the saving of Iceland spar and rectilinear passage of the light through the system. In addition, great sensitivity would be secured because a practically vanishing line would be obtained between the halves of the field, as the small strip of spar need not be over 0.1 mm thick. The device was perfected and used by Bates, who subsequently succeeded in combining the two cells into one. The Brace sensitive-strip spectropolarizing system is the most sensitive yet devised. However, it is rather fragile, and its use is not recommended except in work requiring the highest obtainable precision.

(c) TESTS OF POLARISCOPES WITH CIRCULAR SCALES

Polariscopes with circular scales will be accepted for test (see test-fee schedule 421, p. 553). A thorough examination is given all optical parts. The scale will be checked for as many points as desired.

(d) NATIONAL BUREAU OF STANDARDS EQUIPMENT

Polariscopes with circular scales for measuring absolute rotations are now made by most polariscope builders. The National Bureau of Standards is fortunate in having a large Schmidt and Haensch precision instrument with a silver scale reading to 0.001°. The Lippich polarizing system is unusually good, the larger nicol having an available opening of 15 mm. The instrument is mounted on a cast-iron base 1 m in length. Owing to the necessity for accurately controlling the temperature while measuring rotations, a large air thermostat, consisting of a wooden box 40 by 55 by 60 cm covered with asbestos, is mounted between the polarizing and analyzing systems. Access is had to the interior by means of a small door in the side. The room in which the instrument is located is thermostated at a temperature approximately 1 degree lower than is desired in the polariscope air bath. An electric heater then brings the bath temperature up to the proper point and maintains it there. The heater is made of fine resistance wire wound around a large framework which fits inside. the box. See figs. 26 and 27, p. 100-101. The current is controlled by means of an electronic relay operated by a mercury contact, which in turn is operated by toluene contained in a series of glass tubes so constructed and placed as to give a maximum change of volume in a minimum of time when a small temperature change takes place. The air is kept constantly stirred by a small fan. The temperature remains constant to 0.01° C. No mechanical relay of any kind is used, and consequently there is no trouble from relays sticking.

A large Weiss electromagnet, figure 8, equipped with a suitable polariscope is available for the study of magnetic rotation. This magnet is cooled by water circulation, thereby permitted continuous use even when heavy currents are employed.

2. LIGHT SOURCES FOR CIRCULAR-SCALE POLARISCOPES

(a) GENERAL

In accordance with the rapidly increasing use of polarimetry in commercial and scientific work, there has arisen a demand for greater accuracy. The largest source of error in precision measurements is in the light sources. The production and utilization of suitable light sources is by far the most difficult problem with which the polari

scopist must cope, and it therefore receives continuous study at this Bureau. For many years any sodium source was considered suitable for this work. Then came the so-called light filters of Lippich [22] and Landolt [23]. These filters, however, are open to two severe criticisms: (1) The efficiency of the purification is a function of the intensity of the source, and (2) the available light is reduced by absorption in the liquids used in the cells.

Subsequently, spectrum filtration came into use for precision work. In this method, light from an intense source is passed through an optical system containing a dispersing medium, and only the desired wave lengths are permitted to enter the polariscope.

(b) SODIUM

(1) FLAME. Until recently the sodium lines have been the one intense source with which a large percentage of the precision work has been done. This source has been used for determining practically all of the polarimetric constants, including the standardization of quartz control plates. Unfortunately, the two sodium lines are difficult to separate from the remainder of the spectrum, and as the flame is intense there is danger of one or the other of the lines reversing. In 1906 [24] a careful study of spectrum lines as light sources for polarimetric work was made at this Bureau. It was found that the lack of intensity, the high dispersion necessary for purification, the presence of other lines of considerable intensity in the neighborhood of D, and D2, as well as the unstable line structure under certain conditions, render this source far from satisfactory. Owing to the great precision

required in present-day polariscopic measurements, intensity is of paramount importance. Even in commercial instruments the halfshade angle is generally not over 10 degrees, which means that the

polarization plane of the analyzing nicol is practically at right angles to the planes of the polarizing system and less than 1 percent of the light is transmitted. Thus only a very small fraction of the incident light ever reaches the eye of the observer. After an extensive investigation, the Bureau has found that for an intense sodium flame source the best results are obtained by feeding some form of fused Na2CO3 into an oxyhydrogen flame. In utilizing this source a rod of this substance is placed in the flame at one of the positions shown in figure 9. The line structure of the sodium lines obtained by this method was studied with an echelon spectroscope. The lines were found to be extremely sharp and to differ from the arc spectra by the absence of the characteristic haziness in the edges of the lines. If displacements as large as 5 A had occurred, they would have been readily detected. The broadening was at all times symmetrical. In this respect the observations agree with the more recent work of other observers, but are contradictory to the results obtained by Ebert [25], which have been accepted in polariscopic work. By careful manipulation of the flame in position 3, reversals of the lines took place, D2 preceded D, in this respect.

FIGURE 9.-Oxyhydrogen sodium flame studied at the National Bureau of

Standards.

TABLE 1.-Structure of sodium lines at different intensities

It is thus evident that so far as their line structure is concerned, the sodium lines can be depended upon to give a sufficiently definite optical center of gravity up to the point of reversal in position 3 (fig. 9). However, very noticeable variations in polariscopic measurements are likely to be observed with sodium sources at different intensities. These variations, it is believed, are not due to changes in the line structure of the source, but to the difficulty of excluding all impurities in the light, even with a very narrow slit and a dispersion sufficient to separate D1 and D2. This extraneous light constitutes a different percentage of the total illumination whenever the intensity of the source varies. It may, therefore, appreciably change the optical center of gravity, thereby giving a different apparent rotation when

the field appears dim than when it appears bright. Aside from the color and stability of the sodium lines up to reversal, there is little in their favor as a polariscopic source. The flame requires the constant attention of an assistant, and even in reversal the intensity is not nearly sufficient to permit the use of the greatest sensitivity of a good polarizing system. The sodium flame sources have been summarized by Landolt [26] in table 2.

TABLE 2.-Optical center of gravity of sodium light sources, Landolt [26]

Various types of sodium lamps have been designed to give the greatest possible intensity consistent with a minimum of attention. In the Pribram [27] lamp, fused salt in platinum boats is exposed to a Bunsen flame. When the supply in one boat is exhausted, the boat is withdrawn and a second containing a fresh supply is quickly introduced. Fairly constant illumination is obtained for a long time. In the Schmidt & Haensch lamp, fine platinum wires are bent and inserted into a spoon, the melted salt being drawn up to the point by capillarity to the hottest portion of the flame. In the Landolt [28] type, a Muencke burner (Bunsen lamp with conical wire-gauze top and sufficiently strong air supply to cause the inner dark cone of the flame to disappear) is used. Exposed to the flame are two heavy nickel wires, around the middle of which nickel gauze is wrapped. The gauze is charged by immersing in melted salt. An intense flame is obtained. The Zeiss lamp is a simple and convenient type. Pumice stone saturated with salt is exposed to a Bunsen flame. The position of the pumice stone with respect to the flame is important and is easily controlled by means of a thumbscrew. For a very intense sodium flame the method of molded sticks of fused sodium carbonate previously described, is the best.

(2) ELECTRIC SODIUM LAMP.-Two forms of electric sodium lamps with somewhat different characteristics have recently become available. They are the General Electric sodium vapor lamp [29] made in this country (fig. 10A), and the Osram sodium vapor lamp [30] made in Europe (fig. 10B).

The General Electric lamp must be run from a special transformer housed in the base. The bulb has considerable area which is uniformly illuminated. It must be used on alternating current and is rated at 60 watts.

The Osram lamp, on the other hand, has a comparatively small working area but has a higher intrinsic brightness. It is constructed to operate either from alternating or from direct current (not both, however, as the electrodes are somewhat different in the two types). It uses about 1.4 amperes and may be plugged directly into the power