Point a standard camera at a finished computer chip and all you see is a shiny, featureless surface. Nothing is wrong with the camera. Visible light simply cannot get past the surface it lands on. A micro crack inside the chip, a connection that didn't line up right, a tiny gap hidden under the casing. All invisible, and all still there.
A growing number of inspection lines are solving this problem with shortwave infrared light, SWIR for short. SWIR is not exotic physics. Engineers have been quietly building tools around this wavelength band for two decades. What has changed is how many industries now need to see through their materials, not just at them.
A Camera That Sees Through What Looks Solid
SWIR sits in the electromagnetic spectrum between visible light and thermal infrared, generally between 0.9 and 2.5 micrometers (figure 1). The exact boundaries shift slightly depending on which standards body or vendor you ask. What matters more than the label is the behavior. Like visible light, SWIR photons reflect off objects rather than radiate from them. That single fact is why SWIR images look photographic, with sharp edges and real shadows, instead of the soft blur produced by a thermal camera reading heat signatures.

Figure 1. SWIR's place in the spectrum.
What makes SWIR genuinely useful is which materials let it through. Silicon, opaque to visible light, becomes almost transparent to SWIR. So does ordinary glass, most plastics, and even some liquids. Water absorbs strongly around 1450 nanometers, which turns moisture almost black in a SWIR image. A spot on an apple that is invisible under supermarket lighting shows up instantly (figure 2). A fill line inside a sealed plastic bottle becomes easy to read. A crack buried inside a silicon chip stops hiding.

Figure 2. The same apples in three views. Only SWIR reveals the bruising invisible to the eye and to a standard monochrome camera. Credit: Stemmer-imaging.
How SWIR Light Is Made
Two separate pieces of hardware have to work together to make a SWIR image: a source that produces the light, and a sensor that can see it. Ordinary silicon camera sensors go blind above about one micrometer, so SWIR imaging depends on a different detector material entirely, usually indium gallium arsenide, known as InGaAs, or mercury cadmium telluride for the longer end of the band.
LEDs built from indium gallium arsenide phosphide on an indium phosphide substrate can be tuned to emit at specific SWIR wavelengths, the same family of semiconductor materials used in telecom laser diodes. Erbium doped fiber lasers, a technology borrowed directly from fiber optic communications, produce a strong, eye safe beam at 1550 nanometers. Broadband sources such as tungsten halogen lamps work too. Any hot object radiates across a wide spectrum, so a simple halogen bulb naturally puts out a healthy amount of SWIR alongside its visible light, with no special engineering required. Outdoors at night, no source is needed at all. Starlight and airglow, a faint natural emission high in the atmosphere, provide enough ambient SWIR for passive imaging without a single watt of illumination.
Why It Is Cheaper to Deploy Than It Sounds
SWIR's biggest practical advantage is economic. Ultraviolet and thermal infrared systems both need exotic lens materials, because ordinary glass blocks those wavelengths. SWIR does not have that problem. Standard optical glass transmits SWIR wavelengths well, so SWIR lenses can be manufactured with the same equipment and techniques already used for visible light optics. That keeps a SWIR inspection tool closer in cost to a normal industrial lens than to a thermal imaging system, and it means SWIR optics can share windows, filters, and enclosures with existing visible light hardware.
That manufacturing advantage does not mean any lens will do. Standard visible light camera lenses are not suitable for SWIR inspection. At SWIR wavelengths, chromatic aberration increases significantly, transmission drops, and anti-reflection coatings become ineffective. That is why SWIR inspection systems need lenses built specifically for the job, using glass types and coatings optimized for that part of the spectrum.
The other implementation detail worth knowing is heat. SWIR LEDs run hotter than their visible light counterparts, and that heat can shift the emitted wavelength or add noise to the image. The standard fix is strobing: pulsing the SWIR light only while the camera shutter is open. That lets a system run its LEDs at higher brightness for a brief instant, while letting it cool off between exposures and avoid overheating.
SWIR shows up across a surprising range of industries. Food sorting, recycling, art conservation, and defense night vision (figure 3) are all real, established, and growing applications. But the industry currently pulling SWIR fastest into the mainstream is semiconductor inspection. As chips get denser and packaging becomes more three dimensional, defects increasingly hide beneath the surface.

Figure 3. A parking lot at night: nearly unreadable in visible light, fully legible in SWIR, without any active illumination. Credit: wavelength opto-electronic.
SWIR in Semiconductor Industry
It starts with the wafer, where cutting introduces tiny cracks invisible under normal light. Left alone, these cracks tend to spread once the wafer moves through later heating and bonding steps, so catching them early in an inspection process saves a lot of wasted work downstream.
The same trick works once a chip is mounted onto its board. Tiny air pockets can form in the glue or solder holding it in place, weakening the connection and trapping heat. Because the chip is transparent to SWIR, a camera can look straight through it to check that hidden layer without touching the part. It even works on a chip once it is fully sealed inside its protective shell, letting inspectors check the internal wiring without cracking anything open (figure 4).

Figure 4. SWIR semiconductor inspection process. Credit: wavelength opto-electronic.
Power chips used in electric vehicles face the strictest version of this test, since carmakers set exact limits on how much hidden gap they will accept, and SWIR is now used to check every part on the line rather than a sample. The same principle applies to solar cells, cut from the same kind of silicon and checked at similarly high volume.
As chip manufacturing scales to meet AI demand, more chips move through more inspection stations, and each one is a candidate for a SWIR upgrade. It is telling that onsemi, one of the larger chipmakers, acquired SWIR Vision Systems®, a SWIR detector company back in 2024, rather than simply buying its cameras off the shelf.
