Every photodetector does the same basic job: absorb photons, separate the resulting charges, and deliver them to an external circuit as current or voltage. The differences between detector types — photodiode, photoconductor, phototransistor, bolometer — lie in how they handle those three steps, and in what they sacrifice to optimize speed, sensitivity, spectral range, or noise.
Excitons enter the picture at the very first step. In many semiconductors, especially low-dimensional and organic materials, absorbed photons do not immediately create free electrons and holes. They create excitons. Whether those excitons dissociate into free carriers, transfer their energy, or recombine determines whether the detector produces a signal at all. In bulk silicon, exciton binding energy is small and thermal energy at room temperature dissociates excitons almost immediately, so detector design can largely ignore them. In organic semiconductors, monolayer transition metal dichalcogenides, and quantum dots, binding energies are large, and exciton physics dominates.
That makes excitonic photodetectors both promising and subtle. They can be spectrally sharp, polarization-sensitive, gate-tunable, and built from atomically thin layers. They can also be slow, inefficient, or noisy if exciton dissociation and charge extraction are not engineered carefully.
Absorption: the exciton advantage
The strength of excitonic absorption is one of the field’s most attractive properties. A monolayer of a transition metal dichalcogenide such as molybdenum disulfide or tungsten diselenide can absorb a substantial fraction of incident light at its exciton resonance despite being less than a nanometer thick. That is possible because the oscillator strength is concentrated into a narrow exciton peak rather than spread across a broad band continuum. Quantum dots and organic molecules similarly concentrate absorption into excitonic transitions.
For a detector, strong absorption in a thin layer is valuable. It means the active region can be extremely thin, which reduces transit time and can improve speed. It also means the detector can be transparent at other wavelengths, stacked with other layers, or integrated on top of silicon readout circuitry without requiring a thick absorber. In imaging arrays, thin absorbers can reduce crosstalk between pixels.
The trade-off is spectral breadth. A sharp exciton resonance gives a narrowband detector, which is excellent for spectroscopy, filtering, or wavelength-selective sensing, but poor for broadband imaging unless multiple materials or broadened absorption are combined. Many practical detectors therefore use excitonic materials not as the sole absorber but as a sensitizer or as part of a heterostructure that broadens the response while retaining some of the excitonic selectivity.
Dissociation: from bound pair to free carriers
An exciton is neutral. A detector needs charge. The conversion step — exciton dissociation — is where many excitonic detectors succeed or fail. Dissociation requires an electric field, a heterojunction with a type-II band alignment, or enough thermal energy to overcome the binding energy. In devices with large binding energies, dissociation is not automatic.
Three strategies dominate. The first is a built-in field from a p-n junction or Schottky barrier, which pulls the electron and hole apart. The second is a donor-acceptor heterojunction, familiar from organic photovoltaics, where the energy offset at the interface provides the driving force for charge separation. The third is a van der Waals heterostructure of two TMD monolayers with staggered bands, so that the electron transfers to one layer and the hole to the other, forming an interlayer exciton that can then be dissociated by a field or collected at contacts.
Each strategy has costs. A strong field improves dissociation but can increase dark current — the current that flows without light — which adds noise. A heterojunction improves separation but can introduce recombination at the interface if the materials are not well matched. A van der Waals stack offers clean interfaces but requires precise alignment and transfer, and interlayer excitons, while long-lived, can be slow to dissociate.
The design lesson is that absorption and dissociation cannot be optimized separately. A material that absorbs strongly but dissociates poorly makes a poor detector. The best devices co-design the absorber, the junction, and the contacts.
Transport and collection: getting charges out
Once dissociated, electrons and holes must travel to the contacts before they recombine. In bulk semiconductors, this is described by mobility and lifetime. In excitonic and low-dimensional devices, additional effects matter: excitons can diffuse before dissociating, charges can be trapped at defects or interfaces, and in atomically thin devices the contacts can dominate the entire device resistance.
Exciton diffusion length matters because it sets how far an exciton can travel to reach a dissociation site. In high-quality TMD monolayers, exciton diffusion can extend over distances that are large compared with device features, which is useful for funneling excitons to junctions. In disordered organic films, diffusion lengths are typically much shorter, so the dissociation interface must be within a short distance of where the exciton is created — the reason bulk heterojunctions with nanoscale intermixing became standard in organic photovoltaics and detectors.
Contact engineering is equally important. A monolayer device with poor contacts will show high resistance, slow response, and nonlinear behavior that can be mistaken for intrinsic material properties. Edge contacts, graphene contacts, and careful choice of metal work function are all used to reduce contact barriers. In phototransistors, a gate electrode modulates the channel conductivity, providing gain: one absorbed photon can modulate many carriers, producing large responsivity at the cost of slower response and higher noise.
Figures of merit: what to compare
Photodetector papers report a battery of metrics, and they are not interchangeable. Responsivity, in amps per watt, measures signal per incident optical power. External quantum efficiency measures electrons collected per photon incident. Detectivity, often written as D-star, normalizes sensitivity by area and bandwidth and is the closest single number to “how weak a signal can this detect,” but it depends on assumptions about noise and measurement bandwidth.
Response time and bandwidth describe how quickly the detector can follow a changing signal. Dark current and noise equivalent power describe the floor below which signals are lost. Linear dynamic range describes the span of light levels over which the response remains proportional to input. Spectral responsivity describes how the response varies with wavelength.
Excitonic detectors often excel on one or two metrics while trailing on others. A phototransistor based on a TMD or quantum dot film can show very high responsivity because of photoconductive gain, but its response time may be seconds or milliseconds, far too slow for communications. A p-n junction TMD detector can be fast but have modest responsivity because it lacks gain. A narrowband excitonic detector can have excellent spectral selectivity but low broadband efficiency.
Readers should therefore be skeptical of single-metric headlines. A claim of “record responsivity” without response time, dark current, and measurement conditions is incomplete. The most useful comparisons specify wavelength, bias, light power, device area, and bandwidth — and report whether the measurement was steady-state or time-resolved.
Gain versus speed: the central trade-off
The gain-speed trade-off is the most important design tension in photodetectors. Gain arises when one photon leads to many carriers passing through the circuit, typically because one carrier type is trapped and the other recirculates. That boosts signal but slows the device, because the trapped carrier must eventually be released. It also adds noise, because trapping is a stochastic process.
Photodiodes, which operate without gain, are fast and low-noise but have responsivity limited by quantum efficiency. Photoconductors and phototransistors can have large gain but are slower and noisier. Avalanche photodiodes achieve gain through impact ionization, which is fast but requires high bias and adds excess noise.
Excitonic materials offer new ways to tune this trade-off. Long-lived interlayer excitons can provide a reservoir that extends response, which is useful for memory or sensing but harmful for high-speed detection. Gate-tunable TMD devices can switch between photodiode and phototransistor operation, offering a detector whose gain can be adjusted electrically. Quantum dot detectors can combine strong absorption with tunable spectral response by changing dot size, but charge transport between dots is often slow unless the dots are closely coupled.
There is no universally best point on this trade-off. A sensor for slow environmental monitoring values sensitivity over speed. A receiver for optical communications values bandwidth above all else. A scientific imager values low noise and linearity. The right question is not “which detector is best” but “which trade-off fits the application.”
Polarization, valley, and spectral selectivity
One of the more distinctive promises of excitonic detectors is selectivity beyond intensity. In monolayer TMDs, excitons in different valleys couple to different circular polarizations of light, a consequence of valley-contrasting optical selection rules. In principle this allows a detector that distinguishes left- and right-circularly polarized light without external optics. In practice valley polarization is fragile — scattering, defects, and temperature can scramble it — but the effect is real and actively studied for polarization-sensitive imaging and information processing.
Anisotropic materials such as black phosphorus and some organic crystals show polarization-dependent absorption along different crystal axes, enabling linear polarization sensitivity in a single device. Strain and dielectric engineering can shift exciton resonances, offering a route to electrically or mechanically tunable spectral response. Stacking different TMDs or quantum dots with different bandgaps can create multispectral detectors that separate colors without filters.
These capabilities are attractive for compact imaging, remote sensing, and machine vision, where bulky filter arrays and polarizers add cost and size. The challenge is to preserve selectivity while maintaining efficiency, speed, and uniformity across an array. A single-pixel demonstration of valley-sensitive detection is a physics result; an array with uniform polarization response across thousands of pixels is a product.
Integration: on silicon, on flexible substrates, in arrays
A detector is rarely useful alone. It must be integrated with readout electronics, optics, and packaging. Silicon is the dominant readout platform, and integrating excitonic absorbers on silicon is a major research direction. Van der Waals materials can be transferred onto silicon photonics or CMOS without lattice matching, which is a significant advantage over epitaxial III-V materials. Organic and quantum dot films can be coated onto silicon readout circuits at low temperature, enabling large-area imagers.
Flexibility is another integration advantage. TMDs, organics, and quantum dot films can be deposited on plastic substrates, enabling conformable sensors, wearable photodetectors, and curved imagers. The trade-off is that flexible substrates often have poorer thermal management, lower processing temperatures, and greater mechanical stress, all of which affect stability and uniformity.
Array uniformity is the hardest integration problem. A single detector can be hand-optimized. An imaging array needs thousands or millions of pixels with consistent responsivity, dark current, and spectral response. Variations in monolayer growth, transfer, and contact quality translate directly into fixed-pattern noise. This is why many excitonic detector demonstrations remain single-pixel or small-array devices, and why scaling to megapixel imagers remains a long-term goal rather than an imminent product.
Where excitonic detectors fit
The most plausible near-term uses are not replacements for silicon cameras. Silicon is cheap, mature, and excellent across the visible spectrum. Excitonic detectors are more likely to appear where silicon is weak: in the infrared beyond silicon’s bandgap, in narrowband or polarization-sensitive sensing, in flexible or transparent form factors, and in stacked multispectral imagers where thin absorbers are an advantage.
Infrared detection is a particularly interesting case. Extending sensitivity beyond the visible requires narrow-bandgap materials such as indium gallium arsenide, mercury cadmium telluride, or two-dimensional materials such as black phosphorus. Excitonic effects in these materials are less dominant at room temperature because binding energies are smaller, but heterostructure and quantum confinement engineering can still improve absorption and selectivity. In the mid-infrared, thermal detectors and bolometers compete, and the comparison is on noise, cooling requirements, and cost, not on excitonic physics alone.
For readers evaluating a new detector claim, the checklist mirrors that for emitters: What wavelength, bias, and power? What are responsivity, detectivity, response time, and dark current together? Is the result a single pixel or an array? And was it measured at room temperature under realistic illumination? A detector that answers all of those questions clearly is worth taking seriously.
Sources and further reading
This explainer is based on review-level treatments of photodetectors and low-dimensional optoelectronics. Standard sources include textbooks on semiconductor devices and photonics, review articles in Nature Photonics, Nature Reviews Materials, Advanced Materials, Chemical Reviews, and Applied Physics Reviews on TMD photodetectors, organic photodetectors, quantum dot detectors, and van der Waals heterostructures. Detector figures of merit and measurement practices are covered in device physics textbooks and in review articles on photodetector characterization. Where specific performance numbers appear in the literature they vary widely by material, geometry, and measurement conditions, so this article avoids citing single-device record values as general facts and uses qualitative comparisons instead.