by actually determining the amount of nitrogen remaining in the gas which was weighed.

This might have been done by taking hydrogen from the globe in which it had been weighed, and analyzing in some suitable apparatus. Circumstances made it convenient to adopt a slightly different method. Instead of filling one globe with hydrogen which was to be first weighed and then analyzed, two globes were filled at the same time. They had been exhausted to the same degree. The current of gas divided and went to the two globes in quantities proportional to their capacities; so that the two contained gas of identical quality. One globe was afterwards detached and weighed. The other globe remained permanently connected with the apparatus for determining the impurities in the gas, and was somewhat elaborately guarded against possibility of leakage.*

With this apparatus, the method of determining the amount of nitrogen in the hydrogen was as follows: A vacuous tube containing copper oxide was heated till no more gas was given off. This tube was connected to the globe containing the hydrogen reserved for analysis, and the copper oxide was heated. By noting the change of pressure in the globe the amount of hydrogen withdrawn from it was determined. Then the connection with the globe was closed, the copper oxide was cooled, and the gas remaining was extracted with a Sprengel pump, and transferred to the eudiometer. The structure of the pump made impossible any admixture of air with the gas transferred with it, for the mercury which actuated it all passed through a vacuum trap which was kept exhausted by an auxiliary pump.

When in this way, let us say, one litre of hydrogen had been withdrawn from the part reserved for analysis, and had been reduced to, say, ten cubic centimetres, it was measured in the eudiometer, mixed with oxygen and exploded, and so the amount of hydrogen in the residue was determined. Since cupric oxide is at best but a treacherous material, examination was also made for gas absorbable by alkali, and when absorption occurred, the experiment had to be resumed with a better sample of copper oxide. If the experiment showed no absorption, the gas not hydrogen was considered as nitrogen. Since qualitative examination showed that sulphur and carbon were not present, this supposition was probably justified. It will be noted that it is a matter of indifference whether the oxygen used in the explosion contained nitrogen, but it did not.

In several experiments of this series, nitrogen was found in the hydrogen. But after a considerable volume of hydrogen had been produced, the apparatus was practically free from nitrogen except after breakage. Soon after the use of

*American Journal of Science, 41, 220.

palladium for purifying the hydrogen in the way described, nitrogen could no longer be detected, though sought for repeatedly in a volume as large as two litres. The search for it was then discontinued.

AMOUNTS OF NITROGEN FOUND IN HYDROGEN WHICH WAS WEIGHED:

828 cc.

May 6,

66

1734

.035

.000025

May 8,

640

May 28,

66

798

June 9,
June 20,

66

838

66

1083

.005 .005

.000008

The weights of hydrogen observed in this series of experiments are given in the following table, which shows also the number of different observations of the equilibrium of the balance, counting all the different releases and arrests made at one time as but one observation.

Combining the observations in the way which was thought proper at the time of the experiments, we get the following determinations of the weight of the hydrogen in the globes. The table gives the pressures reduced to 0°, the readings of the thermometers reduced to the scale of 2053, the computed reading of the hydrogen air thermometer, the gross weight and the tare of the globes, their capacities at the temperature of the experiment, and the density computed; the latter includes the correction for nitrogen given on page 59, and is computed for the sea level in

If we increase the mean by one thirty-thousandth,* we get

D = 0.089938 gr.± 0.000007.

6.--HYDROGEN, SECOND METHOD.

In a second series of experiments, the manipulation was precisely like that in the third series of experiments on the density of oxygen, except in the preparation of the hydrogen. No examination of the hydrogen for impurities was made, since

*See note, page 28.

the result of many previous examinations had shown how to obtain the gas with no measurable amount of impurity.

It may be not without interest to consider how small an amount of nitrogen. might be detected when mixed with hydrogen. If the hydrogen should be removed from the mixture by absorption, as carbon dioxide may be removed, it would be possible to detect a very small quantity. My measuring apparatus, for instance, would measure a considerably smaller quantity than the thousandth of a cubic centimetre. If, then, two litres of hydrogen were absorbed by some suppos able reagent for the purpose, a two-millionth of nitrogen would be detected. But the case is different when the amount of hydrogen in a mixture must be determined by explosion with oxygen. In my experiments, two litres of hydrogen were reduced to ten cubic centimetres by absorption by means of copper oxide, but the remaining ten centimetres had to be measured, and the hydrogen determined by explosion, and the nitrogen determined by difference. With sufficient care, one ought to be able to answer for a hundredth of a cubic centimetre; for how much less, I have no means of knowing.

But to detect a hundredth of a cubic centimetre, in the case supposed (and realized), means to detect a two-hundred-thousandth; to this approximation, the examination was carried, in duplicate analyses, with the result that hydrogen containing less than this could be obtained in the usual working of the apparatus, except after fracture. The removal of nitrogen from the apparatus was then tedious, but the result was certain, and hydrogen of such purity is thought to have been used in all the experiments of this series.

Hydrogen was prepared in this series of experiments in very nearly the same way as in the preceding; but some details were different. The hydrogen from the voltameter passed through an almost saturated solution of potassium hydroxide, over incandescent copper, through a tube one metre long and two and a half centimetres in diameter, which was filled with glass beads moistened with a strong solution of lead oxide in potassium hydroxide, then through three tubes of the same dimensions, filled with calcium chloride, with powdered potassium hydroxide, and with phosphorus pentoxide. Here was placed a manometer and a regulating stopcock; all parts were connected by fusion, except where infusible glass had to be connected to the soft glass of the rest of the apparatus.

When the tube containing palladium was to be filled, it was connected by fusion to the tube delivering pure hydrogen, and was heated. When it was so hot

that no absorption would take place, a current was passed through it, and this was continued for half an hour; care was taken to make this current pass through every branch of the connecting tubes so as to expel all air. When the exit of the hydrogen had been closed by fusion, the palladium was cooled, and hydrogen was admitted till its tension was equal to that of the atmosphere. Hydrogen was then again passed through the tube of palladium, escaping into the air for three hours, so as to remove any nitrogen which should have accumulated. When this had been accomplished the exit of the gas and the connection with the voltameter were then closed by fusion. The tube containing palladium had been from the first connected with the globe, but the passage had been stopped by a plug of fusible metal.

8. SECOND METHOD. FILLING GLOBES WITH HYDROGEN.

When the globe and its connections had been exhausted to the desired degree, one of the fusible metal plugs which obstructed the connection with the supply of hydrogen was fused, and hydrogen admitted. When the pressure had become some few centimetres, the connection with the palladium was closed by fusion, and the globe was again exhausted. When the exhaustion was sufficient, a second fusible metal plug was removed, and hydrogen admitted till the atmospheric pres sure was reached. The preliminary exhaustion of the globes was always such that less than a hundred-thousandth of air remained. For the second exhaustion, a vacuum of a thousandth was thought sufficient.

When the globe had been filled with hydrogen, the ice in which it stood was well heaped around it, and the cover replaced over the tank. Readings of the syphon barometer were made as in the case of oxygen. In all cases, the two levels of the mercury to be read were in the same perpendicular, and the scale used stood in contact with the glass tubes of the barometer, so as to minimize the effect of any want of parallelism in the directions of the motions of the reading microscopes.

It will be seen that in all my experiments on the density of gases, their pressure has been determined without any communication between the gas and the atmosphere; contamination was therefore impossible. It was necessary that my work should be so arranged that it could be left at almost any point with little danger of the loss of an experiment in consequence of the interruption.

In the first series of experiments with oxygen and hydrogen, a mano-barometer was used, owing to the conditions of the experiment; but in all other experiments the two central tubes of the manometer, which together constituted a syphon