SCIENCE ROUTE · PHOTON DETECTION · AVALANCHE PHOTODIODE · MEASUREMENT BOUNDARIES

A photon can disappear inside a semiconductor and still leave a measurable electrical trace. In a single-photon avalanche diode, that trace can become a count—but the count is a property of the whole detector system, not a label carried by the photon.

Wait, What? The detected photon is gone

Photodetection often sounds as though a tiny object flies into a sensor and is simply “seen”. The real route is more interesting. When a photon with sufficient energy is absorbed in a semiconductor, its energy can create mobile charge carriers. In an avalanche photodiode, those carriers move through an internal electric field and can trigger further carrier generation through impact ionisation. The original optical event has been converted into a larger electrical response.

There are two ideas that must not be collapsed. A conventional avalanche photodiode can provide internal gain while remaining a proportional detector. A single-photon avalanche diode, or SPAD, is operated in a regime where one suitable carrier can initiate a self-sustaining avalanche that is then quenched and registered as a discrete event. “Avalanche photodiode” therefore names a device family; “one photon becomes one count” belongs specifically to a single-photon counting regime, not to every APD measurement.

Worth My While

This route connects quantum light, semiconductor physics, electronics and metrology. It explains how astronomy, lidar, fluorescence measurements, quantum communication and other low-light systems can turn sparse photons into usable data.

More importantly, it teaches a universal scientific discipline: detector events are not perfect copies of incoming reality. Some photons are missed. Some counts occur without a photon. Some avalanches influence later counts. Timing has uncertainty. A trustworthy measurement therefore includes detector behaviour inside the evidence model.

The Big Question

How does an absorbed photon become a multiplied electrical event, and how can we tell a genuine photon count from detector noise without pretending that efficiency, dark counts, afterpulsing, dead time and timing jitter belong to the photon itself?

Quick Answer

A photon enters a semiconductor and may be absorbed. If its energy creates an electron–hole pair in a region where the device’s electric field can collect and accelerate the carriers, those carriers can gain enough energy between collisions to create additional electron–hole pairs. This multiplication is the avalanche mechanism.

In a linear APD, avalanche multiplication increases the photocurrent while retaining amplitude information. In a SPAD, the device is operated so that an initiating carrier can produce a macroscopic avalanche event. Electronics detect the event and then stop and reset the avalanche. The useful observable is often a count and timestamp, not the size of the original photon’s electrical contribution.

What You Will Learn

• how photon absorption becomes mobile charge in a semiconductor;
• what impact ionisation and avalanche multiplication mean;
• why a linear APD and a SPAD are related but not equivalent;
• why detection efficiency is less than perfect;
• how dark counts, afterpulses, dead time and timing jitter affect evidence;
• why calibration belongs to the detector system rather than to the photon.

Part 1 — Primary Foundation: light can become electricity

Light carries energy. A semiconductor contains electrons arranged in allowed energy states. When a photon is absorbed under suitable conditions, its energy can move an electron into a mobile state and leave behind a corresponding hole. The electron and hole can respond to an electric field and contribute to current.

The first bridge is therefore simple: optical energy → electronic excitation → mobile charge. The photon is not still travelling inside the detector after absorption. Its energy has entered the material system.

Part 2 — Secondary Mechanism: one carrier can make more carriers

In the multiplication region of an avalanche photodiode, charge carriers move under a strong internal electric field. A sufficiently energetic carrier can collide with the lattice in a way that creates an additional electron–hole pair. Those new carriers can also be accelerated and may create further pairs.

This chain reaction is impact-ionisation avalanche multiplication. It provides internal gain: the detector’s electrical response can be much larger than the current associated with the first photo-generated carrier alone.

Part 3 — JC Depth: linear gain and Geiger-mode counting are different jobs

A linear APD is biased below the breakdown condition and provides finite avalanche gain. The output can remain approximately proportional to incident optical power over its validated operating range.

A SPAD is intentionally operated in a regime above breakdown for photon counting. A single initiating carrier can trigger a large avalanche. Once that happens, the avalanche amplitude no longer tells us the original photon energy or the exact number of initiating carriers in a simple proportional way. The event must be quenched, the device must recover, and only then is it ready for the next event.

This distinction protects a common misunderstanding: a SPAD is an excellent event detector precisely because it sacrifices simple amplitude proportionality in exchange for a large, easily discriminated pulse.

Follow One Absorbed Photon

1. Arrival: a photon reaches the detector with a wavelength to which the semiconductor can respond.
2. Transmission or loss: the photon may reflect, pass through or be absorbed. Arrival does not guarantee detection.
3. Absorption: if absorbed appropriately, the photon creates an electron–hole pair.
4. Carrier collection: the device field moves the carriers towards the multiplication region.
5. Impact ionisation: accelerated carriers may generate additional electron–hole pairs.
6. Avalanche: multiplication produces an enhanced electrical response. In SPAD operation this can become a discrete avalanche event.
7. Readout: electronics recognise the event and may assign a timestamp.
8. Quench and recovery: the avalanche is stopped and the detector returns to a sensitive state.
9. Inference: counts and timing are interpreted only after efficiency, noise and recovery behaviour are accounted for.

How Do We Know?

NIST treats single-photon detectors as metrological instruments whose performance requires more than one number. Detection efficiency, dark-count rate, afterpulse probability, dead time, reset behaviour and timing all matter. NIST calibration work has measured single-photon detection efficiency while explicitly correcting for afterpulsing, dark counts and count-rate effects.

NIST’s SPAD research also shows why avalanche charge and quenching matter. Afterpulsing can be reduced by limiting avalanche charge, and detector behaviour at high count rates depends on recovery and electronics. These are not peripheral engineering details. They determine whether a stream of electrical pulses is a faithful representation of incoming light.

Observation vs Inference

Layer	What belongs here
Incoming reality	Photons reach the detector with particular wavelengths, times and spatial distribution.
Physical interaction	Some photons are absorbed and generate charge carriers.
Detector event	An avalanche produces a measurable electrical pulse.
Recorded data	A count, timestamp or analogue signal is stored by electronics.
Inference	Photon flux, timing distribution or source behaviour is estimated after detector corrections.

The Four Main Failure Modes to Remember

1. Missed photons

Detection efficiency is not 100 per cent in ordinary systems. Reflection, incomplete absorption, carrier loss and non-triggering events can all reduce the fraction of incident photons that become recorded counts.

2. Dark counts

A detector can avalanche even when no target photon produced the event. Thermal or other internal processes can generate carriers that trigger false counts. Dark-count rate therefore belongs in any low-light evidence model.

3. Afterpulsing

Charge trapped during one avalanche can be released later and help trigger another event. That means one genuine detection can increase the probability of a later false event. Counts are therefore not always statistically independent.

4. Dead time and timing jitter

After an avalanche, a SPAD needs time to quench and recover. Photons arriving during this interval may not be registered normally. Even detected photons do not receive perfectly exact timestamps; the response has timing uncertainty known as jitter.

Worked Reasoning: twice the counts does not automatically mean twice the light

Suppose a detector records twice as many counts after the source becomes brighter. At low rates and within a calibrated linear regime, that may support roughly twice the incident photon rate. But at higher rates, dead time can make the detector miss events. Afterpulsing can add correlated false events. Background and dark counts may contribute differently between measurements.

The correct reasoning is therefore: compare corrected count rates under a validated detector model. The raw pulse count is an observation; incident photon flux is a measurement inference.

Misconceptions and Repairs

• “One photon always makes one count.” Repair: some photons are missed, and some counts occur without target photons.
• “Every APD is a single-photon detector.” Repair: linear APDs and SPADs operate in different regimes and serve different measurement jobs.
• “A larger avalanche pulse means a more energetic photon.” Repair: in SPAD counting mode the avalanche pulse is set mainly by detector dynamics, not proportional photon-energy readout.
• “Dark counts are photons from darkness.” Repair: they are detector events without the intended photon cause.
• “The detector timestamp is the exact photon arrival time.” Repair: electronics and avalanche formation introduce timing uncertainty.

Deep Science Window: the receiver creates a binary event from an analogue world

A SPAD often converts a continuous physical process into a digital-looking event: click or no click. That simplicity is powerful, but it is created by the detector architecture. The incoming optical field can contain complicated statistics, while the detector adds efficiency, noise, dead time and temporal response.

This is why photon-counting experiments often model the receiver explicitly. The scientific object is not simply “the photon”. It is the photon plus the probability that this detector, at this wavelength, in this state, produces a registered event.

Counterexamples and Model Limits

• A photon can reach the detector yet fail to produce a count.
• A count can occur with no target photon.
• Two closely spaced photons may not produce two resolvable counts because of dead time.
• Afterpulsing can make later events depend on earlier avalanches.
• Detection efficiency changes with wavelength, device design and operating condition.
• SPAD pulse amplitude should not be treated as a simple photon-energy measurement.
• Different detector technologies have different efficiencies, noise and timing behaviour; conclusions should not transfer automatically.

Evidence and Safety Boundaries

This manual explains semiconductor photodetection and avalanche multiplication at a public educational level. It does not provide bias voltages, circuit construction instructions, high-voltage procedures, device fabrication recipes or laser-operating guidance. Detailed detector engineering and laboratory safety belong to trained specialist owners.

KNOW → CONNECT → EXPLAIN → APPLY → CHECK

• KNOW: absorbed light can create mobile electron–hole pairs in a semiconductor.
• CONNECT: impact ionisation multiplies carriers and amplifies the electrical response.
• EXPLAIN: SPAD operation turns a suitable initiating carrier into a discrete avalanche event.
• APPLY: use calibrated counts and timestamps to study a light source.
• CHECK: correct or bound efficiency, dark counts, afterpulsing, dead time, wavelength response and timing jitter.

Checkpoint Questions

1. What happens to the photon when it is absorbed?
2. What is impact ionisation?
3. Why is a SPAD not simply a linear light meter with enormous gain?
4. Name two detector effects that can make raw counts differ from incident photon number.

Answer Key

1. Its energy enters the semiconductor system and can create an electron–hole pair.
2. An energetic carrier creates additional electron–hole pairs through interaction with the semiconductor lattice.
3. In counting mode an avalanche becomes a saturated discrete event that must be quenched and reset, so pulse size is not simply proportional to the initiating photon.
4. Examples include incomplete detection efficiency, dark counts, afterpulsing and dead time.

WHY Questions

• Why does internal multiplication make weak optical events easier to detect?
• Why does a more sensitive detector still need a dark-count measurement?
• Why can dead time distort measurements of a rapidly varying source?
• Why must detector efficiency be calibrated at the relevant wavelength and operating condition?

Singapore and the Wider World

Photon-counting technologies sit underneath modern optical communication, remote sensing, semiconductor research and precision measurement. These fields matter strongly to Singapore’s research and technology ecosystem. For students, the greater value is conceptual: the detector is not a passive window. It is an active physical system that transforms the signal it receives.

eduKateAI Direction Graph

incident photon → absorption probability → electron–hole pair → carrier acceleration → impact-ionisation multiplication → avalanche event → electrical discrimination and timing → quench/recovery → efficiency/noise corrections → bounded photon-count inference.

Where to Go Next

• The Physical World for photons, semiconductors, electric fields and energy transfer.
• Scientific Inquiry and Evidence for calibration, uncertainty and receiver effects.
• One Bolometer Thermal Pulse to compare photon-energy absorption measured through heating rather than avalanche charge multiplication.
• One Photoacoustic Pressure Pulse to compare absorbed light converted into heat and sound.

Authoritative and Current Sources

• NIST: Single-Photon Detectors, updated 2022 — SPAD performance, dark counts and afterpulsing.
• NIST: Calibration of Free-Space and Fiber-Coupled Single-Photon Detectors, published 2019 — detection efficiency and corrections for afterpulsing, dark counts and count-rate effects.
• NIST: A Simple Autocorrelation Method for Thoroughly Characterizing Single-Photon Detectors, published 2017 — efficiency, dark counts, afterpulse probability, dead time and reset behaviour.
• NIST: Low-Noise Photon Counting Above 100 Million Counts per Second, published 2021 — avalanche charge and high-rate detector behaviour.

Teaching Guide for Parents, Tutors and Teachers

At Primary level, keep the route to energy conversion: light enters, the material changes electrically, and the sensor produces a signal. At Secondary level, add semiconductors, electron–hole pairs and internal amplification. At JC level, separate linear APD gain from SPAD counting, then introduce efficiency, dark counts, dead time and timing jitter.

The strongest diagnostic question is: “Could this detector produce a count when no target photon arrived, and could a target photon arrive without producing a count?” If the learner answers yes to both and can explain why, they understand the evidence boundary of photon counting.