McKinley Pugh and Faculty Mentor: Dallin Durfee, Physics
Diode lasers are useful in physics because they are relatively cheap and robust, they are available in a number of wavelengths, and they are tunable. However, because diode lasers have large bandwidths compared to atomic resonances, a reflection grating is added outside the laser. This creates the extended cavity in extended cavity diode lasers (ECDLs) and forces the lasers to operate at a narrower line width, one acceptable for use in atomic physics. Unfortunately, because the ECDL has many factors trying to control the wavelength of the laser (e.g. temperature, current, grating angel and position) small changes in the lasers environment can cause it to mode hop. Passive control, such as precisely controlling the current and the temperature, can prevent changes in the environment and prevent mode hops, but ECDLs with very good passive control are expensive and even the best made lasers will eventually drift and mode hop. A better option is some form of active control. We believe we can use frequency noise to predict and prevent mode hops. I have demonstrated that the frequency noise on the laser increases before a mode hop. I have also designed and built a filtering circuit to measure the amount of frequency noise on the laser to use as error signal the move the laser to a stable environment.
For this project, I used a basic red diode laser, equivalent to those used laser pointers, housed in a Thorlabs laser mount. A more expensive laser was not necessary because my goal was to prove frequency noise could be used to improve the stability of an ECDL. A more stable ECDL would simply have a higher of level of stability at the start. A reflection grating was placed in front of the laser in a piezo mount. The position of the grating determined the resonances of extended cavity. Laser modes determine the wavelength of the laser. Those modes in resonance with the extended cavity experienced much more gain than those out of resonance. The reflection grating also reflected different wavelengths as different angles, so the angle of the grating determined which wavelength got reflected back into the laser. Because lasers are non-linear systems, they will put all their power into the mode with the most gain. Thus, the reflection grating angle and position can control the wavelength of the laser. The current, affecting the gain medium of the laser, also has an effect on the wavelength. I controlled the reflection grating position and angle with a piezo amplifier. The current was controlled by a low-noise current driver. I used a sidelock and a lab built PID controller to lock my laser to an optical cavity.
In order to determine if frequency noise predicted mode hops, I observed the laser unlocked. I used two Fabry-Perot cavities. These optical cavities are able to differentiate between very small differences in wavelengths. To observe the condition of the laser, I scanned on length of one cavity. As the length changed, I observed the signal from the photodiode at the end of the cavity. When the laser was single mode and stable, the signal resembled a number of thin, very evenly spaced peaks. As the laser approaches a mode hop, smaller peaks appear between the primary peaks and the primary peaks become smaller. From there, the laser will either mode hop, going to an entirely new set of peaks, or go multimode, with numerous small peaks. The second cavity I used to observe the frequency noise on the laser. Any change in the frequency was read as a change in intensity on the photodiode. Thus, I could observe both the state of the laser and the frequency noise of the laser at the same time. Though any ECDL will mode hop if left alone long enough, I scanned the refection grating to force the laser towards a mode hop so I could observe the process more directly. To observe what frequencies made up the noise, I used an oscilloscope to take the fast Fourier Transform of the signal from the non-scanned cavity.
To create a usable error signal from the frequency noise, I used the photodiode signal from one of the Fabry-Perot cavities. My goal was to create a DC signal that correlated with the level of frequency noise on the laser. I first built a second order high pass filter to get rid of the offset and power-line noise that was not correlated with mode hops. The signal was then sent through a half wave rectifier. I then used a low pass filter to smooth out the signal and leave me with a signal that when up and down as the amplitude of the noise increased and decreased.
This year I designed and built my optical set up. When I first turned on the laser, it would not stay single mode, so I spent some time increasing the laser’s stability with passive components. In the end, the most important changes were fixing the reflection grating more securely in its mount and adjusting the temperature controller to eliminate oscillations in the temperature.
In my research, I found that an increase in the frequency noise on the laser predicted a mode hop. When the laser was unstable, the noise generally followed a 1/f shape, with an anomaly at 120 Hz that we believe is power-line noise. I found that the laser behaved differently depending on the direction from which I approached a mode hop. From one side, the noise increased slowly and steadily, with a very defined 1/f shape, until the laser mode hopped. From the other side, noise still increased before a mode hop, but it increased drastically without much warning and the laser mode hopped very quickly afterward. The 1/f shape indicates that the most important noise is on the lower end of the frequency spectrum.
My circuit to measure the noise generated a signal that moves up and down with the level of noise, but I have run into some trouble with the 120 Hz noise. It is so much bigger than any other noise that it is hard to filter out. Much of my time this semester has been spent trying to eliminate this noise, both by filtering in my circuit and by locating its source.
Discussion and Conclusion
I have demonstrated that frequency noise does work as a predictor for mode hops in ECDLs. However, I have run into some difficulty in converting that noise into a usable error signal. Much of the problem centers on the 120 Hz noise which has an amplitude so much bigger than that of the other noise that it is difficult to filter out. Future work will include attempting to locate the source of this noise and to eliminate it. Once the circuit adequately converts the noise into a useable error signal, the effect of this new feedback system remains to be tested to see how much it really has increased the stability of the ECDL. I have presented on my work at the Frontiers in Optics/ Laser Science APS conference.