If emitters create light and detectors convert it back to electricity, modulators do the work in between: they impress information onto light, steer it, and switch it on and off. In fiber-optic communications, a modulator encodes bits onto a continuous-wave laser. In photonic computing, switches route light between waveguides. In displays and imaging, modulators control intensity and phase pixel by pixel.
Most commercial modulators today are electro-optic: an electrical signal changes the refractive index or absorption of a material, which modulates the light passing through it. Lithium niobate modulators are the workhorse of long-haul communications. Silicon photonic modulators use free-carrier effects. Electro-absorption modulators use the quantum-confined Stark effect to shift an absorption edge.
Excitonic modulators add a different lever. Because excitons interact strongly with light and with each other, a modest change in exciton density, energy, or environment can produce a large change in absorption or refractive index. In the strong-coupling regime, polaritons can exhibit nonlinearities that are orders of magnitude larger than those of conventional nonlinear crystals. That opens the possibility of all-optical switching — light controlling light without an intermediate electrical signal — at powers far below those required by traditional nonlinear optics.
The promise is real, but so are the constraints. Modulators are judged on a demanding set of metrics: speed, extinction ratio, insertion loss, power consumption, footprint, and compatibility with existing photonic platforms. A device that is fast but lossy, or low-power but narrowband, may be scientifically interesting without being commercially useful.
How modulation works: amplitude, phase, and polarization
Modulation can act on three properties of light: amplitude, phase, and polarization. Amplitude modulation changes intensity, which is how most intensity-modulated direct-detection communications systems encode bits. Phase modulation changes the optical phase, which is used in coherent communications and in interferometric switches. Polarization modulation changes the polarization state, which can encode information or route signals.
Most excitonic modulators operate through absorption or refractive index changes near an exciton resonance. Near resonance, absorption is strong and the refractive index varies rapidly with wavelength — a consequence of the Kramers-Kronig relation linking absorption and dispersion. Shifting the exciton energy, broadening it, or bleaching it therefore changes both absorption and index. A device placed in a waveguide or cavity can convert that change into a large modulation of transmitted intensity or phase.
Electrical control can shift exciton energy through the quantum-confined Stark effect, through carrier-induced screening, or through gating that changes the charge density in a monolayer. Optical control can shift exciton energy through exciton-exciton interactions, phase-space filling, or heating. The choice between electrical and optical control determines whether the device is an electro-optic modulator or an all-optical switch, with different applications and constraints.
The Stark effect and gating in 2D materials
In quantum wells and monolayer semiconductors, an electric field perpendicular to the layer shifts the exciton resonance — the quantum-confined Stark effect. In TMD monolayers, the shift can be sizable because the exciton is tightly confined, and it can be tuned with a gate voltage. This provides a compact electro-optic mechanism: a voltage changes the absorption at a fixed wavelength, modulating the transmitted light.
Gating also changes the exciton species present. In a TMD monolayer, adding electrons converts neutral excitons into charged excitons, or trions, which have different energy and oscillator strength. The absorption spectrum therefore changes not just by shifting but by redistributing weight between exciton and trion peaks. This gives an additional modulation mechanism that is distinctive to 2D materials.
The advantage of this approach is footprint. A monolayer modulator can in principle be far thinner and shorter than a lithium niobate or silicon device, because the excitonic effect is strong over a short interaction length, especially when enhanced by a cavity or waveguide. The challenges are insertion loss — a modulator that absorbs too much light even in its “on” state wastes power — and the need for high-quality, uniform monolayers with narrow linewidths. A broad exciton line reduces the modulation contrast because the absorption change is spread over a wide spectral range.
Polariton nonlinearities: why strong coupling matters
In the strong-coupling regime, the relevant excitations are polaritons, not bare excitons. Polariton-polariton interactions arise from the excitonic component, and they can be strong because excitons repel each other and because phase-space filling reduces oscillator strength at high density. The photonic component allows these interactions to be read out as changes in transmission, reflection, or emission.
The practical consequence is a large effective nonlinear refractive index. In conventional nonlinear optics, achieving a pi phase shift typically requires long interaction lengths or high powers. In a polariton microcavity, a much smaller power can produce a measurable blueshift of the polariton resonance, which in turn switches transmission. Laboratory demonstrations have shown all-optical switching with low energies per operation, sometimes described as femtojoule-scale or below, though the exact numbers depend strongly on cavity quality, spot size, and definition of switching energy.
The caveat is that low switching energy in a laboratory microcavity does not automatically translate to a practical switch. The device may require cryogenic temperature, precise wavelength tuning, or a high-quality cavity that is difficult to manufacture. It may have limited bandwidth, high insertion loss, or poor cascadability — the ability to drive a subsequent switch without amplification. For photonic logic or computing, cascadability and fan-out matter as much as switching energy. For communications, what matters is whether the switch can operate at the required data rate with acceptable loss and extinction ratio.
All-optical switching: light controlling light
All-optical switching is the most ambitious application of excitonic nonlinearities. In an all-optical switch, a control beam changes the properties of the material so that a signal beam is transmitted, blocked, or redirected. No electrical conversion is needed, which in principle avoids the speed and power costs of optoelectronic conversion.
There are two broad architectures. In a resonant switch, the control beam shifts a cavity resonance so that the signal beam moves on or off resonance. This can be very sensitive — a small shift produces a large change in transmission — but it is also sensitive to temperature drift and fabrication variations, because the resonance must be precisely aligned. In a nonresonant switch, the control beam changes absorption or index over a broader spectral range, which is more robust but requires larger nonlinearities.
Excitonic systems have demonstrated both architectures in research settings. Polariton switches using GaAs, organic, and perovskite microcavities have shown switching with short pulses and low energies. TMD-based switches have shown gate-tunable and optically controlled modulation. The gap between these demonstrations and a deployable switch is the usual one: operating conditions, integration, and reliability. A switch that works with picosecond pulses in a laboratory may not work with continuous data streams at room temperature in a packaged photonic circuit.
Electro-optic modulators for communications
For fiber-optic communications, the bar is high and well-defined. Commercial modulators operate at tens of gigabaud, with extinction ratios of many decibels, insertion losses of a few decibels or less, and the ability to run continuously at elevated temperatures. Silicon photonic modulators and lithium niobate modulators meet these specifications at scale, with mature packaging and driver electronics.
Excitonic modulators are not yet competing on these terms. Their most plausible near-term role is not as a drop-in replacement for a long-haul modulator but as a compact modulator for short-reach interconnects, on-chip photonics, or specialized wavelengths where silicon is weak. For example, a TMD modulator integrated directly on a silicon waveguide could provide modulation in a footprint far smaller than a silicon Mach-Zehnder modulator, if insertion loss and modulation depth can be made acceptable.
Another near-term role is in free-space and mid-infrared modulation, where conventional modulators are bulkier or less efficient. Excitonic materials with resonances in the infrared, such as black phosphorus or narrow-gap TMDs, could enable compact modulators for sensing and imaging. The evaluation criteria remain the same — speed, extinction ratio, loss, and stability — but the incumbent technology is less entrenched.
Figures of merit: what a modulator must prove
A modulator paper should report, at minimum, the modulation depth or extinction ratio, the insertion loss, the operating wavelength and bandwidth, the driving voltage or optical control power, the speed or bandwidth, and the energy per bit. For all-optical switches, the switching energy, the contrast, the recovery time, and whether the switch is cascadable are essential. For electro-optic modulators, the voltage-length product, the capacitance, and the driver requirements matter because they determine system power consumption.
It is common for a research demonstration to excel on one metric while omitting others. A device with a large modulation depth but high insertion loss may be useless in a link budget. A device with low switching energy but a slow recovery time cannot operate at high data rates. A device that requires cryogenic cooling may be scientifically valuable but not commercially relevant for most communications applications.
Readers should also ask about wavelength stability. A resonant modulator that drifts with temperature will require active stabilization, adding power and complexity. A nonresonant modulator that is less sensitive to drift may be preferable even if its intrinsic modulation depth is smaller.
Integration with silicon photonics
Integration is the decisive challenge for excitonic modulators, as it is for emitters and detectors. Silicon photonics is the dominant platform for integrated photonics because it leverages CMOS manufacturing, offers compact waveguides, and integrates with electronics. Any new modulator material must be compatible with silicon photonics processing or with hybrid integration approaches such as transfer, bonding, or back-end-of-line deposition.
Van der Waals materials have a natural advantage here: they can be transferred onto silicon waveguides without lattice matching, and they do not require high-temperature epitaxy. Organic materials and quantum dots can be coated onto silicon at low temperature. Perovskites can be solution-processed. The challenge is uniformity, adhesion, and long-term stability on silicon, and the need to align the excitonic resonance with the operating wavelength of the silicon circuit.
Cavity enhancement is often used to boost modulation in a short device. A ring resonator or photonic crystal cavity can amplify the effect of a small index or absorption change, reducing the required device length. The trade-off is bandwidth: a high-quality cavity is narrowband and temperature-sensitive. Designers must balance enhancement against bandwidth and stability, just as they do in laser and filter design.
Where excitonic modulators fit
The honest assessment is that excitonic modulators are a research frontier with clear physics advantages and clear engineering hurdles. They are unlikely to displace lithium niobate or silicon modulators in long-haul communications in the near term. They are more likely to appear first in niche applications where size, tunability, or all-optical operation matters more than the full set of commercial specifications: compact on-chip switches, tunable filters, polarization controllers, and scientific instruments.
The longer-term vision is photonic computing and all-optical signal processing, where the ability to switch light with light at low energy could enable new architectures. That vision depends on solving not just switching energy but cascadability, loss, and integration — problems that are as much about systems engineering as about material physics. Progress will be measured not by a single record switching energy but by demonstrations of integrated, room-temperature, cascadable switches operating at realistic data rates.
For readers following the field, the useful questions are: What is the extinction ratio and insertion loss together? What is the speed and the energy per bit? Is it electrical or optical control, and at what temperature? And can it be integrated on silicon photonics at wafer scale? A device that answers those questions convincingly is worth watching, even if it is not yet a product.
Sources and further reading
This article draws on review literature on electro-optic and all-optical modulators, exciton-polariton nonlinearities, and silicon photonics. Useful sources include review articles in Nature Photonics, Nature Reviews Materials, Advanced Optical Materials, and Journal of Lightwave Technology on modulators and switches, polariton nonlinearities, and 2D material photonics, as well as textbooks on photonics and nonlinear optics. Specific switching energies and modulation depths vary widely by platform and measurement method, so this explainer avoids citing single-device record values and instead describes the trade-offs qualitatively. Readers seeking quantitative benchmarks should consult the primary literature for the specific material system and device geometry.