How Photolithography & FinFETs Work: From Silicon Ingot to Nanometre-Scale Transistors
By Neil Harwani • 2 min read • Published 13 Nov 2026
Gate-all-around nanosheet transistors control current better than FinFETs at the newest nodes, but leading-edge wafers are far more expensive — one reason large chips are increasingly split into smaller chiplets.
Student lens
From sand to switch
A transistor is a switch: a voltage on the gate decides whether current flows between source and drain. Chip-making is the art of building billions of these switches on a silicon wafer.
- Grow and slice a pure silicon crystal into wafers.
- Dope regions with impurities (boron, phosphorus) to create p-type and n-type silicon.
- Pattern each layer with photolithography: coat with photoresist, expose through a mask, develop, then etch or deposit.
- Repeat for dozens of layers, then add metal wiring.
Why EUV matters
The smallest feature you can print is limited by the wavelength of light. Extreme ultraviolet (EUV) lithography uses 13.5 nm light, far shorter than the 193 nm deep-UV light used before it, so fewer multi-patterning steps are needed for the finest layers.
Planar → FinFET → Gate-all-around
As transistors shrank, the gate lost control of the channel and current leaked even when the switch was "off". FinFETs wrapped the gate around three sides of a fin; gate-all-around (GAA) nanosheets wrap it around all four. Better control means less leakage at the same size.
Study next: MOS capacitor band diagrams and the threshold voltage $V_T$.
Young-pro lens
The business problem
Every node shrink buys performance and power efficiency, but at a rising price per wafer. For hardware and product leaders, the key question is not "which node is best" but "which node pays back for this chip".
Yield punishes big dies
Defects land randomly on a wafer. A larger die is more likely to contain one, so yield falls as area grows. Murphy's model captures this:
where $A$ is die area and $D_0$ defect density. Try the wafer-yield lab to see cost per good die climb steeply with area.
What this means for strategy
- Chiplets: splitting one large die into several small ones raises yield and lets you mix nodes (leading-edge compute, older-node I/O).
- Custom ASIC vs. off-the-shelf: high non-recurring engineering cost only pays off at volume.
- Supply concentration: very few foundries offer the leading edge, which is a pricing and continuity risk to plan around.
Check current wafer prices and node availability with your foundry or a recent industry source before using figures in a business case.