high frequency and high potential currents. 353

as intensely as tlie other. Reducing the current through thelamps, I may bring the filament in the latter lamp to redness, and,though the filament in the exhausted lamp l is bright, Fig. 186c,the degree of its incandescence is much smaller than in Fig. 1865,when the currents were of a much higher frequency.

In these experiments the gas acts in two opposite ways in de-termining the degree of the incandescence of the filaments, thatis, by convection and bombardment. The higher the frequency andpotential of the currents, the more important becomes the bom-bardment. The convection on the contrary should be the smaller,the higher the frequency. When the currents are steady there ispractically no bombardment, and convection may therefore withsuch currents also considerably modify the degree of incandescenceand produce results similar to those just before shown. Thus, iftwo lamps exactly alike, one exhausted and one not exhausted,are connected in multiple arc or series to a direct-current machine,the filament in the non-exhausted lamp will require a considera-bly greater current to be rendered incandescent. This result isentirely due to convection, and the effect is the more prominentthe thinner the filament. Professor Ayrton and Mr. Kilgoursome time ago published quantitative results concerning thethermal emissivity by radiation and convection in which the ef-fect with thin wires was clearly shown. This effect may be strik-ingly illustrated by preparing a number of small, short, glass tubes,each containing through its axis the thinnest obtainable platinumwire. If these tubes be highly exhausted, a number of themmay be connected in multiple arc to a direct-current machine andall of the wires maybe kept at incandescence with a smaller cur-rent than that required to render incandescent a single one of thewires if the tube be not exhausted. Could the tubes be so highlyexhausted that convection would be nil, then the relative amountsof heat given off by convection and radiation could be deter-mined without the difficulties attending thermal quantitativemeasurements. If a source of electric impulses of high frequencyand very high potential is employed, a still greater number ofthe tubes may be taken and the wires rendered incandescent by acurrent not capable of warming perceptibly a wire of the samesize immersed in air at ordinary pressure, and conveying theenergy to all of them.

I may here describe a result which is still more interesting,and to which I have been led by the observation of these phe-