Modern optical networks and photonic AI chips face a fundamental energy paradox: to keep a circuit "tuned" or a switch "on," we must constantly pump power into it. Traditional silicon photonics rely on the thermo-optic effect, where integrated heaters maintain a specific temperature to shift the refractive index. The moment the power is cut, the circuit loses its state.
To build "set-and-forget" photonic integrated circuits (PICs), researchers are integrating phase-change materials (PCMs)—substances that can switch between amorphous and crystalline states. However, the standard PCMs used in rewritable DVDs, such as $Ge_2Sb_2Te_5$ (GST), are notoriously "lossy" at telecommunication wavelengths. They act like a dark curtain, absorbing the very light they are meant to route.
A new direction in the field, led by researchers at institutions like the University of Washington and MIT, replaces these opaque materials with wide-bandgap, transparent PCMs. By using materials like antimony trisulfide ($Sb_2S_3$) and antimony selenide ($Sb_2Se_3$), they have unlocked non-volatile optical components that maintain high signal integrity without consuming static power.
The Problem with "Lossy" Switching
In a traditional PCM like GST, the transition from an amorphous (disordered) to a crystalline (ordered) state causes a massive change in both refractive index and optical absorption. As Rui Chen and Zhuoran Fang note in Broadband nonvolatile electrically programmable silicon photonic switches, this absorption is a dealbreaker for complex circuits. If every switch in a network dims the light, the signal vanishes after passing through only a few components.
This "insertion loss" limits scalability. For example, in photonic neural networks, where light must pass through banks of "synapses" to perform calculations, high absorption prevents the creation of large-scale weight banks. Jiawei Meng and colleagues emphasize in Electrical programmable multilevel nonvolatile photonic random-access memory that parasitic absorption is the primary bottleneck for scaling photonic RAM (P-RAM).
Transparency through Bandgap Engineering
The breakthrough lies in moving to materials with a wider "bandgap"—the energy gap that electrons must jump to absorb a photon. Materials like $Sb_2S_3$ and $Sb_2Se_3$ have bandgaps high enough that they remain transparent in the near-infrared and telecommunication bands (around 1550 nm), even when they crystallize.
Zhuoran Fang and Jiajiu Zheng demonstrated this in Non‐Volatile Reconfigurable Integrated Photonics Enabled by Broadband Low‐Loss Phase Change Material, showing that $Sb_2S_3$ provides strong phase modulation (changing the speed of light) with almost zero induced loss. This allows for "pure phase shifters"—devices that change the phase of light without changing its intensity. Carlos Ríos and Tian Gu further refined this in Ultra-compact nonvolatile phase shifter based on electrically reprogrammable transparent phase change materials, creating devices that are both ultra-compact and energy-efficient.
Deterministic and Multi-level Control
The goal isn't just binary (on/off) switching; it's precise, analog-like control. By segmenting the PCM or using specialized microheaters, researchers can achieve "multi-level" states, effectively storing multiple bits of data in a single optical cell.
In Non-volatile electrically programmable integrated photonics with a 5-bit operation, Rui Chen and Zhuoran Fang achieved 5-bit (32-level) programming. To overcome the inherent unpredictability of how these materials crystallize, they used a "digital" multi-segment heater design. Instead of trying to partially crystallize a single block of material (which is messy and inconsistent), they switch several small segments independently. This "digital" approach, also explored by Minho Choi in Deterministic quasi-continuous tuning of phase-change material integrated on a high-volume 300-mm silicon photonics platform, ensures that the device state is deterministic and repeatable.
Expanding the Frontier: Cryogenics and Inverse Design
The utility of transparent PCMs extends beyond standard computing. One of the most promising applications is in quantum computing, where components must operate at temperatures near absolute zero. Traditional thermal tuning is impossible here because the heat would destroy the fragile quantum states.
Uthkarsh Adya and Mo Li demonstrated in Non-volatile tuning of cryogenic optical resonators that PCMs can tune silicon resonators at sub-4K temperatures. Once the state is set via a brief electrical pulse, it remains stable without further heating, solving a major scaling hurdle for cryogenic systems.
Finally, the design of these devices is moving away from simple geometric shapes toward "inverse design." In Inverse design of compact nonvolatile reconfigurable silicon photonic devices with phase-change materials, Maoliang Wei and colleagues used algorithms to evolve complex, pixelated structures that incorporate $Sb_2Se_3$. These devices are not only smaller but also more resilient to manufacturing errors, proving that transparent PCMs are ready to move from lab curiosities to foundational components of the next generation of optical hardware.