Brian Neyenhuis and Dr. Dallin Durfee, Physics and Astronomy
Laser diodes are an attractive option for atomic physics research. They are small, available in a wide variety of wavelengths, inexpensive, and require little power. Because diode lasers are typically made in bulk for the computing or telecom industries where linewidth and mode structure are not generally important, a solitary diode usually has a linewidth on the order of GHz and often runs multi-mode. By placing the diode in an extended cavity the finesse of the laser cavity can be increased, narrowing the linewidth of the laser to a few hundred kHz. For even narrower linewidths a diode laser is typically locked to a high-finesse reference cavity. With this method diode lasers with linewidths as narrow as 30 Hz have been reported.
With recent technological advances such as the mode-lock frequency comb, optical frequency atomic clocks may soon become a viable replacement for older microwave frequency standards. Of particular interest for optical frequency standards is the Calcium intercombination line at 657 nm. But to make these optical frequency standards a reality more robust narrow linewidth lasers with the ability to stay locked for long periods of time are needed.
Diode lasers stabilized by means of a high-finesse reference cavity have been previously reported. We present a novel scheme to combine a high speed lock circuit with the ability to scan the laser mode-hop-free over many GHz to increase the long-term stability and make locking and relocking the laser to the correct cavity line easier. We lock a 657 nm diode laser to a 10 kHz resonance of a high-finesse optical cavity with a finesse of 3,000. The Pound-Drever-Hall technique is used to further reduce the laser linewidth and cancel out frequency noise by actively controlling both the position of the diffraction grating forming the extended cavity of the diode and the laser current. A linewidth of 3 kHz is achieved using an electronic loop with a bandwidth of 5 MHz to control the current of the laser.
We use a 657 nm diode laser in a Littrow style extended cavity grating stabilization scheme. The light passes through a phase modulating Electro-Optic Modulator (EOM) acquiring 30 MHz sidebands. This light passes through a polarizing beam splitter and a quarter waveplate and is directed to a high finesse cavity. The light reflected from the cavity passes through the quarter waveplate twice, rotating its polarization by 90 degrees and causing it to be rejected by the polarizing beam cube. This allows us to direct 100 percent of the rejected light to a photo diode where the 30 MHz beat note between the sidebands and the carrier is detected.
The laser beam is collimated only by a single lens mounted directly in front of the diode. No additional beam shaping or collimation was needed to couple to the cavity, although better matching to the TEM00 mode of the cavity could eliminate or greatly reduce the DC background on the photodiode signal, thus eliminating our current need to high-pass filter the signal before it is amplified and demodulated.
To narrow the laser linewidth and provide short term stability the demodulated error signal is fed into a high-speed electronic lock circuit. This circuit provides proportional, differential, and integral gain, each adjustable to better match the individual characteristics of the laser diode. These three gains are summed together and fed into the modulation input of the laser current controller. By modulating the laser current through the current driver rather than directly at the diode, we eliminate the need to bias our control circuit and extend the range of current feedback down to DC. The bandwidth of out lock circuit is not limited by the current controller or the lock circuit but is limited by the characteristics of the diode itself.
To estimate the linewidth the laser is first locked very loosely using only proportional and integral gain. Then the proportional gain is turned up causing the laser to oscillate. As the
proportional gain is increased the oscillations grow until they have reached the full peak-to-peak voltage of our error signal at which point the sine wave will become distorted as we “roll over” the peak of the error signal. This gives us a rough estimation of the voltage corresponding to the full linewidth of the cavity. Then the lock is optimized, adjusting the proportional, differential, and integral gain iteratively until the error signal is minimized. This peak-to-peak voltage is measured and compared to the maximum voltage to estimate the laser’s linewidth. Using this method we estimate our laser’s linewidth at about 3 kHz.
The high-speed lock circuit does an excellent job eliminating high frequency noise and drifts, effectively reducing the linewidth of the laser. But the range over which the laser current can be modulated without hopping to a different longitudinal mode is extremely limited. To extend this range, it is customary to integrate the error signal and feed it directly to the PZT controlling the grating causing the PZT to track any slow drifts in the laser system. This seems to work moderately well at tracking larger frequency excursions, but eventually the PZT and the current feed-forward become mismatched and the laser hops to a different longitudinal mode.
To fix this problem we incorporated a scan balancer into our lock circuit. A fixed ratio of voltages between the current feed-forward and the PZT is determined empirically to allow for the maximum mode-hop free scan range. With this ratio set the laser is able to scan 8 GHz mode hop free. This ratio is set inside the lock circuit so the PZT and the current driver respond to all feedback with the prescribed ratio. Because the PZT control a mechanical joint with strong resonances around 1 kHz a low pass filter is used. Thus the laser current and PZT track slow drifts together allowing the circuit to compensate for larger frequency excursions and reducing the chances of “hopping” out of lock. This scan balancer also allows the laser to be easily scanned mode-hop free from one optical cavity resonance to another making it easier to find the correct cavity resonance.
We have demonstrated a robust lock circuit that improves long term stability and simplifies locking with the addition of a scan balancer. Using this lock circuit we have narrowed the linewidth of a 657 nm diode laser from a few hundred MHz to 3 kHz. Further reduction in linewidth can be achieved with better passive stabilization and we plan to upgrade our cavity to a finesse of 300,000 with the goal to narrow the linewidth to the Hz level.