rests; the oscillations of the pans and the suspended weights are soon stopped by the friction between the disks and the velvet. But when the pan arrests are further moved down 3.5 centimetres, the disks are left perfectly free.

This apparatus for weighing by reversal was first used with the balance made by Becker, but the apparatus for damping oscillations was then different. When the Rueprecht balance was put in place of the Becker balance, the difference in the arrest for the pans made it necessary to modify the apparatus previously used for this purpose, and to readjust the distance of the hooks a and b from the axis of the mechanism, to suit the shorter length of the beam of the new balance. FF are counterpoises which support the weight of the object carrier, and of the objects placed on it.

26.--OXYGEN BY SECOND METHOD.

WEIGHING BY REVERSAL.

All the weighings of this series of determinations were made by Gauss' method of reversals. The globe containing the oxygen was suspended from one of the auxiliary pans of the reversal apparatus just described, and its counterpoise from the opposite hook. Weights were placed on the auxiliary pan, on the side from which the globes were suspended (not on the pan of the balance itself), nearly sufficient to produce equilibrium. Except in two cases, weights smaller than ten milligrammes were not used here; as the weights were removed from observation till the end of the operation of weighing, the falling off of a small weight in closing the doors of the closet might not be noticed at the time, and it was then a question whether it fell off at the closing of the door at the beginning of the weighing, or at the end. But if a weight of ten milligrammes or more fell off, there was no ambiguity.

The globe with its weights, being at the left of the observer, was suspended from the pan of the balance itself by manipulation of the handles shown in Fig. 16. The beam of the balance was then released, and the weights smaller than ten milligrammes which were needed were added to one side or the other as indicated. The balance was then arrested, and the current of dry air introduced into the desiccator, passing through the axis of the reversing mechanism. After a proper interval, the balance was released and three excursions were noted; then the reversing mechanism was used to transfer the globe to the right of the observer, to suspend it to the pan, and to stop its vibrations. The small weights placed on the pan of the balance itself were by hand transferred to the opposite pan, and three excursions were noted. The mean of the apparent weights required in the two positions gives the difference between the weights of the globe and its counterpoise as well as it can be given by a single observation. But in general it will not be the true difference,

It will be observed that currents of air produced outside of the closet of the weighing apparatus have but little effect in disturbing the balance. The openings in the desiccator are small, and are at the same level. It is therefore impossible for a current of air to enter the desiccator near the top, and leave it near the bottom, striking against one of the globes during the passage. If the whole body of air in the small room surrounding the closet becomes less dense, the whole body of air within the closet will ooze out at some opening near the bottom; but this movement will not take place in part through the desiccator.

But if convection currents are produced within the closet itself, they are very likely to enter the desiccator, unless this has a certain symmetrical exposure to them; and if convection currents are produced within the desiccator itself, they are still more likely to disturb the balance to which the globes are suspended.

Suppose, for instance, that the distribution of temperature in the closet is such as to produce currents; a downward current may well enter the desiccator and strike upon one of the globes. Currents in the same direction will also be produced within the desiccator, since its inequalities in temperature will be in the same direction as those of the walls of the closet. The globe on the left hand, let us say, will appear too heavy. If now the positions of the globes be reversed, the currents which enter the desiccator from without are reversed in their effect on the globes, but those which are produced within the desiccator itself will continue for some time, making the right hand globe appear too heavy. These currents, therefore, make the globe appear too heavy in both positions.

As a matter of observation, the convection currents produced within the desiccator are so much the more important that the effect actually noticed in conditions when weighing is possible is attributable entirely to them. When the effect of the currents entering the desiccator from without is noticed, the equilibrium of the balance is so inconstant that weighings have to be postponed. This has occurred but seldom.

Fortunately the effect of the currents set up within the desiccator is easily eliminated. If we wait for, say, half an hour, until we may assume that a new constant condition of temperature has been attained, and again weigh by reversal, the convection currents produced within the desiccator will produce an effect on the apparent difference of weight of the two globes which is opposite to the effect produced in the first weighing. If the conditions of temperature have not materially changed, and if the time used in the manipulation is about the same, the two opposite effects will have about the same magnitude, and will nearly disappear from the mean. The mean of two successive weighings by reversal with an interval of half an hour or an hour ought, therefore, to give very nearly the true difference of weight of the globe and its counterpoise.

That this is the case may be seen from many illustrations in my note-books. I give one in which the globes counterpoised against each other were both closed by fusion, so that their true difference of weight was constant. The volumes were equal on the two sides of the balance, and were 15.77 litres.

TABLE SHOWING THE ELIMINATION OF THE EFFECT OF AIR CURRENTS:

It will be observed that, although successive weighings differ two or three tenths of a milligramme, the means of pairs of successive weighings agree as well as could be desired.

27. OXYGEN BY SECOND METHOD. CHANGE OF SURFACES OF GLOBES.

In this series of experiments, the surfaces of the globes were not touched during the determination of the weight of the globe when empty and that of its weight. when filled with oxygen. In one case, seven successive experiments were finished without touching the globe. Commonly, some trifling accident or the necessity of renewing the lubrication of the stopcock changed the weight of the globe or of its counterpoise; in the series of seven determinations just mentioned, the difference between the weight of the globe and of its counterpoise was determined three times, and the agreement was good.

Although changes in the surface of the globe or of its counterpoise were not to be feared in this series of experiments, it was thought worth while to determine the amount of the changes which might be expected if cleaning were necessary. To eliminate errors which might be caused by changes in the lubrication or by leakage, two globes were used which had been closed by fusion. To one of them was added a flask, also closed by fusion, so as to make the volumes of the masses on the two pans of the balance equal; each measured 15.7667 litres.

The globes had been standing in the balance room for a year. For the first experiment, both were moistened by breathing on them and wiped with a dry

cloth. Their difference of weight was then determined as they hung in the desiccator; it was found to be, as just given, 41.8190 gr.

One of the globes was now washed with distilled water and at once wiped dry and replaced in the desiccator; this was at 12.40 P.M. on April 30. The fol lowing observations were then made :

The mean is 41.8168; so that the washed globe had lost 2.2 milligrammes. On May 11, the same globe was washed again, its counterpoise being simply wiped as before; both were replaced in dry air at 9.30 A.M. The following observations were made:

The mean is 41.8172; so that the globe had lost no more in weight by the second washing soon after the first.

On May 14, the same globe was washed by scouring gently with powdered pumice-stone and distilled water. It was then wiped, passed through the flame of a Bunsen lamp, and replaced in the desiccator at 5 P.M. tions were made:

The following observa

On May 18, at

9 A.M., the same globe was washed with pumice-stone, using hard friction; it was then wiped and put in the desiccator at 12.30 P.M. Weighings were made as follows:

Obviously the first weighing was made too soon. The mean of the other weighings is 41.8158 gr.

The weighing after the first washing shows the well-known loss of material produced by the solvent action of water after glass has been exposed for a long time to the decomposing effect of the air. The next weighing shows that a second washing soon after the first may have a much less solvent action; that of May 15 shows that gentle friction with powdered pumice may not remove anything from a surface already well washed.

If it were safe to argue from the increase from 41.8168 to 41.8175 gr. we might perhaps infer that repeatedly wiping the second globe was gradually producing a loss such as the first suffered from the first washing; but the inference is too insecure.

28. OXYGEN BY SECOND METHOD. PREVENTING LEAKAGE OF GLOBES WHEN EXHAUSTED. Since, with a balance of such gratifying constancy, but a few repetitions of a weighing were necessary, it would have perhaps been safe to trust the stopcocks which had been fused to the globes. But since the manipulation which entirely prevents the leakage of the stopcocks when the globe is exhausted is by no means difficult, it was always used in this series of determinations.

FIG. 20.-Method of preventing leakage of globes when exhausted.

At g, Fig. 20, is the ground joint of the stopcock by which the globe is always connected to other apparatus. The piece bd, which fits on g was reduced to a small diameter at c (the tube at c should have been drawn as at d) and a mark was etched here; the piece was then put in position on the stopcock, and the volume contained between this mark and the key of the stopcock was determined by filling from a burette. This was done once for all. When the globe was to be exhausted, this joint was fused to the joint de, and drawn to a small diameter at d, ready to be closed by fusion when exhausted, and the volume between the mark e and the point d was determined by filling it with water and weighing. The two joints having been thus made ready, b was warmed and coated internally with wax ready for cementing on g. Then the two joints with this wax were carefully weighed by