SWIR Camera Lens for Silicon Wafer Inspection: Using Infrared Transmission for Subsurface Defect and Alignment Imaging

Silicon wafer inspection becomes a fundamentally different optical problem when the information of interest lies beneath the visible surface. Conventional visible-light imaging is highly effective for many surface defects, edges and patterns, but silicon strongly limits transmission through much of the visible spectrum. At wavelengths beyond approximately 1100 nm, however, crystalline silicon becomes increasingly transparent, creating an important inspection window in which short-wave infrared imaging can reveal structures, interfaces and defects that would otherwise remain hidden. This optical property is why SWIR imaging is valuable for semiconductor wafer inspection, backside alignment, bonded-wafer analysis, subsurface feature visualization and other applications where engineers need information from within or through silicon rather than only from its exposed surface.

A SWIR camera lens for silicon wafer inspection must therefore do more than produce a sharp infrared picture. It must support the wavelength region where silicon transmission becomes useful, provide sufficient spatial detail for the relevant defect or alignment feature, match the camera sensor format, and create an inspection geometry suitable for the wafer size and process equipment. The dedicated Kyptec Automation® SWIR Camera Lens collection provides 8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm focal-length options specified for 900–1700 nm imaging, 2 MP resolution, 2/3-inch sensor format, F1.4 aperture and C-Mount, giving machine builders and semiconductor-inspection integrators a focused optical platform for compatible SWIR systems.

Why Silicon Looks Opaque in Visible Light but Becomes Useful for SWIR Transmission Imaging

The usefulness of SWIR for silicon inspection originates in semiconductor physics rather than image-processing software. Silicon has a bandgap corresponding to a wavelength near 1.1 µm. At wavelengths shorter than this region, photons have enough energy to be strongly absorbed by the material. As wavelength moves beyond the absorption edge, silicon can become increasingly transmissive, enabling infrared radiation to propagate through material that appears opaque under visible illumination.

For inspection engineers, this opens an optical path to buried information. Features located beneath a silicon layer can influence transmitted or reflected SWIR intensity and become visible to an appropriately configured camera. Internal patterns of silicon wafers have been demonstrated using transmitted illumination around 1100–1200 nm, illustrating why the SWIR region is so useful for semiconductor inspection.

The precise wavelength should still be selected experimentally because silicon thickness, doping, surface finish, coatings, device structures and illumination geometry all influence transmission. A lens capable of operating across the broader 900–1700 nm SWIR range provides flexibility to optimize the inspection around the actual wafer rather than committing the machine to a single wavelength before testing.

Surface Inspection and Subsurface Inspection Require Different Optical Thinking

A surface defect can often be detected because it changes reflected intensity, texture or edge geometry. A subsurface defect must influence light that has travelled through some portion of the silicon before reaching the detector. This introduces additional factors such as absorption through the wafer, scattering, interface reflections, layer thickness and the optical path between illumination and camera.

For this reason, a visible-light inspection station should not simply be converted to SWIR by replacing the camera. Illumination direction, operating wavelength and lens selection should be reconsidered around the subsurface feature.

If the objective is to inspect an internal pattern through the wafer, transmission illumination may be preferable: illumination is positioned on the opposite side of the wafer and the camera records light that passes through it. If the geometry or process only permits illumination from the camera side, a reflected SWIR arrangement can be evaluated, but the information content may be different.

The lens sits at the end of either path and must preserve the transmitted or reflected structure with sufficient spatial contrast for the inspection algorithm.

Why Transmission Geometry Is Powerful for Internal Silicon Features

In a transmission arrangement, illumination passes through the silicon before entering the lens. Internal structures, thickness changes, defects and interfaces can modify the transmitted intensity, creating contrast at the sensor.

This can be particularly useful when examining features that are inaccessible from the viewing side of the wafer. Instead of mechanically exposing the opposite surface or relying on destructive sectioning, SWIR imaging can provide a non-contact optical view through silicon when the material and wavelength permit sufficient transmission.

Transmission imaging has been demonstrated around 1100 nm for observing internal silicon wafer patterns, supporting its practical relevance for nondestructive semiconductor inspection.

For production equipment, however, the wafer, illumination and camera need precise relative positioning. The light source must cover the required inspection area uniformly, and the SWIR camera lens must provide the appropriate field and image scale without sacrificing the smallest feature required by the alignment or defect-detection algorithm.

Subsurface Defects Do Not All Produce the Same SWIR Signature

The term “subsurface defect” covers many different physical conditions. Cracks, inclusions, voids, particles between bonded layers, damaged regions, internal interfaces and processing irregularities can influence SWIR images in different ways. Some primarily block or scatter transmitted radiation. Others change local geometry or refractive behaviour. Their contrast can also depend strongly on illumination direction.

As a result, a lens should never be selected from the assumption that every internal defect will automatically become highly visible in SWIR. Representative defect samples must be tested.

The correct development question is: at which wavelength and optical geometry does the target defect create enough contrast to be distinguished reliably from normal wafer variation?

Once that condition has been established, focal length and sensor sampling can be designed around the required defect dimensions.

Why Wavelength Selection for Silicon Inspection Should Start Above the Absorption Edge

A 900–1700 nm SWIR camera may detect a wide wavelength span, but wavelengths toward the shorter end of that range do not all penetrate silicon equally. As operation moves beyond approximately 1100 nm, silicon transmission becomes increasingly useful for observing internal structures.

This means that a silicon inspection should not simply use broadband illumination and assume every detected photon contributes equally to subsurface contrast. The wavelength should be selected around the desired penetration, material characteristics and detector response.

For a thin wafer or shallow feature, one spectral region may provide suitable contrast. For greater thickness or different device structures, a longer wavelength may perform more effectively. Testing across several SWIR bands can identify where internal features separate most clearly from the surrounding silicon.

The 900–1700 nm operating range published for Kyptec Automation® SWIR Camera Lenses gives developers room to work within this broader industrial SWIR window.

Backside Alignment: Seeing One Pattern Through Silicon to Register Another

Backside alignment is one of the most valuable consequences of silicon transparency in SWIR. Semiconductor processes may require a feature on one side of the wafer to be aligned accurately with a feature or pattern on the opposite side. In visible light, the silicon substrate can prevent direct observation of both structures simultaneously.

SWIR imaging can make the silicon sufficiently transmissive that buried or opposite-side patterns become visible, allowing the vision system to locate alignment marks and calculate their positional relationship. This principle is used in wafer bonding and other semiconductor alignment processes where features separated by silicon must be registered.

Optical resolution is particularly important here. A system may not need to detect random defects across the entire wafer; instead, it may need highly repeatable localization of a small alignment mark. That shifts the lens-selection priority toward adequate image scale, contrast and geometric consistency in the region where alignment measurements are made.

Alignment Accuracy Is Not the Same as Camera Pixel Count

A camera with more pixels does not automatically produce better wafer alignment. The accuracy with which an algorithm locates a mark depends on the number of pixels representing the feature, edge contrast, lens resolution, focus, illumination uniformity, mechanical stability and calibration.

If an alignment mark occupies only a few pixels, its calculated position can become sensitive to noise and intensity variation. If the optical image is blurred, adding additional sensor pixels may oversample the blur rather than providing meaningful new positional information.

The useful design approach is to establish the required alignment tolerance, determine how large the fiducial should appear at the sensor, and choose the optical geometry accordingly.

For a relatively controlled inspection area, the Kyptec Automation® KL-1412 25 MM SWIR Camera Lens can be evaluated where the wafer feature needs more sensor area than a wide-field configuration would provide. The model belongs to the same 900–1700 nm, 2 MP, 2/3-inch, F1.4, C-Mount SWIR family.

Choosing Between Full-Wafer Coverage and Localized Inspection

A semiconductor inspection station may need to image an entire wafer, a large process region or only a small group of alignment marks. These objectives create very different optical requirements.

Broad coverage helps with wafer presence, coarse alignment, large structural abnormalities and navigation. The trade-off is that the available sensor pixels are distributed across a large physical area. Fine defects therefore occupy fewer pixels.

Localized inspection uses a narrower field, allowing a small region to occupy more of the detector. This can improve feature localization and defect representation but may require mechanical movement or multiple camera positions to inspect a complete wafer.

A production system should define whether the priority is coverage, local detail or a combination of both before selecting focal length.

Wide-Field SWIR Imaging for Wafer Handling and Coarse Alignment

When the system must view a comparatively large wafer region, shorter focal lengths may be appropriate. The Kyptec Automation® KL-1408 8.5 MM SWIR Camera Lens provides the widest focal length currently available in the Kyptec Automation® SWIR portfolio.

Such geometry can be useful where the camera needs to observe a large process area from limited machine height. Potential tasks include locating the wafer, recognizing broad internal structures or establishing a coarse positional reference before a finer inspection stage.

Wide coverage should still be validated carefully at the edge of the sensor. If alignment or defect measurements use the entire image, edge resolution and distortion become important because positional accuracy should not depend strongly on where the feature appears.

Why Wafer Thickness Changes SWIR Transmission

A transmitted SWIR beam travels through the silicon, so increasing wafer thickness increases the optical path within the material. Even above the absorption edge, silicon is not simply an invisible sheet; wavelength-dependent absorption and scattering continue to influence the energy reaching the detector.

As wafer thickness changes, the exposure that works for one sample may produce a weaker or stronger image for another. This matters when the inspection equipment supports multiple wafer types or process stages.

Validation should therefore include the minimum and maximum expected silicon thickness. A wavelength that produces excellent internal contrast through a thin sample may not necessarily provide the same signal margin through a thicker structure.

This is another reason for selecting the operating wavelength experimentally before production settings are frozen.

Why Surface Finish and Interfaces Affect Subsurface Contrast

Polished, textured and processed silicon surfaces can interact differently with SWIR illumination. Interfaces between layers can also produce reflections. These effects may create intensity patterns that are unrelated to the defect the system is trying to identify.

A robust inspection algorithm should therefore be trained or calibrated using good wafers that represent the real range of surface and process variation.

Illumination geometry can be optimized to emphasize the feature of interest while minimizing irrelevant reflections. In transmission systems, mechanical fixtures should also avoid introducing shadows or scattering into the optical path.

The SWIR camera lens should then be positioned so the important wafer region stays within a well-corrected part of the image and receives adequate signal.

Why Focus Can Become More Challenging When Imaging Through Silicon

When an object is viewed through a material, the optical path differs from viewing the same feature directly in air. The apparent position of a buried structure can depend on material thickness and refractive index.

This matters for subsurface silicon inspection because the surface of the wafer and the internal feature may not share exactly the same best-focus condition. If the application needs both simultaneously, aperture and depth of field become important.

The correct focus should therefore be established on the actual feature the machine must inspect, not merely on the visible wafer surface.

For high-precision alignment, repeatability is more important than visual appearance. Focus should be optimized using a representative alignment mark or defect at the real operating wavelength, then mechanically locked and validated across wafer-height tolerance.

Why F1.4 Can Be Valuable When Looking Through Silicon

Transmission through silicon can reduce the amount of optical energy reaching the camera, especially when the wafer, wavelength or internal structures introduce significant attenuation. Short exposure times may also be required if wafers are moving between process steps.

The F1.4 aperture published across the Kyptec Automation® SWIR Camera Lens portfolio provides useful light-gathering capability for compatible systems. A wider available aperture can help collect more of the transmitted SWIR signal, although the final operating setting must balance illumination, depth of field and required image detail.

For semiconductor inspection, this trade-off should be tested rather than assumed. A slightly smaller aperture may provide useful focus margin if the system has sufficient illumination, while a wider aperture may be preferable where signal is the limiting factor.

Why Contrast Through Silicon Matters More Than a Bright Image

A bright internal wafer image is not necessarily a good inspection image. The real objective is separation between the structure of interest and its local background.

For defect detection, this may mean maximizing the intensity difference between a crack and surrounding silicon. For backside alignment, it means creating strong, repeatable edges around fiducial marks. For bonded-wafer inspection, it may mean emphasizing regions where interfaces or trapped particles alter transmission.

Exposure should therefore be adjusted to preserve useful contrast without saturating important areas. If bright regions clip at the sensor's maximum value, structural information is lost. If dark regions fall into noise, subtle internal detail disappears.

The best exposure normally leaves sufficient margin at both ends while maximizing the contrast important to the decision algorithm.

Using 35 mm for Tighter Subsurface Inspection Regions

When the inspection is concentrated on a smaller wafer region, a longer focal length can allocate more of the sensor to the target. The Kyptec Automation® KL-1414 35 MM SWIR Camera Lens provides one such option within the dedicated SWIR range.

A tighter field can be useful for localized bonded-wafer regions, selected alignment structures or semiconductor components where broad scene coverage would waste much of the sensor area.

The benefit comes from image scale. If a fiducial or subsurface defect occupies more pixels, its edges and intensity structure can be measured with greater robustness, provided the lens and sensor preserve the required detail.

This makes focal length a design variable for inspection precision rather than merely a framing preference.

Detecting Particles and Voids Between Bonded Silicon Layers

Wafer bonding introduces another important subsurface inspection challenge. Foreign particles, voids or imperfect bonding regions can occur between surfaces that are inaccessible to direct visible inspection after assembly.

Where the materials and layer structure allow adequate SWIR transmission, internal discontinuities can change the transmitted or scattered signal. SWIR inspection is therefore relevant to identifying particles between bonded wafers and supporting alignment during wafer-packaging processes.

Detection performance depends strongly on defect size and contrast. A small particle requires sufficient spatial sampling, while a shallow intensity difference requires low noise and uniform illumination.

Representative known-good and known-defective bonded samples should therefore be used when establishing the production threshold.

Why Distortion Matters in SWIR Wafer Alignment

For simple visual inspection, modest geometric distortion may sometimes be acceptable. For semiconductor alignment or measurement, it can create systematic positional errors.

If straight coordinates in object space map nonlinearly onto the sensor, the measured distance between wafer features can change according to image position. Software calibration can compensate for stable distortion, but the optical and mechanical system should first provide consistent geometry.

Calibration should cover the entire region in which measurements will be made rather than only the center.

When selecting a SWIR lens for wafer alignment, buyers should therefore consider not only whether the marks are visible, but whether their sensor coordinates remain sufficiently predictable for the required positioning tolerance.

Using 50 mm for Small Alignment Marks and Narrow Inspection Windows

Some semiconductor machines provide generous camera stand-off but require a tightly framed region around a small target. In such layouts, the Kyptec Automation® KL-1416 50 MM SWIR Camera Lens provides the longest focal length in the current Kyptec Automation® SWIR portfolio.

A 50 mm geometry can be considered when the inspection requires a narrow field, stronger target representation or greater stand-off than shorter focal lengths provide.

It should not be treated as inherently more accurate merely because the focal length is longer. Its value appears when the machine geometry allows the relevant subsurface pattern or alignment feature to occupy an appropriate portion of the 2/3-inch sensor.

Production Calibration Should Use Real Semiconductor Structures

Resolution charts and flat reference targets are useful during optical development, but final qualification should use actual wafer structures.

For backside alignment, measure repeated localization of real fiducials through the expected silicon thickness. For subsurface defect inspection, test known cracks, inclusions, voids or interface abnormalities representative of the production limit. For bonded structures, include acceptable process variation as well as known defects.

Repeat the test across the inspection field, through expected focus variation and at the final operating wavelength.

A good production validation should answer two separate questions: Can the system detect the defect? and Can it avoid classifying legitimate process variation as a defect?

Both are required for a commercially useful inspection station.

Why Mechanical Stability Is Critical in Subsurface Alignment

Once a system measures features through silicon, optical performance alone cannot guarantee repeatable alignment. Camera mounting, lens focus, wafer positioning and illumination geometry must remain stable.

Small mechanical changes can shift feature coordinates or alter focus. Vibration may also blur fine patterns during exposure.

OEM designers should therefore use rigid mounting, repeatable wafer fixtures and mechanically secure lens settings. Calibration checks should be incorporated into maintenance so changes can be identified before they affect process yield.

The optical design is successful only when its performance survives the mechanical environment of the semiconductor tool.

Why Kyptec Automation® SWIR Camera Lenses Fit Silicon Inspection Development

Silicon inspection machines can require very different image geometries even when they operate within a similar SWIR spectral region. One station may need a broad view for wafer handling or coarse alignment, while another may require tightly framed subsurface patterns for precise registration or defect analysis.

The Kyptec Automation® SWIR Camera Lens collection addresses this range through five focal lengths—8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm—within a consistent published 900–1700 nm, 2 MP, 2/3-inch, F1.4 and C-Mount platform. The live product information also specifically identifies semiconductor inspection among the intended SWIR application areas.

For OEM machine builders and system integrators, this gives useful flexibility: wavelength requirements can remain centered on silicon transmission while focal length changes according to wafer coverage, target size and camera stand-off. That is a more practical approach than attempting to use one optical geometry for every semiconductor inspection stage.

Frequently Asked Questions About SWIR Camera Lenses for Silicon Wafer Inspection

1. Why can SWIR cameras see through silicon?

Silicon strongly absorbs shorter-wavelength light but becomes considerably more transmissive after its absorption edge near approximately 1.1 µm. This allows SWIR radiation at suitable wavelengths to pass through silicon and carry information about internal or opposite-side structures to the camera. The exact transmission depends on wavelength, thickness, doping and wafer construction, so real samples should always be tested before the production system is specified.

2. What wavelength is best for seeing through a silicon wafer?

There is no single best wavelength for every wafer. Useful transmission begins beyond roughly 1100 nm, and practical systems may operate farther into the SWIR depending on wafer thickness, detector response, coatings and required internal contrast. Around 1100–1200 nm has been demonstrated for internal silicon wafer pattern imaging, but a production system should compare candidate wavelengths using the actual wafer stack.

3. Can SWIR detect defects inside a silicon wafer?

SWIR can reveal certain internal defects or structural variations when they modify transmitted, reflected or scattered infrared radiation sufficiently to create measurable contrast. Detectability depends on the size and nature of the defect, silicon thickness, operating wavelength, illumination geometry, optical resolution and sensor sampling. The required defect class should therefore be demonstrated on representative samples rather than assuming every subsurface flaw will automatically be visible.

4. What is backside wafer alignment using SWIR?

Backside alignment uses the ability of suitable infrared wavelengths to pass through silicon so a vision system can observe alignment features that would otherwise be hidden by the substrate. Patterns on different sides or layers can then be located relative to one another. The process requires stable optical magnification, strong mark contrast and accurate calibration because visibility alone does not guarantee sufficient positional accuracy.

5. Can SWIR be used to inspect bonded silicon wafers?

Yes, where the wafer stack and materials provide sufficient SWIR transmission. Internal interfaces, particles or bonding abnormalities may affect transmitted or scattered intensity and become detectable. SWIR imaging is also relevant to alignment in bonded-wafer processes. Detection limits depend heavily on defect size, wafer thickness, wavelength and optical geometry, so known bonded-wafer samples should be included in qualification testing.

6. Should silicon wafer inspection use reflected or transmitted SWIR illumination?

Transmission illumination is particularly useful when the objective is to observe structures through the wafer because the light deliberately passes through silicon before reaching the camera. Reflection geometry can still be useful for some surface and subsurface conditions when rear-side illumination is impractical. The best arrangement depends on which physical structure must generate contrast and how the semiconductor tool permits illumination to be positioned.

7. Why does silicon wafer thickness affect SWIR image brightness?

A thicker wafer creates a longer optical path through silicon, allowing greater total absorption and potentially more scattering. The same camera exposure may therefore produce different intensity for wafers of different thickness. If one inspection machine supports several wafer types, thickness should be included in wavelength selection, exposure development and validation rather than compensated blindly with gain.

8. Can visible-light machine vision lenses be used for through-silicon inspection?

Mechanical compatibility alone is not enough. The lens should support the SWIR wavelength required to transmit through silicon and should maintain usable imaging performance there. A conventional lens intended mainly for visible wavelengths should not automatically be assumed suitable far beyond the visible range. Kyptec Automation® provides a dedicated SWIR Camera Lens portfolio specified for 900–1700 nm imaging and compatible 2/3-inch C-Mount systems.

9. How do I choose the focal length for silicon wafer inspection?

Define whether the machine needs full-wafer coverage, a large process region or a small alignment/defect zone. Then establish camera stand-off and the minimum feature size. Shorter focal lengths generally enable broader fields, while longer focal lengths can dedicate more sensor area to smaller regions. The Kyptec Automation® SWIR Camera Lens family spans 8.5 mm through 50 mm, allowing the geometry to be selected according to the actual semiconductor station.

10. Why are my subsurface wafer features visible but blurry?

Possible causes include incorrect focus depth, insufficient lens or sensor resolution, wavelength-dependent focus, motion, excessive field coverage or low contrast caused by weak transmitted signal. Focus specifically on the buried feature rather than merely on the wafer surface. Then verify that the feature occupies enough pixels and that exposure provides sufficient contrast without saturation.

11. Can SWIR identify microcracks in silicon?

Certain cracks can alter infrared transmission or scattering and therefore become visible under appropriate SWIR conditions. The practical detection limit depends on crack width, depth, orientation, illumination direction, wavelength and system resolution. Fine cracks should be tested under several illumination geometries because a crack that produces weak contrast in one direction may become more apparent under another.

12. What sensor format should be matched with Kyptec Automation® SWIR lenses for wafer inspection?

The current Kyptec Automation® SWIR Camera Lens portfolio is published for 2/3-inch (φ12) sensor format with C-Mount. Buyers should confirm that their intended SWIR camera's active sensor area is compatible with this lens format and then validate center-to-edge image quality using the actual wafer inspection field.

13. Is 900 nm useful for seeing through silicon if the lens supports 900–1700 nm?

Lens wavelength coverage and silicon transmission are different questions. A lens can support 900 nm optically, but silicon still absorbs strongly below its approximately 1.1 µm band-edge region. For through-silicon imaging, wavelengths above that transition are generally more relevant. The broader 900–1700 nm capability remains useful because the same SWIR lens category can support other imaging conditions and allows the silicon inspection wavelength to be optimized within the usable region.

14. How can I improve contrast of alignment marks viewed through a silicon wafer?

Start by optimizing wavelength and illumination direction, because the mark must first create sufficient physical contrast. Then adjust exposure to avoid clipping, ensure the mark occupies enough sensor pixels, focus at the mark plane rather than only on the wafer surface, and calibrate the imaging geometry. A tighter field using an appropriate Kyptec Automation® SWIR Camera Lens focal length can also allocate more of the sensor to the alignment feature when the machine geometry allows it.

15. What should I validate before buying a SWIR camera lens for semiconductor wafer inspection?

Define the wafer thickness and structure, target wavelength, camera sensor format, C-Mount compatibility, inspection field, working distance, smallest defect or alignment mark, required positional tolerance and illumination arrangement. Then test real wafers containing representative acceptable variation and known defects. Evaluate transmission, contrast, focus, center-to-edge performance and repeatability before standardizing the lens across production equipment. The Kyptec Automation® SWIR Camera Lens portfolio provides five focal lengths within one dedicated SWIR platform, allowing this qualification to be matched to different semiconductor inspection geometries.

Conclusion

The ability to inspect through silicon is one of the most technically valuable uses of SWIR imaging in semiconductor manufacturing. Once wavelength moves beyond silicon's strong shorter-wavelength absorption region near approximately 1.1 µm, transmitted infrared radiation can carry information about internal patterns, backside features, bonded interfaces and certain subsurface defects that remain inaccessible to ordinary visible inspection. This makes SWIR useful not simply as another imaging spectrum, but as a practical nondestructive route to information located within or behind silicon.

Reliable inspection still requires much more than choosing a wavelength above 1100 nm. Silicon thickness, device structure, illumination geometry, defect type, transmitted signal, sensor sampling, focal length, focus depth and mechanical stability all determine whether a buried feature becomes sufficiently clear for an automated decision. Full-wafer visualization and micron-scale alignment also represent fundamentally different optical objectives, which is why focal length and image scale must be selected from the inspection requirement rather than standardized blindly.

The Kyptec Automation® SWIR Camera Lens collection gives semiconductor OEMs and machine-vision integrators a focused family of 8.5 mm, 12.5 mm, 25 mm, 35 mm and 50 mm lenses specified across 900–1700 nm with 2 MP resolution, 2/3-inch format, F1.4 aperture and C-Mount. This combination allows developers to move from wider wafer views to tighter subsurface or alignment inspection fields while remaining within a dedicated SWIR optical family. By pairing the appropriate Kyptec Automation® SWIR Camera Lens with a validated silicon-transmission wavelength and production-representative wafer samples, an inspection system can be engineered around the internal information the semiconductor process actually needs rather than only what is visible at the surface.