Optical & Photonic RF

Dark Current

Q: How much does photodiode dark current change with temperature?

For most semiconductor photodiodes the thermally generated component of dark current roughly doubles for every 8 to 10 degrees C rise in junction temperature. An InGaAs PIN diode with 1 nA of dark current at 25 C can climb to 8 to 16 nA at 55 C, and over 100 nA at 85 C. Diffusion-limited devices follow an even steeper exponential set by the bandgap, while generation-recombination and surface leakage components scale more gradually. Because of this strong temperature dependence, low-noise optical receivers and single-photon detectors are often thermoelectrically cooled to minus 20 C or colder, which can cut dark current by one to two orders of magnitude and substantially lower the receiver noise floor.

Q: Why does dark current matter more for an avalanche photodiode than a PIN diode?

In an avalanche photodiode the dark current has two parts: an unmultiplied surface leakage term and a bulk term that is amplified by the same avalanche gain M as the signal. The multiplied dark current grows with gain and, worse, the excess-noise factor F(M) means the shot noise it contributes rises faster than M itself, approximately as M to the power 2 plus x where x is the excess-noise exponent (0.2 to 0.5 for InGaAs, 0.02 to 0.1 for silicon). At high gain the multiplied dark current can dominate the noise budget, so APD optical receivers have an optimum gain that balances signal multiplication against multiplied dark-current shot noise. A PIN diode has no multiplication, so its dark current contributes only its own shot noise with no gain penalty.

Q: How is dark current measured on a photodiode?

Dark current is measured by placing the photodiode in complete darkness (a light-tight fixture or capped package), applying the specified reverse bias, and reading the steady-state current with a sensitive source-measure unit or picoammeter after allowing the device to settle. Specifications are typically given at a fixed reverse voltage such as 5 V or at 90 percent of breakdown for an APD, and at 25 C case temperature. Care is needed to suppress stray light, triboelectric and cable leakage, and to let the junction reach thermal equilibrium, since dark current can drift for seconds to minutes after bias is applied. Reporting current density in nanoamps per square centimeter normalizes for active-area differences between detectors.

/dahrk KUR-uhnt/

Even with the optical input completely blocked, a reverse-biased photodiode passes a small leakage current driven by thermally generated carriers, surface states, and tunneling. This dark current sits on top of the photocurrent and adds shot noise, setting the noise floor of an optical receiver and degrading sensitivity in RF-over-fiber and analog photonic links. Values range from sub-picoamp in cooled silicon to a few nanoamps for a room-temperature InGaAs PIN photodiode, and the bulk component roughly doubles every 8 to 10 degrees C. In an avalanche photodiode the bulk term is amplified by the avalanche gain, so dark current often dictates the optimum operating gain.

Category: Optical & Photonic RF

Typical (InGaAs PIN, 25 C): 0.5 to 5 nA

Temperature behavior: ×2 per 8 to 10 °C

Where Dark Current Comes From

When a photodiode is reverse biased and kept in darkness, the depletion region still generates a steady trickle of carriers. Three mechanisms dominate. Diffusion current arises when minority carriers in the neutral regions diffuse into the depletion region; it follows the intrinsic-carrier-concentration-squared term and is the limiting floor in well-passivated silicon. Generation-recombination current comes from mid-gap traps within the depletion region itself and scales with the depletion volume and the intrinsic carrier concentration. Surface leakage flows along the periphery of the junction through surface states and contamination, and is the term most sensitive to passivation quality and humidity. The relative weight of these terms determines how strongly a given detector reacts to temperature and reverse voltage.

For narrow-bandgap materials such as InGaAs (used at 1310 and 1550 nm) and germanium, the intrinsic carrier concentration is far higher than in silicon, so dark current is orders of magnitude larger at the same temperature. A silicon photodiode may show 1 to 50 pA, while an InGaAs device of similar area shows 0.5 to 5 nA and a germanium device tens to hundreds of nanoamps. This is why long-wavelength fiber receivers benefit so strongly from cooling, and why mid-wave and long-wave infrared detectors are almost always operated cryogenically. The strong temperature dependence is the practical signature engineers use to confirm that a measured leakage is genuine bulk dark current rather than stray light or instrument leakage.

The Dark-Current and Shot-Noise Equations

Diode dark current (reverse bias, no light):
I_d ≈ I_s × [exp(qV / nkT) − 1] → I_d ≈ −I_s (for V < 0)

Shot noise from dark current:
i_n² = 2 × q × I_d × B (A²)

APD multiplied dark-current noise:
i_n² = 2 × q × (I_ds + I_db × M² × F(M)) × B

Where I_s = reverse saturation current, q = 1.602 × 10⁻¹⁹ C, n = ideality factor, k = Boltzmann constant, T = junction temperature (K), B = bandwidth, I_ds = surface (unmultiplied) leakage, I_db = bulk (multiplied) leakage, M = avalanche gain, F(M) = excess-noise factor. Example: I_d = 2 nA over B = 1 GHz → i_n = √(2×1.602×10⁻¹⁹×2×10⁻⁹×10⁹) ≈ 0.80 nA RMS.

Dark Current by Detector Material

Detector	Typical wavelength	Dark current (25 °C)	Multiplied?	Common use
Silicon PIN	400 to 1000 nm	1 to 50 pA	No	Visible, short-reach fiber
Silicon APD	400 to 1000 nm	0.1 to 5 nA at M=100	Yes (low F)	LiDAR, single-photon
InGaAs PIN	1100 to 1650 nm	0.5 to 5 nA	No	Telecom, RF-over-fiber
InGaAs APD	1100 to 1650 nm	5 to 50 nA at M=10	Yes (high F)	Long-haul receivers
Germanium PIN	800 to 1600 nm	50 to 500 nA	No	Legacy / low cost

Common Questions

Frequently Asked Questions

How much does photodiode dark current change with temperature?

The thermally generated bulk component roughly doubles for every 8 to 10 °C of junction temperature rise. An InGaAs PIN at 1 nA at 25 °C can reach 8 to 16 nA at 55 °C and over 100 nA at 85 °C. Surface leakage scales more gently, while diffusion-limited devices follow a steeper bandgap-set exponential. Cooling to minus 20 °C or colder cuts dark current by one to two orders of magnitude, which is why low-noise and single-photon receivers are thermoelectrically cooled.

Why does dark current matter more for an avalanche photodiode than a PIN diode?

An APD splits dark current into an unmultiplied surface term and a bulk term amplified by the avalanche gain M. The multiplied term's shot noise grows roughly as M^2+x, where x is the excess-noise exponent (0.2 to 0.5 for InGaAs, 0.02 to 0.1 for silicon), so it can dominate the noise budget at high gain. This creates an optimum operating gain. A PIN diode has no multiplication, so its dark current adds only its own shot noise with no gain penalty.

How is dark current measured on a photodiode?

Place the device in a light-tight fixture, apply the specified reverse bias (for example 5 V, or 90% of breakdown for an APD), and read the steady-state current with a picoammeter or source-measure unit after the junction settles. Specs are usually quoted at 25 °C and often as current density in nA/cm² to normalize for active area. Suppress stray light and cable leakage, and allow seconds to minutes for the reading to stabilize.

Related Terms

← DAPS Handover Dark Fiber →