somewhat in the direction at right angles thereto. It was apprehended that the air might tend to aggregate in masses as soon as it departed from the perpendicular direction. This disposition was intended to quicken the velocity a little at that point and make the change of direction as abrupt as possible. This involved some loss of head, more than was necessary, as I am inclined to think. The flaring of the passage, after assuming the horizontal direction, was designed to utilize the vis viva lost by the water in passing from the higher to the lower velocity. The dimensions of the pit did not admit of so long a horizontal passage as was desirable. A glass gauge was first applied to the air chamber, as in the arrangement already described. The drenching to which it would subject an observer, however, decided me to abandon this, and adopt an arrangement bringing all the essential facts of the process under the eye of a single observer above ground. An opening, g, was made in the side of the air chamber just 21 inches below the top, opening into the ascending shaft. The filling of the air chamber to the depth of 21 inches was announced by the appearance of large bubbles in the chamber D, followed in a few seconds by a violent commotion therein. By raising the lever, h, the air accumulated in the chamber was discharged, the cessation of the loud hissing indicating the complete emptying of the chamber. The water was discharged from the chamber Dover a weir 5.56 feet long. A gauge in this chamber indicated the depth on the weir, and, in connection with another in the chamber A, the fall. The fall, it is hardly necessary to observe, is the height of the surface in the chamber A above that in the chamber D. The pressure of the air is determined by the height in the latter above that in the air chamber. The leakage, which was considerable, was readily determined by observing the rate at which the water fell in the chamber D after shutting off its admission. The fall was varied by adjustable weir plank. At each change of weir plank, the reading of the gauge was taken when the water was exactly at the crest of the weir. This reading subtracted from that taken during the flow of the water gave the depth on the weir. No microscopic nicety, but only substantial accuracy was attempted in any of these observations. The tank and the lower part of the ascending shaft were of four inch pine plank. The rest of three and two inch plank, all confined by an immense number of screw and lag bolts with large cast iron washers. The timber was put in quite dry and joints paid with white lead and oil in the expectation that the swelling would make all tight. This expectation was not fully realized, the great pressure starting some of the joints and causing considerable leakage, amounting in some cases to nearly half a cubic foot per second. This was, of course, added to the quantity passing the weir. As originally constructed, the upper part of the shaft was provided with a siphon, as in previous arrangements. Meeting with great difficulty in exhausting the air, I decided to take it out and rebuild it on a different plan, but in the meantime took occasion to make an experiment which I had contemplated from the first, viz., to ascertain whether the introduction of the air could not be equally well accomplished by giving the water a slight fall at its entrance. The readiness with which water in violent agitation breaks into foam is familiar to all, as well as the fact that foam is nothing more than water impregnated with minute air bubbles. The method adopted was the one indicated in the sketch, viz., allowing the water to fall into the descending shaft over a barrier of stop-plank. The complete success of the experiment made it apparent that the siphon might henceforth be dismissed from the method, although it would undoubtedly have some advantage as regards efficiency, especially with a low head. In the theory of the subject formerly presented I assumed 12 inches per second as the velocity with which bubbles of air rise in still water. This was founded on experiments with bubbles formed by air escaping through orifices. In the process above described the air appears to be broken up into finer bubbles, which do not rise so fast. I found that a velocity of 0.75 ft. per sec. in the descending shaft brought an appreciable quantity of air into the chamber; 0.86 ft. brought it in freely. The mode of conducting experiments on compression was as follows: The water being turned on and flowing in full volume over the weir, the air chamber being empty, the escape valve was closed by the lever, h, and at the same instant a stop watch was started. The gauges in A and D were observed and recorded; then the eye was fixed on the surface in D. On the appearance of large air bubbles here, the watch was stopped. It must be observed that when the velocity in the descending shaft much exceeded three feet per second large quantities of air were carried past the air chamber, and rose in the ascending shaft. There was no danger, however, of confounding these bubbles with those announcing the filling of the air chamber, the latter being larger and immediately followed by a great commotion in the basin D. The passage of air past the air chamber disposes of one difficulty that has been predicted of this method, viz., that the air would tend to aggregate in large masses. No tendency of this kind was observable in the air coming up the ascending shaft, though it had traversed a distance of over 70 feet. It reached the surface in fine bubbles, very nearly uniform in size. The capacity of the air chamber, filled to a depth of 21 inches, was 71.19 cubic feet. The following is an example of the computations : Take experiment No. 1. Fall, 4.07 feet; depth on weir, 0·63 ft. ; quantity of water, 1.665 × (5.56 0.2 × 0·63) + (leakage) 0.41 = 9.45 cubic feet per second. Pressure of air, 27.58 ft. water. Power of 1 cub. ft. air-Admission, 27.58 62·3 =1780·5 ft. lbs. Time, from closing of escape valve to appearance of bubbles, 2 m. 48 s. Velocity of water in ascending shaft, 9.45 = 1·18 do. of air, 2·18 ft. 8 No account was taken of the leakage of air, which was slight. The assumption that the mean pressure during expansion is half the initial pressure rather over-estimates the power of the air. On the other hand, taking an atmosphere at 34 feet water (its value at the sea level) leads to a greater error in the opposite direction. The bottom of the pit is about 750 feet above the sea level. A large number of experiments were made before discovering the most advantageous conditions of working. Table I contains all the experiments covered by the dates therein given. The best efficiency is 52 per cent., being about double the best result obtained with the apparatus, Fig. 2. This percentage would, of course, be entirely unsatisfactory in a practical motor; but an air-compressing system on a practical scale would exceed this model, in dimensions, in greater ratio than this exceeds the apparatus represented in Fig. 2. The object of these experiments is to ascertain what could be expected from such a system. Let us first see how these results accord with the theory of the subject. TABLE I. Results of Experiments in Compressing Air by causing it to mix with a Descending Current of Water. The losses of power, or the losses of head, which are the same thing, the head being taken to represent the power in any given experiment, are of three kinds: First, the loss at the entrance, or the fall required to impregnate the air with bubbles. This did not admit of very close measurement, but, as near as could be estimated, did not differ materially from one foot in all the experiments. Second, the resistances to movement. In experiment No. 5, with no air passing, a velocity of 4.49 feet per second in the descending shaft corresponds to a head of 0.98 feet wholly absorbed in resistances to movement. This gives the means of finding the resistance to movement for any other velocity, such resistance being represented by a head which is proportional to the square of the velocity. Third, a certain loss is occasioned by the fact that the air does not move downward with the same velocity as the water. The air bubbles, as I find, would rise in still water at the rate of 0.75 foot per second. When the water, therefore, is moving downward at the rate of 3 feet per second the air goes only 2·25. There is, then, a loss of power of 0.75 25 per cent. of the head |