A device that turns light into an electrical signal. A absorbs only a few percent of the light that hits it, but it responds quickly, can be tuned by a and can be placed on surfaces an ordinary detector cannot reach, such as a flexible sheet or an optical chip.
Going deeper
Three ways a 2D layer turns light into a signal. Trapping one carrier gives gain but slows the response; a junction is fast but yields at most one electron per absorbed photon; and in gapless graphene, heating and the thermoelectric effect often dominate.
Three ways to turn light into current
Absorbed light creates pairs, and a detector has to separate them before they recombine. In photoconductive detectors a bias pulls the pairs apart and the conductance rises while the light is on. If one carrier is caught in a trap – a defect, a or an adsorbed molecule – the other can cross the channel many times before recombination, so each moves many electrons through the circuit.
In detectors the built-in field of a p–n junction or a contact separates the pair without any bias. They are fast and quiet in the dark, but give at most one electron per absorbed photon. In photothermal detectors the light mainly heats the material: a bolometer reads the change in resistance, and a photothermoelectric detector reads the voltage from a temperature difference across a junction – the dominant mechanism in many graphene devices.
The gain–speed trade-off
Photoconductive gain is roughly the time the trapped carrier stays trapped divided by the time the free carrier takes to cross the channel. Long-lived traps therefore give enormous responsivities – graphene coated with light-absorbing has reached about 107 A/W – but the same traps set the response time, often milliseconds to seconds, and the gain collapses at higher light levels as traps fill.
A detector that is very sensitive to faint, steady light can thus be useless for video or an optical data link, where speed matters more. Junction devices and graphene detectors built into answer in picoseconds and handle signals of tens of gigahertz, at far lower responsivity.
Comparing detectors fairly
Responsivity – current per watt of light – depends on which watts are counted. Many reports divide by the power falling on the small rather than the whole beam, which flatters small devices. Specific detectivity, D*, normalises the noise to area and bandwidth, but is often estimated by assuming the noise is shot noise from the dark current alone, which overstates it when low-frequency noise dominates, as it often does in 2D devices.
A fair comparison needs the mechanism, the wavelength, the light intensity at which responsivity was measured, the bias and gate voltages, the rise time or bandwidth and a measured noise spectrum. Internal quantum efficiency – electrons per absorbed photon – also matters, because a single layer absorbs only a few percent of the light that reaches it.
For specialists
A device converting incident photons into current or voltage. A 2D channel trades absorption per pass for gate tunability, short transit times and transfer onto arbitrary or waveguides; the response may be photoconductive with gain, photovoltaic at a junction, or bolometric. Responsivity alone says little – it is only comparable alongside the mechanism, the bandwidth and the illuminated area.