Hydrofluoric Acid (HF): Selectivity Limits

HF can remove silicon dioxide faster than crystalline silicon in some oxide cleans, but this selectivity depends on the chemistry and stack. Fluoride-containing species break the existing silicon–oxygen network and form soluble products. An oxidizing mixture can also remove silicon, as shown in a study of HF and oxidant solutions; the response of silicon nitride, metals, and other films depends on the full chemistry and surface condition. HF-based steps include selected sacrificial-oxide strips and native-oxide cleans before interface-sensitive processes. Unlike wet etches that first oxidize a surface and then dissolve its oxide, HF can attack an existing oxide network directly [T1].

Where HF Acts in a Modern Flow

In a production flow, HF-based chemistry is one option when an oxide must be removed while neighboring materials are preserved. The selected stack determines whether it is suitable:

1. Sacrificial-oxide removal: an HF-based wet step can remove an oxide film when the neighboring materials tolerate the selected chemistry.
2. Oxide deglaze: a buffered oxide etch can remove residual hard-mask oxide while the underlying film and endpoint are monitored for loss.
3. Contrasting nitride strip: hot phosphoric acid removes silicon nitride preferentially to oxide. This is a different wet chemistry, useful for comparing which material each step targets.

Physics & Mechanism

Fluoride-containing species react with the Si–O network and allow silicon-containing reaction products to enter the liquid. The existing oxide can be removed without first growing a new oxide on the surface [T1]. Wet exposure generally reaches accessible surfaces from more than one direction; actual removal rates and selectivity depend on the solution, film composition, and transport near the surface. This makes stack compatibility and feature geometry part of the same process decision.

Process Principles

Process engineers tune HF chemistry along three axes, and each direction has a reason.

• Buffering vs. rate drift: buffered oxide etch (BOE) combines HF with ammonium fluoride to moderate changes in etchant activity and improve process control; mask compatibility still requires separate verification [T1].
• Dilution vs. material loss: dilution changes removal rates and the time available to control an oxide clean, but it does not by itself create a reliable stop on every underlying film.
• Temperature and wetting vs. uniformity: temperature and liquid access can change removal rates across a patterned surface. Their effects must be checked for the actual chemistry and geometry.

Challenges & Failure Modes

• Lateral removal: wet chemistry can reach oxide under a patterned opening and cause undercut. The extent depends on access, film properties, and time, so critical dimensions need their own verification.
• Neighboring-film sensitivity: a mask defect or an incompatible exposed material can turn a selective oxide clean into unwanted material loss.
• Residue and contamination: by-products and mobile ionic residues must be rinsed away completely; reliability of the underlying dielectric depends on rigorous deionized-water rinses and chemical cleanliness.
• Safety and handling: HF penetrates skin and attacks calcium in the body; fab handling depends on dedicated dispensing, neutralization, and monitoring — a real operational constraint, not a footnote.

From Principle to Production Flow

The 40nm BSI CMOS Image Sensor flow contains an oxide Wet Deglaze Etch followed by a SiN Strip. Their names distinguish the target materials but do not establish the actual chemistry: HF-based oxide removal and hot-phosphoric nitride removal are examples to compare, not recipes assigned to these steps. From the 40nm flow overview you can compare the two targets. For broader context, see wet etching and dilute HF.

Technology Node Evolution

As devices introduce more sensitive interfaces and three-dimensional features, oxide removal has to balance completeness against loss of neighboring films and changes to surface condition. HF-based cleans remain useful in selected stacks, but the chemistry, liquid access, and subsequent interface requirements must be evaluated together; no single HF clean applies across all nodes or materials.

Related Processes

HF-based oxide removal is one example of chemical selectivity. Dry etching adds directional control for patterned features; chemical mechanical planarization levels a surface and requires a separately chosen post-polish clean. Etch rate helps compare removal of target and neighboring films.

Future Outlook

Current research examines how fluoride-based surface reactions can be controlled at new material interfaces and how smaller, more precisely delivered wet cleans affect uniformity and waste. Each application still needs its own material-compatibility and endpoint evidence; the value of HF is selective chemistry under a defined set of conditions.