Etching and Deposition

A chip is a stack of many thin films, each doing a specific electrical or structural job. Deposition is how those films are formed with the thickness, uniformity, and composition each job requires.

• Conductors: metal films carry current — the contacts to devices and the wiring that links them.
• Insulators (dielectrics): films that electrically separate conductors and devices so signals do not short together.
• Semiconducting materials: semiconductor films (sometimes grown crystalline by epitaxy) that form or extend the active regions of devices.
• Barrier layers: thin films that block one material from diffusing into another — for example, keeping metal atoms out of the surrounding dielectric.
• Liners and adhesion layers: films that help the next material stick and deposit evenly on the surface below.
• Interconnect-related films: the combined stack of conductors, barriers, and dielectrics used to build multi-level wiring.

There is no single best deposition method. Each forms a film in a different way, with different strengths — so fabs pick the method that fits the material, the shape, and the thickness control a layer needs. (Specific material/process pairings are process-dependent and not listed here.)

• CVD (chemical vapour deposition) — What: a film grown from chemistry. How: precursor gases flow over the heated wafer and react at the surface, depositing a solid film (precursor gases → surface reaction → film formation). Materials: many conductors, insulators, and semiconductors depending on the chemistry. Why: good throughput and often good conformality. Limitations: needs suitable precursor chemistry, often elevated temperature, and byproducts to manage.
• PVD (physical vapour deposition) — What: a film formed by physically moving material. How: atoms are knocked off a solid source (sputtering) or boiled off it (evaporation) in vacuum and travel to the wafer. Materials: metals and some other materials. Why: simple, fast, high-purity films — common for metal layers. Limitations: largely line-of-sight, so it coats steep, narrow features poorly (limited conformality).
• ALD (atomic layer deposition) — What: film grown one atomic layer at a time. How: two precursors are pulsed alternately; each reaction is self-limiting, so exactly one layer forms per cycle (covered in its own deep dive). Materials: high-quality thin dielectrics and other films. Why: unmatched thickness control and conformality. Limitations: slow, because growth is cycle-by-cycle.
• Epitaxy — What: growth of a crystalline film aligned to the wafer's crystal. How: atoms arrive and arrange onto the existing lattice so the film continues the crystal structure. Materials: semiconductor layers. Why: gives device-quality single-crystal material. Limitations: demanding conditions (high purity, often high temperature) and constraints from the underlying crystal.

Chemical vapour deposition grows a film from gas-phase chemistry: precursor gases are delivered to a hot wafer, react at the surface, and leave a solid film behind while volatile byproducts are pumped away.

• Conformality: because the film grows from surface reactions rather than line-of-sight arrival, CVD can coat vertical walls and feature bottoms far better than PVD — valuable for 3D shapes.
• Uniformity: gas flow, temperature, and pressure must be balanced so the film grows at the same rate everywhere on the wafer.
• Temperature: reaction rate and film quality depend strongly on temperature; higher temperatures often improve quality but limit what underlying layers can tolerate.
• Pressure: chamber pressure affects how gases move and react; low-pressure and plasma-assisted variants exist to control rate, uniformity, and allowable temperature.
• Reaction chemistry: the choice of precursors sets which material forms, the byproducts produced, and the conditions required — the chemistry is central, not incidental.

Physical vapour deposition moves material physically, not chemically: atoms leave a solid source and travel through vacuum to condense on the wafer.

• Sputtering: energetic ions strike a target, knocking atoms loose; those atoms fly to the wafer and build up a film. It is the workhorse for many metal layers.
• Evaporation: the source material is heated until it vaporises, and the vapour condenses on the cooler wafer.
• Where PVD is useful: fast, high-purity metal films on relatively open surfaces — for example, blanket metal layers and seed layers.
• Limitations: PVD is largely line-of-sight, so it thins or leaves gaps on steep sidewalls and at the bottoms of narrow, deep features — poor conformality compared with CVD or ALD.

Atomic layer deposition earns its own section because it works differently from the others: it grows a film in discrete cycles, and each cycle adds at most one atomic layer.

• Self-limiting reactions: a first precursor is pulsed in and reacts with the surface until every available site is taken — then it stops on its own, because there is nothing left to react with. The chamber is purged, a second precursor is pulsed to complete the layer, and it too self-limits.
• Cycle-by-cycle growth: repeating this pulse–purge–pulse–purge cycle adds one atomic layer at a time, so final thickness is set simply by counting cycles.
• Conformality: because each step saturates every exposed surface equally — including deep, narrow walls — ALD coats complex 3D features with near-perfect evenness.
• Atomic-scale thickness control: growth per cycle is essentially fixed, giving sub-nanometre control that other methods cannot match.
• Why it is valuable for complex 3D structures: modern devices have extremely narrow, deep, and folded features; ALD's self-limiting, conformal growth is often the only way to line or fill them uniformly — at the cost of being slow.

Epitaxy is deposition that preserves crystal order: instead of an amorphous or polycrystalline film, it grows a single-crystal layer that continues the wafer's own lattice.

• Crystal growth: arriving atoms settle into positions that extend the existing crystal, so the new film shares the substrate's orderly lattice.
• Substrate relationship: the underlying crystal acts as a template; the film's structure (and any strain) depends on how well the two lattices match.
• Why crystal quality matters: defects and dislocations in the grown layer degrade device performance, so epitaxy demands very clean, controlled conditions.
• Device applications: epitaxial layers provide device-quality semiconductor material — for example, engineered channel or source/drain regions and high-quality starting layers for devices.

Beyond simply adding material, a production film has to meet many quality requirements at once. These are the properties process engineers actually tune.

• Conformality: how uniformly a film coats 3D topography — ideally the same thickness on the top, the sidewalls, and the bottom of a feature. It is the single biggest differentiator between deposition methods.
• Step coverage: a practical measure of conformality — the ratio of the film's thickness on a sidewall or feature bottom to its thickness on the flat top. Poor step coverage leaves thin spots or voids where a feature bends.
• Nucleation: film growth starts as atoms gather into initial clusters (nuclei) on the surface. How readily and evenly nuclei form decides how thin a film can be before it is continuous, and how rough it ends up.
• Growth mechanisms: films can build up island-by-island, layer-by-layer, or a mix. The mode affects density, roughness, and the minimum continuous thickness achievable.
• Film stress: every deposited film carries intrinsic and thermal stress (tensile or compressive). Too much stress bows the wafer, cracks the film, or peels it off, so stress is deliberately managed.
• Composition control: hitting the right stoichiometry and purity — small deviations shift electrical and mechanical behaviour, so composition is monitored as tightly as thickness.
• Interface quality: the boundary between two films (how abrupt and clean it is, whether an unwanted reaction occurred) strongly affects the device — the gate-dielectric-to-channel interface is a classic example.
• Wafer uniformity: consistency of thickness and composition across the whole wafer and wafer-to-wafer; non-uniformity means devices behave differently depending on where they sit.

A side-by-side comparison of the three most common thin-film methods. Read it as a map of trade-offs, not a ranking — each is the right tool for different layers.

Etching removes selected material from the wafer — normally the material left exposed by openings in a resist or hard mask. It splits into two broad families with different behaviour.

• Wet etching — a chemical reaction in a liquid dissolves and removes the target material (chemical reaction → material dissolution/removal). It is often highly selective and gentle, but usually isotropic (etches sideways as well as down), which limits how fine a feature it can hold. Selectivity, isotropy, surface compatibility, and bath control (concentration, temperature, time) are the levers. (No specific chemistries are given here.)
• Dry etching — a plasma of reactive species removes material, and can be made strongly directional. It combines chemistry (reactive species attack the material) with physics (ions accelerated toward the wafer strike it vertically), enabling anisotropic profiles with nearly vertical walls.
• Reactive-ion etching (RIE) is the classic example: a plasma supplies reactive species while an electric field drives ions downward, so the etch proceeds mostly straight down. Plasma etching more broadly spans more-chemical (less directional) to more-physical (more directional) regimes.
• Isotropic vs anisotropic: isotropic etches remove material equally in all directions (rounded, undercut profiles); anisotropic etches remove mainly downward (straight sidewalls). Fine, dense patterns need anisotropy, which is why dry etch dominates there.

Selectivity answers one question: how well can we remove the target material without damaging the other materials around it?

• Target material: the film you actually want to remove.
• Mask: the resist or hard mask on top that must survive the etch so the pattern is preserved.
• Underlying layer: the film beneath that the etch should stop on, not cut into.
• High selectivity means the etch removes the target much faster than it removes the mask or the underlayer — giving a wide margin to stop at the right place.

The shape an etch leaves behind — its profile — matters as much as how much it removes. Profile and depth-dependent effects decide whether a feature is usable.

• Vertical sidewalls: straight-down walls from a strongly anisotropic etch — needed for fine, densely packed features.
• Tapered sidewalls: sloped walls, sometimes deliberate (to help a later film fill or coat the feature).
• Undercut: unwanted sideways etching beneath the mask edge, which shrinks or distorts the feature — common with isotropic etching.
• Aspect ratio: depth divided by width; as it rises, reactants struggle to reach the bottom and byproducts struggle to leave, making deep narrow etches much harder.
• Microloading: etch rate depends on how much exposed area or how many features are nearby — dense regions can etch at a different rate than isolated ones.
• Aspect-ratio-dependent etching (ARDE): narrow, high-aspect-ratio features etch more slowly than wide ones, so features of different sizes reach different depths in the same time — a key challenge for 3D structures.

A production etch is a balance of chemistry and physics tuned to remove one material, in one direction, to one depth. These are the knobs and effects engineers work with.

• Anisotropy: how directional the etch is. It comes from balancing a chemical (isotropic) component against an ion-driven (vertical) component — more ion drive gives straighter walls.
• Selectivity: removing the target much faster than the mask or the underlayer, achieved mainly through the choice of chemistry.
• Etch rate: how fast material is removed. It must be uniform and repeatable across the wafer, not merely fast — a fast but uneven etch is useless.
• Plasma chemistry: the gas mix sets which reactive species form, what they attack, and whether byproducts are volatile enough to leave — it largely determines both selectivity and rate.
• Ion energy: the energy of ions hitting the surface. Higher energy increases directionality and physical removal, but risks surface damage and can lower selectivity.
• Charging effects: insulating surfaces and deep features can accumulate electric charge that deflects incoming ions, distorting profiles (notching or bowing) — a real limit at high aspect ratios.
• Aspect-ratio-dependent etching (ARDE): deep, narrow features etch more slowly because reactants struggle to reach the bottom and byproducts struggle to escape.
• Microloading: the local etch rate depends on how much exposed area is nearby, so dense and isolated regions can etch differently.
• Endpoint detection: sensing the moment a layer is cleared — often by watching the plasma's optical emission change — so the etch stops at the right depth instead of over-etching into the layer below.

Deposition and etch tools drift — chambers coat up, parts wear, gases vary. Fabs hold results steady by measuring, monitoring, and feeding corrections back, so the process is a controlled loop rather than a fixed recipe.

• Sensors: in-situ sensors track plasma conditions, gas flows, pressure, temperature, and RF power in real time during the process.
• Endpoint detection: optical-emission or interferometric signals detect the transition to a new layer, so etch or deposition stops at exactly the right point.
• Statistical process control (SPC): measured outputs are tracked on control charts so drift is caught and corrected before it produces out-of-spec wafers.
• Metrology: inline and offline measurement of thickness, critical dimension, profile, and defects supplies the data the control system acts on.
• Feedback (run-to-run / advanced process control): corrections computed from metrology adjust the settings of the next run to re-centre the process.
• Chamber monitoring: the chamber's own condition (wall deposits, part wear, and matching between supposedly identical chambers) is tracked, because the chamber state itself changes the result.

Lithography, deposition, etching, and metrology are not separate topics in practice — they are one repeating loop, and repeating it is what builds complex structures.

1. Lithography — where→
2. Deposition — add material→
3. Etching — remove material→
4. Metrology — measure & control→
5. Repeat → complex 3D structures

• Each pass of the loop builds one part of one layer: lithography places the pattern, deposition supplies the material, etching shapes it, and metrology confirms every pass is in spec.
• A chip is hundreds of these passes stacked and aligned — remove any one of the four and the structure cannot be built reliably.
• Repeating and combining the steps is precisely what creates 3D transistors (fins, then nanosheets) and multi-level interconnect — structures no single step could ever make on its own.
• As flat scaling slows, more progress comes from building vertically, which leans even harder on conformal deposition, high-aspect-ratio and selective etching, and tight metrology working together.

The topics below separate what is already in high-volume production from what is emerging or ramping — a distinction worth keeping clear.

• Atomic-scale process control: atomic layer etching (ALE) removes material in self-limiting cycles — the etch counterpart of ALD — for near-atomic precision. ALD is in production; ALE is emerging and expanding.
• High-aspect-ratio etching: pushing ever deeper, narrower, straighter etches (as in 3D memory) against ARDE, charging, and profile control — a continual engineering frontier.
• 3D transistor structures: FinFETs are in production; device architecture keeps evolving toward more vertical, more folded shapes.
• Gate-all-around (GAA) structures: nanosheet transistors ramping at the leading edge, which depend on extremely selective etch and deposition to remove sacrificial layers from between stacked sheets.
• Backside processing: routing power (and eventually more) on the wafer's back side to relieve front-side congestion — emerging and beginning to ramp.
• New materials: alternative channel, contact, and dielectric materials are under active research to extend device performance.
• Selective deposition (area-selective deposition): growing a film only where it is wanted, guided by surface chemistry, potentially removing patterning steps — emerging.
• Selective etching: removing one material while barely touching an adjacent one, at very high selectivity — some in production and a key enabler for GAA, being pushed much further.