64 PROCEEDINGS OF THE AMERICAN ACADEMY. 



absent, B the induction flux through the central cross section of the 

 rod under investigation, and m denotes the ratio of the length of the 

 rod to its diameter. Figure 4 gives two B-H curves for this material. 



In Experiment D, we studied the behavior of a rod of fine annealed 

 tool steel 0.794 cms. in diameter for which the final value of the mag- 

 netization vector, /, is about 1540. This was furnished with a test 

 coil of 21 turns wound about its centre and was tested first when it 

 had a length of 91.3 cms. and again when it had been shortened to 

 21.8 cms. At excitations of 1300, 1800, and 2650 gausses the fluxes 

 did not differ from each other by more than one fifth of one per cent 

 of either. At 900 gausses, however, the shorter rod had a flux quite 

 one per cent below that of the longer one. At 2700 gausses, the 

 value of B was about 21960. 



A rod of Special Magnet Steel, 0.805 cms. in diameter and origi- 

 nally 124 diameters long, gave upon being shortened to about 30 

 diameters, one half of one per cent less flux in a field of 1260 gausses, 

 but in fields of about 1750 and 2600 gausses the fluxes in the case of 

 the short piece were indistinguishable from the corresponding fluxes 

 for the longer one. This was Experiment E. 



It seems to be true, therefore, that in properly constructed solenoids 

 with uniform fields above 2500 gausses, we may safely use rods only 

 25 or 30 diameters long with the expectation of finding that the fluxes 

 through their meridian sections are the same, within the limits of 

 ordinary careful laboratory practice, as if the pieces were infinitely 

 long. At low excitations, however, the corrections for the ends of 

 pieces as short as this would, of course, be very large. The magnetiz- 

 ing solenoid should be about 25 diameters of its own coil longer than 

 the test piece. For ordinary work, where an accuracy of one half of 

 one per cent will suffice, test pieces only 15 diameters long will often 

 serve at excitations above 2500 gausses. 



I wish to express my great obligation to the Trustees of the Bache 

 Fund of the National Academy of Sciences for the loan of some of the 

 apparatus used in making the observations mentioned in this paper. 



The Jefferson Physical Laboratory, 

 Cambridge, Mass. 



