A frequency-stabilised diode laser and its application to atomic spectroscopy of caesium : a thesis presented in partial fulfilment of the requirements for degree of PhD in Physics at Massey University

Abstract

This thesis examines the hypothesis that a commercially available single-chip AlGaAs diode laser can be rendered suitable for atomic physics research through the addition of external circuitry alone. This circuitry is discussed and the frequency-stabilised laser is applied to some interferometric and spectroscopic studies. These studies also include the development of theoretical models. The design and construction of an ultra-stable constant current source is discussed, along with an evaluation of its effectiveness. The current source is capable of supplying up to 100-mA de with the addition of a ramp current of up to 20 rnA to allow frequency scanning of the laser, and exhibits short-term fluctuations of ±8 JlA with a drift of less than 2 J1A in ten hours. A temperature controller, capable of both heating and cooling the laser, is described. This can maintain temperature stability to ±1 mdeg. A data acquisition and control unit is designed and constructed to enable the interfacing of both the laser and other measurement apparatus to a micro-computer. A system for automatically measuring wavelength maps as a function of injection current and laser temperature is developed, utilising the data acquisition and control unit in conjunction with a grating spectrometer. This is applied to four diode lasers. Linearity measurements are made of the frequency scan with injection current. The linewidth of one laser is measured as a function of its output power by heterodyning its output with that of a second laser operated at fixed injection current and temperature, and observing the beat note on a rf spectrum analyser. The linewidth can be reduced to below 10 MHz. The absolute frequency stability of the laser is measured by monitoring the change in absorbed power over time when the laser is tuned to the D2 transition of caesium. The drift is measured to be 10 kHz/s over five hours, and is attributed largely to temperature drift. An analysis is presented which allows the spectral linewidth of a light source with a Lorentzian or Gaussian profile to be calculated from the measured fringe visibility of the transmission fringes of a high-finesse scanning Fabry-Perot interferometer when these fringes are not fully resolved. The method is also applied to truncated spectral profiles, which may provide a more appropriate model for the spectral profile of some lasers, and the differences are characterised. The analysis is then extended to interferometers with low finesse and also to include the Voigt profile. An experimental verification of the analysis is made using a diode laser whose spectral profile is demonstrated to be Lorentzian in shape. The frequency-stabilised diode laser is applied to the study of saturated absorption spectroscopy of the caesium D2 line at 852.1 nm. Both single-beam and two-beam experiments are performed using an absorption cell containing caesium vapour with no buffer gas. A theoretical model is developed, based on rate equations for the population densities of the 48 sublevels involved in the transition, including a rate for ground state relaxation due to transit of atoms through the laser beam. The model facilitates the analysis of the effects of optical pumping and extent to which the ground state relaxation rate limits the optical pumping. For the medium-resolution single-beam experiments the laser is shown to be a suitable source for resolving the Doppler-broadened transitions, and good agreement is obtained with the model for sufficiently low laser powers. However, for the sub-Doppler two-beam experiments the limited frequency stability is revealed. An analysis of the lock-in detection process for the two-beam experiments is made, emphasising the relationships between the detection frequency, the ground state relaxation rate, and the shape of the detected signal. It is discovered that if the detection frequency is of the order of the ground state relaxation rate then a simple subtraction of the Doppler-broadened background from the probe absorption signal does not adequately model the lock-in detection process. A technique is described to experimentally determine the ground state relaxation rate from the detected signal as a function of the lock-in phase.