Photonics for AI and AI for Photonics: Material and Characteristics Integration

*Sunil Sharma and Lokesh Tharani*

## **Abstract**

We are living in the technological era, where everything is integrated with each other. If we are discussing regarding communication, it is integrated with one or two technologies. If we are discussing regarding automation, discussing regarding Image processing, discussing regarding embedded system, they all are integrated with a combination of technologies. Correspondingly Artificial Intelligence (AI) and Photonics are also integrated with each other. Now a day as AI is utilizing with photonics in abundant fields as well photonics is also serving AI to facilitate ultrafast AI networks to offer a novel class of Information Processing Machines (IPM). This chapter is based on identification and implementation of photonics for AI utility and AI for photonics. In this category a Dual core Photonics crystal fiber (PCF) is proposed which serve to identify infected cells of human being along with the integration of AI. This proposed design of PCF is providing relative sensitivity and confinement loss in an optimized manner with the impact of AI. Here potency of AI as well as of Photonics is explained to serve their applications related to each other.

**Keywords:** Artificial Intelligence, Fiber Optics and Photonics, Optical Networks, Photonic Crystal Fiber, Integration

### **1. Introduction**

Latest technological development in photonics has multiplied only due to integration of photonic platform with conception of Opto-electronic elements [1]. The Photonic Integrated Circuits (PICs) [2] have facilitated the ultrafast Artificial Neural Networks (ANN) [3], to propose a novel class of Information Processing Machines (IPM) [4]. There are number of reasons available which reveals that photonics is somewhere associated with AI. In this direction the latest example can be considered as development of Neuro-morphic [5] electronics, which shows that *high processor delay* can be eliminated by offering a consequent technology to extend the vicinity of AI. It offer sub-nanosecond [5] delay and consequently conquers challenges in terms of present and future aspects.

This latest developed technology 'Neuro-morphic electronics system' [5] is integrated with most recognized technology which is known as semiconductor photonics. It is composed of third and fifth group of elements i.e. GaAs and InP [2].

**Figure 1.** *GaAs & InP Composed Photonic Integrated Circuits [2].*

Below **Figure 1** represents the photonic integrated technology indicating fabrication, characteristics like growing and mixing of GaAs and InP materials to provide efficient, robust, and monolithic optoelectronic integration platform. It was developed and observed by Sandia National laboratory services.

The developmental growth of photonic crystals, components and meta-materials [6] lead to the advancement of photonics in the area of designing, modeling and technological integration. This kind of integration investigates AI with photonics. This promising domain is someway sustained by 'photonic materials' [7] which assist to find out and intend innovative applications of AI. It should be noted down that how photonics is contributing for the implementation of AI tools and techniques.

The contributing field of photonics towards AI includes Neuro-morphic electronic system, Optical Neural Network (ONN), Nano Photonics,

#### **Figure 2.**

*Proposed dual cores PCF with different mode indexes.*

*Photonics for AI and AI for Photonics: Material and Characteristics Integration DOI: http://dx.doi.org/10.5772/intechopen.97781*

meta-materials, optical sensing, optical imaging [8], optical computing, Information Processing Machines etc. These above mentioned optics emerging domains can be integrated with AI tools [9] to enhance the efficiency and performance of these systems.

**Figure 2** represents the design structure of dual core silica PCF with an effective index mode of 1.4053. By changing the mode index value we can have light confinement variation which is shown below in **Figure 3 (a & b)**.

As shown below in **Figure 4** indicates contribution of photonics in terms of machine intelligence with Neuro-morphic computing along with Optical neural network and optical sensing for AI technology. These latest technologies helped AI to diagnose critical disease.

**Figure 3.**

*(a, b) Light Confinement through proposed design for different values of Index Modes (c) Identification of infected cells with AI (d) Relative sensitivity (e) confinement loss of proposed design.*

**Figure 4.**

*Contributing field of photonics for AI (a) AI with Neuro-morphic computing (b) Optical Neural Network (c) Optical sensing and computing [5].*

### **2. Photonics materials and their characteristics for AI**

We all are witnessing an inconceivable age of drastically development in applications that necessitate expansion in AI [10]. If we are discussing about the ingenious novel outcomes that are gradually trending towards the market place and many more are preferred and expected. Fiber Optics & Photonic materials [11] are widely used for these products like new display, personalized mobile devices, novel sensors, and new information processing machining products for both storage and data processing. It is trending in very clear manner that the areas of Fiber Optics & Photonic materials are fundamental technologies for the globe. Inventing and uncovering new materials [12] in the Fiber Optics & Photonics domain will be exceedingly critical to see more and more novel outcomes to improve normal people's lives.

Materials that have been exposed at the crucial point of life, always changes the history of human being along with the country. Materials that are used to senses, materials that are used to stores, materials that can be used as energy efficient, some translucent materials which can be folded easily and some materials that are manufacturable at low cost. New discovered materials such as doped silica materials [13], resistance changing materials and spontaneously magnetize and polarize materials have been discovered and using widely for AI integration and their applications.

In the line of discovery of new materials, the Picometer [7] can also be considered as a vibrant example in the field of atomic structures. There are numerous atomic structures available that were simulated and their data were utilized for AI analysis to identify artificially controlled 'oxygen octahedral rotation' (OOR) patterns as shown in below **Figure 5**.

*Photonics for AI and AI for Photonics: Material and Characteristics Integration DOI: http://dx.doi.org/10.5772/intechopen.97781*

**Figure 5.** *Oxygen Octahedral Rotations and its characteristics [7].*

It was used as Disorder-Driven Metal–Insulator Transition in Crystalline Vacancy-Rich Ge-Sb-Te Phase-Change Materials [7].

#### **2.1 New Investigative Materials for AI**

The discovery and development in the new materials plays an important role in the technological progress. As we have already seen that how silica has revolutionized the microelectronics industry. Materials discovery and design efforts require interplay between materials prediction, synthesis and characterization [12] have increased applications of computational tools and techniques, increased generation of material's databases, and accelerated advances in experimental methods significantly. Some of them are composed of three special elements i.e. germanium, antimony and tellurium which is defined as Ge-Sb-Te alloy [2] and can be termed as phase-change memory materials. This alloy is selected from the group of chalcogenide glass (As2Se3) [12] which can be used in rewritable optical discs.

The above mentioned **Figure 6** is used as a non-volatile quasi-continuously reprogrammable platform. This phase-change memory material rapidly changes its atomic structure from crystalline to solid amorphous when swiftly melted in presence of temperature. These kinds of materials are widely used in 'electronic memory' applications of AI tools such as *data storage*. Even though there are countless integration is possible with Ge-Sb-Te alloy, the new material **GST467** [6] revealed by CAMEO (Closed-Loop Autonomous System for Materials Exploration and Optimization) is most favorable for phase-changing applications.

CAMEO found the best Ge-Sb-Te alloy that had the largest difference in "optical contrast" [6]. GST467 also found applications in photonic switching devices that can be used to control the direction of light in given circuit. These devices can also be utilized in Neuro-morphic computing [5], which is an emerging field focusing on development of devices which imitate the formation and role of neurons in human brain. Materials science or solid-state physics is plagued by the 'curse of dimensionality'.

#### **3. AI for photonics**

When the words "artificial intelligence" (AI) comes to mind, our first thoughts may be of super-smart computers or robots that perform tasks without needing any help from humans.

#### **Figure 6.**

*GST467 with AI (a) Schematic cross-section of the hybrid waveguide. (b)&(c) Fundamental quasi-transversal electric (TE) mode profiles of the hybrid waveguide at 1550 nm for (d) complex refractive index of GST and GST as a function of wavelength. (e) XRD data of GST [6].*

A multi-institutional team of research scholars from National Institute of Standards and Technology (NIST) [6] have developed an AI algorithm known as CAMEO. It was used for the discovery of potentially applicable new photonic material without any additional preparation and efforts from the scientist. These AI systems helped to reduce the trial-and-error time which generally scientists use up in the lab. Along with this these systems maximizes the productivity and efficiency of their research work. Another research scientists team at POSTECH (Pohang University of Science and Technology) [7] got succeed in creating a novel substance that generates electricity by effect of polarization at room temperature. The variation so observed would be confirmed in crystal structure by analysis of deep neural network. The above mentioned examples revealed the techniques behind making materials used in new memory devices by using artificial intelligence. So it is very much clear that the use of modern computational techniques like AI can be used to improve the rate of discovery of these new photonics materials and vice versa. Helping scientists in reaching their outcomes more efficiently and quickly by performing only few experiments with limited resources. All these things became possible only because of integration of AI and Photonics.

The optical properties are typically calculated by using Maxwell's Equations [13]. The desired optical response can be obtained by adjusting the initial design and performing multiple simulations until the outcome is achieved. Despite designing issues AI can help optics and nano photonics in different tasks, for example AI used to estimate the optical properties of black carbon fractal aggregates. Another

example is reported where they combines finite element simulations and clustering for the identification of photonic modes [14] with large local field energies and specific spatial properties. It is shown that the combination of machine learning with photonics [15] can revolutionize one of the most important fields in optical imaging.
