Photon Detection Efficiency

Photon detection efficiency (PDE) is a key performance figure for single-photon detectors. It is the probability that an incident photon produces a registered count (a “click”). This probability depends on several factors:

• Optical coupling losses: Some photons are lost before reaching the sensitive region, e.g. by parasitic reflections at an optical window or the detector surface.
• Fill factor: In detector arrays, some photons may strike inactive regions of the chip.
• Absorption efficiency: Some photons are not absorbed in the active region, particularly if the absorption coefficient at the chosen wavelength is small. They may also be absorbed in regions where carriers recombine before contributing.
• Avalanche triggering: In a Geiger-mode avalanche photodiode (SPAD), generating an electron–hole pair does not always trigger an avalanche. The trigger probability approaches unity at high excess bias but may be significantly below 100% near breakdown. (Some materials do not allow operation substantially above breakdown without excessive dark count rates.)

By definition, PDE is evaluated when the detector is ready to detect — that is, not during a “dead time” following a previous avalanche. In practice, the effective detection efficiency is reduced by dead time, hold-off, or gating duty cycle.

In some fields (e.g. superconducting nanowire detectors), the term system detection efficiency (SDE) is used to include all coupling optics and packaging losses. PDE is then the detector-intrinsic efficiency.

Most of the loss mechanisms explained above are encompassed in the conventional quantum efficiency of a photodetector. However, quantum efficiency refers only to the conversion of absorbed photons into charge carriers. PDE goes further by including the avalanche trigger probability and, for arrays, the fill factor. As a result, the PDE is generally lower than the quantum efficiency. This discrepancy is especially important for longer-wavelength SPADs, where the excess bias cannot be raised high enough for near-unity avalanche triggering.

High PDE values are achievable only within wavelength ranges where the absorption coefficient is favorable:

• Silicon SPADs: At suitable wavelengths (≈400–1000 nm), PDE values >50% are common; optimized devices can exceed 60%.
• InGaAs/InP SPADs: At telecom wavelengths (≈1.3–1.55 µm), typical PDE values are ≈10–30% under practical conditions.
• Ge-on-Si SPADs: Research devices show ≈20–40% PDE around 1.3–1.55 µm.
• Superconducting nanowire single-photon detectors (SNSPDs): PDE can exceed 80% across wide wavelength ranges.

The optimization of photon detection efficiency often involves trade-offs with other performance factors. For example, raising the bias voltage of SPADs improves efficiency, but also raises the dark count rate and afterpulsing. Therefore, fair comparisons of devices or technologies should look at efficiencies only under well-defined conditions.

Generally, the detection efficiency is substantially wavelength-dependent. For some nanostructured detectors (e.g. superconducting nanowires), PDE depends also on the polarization of light due to anisotropic absorption.

Measuring PDE is more challenging than measuring responsivity in linear-mode detectors. Two main approaches are used: