Second International Energy 2030 Conference, Abu Dhabi (2008) 10 pp.
10. M.S. Butler, C.W. Moran, P.B. Sunderland, R.L. Axelbaum, Fire Hazards of Small Leaks in
Hydrogen Systems, Poster, International Association of Fire Safety Science, Karlsruhe
(2008).
11. C.W. Moran, M.S. Butler, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Observations of
Normal Gravity Flames that are Weaker than Microgravity Flame Balls, Poster, 32nd
International Symposium on Combustion, Montreal (2008).
12. P.B. Sunderland, Fire Hazards of Small Hydrogen Leaks, 3rd European Summer School on
Hydrogen Safety, Ulster (2008) 34 pp.
13. P.B. Sunderland, Pressure Relief Devices for Hydrogen Vehicles, 3rd European Summer
School on Hydrogen Safety, Ulster (2008) 20 pp.
14. V.R. Lecoustre, C.W. Moran, P.B. Sunderland, B.H. Chao, R.L. Axelbaum, Experimental
and Numerical Investigation of Extremely Weak Hydrogen Diffusion Flames, 6th U.S.
National Combustion Meeting, Ann Arbor (2009) 12 pp.
15. P.B. Sunderland, American Perspectives and Regulations for Hydrogen Vehicles, Progress in
Hydrogen Safety Short Course, Ulster (2009) 20 pp.
16. P.B. Sunderland, Hydrogen Flame Quenching Limits, Extinction, and Materials
Degradation, Progress in Hydrogen Safety Short Course, Ulster (2009) 33 pp.

6. Oral Presentations
1. N.R. Morton, Quenching Limits and Materials Degradation of Hydrogen Diffusion Flames,
5th U.S. Combustion Meeting, San Diego, March 27, 2007.
2. K.B. Lim, An Asymptotic Analysis of Spontaneous Ignition of Hydrogen Jets, 5th U.S.
Combustion Meeting, San Diego, March 28, 2007.
3. P.B. Sunderland, Quenching, Blowoff, and Spontaneous Ignition Limits of Hydrogen Flames
and Jets, Oral Presentation, NIST Annual Fire Conference, Gaithersburg, April 4, 2007.
4. P.B. Sunderland, Fire Hazards of Small Hydrogen Leaks, SAE World Congress, Detroit,
April 16, 2007.
5. M.S. Butler, Quenching Limits of Hydrogen Diffusion Flames, Eastern States Section of the
Combustion Institute, Charlottesville. October 24, 2007.
6. P. B. Sunderland, Quenching Limits and Materials Degradation of Flaming Hydrogen Leaks,
Department of Mechanical Engineering, University of Texas at Austin, March 18, 2008.
7. P. B. Sunderland, Fire Hazards of Hydrogen Leaks, Austin / San Antonio SFPE Chapter, San
Marcos, Texas, March 19, 2008.
8. P.B. Sunderland, Fire Hazards of Hydrogen Leaks, National Institute of Standards and
Technology Annual Fire Conference, Gaithersburg, March 31, 2008.
9. M.S. Butler, Flame Quenching Limits of Hydrogen Leaks, SAE World Congress, Detroit,
April 15, 2008.
10. P.B. Sunderland, Pressure Relief Devices for Hydrogen Vehicles, 3rd European Summer
School on Hydrogen Safety, Belfast, July 23, 2008.

4
11. P.B. Sunderland, Fire Hazards of Small Hydrogen Leaks, 3rd European Summer School on
Hydrogen Safety, Belfast, July 23, 2008.
12. P.B. Sunderland, Fire Hazards of Vehicles Powered by Alternative Fuels Including
Hydrogen, CNG, and Electricity, DC Section of SAE, U.S. Department of Transportation,
Washington, Oct. 23, 2008.
13. P.B. Sunderland, Fire Hazards of Small Hydrogen Leaks, Johns Hopkins University, Dept. of
Mechanical Engineering, Baltimore, Oct. 24, 2008.
14. V.R. Lecoustre, Experimental and Numerical Investigation of Hydrogen Flame Quenching
Limits, National Institute of Standards and Technology Annual Fire Conference,
Gaithersburg, April 29, 2009.
15. P.B. Sunderland, American Perspectives and Regulations for Hydrogen Vehicles, Progress in
Hydrogen Safety Short Course, Belfast, April 30, 2009.
16. P.B. Sunderland, Hydrogen Flame Quenching Limits, Extinction, and Materials
Degradation, Progress in Hydrogen Safety Short Course, Belfast, May 1, 2009.
17. V.R. Lecoustre, Experimental and Numerical Investigation of Extremely Weak Hydrogen
Diffusion Flames, 6th U.S. National Combustion Meeting, Ann Arbor, May 18, 2009.

7. Theses and Dissertations

1. Michael S. Butler, Flame Quenching and Materials Degradation of Hydrogen Leaks, M.S.
Thesis, Department of Mechanical and Aerospace Engineering, Washington University in St.
Louis, 50 pp., May, 2008.
2. Vivien R. Lecoustre, Numerical Investigations of Gaseous Spherical Diffusion Flames, Ph.D.
Dissertation, UMD Department of Mechanical Engineering, anticipated December, 2009.
3. Kian Boon Lim, An Asymptotic Analysis of Spontaneous Ignition of Hydrogen Jets, M.S.
Thesis, UMD Department of Fire Protection Engineering, 64 pp., May, 2007.
4. Christopher W. Moran, Flame Quenching Limits of Hydrogen Leaks, M.S. Thesis, UMD
Department of Fire Protection Engineering, 48 pp., May, 2008.
5. Nicholas R. Morton, Quenching Limits and Materials Degradation of Hydrogen Diffusion
Flames, M.S. Thesis, UMD Department of Fire Protection Engineering, 49 pp., May, 2008.

8. Service and Award

1. Peter B. Sunderland, Excellence in Oral Presentation, for Fire Hazards of Small Hydrogen
Leaks, SAE World Congress, Detroit, 2007.
2. Peter B. Sunderland, Editorial Co-Chair, Hydrogen Vehicle Fire Safety, SAE World
Congress, 2008.

5
 APPENDIX A

Reprints of Publications
 2007 Fall Technical Meeting
Eastern States Section of the Combustion Institute
University of Virginia
October 21-24, 2007

Quenching limits of hydrogen diffusion flames

M.S. Butler1, C.W. Moran 2, P.B. Sunderland 2, R.L. Axelbaum1
1
Dept. of Energy, Environmental, and Chemical Engineering, Washington Univ., St. Louis MO
63130, USA
2
Dept. of Fire Protection Engineering, Univ. of Maryland, College Park MD 20742, USA

This study examines the types of hydrogen leaks that can support combustion. Hydrogen,
methane, and propane diffusion flames on round burners and leaky compression fittings were
observed. Measurements included limits of quenching and blowoff for round burners with
diameters of 0.006 - 3.175 mm. The measured mass flow rates at the quenching limits were found
to be generally independent of burner diameter. In terms of mass flow rate, hydrogen had the
lowest quenching limit and the highest blowoff limit of the fuels considered. The quenching limits
for hydrogen diffusion flames on round burners with thick walls were found to be higher than for
thin walls. The minimu m mass flow rate of hydrogen that can support combustion from a leaking
compression fitting was found to be independent of pressure and about an order of magnitude
lower than the corresponding methane and propane flow rates. The implications for fire safety are
discussed.

Introduction
Concerns about the emissions of greenhouse gases have led to extensive consideration of
hydrogen as an energy carrier. Hydrogen presents several unusual fire hazards, including high
leak propensity, ease of ignition, and invisible flames. The scenario of interest in this work is that
a small leak in a hydrogen system might ignite, support a flame that is difficult to detect, and
degrade containment materials to the point of a catastrophic failure. This study includes
experiments and analysis to identify which hydrogen leaks can support flames.
A Department of Energy report [1] found that hydrogen containment was the chief safety
concern associated with using hydrogen as a transportation fuel and documented several
catastrophic hydrogen fires.
Quenching and blowoff limits bound the leak flow rates that can support combustion.
Measurements of propane quenching and blowoff flow rates were made by Matta et al. [2].
Quenching limits for methane were performed by Cheng et al. [3]. Blowoff measurements for
hydrogen, methane and propane were performed by Kalghatgi [4].
Research has been done in evaluating leak flow rates of hydrogen, methane, and propane. Swain
and Swain [5] modeled and measured leak rates for diffusion, laminar, and turbulent flow
regimes. They found that combustible mixtures in an enclosed space resulted more quickly for
propane and hydrogen leaks than for methane leaks. Their supply pressures were the same for all
fuels.
Thus motivated, the objectives of this work are to measure limits of sustained combustion, both
at quenching and blowoff, for hydrogen on round burners and lower flaming limits for hydrogen,
methane and propane on leaky compression fittings.
Flame Quench Scaling
A scaling analysis was developed to interpret measured flame quenching limits. These limits are
the minimum flow rates required to support a diffusion flame.
The stoichiometric length Lf of laminar gas jet diffusion flames on round burners is:
Lf / d = a Re = a ρ u0 d / μ, (1)
where d is the burner inside diameter, a is a dimensionless fuel-specific empirical constant, Re is
Reynolds number, u0 is the average fuel velocity in the burner, ρ is fuel density, and μ is fuel
dynamic viscosity. The scaling of Eq. (1) arises from many theoretical and experimental studies,
including Roper [6], Sunderland et al. [7], and references cited therein. Constant a here is
assigned values measured by Sunderland et al. [7], as listed in Table 1.
The base of an attached jet diffusion flame is quenched by the burner. Its standoff distance can
be approximated as one half of the quenching distance of a stoichiometric premixed flame. Such
quenching distances typically are reported as the minimum tube diameter, Lq , through which a
premixed flame can pass. It is assumed here that a jet flame can be supported only if its
stoichiometric length is greater than half this quenching distance:
Lf ≥ Lq / 2. (2)
Measurements of Lq , shown in Table 1, are taken from Kanury [8]. When co mbined, Eqs. (1) and
(2) predict the following fuel flow rate, mfuel , at the quenching limit:
mfuel = π ρ u0 d2 / 4 = π Lq μ / ( 8 a ). (3)
Equation (3) indicates that the fuel mass flow rate at the quenching limit is a fuel property that is
independent of burner diameter. When values of Lq , μ, and a from Table 1 are inserted into Eq.
(3), the predicted fuel flow rates at quenching shown in Table 1 are obtained.
Table 1: Selected fuel properties of hydrogen, methane, and propane. Values
for a are from Sunde rland et al. [7], Lq and SL are from Kanury [8], and μ is
from Weast and Astle [9].
Fuel a Lq [mm] SL [cm/s] μ [g/m-s] mfuel [mg/s]
predicted
H2 0.236 0.51 291 8.76e-3 0.008
CH4 0.136 2.3 37.3 1.09e-2 0.085
C3 H8 0.108 1.78 42.9 7.95e-3 0.063

Experime ntal
Two different burner configurations are considered here: round burners and leaky compression
fittings.
There is generally a limited range of flow rates for which a flame can be established on the
present burners. Above this range, the flow is said to be above its blowoff limit. This limit occurs
when the flow velocity in the flammable region of the flame becomes greater than the burning
velocity of the mixture. Below this range, the flow is said to be below the quenching limit.
Quenching occurs when there is too much heat loss for combustion to be sustained.
Round Burners
Three different types of round burners were considered for this work: tube burners, pinhole
burners, and curved-wall pinhole burners. Each had a range of flow passage diameters. All tests
were performed at ambient lab pressure, and flow was delivered via a pressure regulator and a
flow controlling valve.
The tube burners were made from stainless steel hypodermic tubes, oriented vertically. Tube
inside diameters were 0.006, 0.016, 0.033, 0.047, and 0.087 mm. These burners resemble
microinjectors that may be used in future, small scale microelectromechanical power generators
[7].
The pinhole burners were stainless steel nozzles that are manufactured for solid-stream spray
generation. The top of all the burners except the two smallest ones is a curved surface with a hole
passing through its axis as shown in Figure 1. The two smallest burners had holes in planar, not
curved, surfaces. It is expected that the variation of quenching flow rate as a result of this slight
curvature is on the order of the error associated with measurement. The hole diameters were
0.008, 0.127, 0.356, 0.530, 0.711, 0.838, 1.016, 1.397, 1.778, 2.388, and 3.175 mm.
The curved-wall pinhole burners were constructed of stainless steel tubes with two outside
diameters: 1.59 and 6.35 mm. A radial hole was drilled in each tube. Hole sizes were 0.406,
0.533, 0.737, 0.864, and 1.016 mm (small tubes) and 0.406, 1.753, 2.464 and 3.124 mm (large
tubes).
Figure 1 shows a false-color image of a hydrogen flame on a 0.356 mm pinhole burner just prior
to extinction. The photo was taken in a dimly lit room with an f- number of 4.2, ƒ/4.2, and an
exposure time of 30 s. The flame was still very faint after using these settings, so a digital
contrast enhancement was applied. The hydrogen quenching distance of Table 1 suggests that
steady hydrogen diffusion flames should be anchored about 0.25 mm above the burner surface.
Figure 1 shows this is reasonable for the present flames near their quenching limits.
Certain complications are unique to these experiments. The present small flow rates could not be
accurately measured using conventional flow meters. Instead, a glass soap-bubble meter was
installed upstream of the burners. Tests performed with varying air moisture contents found

Figure 1: False -color image of hydrogen flame over a

0.356 mm orifice.
quenching flow rate to be generally independent of air moisture content. Hydrogen flame
detection was complicated by their dimness, especially at low flow rates, but this was resolved
by using thermocouples to detect flames. K-type thermocouples were chosen to avoid catalysis,
and were placed several flame lengths above the flames to avoid disturbances.
Quenching flow rates were measured by first establishing a flame slightly larger t han that shown
in Figure 1. The flow rate was then decreased slowly until flame extinction. A soap bubble was
then introduced for flow rate measurement. Each burner was allowed to warm to just above room
temperature to prevent water condensation on the burner surface. This is because water
condensation was found to disturb flow from the small burners, sometimes extinguishing the
flames. Tests performed with different burner temperatures found quenching flow rate to be
largely independent of burner temperature.
The fuel flow rate at blowoff was measured by first establishing a stable, large flame. The flow
rate was then increased until the flame first lifted and then extinguished. For these tests the
flames were detected visually.
Leaky Fittings
Quenching limits were also measured for leaky compression fittings. These involved 6.35 mm
outside diameter stainless steel tubes. Each tube entered a tube union fitting, which was a
stainless steel Swagelok® compression fitting. The end of the union opposite the tube was
sealed. This configuration simulates a possible unintended leak in a fuel line. One fitting (burner
1) was intentionally overtightened to introduce a leak. The fitting was originally made correctly
using the manufacturer's instructions and an application of soap produced no bubbles for
hydrogen at 5.52 bar. The fitting was then tightened an additional 0.75 turns. Burner 2 had a
front ferrule that was intentionally scratched to simulate possible damage during installation, but
was otherwise made according to manufacturer instructions. Both overtightened and scratched
fittings are occasionally encountered in plumbing systems.
For the leaky fitting tests, pressure was varied to find quenching limits. Flow was measured
using a rotameter that was calibrated with a soap bubble meter. For methane and propane, the
minimum flaming flow rate was found by slowly increasing the pressure until an external pilot
flame produced a visible flame near the leak. The experiments were done in a dim room so that
methane and propane flames were readily visible. Thermocouples verified that even the smallest
methane and propane flames were visible.
For hydrogen an audible pop was always found to occur at ignition. For the lowest flow rates,
this pop was followed by extinction when the external flame was removed. Such extinction was
confirmed with thermocouples. At slightly higher pressures, this pop was followed by steady
burning. Near the quenching limits, hydrogen flames were never visible, even under darkened
laboratory conditions.
Several orientations were tested to see the effect of burner orientation on the minimum flow rate.
Each test was repeated to establish repeatability.
Results
Round Burners
Figure 2 shows a summary of all the limit data for tube burners. The blowoff flo w rates increase
with tube burner diameter, and the quenching flow rates are relatively flat. The
 1000

Blowoff Limits
100

Mass Flow Rate (mg/s) 10

Hydrogen
Hydrogen-Kalghatgi
Hydrogen
1 Propane-Matta
Propane-Kalghatgi
Propane_Matta
Methane-Kalghatgi
0.1 Methane-Cheng

0.01
Quenching Limits

0.001
0 0.5 1 1.5 2 2.5 3

Tube Diameter (mm)

Figure 2: Plot of quenching and blowoff limits versus tube diameter.

limits of combustion in Figure 2 are much wider for hydrogen than for propane or methane. The
hydrogen quenching limits are about an order of magnitude lower than those for methane and
propane, while hydrogen blowoff limits are about an order higher than those of the others. The
data from Matta et al [2], Cheng et al. [3], and Kalghatgi [4] show that the limits for methane and
propane are very similar. For all fuels, the quenching limits are nearly independent of burner
diameter, whereas blowoff limits increase with increasing diameter. The independence of
quenching limit from burner diameter is consistent with the analysis presented above, see Eq.
(3).
Figure 3 shows the present hydrogen quenching limits. Data are included for all three types of
burners. The theory of Eq. (3) predicts that the quenching flow rate is independent of burner
diameter. It appears from Figure 3 that the scaling analysis did roughly predict the average
quenching flow rate for hydrogen, although there is clearly another mechanism not accounted for
in the calculation.
The tube burners show the same trend as the data from Matta et al. [2]; there is a slight increase
in quenching flow rate with increasing burner size. It is reasonable to conclude that there is more
heat loss to the burner with increasing burner size; with more heat loss, the quenching flow rate
will increase to account for the weaker flame.
The pinhole burners show the upper limit for quenching flow rates. These flows are around 0.008
mg/s, as predicted. This plot does not exhibit the same trend as the tube burners because
approximately the same amount of heat loss occurs independent of burner size. The hump in
Figure 3 for orifice burners is under investigation.
The effect of curvature was investigated by using the curved-wall pinhole burners. It was
 0.012

Quenching Mass Flow Rate (mg/s)

0.01

0.008

0.006

0.004

Tube burner
0.002 1.59 mm curved-wall pinhole burner
Pinhole burner
6.35 mm curved-wall pinhole burner
0
0 0.5 1 1.5 2 2.5 3 3.5
Burner Diameter (mm)
Figure 3: Plot of quenching mass flow rates versus burner
diameter for hydrogen.

expected that with increasing curvature, a flame would experience less wall heat loss. This
should result in a stronger flame and a lower quenching flow rate. This is exactly what was
observed in the two curved-wall burners. The 6.35 mm burners act more like pinhole burners,
whereas the 1.59 mm burners act more like tube burners; the 6.35 mm curved-wall burners have
higher average quenching flow rates than the 1.59 mm curved-wall burners.
Both the tube and pinhole burners have an increase in quenching flow rate for the smallest
burners. This is because the velocity at the burner outlet is large enough to weaken the stability
of the flame.
Leaky Fittings
The leaky fittings were observed for hydrogen, methane, and propane in vertical, inverted, and
horizontal orientations. Figure 4 shows the effect of burner orientation on the minimum flaming
flow rate for Burner 1. It is seen that the orientation of the leak does not have an effect on the
quenching limit of hydrogen because these flames were so small at their limits. Burner
orientation did have an effect on propane and methane with a horizontal configuration requiring
the highest flow rate and inverted orientation requiring the lowest. The inverted orientation
required the lowest flow rate of all fuels as this kept the flame the farthest away from the burner
so that less heat was lost to the burner.
 0.6
6.3mm Burner 1 Vertical
0.5 6.3mm Burner 1 Inverted

Minimum Flowrate (mg/s)

6.3mm Burner 1 Horizontal
0.4

0.3

0.2

0.1

0.0
Hydrogen Methane Propane

Figure 4. Minimum flaming flow rate for Burner 1 in vertical, horizontal

and inverted orientations
Figure 5 shows the minimum flaming flow rate for hydrogen, propane, and methane for the
vertical burner orientation. The minimum flow rate necessary for sustained flaming is plotted
with respect to pressure. The data in Figure 5 were obtained by adjusting the torque on a leaky
fitting. Thus the data at higher pressures result from a fitting with a higher torque.
Within experimental uncertainties, the data of Figure 5 are independent of pressure for all three
fuels. The mean minimum flow rate for a flaming hydrogen leak is 0.040 mg/s. The hydrogen
flow rate was about an order of magnitude lower than for the other fuels due to its low quenching
distance and low molecular weight. This behavior is similar to that seen in the round burners
above, in Figure 2.

0.5

0.4
Minimum Flowrate (mg/s)

0.3

Hydrogen
Propane
0.2 Methane
Series4
Series5
Series6
Linear (Series4)
0.1 Linear (Series6)
Linear (Series5)

0
0 2 4 6 8
Pressure (bar)
Figure 5. Minimum flaming flow rate versus pressure in the
vertical orientation.
The minimum hydrogen flaming flow rate for round burners (Figure 2) is about an order of
magnitude lower than that for leaky fittings (Figure 5). This is attributed to additional burner heat
losses in the leaky fittings, where flames burn near concave metal surfaces.
Conclusions
The quenching and blowoff limits for hydrogen on round burners were measured and compared
to published values for other gases. Hydrogen diffusion flames on tube burners had much wider
limits of combustion than propane and methane. It has been shown that the quenching flow rates
for all these gases are largely independent of burner size.
For hydrogen the quenching flow rate was found to depend on burner type, owing to differences
in wall heat loss. Tube burners had the lowest quenching flow rates, orifice burners had the
highest, and radial burners fell in-between, as expected.
The minimum flow rate necessary for sustaining a hydrogen flame at a leaky compression fitting
is 0.040 mg/s. This is about an order of magnitude lower than for propane or methane. The
minimum mass flow rate for all fuels is nearly independent of pressure.
The round burner and leaky compression fitting experiments yielded similar results in that they
both found that the lower limits of combustion for hydrogen were about an order of magnitude
lower than those for propane and methane. The two experiments differed in that their lower
combustion limits were off by an order of magnitude owing to different amounts of heat loss.
Acknowledge ments
This work is supported by NIST under the technical direction of J. Yang. Helpful discussions
with B.H. Chao are appreciated.

References

[1] L.C. Cad wallader, J.S. Herring, Safety Issues with Hydrogen as a Vehicle Fuel, Report INEEL/ EXT-99-00522
prepared for U.S. Depart ment of Energy (1999).
[2] L.M . Matta, Y. Neu meier, B. Lemon, B.T. Zinn, Characteristics of Microscale Diffusion Flames, Proc.
Co mbust. Inst. 29 (2002) 933-938.
[3] T.S. Cheng, C.P. Chen, C.S. Chen, Y.H. Li, C.Y. Wu, Y.C. Chao, Characteristics of Microjet Methane
Diffusion Flames, Co mbust. Theory Modeling 10 (2006) 861-881).
[4] G.T. Kalghatgi, Blow-Out Stability of Gaseous Jet Diffusion Flames. Part I: In Still Air, Combustion Science
and Technology, 26:5, 233-239 (1981).
[5] M.R. Swain, M.N. Swain, A Comparison of H2, C H4, and C3H8 Fuel Leakage in Residential Settings, Int. J.
Hydrogen Energy 17:807-815 (1992).
[6] F.G. Roper, The Prediction of Laminar Jet Diffusion Flame Sizes: Part 1. Theoretical Model, Combust. Flame
29:219-226 (1977).
[7] P.B. Sunderland, B.J. Mendelson, Z.G. Yuan, D.L. Urban, Shapes of Buoyant and Nonbuoyant Laminar Jet
Diffusion Flames, Co mbust. Flame 116:376-386 (1999).
[8] A.M. Kanury, Introduction to Combustion Phenomena. Gordon and Breach, New Yo rk, p. 131 (1975).
[9] R.C. Weast, M.J. Astle, CRC Handbook of Chemistry and Physics, 59th edition, CRC Press, Inc. West Palm
Beach, p. F-58 (1979).
[10] B.R. Munson, D.F. Young, T.J. Okiishi, Fundamentals of Fluid Mechanics, Wiley, Hoboken, NJ, p. 359
(2002).
[11] N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. A xelbaum, Quenching Limits and Materials Degradation of
Hydrogen Diffusion Flames, 5th U.S. Co mbustion Meeting, San Diego (2007).
[12] M.Y. Bahadori, U. Hegde, L. Zhou, Liftoff and Blowoff of Jet Diffusion Flames, Paper AIAA-2002-1078,
Aerospace Sciences Meeting, Reno, Nevada, January 14-17, 2002.
Paper # A36 Topic: Laminar Flames

5th US Combustion Meeting

Organized by the Western States Section of the Combustion Institute
and Hosted by the University of California at San Diego
March 25-28, 2007

An asymptotic analysis of spontaneous ignition of hydrogen jets

Kian Boon Lim1, Beei-Huan Chao2, Peter B. Sunderland1, Richard L. Axelbaum3
1
Dept. of Fire Protection Engineering, Univ. of Maryland, College Park, MD 20742, USA
2
Dept. of Mechanical Engineering, Univ. of Hawaii, Manoa, HI 96822, USA
3
Dept. of Mechanical and Aerospace Engineering, Washington Univ., St. Louis, MO 63130, USA

Analytical work is presented for the spontaneous ignition of a hydrogen jet emanating from a slot
into air. A similarity solution of the flow field was obtained. This was combined with the species
and energy conservation equations, which were solved using activation energy asymptotics. Limits
of spontaneous ignition were identified as functions of slot width, flow rate, and temperatures of
the hydrogen jet and ambient air. Two scenarios are examined: a cool jet flowing into a hot
ambient and a hot jet flowing into a cool ambient. For both scenarios, ignition is favored with an
increase of either the ambient temperature or the hydrogen supply temperature. Moreover, for the
hot ambient scenario, a decrease in fuel Lewis number also promotes ignition. The Lewis number
of the oxidizer only has a weak effect on ignition. Because spontaneous ignition is very sensitive
to temperature, ignition is predicted to occur near the edge of the jet if the hydrogen is cooler than
the air and on the centerline if the hydrogen is hotter than the air.

Nomenclature
aT constant representing the temperature increase through reaction
B pre-exponential factor
cp specific heat at constant pressure
D mass diffusion coefficient
Da Damköhler number
D˜ a reduced Damköhler number
E activation temperature
f nondimensional streamfunction
Le Lewis number
p pressure
Pr Prandtl number
qF heat of combustion per unit mass of fuel
R ideal gas constant
Sc Schmidt number
T temperature
u flow velocity in the x (streamwise) direction
v flow velocity in the y (transverse) direction
W molecular weight
x streamwise spatial coordinate
X ignition location in the x direction from the exit of the slot
x0 distance from the fictitious origin of the jet to the exit of the crack
y transverse spatial coordinate (y = 0 along the centerline)
Y mass fraction

1
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

Greek
α parameter defined as α = (T˜∞ − T˜0 ) /Y˜F,0
ε small parameter, defined as ε = T˜ 2 / E˜ , used for asymptotic expansion
∞
φO perturbation of oxidizer concentration in the inner, reaction region
θ perturbation of temperature in the inner, reaction region
γ parameter defined as γ = (T˜ − T˜ ) /Y˜
0 ∞ O,∞
λ thermal conductivity
μ viscosity
ν stoichiometric coefficient
ρ gas density
ψ streamfunction
η similarity variable
ξ stretched spatial coordinate, defined in Eq. (24)
ζ stretched spatial coordinate in the inner, reaction region (inner variable)

Subscripts
0 value of variables at the exit of the crack
F fuel
f frozen solution
O oxidizer
T temperature
∞ ambient condition

Superscripts
~ nondimensional quantity
∧ rescaled nondimensional quantity

1. Introduction

Hydrogen jets flowing into heated air ignite spontaneously at an air temperature of 943 K [1].
This is cooler than for most other fuels [2], including gasoline and methane, and is not much
higher than the autoignition temperature of stoichiometric hydrogen/air mixtures, 858 K [1].
Hydrogen is under consideration as a major new fuel carrier. Occasional unintended hydrogen
leaks will be unavoidable, and some may involve heated hydrogen and/or air. Thus an improved
understanding of limits of spontaneous ignition of hydrogen jets is sought here, with the aid of
activation-energy asymptotics.
Asymptotic flame theories can provide valuable insights into combustion reactions [4-6]. Im
et al. [7,8] analyzed thermal ignition in supersonic hydrogen/air mixing layers and obtained
ignition characteristics over a wide range of conditions. An investigation of different combustion
regimes by Damköhler-number and activation-energy asymptotics in a stagnant mixing layer was
performed by Lee and Chung [9].
Zheng and Law [10] identified ignition limits of premixed hydrogen-air flames where
ignition was by heated counterflow. Toro et al. [11] examined in detail the structure of laminar
hydrogen jet flames both experimentally and numerically. Chaos et al. [12] examined Lewis-
number effects in unsteady laminar hydrogen jet flames. Liu and Pei [13] examined autoignition
and explosion limits of hydrogen-oxygen mixtures in homogenous systems.
The present analysis considers the spontaneous ignition of a jet of hydrogen or other gaseous
fuel leaking through a crack into air. The crack is taken to be straight and long, yielding a two-
dimensional flow field. The ignition analysis identifies limits of spontaneous ignition.

2
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

The objectives of this work are to:

1) develop a model of spontaneous ignition
for two cases: a cool fuel jet flowing into
heated air and a heated fuel jet flowing into
cool air, T∞ Reaction
Air zone
2) identify limits of spontaneous ignition as x
functions of slot width, flow rate, fuel
Lewis number, and temperatures of the fuel
jet and the ambient air, and
3) identify the location of ignition. y
uo
2. Formulation To
Fuel
The problem of interest is a steady, isobaric Figure 1: Schematic of slot and hydrogen leak.
laminar jet of fuel (e.g. hydrogen) at
temperature T0 issuing from a rectangular slot into an oxidizing environment (e.g. atmospheric
air) at a temperature of T∞ , as shown schematically in Fig. 1. Spontaneous ignition occurs when
either T0 or T∞ is sufficiently high that the weak reaction between the fuel and the oxidizer
transitions to a vigorous burning flame. This study analyzes the ignition state as a function of
various physical properties including Lewis number, T0, T∞ , the flow velocity at the slot exit, u0,
and the width of the slot. The slot is considered sufficiently long that end effects are negligible.
The reaction chemistry is simulated by a single-step, overall, irreversible reaction with second
order Arrhenius kinetics and a high activation energy.
With the above problem definition, conservation of mass and momentum are given by
∂ ( ρ u) ∂ ( ρ v)
+ =0 , (1)
∂x ∂y
∂u ∂ u ∂ ⎜⎛ ∂u ⎟⎞
ρu + ρv − μ =0 , (2)
∂x ∂ y ∂ y ⎝ ∂ y⎠
which are to be solved subject to the boundary conditions
y = 0 , x = x0 : u = u0 ; x > x0 : ∂ u/∂ y = 0 , v=0 , (3)

y →∞ : u →0 . (4)
Introducing a streamfunction ψ of the form
ρ u = ρ∞ u0 (∂ψ /∂ y ) , ρ v = − ρ∞ u0 (∂ψ /∂ x ) , (5)

such that the continuity equation is satisfied, transforming the coordinates from (x, y) to ( x˜ , η)
where

x˜ = x / x0 , η = ρ∞ u0 /(6 μ∞ x0 ) x˜−2/3 ∫0y ( ρ / ρ∞ )d y , (6)

3
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

and defining

ψ = 6 μ∞ x0 /( ρ∞ u0 ) x˜1/3 f (η ) , (7)
Eqs. (2) – (4) are transformed to

( f ′′′ /2) + f f ′′′ + f ′2 = 0 . (8)

η=0 : f = f ′′ = 0 f ′ =1 ; η →∞ : f ′→0 . (9)

A similarity solution is assumed to exist so that f is a function of η only. Solving Eqs. (8) – (9)
following Schlichting [14] and Bickey [15] yields f = tanh η.

Applying the coordinate transformation and the solution of the momentum equation to the
energy and species conservation equations, we obtain

1 ∂2 T˜ ∂ T˜ 2 η ) x˜ ∂ T˜ = −Da x˜ 4/3Y˜ Y˜ T˜−1 exp(− E˜ /T˜ )

+ 2(tanh η ) − 6(sech F O , (10)
Pr ∂η2 ∂η ∂ x˜

1 ∂2 Y˜F ∂ Y˜F 2 η ) x˜ ∂ Y˜F = Da x˜ 4/3Y˜ Y˜ T˜−1 exp(− E˜ /T˜ )

+ 2(tanh η ) − 6(sech F O , (11)
Sc F ∂η2 ∂η ∂ x˜

1 ∂2 Y˜O ∂ Y˜O 2 η ) x˜ ∂ Y˜O = Da x˜ 4/3Y˜ Y˜ T˜−1 exp(− E˜ /T˜ )

+ 2(tanh η ) − 6(sech F O , (12)
ScO ∂η2 ∂η ∂ x˜

which are to be solved subject to

η = 0 , x˜ =1 : T˜ = T˜0 , Y˜F = Y˜F,0 , Y˜O = 0 , (13)

η = 0 , x˜ >1 : ∂ T˜ /∂η = ∂ Y˜F /∂η = ∂ Y˜O /∂η = 0 , (14)

η →∞ : T˜ → T˜∞ , Y˜F → 0 , Y˜O → Y˜O,∞ . (15)

In the above,
c T ν W 6 x0 νO c p p B
T˜ = p , Y˜F = YF , Y˜O = F F YO , Da =
qF νOWO WF qF Ru0
c E μ μ λ /c p
E˜ = p , Pr = , Sc j = , Le j =
ρ Dj .
qF λ /c p ρ Dj

The values of cp, ρ λ, ρ μ, and ρ2 Dj are considered constant. The ideal gas equation of state has
been adopted here.

In the non-reactive limit, solving Eqs. (10) – (15) gives the frozen solutions,

T˜f = T˜∞ + ( T˜0 − T˜∞ )(sech2Pr η )/x˜1/3 , (16)

4
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

Y˜F, f = Y˜F,0 (sech2ScF η )/ x˜1/3 , (17)

Y˜O, f = Y˜O,∞ [1− (sech2ScO η )/ x˜1/3 ] . (18)

2.1. Cool jet flowing into a hot ambient (T∞ > T0)
In the presence of a weak reaction, the temperature is increased from its frozen value by a
small, O (ε) amount where ε = T˜∞2 / E˜ while the reactant concentrations are reduced from their
respective frozen values by an O (ε) amount. Because ignition is primarily controlled by
temperature, ignition occurs near η → ∞ if successful. Away from this high temperature region,
the reaction is frozen. In the outer, chemically frozen region, the outer solutions are similar to
Eqs. (16) – (18) but with an O (ε) change in their values. In the inner, reactive region, defining a
stretched inner variable as

ζ = Y˜F,0 (sech η )2Pr /(ε x˜1/3 ) , (19)

and substituting into Eqs. (10) – (12) yields, when LeF is sufficiently smaller than unity, as for
hydrogen,

ζ 2 (∂2 θ /∂ζ 2 ) = − D˜ a ζ LeF exp(θ − αζ ) , (20)

where α = ( T˜∞ − T˜0 )/Y˜F,0 and the reduced Damköhler number is defined as

D˜ a = Da Y˜O,∞ x˜ 4 /3[Y˜F,0 /(ε x˜1/3 )]1− LeF exp(− E˜ /T˜∞ )/(4Pr T˜∞ ) . (21)

The boundary conditions required to solve this equation can be found by matching the inner and
outer solutions as
ζ=0 : θ=0 ; ζ →∞ : ∂θ /∂ζ → 0 , θ → aT (22)

For the case of LeF close to unity, Eq. (20) is modified to

ζ 2 (∂2 θ /∂ζ 2 ) = − D˜ a(ζ − θ )exp(θ − αζ ) . (23)

2.2. Hot jet flowing into a cool ambient (T0 > T∞)
For the case of a hot jet issuing into a cold ambient, any ignition will occur near the jet
centerline, η = 0. Moreover, because the jet will be cooled by the cold ambient gas along the
flow, ignition is expected to occur near the slot exit. The analysis is similar to that in Section 2.1,
except that ε = T˜02 / E˜ and the inner variables are defined as

ζ = η PrY˜O, ∞ /ε , ξ = Y˜O,∞ ( x˜ −1)/(3ε ) , (24)

leading to

∂2 θ ∂θ 1 ∂2 φO ∂φ
− 2 + −2 O = 0 , (25)
∂ζ 2 ∂ξ LeO ∂ζ 2 ∂ξ

5
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

∂2 θ ∂θ
− 2 = − D˜ a( φO + LeO ζ 2 + ξ )exp[θ − β (ζ 2 + ξ )] , (26)
∂ζ 2 ∂ξ
with the initial and boundary conditions
θ (ξ = 0) = φO (ξ = 0) = 0 , (27)

ζ = 0 and ζ → ∞ , ξ > 0 : ∂θ /∂ζ = ∂φO /∂ζ = 0 , (28)

where γ = ( T%0 − T%∞ ) / Y%O ,∞ and the reduced Damköhler number is

D˜ a = ε[Da Y˜F,0 /( T˜0 Y˜O,∞ )]exp(− E˜ /T˜0 ) . (29)

Ignition is considered successful when the heat generation through reaction is sufficient to
compensate the heat loss from the jet to the ambient at any location, and the ignition criterion is
given by
(∂θ /∂ξ )ζ =0 ≥ β or ∂θ /∂ζ ≥ 2 βζ at any ξ . (30)

3. Results

3.1. Cool jet flowing into a hot ambient (T∞ > T0)
Equations (20) and (23), subject to Eqs. (21) – (22), were solved by a fourth order Runge-
Kutta method. The results are shown in Fig. 2, a plot of the reaction temperature increase versus
reduced Damköhler number. This reveals the lower and middle branches of an S-curve [4]. In
each such curve, there is a maximum value of D˜ a above which a solution does not exist. For
values of D˜ a smaller than this critical value, there are two solutions for each D˜ a . The critical
value of D˜ a represents the transition
10
from weak reaction to vigorous burning,
and is defined as the ignition state. The 9
lower branch, showing an increase of 8
temperature with higher reaction rate, is
7
the physically realistic branch. The
middle branch represents conditions that 6
are not physically possible. Spontaneous
aT

5
ignition is predicted for any D˜ a greater
4
than this critical value. α = 0.03
3 0.02
Three curves are included in Fig. 2, 0.01
each with a different value of 2
α = (T˜∞ − T˜0 ) / Y˜F,0 . Fig. 2 indicates that a 1
~
decrease in α reduces the critical Da at 0
ignition, which means that ignition is 0 0.01 0.02 0.03 0.04 0.05
%
Da
favored. Such a decrease can be
accomplished either by increasing the Figure 2: aT versus D% a for varying α, with LeF = 0.6
reactant mass fraction in the fuel supply, and constant T%∞ .

6
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

Y˜F,0 , or by increasing the jet temperature, T˜0 . Both findings are physically realistic.
Parameter α also can be changed by variations in the ambient temperature, T˜∞ , but this
~ ~
changes Da simultaneously. To investigate the effects of T˜∞ variations at fixed Da requires a
rescaling. The rescaling is performed here by specifying a reference value of T˜∞ as Tˆ∞ , defining
rescaled parameters εˆ = ( Tˆ∞ )2 / E˜ , aˆT = ( T˜∞2 /Tˆ∞2 )aT and

Dˆ a = Da Y˜O,∞ x˜ 4 /3[Y˜F,0 /(εˆ x˜1/3 )](1− LeF ) exp(− E˜ /Tˆ∞ )/(4Pr Tˆ∞ ) , (31)

and plotting the results in terms of rescaled variables aˆT and Dˆ a . The results are shown in
Fig. 3. Here an increase in T˜∞ , which increases α without changing Dˆ a , is seen to favor ignition.
This also is physically realistic because more heat is transferred to the cold fuel flow at a higher
rate when the ambient is at a higher temperature. By the same reason, when the kinetic data are
unchanged, an increased T˜∞ yields ignition to occur nearer the edge of the jet. Note that the
Damköhler number shown after Eq. (15) is a function of the radial distance from the virtual
origin of the jet.
The influence of reactant Lewis number on ignition is considered in Fig. 4. Here LeF is
defined as the mixture thermal diffusivity divided by the mass diffusivity of fuel into the
mixture. A decrease in LeF for fixed Dˆ a is seen to favor ignition. This occurs because a smaller
LeF implies that fuel species diffuse more quickly into the hot oxidizer. A fuel such as a mixture
of hydrogen and nitrogen has a small LeF. Nayagam and Williams [16] found that in a one-
dimensional model of steady motion of edges of reaction sheets, increasing the Lewis number
decreases the propagation velocity at small Damköhler numbers.

14 10

12
8
10
0.03 0.02 α = 0.01 6
8
aT
aˆT

6 4

4
2
2 0.6 0.8 Le F = 1.0

0 0
1E-9 1E-7 1E-5 1E-3 1E-1 0.001 0.01 0.1
D̂a D̂a
Figure 3: aˆT versus D̂a for varying α, Figure 4 : aT versus D̂a for varying LeF , with
with LeF = 0.6 . α = 0.02, T˜0 = 0.0358, and E˜ = 1.79 .

7
 5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

3.2. Hot jet flowing into a cool ambient (T0 > T∞)
Eqs. (25) – (28) were solved by the Crank-Nicholson method and the resulting matrix was
inverted by LU decomposition. Selected results are shown in Figs. 5 and 6. In Fig. 5, θmax
represents the maximum value of temperature increase through reaction before ignition occurs.
The corresponding ignition location, X, is shown in Fig. 6.
Three curves are included in Figs. 5 and 6, each with a different value of γ, where
γ = ( T%0 − T%∞ ) / Y%O ,∞ . These curves do not have the shape of an S-curve because the solutions are
derived from a partial differential equation, in contrast to the ordinary differential equation of the
~
cool jet. On each curve, by increasing the reaction rate, Da , a smaller temperature increase and
a shorter ignition location is observed before ignition, as is reasonable to expect. A higher value
~
of Da yields an increased heat generation rate, which compensates for some heat loss from the
hot jet to the cold ambient, favors ignition, and moves the point of ignition closer to the jet exit.
~
In contrast, a reduction in Da weakens the reaction and makes ignition more difficult such that
both θmax and X increase. Although an increase of X provides longer residence time for the
~
reaction so that ignition can occur at a smaller Da , the reaction rate decreases with X because
the jet is cooled by the cold ambient, as can be seen from the reaction term of Eq. (26). A sharp
~
increase in θmax and X , as shown the low Da side of the γ = 0.3 curves in Figs. 5 and 6, means
that the reduction of reaction rate dominates over the residence time increase, and defines the
~
smallest Da for which ignition occurs.
~
Figures 5 and 6 also indicate that a decrease in γ for any fixed Da favors ignition, as
ignition occurs at a lower temperature increase, θmax, and at a shorter ignition location, X. More
~
importantly, a decrease in γ permits ignition at a lower value of Da . Such a decrease can be

1.0 1.8
0.3
1.6 0.3

1.4

0.8 1.2
1
X
m ax

0.8
θ

0.6 0.2 0.6 0.2

0.4
γ =0.1 γ = 0.1
0.2
0.4 0
0 1 2 3 0 1 2 3 4 5
%
Da %
Da
Figure 5 : θ max % for varying γ, with
versus Da % for varying γ, with
Figure 6 : X versus Da
Leo=1.0 and constant T%0 . Leo = 1 .

8
 5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

accomplished either by increasing the reactant mass fraction in the oxidizer supply, Y˜O,∞ , or by
increasing the ambient temperature, T˜∞ . Both findings are physically realistic.
~
Parameter γ also can be changed by variations in the jet temperature, T˜0 , but this changes Da
~
simultaneously. To investigate the effects of T˜0 variations at fixed Da requires a rescaling
similar to that performed in Section 3.1. The rescaling is performed here by specifying a
reference value of T˜0 as Tˆ0 , defining rescaled parameters εˆ = ( Tˆ0 )2 / E˜ , θˆmax = ( T˜0 /Tˆ0 )2θmax and

Dˆ a = εˆ [Da Y˜F,0 /( Tˆ0 Y˜O,∞ )]exp(− E˜ /Tˆ0 ) , (32)

and plotting the results in terms of rescaled variables θˆmax and Dˆ a . The results are shown in
Figs. 7 and 8. Here an increase in T˜0 , which increases γ without changing Dˆ a , is seen to favor
ignition because ignition can occur at a lower reaction rate, or lower Dˆ a . This also is physically
realistic. Ignition is predicted to occur near the centerline if the fuel is hotter than the air because
this is where the highest temperature is attained.
The effects of oxidizer Lewis number on spontaneous ignition are considered in Fig. 9. This
plot shows that decreased LeO makes ignition more difficult. For an increase in the mass
~
diffusivity of the oxidizer (or a decreased LeO) at a fixed value of Da , θmax increases. In
~
addition, the minimum Da for ignition increases with decreased LeO. This differs from the
ignition behavior with respect to fuel Lewis number in the cool jet case (Fig. 4). In a cool jet,
there is unlimited heat transfer from the hot ambient gas to preheat the fuel so that a higher fuel
diffusion rate (lower LeF) results in a higher fuel concentration in the reaction region, more heat
generation through the reaction and, hence, easier ignition. In the hot jet, only limited heat is

1.0 1.0
0.3
0.8

0.8
0.6
θˆmax

0.2 0.3
0.4
0.6
0.2
γ = 0.1 0.2
γ = 0.1

0.4 0.0
0.001 0.01 0.1 1 10 0.001 0.01 0.1 1 10
D̂a D̂a
Figure 7 : θˆmax versus D̂a for varying γ, with Figure 8 : X versus D̂a for varying γ , with
Leo = 1 . Leo = 1 .

9
 5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

0.8 0.8

0.7 Le o = 0.9
Le o = 0.9
0.6 1.0
0.7 1.0
0.5 1.1
max

X
1.1 0.4
θ

0.3
0.6
0.2

0.1
0.5 0.0
0 1 2 3 4 5 0 1 2 3 4 5
%
Da %
Da
Figure 9 : θ max versus %
Da for varying Leo , Figure 10 : X versus % for varying Leo,, with
Da
with γ = 0.3 . γ = 0.3 .

available from the fuel flow. An increased oxidizer mass diffusivity increases the transport rate
of oxidizer to the center of the jet, thus requiring more heat to preheat the oxidizer, decreasing
the temperature in the hot zone, and making ignition more difficult. Furthermore, unlike the cool
jet case, the Lewis number of the oxidizer only has a weak effect on the ignition state (see Fig. 9)
because ignition occurs near the jet exit if successful. In the reaction region, the flow velocity is
high such that streamwise convection dominates over transverse diffusion. Moreover, because
LeO is close to unity for oxygen in air, the effect of LeO in a hot-hydrogen cold-air system is
secondary. As in Fig. 8, the effect of LeO on the ignition location is similar to that on the ignition
state, shown in Fig. 10.

4. Conclusions

The spontaneous ignition of a hydrogen jet emanating from a slot into air has been
considered analytically. A similarity solution of the flow field was obtained, which was
combined with the species and energy conservation equations. Solutions were found using
activation energy asymptotics.
The analysis yielded limiting conditions for spontaneous ignition of fuel jets. For a cool jet
flowing into a hot ambient, ignition is found to be a strong function of ambient temperature and
fuel Lewis number. Ignition was favored by an increase in ambient temperature or a decrease in
Lewis number. For the hot jet scenario, ignition was significantly affected by the jet temperature,
but only weakly affected by the oxidizer Lewis number.
Because spontaneous ignition is very sensitive to temperature, ignition is predicted to occur
near the edge of the jet if the fuel is cooler than the air and on the centerline if the fuel is hotter
than the air.
The present model can be extended to studies of flame extinction and to circular jet
configurations.

10
5th US Combustion Meeting – Paper # A36 Topic: Laminar Flames

Acknowledgment

This work was supported by NIST grant 60NANB5D1209 under the technical management of
J. Yang.

References
[1] CRC Handbook of Chemistry and Physics, R.C. Weast, ed., CRC Press, West Palm Beach (1979).
[2] E.C. Magison, Electrical Instruments in Hazardous Locations, 2nd Edition, Instrument Society of America,
Pittsburg, Pennsylvania, p304, 1972.
[3] J. Hord, Is Hydrogen a Safe Fuel?, International journal of hydrogen energy, 3:157, 1978.
[4] F.A. Williams, Combustion Theory, 2nd Edition, Addison-Wesley Publishing Company, Redwood City, CA,
p76-84, 1985.
[5] N. Peters, Turbulent Combustion, Cambridge University Press, Cambridge, England, p178-186, 2000.
[6] A. Liñán, F.A. Williams, Fundamental Aspects of Combustion, Vol. 34 of Oxford Engineering Science Series,
Oxford University Press, New York, p57-81,1993.
[7] H.G. Im, S.R. Lee, C.K. Law., Ignition in the Supersonic Hydrogen/Air Mixing Layer with Reduced Reaction
Mechanisms, AIAA-1994-548 Aerospace Sciences Meeting and Exhibit, 32nd, Reno, NV, Jan 10-13, 1994
[8] H.G. Im, B.H. Chao, J.K. Bechtold, C.K. Law, Analysis of Thermal Ignition in the Supersonic Mixing Layer,
AIAA Journal 32, p341-349, 1994.
[9] S.R. Lee, S.H. Chung, On the structure of Hydrogen Diffusion Flames with Reduced Kinetic Mechanisms,
Combustion Science Tech., 96, p247.
[10] X.L. Zheng, C.K. Law, Ignition of Premixed Hydrogen/Air by Heated Counterflow under Reduced and
Elevated Pressures, Combustion and Flame p136-168, 2004.
[11] V.V. Toro, A.V. Mokhov, H.B. Levinsky, M.D. Smooke, Combined Experimental and Computational Study
of Laminar, Axisymmetric Hydrogen-air Diffusion Flames, Proc. Combust. Inst. 30:485, 2005.
[12] M. Chaos, R.-H Chen, E.J. Welle, W.L., Roberts, Fuel Lewis Number Effects in Unsteady Burke Schumann
Hydrogen Flames, Combust. Sci. Tech. 177:75, 2005.
[13] Y., Liu, P., Pei , Asymptotic analysis on autoignition and explosion limits of hydrogen-oxygen mixtures in
homogeneous systems, International journal of hydrogen energy , Vol. 31, No. 5, pp. 639-647, 2006.
[14] H. Schlichting, Boundary-Layer Theory, Seventh Edition, McGraw-Hill, Inc.,p. 127-200, 1979.
[15] W. Bickey, The Plane Jet, Phil. Mag. Ser. 7, 23: 727, 1939.
[16] V. Nayagam, F.A. Williams, Lewis-number effects on edge-flame propagation, Journal of fluid mechanics,
458: 219-228, 2002.

11
 NHA Hydrogen Conference
San Antonio, Texas
March 21, 2007

FIRE HAZARDS OF SMALL HYDROGEN LEAKS

N.R. Morton1, P.B. Sunderland1, B.H. Chao2, R.L. Axelbaum3

This study examines the types of hydrogen leaks that can support combustion
and the corrosive effects of hydrogen flame exposure to aluminum, galvanized
steel, stainless steel, and SiC fibers. Hydrogen, methane, and propane diffusion
flames on round burners were observed. Measurements included limits of
quenching and blowoff for burners with diameters of 0.36 — 1.78 mm. The
measured mass flowrates at the quenching limits were found to be independent
of burner diameter. In terms of mass flowrates, hydrogen had the lowest
quenching limit and the highest blowoff limit of the fuels considered. Hydrogen
flames were found to be more corrosive than methane flames to aluminum,
galvanized steel, and SiC fibers.

1. Introduction
Concerns about the emissions of greenhouse gases have led to extensive
consideration of hydrogen as an energy carrier. Hydrogen presents several
unusual fire hazards, including high leak propensity, ease of ignition, and
invisible flames. The scenario of interest in this work is that a small leak in a
hydrogen system might ignite, support a flame that is difficult to detect, and
degrade containment materials to the point of a catastrophic failure. This study
includes experiments and analysis to identify which hydrogen leaks can support
flames. Material degradation by hydrogen and methane flames also is examined
here.
A Department of Energy report [1] found that hydrogen containment was the
chief safety concern associated with using hydrogen as a transportation fuel and
documented several catastrophic hydrogen fires.
Quenching and blowoff limits bound the leak flowrates that can support
combustion. Measurements of propane quenching and blowoff flowrates were
made by Matta et al. [2]. Quenching limits for methane were performed by Cheng
et al. [3].
Research has been done in evaluating leak flow rates of hydrogen, methane, and
propane. Swain and Swain [4] modeled and measured leak rates for diffusion,
laminar, and turbulent flow regimes. They found that combustible mixtures in an
enclosed space resulted more quickly for propane and hydrogen leaks than for
methane leaks. Their supply pressures were the same for all fuels.
Khan et al. [5] examined the effects of raised temperatures on carbon fabric/epoxy
composites, a likely material for high pressure storage tanks. Pehr [6] discusses
1
Dept. of Fire Protection Engineering, Univ. of Maryland, College Park MD
2
Dept. of Mechanical Engineering, Univ. of Hawaii, Manoa HI
3
Dept. of Mechanical and Aerospace Engineering, Washington Univ., St. Louis MO
1
some of the issues associated with hydrogen containment. Utgikar and Thiesen [7]
discuss the impact of hydrogen on materials, and the safety of hydrogen fuel
tanks.
Thus motivated, the objectives of this work are to (1) measure limits of flaming
(at quenching and blowoff) for hydrogen, methane, and propane issuing from
circular burners of various sizes, and (2) examine material degradation arising
from exposure to hydrogen and methane diffusion flames.

2. Flame Quench Scaling

A scaling analysis was developed to interpret measured flame quenching limits.
These limits are the minimum flow rates required to support a diffusion flame.
This analysis also yields a dimensionless crack parameter that indicates how close
a given leak is to the quenching limit.
The stoichiometric length Lf of laminar gas jet diffusion flames on round burners
is:
Lf / d = a Re = a ρ u0 d / μ , (1)
where d is burner inside diameter, a is a dimensionless fuel-specific empirical
constant, Re is Reynolds number, u0 is the average fuel velocity in the burner, ρ is
fuel density, and μ is fuel dynamic viscosity. The scaling of Eq. (1) arises from
many theoretical and experimental studies, including Roper [8], Sunderland et al.
[9], and references cited therein. Constant a here is assigned values measured by
Sunderland et al. [9], as listed in Table 1.

Table 1: Selected fuel properties of hydrogen, methane, and

propane. Values for a are from Sunderland et al. [9], Lq and SL are
from Kanury [10], and μ is from Weast and Astle [11].
mfuel [mg/s] mfuel [mg/s]
Fuel a Lq [mm] SL [cm/s] μ [g/m-s]
predicted measured
H2 0.236 0.51 291 8.76e-3 0.008 0.021
CH4 0.136 2.3 37.3 1.09e-2 0.085 0.112
C3H8 0.108 1.78 42.9 7.95e-3 0.063 0.031

The base of an attached jet diffusion flame is quenched by the burner. Its standoff
distance can be approximated as one half of the quenching distance of a
stoichiometric premixed flame. Such quenching distances typically are reported as
the minimum tube diameter, Lq, through which a premixed flame can pass. It is
assumed here that a jet flame can be supported only if its stoichiometric length is
greater than half this quenching distance:
Lf ≥ Lq / 2 to support a flame.
(2)
Measurements of Lq, shown in Table 1, are taken from Kanury [10]. When
combined, Eqs. (1) and (2) predict the following fuel flowrate, mfuel, at the
quenching limit:

2
 mfuel = π ρ u0 d2 / 4 = π Lq µ / ( 8 a ) . (3)
Equation (3) indicates that the fuel mass flow rate at the quenching limit is a fuel
property that is independent of burner diameter. When values of Lq, µ, and a from
Table 1 are inserted into Eq. (3), the predicted fuel flowrates at quenching shown
in Table 1 are obtained.
A crack parameter can now be derived. Assuming fully-developed,
incompressible laminar flow in the burner,
u0 = d2 ∆p / (32 μ Lb ) , (4)
where Δp is the pressure drop across the burner and Lb is the burner flow passage
length [12]. Equation (4) is valid for many laboratory burners. However,
compressed hydrogen storage systems at pressures of up to 350 bar require more
advanced models of leak flowrates such as those in [4].
Combining Eqs. (1), (2), and (4) yields, for leaks that are fully-developed,
incompressible and laminar,
CP = a ρ d4 ∆p / (16 μ2 Lb Lq ) ≥ 1 to avoid flame quenching, (5)
where CP is the dimensionless quenching crack parameter.

3. Experimental
3.1 Quenching and Blowoff
Quenching and blowoff limits of hydrogen, methane, and propane diffusion
flames were measured. These tests involved five burners with inside diameters of
0.356, 0.711, 0.838, 1.397, and 1.778 mm. All tests were performed in quiescent
air at 1.01 bar.
The burners were stainless steel nozzles that are manufactured for spray
generation. The top of each burner is a curved surface with a hole passing through
its axis. Fuel was delivered to each burner via a pressure regulator, a metering
valve, and a rotameter.
For the quenching limit measurements, a flame approximately 5 mm long was
ignited. The flow was then reduced until the flame extinguished. This was done
several times for each burner and each fuel. The flames were small enough, and
the experiments were done quickly enough, that there was no noticeable increase
in the temperature of the burners. Measurement of the hydrogen quenching limits
required special care, as small hydrogen flames are nearly invisible even in a
darkened lab. Methods to identify quenching for hydrogen flames included
passing paper above the burner and increasing the flowrate.
Also measured were the blowoff limits of each fuel for each burner. Blowoff
limits were measured by igniting a flame and then increasing the flow rate until
the flame lifted off and extinguished. The tests were performed quickly to ensure
burner temperatures remained close to ambient.

3
Blowoff occurs when velocities in the flammable regions exceed the burning
velocity. For laminar flames the relevant burning velocity is the laminar flame
speed, which is shown in Table 1, however most of the present flames were
turbulent just before blowoff.
3.2 Materials Degradation
Materials degradation tests were performed on specimens of six different
materials: aluminum alloy 1100, galvanized 1006-1008 carbon steel, 304 stainless
steel, 316 stainless steel, SiC yarn and SiC filament. These materials were chosen
owing to their common use in gas storage systems. The specimens were
approximately 100 mm long, with diameters as given in Table 2.
Table 2: Wire and fiber specimen
diameters.
Material Diameter (mm)
Aluminum Alloy 1100 1.01
Galvanized 1006-1008 Carbon Steel 1.04
304 Stainless Steel 1.04
316 Stainless Steel 1.01
SiC yarn 1.14
SiC filament 0.015

The burners for these tests were stainless steel tubes with inside diameters of 2.43
mm. The flames were approximately 15 mm long, and are shown in Fig. 1. These
images were recorded using a Nikon D100 Digital Camera with a 60 mm focal
length lens, and with ISO 1600, direct sunlight white balance, 50 ms shutter time,
and f/3.8.
The samples were installed horizontally in hydrogen and methane diffusion
flames at a height of 7 mm. This height was near the flame mid-height and was
low enough in the methane flame to avoid soot deposition.

4. Results
The images of Fig. 1 show sample hydrogen and methane diffusion flames. The
methane flame exhibits the familiar blue and yellow regions of hydrocarbon
diffusion flames. The hydrogen flame is much dimmer and is visible only in a
darkened room.
4.1 Quenching and Blowoff Limits
The measured fuel mass flowrates at quenching and blowoff are presented in Fig.
2. Results are shown for hydrogen, methane, and propane and are plotted as a
function of burner diameter.

4
 15 mm
A B

Figure 1: Color images of (A) hydrogen flame and (B) methane flame.

Figure 2 shows that burner mass flowrate at the quenching limit is independent of
burner diameter. This finding is supported by the prediction of Eq. (3). Mass
flowrates at the quenching limits increase from hydrogen to propane to methane.
Results averaged for all burner diameters are shown in Table 1. The predictions of
Eq. (3), also given in Table 1, capture the trends of the quenching experiments. It
may be possible to improve the agreement by using different published values of
quenching distances of premixed flames. The prediction may also be improved by
using the available standoff distances of these fuels, instead of quenching
distance, as the length scale in the analysis.
Matta et al. [3] measured quenching limits for propane, and found that flowrate is
nearly independent of burner tube diameter. The prediction [3] uses to correlate
the quenching data uses the standoff distance as the length scale for the analysis.
Ref. [3] also noted that the predicted flow velocity for the flammable mixtures
will be larger than the local flame speed at blowoff. The measurements from the
present study were found to be lower for the quenching regime, but similar for
blowoff.
Cheng et al. [2] measured quenching velocities for methane, and makes use of
flame length correlations and measurements of standoff distance to predict when
quenching will occur. The measurements from the present study were found to be
smaller then the quenching measurements and predictions from [2].
The blowoff measurements in Fig. 2 show that mass flowrate at blowoff increases
with burner diameter. Blowoff mass flowrates increase from methane to propane
to hydrogen. This is qualitatively supported by the laminar flame speeds shown in
Table 1.

5
 1000 1000

100
100

10
Fuel Quenching Blowoff
10 H2
CH4
m fuel (mg/s)

C3H8

u0 (m/s)
Matta [CH4]
1
Cheng [C3H8]
Fuel Quenching Blowoff
H2
1 CH4
C3H8
Matta [CH4]
0.1
Cheng [C3H8]

0.1
0.01

0.01 0.001
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Burner Diameter (mm) Burner Diameter (mm)

Figure 2. Measured fuel mass flowrate at Figure 3. Measured fuel velocity at the
the quenching and blowoff limits versus quenching and blowoff limits versus
burner diameter. The lines are the fits of burner diameter. The curves shown are fits
the present experiments. of the present experiments.

Figure 3 shows the same measurements and correlations of Fig. 2 when the
ordinate is changed to fuel velocity. This figure suggests a regime may exist at the
smallest burner diameters where the blowoff limit is lower than the quenching
limit. Burners smaller than those considered here will need to be tested to further
evaluate this.
4.1 Material Degradation
Aluminum alloy 1100 showed very different effects when exposed to hydrogen
and methane flames. Figure 4 includes images of the aluminum samples after the
8 hour exposure. The hydrogen flame caused severe warping, as well as
noticeable oxidation of the aluminum wire, after one hour. As the test continued,
the distortion became more and more severe, as did the degree of oxidation, until
the aluminum wire failed. The wire in the methane flame for the same exposure
time did not reveal these effects. There is some slight discoloration where some
soot deposited from the methane flame, but nothing approaching what was
observed for they hydrogen exposure.

6
 B

Figure 4: Aluminum wires following 8 hours of exposure to (A)

hydrogen flame and (B) methane flame.

Another material that showed significant differences upon flame exposure is

galvanized carbon steel. The sample exposed to a hydrogen flame showed more
significant corrosion than the sample in the methane flame, see Fig. 5.

A B

Figure 5: Galvanized 1006-1008 Carbon steel following one hour of

exposure to (A) hydrogen flame and (B) methane flame. The scale
markings are in mm.

A test of one hour exposure of the fiber yarn showed that it performed similarly in
both hydrogen and methane flames. Several individual filaments failed during
both exposures, but most remained intact.
Individual SiC filaments were observed to burn through during exposure to either
hydrogen or methane flames. Filaments in the hydrogen and methane flames were
observed to fail in 15 and 116 minutes, respectively.

5. Conclusions
The quenching and blowoff limits for hydrogen, methane, and propane have been
measured for small round burners. Materials degradation of exposure to hydrogen
and methane diffusion flames was observed. The conclusions of this study are:
The measured fuel mass flow rate at the quenching limits is independent of burner
diameter. This is consistent with a simple scaling analysis based on a premixed
flame quenching distance.
Hydrogen has a lower mass flowrate at quenching and a higher mass flowrate at
blowoff than either methane or propane.

7
Hydrogen flames caused faster corrosion than methane flames on aluminum alloy
1100, galvanized steel, and SiC filaments.

6. Acknowledgements
This work was supported by NIST grant 60NANB5D1209 under the technical
management of J. Yang. The assistance of K.B. Lim, V. Lecoustre, and C. Moran
is appreciated.

7. References
[1] Cadwallader, L.C. and Herring, J.S., Safety Issues with Hydrogen as a Vehicle Fuel,
Report INEEL/EXT-99-00522, prepared for U.S. Department of Energy (1999).
[2] Matta, L.M., Neumeier, Y., Lemon, B., Zinn, B.T., Characteristics of Microscale
Diffusion Flames. Proceedings of the Combustion Institute v. 29. n.1, p 933-938 (2002).
[3] Cheng, T.S., Chen, C.P., Chen, C.S., Li, Y.H., Wu, C.Y., Chao, Y.C., Characteristics of
Microjet Methane Diffusion Flames. Combustion Theory and Modeling v.10, n. 5, p.
861-881 (2006).
[4] Swain, M.R., Swain, M.N., A Comparison of H2, CH4, and C3H8 Fuel Leakage in
Residential Settings, Int. J Hydrogen Energy 17:807-815 (1992).
[5] Khan, R., Kan, Z., Al-Sulaiman F., Merah, N., Fatigue Life Estimates in Woven Carbon
Fabric/Epoxy Composites at Non-Ambient Temperatures. Journal of Composite
Materials Vol. 36, No. 22, p. 2517-2535 (2002).
[6] Pehr, K. Aspects of Safety and Acceptance of LH2 Tank Systems in Passenger Cars. Int.
J. Hydrogen Energy Vol. 21, No. 5, pp. 387-395 (2006).
[7] Utgikar, V.P., Thiesen, T., Safety of compressed hydrogen fuel tanks: Leakage from
stationary vehicles. Technology in Society 27:315-320 (2005).
[8] Roper, F.G., The Prediction of Laminar Jet Diffusion Flame Sizes: Part I. Theoretical
Model. Combust. Flame 29:219-226 (1977).
[9] Sunderland, P.B., Mendelson, B.J., Yuan, Z.-G., Urban, D.L., Shapes of Buoyant and
Nonbuoyant Laminar Jet Diffusion Flames, Combust. Flame 116:376-386 (1999).
[10] Kanury, A.M., Introduction to Combustion Phenomena. Gordon and Breach, New York,
p. 131 (1975).
[11] Weast, R.C., Astle, M.J., CRC Handbook of Chemistry and Physics, 59th edition, CRC
Press, Inc. West Palm Beach, p. F-58 (1979).
[12] Munson, B.R., Young, D.F., Okiishi, T.H., Fundamentals of Fluid Mechanics, Wiley,
Hoboken, NJ, p. 359 (2002).
[13] Odegard, Ben C., Thomas, George J., Testing of High Pressure Hydrogen Composite
Tanks. Proceedings of the 2001 DOE Hydrogen Program Review. Livermore, CA, USA.
[14] Bahadori, M. Y., Hegde, U., Zhou, L., Liftoff and Blowoff of Jet Diffusion Flames, Paper
AIAA-2002-1078, Aerospace Sciences Meeting, Reno, Nevada, January 14-17, 2002.

8
 2007-01-0429

Fire Hazards of Small Hydrogen Leaks

N.R. Morton, P.B. Sunderland
University of Maryland

B.H. Chao
University of Hawaii

R.L. Axelbaum
Washington University

Copyright © 2007 SAE International

ABSTRACT hydrogen system to perform a cost-benefit analysis of

protection against leaks.
This study examines the types of hydrogen leaks that can
support combustion and the effects on stainless steel of Research has been done in quantifying leak flow rates,
long term hydrogen flame exposure. Experimental and comparing hydrogen to methane and propane. Swain
analytical work is presented. Hydrogen diffusion flames and Swain (1992) modeled and measured leak rates for
on round burners were observed. Measurements diffusion, laminar, and turbulent flow regimes. They found
included limits of quenching, blowoff, and piloted ignition that combustible mixtures in an enclosed space resulted
for burners with diameters of 0.36 - 1.78 mm. Results are more quickly for propane and hydrogen leaks than for
compared to measurements for methane and propane. A methane leaks. However their supply pressures were the
dimensionless crack parameter was identified to correlate same for all fuels and thus did not reflect plans for
the quenching limit measurements. Flow rates were hydrogen systems in vehicles with pressures of up to 700
0.019 - 40 mg/s for hydrogen, 0.12 - 64 mg/s for bar.
methane, and 0.03 - 220 mg/s for propane. Hydrogen
flames were found to be corrosive to 316 stainless steel The present research seeks to determine the relative fire
tubing. hazards of small hydrogen leaks compared to those of
methane and propane. The modeling and
INTRODUCTION experimentation focus primarily on small burners and
flames near the quenching limit. Experiments also
Concerns about the emissions of greenhouse gases consider the corrosive effects associated with the
have led to extensive consideration of hydrogen as an exposure of 316 stainless steel to hydrogen and methane
energy carrier. Hydrogen presents several unusual fire flames.
hazards, including high leak propensity, ease of ignition,
and invisible flames. This research concerns experiments Thus motivated, the objectives of this work are to (1)
and analysis to identify which hydrogen leaks can measure limits of flaming (at ignition, quenching and
support flames. A small leak in a hydrogen system could blowoff) for hydrogen, methane, and propane issuing
ignite easily, support a flame that is difficult to detect, and from circular burners of various sizes, and (2) examine
degrade containment materials to the point of a material degradation arising from hydrogen and methane
catastrophic failure. Hydrogen leaks can develop in diffusion flames.
pressure vessels, piping, seals, valves, pressure
regulators, and pressure relief devices. FLAME QUENCHING THEORY

A Department of Energy report (Cadwallader and A theoretical model was developed to predict flame
Herring, 1999) found that hydrogen containment was the quenching limits. These limits are the minimum flow rates
chief safety concern associated with using hydrogen as a sufficient to support a diffusion flame. This theory also
transportation fuel. This report documents several yields a dimensionless crack parameter that indicates
catastrophic hydrogen fires. how close a given leak is to the quenching limit.

Research in hydrogen combustion has increased The stoichiometric length Lf of a laminar gas jet diffusion
recently, but no study to date has characterized the types flames on a round burner is:
of hydrogen leaks that can support a flame. Absent such
information, it may be difficult for the designers of a Lf / d = a Re = a ρ u0 d / μ , (1)
where d is burner inside diameter, a is a dimensionless EXPERIMENTAL
fuel-specific empirical constant, Re is Reynolds number,
u0 is the average fuel velocity in the burner, ρ is fuel Two types of measurements were made. Quenching and
density, and μ is fuel dynamic viscosity. The scaling of blowoff limits of small-scale hydrogen, methane, and
Eq. (1) arises from many theoretical and experimental propane flames were measured. These tests involved
studies, including Roper (1977), Sunderland et al. (1996), five hemispherical burners of different diameters.
and references cited therein. Constant a here is assigned Materials degradation tests were also performed, using
values measured by Sunderland et al. (1996), as listed in tube burners. Flowrates were measured with calibrated
Table 1. rotameters.
Table 1: Selected fuel properties of hydrogen, methane, and propane. The quenching and blowoff limit burners are
Values for a are from Sunderland et al. (1996), Lq and SL are from
Kanury (1975), and μ is from Weast and Astle (1979). hemispherical stainless steel nozzles that are
manufactured for spray generation. At the apex of the
hemisphere is a drilled hole of the specified diameter.
Fuel a Lq [mm] SL [cm/s] μ [g/m-s]
H2 0.236 0.51 291 8.76e-3 Fuel flow was commenced and ignited, creating a flame
CH4 0.136 2.3 37.3 1.09e-2 approximately 5 mm in size. The flow was then reduced
C3H8 0.108 1.78 42.9 7.95e-3 until the flame extinguished. This was done several times
for each burner and each fuel. The flames were small
enough, and the experiment was done quickly enough,
The base of an attached jet diffusion flame is quenched that there was no noticeable increase in the temperature
by the burner. Its standoff distance can be approximated of the stainless steel burners.
as one half of the quenching distance of a stoichiometric
premixed flame. Such quenching distances typically are Inverted burns also were performed, in which the jet
reported as the minimum tube diameter, Lq , through direction was downward. Hydrogen performed essentially
which a premixed flame can pass. It is assumed here that the same; the quench limit was largely independent of
a jet flame can be supported only if its stoichiometric burner orientation. Methane required less fuel to sustain
length is greater than half this quenching distance: a flame in the inverted position, and propane required a
significantly larger flowrate to sustain an inverted flame.
Lf ≥ Lq / 2 to support a flame. (2)
Also measured were the blowoff flows of each fuel for
each burner. Blowoff is achieved when the flammable
Measurements of Lq , shown in Table 1, are taken from regions flow faster than the laminar flame speed, which is
Kanury (1975). When combined, Eqs. (1) and (2) predict shown in Table 1. Blowoff limits were measured by
the following fuel flowrate, mfuel, at the quenching limit: igniting a flow of fuel and then increasing the flow rate
until the flame lifted off and extinguished.
mfuel = π ρ u0 d2 / 4 = π Lq μ / ( 8 a ) . (3)
Tests were also conducted to determine the corrosive
Eq. (3) indicates that the fuel mass flow rate at the effects of these flames on 316 stainless steel. For these
quenching limit is a fuel property that is independent of tests the flames considered were those of Fig. 1.
burner diameter.

A crack parameter can now be derived. Assuming fully-

developed laminar flow in the burner,
15 mm

u0 = d2 ∆p / (32 μ Lb ) , (4)

where Δp is the pressure drop across the burner and Lb

is the burner flow passage length (Munson et al., 2002).

Combining Eqs. (1), (2), and (4) yields

CP = a ρ d4 ∆p / (16 μ2 Lb Lq) ≥ 1 to support a flame, (5)

where CP is the dimensionless crack parameter. A B

Figure 1: Color images of a representative hydrogen flame (A) and

methane flame (B).
RESULTS by using different published values of quenching
distances of premixed flames.
Typical hydrocarbon flames burn much brighter than do
hydrogen flames, as shown in Fig. 1. Fig. 2 demonstrates the important finding that for each
fuel there is a critical mass flowrate below which
The results of the quenching and blowoff studies are combustion is impossible. Using these results, given the
presented in Figs. 2 and 3. These figures also include the flowrate of a leak, it is readily known whether the leak is
predictions of quenching limits from Eq. (3). Figure 2 flammable.
shows the fuel mass flow rate versus diameter, while
Figure 3 shows fuel velocity versus diameter for the same The blowoff data show that methane will reach blowoff at
tests. the lowest mass flowrate, propane next, and hydrogen
will require the most mass flow to achieve blowoff. These
1000 observations are qualitatively supported by the laminar
Fuel Quenching Blowoff flame speeds shown in Table 1.
H2 □ ■
CH4 Δ ▲
Figure 3 shows the same measurements of Fig. 2 where
C3H8 % &
the ordinate is changed to fuel velocity. This figure
100
suggests a regime may exist at the smallest burner
diameters where the blowoff limit is lower than the
quenching limit. Burners smaller than those considered
H2 CH4 here will need to be tested to further evaluate this.
10

C3H8 1000
H2
m fuel (mg/s)

1
100

CH4
0.1 CH4

C3H8 10 C3H8

0.01 H2
u0 (m/s)

CH4
0.001
0.0 0.5 1.0 1.5 2.0 0.1
H2
d (mm)

C3H8
Figure 2: Measured fuel mass flowrate at the quenching and blowoff
limits versus burner diameter. The lower set of three lines are the
0.01
quenching limit theory of Eq. (3) for the fuels as shown. The upper set of Fuel Quenching Blowoff
three lines are the best fits of the blowoff measurements for the fuels as H2 □ ■
shown. CH4 Δ ▲
C3H8 % &

0.001
0.0 0.5 1.0 1.5 2.0
Figure 2 shows that burner mass flowrate at the d (mm)
quenching limit is independent of burner diameter. This
finding is supported by the present theory. Hydrogen
Figure 3: Measured fuel velocity at the quenching and blowoff limits
requires the smallest mass flowrate, which is expected
versus burner diameter. The curves shown are the quenching limit
given the wide flammability limits of hydrogen compared theory of Eq. (3) for the fuels as shown. The lines are the best fits of the
to methane and propane. Propane requires slightly blowoff measurements for the fuels as shown.
higher mass flow rates, and methane requires the
highest. The theory of Eq. (3) captures the trends of the
quenching experiments, but its predicted quenching limits
do not agree very well with measurements except for
methane. It may be possible to improve this agreement
MATERIAL DEGRADATION additional corrosion on the hydrogen burner is believed to
arise because hydrogen flames have relatively short
Two stainless steel burners were fabricated from 316 standoff distances and thus result in increased material
stainless steel. One supported a hydrogen flame and the temperatures.
other a methane flame, as shown in Fig. 1. Figure 4
shows the burners in their pre-test conditions. CONCLUSIONS

The two flames were burned for 355 hours. The flames The quenching and blowoff limits for hydrogen, methane,
burned continuously except for brief periods when and propane have been modeled and measured for small
images were recorded. Figure 5 shows the burners after round burners. The fuel mass flow rate at the quenching
the test. limits was found to be independent of burner diameter.
The fuel mass flow rate at blowoff was found to be
considerably higher for hydrogen than for methane or
propane.

Hydrogen flames were found to cause more corrosion of

316 stainless steel than methane for similar exposure
5 mm

conditions.

ACKNOWLEDGMENTS

This work was supported by NIST grant 60NANB5D1209

under the technical management of J. Yang. The
assistance of K.B. Lim, V. Lecoustre, and C. Moran is
appreciated.

A B REFERENCES

Cadwallader, L.C. and Herring, J.S., Safety Issues with

Hydrogen as a Vehicle Fuel, Report INEEL/EXT-99-
Figure 4: Color image of 316 stainless steel tube burners before the 00522, prepared for U.S. Department of Energy (1999).
burn test. Burners A and B are for hydrogen and methane respectively.
Kanury, A.M., Introduction to Combustion Phenomena.
Gordon and Breach, New York, p. 131 (1975).

Munson, B.R., Young, D.F., Okiishi, T.H., Fundamentals

of Fluid Mechanics, Wiley, Hoboken, NJ, p. 359 (2002).

Roper, F.G., The Prediction of Laminar Jet Diffusion

5 mm

Flame Sizes: Part I. Theoretical Model. Combust. Flame

29:219-226 (1977).

Sunderland, P.B., Mendelson, B.J., Yuan, Z.-G., Urban,

D.L., Shapes of Buoyant and Nonbuoyant Laminar Jet
Diffusion Flames, Combust. Flame 116:376-386 (1999).

Swain, M.R., Swain, M.N., A Comparison of H2, CH4, and

C3H8 Fuel Leakage in Residential Settings, Int. J
Hydrogen Energy 17:807-815 (1992).
A B
Weast, R.C., Astle, M.J., CRC Handbook of Chemistry
and Physics, 59th edition, CRC Press, Inc. West Palm
Figure 5: Color image of the burners of Fig. 4 following 355 hours of
Beach, p. F-58 (1979).
burning. Burners A and B supported hydrogen and methane flames
respectively.
CONTACT

The corresponding author for this work is Peter B.

At the end of the test, there was noticeably more Sunderland, [email protected], www.enfp.umd.edu, (301)
corrosion on the hydrogen burner, see Fig. 5. The 405-3095.
 Combustion Theory and Modelling
Vol. 12, No. 6, 2008, 1179–1196

A theoretical study of spontaneous ignition of fuel jets in an oxidizing

ambient with emphasis on hydrogen jets
K.B. Lima , B.H. Chaob∗ , P.B. Sunderlanda , and R.L. Axelbaumc
a
Department of Fire Protection Engineering, University of Maryland, College Park, MD 20742, USA;
b
Department of Mechanical Engineering, University of Hawaii at Manoa, Honolulu, HI 96822, USA;
c
Department of Energy, Environmental and Chemical Engineering, Washington University, St. Louis,
MO 63130, USA
(Received 4 October 2007; final version received 23 April 2008)

An analysis was performed for the spontaneous ignition of a hydrogen (or other gaseous fuel)
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

jet emanating from a slot into an oxidizing ambient (e.g., air). A similarity solution of the flow
field was obtained. This was combined with the species and energy conservation equations,
which were solved using activation energy asymptotics. Limits of spontaneous ignition were
identified as functions of slot width, flow rate, and temperatures of the hydrogen jet and ambient
gas. Two scenarios are examined: a cool jet flowing into a hot ambient and a hot jet flowing into
a cool ambient. For both scenarios, ignition is favored with an increase of either the ambient
temperature or the hydrogen supply temperature. Moreover, for the hot ambient scenario, a
decrease in fuel Lewis number also promotes ignition. The Lewis number of the oxidizer only
has a weak effect on ignition. Because spontaneous ignition is very sensitive to temperature,
ignition is expected to occur near the edge of the jet if the hydrogen is cooler than the ambient
gas and near the centerline if the hydrogen is hotter than the ambient gas.
Keywords: Spontaneous ignition; Hydrogen jets; Activation energy asymptotics; Lewis num-
bers; Ignition location

Nomenclature
aT constant representing the temperature increase through reaction
aF constant representing the fuel consumption through reaction
B pre-exponential factor
cp specific heat at constant pressure
D mass diffusion coefficient
Da Damköhler number
D̃a reduced Damköhler number
D̄a rescaled reduced Damköhler number defined in Eq. (44)
E activation temperature
f nondimensional streamfunction
h half width of the slot
Le Lewis number
n reaction order

∗
Corresponding author. E-mail: [email protected]

ISSN: 1364-7830 print / 1741-3559 online

C 2008 Taylor & Francis
DOI: 10.1080/13647830802315095
http://www.tandf.co.uk/journals
 1180 K.B. Lim et al.

p pressure
Pr Prandtl number
qF heat of combustion per unit mass of fuel
R ideal gas constant
Sc Schmidt number
T temperature
T̂ reference temperature
u flow velocity in the x (streamwise) direction
v flow velocity in the y (transverse) direction
W molecular weight
x streamwise spatial coordinate
x0 distance from the virtual origin of the jet to the exit of the slot
y transverse spatial coordinate (y= 0 along the centerline)
Y mass fraction
Z mixture fraction
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Greek symbols
α parameter defined as α = ( T̃∞ − T̃0 )/ỸF,0
β parameter defined as β = ( T̃0 − T̃∞ )/ỸO,∞
ε small parameter used for asymptotic expansion
φ perturbation of species concentration in the reaction region
θ perturbation of temperature in the reaction region
λ thermal conductivity
µ viscosity
ν stoichiometric coefficient
ρ gas density
ψ streamfunction
η similarity variable
σ stoichiometric oxidizer to fuel mass ratio
ξ stretched spatial coordinate in the reaction region along the streamwise direction for the hot
jet scenario
ζ stretched spatial coordinate in the reaction region along the transverse direction

Subscripts
0 value of variables at the exit of the slot
F fuel
f frozen solution
I ignition state
O oxidizer
∞ ambient condition

Superscripts
∼ nondimensional quantity
∧ rescaled nondimensional quantity based on T̂
 Combustion Theory and Modelling 1181

1. Introduction
Concerns about the emissions of greenhouse gases and supply of oil have led to extensive
consideration of hydrogen as a major fuel carrier. Nonetheless, hydrogen presents several unusual
fire hazards, including high leak propensity, ease of ignition, and invisible flames. For example,
heated air jets flowing into hydrogen will ignite spontaneously at an air temperature of 913 K [1].
This is cooler than for other fuels [1, 2], including gasoline and methane, and is not much higher
than the autoignition temperature of stoichiometric hydrogen/air mixtures, 858 K [3]. Occasional
unintended hydrogen leaks will be unavoidable, and some may involve heated hydrogen and/or
air. Thus an improved understanding of limits of spontaneous ignition of hydrogen jets is sought
here, with the aid of activation-energy asymptotics.
Zheng and Law [4] identified ignition limits of premixed hydrogen–air flames where ignition
was by a heated counterflowing stream. The ignition limits of non-premixed hydrogen–air jets
will be different from those of premixed counterflowing streams because of the non-premixed
nature and the absence of strain due to counterflow heating.
Thermal ignition in subsonic [5] and supersonic [6, 7] mixing layers has been studied by
activation energy asymptotics and numerically with one-step reaction as well as reduced mecha-
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

nisms. Im et al. [6] employed a reduced mechanism to analyze thermal ignition in a supersonic
hydrogen/air mixing layer and obtained ignition characteristics over a wide range of conditions.
These findings are particularly applicable to scramjets. Lee and Chung [8] investigated the dif-
ferent combustion regimes in a stagnant mixing layer by employing Damköhler-number and
activation-energy asymptotics, based on an eight-step reduced mechanism. Law and co-workers
[9, 10] investigated the ignition of hydrogen and air in a mixing layer by using reduced reaction
mechanisms that they developed.
Toro et al. [11] examined in detail the structure of laminar hydrogen jet flames both experi-
mentally and numerically. Chaos et al. [12] examined Lewis-number effects in unsteady laminar
hydrogen jet flames, which will have different effects compared to steady laminar hydrogen jet
flames. Liu and Pei [13] examined autoignition and explosion limits of hydrogen–oxygen mix-
tures in homogeneous systems, which involved reduced mechanisms. Dryer et al. [14] examined
spontaneous ignition of pressurized releases of hydrogen and natural gas into air. This investiga-
tion involved multi-dimensional transient flows with shock formation, reflection and interactions
that resulted in the transition to turbulent jet nonpremixed combustion. This is a different aspect
of risk, associated with rapid failures of pressure vessels, as compared to the scenario being con-
sidered here, which involves small leaks/cracks that are undetected, and that ignite spontaneously
when the limits are reached.
The present study applies asymptotic analysis to investigate the spontaneous ignition of a
laminar jet of hydrogen, or another gaseous fuel, leaking through a crack into an oxidizing
ambient. Asymptotic flame theories can provide valuable insights into combustion reactions [15–
18]. Quantitative and predictive derivations can be made using the concept of distinguished limits
in activation energy asymptotics. Based largely on the concept of Zel’dovich number, asymptotic
analysis enables evaluation of temperature effects on reaction rates despite the narrowness of the
reaction zone relative to the transport zones of the flame structure.
The crack is taken to be straight and long, yielding a two-dimensional flow field. The ignition
analysis identifies limits of spontaneous ignition. Unfortunately, comparisons with experiments
are not possible because no study to date of spontaneous ignition of nonpremixed jets has measured
the key phenomena analyzed here.
The objectives of this work are to: (1) develop a model of spontaneous ignition for two cases:
a cool fuel jet flowing into heated ambient gas and a heated fuel jet flowing into cool ambient
gas; (2) identify limits of spontaneous ignition as functions of slot width, flow rate, fuel Lewis
 1182 K.B. Lim et al.

number, and temperatures of the fuel jet and the ambient gas; and (3) identify the location of
ignition.

2. Formulation
The problem of interest is a steady, isobaric laminar jet of fuel (e.g. hydrogen) at temperature
T0 issuing from a rectangular slot into an oxidizing environment (e.g. atmospheric air) at a
temperature of T∞ , as shown schematically in Figure 1. Spontaneous ignition occurs when either
T0 or T∞ is sufficiently high that the weak reaction between the fuel and the oxidizer transitions to
a vigorous burning flame. This study analyzes the ignition state as a function of various physical
properties including Lewis number, T0 , T∞ , the flow velocity at the slot exit, u0 , and the width
of the slot. The slot is considered sufficiently long that end effects are negligible. The reaction
chemistry is assumed to follow a single-step, overall, irreversible reaction with second order
Arrhenius kinetics and a high activation energy.
With the above problem definition, conservation of mass and momentum are given by
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

∂ (ρ u) ∂ (ρ v)
+ = 0, (1)
∂x ∂y


∂u ∂u ∂ ∂u
ρu +ρv − µ = 0, (2)
∂x ∂y ∂y ∂y

Figure 1. Schematic of slot and fuel (e.g. hydrogen) leak for the (a) hot ambient and (b) hot jet cases.
 Combustion Theory and Modelling 1183

which are to be solved subject to the boundary conditions

y = 0, x = x0 : u = u0 , v = 0; x > x0 : ∂u/∂y = 0, v = 0, (3)

y → ∞ : u → 0. (4)

Introducing a streamfunction ψ of the form

ρ u = ρ∞ u0 (∂ ψ/∂ y ), ρ v = − ρ∞ u0 (∂ ψ/∂ x ), (5)

such that Eq. (1) is satisfied, transforming the coordinates from (x, y) to (x̃, η) where
 y
−2/3
x̃ = x/x0 , η = [ ρ∞ u0 /( 6 µ∞ x0 ) ] 1/2
x̃ (ρ/ρ∞ )dy, (6)
0

and defining
ψ = [ 6 µ∞ x0 /( ρ∞ u0 ) ]1/2 x̃ 1/3 f ( η ), (7)
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Equations (2)–(4) are transformed to

( f  /2 ) + f f  + f 2 = 0, (8)
  
η = 0: f = f = 0, f = 1; η → ∞ : f → 0. (9)

The notations used in this study are listed in the Nomenclature section. A similarity solution is
assumed to exist so that f is a function of η only. Solving Equations (8)–(9) following Bickey [19]
and Schlichting [20] yields f = tanh η. In these equations, x0 is the value of x at the slot exit from
the virtual origin of the jet, which can be determined by the conservation of the x-momentum
across the slot exit, given by
 2
h  
x0 = 3 ρ02 (ux=x0− )2 dy / 32u30 ρ∞ µ∞ . (10)
−h

Applying the coordinate transformation and the solution of the momentum equation to the energy
and species conservation equations, we obtain

1 ∂ 2 T̃ ∂ T̃ ∂ T̃
2
+ 2 ( tanh η ) − 6 ( sech2 η ) x̃ = − Da x̃ 4/3 ỸF ỸO T̃ −1 exp ( −Ẽ/T̃ ), (11)
Pr ∂ η ∂η ∂ x̃
1 ∂ 2 ỸF ∂ ỸF ∂ ỸF
+ 2 ( tanh η ) − 6 ( sech2 η ) x̃ = Da x̃ 4/3 ỸF ỸO T̃ −1 exp ( −Ẽ/T̃ ), (12)
ScF ∂ η2 ∂η ∂ x̃
1 ∂ 2 ỸO ∂ ỸO ∂ ỸO
2
+ 2 ( tanh η ) − 6 ( sech2 η ) x̃ = Da x̃ 4/3 ỸF ỸO T̃ −1 exp ( −Ẽ/T̃ ), (13)
ScO ∂ η ∂η ∂ x̃

which are to be solved subject to

η = 0, x̃ = 1 : T̃ = T̃0 , ỸF = ỸF,0 , ỸO = 0, (14)

η = 0, x̃ > 1 : ∂ T̃ /∂ η = ∂ ỸF /∂ η = ∂ ỸO /∂ η = 0, (15)
η→∞: T̃ → T̃∞ , ỸF → 0, ỸO → ỸO,∞ . (16)
 1184 K.B. Lim et al.

In the above,

cp T YO 6 x0 νO cp p B cp E
T̃ = , ỸF = YF , ỸO = , Da = , Ẽ = ,
qF σ WF qF R u0 qF
µ µ λ/cp νO WO
Pr = , Scj = , Lej = ,σ = .
λ/cp ρ Dj ρ Dj νF WF

The values of cp, ρ λ, ρ µ, and ρ 2 Dj are considered constant. The ideal gas equation of state has
been adopted in the derivation of Equations (11)–(13).
In the non-reactive limit, solving Equations (11)–(13), but without the reaction terms, subject
to Equations (14)–(16) gives the frozen solutions,

T̃f = T̃∞ + (T̃0 − T̃∞ ) ( sech2 Pr η ) / x̃ 1/3 , (17)

ỸF,f = ỸF,0 (sech2 ScF η )/x̃ 1/3 , (18)
ỸO,f = ỸO,∞ [1 − (sech
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

2 ScO 1/3
η )/x̃ ]. (19)

2.1. Cool jet flowing into a hot ambient (T∞ >T0 )

In the presence of a weak reaction, the temperature is increased from its frozen value by a small,
O(ε) amount where ε = T̃∞ 2
/Ẽ while the reactant concentrations are reduced from their respective
frozen values by an O(ε) amount. Because of the high activation energy, ignition is primarily
controlled by temperature and occurs near η → ∞ if successful. Away from this high temperature
region, the reaction is negligible and can be considered frozen. In the outer, chemically frozen
region, the solutions are similar to Equations (17)–(19) but with an O(ε) change in their values.
With the application of Equations (14) and (15), the solutions of T̃ and ỸF are given by

T̃ = ( T̃∞ + ε aT ) − [ ( T̃∞ − T̃0 ) + ε aT ] ( sech2 Pr η ) / x̃ 1/3 + O ( ε2 ), (20)

ỸF = ε aF + ( ỸF,0 − ε aF ) ( sech 2 ScF
η ) / x̃ 1/3
+ O ( ε ). 2
(21)

The solution of ỸO is not important to the analysis.

In the inner, reactive region, a stretched spatial inner variable is defined as

ζ = ỸF,0 ( sech η )2Pr /( ε x̃ 1/3 ), (22)

while the solutions are expanded as

T̃ = T̃∞ + ε ( θ − α ζ ) + O ( ε2 ), (23)
1/3 1 − LeF
ỸF = ε LeF
( ỸF,0 /x̃ ) ζ LeF
+ O ( ε ), (24)
ỸO = ỸO,∞ + O ( ε ). (25)

Equation (24) is obtained by considering LeF to be sufficiently smaller than unity, as for hydrogen.
Substitution of Equations (22)–(25) into Equation (11) and expanding in orders of ε yield

ζ 2 ( ∂ 2 θ/∂ ζ 2 ) = − D̃a ζ LeF exp ( θ − α ζ ), (26)

 Combustion Theory and Modelling 1185

where α = ( T̃∞ − T̃0 )/ỸF,0 is the parameter indicating the effect of temperature difference and
D̃a is the reduced Damköhler number defined as

D̃a = Da ỸO,∞ x̃ 4/3 [ ỸF,0 /( ε x̃ 1/3 ) ]1 − LeF exp ( −Ẽ/T̃∞ )/( 4 Pr T̃∞ ). (27)

The boundary conditions required to solve this equation can be found by applying Equations (22)
and (23) to Equation (16), and matching the inner and outer solutions to yield

ζ = 0: θ = 0; ζ → ∞ : ∂ θ/∂ ζ → 0, θ → aT . (28)

For the case of LeF close to unity, as for most of gaseous fuels, Equation (24) is modified to

ỸF = ε (φF + ζ ) + O (ε2 ), (29)

Substitution of Equations (22), (23), (25) and (29) into Equations (11) and (12) gives
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

ζ 2 ( ∂ 2 θ/∂ ζ 2 ) = − D̃a ( φF + ζ ) exp ( θ − α ζ ), (30)

∂ ( θ + φF )/∂ ζ = 0,
2 2
(31)

where D̃a is given by Equation (27) with LeF = 1. Applying the boundary conditions of
Equation (16) and the conditions obtained by matching the inner and outer solutions as before,
we obtain Equation (28) and φ F = – θ , which can be applied to Equation (30) to yield

ζ 2 ( ∂ 2 θ/∂ ζ 2 ) = − D̃a ( ζ − θ ) exp ( θ − α ζ ). (32)

2.2. Hot jet flowing into a cool ambient (T0 >T∞ )

For the case of a hot jet issuing into a cold ambient, any ignition will occur near the jet centerline,
η = 0. Moreover, because the jet will be cooled by the cold ambient gas along the flow, ignition
is expected to occur near the slot exit. The analysis is similar to that in Section 2.1, except that
ε = T̃02 /Ẽ, the inner spatial variables are defined as

ζ = η ( P r ỸO,∞ /ε )1/2 , ξ = ỸO,∞ ( x̃ − 1 )/(3 ε), (33)

and the inner solutions are expanded as

T̃ = T̃0 + ε [θ − β (ζ 2 + ξ )] + O (ε2 ), (34)

ỸF = ỸF,0 + O ( ε ), (35)
ỸO = ε (φO + LeO ζ + ξ ) + O (ε ).
2 2
(36)

Substitution of Equations (33)–(36) into Equations (11) and (13) leads to

∂2 θ ∂θ 1 ∂ 2 φO ∂ φO
2
− 2 + 2
−2 = 0, (37)
∂ζ ∂ξ LeO ∂ ζ ∂ξ
∂2 θ ∂θ
2
−2 = − D̃a ( φO + LeO ζ 2 + ξ ) exp [ θ − β ( ζ 2 + ξ ) ], (38)
∂ζ ∂ξ
 1186 K.B. Lim et al.

with the initial and boundary conditions

ξ = 0: θ = φO = 0, (39)
ζ = 0 and ζ → ∞, ξ > 0: ∂ θ/∂ ζ = ∂ φO /∂ ζ = 0, (40)

where β = ( T̃0 − T̃∞ )/ỸO,∞ and D̃a is

D̃a = ε [Da ỸF,0 /(T̃0 ỸO,∞ )] exp (−Ẽ/T̃0 ). (41)

Ignition is considered successful when the heat generation through reaction is sufficient to
compensate for the heat loss from the jet to the ambient at any location, i.e. when the reaction is
self-sustainable. The ignition criterion is then defined as

( ∂ θ/∂ ξ )ζ = 0 ≥ β or ∂ θ/∂ ζ ≥ 2 β ζ at any ξ . (42)

Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

The smallest value of ξ that satisfies either of these conditions represents the first point that
thermal runaway would occur and is identified as the location of ignition.

3. Results and Discussion

3.1. Cool jet flowing into a hot ambient (T∞ >T0 )
For this case, Equation (26) subject to Equation (28) was solved by a fourth order Runge-Kutta
method. The results are shown in Figure 2, a plot of the reaction temperature increase, aT , versus
reduced Damköhler number, D̃a, for selected values of α = ( T̃∞ − T̃0 )/ỸF,0 and LeF = 0.6.
This reveals the lower and middle branches of an S-shaped ignition/extinction curve [3,15,16].
In each such curve, there is a maximum value of D̃a above which a solution does not exist.
For values of D̃a smaller than this critical value, there are two solutions for each D̃a. The
critical value of D̃a represents the transition from weak reaction to vigorous burning, and is
defined as the ignition state, D̃aI . The lower branch, showing an increase of temperature with
higher reaction rate, is the physically realistic branch. The middle branch represents conditions
that are not physically possible. Spontaneous ignition is successful for any D̃a greater than
D̃aI .
Figure 2 indicates that a decrease in α reduces D̃aI , which means that ignition is favored.
Such a decrease can be accomplished by increasing the jet temperature, T̃0 , qF /cp . For these cases,
reaction is enhanced when α is reduced such that it is easier to ignite the reactants. A fuel such
as hydrogen (which has qF /cp of roughly 78 600 K in air), is hence more ignitable as compared
to fuels such as methane (qF /cp = 36 800 K) and isooctane (qF /cp = 33 800 K) when the flow
conditions are the same. As an example, for YF,0 =1, T0 = 300 K and T∞ = 1000 K, the values of
α are 0.0089, 0.019 and 0.021 for hydrogen, methane and isooctane, respectively.
Parameter α also can be changed by variations in the ambient temperature, T̃∞ , or the reactant
mass fraction in the fuel supply, ỸF,0 , but these changes result in a simultaneous change of D̃a
as can be seen from Equation (27). To investigate the effects of T̃∞ at fixed D̃a, a rescaling
is required. The rescaling is performed by specifying a reference value of T̃∞ as T̂∞ , defining
rescaled parameters ε̂ = T̂∞2
/Ẽ, âT = ( T̃∞ /T̂∞ )2 aT and

D̂a = Da ỸO,∞ x̃ 4/3 [ỸF,0 /(ε̂ x̃ 1/3 )]( 1 − LeF ) exp (−Ẽ/T̂∞ )/(4 Pr T̂∞ ). (43)
 Combustion Theory and Modelling 1187
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Figure 2. Reaction temperature increase aT versus reduced Damköhler number D̃a for selected values of
α = ( T̃∞ − T̃0 )/ỸF,0 with LeF = 0.6. The plot is for a cool jet flowing into a hot ambient.

and plotting the results in terms of rescaled variables âT and D̂a. In this section, T̃0 = 0.00382,
Ẽ = 0.127, ỸF,0 = 1 and T̂∞ = 0.01382 were used to exhibit the ignition behavior. These data
correspond to T0 = 300 K, E = 10 000 K, T∞,ref = 1090 K for hydrogen. The rescaled results
are shown in Figure 3. Here an increase in T̃∞ , which increases α without changing D̂a, is seen
to favor ignition. This also is physically realistic because more heat is transferred to the cold fuel

Figure 3. Rescaled plot of âT versus D̂a from Figure 2 when changes of α is caused by variations in T̃∞ ,
with T̃0 = 0.00382, Ẽ = 0.127, ỸF,0 = 1, and T̂∞ = 0.01382.
 1188 K.B. Lim et al.

flow at a higher rate when the ambient is at a higher temperature. For the same reason, when the
kinetic data are unchanged, an increased T̃∞ yields ignition closer to the exit of the slot. Note that
the reduced Damköhler number shown in Equation (27) is a function of the axial distance from
the virtual origin of the jet.
When the variation of α is caused by the change of ỸF,0 , a similar rescaling of D̃a by defining
a reference value of ỸF,0 is necessary. The result is qualitatively the same as that presented in
Figure 2, which is expected because an increase in ỸF,0 (a decrease in α) results in a greater fuel
concentration in the reaction region and a higher reaction rate.
To investigate the influence of fuel Lewis number, LeF , on ignition, a reference fuel Lewis
number is necessary. As can be seen after Equation (16), LeF is defined as the mixture thermal
diffusivity divided by the mass diffusivity of fuel into the mixture. Specifying LeF = 1 as the
reference LeF , we define a second rescaled reduced Damköhler number from Equation (43) as

D̄a = Da ỸO,∞ x̃ 4/3 exp (−Ẽ/T̂∞ )/(4 Pr T̂∞ ). (44)

Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

The results for α = 0.02 are presented in Figure 4 by plotting aT versus D̄a. The curve for
LeF = 1 was obtained by solving Equation (32) subject to Equation (28). For demonstration
purpose, a value of x̃ = 2 also is specified whenever Equation (44) is used. Only quantitative
differences exist when another value of x̃ is adopted. Figure 4 shows that a decrease in LeF for
fixed D̄a favors ignition. This occurs because a smaller LeF implies that fuel species diffuse more
quickly into the hot oxidizer. A fuel such as hydrogen, which has LeF of 0.6 or less in mixtures
of nitrogen, is hence more ignitable as compared to fuels such as isooctane and methane, which
have higher LeF , under the same conditions. Nayagam and Williams [21] found that in a one-
dimensional model of steady motion of edges of reaction sheets, increasing the Lewis number
decreases the propagation velocity at small Damköhler numbers. This indicates that the reaction
rate is raised when LeF is reduced, which agrees with the findings of Figure 4.

Figure 4. Reaction temperature increase aT versus reduced Damköhler number D̄a for selected values of
LeF , with α = 0.02, and x̃ = 2, cool jet flowing into hot ambient.
 Combustion Theory and Modelling 1189
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Figure 5. Variation of the ignition state, D̄aI , versus α (which is changed by variations in T̃0 ) for selected
values of LeF , with Ẽ= 0.127, T̂∞ = 0.01382, x̃= 2 and T̃∞ = 0.02382, cool jet flowing into hot ambient.

The ignition states for the three values of LeF adopted in Figure 4 are shown in Figure 5
by plotting the reduced Damköhler number at ignition, D̄aI , versus α with fixed T̃∞ = 0.02382.
Although the selection of T̃∞ may be quantitatively unrealistic for values of α greater than 0.02,
the proper qualitative behavior is exhibited. The value is chosen such that Figures 4 and 5 are
plotted using the same parameters. The region above the curve for any LeF is the region in which
ignition is successful and a flame can be established while the region below the curve is the region
where ignition fails. The plot shows that ignition is favored when α is decreased since D̄aI is
reduced, as discussed earlier. Moreover, as LeF is decreased, the ignitable region is broadened
such that ignition occurs more easily. This figure again shows that hydrogen is more diffusive and,
hence, more ignitable (more dangerous from a fire safety perspective) than other hydrocarbon
fuels, which is consistent with the findings shown in Figure 4.
For cases where the variation of α is caused by a change of T̃∞ , the ignition state is presented
in Figure 6, which shows that ignition is favored for larger α and smaller LeF , as was seen in
Figures 3–5. The effect of the reaction order of the fuel, which can be obtained by replacing ỸF
with ỸFnF in the reaction term of Equations (11)–(13), also has been studied. The results show that
ignition is favored when the reaction order is increased because of an increased reaction rate. Its
effect, however, is much weaker as compared to other effects already discussed.

3.2. Hot jet flowing into a cool ambient (T0 > T∞ )

For this scenario, ignition is expected to occur near the centerline because the fuel is hotter than
the ambient gas. To obtain the ignition condition, Equations (37)–(40) were solved by the Crank-
Nicholson method and the resulting matrix was inverted by LU decomposition. The solution of
 1190 K.B. Lim et al.
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Figure 6. Variation of the ignition state, D̄aI , versus α corresponding to Figure 5 when changes of α is
caused by variations in T̃∞ .

θ , the temperature increase through reaction, is plotted versus ζ in Figure 7 for selected values
of ξ . Here β = 0.3, D̃a = 0.5 and LeO = 1, where β = ( T̃0 − T̃∞ )/ỸO,∞ and D̃a is given by
Equation (41). At the jet exit, ξ = 0, there is no reaction so θ = 0. With the increase of ξ , reaction
takes place and θ increases. For this case, a maximum θ is developed away from but close to
the centerline of the jet, ζ = 0. Although the reaction is very sensitive to temperature variations,
it also needs the oxidizer. When moving away from ζ = 0, the temperature reduces while the
oxidizer concentration increases. The maximum value of θ then is located where the maximum
reaction rate occurs through the competition of these two effects. At ξ = 1.153, the reaction is
sufficiently strong and the heat generation is sufficiently high to overcome the heat loss to the cold
ambient such that one of the conditions in Equation (42) is satisfied and ignition is successful.
The value of ξ = 1.153 is then identified as the ignition location ξ I and the maximum value
of θ at ξ I is defined as θ max . Beyond ξ I = 1.153, θ continues to increase, reaches a maximum
value, and then gradually drops to 0 as ξ → ∞. That is, the temperature continues to increase,
reaches its maximum, and decreases to T∞ as ξ → ∞. Thermal runaway is not observed because
of the strong heat loss from the reaction region to the ambient, given by the exp [− β ( ζ 2 + ξ )]
term of Equation (38). This part of the solution is unrealistic because once ignition is successful
at ξ I , a flame is established and the weak-reactive assumption is no longer valid. It is expected
that thermal runaway would be predicted by a transient analysis of the temporal evolution of
temperature at ξ I = 1.153.
If ignition fails, ( ∂ θ /∂ ξ )ζ =0 first increases and then decreases such that ( ∂ θ/∂ ξ )ζ = 0 ≥ β
can never be satisfied, because the heat generation is not high enough to overcome the heat loss to
the ambient. For smaller values of D̃a (but still large enough to yield ignition), θ max is established
away from ζ = 0 in the beginning, and later shifts to the centerline before ignition occurs. The
 Combustion Theory and Modelling 1191
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Figure 7. Evolution of the temperature increase through reaction, θ, versus ζ for some values of ξ until
ignition. The plot is for a hot jet flowing into a cool ambient with β = 0.3, D̃a = 0.5 and LeO = 1.

reason for this is that the values of ξ I are relatively large and sufficient oxidizer is transported to
the centerline such that ignition is dominated by the high temperature at the centerline.
Results of ignition conditions for three values of β are shown in Figures 8 and 9 by plotting
θ max and ξ I , respectively, versus D̃a for LeO = 1. On each curve, by increasing the reaction
rate, i.e. D̃a, a smaller temperature increase and a shorter ignition location are observed before
ignition. A higher value of D̃a yields an increased heat generation rate, which further compensates
for heat loss from the hot jet to the cold ambient and favors ignition. The stronger reaction rate

Figure 8. Reaction temperature increase before ignition, θ max , versus reduced Damköhler number D̃a for
selected values of β = ( T̃0 − T̃∞ )/ỸO,∞ and LeO = 1, hot jet flowing into cool ambient.
 1192 K.B. Lim et al.
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

Figure 9. Ignition location ξ I corresponding to Figure 8.

also brings the point of ignition closer to the jet exit. In contrast, a reduction in D̃a weakens
the reaction and makes ignition more difficult such that both θ max and ξ I increase. Although an
increase of ξ provides longer residence time for the reaction so that ignition can occur at a smaller
D̃a, the reaction rate decreases with ξ because the jet is cooled by the cold ambient, as can be
seen from the reaction term of Equation (38). A sharp increase in θ max and ξ I for lower values
of D̃a on each β curve in Figures 8 and 9 means that the reduction of reaction rate dominates
over the residence time increase, and defines the smallest D̃a for which ignition is possible. The

Figure 10. Rescaled plot of ξ̂I versus D̂a from Figure 9 when changes of β is caused by variations in T̃0 ,
with T̃∞ = 0.00382, Ẽ = 0.127, ỸO,∞ = 0.0291, and T̂0 = 0.00673.
 Combustion Theory and Modelling 1193

minimum value of D̃a below which ignition fails is then identified as the ignition state, D̃aI .
Ignition is successful for all D̃a > D̃aI .
Figures 8 and 9 also indicate that a decrease in β for any fixed D̃a favors ignition, as
ignition occurs at a lower temperature increase, θ max , and at a shorter ignition location, ξ I . More
importantly, a decrease in β permits ignition at a lower value of D̃a. Such a decrease can be
accomplished either by increasing the ambient temperature, T̃∞ or qF /cp . Both findings are
physically realistic as in the cold jet scenario.
Parameter β also can be changed by variations of the jet temperature, T̃0 , or the reactant mass
fraction in the oxidizer supply, ỸO,∞ , but these changes result in a simultaneous change of D̃a. To
investigate the effects variations in T̃0 at fixed D̃a requires a rescaling similar to that performed in
Section 3.1. The rescaling is performed here by specifying a reference value of T̃0 as T̂0 , defining
rescaled parameters ε̂ = T̂02 /Ẽ, ξ̂I = ( T̃0 /T̂0 )2 ξI and

D̂a = ε̂ [Da ỸF,0 /(T̂0 ỸO,∞ )] exp (−Ẽ/T̂0 ), (45)

Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

and plotting the results in terms of rescaled variables ξ̂I and D̂a corresponding to Figure 9. In
the following discussion, T̃∞ = 0.00382, Ẽ = 0.127, ỸO,∞ = 0.0291 and T̂0 = 0.00673 have
been used to predict the ignition characteristics. For hydrogen, these correspond to T∞ = 300 K,
E = 10 000 K, YO,∞ = 0.233, and T0,ref = 530 K (β ref = 0.1). The results are shown in Figure
10. Here an increase in T̃0 , which increases β without changing D̂a, is seen to favor ignition
because ignition can occur at a lower reaction rate, or lower D̂a, as expected and in agreement with
Figure 4. A plot corresponding to Figure 8 also can be performed but will not be included because
the results are qualitatively similar to Figure 10.
As for the cold jet scenario, when the variation of β is caused by the change of ỸO,∞ , a
rescaling of D̃a by defining a reference value of ỸO,∞ , similar to that of Equation (45), is needed.
The results are qualitatively the same as those presented in Figures 8 and 9 because an increase
in ỸO,∞ (a decrease in β) results in a greater oxidizer concentration in the reaction region.

Figure 11. Effect of LeO on ignition for a hot jet flowing into cool ambient is exhibited by plotting the
ignition location ξ I versus reduced Damköhler number D̃a with β = 0.3.
 1194 K.B. Lim et al.

The effects of oxidizer Lewis number on spontaneous ignition are considered in Figure 11
by plotting ξ I versus D̃a for various LeO . This plot shows that a decrease in LeO makes ignition
more difficult. For an increase in the mass diffusivity of the oxidizer (or a decreased LeO ) at a
fixed value of D̃a, ξ I (and θ max ) increases. In addition, the minimum D̃a for ignition increases
with decreased LeO such that a higher reaction intensity is necessary to ignite the reactants. This
is contrary to the ignition behavior with respect to fuel Lewis number in the cool jet case (Figure
4). In a cool jet, there is unlimited heat transfer from the hot ambient gas to preheat the fuel so
that a higher fuel diffusion rate (lower LeF ) results in a higher fuel concentration in the reaction
region, more heat generation through the reaction and, hence, easier ignition. In the hot jet, only
limited heat is available from the fuel flow. An increased oxidizer mass diffusivity (lower LeO )
increases the transport rate of oxidizer to the center of the jet, thus requiring more heat to preheat
the oxidizer, decreasing the temperature in the hot zone, and making ignition more difficult. In
addition, unlike the cool jet for which LeF has a significant effect on ignition, the Lewis number
of the oxidizer only has a weak effect on the ignition state (see Figure 11) because ignition
occurs near the exit plane, if successful. In the reaction region, the flow velocity is high such that
streamwise convection dominates over transverse diffusion. This observation, when combined
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

with the fact that LeO is close to unity for oxygen in air, indicates that the effect of LeO in a
hot-hydrogen cold-air system is secondary.
The ignition states for the hot jet scenario are shown in Figure 12 by plotting the reduced
Damköhler number at ignition, D̃aI , versus β when the variation of β is through the change
of T̃∞ , for LeO = 1. Because the effect of LeO is known from Figure 11 and is less important,
it is not included in Figure 12. Instead, the effect of the oxidizer reaction order, nO , which is
obtained by replacing the ỸO in the reaction term of Equations (11)–(13) with ỸOnO , is studied.

Figure 12. Variation of the ignition state, D̃aI , versus β (which is changed by variations in T̃∞ ) for selected
values of the oxidizer reaction order, nO , hot jet flowing into cool ambient.
 Combustion Theory and Modelling 1195

The region above the curve for any nO is the region in which ignition is successful, and a flame
can be established, while the region below the curve is the region where ignition fails. The plot
shows that ignition is favored when β is decreased since D̃aI is reduced, in agreement with that
discussed in Figure 5. Figure 12 also shows that ignition is favored when the oxidizer reaction
order is increased, as the region that is ignitable is broadened. This is a result of increased reaction
rate, or the heat generation rate. For cases when the variation of β is caused by the change of
T̃0 , rescaling can be performed using Equation (45) as in Figure 10 and the result is qualitatively
similar to Figure 6.
For both the cool and hot jet scenarios, the dependence of ignition on the slot width, h, can
be identified through the x0 /u0 term given in the definition of Da after Equation (16), where x0 is
given in Equation (10). Equation (10) indicates that x0 increases with h, but the exact expression
requires knowledge of the flow field in the slot. It can be shown that for a choked uniform flow
driven by a high pressure gradient, and a fully-developed parabolic flow induced by a lower
pressure gradient, x0 /u0 (and all different forms of reduced Damköhler number) is proportional
to h2 . Thus an increase in slot width favors ignition because there is more fuel in the jet.
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

4. Conclusions
The spontaneous ignition of a fuel jet, with emphasis on a hydrogen jet, emanating from a slot
into an oxidizing ambient has been considered analytically. A similarity solution of the flowfield
was obtained, which was then applied to the species and energy conservation equations. Solutions
of the ignition conditions were found using activation energy asymptotics. Because spontaneous
ignition is very sensitive to temperature, ignition is expected to occur near the edge of the jet if
the fuel is cooler than the ambient gas, and on the centerline if the fuel is hotter than the ambient
gas.
For a cool jet flowing into a hot ambient, ignition is found to be a strong function of ambient
temperature, initial fuel concentration and fuel Lewis number, but a weak function of the reaction
order of the fuel. Ignition was favored by an increase in ambient temperature and initial fuel
concentration or a decrease in Lewis number. For the hot jet scenario, ignition was significantly
affected by the jet temperature, moderately affected by the reaction order of the oxidizer, but only
weakly affected by the oxidizer Lewis number.
The value of the mixture fraction Z at which ignition occurs can be extracted from the solutions
as Z = ( σ YF − YO + YO,∞ )/( σ YF,0 + YO,∞ ). In the first scenario of cool jet flowing into a
hot ambient, ignition occurs at Z → 0 since it occurs at the jet edge where YO → YO,∞ and
YF → 0. In the second scenario of the hot jet, ignition occurs at the jet centerline, where Z → 1
since YO → 0 and YF → YF,0 .
The present model can be extended to studies of flame extinction and to circular jet config-
urations. When experimental data becomes available, parametric comparisons can be made to
establish reaction rates for use in the present model.

Acknowledgment
This work was supported by NIST grant 60NANB5D1209 under the technical management of Dr. J. Yang.

References
[1] J.M. Kuchta, Investigation of fire and explosion accidents in the chemical, mining, and fuel-related
industries—a manual, U.S. Bureau of Mines, Bulletin 680 (1985).
[2] J. Hord, Is hydrogen a safe fuel?, International Journal of Hydrogen Energy 3 (1978), 157–176.
 1196 K.B. Lim et al.

[3] A.M. Kanury, Introduction to Combustion Phenomena (Amsterdam: Gordon and Breach Publishers)
(1975).
[4] X.L. Zheng, and C.K. Law, Ignition of premixed hydrogen/air by heated counterflow under reduced
and elevated pressures, Combustion and Flame 136 (2004), 168–179.
[5] C.E. Grosch, and T.L. Jackson, Ignition and structure of a laminar diffusion flame in a compressible
mixing layer with finite rate chemistry, Physics of Fluids A 3 (1991), 3087–3097.
[6] H.G. Im, S.R. Lee, and C.K. Law, Ignition in the supersonic hydrogen/air mixing layer with reduced
reaction mechanisms, Paper No. AIAA-1994-548, 32nd Aerospace Sciences Meeting and Exhibit,
Reno, NV, (1994), 10–13 January.
[7] H.G. Im, B.H. Chao, J.K. Bechtold, and C.K. Law, Analysis of thermal ignition in the supersonic
mixing layer, AIAA Journal 32 (1994), 341–349.
[8] S.R. Lee, and S.H. Chung, On the structure of hydrogen diffusion flames with reduced kinetic mecha-
nisms, Combustion Science and Technology 96 (1994), 247–277.
[9] B.T. Helenbrook, and C.K. Law, Ignition of hydrogen and oxygen in counterflow at high pressures,
Proceedings of Combustion Institute 26 (1996), 815–822.
[10] H.G. Im, B.T. Helenbrook, S.R. Lee, and C.K. Law, Ignition in the supersonic hydrogen/air mixing
layer with reduced reaction mechanisms, Journal of Fluid Mechanics 322 (1996), 275–296.
[11] V.V. Toro, A.V. Mokhov, H.B. Levinsky, and M.D. Smooke, Combined experimental and computational
study of laminar, axisymmetric hydrogen–air diffusion flames, Proceedings of Combustion Institute 30
Downloaded By: [Chao, B. H.] At: 19:40 28 November 2008

(2005), 485–492.
[12] M. Chaos, R.-H. Chen, E.J. Welle, and W.L. Roberts, Fuel Lewis number effects in unsteady Burke
Schumann hydrogen flames, Combustion Science and Technology 177 (2005), 75–88.
[13] Y. Liu, and P. Pei, Asymptotic analysis on autoignition and explosion limits of hydrogen–oxygen
mixtures in homogeneous systems, International Journal of Hydrogen Energy 31 (2006), 639–647.
[14] F.L. Dryer, M. Chao, Z. Zhao, J.N. Stein, J.Y. Alpert, and C.J. Homer, Spontaneous ignition of
pressurized releases of hydrogen and natural gas into air, Combustion Science and Technology 179
(2007), 663–694.
[15] F.A. Williams, Combustion Theory, 2nd Edition (Redwood City: Addison-Wesley) (1985).
[16] C.K. Law, Combustion Physics (New York: Cambridge University Press) (2006).
[17] N. Peters, Turbulent Combustion, (Cambridge: Cambridge University Press) (2000).
[18] A. Liñán, and F.A. Williams, Fundamental Aspects of Combustion, Vol. 34 of Oxford Engineering
Science Series (New York: Oxford University Press) (1993).
[19] W. Bickey, The plane jet, Philosophical Magazine Series 7, 23 (1939), pp. 727–731.
[20] H. Schlichting, Boundary-Layer Theory, 7th Edition (New York: McGraw-Hill) (1979).
[21] V. Nayagam, and F.A. Williams, Lewis-number effects on edge-flame propagation, Journal of Fluid
Mechanics 458 (2002), 219–228.
 2008-01-0726

Flame Quenching Limits of Hydrogen Leaks

M.S. Butler, R.L. Axelbaum
Dept. of Energy, Environmental and Chemical Engineering, Washington Univ., St. Louis MO 63130, USA

C.W. Moran, P.B. Sunderland

Dept. of Fire Protection Engineering, Univ. of Maryland, College Park MD 20742, USA

Copyright © 2007 SAE International

ABSTRACT solid, and it can be transported using pipelines, tankers

This study examines the types of hydrogen leaks that or rail trucks over long distances [4]. It is also projected
can support combustion. Hydrogen, methane, and that hydrogen production can be produced efficiently
propane diffusion flames on round burners and leaky using renewable methods such as wind [5].
compression fittings were observed. Measurements
included limits of quenching and blowoff for round There are, however, several hazards associated with
burners with diameters of 0.006 - 3.18 mm. The hydrogen. Applications for the gas will likely shift to
measured mass flow rates at the quenching limits were include commercial applications; these hold more risk
found to be generally independent of burner diameter at than the current, tightly-regulated industrial applications,
relatively large burner diameters. In terms of mass flow thus codes and standards must be updated [6].
rate, hydrogen had the lowest quenching limit and the Experimental and numerical work has shown that there
highest blowoff limit of the fuels considered, which are some obvious safety risks in hydrogen explosions
means that there are high and low flow rates where and high pressure gas releases [7,8]. Also, hydrogen
hydrogen is able to support a flame while methane and itself presents several unusual fire hazards, including
propane are not able to. The quenching limits for high leak propensity, high blowoff limits, ease of ignition,
hydrogen diffusion flames on round burners with thick and invisible flames. These characteristics have been
walls were found to be higher than for thin walls. The investigated in the context of micro diffusion flames.
limits were also found to be independent of burner
orientation; leaks with low flow rates are able to support There are three classifications for laminar jet flames:
flames independent of their orientation. The minimum diffusion controlled flames such as the Burke-Schumann
mass flow rate of hydrogen that can support combustion flame [9], diffusion and momentum controlled flames
from a leaking compression fitting was found to be such as the Roper flame [10] and micro diffusion flames
0.028 mg/s. This flow was independent of pressure (up controlled by momentum and diffusion [11]. Extensive
to 131 bar) and about an order of magnitude lower than research has been done on the first two types, but only a
the corresponding methane and propane flow rates. The limited amount has been done on micro diffusion flames.
implications for fire safety are discussed. Baker et al. [12] devised a flame height expression for
purely diffusion controlled flames capable of accurately
INTRODUCTION predicting micro-slot diffusion flame heights. Useful
Concerns about the emissions of greenhouse gases parameters for characterizing dominant flame
have led to extensive consideration of hydrogen as an mechanisms were defined: a diffusion-buoyancy and a
energy carrier. Currently, oil constitutes about 33% of diffusion-momentum parameter. This work extended the
the world’s primary energy, and it is primarily produced investigations of Roper et al. [13] to smaller slot sizes.
in countries that are replete with political instability as Ban et al. [11] investigated flames established on
viewed from a western point of view [1]. With crude oil circular burners with inner diameters of 0.15, 0.25, and
approaching $100/barrel, supply is not keeping up with 0.40 mm. The experiments worked to verify the
demand and political instability makes the price even predicted flame shape of laminar flames (Re = 1 - 10) for
more tenuous. In addition, our society’s dependence on three fuels: ethane, ethylene and acetylene. The work
fossil fuels possibly results in global warming, acid rain, found that buoyancy effects are negligible for small
etc. It is also predicted that crude oil supplies will run dry flames; nearly spherical flames were established that
by 2043 [2]. Hydrogen as an energy carrier can help were unaltered by rotating the burner. An in depth
mitigate these issues [3]. Some benefits of hydrogen analysis of a micro diffusion hydrogen flame was
include that it can be converted to electricity with a high performed by Cheng et al. [14]; their numerical solution
efficiency, its raw materials and the majority of its for species was compared with experimental data. It was
products are water, it can be stored as liquid, gas or found that buoyancy effects were insignificant.
Nakamura et al. [15] numerically studied methane micro has also been noted that hydrogen flames have weak
diffusion flames sitting on a circular burner of a diameter luminosity [14], hence a hydrogen leak sustaining a
less than 1 mm. They found that small flames have the flame would be difficult to detect by the human eye.
same, nearly spherical structure as those in
microgravity, citing weak buoyancy forces. Also, the Considering these properties of hydrogen, the scenario
existence of a minimum flame size necessary for of interest in this work is that a small, easily ignitable
combustion was predicted. leak in a hydrogen system might ignite, support a flame
that is difficult to detect, and degrade containment
Quenching and blowoff limits bound the leak flow rates materials to the point of a catastrophic failure. This study
that can support combustion. Matta et al. [16] found that includes experiments and analysis to identify which
a flame is not able to exist when its predicted length is hydrogen leaks can support flames. Thus motivated, the
less than the measured standoff distance. Experiments objectives of this work are to measure limits of sustained
verified this analytical method of finding the quenching combustion, both at quenching and blowoff, for
flow rate by establishing a propane flame over a hydrogen on round burners and lower flaming limits for
hypodermic, stainless steel tube and decreasing the fuel hydrogen, methane and propane on leaky compression
flow rate until extinction occurred. They also measured fittings.
blowoff. Cheng et al. [17] performed similar quenching
experiments and included a dominant mechanism FLAME QUENCH SCALING
similar to that of Baker et al. [12], except using methane A scaling analysis was developed to interpret measured
as fuel. Kalghatgi et al. [18] measured blowoff limits for a flame quenching limits. These limits are the minimum
variety of fuels including propane, methane and flow rates required to support a diffusion flame.
hydrogen on circular burners. These studies show that
hydrogen blowoff limits are higher than those for The stoichiometric length Lf of laminar gas jet diffusion
methane and propane. For a given leak size, there is a flames on round burners is:
range of mass flow rates where hydrogen is able to
support a stable flame but methane and propane would Lf / d = a Re = a ρ u0 d / μ, (1)
be blown off.
where d is the burner inside diameter, a is a
In order to understand how much gas is flowing through dimensionless fuel-specific empirical constant, Re is
a given leak, research has been done in evaluating leak Reynolds number, u0 is the average fuel velocity in the
flow rates of hydrogen, methane, and propane. Swain burner, ρ is fuel density, and μ is fuel dynamic viscosity.
and Swain [19] modeled and measured leak rates for The scaling of Eq. 1 arises from many theoretical and
diffusion, laminar, and turbulent flow regimes. They experimental studies, including Roper [10], Sunderland
found that combustible mixtures in an enclosed space et al. [24], and references cited therein. Constant a here
resulted more quickly for propane and hydrogen leaks is assigned values measured by Sunderland et al. [24],
than for methane leaks. It was also found that hydrogen as listed in Table 1.
had significantly higher volumetric flow rates through
leaks than did methane and propane. Their supply The base of an attached jet diffusion flame is quenched
pressures were the same for all fuels. by the burner. Its standoff distance can be approximated
as one half of the quenching distance of a stoichiometric
Another risk associated with hydrogen is that it is often premixed flame. Such quenching distances typically are
stored at high pressures in spite of its high propensity to reported as the minimum tube diameter, Lq, through
leak. It is predicted that hydrogen will need to be stored which a premixed flame can pass. It is assumed here
at up to 40 MPa to gain the efficiency desired [22]. that a jet flame can be supported only if its stoichiometric
Hydrogen’s safety hazards resulted in a Department of length is greater than half this quenching distance:
Energy report [23] finding that hydrogen containment
was the chief safety concern associated with using Lf ≥ Lq / 2. (2)
hydrogen as a transportation fuel and documented
several catastrophic hydrogen fires. One such accident Measurements of Lq, shown in Table 1, are taken from
that displays hydrogen’s difficulty in storage is a
pressurized hydrogen tank rupture in Hanau, Frankfurt in
1991. Heavy damage was done to the rest of the plant Table 1: Selected fuel properties of hydrogen, methane, and
propane. Values for a are from Sunderland et al. [7], Lq and SL
[20].
(laminar burning velocity) are from Kanury [8], and μ is from
Weast and Astle [9].
Other characteristics of hydrogen that make it more
prone to risk are that it is also easier to ignite than most Fuel a Lq SL μ mfuel
other common fuels, and, once ignited, it is difficult to [mm] [cm/s] [g/m-s] [mg/s]
detect. The minimum ignition energy for hydrogen is an predicted
order of magnitude lower than those for methane and H2 0.236 0.51 291 8.76e-3 0.008
propane [21]. This means that a small spark might be CH4 0.136 2.3 37.3 1.09e-2 0.085
able to ignite a hydrogen leak, whereas the same spark C3H8 0.108 1.78 42.9 7.95e-3 0.063
might not be able to ignite a methane or propane leak. It
Kanury [25]. When combined, Eqs. (1) and (2) predict radial hole was drilled in each tube. Hole sizes were
the following fuel flow rate, mfuel , at the quenching limit: 0.41, 0.53, 0.74, 0.86, and 1.02 mm (small tubes) and
0.41, 1.75, 2.46 and 3.12 mm (large tubes).
mfuel = π ρ u0 d2 / 4 = π Lq μ / ( 8 a ). (3)
Certain complications are unique to these experiments.
Eq. (3) indicates that the fuel mass flow rate at the The present small flow rates could not be accurately
quenching limit is a fuel property that is independent of measured using conventional flow meters. Instead, for
burner diameter. When values of Lq, μ, and a from tests with round burners a glass soap-bubble meter was
Table 1 are inserted into Eq. (3), the predicted fuel mass installed upstream of the burners. Tests performed with
flow rates at quenching are obtained. varying air moisture content found quenching flow rate to
be generally independent of air moisture content in the
EXPERIMENTAL range of 46% to 90% relative humidity. Hydrogen flame
Two different burner configurations are considered here: detection was complicated by their dimness, especially
round burners and leaky compression fittings. at low flow rates, but this was resolved by using
thermocouples to detect flames. K-type thermocouples
There is generally a limited range of flow rates for which were chosen to avoid catalysis, and were placed several
a flame can be established on the present burners. flame lengths above the flames to avoid disturbances.
Below this range, the flow is said to be below the
quenching limit. Quenching occurs when there is too Quenching flow rates were measured by first
much heat loss for combustion to be sustained. Above establishing a small flame, decreasing the flow rate until
this range, the flow is said to be above its blowoff limit. flame extinction and then introducing a soap bubble for
This limit occurs when the flow velocity in the flammable flow rate measurement. Each burner was allowed to
region of the flame becomes greater than the burning warm to just above room temperature to prevent water
velocity of the mixture. condensation on the burner surface. This is because
water condensation was found to disturb flow from the
Round Burners small burners, sometimes extinguishing the flames.
Three different types of round burners were considered Tests performed with different burner temperatures
for this work: tube burners, pinhole burners, and curved- found quenching flow rate to be largely independent of
wall pinhole burners, as displayed in Fig. 1. Each had a burner temperature.
range of flow passage diameters. All tests were
performed at ambient lab pressure, and flow was Burner temperature was not an issue during blowoff
delivered via a pressure regulator and a flow controlling measurements because the burners remained relatively
valve. cool. The fuel flow rate at blowoff was measured in
much the same way as quenching limits were. A stable,
The tube burners were made from stainless steel large flame was first established; the flow rate was then
hypodermic tubes, oriented vertically. Tube inside increased by increasing the upstream pressure via a
diameters were 0.006, 0.016, 0.033, 0.047, and pressure regulator until the flame first lifted and then
0.087 mm. These burners resemble microinjectors that extinguished. For these tests the flames were detected
may be used in future, small scale microelectro- visually. Hearing protection was used for blowoff tests
mechanical power generators [16]. The pinhole burners on larger burners.
were stainless steel nozzles that are manufactured for
solid-stream spray generation. The top of each burner To complete the investigation on round burners,
(except the two smallest ones) is a curved surface with a experiments were performed to consider buoyancy
hole passing through its axis, as shown in Fig. 2. The effects. Quenching flow rates were found for both
two smallest burners had holes in planar, not curved, pinhole and tube burners in the horizontal (rotated 90°)
surfaces. It is expected that the variation of quenching and inverted orientations (rotated 180°). The limits were
flow rate as a result of this slight curvature is on the
order of the error associated with measurements. The
hole diameters were 0.008, 0.13, 0.36, 0.53, 0.71, 0.84,
1.01, 1.40, 1.78, 2.39, and 3.18 mm. The curved-wall 0.356 mm
pinhole burners were constructed of stainless steel
tubes with two outside diameters: 1.59 and 6.35 mm. A

Pinhole Burner Curved-wall Burner Tube Burner Figure 2: Contrast-enhanced image of hydrogen
flame over a 0.356 mm pinhole burner (original in
Figure 1: Burners used to find quenching limits. color).
measured in the same manner. were cooled to room temperature prior to each test.

Leaky Fittings For hydrogen the flames generally were not visible even
Quenching limits were also measured for leaky in dark conditions. Existence or absence of persistent
compression fittings. Leaks were intentionally flames was determined with a thermocouple about 2 cm
introduced. This simulates an unintended leak in a above the burner. Unlike methane and propane, for
defective fuel line. Presently no regulations exist for hydrogen a pop was always heard at ignition. For a flow
allowable leak rates in hydrogen plumbing systems. rate (i.e., a pressure) below the quenching limit the pop
was followed by extinction when the external flame was
The leaky fittings involved 6.35 mm outside diameter removed. At slightly higher pressures, this pop was
stainless steel tubes. Each tube entered a Swagelok® followed by a persistent flame.
stainless steel tube union compression fitting. The end
of the union opposite the tube was sealed such that the For methane and propane, quenching limits were
only flowing gas was that which went through the leak. identified visually – tests with thermocouples confirmed
Leaks were introduced three ways into fittings that these flames to be visible, even when only slightly above
previously were made according to manufacturer their quenching limits. Most tests were performed with
instructions and found to be leak tight. The first was by the burner in the vertical orientation, with the leak at the
reducing the torque on the threaded nut. The second top end of the 6.35 mm tube. Horizontal and inverted
was by tightening the threaded nut by an additional orientations also were considered for some tests.
0.75 turns. The third was by scratching the front ferrule
sealing surface. All three types of leaks are occasionally RESULTS
encountered in plumbing systems. The present Round Burners
quenching limits results were the same for all three The weak luminosity of small hydrogen flames was
types of leaks, so only results from the reduced torque immediately confirmed. Figure 2 shows an image of a
fittings are presented here. hydrogen flame on a 0.36 mm pinhole burner just prior
to extinction. The photo was taken in a dimly lit room
Quenching limits were found by adjusting the torque on with an f-number of 4.2, ƒ/4.2, and an exposure time of
the threaded nut and the upstream pressure. For 30 s. The image was still very faint with these settings,
pressures above (and below) the quenching limit, an so a digital contrast enhancement was applied. The
external flame impinging on the leak region did (and did hydrogen quenching distance of Table 1 suggests that
not) result in a persistent flame at the leak. Upstream steady hydrogen diffusion flames should be anchored
pressure was controlled with a pressure regulator in the about 0.25 mm above the burner surface. Figure 2
range of 1.7 – 131 barg (24 – 1900 psig). When a shows this is reasonable for the present flames near
quenching limit was established, any flame was their quenching limits.
extinguished and a tube was installed over the fitting
such that the leak flow passed into a soap bubble meter Figure 3 shows a summary of all the limit data for tube
at laboratory pressure. The quenching flow rate was burners. The blowoff flow rates increase with tube
then measured with the bubble meter. burner diameter, and the quenching flow rates are
relatively flat. The limits of combustion in Fig. 3 are much
All tests were performed in a dim room. The burners
0.016
1000
Tube burner
Blowoff Limits 0.014 1.59 mm curved-wall pinhole burner
Quenching Mass Flow Rate (mg/s)

Pinhole burner
100 6.35 mm curved-wall pinhole burner
0.012
Mass Flow Rate (mg/s)

10 0.01
Hydrogen-Kalghatgi
Hydrogen-This work
Propane-Matta 0.008
1 Propane-Kalghatgi
Methane-Kalghatgi 0.006
Methane-Cheng
0.1
0.004

Quenching Limits 0.002

0.01

0
0.001 0 0.5 1 1.5 2 2.5 3 3.5
0 0.5 1 1.5 2 2.5 3
Burner Diameter (mm)
Tube Diameter (mm) Figure 4: Quenching mass flow rates versus burner diameter
Figure 3: Quenching and blowoff limits versus tube diameter. for hydrogen.
wider for hydrogen than for propane or methane. The than heat lost to the pinhole burners. This acts as a pre-
hydrogen quenching limits are about an order of heating effect which stabilizes the flame.
magnitude lower than those for methane and propane,
while hydrogen blowoff limits are about an order of There also appears to be varying amounts of heat loss
magnitude higher than those of the others. For given a as a function of burner diameter. The tube burners show
leak size there is a range of low and high flow rates that the same trend as the data from Matta et al. [8]; there is
hydrogen would be able to support a flame while an increase in quenching flow rate with increasing
propane and methane would be either quenched or burner size. It is reasonable to conclude that there is
blown off. The data from Matta et al. [16], Cheng et al. more heat loss to the burner with increasing burner size;
[14], and Kalghatgi [18] show that the limits for methane with more heat loss, the quenching flow rate will
and propane are very similar. For all fuels, the increase to account for the weaker flame. The pinhole
quenching limits are nearly independent of burner burners do not exhibit the same trend as the tube
diameter, whereas blowoff limits increase with burners because approximately the same amount of
increasing diameter. The independence of quenching heat loss occurs independent of burner size. The reason
limit from burner diameter is consistent with the analysis for the local maximum in Fig. 4 for pinhole burners is
presented above, see Eq. (3). under investigation, however at large burner diameters,
the quenching mass flow rate for both pinhole and tube
Figure 4 shows the present hydrogen quenching limits. burners approaches 0.008 mg/s as predicted by the
Data are included for all three types of burners. The scaling analysis.
theory of Eq. (3) predicts that the quenching flow rate is
independent of burner diameter. Figure 4 indicates that The effect of curvature was investigated by using the
the scaling analysis roughly predicts the average curved-wall pinhole burners. It was expected that with
quenching flow rate for hydrogen especially at burner increasing curvature, a flame would experience less wall
diameters greater than 2 mm, although there is clearly heat loss. This should result in a stronger flame and a
another mechanism not accounted for in the calculation. lower quenching flow rate. This is exactly what was
What is clear, though, is that despite varying amounts of observed in the two curved-wall burners. The 6.35 mm
heat loss, all burners displayed hydrogen’s ability to burners act more like pinhole burners, whereas the
sustain flames at very low mass flow rates. 1.59 mm burners act more like tube burners; the
6.35 mm curved-wall burners have higher average
Heat loss is likely responsible for the differences quenching flow rates than the 1.59 mm curved-wall
between the types of burners. The pinhole burners show burners. Although curvature did affect the quenching
the upper limit for quenching flow rates while the tube flow rate, the results showed that, despite the amount of
burners bound the lower limits. This is because the curvature, a hydrogen leak can support combustion at
pinhole burners have more heat loss than tube burners. very low mass flow rates.
An alternative explanation is that heat lost to the tube
burners is more likely to be transferred to the fuel flow Both the tube and pinhole burners have an increase in

0.008 0.012
Quenching Mass Flow Rate (mg/s)

Quenching Mass Flow Rate (mg/s)

0.01
0.007

0.008
0.006

0.006

0.005
0.004
Horizontal Horizontal
Inverted Inverted
0.004 Vertical Vertical
0.002

0.003 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5 3 3.5
Tube Diameter (mm) Pinhole Burner Diameter (mm)

Figure 5: Quenching mass flow rates for several orientations Figure 6: Quenching mass flow rates for several orientations
of tube burners. of pinhole burners.
 6.33 mm

Hydrogen Methane Propane

Figure 7: Flames near extinction on leaky compression fittings (original in color).

quenching flow rate for the smallest burners. This is suggests that the flame structure should not change with
because the velocity at the burner outlet is large enough varying orientation, which is consistent with the
to weaken the flame. experimental results.

It was found from Figs. 5 and 6 that the quenching flow Leaky Fittings
rate for pinhole and tube burners was independent of Figure 7 includes images of flames on leaky
orientation indicating that the flow is not controlled by compression fittings just above extinction. It is apparent
buoyancy. There are two characteristics that can affect that the hydrogen flame is significantly smaller than
quenching: the flow field and heat loss. Earlier tests those for methane and propane. This seems to indicate
found that quenching flow rate was independent of that much less hydrogen must be escaping through the
burner temperature. During these experiments, the leak to sustain combustion. This was validated in the
burner was found to be warmer during non-vertical tests. measurements.
If burner temperature does not affect quenching and it is
the case that the quenching limit is not changing with Figure 8 shows the measured quenching limits for
orientation, then the flow field must also be constant with hydrogen, propane, and methane at a leaky
varying orientation. Hence, it was shown that flames compression fitting in the vertical orientation. For each
near extinction are driven by mechanisms other than fuel the data at increased pressure are associated with
buoyancy. This shows that a hydrogen leak is able to increased torque on the fitting. The upper limit on
support a flame at very low mass flow rates regardless pressure for propane is lower than for the others
of orientation. The weak effect of buoyancy in these because the vapor pressure of propane at 25 °C is 9.1
flames is supported by a nondimensional analysis. bar (142 psia).

The Froude numbers for the present pinhole and tube Within experimental uncertainties, the data of Fig. 8 are
burner quenching limit flames were found to be at least independent of pressure for each fuel. This indicates
0.1. Flames with these Froude numbers are within the that, as expected, at fixed fuel mass flow rate the
transitional to buoyancy-controlled regime [11]. This
0.4
0.4

Vertical
0.3
Minimum Flowrate (mg/s)

0.378 mg/s Horizontal

Minimum Flowrate (mg/s)

0.3
Inverted
0.336 mg/s

Hydrogen
0.2 Methane 0.2
Propane
h
m
p
0.1 0.028 mg/s Linear (h) 0.1
Linear (m)
Linear (p)

0
1 10 100 1000 0
Pressure (bar) Hydrogen Me thane Propane

Figure 8: Quenching limit (minimum flaming flow rate) versus

upstream pressure for a leaky compression fitting in the Figure 9: Minimum flaming flow rate for a leaky compression
vertical orientation. The horizontal lines are means of the fitting in vertical, horizontal and inverted orientations.
measured quenching limits. Upstream pressure was about 4 bar in all cases.
upstream pressure has little or no effect on the velocity propane and methane. The two experiments differed in
profile of the jet entering the surrounding air. The mean that their lower combustion limits were off by an order of
hydrogen flow rate, 0.028 mg/s, is about an order of magnitude owing to increased heat losses in the leaky
magnitude lower than for the other fuels due to its low compression fittings. In both cases, hydrogen was able
quenching distance and low molecular weight. This to sustain flames at much lower mass flow rates than
behavior is similar to that seen in the round burners methane and propane.
above, in Fig. 3. For these leaks, there is a range of
pressures where hydrogen is able to support a flame ACKNOWLEDGMENTS
while propane and methane flames would be quenched. This work is supported by NIST under the technical
direction of J. Yang. Helpful discussions with B.H. Chao
The minimum hydrogen flaming flow rate for round are appreciated.
burners (Fig. 3) is about an order of magnitude lower
than that for leaky fittings (Fig. 8). This is attributed to REFERENCES
additional burner heat losses in the leaky fittings, where 1. Marban G, Valdes-Solis T. Towards the hydrogen
the flames burn near concave metal surfaces. economy? Int. J. Hydrogen Energy 2007;32:1625-
1637.
Figure 9 shows the effect of burner orientation on the 2. Kruger P. Alternative energy resources. Hoboken,
minimum flaming flow rate. Burner orientation has little NJ: John Wiley & Sons, Inc., 2006.
or no effect on the quenching limit of hydrogen because
3. Yamawaki M, Nishihara T, Inagaki Y, Minato K,
these flames were so small at their limits. Burner
Oigawa H, Onuki K, Hino R, Ogawa M. Application
orientation did have an effect on propane and methane
of nuclear energy for environmentally friendly
quenching limits, the inverted configuration requiring the
lowest flow rate. The inverted orientation minimized the hydrogen generation. Int. J. Hydrogen Energy
impingement of flames on metal surfaces. 2007;32:2719-2725.
4. Sherif SA, Barbir F, Veziroglu TN. Wind energy
CONCLUSIONS and the hydrogen economy - review of the
The quenching and blowoff limits for hydrogen on round technology. Solar Energy 2005;78:647-660.
burners were measured and compared with published 5. Von Jouanne A, Husain I, Wallace A, Yokochi A.
values for other gases. Hydrogen diffusion flames on Gone with the wind. IEEE Industry Applications
tube burners had much wider limits of combustion than Magazine July/Aug 2005:12-19.
propane and methane. There was a high range and a 6. MacIntyre I, Tchouvelev AV, Hay DR, Wong J,
low range of flow rates where hydrogen was able to Grant J, Benard P. Canadian hydrogen safety
sustain a flame while propane and methane would be program. Int J Hydrogen Energy 2007;32;2134-
blown off or quenched. It has been shown that the 2143.
quenching flow rates for all these gases are largely 7. Takeno K, Okabayashi K, Kouchi A, Nonaka T,
independent of burner size especially at relatively large Hashiguchi K, Chitose K. Dispersion and
burner diameters. explosion field tests for 40 MPa pressurized
hydrogen. Int J Hydrogen Energy 2007;32:2144-
The quenching flow rate was found to depend on burner 2153.
type, owing to differences in wall heat loss. Tube 8. Dorofeev SB. Evaluation of safety distances
burners had the lowest quenching flow rates, pinhole related to unconfined hydrogen explosions. Int J
burners had the highest, and curved wall burners fell in- Hydrogen Energy 2007;32:2118-2124.
between, as expected. Although the type of burner did 9. Burke SP, Schuman TEW. Diffusion flames. Ind.
affect the quenching mass flow rate, the results still Eng. Chem. 1928;20:998-1004.
showed that a hydrogen leak can sustain a flame at 10. Roper FG. The prediction of laminar jet diffusion
much lower flow rates than leaks of other common fuels. flame sizes: Part 1. Theoretical model. Combust.
Flame 1977;29:219-226.
Both types of burners were found to be largely 11. Ban H, Venkatesh S, Saito K. Convection-diffusion
independent of burner orientation. As a result, a leak in a
controlled laminar micro flames. Transactions of
hydrogen system would be able to sustain a flame at low
the ASME 1994;116:954-959.
flow rates independent of the orientation of the leak.
12. Baker J, Calvert ME, Murphy DW. Structure and
dynamics of laminar jet micro-slot diffusion flames.
The minimum flow rate necessary for sustaining a
Journal of Heat Transfer 2002;124:783-790.
hydrogen flame at a leaky compression fitting is
0.028 mg/s. This is about an order of magnitude lower 13. Roper FG, Smith C, Cunningham AC. The
than for propane or methane. The minimum mass flow prediction of laminar jet diffusion flame sizes: Part
rate for all fuels is independent of upstream pressure. II. Experimental Verification 1977;29(3):227-234.
14. Cheng TS, Chao Y-C, Wu C-Y, Li Y-H, Nakamura
The round burner and leaky compression fitting Y, Lee K-Y, Yuan T, Leu TS. Experimental and
experiments yielded similar results in that they both numerical investigation of microscale hydrogen
found that the lower limits of combustion for hydrogen diffusion flames. Proceedings of the Combustion
were about an order of magnitude lower than those for Institute 2005;30:2489-2497.
 15. Nakamura Y, Yamashita H, Saito K. A numerical
study on extinction behaviour of laminar micro-
diffusion flames. Combustion Theory and
Modelling 2006;10(6):927-938.
16. Matta LM, Neumeier Y, Lemon B, Zinn BT.
Characteristics of microscale diffusion flames.
Proc. Combust. Inst. 2002; 29:933-938.
17. Cheng TS, Chen CP, Chen CS, Li YH, Wu CY,
Chao YC. Characteristics of microjet methane
diffusion flames. Combust. Theory Modeling
2006;10:861-881.
18. Kalghatgi GT. Blow-out stability of gaseous jet
diffusion flames. Part I: In still air. Combustion
Science and Technology 1981;26:5:233-239.
19. Swain MR, Swain MN. A comparison of H2, CH4,
and C3H8 fuel leakage in residential settings. Int J
Hydrogen Energy 1992;17:807-815.
20. Rigas F, Sklavounos S. Evaluation of hazards
associated with hydrogen storage facilities. Int J
Hydrogen Energy 2005;30:1501-1510.
21. Babrauskas V. Ignition handbook, 1st edition.
Issaquah, WA: Fire Science Publishers, 2003.
22. Takeno K, Okabayashi K, Kouchi A, Nonaka T,
Hashiguchi K, Chitose K. Dispersion and
explosion field tests for 40 MPa pressurized
hydrogen. Int J Hydrogen Energy
2007;32(13):2144-2153.
23. Cadwallader LC, Herring JS. Safety issues with
hydrogen as a vehicle fuel. Report INEEL/EXT-99-
00522 prepared for U.S. Department of Energy
1999.
24. Sunderland PB, Mendelson BJ, Yuan ZG, Urban
DL. Shapes of buoyant and nonbuoyant laminar
jet diffusion flames. Combust. Flame
1999;116:376-386.
25. Kanury AM. Introduction to combustion
phenomena. New York: Gordon and Breach,
1975. p. 131.
26. Weast RC, Astle MJ. CRC Handbook of Chemistry
and Physics, 59th edition. West Palm Beach: CRC
Press, Inc., 1979. p. F-58.
27. Munson BR, Young DF, Okiishi TJ. Fundamentals
of fluid mechanics. Hoboken, NJ: Wiley, 2002. p.
359.
28. Morton NR, Sunderland PB, Chao BH, Axelbaum
RL. Quenching limits and materials degradation of
hydrogen diffusion flames. 5th U.S. Combustion
Meeting, San Diego 2007.
29. Bahadori MY, Hegde U, Zhou L. Liftoff and blowoff
of jet diffusion flames. Paper AIAA-2002-1078,
Aerospace Sciences Meeting, Reno, Nevada,
January 14-17 2002.

CONTACT
Peter B. Sunderland, [email protected].
 WASHINGTON UNIVERSITY
SCHOOL OF ENGINEERING AND APPLIED SCIENCE
DEPARTMENT OF MECHANICAL, AEROSPACE, AND
STRUCTURAL ENGINEERING

FLAME QUENCHING LIMITS AND MATERIALS

DEGRADATION OF HYDROGEN LEAKS
by
Michael S. Butler, Jr.
Prepared under the direction of Professor R.L. Axelbaum

A thesis presented to the School of Engineering of

Washington University in partial fulfillment of the
requirements for the degree of
MASTER OF SCIENCE

May 2008
Saint Louis, Missouri
 WASHINGTON UNIVERSITY
SCHOOL OF ENGINEERING AND APPLIED SCIENCE
DEPARTMENT OF MECHANICAL, AEROSPACE, AND
STRUCTURAL ENGINEERING

ABSTRACT

FLAME QUENCHING LIMITS AND MATERIALS

DEGRADATION OF HYDROGEN LEAKS
by
Michael S. Butler, Jr.

ADVISOR: R.L. Axelbaum

May 2008
St. Louis, Missouri

This study examines the types of hydrogen leaks that can support combustion and their effects
on surrounding material. Hydrogen, methane, and propane diffusion flames on round burners
and leaky compression fittings were observed. Measurements included limits of quenching and
blowoff for round burners with diameters of 0.006 - 3.18 mm. The measured mass flow rates at
the quenching limits were found to be generally independent of burner diameter at burner
diameters between 1.5 mm and 3.2 mm. In terms of mass flow rate, hydrogen had a lower
quenching limit and a higher blowoff limit than either methane or propane. Hydrogen flames
were the weakest flames ever recorded with mass flow rates as low as 0.0039 mg/s, which
translates to a thermal power output of 0.55 W. The quenching limits for diffusion flames on
round burners with thick walls were found to be higher than for thin walls. The limits were also
found to be independent of burner orientation. In other words leaks with low flow rates are able
to support flames independent of their orientation. The minimum mass flow rate of hydrogen
that can support combustion from a leaking compression fitting was found to be 0.028 mg/s.
This flow was independent of pressure (up to 131 bar) and about an order of magnitude lower
than the corresponding methane and propane mass flow rates. The hydrogen embrittlement
effects on 316 stainless steel and 1100 aluminum tubes with flames supported over radial holes
were also investigated. Micrograph images of embrittled and as-received material were
compared, and the degraded material was found to have defects that were either multi-phase
material or microvoids in the material. Each of these phenomena would result in decreased
fatigue resistance in a real life system. The implications for fire safety are discussed.
Contents
List of Tables .......................................................................................................... iii

List of Figures ......................................................................................................... iv

Acknowledgements................................................................................................. vi

1 Why Hydrogen? ..................................................................................................1

2 Hydrogen Leak Quenching Limits ................................................................... 4

2.1 Introduction .................................................................................................................. 4
2.2 Flame Quench Scaling ................................................................................................. 7
2.3 Experimental ................................................................................................................. 9
2.3.1 Round Burners ................................................................................................ 9
2.3.2 Leaky Compression Fittings ........................................................................ 11
2.4 Results ......................................................................................................................... 13
2.4.1 Round Burner Results .................................................................................. 13
2.4.2 Weakest Flames ............................................................................................. 19
2.4.3 Numerical Solution ....................................................................................... 21
2.4.4 Leaky Fittings Results................................................................................... 23
2.5 Conclusions ................................................................................................................. 26

3 Materials Degradation From A Sustained Leak Flame ....................................28

3.1 Introduction ................................................................................................................ 28
3.2 Experimental ............................................................................................................... 32
3.2.1 Burner Configuration and Temperature Measurement ........................... 32
3.1.3 Materials Characterization ........................................................................... 33
3.2 Results ......................................................................................................................... 34
3.2.1 Temperature Measurement ......................................................................... 34
3.2.2 Numerical Simulation .......................................................................................
3.2.3 Materials Characterization ............................................................................. 9
3.3 Conclusion ................................................................................................................... 10

4 Future Work ....................................................................................................... 7

References .......................................................................................................22

Vita .......................................................................................................23

ii
List of Tables
Table 2.1: Gas Characteristics.................................................................................................... 8
Table 2.2: Hole diameters.......................................................................................................... 10

iii
List of Figures
Figure 2.1 Burners ........................................................................................................................ 9
Figure 2.2 Contrast-enhanced hydrogen flame ...................................................................... 13
Figure 2.3 Quenching and blowoff limits ..................................................................... 14
Figure 2.4 Quenching mass flow rate vs. diameter ....................................................... 15
Figure 2.5 Effect of orientation on tube burners .......................................................... 17
Figure 2.6 Effect of orientation on pinhole burners ..................................................... 18
Figure 2.7 Choked flow rates through a leak ................................................................ 19
Figure 2.8 Weakest flames............................................................................................ 21
Figure 2.9 Numerical grid system ................................................................................. 22
Figure 2.10 OH contours compared to real flame ........................................................ 23
Figure 2.11 Flames near extinction............................................................................... 24
Figure 2.12 Quenching limits for leaky fittings ............................................................. 24
Figure 2.13 Effect of orientation on leaky fittings ........................................................ 25
Figure 2.14 Effect of burner size on leaky fittings ........................................................ 26
Figure 3.1 Flame temperature profile for pinhole burner ............................................. 37
Figure 3.2 Flame temperature profile for tube burner .................................................. 38
Figure 3.3 Numerical solution vs. experimental for pinhole burner .............................. 39
Figure 3.4 Numerical solution vs. experimental for tube burner ................................... 40
Figure 3.5 Experimental and numerical compared to real flame ................................... 41
Figure 3.6 Control aluminum specimen ....................................................................... 43
Figure 3.7 Vertical flame aluminum specimen .............................................................. 43
Figure 3.8 Inverted flame aluminum specimen ............................................................. 43
Figure 3.9 Control steel specimen ................................................................................ 44
Figure 3.10 Vertical flame steel specimen ..................................................................... 45
Figure 3.11 Inverted flame steel specimen.................................................................... 45

iv
Acknowledgements
Thanks to the guidance and support of Dr. Richard Axelbaum for giving a physics
student a chance to learn a new field, to my parents for teaching me that I can, to my
grandma for not letting me go hungry, and to my master’s thesis committee for their
considerable time and effort. Considerable work was done by Chris Moran and Peter
Sunderland from University of Maryland with their studies on leaky compression
fittings. All of the work presented here on leaky compression fittings is theirs. Thanks
to Xiaofeng Zheng for his work on the numerical simulations. This work was
supported financially by the National Institute of Standards and Technology and the
NASA-Missouri Space Grant Program.

Michael S. Butler, Jr.

Washington University in St. Louis

May 2008

v
 1

Chapter 1

Why Hydrogen?
Concerns about the emissions of greenhouse gases have led to extensive
consideration of hydrogen as an energy carrier. Oil supplies 33% of the world’s primary
energy [1], and hydrogen could help mitigate concerns about fossil fuel consumption
[2]. Some benefits of hydrogen include that its products of combustion are mainly
water, it can be stored as liquid, gas or solid, and it can be transported using pipelines,
tankers or rail trucks over long distances [3]. It is also projected that hydrogen
production can be produced efficiently using nuclear sources or renewable methods
such as wind [4].

There are also many other advantages of hydrogen as an energy carrier.

Hydrogen might some day be able to relieve the increasing cost of fossil fuels. It is
probable that the cost of fossil fuels will continue to increase with dwindling supplies
while the cost of hydrogen will decrease with advances in technology. Hydrogen can be
produced from water, which is indigenous or easily accessible and relatively cheap to
every nation. When the world gets to the point when avoiding carbon-containing
resources is cost efficient, there will be an essentially infinite amount of raw materials
for hydrogen production. Hence, this can be an opportunity to achieve wider global
distribution of a sustainable energy supply. This would benefit developing nations that
do not currently have access to oil fields as well as developed nations that rely on
international fuel supplies. As an example, the US sacrifices about $1 billion per week
from its economy in the cost of purchasing foreign energy. Also, hydrogen has the
highest specific energy of the chemical fuels adding to its potential efficiency. In
regards to its environmental benefits, hydrogen can help inhibit the growth of
greenhouse gas emission while stimulating economic growth in a new industry for years
 2
to come [5]. Although many of these advantages are only foreseeable in the future,
there are some current applications of hydrogen.

Hydrogen is currently economically feasible in several industries, and hydrogen

production in the US has already reached 3 billion cubic feet per year. The applications
that currently require this hydrogen include industrial processes that use hydrogen as a
chemical feed-stock. It is used as a reducing agent in the mineral industry, as a
hydrogenation agent in the petroleum industry, and as a bonding agent in the chemical
industry. It is also used as a fuel in high-temperature welding and for space exploration.
It can be used in future applications as a fuel for power plant fuel-cell generation of
electricity, as a coolant in super-conductor technology, and as the previously noted
transportation fuel to replace oil [5].

There are, however, several unusual hazards associated with hydrogen, and as
applications for hydrogen shift to include the commercial sector there will be greater
risk than in the current, tightly-regulated industrial sector. Thus, codes and standards
must be updated [6]. Experimental and numerical results have shown that hydrogen
poses significant safety risks with respect to explosions and high pressure gas release [7].
In addition, hydrogen has unique properties that lead to unusual fire safety hazards,
including high leak propensity, high blowoff limits, material embrittlement tendencies,
ease of ignition, and invisible flames. This combination of properties suggests that
small, easily ignitable hydrogen leaks may support combustion that could go unnoticed
and degrade containment materials.

To investigate this scenario, two sets of experiments were performed exposing

the additional risks associated with hydrogen. First, the quenching limits and blowoff
limits were found for hydrogen diffusion flames on several types of round hole burners.
A quenching limit is the lowest hydrogen flow rate where a flame can still be sustained,
and a blowoff limit is the maximum amount of flow where a flame can exist. Also,
quenching limits were found for leaky Swagelok fittings. Next, a material degradation
 3
study was performed to show the long-term effects of having a stable, hydrogen leak
flame. A small, nearly invisible hydrogen flame was established on an aluminum and a
steel tube and allowed to exist for over a month. The effects of the degradation on the
metal were then examined.
 4
Chapter 2

Hydrogen Leak Quenching Limits

2.1 Introduction
This work was done in conjunction with colleagues at the University of
Maryland’s Department of Fire Protection Engineering. Our goal was to evaluate the
increased fire safety risks associated with hydrogen as compared to other common
gaseous fuels such as propane and methane. Hydrogen has several unique
characteristics that make it especially prone to safety hazards compared to other fuels; it
has a high propensity for leakage, broad flammability limits, and the flame has low
luminosity. One consequence of these characteristics is that a small crack in a hydrogen
system can mean a large flow rate as compared to the scenario of the same sized crack
in a propane or methane system. also, because of the broad flammability limits, a
hydrogen and air mixture is more likely to be combustible than if it were a mixture of
propane or methane in air. Once a leak is established, it will be difficult to detect
because of its low luminosity, especially if the flame size is small.

These small hydrogen flames are important to analyze because they can be
potentially dangerous despite their size. The “leak flame” could degrade the material
around the crack and make it bigger, and they can be an ignition source in the presence
of other fuels. The “leak flames” that are under consideration are very similar to flames
that have already been studied, micro diffusion flames.

Only limited results are available for micro diffusion flames. The studies
employed a variety of approaches including numerical techniques and varying burner
sizes, burner shapes, and fuel gases to characterize micro diffusion flames. Micro-slot
 5
burners were used by Baker et al. [9] to verify a flame height expression for purely
diffusion controlled flames, and defined useful parameters for characterizing dominant
flame mechanisms. Their diffusion-buoyancy and diffusion-momentum parameters
were found to be in the range of flames dominated by diffusion, i.e. not buoyancy. This
work extended the investigations of Roper et al. [10] to smaller slot sizes. Micro
diffusion flames were also studied on circular burners. Ban et al. [11] investigated
flames established on such burners with inner diameters of 0.15, 0.25, and 0.40 mm.
The experimental results from these small burners verified the predicted flame shapes,
which were found assuming weak gravitational effects, of laminar flames (Re = 1-10) for
three fuels: ethane, ethylene and acetylene. Weak buoyancy effects were further proven
for low flow rate, nearly spherical flames that were unaltered when the burner was
rotated. The numerical results conducted by Cheng et al. [12] supported these findings.
Simulations for hydrogen micro diffusion flames were performed that yielded
temperature contours and species profiles. The group concluded that buoyancy effects
were insignificant for such flames, as indicated by their shape and a dimensionless
number analysis. Nakamura et al. [13] simulated methane micro diffusion flames
supported on circular burners of diameters less than 1 mm. They, too, found the nearly
spherical flame structure associated with weak buoyancy forces. The group was also
able to find the presence of a minimum flame size necessary for combustion.

Micro diffusion flames could be found in hydrogen plumbing systems as a result

of the ignition of leaks from cracks or holes in fittings, tubing, or any other hydrogen-
containing device. In order to relate the previous studies where flow rate was the
independent variable to systems where hydrogen pressure is the controlling variable,
Lee et al. [14] conducted leak rate experiments on micromachined orifices of different
sizes and shapes. They examined the differences in flow rates among circular, square,
and elliptical slit orifices as a function of pressure. In most cases, the flow was choked,
which yielded flows higher than those of the helium signature test. Helium was used as
a safer substitute for hydrogen. Another type of leak was investigated by Schefer et al.
[15] where the flow was pressure-driven convection and permeation through metals.
They obtained analytical relationships for flow rates of choked flows, subsonic laminar
 6
flows, and turbulent flows. Hydrogen leakage in threaded pipe fittings was also
considered. Ge and Sutton [16] found that a larger tightening torque is less important
in leak prevention than choice and proper application of thread sealant. The tests were
run up to 7000 kPa.

Swain and Swain [17] evaluated the safety risks associated with leaks of
hydrogen, methane, and propane in the case where the leaks occur in an enclosed area.
They modeled and measured leak rates for diffusion, laminar, and turbulent flow
regimes and found that, for a given supply pressure, combustible mixtures resulted
more quickly for propane and hydrogen leaks than for methane leaks. It was also found
that hydrogen had significantly higher volumetric flow rates than methane or propane.

Because of the possibility of higher flow rates, blowoff limits have been studied.
They were measured by Kalghatgi et al. [18] for a variety of fuels including propane,
methane and hydrogen on circular burners. The limits were measured by establishing a
flame over a stainless steel, hypodermic tube and increasing the fuel flow rate until the
flame went out. Quenching flow rates were obtained experimentally by Matta et al. [19]
for propane by establishing a flame over the same type of burner and decreasing the
fuel flow rate until the flame extinguished. The same quenching limit experiments were
performed by Cheng et al. [20] for methane. Both groups found that quenching mass
flow rate was largely independent of burner diameter. As a result, the ability for a flame
to exist at a leak is not a function of the leak size, but rather how much fuel is being
emitted by the leak. Since hydrogen systems are projected to contain pressures up to 40
MPa, it is critical to be able to quantify the quenching limit of hydrogen.

Other characteristics of hydrogen that make it more prone to risk are that it is
easier to ignite than most other common fuels, and, once ignited, it is difficult to detect.
The minimum ignition energy for hydrogen is an order of magnitude lower than those
for methane and propane [16]. This implies that a small spark discharge that would be
too weak to ignite a methane or propane leak might be able to ignite a hydrogen leak.
 7
Hydrogen flames also have minimal luminosity [12], hence a hydrogen leak sustaining a
flame would be difficult to detect by the human eye.

Given that these conditions might exist, it is also a possibility that the hydrogen
leak flame can degrade the surrounding material and result in a catastrophic failure.
Khan et al.[21] examined the effects of raised temperatures on carbon fabric/epoxy
composites, a likely material for high pressure hydrogen storage tanks. Pehr [22]
discusses some of the issues associated with hydrogen containment. Utgikar and
Thiesen [23] discuss the impact of hydrogen on materials, and the safety of hydrogen
fuel tanks. These studies have found that hydrogen has unique degradation properties.

Considering these properties, the scenario of interest in this work is that a small
leak in a hydrogen system might ignite, support a flame that is difficult to detect, and
degrade containment materials to the point of a catastrophic failure. This study includes
experiments and analysis to identify hydrogen leaks that can support flames. Thus
motivated, the objectives of this work are to measure limits of sustained combustion,
both at quenching and blowoff, for hydrogen on round burners and quenching limits
for hydrogen, methane and propane on leaky compression fittings.

2.2 Flame Quench Scaling

A scaling analysis is presented to interpret measured flame quenching limits.
These limits are the minimum flow rates required to support a diffusion flame. The
stoichiometric length Lf of laminar gas jet diffusion flames on round burners is:

Lf / d = a Re = a ρ u0 d / μ , (1)

where d is the burner inside diameter, a is a dimensionless fuel-specific empirical

constant, Re is Reynolds number, u0 is the average fuel velocity in the burner, ρ is fuel
density, and μ is fuel dynamic viscosity. The scaling of Eq. (1) arises from many
 8
theoretical and experimental studies, including Roper [10], Sunderland et al. [25], and
references cited therein. The constant a is fuel specific and has been measured by
Sunderland et al. [25], as listed in Table 2.1.

The base of an attached jet diffusion flame is quenched by the burner. Its
standoff distance can be approximated as one half of the quenching distance of a
stoichiometric premixed flame. Such quenching distances typically are reported as the
minimum tube diameter, Lq, through which a premixed flame can pass. It is assumed
here that a jet flame can be supported only if its stoichiometric length is greater than
half this quenching distance:

Lf ≥ Lq / 2 . (2)

Table 2.1 Measurements of Lq, shown in Table 1, are taken from Kanury [26]. Selected fuel
properties of hydrogen, methane, and propane. Values for a are from Sunderland et al. [25], La
and SL (laminar burning velocity) are from Kanury [26], and μ is from Weast and Astle [27].

Fuel a Lq SL μ mfuel
[mm] [cm/s] [g/m-s] [mg/s]
predicted
H2 0.236 0.51 291 8.76e-3 0.008
CH4 0.136 2.3 37.3 1.09e-2 0.085
C3H8 0.108 1.78 42.9 7.95e-3 0.063

Measurements of Lq, shown in Table 2.1, are taken from Kanury [26]. When combined,
Eqs. (1) and (2) predict the following fuel flow rate, mfuel , at the quenching limit:

mfuel = π ρ u0 d2 / 4 = π Lq μ / ( 8 a ) . (3)

Eq. (3) indicates that the fuel mass flow rate at the quenching limit is a fuel property
that is independent of burner diameter. When values of Lq, μ, and a from Table 2.1 are
inserted into Eq. (3), the predicted fuel mass flow rates at quenching are obtained and
are listed in the table.
 9
2.3 Experimental
Two different burner configurations are considered here: round hole burners
and leaky compression fittings. There is generally a limited range of flow rates for
which a flame can be established on the present burners. Below this range, the flow is
said to be below the quenching limit. Quenching occurs when there is too much heat
loss for combustion to be sustained. On the other extreme, blowoff limit occurs when
the flow rate is reached beyond which the flame blows off the burner.

2.3.1 Round Hole Burners

Three different types of burners were considered for this work: tube burners,
pinhole burners, and curved-wall pinhole burners, as displayed in Fig. 2.1. Each had a
range of hole diameters. The flow was delivered via a pressure regulator and a flow
control valve, and all tests were performed at ambient lab pressure.

Pinhole Burner Curved-wall Burners Tube Burner

Figure 2.1 Burners used to find quenching limits. Inside diameters for pinhole burners were.
Inside diameters for curved-wall burners were mm (small tubes) and mm (large tubes). Tube
inside diameters were mm.

The tube burners were made from stainless steel hypodermic tubes. Tube inside
diameters ranged from 0.051 – 2.21 mm. These burners resemble microinjectors that
may be used in future small-scale microelectro-mechanical power generators [14]. The
pinhole burners were stainless steel nozzles that are manufactured for solid-stream spray
generation. The top of each burner (except the two smallest ones) is a slightly curved
 10
surface with a hole passing through its axis, as shown in Fig. 2.2. The two smallest
burners had holes in planar, not curved, surfaces. The hole diameters ranged from
0.008 – 3.18 mm. The curved-wall pinhole burners were constructed of stainless steel
tubes with two outside diameters: 1.59 and 6.35 mm. A radial hole was drilled in each
tube.

Table 2.2 Inside hole diameter on the three types of burners. Measurements are in mm. Large
and small curved-wall burners denote the size of the outer diameter of the tube.

Pinhole Curved-wall Curved-wall Tube

(small) (large)
0.008 0.41 0.41 0.051
0.13 0.53 1.75 0.152
0.36 0.74 2.46 0.406
0.53 0.86 3.12 0.838
0.71 1.02 1.194
0.84 2.21
1.01
1.40
1.78
2.39
3.18

The small flow rates required special flow measurement procedures. For tests
with round burners a glass soap-bubble meter was installed upstream of the burners.
Quenching flow rates were measured by first establishing a small flame, decreasing the
flow rate until the flame extinguished, and then introducing a soap bubble in the meter
for flow rate measurement.

Each burner was allowed to warm to just above room temperature to prevent
water condensation on the burner surface. This was necessary because water
condensation was found to disturb flow from the small burners, sometimes
extinguishing the flames. Tests performed at different burner temperatures, up to about
200 °C, found quenching flow rate to be largely independent of burner temperature.
 11
Tests were also conducted with varying ambient humidity, and quenching flow
rate was found to be generally independent of humidity in the range of 46 – 90%
relative humidity. Observing the existence of a hydrogen flame was complicated by the
minimal luminosity of the flames, especially at low flow rates, but this was resolved by
using a downstream thermocouple to detect the flame. K-type thermocouples were
chosen to avoid catalysis, and were placed several flame lengths above the flames to
avoid disturbances.

Additional quenching experiments were performed to consider buoyancy

effects. Quenching flow rates were found for both pinhole and tube burners in the
vertical, horizontal, and inverted orientations.

Burner temperature was not an issue during blowoff measurements because the
burners remained relatively cool. The fuel flow rate at blowoff was measured in much
the same way as for quenching limits. A stable, large flame was first established; the
flow rate was then increased by increasing the upstream pressure until the flame blew
off. For these tests the flames were detected visually. Hearing protection was used for
blowoff tests on larger burners.

2.3.2 Leaky Fittings

Quenching limits were also measured for leaky compression fittings [31]. Leaks
were intentionally introduced. This simulates an unintended leak in a defective fuel line.
Presently no regulations exist for allowable leak rates in hydrogen plumbing systems.

The experiments employed Swagelok® stainless steel tube union compression

fittings for 3.2, 6.4, and 12.7 mm outside diameter stainless steel tubes. The end of the
union opposite the tube was capped such that the only flowing gas was that which went
through the leak. Leaks were introduced three ways into fittings that previously were
made according to manufacturer instructions and found to be leak tight. The first was
 12
by reducing the torque on the threaded nut. The second was by tightening the
threaded nut by an additional 0.75 turns. The third was by scratching the front ferrule
sealing surface. All three types of leaks are occasionally encountered in plumbing
systems.

Quenching limits were found by adjusting the torque on the threaded nut and
then adjusting the upstream pressure until extinction was observed. For pressures
above (and below) the quenching limit, an external flame impinging on the leak region
did (and did not) result in a persistent flame at the leak. Upstream pressure was
controlled with a pressure regulator in the range of 1.7 – 131 barg (24 – 1900 psig).
When a quenching limit was established, a tube was installed over the fitting such that
the leak flow passed into a soap bubble meter at laboratory pressure. The quenching
flow rate was then measured with the bubble meter. The burners were allowed to cool
to room temperature prior to each test.

For hydrogen, the flames generally were not visible even in dark conditions.
Existence or absence of persistent flames was determined with a thermocouple
positioned 2 cm above the burner. Unlike for methane and propane, a pop was heard at
ignition for hydrogen. For a flow rate (i.e., a pressure) below the quenching limit the
pop was followed by extinction when the external flame was removed. At slightly
higher pressures, this pop was followed by a persistent flame.

For methane and propane, quenching limits were identified visually – tests with
thermocouples confirmed that these flames were visible, even when only slightly above
their quenching limits. Most tests were performed with the burner in the vertical
orientation, with the leak at the top end of the tube. Horizontal and inverted
orientations also were considered for some 6.4 mm burner tests.

2.4 Results
 13
The following results have been divided into those for round hole burners and
those for leaky fittings.

2.4.1 Round Hole Burner Results

Figure 2.2 shows an image of a hydrogen flame on a 0.36 mm pinhole burner

just prior to extinction. The photo was taken in a dimly lit room with ƒ/4.2 and an
exposure time of 30 s. Despite the large aperture and long exposure time, the image is
still very faint. The hydrogen quenching distance of Table 1 suggests that steady
hydrogen diffusion flames should be anchored about 0.25 mm above the burner
surface. Figure 2.2 shows this is reasonable for the present flames near their quenching
limits.

0.356 mm

Figure 2.2 Contrast-enhanced image of hydrogen flame over a 0.356 mm pinhole burner
(original in color).
 14
1000
Blowoff Limits
100

Mass Flow Rate (mg/s)

10
H CH CH
2 4 3 8
This work
1 Matta
Cheng
Kalghatgi
0.1

0.0056 mg/s
Quenching Limits
0.01

0.001
0 0.5 1 1.5 2 2.5 3

Tube Diameter (mm)

Figure 2.3 Quenching and blowoff limits versus tube burner inner diameter. Hydrogen
quenching and blowoff limits include best fit line.

Figure 2.3 shows the limit data for tube burners. The blowoff flow rates
increase with tube burner diameter, and the quenching flow rates are generally
independent of diameter. The limits of combustion in Fig. 2.3 are much wider for
hydrogen than for propane or methane. The hydrogen quenching mass flow rate limits
are about an order of magnitude lower than those for methane and propane, while
hydrogen blowoff mass flow rate limits are about an order of magnitude higher than
those of the others. For a given leak diameter there is a range of low and high mass
flow rates that hydrogen would be able to support a flame while propane and methane
would be either quenched or blown off. The data from Matta et al. [14], Cheng et al.
[12], and Kalghatgi [18] show that the limits for methane and propane are very similar.
For all fuels, the quenching limits are nearly independent of burner diameter, whereas
blowoff limits increase with increasing diameter. The independence of quenching limit
from burner diameter is consistent with the analysis presented above, see Eq. (3).
 15
0.016
Pinhole burner
0.014 6.35 mm curved-wall pinhole burner

Quenching Mass Flow Rate (mg/s)

1.59 mm curved-wall pinhole burner
Tube burner
0.012

0.01

0.008

0.006

0.004

0.002

0
0 0.5 1 1.5 2 2.5 3 3.5
Burner Diameter (mm)

Figure 2.4 Quenching mass flow rates versus burner diameter for hydrogen.

Figure 2.4 shows the hydrogen quenching limits for all three types of burners.
The theory of Eq. (3) predicts that the quenching flow rate is independent of burner
diameter. Figure 2.4 shows that the scaling analysis approximately predicts the average
quenching flow rate for hydrogen, especially at burner diameters greater than 1.5 mm.
For smaller burners, there are clearly other mechanisms that are affecting quenching
limits that are not accounted for in the simple model.

Despite varying amounts of heat loss, the different burners display hydrogen’s
ability to sustain flames at very low mass flow rates. Nonetheless, heat loss contributes
to the differences in limits for different round hole burners. For small burner
diameters, the pinhole burners show the highest quenching flow rates, while the tube
burners show the lowest. Pinhole burners have more heat loss than tube burners due to
the geometry and the mass of the burner. Additionally, heat lost to the tube burners is
more likely to be transferred to the incoming fuel and oxidizer flows than heat lost to
the pinhole burners. This is not a loss as it acts to pre-heat the reactants, so that the
primary losses are due to radiation from the tube which are small as the tube
 16
temperature is near ambient. The results of the pinhole burners were similar to those
observed by previous investigators.

The tube burners show the same trend as the data from Matta et al. [19] for
propane in that there is an increase in quenching flow rate with increasing burner size.
Heat loss rate increases with increasing burner diameter; with more heat loss, the
quenching flow rate must increase to account for the weaker flame. The pinhole
burners do not exhibit the same trend as the tube burners because approximately the
same amount of heat loss occurs independent of burner hole size. The reason for the
local maximum in Fig. 2.4 for pinhole burners is under investigation, however at large
burner diameters, the quenching mass flow rate for both pinhole and tube burners
approaches 0.008 mg/s, as predicted by the scaling analysis.

The effect of curvature on heat loss was investigated by using the curved-wall
pinhole burners. It is expected that with increasing curvature, a flame will experience
less wall heat loss. This should result in a stronger flame and a lower quenching flow
rate. This is exactly what was observed for the two curved-wall burners. The 6.35 mm
burners act more like pinhole burners, whereas the 1.59 mm burners act more like tube
burners, and thus the 6.35 mm curved-wall burners have higher average quenching flow
rates than the 1.59 mm curved-wall burners.

The quenching mass flow rate for hole diameters less than 0.2 mm is actually
higher than that for 0.2 mm. This is true for both the pinhole and tube burners. We
believe that the flow field is affecting the quenching limits at these very small sizes. In
both cases, the miniscule inner diameters of the burners drastically increase the flow
velocities exiting the burner. While in the larger sizes, diffusion dominates at the
standoff distance, while that may not be so for the smallest diameters.
 17
0.008

Quenching Mass Flow Rate (mg/s)

0.007

0.006

0.005

Horizontal
Inverted
0.004 Vertical

0.003
0 0.5 1 1.5 2 2.5
Tube Diameter (mm)

Figure 2.5 Quenching mass flow rates for several orientations of tube burners.

The effect of burner orientation was investigated in Figs. 2.5 and 2.6. The
results reveal that the quenching flow rate for pinhole and tube burners is independent
of orientation. Our earlier studies found that quenching flow rate was independent of
burner temperature in the range of 20 °C to 200 °C. During these experiments, the
burner was found to be warmer during non-vertical tests. If burner temperature does
not affect temperature and it is the case that the quenching limit is not changing with
orientation, then the flow field must also be constant with varying orientation. Hence,
it was shown that flames near extinction are driven by mechanisms other than
buoyancy. The Froude numbers were calculated from:
Fr=ue2/gd, (4)
where ue is the fuel velocity at the burner port, g is the acceleration of gravity, and d is
the burner diameter. The values for the present pinhole and tube burner quenching
limit flames were found to be in the range of 0.17 to 3.9 E+10. Flames with the
smallest Froude numbers are within the transitional to buoyancy-controlled regime [11];
typical buoyancy-controlled diffusion flames have Fr less than 0.1. This suggests that
the flame structure at the quenching limits should not change with varying orientation.
 18
Furthermore, small changes in burner temperature have been shown not to affect the
quenching limits, so it is reasonable that the quenching limits are independent of
orientation.

0.012
Quenching Mass Flow Rate (mg/s)

0.01

0.008

0.006

0.004
Horizontal
Inverted
Vertical
0.002

0
0 0.5 1 1.5 2 2.5 3 3.5
Pinhole Burner Diameter (mm)

Figure 2.6 Quenching mass flow rates for several orientations of pinhole burners.

For most of the experiments the pressure drop across the burners was
negligible. Nonetheless, important parameters governing the quenching limit is mass
flow rate, and this can be related to the pressure drop across a small size leak. Thus the
conditions that would lead to a flammable leak can be inferred from the present data.
As an example, an inviscid, choked flow calculation of hydrogen mass flow rates
through a hole of known cross sectional area yields a plot shown in Fig. 2.7. The plot
shows data points where the mass flow rate is equal to the quenching mass flow rate
and assumes that a flame would be quenched at a mass flow rate of 0.008 mg/s. By this
approach the results in this work are applicable to a range of hydrogen systems after
also considering viscous effects, which will be important and will vary with tubing wall
thickness.
 19
60

50

40

Hole Area (μm )

2
30

20

10

0
0 2 4 6 8 10 12
Pressure (bar)

Figure 2.7 Area and pressure given at quenching limit for an inviscid, choked flow. The region
above the curve can sustain a flame, whereas the region below the curve can not sustain a flame.

2.4.2 Weakest Flame

The tube burners in this study had such low quenching flow rates that the
flames supported over them are believed to be the weakest flames ever recorded. The
smallest flame had a thermal output of 0.55 W. The thermal power output was
calculated using
P=mΔH, (5)
where m is the lowest mass flow rate (0.0039 mg/s) and ΔH is the heat of combustion
per mass (141.8 kJ/g).

Previously, Ronney et al. [32] documented flame balls produced under

microgravity conditions with a power output of about 1.3 W, which were believed to be
the weakest flames at that time. Flame balls were first predicted by Zeldovich [32], who
proposed that a solution exists to the steady heat and mass conservation equations
 20
corresponding to a stationary, spherical premixed flame. This solution was termed a
flame ball, and the phenomena was accidentally discovered forty years after Zeldovich’s
work by Ronney et al. in drop tower experiments with lean hydrogen-air mixtures [34].
The microgravity environment was necessary in order to obtain the spherical symmetry
and to avoid extinction brought on by buoyancy. In fact, flame balls have only been
achieved in microgravity conditions with very lean hydrogen mixtures. The group’s
results showed that they were able to obtain flame balls that were considered the
weakest flames ever recorded, at a heat release of about 1.3 W.

The hydrogen diffusion flames in this investigation were very weak, which
opens up the possibility of even weaker flames. In an attempt to achieve even weaker
flames, pure oxygen was used as the oxidizer. By inverting the tube, it became red hot
and helped to stabilize the flame. This is shown in Fig. 2.8 where flames were able to be
sustained with thermal outputs as low as 0.32 W. This shows that in the proper
environment, hydrogen leaks can be established that need very little hydrogen to remain
lit.

1 mm

H2 into O2 H2 into air

0.32 W 0.47 W
a) b)

Figure 2.8 Photos of weakest flames ever recorded; hydrogen flame in a) oxygen and b) air with
a tube burner. ID=0.15 mm, OD = 0.30 mm, ISO 200, f/1.4, 1 s.
 21
2.4.3 Numerical Solution

An attempt was made to numerically verifying the quenching limits. A

computer code, UNICORN, was chosen that was able to model hydrogen combustion
as well as other gaseous fuels. As will be shown, the code was not well suited for the
limit flow situations.

First, a grid was constructed as shown in Fig. 2.9. The grid was altered for
different inner diameters and outer diameters of the burner. Fig. 2.9 also shows that the
domain of the region being investigated is about 10 mm in the radial direction and 20
mm in the vertical direction.

Figure 2.9 Grid system for UNICORN code.

Next, flames were modeled for a series of flow rates near the quenching limit.
The numerical solutions found quenching limits that were too high at smaller pinhole
 22
burner diameters and too low for higher diameters. At higher diameters, the flame
would actually never extinguish - indicating a flaw in the code. At smaller diameters, the
flame would extinguish at flow rates that could easily sustain stable flames under
laboratory conditions. In order to investigate the cause of this discrepancy, photos were
taken and compared to computed OH concentrations.

Figure 2.10 shows the juxtaposition of near limit photos on the left as well as
computed OH concentrations on the right. It was anticipated that the flame would lie
close to where the OH radicals are at their highest concentrations, and the code predicts
the centerline location of the flame reasonably well, particularly at the lower flow rates.
The reason that the code can not even qualitatively predict quenching limits is
unknown.

Hole Diameter: 0.838 mm 1.01 mm

Flow Rate:

15 ccm

10 ccm

Figure 2.10 Photos compared to OH concentrations near quenching. The top row shows flames
with a flow rate of 15 cm3/min and the bottom row is 10 cm3/min. The columns are split up by
inner diameter.
 23

2.4.4 Leaky Fittings Results

Figure 2.11 shows images of hydrogen, methane, and propane flames on leaky
compression fittings in the vertical orientation. These images were recorded slightly
above the quenching limits. The hydrogen flame is significantly smaller than the others
at the quenching limit.

a) b) b)
Figure 2.11 Flames near extinction on leaky compression fittings for a) hydrogen, b) methane,
and c) propane (original in color).

Figure 2.12 shows the measured quenching limits for hydrogen, propane, and
methane for a leaky compression fitting in the vertical orientation. A flame was
established, then the pressure was decreased until the flame extinguished at the
quenching limit. The present quenching limits results were the same for all three types
of leaks, so only results from the reduced torque fittings are presented here. For each
fuel the data at increased pressures are associated with increased torque on the fitting.
The upper limit on pressure for propane is lower than for the others because the vapor
pressure of propane at 25 °C is 9.1 bar (142 psia).
 24
0.4

0.378 mg/s

Minimum Flowrate (mg/s)

0.3

0.336 mg/s

Hydrogen
0.2 Methane
Propane
h
m
p
0.1 0.028 mg/s Linear (h)
Linear (m)
Linear (p)

0
1 10 100 1000
Pressure (bar)

Figure 2.12 Quenching limit (minimum flaming flow rate) versus upstream pressure for a leaky
compression fitting in the vertical orientation. The horizontal lines represent the mean values of
the measured quenching limits for a given fuel.

Within experimental uncertainties, the data of Fig. 2.12 are independent of

pressure for each fuel. This indicates that, as anticipated from the round hole burner
results of Fig. 2.4, the important parameter for the quenching limit is the fuel mass flow
rate and thus the upstream pressure has little or no effect on the limit. For these low
flow rates and small leaks the upstream pressure is not expected to have a significant
effect on the velocity profile of the jet entering the surrounding air. The mean hydrogen
flow rate, 0.028 mg/s, is about an order of magnitude lower than for the other fuels
owing to its low quenching distance and low molecular weight. For these leaks, there is
a range of pressures where hydrogen is able to support a flame while propane and
methane flames would be quenched.

The minimum flow rate to support a hydrogen flame for round burners (Fig.
2.4) is about an order of magnitude lower than that for leaky fittings (Fig. 2.12). This is
attributed to additional burner heat losses in the leaky fittings, where the flames burn
near large concave metal surfaces.
 25
0.4

Vertical
0.3 Horizontal

Minimum Flowrate (mg/s)

Inverted

0.2

0.1

0
Hydrogen M ethane Propane

Figure 2.13 Minimum flaming flow rate for a leaky compression fitting in vertical, horizontal
and inverted orientations. Upstream pressure was about 4 bar in all cases.

Figure 2.13 shows the effect of burner orientation on the quenching flow rate.
Burner orientation has little or no effect on the quenching limit of hydrogen because
these flames are so small at their limits as was shown in section 2.4.2. Burner
orientation did have an effect on propane and methane quenching limits, the inverted
fitting configuration requiring the lowest flow rate. The inverted configuration is such
that the fitting is below the tube, opposite to what is shown in Fig. 2.11. This
orientation minimized the impingement of flames on the surface of the fitting which
allowed less heat to be lost to the fitting.
 26
0.6
3.1mm Burner
6.3mm Burner
0.5
12.6mm Burner

Minimum Flowrate (mg/s)

0.4

0.3

0.2

0.1

0
Hydrogen Methane Propane

Figure 2.14 Minimum flaming flow rate for leaky compression fittings of 3.1 mm, 6.3 mm and
12.6 mm outside diameters. Upstream pressure was approximately 4 bar.

Figure 2.14 shows the effect of the tube fitting size on the minimum flaming
flow rate for fitting in the vertical orientation. The tube fitting size is characterized by
the tube diameter that the fitting is designed for. The tube fitting size played a large role
in the minimum flaming flow rate for all fuels. There is 48% increase from the smallest
to the largest burner for hydrogen, 29% increase for methane and a 45% increase for
propane. This is due to the increased burner surface area that the flame loses heat to. It
could also be due to the fact that the same amount of fuel is evenly distributed over a
larger area hence diluting it. As the tube fitting gets larger, a larger flow is necessary to
sustain combustion.

2.5 Conclusions
The quenching and blowoff limits for hydrogen on round hole burners were
measured and compared with both our data and published values for other gases.
Hydrogen diffusion flames on tube burners had significantly wider mass flow rate limits
of combustion (between quenching and blowoff) than propane and methane. The
 27
quenching flow rates for all these gases are largely independent of burner size,
especially at the larger burner diameters.

The quenching flow rate was found to depend on burner type, owing to
differences in wall heat loss. Tube burners had the lowest quenching flow rates, pinhole
burners had the highest, and curved wall burners fell in-between. Although the type of
burner did affect the quenching mass flow rate, the results showed that a hydrogen leak
can sustain a flame at much lower flow rates than the other fuels for any burner type.
The hydrogen flames investigated were the weakest flames in air ever documented with
a thermal power output of 0.55 W.
Quenching limits for both burner types were found to be largely independent of burner
orientation. As a result, a leak in a hydrogen system would be able to sustain a flame at
low flow rates independent of the orientation of the leak.

The minimum flow rate necessary for sustaining a hydrogen flame on leaky ¼”
tube compression fitting is 0.028 mg/s. This is about an order of magnitude lower than
for propane or methane. The minimum mass flow rate for all fuels is independent of
upstream pressure.

The round burner and leaky compression fitting experiments yielded similar
results in that they both found that the lower limits of combustion for hydrogen were
about an order of magnitude lower than those for propane and methane. The two
experiments differed in that their lower combustion limits differed by an order of
magnitude owing to increased heat losses in the leaky compression fittings. In both
cases, hydrogen was able to sustain flames at much lower mass flow rates than methane
and propane.
 28

Chapter 3

Materials Degradation from a Sustained

Leak Flame

3.1 Introduction

Considering that hydrogen is more likely to be able to sustain a leak flame than
other common gaseous fuels, the effects of a steady, hydrogen leak flame must also be
considered. Furthermore, hydrogen is known to embrittle materials. This can result in
decreased fatigue resistance and failure. These effects can also be increased by when a
material is exposed to hydrogen at high pressures or temperatures. If a fitting or other
component containing hydrogen at high pressures is exposed to a leak flame, the
conditions could be present to degrade the material. Also, because of hydrogen’s low
luminosity, the probability of detecting a leak flame is minimal. If the material is
allowed to degrade for long periods of time, this could possibly result in catastrophic
failure. The scenario under investigation is material degradation due to a sustained
hydrogen leak flame that is allowed to exist for a significant length of time.

Although this application of material degradation may be novel, the study of

hydrogen degradation of materials is not. Nonetheless, the exact methods of hydrogen
degradation are not fully understood. There are several evaluations for material
degradation, but they all have common steps. Due to its low molecular weight,
hydrogen is capable of being adsorbed and absorbed by the material. Marchi et al. [51]
studied permeation rates of hydrogen in steels because this is the expected cause of
hydrogen degradation. They found that there can be significant amounts of hydrogen
 29
that diffuse through steels especially at high temperatures and pressures. There are
many ways that hydrogen can enter a metal including electroplating, acid pickling and
cathodic charging [44]. The hydrogen typically exists as atomic hydrogen once inside
the metal. Minimum energy calculations show that it is advantageous for the atoms to
form clusters. In addition to its localization tendencies, quantum mechanical
calculations show that it is favorable for hydrogen to exist near structure defects and
interfaces, due to their low electron densities. The following are possible next steps in
the degradation process depending upon the type of metal:

• The atomic hydrogen can then form hard, brittle hydrides [45],
• There can be reactions between hydrogen and an impurity or alloy addition (e.g.
2H + O → H2O in copper and 4H + C → CH4 in steel),
• A decrease in the adherence forces can result from the hydrogen clusters
(decohesion),
• Molecular hydrogen can precipitate in the microcracks or microvoids creating
high pressures, which can stabilize those structures [46,49].

The net result of these processes is that materials exposed to hydrogen tend to become
embrittled, which includes loss of ductility and decreased stress at failure, as has been
observed by Wang et al. [36].

Hydrogen embrittlement has been found in different materials including those

that are potentially involved in hydrogen systems of today and those of the future.
Züttel [37] notes that future pressure vessels may contain layers of carbon fiber and
epoxy composites, which have been shown to degrade in a hydrogen environment at
elevated temperatures. Khan et al. [21] studied a woven carbon fabric/epoxy composite
material exposed to hydrogen and found that at increased temperature (tests went up to
150 °C) there was a significant decrease in the fatigue resistance of the material. Under
a sinusoidal load cycling at a frequency of 20 Hz, fatigue resistance decreased by two
orders of magnitude, 2.0E+05 to 2.0E+03 cycles, by increasing the temperature from
 30
25 °C to 100 °C at a maximum stress level of 380 MPa. They also found that fatigue
strength at 10E+6 cycles decreased from 400 MPa at 25 °C to 250 MPa at 150 °C. This
material’s strength properties are clearly negatively affected by even small increases in
temperature above 25 °C.

In addition to hydrogen storage materials, there will inevitably be steels used in

hydrogen systems for plumbing, which have their own risks for material degradation. It
has been shown that hydrogen can have deleterious effects on many steels. Vencill et al.
[37], in attempting to create a typical hydrogen environment in a fuel cell reformer,
found that hydrogen can significantly embrittle some stainless steels including 304L,
310, and 253 MA along with Inconel 625, Inconel 718, and Hastelloy X nickel-based
alloys, and Haynes 188 cobalt-based alloy. They showed that steel that has been
subjected to an environment comparable to that of a fuel cell reformer (625 °C for up
to 1500 hours) can fail at lower strains than the control specimens. A potential
explanation of this was proposed by Oriani [42] in that microvoids can accumulate
precipitated hydrogen at very high pressures. Panasyuk et al. [42] also performed an
analysis of hydrogen systems, and they found that significant pressures within
microcavities can be formed if certain conditions are present. This is one potential
cause for the decrease in ductility found in tensile tests by many investigators [40,41].
The tests show a drastic decrease in area ratio for the hydrogenated material from the
material not exposed to hydrogen, where area ratio is defined as the area of the
specimen before stretching divided by the area afterwards.

These effects can be strongly altered by several variables, including pressure.

Zheng noted that future hydrogen refueling stations may operate at pressures up to 100
MPa, so material selection must take this into consideration. Making matters worse is
that there is a proportional relationship between steel susceptibility to hydrogen
embrittlement and steel strength [45]. Several studies have found that this is a critical
issue because stainless steels showed significant losses in ductility after being exposed to
a high pressure (68.9 MPa) hydrogen environment [46,47].
 31

Since we are considering a leak flame, the effects of elevated temperatures must
be considered as well. Yokogawa et al. [48] performed tensile tests on metals exposed
to elevated temperatures in a hydrogen environment. They found that, depending upon
the steel type, there can be extreme reductions in ductility in samples that were found to
be embrittled by hydrogen.

There is also an increased susceptibility to hydrogen attack when stress is

present in steel. The reason for this is probably due to the increased hydrogen
permeation rate in steel with increased stress. Beck et al. [50] found that this increase is
largely dependent on the amount of stress applied over a range of 0 to 100 kg/mm2.
This means that for even small stresses, say around 5 kg/mm2, the hydrogen permeation
rate and hence hydrogen attack will increase. This permeation rate is probably related
to the findings of Rozenak et al. [53]. They noted that stress can be responsible for
crack advance, phase transformation, and localization of defects in austenitic stainless
steel. Austenitic steels are those typically used in general engineering applications. It
was found that the presence of atomic hydrogen in steel decreased the amount of stress
required for these phenomena to occur.

There are also intrinsic hydrogen embrittlement risks associated with welds that
are acknowledged by safety codes. During welding, hydrogen can be trapped in the
molten metal of the weld, and the process can force hydrogen to diffuse into the heated
portion of the parent material. This leads to embrittlement of both the weld and the
surrounding material. Safety codes require that material testing be conducted 72 hours
after the weld was performed to allow hydrogen cracking to develop [50]. Tsay et al.
[52] found that steel welds containing atomic hydrogen had decreased tensile strengths
compared to those welds that did not contain hydrogen. The same effect was found for
the material surrounding the weld.
 32
The scenario under consideration here is that a leak flame is established and
the surrounding material, whether it be steel or a composite material, becomes degraded
by the hydrogen diffusion flame and could potentially lead to catastrophic failure.

Since the application of this work is meant to be general, the exact placement of
the material in the flame is not known. Therefore, varying degrees of exposure to a leak
flame are considered.

3.2 Experimental

3.2.1 Burner Configuration and Temperature

Measurement

Two experiments were run simultaneously on two different metals, 316 stainless
steel and 1100 aluminum. Three holes were drilled in 3.18 mm outer diameter tubes.
The holes were drilled such that one hole was issuing hydrogen in the upward, vertical
direction, one was oriented horizontally and the third was oriented vertically downward.
The inner diameter of the drilled holes in the steel were 1.0 mm, and 1.1 mm for the
aluminum. Flames were established over each of the holes and the setup was allowed to
run continuously for 44 days for the steel and 73 days for the aluminum.

For safety, the flow system only consisted of a pressure regulator and a flow
restrictor to guard against failure or accidental opening of a valve. A three-sided steel
barrier was constructed to guard against accidental exposure to the experiment. There
were a few days throughout the experiment where the flame was extinguished due to
lack of hydrogen. The experiment was set up as to allow for thermal expansion without
constraint; stress was not desired as a variable in the system.
 33
To understand the environment that material in a hydrogen system might be
exposed to in the case of a leak flame, a temperature profile was measured for a
hydrogen diffusion flame. The flame size was chosen such that its luminosity was
undetectable by the naked eye in normal laboratory lighting and not especially bright in
dim conditions. The flame represents a leak flame that could go undetected in a
hydrogen system.

A B-type thermocouple was used with an inert coating to inhibit catalytic

effects. A radiation correction was imposed to account for radiative losses from the
finite size of the thermocouple bead. The correction assumed a sphere in forced
convection. The radiation correction was as high as 180 °C at the peak temperature
reading but decreased with lower temperatures. This correction is a little larger than
ideal because the bead diameter was 0.2 mm. The thermocouple was translated and the
voltage read using TestPoint data acquisition software. The accuracy of the readings was
validated by directly comparing the voltage reading to an Omega thermocouple
temperature table. The distance between temperature measurements in both the vertical
and horizontal direction was 0.2 mm and was checked using a cathetometer. To
minimize disturbances to the small flames due to air currents, the burner was housed in
a four-sided plexiglass chamber.

3.2.2 Materials Characterization

The tube specimens were cut up and placed into an epoxy that was then allowed
to set. The sample was then polished beginning with a rough grade paper and ending
with an alumina polish. During the experiments, it was observed that the center hole
had a smaller flame size due to head losses. The flow circuit was set up so that
hydrogen was being delivered to both ends of the tubes, but the flow experienced a
pressure drop by the time it got to the center orifice. Thus, the degradation effects
 34
around the center hole were largely ignored because they were not in the same
conditions as their relative outer holes.

3.3 Results

3.3.1 Temperature Measurement

Figure 3.1 shows the results of the flame temperature contours for a pinhole
burner. These results are most applicable to the present materials degradation
experiment in that the surface of the tubes somewhat resembles the slight curvature of
the pinhole burners. The fluid temperature can reach up to 1503 °C. The conditions
0.5 mm from the surface of the burner are over 1200 °C.

12

1600

10

1400
Vertical distance (mm)

1200
Vertical distance (mm)

6
1000

800
4

600

400

-3 -2 -1 0 1 2 3 4
Horizontal distance (mm)

Horizontal distance (mm)

Figure 3.1 Temperature profile of a 70 cm3/min hydrogen flame. The temperature units
are in K.
 35

Temperature results were obtained for a tube burner to evaluate the effect of a
burner with less material near the flame, and are shown in Fig. 3.2. The temperatures
near the burner are significantly greater than for the pinhole burner. This was inferred
from the high temperatures very close to the tube burner. The tube itself also glowed
red indicating its high temperature. The wall temperature of the burner was found by
wrapping a thermocouple around the tip of the tube. The average of the readings was
around 900 °C. The roughness of the isotherms is caused by random ambient
fluctuations of the air within the chamber. Precautions were taken as to minimize these,
but they were difficult to completely
eliminate.

12 2000

1800
10

1600

1400

6
Vertical distance (mm)
Vertical distance (mm)

1200

1000

2
800

0 600

400
-2

-3 -2 -1 0 1 2 3
Horizontal distance (mm)

Horizontal distance (mm)

Figure 3.2 Temperature profile of 70 cm3/min hydrogen diffusion flame issuing from a tube
burner.
 36
2.3.2 Numerical Solution

Similar to the quenching limit numerical survey, numerical results were sought
that would verify the experimental temperature profile. The code was used to model
pinhole burners, which are geometrically similar to the tubes in the present experiment.
There is some curvature of the pinhole burners, but there is also some curvature
involved in the aluminum and steel tube specimens. Figure 3.3 exposes some first
glance issues with the numerical solution. It is immediately apparent that the flame
width, height, and standoff distance are dissimilar. This comparison corroborates the
issues experienced with finding quenching limits numerically. Clearly, if the computer
code is not able to accurately model the flame, then it can not accurately model the
quenching limit. The same problem exists with the tube burners shown in Fig. 3.4.
They are on the same temperature scale, which indicates that the flame temperatures are
different.

1600
8

1400

6
1200

1000

800

2
600

400

-3 -2 -1 0 1 2 3 4

Figure 3.3 Experimental profile of 70cm3/min hydrogen flame on left with numerical solution
on right over pinhole burner.
 37

2000
8

1800

6
1600

1400
4

1200

2
1000

800
0

-2

-3 -2 -1 0 1 2 3

Figure 3.4 Experimental profile of 70 cm3/min hydrogen flame on left with numerical solution
on right over tube burner.

Photographs of flames were taken and compared to the temperature profiles.

Unlike the numerical results, the experimental temperature measurements accurately
represented the basic flame shapes. The numerical flame is significantly smaller than
the actual flame.
 38
6

0
-3 -2 -1 0

-2

-3 -2 -1 0 1 2 3

Figure 3.5 Numerical temperature solution on left and experimental temperature measurements
on right compared to photo. The first row is pinhole burners, and the second is tube burners.

3.3.3 Materials Characterization

The parts of the tube that were studied were several millimeters away from the
hole where the material was at the highest temperatures. Figure 3.6 includes the photos
taken of the aluminum control specimen. In Figs. 3.7 and 3.8, which are for aluminum
in the vertical and downward flame orientation, it is clear that there are dark spots that
don’t exist in the control specimen.
Temperature readings of the material were 490 °C and 640 °C of the vertical
flame and inverted flame specimen, respectively. The temperature readings have limited
accuracy due to the difficulty of the measurements, but it is clear that the temperatures
approach the melting point of aluminum, 660 °C. 1100 aluminum contains .35 wt. %
 39
iron and 0.25 wt. % silicon. By examining aluminum phase diagrams, it was found that
the dark spots are most likely second phase material. Aluminum and iron form Fe3Al
and aluminum and silicon form Si particles within the material at the present
temperatures [55].
To evaluate the effect on the mechanical properties of the material, the
hardness of the samples was measured. The dimensions of a micro depression made
with a known load were found. The control aluminum specimen yielded a hardness of
175 HV while the specimen with the inverted flame had a hardness of 62 HV. Both of
these values are average values over six trials excluding the maximum and minimum
data points because each set had a few outliers. This decrease in hardness is consistent
with the transition from a metal that is in the worked condition to a metal in the
annealed condition. The result is the annealing of the deformation substructure and a
coarsened microstructure. Although, the effects are most likely due to the
temperatures, it can not be ruled out that the effects of hydrogen embrittlement have
not played a role in the degradation. Also, within a hydrogen flame, there are highly
reactive radicals that may also have had an effect.

200 μm 40 μm

Figure 3.6 Microscopic images of the aluminum control specimen.

 40

200 μm 40 μm

Figure 3.7 Microscopic images of an aluminum specimen exposed to a hydrogen,

vertical flame.

200 μm 40 μm

Figure 3.8 Microscopic images of an aluminum specimen after being exposed to a hydrogen,
inverted flame.

Figures 3.9-3.11 show micrographs of the stainless steel specimens. As seen

from Fig. 3.10, which is for the vertical upwards orientation, the steel does not show
significant degradation as compared to the aluminum. Nonetheless, Figs. 3.10 and 3.11,
which is for the vertical flame and inverted flame on the steel, show signs of material
degradation. Considering that the temperature readings were 640 °C and 490 °C, both
cases are likely to have the formation of carbides due to the chromium content [56].
This is most likely what the dark spots are. The same argument applies for the steel in
that the effects are most likely due to temperature, but this experiment allows for the
interaction of metal and radicals in the flame.
 41

200 μm 40 μm

Figure 3.9 Microscopic images of the control, steel specimen.

200 μm 40 μm

Figure 3.10 Microscopic images of the steel specimen exposed to a vertical flame.

200 μm 40 μm

Figure 3.11 Microscopic images of the steel specimen exposed to an inverted hydrogen flame.
 42
Hardness tests were performed to investigate the effects of the material
degradation. The hardness was tested as a function of distance from the surface where
the flame was anchored. The results were all within a normal standard deviation.

3.5 Conclusions
The results show that there is significant degradation in 1100 aluminum
specimens’ hardness results. The stainless steel showed a slight amount of degradation
in that second phase material developed. The hardness tests performed on the stainless
steel showed insignificant effects from the experiment while the aluminum showed a
significant reduction in hardness.
While the results of this study may not be conclusive, it is readily apparent that a
hydrogen leak flame can degrade metals. Aluminum was used as a control experiment,
and while this material may not be used in hydrogen systems, it shows that metals can
be degraded by a hydrogen leak flame. The stainless steel experiment showed visual
signs of degradation but negligible changes in hardness, but it is anticipated that greater
degradation would occur with time.

Chapter 4

Future Work
In regards to the quenching limit investigation, several more issues have been
produced in addition to the ones answered. The study considered only round burners
and leaky Swagelok fittings; there are many other types of burners and leaks that could
represent real world hydrogen leaks. For instance, slot burners could better represent a
leak from a crack considering that a crack will likely have a high aspect ratio. This study
 43
showed that there are different amounts of heat loss depending upon the geometry of
the burner, so different geometries are critical to quenching limits.
The materials degradation study emphasized the significance of the type of
material used in hydrogen systems. Further work is needed to fully investigate the types
of materials that are used in current hydrogen systems and those that are predicted to be
used in the future. Similar experiments can be run using an array of different materials
to show the optimal material for hydrogen systems considering its corrosive properties.
There are several different methods of hydrogen storage that have safety issues that
have not been investigated. In the future, hydrogen sorption techniques could be used
to store hydrogen. Also, if fuel cell vehicles become popular, then fittings within them
have to be gas tight as well as protected in the case of a crash. The time frame of this
experiment was just a few months, but the situation could exist where a leak goes
undetected for years. Thus, questions arise as to the effects of a leak flame over long
periods of time. Additionally, the effects of temperature were lumped with the effects
of hydrogen attack. More experiments are needed to distinguish the two variables.
 44

References
1 Gregorio Marban and Teresa Valdes-Solis. Towards the hydrogen economy? Int
J Hydrogen Energy, 32:1625-1637, February 2007.

2 Michio Yamawaki, Tetsuo Nishihara, Yoshiyuki Inagaki, Kazuo Minato,

Hiroyuki Oigawa, Kaoru Onuki, Ryutaro Hino, Masuro Ogawa. Application of
nuclear energy for environmentally friendly hydrogen generation. Int J Hydrogen
Energy, 32:2719-2725, November 2006.

3 S.A. Sherif, F. Barbir, T.N. Veziroglu. Wind energy and the hydrogen economy
- review of the technology. Solar Energy, 78:647-660, March 2005.

4 Annette Von Jouanne, Iqbal Husain, Alan Wallace, Alex Yokochi. Gone with
the wind. IEEE Industry Applications Magazine, 12-19, July/August 2005.

5 Paul Kruger. Alternative energy resources: the quest for sustainable energy.
John Wiley and Sons, Inc., Hoboken, NJ, 2006, 174-175.

6 I. MacIntyre, A.V. Tchouvelev, D.R. Hay, J. Wong, J. Grant, P. Benard.

Canadian hydrogen safety program. Int J Hydrogen Energy, 32:2134-2143, May
2007.

7 K. Takeno, K. Okabayashi, A. Kouchi, T. Nonaka, K. Hashiguchi, K. Chitose.

Dispersion and explosion field tests for 40 MPa pressurized hydrogen. Int J
Hydrogen Energy, 32;13:2144-2153, June 2007.

8 S.B. Dorofeev. Evaluation of safety distances related to unconfined hydrogen

explosions. Int J Hydrogen Energy, 32:2118-2124, May 2007.

9 John Baker, Mark E. Calvert, David W. Murphy. Structure and dynamics of

laminar jet micro-slot diffusion flames. Journal of Heat Transfer, 124:783-790,
August 2002.

10 F.G. Roper. The prediction of laminar jet diffusion flame sizes: Part 1.
Theoretical model. Combust. Flame, 29:219-226, January 1977.

11 H. Ban, S. Venkatesh, K. Saito. Convection-diffusion controlled laminar micro

flames. Transactions of the ASME, 116:954-959, November 1994.

12 T.S. Cheng, Y.-C. Chao, C.-Y. Wu, Y.-H. Li, Y. Nakamura, K.-Y. Lee, T. Yuan,
T.S. Leu. Experimental and numerical investigation of microscale hydrogen
diffusion flames. Proceedings of the Combustion Institute, 30:2489-2497, 2005.
 45
13 Y. Nakamura, H. Yamashita, K. Saito. A numerical study on extinction
behaviour of laminar micro-diffusion flames. Combustion Theory and Modelling,
10;6:927-938, December 2006.

14 I.D. Lee, O.I. Smith, A.R. Karagozian. Hydrogen and helium leak rates from
micromachined orifices. AIAA Journal, 41;3:457-463, March 2003.

15 R.W. Schefer, W.G. Houf, C. San Marchi, W.P. Chernicoff, L. Englom.

Characterization of leaks from compressed hydrogen dispensing systems and
related components. Int J Hydrogen Energy, 31:1247-1260, August 2006.

16 Xinyu Ge, William H. Sutton. Analysis and test of compressed hydrogen

interface leakage by commercial stainless steel (NPT) fittings. SAE International,
35-47, 2006.

17 M.R. Swain, M.N. Swain. A comparison of H2, CH4, and C3H8 fuel leakage in
residential settings. Int J Hydrogen Energy, 17:807-815, May 1992.

18 Gautam T. Kalghatgi. Blow-out stability of gaseous jet diffusion flames. Part I:

In still air. Combustion Science and Technology, 26;5:233-239, April 1981.

19 L.M. Matta, Y. Neumeier, B. Lemon, B.T. Zinn. Characteristics of microscale

diffusion flames. Proc. Combust. Inst., 29:933-938, 2002.

20 T.S. Cheng, C.-P. Chen, C.-S. Chen, Y.-H. Li, C.-Y. Wu, Y.-C. Chao.
Characteristics of microjet methane diffusion flames. Combust. Theory Modeling,
10:861-881, October 2006.

21 Rehan Khan, Z. Kan, F. Al-Sulaiman, N. Merah. Fatigue life estimates in

woven carbon fabric/epoxy composites at non-ambient temperatures. Journal of
Composite Materials, 36;22:2517-2535, April 2002.

22 K. Pehr. Aspects of safety and acceptance of LH2 tank systems in passenger

cars. Int J Hydrogen Energy, 21;5:387-395, August 1995.

23 Vivek Utgikar, Todd Thiesen. Safety of compressed hydrogen fuel tanks:

Leakage from stationary vehicles. Technology in Society, 27:315-320, 2005.

24 L.C. Cadwallader, J.S. Herring. Safety issues with hydrogen as a vehicle fuel.
Report INEEL/EXT-99-00522 prepared for U.S. Department of Energy, September
1999.

25 P.B. Sunderland, B.J. Mendelson, Z.G. Yuan, D.L. Urban. Shapes of buoyant
and nonbuoyant laminar jet diffusion flames. Combust. Flame, 116:376-386, 1999.
 46
26 A.M. Kanury. Introduction to combustion phenomena. Gordon and Breach, New
York, 1975, p. 131.

27 R.C. Weast, M.J. Astle. CRC Handbook of Chemistry and Physics, CRC Press, Inc.,
West Palm Beach, 59th edition, 1979, P. F-58.

28 B.R. Munson, D.F. Young, T.J. Okiishi. Fundamentals of fluid mechanics. Wiley,
Hoboken, 2002, p. 359.

29 N.R. Morton, P.B. Sunderland, B.H. Chao, R.L. Axelbaum. Quenching limits
and materials degradation of hydrogen diffusion flames. 5th U.S. Combustion
Meeting held in San Diego March 25-28 2007, Paper #25.

30 M.Y. Bahadori, U. Hegde, L. Zhou. Liftoff and blowoff of jet diffusion flames.
Paper AIAA-2002-1078, Aerospace Sciences Meeting held in Reno, Nevada January 14-
17 2002.

31 Chris Moran, Flame quenching limits of hydrogen leaks. M.S. Thesis, University
of Maryland., May 2008.

32 Paul D. Ronney, Ming-Shin Wu, Howard G. Pearlman. Experimental study of

flame balls in space: preliminary results from STS-83. AIAA Journal, 36;8:1361-
1368, August 1998.

33 Ya Zeldovich. Theory of Combustion and Detonation of Gases. Academy of

Science (USSR), Moscow, 1994.

34 P.D. Ronney. Near-limit flame structures at low Lewis Number. Combustion and
Flame, 82:1-14, 1990.

35 P.D. Ronney, K.N. Whaling, A. Abbud-Madrid, J.L. Gatto, and V.L. Pisowicz.
Stationary premixed flames in spherical and cylindrical geometries. AIAA
Journal, 32;3:569-577, 1994.

36 Maoquiu Wang, Eiji Akiyama, Kaneaki Tsuzaki. Effect of hydrogen on the

fracture behavior of high strength steel during slow strain rate test. Corrosion
Science, 49: 4081-4097.

37 Thomas R. Vencill, Walter H. Moy, Yu-Lin Shen, Shailendra B. Rathod.

Thermal embrittlement characterization of metallic alloys in high temperature
hydrogen environments. Materials Science and Technology, 1:189-197, 2006.

38 Andreas Züttel. Hydrogen storage methods. Naturwissenschaften, 91:157-172,

March 2004.
 47
39 John L. Cox. What steel to use at high pressures and temperatures. Chemical
and Metallurgical Engineering, 405-409, August 1933

40 E. Minkovitz, D. Eliezer. Hydrogen-assisted cracking of sensitized 316L

stainless steel. Journal of Materials Science, 16:2507-2511, 1981.

41 Anthony W. Thompson. Hydrogen embrittlement of stainless steels by lithium

hydride. Metallurgical Transactions, 4:2819-2825, December 1973.

42 R.A. Oriani. Hydrogen embrittlement of steels. Annual Reviews in Material

Science, 8:327-357, 1978.

43 V.V. Pansyuk, O.Y. Andreykiv, O.V. Gembara. Hydrogen degradation of

materials under long-term operation of technological equipment. International
Journal of Hydrogen Energy, 25:67-74, 2000.

44 T.J. Carter, L.A. Cornish. Hydrogen in Metal. Engineering Failure Analysis, 8:113-
121, October 1999.

45 Jinyang Zheng, Lei Li, Rui Chen, Ping Xu, Fangming Kai. High pressure steel
storage vessels used in hydrogen refueling station. Journal of Pressure Vessel
Technology, vol. 130, February 2008.

46 M.R. Louthan, G.R. Caskey, J.A. Donovan, D.E. Rawl. Hydrogen

embrittlement of metals. Materials Science and Engineering, 10:357-368, June 1972.

47 D.D. Perlmutter, B.F. Dodge. Effects of hydrogen on properties of metals.

Industrial and Engineering Chemistry, 885-893, May 1956.

48 K. Yokogawa, S. Fukuyama, G. Han and J. He. Hydrogen environment

embrittlement of materials for hydrogen energy service. Hydrogen Energy Progress,
3:2291-2295, 1996.

49 V.I. Tkachov. Effect of hydrogen on the properties of structural steels.

Materials Science, 41;4:107-110, July-August 2005.

50 W. Beck, J. O’M. Bockris, J. McBreen, L. Nanis. Hydrogen permeation in

metals as a function of stress, temperature and dissolved hydrogen
Concentration. Proceedings of the Society of London. Series A, Mathematical and Physical
Sciences, 1;1421:220-235, February 1966.

51 C. San Marchi, B.P. Somerday, S.L. Robinson. Permeability, solubility and

diffusivity of hydrogen isotopes in stainless steels at high gas pressures.
International Journal of Hydrogen Energy, 32:100-116, May 2006.
 48
52 L.W. Tsay, S.C. Yu, San-Der Chyou, D.-Y. Lin. A comparison of hydrogen
embrittlement susceptibility of two austenitic stainless steel welds. Corrosion
Science, 49:4028-4039, April 2007.

53 P. Rozenak, I.M. Robertson, and H.K. Birnbaum. HVEM studies of the effects
of hydrogen on the deformation and fracture of AISI type 316 austenitic
stainless steel. Israel Journal of Technology: Proc. Of the 4th Israel Materials Engineering
Conference, 24:231-236, 1988.

54 M.R. Louthan, Jr. and A.H. Dexter. Hydrogen embrittlement of 1100

aluminum. Metallurgical Transactions A, 6A:1655-1657, August 1975.

55 Metals Handbook, vol. 8: Metallography, Structures and Phase Diagrams, USA, 1973.

56 William F. Smith. Foundations of Materials Science and Engineering, McGraw-Hill,

New York, 2004, p. 497.
 49

Vita
Michael S. Butler, Jr.

Date of Birth December 25, 1983

Place of Birth St. Louis, MO

Degrees B.A. Cum Laude, Physics, May 2006

M.S. Mechanical Engineering, May 2008

Publications M.S. Butler, C.W. Moran, P.B. Sunderland, R.L. Axelbaum, Flame
Quenching Limits of Hydrogen Leaks, SAE World Congress, Detroit,
SAE Paper 2008-01-0726 (2008) 8 pp.

M.S. Butler, C.W. Moran, P.B. Sunderland, R.L. Axelbaum,

Quenching Limits of Hydrogen Diffusion Flames, Eastern States Section
of the Combustion Institute, Charlottesville, October 21-24 (2007).

May 2008
 50

Short Title: Flame Quenching Limits and Materials Degradation of Hydrogen

Leaks
 Author's personal copy

international journal of hydrogen energy 34 (2009) 5174–5182

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/he

Limits for hydrogen leaks that can support stable flames

M.S. Butlera, C.W. Moranb, P.B. Sunderlandb,*, R.L. Axelbauma

a
Department of Energy, Environmental and Chemical Engineering, Washington University, St. Louis, MO 63130, USA
b
Department of Fire Protection Engineering, University of Maryland, College Park, MD 20742, USA

article info abstract

Article history: Quenching and blowoff limits of hydrogen diffusion flames on small burners were
Received 31 January 2009 observed. Four burner types, with diameters as small as 8 mm, were considered: pinhole
Received in revised form burners, curved-wall burners, tube burners, and leaky fittings. In terms of mass flow rate,
4 April 2009 hydrogen had a lower quenching limit and a higher blowoff limit than either methane or
Accepted 5 April 2009 propane. Hydrogen flames at their quenching limits were the weakest flames recorded to
Available online 9 May 2009 date, with mass flow rates and heat release rates as low as 3.9 mg/s and 0.46 W. The
quenching limit for a hydrogen flame at a 6 mm leaky compression fitting was found to be
Keywords: 28 mg/s. This limit was independent of supply pressure (up to 131 bar) and about an order of
Compression fittings magnitude lower than the corresponding limits for methane and propane.
Fire safety ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights
Laminar flames reserved.
Leakage
Quenching limits
Microcombustion
Microflames

1. Introduction in air (2.91 m/s), and the highest heat of combustion

(119.9 kJ/g) [4]. Hydrogen flames are the dimmest of any fuel.
Hydrogen is attractive as an energy carrier for highway vehi- Hydrogen embrittles and attacks metals more than any other
cles. It can power fuel cells or engines with only water vapor fuel.
as exhaust. With oil supplying 33% of the world’s primary Hydrogen may ultimately prove to be no more hazardous
energy, hydrogen could help mitigate concerns over fossil fuel from a fire safety standpoint than gasoline or diesel. However,
consumption if it is produced from renewable energy sources gasoline and diesel have undergone over a century of wide-
or nuclear energy [1,2]. Hydrogen can be stored as a gas, liquid, spread vehicle use, and this has resulted in codes and stan-
or a solid (in metal hydrides), and can be transported using dards that have yielded an acceptable fire safety record.
pipelines, tankers or rail trucks [3]. Further research is necessary if hydrogen is to be rapidly
Hydrogen is an unusual fuel. It has a high leak propensity introduced with a similar safety record. It is expected that
and wide flammability limits, 4–75% by volume [4]. Among existing codes and standards will require updating to ensure
all fuels, hydrogen has the lowest molecular weight, the safe use of hydrogen in highway vehicles [5].
lowest quenching distance (0.51 mm), the smallest ignition The goal of this study is to investigate small hydrogen
energy in air (28 mJ), the lowest auto-ignition temperature by flames–flames that can be characterized as ‘‘leak flames,’’ i.e.,
a heated air jet (640  C), the highest laminar burning velocity small flames that could exist as a result of hydrogen leaking

* Corresponding author. Tel.: þ1 301 405 3095; fax: þ1 301 405 9383.
E-mail address: [email protected] (P.B. Sunderland).
0360-3199/$ – see front matter ª 2009 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved.
doi:10.1016/j.ijhydene.2009.04.012
 Author's personal copy

international journal of hydrogen energy 34 (2009) 5174–5182 5175

from a containment vessel. The scenario of concern is that Cheng et al. [18] for methane. Both studies found that quench-
a small leak in a hydrogen system could ignite, burn unde- ing mass flow rate was largely independent of burner diameter.
tected for a long time, and potentially degrade surrounding Prior to the present work, quenching flow rates had not been
materials or ignite secondary fires. measured for hydrogen. Blowoff limits for diffusion flames on
Several studies have been conducted to evaluate the char- small burners have been measured for a variety of fuels [16,20].
acteristics of hydrogen leaks without combustion. Lee et al. [6] Thus motivated, this study includes experiments and
conducted hydrogen and helium leak rate experiments on analysis to identify which hydrogen leaks can support flames.
micromachined orifices of different sizes and shapes. They A scaling analysis is presented to estimate the fuel mass flow
examined the differences in flow rates among circular, square, rate at the quenching limit. Measured mass flow rates, both at
and elliptical slit orifices as a function of pressure, and in most the quenching and blowoff limits, are presented for hydrogen,
cases the flow was choked. Schefer et al. [7] investigated leaks methane and propane on round-hole burners. Flame quench-
where the flow was due to pressure-driven convection and ing limits for leaky compression fittings are also presented.
permeation through metals. They obtained analytical rela- Further details are available in Butler [21] and Moran [22].
tionships for flow rates of choked flows, subsonic laminar
flows, and turbulent flows. Hydrogen leakage in stainless steel
threaded pipe fittings was considered by Ge and Sutton [8].
2. Scaling for flame quenching limits
They found that a larger tightening torque is less important in
A scaling analysis is presented here for flame quenching
leak prevention than choice and proper application of thread
limits. This is similar to past work of Matta et al. [16]. The
sealant. The tests were run at up to 70 bar and typical hydrogen
stoichiometric length Lf of laminar gas jet diffusion flames on
leak rates through these fittings were found to be 1 mg/s.
round-hole burners is given by:
Studies have also evaluated the risks of hydrogen leaks with
respect to accumulation and explosion of hydrogen [9–11]. Lf =d ¼ a Re ¼ 4mfuel a=ðp mdÞ (1)
Swain and Swain [12] examined risks associated with leaks of
hydrogen, methane, and propane from leaky fittings into an where d is the burner inside diameter, a is a fuel-specific
enclosure. Their measured leak rates showed that for a given coefficient, Re is fuel port Reynolds number, mfuel is the fuel
supply pressure (except at the lowest supply pressures), mass flow rate, and m is fuel dynamic viscosity. The scaling of
hydrogen had a significantly higher volumetric flow rate than Eq. (1) arises from many theoretical and experimental studies,
methane or propane. This behavior is expected for choked including Roper [23], Sunderland et al. [24], and references
flows, owing to the high speed of sound of hydrogen gas. cited therein.
Leak flames resemble the micro diffusion flames that have The base of an attached jet diffusion flame is quenched by
been observed in other laboratories and, while only one the burner and is premixed. The flame’s standoff distance can
previous study specifically evaluated leak flames [13], there be approximated as the standoff distance (i.e., 50% of the
have been many studies of micro diffusion flames [14–19]. quenching distance) of the corresponding stoichiometric
Micro diffusion flames are typically associated with an appli- premixed flame. The 50% modification arises here because
cation, e.g., a microcombustor for power generation. None- premixed flame quenching distances typically are reported as
theless, it is possible that they could arise unexpectedly. For the minimum tube diameter or plate separation distance, Lq,
example, if a fuel leak from a crack or hole in a fitting, tube or through which a premixed flame can pass [4]. It is assumed
storage vessel of a plumbing system were ignited, this could here that a jet flame is above its quenching limit if its stoi-
be characterized as a micro diffusion flame. chiometric length exceeds its standoff distance:
Ban et al. [14] investigated micro diffusion flames that were Lf > Lq =2: (2)
2–3 mm long on round burners with inner diameters of
0.15–0.4 mm. Three fuels were considered: ethane, ethylene and Combining Eqs. (1) and (2) yields the fuel mass flow rate at
acetylene. The experiments agreed with predictions of flame the quenching limit:
shapes in the absence of buoyancy. The flames were nearly
mfuel ¼ pLq m=ð8aÞ (3)
spherical and their shapes were unaltered when burner orien-
tation was changed with respect to gravity. Cheng et al. [17] Eq. (3) predicts that the fuel mass flow rate at the quenching
performed detailed measurements and computations on small limit is a fuel property that is independent of burner diameter,
hydrogen flames burning on similar burners. Buoyancy was which was similarly predicted by Matta et al. [16]. When
found to be insignificant for these flames. Nakamura et al. [19] measured values of Lq, m, and a (see Table 1) are inserted into
simulated methane micro diffusion flames supported on Eq. (3), predicted fuel mass flow rates at the quenching limit
circular burners with diameters less than 1 mm. They, too, found are obtained, as listed in Table 1.
nearly spherical flames, as a result of the weak buoyancy forces.
They also considered quenching limits of the methane flames.
Baker et al. [15] studied micro-slot burners (with port widths of 3. Experimental
0.1–0.76 mm) and developed a flame height expression for purely
diffusion-controlled propane/air nonpremixed flames. 3.1. Round-hole burners
Quenching and blowoff limits refer to flames with the
smallest and highest fuel flow rates for sustained burning. Three types of round-hole burners were considered: pinhole
Matta et al. [16] measured quenching limits for propane on burners, curved-wall burners, and tube burners, as illustrated
small round burners. Similar experiments were performed by in Fig. 1. Each type included various hole diameters, as
 Author's personal copy

5176 international journal of hydrogen energy 34 (2009) 5174–5182

Table 1 – Selected properties of hydrogen, methane, and Table 2 – Hole diameters (mm) for the round-hole
propane and predicted quenching mass flow rates. burners.
Fuel a Lq (mm) SL (cm/s) m (g/m-s) mfuel (mg/s) Pinhole Curved walla Curved wallb Tube

H2 0.236 0.51 291 8.76E-3 8 0.008 0.41 0.41 0.051

CH4 0.136 2.3 37.3 1.09E-2 85 0.13 0.53 1.75 0.152
C3H8 0.108 1.78 42.9 7.95E-3 63 0.36 0.74 2.46 0.406
0.53 0.86 3.12 0.838
Values for a are from [24], Lq and SL (laminar burning velocity) are
0.71 1.02 1.194
from [4], and m is from [25]. The quantity mfuel is predicted from
0.84 2.21
Eq. (3).
1.01
1.40
1.78
tabulated in Table 2. The pinhole burners (such as the one in 2.39
Fig. 2) were stainless steel nozzles that were manufactured for 3.18
solid-stream spray generation. The top surface of each burner a Tube diameter of 1.6 mm.
(except the two smallest ones, which were planar) is a curved b Tube diameter of 6.4 mm.
surface (radius of curvature of 9.6 mm) with a hole passing
through its axis, as depicted in Fig. 1. The pinhole burners
represent pinhole leaks in pressure vessels where the radius Using similar burners, Takahashi et al. [13] also reported
of curvature of the wall is large or infinite. complications of water condensation. Tests performed here at
The curved-wall burners were constructed from stainless different burner temperatures, up to about 200  C, found
steel tubes with two outside diameters: 1.6 and 6.4 mm. A radial quenching flow rate to be largely independent of burner
hole was drilled in the wall of each tube. These burners temperature provided condensation was avoided. Tests were
represent pinhole leaks that might occur in fuel supply tubing. also conducted with varying ambient humidity, and quench-
The tube burners were made from stainless steel hypo- ing limits were found to be generally independent of the
dermic tubes. Although they do not represent practical leak relative humidity of the air in the range of 46–90%.
flames, they are a fundamental configuration that leads to Hydrogen flow rate at blowoff was measured with a soap
a flame from a wall where the radius of curvature approaches bubble meter. A stable flame was established and then the
zero. Their observation should lead to an improved under- flow rate was increased until the flame first lifted and then
standing of microinjectors in small-scale microelectro- extinguished. For blowoff tests the flames were detected
mechanical power generators [16]. visually. Hearing protection was required for the blowoff tests
Fuel was delivered via a pressure regulator and a needle for the larger burners.
valve, and all tests were performed at normal lab pressure and Additional tests were performed to consider buoyancy
temperature. Measuring the small flow rates at the quenching effects. To this end quenching flow rates were found for
limits required special procedures. A glass soap-bubble meter pinhole and tube burners in the vertical, horizontal, and
was installed upstream of the burners. Quenching flow rates inverted orientations.
were measured by establishing a small flame near the
quenching limit and gradually decreasing the flow rate until
the flame extinguished. The fuel flow rate was then measured
using the soap-bubble meter.
Determining the existence of hydrogen flames near their
quenching limits was complicated by their small size and low
luminosity. The flames were detected with a K-type thermo-
couple positioned 10 mm above the burner, yielding detection
that is believed to be more sensitive than available camera
technology could provide.
Burners were allowed to warm slightly above room
temperature to prevent water condensation. This was neces-
sary because condensation was found to disturb the flows
from the small burners, sometimes extinguishing the flames.

Fig. 2 – Image of a hydrogen flame slightly above its

Pinhole Curved-Wall Tube
quenching limit of 7.5 mg/s (original in color). The burner is
Fig. 1 – Schematics of the round-hole burners. Arrows a 0.36 mm pinhole burner, with a surface radius of curvature
show the fuel flow direction. Burner dimensions are of 9.6 mm, and is oriented vertically. The image was
shown in Table 2. recorded in a dim room at f/4.2 with an exposure time of 30 s.
 Author's personal copy

international journal of hydrogen energy 34 (2009) 5174–5182 5177

3.2. Leaky fittings 1000

Quenching limits were also measured for leaky compression

fittings. The tests involved leaks between Swagelok stainless 100
steel tube union compression fittings and stainless steel tubes.
Tube diameters were 3.2, 6.4, and 12.7 mm. The end of each

Mass Flow Rate (mg/s)

H2 CH4 C3H8
union opposite the tube was sealed. New tubes and fittings 10
Present
were assembled according to manufacturer instructions and Matta et al.
were rejected if bubbles appeared in an applied soap water Cheng et al.
1 Kalghatgi et al.
solution when pressurized with hydrogen to 8 bar absolute. ff
wo
Leaks were then introduced in one of three ways: by reducing Blo ng
chi
the torque on the threaded nut, by over tightening the nut, or 0.1 u en
Q
by scratching the front ferrule sealing surface. These types of
leaks are occasionally encountered in plumbing systems. The
three leak types were found to yield the same flame quenching 0.01
limits, so only results for reduced torque fittings are presented
here. Reduced torque fittings allow quenching limits to be
measured for different upstream pressures by adjusting the 0.001
0 0.5 1 1.5 2 2.5 3 3.5
torque, i.e., an increased torque requires an increased Tube Diameter (mm)
upstream pressure at the quenching limit. Most tests were
performed with the fitting in the vertical orientation, with the Fig. 3 – Quenching and blowoff limits for tube burners. Past
leaky fitting at the top end of the tube, but several horizontal measurements shown are from Refs. [16,18,20].
and inverted orientations also were considered.
Quenching limits were found by igniting the fuel (with an
external flame) and then reducing fuel flow rate until extinc- Measured quenching and blowoff limits for hydrogen,
tion. Fuel flow rate was controlled with upstream pressure, methane, and propane on tube burners are shown in Fig. 3.
primarily, and with torque on the threaded nut. Upstream Measurements shown for methane and propane come from
pressure was set with a pressure regulator. When a quenching past work, and were confirmed in our laboratory for repre-
limit was established, a plastic tube was installed over the sentative conditions. Consistent with the prediction of Eq. (3),
fitting such that the leak flow was measured with a down- the mass flow rate at quenching is not strongly dependent on
stream soap bubble meter at laboratory pressure. The burner diameter. The mean quenching limit flow rates are in
quenching limits were obtained with fittings at nearly labo- reasonable agreement with the predictions of Table 1. The
ratory temperature. mean quenching limits for hydrogen are about an order of
The hydrogen flames generally were not visible even in magnitude lower than those of the other fuels. Table 1 and
dark conditions. As with the round-hole burners, the presence Eq. (3) indicate this arises from hydrogen’s short quenching
of hydrogen flames was determined with a thermocouple distance (primarily) and from its large flame length relative to
positioned 10 mm above the flame position. Quenching limits its mass flow rate (i.e., large coefficient a).
for methane and propane were identified visually because The mass flow rate at blowoff, see Fig. 3, increases with
thermocouples confirmed these flames to be visible in all tube diameter because for these flames blowoff occurs when
cases. local velocities exceed local flame speeds. The blowoff limits
Uncertainties for the round-hole and leaky fitting tests are for hydrogen are about an order of magnitude larger than
estimated at 5% for hole diameter, fuel flow rate, and those of the other fuels, largely owing to the high burning
upstream pressure. velocity of hydrogen. Regardless of tube diameter, the limits of
stable combustion of hydrogen are much wider than those for
propane or methane. The limits narrow with decreasing tube
diameter, as was shown for propane in Matta et al. [16], but
4. Results they never meet.
Various existing codes and standards require leak rates
4.1. Round-hole burners below 20 scc/h (0.46 mg/s) for hydrogen and below 200 scc/h
(36 mg/s) for natural gas [13]. For hydrogen this leak rate is well
Fig. 2 shows an image of a hydrogen flame burning slightly below the measured flame quenching limits, whereas for
above its quenching limit on a 0.36 mm pinhole burner. The methane this leak rate is only slightly below the quenching
flame appearance is similar to that of all the quenching limit limits. Referring to the hydrogen measurements in Fig. 3, SAE
hydrogen flames on round-hole burners. The hydrogen J2579 [26] now limits localized hydrogen leaks in highway
quenching distance of Table 1 suggests steady hydrogen diffu- vehicles to a maximum of 5 mg/s.
sion flames should be anchored about 0.26 mm from the burner Fig. 4 shows the hydrogen quenching limits for the three
surface. Fig. 2 indicates that this is reasonable for the hydrogen types of round burners. Fig. 4 indicates that quenching limits
flames near their quenching limits. Because the maximum are affected by burner type and diameter for the smaller
flame dimension is comparable to its standoff distance, this burner diameters. Where the limits are affected by burner
flame closely resembles a flat premixed flame [16]. type, the differences can largely be attributed to varying
 Author's personal copy

5178 international journal of hydrogen energy 34 (2009) 5174–5182

0.012 affecting the quenching limits at these very small sizes. In

particular, the measured quenching flow rates for these two
burners indicate sonic velocities and Reynolds numbers of
0.01 14–130. For all other hole sizes, velocities are well below sonic
Quenching Mass Flow Rate (mg/s)

and Reynolds numbers are small (with a range of 0.4–7).

While the results above are for vertical burner orientation
0.008
(upward flowing fuel), hydrogen quenching limits also were
measured for inverted and horizontal orientations. The
results, shown in Fig. 5, reveal that quenching limits for the
0.006
pinhole and tube burners are not dependent on orientation,
indicating that these limit flames are not affected significantly
0.004
by buoyancy. This can be elucidated by considering their
Froude numbers, given by
pinhole burner
Fr ¼ u2 =ðgdÞ (4)
0.002 1.6 mm curved-wall burner
6.4 mm curved-wall burner where u is the fuel velocity at the burner port and g is the
tube burner
acceleration of gravity. For the present quenching limit
0
0 0.5 1 1.5 2 2.5 3 3.5 flames on pinhole and tube burners, the Fr range is 0.17–39.
Burner Diameter (mm) These are generally in the nonbuoyant regime for micro-
flames of Fr > 1 [14]. Thus, flame structure at the quenching
Fig. 4 – Quenching mass flow rates for hydrogen for various limits should not vary with orientation. Furthermore, as
burner types and diameters. See Fig. 1 for burner mentioned in the experimental section, small changes in
schematics. burner temperature do not affect the quenching limits, so
any effects of orientation on burner heating should not affect
these limits.
amounts of heat loss, as discussed next. An increase in heat The above results demonstrate that flame quenching limits
loss requires a higher quenching flow rate to sustain the depend on fuel type, with burner type, diameter, and orien-
flame. tation having far less significant effects. For a given fuel the
For burner diameters below 1.5 mm, pinhole burners have mass flow rate from the leak is a function of the leak geometry
the highest quenching flow rates, while tube burners have the and the upstream pressure. This relationship depends on
lowest. Pinhole burners have greater heat losses owing to whether the leak regime is diffusional, subsonic laminar,
their shape and mass. Heat conducted to the pinhole burners subsonic turbulent, isentropic choked, supersonic, etc., as
is mostly transferred to the ambient and constitutes a loss of considered by Schefer et al. [7]. While the leaks considered
enthalpy from the system. On the other hand, much of the here would not be considered isentropic, insight into the
heat conducted to the tube burners is transferred to the fuel quenching limits can be obtained by considering isentropic
and oxidizer and thus is not intrinsically lost from the system.
For tube burners the primary mechanism of heat loss is
0.012
expected to be radiation from the tube.
The effects of burner curvature on heat loss were investi-
gated by observing quenching limits on curved-wall burners.
0.01
Quenching Mass Flow Rate (mg/s)

An increase in curvature is expected to reduce wall heat loss.

Indeed the 1.6 mm curved-wall burners yield quenching limits
approaching those of the tube burners, whereas the 6.4 mm 0.008
burners have quenching limits closer to those of the pinhole
burners.
The dependence of quenching limits on burner diameter is 0.006
complicated for small hole diameters, as shown in Fig. 4. For
tube burners the quenching flow rate generally increases with
increasing burner diameter. This also was observed for 0.004
propane by Matta et al. [16], as shown in Fig. 3. This is attrib-
Pinhole Tube
uted to increasing heat loss with increasing burner diameter.
Vertical
In contrast pinhole burners do not exhibit this trend because 0.002
Horizontal
their heat loss is nearly independent of burner hole size. The Inverted
reason for the local maximum in Fig. 4 for pinhole burners is
unknown. For burner diameters above 1.5 mm, the quenching 0
0 0.5 1 1.5 2 2.5 3 3.5
flow rates for both pinhole and tube burners agree reasonably
Burner Diameter (mm)
well with the 8 mg/s prediction of Eq. (3), shown in Table 1.
Quenching mass flow rate increases slightly when hole Fig. 5 – Quenching mass flow rates for hydrogen for
diameters are decreased below 0.1 mm for both pinhole and pinhole and tube burners in horizontal, inverted, and
tube burners (see Fig. 4). It is likely that the flow field is vertical orientations.
 Author's personal copy

international journal of hydrogen energy 34 (2009) 5174–5182 5179

100 methane or propane. Furthermore, at hydrogen’s maximum

anticipated storage pressure of 690 bar (10,000 psi), a hole
diameter of just 0.4 mm is predicted to support a flame.
The limit flames in this study are believed to be among the
weakest steady flames ever recorded. The weakest of these had
Quenching Diameter ( m)

a hydrogen flow rate of 3.9 mg/s (see Fig. 5), which corresponds
10 Methane
to a heat release rate of 0.46 W based on hydrogen’s lower
heating value of 119.9 kJ/g. Previously, Ronney et al. [27] docu-
mented hydrogen flame balls produced under microgravity
Propane conditions with a power output of about 1 W. In subsequent
tests they achieved flames as weak as 0.5 W. It is believed these
1 Hydrogen were the weakest flames recorded until the present work.
Flame balls were first predicted by Zeldovich [28], who
proposed that a solution exists to the steady heat and mass
conservation equations corresponding to a stationary,
spherical premixed flame. The phenomenon was discovered
40 years after Zeldovich’s work by Ronney et al. [29] in drop
0.1 tower experiments with lean hydrogen–air mixtures. The
1 10 100 1000
Pressure (bar) microgravity environment was necessary to obtain the
spherical symmetry and to avoid extinction brought on by
Fig. 6 – Minimum diameter of a circular port for a leak buoyancy. Flame balls have only been achieved in micro-
flame as a function of upstream absolute pressure gravity conditions with very lean hydrogen mixtures. With no
assuming isentropic choked flow. mechanism other than radiative heat loss, and this being low
for hydrogen flames, it was reasonable to assume that these
flames were the weakest possible flames. The present results
choked flows through round holes at the quenching limits. indicate that, despite the possibility of heat loss to the tube,
This could, for example, be relevant to a pinhole leak in a thin the present diffusion flame geometry is actually slightly more
wall tube. Under these conditions the fuel flow rate is linear resistant to extinction than a flame ball. As noted earlier, heat
with respect to leak area A and upstream absolute pressure p0 transfer to the hypodermic tube does not intrinsically imply
as follows [7]: a significant loss of enthalpy as much of this heat preheats the
 1=2  kþ1=2ðk1Þ reactants.
kMW 2
mfuel ¼ Apo (5)
T0 Ru kþ1
4.2. Leaky fittings
where k is fuel specific heat ratio, MW is fuel molecular
weight, T0 is upstream stagnation pressure, and Ru is the Fig. 7 shows images of hydrogen, methane, and propane
universal gas constant. flames on 6.3 mm leaky compression fittings in the vertical
For a given fuel and upstream pressure, Eq. (5) can predict orientation. These images were recorded slightly above the
the hole diameter associated with the quenching limit. The quenching limits. The hydrogen flame is significantly smaller
results of these predictions are shown in Fig. 6 for hydrogen, and dimmer than the others. Near their quenching limits the
methane, and propane. For this plot the quenching limit flow flames do not burn along the entire fitting annulus. For all
rates were taken as the average values for tube burners in three flames the shortest distance between the flame and the
Fig. 3. Each line in Fig. 6 starts at the minimum upstream metal material is approximately 50% of the quenching
pressure for choked flow and ends at the maximum pressure distances given in Table 1. No yellow soot luminosity was
anticipated in alternative fuel vehicles. This plot predicts that visible in any of the flames near their quenching limits. The
for a given storage pressure, hydrogen is susceptible to leak leaks associated with the images of Fig. 7 produced readily
flames for hole diameters that are smaller than those for visible bubbles when soap water solution was applied.

Fig. 7 – Flames near extinction on leaky compression fittings for hydrogen, methane, and propane (original in color). Also
visible is the 6.3 mm tube and the threaded nut. The fitting was vertically oriented and the upstream absolute pressure was
4 bar.
 Author's personal copy

5180 international journal of hydrogen energy 34 (2009) 5174–5182

An audible pop occurred upon hydrogen ignition, but not 0.4

upon ignition of methane or propane. When the external
flame was removed, ignition was quickly followed by extinc-
tion for fuel flow rates below the quenching limit. For fuel flow Vertical

Quenching Mass Flow Rate (mg/s)

rates above the quenching limit, ignition was followed by Horizontal
0.3
a stable flame. Inverted
Fig. 8 shows measured quenching limits for hydrogen,
propane, and methane for 6.3 mm fittings in the vertical
orientation. The data at the higher line pressures were
obtained by increasing the torque on the fitting, thus reducing 0.2
the leak size. The upper limit on line pressure for propane is
lower than that of the others because the vapor pressure of
propane is only 9.1 bar absolute (142 psia) at 25  C.
Within experimental uncertainties, the quenching limits of
Fig. 8 are independent of line pressure for each fuel. This 0.1

suggests that, as anticipated from the round-hole burner

results of Fig. 4, the key parameter controlling quenching
limits is the fuel mass flow rate. For these low flow rates and
small leak sizes the line pressure does not significantly affect 0
the velocity field in the flame region. Hydrogen Methane Propane
The mean hydrogen flow rate at the quenching limit is
Fig. 9 – Hydrogen quenching limits for a 6.3 mm leaky
about an order of magnitude lower than for the other fuels,
compression fitting in vertical, horizontal and inverted
which is consistent with the results of Fig. 3. For leaky fittings,
orientations. The upstream absolute pressure was 4 bar.
the quenching flow rates for hydrogen, methane, and propane
are about an order of magnitude higher than the corre-
sponding flows for tube burners (see Fig. 3). This is attributed
to additional heat loss in the leaky fittings, where the flames limits of propane and methane flames, with the inverted
burn near thick concave metal surfaces. orientation yielding limits about 20% lower than the vertical
The effects of fitting orientation are shown in Fig. 9, and orientation. These flames are larger than the corresponding
reveal that orientation has little or no effect on the quenching hydrogen flames such that buoyancy begins to have a signifi-
limits of the hydrogen flames. This is anticipated as these cant effect on the flow field. The inverted orientation has the
flames are small at their limits and the results are consistent fitting below the tube, and this minimizes the surface area of
with the tube and pinhole burner hydrogen measurements of the flame impingement, leading to a reduced heat loss and
Fig. 5. Nonetheless, fitting orientation did affect the quenching thus lower quenching limits.

0.4 0.6
3.2 mm
6.4 mm
0.5
Quenching Mass Flow Rate (mg/s)

12.7 mm
Quenching Mass Flow Rate (mg/s)

0.378 mg/s
0.3

0.336 mg/s 0.4

Hydrogen
Methane
0.2
Propane 0.3

0.2
0.1
0.028 mg/s

0.1

0
1 10 100 1000
0
Pressure (bar) Hydrogen Methane Propane

Fig. 8 – Quenching limit (minimum flaming flow rate) Fig. 10 – Hydrogen quenching limits for leaky compression
versus upstream absolute pressure for a 6.3 mm leaky fittings for tube diameters of 3.1, 6.3, and 12.6 mm. The
compression fitting in the vertical orientation. The line fits fittings were vertically oriented and the upstream absolute
represent mean values of the measured quenching limits. pressure was 4 bar.
 Author's personal copy

international journal of hydrogen energy 34 (2009) 5174–5182 5181

The effects of fitting diameter on quenching limits are measured. Hydrogen’s low quenching limits, combined with
examined in Fig. 10. Diameter had a significant impact on the its high leak propensity, could present unusual risks in
quenching limits in all cases; on average as tube diameter a hydrogen economy. These risks should be explored in
doubles, the fuel mass flow rate at quenching increases 30%. further research and should be incorporated into the many
This is attributed to increased heat losses associated with relevant safety codes and standards.
larger fittings.
This study has identified new questions that warrant
further study. For example, what corrosive effects do leak
flames have on containment materials? Can permeation leaks Acknowledgments
support flames? Can surface coatings, e.g., intumescent
paints, be applied that will increase the quenching fuel flow This work was supported by NIST under the technical direc-
rate? An improved understanding of hydrogen leaks is tion of J. Yang. The assistance of B.H. Chao and N.R. Morton is
necessary to ensure safe use of hydrogen in the public sector. appreciated.

5. Conclusions references

Flame quenching and blowoff limits were measured for

diffusion flames resulting from fuel issuing from small [1] Marban G, Valdes-Solis T. Towards the hydrogen economy?
burners. Hydrogen fuel was emphasized, and was compared Int J Hydrogen Energy 2007;32:1625–37.
with methane and propane for a variety of burner types that [2] Yamawaki M, Nishihara T, Inagaki Y, Minato K, Oigawa H,
characterized a range of heat losses. The key findings are as Onuki K, et al. Application of nuclear energy for
environmentally friendly hydrogen generation.
follows.
Int J Hydrogen Energy 2007;32:2719–25.
[3] Sherif SA, Barbir F, Veziroglu TN. Wind energy and the
1. Flame quenching limits for round-hole burners were hydrogen economydreview of the technology. Solar Energy
measured for hydrogen, methane, and propane. For 2005;78:647–60.
hydrogen these are in the range of 4–10 mg/s depending on [4] Kanury AM. Introduction to combustion phenomena. New
burner diameter, shape, and orientation. The limits for York: Gordon and Breach; 1975. p. 131.
methane and propane are about 10 times as high. The [5] MacIntyre I, Tchouvelev AV, Hay DR, Wong J, Grant J,
Benard P. Canadian hydrogen safety program. Int J Hydrogen
measurements indicate that many existing safety codes
Energy 2007;32:2134–43.
and standards should revisit their allowable fuel leak rates [6] Lee ID, Smith OI, Karagozian AR. Hydrogen and helium leak
in consideration of these quenching limits. rates from micromachined orifices. AIAA J 2003;41(3):457–63.
2. Flame blowoff limits increase with burner diameter and for [7] Schefer RW, Houf WG, San Marchi C, Chernicoff WP,
hydrogen are about 10 times as high as the corresponding Englom L. Characterization of leaks from compressed
limits for methane and propane. Thus the limits of stable hydrogen dispensing systems and related components.
Int J Hydrogen Energy 2006;31:1247–60.
combustion are much wider for hydrogen than for the other
[8] Ge X, Sutton WH. Analysis and test of compressed hydrogen
fuels considered.
interface leakage by commercial stainless steel (NPT) fittings.
3. For an isentropic choked leak, hydrogen is susceptible to SAE Int 2006:35–47.
leak flames for hole diameters that are smaller than those [9] Dorofeev SB. Evaluation of safety distances related to
for methane or propane. This suggests that for hydrogen unconfined hydrogen explosions. Int J Hydrogen Energy 2007;
stored at 690 bar, a pinhole leak as small as 0.4 mm could 32:2118–24.
support a stable flame. [10] Takeno K, Okabayashi K, Kouchi A, Nonaka T, Hashiguchi K,
Chitose K. Dispersion and explosion field tests for 40 MPa
4. The weakest hydrogen flame found in this study, which
pressurized hydrogen. Int J Hydrogen Energy 2007;32(13):
was generated with a hypodermic needle with an inner 2144–53.
diameter of 0.152 mm, is believed to be the weakest flame [11] Zbikowski M, Makarov D, Molkov V. LES model of large scale
ever observed, with a fuel flow rate of 3.9 mg/s and a heat hydrogen-air planar detonations: verification by the ZND
release rate of 0.46 W. theory. Int J Hydrogen Energy 2008;33:4884–92.
5. The minimum flow rate necessary for sustaining a hydrogen [12] Swain MR, Swain MN. A comparison of H2, CH4, and C3H8fuel
flame on a leaky 6.3 mm tube compression fitting is 28 mg/s. leakage in residential settings. Int J Hydrogen Energy 1992;17:
807–15.
This is about an order of magnitude lower than for propane
[13] Takahashi M, Tamura Y, Suzuki J, Watanabe S. Investigation
or methane. The minimum mass flow rate for all fuels is of the allowable flow rate of hydrogen leakage on receptacle.
independent of upstream (line) pressure, and is sufficient to Detroit: SAE International; April 2008. Paper 2008-01-724.
produce bubbles when a soap-water solution is applied. [14] Ban H, Venkatesh S, Saito K. Convection-diffusion controlled
6. Fitting orientation had about a 20% effect on the quenching laminar micro flames. Trans ASME 1994;116:954–9.
limits of the leaky compression fittings considered here, [15] Baker J, Calvert ME, Murphy DW. Structure and dynamics of
laminar jet micro-slot diffusion flames. J Heat Transfer 2002;
while a doubling of tube size on average results in a 30%
124:783–90.
increase in quenching flow rate.
[16] Matta LM, Neumeier Y, Lemon B, Zinn BT. Characteristics of
microscale diffusion flames. Proc Combust Inst 2002;29:933–8.
This work was undertaken because hydrogen is such an [17] Cheng TS, Chao Y-C, Wu C-Y, Li Y-H, Nakamura Y, Lee K-Y,
unusual fuel and its quenching limits had not previously been et al. Experimental and numerical investigation of
 Author's personal copy

5182 international journal of hydrogen energy 34 (2009) 5174–5182

microscale hydrogen diffusion flames. Proc Combust Inst [24] Sunderland PB, Mendelson BJ, Yuan ZG, Urban DL. Shapes of
2005;30:2489–97. buoyant and nonbuoyant laminar jet diffusion flames.
[18] Cheng TS, Chen CP, Chen CS, Li YH, Wu CY, Chao YC. Combust Flame 1999;116:376–86.
Characteristics of microjet methane diffusion flames. [25] Weast RC, Astle MJ. CRC Handbook of chemistry and physics.
Combust Theory Modeling 2006;10:861–81. 59th ed. West Palm Beach: CRC Press Inc; 1979. p. F-58.
[19] Nakamura Y, Yamashita H, Saito K. A numerical study on [26] SAEJ2579. Recommended practice for general fuel cell vehicle
extinction behaviour of laminar micro-diffusion flames. safety, a surface vehicle recommended practice. Detroit, MI:
Combust Theory Modeling 2006;10(6):927–38. SAE International; January, 2009.
[20] Kalghatgi GT. Blow-out stability of gaseous jet diffusion [27] Ronney PD, Wu MS, Pearlman HG. Structure of flame balls at
flames. Part I: In still air. Combust Sci Technol 1981;26(5): low Lewis-number (SOFBALL): preliminary results from the
233–9. STS-83 and STS-94 space flight experiments. Paper AIAA-
[21] Butler MS. Flame quenching and materials degradation of 1998-463. Reno, Nevada: Aerospace Sciences Meeting;
hydrogen leaks. M.S. Thesis. St. Louis: Washington January 1998.
University; May 2008. 50 pp. [28] Zeldovich Y. Theory of combustion and detonation of gases.
[22] Moran CW. Flame quenching limits of hydrogen leaks. M.S. Moscow, Russia: Academy of Science; 1994.
Thesis. University of Maryland; May 2008. 48 pp. [29] Ronney PD, Wu MS, Pearlman HG. Experimental study of
[23] Roper FG. The prediction of laminar jet diffusion flame sizes: flame balls in space: preliminary results from STS-83. AIAA J
Part 1. Theoretical model. Combust. Flame 1977;29:219–26. 1998;36(8):1361–8.
 Fire Safety of Hydrogen Leaks
M.S. Butler1, C.W. Moran2, P.B. Sunderland2, R.L. Axelbaum1
1
Dept. of Energy, Environmental and Chemical Eng., Washington Univ., St. Louis MO 63130, USA
2
Dept. of Fire Protection Eng., Univ. of Maryland, College Park MD 20742, USA
email: [email protected]

Abstract
Quenching limits of hydrogen diffusion flames on small burners were observed. Four burner types, with
diameters as small as 8 μm, were considered: pinhole burners, curved-wall pinhole burners, tube burners, and leaky
fittings. In terms of mass flow rate, hydrogen had a lower quenching limit and a higher blowoff limit than either
methane or propane. Hydrogen flames at their quenching limits were the weakest flames recorded to date, with mass
flow rates and heat release rates as low as 3.9 μg/s and 0.46 W. The quenching limits were generally independent of
hole diameter and burner orientation, and generally decreased with increased surface curvature. The quenching limit
for a hydrogen flame at a 6 mm leaky compression fitting was found to be 28 μg/s. This limit was independent of
supply pressure (up to 131 bar) and about an order of magnitude lower than the corresponding limits for methane
and propane.

1. Introduction
Hydrogen is attractive as an energy carrier for highway vehicles. It can power fuel cells or engines with only
water vapor as exhaust. Hydrogen could help mitigate concerns over fossil fuel consumption [1,2]. Hydrogen can be
stored as a gas, liquid, or a solid (in metal hydrides), and can be transported using pipelines, tankers or rail trucks
[3]. It is anticipated that hydrogen could be produced efficiently using nuclear sources or renewable sources such as
wind [4].
Hydrogen is an unusual fuel. It has a high leak propensity and wide flammability limits, 4 – 75% by volume
[5]. Hydrogen is the lightest fuel and has the lowest quenching distance (0.51 mm) and smallest ignition energy of
any fuel in air (28 μJ), the lowest auto-ignition temperature of any fuel ignited by a heated air jet (640 ºC), the
highest laminar burning velocity of any fuel in air (2.91 m/s), and the highest heat of combustion (119.9 kJ/g) [5].
Hydrogen flames are the dimmest of any fuel. Hydrogen embrittles metals more than any other fuel does.
Hydrogen may ultimately prove to be no more hazardous from a fire safety standpoint than gasoline or diesel.
However, gasoline and diesel underwent over a century of widespread vehicle use, and this has resulted in codes
and standards that have yielded an acceptable fire safety record. Further research is necessary if hydrogen is to be
rapidly introduced with a similar safety record. It is expected that existing codes and standards will need to be
updated [6] to ensure safe use of hydrogen in vehicles.
The goal of present study is to investigate small hydrogen flames that could be characterized as “leak
flames,” i.e., small flames that could result from hydrogen leaking from a containment vessel. The scenario of
concern is that a small leak in a hydrogen system could ignite, burn undetected for a long period of time, and
potentially degrade surrounding materials, or ignite secondary fires.
Several studies have been conducted to evaluate the characteristics the hydrogen leak itself. Lee et al. [15]
conducted leak rate experiments on micromachined orifices of different sizes and shapes. They examined the
differences in flow rates among circular, square, and elliptical slit orifices as a function of pressure, and in most
cases the flow was choked. Helium was used as a substitute for hydrogen for safety reasons. Schefer et al. [16]
investigated leaks where the flow was due to pressure-driven convection and permeation through metals. They
obtained analytical relationships for flow rates of choked flows, subsonic laminar flows, and turbulent flows.
Hydrogen leakage in stainless steel threaded pipe fittings was considered by Ge and Sutton [17]. They found that a
larger tightening torque is less important in leak prevention than choice and proper application of thread sealant.
The tests were run up to 70 bar and typical hydrogen leak rates through these fittings were found to be 1 μg/s.
Studies have also evaluated the risks of hydrogen leaks with respect to accumulation and explosion of
hydrogen [7,8]. Swain and Swain [18] evaluated the safety risks associated with leaks of hydrogen, methane, and
propane into an enclosed area. They modeled and measured leak rates for diffusion, laminar, and turbulent flow
regimes and found that, for a given supply pressure, hydrogen had a significantly higher volumetric flow rate than
methane or propane. They also found that combustible mixtures resulted more quickly for propane and hydrogen
leaks than for methane leaks.

Abu Dhabi, U.A.E., November 4-5, 2008 1

 Leak flames resemble the micro diffusion flames that have been observed in other laboratories and while only
one previous study specifically evaluated leak flames [20], there have been many studies of micro diffusion flames
[9-14]. Micro diffusion flames are typically associated with an application, e.g., a microcombustor for power
generation. Nonetheless, it is possible that they could arise unexpectedly. For example, if a leak from a crack or hole
in a fitting, tube or storage vessel of a plumbing system were ignited, this could be characterized as a micro
diffusion flame.
Ban et al. [10] investigated micro diffusion flames that were 2 – 3 mm long on round burners with inner
diameters of 0.15 – 0.4 mm. Three fuels were considered: ethane, ethylene and acetylene. The experiments agreed
with predictions of flame shapes in the absence of buoyancy. The flames were nearly spherical and their shapes
were unaltered when burner orientation was changed with respect to gravity. Cheng et al. [11] performed detailed
measurements and computations on small hydrogen flames burning on similar burners. Buoyancy was found to be
insignificant for these flames. Nakamura et al. [12] simulated methane micro diffusion flames supported on circular
burners with diameters less than 1 mm. They, too, found nearly spherical flames, associated with weak buoyancy
forces. They also considered quenching limits of the flames. Baker et al. [9] studied micro-slot burners (with port
widths of 0.1 – 0.76 mm) and developed a flame height expression for purely diffusion-controlled propane/air
nonpremixed flames.
Quenching and blowoff limits refer to flames with the smallest and highest fuel flow rates for sustained
burning. Matta et al. [13] measured quenching limits for propane on small round burners. Similar experiments were
performed by Cheng et al. [14] for methane. Both studies found that quenching mass flow rate was largely
independent of burner diameter. Prior to the present work, quenching flow rates had not been measured for
hydrogen. Blowoff limits for micro diffusion flames have been measured for a variety of fuels by several
researchers [13,19].
Thus motivated, this study includes experiments and analysis to identify which hydrogen leaks can support
flames. A scaling analysis is presented to estimate the fuel mass flow rate at the quenching limit. Measured mass
flow rates, both at the quenching and blowoff limits, are presented for hydrogen, methane and propane on round-
hole burners. Flame quenching limits for leaky compression fittings are also presented.

2. Scaling for Flame Quenching Limits

A scaling analysis is presented here for flame quenching limits. This is similar to past work of Matta et al.
[13]. The stoichiometric length Lf of laminar gas jet diffusion flames on round-hole burners is given by:
Lf / d = a Re = 4 mfuel a / ( π μ d ) , (1)
where d is the burner inside diameter, a is a fuel-specific coefficient, Re is fuel port Reynolds number, mfuel is the
fuel mass flow rate, and μ is fuel dynamic viscosity. The scaling of Eq. (1) arises from many theoretical and
experimental studies, including Roper [21], Sunderland et al. [22], and references cited therein.
The base of an attached jet diffusion flame is quenched by the burner and is premixed. Its standoff distance
can be approximated as one half of the quenching distance of a stoichiometric premixed flame. Quenching distances
typically are reported as the minimum tube diameter or plate separation distance, Lq, through which a premixed
flame can pass. It is assumed here that a jet flame is above its quenching limit if its stoichiometric length exceeds
half this quenching distance:
Lf > Lq / 2 . (2)
Combining Eqs. (1) and (2), yields the fuel mass flow rate at the quenching limit:
mfuel = π Lq μ / ( 8 a ) . (3)
Eq. (3) predicts that the fuel mass flow rate at the
quenching limit is a fuel property that is Table 1
independent of burner diameter, which was similarly Selected properties of hydrogen, methane, and propane and
predicted quenching mass flow rates. Values for a are from [22],
predicted by Ref. [13]. When measured values of Lq,
La and SL (laminar burning velocity) are from [5], and μ is from
μ, and a (see Table 1) are inserted into Eq. (3), the [23]. The quantity mfuel is predicted from Eq. (3).
predicted fuel mass flow rates at the quenching limit
are obtained and are listed in Table 1. Fuel A L S μ m
q L fuel
(mm) (cm/s) (g/m-s) (μg/s)
3. Experimental
H2 0.236 0.51 291 8.76E-3 8
Two different burner configurations are
CH4 0.136 2.3 37.3 1.09E-2 85
considered in this study: round-hole burners and
C3H8 0.108 1.78 42.9 7.95E-3 63
leaky compression fittings.

Abu Dhabi, U.A.E., November 4-5, 2008 2

3.1 Round-Hole Burners
Three types of round-hole burners were considered: pinholes, and curved-wall pinholes, and tubes, as
illustrated in Fig. 1. Each type included various hole diameters, given in Table 2. The pinhole burners were stainless
steel nozzles that were manufactured for solid-stream spray generation. The top surface of each burner (except for
the two smallest ones) is a curved surface (radius of curvature of 4.3 mm) with a hole passing through its axis, as
shown in Fig. 1. The top surfaces of the two smallest pinhole burners were planar, not curved, surfaces. The pinhole
burners represent pinhole leaks in pressure vessels where the radius of curvature of the wall is large or infinite.
The curved-wall pinhole burners were constructed from stainless steel tubes with two outside diameters: 1.6
and 6.4 mm. A radial hole was drilled in each tube. These burners represent pinhole leaks that might occur in fuel
supply tubing.
The tube burners were made from stainless steel hypodermic tubes. Although they do not represent practical
leak flames, they are a fundamental configuration that leads to a flame from a wall where the radius of curvature
approaches zero. Their observation should also lead to an improved understanding of microinjectors in small-scale
microelectro-mechanical power generators [13].
Fuel was delivered via a pressure regulator and a needle valve, and all tests were performed at normal lab
pressure and temperature. Measuring the small flow rates at the quenching limits required special procedures. A
glass soap-bubble meter was installed upstream of the burners. Quenching flow rates were measured by establishing
a small flame near the quenching limit and gradually decreasing the flow rate until the flame extinguished. The fuel
flow rate was then measured using the soap-bubble meter.
Determining the existence of hydrogen flames near their quenching limits was complicated by their small size
and low luminosity. These flames were detected with a K-type thermocouple positioned 10 mm above each burner.
Burners were allowed to warm slightly above room temperature to prevent water condensation. This was
necessary because water condensation was found to disturb the flows from the small burners, sometimes
extinguishing the flames. Using similar burners, Takahashi et al. [20] also reported complications of water
condensation. Tests performed at different burner temperatures, up to about 200 °C, found quenching flow rate to
be largely independent of burner temperature provided condensation was avoided. Tests were also conducted with
varying ambient humidity, and quenching limits were found to be generally independent of the relative humidity of
the air in the range of 46 – 90%.
Hydrogen flow rate at blowoff was measured with a soap bubble meter. A stable flame was established and
then the flow rate was increased until the flame first lifted and then extinguished. For blowoff tests the flames were
detected visually. Hearing protection was required for the blowoff tests for the larger burners.
Additional tests were performed to consider buoyancy effects. To this end. quenching flow rates were found
for pinhole and tube burners in the vertical, horizontal, and inverted orientations.

3.2 Leaky Fittings

Quenching limits were also measured for leaky
compression fittings. The tests involved leaks between Table 2
Swagelok® stainless steel tube union compression fittings and Hole diameters (mm) for the present round-hole
stainless steel tubes. Tube diameters were 3.2, 6.4, and burners.
12.7 mm. The end of each union opposite the tube was sealed.
New tubes and fittings were assembled according to Pinhol Curve Curved Tube
manufacturer instructions and were rejected if bubbles e d walla wallb
0.008 0.41 0.41 0.051
appeared in an applied soap water solution when pressurized 0.13 0.53 1.75 0.152
with hydrogen to 8 bar absolute. Leaks were then introduced 0.36 0.74 2.46 0.406
in one of three ways: by reducing the torque on the threaded 0.53 0.86 3.12 0.838
0.71 1.02 1.194
0.84 2.21
1.01
1.40
Curved-wall Burners Tube Burner
1.78
Pinhole Burner
2.39
3.18
a
Fig. 1. Schematics of the present round-hole burners. Arrows For tube diameter of 1.6 mm.
show the fuel flow direction. Burner dimensions are given in b
For tube diameter of 6.4 mm.
Table 2.

Abu Dhabi, U.A.E., November 4-5, 2008 3

nut, by over tightening the nut, or by scratching the front ferrule sealing surface. These types of leaks are
occasionally encountered in plumbing systems. The three leak types were found to yield the same flame quenching
limits, so only results for reduced torque fittings are presented here. Reduced torque fittings allow quenching limits
to be measured for different upstream pressures by adjusting the torque, i.e., an increased torque requires an
increased upstream pressure at the quenching limit. Most tests were performed with the burner in the vertical
orientation, with the leaky fitting at the top end of the tube. Horizontal and inverted orientations also were
considered for some tests.
Quenching limits were found by igniting the fuel (with an external flame) and then reducing fuel flow rate
until extinction. Fuel flow rate was controlled with upstream pressure, primarily, and with torque on the threaded
nut. Upstream absolute pressure was set with a pressure regulator. When a quenching limit was established, any
flame was extinguished and a plastic tube was installed over the fitting such that the leak flow was measured with a
downstream soap bubble meter at laboratory pressure. The quenching limits were obtained with burners at nearly
laboratory temperature.
The hydrogen flames generally were not visible even in dark conditions. As with the round-hole burners, the
presence of hydrogen flames was determined with a thermocouple positioned 10 mm above the flame position.
Quenching limits for methane and propane were identified visually because thermocouples confirmed these flames
to be visible in all cases.
Uncertainties for the present round-hole and leaky fitting tests are estimated at ±5% for hole diameter, fuel
flow rate, and upstream pressure.

4. Results
4.1 Round-Hole Burners
Figure 2 shows an image of a hydrogen flame burning slightly above its quenching limit on a 0.36 mm
pinhole burner. The flame appearance is similar to that of all the quenching limit hydrogen flames on round-hole
burners. The hydrogen quenching distance of Table 1 suggests steady hydrogen diffusion flames should be
anchored about 0.26 mm from the burner surface. Figure 2 indicates this is reasonable for the present hydrogen
flames near their quenching limits. Because the maximum flame dimension is comparable to its standoff distance,
this flame closely resembles a flat premixed flame [13].
Measured quenching limits for hydrogen,
methane, and propane on tube burners are shown in 1000
Fig. 3. Consistent with Eq. (3), for each fuel the
mass flow rate at quenching is nearly independent Blowoff Limits
of burner diameter. The mean quenching limit flow 100
rates are shown and are in reasonable agreement
Mass Flow Rate (mg/s)

10
H CH CH
2 4 3 8
This work
0.356 mm 1 Matta
Cheng
Kalghatgi
0.1

0.0056 mg/s
Quenching Limits
0.01

0.001
0 0.5 1 1.5 2 2.5 3

Fig. 2. Color image of a hydrogen flame slightly Tube Diameter (mm)

above its quenching limit of 7.5 μg/s. The burner is
a 0.36 mm pinhole burner and is oriented vertically. Fig. 3. Quenching and blowoff limits for tube burners. Hydrogen
The image was recorded in a dimly lit room at ƒ/4.2 quenching and blowoff limits include best fit lines. Past
with an exposure time of 30 s. measurements shown are from Refs. [13,14,19].

Abu Dhabi, U.A.E., November 4-5, 2008 4

with the predictions of Table 1. The mean quenching limits for hydrogen are about an order of magnitude lower
than those of the other fuels. Table 1 and Eq. (3) indicate this arises from hydrogen’s short quenching distance
(primarily) and from its large flame length relative to its mass flow rate (i.e., large coefficient a).
Figure 3 also shows measured blowoff limits. The mass flow rate at blowoff increases with tube diameter
because for these flames blowoff occurs when the local velocity exceeds the local premixed flame speed. The
blowoff limits for hydrogen are about an order of magnitude larger than those of the other fuels, largely owing to
the high burning velocity of hydrogen. Regardless of tube diameter, the limits of stable combustion of hydrogen are
much wider than those for propane or methane. The limits narrow with decreasing tube diameter, as was shown for
propane in Ref. [13], but they never meet.
Figure 4 shows the hydrogen quenching limits for the three types of round burners. Figure 4 indicates that, in
general, quenching limits are somewhat affected by burner type and diameter for the smaller burner diameters.
Where the limits are affected by burner type, the differences can largely be attributed to varying amounts of heat
loss, as discussed next. An increase in heat loss requires a higher quenching flow rate to sustain the flame.
For burner diameters below 1.5 mm, pinhole burners have the highest quenching flow rates, while tube
burners have the lowest. Pinhole burners have greater heat losses owing to their shape and mass. Heat conducted to
the pinhole burners is mostly transferred to the ambient and constitutes a loss of enthalpy from the system. On the
other hand, much of the heat conducted to the tube burners is transferred to the fuel and oxidizer and thus is not
intrinsically lost from the system. For tube burners the primary heat losses are expected to be radiation from the
tube.
The effects of burner curvature on heat loss were investigated by observing curved-wall pinhole burners. An
increase in curvature is expected to reduce wall heat loss. Indeed the 1.59 mm curved-wall pinhole burners yield
quenching limits approaching those of the tube burners, whereas the 6.35 mm burners have quenching limits closer
to those of the pinhole burners.
Burner diameter has interesting effects on the quenching limits of Fig. 4. For tube burners quenching flow
rate generally increases with increasing burner diameter. This also was observed for propane by Matta et al. [13], as
shown in Fig. 3. This is attributed to increasing heat loss with increasing burner diameter. In contrast pinhole
burners do not exhibit this trend because their heat loss is nearly independent of burner hole size. The reason for the
local maximum in Fig. 4 for pinhole burners is unknown. For burner diameters above 1.5 mm, the quenching flow
rates for both pinhole and tube burners agree with the 8 μg/s prediction of Table 1.
Quenching mass flow rates show a small increase for hole diameters less than 0.2 mm for both pinhole and
tube burners. It is likely that the flow field is affecting the quenching limits at these very small sizes. In particular,
the measured quenching flow rates for these two burners indicate sonic velocities. For the next larger hole sizes
velocities are well below sonic. While exit velocities
are high, the hole size is sufficiently small that the 0.016
Reynolds numbers are all small and less than 10. Pinhole burner
While the results above are for vertical burner 0.014 6.35 mm curved-wall pinhole burner
Quenching Mass Flow Rate (mg/s)

1.59 mm curved-wall pinhole burner

orientation (upward flowing fuel), hydrogen Tube burner
quenching limits also were measured for inverted 0.012
and horizontal orientations. The results, shown in
Figs. 5 and 6, reveal that quenching limits for the 0.01
pinhole and tube burners are not dependent on
orientation. Thus these limit flames are not affected 0.008
significantly by buoyancy. This can be elucidated by
considering their Froude numbers, given by 0.006
Fr = u2 / ( g d ) , (4)
where u is the fuel velocity at the burner port and g 0.004
is the acceleration of gravity. For the present
quenching limit flames on pinhole and tube burners, 0.002
the Fr range is 0.17 – 39. These are not in the
buoyancy-controlled regime of Fr ≤ 0.1 [10]. Thus 0
flame structure at the quenching limits should not 0 0.5 1 1.5 2 2.5 3 3.5
vary with orientation. Furthermore, as mentioned in Burner Diameter (mm)
the experimental section, small changes in burner
temperature do not affect the quenching limits, so Fig. 4. Quenching mass flow rates for hydrogen for various burner
any effects of orientation on burner heating should types and diameters.

Abu Dhabi, U.A.E., November 4-5, 2008 5

not affect these limits.
The above results demonstrate that flame quenching limits depend on fuel type and mass flow rate, with
burner type, diameter, and orientation having far less significant effects. For a given fuel the mass flow rate out of
the leak is a function of the leak geometry and the upstream pressure. This relationship depends whether the leak
regime is diffusional, subsonic laminar, subsonic turbulent, isentropic choked, supersonic, etc., as considered by
Schefer et al. [16]. While the leaks considered here would not be considered isentropic, insight into the present
quenching limits can be obtained by considering isentropic choked flows through round holes at the quenching
limits. This could, for example, be relevant to a pinhole leak in a thin wall tube. Under these conditions the fuel
flow rate is linear with respect to leak area A and upstream absolute pressure p0 as follows [16]:
1/ 2 k +1
⎛ k MW ⎞ ⎛ 2 ⎞ 2( k −1)
m fuel = A po ⎜⎜ ⎟⎟ ⎜ ⎟ , (5)
⎝ T0 Ru ⎠ ⎝ k +1⎠

where k is fuel specific heat ratio, MW is fuel molecular weight, T0 is upstream stagnation pressure, and Ru is the
universal gas constant.
Insight into the present quenching limits is obtained by considering isentropic choked flows through round
holes at the quenching limits. For a given fuel and upstream pressure, Eq. (5) can predict the hole diameter (the
quenching diameter) associated with the quenching limit. The results of these predictions are shown in Fig. 7. For
this plot the quenching limit flow rates were taken as the average values for tube burners in Fig. 3. Each line in Fig.
7 starts at the minimum upstream pressure for choked flow and ends at the maximum pressure anticipated in
alternative fuel vehicles. This plot predicts that for a given storage pressure, hydrogen is susceptible to flaming
leaks for hole diameters that are smaller than those for methane or propane. Furthermore, at hydrogen’s maximum
anticipated storage pressure of 690 bar (10,000 psi), a hole diameter of just 0.4 μm is predicted to support a flame.
The limit flames in this study are believed to be among the weakest steady flames ever recorded. The weakest
of these had a hydrogen flow rate of 3.9 μg/s (see Fig. 5), or a heat release rate of 0.46 W based on hydrogen’s
lower heating value of 119.9 kJ/g. Previously, Ronney et al. [24] documented hydrogen flame balls produced under
microgravity conditions with a power output of about 1 W. In subsequent tests they achieved flames as weak as
0.5 W. These were believed to be the weakest flames ever recorded.
Flame balls were first predicted by Zeldovich [25], who proposed that a solution exists to the steady heat and
mass conservation equations corresponding to a stationary, spherical premixed flame. This solution was termed a
flame ball, and the phenomena was discovered forty years after Zeldovich’s work by Ronney et al. [26] in drop
tower experiments with lean hydrogen-air mixtures. The microgravity environment was necessary to obtain the

0.008 0.012
Quenching Mass Flow Rate (mg/s)
Quenching Mass Flow Rate (mg/s)

0.01
0.007

0.008
0.006

0.006

0.005
0.004
Horizontal Horizontal
Inverted Inverted
0.004 Vertical Vertical
0.002

0.003 0
0 0.5 1 1.5 2 2.5 0 0.5 1 1.5 2 2.5 3 3.5
Tube Diameter (mm) Pinhole Burner Diameter (mm)

Fig. 5. Quenching mass flow rates for tube burners in Fig. 6: Quenching mass flow rates for pinhole burners in
horizontal, inverted, and vertical orientations. horizontal, inverted, and vertical orientations.

Abu Dhabi, U.A.E., November 4-5, 2008 6

spherical symmetry and to avoid extinction brought on by buoyancy. Flame balls have only been achieved in
microgravity conditions with very lean hydrogen mixtures. With no mechanism other than radiative heat loss, and
this being low for hydrogen flames, it was reasonable to assume that these flames were the weakest possible flames.
The present results indicate that, despite the possibility of heat loss to the tube, the present diffusion flame geometry
is actually slightly more resistant to extinction than a flame ball. As noted earlier, heat transfer to the hypodermic
tube does not intrinsically imply a significant loss of enthalpy as most of this heat preheats the reactants.

4.2 Leaky fittings

Figure 8 shows photographs of hydrogen, methane, and propane flames on 6.3 mm leaky compression
fittings in the vertical orientation. These images were recorded slightly above the quenching limits. Near their
quenching limits the flames did not burn along the entire fitting annulus. The hydrogen flame is significantly
smaller and dimmer than the others. For all three flames the shortest distance between the flame and the metal
material is approximately 50% of the quenching distance given in Table 1. No yellow soot luminosity was visible in
any of the present flames near their quenching limits. The leaks associated with the images of Fig. 8 produced
readily visible bubbles when soap water solution was applied.
An audible pop occurred upon hydrogen ignition, but not upon ignition of methane or propane. When the
external flame was removed, ignition was quickly followed by extinction for fuel flow rates below the quenching
limit. For fuel flow rates above the quenching limit,
ignition was followed by a stable flame.
Figure 9 shows measured quenching limits for
hydrogen, propane, and methane for 6.3 mm fittings
in the vertical orientation. The data at the higher line
pressures were obtained by increasing the torque on 5 mm
the fitting, and thus reducing the leak size. The upper Hydrogen Methane Propane
limit on line pressure for propane is lower than that
of the others because the vapor pressure of propane
is only 9.1 bar (142 psia) at 25 °C. Fig. 8. Flames near extinction on leaky compression fittings for
Within experimental uncertainties, the hydrogen, methane, and propane (original in color). Also visible
quenching limits of Fig. 9 are independent of line here is the 6.3 mm tube and the threaded nut. The fitting was
pressure for each fuel. This suggests that, as vertically oriented and the upstream absolute pressure was
4 bar.

100 0.4
Quenching Mass Flow Rate (mg/s)

0.378 mg/s
Quenching Diameter ( μm)

0.3
10 Methane
0.336 mg/s
Hydrogen
Methane
Propane 0.2 Propane
h
m
1 Hydrogen p
Linear (h)
0.1 0.028 mg/s Linear (m)
Linear (p)

0.1 0
1 10 100 1000 1 10 100 1000
Pressure (bar) Pressure (bar)

Fig. 9. Quenching limit (minimum flaming flow rate) versus

Fig. 7. Minimum diameter of a circular port for flaming upstream absolute pressure for a 6.3 mm leaky compression
(quenching diameter) as a function of upstream absolute fitting in the vertical orientation. The line fits represent mean
pressure assuming isentropic choked flow. values of the measured quenching limits.

Abu Dhabi, U.A.E., November 4-5, 2008 7

 anticipated from the round-hole burner results of Fig. 4, the key parameter controlling quenching limits is the fuel
mass flow rate. For these low flow rates and small leak sizes the line pressure does not have a significant effect on
the velocity field in the flame region.
The mean hydrogen flow rate at the quenching limit is about an order of magnitude lower than for the other
fuels, which is consistent with the results of Fig. 3. For leaky fittings, the quenching flow rates for hydrogen,
methane, and propane are about an order of magnitude higher than the corresponding flows for round-hole burners
(Fig. 4). This is attributed to additional burner heat losses in the leaky fittings, where the flames burn near concave
metal surfaces.
The effects of orientation of leaky fittings are investigated in Fig. 10. Fitting orientation had little or no effect
on the quenching limits of hydrogen flames. This is anticipated as these flames are small at their limits and is the
results are consistent with the round-hole hydrogen measurements of Figs. 5 and 6. Nonetheless, burner orientation
did have an effect on the quenching limits of propane and methane flames, with the inverted limits averaging 20%
lower than the vertical limits. These flames are larger than the corresponding hydrogen flames such that buoyancy is
beginning to have a significant effect on the flow field. The inverted orientation has the fitting below the tube, and
this minimizes the surface area of the flame impingement, leading to a reduced heat loss and thus lower quenching
limits.
The effects of fitting diameter on quenching limits are examined in Fig. 11. Diameter had a significant impact
on the quenching limits in all cases; on average as tube diameter doubles, the fuel mass flow rate at quenching
increases 30%. This is attributed to increased heat losses associated with larger fittings.
This study has identified new questions that warrant further study. For example, what corrosive effects do
leak flames have on containment materials? Can permeation leaks support flames? Can surface coatings, e.g.,
intumescent paints, be applied that will increase the fuel flow rate at quenching? An improved understanding of
hydrogen leaks is necessary to ensure safe use of hydrogen in the commercial sector.

0.4 0.6
3.2 mm
6.4 mm
Vertical 0.5 12.7 mm
Quenching Mass Flow Rate (mg/s)
Quenching Mass Flow Rate (mg/s)

0.3 Horizontal
Inverted
0.4

0.2 0.3

0.2

0.1
0.1

0 0
Hydrogen Methane Propane Hydrogen Methane Propane

Fig. 11. Quenching limits for leaky compression fittings for

Fig. 10. Quenching limits for a 6.3 mm leaky compression tube diameters of 3.1 mm, 6.3 mm and 12.6 mm. The fittings
fitting in vertical, horizontal and inverted orientations. The were vertically oriented and upstream absolute pressure was
upstream absolute pressure was 4 bar. 4 bar.
5. Conclusions
Flame quenching and blowoff limits were measured for diffusion flames results from fuel issuing out of small
burners. Hydrogen fuel was emphasized, and was compared with methane and propane for a variety of burner types
that characterized a range of heat losses. The key findings are as follows:
1. Flame quenching limits for round-hole burners are nearly independent of burner diameter. These limits are
reasonably predicted by a simple model presented here. The quenching limit mass flow rate of hydrogen,
5.6 μg/s, which is only is about one-tenth of that of methane and propane.

Abu Dhabi, U.A.E., November 4-5, 2008 8

2. Existing codes and standards require leak rates below 200 mL/h (36 μg/s) for natural gas and 20 mL/h
(0.46 μg/s) for hydrogen [20]. The measurements show that these flow rates are well below the associated
quenching limits.
3. The mass flow rate at blowoff increases with burner diameter, and for hydrogen is about 10 times that for
methane and propane for the same diameter burner. Thus the limits of stable combustion are much wider for
hydrogen than for the other fuels.
4. For round-hole burners, the quenching flow rate was found to depend weakly on burner type, and to be
independent of burner orientation. These observations are consistent with known wall heat loss and buoyancy
effects.
5. For an isentropic choked leak, hydrogen is susceptible to flaming leaks for hole diameters that are smaller
than those for methane or propane. This suggests that for hydrogen stored at 690 bar, a pinhole leak as small
as 0.4 μm could support a stable flame.
6. The weakest hydrogen flame found in this study, which was generated with a hypodermic needle with an
inner diameter of 0.152 mm, is believed to be the weakest flame ever observed, with a fuel flow rate of
3.9 μg/s and a heat release rate of 0.46 W.
7. The minimum flow rate necessary for sustaining a hydrogen flame on a leaky 6.3 mm tube compression
fitting is 28 μg/s. This is about an order of magnitude lower than for propane or methane. The minimum mass
flow rate for all fuels is independent of upstream (line) pressure, and is sufficient to produce bubbles when a
soap-water solution is applied.
8. Fitting orientation had about a 20% effect on the quenching limits of the leaky compression fittings
considered here, while a doubling of tube size on average results in a 30% increase in quenching flow rate.

Acknowledgments
This work is supported by NIST under the technical direction of J. Yang. The assistance of B.H. Chao and
N.R. Morton is appreciated.

References
[1] Marban G, Valdes-Solis T. Towards the hydrogen economy? Int J Hydrogen Energy 2007;32:1625-1637.
[2] Yamawaki M, Nishihara T, Inagaki Y, Minato K, Oigawa H, Onuki K, Hino R, Ogawa M. Application of
nuclear energy for environmentally friendly hydrogen generation. Int J Hydrogen Energy 2007;32:2719-
2725.
[3] Sherif SA, Barbir F, Veziroglu TN. Wind energy and the hydrogen economy - review of the technology.
Solar Energy 2005;78:647-660.
[4] Von Jouanne A, Husain I, Wallace A, Yokochi A. Gone with the wind. IEEE Industry Applications
Magazine July/Aug 2005:12-19.
[5] Kanury AM. Introduction to combustion phenomena. New York: Gordon and Breach, 1975. p. 131.
[6] MacIntyre I, Tchouvelev AV, Hay DR, Wong J, Grant J, Benard P. Canadian hydrogen safety program. Int J
Hydrogen Energy 2007;32;2134-2143.
[7] Takeno K, Okabayashi K, Kouchi A, Nonaka T, Hashiguchi K, Chitose K. Dispersion and explosion field
tests for 40 MPa pressurized hydrogen. Int J Hydrogen Energy 2007;32(13):2144-2153.
[8] Dorofeev SB. Evaluation of safety distances related to unconfined hydrogen explosions. Int J Hydrogen
Energy 2007;32:2118-2124.
[9] Baker J, Calvert ME, Murphy DW. Structure and dynamics of laminar jet micro-slot diffusion flames.
Journal of Heat Transfer 2002;124:783-790.
[10] Ban H, Venkatesh S, Saito K. Convection-diffusion controlled laminar micro flames. Transactions of the
ASME 1994;116:954-959.
[11] Cheng TS, Chao Y-C, Wu C-Y, Li Y-H, Nakamura Y, Lee K-Y, Yuan T, Leu TS. Experimental and
numerical investigation of microscale hydrogen diffusion flames. Proceedings of the Combustion Institute
2005;30:2489-2497.
[12] Nakamura Y, Yamashita H, Saito K. A numerical study on extinction behaviour of laminar micro-diffusion
flames. Combustion Theory and Modelling 2006;10(6):927-938.
[13] Matta LM, Neumeier Y, Lemon B, Zinn BT. Characteristics of microscale diffusion flames. Proc. Combust.
Inst. 2002; 29:933-938.
[14] Cheng TS, Chen CP, Chen CS, Li YH, Wu CY, Chao YC. Characteristics of microjet methane diffusion
flames. Combust. Theory Modeling 2006;10:861-881.

Abu Dhabi, U.A.E., November 4-5, 2008 9

[15] Lee ID, Smith OI, Karagozian AR. Hydrogen and helium leak rates from micromachined orifices. AIAA
Journal 2003;41(3):457-463.
[16] Schefer RW, Houf WG, San Marchi C, Chernicoff WP, Englom L. Characterization of leaks from
compressed hydrogen dispensing systems and related components. Int J Hydrogen Energy 2006;31:1247-
1260.
[17] Ge X, Sutton WH. Analysis and test of compressed hydrogen interface leakage by commercial stainless steel
(NPT) fittings. SAE International 2006:35-47.
[18] Swain MR, Swain MN. A comparison of H2, CH4, and C3H8 fuel leakage in residential settings. Int J
Hydrogen Energy 1992;17:807-815.
[19] Kalghatgi GT. Blow-out stability of gaseous jet diffusion flames. Part I: In still air. Combustion Science and
Technology 1981;26:5:233-239.
[20] Takahashi M, Tamura Y., Suzuki J, Watanabe S, Investigation of the allowable flow rate of hydrogen
leakage on receptacle. Paper 2008-01-724, SAE International, Detroit, April 2008.
[21] Roper FG. The prediction of laminar jet diffusion flame sizes: Part 1. Theoretical model. Combust. Flame
1977;29:219-226.
[22] Sunderland PB, Mendelson BJ, Yuan ZG, Urban DL. Shapes of buoyant and nonbuoyant laminar jet
diffusion flames. Combust. Flame 1999;116:376-386.
[23] Weast RC, Astle MJ. CRC Handbook of Chemistry and Physics, 59th edition. West Palm Beach: CRC Press,
Inc., 1979. p. F-58.
[24] Ronney PD, Wu MS, Pearlman HG. Structure of flame balls at low lewis-number (SOFBALL): preliminary
results from the STS-83 and STS-94 space flight experiments. Paper AIAA-1998-463, Aerospace Sciences
Meeting, Reno, Nevada, January, 1998.
[25] Zeldovich Y. Theory of Combustion and Detonation of Gases. Moscow, Russia: Academy of Science, 1994.
[26] Ronney PD, Wu MS, Pearlman HG. Experimental study of flame balls in space: preliminary results from
STS-83. AIAA Journal 1998;36(8):1361-1368.

Abu Dhabi, U.A.E., November 4-5, 2008 10

 Experimental and Numerical Investigation of Extremely Weak Hydrogen Diffusion Flames

V.R. Lecoustre1, C.W Moran1, P.B. Sunderland1,*, B.H. Chao2, R.L. Axelbaum3
1 2
Dept. of Fire Protection Engineering Dept. of Mechanical Engineering
University of Maryland University of Hawaii,
College Park MD 20742 Honolulu HI 96822
3
Dept. of Energy, Environmental
and Chemical Engineering
Washington University
St. Louis MO 63130
Abstract
Hydrogen micro diffusion flames were studied experimentally and numerically. The experiments involved gas jets of
hydrogen flowing downward from stainless steel tube burners with inside diameters of 0.15 mm. At their quenching
limits, these flames had heat release rates of 0.46 and 0.25 W in air and in oxygen, respectively. The corresponding
hydrogen flow rates were 3.9 and 2.1 mg/s. These are the weakest flames ever observed, and are even weaker than
microgravity flame balls. The flames were modeled numerically as spherical diffusion flames with detailed chemistry
and transport. The modeling results confirmed the quenching limits and revealed high rates of reactant leakage near the
limits. The model predicted double peaked heat release rate profiles for some flames. The results may be of use in future
microcombustor designs.

Introduction compared them to the shapes predicted analytically. They

The use of microcombustion processes for electric found that buoyancy effects are negligible for small
power generation has potential advantages over flames and their shapes tend to be spherical. At this size,
conventional electrochemical batteries, in terms of power the effects of diffusion are comparable to the effects of
generation per unit volume and energy storage per unit convection. Matta et al. [7] reached the same conclusion
mass [1]. The increasing needs and demand for smaller considering a preheated weak propane flame issued at a
scale and higher density power sources motivates the rate of 0.31 cc/min from a 101 µm tube and equivalent to
development of micropower devices. 0.49 W.
Recent developments in silicon technologies and Cheng et al. [8] characterized microjet methane
development of new materials have led to opportunities of diffusion flames issuing from burners with inner
creating highly effective small thermoelectric and diameters ranging from 186 – 778 µm. They compared
photovoltaic power generators [2]. Their use opens new the measured flame shapes with those predicted using a
prospects in creation of micro scale energy supply 2-D code with detailed chemistry model based on GRI-
devices. Recent developments Micro-Electro-Mechanical Mech 3.0. They observed that near extinction, the flame
Systems (MEMS) [3] have enabled new microcombustor burns in a diffusion mode. Hydrogen microscale flames
designs. Microcombustors can be defined as reactors with were also studied both experimentally and numerically
characteristic length scales on the order of millimeters or [9]. Those flames were produced from small tubes (0.2
less. and 0.48 mm) and were spherical, with a radius of 1 mm.
Laminar diffusion flames can be classified in three This work observed that the increased diffusivity of H2
categories. The first one corresponds to diffusion leads to lean-burn conditions for the flame.
controlled flames, which can be accurately modeled using Nakamura et al. [10] numerically studied methane
Burke-Schumann theory [4]. The second category micro diffusion flames on circular burners of diameter
comprises flames that are controlled by momentum or less than 1 mm. They found that small flames have the
buoyancy, such as the Roper flame [5]. The third category same, nearly spherical, structure as those in microgravity,
corresponds to micro diffusion flames, where momentum with similar species and temperature distributions. The
and diffusion effects are comparable. spherical shape occurs when the Peclet number, Pe, of the
Recent attention has been dedicated to microflames. flame is lower than 5. The Peclet is defined as the ratio of
Ban et al. [6] observed flame shapes of small hydrocarbon the mean exit velocity Ue to the diffusion velocity UD.
laminar diffusion flames issuing from small fuel jets and This number is defined as:

* Corresponding author: [email protected]

Proceedings of the 6th U.S. National Combustion Meeting
 U el D
Pe = , (1)
D
with lD and D being the characteristic diffusion length and
the mass diffusivity coefficient, respectively.
Nakurama et al. [10] predicted a near quenching
methane-air micro-flame releasing 0.5 W. The existence
of a minimum flame size below which no flame exists
was predicted, the minimum flame height being
comparable to the reaction layer thickness. This is

WE
confirmed by Matta et al [7], who concluded that a micro-
flame must be longer than its quenching distance to exist.
The ability to burn weak but stable flames is
important in the design of microcombustors. In some
microcombustors, weak flames may be suitable as
permanent pilots, thus replacing electrical ignitors. Weak
flames also can increase the turndown ratios of
microcombustors.
Ronney et al. [11] observed the burning of spherical
flame balls in microgravity during the STS-83/MSL-1
Space Shuttle mission. Heat released by those flames was
as low as 1 W. Subsequently they achieved flames as
weak as 0.5 W. It is believed these were the weakest
flames recorded until the present work.
Butler et al. [12] examined quenching limits
associated with various fuels and burner configurations.
Their weakest flames involved hydrogen issuing
downward into air from a hypodermic tube with an inside
diameter of 0.15 mm. These flames, and a related flame
burning in oxygen, are the subject of this study.
This work aims to experimentally observe weak 0.5 mm
hydrogen micro-flames at their quenching limits. This
work also investigates these and related flames
numerically to study their structure in detail. Special Fig. 1. Color images of hydrogen flames at their quenching
attention on the effects of the burner on the flame limits burning in air (left) and oxygen (right).
structure is carried. The numerical code used here is a
one-dimensional, time accurate spherical diffusion flame
code with detailed hydrogen/air combustion model and the flames are smaller than the smallest letters on a U.S.
transport properties. dime. The flames are hazy, suggesting distributed reaction
zones rather than thin flame sheets.
Experimental The heat release rates associated with the mass flow
The experiments involved hydrogen non-premixed rates are provided in Table 1, assuming complete
flames burning in quiescent air and weakly combustion and a heat of combustion of 120 kJ/g. Also
counterflowing oxygen. Experimental details are reported presented in Table 1 are the Reynolds numbers, Re, at the
in Butler et al. [12]. The burners used were made of burner exit based on hydrogen properties at laboratory
stainless steel with inside and outside diameters of 0.15 conditions. At these low Reynolds numbers, the hydrogen
and 0.30 mm and with downward orientation. Hydrogen flow in the hypodermic tube has similar conditions to
flow rates were measured with a soap bubble meter. The those experienced in a Stokes flow, where viscous forces
hydrogen flames were not visible but were detected with a
thermocouple 1 cm above the burner tip. Following
ignition, the hydrogen flow rate was slowly reduced until Table 1. Summary of the hydrogen flames at their quenching
each flame extinguished at its quenching limit. limits.
Images of the two hydrogen flames at their quenching
limits are shown in Fig. 1. The test conditions for these Oxidizer Flow rate Ue Re Fr Pe Power
flames are given in Table 1. The images were recorded μg/s m/s - - - W
with a Nikon D100 camera at ISO 200, f/1.4, and 30 s. air 3.9 2.5 3.96 65 5.3 0.46
The word “WE” is included at flame scale to show that O2 2.1 1.4 2.13 36 3.0 0.25

2
are very important. The velocities of the issued hydrogen velocity, ρ is the gas density, cp is the averaged specific
are 2.5 m/s and 1.4 m/s for H2/air and H2/O2, respectively. heat at constant pressure, λ is the heat conductivity of the
Froude number is defined here as Fr = Ue / ( g d )1/2, mixture, hk is the specific enthalpy of the kth species, cp,k is
with Ue the mean exit velocity, g = 9.81 m/s2 and d the the specific heat of the kth species, Vk is the diffusion
burner inside diameter. The relatively high Froude velocity of kth species, ωk is the production rate of the kth
numbers in Table 1 reveal the flames to be minimally species, K is the total number of species considered in the
affected by gravity. Examination of the Peclet number, reaction mechanism, and Ra is a term expressing the local
Pe, assesses the dominance of diffusion over momentum. radiative heat loss rate.
From Eq. (1), it can be derived: Conventional finite difference techniques with non-
⎛l ⎞ uniform mesh spacing were adopted for the discretization
Pe = Re D ⎜⎜ D ⎟⎟ Sc , (2) of the differential equations. The transient terms were
⎝D⎠
expressed by a forward difference formula, the diffusive
with Re the Reynolds number associated with the
terms by a central difference formula, and the convective
conditions at the burner outlet, lD the characteristic
terms by an upwind difference formula. The discretized
diffusion length, D the burner diameter and Sc the
equations were solved using Sandia’s Twopoint package
Schmidt number of the fuel, taken here as 0.204 for
[16], which uses a modified damped Newton’s method to
H2/air and 0.22 for H2/O2 [13]. If the characteristic
solve transient and steady-state boundary value problems.
diffusion length is taken as the same order of magnitude
The chemical reaction rates, the thermodynamic
as the flame size, 1 mm, the Peclet numbers are 5.3 and 3
properties, and the transport properties were evaluated by
for the H2/air and H2/O2 flames, respectively. According
CHEMKIN and the Transport library [15,17]. The net
to Han [6], flames with such small Peclet numbers are
spherical (or hemispherical here). chemical production rate of each species, ωk, results from
a competition between all the chemical reactions
involving that species and is calculated using the modified
Numerical Methods
Laminar spherical hydrogen flames at various flow Arrhenius form:
rates and issuing from different burners were numerically ⎛ Ea ⎞
k f = ΒT b exp⎜ − ⎟ , (6)
modeled using a one-dimensional solver, allowing the ⎝ RT ⎠
modeling of steady state and transient laminar spherical where B is the pre-exponential factor, R is the ideal gas
diffusion flames. The use of spherical coordinates is constant and Ea is the activation energy of the reaction. B,
motivated by the low values of the Peclet number for the R, and Ea were given by the reaction mechanism and
experimental flames, see Table 1. written into an input file read by the code. The gas phase
The solver was modified from the PREMIX [14] chemistry model used was based on GRI 3.0 mechanism
code, which was originally developed for the study of [18]. Reactions considered are reported in Table 2 along
one-dimensional freely propagating and burner-stabilized with the corresponding parameters for Eq. (6). Mixture-
premixed laminar flames and which is a part of Sandia’s average diffusion was considered.
CHEMKIN package [15]. The code features include the Modified transport properties were introduced to
use of detailed chemistry and detailed transport properties improve agreement with observed flame sizes. Using the
using CHEMKIN II format. The equations numerically same code, Santa et al. [19] have shown that increasing
solved are: the diffusivities of all species and heat by 30% yields
∂ρ ∂ higher peak temperatures and smaller flames. This
A(r ) + (m& ) = 0 , (3)
∂t ∂r strategy was maintained in the present simulations.
∂T ∂T 1 ∂ ∂T Local scalar dissipation rate was computed using the
ρA(r ) + m& − ( A( r )λ )+ following definition:
∂t ∂r c p ∂r ∂r
(4)
K 2
A( r ) ∂T
cp ∑ ( ρc p , k Yk V k
∂r
+ hk ω& k ) + Ra = 0, ⎛ dZ (r )
χ (r ) = 2α mix (r )⎜⎜
⎞
⎟ ,
⎟
(7)
⎝ dr
k =1
r =r ⎠

∂Yk ∂Y ∂ The diffusivity coefficient αmix in Eq. (7) is the local

ρ A(r ) + m& k + ( A( r ) ρ Yk Vk ) thermal diffusivity of the mixture. The mixture fraction
∂t ∂r ∂r (5)
was computed from the local mass fractions of oxygen
− A( r ) M k ω& k = 0, k = 1,2,..., K , and hydrogen atoms. The mixture fraction is expressed as:

where A(r) is the surface area of the sphere of radius r, T YH (r ) + 2(YO ,∞ − YO (r ) )

is the local temperature, Yk is the mass fraction of the kth Z (r ) = . (8)
species, Wk is the molecular weight of the kth species, t is 1 + 2YO ,∞
time, r is the radial spatial coordinate, U is the radial flow

3
 The burner surface was defined as the inner boundary boundary such that it did not interfere with the flame,
of the computational domain. At this location, an simulating an infinite boundary. The gradients at the outer
adiabatic boundary condition was considered, allowing boundary were negligible. The pressure was held constant
the burner temperature to increase. Cheng [9] estimated over the domain at 0.98 bar. For all the simulations
burner temperatures of about 1010 K based on the burner performed, heat losses by radiation were neglected.
glowing. Moreover, this condition frees the study from Cheng et al. [20] showed that the effects of radiation on
effects of heat losses to the burner and emphasizes the flame structure are small for low Reynolds number
effects due to the burner presence. Cheng et al. [20] flames. The flames modeled were purely adiabatic.
showed that for small flames with low Reynolds numbers, The computational domain for the predictions of
variable wall temperature conditions yield better diffusion hydrogen micro-flames ranged from the burner
agreement between computed and measured data. exit (1 μm, 75 μm, 300 μm or 3.175 mm) to 150 cm. For
The mass flux allowed backward species diffusion all cases considered, the position of the outer boundary
into the burner, conditions that were adopted by Chang et was far enough from the flame that the temperature
al. [9]. Under these conditions, the mass fraction of the kth gradient there was negligible. Non-uniform grids were
species at the burner surface was expressed by: generated, with clustered mesh points near the reacting
εk zone and coarse points near the outer boundary. Adaptive
Yk = , (9) mesh point addition was used to reduce the gradients and
ρVk A
1+ curvature, improving accuracy. New grid points were
m& b
automatically added until values of gradients and
with εk as the input mass fraction of the kth species of the curvature fell below user specified limits. Further addition
initial mixture. of grid points did not affect the solution. Typically the
At the outer boundary, temperature was assigned a computational domain consisted of 300 points.
constant value. Species mass fractions at this boundary Steady state solutions of H2/air and H2/O2 were
were set to as either air or oxygen. For all the simulations, considered. For the study of quenching limits, the
special attention was dedicated to the position of the outer following process was observed. A first solution based on
the characteristic of the flame studied was generated by
the code with a moderate mass flow. Then this solution
Table 2. H2-O2 reactions mechanisms considered for the was used as the starting condition for a new simulation
simulations. From GRI Mech. 3.0 [18]. using the same parameters but with a lower mass flow
rate. The process was reiterated until extinction was
Name Reaction b Ea defined. Extinction was defined as a solution of the flame
(gmol.cm­3.s­1) (cal.gmol­1)
R1 2O+M → O2+M 1.20E+17 -1.0 0
where no reaction was observed and with the absence of a
R2 O+H+M → OH+M 5.00E+17 -1.0 0.0 significant raise of temperature.
R3 O+H2 → H+OH 3.87E+04 2.7 6260
R4 O+HO2 → OH+O2 2.00E+13 0.0 0 Numerical Results
R5 O+H2O2 → OH+HO2 9.63E+06 2.0 4000 Numerical predictions of steady state hydrogen flame
R33 H+O2+M → HO2+M 2.80E+18 -0.86 0 were performed considering four different burner radii,
R34 H+2O2 → HO2+O2 2.08E+19 -1.24 0
R35 H+O2+H2O → HO2+H2O 1.13E+19 -0.76 0
1μm, 75 μm, 300 μm or 3.175 mm and two different
R36 H+O2+N2 → HO2+N2 2.60E+19 -1.24 0 oxidizers, quiescent air and quiescent oxygen. Predictions
R37 H+O2+Ar → HO2+Ar 7.00E+17 -0.8 0 of flames were realized using the 1-D spherical code
R38 H+O2 → O+OH 2.65E+16 -0.67 17041 considering adiabatic conditions in temperature at the
R39 2H+M → H2+M 1.00E+18 -1.0 0 burner surface and neglecting radiation.
R40 2H+H2 → 2H2 9.00E+16 -0.60 0 Numerical investigations of quenching limits were
R41 2H+H2O → H2+H2O 6.00E+19 -1.25 0
performed first to characterize the lowest hydrogen mass
R43 H+OH+M → H2O+M 2.20E+22 -2.00 0
R44 H+HO2 → O+H2O 3.97E+12 0.0 671
flow sustaining a flame and to compare it with the
R45 H+HO2 → O2+H2 4.48E+13 0.0 1068 experimental results. The radius of the burner was taken
R46 H+HO2 → 2OH 8.40E+13 0.0 635 at 75 μm for the cases H2/air and H2/O2, matching the
R47 H+H2O2 → HO2+H2 1.21E+07 2.0 5200 dimension of the hypodermic tube used in both
R48 H+H2O2 → OH+H2O 1.00E+13 0.0 3600 experiments. Numerically, the lowest mass flow rate was
R84 OH+H2 → H+H2O 2.16E+08 1.51 3430 achieved by decreasing hydrogen flow rate and reusing
R85 2OH+M → H2O2+M 7.40E+13 -0.37 0
R86 2OH → O+H2O 3.57E+04 2.4 -2110
the former steady-state solution as initial guess for the
R87 OH+HO2 → O2+H2O 1.45E+13 0.0 -500 solver. This sped convergence.
R88 OH+H2O2 → HO2+H2O 2.00E+12 0.0 427 The lowest hydrogen mass flow rates sustaining
R89 OH+H2O2 → HO2+H2O 1.70E+18 0.0 29410 flames were numerically predicted to be 3.65 and
R115 2HO2 → O2+H2O2 1.30E+11 0.0 -1630 2.67 μg/s for the H2/air and H2/O2 flames, respectively.
R116 2HO2 → O2+H2O2 4.20E+14 0.0 12000 The heat release rates associated with those mass flow

4
rates are 0.41 and 0.31 W. These heat release rates were burner and 1.18 s-1 for the 1 μm radius burner. Due to the
obtained by numerical integration of the local heat release small flame radius, about 200 μm for both, the heat
rate computed by the code using CHEMKIN library. release rate density is very high, peaking at 3800 and
These values are in reasonable agreement with the 8000 W/cm3, respectively.
measured mass flow rates and heat release rates of The smaller burner considered presents the highest
Table 1. scalar dissipation rate value prior extinction. Its predicted
In Fig. 2, predicted H2/air flame structure issued prior surface temperature is 300 K, which implies that the
extinction and from a 75 μm radius is plotted and burner does not remove any heat from the system.
compared with the structure of a H2/air flame issuing Moreover, at the burner surface, the mass fraction of H2 is
from a 1 μm radius burner, at a mass flow rate of nearly unity, which is not the case for the 75 μm radius
3.49 μg/s. This mass flow rate corresponds to the lowest burner.
mass flow rate sustaining a flame issuing from a spherical The oxidizer presence is important near the burner for
burner with a radius of 1 μm. For this configuration, the the 75 µm burner. The oxygen and intermediate species
smallest heat released rate predicted is 0.4 W. mass fractions at the burner surface are high. The mass
Superimposing the two data set shows similarities fraction of H2 drops considerably outside the burner. For
shared by the two flames. A common region is located the 75 μm radius burner, H2 mass fraction at the burner
beyond 75 μm, corresponding to the radius of the burner. surface is lower than 1%. The reaction zone exists under
Figure 2 shows the flame truncation induced by the lean conditions. This was also observed by Chang et al.
physical presence of the burner. Both flames present the [9].
same peak temperature of 1290 K. Scalar dissipation rates The main reactions contributing to the heat release
at the flame location are 0.98 s-1 for the 75 μm radius rate for the 75 μm radius burner are:

0
10 1500 0.03
rb=75 μm rb = 75 μm
O2

Rate of progress (mol/cm -s)

-1
10 T

3
H 2O
Temperature (K)

0.02
Mass Fraction

-2
10 H2 1000
R84
O
-3
10
OH R46
H
-4
0.01
10 500 R3
HO2
R38
R35
10-5
H 2 O2 R45 R36 0
-6
10 0
0 100 200 300 400 0 100 200 300 400
Radius (μm) Radius (μm)

0.03 1500 4000

rb = 75 μm rb = 75 μm
Heat Release Rate (W/cm )
Production rate (mol/cm -s)

0.02 H 2O T
3

H
Temperature (K)

0.01 1000 QC

2000
0 O
OH R84

-0.01 O2 500
R35
H2 R46

-0.02 R45
R36
0
R3
R38
-0.03 0
0 100 200 300 400 0 100 200 300 400
Radius (μm) Radius (μm)

Fig. 2. From top left, clockwise, plot of the species mass fraction, elementary reactions rate of progress, elementary reactions heat
release rate and species production rate for H2/Air flames at the onset of kinetic extinction. Burner radii are 75 μm plain line (1 μm
dashed), supplied mass flow rate is set at 3.65 μg/s (3.49 μg/s), Qc is 0.41 W (0.40 W).

5
 R84 OH+H2 → H2O+H , R46 H+HO2 → 2OH ,
R35 H+O2+H2O → HO2+H2O , and to a lesser measure,
R46 H+HO2 → 2OH , R45 H+HO2 → H2 + O2 , and
R45 H+HO2 → O2+H2 , and R36 H+O2+N2 → HO2+N2 .
R36 H+O2+N2 → HO2+N2. The main reactions predicted for the 1 μm radius
There is a significant endothermic reaction, the chain burner flame are the same, except their order of
branching reaction importance are different. The main reaction, in terms of
R38 H+O2 → OH+O , heat release rate is
which consumes part of the energy release. R46 H+HO2 → 2OH ,
The major contributors to the heat release rate are not followed by
necessarily the reactions with the highest rates of R45 H+HO2 → H2 + O2 ,
progress. Indeed, although reaction R35 H+O2+H2O → HO2+H2O , and
R84 OH+H2 → H2O+H R84 OH+H2 → H2O+H ,
shows the highest rate of progress and the highest heat all are reactions involving the consumption of H radical
release rate, reaction except for reaction R84. In terms of rate of progress, the
R3 O+H2 → H+OH , important reactions are
which does not contribute to the heat release rate, has the R84 OH+H2 → H2O+H ,
second highest rate of progress. The third highest belongs R46 H+HO2 → 2OH ,
to reaction R35 H+O2+H2O → HO2+H2O ,
R38 H+O2 → OH+O , R45 H+HO2 → H2 + O2, and
followed by chain termination reaction R36 H+O2+N2 → HO2+N2 .
R35 H+O2+H2O → HO2+H2O , Note that reactions
chain branching R3 O+H2 → H+OH and

0
10 1500 0.1
O2
T

Rate of progress (mol/cm -s)

-1
10 Burner 0.08

3
Burner H 2O surface
surface
Temperature (K)
Mass Fraction

-2
10 1000
O 0.06
R84
-3
10 H2 OH
0.04
R38
-4
10 HO2 500
R3 0.02
H
-5 R35
10 R46
R4
H 2 O2 R34
0
R87
-6
10 0
0 100 200 300 400 0 100 200 300 400
Radius (μm) Radius (μm)

1500 12000
0.1
T
Heat Release Rate (W/cm )
Production rate (mol/cm -s)

Burner QC
3

Burner H 2O surface
8000
0.05 surface
Temperature (K)

1000
H
O R84
0 4000
OH
R87 R35
500 R4
O2 R34
-0.05 0
H2 R45
R3

R38
-0.1
0 -4000
0 100 200 300 400 0 100 200 300 400
Radius (μm) Radius (μm)
Fig. 3. From top left, clockwise, plot of the species mass fraction, elementary reactions rate of progress, elementary reactions heat
release rate and species production rate for H2/O2 flame at the onset of kinetic extinction. Burner radius is 75 μm, hydrogen mass flow
rate is 2.67 μg/s, and Qc is 0.31 W.

6
 R38 H+O2 → OH+O closer to the burner to eventually being so close that any
have negative rates of progress close to the burner. further decrease in mass flow rate does not change the
The structure of a weak hydrogen flame burning in flame position, the burner preventing the flame from
oxygen at the onset of extinction was numerically moving inward. This is observed in Fig. 2 for the burner
investigated and is shown in Fig. 3. The lowest possible radius of 3.175 mm for mass flow rate lower than
flow rate sustaining a steady state flame is found to be 0.1 mg/s. This has for effect to truncate the flame and to
2.67 μg/s. At this flow rate, the peak temperature is accelerate flame extinction.
1400 K and the heat generated is 0.31 W. The scalar Extinction is observed for a peak temperature of about
dissipation rate at the peak temperature is 2.43 s-1. This 1300 K. This value is common for the four burner radii
value is twice the scalar dissipation rate of the extinction considered. Reducing the mass flow rate below 0.1 mg/s
limit H2/air flame generated by the 1 μm radius burner. leads to a similar drop in peak temperature for burner
For the same burner size, higher oxygen and OH radii of 1, 75 and 300 μm. The 1 μm radius burner does
radical leakage at the flame location is observed for the not affect the flame, as observed in Fig. 2, and thus can be
system H2/O2 than for the H2/air system. However H2/O2 taken as reference. Therefore, the mass flow rate of
flames present faster chemistry, with peak values of 3.49 μg/s is the lowest mass flow rate possible regardless
species production rate 5 times higher than those of burner size. The flame produced generates a heat
predicted for the H2/air flame. This is due to an increase release rate of 0.4 W, which is the weakest flame
of reaction rates of progress. predicted numerically.
The main chemical reactions contributing to the heat We aim to investigate the structure differences as the
release rate for H2/O2 flames at the onset of extinction are: regime change from a well developed flame with a mass
R84 OH+H2 → H2O+H , flow rate high enough to be free from the burner effects to
R35 H+O2+H2O → HO2+H2O , a flame to a flame at the onset of extinction. Figure 2
R4 O+HO2 → OH+O2 , represents the latter.
R87 OH+HO2 → O2+H2O , Figure 5 plots the temperature, local heat release rate
R34 H+2O2 → HO2+O2 , and and rate of progress, of main reactions, species net
R45 H+HO2 → O2+H2 production rate and species distributions for a spherical
for the exothermic reactions, and diffusion flame of hydrogen issuing from a 3.175 mm
R3 O+H2 → H+OH and radius burner at a rate of 10 mg/s into quiescent air. This
R38 H+O2 → O+OH flame is typical of a large and strong diffusion flame. This
for the endothermic ones. The important reactions by flow rate produces a relatively large diffusion flame, with
decreasing rate of progress, are a peak temperature of 2303 K located 22 cm from the
R84 OH+H2 → H2O+H , burner center. The predicted heat release rate is 1275 W.
R38 H+O2 → O+OH (which is positive), The scalar dissipation rate at the peak temperature is
R3 O+H2 → H+OH , 5.2e-6 s-1, characterizing a very low mixing rate,
R35 H+O2+H2O → HO2+H2O, corresponding to a large Damköhler number. The high
R46 H+HO2 → 2OH ,
R4 O+HO2 → OH+O2 , and
R34 H+2O2 → HO2+O2 . 2500 10
2

The H2/O2 flame has a slightly higher temperature at the

onset of extinction (1400 versus 1290 K) than the H2/air
flame. However, its flame extent is smaller than that of
101
the H2/air flame. The location of the peak temperature is
Flame radius (cm)
located at a radius of 130 μm for the H2/O2 flame and at 2000
almost 200 μm for the H2/air flame.
Tmax (K)

The impacts of the mass flow rate and of the burner 3.175 mm 10
0

size were studied for the H2/air flame. Their effects on rb = 3.175 mm
flame temperature, flame position, scalar dissipation rate 1500
were investigated and plotted in Fig. 4. Four different 10-1
1 μm 300 μm
burner radii were considered: 3.175 mm, 300 μm, 75 μm
and 1 μm. rb = 300 μm
75 μm 1 μm
At high mass flow rates, flame temperature and radius 75 μm

are independent of the burner radius. The flame

-2
1000 -3 -2 -1 0 10
1
10 10 10 10 10
temperature plateaus at about 2300 K for mass flow rates mH2 (mg/s)
higher than 0.3 mg/s. Decreasing the mass flow rate Fig. 4. Evolution of predicted H2/Air flame temperature and
reduces the flame radius, the flame moving closer to the flame radius with supplied mass flow rate issued from 3.175
burner. As the flow rate diminishes, the flame moves mm, 300 μm, 75 μm and 1 μm radius burners.

7
flow rate prevents the burner temperature from rising, hydrogen-oxygen combustion. In terms of rate of
keeping it constant at 300 K. The H2 and O2 progress, the importance of the reactions varies.
concentrations near the reaction zone are very low, Exothermic reaction
indicating negligible leakage of fuel or oxidizer. R84 OH+H2 → H2O+H
The overall heat release rate presents two peaks, is the reaction having the highest rate of progress,
instead of one as seen in Figs. 2 and 3. The bigger peak is followed by
near the location of peak temperature, while the smaller R38 H+O2 → O+OH .
peak is close to the burner. The first peak corresponds to The rate of progress of these two reactions is 4 times
the location of water formation, with the reaction higher than the rate of progress of other reactions, namely
R43 H+OH+M → H2O+M R3 O+H2 → H+OH ,
as the major contributor. Other less important reactions at R35 H+O2+H2O → HO2+H2O ,
this location are the exothermic reactions R89 OH+H2O2 → HO2+H2O , and
R84 OH+H2 → H2O+H , R43 H+OH+M → H2O+M .
R85 2OH+M → H2O2+M , Molecular oxygen is consumed by reactions R35 and
R35 H+O2+H2O → HO2+H2O , R38. Molecular hydrogen is consumed by reaction R84.
R89 OH+H2O2 → HO2+H2O and The radical OH is consumed within the inner part of the
R87 OH+HO2 → O2+H2O flame and is produced in the outer part of the flame.
and the endothermic reactions The smaller peak of heat release rate corresponds to
R86 2OH → O+H2O and the contributions of the reactions
R38 H+O2 → O+OH , R40 2H+H2 → 2H2 and
which is one of the most important branching reactions of R41 2H+H2O → H2+H2O .

0
10 2500 8E-08
T

Rate of progress (mol/cm -s)

-1
10 H 2O

3
2000 R38
O2 R84
H2
Temperature (K)
Mass Fraction

-2
10 4E-08
OH 1500
-3
10 R3
R89
1000 R41 R43 R35
O
-4
10 0
R87

500
10-5 H R86

10-6 0 -4E-08
0 10 20 30 40 50 0 10 20 30 40 50
Radius (cm) Radius (cm)

1E-07 2500 0.02

T
Heat Release Rate (W/cm )
Production rate (mol/cm3-s)

H 2O
3

2000 QC 0.015
5E-08
Temperature (K)

H
1500 0.01
O R43
0
OH
O2 1000 R84 0.005
R85
R35
-5E-08 R89
500 R41 0
H2 R86 R87

R38
-1E-07 0 -0.005
0 10 20 30 40 50 0 10 20 30 40 50
Radius (cm) Radius (cm)
Fig. 5. From top left, clockwise, plot of the species mass fraction, elementary reactions rate of progress, elementary reactions heat
release rate and species production rate for H2/Air flame. Burner radius is 3.175 mm, supplied mass flow rate is set at 10 mg/s, Qc is
1275 W.

8
At this location H radicals recombine into H2. Hydrogen magnitude larger than the scalar dissipation rate observed
atom radical is mostly produced in the zone of high at the reaction zone in Fig. 5.
temperatures, with its production peaking near the flame The low flow rates of the flame in Fig. 6 induce
location, located at a reasonable distance from its profound changes on the flame structure. The flame is
consumption location, implying that H radicals diffuse very close to the, the peak temperature being found about
upstream before being recombined into H2. 400 μm from the burner surface. The temperature profile
Figure 6 plots the temperatures, local heat release of the flame is relatively flat, compared to the profiles
rates, rates of progress, species net production rates, and exhibited by flames at the onset of kinetic extinction.
species distributions for a spherical diffusion flame of Burner surface temperature is close to 2000 K.
hydrogen issuing from a 3.175 mm radius burner at a rate Significant oxygen leakage is observed for this flame.
of 0.1 mg/s into quiescent air. This flame is at the turning The mass fraction of hydrogen is very low, less than
point of the burning regime and corresponds to the 0.1%. This is less than the hydrogen mass fraction
dramatic change of slope evident in Fig. 4. observed in Fig. 2, which is about 1%. The mass fraction
This flame shows a different structure than the large of OH radical is high, about 10 times those of the H2/air
and strong diffusion flame of Fig. 5. The flame in Fig. 6 flames at the onset of kinetic extinction.
has a peak temperature of 2059 K, which is far from the The heat release rate curve presents only 1 peak,
temperature observed at kinetic extinction, typically which is located very close to the burner. This
1300 K for H2/air. The predicted heat release rate is corresponds to the production of water and H radical. The
12.52 W, which is still significant, compare to the flames main heat release rate exothermic contributors are
of Figs. 2 and 3. The scalar dissipation rate at the peak reactions
temperature is 1.6e-2 s-1, which is about four orders of R84 OH+H2 → H2O+H ,

100 2500 Burner 0.003

Burner
surface surface

Rate of progress (mol/cm -s)

10-1 H 2O

3
2000 T
O2
0.002
Temperature (K)
Mass Fraction

-2
10 OH
1500
O R84
10-3
0.001
1000
-4 H R38
10 H2

R3
500 R35
10-5
R43 0
HO2 R86
-6
10 0
3000 3500 4000 4500 3000 3500 4000 4500
Radius (μm) Radius (μm)

0.002 Burner 2500 Burner 200

surface surface
Heat Release Rate (W/cm )
Production rate (mol/cm3-s)

H 2000 150
0.001 T
H 2O
Temperature (K)

1500 100
OH
0
O QC
1000 50
O2 R84

-0.001 H2 R35
500 R43 0
R3
R86
R38
-0.002 0 -50
3000 3500 4000 4500 3000 3500 4000 4500
Radius (μm) Radius (μm)
Fig. 6. From top left, clockwise, plot of the species mass fraction, elementary reactions rate of progress, elementary reactions heat
release rate and species production rate for H2/Air flame presenting important leakage. Burner radius is 3.175 mm, supplied mass flow
rate is 100 μg/s, and Qc is 12.52 W.

9
 R35 H+O2+H2O → HO2+H2O) , and ⎛ Ea ⎞
Da ∝ χ st −1 exp⎜⎜ − ⎟
⎟ , (11)
R43 H+OH+M → H2O+M . ⎝ RTad ⎠
The main endothermic reactions are
where Ea is energy of activation, Tad is the adiabatic flame
R38 H+O2 → O+OH , temperature and R the universal gas constant.
R86 2OH → O+H2O and Because reducing the mass flow affects only the scalar
R3 O+H2 → H+OH .
dissipation rate, we can write:
Note that R86 has a negative rate of progress, while
χ st ∝ m −2 , (12)
reaction R84 has the highest rate of progress. The
dominant reactions remain the same as those for the This is the trend observed in Fig. 7.
strong diffusion flame shown in Fig. 5, although their Plotting the peak temperature against the scalar
rates are lower because the flame is truncated by the dissipation rate yields a S-curve as illustrated in Fig. 8. At
burner. low scalar dissipation rates, the peak temperature remains
Kinetic extinction is achieved as the mass flow rate is constant and nearly equal to the adiabatic H2/air flame
gradually decreased. This diminishes the flame temperature. At this point, there is no reactant leakage
temperature and the flame radius, as seen in Fig. 4. This across the flame.
also increases the scalar dissipation rate at the flame Reducing the mass flow rate leads to an increase of
location. The evolution of scalar dissipation rate with the scalar dissipation rate and hence to a decrease of the
mass flow rate can be observed in Fig. 7. The scalar Damköhler number. When the scalar dissipation rate is
dissipation rate is increased by a factor of 106 as the flow greater than 10-3 s-1, the flame peak temperature starts to
is decreased by a factor 103. diminish. Increasing the scalar dissipation rate leads to
Mills and Matalon [21] have shown that the reduced flame size reductions.
Damköhler number Da for adiabatic spherical diffusion At some point the flame cannot move closer to the
flame scales with the mass flow rate to the power two: burner even when the mass flow decreases. When this
Da ∝ m& 2 . (10) occurs, the flame position is not affected by diminishing
the mass flow rate. However this decreases the velocity at
Therefore a decrease of mass flow rate leads to a decrease the flame location, decreasing the strength of the mixing
in Damköhler number, meaning that the characteristic and thus the value of the scalar dissipation rate. This
time of transport is reduced to eventually be of the same creates this backward branch, which can be taken as the
order of magnitude as the chemistry time scale, leading to middle branch of the S-curve. This part of the curve
kinetic extinction. corresponds to a flame structure with an important
Scalar dissipation rate scales as the inverse of a oxidizer leakage, as seen in Fig. 6. The smaller burners
characteristic transport time. If one models the reaction present oxidizer leakage, but in lesser rates. For the
rate with the Arrhenius law, the Damköhler number can smallest burner considered, quenching occurs at the
be expressed by: turning point of the S-curve, presenting the highest scalar
dissipation rate, characteristic of kinetic extinction.

101 2500
1 μm

100
75 μm
300 μm
10-1
2000
χflame (s )

Tmax (K)
-1

-2
10
1 μm

10 -3 3.175 mm
3.175 mm 300 μm
1500
10-4
75 μm
-5
10

10-6 -3 -2 -1 0 1
1000 -6
10 10
-5
10
-4
10
-3
10
-2
10
-1
10
0
10
1
10 10 10 10 10
χflame (s )
-1
mH2 (mg/s)
Fig. 7. Evolution of predicted scalar dissipation rate of H2/Air Fig. 8. Evolution of predicted peak temperature of H2/Air
flame with supplied mass flow rate issued from 3.175 mm, 300 flame with scalar dissipation rate at flame location for 3.175
μm, 75 μm and 1 μm radius burners. mm, 300 μm, 75 μm and 1 μm burner radii.

10
Conclusions Acknowledgments
Hydrogen micro flames were studied both This work was co-funded by the National Aeronautics
experimentally and numerically based on a spherical and Space Administration (D.P. Stocker, technical
model. The weakest flames observed experimentally have contact) and by the National Institute of Standards and
mass flow rates as low as 3.9 μg/s for the H2/air flames Technology (J. Yang, technical contact). The authors
and 2.1 μg/s for the H2/O2 flames, corresponding to heat thank M.S. Butler for his help with the experiments.
release rates of 0.46 and 0.25 W, respectively.
The weakest flames observed numerically, with a References
burner radius matching the experimental dimension, have [1] P.D. Ronney, http://carambola.usc.edu/
mass flow rates of 3.65 μg/s for the H2/air flame and 2.67 Research/MicroFIRE/.
μg/s for the H2/O2 flame, corresponding to heat release [2] A. Carlos Fernandez-Pello, Micropower generation
rates of 0.4 and 0.31 W, respectively. They are in using combustion: issued and approaches.
reasonable agreement with the measured quenching Proceedings of the Combustion Institute 29 (2002)
limits. pp. 883-899.
Kinetic extinction, arising from hydrogen flow rate [3] S.A. Jacobson, A.H. Epstein, An informal survey of
reduction, was numerically observed for adiabatic flames. power MEMS. The International Symposium on
For the larger burners, the presence of the burner prevents Micro-mechanical Engineering (2003)
the flames from moving inward when flow rate is [4] Burke SP, Schuman. Diffusion flames. Ind. Eng.
reduced, resulting in truncated flames. This yields Chem. 20 (1928) 998-1004.
extinction at higher flow rates even under adiabatic [5] Roper FG. The prediction of laminar jet diffusion
conditions and high rates of oxidizer leakage across the flame sizes: Part 1. Theoretical model. Combust.
flame. Flame 29 (1977) 219-226.
The flames are predicted to extinguish at high level of [6] H. Han, S. Venkatesh and K. Saito, Convection-
scalar dissipation rate, of about 1 s-1 for H2/air flames and Diffusion Controlled Laminar Micro Flames,
2 s-1 for H2/O2 flames, which is a factor of 106 higher than Journal of Heat Transfer 954 (1994) pp. 954-959.
for the H2/air diffusion flame at highest flow rate [7] L.M. Matta, Y. Neumeir, B. Lemon and B.T. Zinn,
considered. Because heat losses are neglected in this Characteristic of microscale diffusion flames.
study, the flames predicted are at the onset of kinetic Proceedings of the Combution Institute 29 (2002)
extinction. pp. 933-939.
The main reactions contributing to the flame heat [8] T.S. Cheng, C.-P. Chen, C.-S. Chen, Y.-H. Li, C.-Y.
release rate differ with the Damköhler number. Flames Wu and Y.C. Chao, Characteristics of Microjet
with high Damköhler number present a double peaked methane diffusion flames. Combust. Sci. and Tech.
heat release rate profile. The main exothermic reactions 10 (2006) 861-881.
are, in order of contribution to the heat release rate: the [9] T.S. Cheng, Y.-C. Chao, C.-Y. Wu, Y.-H. Li, Y.
chain terminating reaction H+OH+M → H2O+M and the Nakamura, K.-Y. Lee, T. Yuan, T.S. Leu,
chain propagating reaction OH+H2 → H2O+H for the Experimental and numerical investigation of
bigger peak and the chain-terminating reaction 2H+H2O microscale hydrogen diffusion flames. Proceedings
→ H2+H2O and the chain propagating reaction OH+H2 → of the Combustion Institute 30 (2005) pp. 2489-
H2O+H the for the smaller peak. 2497.
Flames with low Damköhler number do not present a [10] Nakamura Y, Yamashita H, Saito K. A numerical
double peaked heat release rate profile but a single peak study on extinction behaviour of laminar
profile, with the maximum heat release rate close to the microdiffusion flames. Combustion Theory and
burner. The main exothermic reactions are, in order of Modeling 10 (2006) 927-938.
contribution to the heat release rate: the chain propagating [11] P.D. Ronney, M.S. Wu, K.J. Weilandand H.G
reactions H+O2+H2O → HO2+H2O and OH+H2 → Pearlman, Flame Ball Experiments in Space:
H2O+H, and the chain-branching reaction H+HO2 → Preliminary Results from STS-83. AIAA Journal 36
2OH. The hydroperoxy radical (HO2) plays an important (1998) 1361-1368.
role in those flames. [12] M.S. Butler, C.W. Moran, P.B. Sunderland, R.L.
The main endothermic reaction, which absorbs a part Axelbaum, Flame Quenching Limits of Hydrogen
of the heat released by exothermic reactions, is the chain- Leaks, Int. J. Hydrogen Energy, submitted.
branching reaction H+O2 → O+OH. Its contribution [13] B.J. Lee, S.H. Chung, Stabilization of Lifted
remains the same regardless of the Damköhler number. Tribrachial Flames in Nonpremixed Jet. Combust.
Flame 109 (1997) 163-172.
[14] R.J. Kee, J.F. Grear, M.D. Smooke, J.A. Miller, E.
Meeks, A program for modeling steady, laminar,

11
 one-dimensional premixed flames, Sandia National
Laboratories Report No. SAND85-8240 1987.
[15] R.J. Kee, F.M. Rupley, J.A. Miller, E. Meeks,
CHEMKIN-III: A Fortran Chemical Kinetics
Package for the analysis of Gasphase Chemical and
Plasma Kinetics, Sandia National Laboratories
Report No.SAND96-8216.
[16] J.F. Grcar, The twopoint program for boundary
value problems. SAND91- 8230, Sandia National
Laboratories (1996).
[17] R.J. Kee, G. Dixon-Lewis, J. Warnatz, M.E. Coltrin,
J.A. Miller, A Fortran Computer Code Package for
the Evaluation of Gas Phase Multicomponent
Transport Properties. Report SAND86-8246,
Sandia National Laboratories (1988).
[18] Gregory P. Smith, David M. Golden, Michael
Frenklach, Nigel W. Moriarty, Boris Eiteneer,
Mikhail Goldenberg, C. Thomas Bowman, Ronald
K. Hanson, Soonho Song, William C. Gardiner, Jr.,
Vitali V. Lissianski, and Zhiwei Qin
http://www.me.berkeley.edu/gri_mech/
[19] K.J. Santa, Z. Sun, B.H. Chao, P.B. Sunderland,
R.L. Axelbaum, D.L. Urban, D.P. Stocker,
Numerical and experimental observations of
spherical diffusion flames, Combust. Theory and
Modelling 11 (2007) 639-652.
[20] T.S. Cheng, C.-Y. Wu, C.-P. Chen, Y.-H. Li, Y.-C.
Chao, T. Yuan, T.S. Leu, Detailed measurements
and assessment of laminar hydrogen jet diffusion
flames. Combust. Flame 146 (2006) 268.
[21] K. Mills, M. Mantalon, Burner Generated Spherical
Diffusion Flames. Combust. Sci. and Tech. 129
(1997) 295-319.

12
 ABSTRACT

Title of Document: AN ASYMPTOTIC ANALYSIS OF

SPONTANEOUS IGNITION OF HYDROGEN
JETS

Kian Boon Lim, M.S., 2007

Directed By: Assistant Professor Peter B. Sunderland,

Department Of Fire Protection Engineering

Analytical work is presented for the spontaneous ignition of a hydrogen jet emanating

from a slot into air. A similarity solution of the flowfield was obtained. This was

combined with the species and energy conservation equations, which were solved

using activation energy asymptotics. Limits of spontaneous ignition were identified as

functions of slot width, flow rate, and temperatures of the hydrogen jet and ambient

air. Two scenarios are examined: a cool jet flowing into a hot ambient and a hot jet

flowing into a cool ambient. For both scenarios, ignition is favored with an increase

of either the ambient temperature or the hydrogen supply temperature. Moreover, for

the hot ambient scenario, a decrease in local fuel Lewis number also promotes

ignition. The Lewis number of the oxidizer only has a weak effect on ignition.

Because spontaneous ignition is very sensitive to temperature, ignition is predicted to

occur near the edge of the jet if the hydrogen is cooler than the air and on the

centerline if the hydrogen is hotter than the air.

 AN ASYMPTOTIC ANALYSIS OF SPONTANEOUS IGNITION
OF HYDROGEN JETS

By

KIAN BOON LIM

Thesis submitted to the Faculty of the Graduate School of the

University of Maryland, College Park, in partial fulfillment
of the requirements for the degree of
Master of Science
2007

Advisory Committee:
Peter B. Sunderland, Chair
André W. Marshall
Arnaud C. Trouvé
© Copyright by
Kian Boon Lim
2007
 Dedication

To my dearest wife Jolene and precious son Zaccaeus. And to God, for His grace and

providence, and for watching over me every step of the way.

iv
 Acknowledgements

This work was supported by NIST under grant 60NANB5D1209, the Fulbright

Fellowship program under grant number 15052695, and the sponsorship of the

Singapore Civil Defense Force.

First and foremost, I would like to express my heartfelt gratitude to Professor Peter

Sunderland, my advisor, for his guidance, encouragement and genuine concern

throughout my stay in UMD. His progressive thinking and openness has certainly

made things so much easier. He is certainly the coolest Professor that I have the

privilege of working with. Special thanks also to Professor Beei-Huan Chao, from the

University of Hawaii, for his patience and guidance for the research work. I cannot

imagine getting through the mind-boggling and complex mathematics and derivations

without his great help. It is indeed an honor to learn from the Master himself. Also

thanks to Professor Richard Axelbaum, from Washington University in Missouri, for

his insights and comments on my work, which helped in refining and enhancing this

thesis.

I would also like to thank my advisory committee, Professor Arnaud Trouvé and

Professor André Marshall for their valuable time and suggestions.

My sincere thanks to Vivien Lecoustre “The French Meister” for the help he had

given for my research, the French that he taught me (quelle belle langue), and the

friendship we have forged.

v
 Table of Contents

Dedication .................................................................................................................... iv
Acknowledgements....................................................................................................... v
Table of Contents......................................................................................................... vi
List of Tables .............................................................................................................. vii
List of Figures ............................................................................................................ viii
Nomenclature............................................................................................................... xi
Chapter 1: Introduction ................................................................................................. 1
1.1 Motivation for Project......................................................................................... 1
1.2 Literature Review................................................................................................ 1
1.3 Combustion S-curve............................................................................................ 3
1.4 Activation Energy Asymptotics.......................................................................... 4
1.5 Project Objectives ............................................................................................... 7
Chapter 2: Formulation ................................................................................................. 8
2.1 Introduction......................................................................................................... 8
2.2 Assumptions........................................................................................................ 9
2.2 Assumptions...................................................................................................... 10
2.3 Slot Flowfield.................................................................................................... 10
2.4 Conservation Equations and Boundary Conditions in the Jet........................... 13
2.5 Coordinate Transformation and the Solution of Momentum Equation ............ 14
2.6 Nondimensionalizing the Energy and Species Equations................................. 19
2.7 Frozen Solution................................................................................................. 22
2.8 Summary of Formulation.................................................................................. 24
Chapter 3: Ignition Analysis ....................................................................................... 25
3.1 Description of Analysis..................................................................................... 25
3.2 Ignition Analysis I : T∞ > T0 , LeF < 1 ........................................................... 25
3.3 Ignition Analysis II : T∞ > T0 , LeF ≈ 1........................................................... 34
3.4 Ignition Analysis III : T0 > T∞ , x
≈ 1 , Thicker Reaction Region .................... 38
3.5 Analysis Scenarios ............................................................................................ 43
Chapter 4: Results and Discussions ............................................................................ 47
4.1 Recapitulation ................................................................................................... 47
4.2 Cool jet flowing into a hot ambient (T∞ > T0) ................................................. 47
4.2.1 Ignition states that separate the ignitable and non-flammable regions .... 51
4.3 Hot jet flowing into a cool ambient (T0 > T∞) ................................................. 54
4.3.1 Ignition states that separate the ignitable and non-flammable regions .... 60
Chapter 5: Conclusions ............................................................................................... 61
Chapter 5: Conclusions ............................................................................................... 62
5.1 Conclusions....................................................................................................... 62
Bibliography ............................................................................................................... 64

vi
 List of Tables

Table 1 Values of α and specific heat of combustion (qF) from [4]. 49

vii
 List of Figures

Fig. 1.1 Combustion S-Curve reproduced from [18] 4

Fig. 2.1 Schematic of slot and hydrogen leak for the hot ambient case 9

Fig. 2.2 Schematic of slot and hydrogen leak for the hot jet case 9

Fig. 3.1 Schematic of inner and outer regions for the hot ambient case 43

Fig. 3.2 Schematic of inner and outer regions for the hot jet case 45

Fig. 4.1
for varying α (which is changed by variations in

aT versus Da
Da
T˜0 or Y˜F,0 ), with LeF = 0.6 and constant T
∞ = 0.0123,

hot ambient. 48

Fig. 4.2 Rescaled plot of aˆT versus Da (rescaled D~a values due to
D̂a
changes in T˜∞ ) for varying α (which is changed by variations

in T˜∞ ), with LeF = 0.6 , T˜0 = 0.00226, E˜ = 1.79,

Y
F ,0 = 1, and α =0.01 as reference value, hot ambient. 50

Fig. 4.3 aT versus Da (rescaled D~a values due to changes in LeF)

D̂a
for varying LeF , with α = 0.02, T˜0 = 0.00226, and

E˜ = 1.79 , hot ambient. 51

Fig. 4.4 α (which is changed by variations in T˜0 or Y˜F,0 ) versus Da

D̂a
(rescaled D~a values due to changes in LeF) for varying LeF,

with LeF = 1 as reference and constant T
∞ = 0.0123,

hot ambient. 52

Fig. 4.5 α (which is changed by variations in T˜∞ ) versus Da (rescaled

D̂a
~
Da values due to changes in T˜∞ ) for varying LeF, and T˜0

viii
 = 0.00226, hot ambient. 53

Fig. 4.6 α (which is changed by variations in T˜∞ ) versus Da (rescaled D~a

D̂a
values due to changes in T˜∞ ) for varying no at LeF = 1 and T˜0 =

0.00226, hot ambient. 53

Fig. 4.7 θ max versus
for varying β (which is changed by variations

Da
Da
in T˜∞ or Y˜O,∞ ) at Leo = 1 and constant T˜0 = 0.0388, hot jet. 54

Fig. 4.8
for varying β (which is changed by variations in

X versus Da
Da
T˜∞ or Y˜O,∞ ) , with Leo = 1 and constant T˜0 =0.0388, hot jet. 55

Fig. 4.9 Rescaled plot of θˆmax versus Da (rescaled D~a values due to
D̂a
changes in T˜0 ) for varying β (which is changed by variations

in T˜0 ), with Leo = 1, T
∞ =0.0358, E˜ = 1.79, Y˜O,∞ = 0.029,

and β = 0.1 as reference value , hot jet. 57

Fig. 4.10 X versus Da (rescaled D~a values due to changes in T˜0 )

D̂a
for varying β (which is changed by variations in T˜0 ), with

Leo = 1, T
∞ =0.0358, E˜ = 1.79, Y˜O,∞ = 0.029, and β = 0.1

as reference value , hot jet. 58

Fig. 4.11 θ max versus
for varying

Da Leo , with β = 0.3 ,
Da
and constant T˜0 = 0.0388, hot jet. 59

Fig. 4.12
for varying Leo , with β = 0.3 ,

X versus Da
Da
and constant T˜0 = 0.0388, hot jet. 60

Fig. 4.13 β (which is changed by variations in T˜∞ or Y˜O,∞ ) versus Da

Da
for varying no, and constant T˜0 = 0.0388, hot jet. 61

ix
Fig. 4.14 β (which is changed by variations in T˜0 ) versus Da (rescaled
D̂a
~
Da values due to changes in T˜0 )for varying no, with Leo = 1,

T
∞ =0.0358, E˜ = 1.79, Y˜O,∞ = 0.029, and β = 0.1 as reference

value, hot jet. 61

x
 Nomenclature

aT constant representing the temperature increase through reaction

B pre-exponential factor

cj molar concentration of species j

cp specific heat at constant pressure

cv specific heat at constant volume

Da Damköhler number

D˜ a reduced Damköhler number; flow time divided by chemical time

DaE Damköhler number at extinction

DaI Damköhler number at ignition

Dj mass diffusion coefficient of species j

E extinction state (when used in Fig.1)

E activation temperature

F fuel

f nondimensional stream function

h half width of the slot

I ignition state

Lej local Lewis number of species j; thermal diffusivity of mixture divided by

mass diffusivity of species j into mixture

F perturbation of the fuel Lewis number from unity

xi
M linear momentum per unit depth of the slot

O oxidizer

p pressure

pa atmospheric pressure

P combustion products

Pr local Prandtl number; mixture viscosity divided by mixture thermal diffusivity

qF heat of combustion per unit mass of the fuel

R ideal gas constant (R°/W*)

R° universal gas constant

Scj local Schmidt number of species j; mixture viscosity divided by mass

diffusivity of species j into mixture

T temperature

u flow velocity in the x (streamwise) direction

v flow velocity in the y (transverse) direction

w thickness of the wall

W* average molecular weight

Wj molecular weight of species j

x streamwise spatial coordinate

x0 distance between the virtual origin of the jet to the exit of the slot

y transverse spatial coordinate (y = 0 on the plane of symmetry)

Yj mass fraction of species j

xii
 Greek Symbols

α parameter defined as α = (T˜∞ − T˜0 ) / Y˜F,0

β parameter defined as β = (T˜0 − T˜∞ ) / Y˜O,∞

ε small parameter, defined as ε = T˜∞2 / E˜ , used for asymptotic expansion

φF perturbation of fuel concentration in the inner, reaction region (because of the

weak reaction)

ΦF perturbation of fuel concentration in the outer, chemically inert region

(because of the weak reaction)

φO perturbation of oxidizer concentration in the inner, reaction region

ΦO perturbation of oxidizer concentration in the outer, chemically inert region

(because of the weak reaction)

γ specific heat ratio (cp/cv)

η similarity variable

λ thermal conductivity

µ dynamic viscosity

νj stoichiometric coefficient of species j

θ perturbation of temperature in the inner, reaction region (because of the weak

reaction)

Θ perturbation of temperature in the outer, chemically inert region (because of

the weak reaction)

xiii
ρ gas density

σ parameter defined as σ = Y˜F,0 / x˜ 1/ 3

ω reaction rate function

ξ spatial coordinate defined as ξ = sech 2 η

ψ streamfunction

ζ stretched spatial coordinate in the inner, reaction region (inner variable)

ζ rescaled inner spatial coordinate defined as ζ = σζ

Subscripts

0 value of variables at the exit of the slot

f frozen solution

F fuel

i value of variables in the fuel supply

O oxidizer

P combustion products

T temperature

∞ value of variables at the ambient

Superscripts

nj reaction order of species j

~ nondimensional quantity

∧ rescaled nondimensional quantity

xiv
Chapter 1: Introduction

1.1 Motivation for Project

Concerns about the emissions of greenhouse gases have led to extensive

consideration of hydrogen as a major fuel carrier. Hydrogen presents several unusual

fire hazards, including high leak propensity, ease of ignition, and invisible flames.

Heated air jets flowing into hydrogen ignite spontaneously at an air temperature of

943 K [1]. This is cooler than for most other fuels [2,3], including gasoline and

methane, and is not much higher than the autoignition temperature of stoichiometric

hydrogen/air mixtures, 858 K [4]. Occasional unintended hydrogen leaks will be

unavoidable, and some may involve heated hydrogen and/or air. Thus an improved

understanding of limits of spontaneous ignition of hydrogen jets is sought here, with

the aid of activation-energy asymptotics.

1.2 Literature Review

Asymptotic flame theories can provide valuable insights into combustion

reactions [5-7]. Quantitative and predictive derivations can be made using the concept

of distinguished limits in activation energy asymptotics. Based largely on the concept

of Zel’dovich number, asymptotic analysis enables derivation and establishing of

temperature effects on reaction rates despite the narrowness of the reaction zone

relative to the preheat zone of the laminar flame structure.

Im et al. [8,9] analyzed thermal ignition in supersonic hydrogen/air mixing layers

and obtained ignition characteristics over a wide range of conditions. The findings

were however based on reduced mechanisms and supersonic flows, which are more

1
applicable for scramjets. An investigation of different combustion regimes by

Damköhler-number and activation-energy asymptotics in a stagnant mixing layer,

based on an eight-step reduced mechanisms was performed by Lee and Chung [10].

Helenbrook and Law investigated the ignition of hydrogen/air mixing layer with

reduced reaction mechanisms which they developed [11,12]. However, no single-step

reaction mechanism was developed. Compared to a single-step, overall, irreversible

reaction with second order Arrhenius kinetics and a high activation energy (which

was used in the research), the reduced mechanism has different assumptions in length

scales in determining the reaction rates, which will yield different results with

emphasis on temperature dependence for thermal runaway. This is because the

emphasis for reduced mechanisms is on the role of chemical kinetic mechanisms,

involving chain-branching and termination reactions, in effecting a non-linear

feedback in the concentrations of certain radicals and consequently, thermal runaway.

Zheng and Law [13] identified ignition limits of premixed hydrogen-air flames

where ignition was by heated counterflow. Ignition limits of non-premixed hydrogen-

air flames from jets will be different because of the non-premixed combustion mode

and the absence of strain due to counterflow heating. Toro et al. [14] examined in

detail the structure of laminar hydrogen jet flames both experimentally and

numerically. For completeness, analytical results should be obtained to enable

comparison with experimental and numerical results, under the same conditions and

scenarios. Chaos et al. [15] examined Lewis-number effects in unsteady laminar

hydrogen jet flames, which will have different effects compared to a steady laminar

hydrogen jet flame. Liu and Pei [16] examined autoignition and explosion limits of

2
hydrogen-oxygen mixtures in homogenous systems, which involved reduced

mechanisms. Dryer et al. [17] examined spontaneous ignition of pressurized releases

of hydrogen and natural gas into air, which investigated multi-dimensional transient

flows involving shock formation, reflection and interactions which resulted in the

transition to turbulent jet diffusive combustion. This is a different aspect of risk

associated with rapid failures of compressed storages, as compared to the scenario

being considered here, involving small leaks/cracks that are undetected, and that

ignite spontaneously when the limits are reached.

1.3 Combustion S-curve

In understanding and analyzing the limits of spontaneous ignition of hydrogen

jets, it is important to appreciate the fundamentals of flame ignition and extinction,

which can be characterized and explained by the famous combustion S-curve

[5,7,18]. The S-curve, as shown in Fig. 1.1, comprises 3 branches, the lower, middle

and upper branches. The y-axis represents the reaction temperature, and the x-axis

represents the Damköhler-number (Da). Starting with the left end of the lower

branch, at Da = 0, we have the chemically frozen flow limit. By increasing Da along

this branch, every possible weakly reacting state that the system can have was

covered. DaI represents the ignition Da, at which weak reactions transition to

vigorous burning with a sudden jump to the upper branch. Anywhere beyond this Da

will result in spontaneous combustion. We define this point as the ignition state.

Conversely, as we decrease the Da for an intense burning flame on the upper branch

to the point DaE, there will be another jump of the temperature down to the lower

branch. This point is defined as the extinction Da.

3
 Physically, the existence of turning points implies that there exist states for which

the chemical reaction rate cannot balance the heat transport rate in steady state. Thus

for the lower branch, beyond DaI, the chemical heat is generated so fast in the

reaction zone that it cannot be transported away in steady manner. The middle branch

is never observed because it has a negative slope which implies that reaction

temperature decreases as Da increases, which is physically unrealistic.

Fig. 1.1: Combustion S-Curve reproduced from [18].

1.4 Activation Energy Asymptotics

There are a number of possible approaches to modeling the influences of finite-

rate chemistry on diffusion flames. Known rates of elementary reaction steps may be

employed in the full set of conservation equations, with solutions sought by

4
numerical integration and computational fluid dynamics (CFD). While CFD is like an

experiment, with only one condition considered at a time, other modeling such as

activation-energy asymptotics (AEA) can quickly identify trends and give additional

physical insights. Complexities of diffusion flame problems motivates searches for

simplifications of the chemical kinetics [19]. With simplified chemical kinetics,

perturbation methods [20] are attractive for improving understanding and also for

seeking quantitative comparison with experiment results. Two types of perturbation

approaches have been developed, Damköhler-number asymptotics and AEA. In the

former, the ratio of a flow time to reaction time, one of the dimensionless groups

introduced by Damköhler [21] is treated as a large parameter. And in the latter, the

ratio of the activation energy to the thermal energy, emphasized as important by

Frank-Kamenetskii [22], is taken to be large. Damköhler-number asymptotics can

provide estimates of reaction zone broadening in near-equilibrium situations [23,24],

and also affords possibilities of investigating other regimes [25].

Analyses of phenomena such as sharp ignition and extinction events cannot be

performed on the basis of Damköhler-number asymptotics, but they can be treated by

activation-energy asymptotics. Moreover, activation-energy asymptotics may lead to

results valid for all Damköhler numbers, and therefore results of Damköhler-number

asymptotics may be extracted from those of activation-energy asymptotics.

Activation-energy asymptotics is the more general of the two types of perturbation

approaches. As compared to CFD, AEA was selected for our study as it provides the

complete physics of the problem instead of just providing exact solutions for

5
individual points, and provides a good representation of trends and limit behavior,

which is sought here.

In the research, the AEA approach which was adopted to derive the exact

solutions to the Navier-Stokes Equations can be summarized as:

1) Solutions of the flowfield for the scenario of hydrogen jet emanating from a

rectangular slot were derived. A similarity solution of the non-reacting flowfield

is obtained, and then used in the energy and species conservation equations.

Coordinate transformation is necessary to investigate the effects of perturbations

because the reaction zone of concern is a very small thin one.

2) Frozen solutions were then obtained. These represent the solutions in very low Da

regime whereby there are no reactions.

3) Outer solutions were derived. These solutions deviate from the frozen solution by

a small amount due to perturbation, and are present in outer regions where there is

no reaction due to the low temperature. Before ignition, there are only weak

reactions.

4) Inner solutions were derived for the reaction zone where weak chemical reactions

occur.

5) Matching of inner and outer solutions is then performed to determine the

conditions whereby ignition can occur. Ignition can occur when the heat

generated from the chemical reaction is sufficient to overcome heat losses. The

results are presented in terms of Da, To, T∞ , LeF, and Leo.

6
1.5 Project Objectives

The present analysis considers the spontaneous ignition of a jet of hydrogen or

other gaseous fuel leaking through a slot into air. The slot is taken to be straight and

long, yielding a two-dimensional flow field. The ignition analysis identifies limits of

spontaneous ignition.

The objectives of this work are to:

1) develop a model of spontaneous ignition for two cases: a cool fuel jet flowing into

heated air and a heated fuel jet flowing into cool air,

2) identify limits of spontaneous ignition as functions of slot width, flow rate, fuel

Lewis number, and temperatures of the fuel jet and the ambient air, and

3) identify the location of ignition.

7
Chapter 2: Formulation

2.1 Introduction

The problem of interest is a steady, isobaric laminar jet of fuel (e.g. hydrogen) at

temperature T0 issuing from a rectangular slot into an oxidizing environment (e.g.

atmospheric air) at a temperature of T∞ , as shown schematically in Figs. 2.1 and 2.2

for the two scenarios. Spontaneous ignition occurs when either T0 or T∞ is

sufficiently high that the weak reaction between the fuel and the oxidizer transitions

to a vigorous burning flame. This study analyzes the ignition state as a function of

various physical properties including Lewis number, T0, T∞ , the flow velocity at the

slot exit, u0, and the width of the slot. The slot is considered sufficiently long that end

effects are negligible. The reaction chemistry is simulated by a single-step, overall,

irreversible reaction with second order Arrhenius kinetics and a high activation

energy.

The formulation that follows is an exact solution of the conservation of mass,

momentum, energy and species. The key assumptions are boundary layer behavior

∂ ∂ and single-step chemistry. A similarity solution of the non-reacting

(  )
∂ y ∂ x

flowfield is obtained. This is then used in the conservation of energy and species

equations.

8
 Outer
Stoichiometric Region
contour

Reaction
zone

T∞ Ignition
x
Air
y

uo
Fuel
To
Fig. 2.1: Schematic of slot and
hydrogen leak for hot ambient case

Stoichiometric
contour
T∞ Reaction
zone x Ignition
Air
y

uo
Fuel
To
Fig. 2.2: Schematic of slot and
hydrogen leak for hot jet case

9
2.2 Assumptions

The assumptions are as follow:

(1) steady, 2D flow, negligible body force

(2) reaction follows Arrhenius kinetics

(3) high activation energy reaction

(4) isobaric flow

(5) symmetric with respect to the plane of symmetry

(6) cp, ρ λ, ρ µ, and ρ2 Di do not vary with position

These assumptions are reasonable in light of the mathematical simplicity they

introduce. Similar assumptions are commonly invoked in this type of analysis.

[11,12]

2.3 Slot Flowfield

The reaction is

νF F + νO O → νP P (2.1)

The flowfield is described by conservation of mass,

∂ ( ρ u ) ∂ (ρ v )
∂x + ∂y =0 (2.2)

conservation of momentum, x-direction

∂u ∂ u ∂   ∂u 2 ∂ u ∂ v  ∂  ∂u ∂v 
ρ u ∂ x + ρ v ∂ y − ∂ x µ 2 ∂ x − 3  ∂ x + ∂ y    − ∂ y  µ  ∂ y + ∂ x   = − ∂ x
∂p
(2.3)
       

and conservation of momentum, y-direction

∂v ∂ v ∂  ∂ v ∂u  ∂   ∂ v 2 ∂ u ∂ v  
ρ u ∂ x + ρ v ∂ y − ∂ x  µ  ∂ x + ∂ y   − ∂y µ 2 ∂ y − 3  ∂ x + ∂ y    = − ∂ y
∂p
(2.4)
       

Within the slot, the only velocity component is u along the x direction : v = 0

10
∂( ρ u)
∂x =0 Therefore, ρ u is independent of xo (2.5)

Conservation of momentum becomes

∂u 4 ∂ ∂u
ρ u ∂ x − 3 ∂ x  µ ∂ x  − ∂ y µ ∂ y  = − ∂ x
∂ ∂u ∂p
(2.6)
   

∂  ∂u 2 ∂  ∂u ∂p
    (2.7)
∂ x  µ ∂ y  − 3 ∂ y  µ ∂ x = ∂ y

y=0 : ∂ u/ ∂ x = 0

Case I Uniform flow

A uniform flow is possible when the effect of viscosity is negligible (µ ≈ 0).

At any x, u is uniform for all y

∂p
∂y = 0 leading to p = p( x )

du dp
ρu d x = − d x leading to ρu 2 = c − p (Bernoulli Equation)

In the tank : u = 0 , p = pi , therefore c = pi

At the exit : u = u0 , p = p0 , ρ = ρ0 therefore, ρ0 u02 = pi − p0 or

u0 = ( pi − p0 )/ ρ 0 (2.8)

h 2 h h
M = ∫−h ρ0 u0 d y = ∫−h (pi − p0 )d y = (pi − p0 ) ∫− h d y = 2 h( pi − p0 ) (2.9)

For this case, there is no friction. The flow is supported by the expansion caused by

the pressure reduction so ρ cannot be considered constant. If the pressure difference

is large, the flow is choked (Ma = 1 at the exit).

If there is no heat addition or generation during the expansion, the gas

temperature density and velocity at the exit (T0, ρ0 and u0) are (isentropic

compressible flow)

11
T0 / Ti = 2 /( γ + 1) , γ = cp/cv (2.10)

p0 / pi = [ 2/( γ +1 )]γ /( γ −1) (2.11)

ρ0 / ρ i = [2/( γ + 1)]1/( γ − 1) (2.12)

An expansion wave exists if pa < p0.

Case II Fully developed flow in a channel : u = u ( y )

The flow velocity is relatively low so the flow can be considered incompressible and

isothermal

∂p
∂y = 0 leading to p = p( x ) (µ = µi = µ0 = constant when ρ = constant)

d  du  d p
  (2.13)
d y µ0 d y  = d x

since lefthand side is a function of y only and righthand side is a function of x only,

dp
= constant.
dx

d p/ d x = −( pi − p0 )/ w w is the thickness of the wall, p0 = pa (2.14)

du 1 dp pi − p0 p −p
d y = µ0 d x y + c1 = − µ0 w y + c1 and u = − 2iµ w0 y 2 + c1 y + c2 (2.15)
0

du
At y = 0 (centerline) : d y = 0

p −p
therefore c1 = 0 leading to u = − 2iµ w0 y 2 + c2 (2.16)
0

At y = h (channel wall) : u = 0

p −p p −p
therefore c2 = 2iµ w0 h2 leading to u = 2iµ w0 (h2 − y 2 ) (2.17)
0 0

12
 h
⌠ 2
p −p 2 h
M = ∫−h ρ0 u2 d y =  ρ0  i (h − y 2 )  d y = ρ0  i 0  ∫−h (h4 − 2 h2 y 2 + y 4 )d y
h p − p0 2
(2.18)
⌡− h  2 µ 0 w   2 µ0 w 

2 2 2
p − p  2 1 h p − p  2 1 4 p − p 
= ρ0  i 0  ( h4 y− h2 y 3 + y5 ) = ρ0  i 0  [h4 ( 2h)− h2 (2h3 )+ (2h5 )] = ρ0 h 5  i 0 
 2 µ0 w  3 5 −h  2 µ0 w  3 5 15  µ0 w 

p −p
At y = 0 : u = u0 ∴ u0 = 2iµ w0 h2
0

2.4 Conservation Equations and Boundary Conditions in the Jet

(a) Conservation equations

p = ρR T or ρ = p /( R T ) ; R = ideal gas constant , p = constant (isobaric

flow)

∂ ( ρ u ) ∂ (ρ v )
∂x + ∂y =0 (2.19)

∂p
∂y = 0 leading to p = p( x )

∂u ∂u ∂
ρ u ∂ x + ρ v ∂ y − ∂ y  µ ∂ y  = − ∂ x = − d x = 0
∂u ∂p dp
(At any x, as y → ± ∞ , u → 0 therefore
 

dp
d x =0 )

Conservation of energy is,

∂T ∂T ∂ ∂T
ρ u cp ∂ x + ρ v cp ∂ y − ∂ y λ ∂ y  = ν FW F qFω (2.20)
 

Conservation of fuel is,

∂Y ∂Y ∂Y
ρ u ∂ xF + ρ v ∂ yF − ∂ y ρ DF ∂ yF  = − νF WF ω
∂
(2.21)
 

13
Conservation of oxidizer is,

∂Y ∂Y ∂Y
ρ u ∂ xO + ρ v ∂ yO − ∂ y  ρ DO ∂ yO  = − νO WO ω
∂
(2.22)
 

Conservation of oxidizer is,

ω = BcFnF cOnO T nT exp ( − E / T ) = B ( ρ YF / WF )nF ( ρ YO / WO )nO T nT exp ( − E / T )

n nO n nO
= ( B/ WF F WO ) ρnF +n O YF F YO T nT exp ( − E / T ) (2.23)

(b) Boundary and interface conditions

Let x = 0 be the virtual origin of the jet and x = x 0 be the exit of the jet.

−
x = x0 , −h< y < h : T = T0 , YF = Y F,0 , YO = 0 , u = u0 , v=0

y = 0 , x = x 0+ : T = T0 , YF = Y F,0 , YO = 0 , u = u0 , y = 0 , x > x0 :

∂ T / ∂ y = ∂ Y F/ ∂ y = ∂ YO / ∂ y = ∂ u/ ∂ y = 0 , v=0

y→ ∞ : T → T∞ , YF → 0 , YO → YO,∞ , u→0

2.5 Coordinate Transformation and the Solution of Momentum Equation

A stream function ψ is defined such that the continuity equation is satisfied :

ρu ∂ψ ρv ∂ψ
ρ∞ u0 = ∂ y , ρ∞ u0 = − ∂ x (2.24)

A new coordinate system is defined using similarity variables [26,27].

α1 ⌠ y ρ
η= d y′ , ψ = α 2 x 1/ 3 f (η ) , x˜ = x / x0 (2.25)
x 2/ 3 ⌡0 ρ∞

(α1 and α2 are constants that are defined later to simplify the expression)

14
 It is assumed that a similarity solution exists so that f is a function of η only.

Coordinate transformation from (x, y) to (x, η) yields

∂ ∂ ∂ x ∂ ∂η ∂  2 α ⌠
y
ρ α ∂  ⌠
y
ρ  ∂
= + = +− 1  d y ′ + 2 1/ 3   d y′  
∂ x ∂ x ∂ x ∂ η ∂ x ∂ x  3 x 5 / 3 
⌡0 ρ∞ x ∂ x  
⌡0 ρ∞ ∂η

∂ 2η ∂ α1  ∂ ⌠ ρ ∂
y

= − + 2/3  d y′ (2.26)

∂ x 3 x ∂ η x  ∂ x ⌡0 ρ∞ ∂η


∂ ∂ ∂ x ∂ ∂ η ∂  α1 ∂  ⌠ y ρ  α1 ρ ∂
∂ y ∂ x ∂ y + ∂ η ∂ y = ∂ η  x 2/ 3 ∂ y  ⌡0 ρ∞ d y ′   = x 2/ 3 ρ∞ ∂ η
= (2.27)

ρu ∂ψ α1 ρ ∂ α α ρ df
ρ∞ u0 = ∂ y = x 2/ 3 ρ∞ ∂ η α 2 x
1/ 3
[ x ]
f (η ) = 11/ 32 ρ d η
∞
(2.28)

u x 1/ 3 d f u df
We define α1α 2 = x01/ 3 or α2 = x 01/ 3 / α1 such that u = 01/ 3 d η or u = ˜ 1/03 d η
0 x x

ρv ∂ψ ∂  x01/ 3 1/ 3  2 ∂  x01/ 3 1/ 3 
=− =−  x f ( η )  + η  x f (η ) 
ρ∞ u0 ∂x ∂ x  α1  3 x ∂ η  α1 
 ∂ ⌠y ρ  ∂  x01/ 3 1/ 3 
−α1 x −2/ 3   ′ x f (η ) 
 ∂ η  α1
d y
 ∂ x ⌡0 ρ∞ 
 

x01/ 3  d f  x01/ 3  ∂ ⌠ ρ d f
y

=− 2/3 
f − 2 η  − 
1/ 3 
d y′  (2.29)
3 α1 x  
d η  x  ∂ x ⌡0 ρ∞  dη


∂ u ∂  x 01/ 3 u0 d f  2 η ∂  x 01/ 3 u0 d f  α1  ∂ ⌠ y ρ  ∂  x 01/ 3 u0 d f 

=   −   + 
∂ x ∂ x  x 1/ 3 dη  3 x ∂ η  x 1/ 3 d η x 2/ 3  ∂ x ⌡0 ρ∞ d y ′  ∂ η  1/ 3 d η
  x 

x 01/ 3 u0  d f d 2 f  α1 x 01/ 3 u0  ∂ ⌠y ρ  d2 f
=−  + 2 η  +  d y 
′ (2.30)
3 x 4/ 3  d η d η2   ∂ x ⌡0 ρ ∞
x 2
 dη

∂u α 1 ρ ∂  x 01/ 3 u0 d f  α 1 x 01/ 3 u0 ρ d2 f
∂ y x 2/ 3 ρ∞ ∂ η  x 1/ 3 dη  =
= x ρ∞ dη2 (2.31)

∂  ∂ u  α1 ρ ∂  α1 x01/ 3 u0 ρ µ d 2 f  α12 x01/ 3 u0 ρ 2 µ d 3 f

µ =  =
∂ y  ∂ y  x 2/ 3 ρ ∞ ∂ η  x ρ∞ d η 2  x5/ 3 ρ∞ 2 d η 3
α x u ρ ρ ∞ µ∞ d f α1 x01/ 3 u0
2 1/ 3 3 2
ρ µ∞ d 3 f
= 1 50/ 3 0 = (2.32)
x ρ∞ 2 d η 3 x5 / 3 ρ∞ d η 3

15
(It is assumed that ρ µ = ρ∞ µ∞ =constant)

Momentum equation is given by,

∂u ∂u ∂
ρ u ∂ x + ρ v ∂ y − ∂ y  µ ∂ y  = 0
∂u
(2.33)
 

ρ x 01/ 3 u0 d f  x01/ 3 u0  d f d2 f  α 1 x01/ 3 u0  ∂ ⌠ y ρ  d2 f 

 −  + 2 η  +  d y ′  
x 1/ 3 d η  3 x 4/ 3  dη dη2  x  ∂ x ⌡0 ρ ∞  dη 2 

 ρ u x 1/ 3  d f  ρ ∞ u0 x01/ 3  ∂ ⌠ ρ
y
 d f  α1 x01/ 3 u0 ρ d 2 f
+  − ∞ 0 20/ 3  f − 2η   d η 

 − d y ′
x1/ 3   ρ∞ d η 2
 3α1 x  dη   ∂ x ⌡0 ρ∞   x
α 2 x 1/ 3 u ρ µ∞ d 3 f − ρ u02 x02 / 3   d f  2 + f d2 f + α 2 3 µ∞ d3 f 
− 1 5/0 3 0 
3x 5/ 3   d η d η2 1 =0 (2.34)
x ρ dη3
∞
ρ∞ u0 x01/ 3 d η3 

3 µ∞ 1 ρ∞ u0 x 01/ 3
We define α12 =
2
such that it leads to α1 = 6 µ∞ and
ρ ∞ u0 x 01/ 3

3
1d f d2 f  d f  2
+ f +  =0 (2.35)
2 d η3 d η2  d η

α1 ⌠ y ρ ρ∞ u0 x 01/ 3 1 ⌠ y ρ ρ u y
ρ
d y ′ = 6 µ∞ x0 2/ 3 ⌠
1
η= ⌡ ρ d y ′ = µ ⌡ ρ ⌡ d y′ (2.36)
x 2/ 3 0 ∞ 6 ∞ x 2/ 3 0 ∞ ∞ 0 x˜ 0 ρ∞

x 1/ 3 6 µ∞ 6µ x
ψ = α 2 x 1/ 3 f (η ) = 0α x 1/ 3 f ( η ) = x 01/ 3 x 1/ 3 f ( η ) = ρ ∞u 0 x˜ 1/ 3 f (η ) (2.37)
1 ρ∞ u0 x 01/ 3 ∞ 0

α  ∂ ⌠ ρ  ∂ y
∂ ∂ 2η ∂ ∂ 2η ∂
= − + 21/ 3  d y′  = −
∂ x ∂ x 3 x ∂ η x  ∂ x ⌡0 ρ∞
  ∂ η ∂ ( x0 x
) 3 x0 x
∂ η

ρ ∞ u0 x01/ 3 1  ∂ ⌠ ρ y
 ∂
2/3 2/3 
+  d y′ 
6 µ∞  
x0 x
 ∂ x ⌡0 ρ ∞ 
∂η

1 ∂ 1 2η ∂ ρ∞ u0 1  ∂ ⌠y ρ  ∂
6 µ∞ x 0 x˜ 2/ 3  ∂ x ⌡0 ρ∞
= − + d y′  (2.38)
x 0 ∂ x˜ x 0 3 x˜ ∂η  ∂η

∂ α ρ ∂ ρ u x 1/ 3 1 ρ ∂ ρ ∞ u0 1 ρ ∂
= 2/1 3 = ∞ 0 0 = (2.39)
∂y x ρ∞ ∂ η 6 µ∞ x0 2/ 3 x˜ 2/ 3 ρ ∞ ∂η 6 µ∞ x 0 x˜ 2/ 3 ρ∞ ∂η

16
Boundary conditions

y = 0 , x = x 0+ : u = u0 ⇒ η=0 : d f / d η = 1 (f is independent of x)

y = 0 , x > x0 : ∂ u/ ∂ y = v = 0 ⇒ η = 0 , x > x0 : f = d2 f / d η2 = 0

y→ ∞ : u→0 ⇒ η →∞ : d f / dη →0

Solution in the jet (x > x0 or x˜ > 1 )

d 1 d f d f
3
1d f d2 f  d f  2 1 d3 f d  df
2
+  = f =  d η
0= +f + +f therefore (2.40)
2 dη 3 dη  d η
2 2 d η d η  dη  dη  2 d η
3 2

2
1d f df
+f =c
2 d η2 dη

2
1d f df
η=0 : f = d2 f / d η2 = 0 therefore c = 0 leading to 2 2 + f d η = 0
dη

2 d f 1 d2 f 1 d f 2 1 d  d f
1d f
  df
0= +f = + = + f 2 therefore 2
dη + f = c (2.41)
2 dη 2 dη 2 d η2 2 d η 2 dη  d η 

d f d( aF) dz dF 2 dF
Let z = a η , f=aF(z) , c = a2 then d η = dz d η = a dz a= a dz

d F ( aF)2 dF 2 dF dF
a2 dz + 2 = a2 or dz + F = 1 ⇒ dz = 1− F
2 or =d z
1− F2

f
tanh −1 F = z + c or F = tanh (z + c) or a = tanh ( aη + c) or f = atanh (aη + c)

η=0 : f=0 , d f / dη = 1

tanh c = 0 or c=0 ⇒ f = a tanh ( a η ) and d f/d η = a2 sech2 ( a η )

a2 sech 2 (0 ) = 1 therefore a2 = 1 or a=1

Thus : f = tanh η , d f/d η = sech2 η ,

u = (u0 / x˜ 1/ 3 )( d f / d η ) = u0 (sech 2 η )/ x˜ 1/ 3 (2.42)

17
  x 01/ 3  d f  x 01/ 3  ∂ ⌠y ρ  df 
ρ v = ρ ∞ u0  −  f − 2 η  − 1/ 3  d y′   (2.43)
 3 α1 x 2/ 3  dη  x  ∂ x ⌡ ρ
0 ∞  dη 

 6 µ∞ x 01/ 3  d f 1  ∂ ⌠y ρ  df
= − ρ∞ u0  f − 2η  +  ∂ x ⌡ ρ d y′ d η  (2.44)
 ρ∞ u0 x 0
1/ 3 3 x0 2 / 3 x˜ 2 / 3  d η ˜
x 1/ 3  0 ∞  

 6 µ∞ (tanh η ) − 2η (sech 2 η ) sech 2 η  ∂ ⌠ y ρ 

= − ρ∞ u0  + 1/ 3  d y′  (2.45)
ρ u
 ∞ 0 0 x 3 x˜ 2/ 3 x˜  ∂ x ⌡ ρ
0 ∞ 

Determination of x0 by momentum conservation

∞ h 2
∫−∞ ρ u d y = ∫− h ρ0 u0 d y = M
2

∞
⌠
 u0 
2
 6 µ ∞ x0 2 / 3 ρ ∞ 
M = ∫− ∞ ρ u 2 d y = 
∞
ρ sech 2 η   x
dη 

1/ 3
 

⌡− ∞


x  ρ ∞ u0 ρ 
∞
= u0 6 x0 u0 ρ ∞ µ∞ ∫− ∞ ( sech 4 η ) d η (2.46)

 (sech 2 η )(tanh η ) ∞ 
2 ∞
= u0 6 x0 u0 ρ ∞ µ∞  3 + ∫
3 −∞ (sech 2 η )dη 
 −∞ 

 2 ∞  4
= u0 6 x0 u0 ρ ∞ µ∞  0 + tanh η −∞  = u0 6 x0 u0 ρ ∞ µ∞
 3  3
(2.47)

32 3 2 3 M2
Therefore 3 u0 x 0 ρ ∞ µ∞ = M or x0 = 3 (2.48)
32 u0 ρ∞ µ∞

In summary, the flowfield solution is :

ρ u y
ρ 6 µ∞ x 0 1/ 3 3 M2
η = 6 µ∞ x0 2 / 3 ⌠
1 ˜
⌡0 ρ∞ d y ′ (2.36), ψ = ρ∞ u0 x f (η ) (2.37), x 0 = 3 (2.48)
∞ 0 x˜ 32u0 ρ∞ µ∞

f = tanh η , d f / d η = sech 2 η , u = u0 (sech 2 η )/ x˜ 1/ 3

(2.42)

∂ 1 ∂ 1 2η ∂ ρ∞ u0 1  ∂ ⌠ y ρ  ∂
6 µ∞ x 0 x˜ 2/ 3  ∂ x ⌡0 ρ∞
= − + d y′ (2.38)
∂ x x 0 ∂ x˜ x 0 3 x˜ ∂ η  ∂η

∂ ρ∞ u0 1 ρ ∂
∂ y = 6 µ∞ x 0 x˜ 2/ 3 ρ∞ ∂η (2.39)

 6µ ( tanh η ) − 2η (sech 2 η ) sech 2 η  ∂ ⌠ y ρ 

ρ v = − ρ∞ u0  ρ u ∞x + 1/ 3 
∂ x ⌡ ρ d y′  (2.45)
 ∞ 0 0 3 ˜
x 2/ 3 ˜
x  0 ∞ 

18
2.6 Nondimensionalizing the Energy and Species Equations

The following nondimensional quantities are defined as :

cp T ν W cp E
T˜ = q , Y˜F = Y F , Y˜O = ν FW F YO , E˜ = q (2.49)
F O O F

nT −n F −nO +1 nF +nO −1
6x 0 νOnO B  qF  p 
Damköhler number: Da = nO −1 nO +nF −1  cp  R
(2.50)
νF WF u0  

ρµ
Prandtl number : Pr = λ /µc = ρ λ / c = constant (2.51)
p p

µ ρµ
Schmidt number of species j : Sc j = ρ D = = constant (2.52)
j ρ2 D j

λ/ c Sc
Lewis number of species j : Lej = ρ D p = Prj = constant (2.53)
j

Nondimensionlizing the energy equation:

(1) ρ = p /( R T ) = ( p/ R )( cp / qF )/ T˜ (2.54)

∂T ∂T ∂T ν W B
(2) ρ u cp ∂ x + ρ v cp ∂ y − ∂ y λ ∂ y  = FnF F nO qF ρ nF+ nO Y Fn F YOnO T nT exp (− E/ T )
∂
W F WO

qF  ∂ T
∂ T
1 ∂  ∂ T
 
 ρ u cp + ρ v cp −  µ cp 
cp  ∂x ∂ y Pr ∂ y  ∂ y
nT
 qF 
nO
ν F WF B  ν O WO
 (2.55)
 T
 exp ( − E
/ T
)
nF + nO nF
= qF ρ YF  YO 
WF nF WO nO  ν F WF   cp 

∂ T
∂ T
1 ∂  ∂ T

ρu +ρv − µ 
∂x ∂ y Pr ∂ y  ∂ y 
ν On B O

= O −1
( qF / c p )nT ρ nF + nO Y
F nF Y
O nO T
nT exp ( − E
/ T
) (2.56)
νFn WF nO + nF −1

∂ T˜ 1 ∂ T˜ 1 2 η ∂ T˜ ρ∞ u0 1  ∂ ⌠ y ρ  ∂ T˜ ∂ T˜ ρ ∞ u0 1 ρ ∂ T˜
= −
∂ x x 0 ∂ ˜x x 0 3 x˜ ∂ η + 
6 µ∞ x0 x˜ 2/ 3  ∂ x ⌡0 ρ∞
d y ′  ∂η ; ∂y = 6 µ∞ x 0 x˜ 2/ 3 ρ∞ ∂ η


∂  ∂ T˜  ρ∞ u0 1 ρ ∂  ρ∞ u0 1 ρ ∂ T˜  ρ u0 1 ∂ 2 T˜
 µ
∂y  ∂y  = 
6 µ∞ x 0 x˜ 2 / 3 ρ ∞ ∂ η  µ 6 µ∞ x 0 x˜ 2/ 3 ρ∞ ∂ η  = 6 x 0 x˜ 4 / 3 ∂ η 2 (2.57)

( ρ µ = ρ∞ µ∞ = constant)

19
 ∂ T˜ ∂ T˜ 1 ∂  ∂ T˜ 
ρ u ∂ x + ρ v ∂ y − Pr ∂ y  µ ∂ y  =
 

sech 2 η  1 ∂ T
1 2η ∂ T
ρ∞ u0 1  ∂ ⌠ ρ  ∂ T

y

ρ u0  − + 
2/ 3 
d y′  
x
1/ 3
 x0 ∂ x
x0 3 x
∂ η

6 µ∞ x0 x
 ∂ x ⌡0 ρ ∞ ∂η 
 
1 ρ u0 1 ∂ T 2

 6 µ∞ ( tanh η ) − 2η ( sech η ) sech η ∂ ⌠ ρ
2 2  y
 
− 4/3 2
− ρ ∞ 0
u 2/3
+ 1/ 3  
d y ′

Pr 6 x0 x
∂ η  ρ ∞ u0 x0 3 x
x
 ∂ x ⌡0 ρ ∞  
 ρ∞ u0 1 ρ ∂ T

 
 6 µ∞ x0 x
2 / 3 ρ∞ ∂ η 

ρ u0 ( sech 2 η ) ∂ T
ρ u0 ( tanh η ) ∂ T
1 ρ u0 1 ∂ 2 T

= − −
x0 x
1/ 3 ∂ x
x0 3 x
4 / 3 ∂ η Pr 6 x0 x
4 / 3 ∂ η 2
ρ u0  1 ∂ 2 T
∂ T
∂ T
 (2.58)
=−  + 2 ( tanh η ) − 6 ( sech 2 η ) x

6 x0 x
4 / 3  Pr ∂ η
2
∂η ∂ x


1 ∂ T˜ ∂ T˜ ˜
2
+ 2 (tanh η ) 2 ˜ ∂T
Pr ∂ η2 ∂ η − 6( sech η ) x ∂ x˜

nT
6 x x˜ 4 / 3 νOnO B  qF 
=− 0
ρ u0 ν F O WFnO +nF −1  cp 
n −1 ρ nF+nO Y˜FnF Y˜OnO T˜ nT exp (− E˜ / T˜ )

= − Da x˜ 4/ 3 Y˜F F Y˜O O T˜ nT −nF −nO +1exp (− E˜ / T˜ )

n n
(2.59)

∂Y ∂Y ∂Y ν W B
(3) ρ u ∂ xF + ρ v ∂ yF − ∂∂y ρ DF ∂ yF  = − FnF F nO ρn F+nO YFnF YOnO T nT exp ( − E/ T )
  W F WO

∂ Y
F ∂ Y
∂  ρ DF ∂ Y
F 
ρu +ρv F −  µ 
∂x ∂ y ∂ y µ ∂y 

ν F WF B  ν O WO

nO
 qF 
nT
∂ Y
F ∂ Y
1 ∂  ∂ Y
F 
=− ρ nF + nO
YF nF
 YO   T
 exp ( − E
/ T
) ρu +ρv F − µ 
WF nF WO nO  ν F WF   cp 
∂x ∂ y ScF ∂ y  ∂ y 
ν On B O

=− nO −1 nO + nF −1
( qF / c p ) nT ρ nF + nO Y
F nF Y
O nO T
nT exp ( − E
/ T
) (2.60)
νF WF

By the same analysis as that of (2), we have (change Pr to ScF and the sign of the

righthand side)

1 ∂ Y˜F ∂ Y˜ ∂ Y˜
2
+2(tanh η ) F −6(sech2 η )x˜ F = Dax˜4/ 3 Y˜F F Y˜O O T˜ nT −nF −nO +1 exp( − E
˜ / T˜ )
n n
2 ∂η (2.61)
ScF ∂η ∂ x˜

20
 ∂Y ∂Y ∂Y ν W B
(4) ρ u ∂ xO + ρ v ∂ yO − ∂ y  ρ DO ∂ yO  = − On F O nO ρ nF+nO Y Fn F YOnO T nT exp (− E/ T )
∂
  WF WO

ν O WO  ∂ Y
O ∂ Y
∂  ρ DO ∂ Y
O  
 ρu +ρv O −  µ 
ν F WF  ∂x ∂ y ∂ y µ ∂ y  
nT
 qF 
nO
ν O WO B  ν O WO

 T
 exp ( − E
/ T
)
nF + nO nF
=− qF ρ YF  YO 
WF nF WO nO  ν F WF   cp 

∂ Y
O ∂ Y
1 ∂  ∂ Y
O 
ρu +ρv O − µ 
∂x ∂ y ScO ∂ y  ∂ y 
ν On B
O
(2.62)
=− O −1
( qF / c p ) nT ρ nF + nO Y
F nF Y
O nO T
nT exp ( − E
/ T
)
ν Fn WF nO + nF −1

Similar to (3), we have

1 ∂ Y˜O ∂ Y˜O ˜
2
+ 2(tanh η ) − 6(sech2 η ) x˜ ∂ YO =Da x˜ 4/ 3 Y˜ n F Y˜ nO T˜ nT −nF−nO+1 exp ( − E
˜ / T˜ ) (2.63)
ScO ∂ η2 ∂η ∂ ˜x F O

(5) y = 0 , x = x 0+ : T = T0 , YF = Y F,0 , YO = 0

cp T0
leading to η=0 , x˜ = 1+ : T˜ = T˜0 = q , Y˜F = Y˜ F,0 = Y F,0 , Y˜O = 0
F

y = 0 , x > x0 : ∂ T / ∂ y = ∂ Y F/ ∂ y = ∂ YO / ∂ y = 0

leading to η=0 , x˜ > 1 : ∂ T˜ / ∂η = ∂ Y˜F/ ∂η = ∂ Y˜O / ∂ η = 0

y→ ∞ : T → T∞ , YF → 0 , YO → YO,∞ , u→0

cp T∞ ν W
leading to η →∞ : T˜ → T˜∞ = q , Y˜F → 0 , Y˜O → Y˜O,∞ = ν F W F YO,∞
F O O

(6) In summary

1 ∂ T˜ ∂ T˜ ˜
2
+ 2(tanh η ) 2 ˜ ∂T ˜ 4/ 3 ˜ nF ˜ nO ˜ nT −nF−nO +1exp ( − E˜ / T˜ ) (2.59)
Pr ∂η2 ∂ η − 6( sech η ) x ∂ x˜ = − Dax YF YO T

1 ∂ Y˜F ∂ Y˜F ˜
2
+ 2(tanh η ) − 6(sech2 η ) x˜ ∂ YF = Dax˜ 4/ 3 Y˜ nF Y˜ nO T˜ nT −nF −nO +1 exp( − E
˜ / T˜ ) (2.61)
ScF ∂η2 ∂η ∂ x˜ F O

1 ∂ Y˜O ∂ Y˜O ˜
2
+ 2(tanh η ) − 6(sech2 η ) x˜ ∂ YO =Da x˜ 4/ 3 Y˜ n F Y˜ nO T˜ nT −nF−nO+1 exp ( − E
˜ / T˜ ) (2.63)
ScO ∂ η2 ∂η ∂ ˜x F O

21
η=0 , x˜ = 1 : T˜ = T˜0 , Y˜F = Y˜ F,0 , Y˜O = 0

η=0 , x˜ > 1 : ∂ T˜ / ∂η = ∂ Y˜F/ ∂η = ∂ Y˜O / ∂ η = 0

η →∞ : T˜ → T˜∞ , Y˜F → 0 , Y˜O → Y˜O,∞

2.7 Frozen Solution

In the frozen limit (designated by a subscript “f”), there is no reaction

1 ∂ T˜f ∂ T˜ f ˜
2
(1) Pr + 2(tanh η ) − 6 (sech 2 η ) x˜ ∂ Tf = 0 (2.64)
∂η 2 ∂η ∂ x˜

Since the energy and momentum equations are similar, it is expected that

T˜ f = c1 + Tˆ ( η )/ x˜ 1/ 3

∂ T˜f 1 Tˆ ∂ T˜f 1 dTˆ ∂ 2 T˜ f 1 d2 Tˆ

= − 3 4/ 3 , ∂η = x˜ 1/ 3 d η , =
∂x ˜ x˜ ∂η2 x˜ 1/ 3 dη2

1 ∂ T
f ∂ T
f ∂ T
f

2

2
+ 2 ( tanh η ) − 6 ( sech 2 η ) x

Pr ∂ η ∂η ∂ x

1  1 d T 2 ˆ
 1 d Tˆ  2
 1 Tˆ 
=  1/ 3 2 
+ 2 ( tanh η )  1/ 3  − 6 ( sech η ) x
 − 4 / 3 
Pr  x
d η   x
d η   3 x


1  1 d2 Tˆ d Tˆ  1  d2 Tˆ d[(tanh η ) Tˆ ] 
= 1/ 3  Pr 2 + 2(tanh η ) d η + 2(sech 2 η ) Tˆ  = 3  d η2 + 2Pr dη = 0 (2.65)
x˜  dη  Pr x˜ 1/  

1 d Tˆ d[(tanh η )Tˆ ] d Tˆ
2
Therefore or ˆ
Pr d η2
+2
dη
=0 dη + 2 Pr (tanh η ) T = c2

η=0 : ∂ T˜f / ∂ η = dTˆ / ∂η = 0 ∴ c2 = 0 (tanh 0 = 0) and

d Tˆ ˆ
dη = −2 Pr (tanh η ) T

d Tˆ ( sinh η ) d (cosh η )
ˆ = −2 Pr (tanh η ) dη = − 2 Pr (cosh η ) dη = − 2 Pr (cosh η ) (2.66)
T

22
Therefore  n ( Tˆ ) = −2 Pr  n (cosh η ) + c' or Tˆ = c2 (cosh η )− 2 Pr = c2 ( sech 2 Pr η )

leading to T˜ f = c1 + c2 (sech 2Pr η )/ x˜ 1/ 3

η →∞ : T˜ f → T˜∞ therefore c1 = T˜∞ and T˜ f = T˜∞ + c2 ( sech 2 Pr η )/ x˜ 1/ 3

( sech ( ∞ ) → 0 )

η=0 , x˜ = 1 : T˜ = T˜0 therefore T˜0 = T˜∞ + c2 or c2 = T˜0 − T˜∞ ( sech (0 ) = 1 )

Thus : T˜ f = T˜∞ + ( T˜0 − T˜∞ )(sech 2 Pr η )/ x˜ 1/ 3 (2.67)

∂ 2 Y˜ ∂ Y˜ ∂ Y˜
(2) Sc1
F,f F, f 2 ˜ F, f
+ 2( tanh η )
F ∂η2 ∂η − 6 (sech η ) x ∂ x˜ = 0

Similar to (1), the solution of Y˜F, f is Y˜F, f = c1 + c2 (sech 2ScF η )/ x˜ 1/ 3

η →∞ : Y˜F, f → 0 therefore c1 = 0 and Y˜F, f = c2 (sech 2ScF η )/ x˜ 1/ 3

η=0 , x˜ = 1 : Y˜F, f = Y˜F,0 therefore c2 = Y˜ F,0

Thus : Y˜F, f = Y˜F,0 (sech 2ScF η )/ x˜ 1/ 3 (2.68)

2 ˜ ∂ Y˜O,f ∂ Y˜O, f
1 ∂ YO, f 2 ˜
(3) Sc + 2 (tanh η )
∂η − 6(sech η ) x ∂ x˜ = 0
O ∂ η2

Similar to (2), the solution of Y˜O, f is Y˜O, f = c1 + c2 (sech 2Sc O η )/ x˜ 1/ 3

η →∞ : Y˜O, f → Y˜O,∞ therefore c1 = Y˜O,∞ and Y˜O, f = Y˜O,∞ + c2 (sech 2ScO η )/ x˜ 1/ 3

η=0 , x˜ = 1 : Y˜O, f = 0 therefore 0 = Y˜O,∞ + c2 or c2 = − Y˜O,∞

Thus : Y˜O, f = Y˜O,∞ [1 − (sech 2ScO η )/ x˜ 1/ 3 ] (2.69)

(4) Summary

T˜ f = T˜∞ + ( T˜0 − T˜∞ )(sech 2 Pr η )/ x˜ 1/ 3 (2.67)

23
Y˜F, f = Y˜F,0 (sech 2ScF η )/ x˜ 1/ 3 (2.68)

Y˜O, f = Y˜O,∞ [1 − (sech 2ScO η )/ x˜ 1/ 3 ] (2.69)

2.8 Summary of Formulation

ρ u y
ρ 3 µ∞ x 0 1/ 3 3 M2
η = 3 µ∞ x0 2/ 3 ⌠
1 ˜
⌡ d y′ , ψ= ρ∞ u0 x f (η ) , x0 =
0 ρ∞
∞ 0 x˜ 3
32 u0 ρ∞ µ∞

cp T ν W cp E
T˜ = q , Y˜F = Y F , Y˜O = ν FW F YO , E˜ = q ,
F O O F

nT −n F −nO +1 nF +nO −1
q 
6 x 0 νOnO B  p 
Da = n −1 n +n −1  F
νF O WF O F u0  cp  R

µ ρµ µ ρµ λ / cp Sc
Pr =
λ / cp = ρ λ / cp , Sci =
ρ Di = ρ 2 Di , Lei = ρ D = Pri
i

f = tanh η , d f / d η = sech 2 η , u = u0 (sech 2 η )/ x˜ 1/ 3

1 ∂ T˜ ∂ T˜ ∂ T˜
2
= − Da x˜ 4/ 3 Y˜F F Y˜O O T˜ nT −nF −nO + 1exp ( − E˜ / T˜ )
n n
+ 2 (tanh η ) − 6( sech 2 η ) x˜
Pr ∂η 2 ∂η ∂ x˜

1 ∂ Y˜F ∂ Y˜ ∂ Y˜
2
+ 2 (tanh η ) F − 6 (sech 2 η ) x˜ F = Da x˜ 4 / 3 Y˜F F Y˜O O T˜ nT −nF −nO +1 exp ( − E
˜ / T˜ )
n n
Sc F ∂η 2 ∂η ∂ x˜

1 ∂ Y˜O ∂ Y˜ ∂ Y˜
2
+ 2 (tanh η ) O − 6 (sech 2 η ) x˜ O = Da x˜ 4/ 3 Y˜F F Y˜O O T˜ nT −nF −nO +1 exp ( − E˜ / T˜ )
n n
ScO ∂ η2 ∂η ∂ ˜x

η=0 , x˜ = 1 : T˜ = T˜0 , Y˜F = Y˜ F,0 , Y˜O = 0

η=0 , x˜ > 1 : ∂ T˜ / ∂η = ∂ Y˜F/ ∂η = ∂ Y˜O / ∂ η = 0

η →∞ : T˜ → T˜∞ , Y˜F → 0 , Y˜O → Y˜O,∞

T˜ f = T˜∞ + ( T˜0 − T˜∞ )(sech 2 Pr η )/ x˜ 1/ 3

Y˜F, f = Y˜F,0 (sech 2ScF η )/ x˜ 1/ 3

Y˜O, f = Y˜O,∞ [1 − (sech 2ScO η )/ x˜ 1/ 3 ]

24
Chapter 3: Ignition Analysis

3.1 Description of Analysis

A total of three ignition analyses will be presented in this chapter. The first two

analyses pertains to the cool jet flowing into hot ambient scenario, for two categories

of fuels with local LeF <1 (hydrogen belongs to this category) and LeF ≈ 1 (for fuels

like methane). The third analysis is for the hot jet flowing into cool ambient scenario.

The results and discussion will be based on ignition analysis I (for cool jet case)

and analysis III (for hot jet case), as the main focus is on the spontaneous ignition of

hydrogen, which occur when the temperature and reaction rate conditions in the inner

region (which is the reaction zone) match the outer region (which lies outside the

reaction zone). The second analysis is presented for comparison purposes between

hydrogen and other common fuels like methane and propane.

3.2 Ignition Analysis I : T∞ > T0 , LeF < 1

Because spontaneous ignition is primarily controlled by temperature, ignition

occurs at η → ∞ if successful. This is the outer edge of the jet.

(A) Outer Solutions

In the outer region, there is no reaction because of the low temperature.

Before ignition, there is only weak reaction. All the variables deviate from the frozen

solutions by O ( ε ).

T˜ = T˜ f + εΘ +O (ε2 ) ; T˜ f = T˜∞ + (T˜0 − T˜∞ ) (sech 2 Pr η )/ x˜1/3 = T˜∞ − (T˜∞ − T˜0 ) (sech 2 Pr η )/ x˜1/3

Y˜F = Y˜F, f + εΦF +O (ε2 ) ; Y˜F, f = Y˜F,0 (sech 2 ScF η ) / x˜1/3

25
Y˜O = Y˜O, f + εΦO +O (ε2 ) ; Y˜O, f = Y˜O,∞ [1 − (sech 2 ScO η ) / x˜1/3 ]

η=0 , x˜ = 1 : T˜ = T˜0 , Y˜F = Y˜F,0 , Y˜O = 0 ; η=0 , x˜ > 1 :

∂ T˜ /∂η = ∂ Y˜F /∂η = ∂ Y˜O /∂η = 0

1 ∂2 T˜ ∂ T˜ ˜
2 η ) x˜ ∂ T = 0
(1) Pr + 2 (tanh η )
∂η
− 6 (sech
∂ x˜
∂η2

2 ˜ 2 ∂ [ T˜ f + εΘ +O (ε2 ) ] ˜ 2
1 ∂ [ T f + εΘ +O (ε ) ] 2 η ) x˜ ∂ [ T f + εΘ +O (ε ) ] = 0
+ 2 (tanh η ) − 6 (sech
Pr ∂η2 ∂η ∂ x˜

 1 ∂ 2 T
f ∂ T
f ∂ T
f 

 2
+ 2 ( tanh η ) − 6 ( sech 2 η ) x

 Pr ∂ η ∂η ∂ x

 1 ∂2Θ ∂Θ ∂Θ 
+ε  2
+ 2 ( tanh η ) − 6 ( sech 2 η ) x
2
 + O (ε ) = 0
 Pr ∂ η ∂η ∂ x


2 ˜ ∂ T˜ f ˜
1 ∂ Tf 2 η ) x˜ ∂ T f = 0 , we have
Since Pr + 2 (tanh η )
∂η
− 6 (sech
∂ x˜
∂η2

1 ∂2 Θ ∂Θ ∂Θ
+ 2 (tanh η ) − 6 (sech 2 η ) x˜ ˜ = 0 (3.1)
Pr ∂η2 ∂η ∂x

By the same approach as that for the frozen solution, we obtain

Θ = aT + aT (sech 2 Pr η )/ x˜1/3

η=0 , x˜ = 1 : T˜ = T˜ f + εΘ +O (ε2 ) = T˜0 ∴ Θ=0 ⇒ 0 = aT + aT and

Θ = aT [1 − (sech 2 Pr η )/ x˜1/3 ]

The condition η = 0 , x˜ > 1 : ∂ T˜ /∂η = 0 is automatically satisfied.

Thus : T˜ = T˜ f + εΘ + O (ε2 ) = (T˜∞ + ε aT ) − [ (T˜∞ − T˜0 ) + ε aT ](sech 2 Pr η )/ x˜1/3 +O (ε2 )

2 ˜ ˜ ˜
(2) Sc1 ∂ Y2F + 2 (tanh η ) ∂∂η
YF ∂Y
− 6 (sech 2 η ) x˜ ˜F = 0
∂x
F ∂η

2 ˜ 2 ∂[ Y˜F, f +εΦF +O (ε2 )] ˜ 2

1 ∂ [ YF,f +εΦF +O (ε )] 2 η) x˜ ∂ [ YF, f +εΦF +O (ε ) ] = 0
+2 (tanh η ) −6(sech
ScF ∂η2 ∂η ∂ x˜

26
  1 ∂2 Y˜F, f ∂ Y˜F, f ∂ Y˜F, f 
 + 2 (tanh η ) − 6 (sech 2 η ) x˜ +
 ScF ∂ η2 ∂η ∂ x˜ 
  1 ∂2 Φ 
+ε F + 2 (tanh η ) ∂ΦF − 6 (sech 2 η ) x˜ ∂ΦF +O (ε2 ) = 0
  ScF ∂ η2 ∂η ∂ x 
˜

∂2 Y˜ ∂ Y˜ ∂ Y˜
Since Sc1 F, f
+ 2 (tanh η )
F, f
∂η
F, f
− 6 (sech 2 η ) x˜ ˜ = 0 , we have
∂x
F ∂η2

1 ∂2 Φ F ∂Φ ∂Φ
+ 2 (tanh η ) F − 6 (sech 2 η ) x˜ ˜F = 0 (3.2)
ScF ∂η2 ∂η ∂x

By the same approach as that for the frozen solution, we obtain

ΦF = aF + aF (sech 2 ScF η )/ x˜1/3

η=0 , x˜ = 1 : Y˜F = Y˜F, f + εΦF +O (ε2 ) = Y˜F,0 ∴ ΦF = 0 ⇒ 0 = aF + aF and

ΦF = aF [1 − (sech 2 ScF η )/ x˜1/3 ]

Thus : Y˜F = Y˜F, f + εΦF +O (ε2 ) = ε aF + ( Y˜F,0 − ε aF ) (sech 2 ScF η )/ x˜1/3 +O (ε2 )

2 ˜ ˜ ˜
(3) Sc1 ∂ Y2O + 2 (tanh η ) ∂∂η
YO ∂Y
− 6 (sech 2 η ) x˜ ˜O = 0
∂x
O ∂η

2 ˜ 2 ∂[ Y˜O, f +εΦO +O (ε2 )] ˜ 2

1 ∂ [ YO, f +εΦO +O (ε )] 2 η) x˜ ∂[ YO, f +εΦO +O (ε )] = 0
+ 2(tanh η) − 6(sech
ScO ∂η2 ∂η ∂ x˜

 1 ∂2 Y˜O, f ∂ Y˜O, f ˜ 
 + 2 (tanh η ) − 6 (sech 2 η ) x˜ ∂ YO, f  +ε
 ScO ∂η2 ∂η ∂ x˜ 
  1 ∂2 Φ 
+ε  O + 2 (tanh η ) ∂ΦO − 6 (sech 2 η ) x˜ ∂ΦO + O (ε2 ) = 0
∂ η ∂ ˜ 
  O ∂ η
Sc 2 x 

∂2 Y˜ ∂ Y˜ ∂ Y˜
Since Sc1 O, f
2
+ 2 (tanh η )
O, f
∂η
− 6 (sech 2 η ) x˜
∂
O, f
x˜
= 0 , we have
O ∂η

1 ∂2 ΦO ∂Φ ∂Φ
+ 2 (tanh η ) O − 6 (sech 2 η ) x˜ ˜O = 0 (3.3)
ScO ∂η2 ∂η ∂x

By the same approach as that for the frozen solution, we obtain

ΦO = aO + aO (sech 2 ScO η )/ x˜1/3

27
 η=0 , x˜ = 1 : Y˜O = Y˜O, f + εΦO + O (ε2 ) = 0 ∴ ΦO = 0 ⇒ 0 = aO + aO and

ΦO = aO [1 − (sech 2 ScO η )/ x˜1/3 ]

Thus : Y˜O = Y˜O, f + εΦO + O (ε2 ) = ( Y˜O,∞ + ε aO )[1 − (sech 2 ScO η )/ x˜1/3 ] +O (ε2 )

(4) Summary

T˜ = (T˜∞ + ε aT ) − [ (T˜∞ − T˜0 ) + ε aT ](sech 2 Pr η )/ x˜1/3 +O (ε2 ) (3.4)

Y˜F = ε aF + ( Y˜F,0 − ε aF ) (sech 2 ScF η )/ x˜1/3 +O (ε2 ) ; (3.5)

Y˜O = ( Y˜O,∞ + ε aO )[1 − (sech 2 ScO η )/ x˜1/3 ] +O (ε2 ) (3.6)

(B) Coordination Transformation

Define ξ = sech 2 η then at η = 0 , ξ = 1 ; as η → ∞ , ξ → 0 ;

tanh2 η = 1 − sec h2 η = 1 − ξ ; d (tanh η ) / d η = sech 2 ξ = ξ

∂ ∂ d ξ d (sech 2 η ) ∂ ∂
= = = 2 (sech η )[ − (sech η ) (tanh η ) ] =
∂η ∂ξ d η dη ∂ξ ∂ξ
∂ ∂
= − 2 (sech 2 η ) ( tanh η ) = − 2 (tanh η ) ξ
∂ξ ∂ξ

∂2 ∂  ∂  d (tanh η ) ∂ ∂  ∂ 
= −2 (tanh η )ξ  = −2  ξ + (tanh η ) ξ 
∂η 2 ∂η  ∂ξ   d η ∂ξ ∂η  ∂ξ 

  ∂  ∂  ∂  
= −2 ξ2 + ( tanh η )  − 2 ( tanh η ) ξ ξ  
  ∂ξ  ∂ ξ  ∂ ξ   

∂  ∂ ∂2   ∂2 ∂ ∂
= −2 ξ 2 +4 (tanh2 η ) ξ + ξ 2 2  = 4 ξ (1− ξ ) ξ 2 +  −2 ξ 2 (3.7)
∂ξ  ∂ξ ∂ξ   ∂ξ ∂ξ  ∂ξ

1 ∂2 T˜ ∂ T˜ ˜
2 η ) x˜ ∂ T = − Da x˜ 4 /3 Y˜ nF Y˜ nO T˜ nT −nF −nO +1 exp(− E
˜ /T˜ )
(1) Pr +2 (tanh η )
∂η
−6(sech
∂ x˜ F O
∂η2

1 ∂2 T˜ ∂ T˜ ∂ T˜
+2 (tanh η ) −6(sech 2 η ) x˜ ˜
Pr ∂η 2 ∂η ∂x

1   ∂2 T˜ ∂ T˜  ∂ T˜   ∂ T˜  ∂ T˜
=  4 ξ (1− ξ ) ξ 2 +  − 2ξ 2 2
 +2 (tanh η ) −2 (sech η ) (tanh η ) ∂ξ  −6 ξ x˜ ∂ x˜
Pr   ∂ξ ∂ξ  ∂ξ   

28
 1   ∂2 T˜ ∂ T˜  ˜
2 ∂T ∂ T˜ ∂ T˜
=
Pr 
4 ξ (1− ξ ) ξ 2 +
∂ξ  −2 ξ ∂ξ  −4 ξ (1− ξ ) ∂ξ −6 ξ x˜ ∂ x˜
 ∂ξ  

 1 ∂2 T˜ 1 ∂ T˜  2 2 ∂ T˜ ∂ T˜
= 4 ξ (1− ξ ) ξ 2 + ( −1)
∂ξ  − Pr ξ ∂ξ −6 ξ x˜ ∂ x˜
 Pr ∂ξ Pr 

Thus :

 1 ∂2 T˜ 1 ∂ T˜   2 ∂ T˜ ˜ ∂ T˜  4/3 ˜ n ˜ n ˜ n −n −n +1 ˜ ˜
4 (1−ξ ) ξ2 2 + ( −1)ξ  −ξ  Pr ξ ∂ξ +6 x ∂ x˜  = −Da x˜ YF F YO O T T F O exp(− E /T )
 Pr ∂ξ Pr ∂ξ   

(3.8)

2 ˜ ˜ ˜
(2) Sc1 ∂ Y2F +2 (tanh η ) ∂∂η
YF ∂Y
−6(sech 2 η ) x˜ ˜F = Da x˜4 /3 Y˜FnF Y˜OnO T˜ nT −nF −nO +1 exp (− E˜ /T˜ )
∂x
F ∂η

The equation is similar to (1). By the same approach as that of (1), we have

 1 ∂2 Y˜ 1 ∂ Y˜   2 ∂ Y˜ ∂ Y˜ 
4(1−ξ ) ξ2 2F +( −1)ξ F −ξ  ξ F +6 x˜ ˜F  = Dax˜4/3 Y˜FnF Y˜OnO T˜ nT −nF−nO+1 exp(− E˜ /T˜ )
 ScF ∂ξ ScF ∂ξ   ScF ∂ξ ∂x 

(3.9)

2 ˜ ˜ ˜
(3) Sc1 ∂ Y2O +2 (tanh η ) ∂∂η
YO ∂Y
−6(sech 2 η ) x˜ ˜O = Da x˜4 /3 Y˜FnF Y˜OnO T˜ nT −nF −nO +1 exp (− E˜ /T˜ )
∂x
O ∂η

The equation is similar to (1). By the same approach as that of (1), we have

 1 2 ∂2 Y˜O 1 ∂ Y˜   2 ∂ Y˜O ˜ ∂ Y˜O 

4(1−ξ ) ξ +( −1)ξ O −ξ  ξ +6 x ˜  = Dax˜4/3 Y˜FnF Y˜OnO T˜ nT −nF−nO+1 exp(− E˜ /T˜ )
 O ∂ξ
Sc 2 ScO ∂ξ   ScO ∂ξ ∂x 

(3.10)

(4) η → ∞ : T˜ → T˜∞ , Y˜F → 0 , Y˜O → Y˜O,∞ ⇒ ξ=0 : T˜ = T˜∞ ,

Y˜F = 0 , Y˜O = Y˜O,∞

(5) Frozen solutions

T˜ f = T˜∞ − (T˜∞ − T˜0 ) (sech 2 Pr η )/ x˜1/3 ⇒ T˜ f = T˜∞ − (T˜∞ − T˜0 )ξ Pr/ x˜1/3 (3.11)

Y˜F, f = Y˜F,0 (sech 2 ScF η ) / x˜1/3 ⇒ Y˜F, f = Y˜F,0 ξ ScF / x˜1/3 (3.12)

29
Y˜O, f = Y˜O,∞ [1− (sech 2 ScO η ) / x˜1/3 ] ⇒ Y˜O, f = Y˜O,∞ (1− ξ ScO / x˜1/3 ) (3.13)

(6) Outer solutions :

T˜ = (T˜∞ + ε aT ) −[ (T˜∞ − T˜0 ) + ε aT ]ξ Pr/ x˜1/3 + O (ε2 ) (3.14)

Y˜F = ε aF + ( Y˜F,0 − ε aF )ξ ScF/ x˜1/3 + O (ε2 ) ; (3.15)

Y˜O = ( Y˜O,∞ + ε aO ) (1− ξ ScO/ x˜1/3 ) + O (ε2 ) (3.16)

(7) Summary

 1 ∂2 T˜ 1 ∂ T˜   2 ∂ T˜ ˜ ∂ T˜ 
4(1−ξ ) ξ2 2 +( −1)ξ −ξ  ξ +6 x ˜  =−Dax˜4/3 Y˜FnF Y˜OnO T˜ nT −nF−nO+1 exp(− E˜ /T˜ )
 Pr ∂ξ Pr ∂ξ   Pr ∂ξ ∂x 

 1 ∂2 Y˜ 1 ∂ Y˜   2 ∂ Y˜ ∂ Y˜ 
4(1−ξ ) ξ2 2F +( −1)ξ F −ξ  ξ F +6 x˜ ˜F  = Dax˜4/3 Y˜FnF Y˜OnO T˜ nT −nF−nO+1 exp(−E˜ /T˜ )
 ScF ∂ξ ScF ∂ξ   ScF ∂ξ ∂x 

 1 2 ∂2 Y˜O 1 ∂ Y˜O   2 ∂ Y˜O ˜ ∂ Y˜O  4/3 ˜ n ˜ n ˜ n −n −n +1 ˜ ˜

4(1−ξ ) ξ +( −1)ξ
∂ξ −ξ  Sc ξ ∂ξ +6 x ∂ x˜  =Dax˜ YF F YO O T T F O exp(−E /T )
Sc
 O ∂ξ 2 ScO   O 

ξ=0 : T˜ = T˜∞ , Y˜F = 0 , Y˜O = Y˜O,∞

Frozen solutions : T˜ f = T˜∞ − (T˜∞ − T˜0 )ξ Pr/ x˜1/3 ; Y˜F, f = Y˜F,0 ξ ScF / x˜1/3 ;

Y˜O, f = Y˜O,∞ (1− ξ ScO / x˜1/3 )

Outer solutions : T˜ = (T˜∞ + ε aT ) −[ (T˜∞ − T˜0 ) + ε aT ]ξ Pr/ x˜1/3 + O (ε2 )

Y˜F = ε aF + ( Y˜F,0 − ε aF )ξ ScF/ x˜1/3 + O (ε2 ) ; Y˜O = ( Y˜O,∞ + ε aO ) (1− ξ ScO/ x˜1/3 ) + O (ε2 )

(C) Inner Expansion

Define inner variable : ξ Pr = ε x˜1/3 ζ / Y˜F,0 or ζ = Y˜F,0 ξ Pr /(ε x˜1/3 ) , ε = T˜∞2 /E˜ ,

α = (T˜∞ − T˜0 ) / Y˜F,0 ; ξ=0 : ζ=0 (3.17)

T˜ = T˜ f (ξ Pr = ε x˜1/3 ζ / Y˜F,0 ) + εθ +O (ε2 ) = T˜∞ − (T˜∞ − T˜0 ) (ε x˜1/3 ζ / Y˜F,0 )/ x˜1/3 + εθ +O (ε2 ) = T˜∞ + ε (θ − αζ ) +O (ε2 )

Y˜F = Y˜F, f (ξPr = ε x˜1/3 ζ / Y˜F,0 )+εφF +O (ε2) = Y˜F,0 (ε x˜1/3 ζ / Y˜F,0 )ScF /Pr / x˜1/3 +εφF +O (ε2) = ε LeF ( Y˜F,0 / x˜1/3 )1−LeF ζ LeF +εφF +O (ε2)

For LeF < 1 (for hydrogen, LeF ≈ 0.5) , ε LeF >> ε and

30
 Y˜F = ε LeF ( Y˜F,0 / x˜1/3 )1− LeF ζ LeF + O (ε )

Y˜O = Y˜O, f (ξ Pr = ε x˜1/3 ζ / Y˜F,0 ) + εφO +O (ε2 ) = Y˜O,∞ [1− (ε x˜1/3 ζ / Y˜F,0 )ScO /Pr / x˜1/3 ]+
] + ε φ O + O (ε 2 ) = Y˜O ,∞ + O (ε LeO ;ε )

∂ ∂ ∂ζ Y˜F,0 Pr Pr-1 ∂ ∂ Y˜ ,0 Pr ∂ ∂
= = ξ therefore ξ = Pr F1/3 ξ = Pr ζ or
∂ξ ∂ζ ∂ξ ε x˜1/3 ∂ζ ∂ξ εx˜ ∂ζ ∂ζ

1 ∂ ∂
ξ =ζ
Pr ∂ξ ∂ζ

∂2 ∂  ∂  ∂  Y˜ Pr ∂  Y˜ Pr  ∂ ∂  ∂ 
=   =  F,01/3 ξ Pr-1  = F,01/3  ( Pr −1)ξ Pr−2 + ξ Pr−1   
∂ξ ∂ξ ∂ξ  ∂ξ  ε x˜
2 ∂ζ  ε x˜  ∂ζ ∂ξ ∂ζ  

Y˜F,0 Pr  ˜  
Pr−2 ∂ + ξ Pr−1 YF,0 Pr ξ Pr−1 ∂  ∂  = Pr Pr ζ 2 ∂ + ( Pr −1)ζ ∂
2 
=  ( Pr −1) ξ   
1/3
εx 
˜ ∂ζ εx˜1/3 ∂ζ ∂ζ  2
 ξ  ∂ζ 2 ∂ζ 

1 2 ∂2 2
2 ∂ + ( Pr −1)ζ ∂
ξ = Pr ζ
Pr ∂ξ 2 ∂ζ 2 ∂ζ

1 2 ∂2  1  ∂ ∂2 ∂  1  ∂ ∂2
ξ +  −1ξ = Pr ζ 2 2 + ( Pr - 1)ζ + −1 Pr ζ = Pr ζ 2 2 (3.18)
Pr ∂ξ  Pr  ∂ξ
2 ∂ζ ∂ζ  Pr  ∂ζ ∂ζ

1 ∂ Pr ∂ 1 ∂
ξ = ζ = ζ
Sc j ∂ξ Sc j ∂ζ Le j ∂ζ

1 2 ∂2  1  ∂ Pr  2 ∂
2 ∂   1  ∂
ξ +  − 1  ξ
∂ ξ
=  Pr ζ + ( Pr - 1 ) ζ
∂ ζ  +  Sc − 1  Pr ζ ∂ ζ
Sc j ∂ξ 2  j
Sc  Sc j ∂ζ 2
  j 
Pr  2 ∂2 ∂ 
Pr ζ + ( Pr − Sc j ) ζ
Sc j  ∂ ζ 
=
∂ζ2

1  ∂2  Sc j  ∂  Pr  2 ∂2 ∂ 
 Pr ζ 2 2 + Pr 1− ζ  = Le  ζ + (1− Le j )ζ (3.19)
∂ζ 
=
Le j  ∂ζ  Pr  ∂ζ  j ∂ζ 2

E˜ / T˜ = E˜ /[ T˜∞ + ε ( θ − α ζ ) + O ( ε 2 ) ] = E˜ /{ T˜∞ [ 1 + ( ε / T˜∞ ) ( θ − α ζ ) + O ( ε 2 ) ] }

= (E˜ /T˜∞ ) − ε (E˜ /T˜∞2 ) (θ − αζ ) + (E˜ /T˜∞ ) O (ε2 ) = (E˜ /T˜∞ ) − (θ − αζ ) + O (ε )

exp(− E˜ /T˜ ) = exp[ − (E˜ /T˜∞ ) + (θ − αζ ) + O (ε ) ] = exp(− E˜ /T˜∞ ) exp (θ − αζ ) + O (ε ) (3.20)

(1) ξ = 0 : ζ=0 , T˜ = T˜∞ , Y˜F = 0

T˜ (ζ = 0) = T˜∞ + εθ (ζ = 0) + O (ε2 ) = T˜∞ ∴ θ (ζ = 0) = 0

31
Y˜F (ζ = 0) = ε LeF φF (ζ = 0) + O (ε2 LeF ) = 0 ∴ φF (ζ = 0) = 0

(2)

 1 ∂2 T˜ 1 ∂ T˜   2 ∂ T˜ ∂ T˜ 
4 (1− ξ ) ξ 2 2 + ( −1)ξ  −ξ  ξ +6 x˜ ˜  = − Da x˜ 4 /3 Y˜Fn F Y˜OnO T˜ nT −nF −nO +1 exp(− E˜ /T˜ )
 Pr ∂ξ Pr ∂ξ   Pr ∂ξ ∂x 
 2
∂ T˜ ∂T˜ 1 ∂ T˜   2 ∂ T˜ ∂ T˜ 
4[1− (εζ )1/Pr ] Pr ζ 2 2 + ( Pr - 1)ζ + ( −1) Pr ζ  − (εζ )1/Pr  Pr ζ +6 x˜ ˜ 
 ∂ζ ∂ζ Pr ∂ζ   Pr ∂ζ ∂x 

=−Dax˜4/3[εLeF (Y˜F,0 /x˜1/3 )1−LeF ζ LeF +O(ε)]nF [ Y˜O,∞ +O(ε LeO;ε)]nO [ T˜∞ +O(ε)]nT −nF−nO+1[exp(−E˜ /T˜∞)exp(θ −αζ )+O(ε )]

=−Dax˜4/3 [ε LeF (Y˜F,0 /x˜1/3 )1−LeF ζ LeF +O(ε )]nF [ Y˜O,∞ +O(ε LeO;ε )]nO [ T˜∞ +O(ε )]nT −nF−nO+1[ exp(−E˜ /T˜∞ )exp(θ −αζ )+O(ε )]
(3.21)

Take the leading order terms in ε

∂2 { T˜∞ + ε [ θ − (T˜∞ − T˜0 )ζ/ x˜1/3 ]} ∂2 θ

4 Pr ζ 2 2
= 4 ε Pr ζ 2 2
∂ζ ∂ζ

= −ε LeF n F Da x˜ 4 /3 ( Y˜F,0 / x˜1/3 )(1− LeF )nF Y˜O,∞nO T˜∞nT −nF−nO +1 exp(− E˜ /T˜∞ )ζ LeF nF exp(θ − αζ )

∂2 θ ˜ nO ˜ nT −nF −nO +1 exp (− E˜ /T˜ )

ζ2 = − ε LeF n F−1 Da YO,∞ T∞ ∞ ˜ 4 /3 ˜
x ( YF,0 / x˜1/3 )(1− LeF )nF ζ LeF nF exp (θ − αζ )
∂ζ 2 4 Pr

Da Y˜O,∞nO T˜∞nT −nF−nO +1 exp (− E˜ /T˜∞ ) ˜ 4 /3 ˜ ˜1/3 (1− LeF )nF

Define D˜ a = ε LeF nF−1 4 Pr
x ( YF,0 / x ) (3.22)

= Reduced Damköhler number

∂2 θ
ζ2 = − D˜ aζ LeF nF exp(θ − αζ ) (3.23)
∂ζ 2

(D) Matching

T˜ = (T˜∞ + ε aT ) −[ (T˜∞ − T˜0 ) + ε aT ]ξ Pr/ x˜1/3 +O (ε2 ) = (T˜∞ + ε aT ) −[ (T˜∞ − T˜0 ) + ε aT ](εζ / Y˜F,0 ) +O (ε2 )

= T˜∞ + ε (aT − α ζ ) + O (ε 2 ) = T˜∞ + ε (θ − α ζ ) + O (ε 2 )

ζ →∞

∴ θ (ζ → ∞) = aT or (∂θ /∂ζ )ζ →∞ = 0

32
 (E) Summary

∂2 θ
ζ2 = − D˜ aζ LeF nF exp(θ − αζ ) ; ζ =0 : θ=0 ; ζ →∞ : ∂θ /∂ζ → 0 ,
∂ζ 2

θ → aT

Da Y˜O,∞nO T˜∞nT −nF−nO +1 exp (− E˜ /T˜∞ ) ˜ 4 /3 ˜ ˜1/3 (1− LeF )nF

α = (T˜∞ − T˜0 ) / Y˜F,0 , D˜ a = ε LeF n F−1 x ( YF,0 / x )
4 Pr

, ε = T˜∞2 /E˜

Note: Since there is no differentiation with respect to x˜ , the problem is locally

similar and x˜ can be considered as a constant in the structure equation.

(F) Rescaling

when the change of α is by changing T∞ or YF,0, or nF or LeF is varying

Define Tˆ∞ as the reference value of T˜∞ , nˆF as the reference value of nF ,

YˆF,0 as the reference value of Y˜F,0

εˆ = (Tˆ∞ )2 /E˜ as the reference value of ε ⇒ εˆ /ε = (Tˆ∞ /T˜∞ )2

∧ ∧
ˆ = εˆ Le F nˆF −1 [ Da /( 4 Pr ) ] Y
nO Tˆ nT − nˆF − nO +1 exp ( − E
/ Tˆ ) x
4 / 3 ( Yˆ / x
1/ 3 ) (1 − Le F ) nˆ F
Da O ,∞ ∞ ∞ F ,0

∧ ∧ ∧ ∧

= ( Tˆ∞ / T
∞ ) nT + ( 2 LeF −1) nˆF − nO −1 ( T
∞ 2 / E
)( LeF nF − LeF nF ) T
∞ nF −nˆ F ( YˆF ,0 / Y
F ,0 )(1− Le F ) nˆF ( Y
F ,0 / x1/ 3 )(1− Le F ) nˆF −(1− Le F ) nF exp[ E
( T
∞ 1 − Tˆ∞ 1 )
ˆ

∧ ∧ ∼
T
∞ nF −nˆF ( YˆF ,0 / Y
F ,0 )(1− Le F ) nˆF ( Y
F ,0 / x
1/ 3 )(1− Le F ) nˆF −(1− Le F ) nF exp[ E
( T
∞ −1 − Tˆ∞ −1 ) ] Da (3.24)

ε aT = εˆ aˆT ∴ aˆT = (ε /εˆ ) aT = (T˜∞2 /Tˆ∞2 ) aT

If only T∞ is varying

Dˆ a = (Tˆ∞ /T˜∞ )nT +(2 LeF −1)nF−nO −1 exp[ E˜ (T˜∞−1 − Tˆ∞−1 ) ] D˜ a (3.25)

aˆT = (T˜∞2 /Tˆ∞2 ) aT (3.26)

In the calculations, we

(1) Choose a value of α as the reference value.

33
(2) Find the value of Tˆ∞ based on this reference value of α. (need to specify T˜0 and

Y˜F,0 )

(3) For other values of α, determine T˜∞ using the same values of T˜0 and Y˜F,0 .

(4) Use the values of Tˆ∞ and T˜∞ to determine Dˆ a .

ˆ
If only nF is varying, Dˆ a = (T˜∞2 /E˜ ) LeF ( nˆF −nF ) T˜∞nF −nF ( Y˜F,0 / x˜1/3 )(1− LeF )( nˆF −nF ) D˜ a (3.27)

If only YF,0 is varying, Dˆ a = ( YˆF,0 / Y˜F,0 )(1− LeF )nF D˜ a (3.28)

∧ ∧ ∧ ∼
ˆ = ( T
2 / E
)( Le
If only LeF is varying, Da F − LeF ) nF
( YˆF ,0 / Y
F ,0 )(1− LeF ) nF ( Y
F ,0 / x
1/3 )( LeF + LeF ) nF Da (3.29)
∞

3.3 Ignition Analysis II : T∞ > T0 , LeF ≈ 1

As in I, ignition occurs at η → ∞ if successful.

(A) Coordination Transformation and Outer Solutions (From Analysis I)

Define ξ = sech 2 η then at η = 0 , ξ = 1 ; as η → ∞ , ξ = 0

 1 ∂2 T˜ 1 ∂ T˜   2 ∂ T˜ ∂ T˜ 
4(1− ξ ) Pr ξ2 2 +(Pr −1)ξ ∂ξ  − ξ  Pr ξ ∂ξ +6 x˜ ˜  =−Dax˜4/ 3 Y˜F F Y˜O O T˜ nT −nF −nO +1 exp( − E
˜ / T˜ )
n n
 ∂ξ   ∂x 
 1 ∂2Y˜ 1 ∂ Y˜   2 ∂ Y˜ ∂Y˜ 
4(1− ξ) Sc ξ2 2F +(Sc −1)ξ ∂ξF  − ξ Sc ξ ∂ξF +6x˜ ˜F  =Dax˜4/ 3Y˜F F Y˜O O T˜nT −nF−nO+1exp(− E˜ / T˜ )
n n
 F ∂ξ F   F ∂ x 

ξ=0 : T˜ = T˜∞ , Y˜F = 0

Outer solutions

T˜ = ( T˜∞ + ε aT ) − [( T˜∞ − T˜0 ) + ε aT ] ξPr/ x˜ 1/ 3 + O ( ε 2 )

Y˜F = ε aF + ( Y˜F,0 − ε aF )ξ ScF/ x˜ 1/ 3 + O ( ε 2 )

Y˜O = ( Y˜O,∞ + ε aO )(1− ξScO / x˜ 1/ 3 ) + O ( ε 2 )

34
(B) Inner Expansion

Define inner variable : ξPr = εζ or ζ = ξ Pr / ε , ε = T˜∞ 2 / E˜ ;

ξ=0 : ζ=0

From Analysis I

T˜ = T˜∞ + ε [ θ − ( T˜∞ − T˜0 )ζ/ x˜ 1/ 3 ]+ O ( ε 2 ) ; Y˜F = ε LeF ( Y˜F,0 ζ LeF / x˜ 1/ 3 ) + εφ F + O ( ε 2 )

1 ∂ ∂ 1 2 ∂2 1 ∂ 2 ∂
2
Pr ξ ∂ ξ = ζ ∂ζ ; ξ + ( −1)ξ
Pr ∂ξ 2 Pr ∂ ξ = Prζ ∂ζ 2

1 ∂ 1 ∂ 1 2 ∂2 1 ∂ Pr  2 ∂ 2 ∂ 
Sc F ξ ∂ ξ = LeF ζ ∂ζ , ξ + (
Sc F ∂ξ 2 ScF
− 1) ξ ∂ξ LeF ζ ∂ ζ 2 + (1− LeF )ζ ∂ζ 
=

∂2 θ
4ε Prζ2 = −DaY˜O,∞nO T˜∞nT −nF−nO +1exp(−E˜/ T˜∞ )x˜4/ 3[εLeF (Y˜F,0ζLeF / x˜1/ 3)+ εφF]nF exp[θ −(T˜∞ − T˜0 )ζ/ x˜1/ 3]+O(ε )
∂ζ 2

ζ=0 : θ=0 , φF = 0

Expand LeF as LeF = 1+ ε  F + O ( ε 2 ) , then for any variable A

A LeF = exp [ n ( A LeF ) ] = exp ( LeF n A ) = exp [(1 + ε  F +  ) n A ] =

= exp (  n A )exp ( ε F  n A +  ) = A ( 1 + ε F  n A +  ) (3.30)

Thus : Y˜F = ε LeF ( Y˜F,0 ζ LeF / x˜ 1/ 3 ) + εφ F + O ( ε 2 ) = ε (φ F + Y˜F,0 ζ / x˜ 1/ 3 ) + O ( ε 2 )

1 2 ∂2 1 ∂ ∂ 2 Y˜F
ξ +( − 1) ξ = Pr ζ 2 +O (ε ) (3.31)
Sc F ∂ξ 2 ScF ∂ξ ∂ζ 2

2
(1) 4 ε Prζ 2 ∂ θ2 = − DaY˜O,∞nO T˜∞nT −n F−nO +1 exp (− E˜ / T˜∞ ) x˜ 4/ 3
∂ζ

3
[ ε LeF ( Y˜ F,0 ζ LeF / x˜ 1/ 3 ) + εφ F]n F exp [ θ − ( T˜∞ − T˜0 )ζ/ x˜ 1/ 3 ]+ O( ε )

2 ∂2θ Da Y˜ O , ∞ n O T˜ ∞ n T − n F − n O + 1 exp ( − E˜ / T˜ ∞ ) 4/ 3
ζ = − 4 ε Pr x˜
∂ζ2
[ ε ( φ F + Y˜ F , 0 ζ / x˜ 1/ 3
) ]n F
exp [ θ − ( T˜ ∞ − T˜ 0 ) ζ / x˜ 1 / 3
]+ O ( ε )

35
 DaY˜O,∞nO T˜∞nT −nF −nO +1exp(− E
˜ / T˜∞)
= − εnF−1 4Pr x˜4/ 3(φF +Y˜F,0ζ/ x˜1/ 3 )nF exp[θ +(T˜0 −T˜∞)ζ/ ˜x1/ 3]+O(ε ) (3.32)

DaY˜O,∞ nO T˜∞ nT −n F−nO +1 exp (− E˜ / T˜∞ ) 4/ 3

From Section 3.2, D˜ a = ε nF−1 4 Pr x˜

∂2θ
The leading order terms are ζ2 = − D˜ a( φF + Y˜F,0 ζ / x˜ 1/ 3 )n F exp [ θ − ( T˜∞ − T˜0 )ζ/ x˜ 1/ 3 ]
∂ζ 2

 1 2 ∂ 2 T˜ 1 ∂ T˜   2 ∂ T˜ ∂ T˜ 
(2)  4(1− ξ )  Pr ξ 2 + ( Pr −1)ξ ∂ξ  − ξ  Pr ξ ∂ξ + 6 x˜ ∂ x˜ 
  ∂ξ   

  1 ∂ 2 Y˜F 1 ∂ Y˜   2 ∂ Y˜ ∂ Y˜  
+  4(1− ξ )  Sc ξ 2 2 + ( Sc − 1) ξ ∂ξF  − ξ  Sc ξ ∂ ξF + 6 x˜ ˜F   = 0
  F ∂ξ F   F ∂x 

From Section 3.2,

 1 ∂2 T˜ 1 ∂ T˜   2 ∂ T˜ ∂ T˜  2
2 ∂ θ + O ( ε 1+(1/ Pr ) ;ε 2 )
4(1− ξ )  Pr ξ2 2 + ( Pr −1)ξ ∂ξ  − ξ  Pr ξ ∂ξ + 6 x˜ ∂ x˜  = 4 ε Prζ
 ∂ξ    ∂ζ 2

 1 ∂ 2 Y˜F 1 ∂ Y˜   2 ∂ Y˜ ∂ Y˜ 
4(1− ξ )  Sc ξ2 2 + ( Sc − 1)ξ ∂ ξF  − ξ  Sc ξ ∂ξF + 6 x˜ ˜F 
 F ∂ξ F   F ∂x 

Pr  ∂ Y˜F ∂ Y˜F   2 ∂ Y˜F ˜ 

2
= 4[1− ( εζ )1/ Pr ] Le ζ 2 + ( 1- LeF )ζ  −( ε ζ )1/ Pr
 ζ + 6 ˜ ∂ YF 
x
F ∂ζ 2 ∂ζ   LeF ∂ζ ∂ x˜ 

∂ 2 [ ε (φ F + Y˜F,0 ζ / x˜ 1/ 3 )] ∂ 2 φF
= 4 Pr ζ 2 2 + O ( ε 2;ε 1+(1/ Pr) ) = 4 ε Pr ζ 2 + O ( ε 2;ε 1+(1/ Pr) )
∂ζ ∂ζ 2

∂2θ 2
1+(1/ Pr ) ) + 4 ε Prζ 2 ∂ φF + O ( ε 2 ;ε 1+(1/ Pr ) ) = 0 ∂ 2 ( θ + φF )
Thus 4 ε Prζ 2 + O ( ε or =0
∂ζ 2 ∂ζ 2 ∂ζ 2

⇒ θ + φF = c ζ + c2

ζ=0 : θ = φF = 0 ∴ c2 = 0 and θ + φF = c ζ

(3) Summary

∂2θ
ζ2 = − D˜ a( φF + Y˜F,0 ζ / x˜ 1/ 3 )n F exp [ θ − ( T˜∞ − T˜0 )ζ/ x˜ 1/ 3 ] ; ζ=0 : θ=0
∂ζ 2

θ + φF = c ζ where c still needs to be determined.

36
(C) Matching

T˜ = ( T˜∞ + ε aT ) − [( T˜∞ − T˜0 ) + ε aT ] ξPr/ x˜ 1/ 3 + O ( ε 2 ) = ( T˜∞ + ε aT ) −[( T˜∞ − T˜0 )+ ε aT ]( εζ )/ x˜ 1/ 3 +O ( ε 2 )

= T˜∞ + ε [ aT − ( T˜∞ − T˜0 )ζ/ x˜ 1/ 3 ]+ O ( ε 2 ) = T˜∞ + ε [θ − ( T˜∞ − T˜0 ) ζ/ x˜ 1/ 3 ] + O ( ε 2 )

ζ→∞

∴ θ ( ζ → ∞) = aT or ( ∂θ / ∂ζ )ζ → ∞ = 0

Y˜F = ε aF + ( Y˜F,0 − ε aF )ξ ScF/ x˜ 1/ 3 + O ( ε 2 ) = ε aF + ( Y˜F,0 − ε aF )( ξ Pr )Sc F / Pr/ x˜ 1/ 3 + O ( ε 2 )

= ε aF +( Y˜F,0 − ε aF )( εζ )LeF/ x˜ 1/ 3 +O ( ε2 ) = ε aF +( Y˜F,0 − ε aF )( εζ )[1+ ε F n( εζ ) +]/ x˜ 1/ 3 +O( ε2 )

= ε ( aF + Y˜F,0 ζ/ x˜ 1/ 3 ) + O ( ε 2 ) = ε ( φF + Y˜F,0 ζ / x˜ 1/ 3 )+ O ( ε 2 )
ζ→∞

∴ φF (ζ → ∞) = aF

From (B) : θ + φF = c ζ

ζ→∞ : θ → aT , φF → aF ∴ aT + aF = c ζ ⇒ aT + aF = c = 0 or

aF = – aT ⇒ θ + φF = 0 or φF = – θ

(E) Summary

∂2θ
ζ2 = − D˜ a(− θ + Y˜F,0 ζ / x˜ 1/ 3 )n F exp [ θ − ( T˜∞ − T˜0 )ζ/ x˜ 1/ 3 ] ; ζ=0 : θ=0 ;
∂ζ 2

ζ→∞ : d θ / dζ → 0 , θ → aT

Rescaling

From Section 3.2, σ = Y˜F,0 / x˜ 1/ 3 , ζ = σζ , β = ( T˜∞ − T˜0 )/ Y˜F,0 , then

∂2θ ˜ a( − θ + ζ )nF exp ( θ − β ζ )

ζ2 =−D ; ζ =0 : θ=0 ;
∂ζ 2

ζ →∞ : ∂θ / ∂ζ → 0 , θ → aT

Note: Since there is no differentiation with respect to x˜ , the problem is locally

similar and x˜ can be considered as a constant in the structure equation.

37
3.4 Ignition Analysis III : T0 > T∞ , x
≈ 1, Thicker Reaction Region

Because spontaneous ignition is primarily controlled by temperature, ignition occurs

near η = 0. This is the centerline of the jet.

(A) Outer Solutions

The outer solution is as follows

T˜ = T˜∞ +[( T˜0 − T˜∞ ) + ε aT ]( sech 2 Pr η )/ x˜ 1/ 3 + O ( ε 2 ) (3.33)

Y˜F = ( Y˜ F,0 − ε aF )(sech 2Sc F η )/ x˜ 1/ 3 + O ( ε 2 ) ; (3.34)

Y˜O = Y˜O,∞ − ( Y˜O,∞ + ε aO )(sech 2Sc O η )/ x˜ 1/ 3 + O ( ε 2 ) (3.35)

(B) Inner Expansion

If successful, ignition occurs in an inner region near η = 0 and x˜ ≈ 1

Since d (sech η )/ d η = 0 at η = 0, it seems that the reaction region should be thicker

than O ( ε ).

Define inner variables η = ε /( Pr Y˜O,∞ ) ζ , x˜ = 1+ ε ( 3ξ / Y˜O,∞ ) , ε = T˜0 2/ E˜ ,

β = (T˜0 − T˜∞ ) / Y˜O,∞ ; when η = 0 : ζ = 0

⇒ η2 = εζ 2 /( Pr Y˜O,∞ ) , ∂η = ε /( Pr Y˜O,∞ ) ∂ζ , ∂η2 = [ ε /( Pr Y˜O,∞ )] ∂ζ 2 ,

∂ x˜ = ( 3ε / Y˜O,∞ )∂ ξ

sech η = 1− ( η2 / 2) +  = 1− ε [ζ 2 /( 2 Pr Y˜O,∞ )] + O ( ε 2 )

sech2 η ={1− ε[ζ2/( 2PrY˜O,∞ )] +O(ε2 )}2 =1−2ε [ζ 2/[ 2PrY˜O,∞ )]+O(ε 2) =1− ε [ζ2 /( PrY˜O,∞ )]+O(ε 2)

sech2Pr η ={1− ε[ζ 2/( 2PrY˜O,∞)]+O(ε2 )}2Pr =1− ε 2Pr[ζ2/( 2PrY˜O,∞)]+O(ε2)= 1−ε (ζ2/ Y˜O,∞ )+O(ε 2)

sech 2Sc i η = {1− ε [ ζ 2 /( 2 Pr Y˜O, ∞ )]+ O ( ε 2 )}2Sc i = 1− ε 2Sci [ζ 2 /( 2 Pr Y˜O,∞ )]+ O ( ε 2 )

= 1− ε ( Sci / Pr )( ζ 2 / Y˜O, ∞ )+ O ( ε 2 ) = 1− ε Lei ( ζ 2 / Y˜O,∞ ) + O ( ε 2 ) (3.36)

38
tanh η = η − (η 3/ 3)+  = ε /( Pr Y˜O,∞ ) ζ + O ( ε 3/ 2 ) (3.37)

1/ x˜ 1/ 3 = 1/[ 1+ ε (3ξ / Y˜O,∞ )]1/ 3 = 1/[ 1+ ε ( ξ / Y˜O,∞ ) + O ( ε 2 )] = 1− ε ( ξ / Y˜O,∞ )+ O ( ε 2 ) (3.38)

T˜ = T˜ f [η = ε /( Pr Y˜O,∞ ) ζ , x˜ = 1 + 3ε ( ξ / Y˜O,∞ )]+ ε θ + O ( ε 2 )

= T˜∞ + ( T˜0 − T˜∞ )[1 − ε (ζ 2 / Y˜O,∞ ) + O( ε 2 )][1 − ε ( ξ / Y˜O,∞ ) + O( ε 2 )] + εθ + O( ε 2 )

= T˜0 + ε {θ −(T˜0 − T˜∞ )[(ζ2 / Y˜O,∞ )+(ξ/ Y˜O,∞ )]}+O( ε2 )= T˜0 + ε {θ −[(T˜0 − T˜∞)/ Y˜O,∞ ](ζ2 + ξ)}+O(ε2 )

= T˜0 + ε [ θ − β (ζ 2 + ξ ) ]+ O (ε2 ) (3.39)

Y˜F = Y˜ F, f [ η = ε /( Pr Y˜O, ∞ ) ζ , x˜ = 1 + 3 ε ( ξ / Y˜O,∞ )] + εφ F + O ( ε 2 )

= Y˜F,0 [1− ε LeF (ζ 2 / Y˜O,∞ ) + O( ε 2 )][1 − ε ( ξ/ Y˜O,∞ ) +O ( ε 2 )] + εφF + O( ε 2 ) = Y˜F,0 + O( ε ) (3.40)

Y˜O = Y˜O, f [ η = ε /( PrY˜O, ∞ ) ζ , x˜ = 1 + 3 ε (ξ / Y˜O, ∞ )] + εφO +O ( ε 2 )

= Y˜O,∞ {1 −[1 − ε LeO ( ζ 2 / Y˜O,∞ ) + O( ε 2 )][1 − ε ( ξ/ Y˜O,∞ )+ O ( ε 2 )]}+ ε φO + O( ε 2 )

= ε {φO + Y˜O,∞[LeO (ζ 2 / Y˜O,∞ ) + (ξ / Y˜O, ∞ )]}+ O( ε 2 ) = ε ( φO + LeO ζ 2 + ξ ) + O( ε 2 ) (3.41)

E˜ /T˜ = E˜ /{ T˜0 + ε [ θ − β (ζ 2 + ξ ) ]+ O (ε2 )} = E˜ /〈 T˜0 {1 + ε [ θ − β (ζ 2 + ξ ) ]/T˜0 + O (ε2 )} 〉

= (E˜ /T˜0 ){1 − ε [ θ − β (ζ 2 + ξ ) ]/T˜0 + O (ε2 )} = (E˜ /T˜0 ) − ε (E˜ /T˜02 )[ θ − β (ζ 2 + ξ ) ]+ (E˜ /T˜0 )O (ε2 )

= (E˜ /T˜0 ) −[ θ − β (ζ 2 + ξ ) ]+ O (ε )

exp(− E˜ /T˜ ) = exp{− (E˜ /T˜0 ) −[ θ − β (ζ 2 + ξ ) ]+ O (ε )} = exp (− E˜ /T˜0 ) exp[ θ − β (ζ 2 + ξ ) ]+ O (ε ) (3.42)

(1) η = 0 , x˜ = 1 : T˜ = T˜0 , Y˜O = 0 ⇒ ζ=0 , ξ=0 : θ = φO = 0

(2) η = 0 , x˜ > 1 : ∂ T˜ / ∂ η = ∂ Y˜O / ∂ η = 0 ⇒ ζ=0 , ξ>0 :

∂ θ / ∂ ζ = ∂ φO / ∂ ζ = 0

1 ∂ 2 T˜ ∂ T˜2 ˜ ∂ T˜  
1 O ∂ 2 Y˜ O 2 ∂ Y˜ ∂ Y˜ 
(3)  Pr ˜ O
2 + 2(tanh η ) ∂η − 6(sech η ) x ∂ x˜  +  ScO ∂ η2 + 2(tanh η ) ∂ η − 6(sech η ) x ∂ x˜  = 0
 ∂η   

1 ∂2 { T˜0 +ε [ θ − β (ζ 2 +ξ ) ]+O (ε2 )} ∂ { T˜0 +ε [ θ − β (ζ 2 +ξ ) ]+O (ε2 )}

˜ +2[ ε /(Pr Y˜O,∞ ) ζ +O (ε3/2 )]
Pr [ ε /(Pr YO,∞ )]∂ζ 2
ε /(Pr Y˜O,∞ ) ∂ζ

39
 ∂ { T˜0 + ε [ θ − β (ζ 2 + ξ ) ]+ O (ε2 )}
−6 {1 − ε [ ζ 2 /(Pr Y˜O,∞ ) ]+ O (ε2 )}[1 + 3 ε (ξ / Y˜O,∞ ) ]
(3 ε / Y˜O,∞ )∂ξ

2 2 2 2 2
1 ∂ [ ε ( φO + LeO ζ + ξ ) + O( ε )] 3/ 2 ∂ [ ε ( φO + LeO ζ + ξ ) + O( ε )]
+ Sc + 2[ ε / Pr ζ +O ( ε )]
O [ ε /( Pr Y˜O,∞ )] ∂ ζ 2
ε /( Pr Y˜O,∞ ) ∂ ζ

∂ [ ε (φO + LeO ζ 2 + ξ ) + O( ε 2 )]
− 6 {1 − ε [ζ 2 /( PrY˜O, ∞ )]+ O( ε 2 )}[1 + 3 ε ( ξ / Y˜O,∞ )] =0 (3.43)
( 3ε / Y˜O,∞ ) ∂ ξ

The leading order terms are

 ∂2 [ θ − β (ζ 2 + ξ ) ] ∂ [ θ − β (ζ 2 + ξ ) ] Pr ∂2 (φO + LeO ζ 2 + ξ ) ∂ (φO + LeO ζ 2 + ξ ) 

Y˜O,∞  − 2 + − 2 =0
 ∂ζ 2 ∂ξ ScO ∂ζ 2 ∂ξ 

∂2 θ ∂θ 1 ∂2 φO  ∂φ
−2 β −2 +2 β +  + 2 LeO  − 2 O − 2 = 0 or
∂ζ 2 ∂ξ LeO  ∂ζ 2
 ∂ξ

∂2θ ∂θ 2
1 ∂ φO ∂φ
− 2 + −2 O =0
∂ζ 2 ∂ ξ LeO ∂ζ 2 ∂ξ

1 ∂ T˜ ∂ T˜ ∂ T˜
2
(4) Pr 2 + 2 (tanh η ) ∂ η − 6( sech 2 η ) x˜ ∂ x˜ = − Da x˜ 4/ 3 Y˜FnF Y˜OnO T˜ nT −nF −nO +1exp (− E˜ / T˜ )
∂η

∂ 2θ ∂θ 
Y˜ O ,∞  ∂ ζ 2 − 2 ∂ ξ + O (ε ) 
 

 4/ 3 [ Y˜ nF 2 2 nO ˜ n T −n F −n O +1
 = − Da[ 1 + O ( ε ) ] F,0 + O ( ε ) ] [ ε ( φO + LeO ζ + ξ ) + O ( ε )] [ T0 + O ( ε ) ]


⋅exp(− E˜ /T˜0 ) exp[ θ − β (ζ 2 + ξ ) ]+O (ε )

= −εnO Da Y˜F,0nF T˜0nT −nF−nO +1 exp(− E˜ /T˜0 ) (φO + LeO ζ 2 + ξ )nO exp[ θ − β (ζ 2 + ξ ) ]+O (ε )

n −n −n +1
Define D˜ a = εnO Da ( Y˜F,0 F / Y˜O,∞ ) T˜0 T F O exp(− E˜ /T˜0 )
n
(3.44)

= Reduced Damköhler number

∂2 θ ∂θ
Then : −2 = − D˜ a (φO + LeO ζ 2 + ξ )nO exp[ θ − β (ζ 2 + ξ ) ] (3.45)
∂ζ 2 ∂ξ

(C) Matching

T˜ = T˜∞ + [ ( T˜0 − T˜∞ ) + ε aT ]( sech 2 Pr

η )/ x˜ 1/ 3 + O ( ε 2 )

40
= T˜∞ + [ ( T˜0 − T˜∞ ) + ε aT ][ 1 − ε (ζ 2 / Y˜O, ∞ ) ][ 1− ε ( ξ / Y˜O ,∞ ) ] + O ( ε 2 )

= T˜0 + ε { aT −[ (T˜0 − T˜∞ ) / Y˜O,∞ ](ζ 2 + ξ )} + O (ε2 ) = T˜0 + ε [ θ − β (ζ 2 + ξ ) ]+ O (ε2 )

ζ →∞

∴ θ ( ζ → ∞) → aT and ( ∂ θ / ∂ ζ )ζ → ∞ → 0

Y˜O = Y˜O ,∞ − ( Y˜O ,∞ + ε aO )( sech 2 Sc O η )/ x˜ 1/ 3 + O ( ε 2 )

= Y˜O,∞ − ( Y˜O,∞ + ε aO )[1 − ε LeO (ζ 2 / Y˜O, ∞ )][1− ε ( ξ / Y˜O,∞ )]+ O ( ε 2 )

= ε (− aO + LeO ζ 2 + ξ )+ O ( ε 2 ) = ε (φO + LeO ζ 2 + ξ ) +O ( ε 2 )

ζ→∞

∴ φO ( ζ → ∞) → − aO and ( ∂φO / ∂ ζ )η→ ∞ → 0

(D) Ignition Criterion

(1) Ignition is successful when ( ∂ T˜ / ∂ x˜ )η =0 ≥ 0

Since η = 0 is in the inner, reactive region, the ignition condition becomes

( ∂ T˜in / ∂ξ )ζ =0 ≥ 0

(∂ T˜in /∂ξ )ζ =0 = 〈 ∂ { T˜0 + ε [ θ − β (ζ 2 + ξ ) ]+O (ε2 )} /∂ξ 〉ζ =0 = ε [ (∂θ /∂ξ )ζ =0 − β ]+ O (ε2 ) ≥ 0

or (∂θ /∂ξ )ζ =0 ≥ β

(2) Ignition is successful when ∂ T˜ / ∂η ≥ 0 at any x˜

In the inner, reactive region, the ignition condition becomes ∂ T˜in /∂ζ ≥ 0 at any ξ

∂ T˜in /∂ζ = ∂ { T˜0 + ε [ θ − β (ζ 2 + ξ ) ]+ O (ε2 )} /∂ζ = ε [ (∂θ /∂ζ ) −2 β ζ ]+ O (ε2 ) ≥ 0

or ∂θ /∂ζ ≥ 2 βζ at any ξ

(E) Initial condition

At the nozzle exit, there is no reaction. Thus

θ (ξ = 0) = φO (ξ = 0) = 0

41
(F) Summary

∂2θ ∂θ 2
1 ∂ φO ∂ φO
−2
∂ζ 2 ∂ ξ LeO ∂ζ 2 − 2 ∂ξ = 0
+

∂2 θ ∂θ
−2 = − D˜ a(φO + LeO ζ 2 + ξ )nO exp[ θ − β (ζ 2 + ξ ) ]
∂ζ 2 ∂ξ

θ (ξ = 0) = φO (ξ = 0) = 0

ζ=0 , ξ>0 : ∂θ / ∂ζ = ∂ φO / ∂ζ = 0

ζ → ∞ : ∂θ / ∂ζ → 0 ; ∂φO / ∂ζ → 0

If LeO = 1, we have φO = − θ and the problem is reduced to

∂2 θ ∂θ
−2 = − D˜ a(ζ 2 + ξ − θ ]nO exp[ θ − β (ζ 2 + ξ ) ]
∂ζ 2 ∂ξ

ζ=0 : ξ=0 : θ =0 , ξ>0 : ∂θ / ∂ζ = 0 ; ζ → ∞ : ∂θ / ∂ζ → 0

Ignition criterion : (∂θ /∂ξ )ζ =0 ≥ β or ∂θ /∂ζ ≥ 2 βζ at any ξ

(G) Rescaling when the change of γ is by changing T0 or YO,∞ , or nO is varying

Define Tˆ0 as the reference value of T˜0 , nˆO as the reference value of nO ,

YˆO,∞ as the reference value of Y˜O,∞

εˆ = Tˆ02 /E˜ as the reference value of ε ⇒ εˆ /ε = (Tˆ0 /T˜0 )2

ˆ
Dˆ a = εˆnO Da ( Y˜F,0nF / YˆO,∞ ) Tˆ0nT −nF −nO +1 exp(− E˜ /Tˆ0 )
ˆ

ˆ ˆ
= (Tˆ0 /T˜0 )nT −nF+nO +1 (T˜0 /E˜ )nO −nO ( Y˜O,∞ / YˆO,∞ ) exp[ E˜ (T˜0−1 − Tˆ0−1 ) ] D˜ a (3.46)

If only T0 is varying: Dˆ a = (Tˆ0 /T˜0 )nT −nF +nO +1 exp[ E˜ (T˜0−1 − Tˆ0−1 ) ] D˜ a (3.47)

In the calculations, we

(1) Choose a value of β as the reference value.

42
(2) Find the value of Tˆ0 based on this reference value of β. (need to specify T˜∞ and

Y˜O,∞ )

(3) For other values of β, determine T˜0 using the same values of T˜∞ and Y˜O,∞ .

(4) Use the values of Tˆ0 and T˜0 to determine Dˆ a .

ˆ
If only nO is varying : Dˆ a = (T˜0 /E˜ )nO −nO D˜ a (3.47)

If only YO,∞ is varying : Dˆ a =( Y˜O,∞ / YˆO,∞ ) D˜ a (3.48)

3.5 Analysis Scenarios

3.5.1 Cool jet flowing into a hot ambient (T∞ > T0)

(With reference to Section 3.2 Ignition Analysis I) In the presence of a weak reaction,

the temperature is increased from its frozen value by a small, O (ε) amount where

ε = T˜∞2 / E˜ while the reactant concentrations are reduced from their respective frozen

values by an O (ε) amount. Because ignition is primarily controlled by temperature,

ignition occurs near η → ∞ if successful. Fig.3.1 shows a schematic of the inner and

outer regions for the scenario.

Outer
Region

Inner
Region
η= 0
T∞ η= ∞
x
Air
y

Fuel
To
Fig. 3.1: Schematic of inner and outer
regions for cool jet, hot ambient case.
43
 Away from this high temperature region, the reaction is frozen. In the outer,

chemically frozen region, the outer solutions are similar to Eqns. 3.4 – 3.6, but with

an O (ε) change in their values. In the inner, reactive region, defining a stretched inner

variable as

ζ = Y˜F,0 (sech η )2Pr /(ε x˜1/3 ) (3.49)

and substituting into Parts (C) to (E) of Ignition Analysis I, yields, when LeF is

sufficiently smaller than unity, as for hydrogen,

ζ 2 (∂2 θ /∂ζ 2 ) = − D˜ a ζ LeF exp(θ − αζ ) , (3.50)

where α = ( T˜∞ − T˜0 )/Y˜F,0 and the reduced Damköhler number is defined as

D˜ a = Da Y˜O,∞ x˜ 4/3[Y˜F,0 /(ε x˜1/3 )]1−LeF exp(− E˜ /T˜∞ )/(4Pr T˜∞ ) . (3.51)

The boundary conditions required to solve this equation can be found by matching the

inner and outer solutions as

ζ=0 : θ=0 ; ζ →∞ : ∂θ /∂ζ → 0 , θ → aT (3.52)

For the case of LeF close to unity, Eqn. (3.2) is modified to

ζ 2 (∂2 θ /∂ζ 2 ) = − D˜ a (ζ − θ )exp(θ − αζ ) . (3.53)

44
3.5.2 Hot jet flowing into a cool ambient (T0 > T∞)

(With reference to Section 3.4 Ignition Analysis III) For the case of a hot jet issuing

into a cold ambient, any ignition will occur near the jet centerline, η = 0. Fig.3.2

shows a schematic of the inner and outer regions for the scenario.

ζ= ∞
Outer
Region

Inner
T∞ Region
x
Air
y

η= ∞
0
Fuel
To
Fig. 3.2: Schematic of inner and outer
regions for hot jet, cool ambient case.

Moreover, because the jet will be cooled by the cold ambient gas along the flow,

ignition is expected to occur near the slot exit. The analysis is similar to that in

Section 3.1, except that ε = T˜02 / E˜ and the inner variables are defined as

ζ = η PrY˜O,∞ /ε , ξ = Y˜O,∞ ( x˜ −1)/(3ε ) , (3.54)

leading to

∂2 θ ∂θ 1 ∂2 φO ∂φ
− 2 + −2 O = 0 , (3.55)
∂ζ 2 ∂ξ LeO ∂ζ 2 ∂ξ

45
∂2 θ ∂θ
− 2 = − D˜ a ( φO + LeO ζ 2 + ξ )exp[θ − β (ζ 2 + ξ )] , (3.56)
∂ζ 2 ∂ξ

with the initial and boundary conditions

θ (ξ = 0) = φO (ξ = 0) = 0 ,

ζ = 0 and ζ → ∞ , ξ > 0 : ∂θ /∂ζ = ∂φO /∂ζ = 0 ,

where β = ( T
0 − T
∞ ) / Y
O ,∞ and the reduced Damköhler number is

D˜ a = ε[ Da Y˜F,0 /( T˜0 Y˜O,∞ )]exp(− E˜ /T˜0 ) . (3.57)

Ignition is considered successful when the heat generation through reaction is

sufficient to compensate the heat loss from the jet to the ambient at any location, and

the ignition criterion is given by

(∂θ /∂ξ )ζ = 0 ≥ β or ∂θ /∂ζ ≥ 2 βζ at any ξ . (3.58)

46
Chapter 4: Results and Discussions

4.1 Recapitulation

Limits of spontaneous ignition were identified as functions of slot width, flow

rate, and temperatures of the hydrogen jet and ambient air for the two scenarios of a

cool jet flowing into a hot ambient and a hot jet flowing into a cool ambient.

Specifically, Sections 3.2 (for the cool jet case) and 3.4 (for the hot jet case) were

referenced for the analysis of hydrogen due to its low local LeF of 0.6. Equations

balancing diffusive terms with reaction terms were obtained for the cool jet case

(Eqn. 3.50), and equations balancing the transverse diffusive, streamwise convective

terms and reaction terms were obtained for the hot jet case (Eqn. 3.56). The solutions

from the equations are discussed in greater detail in the following sections.

4.2 Cool jet flowing into a hot ambient (T∞ > T0)

Equations (3.50) and (3.53), subject to Eqns. (3.51) – (3.52), were solved by a

fourth order Runge-Kutta method (using computer codes). The results are shown in

Fig. 4.1, a plot of the reaction temperature increase (aT) versus reduced Damköhler

number. This reveals the lower and middle branches of an S-curve [5]. In each such

curve, there is a maximum value of D˜ a above which a solution does not exist. For

values of D˜ a smaller than this ignition D˜ a , there are two solutions. This represents

the transition from weak reaction to vigorous burning, and is defined as the ignition

state. The lower branch, showing an increase of temperature with higher reaction rate,

47
is the physically realistic branch. The middle branch represents conditions that are not

physically possible, owing to the negative slope. Spontaneous ignition is predicted for

any D˜ a greater than this critical value.

Three curves are included in Fig. 4.1, each with a different value of α , where

~ ~ ~
α = (T∞ − T0 ) / YF , 0 (4.1)

10

6
aT

4
α = 0.005 0.01 0.02 0.03

0
0 0.01 0.02 0.03 0.04 0.05


Da

Fig. 4.1: aT versus Da

Da
for varying α (which is changed by
variations in T˜0 or Y˜F,0 ) , with LeF = 0.6 and constant T
∞ =
0.0123, hot ambient. Definitions of Da
and aT are in Section
Da
3.2(Part C-2).

~
Fig. 4.1 indicates that a decrease in α reduces the critical Da at ignition, which

means that ignition is favored. Such a decrease can be accomplished either by

increasing the reactant mass fraction in the fuel supply, Y˜F,0 , or by increasing the jet

temperature, T˜0 . Both findings are physically realistic. A fuel such as hydrogen

48
(which has LeF of 0.6 or less in mixtures with N2, is hence more ignitable as

compared to fuels such as isooctane and methane (which have higher LeF, under the

same conditions). Table 1 lists the typical values of α and qF for hydrogen and 2 other

fuels.

Properties/Fuel Hydrogen (H2) Methane (CH4) Iso-octane (C8H18)

Typical α 0.0064 0.024 0.024
qF (kJ/g) 120.9 54.8 46.4
Table 1: Values of α and specific heat of combustion (qF) from [4].

Parameter α also can be changed by variations in the ambient temperature, T˜∞ , but

~
this changes Da simultaneously. To investigate the effects of T˜∞ variations at fixed

~
Da requires a rescaling. The rescaling is performed here by specifying a reference

value of T˜∞ as Tˆ∞ , defining rescaled parameters:

εˆ = ( Tˆ∞ )2 / E˜ , (4.2)

aˆT = ( T˜∞2 /Tˆ∞2 )aT and (4.3)

~
~ ~ ~
Dˆ a = DaYO ,∞ ~
x 4 / 3 [YF ,0 /(εˆ ~
x 1 / 3 )](1− LeF ) exp(− E / Tˆ∞ ) /(4Pr Tˆ∞ ) (4.4)

and plotting the results in terms of rescaled variables aˆT and Dˆ a . The results are

shown in Fig. 4.2. Here an increase in T˜∞ , which increases α without changing Dˆ a , is

seen to favor ignition. This also is physically realistic because more heat is transferred

to the cold fuel flow at a higher rate when the ambient is at a higher temperature. By

the same reason, when the kinetic data are unchanged, an increased T˜∞ yields ignition

to occur nearer the edge of the jet. Note that the Damköhler number shown is a

49
function of the axial distance from the virtual origin of the jet (see Eqn. 2.50).

10

6
aˆT

2
α = 0.03 0.02 0.01 0.005

0
1E-9 1E-7 1E-5 1E-3 1E-1 1E1
D̂a
Fig. 4.2: Rescaled plot of aˆT versus Da (rescaled D~a
D̂a
values due to changes in T˜∞ ) for varying α (which is
changed by variations in T˜∞ ), with LeF = 0.6,
T˜0 = 0.00226, E˜ = 1.79, Y
F ,0 = 1, and α = 0.01 as
reference, hot ambient.

The influence of reactant Lewis number on ignition is considered in Fig. 4.3. Here

LeF is defined as the mixture thermal diffusivity divided by the mass diffusivity of

fuel into the mixture. A decrease in LeF for fixed Dˆ a is seen to favor ignition. This

occurs because a smaller LeF implies that fuel species diffuse more quickly into the

hot oxidizer. A fuel such as a mixture of hydrogen and nitrogen has a small LeF, is

hence easier to ignite than other fuels of higher LeF at the same conditions. Nayagam

and Williams [23] found that in a one-dimensional model of steady motion of edges

of reaction sheets, increasing the Lewis number decreases the propagation velocity at

small Damköhler numbers.

50
 10

6
aT

4
Le F = 0.6 0.8 1

0
0 0.004 0.008 0.012
Da
Fig. 4.3: aT versus Da
D̂a
(rescaled D~a values due to
D̂a
changes in LeF) for varying LeF, with α = 0.02
T˜0 = 0.00226 , E˜ = 1.79, hot ambient.

4.2.1 Ignition states that separate the ignitable and non-flammable regions

Plots of ignition Dˆ a versus α under several conditions for the cool jet scenario

are shown in Figs. 4.4 - 4.6. In Fig. 4.4, ignition Dˆ a versus α is plotted for varying

LeF, with LeF =1 as the reference. The plot shows that as the LeF decreases, ignition is

favored (similar to explanation for Fig. 4.3) for decreasing α (at fixed T˜∞ ), resulting

in a larger ignitable region. This is consistent with the findings shown in Fig. 4.2.

51
 0.03
Non-flammable

0.02

Le F = 0.6 0.8 1
α

0.01

Ignitable

0
0 0.005 0.01

ˆ a
DDa
ˆ
Fig. 4.4: α (which is changed by
˜ ~
variations in T˜0 or
YF,0 ) versus Da (rescaled Da values due to changes
D̂a
in LeF) for varying LeF, with LeF = 1 as reference and
constant T
∞ = 0.0123, hot ambient.

In Fig. 4.5, ignition Dˆ a versus α is plotted for varying LeF. The plot shows that as

the LeF decreases, ignition is favored (similar to explanation for Fig. 4.3) for

increasing α (due to increases in T˜∞ ), resulting in a larger ignitable region, which is

physically realistic.

In Fig. 4.6, ignition Dˆ a versus α is plotted for varying nF. The plot shows that as

the nF increases, ignition is favored due to an increased reaction rate, resulting in a

larger ignitable region. The effect of increasing nF on ignition, however, is weaker as

compared to the effects of reducing LeF or increasing T˜∞ .

52
 0.03

0.02
Ignitable
1
α

0.8
Le F = 0.6
0.01 Da T˜0
D̂a
Non-flammable

0
0.0001 0.001 0.01 0.1 1
Da
D̂a Davariations in T˜∞ ) versus
Fig. 4.5: α (which αis changed by
~ D̂a
Da (rescaled Da values due to changes in T˜ ) for varying
∞
D̂a
LeF, and T˜0 = 0.00226, hot ambient.
0.03

Ignitable

0.02
α

n F = 0.8
0.01
n F = 1.2
Non-flammable

nF = 1

0
1E-6 1E-4 1E-2 1E+0
D̂a
Fig. 4.6: α (which is changed by variations in T˜∞ )
versus Da (rescaled D~a values due to changes in T˜∞ )
D̂a
for varying no at LeF = 0.6, and T˜0 = 0.00226, hot
ambient.
53
4.3 Hot jet flowing into a cool ambient (T0 > T∞)

Eqns. (3.55) – (3.56) were solved by the Crank-Nicholson method and the

resulting matrix was inverted by LU decomposition (using computer codes). Selected

results are shown in Figs. 4.7 and 4.8. In Fig. 4.7, θmax represents the maximum

value of temperature increase through reaction before ignition occurs. The

corresponding ignition location, X, is shown in Fig. 4.8.

Three curves are included in Figs. 4.7 and 4.8, each with a different value of β,

where

β = ( T
0 − T
∞ ) / Y
O ,∞ (4.5)

2.0

1.8

1.6

1.4
max

1.2
θ

1.0 0.3
0.2
0.8
β =0.1
0.6

0.4
0.4
00 11 2 3


Da
Da
Fig. 4.7: θ max versus Da

for varying β (which is
Da
changed by variations in T˜∞ or Y˜O,∞ ) at Leo = 1, and
constant T˜0 = 0.0388, hot jet.

54
 2

1.5

X
1 β = 0.1 0.2 0.3

0.5

0
00 1
1 2
2 3
3


Da
Da
Fig. 4.8: X versus Da
Da


for varying β (which is
changed by variations in T˜∞ or Y˜O,∞ ) at Leo = 1, and
constant T˜0 = 0.0388, hot jet.

These curves do not have the shape of an S-curve because the solutions are

derived from a partial differential equation (since it is dependent on both transverse

and streamwise stretched coordinates), in contrast to the ordinary differential equation

of the cool jet (which is only dependent on the transverse diffusive term). On each
~
curve, by increasing the reaction rate, Da , a smaller temperature increase and a

shorter ignition location is observed before ignition, as is reasonable to expect. A

~
higher value of Da yields an increased heat generation rate, which compensates for

some heat loss from the hot jet to the cold ambient, favors ignition, and moves the
~
point of ignition closer to the jet exit. In contrast, a reduction in Da weakens the

reaction and makes ignition more difficult such that both θmax and X increase.

55
Although an increase of X provides longer residence time for the reaction so that
~
ignition can occur at a smaller Da , the reaction rate decreases with X because the jet

is cooled by the cold ambient, as can be seen from the reaction term of Eq. (3.56). A
~
sharp increase in θmax and X , as shown the low Da side of the curves in Figs. 4.7

and 4.8, means that the reduction of reaction rate dominates over the residence time
~
increase, and defines the smallest Da for which ignition occurs.

~
Figures 4.7 and 4.8 also indicate that a decrease in β for any fixed Da favors

ignition, as ignition occurs at a lower temperature increase, θmax, and at a shorter

ignition location, X. More importantly, a decrease in β permits ignition at a lower

~
value of Da . Such a decrease can be accomplished either by increasing the reactant

mass fraction in the oxidizer supply, Y˜O,∞ , or by increasing the ambient temperature,

T˜∞ . Both findings are physically realistic.

Parameter β also can be changed by variations in the jet temperature, T˜0 , but this

~ ~
changes Da simultaneously. To investigate the effects of T˜0 variations at fixed Da

requires a rescaling similar to that performed for the cool jet scenario. The rescaling

is performed here by specifying a reference value of T˜0 as Tˆ0 , defining rescaled

parameters :

εˆ = ( Tˆ0 )2 / E˜ , (4.6)

θˆmax = ( T˜0 /Tˆ0 )2θmax and (4.7)

Dˆ a = εˆ [Da Y˜F,0 /( Tˆ0 Y˜O,∞ )]exp( − E˜ /Tˆ0 ) (4.8)

56
and plotting the results in terms of rescaled variables θˆmax and Dˆ a . The results are

shown in Figs. 4.9 and 4.10. Here an increase in T˜0 , which increases β without

changing Dˆ a , is seen to favor ignition because ignition can occur at a lower reaction

rate, or lower Dˆ a . This also is physically realistic. Ignition is predicted to occur near

the centerline if the fuel is hotter than the air because this is where the highest

temperature is attained.

2.0

1.8
β = 0.3 0.2 0.1
1.6

1.4
θˆmax

1.2

1.0

0.8

0.6

0.4
0.001 0.01 0.1 1
Da
D̂a
θˆ
Fig. 4.9: Rescaled plot of max versus Da (rescaled
~ D̂a
Da values due to changes in T˜0 ) for varying
β (which is changed by variations in T˜0 ), with Leo =
1, T
∞ = 0.0358, E˜ = 1.79, Y˜O,∞ = 0.029, and β = 0.1 as
reference value, hot jet.

57
 2

1.5

1 β = 0.3 0.2 0.1

X

0.5

0
0.001 0.01 0.1 1 10
Da
D̂a
Fig. 4.10: X versus Da (rescaled D~a values due to
D̂a
changes in T˜0 ) for varying β (which is changed by
variations in T˜0 ), with Leo = 1, T
∞ = 0.0358,
E˜ = 1.79, Y˜O,∞ = 0.029, and β = 0.1 as reference value,
hot jet.

The effects of oxidizer Lewis number on spontaneous ignition are considered in

Fig. 4.11. This plot shows that decreased LeO makes ignition more difficult. For an

increase in the mass diffusivity of the oxidizer (or a decreased LeO) at a fixed value of

~ ~
Da , θmax increases. In addition, the minimum Da for ignition increases with

decreased LeO. This differs from the ignition behavior with respect to fuel Lewis

number in the cool jet case (Fig. 4.3). In a cool jet, there is unlimited heat transfer

from the hot ambient gas to preheat the fuel so that a higher fuel diffusion rate (lower

58
LeF) results in a higher fuel concentration in the reaction region, more heat generation

through the reaction and, hence, easier ignition.

In the hot jet, only limited heat is available from the fuel flow. An increased

oxidizer mass diffusivity increases the transport rate of oxidizer to the center of the

jet, thus requiring more heat to preheat the oxidizer, decreasing the temperature in the

hot zone, and making ignition more difficult. Furthermore, unlike the cool jet case,

the Lewis number of the oxidizer only has a weak effect on the ignition state (see Fig.

4.11) because ignition occurs near the jet exit if successful. In the reaction region, the

flow velocity is high such that streamwise convection dominates over transverse

diffusion. Moreover, because LeO is close to unity for oxygen in air, the effect of LeO

in a hot-hydrogen cold-air system is secondary. As in Fig. 4.10, the effect of LeO on

the ignition location is similar to that on the ignition state, shown in Fig. 4.12.

1.8

1.6

1.4
max

1.2
θ

1
0.9
1
Le o = 1.1
0.8

0.6
0 1 2


Da
Da

for varying Leo, with
Fig. 4.11: θ max versus Da
Da
˜
β = 0.3 and constant 0 = 0.0388, hot jet.
T
59
 2.0

1.6

1.2

0.9
0.8 1

Le o = 1.1

0.4
0 1 2


DaDa
Fig. 4.12: X versus Da
for varying Leo, with
Da
˜
β = 0.3 and constant T0 = 0.0388, hot jet.

4.3.1 Ignition states that separate the ignitable and non-flammable regions

Plots of ignition Dˆ a versus β under several conditions for the hot jet scenario are

shown on Figs. 4.13 – 4.14. In Fig. 4.13, ignition Dˆ a versus β was plotted for

varying no and constant T˜0 . The plot shows that as the oxidizer reaction order

increases, ignition is favored, resulting in a larger ignitable region, which is

physically realistic.

In Fig. 4.14, ignition Dˆ a versus β was plotted for varying no and T˜0 . The plot

shows that as the no and T˜0 increases, ignition is favored, resulting in a larger

ignitable region covering lower values of Dˆ a .

60
 0.3

Non-flammable

0.2

n o = 1.1 1 0.9
β

0.1
Ignitable

0
0 0.5 1 1.5 2


Da
Fig. 4.13: β (which is changed by variations in T˜∞ or
Y˜O,∞ ) versus Da for varying no, and constant T˜0 =


Da
0.0388, hot jet.
0.3

Ignitable

0.9

0.2 1
n ο = 1.1
β

0.1
Non-flammable

0
1E-4 1E-2 1E+0
D̂a
Fig. 4.14: β (which is changed by variations in T˜0 )
versus Da (rescaled D~a values due to changes in T˜0 )
D̂a
for varying no, with Leo = 1, T
∞ = 0.0358, E˜ = 1.79,
Y˜O,∞ = 0.029, and β = 0.1 as reference value, hot jet.

61
Chapter 5: Conclusions

5.1 Conclusions

The spontaneous ignition of a hydrogen jet emanating from a slot into air has been

considered analytically. A similarity solution of the flowfield was obtained for the

scenario of cool jet flowing into hot ambient, which was then combined with the

species and energy conservation equations. Solutions were found using activation

energy asymptotics.

The analysis yielded limiting conditions for spontaneous ignition of fuel jets. For

a cool jet flowing into a hot ambient, ignition is found to be a strong function of

ambient temperature and fuel Lewis number. Ignition was favored by an increase in

ambient temperature or a decrease in Lewis number. For the hot jet scenario, ignition

was significantly affected by the jet temperature, but only weakly affected by the

oxidizer Lewis number.

Because spontaneous ignition is very sensitive to temperature, ignition is

predicted to occur near the edge of the jet if the fuel is cooler than the air and on the

centerline if the fuel is hotter than the air.

The value of the mixture fraction Z at which ignition occurs can be extracted from

σ YF + (Yo,∞ − Yo )
the AEA solutions as Z = .
σ YF ,o + Yo,∞

62
 In the first scenario of cool jet flowing into a hot ambient, ignition occurs at

Z
0 since it occurs at the jet edge where Y
O → Y
O ,∞ and YF is very small and of the

order ε. In the second scenario of the hot jet, ignition occurs at the jet centerline,

where Z
1 since Y
O → 0 and YF → YF ,0 .

The present model can be extended to studies of flame extinction and to circular

jet configurations. When experimental data becomes available, parametric

comparisons can also be made to establish reaction rates for use in the present model.

63
 Bibliography

[1] E.C. Magison, Electrical Instruments in Hazardous Locations, 2nd Edition,

Instrument Society of America, Pittsburg, Pennsylvania, pp.304, 1972.

[2] J. Hord, Is Hydrogen a Safe Fuel?, International journal of hydrogen energy,

3:157, 1978.

[3] R.C. Weast, CRC Handbook of Chemistry and Physics, CRC Press, West Palm

Beach (1979).

[4] A.M. Kanury, Introduction to Combustion Phenomena, Gordon and Breach

Publishers, Amsterdam (1975).

[5] F.A. Williams, Combustion Theory, 2nd Edition, Addison-Wesley Publishing

Company, Redwood City, CA, pp.76-84, 1985.

[6] N. Peters, Turbulent Combustion, Cambridge University Press, Cambridge,

England, pp.178-186, 2000.

[7] A. Liñán, F.A. Williams, Fundamental Aspects of Combustion, Vol. 34 of

Oxford Engineering Science Series, Oxford University Press, New York, pp.57-

81,1993.

[8] H.G. Im, S.R. Lee, C.K. Law., Ignition in the Supersonic Hydrogen/Air Mixing

Layer with Reduced Reaction Mechanisms, AIAA-1994-548 Aerospace

Sciences Meeting and Exhibit, 32nd, Reno, NV, Jan 10-13, 1994

[9] H.G. Im, B.H. Chao, J.K. Bechtold, C.K. Law, Analysis of Thermal Ignition in

the Supersonic Mixing Layer, AIAA J 32, pp.341-349, 1994.

64
[10] S.R. Lee, S.H. Chung, On the structure of Hydrogen Diffusion Flames with

Reduced Kinetic Mechanisms, Combust. Sci. Tech., 96, pp.247.

[11] B.T Helenbrook, C.K. Law, Ignition of Hydrogen and Oxygen in Counterflow at

High Pressures, Proc. Combust. Inst., pp.815-822, 1996.

[12] H.G. Im, B.T Helenbrook, S.R. Lee, C.K. Law, Ignition in the supersonic

hydrogen/air mixing layer with reduced reaction mechanisms, J. Fluid Mech.,

vol 322, pp.275-296, 1996.

[13] X.L. Zheng, C.K. Law, Ignition of Premixed Hydrogen/Air by Heated

Counterflow under Reduced and Elevated Pressures, Combust. Flame, pp.136-

168, 2004.

[14] V.V. Toro, A.V. Mokhov, H.B. Levinsky, M.D. Smooke, Combined

Experimental and Computational Study of Laminar, Axisymmetric Hydrogen-air

Diffusion Flames, Proc. Combust. Inst. 30:485, 2005.

[15] M. Chaos, R.-H Chen, E.J. Welle, W.L., Roberts, Fuel Lewis Number Effects in

Unsteady Burke Schumann Hydrogen Flames, Combust. Sci. Tech. 177:75,

2005.

[16] Y., Liu, P., Pei , Asymptotic analysis on autoignition and explosion limits of

hydrogen-oxygen mixtures in homogeneous systems, International J. Hydrogen

Energy , Vol. 31, No. 5, pp.639-647, 2006.

[17] F.L. Dryer, M.Chao, Z. Zhao, J.N. Stein, J.Y. Alpert, C.J. Homer, Spontaneous

Ignition of Pressurized Releases of Hydrogen and Natural Gas into Air,

Combust. Sci. Tech., v179, pp.663-694, 2007.

65
[18] C.K. Law, Combustion Physics, C.K. Law, Cambridge University Press, New

York, pp.313-315, 2006.

[19] R.W. Bilger, Combust. Flame, 30, pp.227-228, 1977.

[20] M.H. Holmes, Introduction to Perturbation Methods, p1-5 ,Springer, New York,

1994.

[21] G. Damköhler, Zeit.Elektrochem. 42, 846, 1936.

[22] D.A. Frank-Kamenetskii, Diffusion and Heat Transfer in Chemical Kinetics,

New York, Plenum Press, 64-72, 1969.

[23] D.R Kasoy and F.A. Williams, Phys Fluids 11, 1343, 1968.

[24] D.R Kasoy, M.K., Liu and F.A. Williams, Phys Fluids 12, 265, 1969.

[25] F.E. Fendell, J. Fluid Mech.21, 281, 1965.

[26] H. Schlichting, Boundary-Layer Theory, Seventh Edition, McGraw-Hill, Inc.,p.

127-200, 1979.

[27] W. Bickey, The Plane Jet, Phil. Mag. Ser. 7, 23: 727, 1939.

[28] V. Nayagam, F.A. Williams, Lewis-number effects on edge-flame propagation,

Journal of fluid mechanics, 458: 219-228, 2002.

66
 ABSTRACT

Title of Document: FLAME QUENCHING LIMITS OF

HYDROGEN LEAKS

Christopher W. Moran, Master of Science

in Fire Protection Engineering, 2008

Directed By: Assistant Professor, Peter B. Sunderland,

Department of Fire Protection Engineering

This study examines flame quenching limits of hydrogen leaks in compression

fittings and tube burners. Experimental work is presented. Measurements included

ignition limits for leaking compression fittings on tubes of 3.16-12.66 mm in diameter

and the ignition and quenching limits of tube burners with diameters of 0.15 –

0.56 mm. Minimum ignition flow rates of 0.028 mg/s for hydrogen, 0.378 mg/s for

methane, and 0.336 mg/s for propane were found in the compression fitting

experiments. The upstream pressure does not play a role in the ignition flowrate

limit. The minimum quenching limits of hydrogen found in tube burners were 2.1

and 3.85 μg/s in oxygen and air, respectively. These correspond to heat release rates

of 0.252 and 0.463 W, respectively, the former being the weakest observed flame

ever.
 FLAME QUENCHING LIMITS OF HYDROGEN LEAKS

By

Christopher W. Moran

Thesis submitted to the Faculty of the Graduate School of the

University of Maryland, College Park, in partial fulfillment
of the requirements for the degree of
Master of Science
2008

Advisory Committee:
Assistant Professor Peter B. Sunderland, Chair
Professor Marino di Marzo
Associate Professor Arnaud Trouvé
 © Copyright by
Christopher W. Moran
2008

ii
 Acknowledgements

This work was supported by NIST (Grant 60NANB5D1209) under the

technical management of J. Yang. I also very much appreciate the support that was

offered me through the Clark School of Engineering Distinguished Graduate

Fellowship that allowed me to work and study here.

I would like to express my gratitude to my advisor Dr. Peter Sunderland, for

his guidance, support, and encouragement. Never before have I met a professor as

enthusiastic about teaching and helping his students. It has been my privilege.

I also thank Dr. James Milke, for guiding me through the program and

involving me in the department. His enthusiasm and guidance while I considered the

pursuit of this degree was invaluable.

I thank my advisory committee, Dr. Arnaud Trouvé and Dr. Marino di Marzo

for their time, input and help.

The faculty of the Department of Fire Protection Engineering has been

exceptional for their help and support. I would not have been able to get through my

years here without them.

I would also like to thank my colleagues, for their assistance, reminders, and

friendship. Specifically, Richard Axelbaum and Michael Butler of Washington

University were instrumental in the completion of this work. Beei-Huan Chao (of

University of Hawaii) assisted with the choked-flow calculations.

Lastly, I would like to thank my parents for their encouragement and support.

I never would have been able to have the college experience that I did without them.

ii
Table of Contents
Acknowledgements....................................................................................................... ii
List of Tables ............................................................................................................... iv
List of Figures ............................................................................................................... v
Chapter 1: Introduction ................................................................................................ 1
1.1 Hydrogen Chemistry........................................................................................... 1
1.2 Hydrogen Economy ............................................................................................ 2
1.3 Hydrogen Fire Safety.......................................................................................... 4
1.4 Objectives ........................................................................................................... 5
Chapter 2: Flame Quenching Limits of Leaky Compression Fittings ......................... 7
2.1 Experimental Background and Procedure .......................................................... 7
2.1.1 Experimental Introduction ........................................................................... 7
2.1.2 Fittings ....................................................................................................... 10
2.1.3 Flow System ............................................................................................... 13
2.1.4 Lab Safety................................................................................................... 14
2.1.5 Procedures ................................................................................................. 15
2.1.6 Sources of Error......................................................................................... 17
2.2 Results............................................................................................................... 18
2.2.1 Ignition Flowrate Limits ............................................................................ 19
2.2.2 Effects of Fitting Orientation ..................................................................... 24
2.2.3 Effects of Leak Type ................................................................................... 25
2.2.4 Effects of Tube Diameter ........................................................................... 26
Chapter 3: Weak Flames............................................................................................ 30
3.1 Experimental Background and Procedure ........................................................ 30
3.1.1 Experimental Introduction ......................................................................... 30
3.1.2 Burners....................................................................................................... 31
3.1.3 Flow System ............................................................................................... 32
3.1.4 Procedures ................................................................................................. 35
3.1.5 Sources of Error......................................................................................... 36
3.2 Results............................................................................................................... 37
3.2.1 Ignition Flowrate Limits ............................................................................ 38
3.2.2 Effects of Orientation on Ignition Flowrate Limits.................................... 39
3.2.3 Quenching Flowrate .................................................................................. 41
3.2.4 Weakest Flame Ever Observed .................................................................. 43
Chapter 4: Conclusions .............................................................................................. 46
4.1 Compression Fittings ........................................................................................ 46
4.2 Weak Flames..................................................................................................... 47
Appendix A: Fuel Properties ..................................................................................... 49
References................................................................................................................... 50

iii
List of Tables

Table 2.1. Compression Fitting Burner Summary ..................................................... 10

Table 2.2. Round Orifice Diameters .......................................................................... 23
Table 3.1. Tube Burner Summary.............................................................................. 32
Table 3.2. Volumetric Flowrates and Exit Velocities................................................ 34
Table 3.3. Ignition Flowrate Limits for Burner Heights and Oxygen Velocities. ..... 38
Table 3.4. Quencing Flowrates and Heat Release Rates. .......................................... 41
Table A.1. Fuel Properties ......................................................................................... 49

iv
List of Figures
Fig. 2.1 Image of Fitting Assembly 11
Fig. 2.2 Image of Weak Flame on Platinum Burner 12
Fig. 2.3 Hydrogen Flow System Schematic 14
Fig. 2.4 Image of Fitting Assembly and Gas Collection Device 17
Fig. 2.5 Image of Flames on Burner 8L 19
Fig. 2.6 Image of Flames on Burner 4L 19
Fig. 2.7 Image of Flames on Burner 2L 19
Fig. 2.8 Plot of mfuel versus Upstream Pressure 21
Fig. 2.9 Plot of Vfuel versus Upstream Pressure 22
Fig. 2.10 Plot of Leak Diameter versus Upstream Pressure 24
Fig. 2.11 Plot of mfuel versus Orientation 25
Fig. 2.12 Plot of mfuel versus Leak Type 26
Fig. 2.13 Plot of mfuel versus Burner Diameter 27
Fig. 2.14 Image of Hydrogen Flames 28
Fig. 2.15 Image of Methane Flames 28
Fig. 2.16 Image of Propane Flames 29
Fig. 3.1 Image of Coflow Burner 33
Fig. 3.2 Hydrogen Flow System Schematic 34
Fig. 3.3 Oxygen Flow System Schematic 35
Fig. 3.4 Plot of mfuel versus Diameter 39
Fig. 3.5 Plot of mfuel versus Orientation 40
Fig. 3.6 Plot of Quenching mfuel versus Diameter 42
Fig. 3.7 Plot of Power versus Diameter 43
Fig. 3.8 Image of Weakest Flame 44
Fig. 3.9 Image of Weak Flame on Platinum Burner 45

v
Chapter 1: Introduction
1.1 Hydrogen Chemistry
Hydrogen is a very unusual fuel. It is the lightest fuel and has very wide flammability limits, 4-

75% by volume (Kanury, 1975). Hydrogen has the lowest quenching distance (0.51 mm), the

smallest ignition energy of any fuel in air (0.028 mJ), the lowest auto-ignition temperature of any

fuel ignited by a heated air jet (640 ºC), the highest laminar burning velocity of any fuel in air

(2.91 m/s), and the highest heat of combustion (119.9 MJ/kg) (Kanury, 1975). Hydrogen flames

are the dimmest of any fuel. Hydrogen also embrittles many metals more than any other fuel.

Despite these unusual aspects of hydrogen, it is attractive as an energy carrier. It can be

produced from water and electricity. It can power fuel cells or engines with only water vapor as

exhaust. Its combustion produces no carbon dioxide.

Hydrogen may not be any more intrinsically hazardous from a fire safety standpoint than

gasoline or diesel. However, over a century of experience with these traditional fuels in

widespread vehicle use has resulted in good fire safety records. Further research is necessary if

hydrogen vehicles are to be introduced with a similar safety record.

One of the most attractive aspects of using hydrogen as an alternative fuel source is that there are

no carbon products produced. No greenhouse gases are produced during the reaction. The lack

of carbon in the reaction eliminates the main source of radiation and visible light as compared to

hydrocarbon flames (Bregeon et al., 1978). This makes detecting hydrogen flames especially

difficult. The flames are also typically hotter than hydrocarbon flames as the adiabatic flame

temperature is approximately 2400 K (Turns, 2000).

Although hydrogen leaks can be extremely dangerous, there are aspects of hydrogen that are

1
safer than other combustible gases. Because hydrogen is so light, it rises at 20 m/s, much faster

than most combustible gases and 6 times faster than natural gas (DOE, 2007). Hydrogen also

diffuses 3.8 times faster than natural gas (DOE, 2007). This makes it more difficult for the gas

to collect and reach flammable concentrations in open spaces. If the gas is not contained, it

diffuses too quickly to reach explosive concentrations. It is much more likely that the hydrogen

leak would be ignited near the source and form a flame as opposed to forming an explosive

mixture.

Hydrogen can cause steel embrittlement and permeation leaks. While permeation leaks are

generally not flammable, they do present a challenge in the storage of hydrogen. Hydrogen's low

molecular weight motivates its storage at higher pressures (or liquification). Due to hydrogen's

low molecular weight, it has the highest volumetric leak propensity of any fuel.

1.2 Hydrogen Economy

As gasoline prices climb and concerns grow about the extensive release of greenhouse gases into

the atmosphere, hydrogen has emerged as one of the leading alternative fuel sources. Oil

currently supplies 33% of the world’s primary energy (Marban and Valdes-Solis, 2007).

Hydrogen as an energy carrier can help mitigate concerns about fossil fuel consumption

(Yamawaki et al., 2007). Some benefits of hydrogen are that it can be converted to electricity

with a high efficiency, its products of combustion are mainly water, it can be stored as liquid, gas

or solid, and it can be transported using pipelines, tankers or rail trucks over long distances

(Sherif et al., 2005). It is also projected that hydrogen can be produced efficiently using nuclear

sources or renewable methods such as wind (Von Jouanne et al., 2005). Hydrogen, however, is

also unique in that it has unusual fire hazards that must be taken into account before widespread

2
use of this fuel can be considered safe.

The United States government has experienced itself over the past few years to moving towards

a hydrogen fueled economy. In 2003, President Bush announced the Hydrogen Fuel Initiative to

accelerate the research and development of hydrogen for the use in transportation (DOE, 2007).

It was thought that widespread use of hydrogen could help reduce the dependence on foreign oil

as well as reduce pollutants. In 2005, the Energy Policy Act was created to reinforce

government support for hydrogen and alternative fuel technology (DOE, 2007). The Advanced

Energy Initiative was enacted in 2006 to again help accelerate research with the potential to

reduce near term oil use and advance activities under the Hydrogen Fuel Initiative (DOE, 2007).

Presently, however, there are few codes or standards to regulate the hydrogen transportation

industry. These codes and standards are currently being experienced.

Hydrogen is being considered as a fuel source for transportation purposes largely owing to its

lack of carbon emissions after combustion. Veziroglu and Babir (1992) argue that hydrogen is

one of the safest fuels owing to its high diffusivity and buoyancy. In case of a leak, hydrogen

will be released into the atmosphere as opposed to gasoline, which will pool and presents a

longer danger. Hydrogen also radiates much less heat as it does not produce carbon compounds.

This means that people are rarely burned by a hydrogen flame unless they are in the flame.

Lovins (2003) compares hydrogen to gasoline and natural gas as an energy source and states that

"a good fuel cell system is about 50-70% efficient, hydrogen to electricity, while a typical car

engine's efficiency from gasoline to output shaft averages only about 15-17%."

Bossel and Eliasson (2002) believe that the upstream energy costs of a hydrogen economy have

not been adequately assessed. The energy necessary to form hydrogen either through electrolysis

or the conversion of another gas does not make it efficient enough to use as an energy carrier.

3
They state that "the production, packaging, storage, transfer, and delivery of the gas are so

energy consuming that other solutions must be considered."

1.3 Hydrogen Fire Safety

Owing to its low molecular weight, hydrogen is often stored at high pressures. It is predicted that

hydrogen will need to be stored at up to 40 MPa to be an effective vehicle fuel (Takeno et al.,

2007). Hydrogen’s safety hazards resulted in a Department of Energy report (Cadwallader and

Herring, 1999) finding that hydrogen containment was the chief safety concern associated with

using hydrogen as a transportation fuel. One characteristic of hydrogen that makes it more prone

to risk is that it is easier to ignite than most other common fuels. The minimum ignition energy

for hydrogen is an order of magnitude lower than those for methane and propane (Ge and Sutton,

2006). This means that a small spark might be able to ignite a hydrogen leak, whereas the same

spark might not be able to ignite a methane or propane leak. It has also been noted that hydrogen

flames have weak luminosity (Cheng et al., 2005), hence a hydrogen leak sustaining a flame

would be difficult to detect by the human eye.

Swain and Swain (1992) did a comparison study of hydrogen, methane and propane fuel leakage

in a residential setting. Only pressures less than 0.965 bar (14 psi) were used in the study. In

their research, they found three times more hydrogen volume than methane would escape

through a leak. As relatively high leakage rates can lead to combustible fuel clouds, they

discovered that propane produces a cloud much faster than hydrogen or methane.

Hydrogen is odorless and colorless, which makes detecting hydrogen leaks extremely difficult.

Odorants are not used with hydrogen as no known odorant is light enough to diffuse as fast as

hydrogen and odorants poison fuel cells. Hydrogen is nontoxic although death can occur due to

4
asphyxiation if too much oxygen is displaced.

Hydrogen's unique fire hazards present a challenge to firefighters. Currently, firefighters are

taught to detect hydrogen flames with a straw broom. Thermal imaging firefighting cameras are

effective but not all departments have them. When near a possible hydrogen leak, firefighters

listen for leaking gas, look for heat shimmering, and use their detection methods. If a leak is

found, they are taught to stop the flow if possible. If they are unable to stop the flow, then the

flame is allowed to consume the gas supply if safe. Extinguishing the flame without stopping the

leak can result in an explosive mixture.

1.4 Objectives
Codes and standards are currently being experienced for hydrogen system fire safety. One issue

of interest is permissible leak rates. However, prior to the present work no measurements had

been performed to establish the maximum leak rate for nonflammable conditions.

Thus motivated, the objects of this work are:

1. Measure the minimum flowrates that are necessary to support flames from compression

fittings in air. Three different gases were used in the experiments; hydrogen, methane

and propane.

2. Identify the minimum ignition and quenching hydrogen flowrates possible for any choice

of burner and oxidizer. Determine whether this produces the weakest flame ever

observed.

The results of this work are expected to be helpful in the creation of the codes and standards

governing the safe use of hydrogen and in the field of micro-combustion. The data found during

these experiments gives the ignition flowrate limits for leaky compression fittings, which can be

5
used when deciding the minimum allowable leaks in a commercial application. Micro-

combustion is a growing field as electronics get smaller and batteries make up most of the device

weight. A lighter, more efficient (more environmentally friendly) power supply can be found

using micro-combustors (Federici, 2006).

6
Chapter 2: Flame Quenching Limits of Leaky Compression Fittings
2.1 Experimental Background and Procedure
As many high pressure systems are made with compression fittings, it is important to determine

the possible leakage scenarios and the ignition flowrate limits associated with these fittings.

Three different fuels were used to determine the ignition flowrate limits for a number of different

possible failure scenarios. The first, hydrogen, is being considered for use in alternative fuel

source applications but many of the dangers are still unknown. The second gas is methane. This

gas is commonly used as a fuel in many applications and its properties are better known. In

many cases, it is known as natural gas and is used to heat homes and power vehicles. Methane's

advantage over other hydrocarbons is that it produces less CO2 when combusted. The third gas

used is propane. Propane is most commonly used for cooking on grills and portable stoves. It is

also used in refrigeration and as an automotive fuel.

Although the main purpose of the experimentation was to find the flame quenching limits for

hydrogen, the same tests were also done on methane and propane to have a comparison of the

different fuels as they are used in many of the same applications.

2.1.1 Experimental Introduction

Quenching and blowoff limits bound the leak flowrates that can support combustion. Matta et al.

(2002) found that a propane flame is not able to exist when its predicted length is less than the

measured standoff distance. Experiments verified this analytical method of finding the

quenching flowrate by establishing a propane flame over a hypodermic, stainless steel tube and

decreasing the fuel flowrate until extinction. It was also found that the flowrate at quenching is

practically independent of the tube diameter. Work in this laboratory has extended this work to

hydrogen fuel and diverse burners (Butler et al., 2008).

7
There are three classifications for laminar jet flames: diffusion controlled flames (e.g., Burke and

Schuman, 1928), diffusion and momentum controlled flames (e.g., Roper et al., 1977) and micro

diffusion flames controlled by momentum and diffusion (e.g., Ban et al., 1994). Extensive

research has been done on the first two types, but only a limited amount has been done on micro

diffusion flames. Baker et al. (2002) devised a flame height expression for purely diffusion

controlled flames capable of accurately predicting micro-slot diffusion flame heights. Useful

parameters for characterizing dominant flame mechanisms were defined: a diffusion-buoyancy

and a diffusion-momentum parameter. This work extended the investigation of

Roper et al. (1977) to smaller slot sizes. Ban et al. (1994) investigated flames established on

circular burners with inner diameters of 0.15, 0.25, and 0.40 mm. The experiments worked to

verify the predicted flame shapes of laminar flames for three fuels: ethane, ethylene and

acetylene. The work found that buoyancy effects are negligible for small flames; nearly spherical

flames were established that were unaltered by rotating the burner. An in depth analysis of a

micro diffusion hydrogen flame was performed by Cheng et al. (2005); their numerical solution

for species was compared with experimental data. It was found that buoyancy effects were

insignificant. Nakamura et al. (2006) numerically studied methane micro diffusion flames on a

circular burner of diameter less than 1 mm. They found that small flames have the same, nearly

spherical structure as those in microgravity, citing weak buoyancy forces. Also, the existence of

a minimum flame size necessary for combustion was predicted.

Lee et al. (2003) conducted leak rate experiments on micro-machined orifices of different sizes

and shapes. They examined the differences in flowrates among square and elliptical slit orifices.

In almost every case, the flow was choked in the orifice, which caused large underestimates in

the flow using the helium signature test. Schefer et al. (2006) also presented equations to

8
calculate choked flow, as well as subsonic laminar and turbulent flows. They considered leaks

owing to pressure-driven convection and permeation through metals.

Research has been conducted on hydrogen interface leakage in national pipe thread (NPT)

fittings by Ge and Sutton (2006). They found that the best threaded fittings investigated leaked

hydrogen with leak rates of 1 μg/s. Leak rates under non-ideal conditions were far higher. Ge

and Sutton (2006) found that a larger tightening torque is not as important in sealing the threads

as the Teflon material and properties. The tests were run at a pressure drop of 70 bar. They

determined that two wraps of Teflon experiences had better performance than one wrap and the

Swagelok™ anaerobic pipe thread sealant out performed the Teflon.

Compression fittings are typically used with gases at high pressure. Compression fittings are a

reliable, easy to use method that allows fittings to be taken apart and attached with ease. A large

benefit of using compression fittings over NPT fittings is the lack of Teflon tape that is necessary

to prevent leakage. Any time an NPT fitting is taken apart, it must be cleaned and rewrapped

before it can be used again.

Swagelok has made available product test reports that give information on the leakage of their

fittings. Product test report (PTR) 396 shows for 144 samples of 6 and 12 mm fittings tested at

310 and 200 bar nitrogen, respectively, there were no detectable leaks found after 10 minutes. In

PTR-865, the fittings were tested again at 512 and 312 bar nitrogen, respectively, and leaked less

than one bubble per minute during the 10 minute test period. Leaks were detected by

submerging the fittings under water and watching for bubbles. Using hydrogen instead of

nitrogen would dramatically increase any leak rates.

9
2.1.2 Fittings
The compression fitting experiments were conducted using threes types of leaky fittings of three

different diameters. Each leaky fitting was made by attaching a steel tube to a Swagelok union

and capping the other end of the union. The leaks occurred at the connection between the steel

tube and the union. A sample leaky fitting can be seen in Fig. 2.1. Compression fittings with

outside tube diameters of 3.16, 6.33 and 12.66 mm were used. The leaks were caused using

three different methods: loosened, over-tightened, and scratched fittings. These methods were

chosen as possible leak modes that compression fittings used in commercial applications would

experience. All of the fittings used can be found in Table 2.1.

Table 2.1. Leaky fittings used in compression fitting experiments

Tube Diameter
Fitting Number Description
in mm
Fitting 8L 1/8 3.16 Correctly made, then loosened
Fitting 4L 1/4 6.33 Correctly made, then loosened
Fitting 4T 1/4 6.33 Correctly made, then over-tightened
Fitting 4S 1/4 6.33 Correctly made, then scratched
Fitting 2L 1/2 12.66 Correctly made, then loosened
.

10
 Figure 2.1. An example fitting assembly is shown. (A) Stainless steel plug for Swagelok
tube fitting. (B) Swagelok union. (C) Swagelok nut. (D) Steel tube.

Figure 2.2 shows a cross-sectional view of a compression fitting. The steel tube, nut, fitting

body and ferrules are shown. Leak paths seen in the experiments are shown with the dotted line.

The leak travels from the tube exit, around the front and back ferrule, and exits between the nut

and the tube. Leaks were not seen coming from threads of the compression fitting.

11
 Figure 2.2. Cross-sectional view of the fitting connection with proposed leak path. (A)
Steel tube. (B) Swagelok nut. (C) Back ferrule. (D) Front ferrule. (E) Fitting body.

Loosened Fittings
This method simulates a fitting that has not been retightened properly and then pressurized.

Each fitting was made using the manufacturer's directions and confirmed to be leak free at

6.89 bar (100 psi) using hydrogen. If no leaks were present, the fitting was then loosened until

the ignition flowrate limit was found. This was done at several different pressures to show that

there is no change in minimum flowrate necessary for combustion for different pressures.

Over-tightened Fittings
This method simulates a fitting that has been over-tightened. Over-tightening damages the

ferrule in the compression fitting, allowing leaks to occur. This method was done only for the

6.33 mm fitting. It is difficult to over tighten a 12.66 mm fitting as it is so large that a high

torque is required. The fittings were made using the manufacturer's directions and confirmed to

be leak free at 6.89 bar (100 psi) using hydrogen. The fitting between the steel tube and union

was taken apart and then retightened to finger tight. The fitting was then tightened one full turn

12
(3/4 turn past manufacturer's instructions). The fitting was then taken apart again and tightened

until the ignition flowrate limit was found for several different pressures.

Scratched
This method simulates a ferrule being damaged by insertion into a fitting and was done only for

the 6.33 mm fitting. The fittings were made using the manufacturer's directions and confirmed

to be leak free at 6.89 bar (100 psi) using hydrogen. The ferrule on the steel tube was then

scratched using a small triangular file. The fitting was then reassembled properly and the

ignition flowrate limit was found by slowly tightening the fitting.

2.1.3 Flow System

Methane and propane were used in the experiments as well as hydrogen. Each test was

completed using all three fuels so that comparisons could be made. Hydrogen, methane, and

propane were tested in pressure ranges of 1-135 bar, 1-100 bar, and 1-7 bar, respectively.

Propane could not be tested at pressures higher than 7 bar as this is its vapor pressure at room

temperature.

The fuel flow system is shown in Fig. 2.3. For safety reasons, the flow was passed through a

filter and relief valve to keep the downstream pressure below 6.89 bar in the event of any failure

of the bottle pressure regulator. When experiments were done at pressures higher than this, the

filter and relief valve were temporarily taken out of the system. The upstream pressure was

controlled by a pressure regulator located just before the leaky fitting.

13
 A – Gas Cylinder G – Ball Valve
B D
C E B – Valve H – Pressure Regulator

F C – Pressure Regulator I – Rubber Stopper

D – Needle Valve J – Leaky Fitting

L J E – Filter K – Open Tube

I F – Relief Valve (6.89 bar) L – Bubble Meter

H G
A K

Figure 2.3. Experimental setup of flow system.

2.1.4 Lab Safety

Fire safety is one of the primary concerns with the use of hydrogen. Methane and propane

exhibit many normal characteristics of hydrocarbon flames, such as high radiant energy and

brightly visible flames. Hydrogen on the other hand does not have these characteristics. Larger

hydrogen flames can be seen in a dim room but flames near the quenching limit are not visible

even in a darkened room. Due to this, the following methods of flame detection were compiled

to allow for safe flame detection.

1. Hot Plume Check with Thermocouple: This is the safest method and the method that was

primarily used during these experiments. A thermocouple is slowly inserted above the

flame region to check for a change in temperature owing to the hot plume. This method

allows the researcher to get close to the flame region without burning. At the quenching

limits, temperature rises on the order of 15-50 ºC were observed. A Fluke 179 with

thermocouple probe was used to measure temperature. The thermocouple was a type K.

2. Thin Paper Check: This method involves slowly inserting a piece of paper into the flame

region and looking for ignition. This method works better for larger flames as small

14
 drafts can extinguish the smaller flames. Also, this method should not be used when

combusting in a pure oxygen environment as any paper in the oxygen will burn violently.

As the automatic detection of hydrogen is difficult and there are no odorants available with

hydrogen, it was important to frequently check for leaks in the system. Excess hydrogen buildup

in an enclosed space can lead to an explosion. To prevent this, the lab was well ventilated so that

any hydrogen buildup would have been difficult. Also the flowrates used during the experiments

were so small that a potential explosive buildup was very unlikely.

2.1.5 Procedures
A Nikon D100 camera with a 50 mm lens was used to photograph the different experiments. In

most cases, extender rings were used to allow closer photographs of the experiments. The

camera was connected to a computer and controlled using the Nikon capture control software.

This allowed many camera settings to be controlled on the computer and direct transfer of

pictures from the camera.

Before each test, the system was pressurized and tested for leaks. This was done by squirting a

small amount of soap water on each fitting. If any bubbles were seen, the fitting was tightened

until the leak stopped. The only exception to this was the fitting between the steel tube and

union that was purposely made to leak. This detection method allowed for quick discovery of

leaks as well as being an accurate detector of small leaks.

To determine the minimum pressure and flow necessary for a sustained flame, the fitting was

tightened and the pressure was slowly raised until a flame occurred. The upstream pressure in

the system was controlled by the pressure regulator located before the fitting. Ignition of the fuel

was caused by a butane lighter held briefly to the fitting. It was considered to be a sustained

15
ignition if the flame lasted a minimum of 10 s. After a successful ignition of any flame, the steel

tube and fitting were allowed to cool to room temperature before the next test was attempted.

This kept the gas from being preheated in the fitting. Also, as the metal fitting heats, it expands,

causing the leak path to shrink and changing the minimum flowrate. Allowing the fitting to cool

between tests minimized these sources of error.

For each test, the fitting was imaged against a black background to make the flames more visible.

Flames for the propane and methane tests were visible to the naked eye and ignition was

determined visually. This was backed up by the use of a thermocouple to determine if a flame

was present. The hydrogen flames were too small and dim to be seen by eye and therefore only

the thermocouple was used to determine if a flame existed. The thermocouple was placed

slightly off and above the flaming region so that it would not interfere with the flame. When the

hydrogen gas ignited, an audible pop was heard that signaled ignition. In some cases however,

the fuel ignited but the flame was immediately extinguished. Several fitting orientations were

tested to see the effect fitting orientation had on the ignition flowrate limit. Flames were never

observed inside the fitting owing to the quenching limits of each fuel. Each test was repeated to

establish repeatability.

To measure the flowrate from the leaky fittings, an apparatus was built to capture the escaping

gas. This can be seen in Fig. 2.4. An open tube was created that fit over the leaky fitting

assembly and created an air tight seal that allowed the gas to be collected and sent to a bubble

meter. All flames were extinguished and the fittings were allowed to cool before any flow

measurements were taken. As the gas traveled to the bubble meter, it was bubbled through water

to give the gas 100% relative humidity. This was done so that partial humidification in the

bubble meter did not interfere with the results. A humidity adjustment of 2.645% was then made

16
when calculating the flowrate (Incropera, 2002). In the bubble meter, the time for the bubble to

travel 10 cc was clocked and recorded.

Figure 2.4. The fitting assembly and the gas capture apparatus.

Equation 2.1 shows the conversion of gas volumetric flow to mass flow.

pMW
M = 0.97355Q (2.1)
Ru T
where M is the mass flowrate, Q is the volumetric flowrate, p is ambient pressure, MW is fuel

molecular weight, Ru is the universal gas constant, and T is ambient temperature. Measurements

for the laboratory temperature and pressure were measured when tests were run. Ambient

pressures were obtained online (www.weather.com).

2.1.6 Sources of Error

There is some uncertainty with each flow measurement. The volumetric flowrate was timed

using a stopwatch. Most flows were small enough that being slightly off would not affect the

measurement significantly. All flows were measured multiple times so that the results could be

17
averaged and uncertainty lowered.

One issue in the lab was the inability to control the airflow. The ventilation in the lab space

caused air currents and it was not possible to close or turn off the air circulation system. A draft

curtain was deployed around the experiment to try to minimize these air currents. These air

currents were sometimes problematic when attempting to find the ignition limit flowrates. Care

was taken to limit the air currents around the experiments but it is possible that lower ignition

flowrates could be obtained in a more controlled environment.

2.2 Results
Figure 2.5 shows that the hydrogen flames at quenching are much smaller than the methane and

propane flames. The hydrogen flames were also much dimmer than the other flames.

Figures 2.5, 2.6 and 2.7 show images of hydrogen, methane, and propane flames on leaky

compression fittings in the vertical orientation for the different sized fittings. The images were

recorded slightly above the quenching limits. The hydrogen flame is significantly smaller than

the methane and propane flames in each figure. This is due to the smaller flowrate as hydrogen

has a lower quenching limit. The quenching distance of hydrogen is approximately a quarter of

the distance of methane and propane. This allows the flame to burn much closer to the metal.

As the fitting size increases, in can be seen that the flame size also increases.

The flames were isolated to one side of the fitting where the most damage to the ferrule occurred.

This is most evident in the hydrogen photo as the flame is much smaller. The photos below are

from the loosened fittings. These flowrates were slightly larger than those from the scratched

and over-tightened fittings. This could be seen from the increased flowrate in Fig. 2.12.

18
 Figure 2.5. Color images of Fitting 8L flames for each fuel. Each fuel is flowing at the
minimum flowrate possible for ignition. The camera settings for hydrogen were 20 s,
F1.4, ISO 1600. The camera settings for methane/propane were 1 s, F1.4, ISO 800.

Figure 2.6. Color images of Fitting 4L flames for each fuel. Each fuel is flowing at the
minimum flowrate possible for ignition. The camera settings for hydrogen were 20 s,
F1.4, ISO 1600. The camera settings for methane/propane were 1/10 s, F1.4, ISO 800.

Figure 2.7. Color images of Fitting 2L flames for each fuel. Each fuel is flowing at the
minimum flowrate possible for ignition. The camera settings for hydrogen were 20 s,
F1.4, ISO 1600. The camera settings for methane/propane were 1/10 s, F1.4, ISO 800.

2.2.1 Ignition Flowrate Limits

Figure 2.8 shows the measured ignition flowrates for hydrogen, methane, and propane for the

vertical orientation. The minimum flowrate necessary for sustained ignition is plotted versus

pressure. For each fuel, the measurements at increased pressures are associated with an increase

19
in torque on the fitting. The upper limit on pressure for propane is lower than that of the other

gases because the vapor pressure of propane at room temperature is 9.1 bar (142 psia).

Within experimental uncertainties, the ignition limits of Fig. 2.8 are independent of pressure for

each fuel. This indicates that, as expected, at a fixed fuel mass flowrate the upstream pressure

has little or no effect on the velocity profile of the jet entering the surrounding air. The mean

hydrogen flowrate, 28 μg/s, is about an order of magnitude lower than for the other fuels owing

to its low quenching distance and low molecular weight. The mean methane and propane

flowrate limits were 378 and 336 μg/s, respectively. The quenching distance of hydrogen is

approximately one fourth that of methane and propane and its molecular weight is an order of

magnitude smaller than that of both propane and methane. This explains why the hydrogen

flowrate is an order of magnitude smaller than that of both methane and propane. Butler et al.

(2007, 2008) found the same behavior in their experiments using round burners. This is an

important finding in that it shows that the upstream pressure does not play a role in the ignition

flowrate limit.

20
 0.4

0.378 mg/s

Minimum Flowrate (mg/s)

0.3

0.336 mg/s

Hydrogen
0.2 Methane
Propane
h
m
p
0.1 0.028 mg/s Linear (h)
Linear (m)
Linear (p)

0
1 10 100 1000
Pressure (bar)
Figure 2.8. Ignition mass flowrate limit versus upstream pressure in the vertical
orientation.

The measured ignition mass flowrates of Fig. 2.8 were converted to volumetric flowrates using

the measured temperature and pressure of the laboratory that day. The resulting volumetric

flowrates are plotted in Fig. 2.9 with respect to upstream pressure. Within experimental

uncertainties, the minimum fuel volumetric flowrates of Fig. 2.9 are independent of pressure.

Propane requires the lowest volumetric flowrate for ignition while methane requires the highest.

The volumetric flowrates for hydrogen, methane and propane are 0.337, 0.581, and 0.187 mL/s,

respectively. While hydrogen has the lowest mass flowrate necessary for ignition, propane has

the lowest volumetric flowrate for ignition.

21
 0.7

0.6

Minimum Flowrate (mL/s)

0.581 mL/s Hydrogen
0.5
Methane
0.337 mL/s Propane
0.4
H
M
0.3 P
Linear (M)
0.2 Linear (H)
Linear (P)
0.187 mL/s
0.1

0.0
1 10 100 1000
Pressure (bar)
Figure 2.9. Ignition volumetric flowrate limit versus upstream pressure in the
vertical orientation.

The pressure and flowrate measurements at the ignition limits can provide insight into the

associated leak paths. For simplicity, it is assumed here that the leak paths are choked round

orifices. Using equations from Munson et al. (2006) for isentropic compressible flow, the orifice

areas and diameters were found. Equation (2.2) was used to find the static pressure P in the

orifice, where P0 is the stagnation (upstream) pressure, and k is fuel specific heat ratio.

k
P ⎛ 2 ⎞ k −1
=⎜ ⎟ (2.2)
P0 ⎝ k + 1 ⎠
Equation (2.3) was used to find the static temperature in the orifice, T, where T0 is the stagnation

(upstream) temperature. As there is a decrease in static pressure, there is a decrease in static

temperature in the orifice. This static temperature was used to find the speed of sound, c, for the

fuel using Eq. (2.4) where Ru is the universal gas constant and MW is the fuel molecular weight.

T 2
= (2.3)
T0 k + 1
Ru Tk
c= (2.4)
MW

22
After finding the static pressure and temperature in the orifice, the density of the fuel, ρ, can be

found using the ideal gas law using Eq. (2.5). The orifice area can be found from Eq. (2.6),

where M is fuel mass flowrate at the ignition limit. The minimum and maximum calculated leak

areas and leak diameters are given in Table 2.2.

PMW
ρ= (2.5)
Ru T
M = ρcA (2.6)

Table 2.2. Round orifice diameters assuming choked flow

Upstream Speed of Mass Flowrate Leak Area Leak Diameter
Fuel
Pressure (Bar) Sound (m/s) (mg/s) (μm2) (μm)
344.7 0.687 0.936
Hydrogen 1193.1 0.028
4.14 57.3 8.54
206.8 5.81 2.72
Methane 416.4 0.378
4.14 290.5 19.2
6.89 103.3 11.5
Propane 244.2 0.336
4.14 172.1 14.8

Figure 2.10 shows a plot of leak diameter for the minimum ignition flowrate versus the upstream

pressure assuming chocked flow. The diameter of the leak decreases sharply with an increase in

pressure. Hydrogen requires the smallest leak diameters to reach its minimum ignition flowrate

while methane requires the largest leak diameter. The curves in Fig. 2.10 span the typical ranges

of fuel pressures for vehicle tanks.

23
 100
1 10 100 1000

Methane

Leak Diameter (um)

10
Propane

Hydrogen
1

0
Upstream Pressure (bar)

Figure 2.10. Leak diameter for minimum ignition flowrate versus upstream
pressure assuming choked flow.

2.2.2 Effects of Fitting Orientation

Figure 2.11 shows the effect of fitting orientation on the ignition flowrate limit for Fitting 4L.

The orientation of the leak does not have an effect on the flowrate of hydrogen owing to its lower

quenching limit and ease of ignition. Fitting orientation did have an effect on propane and

methane with the horizontal configuration requiring the highest flowrate and inverted orientation

requiring the lowest. The inverted orientation required the lowest flowrate of each fuels as this

kept the flame away from the fitting so that less heat was lost to the surrounding metal. The

horizontal and vertical orientations gave a large surface to absorb the heat from the flame

causing a larger flowrate to be necessary for sustained ignition. This can be seen in Fig. 2.5

where the flames are directly below the compression fitting and lose a significant portion of their

energy through convection and radiation.

24
 0.4

Vertical
0.3

Minimum Flowrate (mg/s)

Horizontal
Inverted

0.2

0.1

0
Hydrogen Methane Propane

Figure 2.11. Minimum flaming flow rate for Fitting 4L in vertical, horizontal and
inverted orientations.

2.2.3 Effects of Leak Type

Figure 2.12 shows the effect of the different leaky fittings on the ignition flowrate limits. The

loosened fittings had the highest flowrate for ignition among the leakage configurations. The

flowrates for the loosened fittings had a 26% higher flowrate than the over-tightened

configuration. The flowrates for the over-tightened and scratched fittings were comparable for

hydrogen. Only a slight difference was found and this is within the experimental uncertainties.

Methane and propane followed the same trend with the loosened fitting having the highest

flowrate and the scratched fitting having the lowest.

The trends for the methane and propane ignition flowrates are as expected. In the loosened

fitting, there is no damage to the ferrule and therefore no easy leakage path for the gas. This

means that the gas exits over a wider area surrounding the tube so that a larger flowrate is

necessary to sustain the flame. The next higher flowrate was from the over-tightened fitting.

25
This method damages the ferrule, allowing for a small leak. This leak can spread before exiting

the ferrule, making it more difficult to ignite. However, the damage is confined to a small area

on the ferrule so that leak will not spread as much as the loosened fitting. The smallest flowrate

came from the scratched fitting. The leak was caused by filing a straight path from one end of

the ferrule to the other. This gave a very specific leak path for the gas to follow. As the gas

exited the fitting in one location, it was less spread out than the other two configurations. This

allowed for a smaller ignition flowrate limit. The less the gas spreads before exiting the fitting,

the smaller the flowrate necessary for ignition.

0.5

Loose
0.4
Overtightened
Minimum Flowrate (mg/s)

Scratched

0.3

0.2

0.1

0
Hydrogen Methane Propane

Figure 2.12. Minimum flaming flowrate for each gas with different kinds of leaks.

2.2.4 Effects of Tube Diameter

Figure 2.13 shows a comparison of flowrates against fitting size. The trends seen are as expected

for each fuel. The smallest flowrate came from the smallest fitting while the largest flowrate

came from the largest fitting. The leak types for each size fitting were the same so that the only

difference was the fitting diameter. It can then be expected that the smaller fittings would need a

26
smaller flowrate for ignition as there is less surrounding metal to transfer heat away from the

reaction. Each larger fitting size has four times as much surface area as the fitting below it

allowing for more heat loss to the surroundings. For the most part, it is approximately a 20%

increase in flowrate as the fitting size doubles. The only difference was a 50% increase in

flowrate for hydrogen and propane when going from fitting 8L to 4L. Also, in the smaller

fittings, the gas will leak in a smaller area allowing for a higher concentration of fuel. For the

larger fittings, the gas has a larger leakage area, making for a smaller concentration of fuel.

0.6
3.1 mm Fitting
6.3 mm Fitting
0.5
12.6 mm Fitting
Minimum Flowrate (mg/s)

0.4

0.3

0.2

0.1

0
Hydrogen Methane Propane

Figure 2.13. Minimum flaming flowrate for each gas at differently sized
compression fittings.
Hydrogen flames in Fig. 2.14 are all similar regardless of the fitting size. The flowrates for each

fitting diameter were similar in size and would correspond to flames of approximately the same

size. The flames were not visible to the naked eye and an extended exposure at a high ISO was

necessary to obtain a picture of them. Each flame is approximately 1.35 mm in diameter.

27
 Figure 2.14. Color images of hydrogen flames in vertical position for different sized
fittings. Each flame is at its minimum flowrate. The camera settings are 20 s, F1.4, ISO
1600.
In Fig. 2.15 and 2.16 the methane and propane flames are all approximately the same size even

though Fig. 2.13 showed the flowrate increased with increasing fitting size. As the flame width

and height are the same, there could be an increase in flame depth that is not readily visible

through the image. The flowrates also are possibly not large enough that an increase in flame

size would be visible. Each flame is approximately 3.85 mm wide and 3.85 mm from the flame

tip to the base.

Figure 2.15. Color images of methane flames in vertical position for different sized
fittings. Each flame is at its minimum flowrate. The camera settings are 1/10 s, F1.4,
ISO 800.

28
Figure 2.16. Color images of propane flames in vertical position for different sized
fittings. Each flame is at its minimum flowrate. The camera settings are1/10 s, F1.4,
ISO 800.

29
Chapter 3: Weak Flames
3.1 Experimental Background and Procedure
Weak flames have an important role in the field of micro-combustors. These weak flames hold

the ability to replace electrical igniters. The ability to produce weak but stable flames can

increase the turndown ratio of micro-combustors. The weak flames are also important in

fundamental combustion models. If these simple flames can be modeled correctly, then the

results can be expanded to the extinction behavior of larger, more complex flames.

3.1.1 Experimental Introduction

Ronney et al. (1998) did experiments in microgravity using weak fuel pre-mixtures to produce

flame balls. A flame ball is a steady, spherical flame in a premixed gas. Fuel and oxygen diffuse

in while the combustion products and heat diffuse out. Ronney et al. (1998) found that the flame

balls released 1.0-1.8 W per ball. Later, flame balls with energy releases as low as 0.5 W were

recorded (Philips, 2003). At the time, they were the weakest flames ever recorded.

Butler et al. (2007, 2008) did experiments on curved wall, pinhole and tube burners measuring

quenching flowrates. They found that the lowest flowrates occurred for tube burners and had a

minimum for burners with an internal diameter of approximately 0.15 mm. They later showed

that tube burners in an inverted configuration allowed for the smallest necessary flow to sustain

combustion.

Hydrogen is being used in micro-electrical-mechanical systems (MEMS) that are found in many

commercial applications. Hydrogen is the fuel of choice for these systems owing to its high

heating value, rapid rate of vaporization, fast diffusion velocity, short reaction time, and high

flame speed (Yang et al., 2002). Yang et al. (2002) found that hydrogen fuels are 24 times more

30
powerful than state of the art lithium batteries of the same size. Hydrogen is also cheaper,

provides a more constant voltage, has no memory effect and instant recharge capability (Yang et

al., 2002). Zhang et al. (2006) also noted that the quenching distance of hydrogen is much

shorter than that of hydrocarbons and the combustion of hydrogen is more stable than other fuels.

Zhang et al. (2006) performed experiments where hydrogen was mixed with hydrocarbons to

lower the quenching distance and increase the flame speed of the mixture. They found that

hydrogen was able to ignite but that pure methane and ethane were not unless platinum was

present. The main issues when working with micro-combustion are the thermal and radical

quenching effects. The high surface-to-volume ratio of micro-combustion devices poses

challenges to researchers (Chen et al., 2007). These issues can be reduced or eliminated by

increasing the wall temperature or preventing heat losses to the wall (Fernandez-Pello, 2002).

As the increase of the surface-to-volume ratio of the combustor becomes a problem for gas-phase

combustion, it begins to favor catalytic combustion. This reaction is typically slower than a gas-

phase reaction and heat loss is still a problem, but the increase in surface area and lower

temperatures of the catalytic reaction may allow easier implementation (Fernandez-Pello, 2002).

These catalytic systems are typically easier to start, self-sustaining at leaner fuel/air ratios and

can be designed with no moving parts (Federici et al., 2006).

3.1.2 Burners
Experiments were performed to attempt to observe the weakest flame observed to date. Tube

burners were used as Butler et al. (2007, 2008) found that tube burners in the inverted position

allowed for the smallest flowrates with sustained combustion. The three burners used during the

experiment can be found in Table 3.1. Steel hypodermic tubes (Small Parts, Inc.) were obtained

31
with an inside diameter of 0.152 mm. Platinum hypodermic tubes (Goodfellow Cambridge

Limited) were obtained with an inside diameter of 0.16 mm. Graphite/polyimide ferrules (Small

Parts, Inc) were used to attach the steel and platinum tubes to a 3.16 mm Swagelok fitting. This

allowed for easy switching of burners as well as leak-proof connections. Each burner was tested

at 5.52 bar (80 psi) with hydrogen for leaks. The ferrules maximum operating pressure was

6.89 bar (100 psi). Soap water was placed along the connections to test for leaks.

Tests were run with an inverted burner (downward discharge) using air as an oxidizer as well as

pure oxygen. Burner SS56 was used during these tests. It was important to determine the

advantage of burning hydrogen in a pure oxygen environment over combustion in ambient air.

Table 3.1. Tube burners used during smallest flame tests. The steel tubes were 152 mm
in length and the platinum tube was 100 mm in length.
Name Material ID (mm) OD (mm)
SS56 Stainless Steel 304 0.559 1.57
SS15 Stainless Steel 316 0.152 0.305
PT16 Platinum (99.95% Pure) 0.16 0.4

3.1.3 Flow System

As oxygen was found to allow smaller flames, it was primarily used as the oxidizer in the weak

flame experiments. The oxygen used was 99.994% purity from Airgas. To provide a steady,

laminar flow of oxygen to the flame, it was supplied through a coflow burner. Oxygen was run

through a ceramic honeycomb in the coflow burner to make the flow laminar. Tests were run

with oxygen flowing at several velocities from the coflow burner to determine the velocity that

allowed for the minimum ignition and quenching limits. Velocities of 4.5, 9, and 18 cm/s were

used. The base value of 9 cm/s was taken from the literature (Santoro, 1987) and then halved

and doubled to find the most beneficial velocity.

32
To determine the velocity at which the air exited the coflow burner, the volumetric flowrate

measured by a flowmeter was divided by the coflow burner area. The coflow burner had a

diameter of 102 mm. The center of the coflow burner was hollow and can be used to flow fuel.

It was capped off underneath and not used in these experiments. The top of the burner can be

seen in Fig. 3.1. The oxygen flow created a pure oxygen atmosphere above the burner. The

volumetric flowrates and corresponding oxygen velocities can be found in Table 3.2.

UNUSED

HONEYCOMB

Figure 3.1. Color image of top view of coflow burner. Internal diameter of the ceramic
honeycomb is 102 mm and outside diameter of central tube is 16 mm.
In addition to running tests at several different oxygen velocities, the burner height above the

coflow burner was varied to determine the height that allowed for the smallest flowrate. Each

height measurement was made at each oxygen velocity so that a variety of values could be

analyzed. Heights of 10, 20, and 30 mm above the coflow burner were measured.

33
 Table 3.2. Volumetric flowrates and corresponding exit velocities of oxygen.
Volumetric Flowrate (m3/s) Oxygen Flow Velocity (m/s)
3.59E-4 0.045
7.17E-4 0.09
1.43E-3 0.18
The hydrogen flow system (Fig. 3.2) for the smallest flame experiments was almost identical to

the setup used in the compression fitting experiments. The main difference was that a very

sensitive needle valve with very low flow coefficient was placed before the burner to allow fine

changes in flow. This was critical with the small flowrates that were achieved. Experiments

were run at 2.76 bar (40 psi) into the flow valve. The oxygen system (Fig. 3.3) ran from the

pressure regulator on the tank, past a relief valve and through a flowmeter. The flowmeter was

used to determine the velocity of the oxygen exiting the coflow burner. The flowmeter used was

an Omega, model FL-2063-NV.

A – Gas Cylinder G – Ball Valve

B D
C E B – Valve H – Pressure Regulator

F C – Pressure Regulator I – Needle Valve

D – Needle Valve J – Rubber Stopper
K G
M E – Filter K – Leaky Fitting
J F – Relief Valve (6.89 bar) L – Open Tube

L I H M – Bubble Meter

Figure 3.2. Experimental setup of hydrogen flow system.

34
 A – Gas Cylinder
B D B – Valve
C
C – Pressure Regulator
E
D – Needle Valve
E – Relief Valve (6.89 bar)
H F – Ball Valve
G – Flowmeter
F
H – Coflow Burner
G
A
Figure 3.3. Experimental setup of oxygen flow system.

3.1.4 Procedures
After the optimum height and coflow burner speeds were determined, tests were run using

burners SS15 and PT16. Care had to be taken when igniting the flame with the butane lighter as

the flame from the lighter could ignite the stainless steel in burner SS15 in the pure oxygen

atmosphere. The hydrogen flames were too small for a pop to be heard upon ignition as in the

compression fitting experiments. Ignition was determined by placing the thermocouple slightly

to the side and above the burner and looking for a temperature rise. If flames were sustained for

more than 10 s, the test was considered successful. The burners were photographed against a

black background with the lights off for image contrast. Photographs were taken of the weak

flames using the Nikon D100. Most pictures required extended exposure times.

The mass flowrate was found using the same method as the compression fitting experiments.

The gas exiting the burner was captured and run through a bubble meter where the volumetric

flowrate could be measured. Due to the extremely low flowrate, the gas was not bubbled

through water on the way to the bubble meter as in the compression fitting experiments. The

bubbling caused fluctuations in the measured volumetric flowrate. The power heat release rate

was found by multiplying the hydrogen mass flowrate by the lower heating value. The lower

35
heating value was taken from literature as 119.9 kJ/g (Kanury, 1975).

It is expected that not all of the hydrogen flowing from the burner was combusted. This means

that the power measured by using the mass flowrate could be higher than the actual heat release

rate. However, as there was no means to verify this, the power measured is the power output as

if all the hydrogen were converted to water. Tests were repeated to show repeatability.

3.1.5 Sources of Error

The weak flame experiments were also susceptible to similar errors as the compression fitting

experiments concerning the airflow in the laboratory and timing of the volumetric flowrate. The

airflow in the laboratory was not a problem for the experiments in the oxygen environment as the

oxygen velocity negated the air currents. The airflow did still cause issues when combusting in

the ambient air. These sources of error were controlled and minimized as much as possible as

explained previously.

There are several potential sources of error during the experimental measuring of the gas

flowrates. The humidity of the lab is one potential source of error affecting the minimum

ignition and quenching flowrates of the fuels. It has been shown that that humidity has no

significant influence on the minimum ignition energy on a hydrogen-air mixture

(Ono et al., 2007), however this did not show the effect of humidity on ignition or quenching

flowrate limits. When measuring the flowrate for the compression fitting experiments, the gas

was bubbled through water so that the humidity of the gas would not affect the volumetric

flowrate measurement. This was not possible during the weak flame experiments as it caused

too many fluctuations during the measurement.

One issue that arose during the weak flame experiments was the ignition of the SS15 burners.

Several burners caught fire during the experiments and had to be replaced. This occurred while

36
trying to ignite the hydrogen exiting the burner in the pure oxygen environment. This was not a

problem when the burner was first introduced to the pure oxygen environment, but only after

extended periods of time in the oxygen. Ignition of the fuel occurred by a butane lighter, which

burns much hotter when it enters the pure oxygen atmosphere. If the lighter was brought too

close to the burner, there were instances where the burner ignited and was not extinguished until

both the hydrogen and oxygen flows were stopped. There were also several cases where the

burner would glow as if from the hydrogen combustion heating but would continue even after

the hydrogen flow was stopped. This made finding the ignition and quenching flowrate limits

difficult as it was not known whether only a flame existed or a flame with the metal reaction. It

is possible that the quenching flowrate for the stainless steel burner was lower owing to a

chemical reaction between the iron and oxygen. This would allow for a smaller hydrogen

flowrate to be found than would occur without metal oxidation. To avoid this, anytime a

measurement was taken, the flow was backed off afterwards to ensure that hydrogen combustion

was the only reaction taking place. The best method to avoid this was to frequently replace the

stainless steel burners. However, using the methods and tools available, it was impossible to tell

whether the measurements were only due to the flame or a combination of the flame and metal

oxidation on burner SS15. The platinum burner did not have this problem as it is a noble metal

and does not oxidize.

3.2 Results
Data collected from the weak flame experiments was used to investigate the ignition and

quenching limits of hydrogen in both air and oxygen. Hypodermic stainless steel and platinum

tubes were used to observe the weakest flames in the world to date.

37
3.2.1 Ignition Flowrate Limits
The oxygen versus air experiments (done with burner SS56) showed that a smaller quenching

flowrate can be obtained by flowing hydrogen into pure oxygen. It was found that combustion in

pure oxygen gave a 30% reduction in the minimum mass flowrate compared to air. The

hydrogen quenching limit in air was 5.64 μg/s while in oxygen it was 3.97 μg/s. This is expected

as burning in pure oxygen results in higher flame temperatures, which raises the reaction rate.

Table 3.3 shows the different ignition flowrate limits that were obtained at various burner heights

and oxygen velocities. For each oxygen velocity, the highest mass flowrate corresponded to the

highest burner height. This is likely due to air mixing in with the oxygen flow at this height,

especially at the weakest oxygen velocity. Diluting the oxygen flow would increase the

minimum fuel flowrate necessary for ignition.

Table 3.3. Burner height and oxygen velocity versus ignition flowrate limits of
hydrogen for Burner SS15.
Burner Height Mass Flowrate
(mm) (μg/s)
Oxygen Velocity = 45 mm/s
10 5.28
20 5.15
30 6.37
Oxygen Velocity = 90 mm/s
10 5.18
20 5.05
30 5.33
Oxygen Velocity = 180 mm/s
10 4.85
20 5.02
30 5.12
Figure 3.4 shows the ignition flowrate limits of hydrogen found in both oxygen and air. Also

shown are the quenching limits of hydrogen in air found by Butler et al. (2007, 2008). As

expected, the ignition flowrate limits of hydrogen in air are larger than the quenching limits

found by Butler et al. (2007, 2008). A higher flowrate is necessary to have sustained ignition as

there is less preheating of the burner. The minimum ignition flowrate in air was 4.67 μg/s while

38
the quenching limit was 3.93 μg/s. As Fig. 3.4 shows, a 16% higher flowrate is necessary to

have sustained ignition than is needed at the quenching limit. The ignition flowrate limit of

hydrogen in pure oxygen was lower than the hydrogen quenching limit in air. As the flame

temperature is higher in oxygen, the reaction is faster allowing for the smaller flowrate. The

minimum ignition flowrate in oxygen was 2.7 μg/s. This is a 45% lower flowrate than the

quenching limit in air and 73% lower than the ignition flowrate limit in air.

8
Minimum Ignitable Flowrate (μg/s)

5
Oxygen
4 Air
Butler et al. (2008)
3

0
0 0.5 1 1.5 2 2.5
Burner ID (mm)
Figure 3.4. Ignition flowrate limits of hydrogen in air and oxygen with burner in
vertically downward position. Butler et al (2007, 2008) data is the quenching limit in
air.

3.2.2 Effects of Orientation on Ignition Flowrate Limits

Figure 3.5 shows the ignition flowrate limits in both oxygen and air for different tube burner

orientations. The initial hypothesis was that the lowest flowrate would occur when the burner

was facing upwards as there is less surrounding metal to take heat away from the reaction.

However, when the burner is facing downwards, a portion of the heat is being used to preheat the

unburned hydrogen. Thus this becomes a heat-recirculating burner, for which temperatures

39
exceeding the adiabatic flame temperature are possible. Raising the temperature of the hydrogen

supply allows for a higher temperature in the flame and a faster reaction so that a smaller

flowrate can be achieved. The horizontal configuration required the highest flowrate as it does

not preheat the gas as effectively.

There is a difference between the flowrate trends necessary for sustained ignition in air and

oxygen. The lowest flowrate for both comes in the downward orientation. However, the

horizontal orientation has the highest flowrate for ignition in oxygen while the upward

orientation has the highest flowrate for ignition in air. Butler et al. (2007, 2008) found that the

horizontal configuration gave the highest quenching flowrate in air. As the ignition flowrate

limit in oxygen follows this trend, it is likely that the higher minimum ignition flowrate for air in

the upward orientation is due to error, although it is within the range of experimental uncertainty.

6
Oxygen
Minimum Ignitable Flowrate (μg/s)

5 Air

0
Inverted Horizontal Vertical

Figure 3.5. Ignition flowrate limits of hydrogen in oxygen and air for different burner
orientations.

40
3.2.3 Quenching Flowrate
The quenching flowrates for burners SS15 and PT16 can be found in Table 3.4. These were the

only two burners used when attempting to create the weakest, sustained flame. Butler et al.

(2007, 2008) found that the lowest quenching flowrate came from tube burners in the downward

configuration of approximately 0.15 mm internal diameter. From experimentation, minimum

quenching flowrates of 2.1 μg/s were found, which corresponds to a power output of 0.252 W.

Ronney et al. (1998) obtained a minimum power of approximately 1 W in his SOFBALL

experiments, later achieving a power of 0.5 W (Phillips, 2003). These were done in a

microgravity environment and recorded on video. Butler et al. (2007, 2008) found a minimum

heat release rate of 0.55 W in his tube burner quenching experiments. It can be seen that the

quenching limit in oxygen provided a flowrate approximately half that of Butler et al. (2007,

2008) found. This allowed for a heat release rate of 0.252 W, the weakest flame ever observed.

Table 3.4. Hydrogen quenching flowrates in oxygen and corresponding power output.
Burner Quenching Flowrate (μg/s) Power (W)
SS15 2.1 0.252
PT16 2.3 0.276
A slight difference in the measured flowrate for the platinum and stainless steel burner can be

seen. Although there is an 8 μm difference between the internal diameters of the burners, the

main reason that different flowrates were found is attributed to the properties of the burners.

Platinum and stainless steel 316 have thermal conductivities of 71.6 and 16.3 W/m-K,

respectively (Incropera, 2002). Platinum transfers the heat away from the burner tip much better

than stainless steel so that more heat is lost from the reaction region. There is a delicate balance

between the combustion heating and the heat loss to the surroundings. If the heat loss is too

great, the reaction rate will slow and the flame will extinguish.

41
 8

Quenching Flowrate (μg/s)

6

5
Oxygen
4 Air
Butler et al. (2008)
3

0
0 0.5 1 1.5 2 2.5
Burner ID (mm)
Figure 3.6. Quenching flowrates of hydrogen in air and oxygen with burner in vertically
downward position. Butler et al (2007, 2008) data is quenching limit of air.

Figure 3.6 shows the hydrogen quenching limits obtained in both air and oxygen. The data from

Butler et al. (2007, 2008) is also plotted. The quenching flowrates in air were the same within

experimental uncertainties as the Butler et al. (2007, 2008) data. Quenching flowrates of 3.85

and 3.94 μg/s were found in air for burners SS15 and PT16, respectively. The hydrogen

quenching flowrate in oxygen was much lower than that in air. This was expected as the

previous data has shown that the minimum ignition and quenching flowrates are lower in oxygen

than air. There was a 40-45% drop in quenching flowrate from air to oxygen.

42
 1.0

0.9

0.8

0.7

Power (W) 0.6

Oxygen
Air
0.5
Butler et al. (2008)
0.4

0.3

0.2

0.1

0.0
0 0.5 1 1.5 2 2.5
Diameter (mm)
Figure 3.7. Heat release rate in air and oxygen with burner in vertically downward
position. Butler et al (2007, 2008) data is for air only.

Figure 3.7 shows the effects of burner diameter on the heat release rate. The heat release rate is

directly proportional to the mass flowrate from Fig. 3.6. Butler et al. (2007, 2008) showed that

an increase in burner diameter causes the heat release rate to increase owing to an increase in

quenching flowrate. A minimum in terms of flowrate and power occurs at an approximate

internal diameter of 0.15 mm.

3.2.4 Weakest Flame Ever Observed

Representative quenching limit flames from burner SS15 can be seen in Fig. 3.8. These are

images of the world's weakest observed flame to date. The flame diameter is approximately that

of the burner for the H2/O2 reaction. A faint outline of the flame for the H2/air reaction is visible

beneath the burner. It is difficult to maintain the flame at the quenching limit in the H2/air

reaction due to the air currents in the laboratory. The H2/O2 reaction did not have this problem

as the oxygen velocity negated any airflow currents. The glowing in the H2/air reaction is

43
attributed to the higher flowrate exiting the burner.

Figure 3.8. Color image of world’s weakest flame using burner SS15 for hydrogen in
oxygen (left) and air (right). The H2/O2 flame has a power of 0.252 W while the H2/air
flame has a power of 0.463 W. The camera settings were 30 s, F2.8, ISO 800.
Flames from the burner PT16 are shown in Fig. 3.9. These were taken just before the quenching

flowrate. The H2/O2 flame is approximately the same size as the internal diameter of the

platinum burner. The H2/air flame is slightly larger, which would correspond to the higher

quenching flowrate. A slight glow can be seen coming off the burner from the H2/O2 flame.

This could be due to a surface reaction on the burner or the heating of the burner due to the

flame.

44
Figure 3.9. Color image of weak flame using burner PT16 for hydrogen in oxygen (left)
and air (right). The H2/O2 flame has a power of 0.276 W while the H2/air flame has a
power of 0.474 W. The camera settings were 30 s, F1.4, ISO 200.

45
Chapter 4: Conclusions
This was a study of the ignition and quenching flowrates of hydrogen flames from compression

fittings and tube burners using both air and oxygen as oxidizers. Methane and propane were also

used as fuels in the compression fitting experiments to give a scale to the hydrogen limits and

find their ignition limits. This work is expected to be useful in the creation of hydrogen safety

codes and in the field of micro-combustors.

4.1 Compression Fittings

The minimum flowrates necessary for igniting and sustaining hydrogen, methane, and propane

flames on a leaky compression fitting is 0.028, 0.378, and 0.336 mg/s, respectively. This was

due to hydrogen's lower molecular weight and smaller quenching distance. It was found that the

ignition flowrate limit of each fuel was independent of upstream pressure for pressures of 1-100

bar (1-7 bar for propane).

The leaky fitting orientation has no statistically significant effect on the ignition flowrate limit of

hydrogen. Burner orientation did play a significant role in the minimum flowrate for propane

and methane. The lowest flowrate occurred with the burner in an inverted orientation and the

highest flowrate with the burner in a vertical orientation. Orientation had an effect due to the

amount of surrounding metal that the flames impinged upon. The less heat lost to the

surroundings, the lower the flowrate was.

For each fuel, the leaky fitting with the scratched ferrule had the lowest flowrate necessary for

sustained ignition. This burner had the straightest path for the fuel to flow before exiting the

fitting, which allowed for smaller flowrates. The burners with the loosened fittings had the

highest flowrate necessary for sustained ignition. There was a larger area for the fuel to spread

46
before exiting the fitting, which made for a higher flowrate to sustain ignition.

The ignition flowrate limit increased with increasing burner size. As the burners got larger, the

flames impinged on a wider, thicker surface. As more heat was drawn from the flame, a higher

flowrate was necessary for sustained ignition. Also, the flow was less concentrated at the fitting

exit for the larger burners, which made for larger flowrates.

In every case, the hydrogen flames were too small and dim to see with the naked eye. Photos of

the hydrogen flames required extended exposures at a high ISO. When the lighter was brought

near the burner, a pop typically signaled that ignition of the hydrogen had occurred. The

methane and propane limit flames were visible for every burner but did not omit a popping sound

upon ignition.

Applying a soap water solution is an effective way to check for hydrogen leaks at fittings. Leak

rates above the quenching limits produce visible bubbles in the solution.

4.2 Weak Flames

The weakest flames were obtained when the burner height above the coflow burner was 10 mm

with an oxygen velocity of 0.18 m/s. The optimal burner orientation was downwards. The

ignition and quenching flowrates were lower in a pure oxygen environment than in air. As in the

compression fitting experiments, the hydrogen flame was not visible to the naked eye and could

only be seen using extended exposures with the camera.

This study led to the discovery of the weakest flames observed to date. The weakest flame ever

recorded was found using a stainless steel burner with an internal diameter of 0.152 mm. A

quenching flowrate of 2.1 μg/s was found in a pure oxygen atmosphere. This corresponds to a

power output of 0.252 W, half the size of the previous weakest observed flame. In comparison, a

47
birthday candle flame produces 50 W. The platinum burner with an internal diameter of

0.16 mm gave a quenching flowrate of 2.3 μg/s, which corresponds to a power of 0.28 W. This

is slightly higher than that found with the stainless steel burner and could be the result of a lack

of surface reactions on the burner.

MEMS systems can benefit from these flames owing to the high turndown ratios and use of pilot

flames. These flames have such low flowrates that they can be used as pilot flames, removing

the need for electrical igniters. Using a weak hydrogen flame, especially in an oxygen

atmosphere, would allow these systems to be made even smaller.

48
Appendix A: Fuel Properties

Hydrogen
Molecular Weight 2.016 g/mol
Lower Heating Value 119.9 MJ/kg
Spontaneous Ignition Temperature 571.1 °C
Maximum Flame Speed 291.19 cm/s
Quenching Distance 0.51 mm
Specific Heat Ratio 1.41
Methane
Molecular Weight 16.04 g/mol
Lower Heating Value 50 MJ/kg
Spontaneous Ignition Temperature 632.2 °C
Maximum Flame Speed 37.71 cm/s
Quenching Distance 2.03 mm
Specific Heat Ratio 1.31
Propane
Molecular Weight 44.096 g/mol
Lower Heating Value 46.3 MJ/kg
Spontaneous Ignition Temperature 504.4 °C
Maximum Flame Speed 42.89 cm/s
Quenching Distance 1.78 mm
Specific Heat Ratio 1.15
Table A.1. Values obtained from Appendix C of SFPE Handbook (2002)

49
References
Baker J., Calvert M.E., Murphy D.W. Structure and dynamics of laminar jet micro-slot diffusion
flames. Journal of Heat Transfer 2002; 124:783-790.

Ban H., Venkatesh S., Saito K. Convection-diffusion controlled laminar micro flames.
Transactions of the ASME 1994;116:954-959.

Bossel, U.B., Eliasson, B. The Future of the Hydrogen Economy: Bright or Bleak?. Proc. Of
Fuel Cells World, Lucerne, Switzerland, July 2002.

Bregeon B., Gordon A.S., and Williams, F.A. Near-Limit Downward Propagation of Hydrogen
and Methane Flames in Oxygen-Nitrogen Mixtures. Combustion and Flame, 1978; 33:33-
45.

Burke S.P., Schuman T.E.W. Diffusion flames. Industrial and Engineering Chemical Research.
1928; 20:998-1004.

Butler M.S., Axelbaum R.L., Moran C.W., Sunderland P.B. Flame Quenching Limits of
Hydrogen Diffusion Flames, Eastern States Section of the Combustion Institute,
Charlottesville, 2007.

Butler M.S., Axelbaum R.L., Moran C.W., Sunderland P.B. Flame Quenching Limits of
Hydrogen Leaks, SAE World Congress, Detroit, SAE Paper 2008-01-0726, 2008.

Cadwallader L.C., Herring J.S. Safety issues with hydrogen as a vehicle fuel. Report
INEEL/EXT-99-00522 prepared for U.S. Department of Energy 1999.

Chao Y.C., Chen G.B., Hsu C.J., Leu T.S., Wu C.Y., Cheng T.S. Operational characteristics of
catalytic combustion in a platinum microtube. Combustion Science and Technology,
2004; 176:1755-1777.

Chen G.B., Chen C.P., Wu C.Y., Chao Y.C. Effects of catalytic walls on hydrogen/air
combustion inside a micro-tube. Applied Catalysis A: General, 2007; 332:89-97.

Cheng T.S., Chao Y-C., Wu C-Y., Li Y-H., Nakamura Y., Lee K-Y., Yuan T., Leu T.S.
Experimental and numerical investigation of microscale hydrogen diffusion flames.
Proceedings of the Combustion Institute, 2005; 30:2489-2497.

Federici J.A., Norton D.G., Bruggemann T., Voit K.W., Wetzel E.D., Vlachos D.G. Catalytic
microcombustors with integrated thermoelectric elements for portable power production.
Journal of Power Sources, 2006; 161:1469-1478.

Fernandez-Pello A.C. Micropower generation using combustion: issues and approaches.

Proceedings of the Combustion Institute, 2002; 29:883-899.

50
Ge X., Sutton W.H. Analysis and test of compressed hydrogen interface leakage by commercial
stainless steel (NPT) fittings. SAE International, 2006:35-47.

Incropera F.P., DeWitt D.P. Fundamentals of Heat and Mass Transfer. Hoboken, NJ: Wiley,
2002: 906-907.

Jacobson A.S., Epstein A.H. An informal survey of power MEMS. The International
Symposium on Micro-Mechanical Engineering, December 2003.

Jouanne A., Husain I., Wallace A., Yokochi A. Gone with the wind. IEEE Industry Applications
Magazine July/Aug 2005:12-19.

Kanury A.M. Introduction to Combustion Phenomena. New York, NY: Gordon and Breach,
1975: 130-131.

Lee I.D., Smith O.I., Karagozian A.R. Hydrogen and helium leak rates from micromachined
orifices. AIAA Journal, 2003; 41(3):457-463.

Lovins, A.B. Twenty Hydrogen Myths, Final Report. Rocky Mountain Institute Snowmass, CO,
2003.

Marban G., Valdes-Solis T. Towards the hydrogen economy? International Journal of Hydrogen
Energy 2007; 32:1625-1637.

Matta L.M., Neumeier Y., Lemon B., Zinn B.T. Characteristics of microscale diffusion flames.
Proceedings of the Combustion Institute, 2002; 29:933-938.

Munson B.R., Young D.F., Okiishi T.H. Fundamentals of Fluid Mechanics. Hoboken, NJ:
Wiley, 2006.

Nakamura Y., Yamashita H., Saito K. A numerical study on extinction behaviour of laminar
micro-diffusion flames. Combustion Theory and Modeling, 2006; 10(6):927-938.

Ono R., Nifuku M., Fujiwara S., Horiguchi S., Oda T. Minimum Ignition Energy of Hydrogen-
Air Mixtures: Effects of Humidity and Spark Duration. Journal of Electrostatics, 2007;
65(2):87-93.

Peterson, R.B. Small Packages. Mechanical Engineering Magazine 2001.

Phillips T. NASA, 2003. <http://spaceresearch.nasa.gov/general_info/31jan_kelly_lite.html>

Ronney P.D., Wu M.S., Pearlman H.G. Structure of Flame Balls at Low Lewis-Number
(SOFBALL): Preliminary Results from the STS-83 and STS-94 Space Flight
Experiments. Aerospace Sciences Meeting and Exhibit, Reno (1998).

Ronney P.D., Wu M.S., Pearlman H.G. Experimental study of flame balls in space: preliminary

51
 results from STS-83. AIAA Journal, 1998; 36(8): 1361-1368.

Roper F.G. The prediction of laminar jet diffusion flame sizes: Part 1. Theoretical model.
Combustion and Flame, 1977; 29:219-226.

Santoro R.J., Yeh T.T., Horvath J.J., Semerjian H.G. The Transport and Growth of Soot
Particles in Laminar Diffusion Flames. Combustion Science and Technology, 1987;
53:89-115.

Schefer R.W., Houf W.G., San Marchi C., Chernicoff W.P., Englom L. Characterization of leaks
from compressed hydrogen dispensing systems and related components. International
Journal of Hydrogen Energy, 2006; 31:1247-1260.

Sherif S.A., Barbir F., Veziroglu T.N. Wind energy and the hydrogen economy - review of the
technology. Solar Energy, 2005; 78:647-660.

Swain M.R., Swain M.N. A Comparison of H2, CH4, and C3H8 Fuel Leakage in Residential
Settings. International Journal of Hydrogen Energy, 1992; 17:807-815.

Takeno K., Okabayashi K., Kouchi A., Nonaka T., Hashiguchi K., Chitose K. Dispersion and
explosion field tests for 40 MPa pressurized hydrogen. International Journal of Hydrogen
Energy, 2007; 32(13):2144-2153.

Turns, S.R., An Introduction to Combustion. McGraw-Hill, (2000).

U.S. Department of Energy, Hydrogen Program, Hydrogen and Our Energy Future, DOE/EE-
0320, U.S. Department of Energy 2007.

Veziroglu T.N., Babir F. Hydrogen: The Wonder Fuel. International Journal of Hydrogen
Energy, 1972; 17, 391-404.

Yang W.M., Chou S.K., Shu C., Li Z.W., Xue H. Combustion in micro-cylindrical combustors
with and without a backward facing step. Applied Thermal Engineering, 2002; 22:1777-
1787 (2002).

Yamawaki M., Nishihara T., Inagaki Y, Minato K., Oigawa H., Onuki K., Hino R., Ogawa M.
Application of nuclear energy for environmentally friendly hydrogen generation.
International Journal of Hydrogen Energy, 2007; 32:2719-2725.

Zhang Y., Zhou J., Yang W., Liu M., Cen K.. Effects of hydrogen addition on methane catalytic
combustion in a microtube. International Journal of Hydrogen Energy, 2006; 32:1286-
1293.

52