Project code for the IAI5101 (intro to ML for Sci & Eng) project: Automated Mode-locking in all-fibre femtosecond laser using ML.
Mode-locking is a technique for generating ultrashort pulses from lasers, which are then called Mode-locked lasers. Mode-locking in a laser cavity can be initiated by either an active or a passive method. Active mode-locking is carried out by using an acousto-optic or electro-optic modulator inside the cavity [1]. The modulator frequency is matched with the laser cavity repetition rate to generate ultrashort pulses [2]. The disadvantage of this method is that the laser pulses generated by active mode-locking are limited in pulse width (µs, ns) because of the speed limitation of the modulator. The most common and reliable method to generate ultrashort pulses is through passive mode-locking. Passive mode-locking is achieved using suitable saturable absorbers such as semiconductor saturable absorber mirrors (SESAM), carbon nanotubes and graphene in the fibre laser cavity and can generate laser pulses in fractions of a nanosecond or even a picosecond [1]. However, the pulse width and energy have been limited by the response time and saturation energy (intensity) of the saturable absorber. There was therefore a need to overcome the limitations of real saturable absorbers using an artificial saturable absorber based on nonlinear polarization evolution (NPE). Controlling the NPE of the signal inside the laser cavity in such a way as to act as a saturable absorber has been one of the most widely used techniques in research labs in the past few decades. Usually, mechanical polarization controllers are used to initiate and manually control the NPE-based mode-locking technique [3]. Although the mechanical saturable absorber is important to initiate mode-locking, it does not guarantee its stability. If the laser is automatically mode-locked when it is switched on, self-starting mode-locking is achieved. However, self-starting mode-locking is not always achieved due to external perturbations and parasitic intracavity reflections. In such cases of non-self-starting mode-locking, the mechanical polarization controllers must be readjusted to force the cavity to generate ultrashort pulses. This is done by manually adjusting the angle of the polarization controller and simultaneously monitoring the pulse profile or spectrum using a fast oscilloscope with an embedded photodetector or an optical spectrum analyzer. The whole procedure is manual and time-consuming and requires the use of expensive equipment. Also, the stability and environmental sensitivity of light polarization hinders the use of NPE for mode-locking outside research labs. The objective of this project is to implement an automated mode-locked femtosecond fibre laser setup with an electrically actuated polarization controller (EPC) in combination with a microcontroller system that will use a suitable machine learning technique to achieve fast, stable, and selective mode-locking.
Improve the Mode-locking stability and allow for selective locking of desired pulsed operation in an all-fibre femtosecond laser using suitable Machine learning technique.
- Achieve fast, stable, and self-starting mode-locking.
- Ability to selectively lock onto desired pulsation regimes, mainly fundamental mode-locking (FML) and Harmonic mode-locking (HML).
- Ability to recover back lost mode-locking in case of detachments.
In recent years, different approaches to automate mode-locking (AML) have been proposed. One was nonlinear polarization rotation-based mode-locking, where piezoelectric squeezers were employed as polarization state controllers with an all-fibre division-of-amplitude polarimeter, with a feedback loop used to detect and measure polarization state changes [4]. Another was nonlinear polarization rotation in ytterbium-doped fibre lasers, where electronic polarization controllers were used to control the polarization state and a control system based on photodiode and high-speed counter to detect mode-locking [5]. In yet another, NPE-based mode-locking in fibre lasers involved using liquid crystal-based polarization state controllers along with a radio-frequency spectrum analyzer to detect and automate the mode-locking procedure [6]. Most recently, an all-fibre motorized polarization state controller along with a commercial polarimeter to detect the sudden change in the first stoked parameter was employed to automate the mode-locking procedure in a nonlinear polarization rotation-based similariton erbium-doped fibre laser [7]. However, the above-mentioned works are slow in achieving the mode-locked operation, it takes long time to switch from noise to a mode-locking regime. Moreover, it doesn’t offer the discrimination criteria for fundamental mode-locking (FML), and harmonic mode-locking (HML) operating regimes and ability to switch from one operational regime to another. An intelligent laser should have the ability to automatically recover back the lost mode-locking point in case of detachments. Recently, Pu G Q, et al [8], was able to achieve automatic mode lock onto various operational regimes including FML, HML, QS and QML using human like algorithm. They were able to achieve mode locking with short start up time as short as 0.22sec and recovery time (recover back the lost mode locking in case of detachments) within 14. 8msec. However, we believe, the automation procedure can be further improved using machine learning techniques to acquire faster mode locking times, enable better stability and robustness to ultrafast lasers. To address all these dilemmas surrounding automation procedures of smart lasers, machine learning and deep learning algorithms has seen tremendous interest and growth in the recent decade [9].
Mode-locking in the femtosecond fibre laser cavity will be achieved by changing the polarization state of light propagating through the fibre laser cavity using an electronic polarization controller (General Photonics PolaRITE III) in combination with a microcontroller system. Theoretically, whenever there is a mode-locking happening in a fibre laser, there is an abrupt change in the first stokes parameter (S1). Detecting this abrupt change in S1 parameter will be done using output polarization measurements using a polarization beam splitter and photodiodes. The analog outputs from the photodiodes will be monitored for polarization state change using the microcontroller, based on which a feedback signal will be used to force the EPC to mode-lock.
- F.X. Kärtner, J.A. der Au, and U. Keller, “Mode-locking with slow and fast saturable absorbers — what’s the difference?” IEEE J. Sel. Topics Quantum Electron., 4 (2), 159–168, March/April (1998).
- H. A. Haus, “Theory of mode locking with a slow saturable absorber,” IEEE J. Quantum Electron 11, 736–746 (1975).
- M.A. Abdelalim, Y. Logvin, D.A. Khalil, and H. Anis, “Properties and stability limits of an optimized mode-locked Yb-doped femtosecond fiber laser,” Opt. Express 17, 2264–2279 (2009).
- T. Hellwig, T. Walbaum, P. Groß, and C. Fallnich, “Automated characterization and alignment of passively mode-locked fiber lasers based on nonlinear polarization rotation,” Appl. Phys. B 101 (3), 565–570 (2010).
- X. Shen, W. Li, M. Yan, and H. Zeng, “Electronic control of nonlinear-polarization-rotation mode locking in Yb-doped fiber lasers,” Opt. Lett. 37 (16), 3426–3428 (2012).
- D. Radnatarov, S. Khripunov, S. Kobtsev, A. Ivanenko, and S. Kukarin, “Automatic electronic-controlled mode locking self-start in fibre lasers with non-linear polarization evolution,” Opt. Express 21 (18), 20626–20631 (2013).
- M. Olivier, M.-D. Gagnon, and M. Piché, “Automated mode locking in nonlinear polarization rotation fiber lasers by detection of a discontinuous jump in the polarization state,” Opt. Express 23, (2015).
- Pu G, Yi L, Zhang L, et al. Intelligent programmable mode-locked fiber laser with a human-like algorithm. Optica, 2019, 6: 362–369.
- G. Pu, L. Zhang, W. Hu, and L. Yi, “Automatic mode-locking fiber lasers: progress and perspectives,” Sci. China Inf. Sci. 63, 160404 (2020).