The effects of atmosphere composition on the rate of opposed-flow flame spread (Sf) over thin solid fuel beds at microgravity (µg) were measured and compared to compared to earth gravity (1g) results and theoretical predictions. Two modifications to standard atmospheres were considered. First, the effects of sub-flammability-limit concentrations of a gaseous fuel (CO or CH4) were measured and compared to an existing theoretical model that was extended to µg conditions. The agreement between the model and experiment is reasonable considering the simplicity of the model. Notably, both model and experiment show that the effect of added gaseous fuel is greater at µg than 1g. Secondly, the effect of diluent type on Sf was studied by comparing results using He, N2, Ar, CO2 and SF6 diluents. It was found that, in agreement with prior studies in N2 diluent, for He, N2 or Ar diluents, Sf was larger at 1g than µg. In contrast, for CO2 diluent, Sf was slightly lower at 1g than at µg and for SF6 diluent, Sf was much lower at 1g than µg. Moreover, unlike He, N2 and Ar, for CO₂ and especially SF₆ diluents the minimum O₂ concentration required to support flame spread was lower at µg than 1g. For SF₆, the minimum O₂ concentration at µg was even lower than the upward (concurrent-flow) limit. This unusual behavior is proposed to be a result of reabsorption of radiation emitted from the gases, which is significant only for gases with small mean absorption lengths.

Two types of changes to the oxidizing atmosphere are considered in this work. One is the addition of sub-flammability-limit concentrations of a gaseous fuel ("partially premixed" atmospheres). This is of interest because in fires in enclosures, combustion may occur under poorly ventilated conditions, so that oxygen is partially depleted from the air and is replaced by combustible gases such as fuel vapors, H2 or CO. Subsequent fire spread over the solid fuel could occur under conditions of varying oxygen and gaseous fuel content. The significance of flame spread under partially premixed conditions has been noted previously (Beyler, 1984). The second change is the diluent type, which affects the Lewis number of the reactants in the atmosphere, which for oxygen is defined as the thermal diffusivity of the bulk mixture divided by the mass diffusivity of oxygen into the bulk mixture, as well as the radiative properties of the gas. It is well known that both the effects of Lewis number (Greenberg and Ronney, 1993) and radiation (Bhattacharjee and Altenkirch, 1993) may affect flame spread processes. It is of interest to study these diluent effects because in undersea and spaceborne habitations, it is sometimes desirable to use diluent gases other than nitrogen.

Recently Ronney et al. (1995) found that the addition of sub-flammability limit concentrations of combustible gases to an oxidizing atmosphere could increase Sf for thermally-thin fuels at earth gravity substantially - sometimes by a factor of three even at fixed O2 mole fraction. The chemical reaction rates of the premixed fuel was found to have a substantial impact on S_f, even at higher O₂ concentrations where finite-rate chemistry of the nonpremixed fuel does not affect S_f significantly. Furthermore, a surprising result was found for added CO fuel: S_f is actually greater and the most dilute atmosphere capable of supporting combustion is weaker, when for a fixed total number of oxygen atoms in the gas phase, some of the oxygen atoms are in the form of CO rather than O2.

Ronney et al. (1995) showed that these results could be modeled with reasonable accuracy by treating the effect of the gaseous fuel as a thin planar semi-infinite premixed flame sheet located at x = xp < 0, where x is the spatial coordinate parallel to the fuel bed, upstream of the usual nonpremixed flame. This heat source was incorporated into the classical theory of flame spread over thermally-thin fuels developed by deRis (1969) and Delichatsios (1986) where a nonpremixed flame sheet having mixing-limited heat release (i.e., infinitely fast reaction rate between the solid fuel vapors and oxygen) touches the fuel bed at x = 0 (i.e., xn = 0). For non-merged flames, Ronney et al. (1995) derived the relation

(1),

where l refers to the conductivity, r the density, Cp the specific heat, t the fuel bed thickness, T the temperature, q" the heat release per unit area of the premixed flame front, U the opposed-flow velocity (which was buoyancy-driven since Ronney et al. did not impose any forced convection) and the subscripts g, s, v and ¥ refer to the gas mixture, solid fuel bed, vaporization condition and ambient condition, respectively. Note that the contribution of the nonpremixed (premixed) flame is represented by the first (second) term inside the brackets. From this equation, it can be seen that the added fuel increases q" and thus Sf beyond that of the nonpremixed flame alone, which is consistent with Ronney et al.’s results.

To estimate q", the asymptotic analysis of partially-premixed flames in stagnation-point flow by Hamins et al. (1985) was employed. This analysis considers a nonpremixed flame coupled to a premixed flame that adjusts its position to where q" (which depends strongly on its temperature (Tp) through the Arrhenius term in the reaction rate for the premixed flame) is commensurate with the temperature and concentration gradients at that location. For brevity, the details of the analysis presented in Ronney et al. (1995) are not duplicated here. Solution of Eq. 1 with the relations for q" shows that as U decreases, as would occur if gravity (and thus buoyant convection velocities) were reduced, S_f increases. Therefore, the effect of adding gaseous fuel, which can be significant at 1g, should be still greater at µg. A test of this interesting prediction is a major goal of this investigation.

The non-merged flame model requires a flame separation (xn - xp) of order d º a_g/U, where d is the gas-phase transport length scale and a_g º l_g/r_gC_p,g is the thermal diffusivity. As explained by Hamins et al. (1985), the premixed and nonpremixed flames will merge when the premixed-fuel concentration or the reaction rate of the premixed flame is sufficiently low, which corresponds approximately to the condition (Tf - Tp)/Tp ² RT_p/E_p, where T_p is the temperature of the premixed flame, R is the gas constant and E_p the activation energy of the premixed-flame reaction. This condition indicates that, as expected, the premixed and nonpremixed flame temperatures are nearly the same when the flames merge. In the merged-flame case, the location of both flames is at x = xn and consequently q" = 0. Ronney et al. (1995) showed that, to a first approximation, in the merged-flame case S_f can be estimated using Eq. (1) with q" = 0 and T_f for merged flames given by (Hamins et al. 1985):

(2),

where n is the stoichiometric fuel/(oxygen+inert) mass ratio, Q is the heating value, L is the latent heat of vaporization of the solid fuel, Y_p,° is the ambient mass fraction of premixed fuel, f º n_pY_p,°/Y_O2,° is the equivalence ratio of premixed fuel in the ambient atmosphere, Y_O2,° is the ambient mass fraction of oxygen in the ambient atmosphere and the subscripts n and p refer to the nonpremixed (solid) and premixed (gaseous) fuel, respectively. It is easily verified from Eq. (2) that the effect of gaseous fuel on T_f for the merged flame is minimal under conditions where there is a substantial amount of inert gas in the oxidizing atmosphere (i.e., when n_n is small, as with hydrocarbons burning in air), therefore gaseous fuel has little effect on S_f under most merged-flame conditions. In fact, for some cases T_f may decrease with increasing gaseous fuel concentration if the C_p of the fuel is larger than that of the diluent. This is consistent with the experimental observations of Ronney et al. (1995). Also, since Eq. (2) does not contain any U-dependent terms, under merged-flame conditions, e.g., low f, the effect of added gaseous fuel should be minimal at µg just as it is at 1g.

The effect of diluent type on thin-fuel flame spread at 1g has been studied experimentally by Zhang et al. (1992). These authors found that diluent type has a substantial effect on the applicability of Eq. (1) (with q" = 0). Greenberg and Ronney (1993) showed that these deviations could be explained by the effects of the oxygen Lewis number (Le_O2). It was found that to a first approximation, the effect of Le_O2 on Sf is due solely to its effect on Tf, so that the classical expressions for Sf could be retained if Tf is modified to account for Lewis number effects. The convection environment was predicted not to affect role of Lewis number effects on S_f. This result may seem surprising since it is well known that convection affects the adiabatic temperature of nonpremixed flames with non-unity Lewis numbers (Law and Chung, 1982). However, flame spread rates are controlled by the heat transfer to the fuel bed in the vicinity of the leading edge of the flame, which is inherently a two-dimensional creeping flow problem in which diffusive transport in both streamwise and transverse gradients are of first-order importance and convection effects are negligible (deRis, 1969; Williams, 1976). Fundamentally, this occurs because for thin fuel beds, where heat transfer through the solid phase is negligible, flame spread requires both forward (upstream) heat transfer to propagate the flame as well as transverse heat transfer to the fuel bed to support vaporization of the fuel bed, and these rates must be comparable at steady-state. A boundary-layer configuration where downstream convective transfer matches transverse diffusive transfer is inconsistent with a flame spreading upstream. Consequently, the convective flow in the vicinity of the leading edge of the flame is in general too weak to affect Tf and thus too weak to affect S_f. Convection does of course influence the flame thickness d which, as discussed later, plays an important role in radiation effects.

The favorable comparison of Greenberg and Ronney’s predictions with Zhang et al.’s experiments in view of the creeping-flow assumption indicates that even at 1g, convection does not influence Lewis number effects on flame spread. As a result, one would not expect Lewis number effects (and thus diluent effects) on Sf to be different at 1g and µg, since convection effects are even weaker at µg. This prediction has not been tested experimentally and thus comprises the other major objective of the current work.

Following Ronney et al. (1995), to study the influences of finite rate gas-phase kinetics on partially premixed flame spread, two different gaseous fuels having differing characteristic chemical reaction rates were used, namely CO and CH4. CO was chosen because it is a common product of fires burning in underventilated conditions and thus is of practical importance to partially premixed flame spread which could occur in enclosed spaces. CH4 was chosen as a representative hydrocarbon fragment which could result from pyrolysis of solid fuel beds. These gaseous fuels were added to O2-N2 atmospheres. The resulting Lewis numbers for CO, CH4 and O2 are about 1.0, 0.9 and 1.0 respectively. Since Le and radiation effects are covered separately by changing diluent type in the second part of this work, the intent in this work was to choose fuels and diluents that minimized deviations of Le from unity and variations in radiative properties. O2-N2 atmospheres having O2 mole fractions of 0.18 or 0.30 were employed. The former is close to the minimum O2 concentration and thus finite-rate chemistry effects may be significant for the solid fuel vapors reacting with oxygen (the nonpremixed flame). The latter is far from the limit and thus finite-rate chemistry effects are probably not significant for the non-premixed flame, though finite-rate chemistry is still important for assessing the effect of the partially premixed flame on S_f as discussed in Section 2a. For both fuels and both O₂ mole fractions, f was varied from zero to the lean gas-phase flammability limit.

To study the influences of diluent type, as in Zhang et al. (1992) tests were conducted using He, N2, Ar, CO2 and SF6 diluents. The oxygen Lewis number depends somewhat on the O2 mole fraction; representative values for near-limit conditions are 1.58, 1.04, 0.87, 0.57 and 0.27, respectively, for these diluents. The corresponding Planck mean absorption coefficients, a_p, (a measure of the radiative activity of the gas) at 300K and 1 atm are 0, 0, 0, 28.5 m^-1 and 390 m^-1, respectively. Note that the inverse of a_p is an absorption length (L_P), which is 3.5 cm for CO₂ and 0.24 cm for SF₆. Thus, SF₆ is an extraordinarily strong absorber of its own emitted radiation.

Laboratory Kimwipes were chosen as a fuel for three reasons. First, this fuel produces minimal char which could complicate interpretation of the experimental results. Second, as verified below, this fuel is thin enough (r_st_s = 0.0018 g/cm2) that the fuel behaves as a thermally thin material. Third, because Sf is inversely proportional to r_st_s, for a given atmosphere Sf is higher for this fuel than most other materials commonly used in flame spread experiments. This is advantageous for short-duration µg experiments performed in drop towers.

The estimated uncertainties in the reported values of Sf, O2 mole fraction and total pressure are ±10%, ±1% and ±0.5%, respectively.

To verify that the fuel used can be characterized as thermally-thin, a set of tests was conducted at 1g and µg for one specific atmosphere (33% O₂ in N₂) with varying numbers of fuel sheets. The results (Fig. 6) show that, at least for this mixture, S_f is inversely proportional to the number of sheets (and thus r_st_s) as is required for thermally-thin fuels according to Eq. (1).

It was found that the effect of adding gaseous fuel to the ambient atmosphere was qualitatively similar at 1g and µg in that Sf could be increased significantly by adding gaseous fuel (Figs. 4a - d). For all cases tested, the effect of added gaseous fuel is significantly stronger at µg than 1g, as was hypothesized in Section 2a. For example, at 18% O₂ with CO fuel, S_f increases by a factor of about 2 at 1g as f increases from 0 to 0.37, whereas at µg the increase is a factor of about 5. Also, at both 1g and µg, CO has a greater impact on Sf than does CH₄. This is consistent with the predictions of the model described in Section 2a, and is a result of the higher reaction rates (thus higher q") associated with CO-O₂ chemistry as compared to CH₄-O₂ chemistry. Furthermore, for all but the case having the lowest values of S_f (18% O₂ with CH₄ fuel) case, S_f is actually higher at µg than 1g for large f. To our knowledge, this study is the first to report conditions where S_f is larger at 1g than µg in the absence of a forced flow.

Figures 4a - d show that without premixed fuel, S_f is considerably higher at 1g that µg. This widely thought to be due effects of radiative losses. As explained by Sacksteder and Tien (1994) and others, these losses are larger at µg because the opposed flow velocity U is lower and thus the transport zone thickness d is larger, consequently the total volume over which heat losses are active increases at µg, with a resulting decrease in Sf and the flammable range of atmospheres. The increase in d at µg is confirmed by comparison of Figs. 2a and 2b. An estimate of the impact of radiative losses can be obtained in the following way. The rate of thermal decay dT/dt for radiative loss from a volume of gas at the flame temperature T_f is Q_L/r_gC_p,g, where Q_L is the radiant heat loss per unit volume = 4sa_P(T_f⁴ - T_°⁴) Å 4sa_PT_f⁴ (since T_f⁴ >> T_°⁴) where s is the Stefan-Boltzman constant, thus the radiative time scale is T_f/(dT/dt) = r_gC_p,gT_f/Q_L. The residence time within the flame front is d/U = a_g/U², thus the importance of radiative heat loss scales with the ratio of residence time to radiation time = (a_g/U²)/(r_gC_p,gT_f/Q_L) = 4sa_PT_f³a_g²/l_gU². At 1g, the buoyancy-induced velocities are of the order U Å 1.5(ga_g)^1/3 (deRis, 1969) Å 20 cm/sec for O₂-N₂ atmospheres at 1 atm (deRis, 1969), whereas at µg, U = S_f is typically an order of magnitude lower (see, for example, Fig. 4a), thus it is clear that the effect of radiative losses, which scales as U^-2, will be much more significant at µg. With larger radiative losses, S_f at µg is reduced to well below the corresponding 1g value.

The experimental results shown in Figs. 4a - d can be quantitatively compared with the simple theoretical model employed by Ronney et al. (1995) by substituting S_f itself for the opposed flow velocity (U) at µg (in the prior 1g study U = 1.5(ga_g)^1/3 was employed.) It must be emphasized that no other changes, even property values, were made to the model used by Ronney et al and the model contains no adjustable parameters. Since there is no precise demarcation between non-merged- and merged-flame conditions, both solutions are shown in Figs. 4a - d. Note that in all cases, with gaseous fuel the predicted S_f is larger at µg than at 1g and that for merged-flame conditions, there is little effect of gaseous fuel on S_f. Both of these features are consistent with the discussion in Section 2a.

Figures 4a - d show that the model uniformly underpredicts the increase S_f at µg as f is increased. This can probably be attributed to radiative loss effects which are not considered in the model. At µg, as f increases, S_f increases substantially, and since the impact of radiative losses scales as U^-2 = S_f^-2, radiative losses are much less important at higher f. There is no corresponding change in the impact of radiative losses at 1g since U ~ (ga_g)^1/3 which is much greater than S_f and is essentially constant. In fact, radiative losses are probably unimportant for all 1g flames studied here. As a consequence of this factor, in addition to the predicted increase in the adiabatic S_f at µg as f increases discussed in Section 2a, S_f can be expected to increase significantly more at µg than at 1g as f increases. As evidence of this hypothesis, note that the agreement between model and experiment improves as f increases. While the predictions exhibit only fair agreement with experiment in terms of the actual values of Sf, the agreement is good in terms of the ratios of Sf at 1g and µg at large f, as summarized in Table 1. For example, for 30% O2 in N2 with added CO fuel, at f = 0.2 the experimental spread rates are 3.00 and 3.75 at 1g and µg, respectively, whereas the corresponding theoretical predictions are 2.45 and 3.02; the ratio of S_f at µg to 1g is 1.25 for the experiments and 1.27 for the predictions. The agreement is not nearly as satisfactory where two factors not considered by the model, namely radiative loss and finite-rate chemistry of the non-premixed flame, are strongest, namely 18% O₂ with CH₄ fuel. Radiative losses are strongest for this case since U = S_f is lowest. Finite rate chemistry is most important for this case because the O₂ mole fraction, thus T_f, is lowest.

Figures 5a - e show a surprising effect of the diluent type. For He, N2 and Ar diluents, S_f at µ is always lower than S_f at 1g, which is consistent with numerous prior studies in O2 - N2 atmospheres (e.g., Olson, 1991; Bhattacharjee and Altenkirch, 1993). It was also found that the minimum oxygen concentrations that would support flame spread were lower at 1g than at µg for these diluents, again in agreement with prior studies. Thus, the results for He, N2 and Ar diluents are entirely consistent with prior work. The reason for the higher spread rates and wider flammable range at 1g is likely due to radiative heat losses as discussed in Section 5b.

In contrast to He, Ar and N₂, for CO2 diluent, there is very little difference between the values of S_f at 1g and µg. The minimum O2 concentration is slightly lower at µg than at 1g. Finally, for SF6 diluent, Sf is substantially higher at µg than at 1g for all O₂ concentrations and the minimum O2 concentration is significantly lower at µg (31%) than at 1g (38%). For the O₂-SF₆ atmospheres with low O₂ concentration, the spread rate requires > 2 sec to reach steady-state, thus in this case 5-second drop facility test results are reported also. The µg data at these low O₂ concentrations show a very gradual increase in S_f over time, thus the data from the 5-second drop tower should be considered more reliable. No similar behavior was observed for the other diluents because for SF₆ mixtures at low O₂ concentrations it was necessary to ignite and establish the flames at µg since they could not burn at 1g, whereas in other cases only the transition from 1g to µg spread needed to be accomplished during the drop test.

Two possible mechanisms for this unusual behavior are hypothesized here. The first is a Lewis number effect that is not covered by the model of Greenberg and Ronney (1993). This could happen if the spreading process at 1g is not characterized by creeping flow. Law and Chung’s (1982) analysis predicts that Tf is higher for lower Le and that the effect of Lewis number on Tf decreases as the strength of convection increases. When convection is sufficiently strong, Le does not affect Tf significantly. Therefore, if convection effects were significant at 1g, Tf and thus Sf would be more similar at 1g than µg for varying Le, and furthermore Tf and Sf would be higher at low Le. This is consistent with the experimental observations, however, there is no evidence that convection has a significant effect on the 1g data because Zhang et al. (1992) showed that Greenberg and Ronney’s (1993) model of Le effects on flame spread accurately predicted the experimental results without having to consider flame spread in a mode other than creeping flow.

A second possible mechanism for the observed diluent effects is suggested by the interferometer images of flames in the O₂-SF₆ atmospheres. Figure 7a shows that at 1g, the thickness of the flame in the 42% O₂ - 58% SF₆ atmosphere is a few mm or less, in fact it is so thin that the density gradients are too large to be imaged properly with our interferometer setup, whereas when this flame experiences µg conditions (Fig. 7b) the flame thickness grows rapidly (within 1 sec) to several cm. The increase in flame thickness at µg is of course expected for the reasons mentioned in the previous section, however, the increase for the O₂ - SF₆ atmosphere (compare Figs. 7a and 7b) is much greater than that for the O₂ - N₂ atmosphere (compare Figs. 3a and 3b). This difference cannot be explained based on the estimate d Å a/U; on this basis the ratio of the µg flame thickness (~a_g/S_f) to 1g flame thickness ((~a_g²/g)^1/3) should be smaller for the SF₆ case since S_f is similar for the two cases but a is much smaller for the O₂-SF₆ mixture. For the specific cases shown in Figs. 3b and 6b, representative values of d at µg are (1.4 cm²/sec)/(1.7 cm/sec) Å 0.8 cm and (0.22 cm²/sec)/(1.3 cm/sec) Å 0.17 cm, respectively, using 1000K as a representative average temperature at which to evaluate a. For the O₂-N₂ atmosphere, this estimate is comparable to the thermal thickness seen in the interferogram, however, for the O₂-SF₆ atmosphere, d is underestimated by at least an order of magnitude. Therefore, an additional heat transport mechanism is probably active for the SF₆ case that is unimportant for the N₂ case.

The most plausible source of an additional transport mechanism is reabsorption of thermal radiation emitted by the gases. He, Ar and N2 do not emit thermal radiation and thus the only gaseous species producing significant radiation are the H2O and CO2 combustion products. At the conditions tested in this study, a_p for the combustion products in these diluents are typically 1 m^-1 and thus L_P is much larger than the characteristic size of the flame. Consequently, for these diluents the radiation can be considered optically thin (no reabsorption of emitted radiation) and therefore practically all emitted radiation is lost to the walls of the combustion vessel. However, for CO2 and especially SF6 diluents, L_P is much smaller. For example, at 300K and 1 atm, L_P for a 42% O₂ - 58% SF₆ is about 0.4 cm, and at a representative mean gas temperature of 1000K, L_P is about 6 cm. These values span the apparent thermal thickness seen at µg (Fig. 7b) but are much larger than that seen at 1g (Fig. 7a), which could explain why reabsorption would have a significant effect on S_f at µg but not 1g. The flame front view camera images (not shown) indicate that the visible flame front lies slightly inside (toward the fuel bed) from the region of high temperature gradient (closely spaced fringes) seen in Fig. 7b, thus the flame is characterized by a rapid temperature rise from the ambient atmosphere to the flame front and a much smaller gradient from the flame front to the fuel bed. This would be expected because L_P is a very rapidly increasing function of temperature for SF₆ (Lozinski et al., 1994).

As another test of the radiation hypothesis, a set of experiments were performed for each diluent except Ar at varying pressure (P) for one value of the O2 mole fraction which is well above both the 1g and µg flammability limits for that diluent. As P increases, L_P decreases and thus if the radiation reabsorption hypothesis were correct, the effects of diluent type seen at 1 atm should be stronger (weaker) at higher (lower) P. Results are shown in Fig. 8. At 1g, for all diluents Sf is nearly constant or increases only modestly with P (by a maximum of 40% for He as P increases 8-fold), which is consistent with prior data by Zhang et al. (1992). For reference, according to the deRis (1969) prediction, S_f for thin fuels independent of P. At µg, for He and N₂ diluents, S_f increases with P, which is consistent with prior µg experiments in optically-thin O2-N2 atmospheres (Bhattacharjee and Altenkirch, 1993) and is expected because the impact of optically-thin radiative losses is proportional to 4sa_PT³a_g²/l_gU² ~ P^-1 since a_p ~ P and a_g ~ P^-1. Thus, for He and N₂ diluents the spread rates at 1g and µg tend to converge as P increases. For SF₆ diluent, completely different behavior is observed, suggesting a different mechanism is operative. As P increases, the µg spread rate diverges from the 1g value, which would be consistent with an increasing effect of optically-thick radiation, leading to a reduction in the radiant heat loss and an augmentation of heat transport by radiation. Of course, if P were increased sufficiently, the effect may saturate since little radiation would then escape.

The Lewis number effects described by Greenberg and Ronney (1993) and Law and Chung (1982) are independent of P, whereas the data in Fig. 8 show that the comparison between S_f at 1g and µg is strongly dependent on P. This is further evidence that the observed effect of diluent type cannot be attributed mainly to Lewis number influences. One additional means to discern between the Lewis number and radiation hypotheses would be to perform experiments in O2-Xe atmospheres because these would provide a low Le_O2 but do not emit thermal radiation. Unfortunately, the cost of Xe is prohibitive.

Experiments on flame spread over thin solid fuel beds in varying atmospheres were conducted at earth gravity and microgravity. It was found that when a sub-flammability-limit concentration of a gaseous fuel (CO or CH4) was added to the premixed flame front, S_f increased at both 1g and µg, however, the increase was greater at µg. The observed S_f trends were close to those predicted by a simple model which considers the effect of the premixed fuel to be an increase in the heat flux to the fuel bed caused by a partially-premixed flame front upstream of the usual non-premixed flame front. Experiments without premixed gaseous fuel were conducted in a variety of O2-diluent atmospheres using He, N2, Ar, CO2 and SF6 diluents. He, N2 and Ar results were consistent with prior 1g and µg experiments in that the spread rates were higher and the minimum O2 concentrations were lower at 1g. In contrast, for CO2 diluent these properties were practically the same at 1g and µg and for SF6 the trends observed for He, N2 and Ar were completely reversed. These results are proposed to be an effect of reabsorption of thermal radiation emitted by the gases, which is important only for diluents with sufficiently large absorption coefficients.

In future work, we intend to study the effects of a weak forced flow on S_f at µg. It has been shown experimentally (Olson, 1991) that for flame spread at µg in optically-thin atmospheres, S_f increases as U increases because, as discussed earlier, the impact of optically-thin radiative losses is proportional to U^-2. In contrast, if reabsorption of emitted radiation is an important factor for SF₆ diluent, S_f should decrease with increasing U because d would decrease and thus the size of the region where reabsorption could be beneficial would decrease. Additionally, radiative emissions will be measured by thermopile-type radiometers to determine if the trends with pressure and gravity level proposed here can be verified, and thermocouple-based temperature measurements will be made to quantify the temperature characteristics inferred from the interferometer images.

Figure 7. Interferometer images of spreading flames. Atmosphere: 42% O2 in SF6 at 1 atm. Field of view is 4.2 cm x 3.2 cm.

Figure 8. Flame spread rate vs. pressure for He, N2, SF6 and CO2 diluents at fixed O₂ mole fractions.