Metal additive manufacturing is fundamentally a thermal process. A laser deposits a large amount of energy into a very small region of metal, creating a molten zone that may exist for only milliseconds before solidifying. The temperature, dimensions and cooling behaviour of that molten region influence whether adjacent tracks and layers fuse correctly, how the material solidifies, and how heat builds up in the component. This is why temperature is not merely a secondary process variable. It is one of the most useful windows into what the laser is actually doing to the material.
A thermal camera turns that otherwise difficult-to-observe thermal interaction into spatial data. Instead of returning one temperature from one spot, it can show a two-dimensional temperature field: the intensely hot melt region, recently processed material behind it, steep thermal gradients around it, and cooler material farther away. Depending on the camera, optics and acquisition speed, engineers can extract melt-pool temperature, apparent width and length, cooling behaviour, heat accumulation and changes in thermal conditions from one location or layer to another. NIST research has used high-speed thermography specifically to quantify melt-pool temperatures, temperature gradients and heating and cooling rates in laser powder bed fusion.
The difficult part is that molten and polished metals are some of the hardest industrial targets for infrared thermometry. Their emissivity can be low and variable, their surfaces are highly reflective, the temperatures change extremely quickly, and a powerful processing laser may be operating only fractions of a millimetre from the measurement location. For these reasons, specialised short-wave thermal cameras can be substantially more useful than the conventional 8 to 14 µm thermal imagers commonly used for maintenance, electrical inspection and building diagnostics. NIST identifies emissivity, rapidly changing surface conditions, reflected laser radiation, optical contamination and motion as significant challenges in additive-manufacturing thermography.
How lasers melt metal, from LPBF to DED and laser cladding
In laser-based metal additive manufacturing, a focused laser produces sufficient power density for the workpiece or powder to absorb energy and reach its melting temperature. As the beam moves, the molten material solidifies behind it and becomes part of the component. In powder bed fusion, this sequence is repeated across very thin layers until the three-dimensional part has been built. NIST describes laser powder bed fusion as a process in which a focused, high-power laser repeatedly scans patterns over layers of metal powder, melting and fusing the material into the final part.
Several related terms are used in industry, and they are not all interchangeable.
Laser Powder Bed Fusion, usually abbreviated LPBF or L-PBF, is the broad description for laser-based powder bed fusion of metal. A recoater spreads a thin layer of powder across a build platform, the laser selectively melts the required cross-section, the platform changes position, another powder layer is applied, and the cycle repeats. ISO and ASTM terminology places this process within the powder bed fusion family.
Selective Laser Melting, or SLM, is widely used as a name for essentially this same class of full-melting metal LPBF process. It is best understood as an established industry term rather than as a fundamentally different thermal process from LPBF. Equipment suppliers also use terms such as Direct Metal Laser Solidification and Laser Metal Fusion for closely related implementations.
Directed Energy Deposition, or DED, works differently. Rather than beginning with a complete powder bed, material is delivered into or immediately beside an energy source as the deposition head moves. With laser DED, a laser creates a melt pool on a substrate or previous layer and powder or wire is fed into that molten region. The new material solidifies to form a bead or track. DED is particularly useful for adding material to existing components, repairing parts and building larger structures.
Laser Metal Deposition, or LMD, is a common name for laser-based DED. A laser forms the melt pool while metal feedstock is continuously added. Laser cladding uses very similar equipment and physics, but the objective is usually to apply a metallurgically bonded surface layer for wear resistance, corrosion resistance, dimensional restoration or repair rather than to build an entire three-dimensional component. TRUMPF describes laser metal deposition as a process in which a laser creates a weld pool and metal powder is continuously introduced and melted into it.
All these processes have one crucial feature in common: a small region of liquid metal created by an intense moving heat source.
That region is the melt pool.
YAG lasers, fibre lasers and the melt pool
YAG stands for yttrium aluminium garnet, a crystalline host material used in solid-state lasers. The YAG crystal is doped with laser-active ions that determine the laser's characteristics. Nd:YAG contains neodymium, while Yb:YAG contains ytterbium. Coherent describes YAG as a family of industrial laser gain crystals that can be doped with elements including neodymium and ytterbium.
The most familiar Nd:YAG fundamental wavelength is approximately 1,064 nm. Coherent documentation identifies 1,064 nm as the fundamental Nd:YAG wavelength. Ytterbium-based solid-state systems, including Yb:YAG disk lasers, commonly operate around 1,030 nm.
Modern metal-processing systems also make extensive use of ytterbium-doped fibre lasers. These are not YAG lasers because their laser-active medium is an optical fibre rather than a YAG crystal, but their wavelengths occupy a similar near-infrared region. Current IPG industrial fibre-laser specifications include systems at approximately 1,030 and 1,070 nm, with 1,070 nm particularly common in high-power material processing. IPG also identifies its approximately 1,070 nm single-mode systems as suitable for 3D printing and other precision metal processes.
This is why engineers often encounter the general "one-micron laser" family in welding and additive manufacturing: Nd:YAG near 1,064 nm, Yb:YAG around 1,030 nm, and ytterbium fibre lasers commonly around 1,030 to 1,070 nm. They are different laser architectures, but their process wavelengths are close enough that similar optical and thermal-monitoring considerations often apply.
When one of these beams hits the workpiece, only part of the incident laser energy is absorbed. The absorbed energy creates rapid local heating and eventually melting. The liquid region may be only hundreds of micrometres across in LPBF and may move rapidly with the scanning laser. DED and laser cladding usually involve larger melt pools and slower spatial dynamics. Powder, shielding gas, vapour and hot spatter can all interact with the melt pool. NIST's DED observations show powder entering the laser-heated region, melting on contact and solidifying into the deposit, while hot particles and gas flow can produce additional dynamic phenomena around it.
Melt-pool geometry matters because it provides a practical indicator of the local energy balance. Too little effective energy can leave incomplete fusion between tracks or layers. Excessive energy can create deep vapour depressions and unstable keyhole behaviour that may contribute to pores. NIST research describes lack-of-fusion pores as arising from insufficient melting and distinguishes them from keyhole-related pores formed under high-energy conditions.
Temperature and cooling rate also influence solidification and therefore microstructure. The surrounding thermal gradients contribute to thermal stress and distortion, while heat accumulating over successive scans or layers changes the starting conditions for later material. This is one reason a constant laser setting does not necessarily produce identical thermal behaviour everywhere on a complex part. NIST work on LPBF has specifically examined relationships between thermal behaviour, cooling rate, distortion, residual strain and microstructure.
What a thermal camera actually sees
A radiometric thermal camera measures electromagnetic radiation arriving from the surface and converts that signal into an estimated temperature using its calibration and assumptions about the target and optical path. It does not directly "see heat" in an abstract sense. Each pixel collects radiation over a defined spectral band and viewing angle. Planck's law relates the thermal radiation emitted by an ideal surface to wavelength and temperature, while real materials modify that radiation through their emissivity.
Imagine a thermal image of a laser travelling from left to right across a metal surface. Near the laser position, an engineer might see a small, extremely bright region corresponding to the hottest part of the melt pool. Behind that would be a longer cooling tail, representing recently molten or recently solidified material. Around the track would be lower-temperature zones where heat is conducting into the surrounding metal. Farther away, the material might be below the camera's measurable temperature range and therefore provide little or no useful radiometric signal.
From an appropriately resolved and calibrated image sequence, an engineer may extract peak or average temperature within an area of interest, apparent melt-pool length and width, position of temperature contours, cooling curves, thermal gradients and the persistence of hot regions after the laser passes. Researchers have also used melt-pool imaging to estimate cooling rates, which are important to understanding solidification behaviour.
At a larger spatial scale, thermal imaging can reveal interlayer temperature and heat accumulation. For example, a thin wall built by DED may become progressively hotter as its height increases because there is less surrounding material available to conduct heat away. A change in geometry, scan direction, dwell time or proximity to a previous hot track can similarly alter the thermal response. Such information is useful when developing scan strategies, selecting laser power and travel speed, establishing interpass delays, or determining whether active cooling or preheating is required. NIST identifies traceable temperature measurement, melt-pool dynamics and track formation as important DED metrology targets.
Thermal anomalies can also act as process signatures. A region that is unexpectedly cool might indicate reduced energy absorption, poor fusion or a material-feed problem. An unusually large or persistent hot region could indicate excessive energy input or heat accumulation. Unexpected spatter, a changing melt-pool shape or an abnormal cooling curve can likewise indicate a process departure. Studies at NIST and ORNL have investigated in-process signals for anomaly and defect monitoring, including relationships involving lack of fusion, porosity and other process irregularities.
There is an important qualification, however. A thermal anomaly is not automatically a defect measurement. Different physical mechanisms can produce similar thermal signatures, and some subsurface defects are difficult or impossible to identify reliably with a surface thermal camera alone. Production-quality defect detection therefore requires validated correlations against inspection methods such as metallography or X-ray computed tomography, and often benefits from combining thermal data with other sensors. ORNL notes that existing in-situ sensing can struggle with subsurface porosity and cracking on production LPBF machines.
Why hot reflective metals favour short-wave thermal imaging
Emissivity is one of the central ideas in infrared temperature measurement.
A perfect blackbody has an emissivity of 1. A real metal surface emits only a fraction of the radiation that an ideal blackbody would emit at the same temperature and wavelength. In simplified form, the thermal signal can be thought of as proportional to:
emitted signal ≈ emissivity × blackbody radiation at that temperature.
If the camera assumes the wrong emissivity, it interprets the observed signal as the wrong temperature. The problem becomes particularly difficult with metals because emissivity depends on wavelength, surface condition, oxidation, roughness and viewing geometry. Freshly molten, newly solidified, oxidised and powdered versions of the same alloy can therefore present different radiometric conditions. NIST notes that local effective emissivity in additive manufacturing can change substantially depending on whether the observed surface is powder, solid metal, liquid metal or hot vapour.
A conventional industrial thermal camera often operates in the long-wave infrared region around 8 to 14 µm. That wavelength range works extremely well for many moderate-temperature, relatively high-emissivity targets. Bare metals are a less favourable case. NIST's review of metal emissivity shows a general decline in spectral emissivity as wavelength increases from the visible and near-infrared into the long-wave infrared. Low emissivity also means high reflectivity for an opaque surface, so radiation reflected from surrounding objects can make a greater contribution to what reaches the camera.
A short-wave measurement addresses the problem in two ways.
First, many metals have higher spectral emissivity at shorter wavelengths than they do in the 8 to 14 µm region. That gives the instrument a stronger proportion of target-generated radiation relative to reflected background radiation.
Second, and just as importantly, the relationship between signal and temperature is much steeper at short wavelengths when measuring very hot objects. If an emissivity error causes a given percentage change in measured radiance, a short-wave instrument generally needs a smaller change in indicated temperature to account for that signal difference. NIST consequently states that shorter-wavelength thermography is less sensitive to emissivity errors because it produces a steeper signal-versus-temperature relationship.
This does not mean emissivity stops mattering. It does not. A poorly estimated emissivity can still produce significant temperature error, especially when surface state changes during melting and solidification. Short wavelength simply makes the measurement less vulnerable to a given emissivity uncertainty than a longer-wavelength measurement under comparable high-temperature conditions.
There is also a temperature trade-off. At ordinary industrial temperatures, a surface does not emit much radiation near 0.8 µm, so a camera operating there is not appropriate for measuring a 60°C machine bearing or a 150°C enclosure. At molten-metal temperatures, however, the situation changes dramatically. As temperature rises, substantial thermal radiation appears in the visible and near-infrared portions of the spectrum. NIST thermography systems have used near-infrared measurements for extremely hot LPBF melt-pool phenomena, including calibrated measurements in the 1,600 to 2,800 K region.
That is the practical reason short-wave cameras exist. An 8 to 14 µm camera is often excellent for cooler machinery. A roughly 0.8 to 1 µm camera becomes attractive when the object itself is glowing-hot metal.
Why the laser wavelength and camera wavelength should be separated
Laser processing introduces another complication: the camera is looking directly at, or very close to, a surface illuminated by an extraordinarily intense optical source.
Consider a 1,064 nm Nd:YAG laser. Some of that radiation is absorbed and becomes heat, but some is reflected or scattered. The thermal camera should ideally measure radiation caused by the resulting temperature rather than mistake reflected 1,064 nm laser light for additional thermal emission. The same issue applies to Yb:YAG near 1,030 nm and to fibre lasers around the one-micron region. NIST warns that visible and near-infrared detectors may respond to reflected laser radiation and that even a small amount reaching the detector can add erroneously to the apparent thermal signal.
This is where spectral separation becomes useful.
The PI 08M, for example, has a specified measurement band of 780 to 820 nm. That is separated from a 1,030 nm Yb:YAG laser, a 1,064 nm Nd:YAG laser and common approximately 1,070 nm ytterbium fibre lasers. Current manufacturer specifications list the PI 08M spectral range as 780 to 820 nm.
The camera therefore uses radiation around 800 nm to infer the metal's temperature. It is not attempting to measure the processing laser's output at approximately 1,030 to 1,070 nm. The processing laser creates the thermal state, while the camera observes a different spectral portion of the radiation emitted by the hot workpiece.
The distinction is important because spectral separation alone must not be confused with guaranteed laser rejection. The current PI 08M specification table lists "Optical Filter: No". A real installation must consider the complete spectral response of the detector, lenses, windows and any additional optical elements. NIST specifically cautions that a thermography system viewing a laser process may require sufficiently strong attenuation of the processing-laser wavelength, potentially by many orders of magnitude, both to prevent measurement corruption and to protect the detector.
In other words, 780 to 820 nm is advantageously separated from a 1,064 nm process laser, but an engineer should not assume that wavelength separation by itself makes any camera immune to intense reflected laser radiation. System-level optical design and laser-safety engineering still matter.
The PI 08M as a practical short-wave example
The PI 08M illustrates the design choices involved in a specialised high-temperature metal camera.
Its narrow 780 to820 nm spectral range is aimed at hot-metal and laser-processing measurements. The current technical data specify up to 764 × 480 pixels at 32 Hz, 382 × 288 pixels at 80 Hz, a 72 × 56 pixel high-speed window at 1 kHz, and a 764 × 8 line-scan mode at 1 kHz. The overall calibrated temperature capability extends from 575°C to 1,900°C, although the lower limit changes with acquisition mode. The technical datasheet specifies 575°C at 27 Hz, 625°C at 32 or 80 Hz, and 750°C in the 1 kHz mode.
Those trade-offs are important in additive manufacturing. Full-resolution imaging can be useful for seeing a larger thermal field or studying slower processes, while the small 1 kHz window can follow much faster thermal changes. High-speed LPBF melt-pool research can require even faster instrumentation depending on scan speed, field of view and the phenomenon being studied, so 1 kHz should not be interpreted as universally sufficient for resolving every melt-pool event. NIST experiments, for example, have used high-speed imaging well into the thousands of frames per second for detailed melt-pool research.
For DED, LMD, laser cladding and many laser-welding applications, a camera in this class can be particularly useful because the melt pool and surrounding thermal field are often physically larger and easier to resolve. An engineer could define areas of interest around the melt zone, monitor maximum or average temperatures, observe the cooling trail and compare the thermal signature from one deposited bead to another. The same short-wave approach can also make sense for forging, heat treatment, molten-metal handling and other processes involving sufficiently hot reflective metal.
Just as important is understanding where the PI 08M is not the right measurement tool.
A 1,500°C molten region sits naturally within its intended temperature regime. A 200°C machine housing does not. Nor would it be the obvious camera for monitoring a relatively cool powder bed, chamber wall, optical mount or general machine environment. Those applications may require a camera with a much lower temperature range and usually a longer infrared wavelength. Manufacturer guidance similarly distinguishes short-wave cameras for high-temperature metal from long-wave systems that can measure much lower temperatures.
This is an important general lesson when specifying thermography equipment. The "best" thermal camera is not the one with the shortest wavelength or highest maximum temperature. It is the one whose spectral band, calibrated temperature range, spatial resolution, acquisition speed and optics match the physical measurement.
From process development to alarms and true closed-loop control
Thermal imaging can serve several very different roles in a laser-processing machine, and separating them prevents a common misunderstanding about "process control".
The simplest role is passive monitoring. The camera records temperatures or thermal images while the process operates normally. Engineers analyse the data during parameter development, compare builds, study melt-pool behaviour, optimise laser power and travel speed, validate simulations or investigate why one region of a component behaves differently from another. NIST uses thermography in exactly this type of measurement-science work, including melt-pool geometry and cooling-rate studies.
The next level is alarm or quality monitoring. Software evaluates the thermal data and looks for a condition outside an established envelope. It might flag a peak temperature that is too low, a melt pool that becomes unusually large, an abnormal cooling profile, unexpected heat accumulation or a deposition track whose thermal signature differs significantly from validated production behaviour. That information can trigger an alarm, annotate a build record, stop a process, or identify a region for later inspection. Research programmes at NIST and ORNL have investigated this type of in-situ process-signature and anomaly analysis.
True closed-loop control goes one step further. Measurement data are fed back quickly enough that the process changes automatically in response.
The conceptual loop is straightforward:
Laser heats metal → thermal camera measures the hot zone → controller extracts temperature or melt-pool characteristics → control logic calculates a correction → laser power, travel speed, material feed rate or another controllable variable is adjusted.
The engineering implementation is much harder. The sensor must have sufficient spatial and temporal response. The optical measurement must remain stable. The measured quantity must be related reliably to the physical condition being controlled. Processing and communications must be fast enough. Most importantly, the control algorithm must know how a change in laser power, speed or feed affects the melt pool without creating instability elsewhere in the process. NIST has specifically studied the requirements for responsive control of LPBF melt-pool geometry and notes that control is intended to reduce under-melting, over-melting, residual stress and other quality problems, while also identifying significant barriers to practical implementation.
Simply installing a thermal camera therefore does not make an additive-manufacturing machine closed-loop. The camera is a sensor. Closed-loop operation emerges only when its measurements are calibrated, interpreted and integrated with a controller that can modify the process in real time.
That distinction captures the real reason thermal cameras are valuable in laser metal processing. The laser supplies energy. The melt pool reveals how the material responds. The thermal camera gives engineers a spatial and temporal measurement of that response. Short-wave imaging is especially useful when the target is extremely hot, metallic and reflective, while spectral separation from common 1,030 to 1,070 nm processing lasers can help distinguish thermal emission from the laser illumination itself.
Used carefully, thermal data can progress from simple visualisation to process development, from process development to quality monitoring, and eventually from quality monitoring to feedback control. The central objective is not merely to obtain an impressive image of glowing metal. It is to convert the thermal behaviour of the process into measurable information that engineers can understand, compare and, where the complete control system permits, act upon.