barometer, were used. In this case, the two readings of level were made on the same scale, because the two levels of the mercury were in the same perpendicular.

The time during which occasional readings of pressure were made varied from an hour to three or four hours.

As in the case of the third series of determinations of the density of oxygen, there was no room for the exercise of judgment in combining the observations. The experiments are also affected with a source of constant error, as will be mentioned in more detail; so that there is the less reason for giving more than the pressures and weights observed, together with the density reduced to the sea level at latitude 45°, by the formula

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

D= 0.089970 gr, ± 0.000011.

11. REMARK ON THE RESULTS OF THE FIRST AND SECOND SERIES.

The degree of precision attained in weighing so light a gas as hydrogen in either series of experiments was of course not very great, and the two mean values obtained agree more closely than would be expected. Perhaps they are sufficient to show that the method of weighing hydrogen in globes exhausted with

* See note, page 28.

mercurial pumps will give results not far from the value obtained. But they by no means show that this value is the truth. If we can trust to a method which seems free from objection, and which has been carried out with as much care as can well be used, the value here obtained is not far from one thousandth part too large.

The accidental errors of the measurement of the density of hydrogen by weighing a given volume are considerable. A globe containing 1.8 grammes of hydrogen under standard conditions, and itself weighing not more than twelve hundred grammes, can be obtained, though with difficulty. The gross weight of the globe is then six hundred times the weight of the hydrogen to be determined by its use. It is true that the weight of the globe at any given stage of the experiment can be determined with an error of not much more than one tenth of a milligramme, and that in no long time, provided a pretty elaborate plant be employed. But it has not yet been found easy, I believe, by any one, to secure constancy of the weight of the globe and of the lubrication of its stopcock during the exposure to contact with water, which is necessary if a constant temperature is maintained in the usual way.

It is possible to avoid the use of a stopcock on the globe in which a gas is weighed, and the manipulation for this purpose is not troublesome. I had expected to make use of this method; but a severe explosion broke nearly all my calibrated globes. It is probable that some increase in precision could be obtained by this method. For the globe, after filling with the gas, and after thorough cleansing, could be put in a desiccator, and kept on the balance till it was certain that all effects of contact with water had disappeared. It could then be exhausted without removal from the desiccator, closed by fusion, and again weighed. During both weighings, the globe would have a surface of continuous glass. Still, it is doubtful whether the labor of making a sufficient number of determinations would be well bestowed.

For, unless elaborate precautions are taken, the efficiency of which remains to be proved, mercurial vapor may diffuse into the globe in which the gas is to be weighed. Until, therefore, we can dispense with the use of mercurial pumps and yet produce a vacuum of a few millionths, it seems necessary to devise some method of measuring the density of hydrogen in which the contamination of the gas with the vapor of mercury shall occasion no error.

The weighing of the hydrogen while it is absorbed in palladium, and measuring its volume and pressure in another vessel after expelling it from

the palladium seemed likely to give a more accurate value for the density than could be obtained in any other way now possible. My palladium would absorb about 3.8 grammes, and would give off as much as 3.7 grammes at atmospheric pressure, this quantity measuring over 40 litres. From the increase in net weight alone could be hoped some increase in precision; some from the decrease in gross weight, for the palladium tube with its contents weighed eight hundred grammes, while the larger globes weighed half as much more; some from the small surface of glass exposed; some from the disuse of stopcocks and their lubrication, exposed as it is to so many accidents. But the principal gain hoped for was the fact that mercurial vapor would have no effect on the weight of the hydrogen used in the determination; and its effect on the pressure and volume of the gas is negligible.

Whether the improvement hoped for has actually been secured is now to be submitted to the judgment of those interested.

The use of palladium for the purpose of obtaining pure hydrogen was, so far as I know, first suggested by Chirikoff. Its use for the purpose of accurately weighing hydrogen first suggested itself to me in 1882, so that I discussed the method with other chemists in 1883, and began preparations for the present work. The same method of weighing hydrogen suggested itself independently to Keiser, to whom belongs the credit of first publishing results obtained by it; questions of priority are of slight consequence, but it seems that to me belongs the credit of first inventing the method and beginning to work with it.

13.-HYDROGEN BY NEW METHOD.

APPARATUS FOR MEASUREMENT OF VOLUME.

My stock of palladium foil would absorb about 3.8 grammes of hydrogen, and would give off about 3.7 grammes at atmospheric pressure. A volume sufficient to contain this amount was made up of three globes. They were placed so as to be used like one, while their individual capacities fell within the range of my plant for determination. These three globes were the globes numbered 1, 6, and 7, in the list of globes prepared for weighing gases by Regnault's method mentioned on page 12. Of these, globe No. 6 had not been used before, for the reason mentioned on page 54.

These three globes were placed in three cylinders, a, b, c, Fig. 23, surrounded with finely crushed ice, and these again placed in a larger cylinder, r, 1.2 metres in diameter, and .9 metre high. This tank was surrounded with a non-conducting layer. The globes were now connected to a common inlet tube whose branches led, one to the syphon barometer, one to the self-acting Toepler pump, and one to the place for admitting hydrogen. The latter tube was closed by a fusible metal plug till the proper time.

The tube leading to the pump had to be closed during the introduction of the hydrogen and the measurement of its pressure in such a way that no leakage could occur. A glass stopcock can often be lubricated so that when closed it will not leak, but if opened and closed repeatedly it is liable to leak. In fact a stopcock, in all accurate work, ought to be considered as an instrument for reducing the flow of a current of

gas.

FIG. 23.-Apparatus for receiving hydrogen and measuring its volume and pressure at constant temperature.

The connection between the measuring globes and the air pump was therefore made proof against leakage by using two stopcocks, and filling the space between them with mercury at the atmospheric pressure whenever it was desired to close the passage. The two stopcocks are shown at e, f; a tube inserted at the lowest. point between them led to the stopcock k and the funnel 7, which was full of mercury. If the stopcocks e and f were open and the tube between them free from mercury, a clear passage existed between the air-pump and the globes. When a sufficient vacuum had been obtained by the action of the pump, the stopcocks e and f were closed and k was opened. The pressure of the atmosphere then forced mercury from the funnel to fill the space between e and f; when k was closed. It is

obvious that leakage from the globes towards the air-pump could take place only as mercury was forced out from this closed space through the keys of the stopcocks; which would easily be detected by inspection.

The apparatus met all reasonable expectations until a time when, as was afterwards found, the key of the stopcock e formed an incipient crack, producing capricious and intermittent leakage.

14. -HYDROGEN BY NEW METHOD. CAPACITY OF CONNECTING TUBES.

It is obvious that when a weighed quantity of hydrogen was admitted to these globes, the connecting tubes and one branch of the syphon barometer had to be filled; so that the capacity thus added to the globes had to be determined. Part of these tubes was always at the temperature of the globes themselves, being covered with melting ice, while part was at the temperature of the room, or at the temperature of the water with which the barometer was surrounded. These two parts had to be determined separately.

The tube leading to the barometer being closed by fusion, c, Fig. 24, a three-way stopcock, e, was fused to the tube for admitting hydrogen shown at d. One branch of this was connected to an air-pump, and the other to a gas burette standing in sulphuric acid. Water could not be used, for condensation of its vapor on the cold parts of the connecting tubes would cause error. First, the blank space between the key of the stopcock and the fusible metal at d was determined. This was done by exhausting this space by turning the stopcock into

d

Da

FIG. 24.-Apparatus for determining volume of
connecting tubes.

h

the proper position, during which time the level of the acid outside and inside the burette was made the same, and the level read off. Then the degree of exhaustion was noted, and the key of the stopcock was turned so that the vacuous space was filled with air from the burette, the volume and temperature of which were also noted, after making the adjustment of level again. The volume of the air thus drawn from the burette was obviously equal to the volume of the exhausted space, after applying a correction for the imperfection of the vacuum.

When a few concordant determinations of the blank space had been made, the fusible metal was removed by fusion, and the same determination made of the vol