that it does apply. The absence of some of the hydrogen lines in the solar spectrum has already been noted. That the red and blue lines can be seen is no doubt a consequence of the fact that hydrogen exists in much greater quantities than helium, for it should be noted that the helium lines are not bright, but only insufficiently dark to be observed.

This comparative weakness of some Fraunhofer lines which are very prominent in the flash-spectrum, and are probably due to the high temperature of the portion of the solar disk emitting the correspondent radiation, has been commented upon by Mr. Evershed, whose explanation I consider in the main to be correct. Although a further discussion of some points of detail may be desirable, the matter is independent of scattering, and lies therefore outside the range of this communication.

VICTORIA UNIVERSITY,

Manchester, Eng.

THE REVISION OF ROWLAND'S SYSTEM OF STANDARD

WAVE-LENGTHS1

BY LEWIS E. JEWELL.

In considering the desirability of revising Rowland's system of wave-lengths, and of changing to Michelson's absolute values, it is well to consider carefully what were the probable sources of error entering into the determination of Rowland's or Bell's absolute values, Rowland's "New Table of Standard Wave-Lengths," and Rowland's "Preliminary Table of Solar Spectrum Wave-Lengths." It is also desirable that we should consider what work bearing upon this subject has been done since these tables were constructed, and what material has been accumulated capable of furnishing a basis for more accurate tables; and what work it is desirable to do in the near future, in order that the whole subject may be placed upon a satisfactory basis, with the least friction and the least expenditure of energy. In the discussions which have been in progress for some time regarding the corrections to be applied to relative wave-lengths, and the relations between solar and metallic lines, there are some very important factors which have been almost completely ignored. In the first place, it is well to consider the various errors to which the wave-lengths of lines in the various tables published by Rowland were subject.

The absolute values of wave-length were derived from determinations by Bell, influenced in a measure by determinations of other observers. These were determinations of the wave-length of lines in the solar spectrum, and were corrected, approximately at least, for temperature and air-pressure. Since then better values for the refraction of the air, at different pressures and temperatures, have been determined. The measurements by Bell were also corrected for motion in the line of sight, of the observer's position, as caused by the Earth's rotation upon its axis and revolution around the Sun. in a slightly eccentric ellipse.

1 Paper read in abstract at the Conference on Solar Research, St. Louis, September 22, 1904.

The first determinations of wave-length used in Rowland's early tables were made by Rowland and Koyl, mostly from eye-observations, and were at best only approximate values, and many of the lines taken as standards were of an unsatisfactory character. These measurements were used in the preparation of Rowland's later tables, but were not given much weight, and some of the lines were discarded. Later eye-measurements of relative wave-length were made with both the concave and plane gratings by Rowland and Crew, and afterward by myself from photographic plates taken by Rowland with two or three concave gratings of different values for the grating space. These photographs were upon very fine-grained plates, and the definition in general was very good, but, except in a few cases, no data were given from which the corrections to be applied for temperature and pressure of the air could be derived. Also in the eye-measurements no corrections of this character were made. Rowland himself did not pretend that the measurements were accurate to less than one-hundreth of an Ångström unit, and, such being the case, he had little patience with the idea of making these small corrections; although it is likely that they were responsible for many errors, both systematic and accidental, in his tables, which at times might have been rather greater than one-hundredth of an Ångström

unit.

Having made practically all of the measurements from the photographic plates, and the calculations in the work of reduction, I can speak more positively regarding them. They consisted, for the greater part, of two groups of plates, nearly all taken by Rowland, upon fine-grained emulsions prepared by himself. The first set comprised eighteen plates, 14 in. (35 cm) long by 13 in. (4.8 cm) wide, of overlapping spectra for the determination of relative wave-lengths by the method of coincidences, and consisted of a middle strip containing a portion of the solar spectrum in some particular order, and on each side of it strips of a portion of some overlapping spectrum of another order. These were taken by the aid of a shutter which in one position allowed the middle spectrum to reach the photographic plate, and when the shutter was rotated allowed the spectrum of the outside strips to reach the plate, the middle strip being then covered up by the shutter. Rowland was in the habit of giving one of these strips

two exposures, one both before and after exposing the other strip. He did this under the impression that, if anything happened to the apparatus between the first and second, or between the second and third exposures, the average of the first and third would equal the second exposure. This, however, by no means follows. If the effect that produces the error be of a continuous nature, such as the heating of the grating or slit during the exposure or some such cause, that might be true; but if it were due to the act of turning the shutter in some direction and then back again, or some mechanical movement or jar, this would by no means follow, and the third exposure added to the first might simply introduce an error where there had been none before. Also the first change in the shutter might introduce an error in one direction, and the return of the shutter to its original position might introduce an error in the opposite direction, bringing things back to the original condition; and the first and third exposures might thus very well be coincident, but the second exposure would not be compensated for by the average of the other two.

In photographs taken by myself I have found it better to see to it that the shutter worked easily and then to make only two exposures, turning the shutter and doing everything else to avoid any pressure against the apparatus in any direction, and to avoid jars of any kind to parts of the spectrograph. There is no question whatever but that for such work the shutter should be disconnected from the spectrograph, and made entirely independent in its working, or that the parts should be carefully balanced, work easily, and be operated pneumatically or electrically, so that there can be no displacement due to motion of the shutter or pressure against parts of the apparatus. There is no doubt in my mind that in the plates taken for the measuring of coincidences errors of this kind did probably occur, and at times. may well have amounted to considerably more than one-hundredth of an Ångström unit. It is quite possible that the errors arising might be such as to cause the lines of the second exposure to be always shifted in a given direction with respect to the other two; and the errors might possibly differ as the camera of the spectrograph was moved to different portions of the spectrum, for even a shift in the position of the observer will cause some variations.

In photographs containing solar and metallic arc-spectra upon

the same plate, the solar spectrum was usually taken in the middle, and the arc spectrum outside. Some of these plates were taken by Rowland, and some of them by myself under Rowland's direction; but scarcely any of them contained data for the determination of motion in the line of sight, or for correcting for atmospheric pressure and temperature, where lines of different orders of spectra overlapped. Another remarkably fine set of plates of the solar spectrum alone were taken by Rowland for making his "Photographic Map of the Solar Spectrum;" but only one of them was used in these measurements, and scarcely any of them had data marked upon

them.

The measurements upon both sets of plates mentioned were made by myself, nearly all of them upon a dividing engine constructed by Mr. Schneider under Rowland's direction. The screw of this dividing engine was carefully made by Schneider, and is remarkably free from errors, but the attached microscope was unsatisfactory. The definition of the microscope was not perfect and the field not flat; also some other parts of the dividing engine were not entirely satisfactory. The definition of the plates for coincidences of overlapping solar spectra for the determination of relative wave-lengths were, in general, fairly good, and some of them excellent. The plates for the coincidence of solar and metallic lines were, some of them, good, some bad, and others indifferent. Practically all of them were devoid of data for corrections. One source of trouble was experienced with photographs taken upon thick plates of uneven thickness from the varying focus of different parts of the plate and consequent parallax. Adjustment was made to avoid the trouble as far as possible, and later on means were adopted for nearly getting rid of the trouble. There was, however, always some parallax from curvature of the field of the microscope, but the trouble was avoided and allowed for, as far as possible. The plates for comparison of solar and metallic spectra were 19 in. (48 cm) long by 1 in. (3.2 cm) wide, and bent to the focal curve of the grating.

Notwithstanding the difficulties mentioned, measurements made upon good lines could, in general, be relied upon to two or threethousandths of an Ångström unit, and some good lines to one-thousandth of a unit.