ASML eScan 2200 Multibeam Inspection Tool Revives Mapper Tech
During a presentation by ASML's CEO, Christophe Fouquet, at ITF World 2026 in Antwerp, the eScan 2200 multibeam inspection tool was highlighted. This tool utilizes multibeam technology, reminiscent of the defunct Dutch lithography startup, Mapper Lithography, which had attempted to challenge ASML and EUV technology. Mapper's technology, once considered a failure, is now being reborn with TSMC's backing in massively parallel e-beam lithography.
Beginnings of Electron Beam Direct Write Lithography
Lithography, the process of patterning a wafer, typically uses light, ions, or electrons. While photolithography is common, electron beams offer a compelling alternative. Electron Beam Direct Write (EBDW) lithography uses a highly focused electron beam to directly print chip designs onto a resist-coated wafer.
The first EBDW tools, developed in the 1960s, were based on electron microscopes. They featured an electron gun that generated a single beam, which was then accelerated and focused using electron optics. This beam scanned across the wafer's surface, similar to how a CRT display draws an image.
EBDW's primary advantage is its high resolution. Systems in the 1990s achieved 20-nanometer line widths, and current technology can reach resolutions below 10 nanometers, exceeding the requirements for leading-edge process nodes. Unlike photolithography, which can suffer from "hot spots" where certain mask patterns fail to print, electron beams can draw arbitrary designs, offering greater design flexibility.
Another significant benefit of EBDW is the elimination of photomasks, earning it the moniker "mask-less" lithography. Photomasks for leading-edge processes can cost tens of millions of dollars, representing a substantial cost barrier for chipmakers. Maskless lithography bypasses this expense.
EBDW's Achilles Heel: Throughput
Despite its advantages, EBDW's major drawback is its slow throughput. The single electron beam writes the chip pattern by slowly scanning across the wafer, a process that can take hours or even days. As Moore's Law progresses and designs become more complex, this throughput issue worsens.
Consequently, electron beam lithography is primarily used for producing photomasks. It also finds application in creating prototype chips for clients or small-batch runs in academic labs, where the cost of an expensive mask is not economically justifiable. There are also "maskless" lithography machines, like the MicroWriter, that do not use electron beams but offer similar benefits.
The SCALPEL Project
The technology that would eventually become Mapper originated in the Netherlands. In the late 1990s, Pieter Kruit, a physics professor at Delft University of Technology specializing in electron optics, conceived the idea. Kruit was involved in electron beam projection lithography, specifically "SCALPEL" (SCattering with Angular Limitation Projection Electron Beam Lithography).
SCALPEL, invented in 1989 by AT&T Bell Labs, aimed to mimic photolithography using electron beams. It projected an electron beam onto a scattering mask, which contained the chip design at four times its actual size. The mask was made of a material largely transparent to electrons, with "ink" that scattered electrons at a sharp angle. Electrons either passed through or were scattered. After passing through a magnetic lens and an aperture, the strongly scattered electrons were filtered out, producing a high-contrast image of the chip pattern on the wafer.
Bell Labs worked on SCALPEL for over a decade but failed to make it viable. Kruit, who evaluated a similar system for ASML, identified a fundamental flaw: the aperture forced all electrons through a "crossover" point. Due to electron repulsion, energized electrons at this chokepoint stochastically interfered, causing beam blur that ruined resolution. This blur could not be corrected with ordinary lenses, rendering SCALPEL unusable. Variants of electron projection lithography pursued by Applied Materials and Nikon also proved unsuccessful.
Founding Mapper Lithography
Kruit's experience with SCALPEL led him to develop a new approach to electron beam lithography that avoided its shortcomings. At the time, other e-beam proposals included systems that projected a 1:1 photomask image (which required difficult-to-produce 1:1 masks) and an IBM system that used an array of minicolumns for parallel writing, but faced challenges in producing and controlling a uniform and stable beam array.
Kruit's core concept involved generating many small, parallel electron beamlets. His initial idea proposed millions of beamlets, each representing a pixel in the chip design. Since each beamlet would be low power, they would not interfere with each other. Kruit's graduate students, Bert Kampherbeek and Marco Wieland, developed this concept, naming it "Multiple Aperture Pixel by Pixel Enhancement of Resolution" (MAPPER).
In 2000, with sufficient progress, Mapper Lithography was founded, securing intellectual property licenses from the university and initial funding. Arthur Del Prado, founder of ASM International and a significant figure in the Dutch semiconductor industry, was an early investor.
Building the Mapper Tool
The initial design for Mapper envisioned each electron beamlet having its own individual source. This involved illuminating a standard transparent chromium-quartz photomask with an array of millions of small light spots. An optical lens would then shrink this light pattern before it hit a "converter plate," which transformed the light array into an array of small electron beams directed at the wafer. Each electron "spot" was about 100 nanometers wide, with the entire array measuring approximately 5mm by 20mm, yielding about 20 million beamlets.
However, this setup failed. While the converter plate generated stable electrons, some left the source at incorrect angles or energies. The initial understanding that the converter plate acted as a photocathode, where light directly knocked out electrons, was incorrect. Instead, it functioned as a photodiode feeding a field emitter, where light refilled a pool of excited electrons pulled to the tip by an electric field. This meant that the energy and position of the exiting electrons could not be directly controlled, leading to blur and insufficient switching speed for lithography.
Adding an aperture to filter aberrant electrons resulted in unstable numbers of electrons passing through, making the setup unsuitable. After 1-2 years, Mapper shifted to a different approach: splitting a single electron beam into many parallel beamlets, similar to water flowing through a showerhead. Each beamlet could be individually switched on and off as the entire array scanned a 26mm by 10mm field across the wafer, akin to a CRT gun scanning a TV screen.
Mapper and IMS
This new setup bore resemblance to the IMS Multi-beam mask-writer, made by IMS Nanofabrication in Vienna, which TSMC uses for mask writing. Both start with a single beam and split it into multiple beamlets. However, Mapper's design differed from IMS's.
The IMS mask-writer's demagnifying optics require a crossover point, a chokepoint for electrons, which can lead to column interaction problems with high current beams. The mask-writer compensates by using a relatively low current, sacrificing writing speed, which is acceptable for masks but not for fab lithography. Mapper's system eliminated this crossover, allowing for higher currents but at the cost of increased control complexity.
Building a Prototype: Challenges
Building the Mapper tool presented significant challenges, particularly with the beam blanker array. This custom microchip featured thousands of 1-2 micrometer wide holes, drilled using dry etch. To switch off a beamlet, a voltage was applied around the corresponding hole to deflect the beamlet towards a stopper. This advanced microchip was fabricated by TSMC, with Mapper etching the holes themselves, a process that was both difficult and expensive, costing millions for each iteration.
Other challenges included: * Mechatronics: Traditional linear motors could not be used for wafer positioning due to magnetic fields interfering with e-beams. However, Mapper's continuous scanning, without the violent stops and starts of typical litho tools, meant accelerations were 100 times less than an ASML machine. * Heat: The electron beams constantly striking the wafer generated heat, necessitating active wafer cooling. * Data Velocity: Streaming chip design data to the beam blanker array required rapid data transfer. Copper wires were insufficient, so Mapper adopted glass fiber, which was innovative at the time.
These technical hurdles delayed the product's market entry. The first proof of lithography was achieved in Q4 2005.
Next Generation Lithography and TSMC's Involvement
During this period, there was considerable uncertainty about the next generation lithography technique. EUV showed potential but was one of several candidates. "Next Generation Lithography" conferences highlighted the industry's indecision, with some even advocating for indefinite multi-patterning (like pitch-splitting) rather than a new technology.
TSMC, unwilling to solely bet on EUV, explored alternatives and connected with Mapper. In 2008, TSMC invested in Mapper, seeking a viable EUV alternative. TSMC's R&D staff purchased Mapper's tools, providing feedback and support, particularly interested in how maskless lithography could reduce escalating mask costs. Dr. Burn Lin, TSMC's lithography expert, praised Mapper's achievement of 32-nanometer class patterns in late 2007, validating its low-voltage multibeam resolution. Mapper also placed its tool at CEA-Leti, France's leading semiconductor research facility, as part of the IMAGINE maskless litho research program, to refine the machine for production.
Mapper and TSMC planned to overcome slow throughput through clustering. The target speed was 10 wafers per hour, a significant gap from the 100+ wafers per hour needed to compete with EUV or double-patterning. However, a cluster of ten 10-wafer-per-hour machines could achieve 100 wafers per hour.
EUV's Consolidation and Mapper's Struggles
Mapper's development was hampered by a lack of resources and funding, a common issue in the European deep-tech community for ambitious projects. By 2009, Mapper had an alpha 300-millimeter wafer tool with 110 beams and lower voltage, achieving the correct resolutions at CEA-LETI but with a throughput of only 1 wafer per hour, far short of the 10 needed for clustering.
In 2012, Mapper secured an 80 million euro funding round led by RUSNANO, a state-owned Russian investment fund focused on nanotechnology. This partnership involved Mapper working with Russian research institutes and opening a MEMS Fab in Moscow to produce lens elements.
However, around this time, ASML's EUV technology made significant advancements, particularly with the introduction of a pre-pulse shot to reshape tin droplets, rapidly increasing power levels and EUV throughput. By 2013, Mapper shipped its first prototype pre-production tool, the FLX-1200, to CEA-Leti. This tool had 1,300 beamlets and could pattern at a 32-nanometer half-pitch, suitable for 14-90 nanometer nodes. Its raw throughput was still 1 wafer per hour, with the goal of scaling to 13,000 beamlets for 10 wafers per hour. By then, ASML's EUV light source reached 80-100 watts, processing approximately 60 wafers per hour.
Despite the throughput disparity, Mapper emphasized metrics like wafers per hour per million euro or per square meter of cleanroom footprint, where the smaller FLX-1200 offered advantages. Servicing the FLX-1200 was also easier, requiring access to only one side of the tool, a key feature for clustering that was not possible with EUV.
Bankruptcy and ASML's Acquisition
Even as EUV gained dominance, Mapper's team believed they could bring an economically competitive mainstream lithography machine to market. However, financial challenges intervened. In September 2016, Arthur Del Prado passed away. His estate informed Mapper that funding would cease unless a new lead investor was found.
Finding new funding was difficult. The original pitch of an EUV alternative was hard to sell given EUV's progress and Mapper's substantial funding needs. Other targets, such as tools for prototyping, niche uses, or academia, required over $100 million to commercialize. They also explored personalized chips for security, where maskless lithography could embed unique hardware IP signatures.
Ultimately, Mapper ran out of time and filed for bankruptcy in December 2018. Several firms sought to acquire Mapper's IP, but ASML intervened to keep the technology in the Netherlands. ASML's CEO expressed interest in a deal, and a few months later, ASML acquired Mapper's IP and assets for 75 million euro. While jobs were offered to Mapper's 240-270 employees, the original work on lithography tools was not continued.
Rebirth as Inspection Technology
ASML saw a different application for Mapper's multibeam technology: inspection. In 2016, ASML had acquired Hermes Microvision (HMI), a Taiwanese company specializing in electron beam inspection, for $3.1 billion.
Inspection tools detect defects by comparing wafer patterns to a "golden image" or other parts of the wafer. With EUV, defects became smaller and harder for optical inspection tools to find. Electron beams, with their shorter wavelengths, could detect defects as small as 1-10 nanometers.
The rise of 2.5D and 3D structures also presented challenges for optical inspection, which could not detect defects in the contacts of stacked layers like 3D NAND. E-beam inspection, however, could perform Voltage Contrast inspection, firing primary electrons at the wafer surface and measuring secondary electrons to assess the quality of electrical contacts.
The main problem for e-beam inspection was throughput, as semiconductor lines move very fast. Combining HMI's electron beam inspection with Mapper's multibeam technology offered a potential solution for an inspection tool fast enough for in-line use.
In 2022, ASML launched its first multibeam inspection tool, the eScan 1000, with a 3x3 grid of 9 beams, followed by the eScan 1100 with a 5x5 grid of 25 beams. These tools performed well for customers. The eScan 2200 aims to significantly increase the beam array and, consequently, throughput, with more machines expected in the future.
Conclusion
The eScan tool is part of ASML's "holistic lithography" strategy, which views lithography as more than just the scanner. It integrates other tools, such as computational lithography and optical corrections before exposure, and metrology/inspection tools to find and correct defects after exposure. These components work together to feed data and results back into software for analysis and continuous improvement.
Mapper Lithography's technology, though unsuccessful in competing with ASML's lithography, has found new life within this broader strategy. This demonstrates that in the semiconductor industry, a perceived failure in one era can become a success in another.
Takeaways
- The eScan 2200 uses Mapper’s multibeam electron‑beam technology to boost inspection throughput, turning a former lithography failure into a high‑speed defect‑detection solution.
- Electron Beam Direct Write offers sub‑10 nm resolution and mask‑less flexibility, but its single‑beam architecture suffers from extremely low throughput, limiting its use to mask making and prototyping.
- Mapper’s core innovation was to split a single electron beam into millions of low‑power beamlets, avoiding the crossover‑point blur that doomed earlier projection systems like SCALPEL.
- Despite achieving 32 nm half‑pitch patterns, Mapper could only reach about one wafer per hour, far below the 10‑plus wafers per hour needed to compete with EUV, leading to financial collapse and bankruptcy in 2018.
- After ASML acquired Mapper’s IP, the multibeam approach was repurposed for e‑beam inspection, enabling detection of 1‑10 nm defects in advanced 2.5D/3D structures where optical tools struggle.
Frequently Asked Questions
How does Mapper’s multibeam approach overcome the throughput limitation of traditional electron beam lithography?
The multibeam approach replaces a single scanning electron beam with thousands of parallel beamlets, each writing a small portion of the pattern simultaneously. By eliminating the crossover choke point and allowing higher total current, the system multiplies writing speed, turning the inherently slow EBDW process into a viable inspection tool with acceptable throughput.
Why did ASML repurpose Mapper’s multibeam technology for e‑beam inspection instead of continuing lithography development?
ASML saw that the multibeam system could solve the critical throughput bottleneck in e‑beam inspection, where detecting nanometer‑scale defects in EUV‑produced wafers is essential, while lithography required far higher wafer‑per‑hour rates that Mapper could not meet. By integrating the technology into its holistic lithography strategy, ASML added a fast, high‑resolution inspection capability that complements its EUV scanners.
Who is Asianometry on YouTube?
Asianometry is a YouTube channel that publishes videos on a range of topics. Browse more summaries from this channel below.
Does this page include the full transcript of the video?
Yes, the full transcript for this video is available on this page. Click 'Show transcript' in the sidebar to read it.
Helpful resources related to this video
If you want to practice or explore the concepts discussed in the video, these commonly used tools may help.
Links may be affiliate links. We only include resources that are genuinely relevant to the topic.