Fire in Microgravity
In space, flames don’t extinguish under the same low-oxygen conditions that would put them out on Earth, setting the stage for dangerous flare-ups.
In the late 1980s the American and Soviet space programs were planning longer space missions, some for a half-year or more, and these plans led them to a practical question: If there is an electrical short or other spark, followed by a small spreading fire on the spacecraft, how can it be extinguished in zero-gravity orbit? The first impulse is to create limited-oxygen conditions—to chemically suffocate the spreading flame much like a small kitchen fire is smothered with a blanket. Quite unexpectedly, however, experiments on the Soviet Mir space station using burning plexiglass rods, and the NASA Space Shuttle experiments on flame spread over thin sheets of cellulose (which one of us, Olson, directed), displayed a curious behavior when the oxygen was decreased: The flame front initially shrank as expected, but then all of a sudden, it broke apart into many three-dimensional “flamelets” that resembled hemispherical caps. These cap-shaped flamelets zigged and zagged over the material surface toward the oncoming oxygen in random-looking patterns, even when oxygen levels were reduced lower than would have extinguished an ordinary Earthbound flame. The NASA astronauts and the Mir cosmonauts found that in space the flame successfully fought back against impending oxygen starvation by exploiting shape change, morphing itself from a 2D front into an array of 3D flamelets. The flame had not gone out, and the fire safety implications were frightening.
To study the behavior of these flamelets further, we have tested flames in facilities designed to mimic microgravity conditions at NASA Glenn Research Center, and in 2000 we developed other specialized combustion chambers at Michigan State University. These studies have led to an increased understanding of the behavior of these flamelets and to improved fire safety in space, but also give insights into certain potentially dangerous types of fire spread on Earth.
The flame-to-flamelet phenomenon seen in space experiments has some cousins on the ground. Any such metamorphosis of one flame type to another is called a transition. Other examples of transitions include the ignition or extinction of a flame, or a flame becoming a detonation. Another transition commonly used from ancient civilizations to relatively modern times was to employ smoldering coals from a fire, preserved under ash or carried in an insulated container, to start a new hearth blaze the next day. The tiny flame, reduced to an almost undetectable reaction, would burst into fire upon renewed contact with ambient air.
Such transitions also occur on a large scale. In a forest fire, prevailing winds can fan the flames and produce enough heat to create its own air circulation front, which spreads the fire as a continuous front in the opposite direction to the prevailing wind. Terry Clark at the National Center for Atmospheric Research and his coworkers showed that when the opposing wind dies down, this uniform fire may change into an array of fingering fronts, followed by weakly burning regions. They also created a computer model that correctly reproduced the development of fingers over a scale of a few kilometers, just as in the actual fire.
The study of flame spread began in the 1960s, with Richard S. Magee and Robert F. McAlevy doing experiments at Stevens Institute of Technology in Hoboken, NJ, on flame spread over propellants for military applications in rockets and missiles. Simultaneously, fire-prevention researcher John deRis was theoretically studying flame spread as a Harvard PhD student under coadvisors Howard Emmons (called the father of fire safety research) and George F. Carrier, the much-lauded applied mathematician and fluid dynamicist. In the past half century, flame spread has grown into a fertile branch of combustion research, with applications ranging from large-scale forest fires to domestic, automobile, and industrial fires, to military applications of propellants, to small-scale flame spread over wire bundles. In the 1990s, Howard Baum of the National Institute of Standards and Technology, and another former PhD student of Carrier, devised a theoretical explanation for the flame-to-flamelets transition, creating numerical simulations to show that competition for scarce resources (in this case, oxygen) required the fire to make a compensating geometric adaptation.
In the 1990s an experimental breakthrough occurred when Ory Zik and his collaborators, then at the Weizmann Institute of Science in Israel, produced on-Earth burning patterns nearly identical to the NASA Space Shuttle results. A confined space and controlled airflow were key to his results. Zik’s experiments used flame spread over thin cellulose sheets, which were held in a shallow channel only a few millimeters in height, created by separating a top glass plate and a lower surface on which the sample rested. As the inflowing air was dialed down, the flame eventually broke into numerous smolder fingers that made an intricate pattern of bifurcations and extinctions across the surface. The smolder front managed to survive in its resource-poor environment while consuming a large fraction of the available cellulose fuel.
It turns out that this “near-limit” type of flame spread—existing close to extinction, or quenching—appears in other guises on Earth, although the process often happens in the reverse order of what is seen in space, transitioning instead from flamelets to flame and fire spread. A terrible instance occurred on Swissair Flight 111 on September 2, 1998. After takeoff a small, initially undetected fire originated behind the cockpit bulkhead inside the in-flight entertainment system, possibly due to wire arcing. The insulation blanket was ignited and a creeping flame (possibly a flamelet) spread along it inside a narrow gap. The burn patterns on the pieces recovered later were consistent with small and weak flames that spread slowly, driven by low-velocity ventilation flows between the cabin walls. The initial release of smoke alerted the crew, but they were unable to pinpoint its origin. Eventually, a tiny flame breached a silicone vent cap, allowing a much greater airflow into the narrow channel and leading to a rapidly growing flame. By the time the crew realized the extent of the fire, it was too late and the aircraft no longer responded to cockpit commands. The flight crashed into the Atlantic Ocean, killing all 229 passengers.
The adaptation and survival of a spreading flame, whether on Earth or in space, is therefore a major safety concern. If combustion is somehow started in a confined, narrow channel with a lack of available air, the flames may burn at an undetectable level. As in the case of Swissair Flight 111, when larger amounts of air finally break through, the teetering, near-limit flamelet can grow into a deadly, large-scale fire. Such flamelets thus serve as ignition sources, much like pilot lights in stovetops or matches. Similar to banked coals, if uncontrolled, the nascent flamelets can wreak havoc: In space, and in some cases on Earth, flamelets are the enemy.
Go With the Flow
Flamelets and other smolder-type fires are relatively rare on Earth, and the physics that create that limitation are intrinsic to setting up experiments that mimic flame spread in microgravity.Central to flame spread is the fueling cycle. The solid that is burning obviously is supplying the fuel that the flame consumes, but it’s a circuitous process: The flame heat gasifies the solid surface, which feeds the flame with fuel molecules that mix with oxygen molecules, which then react and release more flame heat, which in turn gasifies more of the solid surface.
Fuel use is related to the two ways that spreading flames can be extinguished. The key value is what’s called the Damköhler number, which is the ratio of the time required for the gasified fuel to flow through the flame and the chemical time needed to burn it. When the Damköhler number approaches 0, the air inflow is so fast that the available time for mixing and burning becomes vanishingly small, and this creates what’s called a blowoff limit. Whenever one blows out a lighted match, this is a blowoff extinction.
On the other hand, if the airflow is slow enough that there is insufficient oxidizer, the Damköhler number approaches infinity, meaning the flame weakens and can be extinguished by conductive and radiative heat losses to surrounding surfaces or the environment. This near-limit extinction manifests itself not as a dramatic blowoff but as a nuanced transition either to smoldering, transient flickering or, in the case of our research, flamelet spread.
There is, however, a catch. On Earth, the blowoff limit is easy to achieve, whereas near-limit extinction is hard. Why is that? On Earth, the flame heat generates a buoyant flow that can supply all the oxidizer it needs. This phenomenon is commonly called air entrainment, and occurs in a fireplace or a wood stove, where the fire draws in oxidizer from the room. When a flame entrains air, it cannot approach the near-limit condition. Thus, on Earth the near-limit state is a rarity, except when conditions suitable for smoldering are attained.
In space, the situation is completely different: Buoyancy requires the balancing force of gravity to exist, so there’s no air entrainment in microgravity. If the air inflow speed is reduced, the flow-through time becomes huge, approaching again the content-rich limit where the Damköhler number approaches infinity. Microgravity allows us to generate the ideal conditions for a bare-knuckle in-flame competition for scarce oxygen. And it also informed how we could control airflow in experiments on Earth to simulate space conditions.
The early 1990s Space Shuttle program opened a window for microgravity combustion research. Outstanding work was done on problems such as how the lack of gravity alters the combustion of fuel droplets and the shape of candle flame, both of which showed remarkable agreement with theoretical predictions. The NASA research program was motivated by spacecraft fire safety for the Shuttle, the International Space Station, and potential Mars missions. However, not all of the NASA-supported research projects actually flew on the Shuttle. Many underwent testing at the NASA Lewis Research Center (renamed after astronaut John Glenn in 1999) in Cleveland, which houses towers designed to drop samples for a duration of either 2.2 or 5.18 seconds, briefly mimicking microgravity conditions.
At NASA-Lewis, two of us (Olson and Miller), aware in the late 1990s of Zik’s groundbreaking research, speculated that it might be possible to construct an on-Earth apparatus to produce the flame-to-flamelets transition close to extinction. Drop tests of these samples in the 1990s showed that flamelets formed almost immediately after the sample went into free fall and the buoyancy-induced entrainment flow had ceased.
Although NASA’s ground-based drop facilities were state-of-the-art, there are two major drawbacks to using them. Drop tests required specialized facilities and equipment, heavy engineering support, and extensive technician participation; thus they are very time-intensive and quite expensive. In addition, the 2.2-second facility provides only about 1 second of viable test time, and the 5.18-second facility about 4 seconds, after the buoyant flow has ceased. One to four seconds is a very short time for transient flamelet formation and spread. Despite these limitations, our research has conducted well over 200, and counting, 5.18-second drop tests. However, flamelets spread slowly, at about 0.1 millimeters per second. Therefore, we needed an on-Earth facility that could run continuous tests for up to tens of minutes.
For that purpose, the Michigan State University Hele-Shaw apparatus was constructed in 2000 (and named for Hele-Shaw flow, which happens between two parallel, flat plates with a tiny gap between them). The apparatus was a marriage of the Zik device, which utilized a narrow gap, and the 2.2-second NASA drop tests. The narrow gap suppresses vertical buoyant flow, and by providing uniform air inflow across the entire sample width, the apparatus makes the initial flame almost perfectly two-dimensional. Our goal was to visualize the transition of a buoyancy-suppressed, perfect 2D flame front into a 3D array of zig-zagging flamelets. After detailed trial-and-error across ranges of flows, gap widths, and methods of sample ignition, the facility produced excellent results.
NASA’s engineers subsequently designed and constructed a modified narrow channel apparatus that was more sophisticated and compact, and also measured exhaust gases while varying the input gas concentration.
Flamelet Life Cycles
Our experimental research using these narrow channel apparatuses and the NASA drop facilities aimed to answer several questions about the basic nature of flamelet spread.
Our first question dealt with flamelet persistence. Are flamelets merely ephemeral entities that flame out fast en route to extinction? Our videos showed that individual flamelets are highly unsteady. Some flamelets extinguished or “died” when other flamelets closed in around them to choke off their oxygen. Other flamelets bifurcated or “gave birth” to new flamelets as fresh surface area became available. Still other flamelets perished almost immediately after “birth.” Certain solo flamelets never gave birth nor did they interact significantly with their neighbors. Individual flamelets possessed an average “life span,” which we measured directly from video and later also statistically analyzed.
Concerning the matter of “birthrates,” we carefully tracked 70 separate flamelets and established that slightly more than half, about 52.5 percent, of the flamelets “gave birth” while the remainder extinguished without bifurcating. For a population to remain steady on average, each bifurcation must result in one viable flamelet, to create a 50 percent persistence rate. However, our examination of the data showed birth rates of consistently more than 50 percent. Our explanation for this apparent “growth rate” was that we did not adequately measure rapid extinctions. Support for our explanation, and for the assertion that flamelets as a group really are steady, came from the fact that the burned surface was consistently 62 percent. No more and no less of the sample was being consumed as the flamelets marched across its surface. (When the air flow was in the same direction as flamelet spread, they propagated more slowly but consumed more of the available fuel, 88 percent.)
Because of this highly unsteady individual flamelet behavior, we formulated three hard requirements for steady flamelet spread as a group: The time scale for monitoring steady state behavior must be larger than the average flamelet life span; the bifurcation and extinction rates must be within a few percent; and a steady-state flamelet population must produce a constant average burned fraction of the sample. If any one of these three criteria was not fulfilled, the flamelet process was not steady state.
Our measurements showed that despite the statistical nature of the bifurcation-extinction flamelet spread process, some features remained unchanged on the individual flamelet level. For example, at a given air inflow rate we measured a near-constant spread rate toward the incoming flow for the individual flamelets. However, when the velocity of the inflow air decreased below 5 centimeters per second, the flamelet spread rate entered a plateau at which it remained until extinction at an opposed air inflow of approximately 0.5 centimeters per second.
Our second question addresses the flamelet distribution pattern, which depends on the overall population as well as individual flamelet characteristics. Individual flamelet behavior is impossible to predict; as a group, however, flamelets showed a uniform distribution. The flamelet population is dictated not only by oxygen supply but also heat losses to the surroundings. The width of flamelet fingers scales with the heat losses, whereas the gap between flamelet fingers scales with both oxygen flow rate and heat losses. The flamelets optimize their spacing in response to the scarce resource, the oxidizer: As the airflow decreased, the flamelets moved further apart in order to survive. We examined the dependence of flamelet population on heat losses to the apparatus surfaces and the surroundings in the narrow channel apparatus. The surface heat losses were varied by changing the gap height; as the gap increased, the number of flamelets decreased. In general, the apparatus produced on the order of ten flamelets per test.
A quantitative method of evaluating the influences of heat losses on extinction examines surface radiative heat losses to the ambient air, and conduction heat losses from the flame to the ambient air and to the substrate. We modeled the ratio of the heat lost by the surface to the heat arriving to the surface from the flame. When this ratio is sufficiently large, flamelet extinction may occur. When we applied heat loss analysis to flamelet spread with 5-millimeter gaps on either side of the sample, we found that the loss ratio increases as the flamelet approaches extinction at low flow velocity. As the flow velocity is decreased, a greater proportion of the flame heat is used to offset heat losses by conduction to the channel walls and radiative losses from the surface. At a sufficiently low inflow velocity, the flamelet can no longer sustain these losses and it finally extinguishes.
Our third question addresses the fact that we have consistently referred to oxygen as the limiting resource. Is this true? To answer this we analyzed the overall burning stoichiometry by systematically evaluated flamelet spread rate, sample thickness, and sample width. The maximum oxygen consumption rate between atmospheric oxygen concentration (21 percent) and extinction concentration (about 14 percent) was 7 percent. Our calculations gave a fuel-rich global stoichiometric ratio—the fuel-to-air ratio needed for complete combustion—of 1.25, indicating that oxygen is indeed the limiting resource, and its rate of supply largely determines the amount of sample consumed. In addition, we measured carbon monoxide in the combustion products, demonstrating fuel-rich conditions as well as potential toxicity.
Our fourth question addresses the fundamental nature of the so-called steady state. Is it truly steady or is it, on a more fundamental level, an oscillation about some fixed value? The latter is often the real world case, and also occurred in our experiments: Our flamelet bifurcation and extinction rate data were unambiguously oscillatory. A bifurcation at one location was soon followed by an extinction, and vice-versa. The oscillation rate had a constant amplitude and period until near the end of the sample when additional space became available. The oscillations showed a sinusoidal trend, hence the oscillations occurred around a mean value. These stable oscillations were still further indication that the mean population was steady and stable. Our data showed that between 120 and 150 seconds, the bifurcation and extinction rates are very similar, after which the bifurcations outnumber extinctions as the population of flamelets grows while the flame approaches the end of the sample.
We have attributed the oscillations in the bifurcation and extinction rates to time lags, during which one of the two processes overcompensates for changes in the other. The time lag phenomenon is innate in the logistic population model, which can be used to characterize the flamelet process. A statistical population described by this logistic equation typically cycles around its so-called carrying capacity—the amount of burnable oxygen supplied divided by the oxygen needed per flamelet—owing to the influence of time delays in response to environmental changes that alter bifurcation and extinction rates. In our tests, the time delays can be attributed to the distance of flamelet spread needed to permit the next bifurcation to occur. Our measured time lag was approximately 25 seconds, which equals the half-life of the average flamelet life span of 37.2 seconds. This time lag value is also one-fourth of the population cycle of 100 seconds.
Physically, the time lag is the time required for one flamelet to influence the bifurcation potential of flamelets around it. To bifurcate, a flamelet must spread laterally to approximately twice its normal width. The lateral spread distance from the flamelet centerline depends on the spread rate, time, and the angle of lateral growth. The lateral growth angle is usually between 15 and 23 degrees, with an upstream spread rate of approximately 0.1 centimeters per second. At this rate, bifurcations should occur at approximately every 2.4 to 3.7 centimeters along the finger flame arc.
The preponderance of our discussion has been about flamelet birth, or bifurcation, which is one of the more fascinating sights in flamelet spread. The angle of lateral flamelet growth depends on the flow velocity perpendicular to the flamelet. As the flamelets widen, the central portion weakens as it becomes more distant from the surrounding edges; thus the flamelet shrinks from the middle. When the weaker portion of the advancing flamelet falls behind the two outer edges, it curves into a concave or cusp shape, with the trailing central cusp regions receiving less oxygen for combustion than the outer regions. If the oxygen concentration is sufficiently low, the flamelet cannot even spread far enough laterally to undergo bifurcation. In this case it will simply extinguish in its original track.
A fundamental question of interest is the possible lack of symmetry in flamelet formation, commonly called hysteresis. Our specific question was, can flamelets reignite a flame front when the oxidizer flow is ramped back up, or will the increased flow blow out the flamelets? We tested numerous flow acceleration rates and determined, in all cases, that the flamelets acted like tiny pilot flames that reignited the flame front. Like the birthing process described above, there was a characteristic lag time associated with breakup and flame reformation. This 15-30 second lag time produced over- and undershoots in the flame and flamelet front speed, after which steady propagation was re-established.
The macroscopic nature of the flamelet and flame spread process is a function of the opposing flow speed. As the flow speed is decreased from approximately 25 centimeters per second, the flame spread rate actually rises because the fuel entering the gas has time to burn in the oxidizer flow. The flame spread rate peaks when the opposed flow is about 12 centimeters per second, at a value of about 0.3 centimeters per second. As the flow speed is further decreased, one sees the formation of cusps, then flamelets, and finally, at about 3 centimeters per second, extinction.
Another important question for the flamelet spread process dealt with the veracity of the simulated microgravity state: Was gravity truly unimportant in the narrow channel apparatus? To answer this question we constructed the Michigan State apparatus so it could be inverted. In principle, the inverted configuration, with the window on the bottom and the heat loss copper plate on top, should produce identical results to the normal configuration. Any observed differences in the test results serve as a measure of the influence of gravity. Our results were nearly (but not exactly) identical. The bandwidth of flow conditions that produced flamelets was slightly narrower, with the flamelets extinguishing at slightly higher flow velocities, suggesting the inverted flamelets were slightly weaker. We determined that the inverted apparatus had incrementally larger heat losses because the flame was closer to the sample and thus also to the copper heat-loss surface, which weakened the flamelets. But the overall behavior of flame-to-flamelet transition and flamelet spread was very much the same regardless of orientation. We concluded that the narrow channel apparatus functions to produce acceptable simulated microgravity conditions: The buoyant flow is largely suppressed by the decreased height of the channel, leaving the other components of the flow to contend with the flamelet spread and structure.
Mind the Gap
Although flamelets have been studied for several decades now, there’s still much left to learn about their dynamics, and the best ways to apply that knowledge to fire safety in microgravity conditions. Our results show that in order to extinguish a flamelet while fuel is still available, the air flow must be suddenly turned up, then shut off entirely. NASA protocols have astronauts turn off ventilation systems as soon as a fire is detected, but to turn it on again to clear smoke once the fire is extinguished. If a flamelet fire goes undetected, this procedure could lead to a sudden, large flareup as these near-invisible flamelets receive an onrush of oxygen.
The danger of fire in space is not a hypothetical one. The Mir space station had two, a smaller one in 1994 and a famously damaging one in 1997. Both involved a faulty oxygen generator. In the 1994 case, a cosmonaut used a handy jumpsuit to extinguish the fire, but none of the crew noticed for a few minutes that the jumpsuit itself was still smoldering. If that fire had flared up before being detected, the danger could have been much greater.
A way to detect these near-invisible fires also would be an important tool for fire prevention, and could perhaps be served by sensors that are sensitive to very low levels of combustion exhaust products. In the meantime, having several different mechanisms to simulate microgravity conditions on Earth will help advance the study of this quiet but potentially deadly type of fire spread, hopefully before long-term, long-distance space missions come to fruition.
- Clark, T., J. Coen, M. A. Jenkins, and D. Packham. 1996. A coupled atmosphere-fire model: Convective feedback on fire-line dynamics. Journal of Applied Meteorology 35:875–901.
- Olson, S. L., F. J. Miller, S. Jahangirian, and I. S. Wichman. 2009. Flame spread over thin fuels in actual and simulated microgravity conditions. Combustion and Flame 156:1214–1226.
- Olson, S. L., F. J. Miller, and I. S. Wichman. 2006. Characterizing fingering flamelets using the logistic model. Combustion Theory and Modelling 10:323–347.
- Wichman, I. S., S. L. Olson, F. J. Miller, and S. A. Tanaya. 2013. Experimental evaluation of flame and flamelet spread over cellulosic materials using the narrow channel apparatus. Fire and Materials 37:503–519.