1: % PI.tex

2: % Photo-ionization paper

3: % Started DML 30.5.03

4: 

5: \documentclass[twocolumn,showpacs,amsmath,amssymb,superscriptaddress]{revtex4}

6: %\documentclass[preprint,showpacs,amsmath,amssymb,superscriptaddress]{revtex4}

7: \usepackage{graphicx}

8: 

9: % Lengths for figures

10: \newlength{\onecolfig}

11: \setlength{\onecolfig}{86mm} % single-column PRA width

12: \newlength{\twocolfig}

13: %\setlength{\twocolfig}{145mm} % (reduced) double-column PRA width

14: \setlength{\twocolfig}{178mm} % double-column PRA width

15: 

16: % Calcium definitions:

17: \newcommand{\ion}[2]{\mbox{$^{#2}$#1$^+$}}

18: \newcommand{\CaI}[1]{\mbox{$^{#1}$Ca}}

19: \newcommand{\CaII}[1]{\ion{Ca}{#1}}

20: \newcommand{\lev}[3]{\mbox{$^{#1}$#2$_{\mbox{\tiny$#3$}}$}}

21: \newcommand{\hfslev}[3]{\mbox{#1$^{\mbox{\tiny$#3$}}_{\mbox{\tiny$#2$}}$}}

22: 

23: % Unit definitions:

24: \newcommand{\unit}[1]{\,\mbox{#1}}

25: \newcommand{\kHz}{\unit{kHz}}

26: \newcommand{\MHz}{\unit{MHz}}

27: \newcommand{\GHz}{\unit{GHz}}

28: \newcommand{\THz}{\unit{THz}}

29: \newcommand{\torr}{\unit{torr}}

30: \newcommand{\mW}{\unit{mW}}

31: \newcommand{\uW}{\unit{$\mu$W}}

32: \newcommand{\mrad}{\unit{mrad}}

33: \newcommand{\mm}{\unit{mm}}

34: \newcommand{\um}{\unit{$\mu$m}}

35: \newcommand{\nm}{\unit{nm}}

36: \newcommand{\K}{\unit{K}}

37: \newcommand{\ms}{\unit{ms}}

38: \newcommand{\us}{\unit{$\mu$s}}

39: \newcommand{\ns}{\unit{ns}}

40: \newcommand{\uA}{\unit{$\mu$A}}

41: \newcommand{\degree}{\mbox{$^{\circ}$}}

42: \newcommand{\degC}{\mbox{\degree{}C}}

43: \newcommand{\G}{\unit{G}}

44: \newcommand{\s}{\unit{s}}

45: 

46: % Other definitions:

47: \newcommand{\citesec}[2]{\cite[\S{}#2]{#1}}   % {} to eliminate gap?

48: \newcommand{\etal}{{\em et al.}}

49: \newcommand{\eg}{{\em e.g.}}

50: \newcommand{\ie}{{\em i.e.}}

51: \newcommand{\viz}{{\em viz.}}

52: \newcommand{\ish}{\mbox{$\sim$}\,}

53: \newcommand{\ltish}{\protect\raisebox{-0.4ex}{$\,\stackrel{<}{\scriptstyle\sim}\,$}}

54: \newcommand{\gtish}{\protect\raisebox{-0.4ex}{$\,\stackrel{>}{\scriptstyle\sim}\,$}}

55: \newcommand{\lr}{\mbox{$\leftrightarrow$}}

56: \newcommand{\bra}[1]{\mbox{$\left< #1 \right|$}}

57: \newcommand{\ket}[1]{\mbox{$\left| #1 \right>$}}

58: \newcommand{\blob}{\mbox{$\bullet$}}

59: \newcommand{\file}[1]{{\small\tt #1}}

60: \newcommand{\wee}[2]{\mbox{$\frac{#1}{#2}$}}

61: \newcommand{\sub}[1]{\mbox{$_{\mbox{\tiny #1}}$}}

62: \newcommand{\diff}[1]{\mbox{\/d$#1$}}

63: 

64: 

65: % English translation:

66: \newenvironment{centre}{\begin{center}}{\end{center}}

67: \newcommand{\centreline}[1]{\centerline{#1}}

68: 

69: \begin{document}

70: \bibliographystyle{apsrev}

71: 

72: \title{Isotope-selective photo-ionization for calcium ion trapping}

73: 

74: \author{D. M. Lucas}

75: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

76: \author{A. Ramos}

77: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

78: \author{J. P. Home}

79: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

80: \author{M. J. McDonnell}

81: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

82: \author{S. Nakayama}

83: \affiliation{Department of Information and Computer Science, Kagoshima University, 1-21-40 Korimoto, Kagoshima 890-0065, JAPAN}

84: \author{J.-P. Stacey}

85: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

86: \author{S. C. Webster}

87: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

88: \author{D. N. Stacey}

89: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

90: \author{A. M. Steane}

91: \affiliation{Department of Physics, University of Oxford, Clarendon Laboratory, Parks Road, Oxford OX1 3PU, U.K.}

92: 

93: \date{\today}

94: 

95: \begin{abstract}

96: We present studies of resonance-enhanced photo-ionization for isotope-selective loading of \CaII{} into a Paul trap.

97: The 4s$^2\,\lev{1}{S}{0}\leftrightarrow$ 4s4p\,\lev{1}{P}{1} transition of neutral calcium is driven by a 423\nm\ laser

98: and the atoms are photo-ionized by a second laser at 389\nm. Isotope-selectivity is achieved by using crossed atomic

99: and laser beams to reduce the Doppler width significantly below the isotope shifts in the 423\nm\ transition. The

100: loading rate of ions into the trap is studied under a range of experimental parameters for the abundant isotope

101: \CaII{40}. Using the fluorescence of the atomic beam at 423\nm\ as a measure of the Ca number density, we estimate a

102: lower limit for the absolute photo-ionization cross-section. We achieve loading and laser-cooling of all the naturally

103: occurring isotopes, without the need for enriched sources. Laser-heating/cooling is observed to enhance the

104: isotope-selectivity. In the case of the rare species \CaII{43} and \CaII{46}, which have not previously been

105: laser-cooled, the loading is not fully isotope-selective but we show that pure crystals of \CaII{43} may nevertheless

106: be obtained. We find that for loading \CaII{40} the 389\nm\ laser may be replaced by an incoherent source.

107: \end{abstract}

108: 

109: \pacs{32.80.Fb, 32.80.Rv, 32.80.Pj}

110: 

111: \maketitle

112: 

113: \section{Introduction}

114: 

115: Resonance-enhanced photo-ionization for loading ion traps was first demonstrated with magnesium and calcium ions by

116: Kj\ae{}rgaard \etal~\cite{00:Kjaergaard}. The same group has recently used crossed-beam Doppler-free excitation of the

117: 272\nm\ 4s$^2\,\lev{1}{S}{0}\leftrightarrow$ 4s5p\,\lev{1}{P}{1} calcium transition, followed by photo-ionization, to

118: load selectively all the naturally occurring isotopes of this element; the loading was not observed directly, but by

119: using charge-exchange to infer the presence of the other isotopes from the fluorescence of laser-cooled \CaII{40} ions

120: which replaced them~\cite{02:Mortensen}. Gulde \etal\ showed that, using the 4s$^2\,\lev{1}{S}{0}\leftrightarrow$

121: 4s4p\,\lev{1}{P}{1} Ca transition at 423\nm, followed by excitation close to the continuum by an ultraviolet photon,

122: photo-ionization is around five orders of magnitude more efficient than conventional electron bombardment

123: ionization~\cite{01:Gulde}.

124: 

125: As pointed out by these authors, photo-ionization has a number of advantages over electron bombardment. Only the

126: desired species is loaded into the ion trap, allowing pure crystals of particular isotopes to be obtained. Since no

127: electron beam is involved, there is no charging of insulating parts of the trap structure, which leads to drifting

128: electric fields that for many experiments must be accurately compensated. The efficiency of the photo-ionization

129: process allows much lower number densities of neutral atoms in the interaction region, greatly reducing the quantities

130: of material sputtered onto the trap electrodes; it has been shown that clean electrode surfaces reduce the heating rate

131: of trapped ions from the motional ground state~\cite{00:Turchette}, an important consideration for quantum logic

132: experiments in ion traps. For experiments in which high-finesse optical cavities are combined with ion traps, the lower

133: atomic beam density reduces the degradation of the mirror surfaces and can avoid the need for a separate ``loading

134: trap''~\cite{03:Mundt,03:Keller}.

135: 

136: Our particular interest lies in using \CaII{43} as an ion-qubit in quantum logic experiments~\cite{97:Steane}. This ion

137: has a number of advantages as a qubit~\cite{03:Lucas,03:SchmidtKaler}; it is also an attractive candidate for a trapped

138: ion optical frequency standard~\cite{93:Plumelle,00:Boshier}. The most obvious difficulty in working with \CaII{43} is

139: its low natural abundance of 0.135\%. In this paper we describe the development of a photo-ionization system capable of

140: loading \CaII{43} into a Paul trap from a natural abundance source. (Isotopically-enriched sources are available,

141: although the maximum enrichment of \ish 80\%~\cite{PC:OakRidge} means that it would still be advantageous to use an

142: isotope-selective method for reliable loading of this isotope.)

143: 

144: We describe the crossed atomic and laser beam setup for excitation of the 423\nm\ 4s$^2\,\lev{1}{S}{0}\leftrightarrow$

145: 4s4p\,\lev{1}{P}{1} transition with a Doppler width narrow compared with the isotope shifts. Spectroscopy of the atomic

146: beam is used to ensure the intensity of the 423\nm\ laser is below saturation (to avoid reducing the

147: isotope-selectivity of this step by power broadening), and also to estimate the number density of atoms in the beam.

148: Photo-ionization from the 4s4p\,\lev{1}{P}{1} level is achieved by a 389\nm\ photon; since we expect the

149: photo-ionization cross-section to be only weakly dependent on the photon energy, we use a non-stabilized diode laser

150: for this step. We study the photo-ionization trap loading rate as a function of the power and detuning of the 423\nm\

151: laser, and as a function of the power of the 389\nm\ laser. We find that the 389\nm\ laser photo-ionizes equivalently

152: above or below lasing threshold, and that it can be replaced by a much cheaper incoherent source if high

153: photo-ionization rates are not required. We estimate the absolute photo-ionization loading efficiency and compare it

154: with that for electron bombardment ionization in our apparatus.

155: 

156: The sensitivity of the loading rate as a function of the 423\nm\ detuning determines the isotope-selectivity which can

157: be attained. However, for very rare species with mass number $X$ a practical limitation is the rate at which

158: charge-exchange replaces the desired, trapped, ions \CaII{X} by the abundant \CaII{40} ions issuing from the oven. We

159: demonstrate loading of all the naturally occurring isotopes, including \CaII{43} and the 0.004\%-abundant \CaII{46}.

160: These two species are laser-cooled for the first time in the experiments reported here. The limited isotope-selectivity

161: attainable for these rare ions means that some form of ``purification'' of the ions loaded is

162: necessary~\cite{96:Alheit,01:Toyoda}; in the case of \CaII{43} we show that this is possible with no loss of the

163: desired species. Hence it should be possible to load arbitrarily large crystals of this important ion.

164: 

165: Level diagrams for \CaI{43} and \CaII{43}, showing the energy levels and transitions relevant to these experiments, are

166: shown in figure~\ref{F:levels}. Isotope shifts in the six transitions shown are summarized in table~\ref{T:isodata},

167: together with the abundances of the six naturally occurring isotopes.

168: 

169: \begin{figure*}

170: \begin{centre}

171: \includegraphics[width=\twocolfig]{levels.eps}

172: \caption{%

173: Energy levels of interest in \CaI{43} (for photo-ionization) and \CaII{43} (for laser-cooling), with hyperfine levels

174: labelled by the total angular momentum $F$. The nuclear spin is $I=\wee{7}{2}$. The largest hyperfine splitting in each

175: term is indicated in MHz; the levels approximately conform to an interval

176: rule~\protect\cite{98a:Nortershauser,98b:Nortershauser}. Diagrams for the even isotopes are the same but with no

177: hyperfine structure.

178: %

179: (a)~\CaI{43}: the lifetime of the 4s4p\,\lev{1}{P}{1} level is 4.49(4)\ns~\cite{93:Mitroy} and the natural linewidth of

180: the 423\nm\ transition 35.4(3)\MHz. Photo-ionization from the 4s4p\,\lev{1}{P}{1} level requires a photon of wavelength

181: less than 389.8\nm.

182: %

183: (b)~\CaII{43}: the natural widths of the ultraviolet transitions are approximately 23\MHz, the lifetimes of the

184: metastable D levels about 1.2\s~\protect\cite{00:Barton}. {\em Inset:\/} Positions of the hyperfine components of the

185: 393\nm\ and 397\nm\ transitions (solid lines, labelled with $F\sub{lower}:F\sub{upper}$) and the isotope shifts of the

186: even isotopes (dotted lines, labelled with the mass number), as a function of detuning $\Delta$ relative to \CaII{40}.

187: The isotope shifts in the infrared transitions (similar in all three transitions) are also shown, without the hyperfine

188: details for \CaII{43}; note the change of scale.

189: }%

190: \label{F:levels}

191: \end{centre}

192: \end{figure*}

193: 

194: \begin{table*}

195: \begin{centre}

196: \begin{tabular}{|cc||c||c|c|} \hline

197: Mass & Natural & \multicolumn{3}{c|}{Isotope shifts (MHz)} \\ \cline{3-5}

198: %

199: number  & abundance & Ca \lev{1}{S}{0}--\lev{1}{P}{1} & Ca$^+$ S--P & Ca$^+$ D--P  \\ %

200:           &         & 423\nm & 397\nm\ [393\nm] & 854\nm\ [850, 866\nm] \\ \hline %

201: 40        &  96.9\% &    0  &    0     &       0    \\ %

202: 42        & 0.647\% &  394  &  425(6)  & $-$2350(4) \\ %

203: 43 (c.g.) & 0.135\% &  612  &  688(17) & $-$3465(4) \\ %

204: 44        &  2.09\% &  774  &  842(3)  & $-$4495(4) \\ %

205: 46        & 0.004\% & 1160  & 1287(4)  & $-$6478(8) \\ %

206: 48        & 0.187\% & 1513  & 1696(6)  & $-$8288(7) \\ \hline %

207: \end{tabular}

208: \end{centre}

209: \caption{Abundances of the naturally occurring calcium isotopes~\cite{02:Coplen}, and isotope shifts for Ca and Ca$^+$

210: transitions relevant to this work. Shifts in the 423\nm\ transition are taken from~\cite{98b:Nortershauser};

211: uncertainties quoted are below 1\MHz. Shifts in the ultraviolet Ca$^+$ transitions were measured

212: in~\cite{92:Martensson}; we quote the (more accurate) measurements for the 397\nm\ transition: the shifts at 393\nm\

213: are all consistent with these. Shifts in the infrared transitions are from~\cite{98a:Nortershauser}; the data given are

214: for the 854\nm\ transition, which are again consistent with those for the other two lines. For the odd isotope, the

215: shift of the centre of gravity of the hyperfine components of each transition is given; note that in the Ca$^+$ S--P

216: lines, the hyperfine structure is significantly

217: larger than the isotope shift, figure~\protect\ref{F:levels}(b).} %

218: \label{T:isodata}

219: \end{table*}

220: 

221: 

222: \section{Apparatus}

223: 

224: \subsection{Ion trap}

225: 

226: The apparatus consists of a linear radio-frequency (r.f., 6.2\MHz) Paul trap in an ultra-high vacuum system (background

227: pressure $<2\times 10^{-11}\torr$), with diode laser systems available to drive all the transitions shown in

228: figure~\ref{F:levels}. Details of the trap are given in~\cite{00:Barton}. Typical axial and radial frequencies used in

229: the present experiments are 170\kHz\ and 500\kHz\ respectively. Auxiliary d.c.\ electrodes are used to compensate stray

230: electric fields to the level of about 1\unit{V/m} using a photon-r.f.\ correlation technique applied to the

231: fluorescence from a single trapped ion~\cite{98:Berkeland}. A magnetic field of \ish 3\unit{G} is applied to prevent

232: optical pumping into magnetic sub-states~\cite{00:Boshier}.

233: 

234: \subsection{Lasers}

235: \label{S:lasers}

236: 

237: The lasers are all of the external-cavity grating-stabilized Littrow design, except the 389\nm\ laser which has no

238: grating stabilization. The blue and ultraviolet laser diodes are all GaN devices, the infrared ones GaAlAs. The

239: grating-stabilized lasers may be locked to tunable, stabilized, optical cavities to reduce linewidths and medium-term

240: frequency drift below 5\MHz. The laser frequencies are set within about 100\MHz\ of the ionic transitions with

241: wavemeters. Doppler-free saturated absorption spectroscopy in a calcium hollow cathode is used to monitor the 423\nm\

242: laser frequency relative to the 4s$^2\,\lev{1}{S}{0}\leftrightarrow$ 4s4p\,\lev{1}{P}{1} atomic transition. The vacuum

243: wavelength of the 389\nm\ free-running laser was measured using a grating spectrograph to be 388.9(1)\nm\ at 20\degC\

244: (with negligible dependence on drive current); above lasing threshold, the spectral width is expected to be \ltish

245: 100\MHz, below threshold \ish 3\nm. For convenience of alignment, the 423\nm\ and 389\nm\ laser beams are both injected

246: into the same single-mode optical fibre to transport them to the trap; this ensures that the beams are well overlapped

247: in the interaction region, which is centred on the trap. The transmission of the fibre is \ish 30\% at 423\nm\ and \ish

248: 10\% at 389\nm. The maximum intensities at the centre of these beams are \ish 50\unit{mW/mm$^2$} at 423\nm\ and \ish

249: 5\unit{mW/mm$^2$} at 389\nm. Typical intensities of the cooling lasers at the trap are: 10\unit{mW/mm$^2$} at 397\nm;

250: 20\unit{mW/mm$^2$} at 393\nm; 6\unit{mW/mm$^2$} at 866\nm; 200\unit{mW/mm$^2$} at 854\nm; 650\unit{mW/mm$^2$} at

251: 850\nm. These intensities are based on the measured spot sizes for the various beams, which range from 40--300\um.

252: 

253: \subsection{Detection system}

254: 

255: Fluorescence from the interaction region at 423\nm\ and 397\nm\ acts as a diagnostic of neutral calcium and calcium

256: ions respectively. The trap region is imaged by a compound lens onto an aperture to reject scattered light; further

257: lenses re-image the trap, via a violet filter, onto a photomultiplier (PMT). A movable beamsplitter allows a portion of

258: the light to be directed to a charge-coupled device (CCD) camera; the camera is used to detect the presence of

259: non-fluorescing ions in the trap (revealed as ``gaps'' in an otherwise regular crystal), and to check whether the ions

260: form a cold crystal or a hot cloud. The net detection efficiency of the PMT system, including the solid angle subtended

261: by the lens and the quantum efficiency of the PMT, was measured to be 0.13(2)\% at 397\nm; the photon count rate

262: observed from a single, cold, \CaII{40} ion under conditions of near-saturation is consistent with this figure. The net

263: detection efficiency at 423\nm\ is $\eta=0.15(3)\%$.

264: 

265: The discriminated pulses from the PMT are counted directly by 10\MHz\ on-board counters in the computer which controls

266: the experiment; two alternately-gated counters are used to eliminate read-out delay. For the large peak photon counting

267: rates encountered in studies of the 423\nm\ fluorescence (up to \ish 10$^6$\unit{s$^{-1}$}), we correct the measured

268: PMT count rate $S'$ for the maximum counting rate $S\sub{max}=10\MHz$, assuming $S\ll S\sub{max}$, to obtain the true

269: count rate $S\approx S'/(1-S'/S\sub{max})$.

270: 

271: \subsection{Calcium beam}

272: \label{S:beam}

273: 

274: The calcium source consists of an oven made from stainless steel tube (diameter 2\mm, wall thickness 0.1\mm) filled

275: with granules of calcium metal, closed by crimping at each end, and with an orifice (estimated area \ish

276: 1\unit{mm$^2$}) filed in the side facing towards the trap. The oven is heated by passing a current (between 3\unit{A}

277: and 6\unit{A}) along its length, which produces a beam of calcium atoms directed towards the trap. Estimated oven

278: temperatures are in the range 500--650\K, depending on the current used (section~\ref{S:oven423}).

279: 

280: The oven is situated 22(2)\mm\ from the centre of the trap. The atomic beam is collimated by the r.f.\ trap electrodes

281: to an effective aperture of length $l=1.7(1)\mm$ along the direction of the 423\nm/389\nm\ laser beams. However, the

282: detection system only images a limited length of the interaction region (inset figure~\ref{F:spectrum423}), so the

283: effective aperture for observation of the 423\nm\ fluorescence is reduced to $l'=0.44(4)\mm$. Since the oven orifice is

284: of comparable dimensions to $l,l'$ the angular divergence of the beam is somewhat larger than these distances would

285: otherwise imply.

286: 

287: 

288: \section{Spectroscopy of the atomic beam}

289: 

290: We describe in this section spectroscopy of the 423\nm\ transition, using the calcium atomic beam from the oven. For

291: optimal isotope-selectivity in the photo-ionization loading, the narrowest possible linewidth of this transition is

292: required. Ideally the linewidth should be small compared with the isotope shifts (table~\ref{T:isodata}), with the

293: lower limit being set by the 35.4(3)\MHz\ natural width. Other contributions to the homogeneous width include power

294: broadening (saturation), transit-time broadening and laser linewidth. Collisional broadening is entirely negligible at

295: the low number densities involved here. The inhomogeneous broadening (Doppler width) is determined by the oven

296: temperature, the collimation angle of the atomic beam and the angle between the laser beam and the atomic beam.

297: 

298: For orthogonal beams, the velocity distribution approximates to a Gaussian whose width compared with that in the oven

299: is determined by the collimation geometry and the oven orifice~\citesec{Bk:Demtroder}{10.1}. For non-orthogonal beams,

300: the velocity distribution broadens and becomes asymmetric, resulting in a broadening and shift when it is convolved

301: with the homogeneous profile.

302: 

303: \subsection{Alignment of the crossed beams}

304: 

305: To ensure orthogonal alignment of the laser and atomic beams, we minimize the width of the 423\nm\ fluorescence

306: spectrum as a function of the angle between the beams. The power of the 423\nm\ laser $P_{423}$ is set safely below the

307: saturation level for maximum sensitivity (see next section). A fluorescence spectrum with the alignment optimized is

308: shown in figure~\ref{F:spectrum423}, together with a simultaneously acquired Doppler-free hollow cathode spectrum. From

309: several such spectra we find that the offset between the \CaI{40} fluorescence peak and the hollow cathode

310: saturated-absorption peak is at most 10\MHz\ (after accounting for a pressure red-shift of up to 13\MHz\ in the hollow

311: cathode lamp due to 6\torr\ of neon buffer gas~\cite{72:Smith}). This indicates a beam misalignment of at most \ish

312: 12\mrad.

313: 

314: The fitted Voigt profile shown has Lorentzian full-width at half-maximum (fwhm) 39\MHz\ and Gaussian fwhm 52\MHz. The

315: transit-time broadening is estimated from the measured laser spot size and the approximate atomic velocity

316: (section~\ref{S:oven423}) to be 2(1)\MHz. The laser linewidth estimated from the error signal when the laser is locked

317: to a stabilized reference cavity is 4(1)\MHz. The Gaussian width is consistent with the collimation aperture

318: $l'=0.44(4)\mm$ and the estimated oven temperature $T\ish 610\K$ if we model the oven as an extended source of size

319: \ish 1\mm.

320: 

321: \begin{figure}

322: \begin{centre}

323: \includegraphics[width=\onecolfig]{09may03d.eps}

324: \caption{%

325: (a)~Fluorescence spectrum from the atomic beam at an oven current of 5.0\unit{A} and a laser power below saturation,

326: $P_{423}\ish 6\uW$. The laser was locked to a tunable, stabilized, optical cavity to minimize drift during the scan;

327: the frequency scale was determined using a second, confocal, cavity (600\MHz\ free spectral range). The best-fit Voigt

328: profile is shown (solid line), where the amplitude, linear baseline, frequency offset, Lorentzian and Gaussian widths

329: were all floated simultaneously. The total fwhm is 76\MHz. The abundances, isotope shifts and hyperfine splittings were

330: given the known values~\protect\cite{98b:Nortershauser,02:Coplen}. {\em Right inset:\/} Magnification of the region

331: containing the three hyperfine components of \CaI{43}, whose expected positions are indicated. The dashed line shows

332: the best fit when the \CaI{43} components are omitted from the theoretical profile. The largest contribution to the

333: signal by \CaI{43} is about 22\% of the total signal above background. {\em Left inset:\/} Fluorescence at \CaI{40}

334: line centre from the interaction region, as observed on the CCD camera. The imaging system aperture limits the length

335: of the 423\nm\ beam which is observed to 0.44(4)\mm.

336: %

337: (b)~Simultaneous Doppler-free saturated absorption signal from the hollow cathode lamp; logarithmic transmission is

338: shown. The Doppler-free peak (fwhm \ish 170\MHz) is superimposed on the Doppler-broadened background (fwhm \ish

339: 2.2\GHz, dashed curve), and is centred within 10\MHz\ of the \CaI{40} fluorescence peak. Peaks due to the other

340: isotopes are too small to be visible.

341: }%

342: \label{F:spectrum423}

343: \end{centre}

344: \end{figure}

345: 

346: \subsection{Saturation broadening}

347: 

348: It is important to avoid saturation broadening of the 423\nm\ transition. Since it is difficult to estimate absolute

349: intensities accurately, and the atomic response is averaged over the laser beam profile, we studied the saturation of

350: the transition empirically, by varying the laser power. Results, obtained after alignment of the crossed beams, are

351: shown in figure~\ref{F:saturation}. Saturation effects begin, both in the peak fluorescence and the Lorentzian

352: linewidth, above a laser power of about 10\uW.

353: 

354: The photo-ionization rate will increase as the population of the 4s4p\,\lev{1}{P}{1} state, \ie, linearly until

355: saturation sets in. Since the atomic beam is much larger than the laser beam, however, the total number of atoms in the

356: upper state will continue to increase even well above saturation, though more slowly with increasing power. The

357: isotope-selectivity will decrease roughly as the square of the Lorentzian linewidth, for fixed Gaussian width (the

358: wings of the Voigt profile are determined overwhelmingly by the Lorentzian component, for the Gaussian width involved

359: here). To maximize the photo-ionization rate without sacrificing isotope-selectivity, we work at powers in the range

360: 5--15\uW\ for photo-ionization loading.

361: 

362: Specifically, the fractional population of the 4s4p\,\lev{1}{P}{1}\,$M_J=0$ upper state as a function of detuning

363: $\Delta_{423}$ at $\lambda=423\nm$ is, for linearly-polarized laser light of intensity $I$,

364: \begin{equation}

365: n\sub{4s4p}(\Delta_{423}) = \frac{\wee{3}{2} I A \beta}{(2\pi\Delta_{423})^2 + \beta^2 + 3 I A \beta} %

366: \label{E:pop4s4p}

367: \end{equation}

368: where $A$ is the Einstein coefficient for the transition ($A=2.23\times 10^8\unit{s$^{-1}$}$; the weak decay route to

369: 4s3d\,\lev{1}{D}{2} may be neglected since the calculated lifetime against this decay~\cite{93:Mitroy} is long compared

370: to the transit time across the laser beam), $\beta=(2\pi\Gamma\sub{L}+A)/2$ with $\Gamma\sub{L}$ the laser linewidth,

371: the intensity is measured in units of a saturation intensity $I\sub{sat}=2\pi h c A/\lambda^3$ and $\Delta_{423},

372: \Gamma\sub{L}$ are in units of Hz. This expression must be integrated over the spatial profile of the laser beam to

373: obtain the expected signal from the atomic beam; the resulting homogeneous fwhm, assuming a circular Gaussian beam and

374: a laser linewidth $\Gamma\sub{L}=4\MHz$, is plotted in figure~\ref{F:saturation}(b).

375: 

376: \begin{figure}

377: \begin{centre}

378: \includegraphics[width=\onecolfig]{p423b.eps}

379: \caption{%

380: Fluorescence and Lorentzian linewidth on the 423\nm\ transition, as a function of laser power $P_{423}$, obtained from

381: fitting 423\nm\ spectra with Voigt profiles (whose Gaussian width was fixed at 51\MHz). The oven current was

382: 4.25\unit{A}. The vertical dashed lines indicate the power at which the intensity at the centre of a beam with 100\um\

383: spot size reaches $I\sub{sat}=2\pi h c A/\lambda^3=3.68\unit{mW/mm$^2$}$.

384: %

385: (a)~Peak (corrected) PMT count rate above background on the \CaI{40} component. A straight line ($y\propto x$) is

386: fitted to the data for $P_{423}<10\uW$; the onset of saturation is visible as the signal drops below this line at high

387: powers.

388: %

389: (b)~Analysed Lorentzian fwhm for the same data. The linewidth starts to increase for power above about 10\uW. The

390: horizontal dashed line indicates the 35\MHz\ natural linewidth. The curve shows the expected homogeneous fwhm, obtained

391: by averaging the atomic response over a circular Gaussian beam with 100\um\ spot size and assuming a 4\MHz\ laser

392: linewidth.

393: }%

394: \label{F:saturation}

395: \end{centre}

396: \end{figure}

397: 

398: 

399: \subsection{Number density and temperature in the atomic beam}

400: \label{S:oven423}

401: 

402: We can use the peak \CaI{40} fluorescence at 423\nm\ to deduce the number density of calcium atoms in the atomic beam,

403: using the known line strength and the measured photon collection efficiency. This in turn permits an estimate of the

404: absolute photo-ionization loading efficiency. By measuring the number density as a function of oven current we can also

405: compare the efficiency of photo-ionization trap loading with the electron bombardment method (which have similar

406: loading rates at very different oven currents).

407: 

408: The peak fluorescence as a function of oven current is shown in figure~\ref{F:oven423}(a). The laser power was set

409: below saturation and under these conditions the peak corrected photon count rate $S$ is related to the number density

410: $N$ of atoms in the ground state by

411: \[ S = \frac{3\lambda^3 A}{8\pi h c} \, P_{423} l' \eta r \cdot {\cal V}(0) \cdot a_{40} N \]

412: where $a_{40}=96.9\%$ is the \CaI{40} abundance, $l'=0.44(4)\mm$ the length of the interaction region observed through

413: the imaging aperture (measured along the laser beam direction), $\eta=0.15(3)\%$ the net collection efficiency of the

414: detection system at $\lambda=423\nm$, $r=1.43(2)$ a geometrical factor arising from the angular distribution of the

415: fluorescence and ${\cal V}(0)$ the value of the fitted Voigt profile at line centre (which is normalized such that

416: $\int_{0}^{\infty} {\cal V}(\nu-\nu_0) \diff{\nu}=1$). The fitted values of ${\cal V}(0)$ show a slight decrease with

417: increasing oven current, due to increasing inhomogeneous width as the temperature rises, but this is no greater than

418: the \ish 10\% fitting uncertainty, and for the purpose of figure~\ref{F:oven423}(a) we use the average value

419: $\overline{{\cal V}(0)}=9.9(1.0)\unit{GHz$^{-1}$}$ so that the number density $N[\unit{m$^{-3}$}] = 1.3(3)\times 10^6

420: S[\unit{s$^{-1}$}]$ and it can be plotted on the same ordinate.

421: 

422: For a small interaction region on axis of the atomic beam, the number density $N$ in the interaction region is related

423: to that in the oven $N_0$ by

424: \[ N = \frac{N_0 \sigma}{4\pi d^2} \]

425: where $\sigma$ is the area of the oven orifice and $d$ the distance from the orifice to the interaction

426: region~\citesec{Bk:Scoles}{4.2}. In our apparatus, $d=22(2)\mm$ but $\sigma$ is not accurately known. However, since

427: number density depends strongly on temperature $T$, we can make a reasonable estimate of $T$ by assuming that

428: $\sigma=1.0(5)\unit{mm$^2$}$, the value suggested by our profile analysis. We then find a temperature which satisfies

429: $p=N_0 k T$ and the known vapour pressure curve $p=p(T)$ simultaneously~\cite{Bk:Barin}. This is shown in

430: figure~\ref{F:oven423}(b), and ranges between about 500\K\ and 650\K. The temperature is useful for estimating the mean

431: velocity along the beam $\overline{v}=\sqrt{\pi k T / 2 M}$ for atoms of mass $M$: $\overline{v}=433(4)\unit{m/s}$ for

432: \CaI{40} at $T=575(10)\K$.

433: 

434: \begin{figure}

435: \begin{centre}

436: \includegraphics[width=\onecolfig]{ovene.eps}

437: \caption{%

438: Calcium number density and oven temperature as a function of oven current.

439: %

440: (a)~Corrected photon count rate above background on the \CaI{40} peak obtained from fitting 423\nm\ spectra with Voigt

441: profiles, where the Lorentzian width was fixed at 41\MHz\ from a spectrum at 4.25\unit{A} oven current. The right-hand

442: ordinate gives the calcium number density in the interaction region deduced from this fluorescence rate as described in

443: the text, and the error bars refer to the number density. A quadratic regression fit is shown (solid line). At

444: 3\unit{A}, the signal was so small that the line shape was poorly determined and the Gaussian width was also fixed for

445: the fit.

446: %

447: (b)~Estimated oven temperature based on the number density in (a), with quadratic regression (solid line). The dashed

448: lines indicate the temperature range implied by the 50\% uncertainty assigned to the oven orifice area

449: $\sigma=1.0(5)\unit{mm$^2$}$.

450: }%

451: \label{F:oven423}

452: \end{centre}

453: \end{figure}

454: 

455: 

456: \subsection{Conclusion}

457: 

458: With the 423\nm\ beam well aligned orthogonal to the atomic beam, and the laser power set below saturation, we observe

459: a linewidth of 76\MHz\ (fwhm), made up of Lorentzian and Gaussian contributions of about 39\MHz\ and 52\MHz\

460: respectively. This compares favourably with the isotope shifts, table~\ref{T:isodata}. We do not use these data to

461: predict relative loading probabilities for the different isotopes, however, since the collimation angle of the atomic

462: beam for loading the trap is greater than that relevant to the 423\nm\ fluorescence ($l>l'$) and the Gaussian

463: contribution to the linewidth will be correspondingly larger. A quantitative estimate of this is difficult since the

464: angular divergence depends on the sizes of the oven orifice and the trap capture region, neither of which is accurately

465: known. Instead, we investigate the actual photo-ionization loading rate as a function of 423\nm\ detuning, as described

466: in the following section.

467: 

468: 

469: \section{Photo-ionization loading studies}

470: 

471: \subsection{Expected form of photo-ionization cross-section}

472: \label{S:Xsection}

473: 

474: Calculations~\cite{PC:vanderHart} of the cross-section $Q$ for photo-ionization from the 4s4p state suggest that, at

475: 389\nm, $Q$ lies in the range 60--280\unit{megabarns} ($1\unit{megabarn} = 10^{-22}\unit{m$^2$}$). The large range

476: results from the existence of an auto-ionizing resonance close to the ionization threshold, whose wavelength is

477: uncertain to \ish 5\nm. Since the energy of the resonance is not well-known, we do not attempt to choose a particular

478: laser wavelength, beyond ensuring that it is below the threshold required for ionization. The width of the resonance is

479: certainly large compared to the laser linewidth, so there is no need for the laser to be carefully

480: frequency-stabilized.

481: 

482: The electric fields present in a Paul trap can allow ionization to occur even for excitation slightly below the

483: continuum limit~\cite{01:Gulde}. For the typical velocity estimated above, atoms cross the photo-ionization laser beams

484: in \ish 0.2\us\ and all atoms are thus able to sample the extrema of the r.f.\ field (period 0.16\us). The peak field

485: varies across the trap, but is at most \ish 10$^5$\unit{V/m}. A rough estimate~\cite{Bk:Kuhn} of the effect of such a

486: field on the ionization threshold may be made by assuming a hydrogenic potential $V(x)=-e/4\pi\epsilon_0 x$ near the

487: continuum and equating this to the potential ($-eEx$) of the electron in the external field $E$; the result is that the

488: ionization limit could be shifted at the furthest to a wavelength of 391.3\nm. However, no significant enhancement of

489: the photo-ionization cross-section is expected by using wavelengths closer to this shifted ionization limit, even for

490: monochromatic light, because the oscillatory electric field blurs out the discrete energy levels of the atom near the

491: continuum and the average transition probability is no higher than that within the continuum~\citesec{Bk:Cowan}{18.6}.

492: 

493: 

494: \subsection{Photo-ionization loading rate measurements}

495: 

496: A typical photo-ionization loading curve is shown in figure~\ref{F:PIload}, where the fluorescence at 397\nm\ from cold

497: \CaII{40} ions in the trap is plotted as a function of time after the photo-ionizing lasers are switched on. As the

498: size of the ion crystal grows, the fluorescence per ion increases due to the effect of r.f.\ micromotion for ions lying

499: off-axis, because they spend time closer to resonance with the red-detuned 397\nm\ cooling laser. However, for the

500: range of crystal sizes in these experiments, the fluorescence per ion is roughly constant and we use the final

501: fluorescence level as a measure of the total number of ions loaded. (The fluorescence was calibrated after a typical

502: load by reducing the trap strength until the ions could be counted on the camera.) For each load, we wait a few seconds

503: after the photo-ionizing lasers are switched off, to allow any hot ions to be cooled and to join the crystal, check

504: that the ions are still crystalline using the camera, and then take the final number of ions $m$ divided by the

505: duration $\tau$ of exposure to the photo-ionizing lasers as our measure of the loading rate $R$. We find that an error

506: $\sqrt{m}/\tau$ is broadly consistent with variations in the loading rate taken under nominally identical conditions

507: (see for example figure~\ref{F:PD423}).

508: 

509: We have studied the loading rate as a function of the power $P_{423}$ and detuning $\Delta_{423}$ of the 423\nm\ laser,

510: and of the power $P_{389}$ of the 389\nm\ laser. This was done at fixed oven current, to keep the atomic number density

511: as constant as possible. For each measurement, the 397\nm\ fluorescence was allowed to increase to a level representing

512: about 40 trapped ions, whereupon the photo-ionizing lasers were blocked. In practice, the number of ions loaded varied

513: between 22(2) and 63(5); the length of photo-ionizing exposure was between 2\s\ and 600\s.

514: 

515: \begin{figure}

516: \begin{centre}

517: \includegraphics[width=\onecolfig]{piload2.eps}

518: \caption{%

519: Typical photo-ionization loading fluorescence curve. The 423\nm\ and 389\nm\ photo-ionizing lasers are switched on at

520: $t=0$ and off at $t=216\s$ (dotted line). The 397\nm\ and 866\nm\ cooling lasers illuminate the ions continuously.

521: Background counts due to scattered 397\nm\ light and (during the loading) 423\nm\ fluorescence and 389\nm\ light have

522: been subtracted from the PMT signal. The step in 397\nm\ fluorescence when the first cold ion appears is arrowed. The

523: large steps in the fluorescence are typical and may result from the details of the cooling dynamics as the number of

524: ions in the trap increases. The dashed line indicates the final fluorescence level, which corresponds to 36(3) ions;

525: the average loading rate of 0.17(3)\unit{ion/s} is obtained by dividing this by the time for which the photo-ionizing

526: lasers are switched on. The conditions were: oven current 4.25\unit{A}, $P_{423}\ish 6\uW, \Delta_{423}=0(5)\MHz,

527: P_{389}=0.93(9)\uW, \Delta_{397}=-68(10)\MHz$.

528: }%

529: \label{F:PIload}

530: \end{centre}

531: \end{figure}

532: 

533: Figure~\ref{F:PD423} shows the dependence of loading rate on 423\nm\ power and detuning. Note that the loading rate

534: could in principle saturate at a different power level to the 423\nm\ fluorescence studied above, since

535: photo-ionization is limited by the width of the 389\nm\ beam (the lifetime of the 4s4p\,\lev{1}{P}{1} level,

536: 4.49(4)\ns, is much shorter than the time an atom spends in the laser beams, \ish 0.2\us) whereas the fluorescence

537: collected is limited by the imaging aperture. Nevertheless, we observe that the loading rate also starts to saturate

538: above a power of $P_{423}\ish 10\uW$ due to saturation of the upper state population. Given the power available in the

539: 423\nm\ laser, the maximum loading rate could be made perhaps an order of magnitude larger by increasing the spot size

540: at the trap to 1\unit{mm}; spot sizes much larger than this, however, would start to reduce the isotope-selectivity

541: because of the increasing atomic beam divergence angle.

542: 

543: The loading rate as a function of $\Delta_{423}$ is of critical importance for isotope-selective loading. We expect the

544: dependence to be modelled by a Voigt profile with similar Lorentzian width to that of the 423\nm\ fluorescence

545: spectrum, but different Gaussian width because of the different atomic beam collimation angles for fluorescence

546: detection and ion loading. The Lorentzian and Gaussian widths are not well-determined if they are both floated, due to

547: the noise in the data, so we fix the Lorentzian width at 39\MHz, from the fluorescence spectrum in

548: figure~\ref{F:spectrum423}, and find a fitted Gaussian width of 71\MHz. This Gaussian width is consistent with a beam

549: collimation angle defined by the electrodes and an oven orifice of size \ish 1\mm. The consequent isotope-selectivity

550: expected is discussed in section~\ref{S:isosel}.

551: 

552: \begin{figure}

553: \begin{centre}

554: \includegraphics[width=\onecolfig]{pd423b.eps}

555: \caption{%

556: (a)~Photo-ionization loading rate as a function of 423\nm\ laser power, with: oven current 4.25\unit{A},

557: $\Delta_{423}=0(10)\MHz, P_{389}=0.93(9)\uW, \Delta_{397}=-80(20)\MHz$. The loading rate starts to saturate for

558: $P_{423}\gtish 10\uW$. The curve is drawn to guide the eye.

559: %

560: (b)~Loading rate versus 423\nm\ detuning, with: oven current 4.25\unit{A}, $P_{423}\ish 6\uW, P_{389}=80(8)\uW,

561: \Delta_{397}=-97(10)\MHz$. A Voigt profile (solid line) is fitted to the data; the frequency offset, amplitude and

562: Gaussian width were floated whilst the Lorentzian width was fixed at 39\MHz\ from the fit in

563: figure~\protect\ref{F:spectrum423}. The fwhm of the fitted curve is 94\MHz. The frequency origin is taken from the

564: fitted offset of a 423\nm\ fluorescence spectrum taken during the same run; the fitted profile here shows a blue-shift

565: of 12\MHz\ which probably results from the slightly different collimation geometry relevant to loading.

566: }%

567: \label{F:PD423}

568: \end{centre}

569: \end{figure}

570: 

571: In contrast to the 423\nm\ bound-bound transition, no saturation effect is expected for the 389\nm\ bound-free ionizing

572: transition, and this is verified for the 389\nm\ powers we can attain (figure~\ref{F:P389}): the dependence of the

573: loading rate on $P_{389}$ is linear over two orders of magnitude. Significantly, there is no change in efficiency

574: whether the 389\nm\ laser is running above or below lasing threshold, despite the fact that the spectral width is

575: expected to be about five orders of magnitude larger (\ish 3\nm) below threshold. This implies a weak dependence on

576: wavelength of the photo-ionization cross-section in this region.

577: 

578: \begin{figure}

579: \begin{centre}

580: \includegraphics[width=\onecolfig]{p389.eps}

581: \caption{%

582: Photo-ionization loading rate as a function of 389\nm\ laser power, with: oven current 4.25\unit{A}, $P_{423}\ish 6\uW,

583: \Delta_{423}=0(5)\MHz, \Delta_{397}=-68(10)\MHz$. No significant difference in efficiency is observed even when the

584: laser is well below lasing threshold (vertical dashed line), indicating that the photo-ionization cross-section is

585: independent of the spectral width of the laser. A straight line ($y\propto x$) is fitted to the data. The single

586: measurement with the ultraviolet LED in place of the 389\nm\ laser is also included (see

587: section~\protect\ref{S:uvLED}).

588: }%

589: \label{F:P389}

590: \end{centre}

591: \end{figure}

592: 

593: 

594: \subsection{Absolute photo-ionization efficiency}

595: 

596: We can deduce a lower limit for the photo-ionization cross-section from the loading rate under conditions of

597: near-saturation on the 423\nm\ transition, when we expect that nearly half the atoms are in the 4s4p\,\lev{1}{P}{1}

598: upper level (at least across that part of the 423\nm\ beam which overlaps the 389\nm\ beam). From the definition of the

599: photo-ionization cross-section $Q$, the number of atoms ionized per unit time is

600: \[ R = Q l \cdot \wee{1}{2} a_{40} N \cdot P_{389} / \varepsilon \]

601: where $l=1.7(1)\mm$ is the length of the interaction region defined by the trap electrodes, $(\wee{1}{2} a_{40} N)$ the

602: number density of \CaI{40} atoms in the 4s4p\,\lev{1}{P}{1} level, $P_{389}$ the 389\nm\ laser power, and $\varepsilon$

603: the 389\nm\ photon energy.

604: 

605: To minimize uncertainty in the number density due to variations in oven conditions, $N$ was determined shortly before

606: measuring the loading rate and found to be $1.7(4)\times 10^{11}\unit{m$^{-3}$}$. At maximum 423\nm\ power, the

607: estimated intensity at the centre of the laser beam was $14(4)I\sub{sat}$ which, for $|\Delta_{423}| < 10\MHz$, gives

608: $n\sub{4s4p}>0.48(1)$ within the half-intensity radius of the beam (equation~\ref{E:pop4s4p}), justifying the

609: assumption of saturation. The loading rate was measured to be $R=4.9(9)\unit{ion/s}$ with $P_{389}=1.1(1)\uW$, giving a

610: cross-section $Q=170(60)\unit{megabarns}$. This is a lower limit since it assumes that every ion produced is trapped

611: and that the capture region of the trap extends across the full width of the atomic beam as far as the r.f.\

612: electrodes, but it lies in the anticipated range (section~\ref{S:Xsection}).%

613: 

614: Alternatively, the loading rate can be expressed in terms of the probability $q$ per atom of being ionized (and

615: trapped) as it crosses the interaction region:

616: \[ R = q \cdot \wee{1}{2} a_{40} N \cdot l w^2 / t \]

617: where $w^2$ is the cross-sectional area of the 389\nm\ beam (which limits the extent of the interaction region), and

618: $t\approx w/v$ is the time an atom of velocity $v$ spends in the interaction region. For the conditions above and the

619: laser beam size $w\approx 0.1\mm$, we have $t\approx 0.2\us$ and find $q\approx 8\times 10^{-7}$. The maximum $q$ we

620: observed (with maximum powers in both the photo-ionizing lasers) was $q\approx 4\times 10^{-5}$, and by eliminating the

621: optical fibre and using the maximum available 389\nm\ power \ish 2\mW, we would in principle be able to obtain a

622: maximum $q\approx 10^{-3}$, \ie, one in 1000 atoms crossing the interaction region would be loaded.

623: 

624: 

625: \subsection{Loading using an ultraviolet LED}

626: \label{S:uvLED}

627: 

628: The observation that the 389\nm\ laser is able to photo-ionize atoms just as efficiently below lasing threshold led us

629: to experiment with an alternative light source for the second step of the photo-ionization, an ultraviolet

630: light-emitting diode (LED). The device tested was a Nichia NSHU590 diode, with nominal output power 500\uW, peak

631: wavelength 375\nm\ and spectral width 12\nm. The chief advantages of this device are that it is about three orders of

632: magnitude cheaper than a wavelength-selected ultraviolet laser diode, and does not require temperature and current

633: stabilization. Measurements with a grating spectrograph indicated that the peak wavelength was in fact between 375\nm\

634: and 390\nm, depending which part of the emission region was focussed on the spectrograph slit, and the spectral width

635: was \ish 25\nm; however at least half of the spectral intensity lies in the range $\lambda<389.8\nm$ useful for

636: photo-ionization. The main drawback of the LED is its extended emission region: it is difficult to focus a large

637: fraction of the power into a region smaller than a few millimetres, unlike a laser beam.

638: 

639: When we first tested the LED in place of the 389\nm\ laser we were surprised to find that it was capable of loading

640: ions into the trap while the 423\nm\ laser was blocked and the oven was switched off! A small number of \CaII{40} ions

641: were loaded, together with several ``dark'' ions; at the same time we noticed that the stray d.c.\ electric field in

642: the trap changed by \ish 10\%, drifting gradually back to its former value over a period of an hour or so. Our

643: interpretation is that the large (\ish 5\mm) patch of light from the LED was irradiating the r.f.\ electrodes, ablating

644: and photo-ionizing calcium and other ions from the surface where they had previously been deposited by the atomic beam.

645: Photo-electric emission could be responsible for charging insulating patches on the electrodes, causing the stray field

646: to change temporarily before the charge leaks away (the photo-electric cut-off wavelength for calcium is 433\nm).

647: Surface ablation is a well-known method of loading ion traps~\cite{02:Blinov}, though it is perhaps surprising at the

648: low light intensities involved here.

649: 

650: To avoid this problem, we imaged the LED onto a 200\um\ pinhole, which was re-imaged at approximately 1:1 magnification

651: into the trap region, and overlapped with the 423\nm\ beam. This eliminated loading of non-\CaII{40} ions and effects

652: on the stray field, but only 0.32(3)\uW of ultraviolet power was then available at the trap. The loading rate under

653: these conditions, 0.06(1)\unit{ion/s}, is plotted in figure~\ref{F:P389} and is similar to that expected with

654: comparable 389\nm\ laser power. This rate is not sufficient for loading the rarer isotopes, where it is necessary to

655: beat losses due to charge-exchange (section~\ref{S:exchange}), but is perfectly adequate for loading small crystals of

656: \CaII{40}. We note that high-power (\ish 150\mW) ultraviolet LEDs are also available, though their cost approaches that

657: of a laser diode.

658: 

659: 

660: \section{Isotope-selective loading}

661: 

662: \subsection{Expected selectivity}

663: \label{S:isosel}

664: 

665: From the data in figure~\ref{F:PD423}(b) we expect good isotope-selectivity, because the net fwhm, 94\MHz, is small

666: compared with the isotope shifts in the 423\nm\ transition. This discrimination has to be set against the very low

667: abundance of some isotopes. In figure~\ref{F:isosel} we plot the calculated probability of loading the different

668: isotopes as a function of detuning $\Delta_{423}$, based on the fitted Voigt profile in figure~\ref{F:PD423}(b). We

669: regard the loading as ``isotope-selective'' for those isotopes where there exists a detuning at which one is more

670: likely to load the desired isotope than any other; it can be seen from figure~\ref{F:isosel} that this is the case for

671: all isotopes except \CaII{43} and \CaII{46}. The most demanding case is \CaII{46}, where the maximum achievable loading

672: probability, relative to the total loading probability, is only 6.0\%. The relative probability of loading \CaII{43} is

673: also plotted in the figure: it reaches a maximum of 21\% at a detuning of 0.623\GHz. This compares with a theoretical

674: maximum of 33\% which would be possible with a Lorentzian width equal to the natural linewidth and negligible Gaussian

675: width; thus not a great deal would be gained by efforts to collimate further the atomic beam.

676: 

677: The detunings of the Doppler-cooling lasers also affect the selectivity, due to the isotope shifts in the ionic

678: transitions (table~\ref{T:isodata}). For example, to cool \CaII{42} efficiently, the 397\nm\ laser needs to be detuned

679: \ish 350\MHz\ to the blue of the \CaII{40} transition, which will tend to heat \CaII{40} ions and expel them from the

680: trap, thus enhancing the selectivity.

681: 

682: \begin{figure}

683: \begin{centre}

684: \includegraphics[width=\onecolfig]{isosel.eps}

685: \caption{%

686: Probability of loading different calcium isotopes as a function of detuning of the 423\nm\ laser, normalized to 1 for

687: \CaII{40} at zero detuning. The curves are generated using Voigt profiles with the same width parameters as the fit in

688: figure~\protect\ref{F:PD423}(b) (except that the Gaussian width is adjusted to take account of the different isotopic

689: masses). The dashed curves for the even isotopes, and the solid curve for \CaII{43} (the three hyperfine components

690: have been summed), are labelled by the mass numbers. The thick solid curve is the sum of the contributions from all the

691: isotopes. The detached solid curve gives the probability of loading \CaII{43} relative to the total loading probability

692: at the same detuning, plotted in the region where this quantity is $>0.05$.

693: }%

694: \label{F:isosel}

695: \end{centre}

696: \end{figure}

697: 

698: \subsection{Charge-exchange loss}

699: \label{S:exchange}

700: 

701: It is possible for an ion \CaII{X} which has been loaded into the trap to undergo charge-exchange with a neutral atom

702: in the atomic beam; since the overwhelming probability is that the neutral atom is \CaI{40} the process is usually:

703: \[ \CaII{X} + \CaI{40} \rightarrow \CaI{X} + \CaII{40} \]

704: and the original ion in the trap is replaced by \CaII{40}. The charge-exchange rate is proportional to the number

705: density in the atomic beam. This process is a limitation for loading more than one ion of a rare isotope: if the loss

706: rate through charge-exchange exceeds the loading rate, then rare ions already trapped are replaced by \CaII{40} faster

707: than subsequent ones can be loaded. If the charge-exchange rate is sufficiently fast, then rare ions will not even be

708: cooled before they are replaced, and the trap will fill up with \CaII{40} before any fluorescence signal from the rare

709: ions is seen.

710: 

711: In practice we find that, providing one is not attempting to load large crystals, charge-exchange is only a limitation

712: for the rarest isotope \CaII{46}. For the typical loading parameters used for loading \CaII{43}, for example, the

713: loading rate is \ish 1\unit{ion/s} while the lifetime against loss due to charge-exchange is \ish 1\unit{min}.

714: \CaII{46} is about thirty times less abundant, so we expect the loss rate to approach the loading rate under the same

715: conditions.

716: 

717: We note that the charge-exchange process was used by Mortensen \etal~\cite{02:Mortensen} to demonstrate

718: isotope-selective loading, but that they were not attempting to retain and to laser-cool the different isotopes;

719: instead they relied on charge-exchange to replace every isotope by \CaII{40} and used the resulting \CaII{40}

720: fluorescence as a measure of the original load size.

721: 

722: 

723: \subsection{Even isotopes}

724: 

725: Loading one of the even isotopes consists of choosing the 423\nm\ detuning for maximum selectivity and setting the

726: appropriate detunings of the 397\nm\ and 866\nm\ lasers used for Doppler-cooling. After loading, we can check for the

727: presence of other isotopes by looking for fluorescence using different detunings of the cooling lasers (the lasers are

728: blocked while the detunings are changed, to avoid possible heating). Phase-sensitive detection with a chopped 866\nm\

729: beam can be used to detect very small fluorescence signals above scattered 397\nm\ background light.

730: 

731: As expected from figure~\ref{F:isosel} we can load small, pure crystals of \CaII{44} and \CaII{48} (and of course

732: \CaII{40}) without difficulty. Predicted peak loading probabilities of \CaII{44} and \CaII{48} are 94\% and 85\%

733: respectively. Practical loading rates can be achieved in spite of the small abundances by using maximum 389\nm\ power

734: (\ish 100\uW) and by increasing the oven current. We concentrate here on the more challenging cases of \CaII{42} (which

735: lies closest to \CaII{40}) and \CaII{46} (which is extremely rare).

736: 

737: The maximum expected probability of loading \CaII{42} is 58\%, at a detuning $\Delta_{423}=0.394\GHz$, with \CaII{40}

738: by far the dominant impurity. However, when we loaded using this detuning, we obtained a crystal consisting of \ish 19

739: \CaII{42} ions and only \ish 5 \CaII{40} ions, \ie, a loading efficiency of about 80\%. We attribute this to the

740: mechanism mentioned above, that the 397\nm\ laser was blue-detuned from the \CaII{40} transition by about 350\MHz, thus

741: tending to heat this species and expel it from the trap; the \CaII{40} ions are only cooled sympathetically by the

742: \CaII{42}~\cite{99:Bowe}. This mixed-species crystal also illustrates the typical charge-exchange rates: after \ish

743: 20\unit{min}, with the photo-ionizing lasers blocked but the oven left on at 4.25\unit{A}, the number of \CaII{40} ions

744: had increased to \ish 18 and there was only a very low signal visible at the \CaII{42} detunings; after a further

745: 30\unit{min}, there was no detectable signal from \CaII{42}. This indicates that the lifetime (per ion) against loss by

746: charge-exchange is \ish 17\unit{min}, for a number density in the atomic beam of $9(2)\times 10^{10}\unit{m$^{-3}$}$.

747: The limited isotope-selectivity achievable could be overcome by selective heating methods~\cite{96:Alheit,01:Toyoda} to

748: obtain a pure crystal of \CaII{42}.

749: 

750: The 0.004\%-abundant \CaII{46} is the most difficult isotope to load, with an expected peak loading probability of

751: 6.0\% (neglecting enhancement by heating as observed for \CaII{42}). To achieve the maximum photo-ionization rate

752: without excessive saturation-broadening of the 423\nm\ transition, we set $P_{423}\ish 180\uW$; we used maximum 389\nm\

753: power (\ish 100\uW) and increased the oven current to 5.0\unit{A}. Using the measured loading rates for \CaI{40} and

754: the number density data we expect a \CaII{46} loading rate of \ish 0.1\unit{ion/s}, with a sacrifice of a factor \ish 3

755: in selectivity because of the high 423\nm\ power. Under these conditions, a fluorescence signal was detected after \ish

756: 17\unit{s} of exposure to the photo-ionizing lasers. The oven was switched off immediately to prevent charge-exchange.

757: The ions could not be made to crystallize, however, indicating that too many impurity ions were present to be cooled

758: sympathetically by the \CaII{46}. We therefore applied a ``tickle'' voltage to one of the trap end-caps, close to the

759: axial resonance frequency, which is less likely to expel the directly-cooled \CaII{46} than other isotopes.

760: Crystallization was then observed, figure~\ref{F:ca46}, and camera images implied that the crystal consisted of a

761: single \CaII{46} ion and several impurity ions (most probably \CaII{40}). The observed isotope shift in the 866\nm\

762: transition was $-6.4(2)\GHz$; although this particular shift has not been measured before, it is in agreement with that

763: expected (table~\ref{T:isodata}) and thus confirms the identity of the ion.

764: 

765: \begin{figure}

766: \begin{centre}

767: \includegraphics[width=\onecolfig]{ca46.eps}

768: \caption{%

769: Fluoresence signal from a single \CaII{46} ion, trapped in the company of several ``dark'' ions (probably \CaII{40}),

770: as a function of 397\nm\ detuning relative to \CaII{40}. Three scans were averaged to produce the plot. The

771: high-frequency ``edge'' in the fluorescence curve, characteristic of scanning past the resonant frequency, has been

772: assigned the known isotope shift of 1287\MHz. The first dip in the fluorescence (arrowed) is the point at which the

773: ions crystallize and the single \CaII{46} ion becomes visible on the camera ({\em inset}). The second dip, at \ish

774: 1200\MHz, is probably due to a two-photon dark resonance between the 866\nm\ and 397\nm\

775: transitions~\protect\cite{92:Siemers}. The crystallization dip, and the movement of the single \CaII{46} ion between

776: discrete positions in the trap after heating and re-crystallization, indicate the presence of other ions.

777: }%

778: \label{F:ca46}

779: \end{centre}

780: \end{figure}

781: 

782: 

783: \subsection{Odd isotope}

784: 

785: The laser requirements for Doppler-cooling \CaII{43} are more demanding than for the even isotopes because of the

786: hyperfine structure, figure~\ref{F:levels}(b). In particular, a separate repumper is necessary on one of the

787: ultraviolet transitions because of the large (3.2\GHz) hyperfine splitting between the \hfslev{S}{1/2}{3} and

788: \hfslev{S}{1/2}{4} levels (where the superscript denotes $F$).

789: 

790: We choose the 393\nm\ $\hfslev{S}{1/2}{4}\leftrightarrow\hfslev{P}{3/2}{5}$ transition for cooling, both because it is

791: an intrinsically strong $F\leftrightarrow F+1$ transition and (thanks to the inverted hyperfine structure) it is

792: blue-detuned from the $\lev{}{S}{1/2}\leftrightarrow\lev{}{P}{3/2}$ transitions in {\em all\/} the even isotopes (inset

793: figure~\ref{F:levels}). This second fact means that any other isotopes that are loaded will tend to be heated.

794: Furthermore, this transition is nearly-cycling (apart from the \ish 6\% decay route to the D levels and off-resonant

795: driving of the 160\MHz-detuned $\hfslev{S}{1/2}{4}\leftrightarrow\hfslev{P}{3/2}{4}$ component), so that the second

796: ultraviolet laser needed for repumping out of the \hfslev{S}{1/2}{3} level can still repump efficiently at low

797: intensities: this is an advantage since it is necessarily red-detuned relative to the even isotopes and can thus cool

798: them. We choose to use the 397\nm\ $\hfslev{S}{1/2}{3}\leftrightarrow\hfslev{P}{1/2}{4}$ transition for this repumping.

799: 

800: Repumping from the \lev{}{D}{3/2} and \lev{}{D}{5/2} manifolds is accomplished with the 850\nm\ and 854\nm\ lasers

801: respectively: the high intensities available in these beams mean that they can repump adequately in spite of the

802: hyperfine structure which spans \ish 400\MHz\ in each transition.

803: 

804: To load we set the appropriate detunings of these four lasers relative to \CaII{40}, and set $\Delta_{423}\approx

805: 600\MHz$. The 423\nm\ power is typically \ish 10\uW, to avoid saturation-broadening, and the 389\nm\ power maximized

806: (\ish 100\uW) for fastest loading. Under these conditions and with an oven current of 5.0\unit{A}, the loading rate is

807: \ish 1\unit{ion/s} and the lifetime against loss through charge-exchange \ish 1\unit{min}. This gives time to detect a

808: single ion of \CaII{43} and switch the oven off before it is replaced by one of \CaII{40}. With this procedure, we can

809: reliably load single ions of \CaII{43} and only rarely find that we have loaded \CaII{40} as well; this indicates that

810: we achieve significant enhancement over the estimated maximum loading probability of only 21\%, presumably because

811: \CaII{40} ions are heated by the 393\nm\ laser and expelled from the trap. The maximum fluorescence rate detected from

812: a single \CaII{43} ion was 23000\unit{s$^{-1}$}, compared with typically 32000\unit{s$^{-1}$} for \CaII{40} (a

813: reduction attributable to less efficient repumping from the D states).

814: 

815: To confirm the identity of the first single ion we loaded by this method, we scanned the frequency of the 397\nm\

816: repumper: at low intensity, the two hyperfine components $\hfslev{S}{1/2}{3}\leftrightarrow\hfslev{P}{1/2}{4}$ and

817: $\hfslev{S}{1/2}{3}\leftrightarrow\hfslev{P}{1/2}{3}$ are well-resolved, figure~\ref{F:ca43}, and have the expected

818: separation. As a further check, the axial resonance frequency was measured and compared with that of \CaII{40}, giving

819: a mass number of 42.8(2).

820: 

821: If the photo-ionizing lasers and oven are left on for longer we find that we load crystals containing about 50\%

822: \CaII{43} and 50\% \CaII{40}. This may occur either because of charge-exchange, or because once \CaII{43} is trapped

823: and cooled, it can cool \CaII{40} sympathetically in spite of the heating effect of the 393\nm\ laser. We discovered

824: that it was possible to purify such mixed crystals by temporarily blocking the 397\nm\ ultraviolet repumper; when this

825: laser is blocked a small fluorescence signal is still visible from \CaII{43} (because of off-resonant repumping on

826: $\hfslev{S}{1/2}{3}\leftrightarrow\hfslev{P}{3/2}{2;3;4}$ by the 3\GHz-detuned 393\nm\ laser) but there is no longer

827: any laser capable of cooling \CaII{40} directly. We assist the \CaII{40} heating by switching on the 866\nm\ repumper,

828: tuned to be resonant with the \CaII{40} transition. One might expect the Coulomb coupling between the different species

829: to make this purification process rather inefficient, but we found on the contrary that it always seems successful at

830: removing the \CaII{40} ions without loss of any \CaII{43}. For example, a mixed crystal consisting of three \CaII{43}

831: ions and eight \CaII{40} ions was purified completely by blocking the 397\nm\ repumper for \ish 8\unit{s}. The

832: technique can also be applied during loading: with the oven and photo-ionizing lasers on, we watch the \CaII{43} ions

833: being loaded on the camera and, as soon as there is evidence of ``dark ions'' in the trap, block the 397\nm\ repumper

834: for a few seconds to purify the crystal, then un-block it to continue loading. In this way, pure \CaII{43} crystals of

835: arbitrary size can be loaded; the inset of figure~\ref{F:ca43} shows a 9-ion linear crystal.

836: 

837: \begin{figure}

838: \begin{centre}

839: \includegraphics[width=\onecolfig]{ca43hfs.eps}

840: \caption{%

841: Fluorescence signal from a single \CaII{43} ion as the frequency of the 397\nm\ ultraviolet repumping laser is scanned.

842: The 397\nm\ intensity was reduced to \ish 0.2\unit{mW/mm$^2$} for this scan, to reduce the widths of the peaks. A

843: double Lorentzian is fitted to the data: the separation between the two components is 582(6)\MHz\ (where a 1\% error is

844: allowed in deducing the frequency scale from a confocal cavity of 300\MHz\ free spectral range), in excellent agreement

845: with the known hyperfine splitting in the \lev{}{P}{1/2} level of 581\MHz. The

846: $\hfslev{S}{1/2}{3}\leftrightarrow\hfslev{P}{1/2}{4}$ component has been assigned the known shift of $-1381\MHz$

847: relative to \CaII{40}.

848: %

849: {\em Inset:\/} A pure crystal of nine \CaII{43} ions, loaded using periodic blocking of the 397\nm\ repumper to remove

850: even-isotope impurity ions, as described in the text. The separation between the closest ions is 13(1)\um.

851: }%

852: \label{F:ca43}

853: \end{centre}

854: \end{figure}

855: 

856: 

857: \section{Photo-ionization and electron bombardment compared}

858: 

859: The limitations of electron bombardment ionization are discussed in~\cite{01:Gulde}. A significant problem in our

860: apparatus was the effect of firing the electron gun on the stray electric fields in the trap. After loading, the static

861: field necessary to compensate the micromotion of the ions was found to decay exponentially with a time constant \ish

862: 6\unit{hours}. For the typical load shown in figure~\ref{F:compensation}(a) the vertical stray field would not have

863: been stable to the 1\unit{V/m} resolution of our compensation method until \ish 28\unit{hours} after the trap was

864: loaded. We ascribe this drift to charge on insulating parts of the trap structure which decays with time, and have not

865: observed this systematic decay after photo-ionization loading. On longer timescales, between loads, we observed that

866: the vertical stray field changed by up to \ish 200\unit{V/m}; this is compared with recent photo-ionization

867: compensation data in figure~\ref{F:compensation}(b). The cause of these long-term changes in the stray field could have

868: been changes in the distribution of material deposited on the trap electrodes when the oven or electron gun was fired:

869: calcium deposited on the stainless steel r.f.\ electrodes could give rise to static fields of this magnitude, due to

870: the difference in work functions between these two metals.

871: 

872: This highlights a further issue: based on the work of the NIST Ion Storage group, we expect that material deposited

873: onto the trap electrodes from the oven and/or electron gun will contribute significantly to motional heating of trapped

874: ions in the sub-Doppler cooling regime relevant to quantum logic experiments~\cite{00:Turchette,02:Rowe}. The estimates

875: of calcium number density and temperature in the atomic beam given in section~\ref{S:oven423}, above, imply a flux of

876: $\ish 4\times 10^9\unit{atom/s/mm$^2$}$ on a surface normal to the atomic beam, at the 6\unit{A} oven current which was

877: necessary for loading the trap using electron bombardment. For the same loading rate, this can be reduced by over four

878: orders of magnitude using photo-ionization. Furthermore, the same photo-ionization loading rate can be maintained even

879: when using a well-collimated atomic beam designed to prevent deposition of atoms on the trap structure.

880: 

881: With maximum power in both photo-ionizing lasers, we measured a loading rate of 1.6(2)\unit{ion/s} at an oven current

882: of 3.25\unit{A}. This compares with an electron bombardment loading rate of \ish 1\unit{ion/s} at an oven current of

883: 6.0\unit{A}. Extrapolating the number density data of figure~\ref{F:oven423}(a) to 6.0\unit{A} oven current indicates

884: that photo-ionization is a factor of $\ish 10^{4}$ times more efficient than our electron bombardment ionization setup.

885: A further two orders of magnitude enhancement of the loading efficiency in this apparatus would be available by (i)

886: increasing the spot sizes of the photo-ionizing lasers to reduce saturation effects on the 423\nm\ transition and (ii)

887: eliminating the optical fibre in the 389\nm\ beam. We note, however, that the per atom efficiency of photo-ionization

888: trap loading can be approached by techniques of surface ionization~\cite{00:Savard}.

889: 

890: The photo-ionization method overcomes the limitations of electron bombardment ionization: the species- and

891: isotope-selectivity ensure that unwanted ions are not trapped and furthermore allow studies of rare isotopes without

892: enriched sources; there is negligible effect due to the loading on the stray fields (at the level of \ish 1\unit{V/m}

893: in this trap); it works well irrespective of radial trap strength~\cite{01:Gulde}, so that compensation of stray fields

894: for different trap conditions is unnecessary; the high efficiency of photo-ionization means that the flux of the atomic

895: beam can be attenuated by many orders of magnitude which reduces immensely problems associated with deposition of

896: material on the trap structure. These advantages must be weighed against the complexities of two additional lasers,

897: compared with an electron gun inside the vacuum system.

898: 

899: \begin{figure}

900: \begin{centre}

901: \includegraphics[width=\onecolfig]{comp.eps}

902: \caption{%

903: (a)~Drift of the stray field versus time after loading with electron bombardment ionization. The magnitude of the

904: vertical field required to compensate horizontal ion micromotion is plotted. The micromotion compensation here was

905: performed by minimizing the width of the $\lev{}{S}{1/2}\leftrightarrow\lev{}{P}{1/2}$ spectrum; we assign an error of

906: $\pm 3\unit{V/m}$. An exponential decay function is fitted to the data (solid line); this has time-constant

907: 6.3(7)\unit{hours} and amplitude 88(4)\unit{V/m}. No such drift is visible after loading by photo-ionization, at the

908: 1\unit{V/m} level.

909: %

910: (b)~Variation of the steady-state stray field versus time, over the course of nearly a year for electron bombardment

911: loading (squares, left) and over the course of 45 days for photo-ionization loading (circles, right). Each point is

912: taken from a different load. For the electron bombardment data the error bar represents the estimated additional drift

913: possible after the final compensation measurement. For the photo-ionization data, the estimated error in compensating

914: the stray field is 1\unit{V/m}; the standard deviation of the data set is 2.8\unit{V/m}, indicating some residual

915: systematic variation (which may be due to the remaining atomic beam flux, or mechanical variations in the trap geometry

916: over time).

917: }%

918: \label{F:compensation}

919: \end{centre}

920: \end{figure}

921: 

922: 

923: \section{Conclusion}

924: 

925: We have presented quantitative studies of photo-ionization ion-trap loading for calcium ions, using the fluorescence

926: from the neutral atoms to estimate the absolute efficiency of this process. We find that, in our trap, this loading

927: method is around four orders of magnitude more efficient than conventional electron impact ionization, comparably to

928: Gulde \etal~\cite{01:Gulde}, and does not give rise to stray electric fields. We have shown that the setup of Gulde

929: \etal\ may be further simplified by eliminating frequency-stabilization for the 389\nm\ laser and, if high loading

930: rates are not required, by replacing this laser with a light-emitting diode. We have optimized the photo-ionization

931: procedure for the purposes of isotope-selection, demonstrating trapping of all the naturally occurring isotopes from a

932: natural abundance source. The rare species \CaII{43} and \CaII{46} have been laser-cooled for the first time.

933: 

934: The odd isotope \CaII{43} is of particular interest for experiments in quantum information processing and optical

935: frequency standards, and we have demonstrated the ability to load pure crystals of this ion. We note, however, that an

936: enriched source would still be advantageous for work with this isotope, since it would reduce any remaining deposition

937: of material from the atomic beam on the trap electrodes to the minimum possible. An interesting feature of

938: photo-ionization compared with electron bombardment ionization is that the laser beams allow it to be

939: spatially-selective: this could be useful for loading specific traps in proposed quantum computing architectures

940: involving microtrap arrays~\cite{00:Cirac}. The spatial selectivity could be extended to three dimensions by crossing

941: the 423\nm\ and 389\nm\ beams. The isotope-selectivity would also be invaluable for loading ``refrigerator''

942: ions~\cite{02:Kielpinski} in schemes where continuous cooling is provided by one isotope, while quantum logic

943: operations are carried out using another isotope: following the estimate in~\cite{02:Blinov}, the decoherence rate, due

944: to off-resonant photon scattering, of a \CaII{43} qubit being cooled sympathetically by a \CaII{42} refrigerator ion is

945: \ish 0.1/s for a trap with a heating rate of $10^3$ motional quanta per second~\cite{00:Turchette}.

946: 

947: Finally, in comparison with the work of Mortensen \etal, it is clear that the 272\nm\

948: 4s$^2\,\lev{1}{S}{0}\leftrightarrow$ 4s5p\,\lev{1}{P}{1} transition is much more favourable for isotope-selective

949: ionization of calcium than the 423\nm\ transition used in this work, since it possesses a narrower natural width,

950: 8.5(5)\MHz~\cite{93:Mitroy}, and isotope shifts about a factor of two larger~\cite{02:Mortensen}. However, this scheme

951: requires a significantly more complex laser source than the blue diode laser used here, and is somewhat less efficient

952: since it involves excitation deeper into the continuum. An unfortunate circumstance at the time of writing is the

953: unavailability of further diodes at 423\nm\ from Nichia Corporation but we are optimistic that, with blue laser diodes

954: under development by several other manufacturers and research laboratories, the situation will improve.

955: 

956: 

957: \begin{acknowledgments}

958: We are extremely grateful to Dr.\ Hugo van der Hart and Ms.\ Claire McKenna for performing calculations of the

959: photo-ionization cross-section. We would also like to thank the ion-trapping group of Prof.\ Dr.\ Rainer Blatt at

960: Universit\"{a}t Innsbruck for many useful discussions, Dr.\ Charles Donald for compiling the electron bombardment

961: ionization data in figure~\ref{F:compensation}, and Wolfgang Kemp at Toptica Photonics for helpful advice. Graham

962: Quelch contributed invaluable technical assistance. The research is supported by the EPSRC, ARDA

963: (P-43513-PH-QCO-02107-1) and the E.U. QGATES network. MM acknowledges the support of the Commonwealth Scholarship and

964: Fellowship Plan. DML is a Royal Society University Research Fellow and wishes to thank Dr.\ Bruce Warrington for {\tt

965: SevenSegmentDisplay} software development.

966: \end{acknowledgments}

967: 

968: \bibliography{calcium}

969: \end{document}

970: