# ALD and self-limiting surface reactions

Canonical: https://semiaifoundry.com/learn/atlas/ald-self-limiting-growth/
Revision: 2026-10-07
Rights: https://semiaifoundry.com/ai-atlas/licenses.html

Fabrication / Repeated operations · Atlas v1.3.0

# ALD and self-limiting surface reactions

Builds a film through separated chemical exposures and controlled surface saturation.

[Explore in Atlas](https://semiaifoundry.com/ai-atlas/#ald-self-limiting-growth)Nanometer films on a wafer; thickness exaggerated

Learning objectiveExplain saturation, separated exposures and the limits of cycle-count control.

## The mechanism

Atomic layer deposition separates precursor reactions in time or space so that surface reactions can saturate. A precursor reacts with available surface sites; excess material and by-products are removed; another exposure completes the reaction cycle. Repetition builds thickness with fine control. The name does not promise one complete atomic plane per cycle: growth depends on chemistry, nucleation and operating conditions. Deep or crowded features still need adequate reactant transport and removal. ALD is valuable where a thin dielectric or liner must follow three-dimensional geometry, but surface selectivity, purity and throughput must be evaluated alongside conformality.

1.  Expose the prepared surface to a precursor that reacts with available sites.
2.  Allow sufficient access into the intended feature so saturation is not restricted to its entrance.
3.  Remove unreacted species and products before an incompatible second exposure.
4.  Complete the complementary surface reaction and repeat the separated cycle.
5.  Measure nucleation, thickness and composition to determine whether the assumed growth regime actually occurred.

## Science and assumptions

### Steady-cycle thickness estimate

$$
t\approx g(N-N_0)
$$

Steady-regime thickness is approximately growth per cycle times applied cycles minus the effective nucleation-delay count.

Use only in the calibrated growth regime with N ≥ N₀; the expression is not a negative-thickness prediction before nucleation.

Symbols and units

t

Film thickness in the established growth regime

nm

g

Measured steady growth per cycle

nm/cycle

N

Applied cycle count

cycles

N₀

Effective nucleation-delay cycle count

cycles

t is thickness in nm, g measured growth per cycle in nm/cycle, N cycle count and N₀ an effective nucleation-delay count.

Assumptions An empirical estimate in a stable growth regime. It cannot be extrapolated through changing chemistry or incomplete saturation.

## Engineering example

Synthetic example: g = 0.08 nm/cycle over 100 steady cycles predicts 8 nm. If early nucleation delays growth by 10 equivalent cycles, the same cycle count instead predicts about 7.2 nm.

## Distinctions and tradeoffs

ALD is not automatically one monolayer per cycle and does not eliminate transport limitations.

Longer exposures and purges can improve saturation in difficult features while reducing tool throughput.

Gate-all-around structures need functional films around channels; high-aspect-ratio memory and interconnect features also make surface access consequential.

## Specifications and interfaces

Control variable

Number of completed cycles

Growth per cycle must be measured for the actual material and regime.

Film output

Thickness and conformality

Nucleation delay may make early-cycle behavior differ from steady growth.

### Interfaces

Prepared surface and precursors Input

Reactive sites and separated exposures

Conformal thin film Output

Accumulated growth with controlled composition and interfaces

## Check your understanding

### Why might a deep feature need longer exposure than a flat witness wafer?

1.  Cycle count alone fixes every local thickness
2.  Reactants must reach and saturate its internal surfaces
3.  Deep features contain no surface sites

Answer and reasoning

Reactants must reach and saturate its internal surfaces

A saturated open surface does not establish sufficient dose and transport at the bottom of a narrow feature.

## Evidence and scope

The numbers illustrate an empirical model and are not a vendor recipe. Material-specific ALD windows and selectivity must be established experimentally.

Use the manufacturing investigation to trace compatible inputs, assembly and operating constraints; no current standalone calculator models this lesson’s full mechanism.

## Sources

-   [Lam Research — Our Processes](https://www.lamresearch.com/products/our-processes/)
-   [Applied Materials — New Ways to Shrink](https://ir.appliedmaterials.com/static-files/f2fbf48d-a761-4c4f-ab4c-73f312ab1ebc)
-   [MIT 6.152J — Micro/Nano Processing Technology lecture notes](https://ocw.mit.edu/courses/6-152j-micro-nano-processing-technology-fall-2005/pages/lecture-notes/)

[Published lesson data](https://semiaifoundry.com/ai-atlas/lesson-data/ald-self-limiting-growth-a831d1b62669b6de.json) · [Atlas releases and licensing](https://semiaifoundry.com/ai-atlas/licenses.html)

## Continue exploring

### Foundations

-   [CVD and PVD: two ways to supply a film](https://semiaifoundry.com/learn/atlas/cvd-pvd-film-transport/)

### Related lessons

-   [Nanosheet release and replacement gates](https://semiaifoundry.com/learn/atlas/nanosheet-gate-last-integration/)
-   [Thermal oxidation and the silicon interface](https://semiaifoundry.com/learn/atlas/thermal-oxidation-interface/)