Photolithography

If chip manufacturing has a single signature move, it’s this one: printing an entire circuit pattern — millions of shapes, some just a few dozen atoms wide — onto a wafer using light, the same way a photograph is developed from a negative. Photolithography is repeated dozens of times per chip, once for nearly every layer, and its precision limits are one of the biggest reasons chip progress has gotten so hard in recent years.

The core idea: printing with light

Photolithography works like a very precise combination of photography and stencil-making. A light-sensitive chemical coating captures a pattern when exposed to light, and then a chemical “developer” washes away the parts that were (or weren’t) exposed, leaving the pattern behind in solid form.

The photolithography sequence

1. Coat — the wafer is spun at high speed while a drop of liquid photoresist is applied, spreading it into an extremely thin, uniform film across the whole surface (spin-coating).
2. Align and expose — the wafer is loaded into a machine called a stepper or scanner, which aligns a mask (also called a reticle) — a glass plate carrying the circuit pattern for this layer — precisely over the wafer, then shines ultraviolet light through it. The light passes through clear areas of the mask and is blocked by opaque areas, projecting the pattern onto the photoresist, typically shrunk down by a lens (commonly 4:1).
3. Develop — a chemical developer is applied. Depending on the type of photoresist, either the exposed regions or the unexposed regions dissolve away, leaving a solid resist pattern that matches (or inverts) the mask pattern.
4. Inspect — the pattern is checked under a microscope or with automated optical/electron-beam inspection before the wafer is allowed to continue to etching or deposition.

    flowchart LR
    A["Bare or coated wafer"] --> B["Spin-coat photoresist"]
    B --> C["Align mask/reticle,<br/>expose with UV light"]
    C --> D["Develop:<br/>dissolve exposed or unexposed resist"]
    D --> E["Resist pattern<br/>on wafer"]
    E --> F["Etch or implant<br/>through resist openings"]
    F --> G["Strip remaining resist"]
  

What the exposure setup looks like

The resolution limit: why smaller features get so hard

Light can’t be focused into an infinitely sharp point — it diffracts, spreading out slightly as it passes through any small opening. That diffraction sets a hard physical floor on how small a feature photolithography can print, described by a version of the Rayleigh criterion:

Here, $\lambda$ is the wavelength of the light used, $NA$ is the numerical aperture of the lens (roughly, how wide a cone of light it can gather and focus — bigger is better), and $k_1$ is a process-dependent factor (always greater than roughly 0.25) that captures how cleverly the exposure is engineered. Smaller resolution numbers are better — they mean finer features.

This formula explains almost the entire history of lithography technology: to print smaller transistors, engineers have spent decades pushing $\lambda$ down (moving from visible light, to UV, to deep-UV lasers at 193 nm, and eventually to 13.5 nm extreme ultraviolet light), pushing $NA$ up (better lenses), and pushing $k_1$ down (clever tricks like multiple-patterning, where one layer is split across several exposures). The most advanced modern lithography uses EUV light and is involved enough to deserve its own page — see EUV Lithography for that deep dive.

What happens after developing

The developed resist pattern isn’t the final structure — it’s a stencil for the next step. Wherever resist has been removed, the wafer surface underneath is exposed and can be etched away or have dopant ions implanted into it, while resist-covered areas are protected. Once that next step is done, the remaining resist is stripped off entirely, ready for the next lithography cycle on the next layer. See Etching and Deposition and Ion Implantation and Doping for what typically happens right after this step.

Key takeaways

• Photolithography prints a circuit pattern onto a wafer by exposing light-sensitive photoresist through a patterned mask, then chemically developing it — much like photography.
• The sequence is coat, expose, develop, inspect, and it repeats for nearly every layer of a chip.
• Resolution is fundamentally limited by light diffraction: $\text{Resolution} = k_1 \cdot \lambda / NA$, so shrinking wavelength and improving lens aperture are the main levers for finer features.
• The developed resist acts as a protective stencil for the etching, deposition, or implantation step that follows.