The above examples will show that it is possible to determine the resistance at 100° C. to better than 1 part in 250,000, and nearly the same order of accuracy is obtainable, with proper precautions, in the vapour of sulphur.

I believe the most difficult point to accurately determine is freezing-point, and for really satisfactory determinations it is advisable to use distilled-water ice; but if this is manufactured in the usual manner there is great danger that some trace of salt may be carried from the freezing-mixture to the ice. One precaution I have found advantageous when using ordinary ice, viz., on adding water to the powdered ice it is advisable to use water resulting from the melting of that ice itself rather than either distilled or tap water.

I will not burden this paper with any further detailed account of the standardization of these thermometers. Between March 4th and July 7th the fixed points were determined on over 30 occasions, in ice, steam, or sulphur vapour, and no variation which would affect temperature measurements between 0° and 100° C. by as much as 2000 C. presented itself during the latter half of these observations, the record and reductions of which alone make a large pile of manuscript.

The two thermometers, with their leads connected as described, were placed at opposite ends of a bridge wire of platinum-silver. During the spring of this year this wire was subjected to a most careful calibration by what was practically Carey Foster's method, and it proved to be more unequal than I had expected. It was therefore re-calibrated by a different method in which a resistance-box was used as a shunt, and the agreement between the results was satisfactory. The whole wire was 80 centim. long and had a total resistance of about 4 ohm. For convenience, and to avoid thermal effects, a similar wire connected with the galvanometer was laid alongside it, and the sliding-piece was fitted with a screw so arranged that a small turn of the screw-head made contact with both wires*.

The wire and contact-maker were covered by a thick copper shield (the screw-head projecting through a narrow slit) passing from end to end of the bridge: thus the temperature of the wire was kept uniform. By means of a vernier the divisions on the scale could be read to millim., which with this wire and thermometers AB and CD indicated at 50° C. a temperature difference of 000915° C.† The temperature

This method of making the connexion is due to Professor Callendar, and is exceedingly effective and convenient.

A thicker wire giving a more open scale was tried at first, but found to be less convenient than the one finally adopted and calibrated.

The re

coefficient of this wire was found to be ⚫00029. sistances of the different parts of the wire were, after correction for the errors of individual coils, &c., merely expressed in terms of the mean box ohm, the absolute value being of no consequence so long as the fixed points were determined in terms of the same standard. The remaining two arms of the bridge were constructed of german-silver. They were wound together, boiled in paraffin, placed in a bottle, and I expended much care in finally adjusting them until equal. Their resistance was about 5 ohms and the galvanometer about 8 ohms, which, assuming the resistance of the thermometers as about 20 ohms each, would give nearly the maximum of sensitiveness. A single storage-cell was always used, and a resistance of 40 ohms was placed in the battery circuit when the thermometers were in ice. A table was then calculated which gave the resistance necessary in the battery circuit when the thermometers were at any temperature in order that C2R (where R is the thermometer resistance) should be constant. Thus the rise in the temperature of the thermometer coils due to current-heating was always the same, and consequent errors were eliminated, a point to which I attach considerable importance.

The value of R-R in thermometer AB was 6.88815; therefore a difference of 1 ohm at 50° C. indicated a difference of 14° 5177 C., and as

the degree value of any bridge-reading at other temperatures can be deduced.

There was no difficulty, in the arrangement above described, in reading with certainty a difference of 1000 C., and, as an illustration, I may mention the fact that if the thermometers were placed in separate hypsometers side by side on the bench and one of the hypsometers was then removed to the ground (about 3 feet below), the difference in the bridge-wire reading thus caused slightly exceeded 4 millim.

Before leaving this portion of the subject I may mention that I carefully compared the thermometer AB with two standards from the International Bureau of Weights and Measures, which passed through my hands this summer on their way to Sydney University, and further with a separate standard sent me by the Bureau in February 1893. As the particulars of this comparison are now in the press †, I • Trans. 1891, A. p. 142.

need not refer to them further than to say that the results were most satisfactory.

When the experiments were in progress the thermometer CD was placed in a hole drilled in the wall of the steel chamber parallel to the side and separated from the inner surface by only in.: thus it assumed the temperature of the walls surrounding the calorimeter. Thermometer AB passed within the calorimeter and was immersed to an inch above its bulb in the contained liquid. The reading on the bridgewire thus gave the difference between the temperature of the calorimeter and the walls of the surrounding chamber.

To ascertain the actual temperature and the constancy of the temperature of the surrounding walls, a second hole, like that already described, was drilled in the steel, in which a mercury thermometer of open scale was placed. The stem, where it protruded above the lid of the tank, was surrounded by a glass tube up which the motor pumped the tank-water, which, passing out at the top, returned to the tank. The stem, therefore, was maintained at a steady temperature, and the fluctuations observed might be considered as solely due to changes in the walls of the steel chamber. The stem of thermometer A (used) at the lower temperatures) was divided into millimetres, about 27 to a degree, and No. II. (used at the higher) about 20 millim. per degree. They were observed through a telescope containing a micrometer-scale giving direct readings to of a millimetre. The tenth of one of these divisions could be estimated, and thus changes of 1000 C. (i. e. about 025 millim.) were easily observable. This was important, as the bridge-wire readings had to be corrected for any movement in the temperature of thermometer CD-that is, the temperature of the walls of the chamber.

о

The stirrer (which consisted of two nearly vertical narrow paddles, reaching from top to bottom of the calorimeter) was so placed as to throw the liquid across the silver flask and coil against the thermometer. This form of stirrer was not the one I should have adopted had the apparatus been designed solely for the experiments which I am now describing. In that case I should have preferred the form described in paper J, which threw the liquid from the bottom to the top of the calorimeter; and I believe that such irregularities as have presented themselves in these experiments, especially with the smallest mass, are due to insufficient stirring. When the calorimeter is full of liquid the nature of the stirring is of less consequence; and as I proposed in the experiments for which the apparatus was designed to completely fill the calorimeter, I adopted the form which it appeared to me would

generate the least heat compatible with efficient mixing. The stirrer was supported in the same manner as that described in a former paper; that is, its bearings were entirely external to the calorimeter except at the base, where, to check vibration, an agate cylinder hung within a ring *.

In place of the 2000 revolutions per minute used in my former work I confined myself, for the reasons above stated, to a comparatively slow rotation of from 500 to 300 revolutions per minute. The revolving shaft was electrically connected with the chronograph in such a manner that the time of every 1000 revolutions was automatically recorded on the tape.

Before commencing the observations on aniline I introduced 299-35 grms. (in vacuo) of water into the calorimeter, and, with a view to ascertaining the water-equivalent, performed a few experiments to determine the rate of rise when a known current was passed through the coil. The point was not of great importance, as I proposed to determine the specific heat of aniline by a method requiring no knowledge of the waterequivalent, which, however, when the specific heat of aniline was known, could be deduced from the experiments, and thus a previous determination by means of water would serve as an independent test of the accuracy of the results. I had not covered the german-silver coil with any insulator, since, aniline being itself an insulator, the precaution was unnecessary; and this freedom to use the naked wire was one of the chief reasons which led to the adoption of aniline as a suitable liquid for my subsequent work. I was therefore aware that I could not expect very satisfactory results from the use of water. The polarization was considerable, as shown by the back E.M.F. at the termination of an experiment, and also by the difficulty experienced in making any accurate determination of the coil-resistance. However, it was subsequently found that the experiments gave very fair results. The temperature was raised from about 5° C. below to 5° C. above the surrounding temperature (i. e., through about 12 centim. of the bridge-wire), and the times of passing the several divisions of the bridge-wire were recorded on the chronograph. Five experiments were performed, and the results of the first three gave a water-equivalent of 80.1, and Nos. 4 and 5 of 79.8. They were thus divided into groups before the results were calculated, as the coil showed decided signs of change in resistance between the two batches of experiments. I had proposed to continue these experiments with a different mass of water, but became alarmed as to the effect on the coil of

A portion of the shaft of the stirrer was constructed of ivory, to diminish its thermal conductivity.

the change that was going on, and decided to withdraw the water, for had the coil become useless from any cause, the whole apparatus would have had to be taken to pieces and the base of the calorimeter unsoldered, &c., involving a delay of at least a week or two.

During the time that the water was in the calorimeter I performed a considerable number of experiments to determine the rise in temperature due to the work done by the stirrer. With this form of stirrer and the low rates, tr2 was practically a constant, where t was the time of rising 1 centim. of the bridge-wire and the rate of revolution per second. The resulting correction was comparatively small, amounting to about of the heat generated by the current during the water-equivalent experiments; thus any small error in the correction introduced when eliminating the heat supplied by the stirrer became unimportant.

After the water was withdrawn, the whole apparatus was kept for three days at a temperature between 40° and 50° C., while dried air was forced continuously through every part of it. At the end of this time a sulphuric-acid bulb (through which the air passed on its exit from the apparatus) showed no increase in weight. I then introduced 294.99 grms. (in vacuo) of aniline into the calorimeter

I will at once state that no great pains were taken to secure an absolutely pure specimen of aniline. For my subsequent experiments the purity of the aniline was a matter of no consequence, as I only required to know the actual capacity of the calorimeter and contents at different temperatures.

Again, as regards the purposes of this investigation I think that it will be of greater service if I am able to supply experimenters with the specific heat of a quality of aniline at all times procurable, rather than if I gave the constants of a quality only to be obtained with difficulty. The sample I used was supplied by Messrs. Harrington Bros., as "pure colourless," and had but a light brown tinge. It was redistilled by me once before use, and I also determined its boiling-point, which was in agreement with that of a specially pure sample which I had examined on a previous occasion f. Further (and this I have always found to be a fair test of the purity), the temperature of the boiling-point did not change during distillation.

From the time that aniline was first introduced into the

*The precautions observed in order that the mass introduced might be accurately determined are fully described in paper J, p. 418. + Trans. 1891 A, p. 64.