July 17, 1913] 



NATURE 



5°7 



overcome by the use of a half-wave plate, as a 

 rotation of this through a given angle is equivalent 

 to a rotation of the Xicol through twice the angle. 



The quarter-wave plate was built up of strips 

 of mica, 2 mm. wide, mounted so that the principal 

 sections of successive strips make an angle of 45 

 with the slit and go° with each other ; the Nicol 

 would thus transmit, say, the red components of 

 the doublets for the odd strips and the violet com- 

 ponents for the even strips. In a photograph of 

 the spectrum the lines would thus have a dentated 

 appearance (Fig. 2), the magnitude of the separa- 

 tion of the components shown in successive strips 

 varying directly with the strength of the field. 



Every conceivable precau- 

 tion appears to have been 

 taken in setting the desired 

 portion of the sun's image on 

 the slit of the spectrograph, 

 and in securing full illumina- 

 tion of the grating in any ex- 

 posure. A valuable check on 

 the observations was obtained 

 by making duplicate exposures 

 with the quarter-wave plate 

 in the normal and inverted 

 positions, which should give 

 displacements of opposite sign 

 if they are caused by a mag- 

 netic field. As a further check, 

 at least one atmospheric line 

 was measured on most of the 

 plates of the first series, but 

 afterwards they were only 

 occasionally measured, as they 

 were invariably found to give 

 no shifts exceeding the errors 

 of measurement. Possible 

 effects of polarisation pro- 

 duced in the spectrograph 

 have also been carefully con- 

 sidered. 



The region of the spectrum 

 selected, A.5800 to \6000, was 

 determined by the considera- 

 tion that the magnetic separa- 

 tion varies directly as the 

 square of the wave-length ; 

 too great a wave-length, how- 

 ever, being undesirable since 

 the average sharpness of the 

 solar lines decreases as the 

 wave-length increases. Numerous difficulties 

 arising from distortion of the ccelostat mirrors of 

 the tower telescope and other causes were success- 

 fully- overcome, and 280 photographs were ob- 

 tained. For purposes of discussion the photographs 

 and measures have been divided into four series. 



For the preliminary observations it was decided 

 to obtain a large number of measures of a few 

 lines rather than a smaller number of measures 

 of many lines. Three lines, showing the largest 

 displacements, were accordingly selected for sys- 

 tematic investigation, namely, A.c;8i2"i39 (Fe,o), 

 A S 828-o 97 (-,0), .\592 9 -8 9 8 (Fe, 2 ). 

 NO. 228l, VOL. 91] 



The measurement of the minute displacements, 

 amounting only to a few thousandths of a milli- 

 metre, presented great difficulties, largely arising 

 from the natural diffuseness of the solar 

 lines. Full details of the individual measures 

 are given in the paper, and discordances ap- 

 pear to have been as faithfully recorded as 

 the measures on which the final conclusions are 

 based. 



The tables and curves show a marked grouping 

 of positive displacements in the northern and of 

 negative displacements in the southern hemisphere, 

 with values decreasing, on the average, from middle 

 latitudes towards the equator or poles. It is 



4° -o o 20 40 6c 80 N 



Fig. 3. — Displacements 

 ol displacements in 

 Vertical Scale : i di 



i. a, A581? ; b, As8a8 ; c. Mean 

 nd AA5812 and 5838 from fourth 



also shown that the displacements were reversed 

 in direction by turning the half-wave plate through 

 45 (equivalent to a rotation of the Nicol through 

 90 ), or by inversion of the compound quarter- 

 wave plate. Hence it is concluded that the light 

 from the red and violet sides of the solar lines in 

 question is circularly or elliptically polarised in 

 opposite directions. In the northern hemisphere 

 the light of the violet component is circularly 

 polarised in the right-handed direction. 



Some of the evidence for the systematic shifts 

 is reproduced in the appended table, giving the 

 mean displacements for \s930 in the first and third 



