Journal of Loss Prevention in the Process Industries 22 (2009) 499–505

Contents lists available at ScienceDirect

Journal of Loss Prevention in the Process Industries

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

The limiting oxygen concentration and flammability limits of

gases and gas mixtures
Isaac A. Zlochower*, Gregory M. Green
Pittsburgh Research Laboratory, National Institute for Occupational Safety and Health, Pittsburgh, PA 15236, USA

a r t i c l e i n f o a b s t r a c t

Article history: This paper presents data on the limiting (minimum) oxygen concentration (LOC), in the presence of added
Received 5 November 2008 N2, of methane (CH4), propane (C3H8), ethylene (C2H4), carbon monoxide (CO), and hydrogen (H2), and some
Received in revised form of their binary mixtures. It also addresses the issue of the flammable concentration (flammability) limits of
27 March 2009
these pure gases in air. The study is based on spark ignited explosions in large, spherical laboratory vessels
Accepted 31 March 2009
(120-L and 20-L) using a 7% pressure-rise criterion for explosion propagation. The results of the study are
compared with the older values which used long flammability tubes with a diameter 5 cm together with
Keywords:
visual evidence of substantial upward propagation. They are also compared to results reported recently
LFL
UFL using a 12-L spherical flask with a visual flame propagation criterion. Finally, they are compared to results
LOC reported in Europe using more modest flammability criteria and smaller chambers.
Flammable The findings reported here show excellent agreement between the 120-L and 12-L results, good
Gas agreement with the 20-L results, and reasonable agreement with the earlier flammability tube values for
Mixture the lower flammability limits. They disagree, however, with the more conservative European values.
These results and those from the 12-L experiments also feature lower LOCs than are given by traditional
flammability tubes. A model for the LOCs of such fuel mixtures based on the Le Chatelier mixture rule for
lower flammable limits is seen to reasonably fit the observed results on binary mixtures and can
accommodate more complex mixtures as well. One such set of ternary mixtures containing CH4 and 1:1
CO:H2 is well fitted by the model.
Published by Elsevier Ltd.

1. Introduction determinations (ASTM International, 2008a, 2008b; British and

European Standard, 2003, 2007). The traditional criterion used in
Starting with basic definitions, the lower and upper flamma- the US, which was the basis of an extensive database of flamma-
bility (or explosibility) limits (LFL and UFL, respectively) are the bility limits, required that flame and explosion propagation be
limiting fuel concentrations in air that can support flame propa- distinguished from ignition phenomena (Coward & Jones, 1952;
gation and lead to an explosion. Fuel concentrations outside those Kuchta, 1985). In order to demonstrate unambiguous flame prop-
limits are non-flammable. The progressive addition of an inert gas agation, U.S. standards required that flame be observed at some
to a fuel–air mixture causes the narrowing of the flammability distance from the ignition source and to have traveled through
range to the point where the two limits coincide. The limiting a significant fraction of the enclosed volume. That requirement was
oxygen concentration (LOC) is the minimum O2 concentration in relaxed in a more recent US standard that featured the use of a 5-L
a mixture of fuel, air, and an inert gas that will propagate flame. In spherical glass flask, and mandated only that flame propagation be
this paper, the inert gas will be nitrogen. In practice, the limits established by evidence of horizontal as well as vertical flame travel
(LFL, UFL, and LOC) represent an average between the neighboring (ASTM International, 2008a). Where ambiguity resulted, the stan-
concentrations inside and outside the experimental flammability dard called for using a larger (12-L) flask with a more extensive
limits (ASTM International, 2008a, 2008b). flame evolution – a flame cone with an arc spanning at least 90 at
There is currently a significant difference of opinion between the top of the flask, as measured from the point of ignition (ASH-
American and European based standard-setting organizations as to RAE, 2007; ASTM International, 2008a). The explosion overpressure
the prescribed test vessels and criteria for flammability and LOC in a closed system was, correspondingly, mandated to be 7%,1 i.e.,

1
* Corresponding author. Tel.: þ1 412 386 4960; fax: þ1 412 386 6595. A 7% pressure increase represents a 1 psi increase starting at 1 bar (14.5 psia) or
E-mail address: iaz0@cdc.gov (I.A. Zlochower). 1 atm (14.7 psia).

0950-4230/$ – see front matter Published by Elsevier Ltd.

doi:10.1016/j.jlp.2009.03.006
500 I.A. Zlochower, G.M. Green / Journal of Loss Prevention in the Process Industries 22 (2009) 499–505

a significant percentage of the initial value (ASTM International, spark electrodes terminated below the center of the sphere and
2008b). Although the exact flame travel distance or overpressure were bent to face each other with a gap of 6 mm. The electrical
required for designation as a true flame propagation and explosion energy so delivered came from a 1300 mF capacitor charged to 300 V
is somewhat subjective and apparatus-dependent, any attempt to which was then discharged through a transformer to generate
minimize such requirements will lead to results that are subjective, a strong spark. This spark was powered by a stored energy on the
apparatus-dependent, and overly conservative. Adding further to capacitor of 58 J based on the equation: stored energy ¼ 1/2CE2,
the problem, the flammability limits and LOCs when traditionally where C is the capacitance and E is the voltage. The actual electrical
determined have no built-in safety factors. In practice, therefore, energy in the spark gap was considerably less, however, because of
such factors must be imposed. For example, NFPA 69 requires that the low efficiency of the transformer circuit (Hertzberg, Conti, &
the fuel concentration only reach 25% of the LFL value (or 60% of Cashdollar, 1985).
the LFL for gaseous systems with automated in-line sensors and The calibrations of the pressure transducers were checked daily
controls) (NFPA 69, 2008). Imposing these factors has been the using the internal shunt calibration resistors provided by the
long-standing practice in the US (MSHA, 1996; NFPA 69, 2008). manufacturer. Samples of the gas mixtures could be collected in
The current standard adopted by the European Union, based on evacuated test tubes through a sampling needle on the side of the
the earlier German standard (DIN 51649, 1986), is a radical departure chamber. These samples were then analyzed by gas chromatography
from the above considerations. In that standard (EN 1839T), a sepa- (GC). During the initial evaluation of the mixing efficiency in the
ration of flame 10 cm above the 0.2–0.5 s induction spark in an 120-L chamber, samples of H2–air mixtures were collected after
8  30 cm cylinder defines a flammable mixture. The same is true if the gases had been added to the chamber and after 2 and 5 min of
the flame extends to 24 cm without separation from the spark mixing by the fan. There was essentially no difference in the
electrodes. The result is basically to substitute an ignitability criterion measured concentrations of H2 for the gas mixtures over this time
for flammability. For a closed spherical or cylindrical system with period, showing that there was good mixing of the gases even before
a minimum volume of 5-L and using such a centrally located induc- the fan was turned on. The reported concentrations are in mole
tion spark, or fuse wire of 10–20 J nominal energy (EN 1839T), the (volume) percent, based on the partial pressure of the component
corresponding criterion is a 5% pressure rise above that given by the relative to the total pressure. It was noted that the addition of gas to
source alone (British and European Standard, 2003). Other European a fully or partially evacuated chamber led to an increase in
researchers go further and insist that any separation of flame from temperature due, presumably, to quasi-adiabatic compression of the
the ignition spark bespeaks a flammable mixture (meaning ignit- prior atmosphere. This temperature rise was allowed to fall to
able). The same applies to an overpressure that is just measurable a steady-state value. That temperature was then used to correct all
(2%) (De Smedt et al., 1999). The LFLs and LOCs resulting from such the component partial pressures to a common (final) temperature
minimal flame propagation criteria are, understandably, lower than using the ideal gas law.
those resulting from the traditional ones since they have a built-in The 20-L chamber, with a pressure rating of over 20 bar, is nearly
safety factor. Despite these concerns the primary purpose of this spherical, having a diameter of 30 cm and a height of 35 cm. The
article is not to critique the European methodology, but to alert the ceramic encased electrodes were located slightly below the mid-
readers to the current existence of incompatible flammability data- height of the chamber, and the exposed, pointed steel tips were
bases; the US database (Coward & Jones, 1952; Kuchta, 1985; NFPA, spaced 6 mm apart. The charging circuit was similar to that used for
2008; Zabetakis, 1965) and the European one (CHEMSAFE, 2007). the 120-L sphere except that an 800 mF capacitor was charged to
More refined techniques often lead to the revision of earlier data and 300 V, giving a nominal stored energy of 36 J. The actual thermal
to progress in the field of study. Revisions based on new arbitrary energy deposited in the chamber, based on the observed pressure
criteria, however, tend to produce confusion. Moreover, the incom- increase due to the spark, was 0.5–1 J. The partial pressures of the
patible databases may lead to the misapplication of traditional safety gaseous components were measured as above or via a sensitive,
factors to the already conservative European data. This caution has temperature-controlled capacitance manometer that had a resolu-
already been issued by Britton (2002) and is echoed in this paper. tion of 0.1 torr (0.13 mbar) over a range of 1000 torr (1.33 bar). This
unit was calibrated by the manufacturer prior to use.
2. Experimental The data from the pressure transducers and the thermocouple
were recorded using a high speed analog to digital (A/D) board in
The flammability data reported here were obtained in the a personal computer (PC). This system can sample the data from
spherical 120-L and 20-L chambers used in earlier studies (Cash- various instrument channels, usually at speeds less than 20 kHz per
dollar, Zlochower, Green, Thomas, & Hertzberg, 2000). The internal channel. An in-house computer software program converted the
diameter of the 120-L chamber is 60 cm and its pressure rating is raw data to engineering units, plotted the data vs. time, and
69 bar. Instrumentation includes a sensitive strain-gauge pressure allowed various data smoothing options. Maximum pressure and
transducer to measure the partial pressures as the gases are added maximum rate of pressure-rise values were obtained from the
and mixed, and a higher capacity strain-gauge pressure transducer pressure vs. time traces. The reproducibility of the flammability
to monitor the explosion pressure. The strain-gauges have data was checked by repeated tests over a period of months and
a response time of 1 ms. The pressure transducers were mounted on years.
the top and bottom flanges of the chamber. A fine wire thermo- The experimental determination of the flammability limits of
couple near the top of the chamber was used to record the steady- the gases studied was done by fully evacuating the chamber and
state temperature after the addition of each gaseous component. then admitting the fuel gas to the predetermined partial pressure
There were also ports with ball valves for connection to a vacuum required to produce the desired final concentration in air. Dry,
pump and vent. An internal non-sparking fan at the bottom of the oil-free air was admitted from a compressed gas cylinder to give the
vessel was used to mix the gases. The gases were added via desired final pressure of 1 bar (the ambient pressure at the eleva-
a manifold and controlled via manual switches in an isolated tion of the laboratory was always less than that final pressure). The
control room where the additions, pressure monitoring, mixing and temperature in the chamber rose during the addition of both the
spark ignition initiation, and PC recording were conducted. fuel and air, as previously noted, and the gas was allowed to come
The spark ignition system consisted of 3-mm diameter metal rod to a steady-state temperature and pressure. The temperature of
electrodes with a durable plastic spacer for rigidity. The pointed the fuel gas was then used to correct the fuel pressure to that
 I.A. Zlochower, G.M. Green / Journal of Loss Prevention in the Process Industries 22 (2009) 499–505 501

corresponding to the final temperature of the mixture according to (Schroder & Molnarne, 2005). The tube values are based on the
the ideal gas law. The final concentration of the fuel gas was then buoyant rise of small flame kernels which represent only a small
calculated from its corrected pressure. Next, the non-sparking fraction of fuel consumed and a correspondingly minimal pressure
internal fan was started and run for at least 4 min to fully mix the rise. Measurements in a 14-L closed sphere gave an LFL of 4.2%
gases. The fan was shut off and the generated turbulence was (Schroder & Molnarne, 2005), but the use of fuse wire ignition
allowed to subside for 1 min. The capacitor spark ignition system rather than the weaker and more localized spark ignition is believed
was then charged to 300 V and discharged through an ignition coil to have unduly influenced the pressure rise associated with partial
to create the high voltage spark. The series of sparks thereby propagation (5%). A more robust explosion pressure-rise criterion
produced lasted for about 0.1 s. (7%) for hydrogen in larger chambers would appear to be better
The initial concentrations of fuel gas chosen were based on the correlated with actual explosion hazards.
literature values for the flammability limits or on prior determi- There is less data for the UFL in the 120-L sphere. However, the
nations using such chambers. Once marginal explosion pressures sharp drop in explosion pressure beyond 75.7% hydrogen (the
were produced, the concentrations were changed in small steps pressure rise is near zero at 76%) argues against a significant
(0.1–0.2%) for the LFL and UFL measurements of methane, propane, dependence of UFL on the above pressure criterion. The UFL is
ethylene, and CO. The concentration step for the LFL of hydrogen estimated as 75.9% H2, the midpoint between these values. Previ-
was 0.5–1% owing to the low sensitivity of the explosion pressure to ously, a value of 76.8% was reported by Cashdollar et al. (2000), but
concentration near the LFL (Cashdollar et al., 2000). that value was obtained in an 8-L chamber. The LFL in that chamber
The LOC measurements were made starting at fuel concentra- was also found to be lower (5%) than in the larger (20-L and 120-L)
tions near the LFL and using oxygen concentrations (from the air chambers. Similarly high UFLs were given in the tests run according
and nitrogen additions) that were near the LFL multiplied by the to the EN 1839 standard (76.6%, tube and 77.0%, closed sphere).
stoichiometric oxygen/fuel ratio (the ‘‘R’’ column in Table 1). The Given the sensitivity of hydrogen flammability to the experimental
oxygen concentration was varied in steps of 0.2% until the marginal conditions and chamber volume, the above relative difference of 1%
explosion conditions were delineated and verified. Then the fuel for the UFL is not surprising.
concentration was changed and the minimum oxygen concentration The data on explosion pressures at different oxygen vs.
again determined. The process was repeated until a global minimum hydrogen concentrations in air–nitrogen mixtures is given in Fig. 1,
oxygen concentration (LOC) with added N2 was established. where the legend symbols represent pressure-rise ranges. In other
words, Fig. 1 and the subsequent figures for LOC data are pressure
3. Results and discussion: single fuels parameter plots for given O2 and fuel concentrations, with the
different symbols representing different pressure-rise levels. Fig. 1
Previous results (Cashdollar et al., 2000) for the LFL of H2 in the shows a minimum oxygen concentration (LOC) of 4.6% that is
120-L sphere using spark ignition showed a slow, linear increase in independent of hydrogen concentration in the range of 10–30%.
explosion pressure with concentration below 7%, followed by an More typically, however, the minimum oxygen concentration for an
accelerated increase above 7%. The pressure rise became steep at 8% inert fuel–air mixture occurs at a narrow range of fuel concentra-
H2. The LFL of H2 is thus quite sensitive to the pressure criterion for tions. Those ‘‘worst-case’’ fuel concentrations define the LOC for the
explosion propagation. The 7% pressure-rise criterion for spark fuel–air mixture containing the inert gaseous additive (e.g. N2).
ignition yielded an LFL of 7% in this chamber. The explosion pressure The results for the LFL of CO are similar to that of H2, but feature
rise also occurred gradually with low H2 concentrations in the 20-L a steeper rise in pressure with concentration. They are also depen-
chamber (below 8% H2), but with a steeper slope. In the 20-L dent on the amount of water vapor present in the mixture. In
chamber, the 7% pressure-rise criterion was satisfied at 6% H2 in air. contrast to the ordinary behavior of water as an effective inhibitor of
With a meter-long upward flame propagation criterion in vertical explosions due to its large heat capacity, here it is an accelerant – at
flammability tubes, the LFL was found to be 4%. The much shorter least at low concentrations. It functions, presumably, by supplying
30 cm tubes used in the EN 1839 (T) method gave an LFL of 3.6% the H and OH free radicals for the branching chain reactions needed

Table 1
Flammability/LOC: 120-L and 20-L closed vessel results vs. 12-L glass sphere and flammability tube.

Fuel (F) Vessel Stoichiometric equation Mole ratio (R) (O2/F) LFL (mol%) UFL (mol%) LOC (N2) (mol%) Explosion criterion
Hydrogen (H2) 120-L 2H2 þ O2 ¼ 2H2O 0.5 7 75.9 4.6 Pressure
20-L 6 4.7 Pressure
Flam.a tube 4 75.0 5.0 Visual

Carbon monoxide (CO) 120-L 2CO þ O2 ¼ 2CO2 0.5 12.2 72.0 5.1 Pressure
12-Lb 12.2 72.5 Visual
Flam.a tube 12.5 74.0 5.5 Visual

Methane (CH4) 120-L CH4 þ 2O2 ¼ CO2 þ2H2O 2 5.0 15.8 11.1 Pressure
20-L 4.9 15.9 10.7 Pressure
12-Lb 4.9 15.8 11.3 Visual
Flam.a tube 5.0 15.0 12.0 Visual

Ethylene (C2H4) 120-L C2H4 þ 3O2 ¼ 2CO2þ2H2O 3 2.7 31.4 8.5 Pressure
12-Lb 2.7 31.5 8.6 Visual
Flam.a tube 2.7 36.0 10.0 Visual

Propane (C3H8) 120-L C3H8 þ 5O2 ¼ 3CO2þ4H2O 5 2.0 9.8 10.7 Pressure
12-Lb 2.0 10.0 10.5 Visual
Flam.a tube 2.1 9.5 11.5 Visual
a
Flammability tube data summarized by Kuchta (1985).
b
The data from a 12-L spherical flask by Kondo et al. (2006, 2008).
502 I.A. Zlochower, G.M. Green / Journal of Loss Prevention in the Process Industries 22 (2009) 499–505

6.0 as the LFL of methane. Both the pressure criterion of 7% and the
>30 existence of a discontinuous change in slope yield the same LFL
15-30 value. The corresponding UFL is 15.8% based on the newer data. The
5.5 7-15
same value is the average of the values reported by Cashdollar et al.
(2000) for the 120-L and 20-L chambers (15.7 and 15.9%, respec-
3-7
tively), while the LFL was 5.0% for both chambers. The 12-L glass
Oxygen,

<3
sphere gave an LFL of 4.9% and a UFL of 15.8% (Kondo, Takizawa,
5.0
Takahashi, & Tokuhashi, 2006).
Fig. 3 shows an LOC of 11.1% O2 for 5.2–5.8% CH4 in air–N2
mixtures in the 120-L sphere, but the LOC was found to be 10.7% O2
4.5
in the 20-L chamber. The 12-L sphere gave a value of 11.3%, as
reported by Kondo et al. (2006). The lower value given by the 20-L
chamber may indicate the need for a higher pressure-rise criterion
4.0 in chambers appreciably smaller than 120-L (Brandes & Ural, 2008).
5 10 15 20 25 30
Past work (Cashdollar et al., 2000) gave an LFL value for propane
Hydrogen,
of 2.05% with a sharp rise in pressure with concentration. The
Fig. 1. LOC data on H2 in air with added N2 that was obtained in the 120-L sphere. pressure drop with concentration near the UFL was more gradual. It
Parametric pressure plot at different O2 vs. H2 concentrations, with the symbols was found to be 9.8% for the 7% pressure criterion. Current results in
representing different pressure-rise ranges. the 120-L sphere are consistent with the earlier findings. The LOC of
propane is seen to be about 10.7% % at 2.5–2.7% propane (Fig. 4). The
for combustion and for rapid flame propagation. The measured LFL is 12-L glass sphere results of Kondo et al. (2006) gave an LFL of 2.0%,
12.2% when the mixture is nearly saturated with water. Similarly, the a UFL of 10.0%, and an LOC of 10.5%, again in agreement with the 120-
UFL is 72% for near saturation. The results in the 12-L glass sphere for L results.
CO in moist air with added N2 (LFL ¼ 12.2%, UFL ¼ 72.5%) that was Ethylene also features a sharp rise in pressure with concentra-
reported by Kondo, Takizawa, Takahashi, Tokuhashi, and Sekiya tion at low concentrations and a slower decline in pressure near the
(2008) agrees with the 120-L values (based on their listed uncer- UFL. The LFL is found to be 2.7% and the UFL is 31.4% using the 7%
tainty of 0.5% for the UFL). Fig. 2 shows an LOC in the 120-L sphere of criterion. The LOC is seen to be 8.5% O2 at 2.9–3.2% ethylene (Fig. 5).
about 5.1% O2 for gas mixtures nearly saturated with water vapor. The 12-L results of Kondo et al. (2006) yielded an LFL of 2.7%, a UFL
That LOC occurs near a 20% CO concentration. of 31.5%, and an LOC ¼ 8.6%.
The above results represent atypical fuels that are lighter and The 120-L and 20-L data are summarized in Table 1 together
more diffusive than O2 (H2), or have branching chain reactions with the reported results from the 12-L sphere (Kondo et al., 2006,
associated with flame propagation that are more difficult to 2008) and flammability tubes (Kuchta, 1985). The reported results
establish (CO). Hydrogen, for example, is more concentrated at the from the 12-L glass sphere are essentially identical to the results
flame front (i.e., the real fuel concentration at the flame front is obtained here in the 120-L sphere despite the great difference in
higher than the nominal value) which clearly impacts the LFL value volume and propagation criterion.
(Cashdollar et al., 2000; Hertzberg, 1989). More typical flammable A good reflection of the difference between the US and European
gases – namely, the common hydrocarbons: methane (CH4), the standards is given by the LFLs of methane in air. Traditionally, this
primary constituent of natural gas; propane (C3H8), the popular value has been listed as 5.0% based on the results of flame propa-
cooking and heating fuel; and ethylene (C2H4), a basic chemical gation studies using a large upward propagation distance in tall,
feedstock for polyethylene and other plastics – were studied as well. wide containers (Britton, 2002; Coward & Jones, 1952), in closed,
Cashdollar et al. (2000) have reported 120-L data on the LFL of larger spherical vessels (120-L and 20-L), and in a very large spherical
methane in air which show a gradual increase in pressure from 4.6 vessel (Cashdollar et al., 2000). The European Standard (Annex E)
to 5.0%. There is little reason, however, to dispute the choice of 5.0% using a 5% pressure-rise criterion in a small vessel (EN 1839B) gave
4.6% as the LFL. Using the flame detachment criterion in a small

6.0
> 30 7-15 <3 15
>30
5.8
15-30
14 7-15
3-7
5.6 <3
Oxygen,

13
Oxygen,

5.4
12

5.2
11

5.0
10.0 15.0 20.0 25.0 30.0 35.0
10
Carbon Monoxide, 5.0 5.2 5.4 5.6 5.8 6.0
LOC = 5.1 Methane,

Fig. 2. 120-L data on the LOC of CO in nearly saturated moist air with added N2. Fig. 3. 120-L data on the LOC of CH4 in air with added N2. Parametric pressure plot at
Parametric pressure plot at different O2 vs. CO concentrations, with the symbols different O2 vs. CH4 concentrations, with the symbols representing different pressure-
representing different pressure-rise levels. rise ranges.
 I.A. Zlochower, G.M. Green / Journal of Loss Prevention in the Process Industries 22 (2009) 499–505 503

11.0 the 120-L sphere and the visual criterion in the 12-L sphere lends
>30 encouragement to the assumption that values have been deter-
15-30
mined that are reasonably independent of vessel size and flame
10.8 7-15
propagation criteria.
3-7
<3 The above argument implies that the LOC values in the older
10.6 literature are not sufficiently conservative and should not be used to
generate safety protocols unless conservative safety factors are
Oxygen,

incorporated. The problem with the older LOC values is also recog-
10.4 nized in the 2008 NFPA 69 standard which requires subtracting 2%
(abs.) from the old values. While that is probably a conservative stance,
10.2
it may involve unnecessary expense in the case of hydrogen – based
on the results obtained in the 120-L sphere. Additional measurements
using large chambers and appropriate flame propagation criteria will
10.0 be needed to form a reliable database of LOC values.
2.0 2.5 3.0 3.5 4.0
Propane
LOC = 10.7 4. Fuel mixtures

Fig. 4. The LOC of propane in air with added N2; 120-L data. Parametric pressure The LFL of fuel mixtures is predicted adequately by the Le
plot at different O2 vs. C3H8 concentrations, with the symbols representing different Chatelier rule for fuel mixtures (Burgess et al., 1982; Coward &
pressure-rise ranges.
Jones, 1952; Kondo et al., 2008; Liekhus, Zlochower, Cashdollar,
Djordjevic, & Loehr, 2000). That rule is less adequate, but still
cylindrical vessel (EN 1839T) yielded a value of 4.4% in this standard
generally conservative, in predicting the UFL of mixtures (Kondo
(British and European Standard, 2003). A similar difference was
et al., 2008). The LOCs of such mixtures, however, have not been
observed for the UFL measurements of methane in air. The European
extensively studied. Prior work on mixtures of hydrogen and
standard methods gave 16.6% and 16.8% values for the tube and
benzene (Thomas, 1996) established the utility of a calculation
‘‘bomb’’ methods, respectively – cited in Annex E, while both the
scheme that will be derived in this paper and used to compare to
120-L and 12-L spheres gave 15.8%. A further confirmation of
the experimental results on binary mixtures of hydrogen, CO, or
the results reported here is given by the studies of Kondo’s group on
ethylene with methane. Mixtures of propane and methane are not
the effect of vessel size and shape on the experimental flammability
considered since their LOCs are too close to produce a significant
limits of methane and propane. They found that cylindrical vessels
variation with composition. The experimental method chosen was
did not yield values that were size independent until quite large
the same as for the single fuels, except that two different fuels were
dimensions were used. Their largest cylinder (45  100 cm) with
mixed together with air and nitrogen to establish the stated
a volume of 160-L gave values quite consistent with those reported
mixtures. Ternary mixtures of 1:1 hydrogen:CO with methane were
here for full upward propagation of a flammable mixture (methane
also studied. Water was not added to the mixtures containing CO
flammability limits ¼ 5.0–15.8%; propane limits ¼ 2.0–9.9%) (Taka-
since the other fuel(s) had the H species needed for efficient CO
hashi, Urano, Tokuhashi, & Kondo, 2003).
combustion. The results are summarized in Table 2.
The different values cited for the LFL of methane in the new
Table 2 shows the experimental and calculated values of the LOC
databases which are based on the European standards (CHEMSAFE,
of CH4 mixtures with H2 and/or CO, or with C2H4. The predicted
2007) should not be interpreted as an argument for the retention of
LOCs involving mixtures of H2 and/or CO with CH4 follow the
older flammability values if new experiments in larger chambers
experimental values but fall to a maximum of 0.9% (abs.) below
using comparable flammability criteria produce new values. That is
them. The calculations to be described in the next section are
the case with experiments reported here, particularly in regards to
therefore seen as realistic but conservative. H2 and CO are unusual
LOC values. These experiments now generate systematically lower
fuels, however (high diffusivity or low reactivity) as noted earlier.
LOC values, as do the results in the 12-L glass sphere (Kondo et al.,
The deviation of the experimental results from the derived formula
2006). This agreement between the pressure-rise criterion used in
is, therefore, not surprising. The one clear example of a mixture of
more typical fuels with widely different LOCs is the mixture of
10.0 methane and ethylene. The calculated values are very close to the
>30 experimental ones with a maximum difference of 0.15% (abs.).
15-30
7-15
9.5
3-7 5. LOC of fuel mixtures
<3

As noted previously, the lower flammability limit (LFL) for

Oxygen,

9.0 mixtures of fuels is adequately given by the Le Chatelier rule:

X
Lmixt: ¼ 1= xi =Li (1)
8.5
Where Lmixt is the LFL of the fuel mixture, Li is the LFL of fuel
component i, xi is the mole fraction of the fuel component.
The LOC of fuel mixtures can be derived from (1) by defining
8.0
2.0 2.5 3.0 3.5 4.0 4.5 5.0 L*mixt and L*i as,
Ethylene, LOC = 8.5
L*i ¼ ðLOCÞi =Ri ; L*mixt ¼ ðLOCÞmixt =R* (2)
Fig. 5. The LOC of ethylene in air with added N2; 120-L data. Parametric pressure plot
for different O2 vs. C2H4 concentrations, with the symbols representing different Where (LOC)i is the experimental value for LOC of component i,
pressure-rise ranges. (LOC)mixt is the calculated LOC for the fuel mixture, Ri is the
504 I.A. Zlochower, G.M. Green / Journal of Loss Prevention in the Process Industries 22 (2009) 499–505

Table 2
LOC: fuel mixtures containing methane: CH4–H2; CH4–CO; CH4–C2H4; CH4–1:1 CO:H2: experimental vs. calculated.

Fuel: % CH4 Methane–H2 Methane–CO Methane–ethylene CH4–1:1 CO:H2

(LOC)exp (LOC)calc (LOC)exp (LOC)calc (LOC)exp (LOC)calc (LOC)exp (LOC)calc

0 4.6 4.6 5.1 5.1 8.6 8.6 4.8 4.8
10 6.5 5.6 6.1 6.1 8.6 8.7
20 7.3 6.7
25 7.7 6.9 7.8 7.4 8.9 9
35 8.3 7.65
40 8.7 8.2
50 9.1 8.64 9.5 9.0 9.3 9.45
60 9.7 9.4
75 10.3 10 10.9 10.2 10.05 10.1
80 10.5 10.3
90 10.9 10.7 11 10.8
100 11.1 11.1 11.1 11.1 11.1 11.1 11.1 11.1

stoichiometric molar ratio of oxygen to fuel i (see Table 1), R* is the results using a 7% pressure-rise criterion with the reported results
P
stoichiometric ratio of oxygen to the fuel mixture, as given by xiRi. in the 12-L glass sphere which uses a visual criterion supports
P the contention that these results are realistic. The issue of the
For L*mixt ¼ 1= xi =L*i (modified Le Chatelier rule),
conflicting American and European flammability standards needs
substituting (2) into the modified Le Chatelier and rearranging
to be resolved so as to provide a consistent and reliable database for
terms gives
flammability.
X X X X
ðLOCÞmixt: ¼ xi R i = xi =L*i ¼ xi R i = xi Ri =ðLOCÞi
Acknowledgement
For a 2-component mixture (a,b):
    This is to acknowledge the important contributions of Kenneth
ðLOCÞmixt: ¼ ðxa Ra þ xb Rb Þ= xa Ra =ðLOCÞa þ xb Rb =ðLOCÞb L. Cashdollar over the years to the flammability and LOC studies in
the 120-L and 20-L chambers, and to his helpful suggestions and
For binary mixtures of methane (m) and hydrogen (h): Rm ¼ 2, advice. Sad to say, Ken recently passed away and we will no longer
Rh ¼ 0.5, xh ¼ 1  xm, LOCm ¼ 11.1, LOCh ¼ 4.6. The result is: have the benefit of his leadership, humanity, and superlative
LOCm–h ¼ (1.5xm þ 0.5)/(0.0715xm þ 0.109). experimental and organizational skills. His many contributions to
For binary mixtures of methane and CO: Rco ¼ 0.5, xco ¼ 1  xm, our organization and the field of explosion prevention will be sorely
LOCco ¼ 5.1. The result is: LOCm–co ¼ (1.5xm þ 0.5)/ missed.
(0.0821xm þ 0.0980).
For binary mixtures of methane and ethylene: Ret ¼ 3,
References
xet ¼ 1  xm, LOCet ¼ 8.6. The result is: LOCm–et ¼ (3 – xm)/
(0.349 – 0.169xm). ASHRAE. (2007). Number designation and safety classification of refrigerants, ANSI/
For ternary mixtures of methane with 1:1 CO:H2, ASHRAE Standard 34-2007. American Society of Heating, Refrigerating, and Air-
xco ¼ xh ¼ (1  xm)/2. The result is: LOCm–co:h ¼ (1.5xm þ 0.5)/ conditioning Engineers (ASHRAE), Atlanta GA.
ASTM International. (2008a). Standard test method for concentration limits of flam-
(0.0768xm þ 0.1034).
mability of chemicals (vapors and gases). E681-04. In: ASTM International annual
book of standards, Vol. 14.02. West Conshohocken, PA: American Society for
The above equations lead to the calculated LOC results shown in Testing and Materials.
ASTM International. (2008b). Standard test methods for limiting oxygen (oxidant)
Table 2. As noted previously, the calculated results for the mixture
concentration in gases and vapors. E2079-07. In: ASTM International annual book
of hydrocarbons accurately predict the experimental values. The of standards, Vol. 14.02. West Conshohocken, PA: American Society for Testing
results for the unusual fuels, H2 and CO, are below the experimental and Materials.
values, but barely outside experimental error. In any case, the Brandes, E., & Ural, E. A. (2008). Towards a global standard for flammability
determination. In Proceedings of the 42nd annual loss prevention symposium –
calculated results are conservative. It remains to be seen how this Global safety congress, paper 2E, April 6–10. New Orleans, LA: American Institute
calculation scheme will fit data on oxygenated and other non- of Chemical Engineers.
hydrocarbon fuels. British and European Standard. (2003). Determination of explosion limits of gases and
vapours. sections 4.1, Method T (BS EN 1839T) and 4.2, Method B (BS EN 1839B).
rue de Stassart, 36, B-1050 Brussels, Belgium: CEN European Committee for
6. Conclusions Standardization.
British and European Standard. (2007). Determination of limiting oxygen concen-
trations of gases and vapours. EN 14756. rue de Stassart, 36, B-1050 Brussels,
The primary finding reported here is that the traditional flam- Belgium: CEN European Committee for Standardization.
mability tube values for the LOCs are too high – at least for N2 Britton, L. G. (2002). Two hundred years of flammable limits. Process Safety Progress,
inerting that was the subject of this study. The differences from the 21, 1–11.
Burgess, D. S., Furno, A. L., Kuchta, J. M., & Mura, K. E. (1982). Flammability of mixed
120-L results are typically less than 1% (e.g. 12 vs. 11.1% for CH4, 11.5
gases. U.S. Bureau of Mines Report of Investigations, 8709. 20 pp.
vs. 10.7 for C3H8), but the difference is 1.5% for ethylene (10 vs. Cashdollar, K. L., Zlochower, I. A., Green, G. M., Thomas, R. A., & Hertzberg, M.
8.5%). These results should be taken into account by the standard- (2000). Flammability of methane, propane, and hydrogen gases. Journal of Loss
Prevention in the Process Industries, 13, 327–340.
setting organizations – or at least be subject to independent
CHEMSAFE. (2007). Database of evaluated safety characteristics. Frankfort/M,
verification. A start in this direction is currently being made by Germany: DECHEMA, BAM, and PTB.
standards committees based in the US (NFPA 69, 2008). Coward, H. F., & Jones, G. W. (1952). Limits of flammability of gases and vapors. U.S.
This study also generally confirms the results of the flamma- Bureau of Mines Bulletin, 503. 155 pp.
De Smedt, G., de Corte, F., Notele, R., & Berghmans, J. (1999). Comparison of two
bility tube measurements on the LFLs of these fuels (hydrogen standard test methods for determining explosion limits of gases at atmospheric
being exceptional). The excellent agreement between the 120-L conditions. Journal of Hazardous Materials, A70, 105–113.
 I.A. Zlochower, G.M. Green / Journal of Loss Prevention in the Process Industries 22 (2009) 499–505 505

DIN (German standard) 51649-1. (1986). Determination of the explosion limits of and hydrogen. Journal of Loss Prevention in the Process Industries, 13,
gases and gas mixtures in air. (Ger.). Berlin, Germany: Deutches Institut fur 377–384.
Normung. MSHA. (March 1996). Safety standards for underground coal mine ventilation: Final
Hertzberg, M. (1989). Selective diffusional demixing: occurrence and size of cellular rule. 30 CFR, Part 75, Section 75.323. In: Federal register, part III. Department of
flames. Progress in Energy and Combustion Science, 15, 203–239. Labor, Mine Safety and Health Administration.
Hertzberg, M., Conti, R. S., & Cashdollar, K. L. (1985). Electrical ignition energies and NFPA 69. (2008). Standard on explosion prevention systems (2008 ed.). Quincy, MA:
thermal autoignition temperatures for evaluating explosion hazards of dust. U.S. National Fire Protection Association.
Bureau of Mines Report of Investigations, 8988. 41 pp. Schroder, V., & Molnarne, M. (2005). Flammability of gas mixtures. Part I: fire
Kondo, S., Takizawa, K., Takahashi, A., & Tokuhashi, K. (2006). Extended Le potential. Journal of Hazardous Materials, 121, 37–44.
Chatelier’s formula and nitrogen dilution effect on the flammability limits. Fire Takahashi, A., Urano, Y., Tokuhashi, K., & Kondo, S. (2003). Effect of vessel size and
Safety Journal, 41, 406–417. shape on experimental flammability limits of gases. Journal of Hazardous
Kondo, S., Takizawa, K., Takahashi, A., Tokuhashi, K., & Sekiya, A. (2008). A study on Materials, 105, 27–37.
flammability limits of fuel mixtures. Journal of Hazardous Materials, 155, 440–448. Thomas, J. K. (1996). Composite minimum oxygen concentrations for benzene–
Kuchta, J. M. (1985). Investigation of fire and explosion accidents in the chemical, mining, hydrogen–nitrogen–air mixtures. WSRC-RP-96-619. Aiken, SC: Westinghouse
and fuel-related industries – a manual. U.S. Bureau of Mines Bulletin, 680. 84 pp. Savannah River Co., Safety Engineering Div..
Liekhus, K. J., Zlochower, I. A., Cashdollar, K. L., Djordjevic, S. M., & Loehr, C. A. Zabetakis, M. G. (1965). Flammability characteristics of combustible gases and
(2000). Flammability of gas mixtures containing volatile organic compounds vapors. U.S. Bureau of Mines Bulletin, 627. 121 pp.