Atomic layer deposition (ALD) is a cyclic thin-film method that separates mutually reactive species in time so each exposure can approach a self-limiting surface reaction. A typical cycle consists of precursor exposure, purge, co-reactant exposure, and purge. Repeating the cycle provides sub-nanometer thickness control and can produce highly conformal films on complex three-dimensional structures. “Atomic layer” describes the control strategy, not a guarantee that every cycle deposits exactly one crystalline monolayer: growth per cycle is chemistry-, temperature-, surface-, and process-dependent and is often sub-monolayer.
ALD grows a film through sequential surface-reaction cycles. A precursor exposure reacts with available surface sites until steric blocking, site consumption, or reaction equilibrium limits further uptake; a purge removes excess precursor and volatile products before the co-reactant arrives. The co-reactant converts the adsorbed species and prepares the surface for another cycle. In an established saturation window, thickness is approximately cycles multiplied by measured growth per cycle, but nucleation delay, surface evolution, precursor decomposition, soft saturation, and chamber memory can make the early or late cycles nonlinear. High-aspect-ratio conformality additionally requires enough dose and time for molecules to reach remote surfaces, plus enough purge to prevent overlap and parasitic CVD.
Atomic layer etching is a useful comparison, but it is not simply ALD played backward. ALE alternates a surface-modification step with a removal step chosen to act preferentially on the modified material. Etch per cycle may be sub-monolayer, one layer, or more depending on material, chemistry, ion energy, temperature, and saturation behavior. Selectivity and low damage must be measured rather than assumed: incomplete modification, spontaneous etching, energetic-ion tails, redeposition, and feature charging can all break the ideal cycle model. ALD and ALE share cyclic surface control, yet their reaction networks and failure mechanisms require separate qualification.
| ALD (deposition) | ALE (etch) | |
|---|---|---|
| Goal | add material | remove material |
| Cycle | precursor → purge → co-reactant → purge | modify → purge → remove → purge |
| Self-limiting because | sites saturate with precursor | only top layer is modified |
| Per cycle | chemistry-specific growth increment | chemistry-specific removal increment |
| Amount set by | cycles × qualified growth per cycle | cycles × qualified etch per cycle |
| Signature strength | high conformality after saturation | controlled, potentially low-damage removal |
| Trade | slow (throughput) | slow (throughput) |
<svg viewBox="0 0 760 470" xmlns="http://www.w3.org/2000/svg" font-family="Segoe UI,Arial,sans-serif"><rect width="760" height="470" fill="#0d1117"/><defs><marker id="ah" markerWidth="8" markerHeight="8" refX="6.5" refY="3" orient="auto"><path d="M0,0 L7,3 L0,6 Z" fill="#8b949e"/></marker><marker id="ay" markerWidth="8" markerHeight="8" refX="6.5" refY="3" orient="auto"><path d="M0,0 L7,3 L0,6 Z" fill="#e0b13a"/></marker></defs><text x="20" y="30" fill="#e6edf3" font-size="19" font-weight="700">ALD & ALE: add or remove material through cyclic surface-limited steps</text><text x="20" y="50" fill="#8b949e" font-size="12.5">Both split a continuous process into self-terminating half-reactions. ALD grows a layer; ALE removes one. Amount = cycles × per-cycle.</text><!-- Panel 1: ALD --><rect x="20" y="66" width="226" height="298" rx="7" fill="#0c141d" stroke="#30363d"/><text x="32" y="87" fill="#34d399" font-size="13.5" font-weight="600">ALD · grow by measured GPC</text><text x="32" y="105" fill="#8b949e" font-size="9" font-weight="600">1 · reactive surface (—OH sites)</text><circle cx="40" cy="111" r="2.2" fill="#f87171"/><circle cx="52" cy="111" r="2.2" fill="#f87171"/><circle cx="64" cy="111" r="2.2" fill="#f87171"/><circle cx="76" cy="111" r="2.2" fill="#f87171"/><circle cx="88" cy="111" r="2.2" fill="#f87171"/><rect x="34" y="114" width="66" height="10" fill="#2b3340"/><text x="110" y="115" fill="#8b949e" font-size="8.5">dangling —OH sites,</text><text x="110" y="126" fill="#8b949e" font-size="8.5">nothing grows alone</text><text x="32" y="150" fill="#38bdf8" font-size="9" font-weight="600">2 · precursor A pulse → saturates</text><circle cx="40" cy="159" r="3" fill="#38bdf8"/><circle cx="52" cy="159" r="3" fill="#38bdf8"/><circle cx="64" cy="159" r="3" fill="#38bdf8"/><circle cx="76" cy="159" r="3" fill="#38bdf8"/><circle cx="88" cy="159" r="3" fill="#38bdf8"/><rect x="34" y="163" width="66" height="10" fill="#2b3340"/><text x="110" y="160" fill="#8b949e" font-size="8.5">each site takes one A,</text><text x="110" y="171" fill="#8b949e" font-size="8.5">then self-limits</text><text x="110" y="182" fill="#6b7280" font-size="8">purge: excess + byproduct out</text><text x="32" y="204" fill="#34d399" font-size="9" font-weight="600">3 · co-reactant B → film increment</text><circle cx="40" cy="211" r="2.2" fill="#f87171"/><circle cx="52" cy="211" r="2.2" fill="#f87171"/><circle cx="64" cy="211" r="2.2" fill="#f87171"/><circle cx="76" cy="211" r="2.2" fill="#f87171"/><circle cx="88" cy="211" r="2.2" fill="#f87171"/><rect x="34" y="214" width="66" height="6" fill="#143524" stroke="#34d399" stroke-width="0.6"/><rect x="34" y="220" width="66" height="8" fill="#2b3340"/><text x="110" y="212" fill="#8b949e" font-size="8.5">B converts A → +1 layer,</text><text x="110" y="223" fill="#8b949e" font-size="8.5">regrows —OH for next</text><line x1="120" y1="248" x2="120" y2="256" stroke="#8b949e" stroke-width="1.2"/><path d="M120,256 L34,256 L34,114" fill="none" stroke="#8b949e" stroke-width="1.1" stroke-dasharray="3 2" marker-end="url(#ah)"/><text x="128" y="260" fill="#c9d1d9" font-size="9" font-weight="600">repeat × N cycles</text><text x="32" y="284" fill="#8b949e" font-size="8.5">thickness = N × growth-per-cycle,</text><text x="32" y="296" fill="#8b949e" font-size="8.5">tuned to the Ångström and perfectly</text><text x="32" y="308" fill="#8b949e" font-size="8.5">conformal into high-aspect trenches.</text><text x="32" y="330" fill="#7cc7f5" font-size="8.5" font-weight="600">Two saturating half-reactions:</text><text x="32" y="342" fill="#8b949e" font-size="8.5">precursor pulse, then co-reactant pulse,</text><text x="32" y="354" fill="#8b949e" font-size="8.5">each stopping once the surface is covered.</text><!-- Panel 2: ALE mirror --><rect x="267" y="66" width="226" height="298" rx="7" fill="#0c141d" stroke="#30363d"/><text x="279" y="87" fill="#e0913a" font-size="13.5" font-weight="600">ALE · cyclic surface removal</text><text x="279" y="105" fill="#8b949e" font-size="9" font-weight="600">1 · film surface to etch</text><rect x="281" y="110" width="66" height="6" fill="#2b3340"/><rect x="281" y="117" width="66" height="6" fill="#2b3340"/><rect x="281" y="124" width="66" height="6" fill="#2b3340"/><text x="357" y="119" fill="#8b949e" font-size="8.5">a film to thin, one</text><text x="357" y="130" fill="#8b949e" font-size="8.5">atomic layer at a time</text><text x="279" y="150" fill="#38bdf8" font-size="9" font-weight="600">2 · modify — Cl adsorbs on top</text><circle cx="287" cy="159" r="2.2" fill="#38bdf8"/><circle cx="299" cy="159" r="2.2" fill="#38bdf8"/><circle cx="311" cy="159" r="2.2" fill="#38bdf8"/><circle cx="323" cy="159" r="2.2" fill="#38bdf8"/><circle cx="335" cy="159" r="2.2" fill="#38bdf8"/><rect x="281" y="162" width="66" height="6" fill="#0a3350" stroke="#38bdf8" stroke-width="0.6"/><rect x="281" y="169" width="66" height="6" fill="#2b3340"/><rect x="281" y="176" width="66" height="6" fill="#2b3340"/><text x="357" y="167" fill="#8b949e" font-size="8.5">top layer only reacts —</text><text x="357" y="178" fill="#8b949e" font-size="8.5">the bulk is untouched</text><text x="279" y="202" fill="#e0b13a" font-size="9" font-weight="600">3 · low-energy removal → −1 layer</text><line x1="289" y1="208" x2="289" y2="217" stroke="#e0b13a" stroke-width="1.1" marker-end="url(#ay)"/><line x1="307" y1="208" x2="307" y2="217" stroke="#e0b13a" stroke-width="1.1" marker-end="url(#ay)"/><line x1="325" y1="208" x2="325" y2="217" stroke="#e0b13a" stroke-width="1.1" marker-end="url(#ay)"/><rect x="281" y="220" width="66" height="6" fill="#2b3340"/><rect x="281" y="227" width="66" height="6" fill="#2b3340"/><text x="357" y="225" fill="#8b949e" font-size="8.5">desorb the modified</text><text x="357" y="236" fill="#8b949e" font-size="8.5">layer — purge, repeat</text><line x1="367" y1="248" x2="367" y2="256" stroke="#8b949e" stroke-width="1.2"/><path d="M367,256 L281,256 L281,110" fill="none" stroke="#8b949e" stroke-width="1.1" stroke-dasharray="3 2" marker-end="url(#ah)"/><text x="375" y="260" fill="#c9d1d9" font-size="9" font-weight="600">repeat × N cycles</text><text x="279" y="284" fill="#8b949e" font-size="8.5">etch depth = N × etch-per-cycle,</text><text x="279" y="296" fill="#8b949e" font-size="8.5">atomic and low-damage — the exact</text><text x="279" y="308" fill="#8b949e" font-size="8.5">mirror of the ALD growth loop.</text><text x="279" y="330" fill="#f0a35a" font-size="8.5" font-weight="600">Two self-limiting half-reactions:</text><text x="279" y="342" fill="#8b949e" font-size="8.5">modify the top layer, then gently remove it,</text><text x="279" y="354" fill="#8b949e" font-size="8.5">each step stopping on its own.</text><!-- Panel 3: why --><rect x="514" y="66" width="226" height="298" rx="7" fill="#0c141d" stroke="#30363d"/><text x="526" y="87" fill="#e6edf3" font-size="13.5" font-weight="600">Why it matters</text><circle cx="532" cy="106" r="3.5" fill="#34d399"/><text x="542" y="109" fill="#e6edf3" font-size="10" font-weight="600">Self-limiting = atomic control</text><text x="542" y="121" fill="#8b949e" font-size="8.5">each half-reaction stops when the surface is</text><text x="542" y="132" fill="#8b949e" font-size="8.5">covered or cleared — you count cycles, not time</text><circle cx="532" cy="152" r="3.5" fill="#38bdf8"/><text x="542" y="155" fill="#e6edf3" font-size="10" font-weight="600">Conformal by nature</text><text x="542" y="167" fill="#8b949e" font-size="8.5">surface-reaction-limited, not flux-limited, so it</text><text x="542" y="178" fill="#8b949e" font-size="8.5">coats every wall of a deep trench equally</text><circle cx="532" cy="198" r="3.5" fill="#e0b13a"/><text x="542" y="201" fill="#e6edf3" font-size="10" font-weight="600">Pulse–purge–pulse–purge</text><text x="542" y="213" fill="#8b949e" font-size="8.5">ALD: precursor + co-reactant. ALE: modify +</text><text x="542" y="224" fill="#8b949e" font-size="8.5">remove. Same rhythm, opposite direction.</text><line x1="526" y1="242" x2="728" y2="242" stroke="#30363d" stroke-width="0.8"/><text x="526" y="260" fill="#f0d9b5" font-size="9.5" font-weight="600">Where it shows up</text><text x="526" y="278" fill="#8b949e" font-size="8.5">· HKMG gate dielectric (HfO₂) — a few Å thick</text><text x="526" y="292" fill="#8b949e" font-size="8.5">· diffusion barriers & liners (TiN, TaN, Co)</text><text x="526" y="306" fill="#8b949e" font-size="8.5">· gate-all-around nanosheet release + recess</text><text x="526" y="320" fill="#8b949e" font-size="8.5">· 3D scaling where thickness must be exact</text><text x="526" y="342" fill="#8b949e" font-size="8.5">Slower than CVD/RIE, so it is used where the</text><text x="526" y="354" fill="#8b949e" font-size="8.5">last atomic layers have to be precisely right.</text><!-- bottom cards --><rect x="20" y="384" width="226" height="70" rx="7" fill="#111a24" stroke="#30363d"/><text x="32" y="404" fill="#e6edf3" font-size="12" font-weight="600">Saturation is the trick</text><text x="32" y="420" fill="#8b949e" font-size="9.5">Each half-reaction ends when the surface is</text><text x="32" y="433" fill="#8b949e" font-size="9.5">fully covered or cleared. That self-termination</text><text x="32" y="446" fill="#8b949e" font-size="9.5">is what gives Ångström-level control.</text><rect x="267" y="384" width="226" height="70" rx="7" fill="#111a24" stroke="#30363d"/><text x="279" y="404" fill="#e6edf3" font-size="12" font-weight="600">Conformal, not line-of-sight</text><text x="279" y="420" fill="#8b949e" font-size="9.5">Because growth is surface-limited, ALD coats</text><text x="279" y="433" fill="#8b949e" font-size="9.5">deep trenches uniformly — where sputter PVD</text><text x="279" y="446" fill="#8b949e" font-size="9.5">and even CVD leave thin, shadowed walls.</text><rect x="514" y="384" width="226" height="70" rx="7" fill="#111a24" stroke="#30363d"/><text x="526" y="404" fill="#e6edf3" font-size="12" font-weight="600">ALD ↔ ALE symmetry</text><text x="526" y="420" fill="#8b949e" font-size="9.5">Mirror processes on the same reactor idea:</text><text x="526" y="433" fill="#8b949e" font-size="9.5">one builds atomic films, the other trims them.</text><text x="526" y="446" fill="#8b949e" font-size="9.5">Together they enable GAA and sub-nm nodes.</text></svg>
The throughput cost of ALD comes from repeated exposure and separation, and high-aspect-ratio structures make those steps longer. Cycle time includes precursor delivery, transport to the most remote surface, surface reaction, evacuation or inert purge, co-reactant delivery, and a second separation step. Spatial ALD, batch processing, multi-station tools, higher precursor utilization, and optimized pumping can improve productivity, but only if saturation, isolation, film properties, and defectivity remain intact. ALD earns its cost where thickness control, composition, interface quality, or conformality materially improves the completed device; it should not be selected from a generic claim that every advanced layer requires it.
Read ALD and ALE through a control-theory lens rather than a 'slow deposition/etch' lens: the whole point is to convert an analog, rate-×-time process — where thickness or depth is the integral of a reaction rate you can never perfectly know — into a digital, count-the-cycles process where the surface saturates and then refuses to change further. Self-limitation is what removes the dependence on flux, time, and geometry all at once, which is why the same idea, run forward or backward, delivers both the conformal films and the damage-free recesses that three-dimensional transistors are built from. The cost you pay for that determinism is cycle time, so the design question at every layer is whether atomic control is worth the throughput — and as devices go vertical, more and more often it is.
ALD Process-Control Atlas
graph TD
A["Define film, substrate, feature, and thermal budget"] --> B["Map precursor and co-reactant saturation"]
B --> C["Verify purge independence and exclude parasitic CVD"]
C --> D["Measure nucleation, growth per cycle, composition, and stress"]
D --> E["Challenge high-aspect-ratio dose and purge"]
E --> F{"Blanket, profile, electrical,<br/>and defect limits pass?"}
F -->|No| B
F -->|Yes| G["Challenge chamber history, source age, and maintenance"]
G --> H["Release control plan"]
Final Perspective
Read ALD through a surface-saturation, transport, purge-separation, nucleation, and film-property lens rather than a one-monolayer-per-cycle lens. Cycle count is a powerful thickness actuator only after saturation, purge independence, stable growth per cycle, representative feature coverage, and the required material properties have all been demonstrated.
Explore 500+ Semiconductor & AI Topics
From EUV lithography to CUDA optimization — search the full knowledge base or chat with our AI assistant.