run through the oil for at least one minute before the flash occurs. It may perhaps seem, at first thought, that a continuous current of air would dilute the vapor to such an extent that the flashing-point must be materially raised, and that this effect must be more marked as the velocity of the current is increased. This is, however, not the case. On the contrary, while a slow, continuous current raises the flashing-point appreciably, a sufficiently rapid one gives nearly the same results as the intermittent method; nor does any further increase in the velocity alter the flashing-point to a sensible extent. It has indeed been found that a large dilution of kerosene-vapor with air is necessary to furnish the conditions for the most violent explosion; and these conditions are also those for the readiest flash by this method of testing. The most explosive mixture, according to Chandler, is formed by nine parts of air to one of vapor. The passage of a large quantity of air through the oil tends, of course, to make the flashing-point higher, by carrying away with it the more volatile portions which determine the flash, and this effect is greater when the quantity of oil is small and the air-current long continued. It is, consequently, necessary in the employment of this method to know the minimum quantity of oil and the maximum duration of air-current which will permit concordant results. These limits have been ascertained in a recent investigation,[1] the results or which are given a little further on.
A tester of still simpler construction than that of Liebermann has also 
Fig. 3.been proposed.[2] It consists, as shown in the cut, of a glass cylinder, closed at one end by a cork, through which a small bent tube, d, c, b, passes. Just within the cork the end of this tube contracts to a small orifice. The other end of the tube connects with a small bellows, or other source of slightly compressed air, the flow of which can be regulated by the pinch-cock e.
Experiments made with cylinders of different dimensions have shown that the best results are obtained when the diameter is between 2·5 and 4 c.m. The length (if only great enough to allow at least the minimum quantity of oil to be used) makes no difference. Cylinders of the same diameter but of different lengths, when filled with oil to within the same distance from the top, all give the same flashing-point. Change in length in such cases is simply equivalent to change in the quantity of oil employed in the test, and it has been proved that the quantity of oil does not affect the determination when it is above a certain minimum.
