NORMAL SULPHATES OF POTASSIUM, RUBIDIUM, AND C^PISIUM. 
469 
of aluminium is about 13’77 millims. which leaves room for a crystal 8‘37 millims. 
thick and an air film of OH5 millim. As the crystals only usually varied 1 or 2 
millims. each side of the thickness mentioned, the amount of under or over compen¬ 
sation was never very large. 
The same thermometers have been employed as were fully described in the previous 
memoir. Their fixed points were carefully redetermined after the completion of the 
determinations. The inner bent thermometer whose bulb was in contact with the 
tripod and whose indications were those accepted, was found to have altered only 
to the extent of 0°‘l, the indications at 0° and 100° in ice and steam, after applying 
the pressure correction for the latter, being 0°'l and 100°'l respectively. Hence the 
interval had remained unchanged, and as only differences of the temperatures are 
employed in calculating the coefficients of expansion, no correction of these latter 
is required for change of interval. 
The usual modus operandi was to expend the greater part of three days in carrying 
out a duplicate pair of determinations, of the linear thermal deformation of any one 
crystal along the direction perpendicular to the two prepared parallel surfaces. The 
afternoon of the first day was employed in adjusting the crystal and the whole 
apparatus so as to afford a suitable field of interference bands. Each of the two 
succeeding days was utilised for the carrying out of a complete series of observations 
of the position and transit of bands for two intervals of temperature, the operations on 
each day occupying 5 to 7 hours, during the whole of which time the author followed 
the bands without intermission. Naturally, the carrying out of sixty-four such 
observations has proved very trying and fatiguing, the observer being continually 
afraid of such highly delicate measurements being vitiated by earth tremors due to 
street traffic or other disturbance, in spite of the rigid mounting of the apj^aratus 
on a slate table. Fortunately, this fear has not proved to have had much foundation, 
as the author’s laboratory is happily situated in an exceptionally quiet part of Oxford 
well removed from the city and the railway. But the experience has shown that the 
observations would have been far more difficult, if not impossible, in a large city with 
a network of underground railways such as London. Although this source of dis¬ 
turbance has been minimised, several observations have been lost, generally after 
spending hours upon them, by the cracking of the crystal under the influence of the 
rise of temperature, slow as it always was in order to avoid this catastrophe. 
The further experience gained during this work indicates that in the case of 
crystals the Abbe method, of calculating the number of bands which pass the point 
of reference between two temperatures from initial and final observations of the 
positions of the bands nearest the reference point, for two wave-lengths, is generally 
inapplicable. The only guarantee that the observation has been a trustworthy one, 
that no disturbance due to any of the causes already referred to has occurred, is 
obtained by carefully following the bands for the whole of the temperature-interval, 
and observing that they maintain their regular distances and exhibit no appreciable 
