relates to sparks without other capacity than lead wires a meter long. Results for carbon (Acheson graphite) are very uncertain on account of the lack of intensity of the carbon spectrum in comparison with compound spectra and the spectra of the impurities (calcium and titanium) present. These foreign spectra were left out of account in the estimate of the ratio of carbon to gas spectra. Values for oxygen with inductance are given as they appeared with no account taken of the shift to infra-red. The mercury spark was about 3 mm long from mercury to mercury through a hole in the side of an auxiliary glass tube. (9) Effect of current density.—Varying the current has a pronounced effect in intensifying the spectra of the electrodes at the relative expense of the spectrum of the intervening gas. The effect is well known and considerable data is available. Varying the current varies the amount of vaporization and ionization, and these factors chiefly govern spectral predominance. Vaporization is, of course, not so much due to a heating of the electrode as a whole as to a skin effect due to the great energy lost in the electrode fall of potential at the surface. Whether the electric current first causes a local heating and this heating vaporizes the metal, or whether the current tears the particles directly from the surface and this electrical vaporization causes the heating afterwards we are unable to state, but evidence for the latter process appears stronger. This energy loss at the surface of an electrode being equal to the product of current and electrode fall, the latter being roughly proportional to the current, we should have vaporization approximately proportional to the square of the current. Since the luminosity of a conducting gas is very nearly proportional to the current, the intensity of the spectrum of the spark or arc atmosphere would be proportional to the current. On the other hand the luminosity of the vapor of the electrodes depends on the variable amount of vapor present as well as on the current. Since the amount of the electrodes vaporized is roughly proportional to the square of the current, the luminosity of the electrode vapor would be proportional, approximately, to the cube of the current. Hence a heavy current would greatly assist the preponderance of metallic over gas spectrum. (10) Effect of electrode temperature.-Raising the temperature of an electrode increases the vapor pressure of the metal, but diminishes the electrode fall of potential," hence the most advantageous temperature will be an intermediate one, varying widely with the substance used. But in the case of nearly all metals of high melting point both coeffi a Cunningham: Phil. Mag., 9, 193; Feb., 1905. cients are very small, and the effect of electrode temperature is negligible. (11) Effect of distance between electrodes.-By varying the length of the column of the conducting gas one may obtain metal and gas spectra in any proportion from pure metal to pure gas. Current, pressure, and distance from an electrode of the observed portion are the chief variable factors to be considered. At high pressures we may have all gradations from an arc to a lightning discharge; at low pressures, everything from an arc to a glow discharge. Current and electrode fall of potential determine the amount of metallic vapor that will appear at a given distance from an electrode. Hence at constant current, pressure and distance apart of electrodes need not be considered, except in so far as they affect electrode fall. Since neither anode fall" nor cathode fall' varies greatly with pressure, at constant current the amount of metallic spectrum showing in a given part of a discharge is practically constant. Even in a low-pressure discharge tube 40 cm long, at a pressure of 0.06 mm, I have found the spectrum of the electrodes (aluminum and magnesium) to be a considerable portion of the canal ray glow spectrum. The effects of varying other conditions than those above discussed appear to be vanishingly small if existent. Varying the e. m. f., for example, has little if any effect so long as current remains constant, account being taken of the very considerable effect on the form of the discharge of varying the control resistance. Nor does the form of the current wave in an intermittent discharge appear to affect the relative intensity of the metallic spectrum. Adding capacity increases the maximum current enormously, but, if the mean current is kept constant, does not appreciably increase the preponderance of the metallic spectrum. The same may be said of the auxiliary spark. Inductance changes only the form of the current wave. Resistance of course would change not only wave form but mean current, but at constant current does not affect metallic preponderance. CONCLUSIONS. The results of this investigation show that the same three general laws (p. 400) hold for the distribution of spectral energy between electrodes and gas as hold for other mixed gases. Relative spectral energy depends upon the relative amounts of the two gases or vapors present and, with moderate excitation, upon the atomic weight. The amount a Skinner: Wied. Ann., 68, p. 752; 1899. of electrode vapor present depends upon current and electrode fall of potential, being a surface effect proportional to the energy used in metal-gas conduction. This electrode fall is again a function of a number of other variables. It is nearly proportional to the current, varies slightly with the pressure on the gas and the temperature of the electrodes. The conditions necessary for obtaining a gas spectrum free from the spectrum of the electrodes used, or for obtaining the spectrum of a given metal as free as possible from the spectrum of the surrounding atmosphere, may then be mapped with some certainty. The general case is considered, namely that of a column of gas forming part of an electric circuit that is otherwise metallic. The conditions favoring a pure gas spectrum are (1) small current density at the electrodes, (2) substance of electrodes to be of some metal like aluminum, copper, or platinum showing small vaporization, (3) electrodes to be at a considerable distance apart and only the middle portions of the discharge are to be used. Gas density, capacity, inductance, series spark, and other conditions have but little effect, and hence may be adapted to other requirements. All three conditions above stated are realized in the Plücker tube, the loss in luminosity caused by the use of small currents being compensated for by the central constriction. In the ordinary arc or spark under the most favorable conditions, it is difficult to obtain a gas spectrum more than 80 per cent pure, whereas by decreasing the current and separating the electrodes, practically 100 per cent purity may easily be realized. To obtain a pure electrode spectrum, we must have (1) large current density and (2) electrodes near together. Oxygen appears to be the most favorable gas to use with the majority of metals, probably on account of the greater cathode and anode fall of potential in this gas. The above conditions are realized in the arc and spark, 95 per cent purity being easily obtainable. Pressure being of little consequence, it may be reduced to give narrow lines, a series spark being added if necessary. Other independent conditions may of course be varied in adaptation to experimental requirements. March, 1905. USE OF WHITE WALLS IN A PHOTOMETRIC LABORATORY. By EDWARD P. HYDE. It is usually considered by photometricians that one of the essential requisites for accurate work in photometry is that the walls of the photometer room be made as nearly "dead black" as possible. It is surprising how long this tradition of photometry has persisted, particularly as it necessitates the exclusive use of a room for photometric measurements, since a room blackened in the customary manner is unfit for any other work. Perhaps it is due to this fact that one so often finds the photometric laboratory in a small room in the basement or garret, or some other place that is unsuited for any other purpose. In all open-flame work, however, much larger errors may result from poor ventilation in a small room than from light reflected from white walls of a larger room, if a proper set of screening diaphragms is used. FIG. 1.-Horizontal section through center of photometer screen. So far as I know nothing has been published as to the magnitude of the error due to white walls when the photometer is protected by black diaphragms, although this arrangement is at present in use to some extent in the photometric laboratory of the Physikalisch-Technische Reichsanstalt of Germany. In the photometric laboratory of the Bureau of Standards the walls of one of the rooms, which is about 29 feet long, 19 feet wide, and 12 feet high, were left white for a year after the construction of the building. Recently they have been given a light terra-cotta finish for decorative purposes, but the following measurements were made while the walls were still white. The photometer is placed near the middle of the room and is supplied with a set of diaphragms covered with black velvet. These diaphragms may be so arranged for any position on the bar that no light, except that reflected from the lamps under test or from the diaphragms themselves can reach the photometer 2214-No. 3-05—9 417 screen, as shown by the accompanying sketches. Fig. 1 is a horizontal section of the photometer through the center of the LummerBrodhun sight-box, showing the arrangement of diaphragms A, B, C, and D. Elevations of these diaphragms perpendicular to the axis of the photometer are given in fig. 2. By reference to fig. 1 it is seen that an eye placed at P could see nothing but black velvet diaphragms and the lamps under test. Conversely, no light could reach the screen except from the lamps or from the diaphragms. In case a Bunsen or Leeson disk is used, extra precautions would have to be taken to prevent light from penetrating to the screen from the front of the box. It is essential that this condition be fulfilled if the walls are to be left white, as any reflected light reaching the screen directly would materially affect the result. Furthermore, it is desirable that all outside light be excluded by the use of dark curtains, as under these circumstances the only light that can reach the photometer screen, apart from the direct light from the lamps, is that part of the light from the two lamps which is reflected by the walls back to the lamps and the diaphragms, and thence to the photometer. In order to determine the magnitude of the errors resulting from this reflected light from the white walls of the photometer room of the Bureau of Standards, the measurements described below were made. The results of these measurements show conclusively that the error is entirely negligible. Before giving the results of the measurements it may be well to say a word or two in regard to the method of measurement employed at the Bureau in the comparison of incandescent lamp standards. On the carriage on the right side of the photometer is placed a well seasoned incandescent lamp, though not ordinarily a standard. This lamp is raised to such a voltage that its color is approximately the same as the color of the lamps to be compared on the other side of the photometer, and it is kept at this voltage throughout the test. The carriage on which this comparison lamp is mounted is then rigidly connected by means of adjustable links to the carriage on which the photometer screen is mounted, and at a suitable distance to produce the desired illumination of the screen. The standard is then mounted on the carriage on the left side of the photometer, and is placed at such a position as to give a balance with the photometer at approximately the middle of the bar. The standard is brought to the proper |