Lock-in thermography is a non-destructive testing technique that finds hidden defects beneath the surface of a material by heating it with a periodically modulated energy source and reading the resulting thermal signature with an infrared camera. Unlike a simple thermal snapshot, the technique extracts both the amplitude and the phase of the temperature oscillation at each pixel, which makes it remarkably sensitive to subsurface features that would be invisible on a raw thermal image. It grew out of the merger of two older fields, photothermal radiometry and infrared imaging, and has since become a workhorse inspection method across aerospace, electronics, energy, and increasingly in medicine.
How the Technique Works
The core idea is to apply energy to a sample in a repeating wave pattern, typically a sinusoidal cycle at a chosen frequency, and then watch how the resulting heat ripples through the material. A lamp, laser array, or other heat source illuminates the surface, cycling on and off (or bright and dim) at a set frequency. An infrared camera records the temperature response over many cycles. Software then performs a mathematical correlation between the known input signal and the camera’s output, extracting two pieces of information at every pixel: how large the temperature swing is (the amplitude image) and how much the temperature wave lags behind the heating wave (the phase image).
Phase images are especially powerful. Because the phase lag depends on what the heat wave encounters on its way through the material, a buried void, crack, or delamination will delay the returning signal differently than sound material does. Phase images also tend to suppress surface artifacts like uneven emissivity or non-uniform heating, making them cleaner and easier to interpret than raw temperature maps. This combination of whole-area optical heating with phase-sensitive detection is what gives lock-in thermography its edge: it can reveal hidden defects quickly and without physical contact.1Revue Générale de Thermique. Lock-in thermography for nondestructive evaluation of materials
The excitation frequency you choose determines how deep the thermal wave penetrates. Lower frequencies produce longer thermal diffusion lengths, reaching deeper into the material but with less sensitivity to fine detail near the surface. Higher frequencies probe shallower depths with better spatial resolution. In practice, an inspector often runs multiple tests at different frequencies to build a depth profile of the part.
Finding Hidden Flaws in Composite Materials
One of the technique’s strongest applications is inspecting carbon-fibre-reinforced polymers and other composites used in aerospace structures. Delaminations, where layers of the composite separate internally, are particularly dangerous because the surface can look flawless while the material is structurally compromised underneath. Lock-in thermography handles this well. In laboratory tests on carbon-fibre panels containing simulated delaminations at various depths, phase images successfully revealed every single embedded flaw, even in difficult scenarios where thin defects overlapped at different depths.2Proceedings of the Institution of Mechanical Engineers, Part G: Journal of Aerospace Engineering. Detection of Delamination in Carbon-Fibre-Reinforced Polymers with Lock-In Thermography
Beyond simply finding the defects, researchers showed they could measure both the lateral size and the depth of each flaw from the phase images alone. The depth estimates matched the known positions well for most specimens, though accuracy dropped when the defects were very deep or when multiple overlapping layers complicated the thermal path. That kind of quantitative output, telling you not just that a flaw exists but where it sits in three dimensions, is what makes the technique attractive for safety-critical structures like aircraft fuselages and wind-turbine blades.
Solar Cells and Microelectronics
The electronics industry has adopted a variant called dark lock-in thermography (DLIT), where instead of shining light on the device, you drive a modulated electrical current through it. Any spot that draws abnormal current, because of a short circuit, a high-resistance defect, or a localized shunt path, heats up slightly. The infrared camera picks that heat up and the lock-in processing pinpoints the location. For solar cells, DLIT can map the spatial distribution of series resistance and shunt resistance across the entire cell surface, turning amplitude and phase images into resistance maps under small-signal conditions.3Journal of Applied Physics. Evaluation of the spatial distribution of series and shunt resistance of a solar cell using dark lock-in thermography
In integrated circuits the demands are even tighter. A modern chip packs billions of transistors into a few square centimeters, so finding a single electrical defect like a short or a resistive open requires micrometer-level resolution. Lock-in thermography adapted to this scale can identify and locate these thermally active electrical defects, making it a valuable step in failure analysis workflows for semiconductor manufacturers.4Materials Science and Engineering: B. Application of lock-in thermography for failure analysis in integrated circuits using quantitative phase shift analysis
The electrically driven approach highlights a broader point: the “lock-in” part of lock-in thermography refers to the signal-processing strategy, not to any single heating method. You can deposit energy with lamps, lasers, ultrasonic transducers, induction coils, or electrical bias, and then use the same phase-extraction math to pull the useful signal out of the noise.
Tracking Fatigue Cracks in Metals
Metals under repeated loading eventually develop fatigue cracks, and catching those cracks early is the difference between a scheduled repair and a catastrophic failure. Lock-in thermography contributes here through thermoelastic stress analysis. When a metal is cyclically loaded, it heats and cools slightly in response to the changing stress field. An infrared camera synchronized to the loading frequency can map those tiny temperature changes across the surface, effectively producing a full-field stress map without gluing on a single strain gauge.
Researchers have used this approach to monitor crack initiation and growth in real time during fatigue tests. In one set of experiments, the thermal stress field remained stable for tens of thousands of loading cycles and then began shifting once damage accumulated enough to alter the local strain pattern, providing an early warning signal well before the specimen failed completely.5PubMed Central. Detection and tracking of cracks based on thermoelastic stress analysis The technique has also been applied to metal-matrix composites, where the measured temperature profiles can be converted into stress values using the material’s thermoelastic coefficient, and from there into crack growth rates.6Strain. Crack‐Tip Stress Field and Fatigue Crack Growth Monitoring Using Infrared Lock‐In Thermography in A359/SiCp Composites
The same lock-in processing concept has been paired with digital image correlation, a technique that tracks surface deformation optically, to study crack propagation in welded joints. By applying the lock-in algorithm to both thermal and optical data simultaneously, both crack length and the stress concentration at the crack tip can be evaluated as a function of loading cycles.7International Journal of Fatigue. Crack growth analysis in welded and non-welded T-joints based on lock-in digital image correlation and thermoelastic stress analysis
Inspecting Protective Coatings
Thermal barrier coatings protect turbine blades and other hot-section components in jet engines and power plants. If the coating debonds from the metal substrate underneath, it can spall off in service, exposing the base metal to temperatures it was never designed to survive. Lock-in thermography has been tested on this problem using specimens made of a nickel-based superalloy coated with zirconia, a common thermal-barrier system. The technique successfully detected debonding defects between the ceramic coating and the metal substrate.8Applied Thermal Engineering. Experimental study on debonding defects detection in thermal barrier coating structure using infrared lock-in thermographic technique
Coating inspection is a demanding case because the layers involved are often very thin and the thermal contrast between a bonded and a debonded region can be subtle. The phase image’s ability to suppress surface noise makes it better suited to this task than a simple infrared snapshot would be.
How Deep Can It See
The depth that lock-in thermography can probe depends on the thermal properties of the material and the excitation frequency. In general, lower frequencies push the thermal wave deeper, but at a cost: longer measurement times and coarser spatial resolution. The “blind frequency” method is a widely used approach where you sweep through several excitation frequencies and identify the frequency at which a defect just becomes invisible in the phase image. That frequency, combined with the material’s thermal diffusivity, gives an estimate of defect depth.
This works well for large, shallow defects, but the physics gets trickier as defects become smaller relative to their depth. A recent modeling study found that the blind frequency method can underestimate defect depth by as much as about 21% when the ratio of defect width to depth drops to 0.5, because lateral heat conduction around the defect blurs the thermal contrast. To compensate, the researchers developed a correction model that reduces the depth-estimation error to within roughly 5%. They also showed that another common approach, the phase difference method, becomes unreliable unless the defect is at least three times wider than the thermal diffusion length at the chosen frequency.9PubMed Central. Quantitative Depth Estimation in Lock-In Thermography: Modeling and Correction of Lateral Heat Conduction Effects
The practical takeaway is that depth estimates from lock-in thermography are reliable for reasonably sized defects but should be treated with caution when the flaw is small compared to how deep it sits. For critical applications, combining multiple frequencies and applying correction factors improves accuracy considerably.
How It Compares to Pulse Thermography
Pulse thermography, where you fire a short burst of energy at the surface and record how the temperature decays, is the other major branch of active thermography. It is faster per measurement because a single flash can be over in milliseconds. Lock-in thermography, by contrast, requires many modulation cycles and therefore longer acquisition times, sometimes minutes per frequency. But the two approaches have different strengths depending on what you are looking for.
A direct comparison using the same total amount of excitation energy found that pulse thermography produces a better signal-to-noise ratio for shallow defects. For deeper defects, the two techniques gave roughly the same signal-to-noise performance.10NDT & E International. Matched excitation energy comparison of the pulse and lock-in thermography NDE techniques Lock-in’s advantage lies in its ability to tune the probing depth by changing the modulation frequency, giving inspectors more control over what layer of the material they are interrogating. Pulse thermography dumps all its energy at once and relies on post-processing to separate depth information from the cooling curve, which can be noisier for deep features.
A third option, frequency-modulated thermography, sweeps the excitation frequency continuously through a range during a single test run. This produces phase images at many frequencies from one experiment, making it considerably faster than running multiple single-frequency lock-in tests in sequence.11NDT & E International. A comparison of the pulsed, lock-in and frequency modulated thermography nondestructive evaluation techniques It trades some signal quality for speed and is gaining traction in production environments where throughput matters.
Physical Resolution Limits
Every imaging technique hits a resolution wall at some point, and lock-in thermography is no exception. Thermal waves behave differently from light waves: they are heavily damped by diffusion, losing amplitude rapidly as they travel. A thermal wave’s amplitude drops by a factor of more than 500 within a single wavelength of propagation. Despite this damping, the governing equation is linear, which means superposition applies and there is a resolution limit analogous to the Abbe diffraction limit in optics. In practical terms, the smallest feature you can detect at a given depth is set by the thermal diffusion length at the frequency you are using.12PubMed Central. Thermodynamic Limits of Spatial Resolution in Active Thermography
This creates an inherent trade-off. To see deeper, you lower the frequency, which increases the thermal diffusion length and therefore coarsens the resolution. To see finer detail, you raise the frequency, which limits how deep you can probe. You cannot simultaneously maximize depth and resolution; the two are fundamentally linked through the physics of heat diffusion. Knowing where those limits fall for a given material helps inspectors choose realistic test parameters rather than chasing detail the technique cannot physically deliver.
Deep Learning for Automated Inspection
Interpreting lock-in thermography images has traditionally been a skilled human task. An experienced operator looks at amplitude and phase maps, mentally accounts for heating non-uniformity and edge effects, and decides whether a contrast anomaly represents a real defect or an artifact. That process is slow and subjective, which makes it an obvious candidate for automation.
Recent work has applied deep learning models to lock-in thermography data, testing whether neural networks can segment defects from phase images more reliably than manual interpretation. A study comparing several approaches found that a U-Net architecture, which processes the spatial information in the phase image, performed best for defect segmentation.13NDT & E International. Spatial and temporal deep learning for defect detection with lock-in thermography The researchers also tested models that used the raw temporal data, the full sequence of infrared frames before phase extraction, but the spatial approach on the already-processed phase image still came out ahead.
Automated defect detection matters most in high-volume production settings, where inspecting every part by hand is impractical. If a neural network can flag suspicious areas on a phase image with high confidence, the human operator only needs to review the flagged regions, cutting inspection time substantially. The challenge, as with most machine-learning applications, is building training datasets that cover the full range of defect types and material conditions the system will encounter in practice.
Emerging Biomedical Uses
Lock-in thermography is slowly finding its way into medical research. The underlying logic transfers neatly: instead of looking for material defects, you look for variations in how tissue transports heat, which reflects local blood flow, tissue composition, and metabolic activity. A lock-in thermal imaging setup designed for dermatological applications demonstrated the ability to detect small variations in tissue thermal properties, including perfusion differences in the skin.14PubMed. A lock-in thermal imaging setup for dermatological applications
In dermatology, mapping blood perfusion non-invasively and without dye injection could be useful for evaluating burn depth, monitoring wound healing, or assessing the margins of skin lesions. The lock-in approach offers better sensitivity to small thermal contrasts than a simple infrared image because the phase extraction filters out background drift and ambient noise. Related work on controlled cooling experiments has explored dynamic infrared thermography as a tool for assessing peripheral vascular function, measuring how quickly the skin rewarms after brief contact with a cold surface.15arXiv. Quantitative Infrared Thermographic Assessment of Hand Cooling Dynamics During Controlled Contact with Metal Plates
Biomedical lock-in thermography is still largely a laboratory technique rather than a clinical routine. The equipment is the same infrared camera and modulated stimulation hardware used in industrial testing, but the stimulation source changes: instead of high-power lamps, you might use a low-power laser or a mild thermal stimulus safe for human tissue. Regulatory approval and clinical validation remain early-stage, but the non-contact, non-ionizing nature of the technique makes it appealing for applications where you want to image tissue function without exposing the patient to radiation or inserting a probe.
When Ultrasound Becomes the Heat Source
A variant worth knowing about replaces the optical heat source with modulated ultrasound. In conventional optical lock-in thermography, the lamps heat the entire surface and you watch how the heat diffuses inward and reflects off subsurface features. In ultrasonic lock-in thermography, also called vibrothermography, the ultrasonic waves travel through the bulk of the material and are selectively converted into heat at locations where mechanical losses are high, such as crack faces, delaminations, and areas of friction. This means the defects themselves generate the heat, rather than the surface.16Revue Générale de Thermique. Lock-in thermography for nondestructive evaluation of materials
The advantage is selectivity. A crack that is tightly closed and thermally invisible to optical excitation may still rub internally under ultrasonic vibration and light up on the infrared image. The disadvantage is that coupling the ultrasound into the part requires physical contact with a transducer, and the energy deposition pattern depends heavily on the part’s geometry and the ultrasonic mode shapes, which can be unpredictable. Both optical and ultrasonic variants coexist in practice, chosen based on the type of defect and the material being inspected.