An optical transistor is a device that uses light to control light, much the way an electronic transistor uses a small electrical signal to switch or amplify a larger one. Despite decades of research and a string of impressive laboratory demonstrations, no optical transistor has yet reached mass production. The gap between what works on a laser table and what fits onto a commercial chip remains wide, but the potential payoff in speed, energy savings, and bandwidth keeps driving the field forward.
What an Optical Transistor Actually Does
In an electronic transistor, a small voltage applied to a gate terminal controls a much larger current flowing between the source and drain. The optical analog replaces those electrical signals with beams of light. A weak “gate” pulse of photons changes how a stronger “source” beam passes through a material or resonator. If the device works well, flipping the gate from off to on dramatically changes the output, and the output carries more optical energy than the gate pulse put in. That last part, gain, is critical. Without gain, optical signals degrade as they pass through successive stages of a circuit, and building complex logic becomes impossible.
Achieving that gain with photons is far harder than with electrons. Photons do not interact with each other in free space. You need a mediating material, some substance whose optical properties change enough when hit by the gate pulse that it reshapes what the source beam does. Finding materials that respond strongly, quickly, and at useful wavelengths is the central engineering challenge of the entire field.
Single-Photon Transistors and the Quantum Extreme
The ultimate test of an optical transistor is whether a single photon can act as the gate. In 2013, researchers at MIT demonstrated a device in which one stored gate photon, trapped inside an atomic ensemble within an optical resonator, produced a fivefold reduction in the transmission of subsequently applied source photons. Without retrieving the gate photon, it could switch several hundred source photons before being lost.1Science. All-Optical Switch and Transistor Gated by One Stored Photon That ratio, hundreds of output photons controlled per single input photon, is exactly the kind of gain that makes a transistor useful rather than merely interesting.
Other groups have pushed single-photon transistor concepts using different physics. One approach exploits Rydberg atoms, atoms excited to very high energy states where they interact strongly over relatively large distances. A single stored Rydberg excitation blocks the transmission of a subsequent target beam through what is known as Rydberg blockade. One experiment using a Förster resonance to enhance this blockade observed a gain of 20, meaning each gate photon suppressed 20 target photons.2PubMed. Single-photon transistor using a Förster resonance A separate proposal showed that a continuous probe beam could be disrupted by a single control photon via van der Waals interactions between Rydberg atoms, offering a potential path to continuous-wave operation rather than pulsed signals.3Physical Review A. Continuous-wave all-optical single-photon transistor based on a Rydberg-atom ensemble
On the theoretical side, cavity-based designs have been proposed in which a resonator’s coupling to a waveguide is tuned in real time so that a single photon of arbitrary shape can be absorbed, stored, and then used to shift the resonator’s frequency enough to let multiple source photons through.4Journal of the Optical Society of America B. Single-photon transistor based on tunable coupling in a cavity quantum electrodynamics system These are still theoretical blueprints more than engineering prototypes, but they suggest that the physics is sound even if the hardware is not yet ready.
The catch with nearly all single-photon transistor demonstrations is the operating environment. Rydberg-atom experiments require ultracold gases, typically a few millionths of a degree above absolute zero. Cavity quantum electrodynamics setups need painstaking alignment and extreme isolation from vibration and thermal noise. None of this is remotely compatible with the inside of a laptop.
Room-Temperature Approaches
Getting optical switching to work under everyday conditions has been a parallel research track, and progress there looks quite different from the quantum experiments. One standout result involves polaritons, hybrid particles that are part light and part matter, formed when photons couple strongly with excitations in a semiconductor. Researchers demonstrated a polaritonic switch at room temperature that operates on a timescale of hundreds of femtoseconds, roughly a hundred times faster than the fastest electronic transistors. The switch works by hitting a polariton condensate with an ultrashort optical control pulse, which scatters the condensate population through a four-wave mixing process and depletes it with a high extinction ratio.5PubMed. Optically Controlled Femtosecond Polariton Switch at Room Temperature Speed like this, pushing toward terahertz switching rates, far outpaces anything electronic circuits can do.
Silicon-based platforms have also seen recent progress. A silicon-organic hybrid switch using a photonic crystal slot nanobeam cavity filled with an organic polymer was demonstrated at a data rate of 20 gigabits per second, consuming only about 85 femtojoules per bit.6Optical Fiber Communication Conference (OFC) 2025. 85 fJ/bit Silicon-Organic Hybrid Kerr All-Optical Switch Based on Photonic Crystal Slot Nanobeam Cavity That energy figure is tiny by electronic standards, and the fact that it uses silicon as a base material is significant because silicon is already the backbone of the semiconductor industry. Any optical transistor that can be built on silicon wafers using existing fabrication tools has a much easier path to commercialization.
Going Small with Plasmonics
One of the persistent barriers to optical devices is size. Light has a wavelength measured in hundreds of nanometers, and conventional optical components tend to be at least that large. Electronic transistors, by contrast, now have features measured in just a few nanometers. This mismatch has historically kept photonics off the chip, restricted to fiber-optic links and other roles where bulk is acceptable.7PubMed. Plasmonics: merging photonics and electronics at nanoscale dimensions
Plasmonics offers one way around this. Surface plasmon polaritons are waves that propagate along the interface between a metal and a dielectric, confined to dimensions well below the free-space wavelength of the light that excited them. By exploiting this confinement, researchers have built an all-optical plasmonic switch with a footprint of just 400 by 90 nanometers. The device uses a phase-change material, a substance that can be toggled between two structural states (amorphous and crystalline) by heating, to achieve a normalized optical transmission contrast of about 98 percent between its on and off states, with an estimated switching energy of only 8 picojoules.8Optics Communications. Ultra-compact all-optical plasmonic switch with triangular resonator and phase-change materials for neuromorphic and advanced optical networks That footprint is small enough to sit alongside electronic transistors on an integrated circuit, although the metal components introduce optical losses that limit how many plasmonic stages you can cascade before the signal fades.
The Energy Equation
Energy consumption is one of the strongest arguments for optical transistors. Electronic circuits waste energy as heat every time charges move through resistive materials, and as transistor counts climb into the billions, that waste adds up. Data centers already account for a meaningful and growing share of global electricity use, and the rise of AI workloads is accelerating the trend.
Photonic approaches can slash switching energy dramatically. The silicon-organic hybrid Kerr switch mentioned earlier operates at 85 femtojoules per bit. Some electro-optic modulators, which sit at the boundary between electronics and photonics, have pushed even further. One demonstration of tightly co-designed electronic and photonic components achieved electrical energy consumption as low as 20 zeptojoules per bit, accomplished by recovering optical energy that conventional modulators waste as heat.9PubMed Central. Energy harvesting optical modulators with sub-attojoule per bit electrical energy consumption That figure, 20 zeptojoules, corresponds to roughly 500 injected electrons per switching event. For context, state-of-the-art electronic transistors involve thousands of electrons per switch.
These numbers come with caveats. The switching energy of an individual device does not account for the energy needed to generate, route, and detect the optical signals in the first place. Laser sources, waveguide coupling, and photodetectors all consume power. But the trend line is clear: at the level of the switch itself, photonics can be extraordinarily efficient.
Two-Dimensional Materials as a New Platform
Beyond silicon and plasmonics, a growing body of work explores two-dimensional materials like graphene and black phosphorus as active layers for optical switches. These atomically thin materials interact strongly with light relative to their thickness and can be integrated onto photonic waveguides in ways that bulk materials cannot. Integrated optical switches built with 2D materials have shown advantages in both speed and energy consumption compared to conventional silicon-only designs.10PubMed Central. Performance of integrated optical switches based on 2D materials and beyond
Graphene, for instance, absorbs about 2.3 percent of incident light per atomic layer, an enormous figure for something one atom thick. That absorption can be tuned electrically or optically, making graphene a natural candidate for modulator and switch applications. Black phosphorus adds the advantage of a tunable bandgap that spans from the visible into the mid-infrared, opening up wavelength ranges that graphene alone does not cover well. The practical challenge is that 2D materials are difficult to grow uniformly at wafer scale and are sensitive to oxidation and contamination. Manufacturing processes that work reliably in a research clean room do not necessarily transfer to a foundry producing millions of chips.
Building Logic Gates from Light
A transistor on its own is just a switch. To do computation, you need logic gates built from combinations of switches. In electronics, this is straightforward because electronic transistors are cheap, reliable, and cascadable. Optical logic gates have been demonstrated but remain far less mature.
A recent design using metalenses at the standard telecommunications wavelength of 1.55 micrometers showed all five fundamental Boolean logic operations: AND, OR, NOT, XNOR, and XOR. The theoretical contrast ratios ranged from about 11.5 dB for NOT up to nearly 40 dB for OR, and the bit rate for each gate type hovered around 1 terabit per second at a 50 percent duty cycle.11Physics Letters A. All-optical logical gates AND, OR, NOT, XNOR and XOR based on metalenses at 1.55 µm Terabit-per-second logic is staggering by electronic standards, but these are theoretical numbers for individual gates under ideal conditions. Stringing many gates together while maintaining signal integrity is the real engineering test, and that has not been convincingly demonstrated yet.
The deeper issue is cascadability. In electronics, the output of one transistor can directly drive the input of the next, indefinitely. In photonics, signals tend to degrade with each stage unless you add amplification, which adds complexity and noise. This is why optical gain, the ability of a transistor to output more light than it receives on its gate, is so important. Without it, every stage in a logic chain bleeds away signal until nothing recognizable comes out the other end.12Optica Publishing Group. A Silicon Optical Transistor
Why There Is No Optical Computer on Your Desk
Given all these promising results, it is fair to ask why optical transistors have not displaced electronic ones. The honest answer is that the obstacles are not just technical but systemic. The semiconductor industry has invested trillions of dollars in electronic fabrication infrastructure over more than half a century. Every new generation of electronic chip benefits from an enormous installed base of design tools, manufacturing equipment, and trained engineers. Optical computing has no comparable ecosystem.
On the technical side, the combination of gain, input-output isolation, and signal buffering that electronic transistors provide almost trivially has proven extremely difficult to replicate in optics. Storing a photon is inherently harder than storing a charge. Photons travel at the speed of light and do not sit still in a register the way electrons do in a capacitor. Optical memory remains a largely unsolved problem. One review of the field noted bluntly that mass-produced optical computers have not materialized because good solutions for optical transistors, optical memory, and many related components still do not exist.13PubMed Central. Optical Computing: Status and Perspectives
That does not mean photonics is irrelevant to computing. Photonic interconnects, optical links that move data between electronic chips, are already used in data centers and high-performance computing clusters. These take advantage of light’s strengths (bandwidth and low transmission loss) without asking it to do the computation itself. There is also growing interest in photonic accelerators for specific workloads like matrix multiplication in neural networks, where light’s parallelism offers a natural advantage. These accelerators use photonic components not as general-purpose transistors but as specialized analog processors.
The Von Neumann Bottleneck and Photonic Memory
Modern computers spend a surprising amount of time and energy simply shuffling data between the processor and memory. This fundamental limitation, rooted in the separation of processing and storage in the standard computer architecture, wastes energy as heat and limits throughput. Photonic approaches to data storage and computation have shown potential to overcome this bottleneck by enabling fast, high-bandwidth data processing directly in the optical domain, reducing the energy lost to resistive heating in electronic buses.14PubMed Central. Photonic (computational) memories: tunable nanophotonics for data storage and computing
Phase-change materials, like the ones used in the plasmonic switch described earlier, are central to this effort. These materials can hold their structural state without power, giving them a non-volatile memory quality. By embedding tiny patches of phase-change material into photonic waveguides, researchers have demonstrated rudimentary optical memory cells that can be written and read with light pulses. The challenge is scaling from isolated memory cells to the billions of bits a practical system requires, all while maintaining speed and energy advantages over electronic memory.
Thermal Crosstalk on Photonic Chips
As photonic circuits grow denser, heat management becomes a serious concern in ways that differ from electronics. Many photonic components, especially ring resonators used as filters and switches, are extremely sensitive to temperature. A shift of even a fraction of a degree can detune a resonator enough to corrupt a signal. When multiple devices sit close together on a chip, the heat generated by one affects its neighbors, a problem called thermal crosstalk.
Researchers have developed predictive models to compensate for this effect. In one study of programmable photonic circuits using microring resonators, the best thermal crosstalk models achieved prediction errors below half a picometer in resonance wavelength shift, and more sophisticated models incorporating distance between components and additional fitting parameters brought errors down further, to around 0.2 to 0.3 picometers.15arXiv. Thermal Crosstalk Modelling and Compensation Methods for Programmable Photonic Integrated Circuits These are impressively small corrections, but they illustrate the precision required when operating photonic devices at scale. Electronic circuits are robust against small temperature variations in comparison. If optical transistors are ever to be packed as densely as their electronic counterparts, solving thermal crosstalk through better materials, smarter layout, and active compensation will be mandatory.
Where the Field Is Headed
The most realistic near-term role for optical transistors is not replacing electronic logic wholesale but filling niches where light has undeniable advantages. Optical interconnects inside and between chips are already happening. Photonic neural-network accelerators are being commercialized by a handful of startups. Optical switches for telecommunications routing, where terabit-per-second throughput matters more than transistor-level logic depth, are a natural fit.
Longer term, the dream of a fully optical processor depends on solving the cascadability and memory problems simultaneously. If a single optical transistor can switch hundreds of photons with one gate photon, as the MIT group showed, then in principle you can chain stages together without electronic regeneration. But doing that reliably, at room temperature, on a chip you can manufacture by the millions, with memory elements that hold their state, and with thermal behavior that does not corrupt neighboring devices, is an engineering challenge that remains open. The physics works. The engineering, at scale, does not yet.