In electronics, switching is almost taken for granted. A transistor uses a small gate voltage to control a larger current. That control is nonlinear: a small change in input can produce a large, sharp change in output. Cascaded transistors build logic gates, and logic gates build computers. The entire edifice of digital computing rests on a reliable, compact, low-energy switch.
In photonics, there is no equally simple switch. Photons in a typical waveguide or fiber pass through each other without interacting. That is wonderful for carrying signals over long distances without crosstalk, but it means there is no direct photonic analogue of the transistor. To make one beam of light control another, you need a material whose optical properties change in response to light. That change is called an optical nonlinearity. In most materials, it is weak. Overcoming that weakness — without paying an unacceptable price in power, size, or speed — is the central problem of all-optical switching.
Exciton-polaritons offer one of the most promising routes around the problem. Because a polariton is part exciton, it interacts with other polaritons far more strongly than bare photons interact. A modest density of polaritons can shift the energy of the polariton mode enough to switch a signal on or off, route it to a different output, or implement a logic operation. This article explains why the nonlinearity problem is hard, how polaritons address it, and what stands between laboratory demonstrations and practical optical logic.
Why photons don’t switch
The root cause is the structure of Maxwell’s equations in a linear medium. In a linear dielectric, the polarization of the material is proportional to the electric field. The superposition principle holds: fields add, and beams pass through each other unchanged. Nonlinear optics arises only when the polarization includes higher-order terms that depend on the square, cube, or higher powers of the field.
In ordinary glass or silicon, those higher-order coefficients are small. To get a large nonlinear phase shift — the kind needed for switching — you typically need one of three things: very high optical intensity, a very long interaction length, or a resonant structure that enhances the field inside the material. High intensity costs energy. Long interaction length costs footprint and latency. Resonant enhancement can help, but it narrows the operating bandwidth and makes the device sensitive to fabrication variations and temperature drift.
Engineers have exploited all three tradeoffs. Nonlinear fibers use kilometers of interaction length. Silicon waveguides exploit tight confinement to boost intensity. Microring resonators and photonic crystal cavities use resonance to enhance the field. Each approach has produced useful devices for specific applications, such as wavelength conversion or pulse shaping. None has yet produced a general-purpose, cascadable optical transistor that can compete with electronics on energy, density, and reliability for logic.
There is also a subtler issue: for logic, a switch must have gain and must restore the signal. A passive nonlinear device that merely modulates a beam may not produce an output strong enough and clean enough to drive the next stage. Without signal restoration, noise and loss accumulate through a cascade, and the circuit fails after a few stages. This is why many demonstrations described as “all-optical switches” are better understood as modulators or routers than as logic gates. A true optical transistor must combine nonlinearity with amplification or regeneration.
What nonlinearity means in practice
It helps to be concrete about what a switch must do. In a simple all-optical switch, a control beam determines whether a signal beam is transmitted or blocked, or whether it exits through one port or another. The figure of merit is often described in terms of the switching energy — how much energy the control pulse must contain — and the switching time — how quickly the device responds and recovers.
Lower switching energy is better, because it determines the total power consumption of a circuit. Faster switching is better, because it determines the clock rate. But the two are often in tension. A highly resonant device can switch with low energy but recovers slowly, because energy stays trapped in the resonator. A non-resonant device can switch quickly but requires high energy. The art of nonlinear photonics is navigating this tradeoff.
A related tradeoff involves footprint. A device that requires a long waveguide to accumulate a sufficient nonlinear phase shift may be too large to integrate at scale. A device that uses a compact resonator may be small but sensitive to fabrication errors. These tradeoffs are why the search for a material with a intrinsically large nonlinearity — one that does not require extreme intensity, length, or resonance — is so important.
Enter the polariton
An exciton-polariton inherits nonlinearity from its excitonic component. When two polaritons are in the same region, their exciton parts interact through Coulomb forces and through phase-space filling — the fact that excitons are composed of fermions and cannot occupy the same state indefinitely. The result is a blueshift: the energy of the polariton mode increases with polariton density. This density-dependent energy shift is the workhorse of polariton switching.
The effect can be surprisingly strong. In polariton microcavities, a relatively modest polariton density can produce an energy shift comparable to the linewidth of the mode. That is enough to move the mode into or out of resonance with a probe beam, switching transmission from high to low or vice versa. Because the polariton effective mass is very light, polaritons can propagate over appreciable distances within their short lifetime, allowing the interaction to be spatially extended.
Polariton systems also offer multiple control knobs. The density can be controlled optically, by pumping with a laser, or electrically, in structures with contacts. The energy landscape can be shaped by patterning the cavity, by applying stress, or by creating optically induced potentials. Researchers have used these knobs to create polariton transistors, in which a gate beam controls the flow of polaritons from source to drain, analogous to an electronic field-effect transistor.
Polariton transistors and logic demonstrations
Laboratory demonstrations over the past decade and a half have established the basic building blocks. In a typical polariton transistor geometry, polaritons are injected at a source, flow along a patterned channel, and are collected at a drain. A gate beam creates a potential barrier or modifies the local energy, controlling whether polaritons reach the drain. The device exhibits transistor-like behavior: a weak gate can control a stronger signal, and the output can be modulated.
Researchers have also demonstrated logic gates, including OR and AND-like operations, by combining polariton flows, and have shown routing and switching of polariton signals in patterned structures. Some experiments have used resonant excitation to achieve switching with low control powers, while others have used non-resonant excitation to create long-lived exciton reservoirs that sustain the potential landscape.
These demonstrations are important proofs of principle. They show that polariton nonlinearity is not a theoretical curiosity but a usable effect for signal control. They do not, by themselves, establish that polariton logic can scale. Most demonstrations involve one or a few devices, operated under carefully tuned laboratory conditions, often at cryogenic temperatures for gallium arsenide systems or at room temperature for wide-bandgap, organic, or two-dimensional material systems with their own constraints. The gap between a single working transistor and a circuit of millions is where most novel computing technologies fail.
The cascadability problem
For optical logic to be useful, gates must be cascadable: the output of one gate must be able to serve as the input to the next without external intervention that erases the advantage. Cascadability imposes several requirements that are often glossed over in single-device papers.
First, the output must have the right wavelength, mode, and coherence to drive the next gate. If each gate shifts the wavelength or degrades the signal, the cascade fails. Second, the output must be strong enough to switch the next gate — that is, the device must have gain, or at least no net loss through the cascade. Third, the logic levels must be restored: a degraded “1” must be cleaned back to a full “1” before it propagates further. Without restoration, noise accumulates.
Polariton devices face particular challenges here because polariton lifetimes are short. A polariton signal decays quickly, so a cascade must be completed within that lifetime or the signal must be converted to a photon and back, or to an electronic signal and back, at each stage. Each conversion adds loss and complexity. Some proposals address this by using polariton condensates as coherent sources that can regenerate signals, or by using hybrid designs in which polaritons handle only the nonlinear step and photons or electrons handle propagation and memory. Whether these approaches can deliver true cascadability at scale is an open research question.
Speed, energy, and footprint
Polariton switching has a compelling speed story. The underlying interaction timescales are fast, often described as picosecond-scale in the literature, which in principle supports very high operating rates. The challenge is not raw speed but energy per operation and device footprint when operated in a practical circuit.
Switching energy depends on how many polaritons are needed to produce a sufficient energy shift, and on how efficiently those polaritons can be created. In resonant schemes, the energy can be low because the pump is efficiently coupled into the mode. In non-resonant schemes, much of the pump energy goes into creating a reservoir of excitons that does not directly contribute to the signal, raising the energy cost. Cavity quality, material choice, and operating temperature all affect this balance, so reported values vary widely and should not be compared without careful attention to what is included in the energy budget.
Footprint is similarly nuanced. A single polariton device can be compact — on the scale of micrometers — but a circuit needs waveguides, couplers, pumps, and control lines, all of which consume area. The relevant comparison is not a single device versus a single transistor, but a functional circuit versus its electronic equivalent, including all overhead. That system-level comparison has not yet been demonstrated for polariton logic.
Room temperature and materials
The materials question is inseparable from the switching question. Early polariton transistor work used gallium arsenide microcavities at cryogenic temperatures, where excitons are stable and cavities can have very high quality factors. For practical computing, room-temperature operation is highly desirable, and the field has made progress with materials whose excitons are more robust.
Gallium nitride, zinc oxide, organic semiconductors, perovskites, and monolayer transition metal dichalcogenides have all been explored for room-temperature polariton devices. Each brings a different combination of binding energy, nonlinearity, cavity quality, and fabrication maturity. Organics, for example, can show strong nonlinearities at room temperature but raise questions about long-term stability under intense illumination. Two-dimensional materials offer strong excitons and compatibility with photonic integration, but large-area uniformity and interface quality remain challenges.
A practical polariton switch will need not only a material that works in a hero device but a process that yields many devices with consistent properties across a wafer. That manufacturing maturity does not yet exist for any polariton platform.
Where polaritons are most likely to matter first
Given the challenges of general-purpose optical logic, it is worth asking where polariton switching is most likely to find a first application. The most plausible near-term uses are not in replacing digital logic but in specialized functions where a strong, fast nonlinearity is the bottleneck.
One candidate is nonlinear activation in photonic neural networks. In a photonic accelerator, linear operations such as matrix multiplication can be done efficiently with interference and wavelength multiplexing, but the nonlinear activation function — which every neural network needs between layers — is hard to implement optically. A polariton device that provides a compact, fast, low-energy nonlinearity could fill that gap, even if it never implements general logic.
Another candidate is ultrafast signal routing and modulation in optical interconnects, where a polariton device switches or modulates a signal at speeds that are difficult for electronics. A third is reservoir computing, where the complex dynamics of a polariton system perform computation without requiring precise, cascadable logic gates.
In each of these cases, the device does not need to satisfy all the requirements of a general-purpose optical computer. It needs to do one job well enough to justify integration with existing electronics and photonics.
What to watch
Progress in polariton switching should be judged not by a single hero metric but by a set of system-level milestones. Can a device operate at room temperature with reproducible properties? Can two or more devices be cascaded without external regeneration? Can the switching energy be measured on a system-level basis, including pump and control overhead, and compared fairly to electronic and photonic alternatives? Can devices be fabricated with wafer-scale uniformity?
Until those milestones are met, polariton switching should be understood as a promising physics platform with demonstrated building blocks, not as a proven computing technology. The physics is genuinely distinctive: polaritons provide a nonlinearity that bare photons lack, at energy scales and speeds that are attractive for computing. The engineering challenge is to turn that nonlinearity into a reliable, manufacturable, cascadable switch.
For readers following this field, the key distinction to keep in mind is between a switch and a logic family. A switch demonstrates that light can control light. A logic family demonstrates that switches can be combined into circuits that compute reliably at scale. Polaritons have cleared the first bar in the laboratory. The second bar is where the field’s future will be decided.
Sources and further reading
This article draws on the review literature on nonlinear optics, polariton devices, and all-optical switching. For nonlinear optics fundamentals, see standard texts on nonlinear optics and reviews in journals such as Reviews of Modern Physics and Nature Photonics. For polariton transistors, logic, and switching demonstrations, see review and perspective articles on exciton-polaritons and polariton devices in Nature Photonics, Nature Reviews Physics, Optica, and Physical Review journals. For context on photonic logic and cascadability requirements, see reviews of photonic computing and photonic neural networks in Nature Photonics and Nature Reviews Electrical Engineering. Specific switching energies, times, and temperatures vary widely by material, cavity, and measurement method; this article therefore avoids asserting a single representative figure and instead describes the tradeoffs and the conditions that determine them. Readers should consult the cited reviews and follow their references to primary experimental papers for device-level numbers.