BURGESS.

Polarized light from incandescent surfaces.-Arago was among the first to observe that the light emitted by incandescent surfaces was partially polarized, the amount of polarized light varying widely for different substances, and increasing with the angle at which the surface was viewed, being nil in a direction normal to the surface. Prevostaye and Desain, and Magnus, showed that the infra red waves are also polarized in the same plane as the light waves. Arago attributed the polarization to refraction, near the surface, of the light coming from molecules below the surface.

Millikan," who has done the most complete work in this field, has pointed out that this view of Arago fails to explain why the light coming from platinum near grazing emergence is practically completely polarized, for such light must to a large extent be surface light. His experiments indicate that the polarization is a phenomenon of refraction, and that all the light suffers refraction at the surface. Millikan has examined the amount of polarization in the light emitted by a large number of substances. The phenomenon is most marked in platinum, silver, and gold, and is very feeble in the readily oxidizable substances, such as iron, and in such substances as glass and porcelain. It is quite marked for iron in the molten state.

Our experiments have been made mainly with a view to determining the amount by which this phenomenon could influence the indications of optical pyrometers that make use of polarizing devices to adjust to equality the light from the incandescent body observed and from the standard comparison light.

In the experiments on platinum the temperature of the horizontal strip of a Joly meldometer, maintained constant by regulating the electric heating current, was measured with a Wanner optical pyrometer, which employs polarizing devices. To eliminate the effect of the sag of the strip the readings were taken with the pyrometer in four positions at right angles. For each temperature the measurements were made at different angles with the normal to the surface. In rotating the pyrometer the portion of the radiating strip viewed was slightly changed, but an examination of the strip showed that the heating was very uniform, the variations not exceeding 4° over a Millikan: Phys. Rev., 3, pp. 81, 177; 1895.

5834-No. 2—05—9

From the above table it will be seen that if the temperature of a platinum surface is measured with a polarizing pyrometer large errors may result if the effect of polarized light is neglected. Thus, if the surface be viewed at an angle of 50 with the normal, two readings may be obtained differing from one another by 80° C. at 1450° C., i. e., the reading may differ from the actual temperature by 40° or thereabouts.

As the light emitted in a direction perpendicular to the surface is not polarized, this source of error can always be avoided by viewing the surface normally. For most substances on which optical pyrometers are used in industrial operations, the amount of polarized light is very feeble compared with platinum, so that its effect is entirely negligible. In some experiments where the incandescent surface of iron. was viewed at an angle of 75° the effect was less than 5°C. In any case,

BURGESS.

the effect may be eliminated by taking readings in four azimuths at intervals of 90°.

Effect of diffuse and reflected light. The ordinary diffuse daylight or light from other sources which is not reflected directly into the pyrometer exerts a comparatively negligible effect on the indications of these instruments, not exceeding 5° C. for a change from a darkened room to bright daylight.

The magnitude of the effect of directly reflected light is very variable, depending upon the reflecting power of the incandescent body observed, and on the temperature, size, and location of the disturbing sources. With a good reflecting surface like polished platinum, which remains a good reflector at high temperatures, the effect may be very great, while with a substance like iron (oxide) it is very much less for the same surrounding sources of light. Again, the greater the difference in temperature between the object viewed and the disturbing light the greater will be the effect.

Some of the effects noted with platinum were as follows: A platinum strip at 800° C. had its temperature apparently increased 15° C. by direct reflected light from an incandescent lamp, 120° C. by a large gas flame, and with the strip mounted within a ring of gas flames an apparent rise of over 300° C. was observed.

With a sheet of iron at 725° C., within the ring and with the gas flames burning low, a change of 35° C. was noted in the indications of the pyrometer, while with the flames burning brightly no measurements could be taken. When the iron was at 1100° C., however,. turning off and on the flames made practically no difference in the temperature readings.

The arrangement with the ring burner approximates a number of industrial operations, where heating is done by radiation from flames. The above results would seem to show that even for poor reflectors, such as iron, clays, etc., considerable errors may be introduced at temperatures below 800°, and that the error becomes small above.

1100.

This source of error may be very nearly eliminated by viewing the object through a tube which cuts off most of the light from surrounding flames. Thus in the instance cited above for iron the error of 35° C. was reduced to 5° C. by this means.

The measurement of very high temperatures.-The temperatures that have been discussed thus far are within the range controlled by thermocouples, calibrated to agree with the gas scale to about 1150° C., which marks at present the upper limit of satisfactory gas thermometry. The thermocouple scale is then extrapolated for 500° or 600° more.

Already there are many operations, such as those carried out in the Moissan furnace, the Goldschmidt thermite process, the production of carbides and metallurgical products in electric furnaces, and many pyrochemical reactions that involve temperatures of 2000 or over. It therefore becomes necessary to establish at least some tentative scale that can be used at these extreme temperatures.

Attempts are being made by Nernst and others to estimate these high temperatures by means of chemical phenomena taking place at high temperatures, but this work is still in a preliminary state. For this purpose, therefore, recourse must be had alone to the extrapolation of the laws of radiation which have been verified throughout the range of measureable temperatures.

Lummer and Pringsheim" have recently taken a single set of observations on an electrically heated carbon tube in an atmosphere of nitrogen, using three radiation methods: Photometric (Wien's law), spectrophotometric (A,,, T=A), and total radiation (Stefan's law), the results agreeing to 20 at 2300 C. absolute.

From our own work it would seem that the radiation laws are still in agreement at the temperature of the arc. Our measurements have given as the black body temperature of the hottest part of the positive crater 3690°, 3680°, and 3720° absolute, as determined with the Holborn-Kurlbaum, Wanner, and Le Chatelier pyrometers, based on the extrapolation of Wien's law. Féry gets for this temperature 3760 by a method based on Stefan's law.

On the basis of these experiments it would seem that the several laws of radiation are in quite satisfactory agreement at the highest attainable temperatures, and thus serve to define the same scale of temperatures.

a Lummer and Pringsheim: Verh. d. Deutsch. Phys. Ges., (5) 1, p. 3; 1903. Waidner and Burgess: Bulletin No. 1, Bureau of Standards; 1904.

C Féry: Comptes Rendus, 134, p. 977; 1902.