Responsivity (R)
Responsivity is how much electrical current a photodetector produces per watt of light falling on it, R = I_ph/P in A/W. It equals ηλ/1.24 with λ in µm, so a detector that converts every photon gives 0.69 A/W at 850 nm and 1.25 A/W at 1550 nm; formula, Si, Ge and InGaAs values, wavelength cutoff, and how it is measured.
Responsivity is the output current a photodetector delivers per unit of optical power incident on it: for a photodiode, the photocurrent in amperes divided by the light power in watts,
where is the photocurrent and the optical power reaching the detector. It is the conversion factor between the optical and the electrical side of every receiver: a detector with A/W turns 1 µW of light into 1 µA of current, and every noise figure that is quoted in optical units, such as the noise-equivalent power, is an electrical noise current divided by .
From quantum efficiency
Each absorbed photon of energy can contribute at most one electron of charge to the external circuit. If a fraction of the incident photons do so, which is the definition of the external quantum efficiency, the responsivity is
The constant µm·W/A is the photon energy in electronvolts times the wavelength in micrometers. With the ceiling is 0.511 A/W at 633 nm, 0.686 A/W at 850 nm, 1.057 A/W at 1310 nm and 1.250 A/W at 1550 nm. The ceiling grows with wavelength because a watt of longer-wavelength light carries more photons: 1 µW at 1550 nm is photons per second, and a photodiode with collects electrons per second from it, a photocurrent of 1.00 µA and a responsivity of 1.00 A/W.
Wavelength dependence and materials
Responsivity rises linearly with wavelength at fixed until the photon energy approaches the bandgap of the absorbing layer, where the absorption coefficient falls, the light passes through the junction unabsorbed, and collapses. The cutoff wavelengths follow from : 1.11 µm for silicon ( eV), 1.65 µm for lattice-matched InGaAs ( eV), and for germanium 1.55 µm at the direct gap of 0.80 eV with a weak indirect tail beyond it. Silicon loses responsivity well before its cutoff because its gap is indirect: the absorption length is about 20 µm at 850 nm and more than 100 µm at 1000 nm (Green 2008), so thin high-speed silicon diodes are markedly less responsive in the near infrared than thick ones. At short wavelengths the loss is the opposite one, with absorption so strong that carriers are generated in the surface layer and recombine before they are collected.
| Detector | Wavelength | Typical | (A/W) |
|---|---|---|---|
| Si PIN | 633 nm | 0.75 – 0.90 | 0.38 – 0.46 |
| Si PIN | 850 nm | 0.75 – 0.90 | 0.51 – 0.62 |
| Ge PIN | 1310 nm | 0.50 – 0.55 | 0.53 – 0.58 |
| Ge PIN | 1550 nm | 0.50 – 0.55 | 0.63 – 0.69 |
| InGaAs PIN | 1310 nm | 0.60 – 0.70 | 0.63 – 0.74 |
| InGaAs PIN | 1550 nm | 0.60 – 0.70 | 0.75 – 0.88 |
| Ideal, | 1550 nm | 1 | 1.25 |
The efficiency ranges are the typical values for each class given by Agrawal (chapter 4); the responsivities are computed from them with the formula above, and devices with good antireflection coatings sit toward the top of each range. Germanium-on-silicon photodetectors in waveguide form reach similar responsivities at 1310 nm, and tensile strain in the germanium film pushes their band edge slightly beyond 1550 nm (Michel et al. 2010).
Gain, amplifiers and other units
An avalanche photodiode multiplies the primary photocurrent by its gain , so its datasheet responsivity is , with the unity-gain value given by the formula above; a gain of 10 on an InGaAs device with of 0.8–0.9 A/W at 1550 nm gives 8–9 A/W, but the multiplied current carries excess noise, and the gain does not change . Amplified detector modules quote responsivity in V/W, which is times the transimpedance gain: a 1 A/W photodiode on a 10 kΩ amplifier gives 10 kV/W, or 10 mV per µW. Responsivity is also a small-signal quantity only while the detector is linear; at high photocurrent, space charge and series resistance make the output saturate, and the ratio of current to power falls below its low-power value.
Measuring it
Responsivity is measured by substitution: the same beam is measured with a calibrated reference power meter and then with the device under test, and the ratio of photocurrent to reference power is . The quantity that counts as must be stated, because it decides what the number includes: power incident on the chip includes the front-surface reflection in , power in the fiber before a pigtailed detector includes the coupling loss, and power in the waveguide for an integrated detector excludes the grating or edge coupler. Overfilling the active area, a reference meter used outside its calibrated wavelength range, and a light source with a broad spectrum near a band edge are the usual sources of error. The photodetector characterization article gives the procedure, and the balanced detector CMRR article shows why the responsivity match between two diodes, not their absolute value, sets the rejection of a balanced receiver.
In a noise budget the responsivity sets the signal and converts every current noise into its optical equivalent: the signal current grows as , its shot noise as , and the fixed noise of the amplifier and dark current not at all, so a higher lowers the noise-equivalent power in direct proportion. Dividing the noise-equivalent power into the square root of the detector area times the bandwidth gives the specific detectivity, , the usual figure of merit for comparing infrared detectors of different sizes. The Photodiode Noise Calculator computes from the quantum efficiency and wavelength, or takes it directly, and carries it through the photocurrent, the noise terms and the SNR.
References: G. P. Agrawal, Fiber-Optic Communication Systems, 4th ed. (Wiley, 2010), chapter 4. B. E. A. Saleh and M. C. Teich, Fundamentals of Photonics, 3rd ed. (Wiley, 2019). M. A. Green, "Self-consistent optical parameters of intrinsic silicon at 300 K including temperature coefficients," Solar Energy Materials and Solar Cells 92, 1305 (2008). J. Michel, J. Liu, and L. C. Kimerling, "High-performance Ge-on-Si photodetectors," Nature Photonics 4, 527 (2010).