A silicon photomultiplier (SiPM) is a solid-state sensor designed to detect and quantify very low levels of light, down to individual photons. For engineers evaluating optical sensing in applications such as LiDAR, medical imaging, or hazard detection, understanding how a silicon photomultiplier works is essential to matching its sensitivity, timing, noise, and dynamic range to the measurement.
Unlike a conventional photodiode, an SiPM uses an array of small avalanche photodiodes operated in Geiger mode. This architecture gives individual photon events a strong electrical signal while allowing the sensor to estimate light intensity from the combined response of many cells. Its performance, however, depends on operating voltage, microcell design, temperature, and the characteristics of the incident light.
How an SiPM detects photons
When silicon absorbs a photon, the photon can transfer energy to an electron, creating an electron–hole pair. A reverse-biased photodiode uses an electric field in its depletion region to move these charge carriers, producing current. The absorption probability depends on the photon’s wavelength, so a silicon sensor’s response is wavelength dependent.
From a SPAD to a microcell array
A single-photon avalanche diode (SPAD) operates with a sufficiently strong electric field that a charge carrier can trigger impact ionization. The resulting avalanche creates a much larger current from the initial charge. Because the avalanche is self-perpetuating, it must be quenched: a series resistor reduces the diode’s voltage below its breakdown voltage, stopping the discharge. The diode then recharges and can detect another photon.
A single SPAD produces a binary event. It signals that an avalanche occurred, but cannot indicate whether one or several photons arrived at that diode at the same time. An SiPM addresses this limitation by combining many independent SPADs, each with its own quenching resistor. Each SPAD-resistor pair is called a microcell.
When a microcell fires, it generates a discrete pulse and then recharges. Other microcells remain ready to detect light. The summed current from the array provides a quasi-analog measure of the incident light: more detected photons generally fire more microcells and produce more output charge. The total charge for an event depends on the number of fired microcells, their gain, and the electron charge.
Key silicon photomultiplier performance parameters
Breakdown voltage and overvoltage
The breakdown voltage is the point at which the electric field allows a Geiger avalanche to occur. SiPMs are operated above this voltage. The difference between the applied bias and breakdown voltage is the overvoltage, which influences several performance characteristics, including gain, photon detection efficiency (PDE), dark count rate, and optical crosstalk. Component data sheets specify the breakdown voltage and recommended operating conditions; these should guide bias selection.
Gain and photon detection efficiency
Gain describes the charge generated by a fired microcell relative to the charge of one electron. It depends on microcell capacitance and overvoltage. Because the output charge is quantized in microcell-sized increments, measurements of low-light pulses can show distinct photoelectron peaks. The spacing between adjacent peaks can be used to determine gain.
PDE is the probability that an incident photon produces an avalanche. It depends on wavelength, overvoltage, and fill factor—the proportion of the sensor area that is sensitive to light. The area between microcells is needed for isolation, quenching resistors, and signal tracks. Larger microcells generally provide a higher fill factor and can increase PDE and gain, but they also bring higher capacitance, longer recovery times, and lower dynamic range.
Noise, crosstalk, and afterpulsing
The principal noise source is dark count rate (DCR), caused mainly by thermally generated carriers that trigger avalanches. These events resemble photon-triggered pulses and contribute noise at the single-photon level. DCR rises with temperature and overvoltage, and scales with active area. Setting a higher detection threshold can reduce false triggers, but also affects which low-light events are registered.
Optical crosstalk occurs when photons emitted during an avalanche trigger a neighboring microcell. It can make a single detected photon appear as a multi-photon event. Its probability generally increases with overvoltage, creating a design trade-off: increasing overvoltage can improve PDE while also increasing crosstalk.
Afterpulsing has a different cause. During an avalanche, charge carriers can become trapped in silicon defects and later be released, potentially triggering another avalanche in the same microcell. An afterpulse during recovery may have less effect because the cell has not fully recharged; delayed events can matter when their rate is high.
Timing, recovery, and dynamic range
Microcell recovery time is the time needed for a fired cell to recharge to its operating voltage. A cell can fire again before it has fully recovered, but its gain will be reduced in proportion to its lower overvoltage. Microcell size affects capacitance and therefore the recharge time: larger cells tend to recover more slowly than smaller ones.
Some SiPM designs provide a separate fast output in addition to the standard anode–cathode output. This signal is formed by the summed, capacitively coupled response of the microcells. It can provide timing information, including the arrival time of the first photon in a pulse, and may be useful for higher-rate or second-photon timing measurements. Its lower capacitance can also be relevant when designing the sensor readout. The availability and characteristics of such an output are device specific.
Dynamic range depends on the total number of microcells, overvoltage, and photon wavelength. At lower light levels, the response is approximately linear. As photon flux rises, more cells are occupied and fewer remain available to fire, so the output becomes nonlinear. For a given sensor area, smaller microcells mean more cells and can extend dynamic range, while larger cells favor fill factor. The expected maximum light level should therefore be considered alongside sensitivity requirements.
Temperature and sensor selection
Temperature affects both breakdown voltage and DCR. A breakdown-voltage shift changes effective overvoltage if the bias is held constant, which can in turn change gain, PDE, and timing. For applications with substantial temperature fluctuations, bias compensation or thermal regulation can help maintain a stable overvoltage. Even with constant overvoltage, DCR changes with temperature because thermal carrier generation changes.
Selection is therefore a system-level decision, not simply a search for the highest PDE or gain. Consider the wavelength range, expected light intensity, timing needs, acceptable noise, temperature conditions, and readout approach together. Also distinguish direct radiation detection from optical detection: an SiPM’s sensitive volume is shallow, and ionization confined to one microcell does not provide a response proportional to deposited energy.
Putting SiPM characteristics into practice
SiPMs offer a compact, rugged, magnetic-field-insensitive alternative to photomultiplier tubes, with high gain at relatively low bias voltage. Their microcell architecture supports both single-photon detection and an estimate of light intensity, but brings design trade-offs among fill factor, recovery, dynamic range, PDE, and noise. For reliable integration, use device-specific data sheets to establish operating bias and performance limits, then evaluate those parameters against the application’s light level, timing, and environmental requirements.