A pulsed dye laser for photo-ionization of magnesium atoms

Transcription

1 A pulsed dye laser for photo-ionization of magnesium atoms Master Thesis realized in March-June 2009 in the Laser Spectroscopy Division of the Max-Planck-Institute for Quantum Optics (Garching) by Camille Estienne, student in the Master of Physics at the Swiss Federal Institute of Technology (Zürich).

2 Supervisor in Zürich: Professor Ursula Keller Supervisor in Garching: Doctor Thomas Udem

3 Aknowledgements Thank you to Valentin Batteiger, Maximilian Hermann, Sebastian Knünz, Guido Saatho, Mariusz Semczuk from the Ions Traps Group and to Karl Linner and Wolfgang Simon from the workshop, for their help and advices during these four months. 3

4 Contents 1 Introduction 6 2 Theory of pulsed operation of dye lasers Laser operation from organic dye molecules Energy levels Triplet state absorption Principle of gain switching and rate equations Principle of gain switching Rate equations Numerical study Resonator design Types of wavelength selective resonators Choice of a resonator A simple model for self-seeding Experimental setup Pump laser and pump optics Pump power and pulse shape Pump optics Pump beam polarization Dye solution, dye cell and circulating pump Dye solution Dye cell Circulating pump Laser resonators Two-mirror cavity Prisms and birefringent crystal Dispersion Spectral narrowing Challenges Highly birefringent crystal Results Littman cavity Spectral narrowing Results Dual-cavity Results

5 4.5 Overview Second-harmonic generation and photo-ionization of magnesium atoms Second-harmonic generation Optimized confocal parameter Results Photo-ionization of magnesium atoms General remarks Proof of principle Conclusion 61 List of Figures 64 List of Tables 68 5

6 Chapter 1 Introduction This thesis reports on the demonstration of a compact and simple laser source for photo-ionization, to be used for ongoing [1, 2] and planned [3] Paul trap experiments. Ion traps are usually loaded by electron impact or by photo-ionization. Electron impact is a simple way to ionize atoms, yet it is a non selective process (other species can be ionized and trapped with the desired ions). This problem is avoided when a resonant multi-step photo-ionization process is used. The atoms are rst selectively excited on resonant transitions and then ionized. Hence, only the desired species will have a signicant ionization probability. This process also allows for isotope selectivity, if the spectral width of resonant transition and the line width of the exciting laser are both smaller than than the isotope shift. Magnesium atoms can be ionized through a two-step process using a dipole allowed transition at 285 nm as resonant excitation step. Unfortunately, light at 285 nm is not easily produced by simple solid state laser systems like diode lasers, while dye lasers can easily be tuned to emit at 570 nm. Then secondharmonic generation can be used to reach the desired wavelength. So far, a single mode continuous-wave dye laser (Coherent 699) was used for this purpose, in combination with a resonant second-harmonic generation step. This system is pumped by a bulky argon ion laser, it contains an elaborate resonator and a resonant enhancement cavity for frequency doubling. Furthermore, it will be used in near future for a lithium ion trap experiment. This thesis reports on the demonstration of a pulsed dye laser system pumped by a compact Nd:YAG laser emitting at 532 nm. The resonator uses rhodamin 6G as ecient laser medium and contains only few frequency selective elements. The high peak power obtainable in pulsed operation allows for ecient single pass frequency doubling without any servo-electronics. Such a pulsed dye laser may also serve as a prototype for a photo-ionization source for a planned beryllium ion trap experiment. Here the resonant transition at 235 nm could be reached by second harmonic generation from a coumarin dye laser, pumped with a solid state laser at 355 nm. 6

7 Chapter 2 Theory of pulsed operation of dye lasers 2.1 Laser operation from organic dye molecules Laser emission from organic molecules (like the one represented in Figure 2.1) was discovered in 1966 by J. R. Lankard and P. Sorokin at IBM Thomas J. Watson Research Center [4, 5, 6]. This discovery soon proved to be an important step in the developement of new laser sources, because of the availability of uorescent organic molecules covering an extremely broad spectrum, ranging from the UV to the IR. Furthermore, each dye molecule can emit over a wide frequency range within which the dye laser can be continuously tuned. This feature is very important for spectroscopic studies. Figure 2.1: Molecular structure of rhodamine 6G (from [6]) Energy levels All uorescent organic molecules resemble one another in the general arrangement of their energy levels (Figure 2.2). The ground state S 0 is a singlet state from which the molecule can be excited to the lowest excited singlet state S 1. Transitions between singlet states are dipole allowed transitions, giving rise 7

8 to the strong absorption bands of uorescent organic molecules. When the molecules decay to the ground state, uorescent light is emitted. Each electronic level of such a big molecule (the rhodamine 6G molecule is represented in Figure 2.1) is made of hundreds of sublevels associated with vibrations and rotations of the whole molecule. The continuous spectrum of uorescent radiation can span tens of nanometers. Non radiative decay from the high vibrational sublevels to the low vibrational sublevels is much faster than the uorescence lifetime. Furthermore, the high vibrational sublevels of the ground state will normally be unoccupied at room temperature. Hence, this system can be analyzed as a quasi-four-level system, and a population inversion can be built between low-lying sublevels of the rst excited singlet state and high-lying sublevels of the singlet ground state. This is the basis for laser action in organic molecules Triplet state absorption A potential problem arises from the existence of triplet states in the spectrum of organic dye molecules [4, 5, 6]. Molecules excited to the rst singlet state can decay, by a non-radiative process, to the lowest triplet state T 1 - a metastable state which lies below the rst excited singlet state. This process is characterized by an intersystem crossing rate k S,T which is an important parameter for laser action. The accumulation of molecules in the lowest triplet state has two consequences which are detrimental to laser action. First, the population of the rst excited singlet state is decreased, as the molecules are 'trapped' in a metastable state, and this increases the lasing threshold. Second, molecules in the lowest triplet state are now available for absorbing radiation corresponding to a triplet-triplet transition, like T 0 T 1 in Figure 2.2. Unfortunately triplettriplet absorption bands often overlap with the uorescence band, which can prevent laser action from occuring altogether. One way to overcome this problem is to pump the active medium with short pulses of radiation. If the threshold for laser action can be reached in a time which is small compared to the inverse of the intersystem crossing rate, the triplet state population can be kept very small. This condition will be satised for t threshold 1/k S,T (where t threshold is the time during which the population inversion rises from zero to the threshold level). For a typical value of k S,T = 10 7 s 1, we need t threshold 100 ns. This condition is easily achieved with a pulsed pump laser delivering pulses of 10 ns. Then all triplet eects can be neglected in a rst approximation. In the following, we assume that this condition is satised. 8

9 Figure 2.2: Energy levels of a typical dye molecule (from [5]). Heavy horizontal lines represent vibrational states. Lighter lines represent rotational states. Excitation is represented by the transition A b. Laser emission is represented by the transition B a. Other transitions (absorption to higher excited singlet states, intersystem crossing, triplet states absorption) represent losses in the laser process. 9

10 2.2 Principle of gain switching and rate equations Laser emission from organic dye molecules was discovered accidentally: Lankard and Sorokin were actually studying stimulated Raman emission from solutions of phtalocyanine dyes, which have a narrow absorption band near the wavelength emitted by a ruby laser. When a pulsed ruby laser was red at a cell containing a phtalocyanine solution, incipient laser action was observed instead of Raman emission [6]. Thus, the rst dye laser happened to be pumped with a pulsed laser source, a method known as gain switching Principle of gain switching Gain switching is a convenient method for obtaining short pulses of laser radiation. A pulsed source of light is used to modulate the gain of the active medium. During the short pump pulse duration, the population inversion in the active medium rises very rapidly above threshold inversion for laser action. A pulse of laser light can build up in the cavity and will quickly deplete the population inversion. As the pump pulse has only a short duration, the laser emission will cease after a short time, until the next pump pulse arrives Rate equations The dynamics of gain switching can be described by a set of two coupled rate equations (see [4, 7, 8]): dn(t) dt dq(t) dt = W (t) B n(t) q(t) n(t) τ (2.1) = V g B n(t) q(t) q(t) t c (2.2) where n(t) is the population inversion density, q(t) is the number of photons in the cavity, W (t) is the number of absorbed pump photons per unit time and per unit volume, τ is the uorescence lifetime, V g is the active volume, t c is the photon cavity lifetime and B is the Einstein coecient for stimulated emission. These equations can be written in a slightly dierent way if we use the threshold condition for the onset of oscillation from [9]: N t h ν l B = hν l t c (2.3) where the term on the left represents the energy delivered to the cavity per unit time by stimulated emission (N t is the threshold inversion, h is the Planck constant and ν l is the lasing frequency), while the term on the right represents the energy lost per unit time due to the nite cavity lifetime. If we susbtitute B = 1/(N t t c ) from Equation 2.3 in Equations 2.1 and 2.2 we obtain: 10

11 dn(t) dt dq(t) dt = W (t) n(t) = V g n(t) N t N t q(t) t c q(t) t c n(t) τ (2.4) q(t) t c (2.5) For simplicity, a Gaussian pulse envelope will be assumed for the pumping pulse: [ ( ) ] 2 t W (t) = W max exp 4 ln 2 (2.6) where t 1/2 is the full width at half maximum Numerical study Equations 2.4 and 2.5 can be solved numerically to obtain some insights on the dynamics of gain switching. A set of parameters corresponding to the experimental work to be presented in the next chapter has been used: a full width at half maximum of 20 ns for the pump pulse, a uorescence lifetime of 6 ns, a threshold population inversion density (calculated from [9]) on the order of cm 3 and a photon cavity lifetime of a few nanoseconds. The behavior of this system below and above threshold is clearly represented by the series of solutions presented in Figure 2.3. The inversion population depletion is very fast when the laser pulse builds up. When W max /n T is not too high, the output laser pulse is shorter than the pump pulse, and it shows a steeply rising slope. This dynamic eect has been used for generating short pulses of laser radiation [7]. Another interesting feature of gain switched oscillators is the appearance of relaxation oscillations for certain values of the parameters. The continuous evolution from a single-pulse regime to a regime with relaxation oscillations can be clearly followed in the series of solutions presented in Figure 2.4. t 1/2 11

12 Figure 2.3: Solutions of the rate equations for an increasing number of pump photons (from left to right and from top to bottom). Cavity length: 15 cm. Active region length: 0.5 cm. Pump pulse width: 20 ns. Blue curve: Pump pulse. Green curve: Population inversion. Red curve: Dye laser pulse. 12

13 Figure 2.4: Solutions of the rate equations for decreasing photon lifetime in terms of a decreasing cavity length (from left to right and from top to bottom). Active region length: 0.5 cm. Pump pulse width: 2 ns. W 0 /n T = 2. Blue curve: Pump pulse. Green curve: Population inversion. Red curve: Dye laser pulse. 13

14 2.3 Resonator design One of the most attractive properties of dye lasers is their tunability. The uorescence band of organic dyes spans several tens of nanometers (Figure 2.5). Within this uorescence band, laser emission can be observed over a wavelength range for which the probability for uorescence is high while the singlet absorption coecient is low (Figure 2.5). If the dye concentration is increased, the absorption will also be increased, and the maximum of the stimulated emission spectrum will be shifted towards longer wavelengths, because the balance of gain and losses becomes more favorable at these wavelengths [4]. Hence it is possible to tune a dye laser by changing the dye concentration or by changing the length of the active volume. But to achieve ne tuning and spectral narrowing it is necessary to introduce wavelength selective elements into the resonator. Figure 2.5: Singlet state absorption and uorescence spectra of rhodamine 6G (from [5]). Yellow line: Laser emission Types of wavelength selective resonators The rst wavelength selective resonator for a dye laser was proposed in 1967 by B. H. Soer and B. B. McFarland [10] who replaced one of the cavity mirrors by a diraction grating. Since then, dierent approaches have been widely used for wavelength selection in dye laser resonators: spatial wavelength separation by dispersion (using prisms) or by diraction (using gratings); interferometric ltering using constructive interferences in an etalon; rotational dispersion by propagation of a polarized light beam in a birefringent material. 14

15 The general idea is to modulate the balance of gain and losses within the stimulated emission band, in order to favour lasing in the selected wavelength interval only. For this purpose, several tuning elements can be used in the same cavity Choice of a resonator Two simple congurations seemed particularly attractive for our purposes. Prisms and birefringent lter The rst conguration consists of two plane mirrors and a series of prisms at Brewster angle together with an anti-reection coated highly birefringent crystal (like in [11]). This ensures minimum surface losses. We can think of a prism followed by a broadband feedback mirror as equivalent to a mirror with a frequency-selective coating. In the same way, a series of prisms arranged for additive dispersion is equivalent to a mirror with a highly frequency-selective coating. Additionally, if they are used at (or near) Brewster angle the prisms will also have an eect on the beam polarization: the losses by reection on the prisms surfaces will be higher for an s-polarized beam, and therefore the lasing threshold will be lower for a p-polarized beam. The laser beam should then be p-polarized. A p-polarized beam will be sensitive to the eect of a birefringent crystal which rotates the polarization direction of dierent frequency components by dierent angles. Hence a birefringent crystal surrounded by prisms at Brewster angle will form a birefringent lter. Grating in grazing incidence A second interesting conguration is a variant of the Littman cavity [13, 14] which uses a plane mirror and a grating in grazing incidence. When a grating is used in Littrow mount (diracted beam collinear with the incident beam) the resolution can be much improved by using a beam expander in the cavity. Early beam expanders consisted of a lens telescope (as in the Haensch design [12] sketched in Figure 2.6) but they were largely replaced by prism beam expanders. When the beam expander is inserted in the cavity, a larger area of the grating is illuminated, and hence the resolution is improved. Furthermore, the beam divergence is reduced, which also improves the wavelength selectivity of the cavity. Additionally, the beam expansion prevents the grating surface from being destroyed by too high laser intensities. It is also possible to illuminate a large area of the grating without any beam expander in the cavity, if the grating is mounted in grazing incidence. In the Littman conguration [13, 14], the rst order diracted beam is returned to the grating by an additional highly reecting tuning mirror (Figure 4.3). Because of the angular dispersion of the diracted beam, only a small wavelength range is reected back along the direction of incidence on the mirror. This back reected beam is diracted again by the grating and it is returned to the dye cell for further amplication. The laser can be tuned by changing the orientation of the 15

16 tuning mirror. The reection from the grating (zeroeth order diracted beam) is used to couple out the energy. Dual-cavity It is possible to lower the lasing threshold of a Littman-type cavity by adding a partially reecting feedback mirror along the path of the zeroeth order diracted beam [15, 16]. In this conguration, the wavelength selective Littman cavity seeds the longer broadband cavity at the chosen wavelength. This mode of operation is known as self-seeding. In this particular case, self-seeding is achieved in a dual-cavity conguration [15, 16]. This conguration consists of two coupled cavities (see Figure 2.8): a long cavity formed by mirrors M1 and M3 (zeroeth order reected beam), and a short cavity formed by mirrors M1 and M2 (rst order diracted beam). Each cavity can lase independently. As the long cavity has no frequency selective elements, it emits broadband laser light. The short cavity is a frequency selective Littman cavity emitting narrow band laser light, but it is not as ecient as the long cavity in extracting energy from the active medium. When the short cavity seeds the long cavity, narrow band laser emission is obtained with a good conversion eciency [15, 16]. In [15, 16] this approach was applied to a gain switched Ti:Sapphire laser. It should be possible to apply the same method to a gain switched dye laser. 16

17 Figure 2.6: Cavity using a grating in Littrow mount and an intracavity telescope for beam expansion (from [12]). Figure 2.7: Cavity using a grating in grazing incidence (from [13]). 17

18 Figure 2.8: Dual-cavity conguration (from [16]). 18

19 2.3.3 A simple model for self-seeding A simple model for self-seeding in a dual-cavity conguration has been proposed by Merriam and Yin [16]. The operation of this system is analyzed in terms of gain switching between two coupled cavities. Assuming that the laser oscillates in a single cavity mode, the rate equation for the number of photons in the cavity (Equation 2.2) can also be written in a slightly dierent way: dq(t) dt = V g σ F n(t) q(t) t c (2.7) where σ is the stimuated emission cross-section and F is the photon ux on the transverse cross-section of the active volume V g. The photon ux F can be expressed in the following way: F = q(t) σ g t c /2 = q(t) σ g c l c (2.8) where σ g is the active volume transverse cross-section, t c is the round-trip time and l c is the cavity length. If we put V g = σ g l g (where l g is the active volume length) and expression 2.8 for F into Equation 2.7, we obtain: dq(t) dt which is the form used in [16]. = l g l c σ c q(t) n(t) q(t) t c (2.9) Following the analysis proposed by Merriam and Yin [16] we can use Equation 2.9 to dene a cavity mode gain G: G = l g l c σ c n(t) 1 t c (2.10) The cavity mode gain G for one cavity can then be considered as a function of the inversion density n(t). If we plot G as a function of n(t) we obtain a straight line whose slope is dened by the relative gain length l g /l c and whose value for zero inversion density depends on the value of t c. These two parameters are characteristic of each cavity. The laser mode gain for a Littman-type cavity and a for a long broadband cavity are plotted in Figures 2.10 and 2.9 for dierent values of the critical parameters. Several important features are common to all these plots. The short Littman cavity has a higher relative gain length and hence a steeper slope. And it has a shorter cavity lifetime, because of the higher losses due to the relatively low grating eciency in rst order 1. The long broadband cavity has a smaller relative gain length but a longer cavity lifetime. With the chosen parameters, the curves cross in the positive gain region. The mode gain for the initial 1 The grating eciency in rst order is around 14 % in our case (measured). This gure has to be squared because the beam hits the grating twice per round-trip in a Littman conguration 19

20 inversion density 2 (represented by the vertical line on the right) is always higher for the short Littman-type cavity. Hence this cavity denes the initial laser mode that seeds the longer broadband cavity. As the inversion is depleted by light amplication, the operating point moves to the left. At some point (the middle vertical line) the cavity mode gain curves cross. The lasing mode can still be amplied by the long broadband cavity, but for some inversion density (represented by the vertical line on the left) the long cavity mode gain goes to zero and laser emission ceases. This simple model does not describe in detail the dynamics of mode competition in the dual-cavity, but it gives a qualitative understanding of the operation of this system. The series of curves represented in Figures 2.9 and 2.10 show the inuence of two parameters: the output coupler reectivity and the broadband cavity length. These two parameters could be varied easily in our setup. When the output coupler reectivity is increased, the curve corresponding to the broadband cavity (blue curve) is translated upward (Figure 2.9). As a consequence, the population inversion density that brings the broadband cavity gain to zero gets lower, and the energy extraction from the gain medium is increased. For a constant value of the output coupler reectivity, the crossing point is shifted towards smaller values of the inversion density when the broadband cavity length is increased (Figure 2.10). But the point where the broadband cavity gain reaches zero does not move signicantly in this case. 2 We have used an initial inversion density n initial = /cm 3 corresponding to the experimental work to be presented in the next chapter (assuming, as in [16], that the initial inversion density is produced by the total pump energy). 20

21 Figure 2.9: Laser mode gain versus population inversion density for increasing output coupler reectivity (from left to right and from top to bottom). Broadband cavity length: 31 cm. Wavelength selective cavity length: 20 cm. Grating eciency in rst order: 14 %. Red curve: Wavelength selective cavity. Blue curve: Broadband cavity. Vertical lines correspond (from right to left) to the initial inversion, the gain switching inversion and the inversion for zero mode gain. 21

22 Figure 2.10: Laser mode gain versus population inversion density for an increasing broadband cavity length (from left to right and from top to bottom). Output coupler reectivity: 20 %. Wavelength selective cavity length: 20 cm. Grating eciency in rst order: 14 %. Red curve: Wavelength selective resonator. Blue curve: Broadband cavity. Vertical lines correspond (from right to left) to the initial inversion, the gain switching inversion and the inversion for zero mode gain. 22

23 Chapter 3 Experimental setup 3.1 Pump laser and pump optics Pump power and pulse shape The pump laser is a compact frequency doubled Q-switched diode pumped Nd- YAG laser (Elforlight, FQ ) at 532 nm. It is triggered externally by a TTL signal and can be operated from single shot to 20 khz repetition rate, with an average power of 0.6 W at 1 khz and 2.5 W at 5 khz. The pulse duration varies between 15 ns and 30 ns, depending on the repetition rate (Figure 3.1 shows the pump pulse for a 100 Hz repetition rate). The energy per pulse and the peak power increase when the repetition rate is decreased (Figures and 3.3) because this gives more time to build up population inversion between the short pumping pulses. As the repetition rate is decreased below 1 khz, the pulse separation approximate the upper state lifetime of the lasing material ( 1 khz in YAG) and nally a limit is reached for increasing the energy per pulse. Figure 3.1: Pump pulse shape (repetition rate: 100 Hz) measured with a fast photodiode (500 MHz bandwith) and an oscilloscope with 100 MHz bandwith. 23

24 Figure 3.2: Energy per pulse of the pump laser for dierent repetition rates (calculated from measured average power and repetition rate). Figure 3.3: Peak power of the pump laser for dierent repetition rates (from the calculated energy per pulse and the measured pulse duration). 24

25 3.1.2 Pump optics The pump beam has a diameter of approximately 2 mm. The dye is pumped transversally (pump beam perpendicular to the output laser beam direction). The pump beam is focused into the dye cell by one or two cylindrical lenses. For optimum pumping we created a line focus in the dye cell by using a telescope consisting of two cylindrical lenses (focal lengths: -30 mm and mm, distance between the lenses: 1.5 cm). The pump beam was focused to a 20 mm long thin line, covering the whole length of the dye cell. The optimum distance between the lenses and the dye cell is not the position where the focus is exactly in the dye cell (Figures 3.4). When the set of two lenses is focused exactly in the dye cell, the amount of amplied spontaneous emission (amplication of uorescence along the direction of the active volume) is so high that it saturates the dye medium, and the dye laser output gets very low. This eect can be clearly seen in Figure 3.4: the output power of the dye laser is strongly reduced when the focus is exactly in the dye cell. At the same time the power of amplied spontaneous emission gets very high (this is not represented in Figure 3.4). For optimum output power of the dye laser, the dye cell has to be set slightly out of focus. The amount of amplied spontaneous emission by the dye is also substantially reduced by removing completely the diverging lens. When only the converging lens is used, the pump light is focused to a line of approximately 0.5 mm length. Two dierent pumping congurations were used: an horizontal pumping conguration (represented in Figures 3.5 and 3.6) and a vertical pumping con- guration (represented in Figures 3.8 and 3.7). For reasons that will become clear in the next section, the vertical pumping conguration is advantageous. 25

26 Figure 3.4: Saturation of the dye medium by amplied spontaneous emission. D: distance between the dye cell and the mount holding the two cylindrical lenses (f = - 30 mm and f = mm). 26

27 3.1.3 Pump beam polarization The pump beam is horizontally polarized. A λ/2 plate was used to rotate the polarization. The eect of the pump beam polarization on the polarization of the output beam from the dye laser is represented in Figures When the pump beam is polarized parallel to the plane containing the pump beam and the dye laser beam, the dipole moment of the molecules are oriented along the resonator axis. In this conguration, the molecules have to rotate before they can radiate along the resonator axis [17]. The resulting angular orientation distribution is isotropic, and the output beam is not polarized. But when the pump beam is polarized perpendicular to the plane containing the pump beam and the dye laser beam, the dipole moment of the molecule is oriented perpendicular to the resonator axis, and the output beam is polarized in the same direction as the pump beam. For our purposes, an horizontally polarized beam is preferred, for higher eciency of the grating and low losses at Brewster angle surfaces. Hence we needed a pump laser beam polarized horizontally and perpendicular to the plane containing the pump beam and the dye laser beam. If we want to keep the laser resonator in an horizontal plane, this is only possible in a vertical pumping conguration. For this purpose, the dye cell was tilted and we used a highlyreective broadband mirror set at 45 o to send the focused pump beam on the dye cell. 27

28 Figure 3.5: Polarization of the output beam when the pump beam is horizontally polarized and the plane of incidence is horizontal. Figure 3.6: Polarization of the output beam when the pump beam is vertically polarized and the plane of incidence is horizontal. 28

29 Figure 3.7: Polarization of the output beam when the pump beam is horizontally polarized and the plane of incidence is vertical. Figure 3.8: Polarization of the output beam when the pump beam is vertically polarized and the plane of incidence is vertical. 29

30 3.2 Dye solution, dye cell and circulating pump Dye solution The active medium is a solution of rhodamine 6G in ethanol. Rhodamine 6G is a red crystalline solid with a molecular weight of grams per mole [18]. In ethanol, its absorption maximum is at 530 nm and its uorescence maximum at 560 nm [18]. When pumped by the second harmonic of a Nd-YAG laser at 532 nm, it emits laser radiation tunable between 555 nm and 585 nm [18]. This is close to the spectral region where the human eye has its maximum wavelength sensitivity. Therefore, the color of the laser output can be clearly seen to vary from yellow-green to orange-red. Initially the concentration was 0.1 g/l ( mol/l) as indicated in [18]. The optimum concentration for emission at 570 nm was found to be 0.15 g/l ( mol/l) from a series of measurements with concentrations varying from 0.03 g/l ( mol/l) to 0.21 g/l ( mol/l). These results are presented in Section Dye cell The dye mixture is contained in a quartz cell with an inner length of 20 mm. The cell is simply a rectangular parallelepiped with no wedge angle. When the dye is illuminated by the pump beam, a strong spot is formed in a direction perpendicular to the cell walls (Figure 3.9), even in the absence of any additional feedback mirror. This spot should not be confused with amplied spontaneous emission, which needs no reecting surfaces and can also be present below threshold. Amplied spontaneous emission gives rise to a broad spot (Figure 3.9) whose direction is xed by the active volume of the dye. Its position can be changed by changing the position of the pump beam focus in the dye cell, but not by tilting the cell. By contrast, the laser beam formed within the dye cell produces a sharper spot whose position changes when the dye cell is tilted. Hence, with their 4 % reectivity, the two parallel quartz walls already form a two-mirror cavity. The spectrum of the laser beam formed within this small cavity is shown in gure It is characterized by peaks separated by 1.2 nm which are probably due to the small dierence in thickness ( 0.1 mm) between the two quartz walls. A wedged cell may have eliminated this undesirable beam. For most measurements, the dye cell was simply tilted, in order to separate the desired laser beam from the unwanted beam formed within the cavity. When the dye cell is tilted enough, the overlap of this beam with the active volume becomes very small and its intensity decreases, but it does not disappear completely. Hence a signicant fraction of the available gain is taken by this unwanted beam. 30

31 Figure 3.9: Amplied spontaneous emission (left) and laser output formed by the dye cell walls (right). Figure 3.10: Spectra of the amplied spontaneous emission and of the laser output formed by the dye cell walls. 31

32 3.2.3 Circulating pump The dye is circulated by means of a small centrifugal pump. At the beginning it produced a lot of bubbles in the dye cell. This problem was solved when the leaks were repaired (Figure 3.11). The pump is still a weak part ot the setup because it produces quite strong vibrations that we could not dampen. The relatively bad performances obtained concerning the spectral width of the laser could be due to these strong vibrations. Figure 3.11: Repairing the circulating pump. 32

33 Chapter 4 Laser resonators 4.1 Two-mirror cavity Lasing was rst observed in a simple cavity consisting of two plane mirrors: a broadband highly reective mirror and a partially reecting mirror for output coupling. An output coupler with 20 % reectivity was chosen from a comparison of the thresholds and slope eciencies obtained with 10 %, 20 % and 30 % reectivity output couplers (Figure 4.1). The threshold pump energy varies between 140 µj and 240 µj per pulse. For 20 ns pulses, this corresponds to a threshold peak power comprised between 7 kw (for a 20 % output coupler) and 12 kw (for a 10 % output coupler). The conversion eciency is 12 % for a 10 % output coupler and 22 % for a 20 % output coupler. 33

34 Figure 4.1: Output power vs. pump power of a two-mirror cavity with dierent output couplers. Horizontal pumping conguration. Repetition rate: 1 khz. Dye concentration: 0.1 g/l. 34

35 4.2 Prisms and birefringent crystal For spectral narrowing of the dye laser output, a series of equilateral SF11 prisms was used. Tuning is then achieved by tilting the highly reective mirror (which is simpler than rotating the prism) Dispersion SF11 glass has a particularly high refractive index. For comparison, wavelength dependance of SF11 and BK7 glasses are plotted in Figure 4.2. The refractive index of SF11 glass at 570 nm is 1.79 (1.52 for BK7). As a consequence, losses for a s-polarized beam are particularly high. The dispersion of SF11 glass is also high, as is indicated by an Abbe number of 25 (compare to 64 for BK7 glass). Figure 4.2: Wavelength dependence of the refractive index of SF11 glass ( red curve) and BK7 glass (blue curve) calculated from Sellmeier equation. From this we can calculate the angle of incidence for minimum deviation: i min = arcsin (sin 30 o n(570 nm)) = o (4.1) and the Brewster angle of incidence: i Brewster = arctan (n(570 nm)) = o (4.2) The prisms are set at minimum deviation. In this conguration, the incident and exit angles are equal, and both are close to Brewster angle. Hence the losses are minimized for a p-polarized beam. This eect is clearly seen with a pumping conguration which initially results in an unpolarized output beam (like in Figures 3.5 and 3.8). The output beam becomes 90 % p-polarized when a prism is inserted in the cavity, because the losses are strongly reduced for this polarization. 35

36 4.2.2 Spectral narrowing Let dα be the angular divergence of the laser beam in the cavity before any prism is inserted. The single pass spectral width dλ for m prisms arranged for additive dispersion at minimum deviation can be calculated from [19]: dλ = n dα 4 m tan i min dn dλ (4.3) where dα is the beam divergence and dn dλ is the dispersion at the wavelength of interest. Formula 4.3 was used for calculating the values given in Table 4.1 for a typical value of the beam angular divergence of 5 mrad 1 [17]. For comparison, the calculated spectral width for a set of three BK7 prisms at minimum deviation is 12 nm. The measured spectral widths are narrower by a factor of ten than the calculated single pass spectral widths (Table 4.1). Indeed Formula 4.3 gives the single pass transfer function of a series of prisms in the case of a passive cavity with no active medium. The linewidth of the gain-switched laser is somewhat smaller due to gain-competition. The best results were obtained when all prisms were set on the tuning mirror side of the dye cell. Number Single pass Measured of prisms spectral width (nm) spectral width (nm) Table 4.1: Comparison of single-pass spectral widths (λ=570 nm, dn dλ = (from Sellmeier equation), dα=5 mrad) and measured spectral width for SF11 prisms at minimum deviation. 1 The actual beam angular divergence has not been measured. 36

37 4.2.3 Challenges Even if the measured spectral widths obtained with a series of prisms looked quite promising, several problems were encountered with this setup. The main problem is that the spectral width measured with a given number of prisms can vary between large limits, for reasons which are not completely understood. For example, the spectral width obtained with one prism at minimum deviation varied between 0.3 nm and 0.9 nm from day to day. These variations could not be correlated clearly with changes in the setup. Another problem concerns the resonator length. For maximum output power, a short cavity is preferred. Hence a compact assembly of three prisms at minimum deviation was built (Figure 4.3). But this set-up cannot be used in combination with the birefringent lter (see Section 4.2.4). Figure 4.3: Compact arrangement of three equilateral SF11 prisms at minimum deviation. 37

38 4.2.4 Highly birefringent crystal A highly birefringent anti-reection coated crystal of yttrium vanadate (YVO 4 ) was inserted in the cavity containing the SF11 prisms. The crystal is 8 mm long with an entrance surface of 5 5 mm. The optical axis is in the plane of the entrance surface. At 570 nm the ordinary refractive index of YVO4 is 2.01 while the extraordinary refractive index is For the equilateral prisms used in this setup, the angle for minimum deviation is close to Brewster angle (Section 4.2.1). Hence the losses suered by the p-polarized component of the laser beam are low, and the laser beam is mainly p-polarized. The intensity of a p-polarized component after a 8 mm thick YVO4 crystal plate is shown in Figure 4.4. Optimum fringe contrast is achieved for an angle φ of 45 o between the optical axis and the direction of polarization. The combination of prisms at Brewster angle and a birefringent crystal with φ=45 o should act as a birefringent lter [20] without any additional polarization selective elements. Figure 4.4: Intensity of a p-polarized component after a 8 mm thick YVO 4 plate. The optical axis is in the plane of the plate. The angle φ between the optical axis and the direction of polarization is 45 o for the red curve, 30 o for the green curve and 20 o for the blue curve. Several congurations were investigated (Figures ). The conguration using three prisms on a single mount was not ecient. The conguration represented in Figure 4.8 is more ecient than the congurations represented in Figures In this case, there is a polarization selective element on each side of the birefringent crystal, and both elements are on the same side of the active medium. Hence the p-polarized component of the beam is selectively enhanced before and after the birefringent crystal, and only this mainly p-polarized beam is amplied by the active medium. 38

39 Figure 4.5: Wavelength selective resonator with one prism at minimum deviation and a birefringent YVO 4 crystal. Figure 4.6: Wavelength selective resonator with two prisms at minimum deviation and a birefringent YVO 4 crystal (two prisms on the same side of the crystal). 39

40 Figure 4.7: Wavelength selective resonator with three prisms at minimum deviation and a birefringent YVO 4 crystal. Figure 4.8: Wavelength selective resonator with two prisms at minimum deviation and a birefringent YVO 4 crystal (one prism on each side of the crystal). 40

41 4.2.5 Results Conversion eciency The results presented here correspond to an horizontal pumping scheme (Figure 3.6). For a two-prisms conguration like the one in Figure 4.8, the threshold pump energy is 350 µj per pulse (Figure 4.9). For 20 ns pulses, this corresponds to 17.5 kw threshold peak power. The conversion eciency is 3 % only. Spectral width Two spectra are shown here. The rst one (Figure 4.10) was obtained in a oneprism conguration (Figure 4.5) and the second one (Figure 4.11) in a two-prism conguration (Figure 4.8). In both cases, transmission fringes of the birefringent lter can be resolved. They are separated by 0.14 nm as predicted from Figure 4.4. The observed width corresponds to the resolution limit of the spectrometer ( 0.1 nm). Unfortunately it has not been possible to select one of those fringes, due to the problems encountered with the prisms. Smaller prisms could be used to allow more round-trips in the cavity. One may also choose a smaller crystal length since the free spectral range of the lter is not well adjusted to the spectral width of the prism laser. Additionally, a polarizer may be included in the cavity (as in [11]). Figure 4.9: Output power vs. pump power for a cavity containing two SF11 prisms at minimum deviation and a 8 mm thick YVO 4 plate. Repetition rate: 1 khz. Dye concentration: 0.1 g/l. 41

42 Figure 4.10: Spectrum of the laser output from a cavity containing one SF11 prism at minimum deviation and a 8 mm thick YVO 4 plate. Figure 4.11: Spectrum of the laser output from a cavity containing two SF11 prisms at minimum deviation and a 8 mm thick YVO 4 plate. 42

43 4.3 Littman cavity In this conguration [13, 14], the resonator consists of three optical elements: a highly reective feedback mirror, a grating in grazing incidence and a highly reective tuning mirror (Figure 4.12). Figure 4.12: Littman cavity. The diraction angles are given by the grating equation: mλ = d (sin θ m sin θ i ), (4.4) where m denotes the diraction order, d is the grating period, θ i and θ m are the incidence angle and the diraction angle of the m-th order respectively. Both angles are measured from the axis of incidence normal to the grating surface. The grating used in the present experiment is a 5 cm long holographic aluminum grating with 2400 groves/mm with optimized modulation depth for the visible range (Optometrics No ). Since the grove spacing is smaller than the wavelength only one diraction order is allowed by the grating equation. The rst negative order is retro-reected by a at aluminum mirror which covers the whole length of the grating Spectral narrowing The angular dispersion of the grating is obtained by the derivative of the grating equation with respect to λ at constant incidence angle θ i : dθ m d λ = m. (4.5) d cos θ m In Littman conguration the grating is used in rst negative order for both passes. Therefore the dispersion adds up to: dθ m,2 dλ = 1 d ( ). (4.6) cos θ m,1 cos θ m,2 The angular dispersion in Littman conguration becomes particularly large, since the diraction angle approaches 90 for the retro-reected beam. 43

44 4.3.2 Results Conversion eciency Threshold and slope eciency of a Littman cavity with horizontal pumping are plotted in Figure The threshold pump energy is 350 µj per pulse. For a 20 ns pulse this corresponds to 17.5 kw threshold peak power. The conversion eciency is 2 % only. The inuence of dye concentration on the conversion eciency was investigated in this pumping conguration. The results are presented in Figure An optimum conversion eciency of 3.6 % is obtained with a concentration of 0.15 g/l ( mol/l). With a repetition rate of 1.7 khz, the conversion eciency could be further improved to 5.4 %. The maximum output power obtained in this conguration is 54 mw (for a repetition rate of 1.7 khz). Figure 4.13: Output power vs. pump power of a Littman cavity. Horizontal pumping conguration. Repetition rate: 1 khz. Dye concentration: 0.1 g/l. 44

45 Figure 4.14: Output power vs. pump power of a Littman cavity for dierent dye concentrations. Horizontal pumping conguration. Repetition rate: 1 khz. 45

46 Threshold and slope eciency are substantially improved when a vertical pumping conguration is used (Figure 4.15). The threshold pump energy is 300 µj per pulse for a pump beam polarization colinear to the dye laser beam and 280 µj per pulse for a pump beam polarization perpendicular to the dye laser beam. For a 20 ns pulse this corresponds to 15 kw and 14 kw threshold peak power respectively. The conversion eciency is 9 % in the rst case and 11 % in the second case. The maximum output power obtained in this conguration is 110 mw (for a repetition rate of 2 khz). Figure 4.15: Output power vs. pump power of a Littman cavity. Vertical pumping conguration. Repetition rate: 1 khz. Dye concentration: 0.15 g/l. 46

47 Spectral width A typical spectrum of the laser output from a Littman cavity is shown in Figure The observed spectral width (less than 0.1 nm) corresponds to the resolution limit of the optical spectrum analyzer. Figure 4.16: Spectrum of the laser output from a Littman cavity. To get a better idea of the spectral characteristics of the laser, a Fabry-Pérot etalon was built with two partially reecting mirrors ( 96 % reectivity at 570 nm). This cavity has a nesse of 77( 2 ). One mirror was mounted on a translation stage. This allows a variation of the free spectral range from 150 GHz (distance between the mirrors: 1 mm) to 6 GHz (distance between the mirrors: 2.5 cm). At 570 nm this corresponds to a free spectral range varying between 0.2 nm and 8 pm. For operation of the Fabry-Pérot interferometer as a spectrometer, the free spectral range of the etalon must be larger than the signal bandwith. Assuming a signal bandwith just below the resolution limit of the optical spectrum analyzer ( 0.1 nm), this condition is satised for a distance of 1 mm between the mirrors. For higher resolution, the distance between the mirrors can be increased, as long as the free spectral range stays larger than the signal bandwith. When this condition is not satised anymore, the interference fringes produced by the Fabry-Perot interferometer get blurred. 2 For two mirrors with the same reectivity R, the nesse F is given by: F = π R 1 R (4.7) 47

48 The laser output is sent through a diverging lens (f=-50 mm) before going through the Fabry-Pérot etalon. Interference fringes are observed on a screen 1.5 m away where their spacing and width can be measured. A qualitative observation of the interference fringes produced by the Fabry- Pérot etalon already gives some interesting information about the spectral features of the laser. First of all, contrasted fringes are still observed with a 6 GHz free spectral range (8 pm at 570 nm). Hence we can assume that the spectral width of the laser is in the GHz range 3. Furthermore, the spectrum does not consist of a single line, but of a series of lines whith relative intensites varying from shot to shot. This feature has been observed by several researchers [5, 6] and is well illustrated by a picture in [6]. These lines can be resolved in the pictures presented in Figure But the main problem revealed by the Fabry-Pérot etalon is the frequency instability of the laser output (Figure 4.17). To reduce this instability, several sources of vibrations have been investigated and controlled: the pump laser fan has been disconnected from the pump laser head and installed on an anti-vibration holder; the pump laser power supply, which is vibrating strongly, has been removed from the optical table; vibrations from the circulating pump have been dampen as much as possible; the mount for the 45 o mirror in the vertical pumping conguration has been made as compact as possible. Nevertheless, uctuations in the frequency output of the laser are still quite strong. A more quantitative estimate has been obtained using the formulas derived in [22]. The spectrum of the laser output consists of close lines which are a few GHz broad. The eective spectral width is broadened to 5-10 GHz due to uncontrolled sources of vibrations. [21]. 3 For isotope selective photo-ionization of magnesium, a spectral width of 1 GHz is needed 48

49 Figure 4.17: Interference fringes produced by the laser output of a Littman-type cavity sent through a Fabry-Pérot etalon (Finesse: 77, Free spectral range: 100 GHz). 49

50 4.4 Dual-cavity A dual-cavity is formed by adding a partially reecting mirror to a Littmantype cavity (Figure 4.18). As described in Section the short frequency selective Littman cavity can seed the long broadband cavity - a process known as self-seeding. The short Littman-type cavity has already been described in the preceding section. The long broadband cavity consists of a highly reecting mirror and an output coupler (zeroeth order reection from the grating). Output couplers with 10 %, 20 % and 30 % reectivity were used in these experiments. Both cavities are coupled through the grating. Figure 4.18: Dual cavity conguration Results Self-seeding The self-seeding process is clearly illustrated in the spectra presented in Figure When the tuning mirror of the Littman cavity is blocked, the long cavity emits broadband laser light (blue curve). When the tuning mirror is unblocked, the broadband spectrum is replaced by a sharp spectrum (red curve) around the frequency dened by the short cavity. The observed width corresponds to the resolution limit of the spectrometer ( 0.1 nm). Conversion eciency Energy extraction from the active medium is also enhanced. Figure 4.20 shows the threshold and slope eciency for dierent values of the output coupler (for comparison, threshold and slope eciency of the simple Littman cavity are also plotted). As predicted by the simple model presented in 2.3.3, the output power increases with increasing values of the output coupler reectivity (see Figure 2.9). For the same dye concentration and pumping conguration, the threshold pump energy is lowered from 350 µj (for a simple Littman cavity) to 300 µj (for a dual-cavity). For 20 ns pulses this corresponds to 15 kw instead of 17.5 kw 50

51 threshold peak power. The conversion eciency is increased from 2 % (for a simple Littman cavity) to 6 % (for a dual-cavity with 30 % output coupler reectivity). It should be noted that these results were obtained with a dye concentration of 0.1 g/l, a repetition rate of 1 khz and in an horizontal pumping conguration. It is most probable that better results would be obtained with a dye concentration of 0.15 g/l, a repetition rate of 2kHz and a vertical pumping conguration. Figure 4.19: Self-seeding of a long broadband cavity by a short wavelength selective cavity. Broadband cavity length: 31 cm. Littman cavity length: 20 cm. Blue curve: spectrum of the laser output from the broadband cavity (tuning mirror blocked). Red curve: spectrum of the laser output from the dual cavity (tuning mirror unblocked). 51

52 Figure 4.20: Output power vs. pump power of a simple Littman cavity and a dual cavity with dierent values of the output coupler reectivity. Broadband cavity length: 31 cm. Littman cavity length: 20 cm. Horizontal pumping conguration. Repetition rate: 1 khz. Dye concentration: 0.1 g/l. 52

53 Spectral width The laser output from the dual-cavity has also been observed with the Fabry- Pérot interferometer. The spectrum shows the same features as with the Littman cavity: several lines a few GHz large and strong uctuations on the order of 5-10 GHz. The lines are sharper and fewer in the dual-cavity conguration. It is possible to select one of them by inserting an aperture or an etalon into the cavity. But this would improve the eective spectral width only if the uncontrolled uctuations in the frequency output could be reduced. 53

54 4.5 Overview A short overview of the results obtained with dierent cavity congurations is presented in Table 4.2. Cavity type Conversion eciency Spectral width Two mirror cavity 22 % 3 nm / 2.7 THz Two prisms + YVO4 3 % 0.2 nm / 185 GHz Littman cavity 2 % 0.01 nm / 10 GHz Dual-cavity 6 % 0.01 nm / 10 GHz Table 4.2: Comparison of conversion eciency and spectral width for dierent cavities. Horizontal pumping conguration. Repetition rate: 1kHz. Dye concentration: 0.1 g/l. Remarks Conversion eciency. In each case, the dye cell has been tilted, in order to separate the laser beam formed by internal reections in the dye cell from the considered laser output (see Section 3.2.2). As a consequence, the conversion eciency is lowered. Also, the results presented here have been obtained in an horizontal pumping conguration, with a 1 khz repetition rate and a dye concentration of 0.1 g/l. These parameters can be optimized for maximum conversion eciency (see Section 4.3.2). With a vertical pumping conguration, a 1.7 khz repetition rate and a dye concentration of 0.15 g/l, a conversion eciency of 11 % was reached in a Littman-type cavity. For this set of parameters, no results are available for the other congurations. Spectral width. In the cavity containing prisms and a highly birefringent crystal, the spectral width is limited by the prisms (see Section 4.2.5). In the Littman-type cavity and in the dual-cavity congurations, the spectral width is limited by uncontrolled frequency uctuations (see Sections and 4.4.1). 54

55 Chapter 5 Second-harmonic generation and photo-ionization of magnesium atoms 5.1 Second-harmonic generation Frequency doubling of the dye laser output is achieved by single-pass conversion in a beta barium borate (BBO) crystal. The good single-pass eciency is due to the high peak power produced by the pulsed dye laser Optimized confocal parameter The fundamental beam at 570 nm is focused in the BBO crystal by a converging lens. An optimized focal length is determined in this section. The confocal parameter is dened by (see Figure 5.1): b = ω 2 0 2π λ (5.1) where ω 0 is the beam waist and λ is the fundamental wavelength. We use the analytical expression given in [23] for the optimized confocal parameter b in the crystal: ( ) B 2 b crystal = l/ B + B 2 where the walk-o parameter B is dened by: (5.2) B(ρ, l, λ) = 1 2 ρ 2π n(φ, 570 nm) l λ (5.3) (ρ is the walk-o angle, n(φ, 570 nm) is the refractive index at 570 nm for the phase-matching angle φ, l is the crystal length, λ is the fundamental wavelength). 55

56 For a BBO crystal and an fundamental beam at 570 nm, the phase-matching angle φ is equal to 43.3 o ( 1 ) and the walk-o angle ρ to 4.8 o ( 2 ). From Equations 5.2 and 5.3 we obtain the optimized confocal parameter in the crystal: b crystal =7.2 mm. The optimized confocal parameter in air is given by: b air = b crystal /n(570)=4.3 mm. From Equation 5.1 we obtain: ω 0 =25 µm. By adjusting the focal length of the focusing lens, the fundamental beam can be focused to the desired spot size ω 0. Using the expression given in [24] we nd an optimal focal length of 14 cm if we assume a collimated beam with radius of 1.25 mm before the crystal. Figure 5.1: Second-harmonic generation of light by a laser beam focused in a non-linear material (from [24]) Results The output of a simple Littman-type cavity is sent into a BBO crystal ( mm). For the rst trials, the Littman cavity was pumped horizontally. The output beam of the dye laser was focused in the crystal with a converging lens (f=150 mm, distance to the crystal center: 16 cm). About 100 µw of light at 285 nm were obtained from 54 mw in the fundamental. The conversion eciency was less than 0.2 % in this case. 1 The phase-matching angle φ (angle between the crystal axis and the direction of propagation of the beam in the crystal) is found from: n(φ, 285 nm) = n ord (570 nm) (5.4) where n ord (570 nm) is the ordinary refractive index at 570 nm and n(φ, 285 nm) is given by: n(φ, 285 nm) = ( cos φ n ord (285 nm) 1 ) 2 ( ) + sin φ 2(5.5) n ext (285 nm) (n ord is the ordinary and n ext the extraordinary refractive index. For BBO: n ord (285 nm) = 1.74 and n ext(285 nm) = 1.60 (from Sellmeier equations)). 2 The walk-o angle ρ is calculated from: ( ) [ n(φ, 285 nm) 2 ρ = arctan 2 1 n ord (285 nm) 2 1 n ext(285 nm) 2 ] sin (2φ) (5.6) 56

57 In the actual setup, the Littman cavity is pumped vertically. For optimum focusing of the fundamental in the non-linear crystal, a second lens (f=200 mm, distance to the crystal center: 10.5 cm) has been added added. In this conguration, about 700 µw of light at 285 nm were obtained from 180 mw in the fundamental 3. The conversion eciency is 0.4 %. 5.2 Photo-ionization of magnesium atoms General remarks Magnesium atoms are resonantly ionized through a two-step process. The resonant excitation step is the 3 1 S P 1 dipole allowed transition at nm. An atom in the 3 1 P 1 level can be ionized by a 285 nm photon from the ionization laser as well as by a 280 nm photon from the cooling laser 4. The saturation intensity for the resonant transition is 4401 W/m 2, which is easily achieved even with a cw-laser. Above saturation intensity the excited state population appoaches 0.5 for both, pulsed and cw-excitation. Once the resonant transition is saturated, the ionisation may be treated as a single photon process. It should then be equally ecient for equal average power of a pulsed and cw laser, unless saturation eects of the second step becomes signicant for a pulsed laser source. The natural bandwith of the 3 1 S P 1 transition is 78.2 MHz and the isotope shift between 26 Mg + and 24 Mg + amounts to 1.4 GHz [21]. The magnesium atoms are evaporated from an atomic oven. The eective linewidth due to Doppler broadening is about 3 GHz. The pulsed dye laser has the advantage that it matches the Doppler width of the resonant transition. In contrast, the continuous-wave dye laser used so far for photo-ionization has to run on a single mode to generate an error signal for the resonant frequency doubling cavity. Its spectral width is few MHz, therefore it can adress only a small set of velocity classes. On the other hand, it is possible to arrange the atomic beam and the photo-ionization laser perpendicular to each other, which reduces the Doppler width. A narrow bandwidth photo-ionization laser may then allow for isotope-selective photo-ionization. Presently, the present spectral width of the frequency doubled pulsed dye laser is too large for this purpose Proof of principle After a rough astigmatism compensation with a set of two cylindrical lenses, the frequency doubled output of the pulsed dye laser was overlapped with the cooling laser beam and directed to the linear Paul trap center. Although the frequency had been adjusted the day before, magnesium ions could be trapped immediately, which already indicates good long term frequency stability, despite the aforementioned short term ucutations on the GHz level. However, dark ions 5 were also loaded, conrming the absence of isotope selectivity. 3 In this case, the laser beam formed by reection on the dye cell walls was superimposed to the laser beam from the Littman cavity, for maximum output power. The laser output was still smoothly tunable. 4 The ionization energy of neutral magnesium is 7.7 ev. The energy of a photon at 285 nm is 4.3 ev. Hence 3.4 ev are needed for the second step (maximum wavelength: 364 nm). 5 Dark ions are most probably 25 Mg and 26 Mg ions. 57

58 Fluorescence photons from trapped Mg + ions were collected with a f/2 imaging system attached to a single photon camera (Quantar Mepsicron II). The imaging system is an electron multiplied CCD camera (Andor DUV A- DV885 LC-VP) allowing single ion resolution. Images of the rst magnesium ions trapped with the frequency doubled pulsed dye laser are presented in Figures 5.2 and

59 Figure 5.2: Magnesium ions loaded in a linear Paul trap with the frequency doubled pulsed dye laser. An increasing number of magnesium ions is trapped after the trap potential is switched on (from left to right and from top to bottom). Intense points at the center of the picture correspond to trapped and cooled 24 Mg ions. Dark sites within the chain correspond to 25 Mg and 26 Mg ions, which are not adressed by the cooling laser. 59

60 Figure 5.3: Crystal of 24 Mg ions loaded in a linear Paul trap with the frequency doubled pulsed dye laser. 60