mark and the globes were covered with ice during the measurement. led to the barometer, and d was for the admission of hydrogen.

The tube e

The valve between the globes and the pump consisted of a U-shaped tube, ƒ g, which could be filled with, or emptied of, mercury by means of the tubes and stopcock shown in connection with it. Suppose the globes to have been exhausted and the pump to be ready to shut off, the stopcock / was opened, when mercury in k was raised by atmospheric pressure, passing first into the vacuum in h, where any air accompanying the mercury would be stopped. From here, the mercury passed into the U tube, fg, and filled it to near the top of the enlargement f. If now the globes are filled with gas, its pressure will force the mercury up the tube g leading to the pump, but leaving the U tube filled above the bend. When pressure was to be measured, the stopcock 7 was opened, and pressure applied to the surface of the mercury in the flask k so as to force mercury up to the mark at ¿.

When it was desired to exhaust the globes again, the mercury was brought to stand in the lower part of the enlargement, and the point o was cut off and replaced with a tube prepared to be closed by fusion. A Bianchi pump was connected to this and the globes were exhausted to three or four millimetres, when the connection was closed by fusion, leaving the glass as shown in the figure. The air-pump was then applied to the flask and the stopcock 7 opened, when the mercury left the bend of the U-shaped tube free. The Toepler pump then continued the exhaustion of the globes.

29.-HYDROGEN BY NEW METHOD. SECOND APPARATUS. CAPACITY OF

CONNECTING TUBES.

Before the valve fg, Fig. 28, was fused to the tube ci, it was sealed off a little below the mark i, and filled with mercury to this mark. Then the capacity of the tubes between this mercury and the fusible metal plugs near the globes was determined just as in the previous case; the values obtained have already been given on page 68. After the determination was completed, the fusible valves were opened, and f g connected by fusion, when the apparatus was ready for use.

30.-HYDROGEN BY NEW METHOD.

SECOND APPARATUS.
PROCESS.

REMAINDER OF THE

The rest of the process was in every respect like that in the preceding series of experiments. The only differences have been noted; the pleasure of working with an apparatus which could leak only by fracture was great. Several verifications were made of the fact that there was no appreciable leakage; it will suffice to give one. The apparatus was exhausted, and the vacuum measured with the

McLeod gauge. Then it was left at rest for twenty-four hours, and the vacuum measured again, with the following result:

December 11.

Left at rest.

Exhausted apparatus; at 9 A.M., vacuum three ten-millionths.

December 12. Measured vacuum left undisturbed since yesterday. Vacuum at 10.30 A.M., three ten-millionths.

It is obvious, therefore, that leakage had been eliminated.

31.-HYDROGEN BY NEW METHOD. SECOND APPARATUS.

REMARKS.

With this apparatus eleven determinations were made, none being lost by reason of accident to the apparatus, in which the probable error of the determination of the density of a gas so light as hydrogen was only one part in eight thousand for a single experiment. The agreement of the results with those of series third and fourth, depending as they do on entirely different calibrations of different apparatus at an interval of months, leads me to hope that they are not seriously in error.

It is a misfortune that it was not possible to make one or two more series of determinations, with new apparatus, or with new determinations of the capacity already involved, so setting at rest any remaining doubt. But to the patience and courage and endurance and powers of recuperation of a single person laboring without assistance at a matter confessedly so difficult, there are limits already too nearly reached.

The nature of the reduction in this series is precisely the same as in the preceding, using only the altered values for capacity of the globes and of the tubes connected with them. The sources of error are the same, except that leakage could take place only by carelessness in constructing the fusible metal plug which closed the apparatus during the exhaustion, or of the connection by which hydrogen was admitted to the globe; which was easy to detect, and did not occur. The other sources of error, also, it is thought, were avoided throughout this series of determi nations.

32.-HYDROGEN BY NEW METHOD.

SECOND APPARATUS. OBSERVATIONS AND
RESULTS.

The observations and results of the fifth series of determinations are given in the following table, which is in all respects like that at page 74, which may be consulted for explanations. The values computed for the density of hydrogen are for the sea level in latitude 45°.

November 23

66

66

23456789 ON

231.9 22.9 59.6 55.0 36.5 43-2574 43.3489 725.40 3.7164 .089861 26 230.7 18.0 60.0 56.3 36.5 43.2574 43.3502 728 13 43.3502 728 13 3-7314 .089877 29 228.5 23.5 60.7 55.9 36.9 43.3498 43.3498 723.05 3.7048 .089870 December 2 240.7 19.0 57.0 53.3 36.9 43.2574 43.3472 726.42 3.7217 .089867 240.6 19.8 57.0 53.1 36.5 43.2574 43.3470 728.05 3.7289 .089839 231.3 20.7 59.8 53.6 36.5 43.2574 43-3495 726.36 3.7219 .089874 239.5 19.5 57.4 53.6 36.5 43.2574 43-3475 725.30 3.7158 .089864 237.3 22.7 58.0 53.5 36.5 43.2574 43-3474 727.71 3.7290 .089883 234.7 22.9 58.8 54.3 36.5 43.2574 43.3482 722.54 3.7004 .089830 261.7 22. I 50.7 46.9 36.5 43.2574 43.3408 722.93 3.7037 .089877 240.6 19.3 57.0 53.2 36.5 43.2574 43.3471 721.90 3.6979 .089851

66

7

9

46

I 2

66

14

18

66

IO

21

II

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

The results of five series of determinations of the density of hydrogen at normal temperature and pressure at the sea level in latitude 45° are:

It is believed that the results of series first and second are affected with some source of constant error. It is supposed that this is due to the entrance of mercurial vapor into the globes in which the hydrogen is weighed. The remaining series of determinations are free from at least this source of error; until further light on the matter is secured, they give the best value of the density of hydrogen which I can obtain. No ground for preference of one result over the others is known except what is to be inferred from the magnitude of the deviations from the mean value in each series, so that they may be given weights as indicated by the probable errors. We then have as a final value :

PART III-ON THE VOLUMETRIC COMPOSITION OF WATER.

1. INTRODUCTION.

A knowledge of the density of oxygen and hydrogen is important, even without reference to the further question of atomic weight. But my principal object was to use these determinations to determine the latter. For this purpose, it is necessary to know also the ratio of the volumes in which the two gases combine to form water, the measurement to be effected in conditions nearly the same as those in which the volume and pressure were measured in the determination of density. This ratio I have attempted to measure by eudiometric methods. My apparatus for the purpose was made about fourteen years ago, but it was never used with oxygen and hydrogen till the measurements which were published in 1891.* Before this, the manipulation of the apparatus had been mastered in a long series of analyses of air. The errors of measurement of a volume of gas with this apparatus depend on the errors of determining temperature, and on the error in reading the level of the mercury in the eudiometer; for the error at the top of the manometric column of mercury is, by the use of suitable means of exact setting, made evanescent in comparison with the other two errors. These two sources of error cannot be well separated. But a discussion of all the analyses made up to a certain date showed that if the probable error of volume in measuring a gas be attributed to either one of these sources, supposing for the moment that the other error did not exist, then the probable error of level was less than the hundredth of a millimetre, or the probable error of temperature was less than the hundredth of a degree. Why therefore any one should call it an objection to my measurements that I read the level to the two-hundredth of a millimetre is incomprehensible.

The reason why my measurements do not agree with those of Scott is not yet explained. He says that the reason is to be found in the presence of ethane in my hydrogen, due to the electrolysis of impure hydroxide; but I never used potassium or sodium hydroxides; as my paper states, I abandoned their use during my preliminary experiments. Carbon dioxide was always sought for, and all which I found * American Journal of Science, 41, 220.

is set down in the table of results in my paper. Nor were my measuring tubes too wide for accurate measurement. The fact that my results were consistent shows that the probable error of a measurement was small, as is computed in the paper.

At the time when the paper was published it was thought, with what still appears to be good reason, that the atomic weight of oxygen is very nearly 15.88. There was also fairly good reason for supposing that the ratio of the densities of the two gases was but very little greater than 15.88; Lord Rayleigh had obtained the value 15.882, and I had obtained the value, 15.879. My value for the volumetric ratio therefore seemed consistent with other data, except with what is known of the divergence of hydrogen and oxygen from Boyle's law.

But the case is now altered. Scott's excellent work* yields the value at 0°, 2.00285. Leduc by two experiments obtains the value, 2.0037. For the ratio of the densities, Lord Rayleigh's discussion of the published data give the densities of the two gases, at Paris, as 1.42961 gr. and .08991 gr., the ratio of which is 15.900. Further, the best result to which I could come from my own experiments, a year ago, is, that the densities at sea level in latitude 45° are 1.4289 and .08987 gr., gr. whose ratio is 15.900. Now if the atomic weight is 15.88 and the ratio of densities is 15.90, the ratio of the volumes is not far from 2.0025.

It is therefore obvious that whatever physical constant is involved in my former determination, the value obtained in the eudiometer by measurement of gases saturated with aqueous vapor cannot be applied to pure oxygen and hydrogen collected in globes whose dimensions are so different from those of the measuring tubes.

Now, it seems probable that, if sources of constant error can be eliminated, an accurate knowledge of the volumetric ratio, combined with the ratio of the densities, would give a value of the atomic ratio which might not be very far behind the methods hitherto employed, either in trustworthiness or even in precision. It was accordingly adjudged worth while to prepare apparatus for three new determinations by three different processes. The apparatus for two of these new processes has been constructed, and the other could be put together for use in a week.

In the first proposed method, the volumes of oxygen and hydrogen which are to be combined, in order to determine the excess of either, were to be measured in the same globes which were used for weighing the gases. Two globes were to be filled with hydrogen while they were surrounded with ice, and the pressure was to be measured. A globe was to be filled in the same way with oxygen. The three globes were to be of such capacities and the pressures were to be so adjusted, that very nearly thirty litres of hydrogen and fifteen litres of oxygen were to be

Philosophical Transactions, 184, A, 543 (1893)