Gandhi Alagappan’s research while affiliated with Institute of High Performance Computing and other places

This article presents an on‐chip optical power limiter that utilizes the thermo‐optical defocusing effect. A pair of input and output waveguides is designed to mimic emitting and receiving antennas. The waveguides are separated by a free‐space region filled with poly‐methyl‐meth‐acrylate (PMMA) material, which has a negative thermal‐optic coefficient that causes a decrease in refractive index with an increase in temperature. As the power in the input waveguide increases, the refractive index of the free‐space region decreases, which in turn increases the radiated beam's divergence angle with respect to input power. The empirical findings demonstrate that the non‐linear divergence angle can be written as θ0+kP${\theta }_0 + kP$, where θ0 represents the divergence angle of the equivalent Gaussian beam, k is a waveguide‐specific constant, and P is the input power. The edge of the receiving waveguide is tapered to adjust the coupling of the divergent beam to the output waveguide. The taper width is optimized to minimize insertion loss. The devices are two orders lengthwise smaller compared to the bulk demonstration, and they exhibit low loss ranging from 0.2 to 10 dB.

This article applies deep learning-accelerated inverse design algorithms and discovers a spectrum of photonic power dividers with exceptional performance metrics despite the simplicity in the design geometry. The deep learning models exhibit high precisions on the order of 10−6 to 10−8 for both TE and TM polarizations of light. These models enable ultrafast search for an empirically describable subspace that simultaneously satisfy compact footprints, ultralow losses, ultrawide bandwidth, and exceptional robustness against fabrication randomness. We demonstrate a spectrum of devices for silicon photonics with programmable power splitting ratios, excess losses as small as 0.14 dB, to the best of our knowledge, the smallest footprints on the scale of sub-λ 2, and low loss bandwidths covering the whole telecommunication spectrum of O, S, E, C, L and U-bands. The robustness of the devices is statistically checked against the fabrication randomness and are numerically verified using the full three-dimensional finite difference time domain calculation.

Integration of single-photon emitters (SPEs) with resonant photonic structures is a promising approach for realizing compact and efficient single-photon sources for quantum communications, computing, and sensing. Efficient interaction between the SPE and the photonic cavity requires that the cavity's resonance matches the SPE’s emission line. Here we demonstrate a new method for tuning silicon nitride (Si3N4) microring cavities via controlled deposition of the cladding layers. Guided by numerical simulations, we deposit silicon dioxide (SiO2) nanolayers onto Si3N4 ridge structures in steps of 50 nm. We show tuning of the cavity resonance exceeding a free spectral range (FSR) of 3.5 nm without degradation of the quality-factor (Q-factor) of the cavity. We then complement this method with localized laser heating for fine-tuning of the cavity. Finally, we verify that the cladding deposition does not alter the position and spectral properties of nanoparticles placed on the cavity, which suggests that our method can be useful for integrating SPEs with photonic structures.

Artificial intelligence (AI) techniques have been spreading in most scientific areas and have become a heated focus in photonics research in recent years. Forward modeling and inverse design using AI can achieve high efficiency and accuracy for photonics components. With AI-assisted electronic circuit design for photonics components, more advanced photonics applications have emerged. Photonics benefit a great deal from AI, and AI, in turn, benefits from photonics by carrying out AI algorithms, such as complicated deep neural networks using photonics components that use photons rather than electrons. Beyond the photonics domain, other related research areas or topics governed by Maxwell’s equations share remarkable similarities in using the help of AI. The studies in computational electromagnetics, the design of microwave devices, as well as their various applications greatly benefit from AI. This article reviews leveraging AI in photonics modeling, simulation, and inverse design; leveraging photonics computing for implementing AI algorithms; and leveraging AI beyond photonics topics, such as microwaves and quantum-related topics.

Integration of single-photon emitters (SPEs) with resonant photonic structures is a promising approach for realizing compact and efficient single-photon sources for quantum communications, computing, and sensing. Efficient interaction between the SPE and the photonic cavity requires that the cavity's resonance matches the SPE emission line. Here we demonstrate a new method for tuning silicon nitride (Si3N4) microring cavities via controlled deposition of the cladding layers. Guided by numerical simulations, we deposit silicon dioxide (SiO2) nanolayers onto Si3N4 ridge structures in steps of 50 nm. We show tuning of the cavity resonance over a free spectral range (FSR) without degradation of the quality-factor (Q-factor) of the cavity. We then complement this method with localized laser heating for fine-tuning of the cavity. Finally, we verify that the cladding deposition does not alter the position of nanoparticles placed on the cavity, which suggests that our method can be useful for integrating SPEs with photonic structures.

Physical fields represent quantities that vary in space and/or time axes. Understanding the distribution of a field pattern is a key element in scientific discoveries and technological developments. In this article, by picking up the electromagnetic field of an optical waveguide as an example, we demonstrate how field patterns can be uncovered using artificial neural networks. The cross section plane of the optical waveguide is discretized into a set of tiny pixels, and the field values are obtained at these pixels. Deep learning model is created by assuming the field values as outputs, and the geometrical dimensions of the waveguide as inputs. The correlation between the field values in the adjacent pixels is established by mean of feedback using a recurrent neural network. The trained deep learning model enables field pattern prediction for the entire (and usual) parameter space for applications in the field of photonics.

An optical waveguide is the fundamental element in a photonic integrated circuit. This paper establishes a universal deep learning representation for the effective refractive index of an optical channel waveguide for the entire and usual parameter space for applications in photonics. The deep learning model is able to make precise predictions for wide spectrum optical wavelengths, dielectric materials of refractive indices varying from 1.45 to 3.8, and a wide range of feasible geometrical parameters of the waveguides. The deep learning model consists of fully connected feedforward neural networks, and rigorous optimization of neural network architecture is carried out. Deep learning models with two and three hidden layers provide rapid convergence with a minimal number of training data points and offer unprecedented precisions that are a few orders better in magnitude than the conventional interpolation techniques.

This article displays the method of constructing deep learning models for optical mode solving, with a minimal number of exact numerical solutions to Maxwell’s equations. We select a silicon nitride channel waveguide and show how the patterns in the effective refractive indices of the fundamental waveguide modes for both polarizations of light, can be uncovered with only 4–16 learning points for the entire parameter space that can be conveniently accessed using existing photo-lithographical and CMOS fabrication techniques. We also illustrate the effect of various transfer functions and neural network layouts to the overall performance of the deep learning model.