Abstract
Experimental and theoretical results are described for the moving flame experiment, where a heat source moves under a horizontal, annular channel filled with liquid mercury. The physical range under discussion consists of the creep flow limit, in which inertial forces are less than viscous forces. A mean surface flow was observed on the mercury. This flow was up to four times greater than the heater speed, and in an opposite direction. The ratio of mean flow to heater speed was theoretically and experimentally found to be proportional to the square of the thermal amplitude, in agreement with a special limit of Davey's earlier results. The basic flow here, as contrasted to those flows analyzed by earlier theories, appears to be generated by surface tension. However, the Reynolds stress mechanism which generates the mean flow is the same. Some experimental flows were so rapid that the range of validity of the present, as well as some earlier theories, was exceeded. The present theory suggests that the higher order nonlinearities may have cancellative effects, which implies that the present and some earlier theories are valid beyond their normal range of validity.