reversal, after which 6 was warmed and fitted to the stopcock of the globe. On e was now fitted the corresponding ground joint f, and this was fused to the air pump. When the exhaustion was completed, d was closed by fusion, the part fused off was removed from f, cleaned, and put on the balance with the globe. From the weight of the glass attached to the globe, together with the volume between the key of the stopcock and the point d, is computed the volume added to the globe. The weight of the glass added is known from the preliminary weighing; it of course simply replaces a certain amount of brass and platinum weights which otherwise would have been needed. The manner in which the addition of this volume was cared for is to be described in the next paragraph.

29.-OXYGEN BY SECOND METHOD.

CORRECTION FOR DIFFERENCE OF VOLUMES.

In all the experiments of this series, the volumes suspended from the opposite pans of the balance were made equal at every weighing. For this For this purpose, some sixty minute flasks were made, of sizes ranging from one cubic centimetre to ten. They were all made nearly equal in weight by loading with mercury, closed by fusion, and provided with platinum hooks. The appearance of one is shown in Fig. 21. Their volumes were then determined by hydrostatic weighings and marked on them by etching, after which they were made rigorously equal in weight when weighed in a vacuum. This was accomplished on an assay balance whose needle moved fifty divisions for a difference of one milligramme, taking account of the density of the air and of the errors of the weights employed.

4.00

FIG. 21. Small flask for equalizing volumes.

When the difference of volume of two masses to be counterpoised against each other had been determined, two of these minute flasks were selected, such that the difference of their volumes was equal to the difference of volumes of the masses to be weighed; adding the small flask to the large volume and conversely, the volumes to be weighed became equal. These flasks had been so accurately adjusted to equality of weight in vacuo by repeated determinations that it was not necessary to take account of their weight or of the minute differences of their weights.

As an example of the simplest application of the equating flasks, a joint, be Fig. 20, weighed approximately 7.77 grammes, and had therefore a volume of 3.11 cubic centimetres. The weights to balance it had a volume of 0.85 cubic centime tres. Adding to these two volumes two flasks having volumes of 1.41 and 3.65 cubic centimetres, we have the two volumes on the opposite pans of the balance 3.11 cc + 1.41 ce = 4.52 cc, and 0.85 cc + 3.65 cc =

4.50 cc.

In the case of an exhausted globe, the number of volumes to be added to

gether is greater, but the principle is the same.

For instance, in weighing globe No. 5 when exhausted and closed as described on page 47, we have the following volumes on the two sides of the balance:

Since the volume of globe 5 is equal to that of its counterpoise when the globe is exhausted, G, = C,, hence the volumes on the opposite sides of the balance are equal, so that the immediate result of the weighings, with no observation of temperature and pressure, and with no computation of correction, gives the true difference of weight between the globe and its counterpoise.

The following table gives the weights obtained of globes filled with oxygen at the same temperature and pressure as that of the hydrogen which was used as a standard of comparison.

The weights of the oxygen found to have the same pressure as the standard volume of hydrogen when at the same temperature need a correction for the difference between the coefficients of expansion of hydrogen and of oxygen. If these coefficients were the same, the equality of pressure observed at ordinary temperatures would also be preserved at the temperature of melting ice. Since the expansion of oxygen is greater than that of hydrogen, the mass of the oxygen required at 0° will be greater than that at ordinary temperatures. If we assume for the coefficients of expansion at constant volume the values .003661 and .003674, we may apply the correction for the mean temperature of the whole series of experiments, which was 13.5° C. The formula for reduction therefore becomes:

by which are obtained the values given in the following table.

32. OXYGEN BY SECOND METHOD. RESULTS.

The following table gives the weights of oxygen contained in the globes in the given conditions, with the weights deduced for one litre of oxygen at normal temperature and pressure at the sea level in latitude 45°.

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

D = 1.42887 gr. ± 0.000048.

33.-OXYGEN BY THIRD METHOD.

In a third series of determinations of the density of oxygen, the globes were surrounded with melting ice while filling with the gas, and the pressure was measured by connecting the globe to the syphon barometer. The surface of the globe was therefore exposed to contact with water; as it was convenient to put the globe in position in the ice before the operation of exhausting the connecting tubes and filling it with oxygen, the contact was somewhat prolonged. Change of weight of the globes during the exposure was therefore inevitable; it was accordingly thought proper to determine the weight of the globe filled with oxygen, and then to exhaust it and determine its weight when empty, and to consider the difference of the two weighings as expressing the weight of the gas removed by the exhaustion. An attempt was made to protect the stopcocks and their lubrication from the action of water by surrounding them with rubber capsules while they lay

in the ice.

The cistern of the barometer was at the level of the centre of the globes during this series of determinations. The reduction of the observations took account of the force of gravity at my laboratory, and of the correction to the length of the scale of my barometer.

34.- -OXYGEN BY THIRD METHOD. FILLING THE GLOBES WITH OXYGEN.

The globe a, Fig. 22, was placed in a cylinder, i, and surrounded with finely crushed ice. This cylinder stood in a large tank, k, also filled with ice, so that the globe was surrounded on all sides with a layer of more than thirty centimetres of ice. To the joint c was fitted the tube which led to a self-acting Toepler pump, to the apparatus for producing oxygen, and to the syphon barometer. The globe and

* See note, page 28.

its connections having been exhausted and the vacuum having been measured, the pump was shut off by means of a stopcock, and oxygen was then admitted till the desired pressure was obtained, when a second stopcock near d was closed. The pressure of the oxygen was then measured. The globe was left in the ice for this purpose for from one to four hours. The stopcock neard was not required to be

tight against any material difference of pressure on the two opposite sides of its key; its only office was to shut off from the globe, during the measurement of pressure, the part of the apparatus whose temperature was somewhat variable. The use of a fusible metal plug at a convenient part of the tube prevented leakage during the exhaustion of the globe and connecting tubes.

35.-OXYGEN BY THIRD METHOD. SOURCES OF OXYGEN.

In seven of the experiments of this series, oxygen was obtained from potassium chlorate; the process was in all respects exactly like that in the previous series. In the other experiments, oxygen was produced by the electrolysis of pure dilute sulphuric acid. The gas issuing from the voltameter passed through a strong solution of potassium hydroxide, where the time of contact was over fifty seconds; then over heated copper oxide, and through drying tubes filled with calcium chloride, pow. dered potassium hydroxide, and phosphorus pentoxide, each of which was one metre long and two and a half centimetres in diameter, All were connected by