Why Can Only a Handful of Companies Make Advanced Photoresist? A Deep Dive from Photochemistry to Semiconductor Supply Chains

An article that actually explains photoresist, rather than merely describing it.

In one sentence: The scanner's job is to project light onto a wafer, but the material that turns that light into a permanent nanoscale pattern is photoresist — in effect, a translator that converts optical information into chemical information, then into a physical pattern, and finally into an electronic device. Over the past four-plus decades, almost every reduction in exposure wavelength has forced this translator to rewrite itself from the ground up. Today only a handful of companies can reliably supply advanced photoresist, and the reason is not mainly formula secrecy — it is the manufacturing consistency and fab-qualification data accumulated over decades, on top of the chemistry.

1. A Trade Dispute That Exposed a Deeper Question

In July 2019, Japan's Ministry of Economy, Trade and Industry (METI) announced tighter export-approval procedures for three materials critical to semiconductor manufacturing bound for South Korea: fluorinated polyimide, photoresist (specifically for advanced processes such as EUV), and high-purity hydrogen fluoride — shifting from bulk export licenses to case-by-case review, with Japanese exporters required to report end-user, product specification, and technical information for every contract. On August 28 that year, Japan went further and removed South Korea from its "white list" of trusted trading partners (a status only 27 countries held at the time), affecting 1,147 controlled items. Japan's official justification was "concerns over export control compliance"; South Korean officials contested that basis. It's worth being precise here: these measures were not an outright export ban — they were a tightening of administrative procedure, and a signal that further restriction was possible.

The episode quickly became a global semiconductor-industry flashpoint — Samsung Electronics and SK hynix were heavily dependent on Japanese supply, and markets briefly worried that advanced chip manufacturing could be disrupted. Most media coverage framed it as a trade dispute. To a materials scientist, though, it exposed a different question:

Why would a chemical that looks like an ordinary liquid become one of the hardest links to replace in the entire global semiconductor supply chain? And more pointedly — why, today, does the ability to reliably supply advanced photoresist still sit with only a handful of companies?

The answer is not simply commercial competition, and not simply patent protection. Understanding it requires starting from something more basic: what happens to that light after exposure ends.

2. The Moment Exposure Ends, the Pattern Should Vanish — Unless Something Translates It

In lithography, the scanner's job is to project a circuit pattern as light onto a wafer coated with photoresist with great precision. But light itself leaves no trace — the instant the shutter closes, the information of "where was illuminated, where was not" should disappear with it. Unless a material exists that can, in that same instant, translate this purely optical information into a difference that subsequent chemical steps can recognize, and that ultimately fixes into a real, physical topography.

That is photoresist's job: a conversion device that translates optical information into chemical information, and chemical information into a physical pattern. Completing that translation requires four classes of components present and working in concert — the resin, which determines which exposed and unexposed regions will dissolve in the developer, and which serves as the physical carrier of the final pattern; the photoacid generator (PAG), which determines what chemical reaction occurs wherever light lands; the protecting-group chemistry built into the resin, which determines exactly where that reaction happens and how it changes the material's solubility; and the solvent and additives, which determine whether the material can be spin-coated evenly into a film tens to hundreds of nanometers thick. The full translation chain runs: exposure → PAG generates acid → post-exposure bake (PEB) → protecting-group cleavage and solubility change → development → pattern formation. Miss any single link, and optical information never completes its translation into a physical pattern.

And the grammar of this translator is locked in from the start by the exposure wavelength itself.

3. Every Reduction in Wavelength Forces This Translator to Reinvent Itself

The relentless shrinking of semiconductor process nodes fundamentally demands shorter exposure wavelengths to achieve higher resolution — from g-line (436 nm) and i-line (365 nm), to KrF (248 nm), to ArF (193 nm, and later ArF immersion), to today's EUV (13.5 nm). On the surface this reads as a history of ever-advancing light sources and optical systems. What a materials scientist sees is a second, equally important — arguably harder — throughline: almost every reduction in wavelength gives birth to an entirely new materials problem. In this industry, every step forward the light source takes forces the material to start over.

The reason lies in absorption spectra. g-line and i-line resists are generally built on diazonaphthoquinone (DNQ)/novolac resin chemistry — a system with moderate absorption across the 365–436 nm band that worked reliably for decades. But compress the wavelength to 193 nm (ArF), and the problem changes: aromatic phenolic polymers such as novolac exhibit extremely strong absorption at 193 nm, an absorption believed to trace directly to the aromatic ring itself. For this translator, that isn't merely a metaphor about "not understanding a new language" — it is literally true: the light cannot penetrate to the bottom of the resist film, so the pattern cannot form accurately through the full film thickness. ArF resists therefore had to abandon aromatic-ring resin backbones entirely, moving to acrylate and alicyclic (adamantane-, norbornene-based) polymer systems containing no aromatic rings — not a local tweak to a formula, but a complete redesign of the resin chemistry.

The moment this translator first underwent a true step change came in the early 1980s. Deep-ultraviolet (DUV) lithography was entering R&D view at the time, and conventional DNQ/novolac chemistry had a quantum yield of roughly "one reaction per photon" — reaching usable exposure sensitivity at shorter wavelengths like 248 nm would have required energy densities high enough to threaten scanner throughput. Chemical amplification, then, was not a lucky accident — it was the inevitable consequence of the process continuing to shrink. In 1982, Hiroshi Ito, C. Grant Willson, and Jean Fréchet at IBM's San Jose research center proposed the concept of "chemical amplification": a poly(p-hydroxystyrene) (PHOST) resin modified with tert-butoxycarbonyl (tBOC) protecting groups, paired with an onium-salt photoacid generator. On exposure, a photon absorbed by the PAG generates one acid molecule; during post-exposure bake, that acid catalytically cleaves the tBOC protecting group while regenerating itself, triggering a cascading chain reaction. Willson's own target at the time was a 30-fold sensitivity improvement; the tBOC system ultimately delivered 100- to 200-fold.

It didn't change the light. It changed how the material makes use of the light — lithographic chemistry, for the first time, moved from a one-to-one photochemical stoichiometric reaction to an amplification mechanism driven by a catalytic cycle. Nearly every advanced photoresist in use today, whether for KrF, ArF, or EUV, traces back to this one year, in this one lab.

This is the first layer of the answer to "why only a handful of companies are left": the number of teams able to master this catalytic-amplification chemistry, and safely carry it across every subsequent wavelength transition, was never large to begin with.

4. Why EUV Suddenly Made This Translator's Job Nearly Impossible Again

EUV lithography uses 13.5 nm extreme-ultraviolet light. By the logic of "shorter wavelength, higher resolution," this should have been a natural next step — but the real bottleneck EUV photoresist runs into isn't chemistry falling behind. It's a more fundamental physical constraint: the higher a photon's energy, the fewer photons a given dose can deliver. By the EUV era, this translator's problem is no longer "not understanding the language" — it's that the "words" it receives are simply too few.

At a given exposure dose, the number of photons actually arriving per unit area is finite, and photon arrival is itself a random (Poisson) process. When too few photons land at a given exposure location, photon shot noise becomes a major source of pattern uncertainty: the initial spatial distribution of the photoacid generator, statistical fluctuation in the exposure process itself, and thermally driven diffusion fluctuations during post-exposure bake all stack together, directly driving up line-edge roughness (LER) and critical-dimension (CD) variability.

This tension isn't new — EUV simply pushed it to a physical limit. The field calls it the RLS trade-off (Resolution–LER–Sensitivity trade-off): resolution, line-edge roughness, and sensitivity are extremely difficult to optimize simultaneously — pushing sensitivity higher (imaging with less light dose) generally means less control and rougher edges; pushing LER lower usually demands a higher exposure dose, sacrificing sensitivity; and pushing resolution higher amplifies the tension between the other two. This isn't an engineering failure at any one company — it's a core, decades-old problem repeatedly discussed at SPIE and other industry conferences, and essentially every resist-formulation paper is, at bottom, searching for an acceptable balance point within this triangle for the process node at hand. At EUV wavelengths, what conventional chemically amplified resist is approaching isn't some formula detail that hasn't been tuned right — it's the physical ceiling of the RLS triangle itself.

Against this backdrop, the industry has begun exploring routes beyond conventional chemical amplification, the most notable being metal-oxide resist. These resists are built around metal-oxide clusters — for instance, organotin oxo-clusters — as their core structure, rather than the conventional polymer-PAG system. The logic: metal atoms present a higher absorption cross-section to EUV photons, absorbing more incident photons and thereby reducing the impact of photon noise on imaging at the same dose. These are molecular-scale metal-oxide cluster structures, not simple dispersed nanoparticles. In 2021, JSR acquired Inpria, a pioneer in this field, for $514 million — a landmark event in this route's move into the mainstream.

This is the second layer of the answer to "why only a handful of companies are left": as conventional chemical amplification approaches its physical limit, the number of teams that can simultaneously master both polymer photochemistry and an entirely different metal-oxide-cluster chemistry is smaller than the generation before it.

5. The Real Moat: Not the Formula, but "Process + Qualification + Data + Time"

Chemical amplification's underlying principle has been public for more than four decades — so why can only a handful of companies still reliably supply advanced photoresist today? The common but inaccurate explanation is "formula secrecy."

The truth is closer to the opposite: the formula can be published, and the process route can be studied and replicated. What cannot be replicated is qualification on the customer's own production line. The real barrier lies outside the formula.

Semiconductor-grade photoresist carries extremely strict specifications for metal-ion contamination and particle count — but what actually costs a supplier its customer's trust is rarely a single batch failing to meet a purity spec. It's the extremely subtle drift between batches — even a small shift in the resin's molecular-weight distribution can be enough to depress yield at the fab. The real enemy of semiconductor materials is change itself, not impurity in isolation.

And switching suppliers is never a procurement decision — it's a process re-qualification, and industry analysis is explicit that this takes twelve months at the very fastest. For a major formula change at a critical process node — a PFAS-replacement iteration, for instance — going from receiving samples to full qualification and volume introduction can take three to six years or longer; even for relatively mature applications, qualification cycles typically run 18 to 24 months; and in reliability-critical settings like automotive-grade semiconductors, qualifying a new photoresist chemistry alone can take two to five years, sitting squarely on a new vehicle program's critical path. Because qualification is measured in years and requires running large volumes of real wafers on an actual production line to accumulate yield data, the industry has naturally developed extremely high customer stickiness — even if a new formula has already "succeeded" in the lab, a fab will not switch suppliers because of that alone.

So what has actually stabilized in this industry is less a set of oligopolies than a tightly locked-in co-evolutionary relationship: fabs don't readily switch resists; resist suppliers don't readily switch resin platforms; resin platforms don't readily switch photoacid generators; and scanner exposure parameters aren't readily retuned for new materials. Every link in the chain has locked itself around the existing state of every other link — and that, more precisely, is what the "moat" actually looks like: Process, customer Qualification, yield and process-window Data, and the Time that qualification itself requires — these four, stacked together, are why this industry hasn't been easily replicated.

One comparison worth correcting here, because it's often oversimplified: it is not the case that "the scanner is easy to replicate, the resist is hard." The opposite is true. ASML holds close to 100% of the EUV lithography-tool market and roughly 90% of the overall lithography-tool market, and its own moat runs just as deep — decades of optics and precision-manufacturing R&D, an exclusive supply of ultra-precision lenses from Zeiss, a network of thousands of precision-component suppliers — no competitor has managed to replicate it. The barriers around the scanner and around the resist are two different kinds of "hard": the scanner's difficulty is rooted mainly in engineering depth and precision-manufacturing supply-chain depth; the resist's difficulty lies in the fact that its real performance can only be proven by running large volumes of actual wafers on a specific customer's line and accumulating yield data — a qualification relationship built customer by customer, which can't be bought and can't be taken apart to study.

6. Supply Chains and Geopolitics: Why a Material Becomes a Strategic Resource

The reason the 2019 Japan-Korea episode shook the world is that it precisely exposed how fragile this co-evolutionary relationship becomes once interrupted. By industry market reports, four Japanese companies — JSR, Tokyo Ohka Kogyo (TOK), Shin-Etsu Chemical, and Fujifilm — together hold the majority of the global semiconductor photoresist market (figures across different reports range roughly from 70% to 90%). At the highest technical tier — EUV photoresist — only three companies currently hold qualification for volume production: TOK, JSR, and Shin-Etsu, all three also Japanese.

This is a separate matter from concentration in the lithography-tool market: ASML's near-100% share of EUV scanners is concentration at the equipment layer; photoresist supply concentrated among a handful of Japanese firms is concentration at the materials layer — and it's the combination of the two that makes advanced-node supply chains genuinely fragile: even with the equipment in hand, without qualified photoresist, exposure simply cannot proceed. A number of economies have, in recent years, continued increasing R&D investment in advanced lithography materials, and domestic-substitution for materials has become a significant thread of industrial policy in several places; but public sources broadly indicate that high-end photoresist — particularly at the ArF-immersion and EUV tiers — still faces long-term challenges in materials chemistry, manufacturing consistency, and customer qualification cycles. This is entirely consistent with the moat logic described above: even with the formula itself in hand, entering an advanced fab's volume-production sequence still requires clearing the qualification and data-accumulation threshold.

ChemAbout Insight

The semiconductor industry is often described as a race over transistor size. At a deeper level, it is equally a race over how a material responds to a beam of light — and every reduction in exposure wavelength forces this translator, which converts optical information into a physical pattern, to reinvent itself. A new photoresist typically needs one to several years of qualification to move from sample stage into volume production at an advanced fab, and what it must withstand along the way is a manufacturing environment where even the most minute batch-to-batch variation can move yield. What's genuinely hard to replicate was never the formula — it's the manufacturing process, process window, and yield data accumulated over those years. What's genuinely scarce was never companies that can make photoresist — it's companies that a fab trusts enough to actually switch to.

---

Evidence Notes

Key Facts

• Invention history: Science History Institute, "Patterning the World: The Rise of Chemically Amplified Photoresists" (Ito/Willson/Fréchet, IBM San Jose, 1982); Ito, H. "Chemical amplification resists: history and development within IBM," IBM Journal of Research and Development, Vol. 41, 1997.
• Chemistry & mechanism: "Resolution, LER and sensitivity limitations of photoresists," SPIE Proceedings art. 69211E; "Study on RLS trade-off resist upgrade for production-ready EUV lithography," SPIE/ResearchGate; NIST, "Modeling the effects of acid amplifiers on photoresist," and related stochastic-effects/LER research.

Industrial Map

• EUV & metal-oxide resist: JSR Corporation, "JSR Agrees to Acquire EUV Pioneer Inpria Corporation," Businesswire, 2021-09-17 (closing 2021-11-11); Inpria Corporation technology overview (metal-oxide photoresist, tin-oxo cluster chemistry).

Regulatory Status

• 2019 Japan-Korea export controls: RIETI, "The impact of export controls on international trade: Evidence from the Japan-Korea trade dispute in the semiconductor industry"; USITC, "The South Korea-Japan Trade Dispute in Context: Semiconductor Manufacturing Chemicals and Concentrated Supply Chains"; C&EN, "Japan hits South Korea with controls on key electronic materials," 2019; Global Trade Alert, "Japan: Government announces export control measures to the Republic of Korea."

Supply Chain Notes

• Supplier concentration & qualification: photoresist market supplier-concentration reports (JSR/TOK/Shin-Etsu/Fujifilm share), synthesized from multiple industry reports; semiconductor-material qualification-cycle analysis (automotive-grade / advanced-node cycles); ASML EUV-scanner market-share coverage.

Sources & Method

• Grading: primary literature, regulatory documents, and peer-reviewed journals take priority; industry and market figures are labeled as third-party estimates. Some market-share figures vary by report methodology and year; the text presents these as ranges / "per reports" rather than single precise figures.