Atomic layer deposition

Atomic layer deposition emerged in 1974 not because someone wanted atomic-precision film growth, but because the conditions aligned: surface chemistry understood self-limiting reactions, vacuum technology could create controlled environments, and electroluminescent displays demanded ultra-thin insulating films with perfect uniformity that conventional deposition couldn't achieve. For decades, thin film manufacturing meant chemical vapor deposition (CVD) or physical vapor deposition (PVD)—continuous processes where precursor gases or evaporated materials coated substrates. These methods worked for many applications but suffered from poor step coverage (film thickness varied on complex 3D structures) and limited thickness control. For microelectronics shrinking toward nanometer scales, these limitations would become fatal.

Tuomo Suntola, a Finnish physicist at Instrumentarium Oy in Helsinki, solved this in 1974 while developing electroluminescent displays for medical imaging. His innovation combined existing knowledge in a new sequence: instead of continuous deposition, use alternating self-limiting reactions. Introduce precursor gas A, wait for it to form a complete monolayer on the surface (self-limiting—no more than one atomic layer reacts), purge excess gas, introduce precursor gas B which reacts only with the monolayer from A, purge again. Repeat. Each cycle deposits exactly one atomic layer, regardless of substrate geometry. Suntola called this Atomic Layer Epitaxy (ALE), later renamed Atomic Layer Deposition (ALD).

The breakthrough was recognizing that surface chemistry could be harnessed for digital film growth—binary on/off reactions rather than continuous analog processes. The key insight came when Suntola switched from high vacuum to inert gas reactors, enabling use of reactive precursors like metal chlorides and water vapor that would be too dangerous in conventional CVD. This made ALD practical for manufacturing rather than laboratory curiosity.

For a decade, ALD remained confined to electroluminescent displays. The process was slow (one atomic layer per cycle, requiring hundreds or thousands of cycles for useful films) and seemed impractical for high-volume manufacturing. But in the early 2000s, semiconductor industry faced crisis: Moore's Law was stalling. Silicon dioxide gate dielectrics had reached their physical limit at about 1.2 nanometers—make them thinner and quantum tunneling causes catastrophic leakage. Continuing transistor miniaturization required replacing silicon dioxide with high-κ dielectrics (materials with higher dielectric constant allowing thicker physical films with same electrical properties). But conventional CVD couldn't deposit high-κ materials like hafnium oxide with the required uniformity and conformality on increasingly complex 3D transistor structures.

ALD became the enabling technology. Intel adopted ALD for hafnium oxide gate dielectrics in their 45nm process node (2007), replacing silicon dioxide for the first time in 40 years of semiconductor manufacturing. The technique's atomic-level precision allowed deposition of uniform 2-3nm high-κ films conforming perfectly to FinFET sidewalls and trenches. TSMC, Samsung, and other manufacturers followed. By 2025, every leading-edge semiconductor process—Intel 18A (1.8nm), TSMC N2 (2nm), Samsung 2nm—relies critically on ALD for multiple layers: gate dielectrics, spacer films, barrier layers, and emerging 2D materials integration. Modern logic chips use ALD for 20+ processing steps.

This was punctuated equilibrium in materials processing. Thin film deposition had evolved incrementally for decades—better CVD, better PVD—then suddenly leaped to atomic-layer control. The catalyst wasn't conceptual—chemists understood self-limiting surface reactions. The catalyst was industrial need meeting available chemistry: Moore's Law demanded precision that only atomic-layer processes could provide, and ALD chemistry had matured sufficiently to scale from research to manufacturing.

The invention demonstrates exaptation twice. Suntola developed ALD for electroluminescent displays—an application that became obsolete as LCD and LED displays dominated. But semiconductor industry repurposed the same technique for an entirely different problem (high-κ gate dielectrics). Then ALD was repurposed again for photovoltaics, batteries, catalysts, and protective coatings. The same layer-by-layer chemistry solving different problems wherever atomic-precision film control matters.

ALD also exhibits strong path-dependence. Once semiconductor manufacturing adopted ALD for gate dielectrics, subsequent process innovations followed that architecture: ALD for spacers, ALD for work function metals, ALD for 2D material integration. Alternative approaches—molecular layer deposition variants, hybrid plasma-thermal ALD—emerged as extensions rather than replacements. The format locked in because fabrication facilities (fabs) invested billions in ALD equipment and process development. Changing fundamental deposition methodology would require replacing entire manufacturing infrastructures.

The biological parallel is nacre (mother of pearl) formation in abalone shells. Like atomic layer deposition which builds films by alternating self-limiting chemical reactions—layer A, purge, layer B, purge, repeat—abalone secrete nacre by alternating deposition of calcium carbonate platelets and organic protein interlayers. Both processes build composite materials with atomic/molecular precision through sequential self-limiting steps. Both achieve superior properties through layered architecture: nacre's 3000x toughness improvement over calcium carbonate alone; ALD films' superior conformality and uniformity over conventional deposition. Both demonstrate that layer-by-layer construction with precise interfacial control creates materials impossible to achieve through continuous bulk processes. The abalone doesn't deposit nacre continuously—it alternates between mineral and protein deposition phases, just as ALD alternates between precursor exposures.

The invention also demonstrates enabling technology and cascading effects. ALD didn't just improve existing devices—it enabled entirely new device architectures that would be impossible otherwise. FinFET transistors (Intel 22nm, 2011) required ALD for conformal gate oxide on vertical fins. Gate-all-around (GAA) nanosheet transistors (TSMC N2, Samsung 2nm, Intel 18A in 2025-2026) are only manufacturable with ALD providing uniform films wrapping completely around nanometer-scale silicon channels. Without ALD, Moore's Law would have stalled at 45nm around 2010. The technique extended semiconductor scaling another 15+ years and enabled the mobile computing revolution, AI accelerators, and advanced packaging technologies.

By 2026, atomic layer deposition is foundational to modern civilization despite public invisibility. Every smartphone, laptop, data center server, and AI chip relies on ALD-deposited films. The technique enabled the continuation of Moore's Law through multiple technology nodes. The invention reached its adjacent possible in 1974 when surface chemistry understanding met electroluminescent display requirements in Helsinki. The human who recognized self-limiting reactions could enable atomic-precision won recognition (Suntola received the Millennium Technology Prize in 2018). But the invention was responding to selection pressure—device scaling demanded atomic-level control. If not Suntola in 1974, then someone else within years, because the conditions had aligned.