164 Lord Kelvin and Mr. J. R. Erskine Murray. [May 30. 



and to be complete across the whole area of 12^ cm. by 5, across 

 which the heat was conducted in that part of the composite slab ; and 

 to give rise to palpably imperfect fitting together of the solid above 

 and below it. We therefore repeated the experiment with the com- 

 posite slab turned upside down, so as to bring the crack in one half 

 of it now to be above the middle, instead of below the middle, as at 

 first. We still found for the composite slab less conductivity in the 

 hot part below the middle than in the cool part above the middle. 

 We inferred that, in respect to thermal conduction through slate 

 across the natural cleavage planes, the thermal conductivity 

 diminishes with increase of tempei-ature. 



5. We next tried a composite square slab of sandstone of the 

 same dimensions as the slate, and we found for it also decisive proof 

 of diminution of thermal conductivity with increase of temperature. 

 We were not troubled by any cracking of the sandstone, with its 

 upper side kept cool by an ice-cold metal plate resting on it, and its 

 lower side heated to probably as much as 300 or 400 C. 



6. After that we made a composite piece, of two small slate 

 columns, each 3'5 cm. square and 6'2 cm. high, with natural cleavage 

 planes vertical, pressed together with thermoelectric junctions as 

 before; but with appliances ( 10 below) for preventing loss or gain 

 of heat across the vertical sides, which the smaller horizontal 

 dimensions (7 cm., 3'5 -cm.) might require, but which were manifestly 

 unnecessary with the larger horizontal dimensions ('25 cm., 25 cm.) 

 of the slabs of slate and sandstone used in our former experiments. 

 The thermal flux lines in the former experiments on slate were per- 

 pendicular to the natural cleavage plaues, but now, with the thermal 

 flux lines parallel to the cleavage planes, we still find the same result, 

 smaller thermal conductivity at the higher temperatures. Numerical 

 results will be stated in 12 below. 



7. Our last experiments were made on a composite piece of 

 Aberdeen granite, made up of two columns, each 6 cm. high and 

 7'6 cm. square, pressed together, with appliances similar to those 

 described in b' ; and, as in all our previous experiments on slate 

 and sandstone, we found less thermal conductivity at higher tem- 

 peratures. The numerical results will be given in 12 below. 



8. The accompanying diagram represents the thermal appliances 

 and thermoelectric arrangement of 6, 7. The columns of slate or 

 granite were placed on supports in a bath of melted tin with about 

 0'2 cm. of their lower ends immersed. The top of each column was 

 kept cool by mercury, and water changed once a minute, as described 

 in 3 above, contained in a tank having the top of the stone column 

 for its bottom and completed by four vertical metal walls fitted into 

 gi^ooves in the stone and made tight against wet mercury by marine 

 glue. 



