Frequency stabilised grating feedback laser diode for atom cooling applications

Transcription

1 Optical and Quantum Electronics 31: 417±430, Ó 1999 Kluwer Academic Publishers. Printed in the Netherlands. 417 Frequency stabilised grating feedback laser diode for atom cooling applications A. G. TRUSCOTT, N. R. HECKENBERG AND H. RUBINSZTEIN-DUNLOP Centre for Laser Science, Department of Physics, The university of Queensland, Brisbane, Queensland 4072, Australia Received 1 December 1997; accepted 24 June 1998 Abstract. The free running linewidth of an external cavity grating feedback diode laser is on the order of a few megahertz and is limited by the mechanical and acoustic vibrations of the external cavity. Such frequency uctuations can be removed by electronic feedback. We present a hybrid stabilisation technique that uses both a Fabry±Perot confocal cavity and an atomic resonance to achieve excellent short and long term frequency stability. The system has been shown to reduce the laser linewidth of an external cavity diode laser by an order of magnitude to 140 khz, while limiting frequency excursions to 60 khz relative to an absolute reference over periods of several hours. The scheme also presents a simple way to frequency o set two lasers many gigahertz apart which should nd a use in atom cooling experiments, where hyper ne ground-state frequency separations are often required. Key words: atom cooling, diode lasers, frequency stabilisation 1. Introduction Developments in semiconductor technology have led to commercially produced, single longitudinal mode tunable laser diodes. While these laser diodes have been shown to be useful for research applications (Camparo 1985), their spectral quality is quite poor. Typical free running linewidths for such a single mode device are around 30 MHz, set by the Schawlow±Townes limit. In order to combat the large linewidth associated with free running laser diodes many stabilisation schemes have been developed. These schemes have been optical or electronic, or a combination of both. The main drawback to an all electronic stabilisation scheme is the wide bandwidth electronics required (Saito et al. 1985; Yamamoto et al. 1985). Optical stabilisation is a more practical solution, especially given the susceptibility of a laser diode to optical feedback. The e ect of optical feedback from an external frequency selective re ector is two fold: it increases the quality factor of the laser's resonator and thus decreases the linewidth, and it provides a preferred oscillation frequency. Many di erent optical feedback stabilisation schemes exist and rely on feedback from Fabry±Perot cavities (Dahmani et al. 1987), phase conjugate

2 418 A. G. TRUSCOTT ET AL. mirrors (Cyr et al. 1991), atomic vapours (Cuneo et al. 1994) or di raction gratings (Macadam et al. 1992). Each scheme has its own virtues, however, the simplest to implement is re ection from a di raction grating in a Littrow con guration. Such a con guration narrows the free running laser linewidth to a few megahertz. Re nements of mechanical designs have led to many improvements in laser stability, however such schemes will always be susceptible to perturbations of the external cavity length whether these perturbations be thermal, acoustic or vibrational in nature. Such perturbations signi cantly broaden the external cavity diode laser linewidth, so an external electronic feedback system is usually used to counteract these perturbations. Previous work on frequency stabilisation of external cavity diode lasers has included electronic feedback systems which use a Fabry±Perot cavity resonance (Maki et al. 1993; Hilico et al. 1994), or an atomic resonance (Maki et al. 1993; Shevy et al. 1993; Utako Tanaka and Tsutomu Yabuzaki 1994; Chuan Xie et al. 1989) as a frequency discriminator. Both these discriminators can produce error signals directly, however this only enables the laser frequency to be locked to the side of the resonance. To lock the laser frequency to the centre of the resonance one must be able to detect small frequency variations about line centre. A common method used to monitor frequency uctuations about the centre of the resonance peak is to detect the phase of a modulated error signal. These modulation techniques (Ho Seong Lee 1990) result in some modulation of the laser frequency that one is trying to stabilise. An external cavity diode laser allows three direct methods of frequency control: grating angle, diode temperature and injection current. Some electronic feedback systems use the piezoelectric transducer (PZT) which controls the grating angle as a feedback controller, however such a system cannot remove vibrational frequencies higher than a few hundred hertz, due to the limited frequency response of the PZT. Also, if the external cavity has a mechanical resonance that is driven by the PZT, the servo cannot correct for frequencies above this resonance (Schremer and Tang 1990). Other feedback systems (Ho Seong Lee 1990) have varied the laser temperature to counter frequency changes, though thermal time constants are usually long and therefore limit the frequencies which can be corrected for. The injection current is a good candidate for a feedback controller (Maki et al. 1993; Hilico et al. 1994) as it has a fast response time and is easy to control. However, the injection current does not tune inde nitely and therefore can not achieve long term stability. In this paper we report on an electronic feedback system that stabilises the frequency of an external cavity diode laser without the use of direct laser frequency modulation, incorporating both a Fabry±Perot resonance

3 LASER DIODE FOR ATOM COOLING APPLICATIONS 419 and an atomic resonance as the frequency reference. This technique provides both short and long term stability: the laser linewidth is reduced from 1.4 MHz to 140 khz and laser frequency excursions are limited to 60 khz relative to an absolute reference over a period of several hours. The scheme also provides accurately controlled tuning of the stabilised laser frequency over a 50 MHz range, as well as a simple method to frequency o set two external cavity diode lasers many gigahertz apart. The combination of these features make this stabilisation technique ideal for laser cooling and trapping. 2. Experimental set-up The central aim of this paper is to present an electronic feedback system that combines well with a grating feedback external cavity diode laser (ECDL). The feedback system uses very little light to achieve a high degree of stabilisation. This idea ts in well with our overall aim for the laser system, that is, to design a system that while robust and stable, allows most of the laser output to be used in atom optics experiments. Furthermore, the system is exible in allowing frequency o sets, fast frequency switching and tuning about the lock point. ECDL design and operating parameters: We use a 50 mw (SDL-5401-G1) single mode laser diode with a centre frequency speci ed by nm. A thermoelectric cooler controls the temperature of the mounting block, which is monitored by a nearby thermistor. A temperature controller using a PID feedback system regulates the temperature of the diode to 200 lk. A holographic di raction grating di racts 10% of the output beam back into the laser. The grating angle can be tuned via a stacked PZT, which allows ne electrical scanning of the laser wavelength. A hermetically sealed aluminium housing isolates the laser and its external cavity from the outside environment. The optical table on which the system sits is not vibrationally isolated, and the experiment in which it will be used is by necessity located near a noisy air conditioning plant. The laser is placed on a thick pad of sorbothane to reduce the vibrational coupling between the laser and the table. An ultra low noise current controller supplies the injection current for the laser diode (Libbrecht and Hall 1993). The system as described has a linewidth of 1.4 MHz when locked to a Rb saturated absorption peak and a long term free running frequency drift of 5 MHz/min. It can be scanned over a range of 9 GHz without a mode hop. The tuning range with mode hops is about 6 nm although this range may contain inaccessible spectral regions. The usable output power is at least 40 mw.

4 420 A. G. TRUSCOTT ET AL OPTICAL SET-UP The optical set-up for the electronic feedback system is shown in Fig. 1. The experimental set-up can be broken into three main parts (I) the Fabry±Perot confocal cavity, (II) the double passed acousto optic modulator AOM and (III) the Rb saturated absorption set-up. Part (I) is responsible for taking out noise frequencies up to 100 khz, while parts (II) and (III) lock the laser to a xed reference (a Rb Doppler free resonance) and allow tuning of the laser frequency about this reference. A half wave plate and polarising beam splitting cube (PBSC) are used to vary the power to the stabilisation system. In general about 1 mw is re ected into the system by the PBSC. The other light required for stabilisation is provided by re ection o an AR coated slide, BS 1, and is on the order of a few hundred microwatts. Part (I): The Fabry±Perot Confocal Cavity. The light re ected o BS 2 and incident on the confocal cavity has a power on the order of 100 lw. The transmitted light from the cavity is monitored by a photodetector PD 1. The Fig. 1. Optical set-up used for frequency stabilisation of an external cavity grating feedback diode laser. M 1 M 7 are mirrors, BS 1 BS 4 are beamsplitters and PD 1 PD 3 are photodetectors.

5 LASER DIODE FOR ATOM COOLING APPLICATIONS 421 cavity resonance as monitored by PD 1 has a height of 800 mv and a slope in its linear region of 100 MHz/V. In order to lock the frequency of the laser to the cavity the voltage to the PZT is changed to bring the cavity into resonance with the laser frequency. The voltage from PD 1 is then integrated with respect to a voltage reference and fed back to the laser injection current. The time constant for this integration is 9 ls and the loop bandwidth is 40 khz. At frequencies below 10 Hz the loop gain is about 70 db, this rolls o as 1/f for higher frequencies. Part (II): The AOM double pass. To achieve a frequency o set into the lock an acousto-optic modulator (AOM) is used. The AOM was not AR coated for 780 nm, but with careful alignment 88% of the input beam could be shifted to the rst order. Unfortunately this angle is directly proportional to the radio frequency (RF) and thus the frequency shift of the light. This spatial displacement of the beam with frequency will cause intensity variations to occur at a xed location as the RF is tuned. In order to eliminate this e ect, an alignment known as a double pass (Camy et al. 1983) is implemented (see Fig. 1). The output beam from the double pass has a frequency m ) 2d (where m is the laser frequency and d the RF) and does not move when the RF is changed. The power in this beam is about 60% of the original beam. In this system the double pass beam (DPB) is used as the pump beam for the saturated absorption set-up, as is explained in the next section. Part (III): Saturated Absorption. Saturated absorption of an atomic species is well understood (Nakayama 1985; Nakayama 1997). However, the experimental set-up used here di ers from the more usually accepted arrangement, in that the pump beam undergoes a frequency shift relative to the probe beam. This optical setup is used to avoid unwanted amplitude modulation of the probe beam. Saturated Absorption with a frequency shifted pump beam. In our saturation absorption set-up we employ two counterpropagating beams; a weak probe beam of frequency m and a strong pump beam of frequency m ) 2d. These two beams intersect in a 10 cm long rubidium cell as shown in Fig. 1. A third beam is also passed through the cell to allow subtraction of the Doppler background from the saturated absorption spectrum. In this con guration only atoms in a velocity class that Doppler shift the pump beam by an amount d and the probe beam by an amount )d will interact with both beams simultaneously. This leads to a saturated absorption spectrum that is shifted by a frequency d and that can be tuned by simply varying d. A lockin technique is used to lock the laser to the centre of a saturated absorption peak. As this method requires a small modulation to generate an

6 422 A. G. TRUSCOTT ET AL. error signal, modulation of the voltage controlled oscillator (VCO) RF is used here to modulate d. Ideally this causes only a small deviation of the laser frequency around the peak and should work well but, in practice, the VCOs available exhibited amplitude modulation as well as frequency modulation when the control voltage was sinusoidally varied. This is of course detrimental, as this amplitude modulation provides a modulation that the lockin ampli er will detect and yet is not related to the frequency of the laser. A way around this problem is to frequency modulate only the pump beam. Spurious amplitude modulation on the pump beam will not e ect the signal as long as the modulation is not large enough to unsaturate the transition. In this con guration a frequency o set of up to d = 50 MHz is possible, while no noticeable e ects are observed from unwanted amplitude modulation of the AOM LOCKING THE LASER FREQUENCY TO A SATURATED ABSORPTION PEAK A lockin technique is used to lock the laser frequency to a saturated absorption peak. A small (80 mv, 40 khz) sinusoidal oscillation is added to the VCO control voltage which in turn modulates the frequency of the pump beam by 200 khz. A lockin ampli er synchronously detects this signal to provide an error signal. This error signal is then integrated and fed back to the piezoelectric element of the confocal cavity. The time constant of the integrator used is s = 0.07 s which limits the loop bandwidth to around 14 Hz. At this stage the laser is locked to the confocal cavity which is in turn locked to an absolute frequency reference. Thus it would seem that the system should stay locked inde nitely, the confocal cavity part of the lock correcting for high frequencies and the saturated absorption part correcting for any absolute long term drifts. Unfortunately this is not the case. Laser diodes do not tune continuously with injection current (Camparo 1985) because the tuning range of a laser diode as a function of current consists of discrete modes. The laser diode system used here has a continuous current tuning range without a mode jump of 450 MHz. These discontinuities in the laser tuning curve limits the locking time of the above stabilisation system. In order to combat this e ect a third feedback loop was added to the stabilisation system. The third feedback loop uses the grating of the external cavity as its feedback element. The tuning range of the grating for the system is up to 9 GHz and so provides for much longer locking times. The third feedback loop uses the integrated signal of the current loop as its error signal. Integrating this signal and feeding it back in the right direction to the

7 LASER DIODE FOR ATOM COOLING APPLICATIONS 423 grating e ectively subtracts the very long term frequency uctuations out of the current loop. The time constant for this third feedback loop is s = 62.8 s. The nal feedback system locking time is only limited by the tuning range of the PZT. In principle this means that in a well air conditioned laboratory one should expect the system to lock inde nitely. A schematic ow diagram for the feedback system is given in Fig 2. Fig. 2. Schematic diagram for three feedback loops used to stabilise the laser diode frequency.

8 424 A. G. TRUSCOTT ET AL. 3. Results 3.1. NOISE SPECTRAL DENSITY In the present application the noise spectral density is determined via sampling of the error signal obtained from the confocal cavity. The slope of the cavity resonance converts any frequency deviation into an amplitude modulation which is then monitored by a fast (up to 1 MHz) photodetector. The contribution of amplitude noise to this signal is negligible. The photodetector used is independent of the feedback system and has a frequency discrimination of 50 MHz/V. The resulting spectral noise density spectrum of both the locked (B) and unlocked (A) laser system is shown in Fig. 3. A direct comparison of curves (A) and (B) show that the feedback loop reduces the laser frequency noise for frequencies below 30 khz. At around 30 khz a mechanical resonance of the Fabry±Perot cavity exists which produces a noise spike. This resonance limits the gain of the loop as oscillation occurs at this frequency when the gain is too large. For frequencies above 30 khz and below 100 khz the noise of the locked system is in fact a little larger than the unlocked system. The noise reduction achieved by the feedback loop is almost 40 db at frequencies below 10 Hz and rolls o as 1/f up to 10 khz. The mechanical resonance of the external cavity laser diode located at around 1 khz is reduced by 20 db. The spikes in the located spectrum are Fig. 3. Spectral noise density spectrum for the locked laser (B) and for the unlocked system (B).

9 LASER DIODE FOR ATOM COOLING APPLICATIONS 425 the result of mains noise and its harmonics. Note that the performance of the system is limited by the noise of the Fabry±Perot frequency discriminator. The noise in this discriminator is comprised of laser amplitude noise (about 1 part in a thousand) and noise in the Fabry±Perot peizo driver. In the present application it is the peizo noise that limits the stability of the system ALLAN VARIANCE The noise spectral density S m f of frequency is an unambiguous description of the oscillator noise and can be used to calculate the Allan variance [19] r 2 y s ˆ 2 pm 0 s 2 Z 1 0 S m f sin 4 pf s df ; 1 where m 0 is the oscillator centre frequency. Using the noise spectral density spectrum Fig. 3 and Equation 1 the square root of the Allan variance for the system can be calculated (Fig. 4). Once again curve (A) is for the unlocked system and curve (B) is for the locked system. Comparing the two curves we see no signi cant di erence for averaging times of 10 )5 s, while for longer averaging times the improvement is considerable. The best frequency stability of 2 10 )12 is attained for an Fig. 4. Square root of the Allan variance calculated from spectral noise density spectrum of an unlocked laser (A) and for the locked system (B).

10 426 A. G. TRUSCOTT ET AL. averaging time of 10 )1 s. The oscillations at lower integration times in curve (B) are due to the large mains frequency spikes and associated harmonics in the noise spectral density MEASUREMENT OF LINEWIDTH In order to measure the linewidth of the system two lasers were locked to the same confocal Fabry±Perot cavity and their heterodyne beatnote measured. The resultant beatnote of the system averaged over a 2 s period had a 3 db width of 200 khz (as shown in Fig. 5), which implies a FWHM of 140 khz for a single laser system. The resolution bandwidth for the measurement was 100 khz FREQUENCY OFFSET LOCKING Frequency o set locking can be achieved using sophisticated electronics (Maki et al. 1993), however a simpler and more useful method is to lock two lasers to the same optical cavity. In this way the tuning of the di erence Fig. 5. Heterodyne beatnote of two lasers stabilised to the same Fabry±Perot cavity. The resolution bandwidth for the measurement was 100 khz. The beatnote has a 3 db width of 200 khz.

11 LASER DIODE FOR ATOM COOLING APPLICATIONS 427 frequency can be achieved by varying the lock point of either laser. For the Fabry±Perot cavity used here this would amount to a frequency tuning of around 50 MHz. Further, this method has the advantage of allowing large frequency o sets since the lasers need not be locked to the same cavity resonance. In the case presented here the free spectral range of the cavity is 3 GHz and thus enables frequency o sets of any multiple of 3 GHz. This type of large frequency o set is hard to reproduce using electronic techniques, since wide band electronics are required. An obvious use for such a system is in the eld of atom cooling and trapping where two optical frequencies are utilised (Chu et al. 1985). The frequency di erence between the two lasers is the ground state hyper ne splitting frequency. In the case of rubidium 85 this would amount to a frequency o set of GHz. This is easily achievable using a confocal cavity of length 4.94 cm, while the error due to length inaccuracies will only be D(FSR) = FSR (DL/L). Thus an error of 100 lm in a 4.94 cm long cavity will only change the cavity free spectral range by 6 MHz which can be easily compensated. The optical set-up for frequency o set locking is shown in Fig. 6 and consists of light from two di erent lasers entering the confocal cavity from opposite directions. The cross-talk between the two systems is around one part in a thousand and is strongly limited by the polarisation quality of the PBSC's. However, this level of noise has no noticeable e ect and is much less than other noise present in the system ESTIMATION OF LONG TERM STABILITY In order to measure the long term stability of the system a second independent absorption spectroscopy setup is used as an absolute reference. The laser frequency is slightly detuned from resonance and thus can be made to sit to one side of an atomic peak. This signal is sampled at 100 Hz to estimate the long term drift of the laser frequency and locking time. The resulting spectrum taken over a six hour period is shown in Fig. 7. The frequency uctuations over this period are at most 60 khz. Thus when the laser is locked one can expect the laser centre frequency to be at least within 130 khz of the atomic resonance, allowing for the 140 khz Lorentzian linewidth of the laser FREQUENCY SWITCHING In many atom trapping applications the frequency of the laser needs to be changed asynchronously. Furthermore, it is usually desirable that the switching time be minimised. The locking method presented here allows fast, controlled switching of the laser frequency. Switching is achieved by quickly

12 428 A. G. TRUSCOTT ET AL. Fig. 6. Optical setup for locking two diode lasers to the same Fabry±Perot cavity. altering the length of the confocal cavity via application of a voltage pulse to the Fabry±Perot PZT. The PZT which controls the cavity length can respond in a time at least as short as 100 ls. The laser frequency will immediately follow the confocal cavity, due to the fast response of its servo loop. To prevent the saturated absorption feedback loop from correcting for this frequency change, a synchronous pulse is sent to the AOM which tunes the saturated absorption locking frequency to the new frequency. In this way frequency shifts up to 50 MHz can be achieved in a switching time of 100 ls. 4. Conclusion The frequency locking scheme presented here uses two frequency discriminators to frequency stabilise an external cavity diode laser. A Fabry±Perot

13 LASER DIODE FOR ATOM COOLING APPLICATIONS 429 Fig. 7. Frequency excursions of the locked laser system obtained by tuning the laser frequency to the side of a saturated absorption peak and monitoring amplitude uctuations. cavity is used to counter frequency uctuations up to 30 khz while a saturated absorption peak provides an absolute frequency reference for the cavity to eliminate any long term drift. The laser linewidth is measured using a heterodyne technique and found to be 140 khz. The long term frequency drift as measured over a six hour period on the side of an independent saturated absorption peak is found to be less than 60 khz, while the square root of the Allan variance for an averaging time of 10 )1 s is almost two orders of magnitude below that of the free running laser. The scheme uses no direct modulation of the diode laser frequency but does allow the system to be tuned up to 50 MHz while still in lock. Furthermore, switching of the laser frequency in 100 ls is possible. The system also allows the implementation of a frequency o set locking scheme that may be found useful in atom cooling and trapping experiments. The technique allows lasers to be locked with gigahertz laser centre frequency di erences, while still allowing megahertz tunability. 5. Acknowledgements This work was supported by the Australian Research Council. We would also like to acknowledge the expert technical assistance of the Physics Department workshops.

14 430 A. G. TRUSCOTT ET AL. References Camparo, J.C., Contemporary Phys. 26(5) 443, Saito, S., O. Nilsson and Y. Yamamoto, IEEE J. Quantum Electron. QE-21(12) 1919, Yamamoto, Y., O. Nilsson and S. Saito, IEEE J. Quantum Electron. QE-21(12) 1919, Dahmani, B., L. Hollberg and R. Drullinger, Opt. Lett. 12(11) 876, Cyr, N., M. Breton, M. Teà tu and S. Threà iault, Opt. Lett. 16(17) 1298, Cuneo, C.J., Je ery J. Maki, and D. H. McIntyre, Appl. Phys. Lett. 64(20) 2625, Macadam, K.B., A. Steinbach and C. Wieman, Am. J. Phys. 60(12) 1098, Je ery J. Maki, N.S. Campbell, C.M. Grande, R.P. Knorpp and D.H. McIntyre, Op. Commun. 102(3,4) 251, Hilico, L., D. Touahri, F. Nez and A. Clarion, Review of Scienti c Instruments , Shevy, Y., J. Kitching and A. Yariv, Opt. Lett. 18(13) 1071, Utako Tanaka and Tsutomu Yabuzaki, Jpn. J. Appl. Phys. 33(3B) 1614, Chuan Xie, Lilei Wang, Li Chen, Linzhen Xie and Yiqiu Wang, Appl. Opt. 28(21) 4552, Ho Seong Lee, Cha Hwan Oh, Sung Hoon Yang and Nak Sam Chung, Review of Scienti c Instruments. 61(9) 2478, Schremer, A.T. and C.L. Tang, IEEE Photonics Technology Letters. 2(1) 3, Libbrecht, K.G. and J.L. Hall, Review of Scienti c Instruments. 64(8) 2133, Camy, G., D. Pinaud, N. Courtier and H. C. Chuan, Revue de Physique Appliquee. 17(5) 357, Shigeru Nakayama, Jpn. J. Appl. Phys. 24(1) 1, Shigeru Nakayama, Physica Scripta. T70 64, Barnes, J.A., A.R. Chi, L.S. Cutler, D.J. Healey, D.B. Leason, T.E. Mcgunigal, J.A. Mullen, W.L. Smith, R.L. Sydnor, R.C.F. Vessot, and G.M.R. Winkler, IEEE Trans. Instr. Measur. Im-20(2) 105, Steven Chu, L. Hollberg, J.E. Bjorkholm, Alex Cable, and A. Ashkin, Phys. Rev. Lett. 55(1) 48, 1985.