Abstract
M.Phil. (Electrical and Electronic Engineering)
A numerical model of an Erbium-Ytterbium co-doped distributed feedback (DFB) fiber laser is developed. The DFB fiber laser is a short length fiber laser whose feedback is distributed throughout the cavity. Its main advantage is its single longitudinal mode operation. The amplifying medium of a DFB fiber laser is a few centimetres long rare earth doped fiber. The feedback is obtained by a fibre Bragg grating printed in the core of the rare earth doped fiber. This type of laser emits naturally in two longitudinal modes. To obtain the single longitudinal mode operation, a π phase shift is introduced in the middle of the grating. Erbium doped DFB fiber lasers present the advantage of emitting single frequency light in the 1550 nm region where telecommunication fibers present the minimum loss. However due to the relatively short length of the gain medium, the number of available Erbium ions is small; as a result pump power absorption is low and the
efficiency of the fiber laser is strongly reduced. The straightforward solution to this problem could be increasing the concentration of Erbium ions. This solution however has the disadvantage of increasing the Erbium ions interactions, thus leading to detrimental effect like cooperative upconversion and excited state absorption, which in term reduce considerably the laser efficiency. The best solution is to use Ytterbium ions as sensitizers along with Erbium ions to enhance the pump absorption, hence the efficiency of the laser.
A model of the DFB fiber laser is an indispensable tool for its design, because it allows one to predict characteristic behaviour that would be both difficult and costly to deduce in laboratory conditions. The model developed in this project is based on rate equations of the Er3+-Yb3+ gain medium and coupled mode equations describing the laser field propagation in the fibre Bragg grating structure. The equations are solved using a quasi-analytical iterative method along with transfer matrix method with appropriate boundary conditions. The quasianalytical method used in this thesis is more robust than numerical solutions because it does not require providing an initial guess on the solution. Furthermore this method is hundreds time faster than the exact numerical solution while giving almost similar results.