Introduction
In advanced semiconductor manufacturing, achieving pristine, atomic-scale cleanliness on silicon surfaces is a fundamental prerequisite for device reliability and performance . Among the array of chemical formulations utilized in modern fabs, dilute hydrofluoric acid (DHF) stands out as one of the most critical reagents . DHF is predominantly employed in front end of line (FEOL) cleaning sequences to remove sacrificial oxides, strip native oxide layers, and passivate exposed silicon surfaces , .
As integration densities increase and device architectures transition from planar transistors to complex three-dimensional structures, managing surface contamination and atomic-scale defects becomes highly challenging . Native silicon dioxide ($SiO_2$), which spontaneously forms upon exposing bare silicon to atmospheric oxygen, represents a major obstacle; it introduces unstable electronic interface states and increases contact resistance . DHF solves this by selectively dissolving these oxide layers while leaving the underlying silicon substrate intact , .
The utility of dilute HF extends far beyond simple oxide removal . It plays a dual role in modern wet clean strategies by acting as both an etchant and a surface-conditioning agent , . By understanding the thermodynamic and kinetic properties of DHF, process engineers can tune surface hydrophobicity, control interfacial layer thickness, and prevent cross-contamination during subsequent high-temperature steps , .
Physics & Mechanism
The chemical behavior of DHF is governed by complex equilibria in aqueous solutions (Engineering Practice). Unlike strong mineral acids, hydrofluoric acid behaves as a weak acid in dilute aqueous solutions, meaning it does not fully dissociate into hydrogen ($H^+$) and fluoride ($F^-$) ions . Instead, a series of thermodynamic associations occur, giving rise to multiple reactive fluorine-bearing species .
Dissociation and Speciation Chemistry
In dilute solutions, the primary dissociation step is expressed as:
$$HF \rightleftharpoons H^+ + F^-$$
At higher concentrations, hydrogen fluoride molecules associate with free fluoride ions to form the bifluoride ion ($HF_2^-$) via homoconjugation :
$$HF + F^- \rightleftharpoons HF_2^-$$
Other higher-order complexes, such as $H_2F_3^-$, can also form as concentration levels vary, but $HF$ and $HF_2^-$ remain the dominant active species in typical dilute HF formulations . The concentration ratios of these species are highly sensitive to the overall solution $pH$ and dilution ratio , .
Reaction Kinetics with Silicon Dioxide
The dissolution of $SiO_2$ in DHF is not a single-step reaction but a coordinated chemical process involving the breaking of silicon-oxygen ($Si-O$) bonds and the formation of highly stable silicon-fluorine ($Si-F$) bonds , . The overall chemical reaction is represented as , :
$$SiO_2 + 6HF \rightarrow H_2SiF_6 + 2H_2O$$
The reaction yields hexafluorosilicic acid ($H_2SiF_6$), which is highly soluble in water and easily rinsed away from the wafer surface , .
On a microscopic level, both molecular $HF$ and the bifluoride complex $HF_2^-$ act as active etchants . Kinetic studies indicate that $HF_2^-$ attacks the oxide surface significantly faster than molecular $HF$ alone . The silicon dioxide matrix is dissolved through nucleophilic attack: the electronegative fluorine species polarize and weaken the $Si-O$ bonds on the surface, facilitating protonation of the oxygen atoms by $H^+$ and subsequent release of water molecules , . This sequence continues until the silicon atom is fully coordinated by six fluorine atoms, forming the soluble $[SiF_6]^{2-}$ complex , .
The etch rate is strongly dependent on the structural properties of the oxide being etched . Thermally grown oxides, which feature highly dense and ordered $Si-O-Si$ networks, exhibit the slowest etch rates in DHF , . Conversely, low-temperature deposited oxides, such as undoped silicate glass (USG) or oxides formed via tetraethyl orthosilicate (TEOS) precursor chemical vapor deposition, have higher porosity, lower density, and varying impurity concentrations (e .g., hydrogen, boron, or phosphorus) . These structural differences lower the activation energy of the etch reaction, leading to significantly accelerated etch rates compared to thermal $SiO_2$ , .
DHF Solution (HF, HF2-, H+)
===========================
│ │ │ (Nucleophilic attack by fluorine species)
▼ ▼ ▼
O ── Si ── O ── Si ── O (Oxide Surface)
───────────────────────────
Si ── Si ── Si (Silicon Substrate)
Surface Passivation and Hydrophobicity
A key attribute of the DHF treatment is the resulting state of the silicon surface . While one might expect the bare silicon surface to remain covered with highly polar $Si-F$ bonds due to the high electronegativity of fluorine, the final surface is actually dominated by hydrogen termination ($Si-H$ bonds) , (Engineering Practice).
This phenomenon is explained by the polarization of the $Si-Si$ backbonds . When a $Si-F$ bond is formed on the surface, the strong electron-withdrawing nature of the fluorine atom polarizes the underlying $Si-Si$ bond, making the subsurface silicon atom susceptible to nucleophilic attack by the remaining hydrogen ions in the solution . Consequently, the fluorine is removed as volatile or soluble species, and the silicon surface is left terminated with stable, non-polar covalent $Si-H$, $Si-H_2$, or $Si-H_3$ groups , (Engineering Practice). This hydrogen passivation yields a highly hydrophobic surface that resists rapid oxidation in cleanroom air, lowering the interface state density and preventing immediate re-contamination before subsequent gate dielectric deposition .
Process Principles
Optimizing the DHF process requires a precise understanding of how key process parameters directionally affect etching performance, selectivity, and surface morphology .
Concentration and pH Tuning
The concentration of DHF directly dictates the availability of active etchant species . Increasing the HF concentration increases the etch rate non-linearly . This non-linear relationship is due to the shift in chemical equilibrium toward the more reactive $HF_2^-$ and higher-order fluoroflux complexes as concentration rises .
To control this reaction rate and maintain high process control, ammonium fluoride ($NH_4F$) can be added to form buffered hydrofluoric acid (BHF), also known as buffered oxide etch (BOE) , , . The addition of $NH_4F$ acts as a pH buffer, stabilizing the concentration of both $HF$ and $HF_2^-$ as fluorine ions are consumed during the etching process , :
$$NH_4F \rightleftharpoons NH_4^+ + F^-$$
This buffering action ensures a constant etch rate over time and prevents local depletion of etchants in high-aspect-ratio features, which is essential for uniform pattern transfer , , . Additionally, buffering the solution to a higher pH minimizes the risk of photoresist lifting and peeling during prolonged wet processing , .
Temperature and Activation Energy
The temperature of the DHF bath is another critical lever , . Wet chemical etching of $SiO_2$ obeys the Arrhenius rate law, where the reaction rate constant is proportional to $\exp(-E_a/kT)$ . Raising the process temperature increases the thermal energy of the system, accelerating the surface reaction kinetics .
However, running at elevated temperatures can degrade process uniformity if the reaction transitions from a reaction-rate-limited regime to a mass-transport-limited regime , . Thus, temperature must be carefully balanced to maintain a controllable, uniform etch across the entire wafer surface (Engineering Practice).
Mass Transport, Agitation, and Surfactants
In deep, high-aspect-ratio structures typical of modern memory and logic devices, the rate-limiting step of the etching process often shifts from surface chemical reaction kinetics to physical diffusion , . As DHF reacts with the oxide deep within a trench or contact hole, the local concentration of active reactants decreases, while the concentration of the byproduct $H_2SiF_6$ increases .
To sustain the etch rate inside confined geometries:
- Agitation and Fluid Dynamics: Implementing active fluid flow or megasonic agitation helps replenish reactants at the wafer surface, though deep features remain heavily reliant on diffusion , .
- Standby and Liquid Film Maintenance: Advanced single-wafer cleaning systems utilize specialized standby liquid delivery to maintain a stable boundary layer, preventing wafer drying and ensuring uniform chemical delivery during transfer steps .
- Surfactants: The high surface tension of pure aqueous DHF can prevent the liquid from fully wetting hydrophobic, high-aspect-ratio features . Introducing specialized surfactants to the DHF solution reduces the surface tension, improving wetting behavior and enhancing etch uniformity across dense macro- and micro-structures .
Challenges & Failure Modes
While DHF is a highly effective processing agent, its chemical and physical interactions present several failure modes that must be carefully managed (Engineering Practice).
STUCTION FAILURE GALVANIC CORROSION
Capillary Force DHF Solution
│ │ │
▼ ▼ ▼
┌──┐ ┌──┐ ┌──────┐
│ │ │ │ │Metal │ ── (Anodic dissolution)
│ │ │ │ └──────┘
_│ └_ _│ └_ _│ └_
///////// ///////// /////////
Silicon Substrate Silicon Substrate
Capillary-Force Pulldown and Stiction
In microelectromechanical systems (MEMS) and advanced multi-gate transistors (such as FinFETs and nanosheets), sacrificial oxides are etched away to release suspended or high-aspect-ratio silicon features , . During the subsequent rinsing and drying steps, the liquid-vapor interface of the evaporating water or rinse solution exerts strong capillary forces on the neighboring structures , (Engineering Practice).
If the mechanical stiffness of the structures is insufficient to overcome these capillary forces, the features will bend and adhere to one another or to the substrate, a failure mode known as stiction or capillary-force pulldown , . Once stiction occurs, strong electrostatic and van der Waals forces keep the structures locked in place, leading to permanent device failure , (Engineering Practice).
Galvanic and Metal Corrosion
Another major challenge in multi-material integration is the compatibility of DHF with metals and silicides , . In advanced nodes, DHF often comes into contact with metals used for contact liners or interconnects, such as aluminum, cobalt, or ruthenium .
Standard DHF solutions exhibit poor selectivity against metals like aluminum, leading to rapid chemical attack and corrosion . Furthermore, if a metal and a semiconductor are simultaneously exposed to DHF, a galvanic cell can form , (Engineering Practice). The difference in electrochemical potential drives electron transfer, resulting in rapid anodic oxidation and subsequent dissolution of either the metal or the silicon at the interface, destroying contact structures and causing electrical opens or high leakage currents , (Engineering Practice).
Particle Re-deposition and Watermarks
During the DHF etch, the silicon surface transitions from hydrophilic (oxide-terminated) to highly hydrophobic (hydrogen-terminated) , (Engineering Practice). While this hydrophobic state is desirable for minimizing native oxide growth, it presents distinct processing risks .
Hydrophobic surfaces have a high affinity for airborne organic contaminants and suspended particles in the rinse bath (Engineering Practice). If the rinsing and drying processes are not perfectly optimized, micro-droplets of water can become pinned on the hydrophobic silicon surface , . As these water droplets evaporate, dissolved oxygen in the liquid causes localized oxidation of the silicon, while dissolved silica and other trace impurities precipitate out, forming localized ring-like residue defects known as watermarks , (Engineering Practice).
Technology Node Evolution
The role and implementation of DHF have undergone significant refinement as the semiconductor industry transitioned from planar devices to advanced three-dimensional architectures .
28nm Planar Node: Interfacial Layer Engineering
At the 28nm planar flow node, the introduction of high-k/metal gate (HKMG) technology fundamentally altered the requirements for surface preparation prior to gate oxide deposition . In these gate-first or gate-last integration schemes, controlling the thickness of the interfacial layer (IFL) between the silicon channel and the high-k dielectric (typically $HfO_2$) was critical to scaling the inversion equivalent thickness ($T_{inv}$) .
A standard DHF-last clean removes the native oxide completely, yielding a hydrogen-terminated surface . While this enables the thinnest possible physical stack and low gate leakage current ($J_g$), direct deposition of high-k materials on a DHF-passivated surface often results in a high density of interface states, leading to severe electron mobility degradation .
To resolve this, engineers developed ozone-water-last treatments . Ozonated deionized water ($O_3/DIW$) provides a controlled oxidative environment that grows an ultra-thin, highly dense, and uniform chemical oxide layer , . This dense chemical IFL acts as a robust barrier against hafnium ($Hf$) diffusion into the silicon substrate during subsequent high-temperature annealing steps, maintaining bulk electron mobility while still permitting aggressive $T_{inv}$ scaling .
| Treatment | Interfacial Layer (IFL) Quality | Inversion Thickness ($T_{inv}$) | Electron Mobility | Gate Leakage ($J_g$) |
|---|---|---|---|---|
| DHF-Last | Bare, H-terminated (no oxide) | Minimal | Degraded (high interface states) | Low |
| SC2-Last | Thicker, low-density oxide | Large | Good | High |
| Ozone-Water-Last | Ultra-thin, high-density oxide | Scaled/Optimized | Excellent | Minimal |
14nm to 7nm Nodes: Three-Dimensional Confinement
The transition to 14nm FinFET and 7nm FinFET architectures introduced severe structural constraints. With the channel wrapped on three sides by the gate, wet cleaning processes had to access narrow, high-aspect-ratio fin trenches without causing structural damage (Engineering Practice).
At these nodes, using standard DHF concentrations posed a major risk of over-etching the shallow trench isolation (STI) oxide surrounding the base of the fins (Engineering Practice). Excessive STI recessing leads to sub-threshold leakage path formation and parasitic capacitance variations (Engineering Practice).
To address this, fabs shifted to ultra-dilute DHF mixtures operated at highly controlled temperatures (Engineering Practice). This approach slowed down the etching kinetics, providing the necessary process window to strip native oxides from the fin sidewalls while minimizing damage to the surrounding isolation oxides and gate spacers .
Related Processes
DHF wet cleans do not operate in isolation; they are deeply integrated with adjacent manufacturing steps (Engineering Practice).
┌────────────────────────┐ ┌────────────────────────┐ ┌────────────────────────┐
│ Photolithography / │ ───> │ DHF Clean / │ ───> │ High-k Dielectric │
│ Patterning Step │ │ Oxide Strip │ │ or Metal Depo │
└────────────────────────┘ └────────────────────────┘ └────────────────────────┘
Photolithography and Patterning
Before a DHF clean is applied, wafer surfaces are patterned using photolithography to expose specific regions for etching . The choice of etchant dilution and chemical makeup is heavily influenced by the masking material . Concentrated HF tends to penetrate and peel off standard photoresists , .
Using dilute DHF or buffered HF minimizes photoresist lifting while maintaining targeted oxide removal rates , . Furthermore, anti-reflective coatings, such as bottom anti-reflective coating (BARC), must be chemically compatible with or completely cleared prior to DHF exposure to prevent chemical shielding and subsequent non-uniform etching (Engineering Practice).
Gate Dielectric and Metal Deposition
As discussed, the primary goal of DHF pre-cleans in HKMG stacks is to prepare the silicon surface for atomic layer deposition (ALD) . The chemical state of the surface directly influences the initial nucleation and incubation delay of the ALD precursors (Engineering Practice). Improperly prepared surfaces can lead to island-like growth, high defect densities, and local variations in the work function of the metal gate stack (Engineering Practice).
Silicide Formation (Salicide)
Prior to the self-aligned silicide (salicide) process, where a metal film (such as nickel-platinum) is reacted with silicon to form low-resistance contacts, a DHF dip is mandatory to clear any native oxide from the source, drain, and gate contact areas .
Any remaining oxide acts as a physical diffusion barrier, preventing the metal-silicon reaction and causing catastrophic increases in contact resistance or complete contact opens . In regions where silicide formation must be blocked, a salicide block layer (typically silicon dioxide or silicon oxynitride) is deposited to prevent metal-silicon interaction, requiring high selectivity from the DHF pre-clean chemistry .
Future Outlook
As the semiconductor industry advances beyond the 3nm node toward Gate-All-Around (GAA) nanosheets and complementary FET (CFET) architectures, the demands on oxide etching and surface preparation will become even more stringent .
Selectivity in Nanosheet Release
In GAA nanosheet integration, alternating layers of silicon ($Si$) and silicon-germanium ($SiGe$) are epitaxially grown (Engineering Practice). The $SiGe$ layers act as sacrificial materials that must be selectively etched away to release the $Si$ nanosheets, which then form the transistor channels .
While DHF is not directly used to etch $SiGe$, it is extensively used to remove protective sacrificial oxides between the nanosheets without attacking the adjacent structural materials . This requires ultra-high selectivity relative to low-k spacer materials, such as silicon carbonitride, which must remain completely intact to prevent parasitic capacitance increases and leakage pathways .
Transition to Vapor-Phase HF (VHF)
To completely eliminate the risk of stiction and capillary-force pulldown in ultra-dense, sub-3nm architectures, there is a strong industry push to transition from wet DHF processes to anhydrous vapor-phase HF (VHF) etching , (Engineering Practice).
VHF etching utilizes a mixture of gaseous HF and a catalyst vapor (such as methanol or water vapor) to react with silicon dioxide, producing volatile byproducts that are carried away in the gas stream , (Engineering Practice). Because the process is entirely dry, it eliminates the liquid-vapor interface and the associated capillary forces, representing a critical pathway for the structural release of nanosheets and advanced MEMS devices , (Engineering Practice).
References
- "REVIEW ARTICLE: Silicon dioxide sacrificial layer etching in surface micromachining" (1997, PAPER).
- "Tinv Scaling and Gate Leakage Reduction for n-Type Metal Oxide Semiconductor Field Effect Transistor with HfSix/HfO2 Gate Stack by Interfacial Layer Formation Using Ozone–Water-Last Treatment" (2008, PAPER).
- "Etch rates for micromachining processing" (1996, PAPER).
- Plummer, J. D., Deal, B. E., & Griffin, P. B., Silicon VLSI Technology: Fundamentals, Practice, and Modeling (2000, TEXTBOOK) .
- Sze, S. M., & Ng, K. K., Physics of Semiconductor Devices (2006, TEXTBOOK) .
- Hu, C., Modern Semiconductor Devices for Integrated Circuits - PN and Metal-Semiconductor Junctions (2010, TEXTBOOK) .
- US-2025323083-A1, "Substrate carrying unit and substrate processing device having the same" (2024, PATENT).
- US-2025210391-A1, "Substrate processing system and substrate processing method" (2023, PATENT).