Photonic crystal fibers (PCFs) are the specialized optical waveguides that led to many interesting applications ranging from nonlinear optical signal processing to high-power fiber amplifiers. In this paper, machine learning techniques are used to compute various optical properties including effective index, effective mode area, dispersion and confinement loss for a solid-core PCF. These machine learning algorithms based on artificial neural networks are able to make accurate predictions of above mentioned optical properties for usual parameter space of wavelength ranging from 0.5-1.8 μm, pitch from 0.8-2.0 μm, diameter by pitch from 0.6-0.9 and number of rings as 4 or 5 in a silica solid-core PCF. We demonstrate the use of simple and fast-training feed-forward artificial neural networks that predicts the output for unknown device parameters faster than conventional numerical simulation techniques. Computation runtimes required with neural networks (for training and testing) and Lumerical Mode Solutions are also compared.

Machine learning is an application of artificial intelligence that focuses on the development of computer algorithms which learn automatically by extracting patterns from the data provided. Machine learning techniques can be efficiently used for a problem with a large number of parameters to be optimized and also where it is infeasible to develop an algorithm of specific instructions for performing the task. Here, we combine the finite element simulations and machine learning techniques for the prediction of mode effective indices, power confinement and coupling length of different integrated photonics devices. Initially, we prepare a dataset using COMSOL Multiphysics and then this data is used for training while optimizing various parameters of the machine learning model. Waveguide width, height, operating wavelength, and other device dimensions are varied to record different modal solution parameters. A detailed study has been carried out for a slot waveguide structure to evaluate different machine learning model parameters including number of layers, number of nodes, choice of activation functions, and others. After training, this model is used to predict the outputs for new input device specifications. This method predicts the output for different device parameters faster than direct numerical simulation techniques. Absolute percentage error of less than 5% in predicting an output has been obtained for slot, strip and directional waveguide coupler designs. This study pave the step towards using machine learning based optimization techniques for integrated silicon photonics devices. Index Terms-Machine learning, neural networks, regression, multilayer perceptron, silicon photonics. I. INTRODUCTION M ACHINE learning (ML) technology is being extensively used in many aspects of modern society: web searches, social networking, smartphones, bioinformatics, robotics, chatbots, and self-driving cars [1]. ML techniques are used to classify or detect objects in images, speech to text conversion, pattern recognition, natural language processing, sentiment analysis and recommendations of products/movies for users based on their search preferences. ML algorithms can be trained to perform exceptionally well when it is difficult to analyze the underlying physics and mathematics of the problem [2]. ML algorithms extract patterns from the raw data provided during the training without being explicitly programmed. The learned patterns can be used to make predictions on some other data of interest. ML systems can be trained more efficiently when massive amount of data is present [3], [4]. Recently, research on the application of ML techniques for optical communication systems and nanophotonic devices is

A full-vectorial numerically efficient Finite Element Method (FEM) based computer code is developed to study complex light-sound interactions in a single mode fiber (SMF). The SBS gain or SBS threshold in a fiber is highly related to the overlap between the optical and acoustic modes. For a typical SMF the acoustic-optic overlap strongly depends on the optical and acoustic mode profiles and it is observed that the acoustic mode is more confined in the core than the optical mode and reported overlap is around 94 % between these fundamental optical and acoustic modes. However, it is shown here that selective co-doping of Aluminum and Germanium in core reduces the acoustic index while keeping the optical index of the same value and thus results in increased acousticoptic overlap of 99.7%. On the other hand, a design of acoustic anti-guide fiber for high-power transmission systems is also proposed, where the overlap between acoustic and optical modes is reduced. Here, we show that by keeping the optical properties same as a standard SMF and introducing a Boron doped 2 nd layer in the cladding, a very low value of 2.7% overlap is achieved. Boron doping in cladding 2 nd layer results in a high acoustic index and acoustic modes shifts in the cladding from the core, allowing much high power delivery through this SMF.

We demonstrate a novel approach to enhance the mode stability through increased effective index difference (Δneff) between the higher-order modes (LP, LP and LP) of a multimode fiber. Fibers with large diameters have bigger effective mode areas (Aeff) and can be useful for high power lasers and amplifiers. However, a large mode area (LMA) results in an increased number of modes that can be more susceptible to mode coupling. The modal effective index difference (Δneff) strongly correlates with mode stability and this increases as the modal order (m) increases. We report here that the mode spacing between the higher order modes can be further enhanced by introducing doped concentric rings in the core. In our work, we have shown a more than 35% increase in the mode spacing between the higher order modes by optimizing the doping profile of a LMA fiber. The proposed design technique is also scalable and can be applied to improve the mode spacing between different higher order modes and their neighboring antisymmetric modes, as necessary.

Citation: Gulistan, A., Rahman, M. M., Ghosh, S. ORCID: 0000-0002-1992-2289 and Rahman, B. M. ORCID: 0000-0001-6384-0961 (2019). Elimination of spurious modes in fullvectorial finite element method based acoustic modal solution.This is the published version of the paper.This version of the publication may differ from the final published version.Permanent repository link: http://openaccess.city.ac.uk/22191/ Link to published version: http://dx.Abstract: Finite element method is a powerful technique for solving a wide range of engineering problems. However, the existence of the spurious solutions in full-vectorial finite element method has been a major problem for both acoustic and optic modal analyses. For emerging photonic devices exploiting light-sound interactions in high index contrast waveguides, this problem is a major limitation. A penalty function is introduced to remove these unwanted spurious modes in acoustic waveguides, which also identifies the acoustic modes more easily. Numerically simulated results also show considerably improved vector mode profiles. The proposed penalty method has been applied for the characterization of low index contrast single mode fiber and also for high index contrast silicon nanowire to demonstrate its effectiveness.