INSTITUTE OF PHYSICS PUBLISHING JOURNAL OF PHYSICS B: ATOMIC, MOLECULAR AND OPTICAL PHYSICS J. Phys. B: At. Mol. Opt. Phys. 35 (2002) 2693–2701 PII: S0953-4075(02)33786-6 Hyperfine structure and isotope shift of transitions in Yb I using UV and deep-UV cw laser light and the angular distribution of fluorescence radiation R Zinkstok, E J van Duijn, S Witte and W Hogervorst Laser Centre Vrije Universiteit, De Boelelaan 1081, 1081 HV Amsterdam, The Netherlands E-mail: rzinkstk@nat.vu.nl Received 7 February 2002, in final form 2 May 2002 Published 6 June 2002 Online at stacks.iop.org/JPhysB/35/2693 Abstract Using the third harmonic of a cw titanium:sapphire laser, the hyperfine structure (HFS) and isotope shift (IS) of three deep-UV transitions of neutral Yb have been measured for the first time. By exploiting the angular distribution of fluorescence radiation, accurate and complete results are obtained for the HFS and IS of the 398.8 nm transition of Yb. From the measured data, normal and specific mass shift as well as field shift values for all transitions considered have been derived. 1. Introduction Ytterbium (Yb) is a lanthanide with atomic number 70, and has closed 4f and 6s subshells in the 4f14 6s2 1 S0 ground state. The natural isotopic composition shows five even isotopes, 168 Yb (0.14%), 170 Yb (3.03%), 172 Yb (21.82%), 174 Yb (31.84%) and 176 Yb (12.73%), and two odd isotopes, 171 Yb (14.31%) and 173 Yb (16.13%) with nuclear spin I = 1/2 and 5/2, respectively. Yb has been studied extensively in various contexts. Apart from a general analysis of the emission spectrum [1] and of highly excited levels [2], data have been collected on Rydberg and auto-ionizing states of Yb and Yb+ [3, 4], and on Stark shifts [5, 6]. Furthermore, Yb was studied in the context of laser isotope separation [7–10] and magneto-optical trapping [11, 12] experiments. Finally, much research has been done on the hyperfine structure (HFS) and isotope shift (IS) of even-parity Yb levels, mostly in the 40 000–50 000 cm−1 region using two-step excitation processes (e.g. [13–15]). The odd-parity energy levels around 38 000 cm−1 , however, have never been investigated in detail, probably due to the high photon energy needed for direct excitation. In this paper, we present data on the HFS and IS in three deep-UV transitions, which are—to our knowledge—studied for the first time. The measurements are performed using the third harmonic of a cw titanium:sapphire (Ti:S) laser, with a linewidth of ≈3 MHz. In addition to a study of these lines, we performed measurements on one of the most extensively studied lines of Yb, i.e. the 398.8 nm first resonance line, corresponding to the 0953-4075/02/122693+09$30.00 © 2002 IOP Publishing Ltd Printed in the UK 2693 2694 R Zinkstok et al transition from the 4f14 6s2 1 S0 ground state to the 4f14 6s6p 1 P1 state. This transition is used in laser cooling experiments [11,12]. Due to several overlapping peaks in the spectrum, resolving the HFS and IS in this transition has proven to be a challenge. Increasingly complex methods have been applied to this issue: interferometric experiments [16], level crossing and anti-level crossing spectroscopy [17–19], high-resolution laser spectroscopy [20, 21], laser sideband techniques combined with photon burst spectroscopy [22] and cooled-atom spectroscopy [23], giving increasingly accurate results. In this paper, we present a simple alternative to obtain accurate results. Using the second harmonic of a Ti:S laser with a linewidth of ≈2 MHz and exploiting the angular distribution of the fluorescence radiation emitted, we have in an elegant way determined the HFS and IS of this transition with an accuracy exceeding that of all previous experiments. 2. Angular distribution of fluorescence radiation It is well known that the dipole radiation emitted by an atom can be non-isotropic. The angular distribution F (θ, φ) depends on the value of m of the transition considered: for m = 0, F (θ, φ) = sin2 (θ ), while for m = ±1, F (θ, φ) = 1 + cos2 (θ ) (see e.g. [24]). When more magnetic sublevels are involved, the contribution of each mi → mf transition is calculated by multiplying the appropriate angular distribution with the relevant (angular) part of the dipole matrix element of that transition. The total angular distribution of the fluorescence radiation is then obtained by summing over all contributions, taking into account the polarization (if present) of the incident light. These distributions of fluorescence radiation have been calculated for all components of the 398.8 nm transition in Yb assuming excitation with linearly polarized light. Since this is a J = 0 → 1 transition, it follows that the even isotopes (which have I = 0) can only decay through m = 0 → 0, so these isotopes show the same anisotropic distribution. The distribution for the odd isotopes, which have a non-zero nuclear spin, is much more isotropic because more decay options are available, including both m = 0 and ±1. The distributions for all isotopes are shown in figure 1. In the figure, the vertical z-axis is parallel to the polarization vector of the incident light. Clearly, the spectrum that will be recorded depends highly on the position of the detector. When the detector observes along the horizontal x-axis, the even isotopes will dominate the signal, while they will be invisible when the detector is placed along the vertical z-axis. Alternatively, the polarization of the incident light can be rotated over 90◦ , as done in the present work. The two different spectra can be subtracted, resulting in a difference spectrum featuring only the odd isotopes. This provides an easy and reliable means of fully resolving the HFS in the 398.8 nm transition of Yb. It is to be noted that this discussion applies only to closed atomic transitions, of which the 398.8 nm transition is an example. When other decay channels are available, more magnetic transitions become possible, which may give rise to a more isotropic angular distribution of the fluorescence radiation. Therefore, it is not expected that this method is applicable to the other transitions presented in this paper. 3. Experimental method and set-up 3.1. Laser system A schematic of the laser system is given in figure 2. The infra-red (IR) light from a cw Ti:S laser (Coherent 899-21) pumped by a 10 W Spectra Physics Millennia X, is divided into two beams by a 50% beamsplitter (BS). The first IR beam is frequency doubled in an external enhancement cavity (EEC) using a LBO nonlinear crystal, Brewster cut for the fundamental wavelength (θ = 90◦ and φ = 33.7◦ ). With 0.8 W of fundamental light, about 320 mW of second harmonic (UV) light is produced, reaching 40% of useful conversion. HFS and IS of several transitions in Yb I 2695 a b 1 0.3 -1 -1 1 1 - 0.3 -1 c d 0.4 0.4 -1 -1 1 1 - 0.4 - 0.4 Figure 1. Calculated angular distributions of fluorescence radiation in the 4f14 6s2 1 S0 –4f14 6s6p 1 P1 transition for (a) all even isotopes, (b) 171 YbF =1/2→1/2 , 173 YbF =5/2→3/2 and 173 YbF =5/2→5/2 , (c) 171 YbF =1/2→3/2 , (d) 173 YbF =5/2→7/2 . The vertical z-axis is parallel to the polarization vector of the incident light. PD PD Servo PB M λ/4 M4 L 25 cm BBO M5 HR deep-UV HRIR HRUV M8 M7 M ML 40 cm HRUV HRIR HRIR LBO M1 M2 BS λ/2 L 25 cm CL 15 cm CL 4 cm HRIR PB M HRUV M4 M3 λ/2 ML 40 cm HRUV λ/4 M PMT PD Servo HR deep-UV AB PD Vacuum chambers BS Ti:S laser HRIR Millennia X Etalon M D oven Computer Figure 2. Schematic of the experimental set-up. The beam of a cw Ti:S laser pumped by a 10 W Millennia X is split into two beams by a 50% BS. One of these beams is frequency doubled inside an EEC consisting of mirrors M1–M4 using an LBO crystal, while the other is enhanced in a second EEC (mirrors M5–M8). The beam reflected off M3 is led into a Hänsch–Couillaud locking device (M mirror, PB polarizing BS, PD photodiode). The second harmonic (UV) beam is imaged through several lenses for beam-shaping (L lens, CL cylindrical lens) and coupled into the second EEC, where the third harmonic is generated by sum-frequency mixing inside a BBO crystal. For both EECs, mode matching is performed by a thin lens (ML). The resulting deep-UV light is collimated by a lens and passes through the vacuum system, where it intersects an AB of Yb. Directly above the point where the beams intersect, a PMT is placed. A small fraction of the fundamental Ti:S beam is led through an etalon for frequency calibration. The etalon signal is measured with a detector (D) and fed to a computer, together with the PMT signal. 2696 R Zinkstok et al The second IR beam is also enhanced in an EEC, which contains a BBO crystal. The UV beam generated in the first EEC is led single pass through this second enhancement cavity, where it overlaps with the IR beam. Inside the BBO crystal, sum frequency mixing is then realized, producing about 5 mW of third harmonic (deep-UV) light. Type I phase matching is realized by rotating the polarization of the UV beam over 90◦ with a λ/2 plate. To keep the cavities in resonance with the laser light, the Hänsch–Couillaud locking technique [25] is used. In principle, the wavelength range of a Ti:S laser is about 690–1100 nm. In our case, phase matching and mirror coating restrictions on the Ti:S laser and EECs yield in practice tunability in the ranges 690–835, 370–415 and 255–278 nm for the fundamental, second harmonic and third harmonic, respectively. Since the linewidth of the Ti:S laser is about 1 MHz, bandwidths of the second and third harmonics are about 2 and 3 MHz respectively. The laser is continuously tunable over about 10 GHz in the deep-UV. Frequency scans are calibrated by passing a small fraction of the fundamental beam through an etalon with a free spectral range of 150±0.3 MHz, while another part of the fundamental beam is led to an ATOS LM-007 lambda-meter as an absolute wavelength reference. 3.2. Vacuum system and atomic beam The atomic beam (AB) is produced by heating a sample of Yb inside a small tantalum oven. This oven is heated to a temperature of about 950 K by electron bombardment using a nearby tungsten wire heated by a large current (≈10 A at ≈17 V). At this temperature the vapour pressure of Yb is high enough to produce an AB of sufficient intensity, emanating from the oven through a small hole. About 30 cm downstream, the atoms pass through a diaphragm 3 mm in diameter, after which a highly collimated beam remains. This ensures that the Doppler broadening of the measured spectral lines is limited to about 10 MHz for the second harmonic measurements, and to about 16 MHz for the third harmonic. The vacuum system consists of two compartments connected by a valve (figure 2). The first compartment contains the oven where the AB is produced, while in the second compartment the LIF measurements are performed. The laser beam passes through this second compartment, where it intersects the AB under an angle of 90◦ to minimize Doppler broadening and shift. Directly above the point where the beams intersect a photomultiplier tube (PMT) is mounted. The fluorescence spot is imaged on the PMT with a lens system. In order to minimize detection of stray light and radiation from the oven, spatial filtering in combination with filters for the wavelength to be detected are employed. Also, the windows of the vacuum compartment are anti-reflection coated for the laser wavelength to minimize stray reflections. The PMT signal is then fed through a discriminator to a counter, which is connected to a computer for analysis. By monitoring both etalon output and photomultiplier signal while scanning the laser, the frequency difference between adjacent spectral components can be determined with a total error smaller than 0.9 MHz, while the absolute wavelength can be measured by the ATOS lambda-meter with a systematic error smaller than 100 MHz. 4. Results and discussion 4.1. Deep-UV transitions Using the set-up described in the previous section, the spectra of several deep-UV transitions in Yb from the 4f14 6s2 1 S0 ground state have been measured. The transitions involved are to the 4f13 5d6s2 1 P1 state at 37 414.59 cm−1 (λ = 267.28 nm), the 4f14 1 S 6s7p 3 P1 state at 38 174.17 cm−1 (λ = 261.96 nm) and the 4f13 5d6s2 3 D1 state at 38 422.36 cm−1 (λ = 260.27 nm). Spectra of these transitions are shown in figure 3; they all show a good signal to noise ratio, on the order of 100–1000. HFS and IS of several transitions in Yb I 2697 174 6000 D1 171 − 1/2 173 − 7/2 176 171 − 3/2 173 − 3/2 Counts 4000 2000 3 172 (a) 0 0 1000 500 1500 174 4000 (b) 3 171 − 3/2 173 − 3/2 170 171 − 1/2 173 − 7/2 2000 173 − 5/2 Counts 172 P1 0 2000 4000 6000 174 0 1 (c) P1 173 − 3/2 176 171 − 3/2 173 − 5/2 173 − 7/2 x100 170 5000 168 171 − 1/2 Counts 172 10000 8000 0 0 2000 4000 6000 Frequency [MHz] Figure 3. Measured spectra of deep-UV transitions in Yb. Peaks are identified by mass number and F -value of the upper state. Shown are spectra of the transitions from the ground state to (a) the 4f13 5d6s2 3 D1 state at 38 422.36 cm−1 , (b) the 4f14 1 S 6s7p 3 P1 state at 38 174.17 cm−1 and (c) the 4f13 5d6s2 1 P1 state at 37 414.59 cm−1 . All isotopes can be identified in these spectra, except for 168 Yb, which is visible only in the spectrum of the 5d6s2 1 P1 state, and for 170 Yb, which is buried under the 172 Yb peak in the 6s2 3 D1 state spectrum. The ISs are shown in table 1. The HFS of these transitions is solved completely: table 2 shows the A and B constants deduced from the measurements (it is to be noted that isotopes with I 1 do not have a B constant). Also shown are the ratios A173 /A171 . These values are nearly equal for all transitions, in accordance with expectations. They agree well with values for this ratio given by other authors, e.g. [13, 15]. 4.2. The 398.8 nm line We used a polarizing beam splitter and a half-wave plate to switch between horizontal and vertical linearly polarized light while the PMT remained fixed in place, observing along the vertical axis (thus defining ‘vertical’ and ‘horizontal’ directions). In this way, pairs of spectra of the 398.8 nm line have been measured, each pair consisting of a horizontal polarization spectrum and a vertical polarization spectrum. Because this transition is exceptionally strong, the PMT current could be detected directly, resulting in a signal-to-noise ratio exceeding 1000. 168 171 − 1/2 170 168 171 − 1/2 3000 174 171 − 3/2 173 − 3/2 173 − 5/2 (b) 0.5 2000 173 − 7/2 1000 x100 176 0 173 − 5/2 1.0 (c) 2000 0.5 3000 171 − 1/2 1000 171 − 3/2 0 173 − 3/2 173 − 7/2 Intensity (a.u.) x100 0 0 Intensity (a.u.) 171 − 3/2 173 − 5/2 176 (a) 0.5 173 − 7/2 174 1.0 172 R Zinkstok et al Intensity (a.u.) 2698 0 0 1000 2000 Frequency (MHz) Figure 4. Measured spectra of the 4f14 6s2 1 S0 –4f14 6s6p 1 P1 transition in Yb. Peaks are identified as in figure 3. (a) Using horizontal polarization, (b) using vertical polarization and (c) the difference spectrum showing only the odd isotopes. Table 1. Measured ISs. IS (MHz) Final state 4f13 5d6s2 1 P Energy level 4f14 6s7p 3 P1 4f13 5d6s2 3 D1 37 414.59 38 174.17 38 422.36 4f14 6s6p 1 P1 4f14 6s6p 3 P1 a 25 068.222 17 992.007 a 1 (cm−1 ) 170–168 908(2) 172–170 842.6(0.6) 80.5(0.9) −696.0(1.8) −659(4) −1368.6(5) −1286.5(5) 174–172 176–174 601.2(1.6) 567(3) 32.3(1.8) 27.5(0.9) −35.5(1.3) −36.2(1.6) −529(2) −1000.3(5) −507.2(1.5) −954.8(5) Data taken from [26]. Examples of both spectra are shown in figure 4. It is to be noted that these spectra are in fair agreement with the angular distribution plots of figure 1. However, there are still even isotopes visible in the vertical polarization spectra, which is due to the finite opening angle of the detector. Therefore, the spectra are subtracted pairwise after normalization, resulting in a difference spectrum showing only the odd isotopes, also shown in figure 4. In this difference spectrum, the HFS is fully resolved. Accurate values for the A and B constants now can be derived directly. These values are included in table 2. From the horizontal polarization spectrum, the IS can be deduced for the even isotopes, while the IS of the odd isotopes is deduced from the difference spectra. All IS data are collected in table 1. The IS data presented here for the 398.8 nm transition are in good agreement with those measured in previous work [16, 20, 22, 23], while the present values have a higher accuracy. For comparison, all values are displayed together in table 3. Our values for the HFS A and B constants also agree well with those reported by others: for comparison, these are given in table 4. Exceptions are those reported in [17], where the sign is not given for A171 , and in [22], HFS and IS of several transitions in Yb I 2699 Table 2. HFS A and B constants of the excited states in MHz. State A171 4f13 5d6s2 1 P1 4f14 6s7p 3 P1 4f13 5d6s2 3 D1 2093(4) 4143(4) −588(2) 4f14 6s6p 1 P1 −210.2(0.8) A173 B173 A173 /A171 −577.6(0.5) 233(2) −0.2760(5) −1147.1(1.4) −12.1(1.1) −0.2769(4) 161.9(1.5) −177(6) −0.275(3) 57.7(0.9) 602.1(1.1) −0.2746(8) Table 3. Comparison of IS data of the 398.8 nm transition of Yb. IS (MHz) 170–168 172–170 Present work −696.0(1.8) −659(4) [16] [20] [22] [23] — — −698(2) — −627(30) −665(10) — −648(8) 174–172 176–174 −529(2) −507.2(1.5) −531(30) −530(4) — −528(3) −470(27) −509(4) — −507(3) Table 4. Comparison of HFS data of the 398.8 nm transition of Yb. A171 A173 B173 Present work −210.2(0.8) 57.7(0.9) 602.1(1.1) [17] [18] [20] [19] [21] [22] [23] 59.7(1.3) 58.5(0.8) 603(17) 590(13) 605(20) 588(2) 216(4) −213(3) −211(1) −213(10) −201(3) −212(3) 58.1(0.3) where a fairly low A-value for 171 Yb is given, while the value for B173 reported by [19] seems to be too low when compared with the others. The overall accuracy of our measurements is higher than that obtained in all other experiments, while we applied a simple and elegant method. 4.3. Isotope shift analysis The IS in a transition between two isotopes with mass numbers A and A can be decomposed into several contributions. These are the normal mass shift (NMS), the specific mass shift (SMS) and the field shift (FS): A −A δνiIS = δνiFS + (MiNMS + MiSMS ). (1) AA The M NMS and M SMS are constants for a given transition. The NMS can easily be calculated using νi MiNMS = (2) 1822.8 where νi is the transition frequency. Consequently, the NMS can be eliminated from equation (1), yielding the reduced isotope shift (RIS). When the RIS of a transition is modified by multiplying it with AA /(A − A), and then plotted against the modified IS of another R Zinkstok et al Modified IS of other transitions [MHz] 2700 1000 500 0 -500 -1400 -1600 -1200 14 -1000 3 Modified IS of the 4f 6s 6p P1 transition [MHz] Figure 5. King lines of the measured transitions from the ground state to: , 4f14 6s7p 3 P1 ; ♦, 4f13 5d6s2 3 D1 and , 4f14 6s6p 1 P1 . ◦, 4f13 5d6s2 1 P1 ; Table 5. Contributions to the IS between 176 Yb and 174 Yb in the measured transitions in MHz. Upper state 4f13 5d6s2 1 P δνNMS 1 4f14 6s7p 3 P1 4f13 5d6s2 3 D1 4f14 6s6p 1 P1 40.2 41.0 41.3 26.9 δνSMS δνFS −307(13) 854(13) −181(10) 168(10) −184(10) 107(10) −103(10) −431(10) transition, a straight line results (King plot). The slope and intercept of such a line relate the FS and SMS of the transitions involved (see [27]). Using the measured IS data a King plot analysis has been made. As reference transition we took the 555.6 nm 6s2 1 S0 → 6s6p 3 P1 transition in Yb, for which [26] gives very accurate data, shown in table 1. Since this is known to be a pure s2 →sp transition [14,27], we can adopt the estimate from [28], δν SMS = (0 ± 0.5)δν NMS , and assume the SMS of this transition to be about zero (the NMS between 174 Yb and 176 Yb is calculated to be 20 MHz using equation (2)). The King lines for the four measured transitions are shown in figure 5. From these King lines, the values for the SMS and the FS have been derived, using the values for the known reference transition at 555.6 nm. These are shown in table 5 for the 174–176 isotope pair. The transition to the 4f13 5d6s2 1 P1 state shows a large negative SMS. This is due to the change in the number of 4f electrons; the momentum of the 4f electrons is strongly coupled to that of the inner-shell d electrons (see e.g. [27]). The positive FS is believed to be a screening effect. The 5d electron in the excited state does not screen the two s electrons as well as the 4f electron in the ground state. This leads to a higher electron density at the nucleus for the excited state, which results in a positive FS (see e.g. [28]). The transition to the 4f14 6s6p 1 P1 state at 398.8 nm shows a negative FS, caused by the decrease of electron density in the s → p transition. The negative SMS in this transition HFS and IS of several transitions in Yb I 2701 is probably due to mixing with a 4f13 5d6s2 state, which exhibits a negative SMS as in the 4f13 5d6s2 1 P1 state [27]. The 4f14 6s7p 3 P1 and 4f13 5d6s2 3 D1 states are mutually perturbing according to [1], which is confirmed by their very similar King lines. The negative SMS in the transitions to these states can be attributed to the change in the number of 4f electrons in the 4f14 6s2 1 S0 → 4f13 5d6s2 3 D1 transition, as before. The positive FS is unexpected for an s2 →sp transition, but it can be due to a screening effect in the 4f14 6s2 1 S0 → 4f13 5d6s2 3 D1 transition, similar to the screening effect discussed above. The SMS and FS values for these two transitions seem to indicate that the 4f13 5d6s2 configuration dominates the IS behaviour of this perturbing pair. 5. Conclusions HFS and IS data have been measured for the first time for three deep-UV transitions in neutral Yb using the third harmonic of a cw Ti:S laser. In addition, accurate measurements have been performed on the 398.8 nm line in neutral Yb. Although a simple technique has been used, exploiting the angular distribution of the fluorescence radiation, the accuracy of our measurements exceeds that of previous experiments, where far more complicated techniques were employed. From a King plot analysis of the data, values for the normal and SMS as well as the FS have been obtained for all measured transitions. References [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] [24] [25] [26] [27] [28] Wyart J F and Camus P 1979 Phys. Scr. 20 43 Camus P, Débarre A and Morillon C 1980 J. Phys. B: At. Mol. Phys. 13 1073 Komarovskii V A and Verolainen Y F 1992 Opt. Spektrosk. 73 855 Xu C B, Xu X Y, Huang W, Xue M and Chen D Y 1994 J. Phys. B: At. Mol. Opt. Phys. 27 3905 Dai C J, Zhang S, Shu X W, Fang D W and Li J 1995 J. Quant. Spectrosc. Radiat. Transfer 53 179 Li J and van Wijngaarden W A 1995 J. Phys. B: At. Mol. Opt. Phys. 28 2559 Park H, Lee J, Lee J and Chang J 1996 Phys. Rev. A 53 1751 Derzhiev V I, Kuznetsov V A, Mikhal’tsov L A, Mushta V M, Sapozhkov A Y, Tkachev A N, Chaushanskii S A and Yakovlenko S I 1996 Quantum Electron. 26 751 Borisov S K, Kuz’mina M A and Mishin V A 1996 J. Russ. Laser Res. 17 332 Sankari M and Suryanarayana M V 1998 J. Phys. B: At. Mol. Opt. Phys. 31 261 Honda K, Takahashi Y, Kuwamoto T, Fujimoto M, Toyoda K, Ishikawa K and Yabuzaki T 1999 Phys. Rev. A 59 R934 Kuwamoto T, Honda K, Takahashi Y and Yabuzaki T 1999 Phys. Rev. A 60 R745 Maier J, Kischkel C S and Baumann M 1991 Z. Phys. D 21 145 Jin W, Horiguchi T, Wakasugi M, Hasegawa T and Yang W 1991 J. Phys. Soc. Japan 60 2896 Kischkel C S, Baumann M and Kümmel E 1991 J. Phys. B: At. Mol. Opt. Phys. 24 4845 Kischkel C S, Baumann M and Kümmel E 1992 J. Phys. B: At. Mol. Opt. Phys. 25 4447 Chaiko Y 1965 Opt. Spectrosc. 18 200 Chaiko Y 1966 Opt. Spectrosc. 20 424 Budick B and Snir J 1969 Phys. Rev. 178 19 Baumann M, Liening H and Lindel H 1977 Phys. Lett. A 59 433 Liening H 1985 Z. Phys. A 320 363 Grundevik P, Gustavsson M, Rosén A and Rydberg S 1979 Z. Phys. A 292 307 Berends R W and Maleki L 1992 J. Opt. Soc. Am. B 9 332 Deilamian K, Gillaspy J D and Kelleher D E 1993 J. Opt. Soc. Am. B 10 789 Loftus T, Bochinski J R and Mossberg T W 2000 Phys. Rev. A 63 023402-1 Davydov A S 1965 Quantum Mechanics (Oxford: Pergamon) p 309 Hänsch T W and Couillaud B 1980 Opt. Commun. 35 441 Clark D L, Cage M E, Lewis D A and Greenlees G W 1979 Phys. Rev. A 20 239 King W H 1984 Isotope Shifts in Atomic Spectra (New York: Plenum) pp 63, 142 Heilig K and Steudel A 1974 At. Data Nucl. Data Tables 14 613