562 Mr. A. P. Carman on tlie 



to be collapsed was closed by end-plugs, and these were 

 supported by an internal rod so as to eliminate end pressures 

 on the tube when the tube was placed under external water 

 pressure. Leakage to the inside of the tube was prevented 

 by wrapping the joints with " friction" insulation tape, and 

 covering the whole joint with asphaltum. End constraints 

 were thus avoided. The experiments described here were 

 on seamless-drawn steel tubes of 1, 2, and 3 inches diameter. 

 The tubes were carefully machined to different thicknesses, 

 so results have been obtained for several values of tjd for 

 each kind of tube. The nickel-steel receptacle was 40 inches 

 long, so that results well beyond the u critical length " could 

 be obtained in most cases. As a matter of interest and of 

 possible importance, experiments were made on the collapse 

 of a series of small tubes of brass, aluminium, hard-rubber, 

 and glass. These results are not given and discussed in 

 this paper, but it may be stated that the curves obtained 

 have similar forms to those for steeL The collapse of the 

 glass tubing was striking, as the tube was reduced to a 

 fairly fine powder upon collapse. The strain was apparently 

 uniform throughout, and all parts seemed to give way at the 

 same time. 



The results of the experiments with steel tubes are given 

 in Tables I., II., and III. These results are also shown by 

 the curves of figs. 1, 2, and 3 (PL XIV,). The number of 

 lobes of the collapsed tube is given in the fourth column of the 

 Tables, and is also indicated by a small numeral at the point on 

 the curve. This is made a matter of record on account of its 

 importance in Southwell's theory, but it is not discussed in 

 the present paper. The data given here are the results of the 

 collapse of over 125 tubes. There were in practically each 

 case two tubes of the given size, and thus an average could 

 be made for each point. This also gave an immediate check 

 on freak collapses. Very few freak collapses, however, occur- 

 red, and these were easily explained by irregularities that 

 appeared upon the inspection of the collapsed tube. While the 

 machine work on these tubes was done with great care by the 

 mechanician of the department, it was impossible always to 

 get the value of tjd with the desired exactness. It was thus 

 necessary to make a correction in the observed collapsing 

 pressure, so as to have sets of results for a curve between 

 pressures and lengths with tjd constant. These corrections 

 were made by interpolation, assuming that the collapsing 



pressure varies as l->) . Since the total variations of (-%) 



in these corrections were small, there was little assumption 



