Welcome to the University of Michigan - Dept. of Mechanical Engineering & Applied Mechanics

Origin of the Dynamic Growth of Vapor Bubbles Associated with Vapor Explosions


An explosive type of vapor bubble growth was observed during pool boiling experiments in microgravity using R-113, where heater surface superheats as high as 70 °C were attained at nucleation. This corresponds to approximately 65 % of the computed superheat limit of the fluid, compared to the approximate 30 % observed at earth gravity for the same system. Photographs and measurement s of the vapor bubble growth provide evidence for rates of growth not accountable by conventional models. The photographs reveal that the liquid-vapor interface of the explosive bubbles become wrinkled and corrugated, leading to the conclusion that some type of instability mechanism is acting.

The classical hydrodynamic instability theories of Landau and Rayleigh-Taylor, used in conjunction with a model of the early growth of spherical vapor bubbles developed by the authors, predict that the early growth should be stable. These theories do not consider the effects of heat transfer at the interface, which is believed responsible for the observed instability of the evaporating surface. This was confirmed by the mechanisms proposed by Prosperetti and Plesset which, although including the effects of heat transfer, required that the unperturbed liquid temperature distribution be known at the moment of onset of the instability. This is generally unknown, so that no comparisons with experiments were possible up to this point. The present pool boiling experiments conducted in microgravity, some of which result in the explosive vapor bubble growth referred to, permit the precise determination of the unperturbed liquid temperature distribution using a model of the early vapor bubble growth along with the measurement of heater surface temperature at nucleation. The limited results to date provide good agreement with the mechanisms proposed by Prosperetti and Plesset.

Experiments and Results

Pool boiling experiments with R-113 were conducted in microgravity in the 132 m evacuated drop tower facility at the NASA Lewis Research Center, which provided 5.18 seconds of free fall with effective body forces on the order of 10--5 g. Photographs of the boiling process were obtained with a D.B. Milliken camera operating at 400 pictures per second. The drop vehicle (shown in Figure 1), vessel and associated electronics were designed to withstand the vacuum conditions and the 50 g impact at the bottom of the drop tower. The test vessel has internal dimensions of approximately 12.5 cm by 14 cm by 27.9 cm, and consists of thick aluminum plates to assist in providing initially uniform internal temperatures. Two heater surfaces are installed in one wall of the test vessel; each consists of a 400 Ĺ thick semi-transparent gold film sputtered on a highly polished quartz substrate, and serves simultaneously as a heater, with an uncertainty of ±4 % in the measurement of the heat flux, and a resistance thermometer, with an overall uncertainty of ±1.0 °C. The heater is rectangular in shape, 1.91 cm by 3.81 cm. Degassed commercial grade R-113 (trichlorotrifluoroethane, ) was used because of its low normal boiling point (47.6 °C), which minimized problems associated with heat loss to the surroundings, and because of its electrical nonconductivity, which made it compatible for direct contact with the thin gold film heater. A pressure transducer was employed to measure the system pressure, with an uncertainty of ±0.172 kPa. Chromel-constantan Teflon sheathed thermocouples with ice reference junctions were used to measure the liquid and vapor temperatures, with a total uncertainty of ±0.05 °C. The desired subcooling was obtained by increasing the system pressure above the saturation pressure corresponding to the initial liquid temperature. In the absence of buoyancy, an initially motionless liquid remains stagnant upon heating until the onset of boiling, and the temperature distribution at the moment of incipient boiling can be determined from a conduction heat transfer analysis. Detailed descriptions of the experimental apparatus and procedure are given in Ervin et al. (1992).

Figure 1: Drop vehicle vessel

As is well known, conventional growing vapor bubbles have a smooth interface. However, under certain conditions with pool boiling in microgravity, the interface took on a rough appearance, accompanied by rapid growth rates. Figure 2 shows a typical roughened hemispherical vapor bubble 17.5 milliseconds after nucleation. This type of roughened surface has not been observed with boiling in earth gravity, most likely due to natural convection.

Figure 2: Enlargement of a growing vapor bubble at 17.5 ms for Experimental Data Code PBMT 1102.800

Two typical growth sequences are presented in the Figure 3 together with the operating conditions. The differences in the wall superheats at nucleation are particularly noteworthy, and are associated with the dramatic differences in the growth rates, especially obvious by comparing the bubble sizes at 7.5 milliseconds. It is of interest that the very first picture A in in the figure below already exhibits a roughened surface, which implies that the onset of the interfacial instability is somewhere between 0 ~ 2.5 milliseconds. It is also to be noted that the bulk liquid is slightly subcooled in Figure 3(a) and considerably subcooled in Figure 3(b).

Figure 3 (a) and (b): Comparisons of early vapor bubble growth behavior under microgravity with different levels of heat flux and subcooling

The growing bubble radius is included, together with the corresponding measurements, all of which have unstable interfaces, as shown in Figure 3(a). The apparent unstable regime exists over a wide interval of time. However, a constraint must be imposed within this regime in that the wavelength cannot exceed the bubble diameter. This is called the ‘k limit’, defined as , for practical reasons as discussed previously by Sturtevant and Shepherd (1982), and is included in Figure 4. The "most unstable regime" is determined by the intersections between the ‘k limit curve’ and the maximum wavenumber curve, giving a range of times between 1.2x10-6 and 3.0x10-4 seconds as the interval of instability. The radii measurements indicated are beyond this interval, demonstrating that the onset of the instability occurred earlier. The corresponding computations were carried out for the case shown in the Figure 3(b), which demonstrated a distinctive stable bubble growth, and the results are presented in Figure 5. The conditions required for the "most unstable regime" are satisfied only marginally, which indicates that the probability for instability will be low, considering that the ‘ k limit’ curve is the lower bound. The result is in fair agreement with the measurements, in which the measured bubble radii are between the computational limits.

Figure 4: Neutral stability diagram and growing vapor bubble radius for R113, experimental data code PUBMT 1102.800, Figure 3 (a).

Figure 5: Neutral stability diagram and growing vapor bubble radius for R113, experimental data code PUBMT 1102.800, Figure 3 (b).


The occurrence of extremely rapid evaporation at a vapor bubble interface has been demonstrated with superheats far below the superheat limit. This occurred with pool boiling in a microgravity environment, and is attributed to the growth of an interfacial instability. Such a process can produce an abrupt pressure rise and be quite destructive if a sufficiently large surface area is present.

Sources of perturbation growth other than an instability can arise at the liquid-vapor interface of a bubble. The present plane instability model including heat transfer reasonably predicts not only the occurrence of explosive bubble growth, but also the wavelength of the unstable interface. Local heat transfer at the interface is an essential mechanism to the instability, where the heat transfer increases at a trough while it decreases at the crest, which can produce differential vapor recoil. This can lead to the distortion of the crests such that liquid particles become entrained with the vapor, departing the interface, and which can then serve to further increase the evaporative flux.

This page was last updated: 26 April 1996 by Patrick M. Fahey, ME Senior

For questions or comments, email pfahey@engin.umich.edu