In short
- Two physical families. Quantum (photon) detectors count photo-generated carriers; thermal detectors measure the heat that absorbed light produces.
- Photodiodes are the workhorse. Linear, fast and efficient, with detectivity D* of about 10¹²–10¹⁴ Jones from the UV to the short-wave IR, mostly at room temperature.
- Low light and long waves. Photomultipliers and single-photon devices reach the photon-counting level; mid- and long-wave IR photon detectors need cooling.
- Thermal detectors are broadband. Bolometers, pyroelectrics and thermopiles respond flatly from UV to far-IR, but are slower and less sensitive (D* about 10⁷–10⁹ Jones).
- Compare with NEP. NEP = √(A)/D*. The calculator below turns D*, area and bandwidth into the noise floor and the signal-to-noise ratio of your measurement.
Optical sensors in radiometry
Optical sensors convert incident radiant energy into an electrical signal. They are the core of radiometry and of most scientific instruments that measure light. They fall into two broad groups: quantum (photon) detectors, such as photodiodes and photomultipliers, and thermal detectors, such as bolometers and pyroelectric sensors.
This review covers each type’s operating principle, spectral range, performance metrics, physical characteristics and applications, and ends with a comparison table. The metrics used throughout are:
Signal-to-noise per unit power, normalised to a 1 cm² area and 1 Hz bandwidth, in Jones (cm·√Hz/W). Higher is better.
The optical power that gives a signal equal to the noise in 1 Hz bandwidth, in W/√Hz. Lower is better.
Output per watt of light, in A/W or V/W. Related to quantum efficiency for photon detectors.
Rise time or time constant. It sets the usable bandwidth and the modulation the detector can follow.
Semiconductor photodiodes (photovoltaic detectors)
Operating principle
A photodiode produces a current when absorbed photons create electron–hole pairs in a semiconductor junction. It can be operated with reverse bias (photoconductive mode: faster, but with more dark current) or at zero bias (photovoltaic mode: lowest noise). The output is linear over many decades, fast and efficient across a broad spectral band.
Spectral sensitivity
The semiconductor’s band gap sets the long-wavelength cut-off. Typical ranges by device class:
- Visible–NIR photodiodes: about 190–1100 nm, with peak responsivity near 800–900 nm of roughly 0.5–0.6 A/W.
- Near-IR photodiodes (0.8–1.8 µm): an older choice for the near IR, with higher dark current than modern SWIR photodiodes.
- SWIR photodiodes: about 900–1700 nm in the standard version, extendable to about 2.5 µm. They are best in the 1–1.7 µm telecom band and have lower dark noise than the older near-IR types.
- Extended-range IR photodiodes (2–5 µm, MWIR): much higher dark current, so they are often thermoelectrically cooled.
- Composition-tuned narrow-gap IR photodiodes: the cut-off can be tuned anywhere from about 1 to 25 µm by the alloy composition, for example to 5 µm or to the 10.6 µm CO₂ laser line. They are usually cooled to 77 K.
Key performance metrics
- Detectivity: visible–NIR photodiodes reach D* of about 10¹²–10¹⁴ Jones in the visible; an advanced device with internal gain reached about 10¹⁴ Jones at 940 nm. SWIR photodiodes reach about 10¹²–10¹³ Jones in the NIR, well above the older near-IR types (about 10¹¹ or lower). Cooled narrow-gap IR photodiodes reach about 10¹⁰–10¹¹ Jones in the mid-IR, roughly 10–50 times better than room-temperature pyroelectric detectors.
- NEP: small visible and SWIR photodiodes reach about 10⁻¹⁵–10⁻¹⁴ W/√Hz.
- Responsivity: a visible–NIR photodiode rises from about 0.2 A/W at 400 nm to about 0.6 A/W at 900 nm. Quantum efficiency can exceed 80–90% near the peak.
- Response time: nanoseconds to microseconds. Small PIN photodiodes rise in under 1 ns (GHz bandwidth); large-area SWIR photodiodes reach tens of MHz.
- Avalanche photodiodes (APDs): internal gain of about 10–100 with an excess-noise factor of about 2–5 and timing jitter of tens of picoseconds. Visible–NIR APDs cover 400–1000 nm and SWIR APDs 1000–1700 nm.
- Single-photon avalanche diodes (SPADs): about 50% photon-detection efficiency at 500 nm, with dark-count rates from a few Hz to kHz depending on cooling.
Physical and electrical characteristics
Photodiodes are compact solid-state devices with active areas of a few millimetres or less. Most work at room temperature; narrow-gap devices for longer wavelengths use thermoelectric or cryogenic cooling to cut dark current. The window is chosen for the band: fused quartz for UV and visible, IR-transmitting crystalline windows for the NIR, and zinc selenide (ZnSe) for the mid-IR. Immersion microlenses can increase the effective area and the D*.
Example applications
- Visible–NIR photodiodes: laser power meters, photometers and solar-irradiance monitors with NIST-traceable calibration, and trap detectors that serve as optical power standards.
- SWIR photodiodes: fiber-optic receivers at 1.3 and 1.55 µm, NIR spectroscopy, OCT and LIDAR.
- Cooled narrow-gap IR photodiodes: thermal imaging, gas analysis and remote sensing. In FTIR spectrometers they offer roughly 100 times the sensitivity of room-temperature pyroelectric detectors, with MHz response.
- APDs and SPADs: fluorescence-lifetime measurement, quantum communication, photon-correlation experiments and LIDAR.
Photon detectors: photoconductive and photoemissive sensors
Photon detectors convert light directly into an electrical signal, either by changing a material’s conductivity or by releasing electrons from a surface. They usually need a bias voltage or a vacuum tube. They are generally more sensitive, especially when cooled, but may need high voltage or cooling and a more complex read-out, and some respond more slowly.
Photoconductive detectors (bulk photoconductors)
In a photoconductor, absorbed photons increase the conductivity of a semiconductor, and a bias voltage turns that change into a current. Room-temperature IR photoconductors cover about 1–3 µm and 1–5 µm, with peak response near 2–3 µm. They run near room temperature, sometimes Peltier-cooled to about −30 °C. The 1–5 µm type reaches D* of about 10⁹ Jones at room temperature: about ten times a pyroelectric detector (about 10⁸) but far below cooled photodiodes. Bandwidths range from tens to hundreds of kHz.
Other photoconductors include extrinsic (impurity-doped) semiconductors, which need cryogenic cooling, and narrow-gap alloy photoconductors once used in thermal cameras, which can exceed pyroelectrics by 10–100 times in D* when cooled but show 1/f noise and need tight temperature control. Quantum-well infrared photodetectors (QWIPs) detect 8–12 µm LWIR at about 70 K with D* of about 10⁹–10¹⁰ Jones and are well suited to large imaging arrays. Quantum-dot, perovskite and 2D-material photoconductors are emerging options for broadband absorption and, potentially, higher operating temperatures.
Because photoconductors are biased, they are noisier than photovoltaic diodes, and cooling suppresses thermally generated carriers. A cooled MWIR photoconductor with a 5.5 µm cut-off at 77 K reaches D* of about 2×10¹¹ Jones. Response times range from microseconds to milliseconds. Optical chopping with lock-in detection helps manage 1/f noise and drift.
Photoemissive detectors (phototubes and PMTs)
In a photoemissive detector, light striking a photocathode in vacuum ejects electrons. A phototube collects them directly. A photomultiplier tube (PMT) multiplies them through about 10 dynode stages, for a gain near 10⁶ electrons per photoelectron.
- Sensitivity: NEP of about 10⁻¹⁶ to 10⁻¹⁸ W/√Hz, enough for single-photon detection in the UV and visible. In photon-counting mode the effective D* can exceed 10¹⁴ Jones; PMTs are usually specified by gain and dark-count rate rather than D*.
- Spectral range: standard alkali photocathodes cover about 185–900 nm; special semiconductor photocathodes extend the range to 1.0–1.7 µm at the cost of higher dark current. Peak quantum efficiency is about 20–30%, higher in the blue for some cathodes.
- Speed and practicalities: fast PMTs produce sub-nanosecond pulses; large tubes show transit-time spreads of a few nanoseconds. They need 500–1500 V, are bulky and fragile and can be damaged by intense light, yet remain the standard for extremely low light levels.
- Applications: scintillation counters for radiation, fluorescence spectrometers, single-photon counting, atmospheric LIDAR, astronomy, Cherenkov telescopes and neutrino observatories. Microchannel-plate PMTs and image intensifiers add sub-nanosecond response and 2D imaging for night vision and streak cameras.
Solid-state photomultipliers are arrays of Geiger-mode APD micro-cells. They run at about 30 V, are compact and insensitive to magnetic fields, and reach single-photon sensitivity with a gain near 10⁶. They are used in PET scanners and portable radiation detectors. Compared with large PMTs they have higher dark counts and a smaller area, and they are used mainly for visible light.
Thermal detectors: bolometers, pyroelectric sensors and thermopiles
Thermal detectors convert absorbed radiation into heat and read the resulting change in a temperature-dependent property. Their spectral response is flat from the UV to the far-IR and terahertz range, which makes them ideal for broadband and long-wavelength radiometry. They are slower, and less sensitive on fast time scales, than cooled photon detectors.
Bolometers
In a bolometer an absorber heats up relative to a heat sink, and a temperature-sensitive resistor (thermistor) measures the change. Early bolometers used blackened platinum strips. Modern uncooled thermal-imaging arrays use MEMS bridge pixels made of resistive materials with a high temperature coefficient of resistance. They need no cooling, although cooling improves sensitivity. The effective spectral range is about 0.1–1000 µm.
- Uncooled arrays (about 300 K): pixels reach an NETD of about 50 mK, an NEP of about 10⁻¹⁰ to 10⁻⁹ W/√Hz and D* of about 10⁸–10⁹ Jones in the 8–14 µm band. That is orders of magnitude below cooled IR photon detectors, but they work at room temperature and are inexpensive in large arrays. Thermal time constants of about 10–20 ms for 17 µm pixels support video at about 30 Hz, but not fast modulation.
- Cooled bolometers (4 K or even 0.1 K): can be limited only by thermal fluctuations or background photon noise. The SPIRE instrument on the Herschel space observatory used transition-edge-sensor bolometers at 0.3 K with NEPs around 10⁻¹⁷ W/√Hz. Cooled bolometers can exceed D* of 10¹² Jones in the far-IR, but they need complex cryogenics and slow read-outs.
Pyroelectric detectors
Pyroelectric crystals have a temperature-dependent electric polarisation. A change in temperature produces a displacement current, which a JFET or op-amp reads. Because they respond only to changes, they need a chopped beam or a moving target. A black coating gives a broadband response from UV to far-IR.
They work at room temperature. A typical FTIR-grade pyroelectric detector has D* of about 2×10⁸ Jones at 10 µm with 1 Hz chopping, roughly two orders of magnitude below a cooled narrow-gap IR photodiode. Thin elements respond in under a millisecond, motion alarms in under 0.1 s, and FTIR use reaches modulation frequencies of a few kHz. Temperature stabilisation limits drift caused by ambient changes.
Thermopile detectors
A thermopile is a chain of thermocouple junctions in series: the hot junctions sit on an absorber and the cold junctions on a reference, and the temperature difference generates a Seebeck voltage. The spectral range is very broad.
The DC output is proportional to absolute power, stable and linear, which is why thermopiles are used for power and irradiance standards. Sensitivity and speed are lower, with D* of about 10⁷–10⁸ Jones. A small 2 mm² thermopile might have an NEP of about 10⁻⁹ W/√Hz (D* about 10⁸) and a response time of tens of milliseconds. The output is tens of microvolts per microwatt, so low-noise amplifiers are needed.
Applications include portable radiometers, heat-flux sensors, pyranometers for solar irradiance, non-contact IR thermometers and low-cost thermal arrays such as a 32×24-pixel thermopile array.
Comparing thermal detectors
At room temperature, pyroelectric detectors usually beat bolometers and thermopiles in sensitivity and speed for modulated signals; pyroelectric arrays sit between cooled photon detectors and uncooled bolometers. Bolometers trade speed for sensitivity through thermal isolation. Pyroelectrics suit modulated signals and thermopiles suit steady DC measurements. All thermal detectors benefit from chopping or averaging.
Typical uses: uncooled bolometer arrays for night vision, surveillance and thermal diagnostics at 7–14 µm (640×480 arrays at 17 µm pitch give video-rate images); superconducting TES bolometers in telescopes that map the cosmic microwave background; pyroelectrics in FTIR spectrometers, NDIR gas analysers and multi-element PIR motion sensors; thermopiles in handheld IR thermometers, laser power meters that absorb the beam, climate-research pyrheliometers and NDIR CO₂ modules.
From D* to noise floor and SNR
Enter the detector’s D*, its active area, the measurement bandwidth and the optical power on the detector. The calculator gives NEP = √A / D*, the noise-equivalent power in your bandwidth (NEP·√Δf) and the signal-to-noise ratio. Pick a class to load a typical D* from the comparison table.
D* assumes detector-limited noise. Amplifier noise, background radiation and 1/f noise below the chopping frequency can raise the real noise floor.
Emerging and advanced optical sensors
Superconducting nanowire single-photon detectors (SNSPDs)
A current-biased superconducting nanowire develops a resistive hot spot when it absorbs a photon, which produces a voltage pulse before the wire resets. SNSPDs operate at 2–4 K. Detection efficiency is about 70–90% at the design wavelength, timing jitter is about 20 ps (20–100 ps depending on the device), dark counts are practically zero when well shielded and the NEP is of the order of 10⁻²⁰ W/√Hz. Typical devices cover about 500–2000 nm, and the range depends on the nanowire material.
Transition-edge sensors (TES) and microwave kinetic inductance detectors (MKIDs)
A TES is a superconducting film held at its transition temperature. An absorbed photon heats it slightly, which causes a sharp rise in resistance, read out by SQUID amplifiers. TES devices operate at about 0.05–0.3 K (cooled to about 50 mK in many set-ups), have near-unity quantum efficiency, resolve photon energy and photon number, and recover in tens to hundreds of microseconds. An MKID detects photons through the shift in the resonant frequency of a superconducting resonator caused by a change in kinetic inductance. MKIDs operate at about 100 mK, and many resonators can be read out together by RF multiplexing, which suits large arrays.
Nanostructured and 2D-material photodetectors
Graphene has a very low heat capacity, so a single absorbed photon can change its resistance or resonant frequency, enabling bolometric photon detection. Solution-processed hybrid perovskite photodetectors combine low dark current and high responsivity, with room-temperature detectivity of about 10¹³–10¹⁴ Jones. Colloidal quantum-dot photodiodes for the short-wave IR, which can be built directly on read-out circuits, reach about 2 µm with detectivity of about 10¹¹ Jones; their stability and noise are still being improved.
Exotic “single-electron” sensors
Skipper-readout image sensors measure each pixel’s charge non-destructively many times, reaching a read noise of about 0.1 e⁻ and allowing individual electrons to be counted. In a single-electron-transistor (SET) bolometer, a photon absorbed in a tiny island coupled through tunnel junctions changes its temperature and therefore the transistor’s conductance. SET bolometers operate below 1 K and are still experimental.
“Copper wire” based sensors
Copper-oxide (CuO) nanobelts grown on a copper wire form a cylindrical photoelectrode that detects light from every direction (0–360°) in a photoelectrochemical cell. In self-powered mode the responsivity is about 83 mA/W, over about 365–850 nm, with a response time of about 1 ms.
Applications of emerging sensors
- Quantum communication and computing, including the read-out of photonic qubits (see photon entanglement in quantum optics).
- Deep-space optical communication.
- Fundamental-physics experiments that need single-photon detection.
- Ultrafast optical signal detection, potentially at THz modulation speeds.
- Flexible and integrated photonics, such as on-chip graphene photodetectors coupled to waveguides.
- Low-cost IR imaging, such as night-vision or SWIR cameras cheaper than today’s.
- Printable or large-area sensor foils.
- Wearable or distributed sensor networks.
Comparative summary of optical sensor types
| Sensor type | Spectral range | Typical D* | Response time | Cooling | Typical applications |
|---|---|---|---|---|---|
| Visible–NIR PIN photodiode | 0.2–1.1 µm | ~10¹³–10¹⁴ Jones (peak ~0.6 µm) | ns–µs | None | Laser power meters, visible radiometry, photometry, imaging arrays |
| SWIR photodiode | 0.9–1.7 µm (extended to 2.6 µm) | ~10¹²–10¹³ Jones (1.5 µm) | ns to tens of ns | Often thermoelectric | 1550 nm receivers, NIR spectroscopy, LIDAR |
| Narrow-gap IR photodiode (LWIR) | 2–12 µm | ~10¹⁰–10¹¹ Jones (77 K, ~10 µm) | ~1 µs (MHz) | Yes, typically 77 K | FTIR, thermal imaging, IR astronomy |
| IR photoconductor (1–5 µm) | 1–5 µm | ~10⁹ Jones (room temp., 3–4 µm) | ~0.1–1 ms | Optional thermoelectric | Legacy IR spectrometers, NDIR gas analysers |
| Cooled MWIR photodiode | 1–5.5 µm | ~10¹¹ Jones (77 K, 5 µm) | ns to 100 ns | Yes, 77 K | MWIR thermal imaging (3–5 µm), tracking sensors |
| Photomultiplier tube (PMT) | 0.115–0.9 µm (to 1.7 µm with special cathodes) | >10¹⁴ Jones (single-photon level) | ~1–5 ns rise | Sometimes slight; often none | Fluorescence, scintillation, photon counting, LIDAR |
| Avalanche photodiode / SPAD | 0.4–1.0 µm | ~10¹³ Jones (linear); single-photon in Geiger mode | <100 ps jitter | Not required; reduces dark counts | Fluorescence lifetime, quantum cryptography, LIDAR, particle physics (solid-state photomultiplier arrays) |
| Uncooled bolometer array | 7–14 µm (broadband to THz) | ~10⁸–10⁹ Jones (300 K, 10 µm) | ~10–20 ms (video rate) | None | Thermal cameras, night vision, thermography |
| Pyroelectric | 0.1–100 µm | ~10⁸ Jones (10 µm, chopped) | ~1 ms (with chopping) | None | FTIR, PIR motion sensors, NDIR gas detectors |
| Thermopile | 0.1–1000 µm | ~10⁷–10⁸ Jones | ~10–30 ms | None | Pyranometers, laser power meters, IR thermometers |
| SNSPD | 0.5–5 µm (material-dependent) | Effectively unlimited (single-photon) | <0.1 ns, 20–100 ps jitter | Yes, 2–4 K | QKD, single-photon experiments, fast optical communication, astronomy |
| TES bolometer | 0.1–100 µm (with absorber) | Effectively unlimited (photon counting, energy-resolving) | ~10–100 µs per event | Yes, 0.05–0.3 K | CMB and X-ray astrophysics, photon-number-resolving quantum optics |
| Perovskite / organic photodiode | 0.3–1.0 µm (into NIR) | ~10¹³ Jones (lab demonstrations) | µs–ms | None | Research imaging, low-cost large-area and flexible sensors |
| CuO nanobelts on copper wire | 0.365–0.85 µm | n/a (responsivity ~83 mA/W) | ~1 ms | None | Experimental omnidirectional sensors, environmental and wearable sensing |
The table shows the central trade-off between sensitivity, speed, wavelength coverage and cooling: single-photon detectors top the sensitivity scale but need cryogenics, while room-temperature thermal detectors are simple and broadband but slower and less sensitive. For more practical selection advice, see your optical detector: 10 tips, and for units and quantities, the basics of light measurement: radiometry and photometry.
Conclusion
Semiconductor photodiodes remain the workhorse of radiometry: linear, fast and efficient from the UV to the short-wave IR. Photon detectors such as photomultipliers and cooled photoconductors and photodiodes cover ultra-low light levels and the mid-IR. Thermal detectors give flat broadband coverage at the cost of speed and sensitivity. Emerging sensors push toward single-photon sensitivity, energy resolution and new form factors. The right choice follows the measurement: wavelength, signal level, bandwidth, and how much cooling and complexity the system can carry. IZAK builds this choice into its multi-channel spectro-radiometer, multi-channel power meter and LED tester.
References
- Photodiode manufacturer, “Photodiode terminology”, application note.
- FTIR instrument manufacturer, “Detectors for Fourier transform spectroscopy”, application note, 2007.
- Sensor-industry trade publication, interview with an IR detector manufacturer, 2021.
- Photonics component supplier, “Photodiode materials overview”, technical tutorial.
- Photodiodes with internal gain, Optics Express 20(5), 5518, 2012.
- FTIR instrument manufacturer, FTIR detector manual, 2007.
- J. Liu et al., review of single-photon detectors, 2024.
- W. Feng et al., CuO nanobelt array photodetector on copper wire, ACS Applied Materials & Interfaces.
- C. L. Tan and H. Mohseni, “Emerging technologies for high performance infrared detectors”, Nanophotonics, 2018.
- J. X. Zhou et al., visible-to-near-infrared organic photodiodes, 2020.
- A. Rogalski and K. Chrzanowski, “Infrared devices and techniques (revision)”, 2009.
- K. Chrzanowski and K. Skorupski, selection of the modulation frequency of thermal radiation, 2008.
- A. Rogalski, “Next decade in infrared detectors”, Proc. SPIE, 2017.
- Optics distributor, “Pyroelectric detectors: comparison of common crystal types”, technical note, 2023.
- Y. Ou et al., uncooled thermal infrared detection near the physical limit, 2024.
Frequently asked questions
QWhat is the difference between D* and NEP?
QWhen should I choose a thermal detector instead of a photodiode?
QWhy do mid- and long-wave IR photon detectors need cooling?
Tzachi Sabati
CEO, IZAK Scientific
Physicist specializing in photonics and quantum technologies, with deep expertise in quantum sensors and advanced optical systems. Leads the Advanced Quantum Lab course at the Technion, bridging academic excellence with industry innovation. At IZAK Scientific, provides cutting-edge photonics-based solutions, developing customized inspection and sensing systems for R&D and production. Passionate about advancing quantum sensing applications and integrating novel technologies to meet industry needs.
