photoresist
**Photoresist** is the radiation-sensitive polymer film that transfers circuit patterns from mask to wafer — it is exposed by light (DUV 193 nm, EUV 13.5 nm) or electrons (e-beam), undergoes a chemical change that makes exposed (positive tone) or unexposed (negative tone) regions soluble in developer, and the remaining pattern acts as an etch mask or implant block for the layer underneath. Every one of the 60–100 patterning steps in a modern chip flow relies on a photoresist, making resist chemistry the molecular interface between the optical system and the silicon.
**Positive vs negative tone.** In positive resist, exposure breaks polymer chains (or generates acid that deprotects them during bake), making exposed regions dissolve in aqueous base developer (TMAH 2.38%). In negative resist, exposure cross-links the polymer, making exposed regions insoluble — unexposed material washes away. Modern logic lithography uses positive-tone resists almost exclusively because they offer higher resolution and better CD control; negative-tone development (NTD) on a positive resist (dissolving the unexposed regions in organic solvent) is used selectively for dense hole/trench patterns where it produces better profile.
**Chemically amplified resists (CAR) — the DUV workhorse.** Since the mid-1990s, all production DUV resists are chemically amplified: a photoacid generator (PAG) absorbs a photon and releases a strong acid (H⁺), which during post-exposure bake (PEB) catalytically deprotects hundreds of polymer sites per acid molecule — amplifying the chemical contrast far beyond what direct photolysis could achieve. The acid diffusion length during PEB sets the fundamental resolution blur:
$$\sigma_{\text{blur}} = \sqrt{2 D_{\text{acid}} \cdot t_{\text{PEB}}}$$
where $D_{\text{acid}}$ is the acid diffusion coefficient (~1–5 nm²/s at typical PEB temperatures) and $t_{\text{PEB}}$ is bake time (60–90 s). At 193 nm immersion with 38–40 nm half-pitch, this blur contributes 3–5 nm of image degradation — manageable. But at EUV's 13 nm HP target, a 3 nm blur is catastrophic, which drives the transition to non-CAR materials.
**The stochastic challenge — photon shot noise.** At EUV wavelengths, each 13.5 nm photon carries 92 eV — roughly 14× more energy than a 193 nm photon. This means fewer photons per unit dose, and Poisson statistics dictate that feature-to-feature dose variation scales as $1/\sqrt{N}$ where $N$ is photons per feature area. For a 10 nm line at typical EUV dose (~30–60 mJ/cm²), only ~30–100 photons land on a single feature — creating measurable line-edge roughness (LER) and local CD variation (LCDU) from pure counting statistics:
$$\text{LER}_{3\sigma} \propto \frac{1}{\sqrt{\text{dose} \cdot \text{area}}} \cdot f(\text{resist chemistry})$$
The "RLS trade-off" (Resolution–LER–Sensitivity) is the fundamental triangle: improving any one metric degrades the others. Higher dose reduces LER but lowers throughput; higher sensitivity improves throughput but increases chemical noise.
| Resist type | Chemistry | Wavelength | Resolution | Sensitivity | Key application |
|---|---|---|---|---|---|
| DNQ/novolac | Diazonaphthoquinone + novolac | g/i-line (436/365 nm) | ~350 nm | High | Legacy, packaging |
| CAR (positive) | PAG + protected polymer (PHS/PMMA) | KrF 248 nm | ~130 nm | Medium-high | Mature nodes |
| CAR (ArF dry) | PAG + methacrylate/cyclo-olefin | ArF 193 nm | ~65 nm | Medium | Via/trench layers |
| CAR (ArF immersion) | Same + topcoat, immersion-compatible | ArF-i 193 nm | ~38 nm (single), ~20 nm (SADP) | Medium | Critical layers N7+ |
| CAR (EUV) | PAG + high-EUV-absorbing polymer | EUV 13.5 nm | ~13 nm | Low (30–80 mJ/cm²) | Logic N5–N3 |
| Metal-oxide (MOx) | Tin/hafnium/zirconium oxide clusters | EUV 13.5 nm | ~8–10 nm | High (15–30 mJ/cm²) | EUV N2+ and High-NA |
| Dry-development resist | Metal-oxide (no wet puddle) | EUV 13.5 nm | ~8 nm | Medium | High-NA, anti-collapse |
**EUV resists — the frontier.** At 13.5 nm wavelength, resist films must be ultra-thin (20–30 nm) to avoid image-blur from electron scattering within the film, yet thick enough to survive the etch below. CAR at EUV faces two problems: (1) low EUV absorption (the polymer is mostly carbon — transparent at 13.5 nm), requiring high dose and low throughput; (2) acid diffusion blur is comparable to feature size. Metal-oxide resists (MOx) solve both: the metal (Sn, Hf, Zr) absorbs EUV 2–3× more efficiently, and the inorganic matrix has near-zero diffusion blur (the exposed region crosslinks in place). Inpria (now acquired by JSR) pioneered tin-oxide EUV resists; they are entering production at N2 and High-NA nodes.
**Dry development — eliminating pattern collapse.** When resist features are tall and narrow (aspect ratio >3:1), surface tension during wet-development rinse pulls adjacent lines together (pattern collapse). Dry-development resists (etch-back development in a plasma chamber rather than wet puddle) eliminate the liquid meniscus entirely, enabling thinner, taller features without collapse — critical for High-NA EUV where resist is only 20–25 nm thick with sub-10 nm features.
```svg
```
**The resist supply chain.** Photoresist manufacturing is one of the most concentrated industries in semiconductors: JSR, TOK (Tokyo Ohka Kogyo), Shin-Etsu Chemical, Sumitomo, and Fujifilm supply >90% of all advanced resists. EUV resist is particularly constrained — only JSR/Inpria (metal-oxide) and TOK/Shin-Etsu (CAR) have production-qualified EUV materials. A single resist formulation takes 3–5 years from R&D to high-volume qualification, making resist supply a potential chokepoint for node transitions.
**What photoresist means for the chip stack.** Resist performance directly gates what lithography can achieve: if the resist can't resolve a feature, no amount of optical engineering helps. The transition from CAR to metal-oxide resists at EUV and High-NA is as consequential as the wavelength change itself — it's the chemistry that ultimately determines whether 8 nm half-pitch is printable in manufacturing. Every CFS simulator page that involves patterning — the Lithography Aerial Image at /lithography and the Etch Simulator at /simulate — ultimately depends on a photoresist step that faithfully captures the aerial image into a physical mask for pattern transfer.