Photomask inspection is the verification of the master pattern, the mask or reticle, before and during its production use.
Miss a defect on a wafer and you lose a die. Miss it on the mask and you lose the lot.
At sub-7nm, a single critical escape can push yield loss past 20% – replicated across every exposure field, on every wafer, before anyone notices.
We’ll cover the full photomask inspection picture: defect taxonomy, inspection methods, EUV’s unique challenges, key tools, and what comes next. This guide is part of our semiconductor manufacturing hub, alongside wafer inspection, metrology, and reliability testing.
Key Notes
A single mask defect replicates across every die, every wafer, every lot it prints.
EUV phase defects are invisible to standard DUV and SEM inspection tools.
AI reduces false positives by 95–99% on existing hardware, no new equipment needed.
Photomask vs. Reticle: Same Thing?
In modern fabs, yes, effectively. A reticle is the photomask used in step-and-repeat or step-and-scan exposure: it carries the pattern for one exposure field, projected (typically at 4x reduction) and repeated across the wafer. “Photomask” is the broader term covering any master pattern plate, including older full-wafer masks exposed in a single shot. Since virtually every mask in production today is a reticle, the industry uses the words interchangeably, and tool vendors do too: KLA’s inspection platforms are named reticle inspection systems.
Everything in this guide applies to both terms. We use “photomask” and “reticle” the way the industry does: as the same object.
Photomask Defect Types – What Inspection Must Find
Not all photomask defects behave the same way. They differ in root cause, printability, detectability, and severity.
The industry classifies them into four main categories:
Pattern (Hard) Defects
Permanent errors in the mask pattern – directly printable in active circuitry because they change transmission or phase in the aerial image.
Types: Extra absorber, missing absorber, line breaks, bridges, pinholes, CD errors
Random: Isolated line break from a local etch glitch, a pinhole from resist processing
Systematic: Global CD error from e-beam calibration that replicates across the entire mask
Particle & Contamination (Soft) Defects
Surface particles, organic contamination, pellicle haze – conditionally printable depending on size and location.
A particle large enough relative to feature size will print
A small particle in the scribe line probably won’t
The catch: “large enough” gets smaller with every node.
Absorber & Film Defects
Coating thickness variation
Phase errors in phase-shift masks (PSMs)
Absorber delamination
These affect CD uniformity and image contrast across the field. Not always catastrophic individually, but cumulatively damaging to parametric yield.
EUV Blank & Multilayer Defects
The hardest category.
EUV masks are built on 40+ alternating Mo/Si multilayer pairs. Buried pits, bumps, or embedded particles in those layers cause phase shifts and linewidth variation – with no amplitude signature in conventional DUV or SEM inspection.
Standard tools can’t see them. They still print.
Catastrophic vs. Tolerable Defects
Mask shops and fabs classify defects by size, type, and electrical impact, not just by physical visibility:
Classification
Examples
Disposition
Catastrophic
Chrome spots, pinholes, clear breaks, bridges in active circuitry, EUV CD collapse
Zero tolerance above specified size threshold
Tolerable
Small defects in scribe lines, dummy fill, non-critical areas below printability threshold
Count-limited acceptance in spec
Manageable (EUV)
Known blank defects that can be pattern-shifted or repaired via focused e-beam
Engineer disposition with verification
The Key Principle
Defect size alone doesn’t determine risk.
Location, layer criticality, and CD impact all factor in. A 30nm defect in a forbidden region of a critical FEOL layer is catastrophic. The same defect in scribe line dummy fill may be irrelevant.
The Photomask Inspection Process
Photomask inspection is a series of verification gates throughout the mask manufacturing and qualification lifecycle.
Blank Inspection
Before patterning, the substrate and EUV multilayer are inspected for surface defects, particles, and (for EUV) phase/amplitude anomalies in the multilayer stack.
For EUV masks, actinic blank inspection at 13.5nm is the gold standard here.
Post-Write / Post-Etch Inspection
First patterned inspection pass after e-beam writing and etch. The goal is catching pattern defects introduced during the write and etch process before any repair attempt.
Repair & Re-Inspection Loop
Detected defects are repaired via focused e-beam deposition (adding material) or etch (removing it). After repair, the mask goes back through inspection to verify the repair was clean and didn’t introduce secondary damage.
AIMS Verification
Aerial Image Measurement System – simulates the actual lithographic image of specific defect sites to confirm that residual or repaired defects won’t print at wafer level.
This is a targeted verification step, not a full-field screen.
Post-Pellicle Inspection
After the protective pellicle is mounted, the mask should ideally be re-inspected.
EUV pellicles are fragile and largely opaque to DUV light, which makes this step technically awkward and a persistent gap in the workflow. Actinic inspection closes part of that gap: through-pellicle inspection at 13.5nm is a designed-in capability of Lasertec’s ACTIS systems, and multiple tools are installed for exactly this purpose, though access remains limited to leading-edge mask shops and fabs.
In-Use Monitoring
Reticles are re-inspected periodically during their production lifetime to catch haze growth, new contamination, or handling damage accumulated during scanner use.
When A Defect Is Found…
The disposition decision (scrap, repair, accept, or pattern-shift) depends on defect type, size, layer criticality, and whether the repair verifies cleanly.
For EUV blank defects, pattern shift (repositioning the layout away from a known blank defect location) is a common mitigation where repair isn’t viable.
Photomask Inspection Methods (How Defects Are Found)
Each inspection method involves a fundamental tradeoff between sensitivity, throughput, and cost. No single method covers everything.
Method
Sensitivity
Throughput
Primary Use
Optical DUV (193 nm)
~50–60 nm defect size
High
Full-field production screening
E-beam inspection
Sub-20 nm
Very low
Critical verification, not full-field
Actinic EUV (13.5 nm)
Phase defects + amplitude
Very low
EUV blank and patterned mask (limited availability)
AIMS
N/A (printability sim)
Single-site
Targeted printability verification
AFM / SEM
Nanometer-scale
Very low
High-res review and metrology
Optical Die-to-Die & Die-to-Database Inspection
The production standard.
DUV optical tools compare either:
Die-to-die: same feature across adjacent chips
Die-to-database: mask pattern against original design intent
Effective sensitivity runs down to roughly 50–60nm defect size under realistic throughput conditions, with leading tools pushing below this.
The tradeoff:
At advanced nodes, practically important defects exist at or near this resolution limit.
E-Beam Inspection
Sub-20nm sensitivity, but orders of magnitude slower than optical.
Too slow for full-field screening of every mask, every time. Used for high-sensitivity verification of specific critical layers or suspect regions flagged by optical tools.
Actinic Inspection (13.5nm)
Optical tools catch amplitude defects but miss shallow pits and phase issues that still print. Actinic patterned mask inspection is in production use (Lasertec’s ACTIS series, deployed since 2019 at leading mask shops and fabs), but systems remain few, costly, and concentrated at the leading edge; for most of the industry, access, not existence, is the constraint.
AI Defect Classification
AI now sits on top of all of these methods, and it materially changes the operational picture.
Machine learning models trained on defect image libraries:
Flag novel anomalies that rule-based algorithms miss entirely
No hardware upgrade required.
Key Photomask Inspection Tools and Equipment
KLA – Optical Patterned Mask Inspection
KLA dominates optical patterned reticle inspection, with the portfolio split by node generation:
Teron 640e: the industry production standard for leading-edge reticle inspection in mask shops, covering 1Xnm-class optical and EUV reticles via 193nm inspection with die-to-database comparison.
Teron SL6xx (SL670e XP): the in-fab side, used by chipmakers for incoming EUV reticle quality checks and periodic re-qualification during production use.
TeraScan 597XRS: the low-cost-of-ownership workhorse for mature nodes (≥32nm design rules), still current for pre- and post-pellicle pattern inspection on those masks.
Lasertec – Actinic EUV Inspection
Lasertec owns the actinic (13.5nm) side of the market, at both ends of the mask lifecycle:
ABICS series: the primary actinic blank inspection system for EUV mask blanks, finding the buried multilayer defects nothing else sees before patterning.
ACTIS A150: actinic patterned mask inspection, in high-volume-manufacturing use at leading mask shops and fabs since its 2019 introduction, including through-pellicle inspection.
ACTIS A300: the High-NA generation (announced late 2023), with the URASHIMA high-brightness source and machine-learning die-to-database reference generation for curvilinear mask inspection.
ZEISS – Aerial Image Verification & SEM Review
ZEISS supplies AIMS platforms for aerial image verification and SEM review tools for high-resolution defect analysis.
Photomask Inspection Market | Where the Investment Is Going
The market reflects where the industry is heading:
EUV mask inspection and pellicle monitoring projected to reach ~$2.22B by 2036
46% surge in EUV mask inspection tool adoption as fabs transition from DUV to EUV
At advanced nodes, sampling-based inspection is no longer viable – full-reticle inspection is becoming the standard expectation
AI Software – Extending What Existing Hardware Can Do
AI photomask inspection platforms deploy on top of existing inspection hardware – KLA, AOI, and other tools – to extend detection capability without new capital equipment:
Near-zero false positives
99%+ detection accuracy
WatchDog detection for unknown defect modes that fall outside configured rule sets
The numbers above are production results, not projections. On a photomask manufacturer’s inspection line, Averroes’ platform reached 99.993% classification accuracy across 1.4 million production images over 13 months, with defect review running 8x faster than the manual process it replaced. The full Tekscend Photomask case study [link: confirm live case-study slug] covers the deployment.
How Much Yield Is Slipping Through Your Inspection Gaps?
See what 99%+ detection catches that your current setup doesn’t
EUV Mask Inspection (The Hardest Problem in Photomask Manufacturing)
EUV mask inspection is harder than DUV mask inspection because EUV masks can carry phase defects, buried in the reflective Mo/Si multilayer, that only 13.5nm light can fully detect, and the tools that inspect at 13.5nm are scarce.
The gap is structural, not incremental.
DUV vs. EUV: A Fundamentally Different Stack
The Actinic Inspection Gap
Only 13.5nm light fully captures EUV phase defects.
The economics of that requirement still bite, but the frame has shifted from existence to access:
Actinic patterned inspection exists and runs in production: Lasertec’s ACTIS systems have been in high-volume-manufacturing use since the early 2020s, including through-pellicle inspection.
The addressable installed base is tiny: a handful of EUV mask shops and leading-edge fabs globally justify the tool economics; everyone else manages without direct actinic access.
Result: for most of the industry, the actinic gap persists as a scarcity problem – few tools, high cost, constrained throughput – rather than a technology absence.
High-NA EUV (Making A Hard Problem Harder)
High-NA EUV compounds the challenge on three fronts:
Anamorphic masks + oblique incidence change how defects manifest in the aerial image
Thicker absorbers make traditional 2D, CD-based defect criteria less reliable
Curvilinear and ILT masks create far more complex edge shapes that inflate nuisance defect counts in ways standard inspection recipes aren’t built to handle
The emerging answer on the actinic side is the ACTIS A300 generation, which pairs High-NA optics with machine-learning die-to-database reference generation to keep curvilinear edge roughness from drowning real defects.
Where the Industry Is Now
In the interim, fabs rely on a patchwork of methods to manage the residual blind spots:
Actinic blank inspection
Wafer print studies
Pattern shift strategies
AFM/SEM review
AI-based classification
It works, but it’s not a complete solution. For where mask quality requirements go from here, see our whitepaper on photomask quality for 3nm and beyond.
Photomask Inspection FAQs
What is the difference between KLA and Lasertec in photomask inspection?
KLA and Lasertec occupy different parts of the photomask inspection market. KLA dominates optical patterned reticle inspection – the Teron series (Teron 640e for leading-edge optical and EUV reticles, Teron SL6xx for in-fab qualification) is the production standard at advanced nodes, with TeraScan serving mature nodes. Lasertec leads on actinic inspection for EUV, with ABICS for blank inspection and the ACTIS series for patterned mask inspection at 13.5nm – the wavelength required to detect phase defects that DUV tools miss.
How big is the photomask inspection market?
The photomask inspection market is growing rapidly, driven by EUV adoption. The EUV mask inspection and pellicle monitoring segment alone is projected to reach approximately $2.22 billion by 2036, with a reported 46% surge in EUV-specific tool adoption already underway as leading fabs transition from DUV to EUV production at advanced nodes.
What does an AI photomask inspection platform do that traditional tools can’t?
AI photomask inspection platforms like Averroes deploy on top of existing inspection hardware (KLA, AOI, and other tools) and deliver capabilities rule-based systems can’t match: 99%+ defect detection accuracy, 95–99% reduction in false positives, and WatchDog anomaly detection that flags novel defect modes outside configured classification schemes. In production at Tekscend Photomask, that meant 99.993% classification accuracy across 1.4 million images and 8x faster defect review – without replacing or upgrading existing capital equipment.
What causes photomask defects during manufacturing?
Photomask defects during manufacturing stem from several process stages – e-beam write errors (dose variation, stitching errors), resist processing issues (scumming, collapse, over/under-development), etch defects (over-etch causing line breaks, under-etch leaving bridges), and handling damage. EUV blank defects originate earlier still.
What is the difference between a photomask and a reticle?
A reticle is the photomask used in step-and-repeat or step-and-scan lithography: it carries one exposure field’s pattern, projected at reduction (typically 4x) and repeated across the wafer. “Photomask” is the broader term, covering any master pattern plate including legacy full-wafer masks. Since virtually all production masks today are reticles, the semiconductor industry uses the two words interchangeably.
Why is EUV mask inspection so difficult?
EUV mask inspection is difficult because EUV masks are reflective, built on 40+ Mo/Si multilayer pairs, and defects buried in that multilayer create phase errors that print on wafers but produce no amplitude signature in conventional DUV or SEM inspection. Fully detecting them requires actinic inspection at EUV’s own 13.5nm wavelength, and actinic tools remain scarce and costly. High-NA EUV adds anamorphic imaging and curvilinear patterns that further complicate defect detection.
Conclusion
Photomask inspection sits at the intersection of the most demanding physics in manufacturing and the most unforgiving yield economics in the industry.
Miss a defect on a wafer, you lose a die. Miss it on the mask, and that defect systematically prints across every lot it touches – at sub-7nm, potentially taking 20%+ yield with it.
The field is under real pressure: EUV phase defects that standard optical tools can’t see, false positive burdens that slow release cycles, fragmented data pipelines that make systemic root cause analysis harder than it should be.
AI is closing several of these gaps. Not by replacing inspection hardware, but by making it significantly more capable. If you’re evaluating how AI can improve defect detection accuracy and reduce false positives on your existing inspection equipment, book a free demo with Averroes.
![Ultimate Guide To Photomask Inspection [2026]](/_next/image?url=https%3A%2F%2F303922.eu15.myftpupload.com%2Fwp-content%2Fuploads%2F2026%2F04%2Fphotomask-inspection-guide.jpg&w=3840&q=75)
Photomask inspection is the verification of the master pattern, the mask or reticle, before and during its production use.
Miss a defect on a wafer and you lose a die. Miss it on the mask and you lose the lot.
At sub-7nm, a single critical escape can push yield loss past 20% – replicated across every exposure field, on every wafer, before anyone notices.
We’ll cover the full photomask inspection picture: defect taxonomy, inspection methods, EUV’s unique challenges, key tools, and what comes next. This guide is part of our semiconductor manufacturing hub, alongside wafer inspection, metrology, and reliability testing.
Key Notes
Photomask vs. Reticle: Same Thing?
In modern fabs, yes, effectively. A reticle is the photomask used in step-and-repeat or step-and-scan exposure: it carries the pattern for one exposure field, projected (typically at 4x reduction) and repeated across the wafer. “Photomask” is the broader term covering any master pattern plate, including older full-wafer masks exposed in a single shot. Since virtually every mask in production today is a reticle, the industry uses the words interchangeably, and tool vendors do too: KLA’s inspection platforms are named reticle inspection systems.
Everything in this guide applies to both terms. We use “photomask” and “reticle” the way the industry does: as the same object.
Photomask Defect Types – What Inspection Must Find
Not all photomask defects behave the same way. They differ in root cause, printability, detectability, and severity.
The industry classifies them into four main categories:
Pattern (Hard) Defects
Permanent errors in the mask pattern – directly printable in active circuitry because they change transmission or phase in the aerial image.
Particle & Contamination (Soft) Defects
Surface particles, organic contamination, pellicle haze – conditionally printable depending on size and location.
The catch: “large enough” gets smaller with every node.
Absorber & Film Defects
These affect CD uniformity and image contrast across the field. Not always catastrophic individually, but cumulatively damaging to parametric yield.
EUV Blank & Multilayer Defects
The hardest category.
EUV masks are built on 40+ alternating Mo/Si multilayer pairs. Buried pits, bumps, or embedded particles in those layers cause phase shifts and linewidth variation – with no amplitude signature in conventional DUV or SEM inspection.
Standard tools can’t see them.
They still print.
Catastrophic vs. Tolerable Defects
Mask shops and fabs classify defects by size, type, and electrical impact, not just by physical visibility:
The Key Principle
Defect size alone doesn’t determine risk.
Location, layer criticality, and CD impact all factor in. A 30nm defect in a forbidden region of a critical FEOL layer is catastrophic. The same defect in scribe line dummy fill may be irrelevant.
The Photomask Inspection Process
Photomask inspection is a series of verification gates throughout the mask manufacturing and qualification lifecycle.
Blank Inspection
Before patterning, the substrate and EUV multilayer are inspected for surface defects, particles, and (for EUV) phase/amplitude anomalies in the multilayer stack.
For EUV masks, actinic blank inspection at 13.5nm is the gold standard here.
Post-Write / Post-Etch Inspection
First patterned inspection pass after e-beam writing and etch. The goal is catching pattern defects introduced during the write and etch process before any repair attempt.
Repair & Re-Inspection Loop
Detected defects are repaired via focused e-beam deposition (adding material) or etch (removing it). After repair, the mask goes back through inspection to verify the repair was clean and didn’t introduce secondary damage.
AIMS Verification
Aerial Image Measurement System – simulates the actual lithographic image of specific defect sites to confirm that residual or repaired defects won’t print at wafer level.
This is a targeted verification step, not a full-field screen.
Post-Pellicle Inspection
After the protective pellicle is mounted, the mask should ideally be re-inspected.
EUV pellicles are fragile and largely opaque to DUV light, which makes this step technically awkward and a persistent gap in the workflow. Actinic inspection closes part of that gap: through-pellicle inspection at 13.5nm is a designed-in capability of Lasertec’s ACTIS systems, and multiple tools are installed for exactly this purpose, though access remains limited to leading-edge mask shops and fabs.
In-Use Monitoring
Reticles are re-inspected periodically during their production lifetime to catch haze growth, new contamination, or handling damage accumulated during scanner use.
When A Defect Is Found…
The disposition decision (scrap, repair, accept, or pattern-shift) depends on defect type, size, layer criticality, and whether the repair verifies cleanly.
For EUV blank defects, pattern shift (repositioning the layout away from a known blank defect location) is a common mitigation where repair isn’t viable.
Photomask Inspection Methods (How Defects Are Found)
Each inspection method involves a fundamental tradeoff between sensitivity, throughput, and cost. No single method covers everything.
Optical Die-to-Die & Die-to-Database Inspection
The production standard.
DUV optical tools compare either:
Effective sensitivity runs down to roughly 50–60nm defect size under realistic throughput conditions, with leading tools pushing below this.
The tradeoff:
At advanced nodes, practically important defects exist at or near this resolution limit.
E-Beam Inspection
Sub-20nm sensitivity, but orders of magnitude slower than optical.
Too slow for full-field screening of every mask, every time. Used for high-sensitivity verification of specific critical layers or suspect regions flagged by optical tools.
Actinic Inspection (13.5nm)
Optical tools catch amplitude defects but miss shallow pits and phase issues that still print. Actinic patterned mask inspection is in production use (Lasertec’s ACTIS series, deployed since 2019 at leading mask shops and fabs), but systems remain few, costly, and concentrated at the leading edge; for most of the industry, access, not existence, is the constraint.
AI Defect Classification
AI now sits on top of all of these methods, and it materially changes the operational picture.
Machine learning models trained on defect image libraries:
No hardware upgrade required.
Key Photomask Inspection Tools and Equipment
KLA – Optical Patterned Mask Inspection
KLA dominates optical patterned reticle inspection, with the portfolio split by node generation:
Lasertec – Actinic EUV Inspection
Lasertec owns the actinic (13.5nm) side of the market, at both ends of the mask lifecycle:
ZEISS – Aerial Image Verification & SEM Review
ZEISS supplies AIMS platforms for aerial image verification and SEM review tools for high-resolution defect analysis.
Photomask Inspection Market | Where the Investment Is Going
The market reflects where the industry is heading:
AI Software – Extending What Existing Hardware Can Do
AI photomask inspection platforms deploy on top of existing inspection hardware – KLA, AOI, and other tools – to extend detection capability without new capital equipment:
The numbers above are production results, not projections. On a photomask manufacturer’s inspection line, Averroes’ platform reached 99.993% classification accuracy across 1.4 million production images over 13 months, with defect review running 8x faster than the manual process it replaced. The full Tekscend Photomask case study [link: confirm live case-study slug] covers the deployment.
How Much Yield Is Slipping Through Your Inspection Gaps?
See what 99%+ detection catches that your current setup doesn’t
EUV Mask Inspection (The Hardest Problem in Photomask Manufacturing)
EUV mask inspection is harder than DUV mask inspection because EUV masks can carry phase defects, buried in the reflective Mo/Si multilayer, that only 13.5nm light can fully detect, and the tools that inspect at 13.5nm are scarce.
The gap is structural, not incremental.
DUV vs. EUV: A Fundamentally Different Stack
The Actinic Inspection Gap
Only 13.5nm light fully captures EUV phase defects.
The economics of that requirement still bite, but the frame has shifted from existence to access:
High-NA EUV (Making A Hard Problem Harder)
High-NA EUV compounds the challenge on three fronts:
The emerging answer on the actinic side is the ACTIS A300 generation, which pairs High-NA optics with machine-learning die-to-database reference generation to keep curvilinear edge roughness from drowning real defects.
Where the Industry Is Now
In the interim, fabs rely on a patchwork of methods to manage the residual blind spots:
It works, but it’s not a complete solution. For where mask quality requirements go from here, see our whitepaper on photomask quality for 3nm and beyond.
Photomask Inspection FAQs
What is the difference between KLA and Lasertec in photomask inspection?
KLA and Lasertec occupy different parts of the photomask inspection market. KLA dominates optical patterned reticle inspection – the Teron series (Teron 640e for leading-edge optical and EUV reticles, Teron SL6xx for in-fab qualification) is the production standard at advanced nodes, with TeraScan serving mature nodes. Lasertec leads on actinic inspection for EUV, with ABICS for blank inspection and the ACTIS series for patterned mask inspection at 13.5nm – the wavelength required to detect phase defects that DUV tools miss.
How big is the photomask inspection market?
The photomask inspection market is growing rapidly, driven by EUV adoption. The EUV mask inspection and pellicle monitoring segment alone is projected to reach approximately $2.22 billion by 2036, with a reported 46% surge in EUV-specific tool adoption already underway as leading fabs transition from DUV to EUV production at advanced nodes.
What does an AI photomask inspection platform do that traditional tools can’t?
AI photomask inspection platforms like Averroes deploy on top of existing inspection hardware (KLA, AOI, and other tools) and deliver capabilities rule-based systems can’t match: 99%+ defect detection accuracy, 95–99% reduction in false positives, and WatchDog anomaly detection that flags novel defect modes outside configured classification schemes. In production at Tekscend Photomask, that meant 99.993% classification accuracy across 1.4 million images and 8x faster defect review – without replacing or upgrading existing capital equipment.
What causes photomask defects during manufacturing?
Photomask defects during manufacturing stem from several process stages – e-beam write errors (dose variation, stitching errors), resist processing issues (scumming, collapse, over/under-development), etch defects (over-etch causing line breaks, under-etch leaving bridges), and handling damage. EUV blank defects originate earlier still.
What is the difference between a photomask and a reticle?
A reticle is the photomask used in step-and-repeat or step-and-scan lithography: it carries one exposure field’s pattern, projected at reduction (typically 4x) and repeated across the wafer. “Photomask” is the broader term, covering any master pattern plate including legacy full-wafer masks. Since virtually all production masks today are reticles, the semiconductor industry uses the two words interchangeably.
Why is EUV mask inspection so difficult?
EUV mask inspection is difficult because EUV masks are reflective, built on 40+ Mo/Si multilayer pairs, and defects buried in that multilayer create phase errors that print on wafers but produce no amplitude signature in conventional DUV or SEM inspection. Fully detecting them requires actinic inspection at EUV’s own 13.5nm wavelength, and actinic tools remain scarce and costly. High-NA EUV adds anamorphic imaging and curvilinear patterns that further complicate defect detection.
Conclusion
Photomask inspection sits at the intersection of the most demanding physics in manufacturing and the most unforgiving yield economics in the industry.
Miss a defect on a wafer, you lose a die. Miss it on the mask, and that defect systematically prints across every lot it touches – at sub-7nm, potentially taking 20%+ yield with it.
The field is under real pressure: EUV phase defects that standard optical tools can’t see, false positive burdens that slow release cycles, fragmented data pipelines that make systemic root cause analysis harder than it should be.
AI is closing several of these gaps. Not by replacing inspection hardware, but by making it significantly more capable. If you’re evaluating how AI can improve defect detection accuracy and reduce false positives on your existing inspection equipment, book a free demo with Averroes.