OF ARTS AND SCIENCES. 263 



To procure a nickel bar as large as the iron ones gave no small 

 trouble. It was not to be found in the market, but some 23 lb. 

 of commercial nickel, in various shapes, put into a new melting- 

 pot by an obliging brass-founder, yielded, after one or two failures, 

 a casting from which a bar 91.5 cm. long and about 2.55 cm. in 

 width and thickness was worked out. The high melting point of 

 nickel makes it a difficult metal to cast in the form of a long bar. 

 A baked sand mould was finally used, and one end of it was ele- 

 vated considerably above the other, in order that the whole might 

 become filled before the metal cooled. Even with this device the 

 success attained was not perfect, and the finished nickel bar was 

 marred by many flaws. A little experience, however, shows that 

 a blow-hole may be of considerable size without deserving to be 

 counted among the more serious causes of error in the determina- 

 tion of thermal conductivity. Borings or cuttings from this nickel 

 bar were analyzed by a chemist, and were found to contain several 

 per cent of iron, and small quantities of other impurities, of which 

 silicon was one. All of the nickel put into the melting-pot was 

 represented when purchased to be fairly pure, but it was discovered 

 too late that two pounds of it was of very poor quality, that is, 

 contained much iron. 



In each of the three bars holes about 0.55 cm. wide and about 

 2.3 cm. deep were bored to receive the thermometer bulbs. The 

 length of each bar was divided into nine nearly equal parts, and 

 one hole was bored at the middle of each part, so that the holes 

 were a little more than 10 cm. distant from one another. Each 

 hole was partly filled with mercury. 



The thermometers used with the bars had bulbs about 0.45 cm. 

 wide and about 2.0 cm. long. They were graduated- for the pur- 

 poses of this investigation on the basis of bulb-immersion only. 

 They were not high grade instruments, and some of them were 

 probably in error nearly 0°.5 at various points between 0° and 100°. 

 At 100° they read from 0°.l to 0.°3 or 0.°4 low. Above 100° the 

 errors were more serious, — thermometer no. 8, which was some- 

 times used with a reading of 150° or higher, having an error of 

 about 2° at that temperature, according to the comparison afterward 

 made with a good thermometer lent me by Professor Holman of 

 the Massachusetts Institute of Technology. 



The conduction experiments were made in May and June, 1890, 

 in the north side of the middle basement of the Jefferson Physical 

 Laboratory. The bars were supported near their ends on thin 



