1: %\documentclass[twocolumn,showpacs,aps]{revtex4}

2: \documentclass[preprint,showpacs,aps]{revtex4}

3: \usepackage{graphicx}

4: 

5: \begin{document}

6: 

7: \title{Absolute frequency measurements of the $D_2$ line

8: and fine-structure interval in $^{39}$K}

9: \author{Ayan Banerjee and Vasant Natarajan}

10: \email{vasant@physics.iisc.ernet.in}

11: \affiliation{Department of Physics, Indian Institute of

12: Science, Bangalore 560 012, INDIA}

13: 

14: \begin{abstract}

15: We report a value for the $D_2$-line frequency of $^{39}$K

16: with 0.25 ppb uncertainty. The frequency is measured using

17: an evacuated ring-cavity resonator whose length is

18: calibrated against a reference laser. The $D_2$ line

19: presents a problem in identifying the line center because

20: the closely-spaced energy levels of the excited state are

21: not resolved. We use computer modelling of the measured

22: spectrum to extract the line center and obtain a value of

23: 391 015 578.040(75) MHz. In conjunction with our previous

24: measurement of the $D_1$ line, we determine the

25: fine-structure interval in the $4P$ state to be 1 729

26: 997.132(90) MHz. The results represent significant

27: improvement over previous values.

28: \end{abstract}

29: \pacs{06.30.Ft, 42.62.Eh, 32.10.Fn}

30: 

31: %06.30.Ft Time and frequency,

32: %42.62.Eh Metrological applications,

33: %32.10.Fn Fine and hyperfine structure

34: 

35: \maketitle

36: 

37: \section{Introduction}

38: Precise measurements of atomic energy levels continue to

39: play an important role in the development of physics. The

40: energy levels of alkali atoms are particularly important

41: because these atoms can be laser cooled to ultra-low

42: temperatures for subsequent use in precision spectroscopy

43: experiments. For example, precise measurements of the $D_1$

44: line in Cs \cite{URH99}, Rb or K \cite{BDN03a}, in

45: conjunction with photon-recoil shift measurements in an

46: atom interferometer \cite{WYC93}, could yield an

47: independent value of the fine-structure constant $\alpha$.

48: In addition, a precise value of the frequency of the $D_2$

49: line in Cs \cite{URH00} is required for

50: atom-interferometric measurements of the local

51: gravitational acceleration \cite{PCC99}.

52: 

53: We have recently developed a technique for measuring the

54: absolute frequencies of optical transitions with sub-MHz

55: precision \cite{BDN03,BRD03}. The frequency is measured

56: using a ring-cavity resonator whose length is calibrated

57: against a reference laser locked to the $D_2$ line of

58: $^{87}$Rb. The frequency of the reference laser is known to

59: an accuracy of 10 kHz \cite{YSJ96}. We have already used

60: this technique to measure the $D_1$ lines in the alkali

61: atoms $^{39}$K, $^{85}$Rb, and $^{87}$Rb with 0.13 ppb

62: uncertainty \cite{BDN03a}, to facilitate measurements of

63: $\alpha$. In this paper, we apply this technique to measure

64: the $D_2$ line of $^{39}$K with an uncertainty of 0.25 ppb.

65: This represents an improvement of more than two orders of

66: magnitude over tabulated values \cite{NIST}. Furthermore,

67: in combination with our previous measurement of the $D_1$

68: line, we obtain the fine-structure interval in the $4P$

69: state of $^{39}$K with very high precision. Knowledge of

70: fine-structure intervals is useful in the study of atomic

71: collisions and relativistic calculations of atomic energy

72: levels. Indeed, a combination of improved methods of

73: calculation and increasingly precise measurements of atomic

74: fine and hyperfine structure (e.g.\ in Li \cite{WAC03}) can

75: lead to a precise value of $\alpha$ \cite{FHH80}.

76: 

77: The $D_2$ line in K presents a problem in determining the

78: line center because transitions to individual hyperfine

79: levels of the excited state are not resolved in

80: conventional saturated-absorption spectroscopy. This is

81: because the various hyperfine levels in the $4P_{3/2}$

82: state lie within 30 MHz of each other \cite{AIV77}, while

83: the natural linewidth is 6 MHz. We have recently

84: demonstrated a technique to resolve such closely-spaced

85: transitions \cite{BAN03}, however, in this work we use

86: computer simulation of the measured spectrum to extract the

87: line center with an accuracy of $\sim$100 kHz.

88: 

89: \section{Experimental details}

90: The experimental schematic is shown in Fig.\ \ref{f1} and

91: has been described extensively in a previous publication

92: \cite{BDN03a}. The four-mirror ring-cavity is placed inside

93: a vacuum chamber and evacuated to a pressure of

94: $\sim$10$^{-2}$ torr to eliminate frequency shifts due to

95: dispersion of air. Lasers 1 and 2 are standard

96: external-cavity diode lasers stabilized using optical

97: feedback from a piezo-mounted grating \cite{BRW01}. The

98: reference laser (Laser1) is locked to the $D_2$ line of

99: $^{87}$Rb using saturated-absorption spectroscopy in a

100: vapor cell, and the cavity is locked to the reference

101: laser. In general, the frequency of the laser to be

102: measured (Laser2) will be offset from the nearest cavity

103: resonance. This offset is compensated by using an

104: acousto-optic modulator (AOM) between Laser2 and the

105: cavity. The AOM is locked to this frequency difference and

106: its frequency is read using a counter. The exact mode

107: number of the cavity is determined by measuring the

108: cavity's free-spectral range using different lock points of

109: the reference laser, as described in Ref.\ \cite{BDN03}.

110: The mode number and the AOM offset together yield the

111: frequency of Laser2. A second AOM kept between the

112: reference laser and its saturated-absorption spectrometer

113: allows us to vary the reference frequency continuously so

114: that the AOM offset measured for Laser2 is always close to

115: a given value. This eliminates potential systematic errors

116: due to changes in the direction of the beam entering the

117: cavity.

118: 

119: The error signals needed for locking the diode lasers are

120: produced by modulating the injection current at a frequency

121: of 20--50 kHz. The error signal is obtained from the

122: saturated-absorption signal by phase-sensitive detection at

123: the third harmonic of the modulation frequency

124: \cite{WAL72}. This produces narrow dispersive signals that

125: are free from effects due to the underlying Doppler profile

126: or intensity fluctuations. The error signal for locking the

127: cavity is only a first-derivative signal since the cavity

128: modes appear against a flat background.

129: 

130: The spectroscopy on $^{39}$K is performed in an ultrahigh

131: vacuum glass cell maintained at a pressure below $10^{-8}$

132: torr by an ion pump. K vapor is produced by heating a

133: getter source \cite{SAES} with a current of 2.6~A. The

134: ultrahigh vacuum environment is necessary to minimize

135: linewidth broadening arising as a result of background

136: collisions. The getter source also gives us control over

137: the amount of K vapor in the cell, which we optimized to

138: obtain the narrowest linewidth.

139: 

140: Fig.\ \ref{f2} shows a typical saturated-absorption

141: spectrum of the $D_2$ line covering the two

142: ground-hyperfine levels. Each peak is actually a

143: convolution of six peaks. However, the individual hyperfine

144: transitions are not resolved because the different

145: hyperfine levels of the excited state lie within 30 MHz of

146: each other. To determine the line center, we therefore

147: performed computer simulations of the spectrum. We first

148: fixed the locations of the 6 peaks (with respect to the

149: line center) according to the known hyperfine shifts

150: \cite{AIV77}. We then set the peak amplitudes to correspond

151: to those obtained in saturated-absorption spectroscopy. We

152: further assumed that the linewidth of all the transitions

153: is the same. We calculated the spectrum for a given value

154: of the line center and linewidth.

155: 

156: Fig.\ \ref{f3} shows a close-up of the measured spectrum

157: (open circles) for $F=2 \rightarrow F'$ transitions. The

158: solid curve is a best fit to the spectrum, obtained with a

159: linewidth of 12.3 MHz. In order to check that this

160: linewidth is reasonable, we tuned the diode laser to the

161: $D_1$ line of K (at 770 nm), where the individual hyperfine

162: transitions are clearly resolved. The linewidth obtained

163: for those transitions was 14 MHz, close to the fit value

164: for the $D_2$ line. To check the linewidth further, we

165: changed the linewidth for the calculated spectrum and held

166: it constant during the fitting. The best-fit calculated

167: spectra for a linewidth of 15 MHz is also shown in Fig.\ 3.

168: Even with this small change, the lineshape deviates from

169: the measured data. The fit linewidth of 12.3 MHz, however,

170: is larger than the natural linewidth of 6 MHz. The primary

171: causes for this increase are power broadening due to the

172: pump beam and a small angle between the counterpropagating

173: pump and probe beams. The effect of stray magnetic fields

174: and background collisions is negligible.

175: 

176: With the best fit to the spectrum in Fig.\ 3, we found that

177: its maximum lies 2.60(10) MHz above the line center.

178: Similar modelling for transitions starting from the $F=1$

179: ground level shows that the maximum in the spectrum lies

180: 12.03(10) MHz below the line center.

181: 

182: \section{Error analysis}

183: The errors in our frequency measurement technique have been

184: discussed extensively in earlier publications

185: \cite{BDN03a,BDN03,BRD03}. We present here a brief overview

186: for the sake of completeness. There are two classes of

187: potential systematic errors that we consider. The first

188: class of errors comes from systematic shifts in the laser

189: frequencies. For the reference laser, changes in the

190: lineshape of the peaks due to optical pumping effects are

191: taken care of by carefully adjusting the pump and probe

192: beam intensities in the saturated-absorption spectrometer

193: (to a ratio of about 3). Shift in its lock point due to

194: peak pulling from neighboring transitions, the underlying

195: Doppler profile, or phase shifts in the feedback loop are

196: minimized by third-harmonic detection for the error signal.

197: Collisional shifts in the Rb vapor cell and the effect of

198: stray magnetic fields are negligible.

199: 

200: The second class of systematic errors is inherent to our

201: technique because we are really comparing the wavelength

202: (and not the frequency) of the two lasers. The most

203: important source of error is dispersion inside the cavity,

204: which is eliminated by using an evacuated cavity. However,

205: there could be wavelength-dependent phase shifts at the

206: dielectric coated mirrors used in the cavity. Such errors

207: can be corrected by repeating the measurement at different

208: cavity lengths. We have shown earlier \cite{BDN03a} that

209: this error is negligible when the unknown laser differs

210: from the reference laser by up to 25 nm. In the current

211: work, the wavelength difference was only 13 nm, therefore

212: the measurements were done at a single cavity length of

213: $\sim$178 mm.

214: 

215: \section{Results}

216: The results of our measurements are listed in Table I. Each

217: value is an average of 50--100 individual measurements, and

218: the quoted error includes our estimate of the systematic

219: error. To check for long-term variations, we repeated these

220: measurements over a period of several weeks. In addition,

221: we used two different lock points of the reference laser,

222: namely the $F=2 \rightarrow F'=(2,3)$ transition and the

223: $F=1 \rightarrow F'=(1,2)$ transition. The frequencies for

224: these transitions differ by 6622.886 MHz. The consistency

225: of our values for the two cases acts as a check on our

226: ability to determine the correct cavity mode number since

227: the cavity free-spectral range is only about 1.7 GHz.

228: Another check on our results is that the measured

229: frequencies for the two transitions should differ by the

230: ground hyperfine splitting in $^{39}$K. Our value of

231: 461.569(210) MHz overlaps very well with the accepted value

232: of 461.720 MHz \cite{AIV77}.

233: 

234: The values in Table I can be combined to yield the

235: hyperfine-free frequency of the $D_2$ line in K

236: \begin{center}

237: $4P_{3/2} - 4S_{1/2}$ : \ \ 391 015 578.040(75) MHz

238: \end{center}

239: This value is consistent with the value of 391 015 585(30)

240: MHz from the National Institute of Standards and Technology

241: energy-level tables \cite{NIST}, but the accuracy is

242: improved by more than two orders of magnitude. The above

243: result can be combined with our earlier measurement of the

244: $D_1$ line \cite{BDN03a} to yield the fine-structure

245: interval in the $D$ line in K

246: \begin{center}

247: $4P_{3/2} - 4P_{1/2}$ : \ \ 1 729 997.132(90) MHz

248: \end{center}

249: This value overlaps at the $1\sigma$ level with our earlier

250: measurement of the fine-structure interval using a

251: home-built wavemeter \cite{BDN04}: 1 729 993 (5) MHz, but

252: again the accuracy is improved considerably.

253: 

254: \section{Conclusion}

255: In conclusion, we have measured the $D_2$-line frequency

256: and fine-structure interval in $^{39}{\rm K}$ with

257: uncertainty less than 100 kHz. The measurement is

258: complicated by the fact that the closely-spaced hyperfine

259: transitions in the line are not resolved in conventional

260: saturated-absorption spectroscopy. We use computer

261: simulation of the measured spectrum to extract the line

262: center. The frequency is measured using a Rb-stabilized

263: ring-cavity resonator that has general applicability to

264: precise frequency measurement of optical transitions

265: \cite{BDN03}. The only other technique of measuring optical

266: frequencies with comparable precision is the frequency-comb

267: method \cite{URH99}. By comparison, our technique is

268: low-cost and easier to implement. In the future, we plan to

269: measure the $D$ lines and fine-structure interval of other

270: alkali atoms such as Li, Na, and Cs. These measurements

271: could play an important role in QED-independent

272: determinations of the fine-structure constant $\alpha$.

273: 

274: This work was supported by the Board of Research in Nuclear

275: Sciences (DAE), and the Department of Science and Technology,

276: Government of India.

277: 

278: \begin{thebibliography}{10}

279: 

280: \bibitem{URH99}

281: T.~Udem, J.~Reichert, R.~Holzwarth, and T.~W. H\"{a}nsch,

282:    Phys. Rev. Lett. \textbf{82}, 3568 (1999).

283: 

284: \bibitem{BDN03a}

285: A.~Banerjee, D.~Das, and V.~Natarajan,  Europhys. Lett.

286: \textbf{65}, 172 (2004).

287: 

288: \bibitem{WYC93}

289: D.~S. Weiss, B.~C. Young, and S.~Chu, Phys. Rev. Lett.

290: \textbf{70},  2706 (1993).

291: 

292: \bibitem{URH00}

293: T.~Udem, J.~Reichert, T.~W. H\"{a}nsch, and M.~Kourogi,

294: Phys. Rev. A

295:   \textbf{62}, 031801R (2000).

296: 

297: \bibitem{PCC99}

298: A.~Peters, K.~Y. Chung, and S.~Chu, Nature \textbf{400},

299: 849 (1999).

300: 

301: \bibitem{BDN03}

302: A.~Banerjee, D.~Das, and V.~Natarajan,  Opt. Lett. \textbf{28},

303: 1579 (2003).

304: 

305: \bibitem{BRD03}

306: A.~Banerjee, U.~D. Rapol, D.~Das, and V.~Natarajan,

307: Europhys. Lett. \textbf{63}, 340 (2003).

308: 

309: \bibitem{YSJ96}

310: J.~Ye, S.~Swartz, P.~Jungner, and J.~L. Hall, Opt. Lett.

311: \textbf{21}, 1280 (1996).

312: 

313: \bibitem{NIST}

314: The latest energy levels are available from the National

315: Institute of Standards \& Technology website -- \\

316: http://physics.nist.gov/PhysRefData

317: 

318: \bibitem{WAC03}

319: J.~Walls, R.~Ashby, J.~Clarke, B.~Lu, and W.~van Wijngaarden, Eur.

320: Phys. J. D \textbf{22}, 159 (2003).

321: 

322: \bibitem{FHH80}

323: W.~E.~Frieze, E.~A.~Hinds, V.~W. Hughes, and F.~M.~J.

324: Pichnick, Phys. Lett. A \textbf{78}, 322 (1980).

325: 

326: \bibitem{AIV77}

327: E.~Arimondo, M.~Inguscio, and P.~Violino, Rev. Mod. Phys.

328: \textbf{49}, 31 (1977).

329: 

330: \bibitem{BAN03}

331: A.~Banerjee and V.~Natarajan, Opt. Lett.  \textbf{28}, 729

332: (2003).

333: 

334: \bibitem{BRW01}

335: A.~Banerjee, U.~D. Rapol, A.~Wasan, and V.~Natarajan, Appl.

336: Phys. Lett. \textbf{79}, 2139 (2001).

337: 

338: \bibitem{WAL72}

339: A.~J. Wallard, J. Phys. E. \textbf{5}, 926 (1972).

340: 

341: \bibitem{SAES}

342: SAES~Getters, Viale Italia 77-1-20020, Lainate (MI), Italy.

343: 

344: \bibitem{BDN04}

345: A.~Banerjee, D.~Das, and V.~Natarajan,  J. Opt. Soc. Am. B

346: \textbf{21}, 79 (2004).

347: \end{thebibliography}

348: 

349: \newpage

350: 

351: \begin{table}

352: \caption{Measured frequencies for the center of different

353: transitions in the $D_2$ line of $^{39}$K. The frequencies

354: were measured with two lock points of the reference laser:

355: Ref.\ $f_1$ denotes the $F=2 \rightarrow F'=(2,3)$

356: transition and Ref.\ $f_2$ denotes the $F=1 \rightarrow

357: F'=(1,2)$ transition. \label{t1}}

358: \begin{tabular}{crr}

359: \hline\hline

360: \multicolumn{1}{c}{\parbox[t]{2.2cm}{Measured \\

361: transition}} &

362: \multicolumn{1}{c}{\parbox[t]{2.5cm}{Frequency (MHz) Ref.\

363: $f_1$}} & \multicolumn{1}{c}{\parbox[t]{2.5cm}{Frequency

364: (MHz) Ref.\ $f_2$}} \\

365: \hline $ F=2 \rightarrow F'$ & 391 015 404.938(150) & 391

366: 015 404.994(150) \\

367:  $ F=1 \rightarrow F'$ & 391 015 866.574(150) &

368: 391 015 866.495(150) \\

369: \hline\hline

370: \end{tabular}

371: \end{table}

372: 

373: \vspace*{6cm}

374: 

375: \newpage

376: 

377: \begin{figure}

378: \includegraphics[width=1\textwidth]{fig1.eps}

379: \caption{ Schematic of the experiment. PZT, piezoelectric

380: transducer; AOM, acousto-optic modulator; vco,

381: voltage-controlled oscillator; LIA, lock-in amplifier;

382: Sat.\ abs., saturated absorption; BS, beam splitter; PD,

383: photodiode; M, mirror.} \label{f1}

384: \end{figure}

385: 

386: \begin{figure}

387: \includegraphics[width=1\textwidth]{fig2.eps}

388: \caption{Typical saturated-absorption spectrum of the $D_2$

389: line in $^{39}$K showing all three sets of transitions. The

390: inverted peak in the center is a ground crossover

391: resonance.} \label{f2}

392: \end{figure}

393: 

394: \begin{figure}

395: \includegraphics[width=1\textwidth]{fig3.eps}

396: \caption{Measured and calculated spectra for $F=2

397: \rightarrow F'$ transitions. The measured spectrum, shown

398: as open circles, is a convolution of 6 peaks. The solid

399: curve is the best fit obtained with a linewidth of 12.3

400: MHz. The dotted curve is the calculated spectrum with a

401: linewidth of 15 MHz. Even with such a small change in

402: linewidth, the curve does not fit the measured spectrum

403: very well.} \label{f3}

404: \end{figure}

405: 

406: \end{document}

407: