0'05 inch in thickness; No. 4, of aluminium, o'002 inch in thickness. These four radiometers were plain on each side, no lampblack being applied. Their appearance is shown in Fig. 6. No. 5 was made of aluminium, iden- | tical with No. 4, but the vanes were lampblacked on each side instead of being bright. Had the vanes pointed radially there could have been no tendency for any one of the flies to move either way, but being inclined, the normal movement, on exposure to radiation, should be in the direction of the arrows-a direction which I called the positive direction.

In Fig. 7 the candle is represented shining on the bulb of the mica-vaned radiometer. The rays of light pass through the first wall without action. They then meet the mica, and that also being transparent, the rays pass through it likewise, and then escape through the opposite side of the bulb as is shown by dotted lines, without absorption and consequently without doing work. But in addition to light the candle is radiating ultra-red dark heat-rays, which in great measure are arrested by the glass, and raise its temperature. The inner surface of the bulb then becomes the surface on which molecular pressure is generated, which may be called the driving surface; this is shown by the shading next the candle. This molecular disturbance presses on the mica-vane which is in front of it, and drives it round in the direction of the arrows as if it were subjected to a bombardment of small shot. The vanes, in fact, may be said to be blown round by what may be likened to a wind, which however is not molar but molecular, inasmuch as there is no wind

dark heat applied in various ways to these five radiometers. The results I obtained led me to think that the kind of dark heat might vary in refrangibility according to its source, and that the rays from hot water, hot glass, and hot metal, might affect the materials composing the vanes in a different manner, and being absorbed by one body and transmitted by another, might cause the positive or negative rotation which I obtained. I immersed the five radiometers in boiling water, and after cooling again immersed them in water only a few degrees above the temperature of the room; the results were similar to those I had previously obtained with water of 70° C. The radiometers were covered successively with hot shades of English, French, and German glass of different thicknesses, and at different degrees of temperature. The bulbs were also heated with a gas or spirit flame, but no uniform results were obtained.

A funnel was then heated in boiling water, and allowed to rest on the five radiometers in succession. They all moved in the positive direction, except the bright aluminium radiometer, which remained stationary. When the funnel was removed, the two aluminium and the thick mica radiometers rotated positively till they were cold. The funnel was allowed to cool. It was then inverted over a radiometer, and steam was passed through for a second or two. The same experiment was repeated with each radiometer. The results were now equally uniform with those of the last experiment, but the rotation was

FIG. 8.

in the sense of an actual transference of gas from one part of the bulb to the other.

In Fig. 8 I have endeavoured to represent part of the action which takes place when the candle shines on the aluminium radiometer. The light passing through the bulb falls on the aluminium plate, and raising its temperature, causes pressure to be exerted on all sides. The molecules rebounding from the face next the glass, cause increased molecular pressure on that side, and produce movement in the direction of the arrows, or positive rotation. As each vane passes the candle it takes up heat, and acquires extra driving energy. As it swings round, the opposite side of the glass acts as a cooler, and by the time the vane has completed the circle, and has radiated away some of its extra heat, it is ready to recommence the cycle of transformation-light, heat, molecular pressure, motion.

Unlike mica, which generates very little pressure on its surface, the aluminium fly carries sufficient driving power to enable it easily to pass the dead centre opposite the candle. Therefore, as soon as the candle has shown on the aluminium radiometer long enough to warm the vanes a little, rotation readily continues.

The action of the pith radiometer is similar to the aluminium, except that the dissipation of pressure from the back surface of the pith will be almost mil. The pith, moreover, being sensitive to the heat-rays, and being a non-conductor, moves quicker than the aluminium, which requires time to get warm throughout.

The agreement between theory and observation, so far, seemed exact. I now tried numerous experiments with

FIG. 9.

negative, the bright aluminium fly moving the best o all, and the pith fly the least.

I repeated the experiment with a thick brass ring, the internal diameter of which was about half that of the bulb (Fig. 9, a) and then with another brass ring a little larger in diameter than the bulb, b, Fig. 9. These rings were each heated to about 400°. With the first the rotation was negative, while in the second all the flies revolved in the positive direction. The two brass rings were made red hot, and held in position till the flies were in rapid movement, when the rings were removed and the hot part of the bulb dipped into cold water, so as to chill the glass quickly, and still keep the fly warm. These experiments proved that when heat is applied round an equatorial ring of the bulbs the rotation is always in the positive direction. The hot ring of glass generates molecular disturbance, which presses towards the centre and strikes the sloping vanes, driving them round as if the wind were blowing on them. In Fig. 10 I have tried to represent this action. The positive movement is independent of the material of which the fly is made, and is only slightly increased or diminished according to the conducting power of the fly for heat. The lighter the weight of the fly to be driven round, the easier it moves, and the heavier the fly the longer it keeps in motion after it is once started.

When heat is applied to either pole of the bulb negative rotation takes place. The molecular pressure proceeding

from a hot pole of the bulb will strike the inner surface of the sloping vanes. and driving them before it, will cause a rotation which appears negative to an observer, although it is really positive to the direction of pressure. Fig. II sufficiently illustrates this mode of action. The heat is supposed to be applied near the centre, and the molecular pressure, radiating on all sides, presses the vanes chiefly on the inner surfaces. The anomalous results obtained when the radiometers were heated with hot glass shades or hot water are thus accounted for. Polar heating gives negative, and equatorial positive, rotation, and when both are applied together by immersion in hot water, the direction of motion is governed by the stronger of these two forces.

In my description of Fig. 7 (p. 534) I showed that the glass heated by the ultra-red rays became hot, and acted on the driving surface, generating molecular pressure, and causing the sloping vanes to turn in the positive direction. At the same time the vanes get warm and become themselves sources of molecular pressure. The amount of molecular pressure thus generated depends on the capacity of the material of the vanes to absorb heat. Thin mica will hold very little, thick mica will hold more, and aluminium will hold most. This extra capacity for heat causes more molecular pressure to proceed from the aluminium and thick mica, and generates a proportionate amount of driving power on the surfaces of the vanes, turning them in the positive direction, and supplementing the action of the equatorial ring of hot glass.

The next subject of investigation was the action of radiation on cones, cylinders, and cup-shaped vanes. A pair of

cone and from the outside, away from the side of the glass, is dissipated without acting, but the pressure between the glass bulb and the side of the cone nearest to it is active; the cones, therefore, are pressed round in the direction of the arrows, and the motion has the appearance of attraction.

Cones being inconvenient in shape, I employed portions of cylinders wherewith to shape the vanes, and I ultimately found that cups were more easily affected by radiation than portions of cylinders, whilst they are more easily fashioned. I found that a four-armed cup-shaped aluminium radiometer, the cups being bright and 10 millims. in diameter, and the radius of the curvature being 6 millims., rotates in the light as well as a flat vaned instrument. I sealed one of these instruments on to the mercury pump. During exhaustion accurate observations were taken of the number of revolutions per minute caused by one or more standard candles 3 inches from the centre of the bulb. I also took observations of pressure, and the exhaustion was carried to a very high point. Fig. 13 shows the curve plotted from these observations, taking the rarefaction of the air in millionths of an atmosphere as abscissæ, and the number of revolutions a minute as ordinates. The curve traced through the dots representing observations illustrates the gradual increase of sensitiveness up to a certain point of rarefaction, and the sudden drop after that point is reached.

To still further investigate the action of dark heat on the vanes, I contrived an apparatus to which I could apply a very intense source of heat always ready in the

thin aluminium disks, cut half across the diameter, were bent into cones and mounted on two arms as a radiometer, the cones facing opposite ways. Several experiments were tried and repeated with cones of different material. The movement which appeared most anomalous was the attraction observed when a candle was allowed to shine on the hollow side of a cone or cup-shaped radiometer, the light being screened off the retreating side. Further experiments, however, showed that the effect of bending the plates, or of making cones of them, is to produce a more favourable presentation to the inner surface of the glass bulb. Radiation falls from the candle on the aluminium; some is reflected and lost, but a portion is absorbed, to be converted into thermometric heat or heat of temperature. Aluminium being a good conductor of heat, and the thickness of metal being insignificant, it becomes equally warm throughout, and a layer of molecular disturbance is formed on each surface of the metal. At a low exhaustion the thickness of this layer is not sufficient to reach from the metal cone to the side of the glass bulb; as the exhaustion increases, this layer extends further from the generating surface, until at a sufficiently high exhaustion the space between the side of the glass bulb and the adjacent portion of the metallic cone is bridged over, and pressure is exerted between the two surfaces. Fig. 12 shows how this pressure will act. The direction of pressure is indicated by dotted lines issuing from the metal cone. The more favourable presentation offered by the cone causes the pressure to be greatest between the glass bulb and the outside of the cone; the pressure from the inside of the

FIG. 12.

same place, the heat not having to pass through glass, and being completely under control as to intensity and time of action. The instrument with which I performed the great number of these experiments is shown in Fig. 14. The cylinder is sealed at the top so as to permit of the highest possible exhaustion. It is drawn off narrow at the end, and a stem is sealed in to hold a needle-point. To the narrow end a fine tube is attached to connect the apparatus to the mercury-pump. Round the needle is placed a ring of fine platinum wire, a a, the ends of which are joined to thicker platinum wires passing through the glass. The fly consists of four square vanes of clear mica, bb, inclined at an angle of 45° to the horizontal plane and supported on light aluminium arms. the vanes is a flat disk of clear mica, cc, having a glass cap in its centre, and easily rotating on a needle-point. The vanes and the mica disk are supported independently of each other on separate needle points, which are held in glass rods, d, d, d. A current of electricity from two Grove's cells, turned on or off by a contact key, gives the power of making the wire ring, a a, red hot when desired.

Above

The normal or positive movement of the disk is in the opposite direction to that of the vanes; thus, if the positive movement of the vanes is in the direction of the hands of a watch, the positive movement of the disk is in the opposite direction. With the apparatus full of air at the ordinary pressure (bar. = 761 millims.) the direction of rotation, both of the vanes and disk, is positive when the platinum wire is ignited. The speed of the vanes is 133 revolutions a minute, and that of the disk i a minute.

exactly alike, both rotating together in the same direction. Up to this pressure and at some distance beyond, the vanes have been gradually diminishing whilst the disk has been increasing in speed. At a pressure of 141 millionths the disk rotates rapidly, positively, but the vanes do not rotate at all. At a little higher exhaustion than the last, viz., at 129 millionths, a great change is observed. The vanes which were still now rotate in the positive direction at a speed of 100 revolutions a minute, whilst the disk rotates as before, but with a little diminished velocity. I have previously shown, in a paper to the Royal Society, that the viscosity of air at a rarefaction of 129 millionths of an atmosphere is only a little less than its viscosity at the normal density, and hence it is certain that the vanes at a speed of 100 revolutions a minute exerts a considerable drag upon the disk when it rotates in the opposite direction.

As the rarefaction increases above this point, the speed of both the disk and vanes increases till those of the latter exceed 600 revolutions a minute.

To carry these experiments to a much higher exhaustion it was necessary to modify the apparatus. The complex apparatus I now employed is shown at Fig. 15. Only the upper part of the pump ab is shown. It has five fall tubes and is fitted with a small radiometer, c, and a McLeod measuring apparatus, de, to enable the degree

FIG. 13.

disk commences to rotate in the same direction as the vanes at a speed of three revolutions a minute.

(At low exhaustions I speak of millimetres of pressure, but at high exhaustions I prefer to count in millionths or an atmosphere.)

At a pressure of 706 millionths of an atmosphere the direction keeps the same as at 1 millim. in each case, but the disk makes ten revolutions and the vanes forty revolutions a minute.

At 294 millionths, the speed of the disk and vanes is

FIG. 14. 2

The

of exhaustion in the apparatus to be ascertained. phosphoric anhydride, for absorbing aqueous vapour, is contained in the horizontal tube f. In order as far as possible to prevent the passage of mercury vapour, three long narrow tubes gg are introduced between the pump and the apparatus to be exhausted; the one nearest the pump is filled with precipitated sulphur, the centre tube contains metallic copper reduced from its oxide, and the third tube phosphoric anhydride. At h is a vacuum-tube containing aluminium wires, and having a capillary bore for examining the spectra of the residual gas. An induction coil and battery are connected with the tube by wires. From the tube h two tubes branch off, one of them, 2, leads to the "viscosity" apparatus contained in the case k, and the other, j, goes to the apparatus to be

exhausted.

The apparatus s, containing the rotating disk and vanes, is sealed to the tube j. The platinum ring is ignited by the battery t. On the top of the ring rests a disk of mica, H, lampblacked on the upper surface; this cuts off direct radiation from the hot ring, and diffuses the heat somewhat over the surface of the black mica. Instead, therefore, of the molecular pressure starting from