The liquid crystals that make phone screens glow are finding a second career: routing light signals inside computer chips. In a burst of recent research published through early 2025, multiple teams have shown that pairing liquid crystal materials with silicon waveguides can produce photonic switches that flip faster and draw less power than many conventional designs. If the approach scales, it could help tame one of the tech industry’s most stubborn problems: the staggering electricity bill for moving data.
Why photonic switching matters now
Data centers worldwide consumed an estimated 460 terawatt-hours of electricity in 2024, according to the International Energy Agency, and that figure is climbing fast as AI workloads multiply. A large share of that energy goes not to computation itself but to shuttling data between processors, memory, and storage over copper wires and conventional optical links. Photonic switches, which route information as pulses of light rather than electrical current, promise to cut those interconnect losses. But today’s leading photonic technologies, built on materials like lithium niobate or indium phosphide, can be expensive to fabricate and difficult to integrate with standard silicon chip manufacturing.
Liquid crystals offer a potential shortcut. They are cheap, well understood from decades of display manufacturing, and highly responsive to electric fields and light. The challenge has always been speed: traditional liquid crystal switching happens on the scale of milliseconds, far too slow for data-center traffic. The new wave of research attacks that limitation from several angles.
Four approaches, four tradeoffs
The most striking speed result comes from a team that engineered ferroelectric nematic liquid crystals as a cladding layer on silicon waveguides. Their study, published in Nature Communications, demonstrated GHz-rate optical phase shifting, pushing liquid crystal devices into the same speed class as lithium niobate modulators used in today’s fiber-optic networks. The key was exploiting the ferroelectric phase of the liquid crystal, which responds to electric fields orders of magnitude faster than the nematic phases used in displays. The paper reports concrete bandwidth and insertion-loss figures measured on a standard silicon photonics platform.
A second design prioritizes energy efficiency over raw speed. Described in a preprint submitted to IEEE Photonics Technology Letters, this silicon electro-optic directional coupler uses a liquid crystal cladding to achieve an ultralow voltage-length product, meaning it needs very little electrical energy to flip a light signal from one waveguide to another. For large-scale optical interconnects where thousands of switches operate at once, shaving microwatts off each switching event adds up to meaningful power savings. Because this work has not yet completed formal peer review, its reported figures should be treated as promising but provisional.
A third line of research tackles idle power. A separate Nature Communications paper demonstrated non-volatile photonic phase shifting using a memristor-integrated resonator: once a phase state is programmed, it holds without drawing continuous current. The study is not specific to liquid crystals, but the principle applies directly. Combining non-volatile memory elements with low-energy liquid crystal switches could, in theory, shrink the total power budget for reconfigurable photonic circuits to a fraction of what today’s designs require.
The most unconventional entry eliminates electronics from the switching process entirely. A peer-reviewed paper in Advanced Photonics (vol. 6, issue 6, 066007), highlighted by the international optics society SPIE, reported sub-nanosecond all-optical switching inside a dye-doped liquid crystal droplet cavity coupled to polymer waveguides. Light controls light, with no electrode charging or discharging to slow things down. The droplet geometry keeps the active region compact enough to fit into dense photonic layouts, though the use of polymer waveguides, which are not standard in telecom-grade chips, raises integration questions.
The gaps that remain
None of these devices has been tested inside a working optical network or data-center switch fabric. Each solves a narrow technical challenge in a controlled lab setting, and several significant unknowns stand between the lab bench and a server rack.
Thermal stability is perhaps the most pressing concern. Conventional liquid crystal displays operate near room temperature, but photonic chips in data centers routinely reach 70 to 85 degrees Celsius under load. Whether ferroelectric nematic or dye-doped formulations maintain their switching properties at those temperatures has not been addressed in any of the published studies. If the materials demand tight thermal control, the cooling overhead could erase the energy savings from low-voltage operation.
Manufacturing compatibility is another open question. Silicon photonics foundries have spent years optimizing processes for materials like silicon nitride and germanium. Introducing liquid crystal layers, whether as claddings or encapsulated droplets, would require new deposition, alignment, and packaging steps. A preprint describing a silicon-organic hybrid thermo-optic switch with nematic liquid crystal cladding demonstrates that integration on a silicon platform is feasible, but thermo-optic switching is inherently slower than electro-optic or all-optical methods, limiting its usefulness for high-speed applications.
The field also lacks a shared benchmark. Researchers working on ferroelectric nematics, directional couplers, memristive resonators, and droplet cavities are each optimizing different figures of merit. Without standardized testing conditions that capture speed, power, footprint, and manufacturability together, it is difficult for network architects to compare these approaches or weigh them against established alternatives from companies like Intel, Ayar Labs, or Lightmatter, which are already shipping or prototyping silicon photonic interconnects for AI data centers.
Where liquid crystal photonics fits in the broader research landscape
As of May 2026, no commercial product based on liquid crystal photonic switching has been announced, and no major foundry has publicly disclosed plans to integrate these materials into a production process. The technology sits firmly in the exploratory research phase.
That said, the peer-reviewed results are genuinely encouraging. GHz-rate electro-optic modulation and sub-nanosecond all-optical switching were not on the liquid crystal roadmap five years ago. The fact that multiple independent groups, using different device architectures, are converging on competitive performance numbers suggests the materials science is sound, even if the engineering path to deployment remains unclear.
For the research to move forward in a meaningful way, the community will need long-term reliability data at elevated temperatures, demonstrations of switch arrays rather than single devices, and head-to-head comparisons with incumbent photonic technologies under realistic traffic loads. Until those milestones are met, liquid crystal photonic switches remain a compelling laboratory result, not yet a data-center solution, but closer to one than most observers expected.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
