The wonderful world of silicon photonics materials: How to choose

By Alex Demkov / La Luce Cristallina

As AI bandwidth and power-efficiency demands accelerate, material choice in silicon photonics has become more critical than ever, driving companies to balance performance, scalability and manufacturability in pursuit of the optimal platform. With so many choices, especially for optical modulators, how can companies evaluate each material to understand its benefits and drawbacks?

Electro-optic polymers enable low-cost, high-volume manufacturing and can be chemically tuned to achieve specific optical properties. However, they struggle with long-term stability under heat, radiation and high frequencies. They can also only be integrated into a far back-end of the line process. Lithium niobate is a trusted workhorse with stable electro-optic properties, but it can pose contamination and integration challenges with other silicon photonics processing flows. Barium titanate, a relatively new entrant in the industry, boasts an exceptional electro-optic coefficient and silicon-foundry compatibility, though its ecosystem and large-scale manufacturing processes are still maturing. Each material offers a distinct balance of performance, cost and manufacturability. Understanding these trade-offs is becoming increasingly critical as AI drives demand for faster, more efficient photonics systems for data communications.

Companies can leverage these materials to support high-bandwidth use cases such as data center interconnects, telecommunications, biosensing and others. Amid these choices, extensive research and careful consideration are crucial in maximizing photonics investments and furthering AI’s integration across the technology ecosystem. Let’s begin by outlining these industry opportunities, then explore each material’s utility across various applications.

Capturing the AI opportunity through photonics materials

AI applications have captured the industry’s attention (and wallet), with the top 5 hyperscalers investing over $250 billion in data center development in 2025. As these AI applications rapidly proliferate, companies require high-speed data center interconnects to transfer massive volumes of training and inference data between campuses. Statistics reflect these trends, with experts estimating the data center interconnect industry’s valuation will balloon to over $25 billion by 2030. The biosensing market is also growing steadily, projected to reach $76.2 billion by 2030. Current and emerging use cases, such as quantum computing and defense applications, also hold the potential to further revolutionize the technology ecosystem.

Meeting these industry demands requires ferroelectric materials (such as barium titanate, lithium niobate, or electro-optic polymers using Pockels or linear electro-optic effect for modulation) that deliver speed, scalability, energy efficiency and reliability across multiple generations of design. These materials will play a critical role in supporting high-capacity use cases as data center and telecommunications operators continue to expand from 400G to 800G, with the broader integration of 1.6-terabit pluggables on the horizon. However, with so many choices, companies must carefully map their specific requirements to the benefits of each material. After all, one size does not fit all for any choice in the technology industry, particularly in photonics.

Rising star: Barium Titanate

As a ferroelectric oxide that can be grown (or transferred) onto silicon, barium titanate offers one of the highest Pockels coefficients among practical materials, equaling ~480 pm/V in sputtered thin-film implementations and ~1,300 pm/V for its highest bulk-crystal tensor component, as compared to ~35 pm/V for bulk lithium niobate. This exceptional electro-optic efficiency enables modulators with significantly lower drive voltages, making them well-suited for dense datacom links, programmable photonic fabrics and compact biosensing devices.

Barium titanate also offers silicon-friendly integration, showing a clear path toward integration with silicon photonics. Recent characterization has shown propagation losses as low as 0.14 dB/cm (at 1550 nm, extracted from racetrack resonators) and 0.7 dB/cm (at 1550 nm, from reflectance measurements of 3-cm waveguides). These low propagation losses enable shorter device lengths and higher integration density, translating into lower overall link power consumption.

However, as a relative newcomer to photonics, its manufacturing ecosystem is still maturing compared to legacy materials. Still, recent advances in RF sputtering and thin-film deposition are accelerating barium titanate’s path toward higher-volume production for telecom and datacom applications. The material’s gigantic Pockels coefficient also enables more longevity through multiple generations of innovation (over 10 years), helping companies prepare for current and future photonics demands. These characteristics make barium titanate best suited for dense short-reach links, programmable photonics switches, optical fabric overlays for AI, high-speed modulators for hyperscale environments and more.

Middle ground: Polymers

Polymers have long represented a flexible, utilitarian approach in silicon photonics. They offer tunable refractive indices and high electro-optic activity (typically 50–150 pm/V depending on formulation). Polymers can be engineered for specific wavelength ranges or device types, making them customizable and adaptable across numerous use cases. This material can also deliver greater cost efficiency than other ferroelectric materials.

Historically, this material has struggled with stability under adverse conditions, including thermal drift, moisture ingress and high optical frequencies. It also faces integration challenges when aiming for full foundry compatibility, hindering high-volume manufacturing. Ideal use cases for polymers include short-reach data communications (such as within data centers) where footprint and drive voltage matter more than reach or linearity. They are also ideal for non-integrated optical modules for consumer applications where cost is the main consideration.

Old reliable: Lithium Niobate

Lithium niobate is the next big thing in the optical industry. The material offers proven performance and a mature manufacturing ecosystem, reflecting its reliability across the photonics industry. Its electro-optic coefficient (~27–35 pm/V) and well-understood fabrication processes make it particularly predictable and stable.

However, the material faces integration and foundry compatibility issues, making it more complex and costly than other alternatives. It can also pose contamination and supply chain security challenges, so companies must carefully evaluate their requirements before fully investing in the material. Primary applications include high-speed modulators for data communication, as well as use in coherent optics and long-reach links where reliability is more important than ultra-compact devices.

Evaluating and balancing the benefits and drawbacks of these materials is crucial to leveraging their full potential amid technological revolutions across numerous sectors. With performance boundaries pushing toward multi-terabit interconnects, this balance will become even more consequential. When considering their choices, companies can ask themselves a few key questions:

●      What is our required bandwidth or data rate (over 100 GHz or less than 50 GHz)?

●      How compact must the device be, and what is our drive-voltage budget?

●      What foundry or wafer process flow are we targeting (Si, SiN, hybrid)?

●      What reliability or lifetime metrics do we need to guarantee?

Companies can maximize their material investments and deliver multi-year performance and cost benefits by evaluating the qualities of these materials and their corresponding use cases, positioning them to support emerging innovations across the broader technology ecosystem. By grounding materials selection in electro-optic coefficients, propagation loss, device scalability and other quantifiable metrics, organizations can keep pace with today’s bandwidth race and tomorrow’s optical evolutions.