High aspect ratio (HAR) processing addresses features whose depth or height greatly exceeds their width. Etching must control the profile and sidewalls of a deep feature, while deposition and filling must manage transport and surface reactions. Through-silicon vias are one three-dimensional integration example: their fabrication includes deep via etching, dielectric and barrier deposition, copper filling, and later planarization . Etch profile and fill integrity are distinct questions that must be assessed for the material stack at hand.
The Four Stages of a Deep Trench's Life
In the 40nm image sensor's deep trench isolation, a trench's production life compresses into four stages — open, etch, fill, planarize — the two HAR pillars handing off to each other.
- Hard-mask opening: an anisotropic oxide hard-mask etch defines where the trench will go.
- Deep silicon etch: reactive ion etching drives the trench vertically into the substrate — several times deeper than wide, sidewalls standing straight.
- Gap fill: liner and bulk dielectric deposition aim to fill the narrow trench; coverage and voids require separate checks.
- Planarization: CMP grinds off the excess and reveals an isolation structure flush with the substrate.
Real cross-sections from the free 40nm flow — the same steps this article's learning route links to.
Process checkpoint
Understand Oxide Hard Mask Etch in context
Stage 1 — the hard-mask opening that defines where the deep trench will go.
Account access is confirmed on the step page.
Physics & Mechanism
Knudsen Diffusion: The Transport Tax at Depth
Inside a HAR trench, the gas mean free path often exceeds the feature width — molecules hit the sidewalls far more often than each other, and transport is governed by Knudsen diffusion. Reactant delivery to the bottom is limited by molecular conductance, and reactant depletion can make conditions at the bottom differ from those near the entrance. A useful assessment compares transport, surface consumption, and by-product removal; neither an aspect-ratio label nor one kinetic regime guarantees a uniform result.
Sticking Coefficients: Stick or Bounce
The sticking coefficient describes the fraction of incident species that adsorb or react at a surface. In the textbook's deposition model, a lower value allows more re-emission and transport into a recessed feature; TEOS-based oxide CVD is given as an example with generally lower sticking than silane-based oxide CVD . This can improve sidewall and bottom coverage under suitable transport and reaction conditions, but the coefficient alone does not guarantee void-free filling.
Ion Directionality: IADF and Passivation
Anisotropic deep etching relies on the plasma sheath's electric field accelerating ions vertically; the narrower the ion angular distribution function (IADF), the more vertical the attack. Sheath collisions broaden the IADF, and oblique ions strike the upper sidewalls, lose energy, and undercut. Some deep-silicon processes alternate etching with polymer passivation, as in the Bosch process. Other chemistries use different sidewall-protection mechanisms. The relevant question is how directional bottom removal is maintained while sidewalls remain protected; no single sequence applies to every deep feature.
Process Principles
The Deposition Side
- Reactor pressure: pressure affects gas density, transport, and reaction conditions. Its effect on coverage must be evaluated with chemistry, temperature, and feature geometry; neither higher nor lower pressure is a universal improvement.
- Precursor chemistry: changes in reactant composition can alter gas-phase and surface reactions. Coverage, composition, and deposition rate must be assessed together; the direction of change is not universal.
- Substrate temperature: temperature changes adsorption, reaction, desorption, and material evolution. Faster local growth can consume reactants near the entrance, while other changes can improve coverage. The resulting profile depends on the chemistry and structure.
The Etch Side
- RF bias power: changing the sheath potential affects ion energy and bottom-surface interactions. It can aid passivation removal, but profile, selectivity, and damage depend on the full plasma and surface conditions; higher bias does not by itself certify better directionality.
- ICP pressure: pressure in an inductively coupled plasma reactor affects collisions and species delivery. Profile and etch rate must be assessed together rather than inferred from pressure alone.
- Passivation and etching: sidewall protection and bottom removal must remain compatible. The balance depends on the chemistry, surfaces, and ion conditions; a gas ratio alone does not establish either protection or successful etching.
Challenges & Failure Modes
- Voids and pinch-off: growth near an entrance can close the opening before the interior is filled. Coverage, profile, and closure must be assessed separately; a buried void can create problems during later processing.
- Aspect-ratio-dependent etching (ARDE, RIE lag): delivery of reactants and removal of by-products can become more difficult in a narrow, deep feature. Together with surface and ion effects, this can change the etch rate or profile. Slower narrow-feature etching is one possible result, not a rule that identifies the mechanism by itself.
- Microtrenching and profile distortion: reflected ions can concentrate removal near bottom edges, while mask evolution can affect the final profile. Their importance depends on the structure and process conditions.
- Charging effects: unequal arrival of electrons and ions can produce local electric fields within a feature. These fields can change ion trajectories and affect the profile or continued etching.
From Principle to Production Flow
The four stages appear in production as adjacent, named steps. The deep trench isolation module of the 40nm BSI CMOS Image Sensor flow: an oxide hard-mask etch opens the window, an anisotropic silicon trench etch drives the trench vertically into the substrate, and nitride fill plus CMP close the module. From the 40nm flow overview you can walk the whole chain with real cross-sections — the public overview provides the sequence and cross-sections, while each teaching page indicates its available reading access.
Technology Node Evolution
| Flow context | HAR questions to investigate |
|---|---|
| 28nm planar | How do isolation fill and contact etch meet their different profile and defect requirements? |
| 14nm FinFET | How do fin isolation and replacement-gate cavities change access to surfaces and the risk of premature closure? |
| 7nm | How do the particular materials and dimensions affect coverage, damage, and integration margins? |
These flow contexts organize learning; they do not assign one deposition or etch method to a node. Atomic layer deposition, flowable deposition, selective growth, and other integration strategies address different coverage or filling tasks. Conformal coating alone does not guarantee void-free gap filling.
Related Processes
Surface preparation, filling, and planarization must be compatible with the particular stack. Wet cleaning, including suitable hydrofluoric-acid processes, can remove selected residues or oxides; material compatibility and drying-related collapse remain separate checks. A subsequent anneal may change film structure or stress, but it does not establish that a buried void has disappeared. CMP removes excess material, with dishing and erosion assessed against pattern density and material selectivity. TEOS, undoped silicate glass, and chemical vapor deposition provide related material and process context.
Future Outlook
Future three-dimensional integration can increase the difficulty of delivering reactants, removing by-products, and maintaining useful sidewalls. New deposition and etch approaches must be assessed against the particular material stack, defect modes, environmental impact, and downstream integration. A technique name or smaller dimension does not establish void-free filling, complete sidewall protection, or electrical reliability.
References
Study of the protrusion of through-silicon vias in dual annealing-CMP processes for 3D integration
Tianjian Liu, Shizhao Wang, Fang Dong, Yang Xi, Yunpeng Zhang, Tao He et al. · Microsystems & Nanoengineering
Silicon VLSI Technology - Full
James D. Plummer, Michael D. Deal, Peter B. Griffin
Silicon VLSI Technology · ISBN 978-0130850379