JOURNAL OF THE OPTICAL SOCIETY OF AMERICA

Wavelength Passed by Hg198 Zeeman-Split Absorption Filter

WALTER G. SCHWEITZER, JR.National Bureau of Standards, Washington, D. C.

(Received 6 May 1963)

We have measured the vacuum wavelength of the 2537-A\ line of Hg19 9 as passed by a Zeeman-split absorp-tion filter. The 6057-A line of Kr8 was used as a standard. An evacuated Fabry-Perot interferometer wasemployed with spacers of 218-, 110-, and 21-mm length. The linewidth obtained by using the Zeeman filter isestimated to be about 5 mK. The measured wavelength 2537.26874i0.00003 is in excellent agreementwith the value obtained by Barger and Kessler for the wavelength emitted by an atomic beam of Hg'98 .

INTRODUCTION

THE use of Zeeman-split absorption lines as verynarrow-pass filters was first suggested several

years ago.' We have been working with Zeeman filtersfor the 2537-A line of Hg'98 and have achieved passbands of about 5 miK. This is about 21 times as narrowas the krypton standard line (6057 A). This Zeemanfilter is, therefore, attractive as a potential standard inits own right. We have previously suggested the 2537-Aabsorption line in an atomic beam of Hg'9 8 as a goodprimary standard of length.2 At the same time wesuggested that the Zeeman filter might serve as afunctional standard which could provide nearly thesame precision as the atomic-beam device but with fargreater simplicity and convenience. This scheme willbe useful only if the wavelengths defined by thesedevices can be measured precisely in terms of thepresent standard of length, and also if the wavelengthdefined by the Zeeman filter is identical, within thedesired precision, to the wavelength defined by theabsorption beam. In this paper we are reporting theresults of the measurement of the wavelength of theline passed by the Zeeman filter relative to the kryptonstandard.

EXPERIMENTAL METHOD

The optical arrangement used is shown in Fig. 1.The 2537-A line from the electrodeless lamp is ab-sorbed and reradiated by the resonance cell, focusedthrough the Zeeman filter, collimated, and passedthrough an evacuated Fabry-Perot interferometer. Thefringes are focused onto the slit of a Gaertner L254large quartz spectrograph. Light from the Kr8" sourceis focused onto the resonance cell and then passedthrough the same optics as the 2537-A line.

The Fabry-Perot talon, consisting of 100-mm-diamfused quartz plates coated with aluminum, separatedby Invar spacers, was mounted in an evacuated cham-ber. It had an effective finesse of about 6 and an effi-ciency of about 35% at 2537 A. This effective finesseincludes the effect of lack of plate flatness on the in-

1 K. G. Kessler and V. G. Schweitzer, Jr., J. Opt. Soc. Am. 49,199 (1959).

2 K. G. Kessler, R. L. Barger, and W. G. Schweitzer, Jr.,"Mercury Atomic Beams for Metrology," in Symposium No. 11InterferomJzetry (Her Majesty's Stationery Office, London, 1960),p. 67.

strumental width. The interferometer design permitteda clear aperture of about 8 cm, but of this only thecentral 5-cm region was used, this being the size of the1-m projection lens.

A thin resonance-fluorescence source for the 2537-Aline was used instead of a discharge tube for tworeasons. This source eliminates interference from anearby, nonresonance line at 2535 A which is not re-solved by this spectrograph at the desired slitwidths.The resonance source also tends to minimize the inter-ference from the 0.2% Hg'99 in the lamp. This pointis elaborated further in the discussion of the results.

The Zeeman filter consisted of a cell 25 mm long ina magnetic field of 900 G. The temperature of the cellwas about 80 'C. The mercury in the cell was of thesame isotopic purity as that in the lamps and resonancecell-99.8% Hg 19 8 , 0.2% Hg1 9 9 .

The krypton lamp was given to us by K. M. Baird,NRC, Ottawa, Canada. The composition of the kryptonwas 99.7% Kr 6 and 0.3% Kr84. This lamp has acapillary 73 mm long, 3.4 mm i.d. and 5 mm o.d. It wasoperated at the triple-point temperature of nitrogenwith a current density of about 3.8 mA/mm2 in thecapillary. The lamp was viewed along the capillary inthe direction cathode-to-anode-to-observer. These con-ditions are within those recommended by the Inter-national Committee on Weights and Measures.'

Care was taken that the same portion of the inter-ferometer was used for both the krypton and mercurylines. This was not difficult to do. It was not possibleto achieve uniform illumination over the entire aper-ture, especially in the case of the krypton source. Theuse of the field lens shown in Fig. 1 helped a great dealin this respect but did not completely eliminate thenonuniformity. Because it was feared that variationin intensity across the fringe patterns might have in-troduced a systematic error, the following check wasmade. Wavelengths were calculated from each ring ofthe interference patterns, and the average of all ex-posures of each ring were then compared with theaverage computed in the usual manner. A systematicerror due to this effect would be expected to result ina systematic difference in wavelengths determined usingdifferent rings. No such systematic differelce was founid.

'Proces-Verbaux Seances Comit6 Inter-Natl. Poids Mesures28, 71 (1960).

1250

VOLUMLE 53, NUMBER 11 NOVEMBER 1963

ZEFMAN-SPLIT ABSORPTION FILT 1F25

L254SPECTROGRAPH

FABRYPEROT

FIG. 1. Optical ar-rangement for measure-ment of Hg198 Zeemanfilter wavelength withkrypton standard.

ZEEMAN Hg198

RESONANCEFILTER / CELL

Hg198

LAMPS(WATER-COOLED-ELECTRODELESS)

The standard deviation for a single determination ofthe average for each ring from the over-all averagewas about 2.5 A for the 218-mm spacer, 6.4 PA for the110-mm spacer, and 13 MA for the 21-mm spacer (seeFig. 2). We conclude from this that nonuniform il-lumination did not contribute any significant syste-matic error to the final wavelengths.

From 12 to 22 exposures were taken with each ofthe three different talons-218.2, 110.4, and 21.5 mmin length. Exposure times were 4 to 10 min with 103a-Oplates for 2537 A and 103a-D plates for 6057 A. Theexposure times were the same for the two lines in eachcase. The short 6talon was used primarily to determinethe dispersion of phase-change correction. Two longspacers were used to guard against possible systematicerrors due to overlapping components.

Eight fringe diameters on each pattern were meas-ured on a vibrating-mirror photoelectric comparator.4

The order number at the center of each fringe patternwas determined with the aid of an IBM 7090 programfor the least-squares method described by Meissner.5

RESULTS AND DISCUSSION

As mentioned above, one of the reasons for using athin resonance cell as a source for the 2537-A line wasto eliminate interference from Hg199. This was not fullyappreciated at the beginning of these experiments andas a consequence, it was necessary to repeat the wholeseries of measurements. In the first experiments, aresonance cell 25 mm in length was used as a source.This cell was optically thick for both the 198 and 199components of the 2537-A line. The wavelength of the2537-A line was measured relative to the Kr8 6 standardby using the two long spacers (218 and 110 mm) andone short spacer (21 mm). The wavelength obtainedwith the medium length spacer was actually the largestof the three, before correcting for dispersion of phasechange. The difference between this and the wave-length obtained from the longest spacer was fairlysmall (only slightly more than 100 A), but well out-side the probable error of measurement. After lookingfor possible causes of a small systematic error of thistype, it was decided that it might well be due to thesmall trace of Hg199 in the cells and lamps. The ratioof intensity of 199 to that of 198 in the electrodeless

M. L. Kuder, J. Res. Natl. Bur. Std. (U. S.) 65C, 1 (1961).5 K. W. Meissner, J. Opt. Soc. Am. 31, 405 (1941).

lamp is quite a bit greater than the ratio of theirabundances because self-absorption is much greaterfor the 198 component. Furthermore, since the reso-nance cell was sufficiently long in this case (25 mm) tobe optically thick, it reradiated the 199 componentsto almost the same extent as the 198 component.

Figure 3 shows the positions of the 199 componentsrelative to the 198 component for the two long spacers.It is difficult to draw a quantitative conclusion fromsuch a figure because the 199 intensities are not known.

40Cg , | 1 ' ' I

218.2 mmSPACER

(22 EXPOSURES)

20

0

-20

-40

60

40110.4 mm

SPACER 20(12 EXPOSURES)

uaA 0

-20

-40

-60

-80

120

100

80

60

4021.5 mmSPACER

(20 EXPOSURES) 20

ILA 0

-20

-40

-60

-80

-100

-120

-1401 2 3 4 5 6 7 8 9 10

RING

FIG. 2. Deviations of averages for individual rings from over-all average (2537-A line, Zeeman filter, and 1-mm resonancecell).

Kr8 6

SOURCE

FIELD LENS

- I I .. I

- I I I I

1251November 1963

W \,ALTER G. SC-IWEITZER, j t.

t9a

1I995 -.4 -3 -.2 -_I 0 1 .2 .3 .4 .5

RELATIVE ORDER NUMBER-218.2 m SPACER

198

l998

199AI I I

-. 5 -. 4 -. 3 -.2 -.1 0 .1 .2RELATIVE OROER NUMBER-110.4mm Si

FIG. 3. Positions of the 199 componentscomponent for the two long spacers. The rih198 line is intended to indicate roughly thfringes-not their shape.

The triangular shape of the 198 linindicate roughly the width of the 198 fshape. There is a tendency in the casifor the apparent position of the 198 friitoward lower order numbers. If the shber were about the same for bothamount to twice the wavelength shifof the two spacers. This would teninconsistency.

The experiment was then repeateda resonance cell only 1 mm thick. TI

TABLE I. Summary of measuremenZeeman-filter wavelength.

25-mm resonance cell)lb WKX)c

8 2537.26876911

1211

812

it2220

1220

2212

2537.268833

2537.268716

1-mm resonance cell

(X)_v2537.268744

2537.268736

2537.268751

Average for 1-mm resonance

2537.26874±0.00003 X

This is the average of the values for both mm spacer with the dispersion of phase-cltained from the 21.5-mm spacer in each case.

0.9 of the 199 energy but only 0.1 of the 198 energyincident upon it. A summary of the results with thetwo different resonance cells is shown in Table I. Inthis table the corrections for dispersion of phase changehave been made but no corrections have been includedfor Doppler or Stark shifts in the krypton source. Theaverage wavelengths (X\a-. number of observations n,standard deviations

af= rE (ai(x),)2/ (11 - ) T,2 i=1

relative to the 198 and the corrections applied for dispersion of phaseangular shape of the change 6, are all included in this table. The consistencye width of the 198 of the values obtained with the 1-mm resonance cell is

well within the standard deviations.The uncertainties shown in Table I are due in part

e is intended to to random errors in measurement, in part to variationsfringes not their in source conditions during measurements, and in part

Me of both spacers to accidental misalignments in the interferometer. Inages to be shifted an effort to make the latter effect as random as possibleift in order num- the interferometer was deliberately readjusted tospacers it would slightly different lengths at frequent intervals duringt for the smaller the course of the measurements. This was done threeI to explain the times during the 12 exposures with the 110-mm spacer,

seven during the 22 exposures with the 218-mm spacer,,this tlme using and five times during the 20 exposures with the 21.5-mmhis cell transmits spacer. The air pressure inside the interferometer dur-ts of Hg

95 ing measurements was less than 5 u. The index of airat this pressure has a negligible effect on the calculatedwavelengths.

No corrections have been included for Doppler and7d Sc Stark shifts in the krypton line because of the un-

±418 23±9 certainty as to the proper value for these corrections.For the conditions under which our krypton source

±428 58±20 was operated, the results of Baird and Smith' predicta very small net shift-of the order of 1 A for the

±31 30±t25 2537-A line. The results of Bruce and Hill7 and ofEnglehard and Terrien' both predict a net shift ofabout +7 A for the 2537-A line.

0f 6 The value 2537.26874±40.00003 A may be compared±t25 2649 with the value 2537.26871±0.00003 A obtained by

Barger and Kessler for the wavelength of this line as±t42 50±20 emitted by a sealed-off atomic beam.9 The difference

between these two values is well within the combined±t51 33±45 standard deviations. It is not likely to represent a realcell difference in wavelength between the Zeeman filter

and the atomic beam because recent direct comparisonshe 218.2- and 110.4 of the wavelength of the Zeeman filter and the absorp-iange correction oh- tion line of an atomic beam indicate that if there is

any difference, it is probably smaller than 2.5,4A.i0

I is the spacer length in mm.b t is the number of observations.o (X1- is the average wavelength, corrected for dispersion of phase

change vith the pair of spacers indicated.

d a is the standard deviation [2 (Xi -(X).,)2/(1 -1)]i, in IA

o 6 is the correction for dispersion of phase change in pA.

I K. M. Baird and D. S. Smith, J. Opt. Soc. Am. 52, 507 (1962).I C. F. Bruce and R. M. Hill, Australian J. Phys. 14, 64 (1961).8 E. Englehar(l and J. Terrien, Rev. Opt. 39, 11 (1960).'R. L. Barger and K. G. Kessler, J. Opt. Soc. Am. 51, 827

(1961).1' W. G. Schweitzer, Jr., and K. G. Kessler (to be published).

99A

3 .4 .5PACER

'a218.221.5

110.421.5

218.2110.4

218.221.5

110.421.5

218.2110.4

.125i2 Sol. 53

-