1887.] Mr A. Campbell on Measurement of Peltier Effect. 399 
Two forms of this arrangement were used. In the first (a) the 
thermopiles were two broad strips (formed of alternate squares of 
the metals), bent into zigzags which fitted one within the other. 
The one zigzag was of iron and zinc, and the other of iron and 
German silver. The iron used here was thin tinplate, and none of 
the metals were annealed. This arrangement was not perfectly sym- 
metrical j the second modification (6) was , however. It consisted of 
four zigzag thermopiles of 18 junctions each, almost identical in 
size and form, made of strips of thin sheet metal, 3 cms. long and 
about 5 mms. broad. Two of them were of FeArg and the other two 
of FeZn. These were arranged in the form of a square, with their 
junctions interlaced, the similar piles being at opposite sides of the 
square. Insulation was ensured by strips of asbestos paper separat- 
ing the junctions from one another. (The insulation was tested both 
at high and low temperatures, and was found to be practically perfect.) 
Clearly this arrangement was perfectly symmetrical. In this case 
the iron used was thin tinplate which had had the tin almost com- 
pletely burnt off it at a red heat ; the zinc and German silver were 
also well annealed. In both cases, the copper boxes or tubes in 
which the piles were heated up were surrounded by asbestos wool 
and fireclay bricks, and the heating was done by a spirit-lamp care- 
fully shut in from air currents. Nearly all the measurements were 
made when the temperatures had become almost 'perfectly steady. 
The battery current, which was from two Tray-Danaells, was kept 
as constant as possible by gradually diminishing the resistance of 
the circuit as the current showed signs of falling. It was measured 
by a mirror galvanometer doubly shunted by copper shunts. In 
using both (a) and (&), the intervals for current, no-current, &c., were 
chosen of such a length that a permanent state of temperature dis- 
tribution had been almost reached before the end of each interval. 
For ( a ) the periods were 60 secs, each, and for ( b ) 90 secs. each. 
The thermopile measurements were made in much the same manner 
as in the case of nickel German silver described above p.e., Peltier 
effect in pile (1) measured by pile (2), and then Peltier effect in pile 
(2) measured by pile (1)]. 
In arrangement (b) the resistance of the galvanometer + FeArg 
thermopiles was 2 - 053 ohms, while that of galvanometer + FeZn 
piles was P963 ohms. This difference of resistance also diminished 
VOL. XIV. 27/1/88 2 o 
