This year's SEMICON Taiwan will be packed with discussions about artificial intelligence and the on-wafer challenges it's creating, where the advanced node and emerging technology worlds collide. And at the center of it all is the need for better process control at a reduced cost-of-ownership (CoO).
Right now, manufacturers are being told to buy high-priced tools designed for the front-end advanced node world - often delivering more performance than needed at a CoO that breaks the bank. It's time to break the cycle and deliver what customers want and need at a CoO that is designed to meet the needs of their business and not simply the needs of the OEM.
Enter Onto Innovation.Take control of your process and break free from costly, unnecessary platforms and patchwork solutions that prevent you from achieving truly connected process control.
Here's how Onto Innovation fights the status quo and puts manufacturers in control.
Advanced Node Logic: Controlling Structures at the Limits of Physics
At the leading edge, 3D architectures are evolving fast. Gate-all-around (GAA) nanosheets require control of individual sheet dimensions, while future complementary field-effect transistor (CFET) structures will intensify 3D metrology requirements. Both demand precise nanowire control with the data richness and enhanced signal-to-noise ratio (SNR) needed to extract more than a single dimension from a single measurement.
Powered by Ai Diffract modeling software, Onto's optical critical dimension (OCD) portfolio provides the precision to identify dimensional drift in complex GAA and CFET structures. Onto's films metrology portfolio measures the ultra-thin oxide, high-k metal gate, and P/N metal stack films that make up these devices. Together, these metrology solutions help manufacturers catch process drift before it becomes a downstream issue.
Advanced Memory: Keeping Pace With Vertical Scaling
As 3D DRAM and 3D NAND scale vertically and device density increases, metrology has to keep pace. Manufacturers face increasingly stringent requirements for high spectrum resolution and sensitivity, and Onto's OCD and films metrology portfolio delivers the tools to stay in control as memory scales.
Silicon Photonics: Controlling a Fundamentally New Kind of Device
Silicon photonics (SiPh) and co-packaged optics (CPO) are scaling toward broader high-volume manufacturing (HVM) adoption. These architectures can integrate laser sources, waveguide-based silicon chips, micro-lenses, and opto-electronic converters, with each introducing unique process-control challenges. Defects can show up anywhere along the optical path - in the V-grooves and micro-lens arrays that couple and align light, the waveguides that route it, the EEL/VCSEL devices that generate it, and the module-level packaging (through glass vias, bump, hybrid bonding, and reflectors) that holds it all together. Any of these defects can negatively impact yield and performance. Onto's metrology and inspection portfolio enables manufacturers to address these key touchpoints with confidence.
Materials Intelligence: Control That Starts Below the Surface
Not every problem appears as a visible defect. Charge trapping, interface states, dopant profiles, and crystalline defects can erode device performance and reliability long before a part reaches test – often beyond the reach of standard metrology recipes. Onto's materials intelligence tools are built for this layer of control: FAaST for charge and interface characterization, CnCV and QUAD for yield-critical electrical insights, Aspect S for material and structure characterization of high aspect ratio (HAR) trenches and TSV, and Celero PL for crystalline defect intelligence. With Onto, manufacturers are able to identify the variations dimensional measurements alone cannot see.
Advanced Packaging: Managing Complexity Where Device Types Converge
Logic, memory, and increasingly photonics come together in an advanced package to act as one system. Process steps like hybrid bonding, TSV, silicon thinning, and copper pad recess and dishing/topography control introduce errors that don't exist in front-end flows. Copper that protrudes or recesses even slightly at a bond interface, for example, can undermine an otherwise perfect hybrid bond; this is why hybrid bonding requires tight control of surface topography, including copper pad recess and dishing/topography, along with inspection designed to catch the sub-micron, low contrast, and non-visible defects standard inspection tools can miss. Interposers, meanwhile, introduce their own set of challenges one layer down, where measuring HAR deep trench capacitors and TSV is as demanding as it is in any 3D memory stack, and where high-speed infrared (HSIR) inspection is needed to catch defects below the surface.
This need for visibility is just as important, if not more so, in an AI package. Each component in an AI package- GPU or CPU, multiple HBM stacks, an interposer, a panel substrate, and CPO - is a process control touchpoint. Onto's portfolio addresses these directly: logic OCD and film metrology for the GPU/CPU, HBM bump and RDL inspection for the memory stack, interposer inspection for the layer connecting them, panel lithography and inspection for the substrate, and SiPh/CPO module inspection wherever optics are integrated - all tied together by process control and analytics software.
Dragonfly G5: One Platform for Multiple Device Types
The Dragonfly G5 system shows what taking control across device types looks like. The system delivers sub-micron defect sensitivity down to 150nm, with best-in-class throughput across front-end, back-end, and advanced packaging applications. That versatility extends beyond traditional wafer-based manufacturing. Specialty and photonics opportunities are evaluated by application, on a single platform that handles wafers and 310×310mm panels, including glass.
This unrivaled versatility is the result of several complementary inspection technologies on one platform, each one designed for a different kind of defect: brightfield and darkfield imaging for standard inspection, 3Di technology for bump metrology, a new illumination mode for sub-micron, low-contrast defects (e.g. CMP, hybrid bonding), Clearfind technology for non-visible defects, and HSIR inspection for sub-surface defects. Outfitted with TrueADC Turing machine learning classification software, the Dragonfly G5 helps manufacturers reduce nuisance defects and improve defect matching and classification accuracy.
The result is a platform that follows manufacturers across device types rather than forcing them to qualify a different tool for each type. The same underlying system supports front-end and back-end inspection, advanced packaging steps like hybrid bonding and 2.5D integration, and selected specialty/photonics inspection applications.
Taking Control, Device by Device
Advanced node logic, advanced memory, silicon photonics, advanced packaging, critical films, and AI packages feature different physics, materials, and failure modes. Onto's approach to each is the same: give manufacturers a clear, connected view of their process so problems are addressed before they become expensive.
That's what taking control of your process really means - not one tool built for one device, but a portfolio designed to give manufacturers visibility and confidence across the full range of their products.
SEMICON Taiwan attendees can find the Onto team at Booth L0728, September 2-4. Stop by and learn how comprehensive process control solutions can help you take control of your process, yield, and future roadmap.