A contact must reach its intended terminal while remaining insulated from the neighboring gate. Conventional placement relies strongly on the landing position and available spacing. Self-aligned integration also uses existing dielectric structures to constrain the opening. This changes the tolerance to geometric overlap; it does not make arbitrary misalignment safe.
Why is the first image not a contact yet?
Step 165 in the 14nm flow forms a SiN hardmask. It prepares a protected surface that later receives the opening pattern. Calling it a completed contact would confuse opposite roles: dielectric separates conductors, while the metal introduced later provides conduction. The hardmask is also only part of the integration scheme. Insulating structures beside the gate and the boundaries retained after etching remain essential.
Process checkpoint
Understand PECVD SiN Hardmask Deposition in context
Distinguish the hardmask from the developing contact opening, then inspect how CMP separates the filled contacts.
Process context for “Self-aligned vs conventional contacts: which boundary can overlap?”: 14nm FinFET · CONTACT_SC · Step 165
Self-aligned contacts: from protected boundaries to conductionWhat does the shallow-trench image teach?
Step 173 records the start of a trench in the upper dielectric. Follow it downward and distinguish that shallow opening from a complete path to a device terminal. The etch must remove material vertically while limiting lateral loss. Directional ion action and sidewall passivation influence that balance. If protective corners erode, an overlap that was intended to be tolerated can instead leave a weak insulating boundary between gate and contact.
Why does metal filling still need CMP?
Further etching, cleaning, interface preparation, and filling turn the cavity into a conductive contact. Filling also leaves excess metal above the dielectric. At step 194, CMP clears that material while retaining metal in the openings. Correctly located contacts can still be bridged if their overburden remains connected.
Read the images as three linked questions: where is protection established, where does an opening begin, and which conductors become separated? Self-alignment depends on the insulating boundary that survives this entire sequence, not simply on whether the final cavity contains metal.
Compare the strategies against the same questions
| Question | A contact relying mainly on placement | A self-aligned contact |
|---|---|---|
| How is the neighboring gate avoided? | Mask position and layout spacing provide the main margin | Retained dielectric boundaries further constrain the etch |
| What matters after a placement error? | Whether the opening reaches an unintended region | Whether the protective boundary remains intact after the offset |
| Can etch control be ignored? | No: the opening profile still affects the connection | No: selectivity and protection loss still matter |
The distinction is not the existence of error. It is whether the structure continues to prevent an unwanted connection when error occurs. SAC can change that tolerance without making every overlap safe. All three illustrations below come from the actual SAC sequence; they are not presented as a measured comparison with a conventional-contact process.
The SiN hardmask covers the existing structure, preparing a protected surface for pattern transfer before contact opening.
The trench has begun in the upper dielectric. Trace its shallow opening downward to distinguish it from a completed contact path.
After CMP, metal remains in contact regions and upper excess has been removed; further etch, clean, and fill steps occur between these views.
If a final structure bridges two conductors, separate protection loss from incomplete clearing of top metal. The first compromises a boundary near the device. The second leaves a conductor above openings that should be separate. Both can produce an electrical short, so the observed failure does not uniquely identify its process origin.
This is why a useful comparison follows the retained insulation and the removal sequence, rather than labeling one approach simply precise and the other self-correcting.
Sources
References
Study of the SiO2-to-Si3N4 etch selectivity mechanism in inductively coupled fluorocarbon plasmas and a comparison with the SiO2-to-Si mechanism
M. Schaepkens, T. Standaert, N. R. Rueger, Pgm Patrick Sebel, G. Oehrlein, J. M. Cook