zotero/storage/GTXVTUXJ/.zotero-ft-cache

BULLETIN No. XLIX.
ON THE CI-IANGE OF WAVE-LDNGTII O F TI-IE IRON LINES I N PASSING FROM THE CENTRE OF THX HUN'S DISC TO THB LIMB,
BY J. EVERSI-IED, F.R.s., AND T. ROYDS, D.Sc.
Following on our researcllcs on the displacements of the solar iron lines at the centre of the sua's disc and at the limb, we now give the results of measures of the change of wave-length of some selected lines in passing from the centre of the disc to tho limb.
It has been shown in ICocla.ilcana1 Observatory Bulletin No. XXXIX that if we select certain lines of iron which in the arc spectrum are not subject to tho density slzift ncar the negative pole, and compare them with
+ the same lines at the centre of the sun's disc, a total of 45 llnes gives an average clisplacement of -I-0.00948,
whilst near the limb the displacement of tho same lines is 0.0150A, showing a cliff~renceof shift, or increase of wave-length, in passing from the centre to a point about one-thirtieth of the raclins inside the limb of 0.005GA. The question whether this increase occurs graclually or suddenly as the limb is approached is s crucial one with regard to the hypothel~iswhich ascribes the limb shift t o s difference of pressure between t11e effective regions of absorption at tlie centre and at the limb ; and the form of the curve representing the change of wave-length between centre a i d limb shoul4 also throw light on the question whether the redward shift at the centre of the disc is due to a downwarcl mavcment raclial to the sni~.
In many inaividual cams the diaerence of wave-length belween centre ancl limb is much greater t l ~ a nt11at indicated above, and in others it is less. In t11e case of the lines at 6301*718mc? 6302.709, which were used by Dun& axad IXalm in their measures of the solar rotation, the 4iRerance cxceeaa O.01A. UiafortunaLely these linea are mongst those which are clisplaced ncap the negative polo in the arc, and we cannot state prccisely what their clisplacement is at the centre of the disc. However, for the purpose oS this investigtxtion, we have used these hnes, and the line at 6280.833, as t11ey are most conveniently situated wi+h regard to tlze ielluwic oxygen lines of the a group, which form standai*dsof reference from which the change of tvave-length in the iron lines may be readlly determined in spectra cxtencling across the gun's clisc. I n aclclition to these lines in the red, it seemed fiesirable also to measure some lines in tho ultra-violet, which probably represent a much
higher level in the s01w atmosphere, and for wliich the limb -centre shift is much smaller. AS tl~creare no
telluric lines in this region, it was necessary t o impress upon the plates emission lines from an arc s]?ectruun:to form atanclarcls of reference.
The methocl is t o project a small solar image on the slit of the spectrograph, centring it precisely BO that the wicle spectrum obtained represents a diameter of the sun. An exposure is then made, sufficient to give a strong image of the spectrum of the more feeble light from the limbs ; and in order to reduce and gracluate the exposwe for the central parts of the image a small strip of metal ifil held in front of the slit ancl moved up and down by hand during the exposure. W ~ t hthis simple clevice only a small amount of practice was ncedecl. to obtain sufficiently uniform density in the photograph.

In regions of the spectrum where there are suitable telluric lines, this procedure is all that is necessary to obtain photographs which will yield the required data. It is only necessary to measure the intervals
separating the selected solar line from one or more telluric lines at the centre and at a number of points equidistant from the centre of the wide spectrum. The solar line will in general be inclined to the telluric
lines by an amount depending on the latitude of the points of intersection of the slit with the sun's limb, thia inclination being of course due to the Doppler shift of the solar rotation ;but in additioiz the measures will show that the line curves slightly towards the red at each end, and it is the nature of this curvature that is the subject of this investigation. A curvature in the same sense due to the action of the grating, being identical
for both solar and telluric lines, does not affect the results. I n the more refrangible parts of the spectrum, where there are no telluric lines, we have found the arc
spectrum of iron to be the most suitable snbstitute, because tlze iron spectrum has been more studied than any other, and it is very necessary to guard against errors due to the displacements of unsymmetrical arc lines near to the poles, which one of us has ascribed t o a density effect. I n the long arc necessary to impress the
plate right across the wide solar spectrum, the wave-length of certain lines will change considerably as one pole or other is approached ;but by selecting symmetrical lines, 11seems safe to assume that the wave-length mill remain constant at all points, especially as the further precaution was taken of moving tlze arc continually up alzd down along the slit during the exposure, and reversing the poles so that half the exposure was made
with the negative pole near the upper end of the slit, and half near the lower end. h difliculty in the use of
the iron arc to determine the shifts of the solax iron lines is that the emission lines fall upon and obliterate the solar lines that it is desired t o measure. This was overcome by the simple process of displaciizg the emission lines through a few angstroms by rotating the grating through an angle of one or two minutes of arc between the exposures on sun and arc. There is no danger in this procedure of changing the inclination of the arc lines.
Amaratus.-The large spectrograph described in Kodaikanal ObservatoryBulletin No. XXXVI was used in this research, the only modification needed being the provisioa of an adjusting and guiding arraizgoment in front of the slit to ensure perfect bisection of the sun's image. A 3-inch lens of 5-feet focus was mounted in the beam of light from the siderostat, and the small solar image (14 mm, diameter) was projected on the slit by a reflecting prism. Immediately covering the slit-plate there was mot~ntecal thin plate of brass with fine circular concentric lines engraved on it of diameters approximating t o those of the greatcat and smallest diameters of the sun's image. A slot about 1m.wide and 16 mm. long was cut in the plate along the vertical diameter of the circles to admit light to the slit, and the plate was then carefully adjusted by screws, 80 $hat the centire of the slot was coincident with the slit. To adjust the image concentrically wit11 the circles, a telescope provided with a collimating lens is mounted so as to view the engraved plate, greatly rnsgi~ifiedby the eye-piece. With the aid of the electric controls of the siderostat, the image viewed with the telescope may be kept as acc-c~ratelcyoncentric during an exposure as the definition of the sun's limb will permit.
Measurement of t7~espectra.-As high resolving power and linear dispersion are necessary in daaling with wave-length differences of less than 0.01A, the third and fourth orders of spectra given by tlze new Anderson
glWatinwg ere used, and these naturally involved rather long exposures both on sun and arc. Sixteen minutes were req~liredto get dense images of the solar limbs in the third order red region, and almost as much in the fourth order violet at L 4380, where the dispersion is 2.4 mm. t o the angstrom ; but in the ultra-violet at h 3928
good images of both sun and arc were obtainecl with six minutes' exposure. In the third order red the linear dispersion is 2'1 mm. to the angstrom, and in the fourth order violet at 3928 it is also slightly over 2 mm. to the angstrom. The resolving power,may be judged from the definition of the very homogeneous oxygen linerj in the a group : the first line of this group, 6276.815 of Rowland, i g barely resolved into two lines with separation 0'0308 ; and the line at 6278.303 is a clearly separated double with components at 6278.276 and 6278.330. Apparently not much would be gained by further increasing the resolving power, since the iron lines suitable for measurement in this region vary from 0108to Og11A in width, and it is this, and the diffusive *characterof their edges, which sets a limit to the accuracy of the measures.
The three lines measured in thig part of the spectrum are all within the group of telluric oxygen lines, and, .are well adapted for the positive on negative method of measurement described in Eodaikanal Observatory
Bulletin No. XXXII. They have a large limb - centre shift, but probably a very small or even negative

centre--arc shift. For the line 6250 it is a small positive shiit but for the other two at 6301 and 6302 it is negative in all the arc aild sun comparisons we have macle, even when the centre of a long arc is used, but the unsymmetrical nature of the lines makes it clifficnltto determine how f a r this may be due to a displacement in

the arc.
- At the other end of the spectrum we have taken, amongst others, the symmetrical Fe lines 3928.075
andl 3930.450, which have a small limb centre shift, but a large centre -- arc slzift. The narrov chromium line at 3928 783 was also measured, and the masrow spark lines of l?e ancl TIat 4385-548 and 4387.007, which

according to Adams have a relatively large limb shift for this part of the spectmm. For all but the red lines, tlze positive on negative method of measuring was fonncl difficult to apply, ancl recourse was had to the

ordinary method of bisection with a spider-thread,

the mean results from the original negative and

from a positive copy.

For the measurement of these spectra, which have a width of about 28mm., a special cross slide provided

with a rnillimetre scale was con~t~uctefdor the micrometer, and for the positive on negative measures a cross

slide had to be provided for both positire and negative images. In either method of measuring, the plate or

FIG. 1.-DISPLACEIMENTOSF FE LTNES ACROSS SUN'S DISU.

mm, 14 13 12 11 10 8
9E=as-t li4m1°b.2

I I
6 4

2

2

Centre

I I

4

6

I I1111
8 10 11 12 13 14 m u w
+ W e ~ltimb
(P = 41°.2

plates have to be moved across the field of view in the direction of the spectrum lines by successive measured intervals from one edge of the spectrum t o the other, the field of view being limited to a narrow strip, either by a mask placed over the plate, or in the positive on negative measures by a mask placed in the eye-piece. The movement is made by a screw bearing on the edge of the sliding plate-carrier.
As the change of wave-length in passing from the centre of the spectrum to the edge is most rapid near the edge, the 14mm. representing the solar radius is d~videdinto 2 mm. intervals as far as 10 mm. lrom the centre, and thence into lmm. intervals, to 13mm. The final measure is made at a point 0 25 mm. within the limb. I n this way, including the measure of the central strip, a series of nineteen cletermii~ationsof the relative positions of the solar and tellnric, or solar and arc lines, is obtained, representing points on the sun7s disc the heliographic co-ordinates of which can be determined from the known position-angle of the slit at the time the plate was exposed. The measured intervals between solar andl reference line, or their diRcrencos, are
then plotted, and a smoot11 curve drawn, from which a small correction to the central interval is obtained, and the other intervals are freed from accidental errors of measurement and possible real irregularities of wavelength. This smoothing process might however have been omitted altogether without seriously aflecting the mean results obtained from several plates : its main purpose is to get a corrected value of the wave-length at the centre of the disc, ancl hence more correct values of the slzifts between limb and centye. There is some
evidence of real, though unsystematic, irregularities of wave-length confirmed by the diflerent lines 011 a pla;te, and amounting at the most to J= 0q002A. We give as an example of this t11e curves in fig. 1, representing the measures of the lines 6280, 6301 and 6302, on the date November 14, 1914. At 2 mm. east of the centre the wave-length of all three lines is in defect, whilst at the centre and at 2 mm. west it is in excess of the values derived from the smooth curve. As different reference lines were used for each of the tl~reeiron lines, this is good evidence of real variation, especially as the lines are best clefined near the centro, and the
variation is about four times greater than the probable error of measurement.
However disconcerting such fortuitous irregularities may be, they are of interest in showing the slight instability of the lines of the solar spectrum, and the need for great caution in interpreting tho results from a single plate in all researches connected with line displacements in solar spectra, ancl especially in solar rotation work.
In the reduction of the measures it is generally necessary t o compute the componsal in tho line of sight of the rotation velocity for each point measured. At times when the solar equator passes througl~,or very near, the centre of the disc, this is not essential, since the plus and minus velocities on the east md west lzalves of the diameter then cancel one another : when, however, the gun's axis is inclined to tho direction of the earth there are appreciable differences, ancl although these are small, it was consiclered wort11 whilo to compute by an appropriate formula the corrections to be applied to the observed displacements in ardor to remove the effect of the solar rotation entirely. A comparison of the observed and computed velocities could. then be used as a check on the general reliability of the measures.
Adopting Adams' formula for the sidereal solar rotation as correctly representing the change of volocity with latitude, the complete formula used for finding the component in the line of sight of the aynodic rotation velocity at any position on the sun's disc is :-
(I 507 cos IQ1 t 0.456 cosS4 -C ) sin A. cos D.
The latitude, $, is determined graphically by plotting the positions of the measured points on t l ~ eappropriate solar chart, of which the Observatory pos$essea an excellent set, carefully constructecl, by the Rev. Father Beaurepaire. 0 is the correction found from Dun&'$ solar tables for converting sidereal into synodical velocity. This is an unnecessary refinement, but its inclusion enables us to compare readily the observed with the computed line shifts. Sine X (the heliographic longitude) is obtained by the formula :-
sine X = central distance x sine angle between sun's axis and slit x secant Q1 D is the inclination of the sun's axis to the line of sight.
I n our experience during the years of 8u~spotminimum, the constants i n Adarns' formula generally give too lmge a value for the solar rotation ; but since we are only concerned here in finding the difference of ahift between points east and west of the centre, any errors in these constants will have no appreciable effect.
In fig. 2 we give a smoothed curve representing the measures of the line 6302, on the date November 24, 1914 when the latitude of intersection of the slit an&limbs was 37O.7, The curve of computed displacements

due to the solar rotation is also given, slightly displaced with regard t o the other curve, to $howtho parallelism of the two near the centre, and their departure as the 11mb shift begins to be manifest. In this particular
plate the results of the rotation shift m e in remarkably close agreement with the computed shlfts from Adnms'
FIG. 2,-UPPER CURV3.I = D I S P L A C ~ M ~ONFT 6302 LINE ACROSS DISu. LOWERCURVE = COMPUT~DDISPLACEMXWT DUB TO SOLAR
ROTATION,

I 1 1 1 1 1 I I I i

I I I I 11rr1

mm.14 13 12 I1 10 8 6 4 2 0 2 4 G 8 10 11 12 13 l4,mm.

East llmb 37O.7.

Coutro

4)W=oa+t

limb 37°*7.

formula. In comparing these, we talc0 tho mean of tlze shifte at correspolzcling poirzt;~cash ~ L X ' L*~WUEIL of tho centre, both for the ob~ervecal nd coml~atcclshifts, ill tl~isway elirninaiing tho limb shifl. In bblo I wo givo the results for the three lines mcssurecl on the abovo cht;t;e. In. Lhe rncan valnc~lfor tho lllroa linos ,Lho lfzrgosi residual is only three in ihe fourth decimal, and t11e moan pcrce~~tagclciflmencc: about $ por conl antler llzo

computed values.

8h@s of iron lznes 6edzusc3n cofitre and I.ilnZi, d1~(3t o c o ~ n p o / n ~ nO$f soZaln vnotation. Iv1A120000. (Cb at kmb= ~ Y ~ . Y . )

1,

. . I Meall percorztage difforonca

-0.82

1

About the same order of accuracy i s shown by the positive on negativc mea8ureil of the three pat@& obtained on November 14, 1914, but in these the mean rotation values are ahout 21; per cent smaller t l ~ n those derived from Adams' formula.
2

These results serve to indicate the accuracy attained In the measurement of these rather wide line$ by the positive on negative method. Measures have also been made of the same lines in other plates by the ordinary metllod, and by another measurer, and whilst the general form of the curve representing the shift across the disc is the same, there is a less satisfactory accordance in tlie rotation shifts. The same remark applies also to the ultra-violet lines measured 111 the orclinary way, tlle chromium line at 3928.783 and tlie Fe and Ti lines at
4385.548 and 4387.007 giving the best results, on account of their narrowness ancl well-defined character.
The stronger ultra-violet lines meas~~reacrie those at 3928.075 and 3930 450. These give very small limb shifts, ancl in. some plates it appears to be a negative shift : the displacement is towards violet lnstead of towards red. As these llnes are rather wide, and the measures are dliflicult, we have some doubt as to the reallty of the negative shifts. I t snggests a variabilxty of the limb shift, of which we have indicat~onsalso in some of the measure$ of the red lines.
ResuZts of the ~?zeusures.-Leaving the question of the negative shifts for further investigation, we now proceed t o give the results of the measures, after elimination of the rotation shlft, and taking the mean. of the east and west displacements. The cli~gramfig. 3 exhibits the mean change of wave-length between the centre of the disc and the limb of the Fe lines 6280, 6301, and 6302. These are derived from lour plates
FIG.&-CHANGE OF WAVE-LENGTH OF FE LINES
BETWEEN CENTREI 03' DISC AND LIMB. Mean of 4 plates,

Centre

I

*4

*5

*6

'7

-8

9

Limb.

measured by the positive on negative method. The abscissls are here given in tenths of the solar radius, and the ordinates in angstroms. The latitudes of the points of intersection of slit and limbs varied between 26O.4 and 41"-2t,he mean being 3 5 O . 6 . The plates were secured on dates November 14 and 24,1914, under practically perfect atmospheric conditions.
The outstanding feature of the curves is the small distance from the centre at which the limb shift begim to manifest itself. This begins at approximately 0.3 of the radius from the centre for all three lines. but

the curve for the 6280 line falls below the otl~etrwo, the increase of wave-length being at first less rapid than for the line 6301. There is also a slight inclication of a decrease of wave-length for tlie line 6301 at the point measurecl nearest to the centre. This decrease does not exceecl 0*0002A,ancl must be considered doubtful. Three of the foar plates measurecl give minus values, and one a plus value.
TWOother plates taken on March 18 and 19, 1915, at latitudes 82O.3 and 8G0-9 (the slit being almost coincident with the sun's axis) were measnrccl by tho orclinary method. Taking the mean results of a positive ancl a, negative image, these show rather larger limb shifts for the lines 6301 ancl 6302, but a smaller shift for the line 6280. The rates of increase between centre and limb are however very similar to those given in the curves. It cannot be said that latitude has any appreciable influence on the limb slzifts, whiclz appealato be the same at all points on the cir~umfc~encofe the sun. But, as already mentioiiod, there is some evidence of variability of the limb shift on plates taken on iliEerent dates, which needs further iavestigntion. FOY instance, a plate talren on December 25, 1914, at latitucle 69' 1,yielcls rernarlcably small limb shifts for all three lines, the values being about one lialf of those usually obtainecl. It may be added that all of the plates were taken under conditions of exceptional purity of ~lrga, lthougll tlie definition of the sun's image may have vasiecl considerably.
Fig. 4 shows similar curves for the lines 43850548 and 4387.007. These lines were selected because &heyhave relaii-vely large limb shifts for this part of the spectrum, ancl. they fall near to the strong Fe line

FIG.4.-CHANGE OF WAVE-LENG'PFI OF ENI-TANCED
L I N l S OF FE AND TI BETWDEN CENTRE AND LIMB.
Mean 01 4 plates.

4385.548 p Fe.

Centre 01

'4

*5

.7

18

-9

Limb

4383-720, a symmetrical line which is readily reversed in tho arc, the absorpt~online forming an excellent line of reference. They are due t o enhanced Fe and enhanced Ti respectively. Four plates were exposed ion October 9, November 19 and 23, 1914,the latitudes of intersection of slit and limb being 27' 3, 5 5 O v 6 , 50°.4 and 51'-7.
Fig. 5 shows the ultra-violet lines measured. The lower curve is the mean result obtained for the hoo strong iron lines 3928.075 and 3930.450. Thege lines, as already stated, have a very small limb shift and
FIG. ~.-WAVE-&ENGTHSOF ULTRA-V1O'J;ET LINES BRTWHFIEIN
CENTRE AND LIMB.

Centre -1

02

*3

4

05

96

.7

-8

9

Limb

a large centre -arc shift. As tlicre is no appreciable difference between them, the mean of the two i s given,,

from two plates. The upper curve in thls diagram is the narrow chromium line 3928,783, the mean of four

plates. The clata for this aiagram are derived from plates taken 011 December 25, 27 and 28,1914, and January 8, 1915. The latitacles range from 40°.4 to 81' 2.

Cornl3a~ingthese ultra-violet lines with those a;t; the red end of the spectrum, wo find the iacrease of

wave-lengtll begins at about the same distance from the centre in both ; but the ult~a-violetlmos, and

especially the strollg lroil lines, show a very muclz amaller rake o f illcrease near the limb than the roll lines.

The enhanced lines in fig. 4 give egsentlally the same type of curve as the red iron lmes.

Discussion of?~~uZts.-IiiKodailcai~aOl bservatory Bulletin No. XLTI the question was raised as to whether

the general movement of reces~ionindicated by the (lisplacement of the Iron lines towards red at;the centre of

the sun's disc was the result of a ciro~ilationof the solar gases in a direction ra&al to the Bun, or was part of

a general movement of recession from the earth. The pre~en~t eriesof measures appears to show that the displacements at the ce1ltl.e are not due to a radial circulation on the sun, but are probably part of a general

ciisplacemelsk increasing from the centye towam the llxnb. If the deacenaing movement at the centre 0%

the disc were due to a circulation radial to the sun, the component of motioi~i n the direction of the earth would diminish from the centre towards the limb in proportion to the cosine of the angular distance from the centre, and this would be apparent in an initial dccrease of wave-lengtll, most marked for those lines having a
large centre-are shift. Tliis decrease would continueuntil counteractedby the increasing limb shift,producing
a asplacement curve across the disc very different in form from that actually found. So far as we have gone, the line 6301 is the only one which shows ally iencleiacy to a clecrease of wave-length, but this line has a very
s-11, or possibly a negative, shift at the celllre of the disc, while the sixong lines in the ultra-violet which
have a mean centre --arc slilft of 0.OIGA exhibit no diminution of wave-length, but a continuous increase from
a p i n t not very far irom the cei1tl.e. In ICoclaikanal Observatory Bulletin No. XXXIX it was shown that the relative slxifts between the limb and
centre cannot be explained by a dlEerence of pressure i11 the effective regions of absorption at the limb and at tile centre of the disc, and the displacement cuyves now found aRorcl aclditioi~alevidence in the same sense, as the following coiisiderations Will zial<eclear.
According to 13alm'r~hypothesis, slC the limb the pail1 of the light through the lower layers is increased in greater ratio t h a i ~that through the upper layers. It is easily seen, however, that owing to the sligl~dt ept11 of t h e reversing layer this effect can only be appreciable very near to the limb. I t is sufficient to calculate the ratio between the tllicdcness of a layer at the bo.tiom of-Lhe reversii~g~ t r a t u mandl one at the top, since for these
extreme layers there is the maximum possible difference. I11 fig. 6 let 0 bo the centre of the sun, and P the
centre of .the disc seen from the earth. A ray of light starts towards the @art11from A on the plzotosphere at

angular distance @ from the centre of the disc, and emerges from the reversing layer at E,at distance 8' from

+ . . . . . the centre of the disc. Then 8' is gxven by the relation

sin 0' = R. in t?/R T

. .

(1)

where R equals the radius of the sun, and T equals the total thickness of the reversing layer. Now

length of path AB of the ray through a thin layer of thickness BC = d at the base of the reversing layer is d/cos 6, and sirmlarly t h e length of path DE through an equally thick layer at the summit is d/coa 4'. The

relative increase in the lowest layer is therefore

cos 0'lcos 8

This ratio is given in the following table, obtained with the help of equation (I),assuming that the sun's radius equals 960", and the total thickness of the reversing layer equals 2".

Whatever assumptions we may make as to the difference of pressure in tho up,per ancl lower layers, and also as to the effect of the difference of path on the pressure displacement of a line, i t is eviilent from this table that no tlifference of path is appreciable except between 85' mc1 90° from the centre, ancl therefore n o change of wave-length can occur except qulte close to the limb. Thus, at about 8 5 O , or 0.996 of the radius from the centre, we might expect a line to begin to be widened on the red side, the violot edge remaining. in it8 normal position owing to the absorption at the highest level. This widening would be of a diEusa cl~aracter, since the rays of light would traverse successive layers of decreasing pressure and t;emperature. The effect of
a higher temperature in the lowest strata would indeed tend to neutralize the incrcasecl absorption due to a longer path, and it is by no means clear that any displacement whatever woulcl be absc?rved. EJowever this
may be, as all the displacement curves agree in showing that the ahift is continued far within the limb, it
cannot be due to the effect of pressure differences. The pressure theory of both limb ancl centre shifts is also
8ficnlt to reconcile with recent views as to the effectivedeptlzs in the reversing layer where absorption occurs, For we now believe that the fainter lines originate at greater deptha than the gtronger lines, conseqnontly they
should give larger shift$at the centre of clisc because of the greater pressure, and smallcr limb -centre
because of the smaller clifferelzce of path for radial and tangential rays. This is just the opposito of w h a ~is
- actually found ; on tlie average the fainter lines have smaller centre arc shifts and. largcr limb -conhe shiftfi~
than the stronger lines. Owing to the remarkable inverse relation between the limb shift and the centre slzift which has been
ghown to exist (Kodaikailal Observatory Bulletin No. XXXIX), we seem justified in assuming tentatively that both are due to a single cause, m done in which pxeaaure playa no part. But the centre shif La have been
interpreted as due t o descending movement, which is greatest in the higher levels, and ilecreases, pogsiljly to
' zero, in the lover levels of the reversing layer ; and this interpretation has been endorsed by Dr. St. John. 9
If the limb shift is also a Doppler effect ws may add the two shifts, and represent the change of ~vave-length between centre and limb as in the diagram fig. 7, in which t11.e upper curve represents the ultra-violet lines 3928 075 and 3930 450, and the lower curve the red line 6301'718, which we will assume has a zero centre-
arc shift. If we accept St. John's conclusions as to the different levels representecl by lines in different spectral
regions, the ultra-violet lines repregent a high level, and therefore show the large descending velocity of which the ordinate 0'016A ia the eqmvalent. The xed line, on the other hand, represents a very low level, and i~ produced in a region where this downward movement is arrested. At the limb, however, there is mzxch less differencein the displacement^, the wave-length of the red line being increased by 0*010A,and the wave-length

" Kodaikanal Observatory Bulletin NOS.XXXVI and XXXIX.

Mt. Wllson Annual Report, p, 266, 1914,

of the violet lines by 0.002 only. The velocities at the limb will then be for the red line 0.5 km/sec, and for the violet lines 1-4km/sec.
FIG.7.

.3

-5

6

7

8

I)

Limb

We thus see that at the limb a great depth of the reversing layer is aRected by the movement, while at %hecentre of the disc only tlze upper layers are in motion. If we could conceive of the Earth exertlng a repulsion on the solar gases, this behaviour of the high-level and low-levellines becomes intelligible ; for at t h e centre of the disc the resistance of the denser layers and of the photosphere itself woulcl arrest any downward movement, while at the limb the entire stratum would be more or less free to move tangentially to the
solar surface. The telocity interpretation of the limb shifts inevitably leads to an Earth effect, even if we assume
combiw~tionof radial and surface currents on the sun to account for the observed displacement curves.
Formulae combining a radial and a surface current can be constructed to Et the displacement curves
satisfactorily, but the surface current must be towards or away from the centre of the disc, a point on the sun which has no significance except in relation to Earth. Also as the radial currents involve an ascending motion, whilst we find a descending movement, it seems futile to discuss these formula
To escape from the hypothesis of an Earth effect, we must assume that the solar line shifts depend upon the direction which the rays of light take in traversing the reversing layer, the shift being least for rays passing normally through it, and most for the tangential rays. The effects of anomalous dispersion will no doubt .appeal to many as giving a probable explanation of the facts, and doubtless the anomalous dispersion theorv w o u l d lend itself admirably t o this problem as to so many othera.

We consider nevertheless that the Doppler hypothesis should not be lightly discarded until indisputable evidence is forthcomingthat anomalous dispersion is really an effective agent in displacing solar lines, The
recent work of Albrecht in regard to this appears to us to be entirely discounted by o u ~measures of close double lines in sun and arc,' and by the subsequent disclosure o f very large overestimates in Rowland's measures of the separations of many solar double lines. The actual close agreement in the sel3arations of arc and
solar double lines seems indeed to rule out anomalous clispersion as an effective agent i n displacing solar liilzes. I t remains to mention a crucial experiment which woulcl decide the quehon whether the displacements
are dlue to motion radial to the Earth, ancl affect only that side of the sun clirected toward the Earth, for by the aid of the planet Venus we can compare the generalmzecl spectrum of a hemisphere of tlie sun t~zuncsd 90' or more from the Earth with the hemisphere facing the Earth. If the displacement of the solar
lines affects only the hemisphere facing the Earth, there will be a clieerence of wave-length betwoon tho lines of the planet's spectrum, when corrected for her motions in the direction of East11 and sun, and those of ordinary sunlight.
I t is probable that with modern instruments the line-of-sight velocity of Venus could be xcaclily
determined with a probable error as small as It; 0.2 kdsec, while the difference of wave-length expectecl corresponds to from 0.6 to 1km/sec, according to tlie lines chosen.
If there is no such difference of wave-length between the light from the two sicles of tho sun, tlmn it will
be necessary to find some explanation of the line shifts other than motion in tho line-of-siglzl.
SUMNhRY.-(1) The first part of this paper describes the method of obtaining spectra repra~entinga diameter of the sun, and the use of the telluric lines at the red end of the spectrumn, ai~ctlhe r e ~ 0 ~ 8 &o1f sthe superposed arc lines of iron at the violet end, as reference lines for determining the displaccment~of the solar lines across the clisc between centre and limb.
(2) The methods of measurement and reduction are described, and the accuracy of the positivo on nogative method of measuring is illustrated by comparing the observed with the calcalatecl shifts clue to the
components in the line-of-sight of the solar rotation movement. (3) Small irregularities disturbing the smoothness of the curve which represents tho changs of wave*
length between centre and limb are indicated. (4) Diagrams a;re given showing the change of wave-length between contre and limb of the rud Sinos
6280.833, 6301.718 and 6302.709, of the violet lines 4385.548 and 4387-007, and of tho, ultra-viol& line8 3928.075,3928.783 and 3930.450.
(5) The form of these curves indicates that the redward shift, or clownward movement, at the centre of the &sc is not due to a radial clrc~~latioofn the solar gases, but is probably part of the gonoral displacemmt
increasing towards the limb. (6) It is shown that a difference of pressure between the effective region of absorption at tho limb and
at the centre of the disc will not account for the displacement carves, since tho, limb shift continues for a long distance within the limb, while a pressure effect could only be appreciable from '99G xnclii from "Ie cen"tro
to the limb.
(7) It is suggested tentatively that both limb shift and centre shift are clue to a single causc, a gonoral movement directed away from the Earth, all over the &sc. This movement affects only the higher parts of the reversing layer at the centre of the disc, because of the resistance offered to a downward movement by the
lower layers or the photosphere : at tlie limb the movement d e c t s the lower strata also, because thero is little resistance to p movement to the solar surface.
(8) It is suggested that observation of the wave-length of the lines in the spectrum of Venus would decide the question whether the shifts are due to a recesaian from the Earth.

THE OBSERVATORY,KODAIRANAL, 4th March 1916.

J. EVERSEED. T. EOYDS.

" The Obsesvatory,"January, 1916

1 [;Pn08,8 CWWI~S

. NADRbS PRINTXD BY TBE STSPERKNTENDENP, O.OVEl3.NMENT PRBSS-1b16.

---__CI