90 BRIDGMAN. ' 



the ends of the wires was then melted and the wires fused together by 

 a small blue gas flame. A flame of the required size may be obtained 

 from a jet issuing from a piece of drawn out glass tubing. The gas to 

 give the blue flame is supplied from a gasometer, which may be im- 

 provised from a couple of old tin cans, filled with a mixture in the 

 right proportions of illuminating gas and air. The joint so formed is 

 very strong mechanically. If the wires are not in satisfactory align- 

 ment they may be made so by hammering. The method is consider- 

 ably superior to any simple electrical method of welding with which I 

 am familiar. Of course there is no objection to smoothing the junc- 

 tion with a little fine emery paper; by this means a joint can be made 

 that cannot be detected by the sense of touch, or of sight, after the 

 wires are enamelled. The wires used were all cut to standard lengths 

 by jigs, so that after enamelling the joints could be located by measure- 

 ments. I have used this same method to make junctions as small as 

 0.004 inch, which are considerably more than two and one half times 

 as hard to make by the ordinary method. 



A great many measurements were made with specimens of the 

 cylindrical form described above, and every effort was made to make 

 the method give good results. The measurements with any one speci- 

 men were usually regular, and would repeat, and were apparently as 

 good as could be asked. The trouble was that the numerical results 

 obtained with different specimens of the same metal did not agree. 

 The lack of agreement was much worse for some metals than others. 

 I finallv came to the conclusion that the method is essentiallv limited 

 by the demands for homogeneity which it imposes on the specimen. 

 Any slight flaw in the casting, resulting in cracks whose dimensions 

 would change with pressure, must change the direction of the lines of 

 flow with pressure. Further, any large scale crystal structure would 

 give different results for different castings. 



The second method was designed to obviate as much as possible 

 the effect of inhomogeneity in the sample. As already mentioned, 

 this is a longitudinal flow method, there being a source and a sink at 

 the ends of the specimen, and the difference of temperature at two 

 fixed points is measured. Since the heat input must cross each section 

 of the specimen, any distortion of the flow by local inhomogeneities 

 must have less effect than in the case of the cylinder, where as an 

 extreme case it is possible that a flaw properly situated might force 

 all the heat input to flow out of the cylinder at one side only. The 

 much smaller dimensions of the longitudinal flow specimens would 

 also seem to reduce the chance of errors from flaws. 



