For isotropic removal of single-crystal silicon in damage-layer etching, wafer thinning, and template release, the ternary acid mixture of hydrofluoric acid, nitric acid, and acetic acid—product designation Hydrofluoric-Nitric-Acetic Acid Etchant (HNA)—is supplied as a pre-blended etchant. The concentrated formulation contains 49% w/w HF, 70% w/w HNO₃, and glacial acetic acid in a 3:5:3 volumetric ratio; alternative volumetric ratios such as 1:3:8 and 2:7:1 are used for diffusion-limited and kinetically limited operating points. The as-received mixture is a clear, fuming liquid and is packaged in fluoropolymer containers to minimize headspace moisture uptake and HF loss. Product specification parameters include HF and HNO₃ assay by titration, trace metal content by inductively coupled plasma mass spectrometry, and particle count by light obscuration. The etch mechanism proceeds by nitric acid oxidation of elemental silicon to silicon dioxide, followed by hydrogen fluoride dissolution of the oxide as hexafluorosilicic acid; acetic acid acts as a solvent-moderator that lowers water activity and reduces the rate of nitric acid ionization.

Composition-Dependent Etch Rate and Transport Regimes in the HNA Ternary System

Etch rate is determined by the position of the formulation on the HF-HNO₃-CH₃COOH ternary composition diagram. In high-HNO₃ formulations, oxidation of silicon is rapid and fluoride transport to the surface becomes rate-limiting; the process is then sensitive to wafer rotation, bubble evolution, and bath viscosity. In high-HF formulations, oxide dissolution is rapid and the rate-limiting step is nitric acid oxidation, so agitation dependence is reduced and the surface tends to remain specular. The balanced 3:5:3 operating point is selected for bulk silicon removal rates in the typical range of 2 µm/min to 10 µm/min at 25°C. Published rate data for silicon span 0.2 µm/min to 20 µm/min at 20°C–25°C, depending on bath age, mask openings, and dopant concentration; heavily boron-doped silicon with concentrations above 5×10¹⁹ cm^-3 exhibits a significant etch rate reduction, a property exploited in p+ etch-stop applications. The apparent activation energy in diffusion-controlled compositions may fall below 20 kJ/mol, whereas kinetically controlled formulations may show higher values. Acetic acid suppresses dissociation of nitric acid by lowering the dielectric constant of the solution, shifting the system from a vigorous exothermic reaction toward a controllable isotropic removal process.

On production immersion tools, the primary sources of thickness non-uniformity are bath temperature drift, incomplete removal of nitrogen oxide bubbles, and progressive loss of HF through evaporation. Wet benches constructed from PVDF or PTFE/PFA are used for batch processing; the product is incompatible with glass, borosilicate, and stainless steel, each of which is attacked by free fluoride and nitric acid. Recirculating filtration through 0.2 µm PTFE membrane cartridges maintains particulate control, while nitrogen sparging or external loop recirculation is applied to disrupt bubble stagnation within tight-pitch cassettes. Equipment safety is governed by SEMI S2 and applicable local codes; all exhaust ducting is scrubbed for NOx and HF acid mist. Bath temperature is maintained within ±0.5°C to limit etch-rate drift under diffusion-limited transport. Production lots with cassette center-to-edge removal differences are corrected by adjusting sparger flow, wafer rotation, or rack position rather than reformulating the acid ratio. A reduction of monitor etch rate by more than 20% relative to fresh bath is commonly used as a replenishment trigger because it corresponds to observable surface roughening; published data for exact replenishment thresholds outside a given tool set are limited. Because dilution can be exothermic, the acid mixture is added to deionized water, not the reverse, and the product is not diluted on the production floor unless the resulting etch-rate shift has been qualified.

Which Lot-to-Lot Metrology Inputs Control the Stop Point for Wet Silicon Thinning?

Endpoint control for HNA thinning is time-based, in contrast to plasma etching in which optical emission or interferometry can provide a real-time signal. The two metrology inputs with the strongest influence are incoming wafer thickness distribution and the measured etch rate of the pre-loaded monitor wafer. Incoming wafer thickness is surveyed by capacitance gauge on a defined sampling plan; after etch, thickness is mapped by Fizeau interferometry or infrared interferometry to calculate total thickness variation and within-wafer non-uniformity. Process engineers establish bath etch rate by etching a monitor wafer of identical crystal orientation and doping under the same rack position and agitation conditions. When the monitor rate falls outside the qualified control limit, the bath is spiked with fresh oxidizer or replaced, rather than extending immersion time. Cassette position effects are significant; edge wafers can show different removal than center wafers due to concentration gradients. A common corrective action is load-slotting with dummy wafers or rotating the cassette halfway through the immersion cycle.

After backgrinding of 200 mm and 300 mm silicon wafers, the residual sub-surface damage layer typically ranges from 5 µm to 15 µm, depending on diamond grit size and spindle parameters. The product is used in a stress-relief etch step to remove this layer before polishing or device processing, with typical removal budgets of 10 µm to 20 µm per side. The etch is isotropic, so lateral removal under edge protection or hard mask openings equals the vertical removal, a key distinction from anisotropic alkaline chemistries. For MEMS release, the same formulation is employed where an isotropic cavity or through-wafer vent is required and sidewall angle is not a critical dimension. Wafer processing is conducted in an ISO 14644-1 Class 5 cleanroom to limit particle adders during transfer. The process sequence is immersion etch, rapid dump rinse, overflow rinse, and spin dry; incomplete rinsing leaves fluoride residue that attacks aluminum pads in subsequent processing. Rinsing to 10–18 MΩ·cm deionized water and drying under filtered air or nitrogen is specified to prevent water spots and residual fluoride corrosion.

When a Low-Temperature Hard Mask Must Withstand Fluoride Attack During Isotropic Silicon Removal

Because the mixture contains free HF, thermal silicon dioxide is not a robust hard mask; it is etched concurrently with the silicon and requires a thickness budget that accounts for both the silicon removal target and the oxide loss. Silicon nitride and silicon carbide hard masks exhibit longer life in HNA, although selectivity is composition- and temperature-dependent. Photoresist masking is generally restricted to thin, hardened resist films and short etches; resist lifting, adhesion loss, and undercutting occur when acid attack at the resist-silicon interface proceeds faster than the vertical silicon etch. For processes requiring precise control of the lateral etch at mask edges, a thin oxide or nitride mask with known undercut bias is qualified by SEM cross-section. Published selectivity ratios for specific photoresist systems are limited and must be generated on the production line; the product is not used where vertical sidewalls are required.

Comparing HNA Against Alkaline Hydroxide and Fluorine-Based Dry Etch Processes

HNA is therefore selected when the process requires high silicon removal rates, mask-defined isotropic release, or removal of subsurface damage across a full wafer batch; alkaline chemistries are selected when crystallographic anisotropic sidewalls are required; dry etching is selected when aspect ratio and critical dimension control dominate. Unlike dilute HF or buffered oxide etchants, which remove silicon dioxide but do not attack elemental silicon at practical rates, HNA combines an oxidizer and a fluoride source, enabling direct silicon removal. Unlike HF-HNO₃ mixtures without acetic acid, the ternary product shows reduced sensitivity to minor temperature fluctuations and improved wetting uniformity across open wafer surfaces. The values in the comparison matrix are typical published operating windows; exact rates and selectivities are tool-, mask-, and crystallographic-orientation-dependent and are verified by coupon tests before process release. The tradeoff is not additive: HNA generates liquid waste containing hexafluorosilicic acid and unreacted HNO₃, requiring neutralization and fluoride precipitation, whereas dry processes generate gaseous effluents scrubbed at the vacuum pump.

Hydrogen fluoride exposure causes severe tissue injury and systemic fluoride toxicity; therefore all handling is performed under local exhaust with acid-resistant gloves, face shield, and protective apron. Calcium gluconate gel is staged at point of use, and emergency response follows ANSI Z358.1 station placement. The product must never be mixed with ammonium hydroxide, hydrogen peroxide, or organic amines; exothermic neutralization and rapid generation of nitrogen oxides may exceed scrubber capacity. Incompatible materials include glass, borosilicate, aluminum, cast iron, and unprotected carbon steel. Waste treatment includes neutralization with calcium hydroxide or calcium chloride followed by precipitation of calcium fluoride and pH adjustment to discharge permit limits. Fluoride concentration in treated effluent is verified by ion-selective electrode per ASTM D1179. Occupational exposure limits for the individual acids are controlled under 29 CFR 1910.1000 and the supplier safety data sheet; the ACGIH TLV-TWA for hydrogen fluoride is 0.5 ppm. Published data for exact etch uniformity after long-term bath aging are limited; a production bath is therefore revalidated by monitor wafer etching after each replenishment event or at shift changes.