As soon as the temperature falls, the mercury contracts, and thus opens the end of the tube B. The gas is thus turned on, and the temperature rises until the regulator again cuts off the gas. In order to prevent the possible extinction of the flame by the regulator, the brass tube which carries the gas to the regulator is connected with the tube which brings the gas from the regulator to the burner by a small brass tap (Fig. 2). This tap forms an adjustable bye-pass, and thus a small flame can be kept burning, even though the regulator be completely shut off. It is obvious that the quantity of gas supplied through the bye-pass must always be less than that required to maintain the desired temperature. This regulator, placed in a beaker of water on a tripod, will maintain the temperature of the water during four or five hours within 0.2 deg. C., and an air bath during six weeks within 0.5 deg. C. To sum up briefly the method of using the regulator:--Being filled with mercury to about 1\2 inch below the T, attach the gas supply as in diagram (Fig. 2), the brass tap being open, and the tube B unclosed by the mercury. Allow the gas to completely expel the air in the apparatus. Push down the tube A so that the end of B is well under the surface of the mercury. Turn off the tap of the bye-pass until the smallest bead of flame is visible. Raise A and B, and allow the temperature to rise until the desired point is attained. Then push the tubes A and B slowly down until the flame is just shut off. The regulator will then keep the temperature at that point. ~Will's Test for Nitro-Cellulose.~--The principle of Dr W. Will's test[A] may be briefly described as follows:--The regularity with which nitro- cellulose decomposes under conditions admitting of the removal of the products of decomposition immediately following their formation is a measure of its stability. As decomposing agent a sufficiently high temperature (135 deg. C.) is employed, the explosive being kept in a constantly changing atmosphere of carbon dioxide, heated to the same temperature: the oxides of nitrogen which result are swept over red-hot copper, and are then reduced to nitrogen, and finally, the rates of evolution of nitrogen are measured and compared. Dr Will considers that the best definition and test of a stable nitro-cellulose is that it should give off at a high temperature equal quantities of nitrogen in equal times. For the purposes of manufacture, it is specially important that the material should be purified to its limit, i.e., the point at which further washing produces no further change in its speed of decomposition measured in the manner described. [Footnote A: W. Will, _Mitt. a. d. Centrallstelle f. Wissench. Techn. Untersuchungen Nuo-Babelsberg Berlin_, 1902 [2], 5-24.]