HF Etching and Silanization: Evidence for the Role of Surface Hydroxyl Groups in Silicon Nitride Resonator Loss

Technical Summary: HF Etching and Silanization: Evidence for the Role of Surface Hydroxyl Groups in Silicon Nitride Resonator Loss

Problem Statement
Silicon nitride (SiNx) nanomechanical resonators are critical components in sensing, quantum technologies, and fundamental physics experiments due to their high mechanical quality factors ($Q$). However, as resonator thickness scales down to the nanometer range, performance becomes increasingly limited by surface-related dissipation. While strategies such as phononic crystal patterns and soft clamping have addressed clamping and bending losses, the intrinsic damping mechanisms bound to the surface remain a primary bottleneck. Previous literature has suggested various culprits for this loss, including free surface charges, defect states (two-level systems), and native oxide layers. The specific role of surface chemistry—particularly the presence of native oxides versus surface functional groups like hydroxyls—requires further elucidation to optimize resonator performance.

Methodology
The authors investigated low-stress, silicon-rich SiNx membranes with nominal tensile stresses of 100 MPa and 200 MPa. The study employed a comprehensive surface treatment strategy involving:

1. Hydrofluoric Acid (HF) Etching: To remove the native oxide layer.
2. Silanization: Using trimethylchlorosilane (TMCS) in gas phase (TMCS(g)) and liquid phase (TMCS(l)) to introduce hydrophobic Si−(CH$_3$)$_3$ termination.
3. Combined Treatments: HF etching followed by TMCS silanization, performed under both ambient air and inert nitrogen (N$_2$) conditions.

The samples underwent extensive characterization before and after treatment:

• Structural and Compositional Analysis: Transmission Electron Microscopy with Electron Energy Loss Spectroscopy (TEM-EELS), X-Ray Photoelectron Spectroscopy (XPS), Rutherford Backscattering (RBS), Elastic Recoil Detection Analysis (ERDA), and X-Ray Reflectivity (XRR) were used to determine layer thickness, elemental composition (N/Si ratios), and native oxide depth.
• Surface Properties: Contact Angle (CA) measurements and Atomic Force Microscopy (AFM) assessed surface energy and roughness.
• Mechanical Performance: Intrinsic quality factors ($Q_{int}$) were measured via laser-Doppler vibrometry in high vacuum. The $Q_{int}$ was derived by eliminating stress-induced dissipation dilution effects.
• Chemical Verification: Nanomechanical Photothermal FTIR (NEMS-FTIR) was utilized to detect chemical changes, specifically the removal of oxide layers and the presence of C-H vibrations from silanization.

Key Results

1. Surface Characterization: Preliminary analysis confirmed a native oxide/oxynitride surface layer of approximately 1–2 nm on the untreated SiNx membranes. The 100 MPa and 200 MPa samples exhibited distinct N/Si ratios and surface energies, with the 100 MPa sample showing higher surface free energy (lower contact angle).
2. Treatment Efficacy:
   • HF Etching: Effectively reduced the oxide thickness and introduced Si−F surface groups. However, it resulted in a modest $Q_{int}$ improvement of 20–25%.
   • TMCS Silanization: Treatments using TMCS (with or without prior HF etching) yielded the most significant enhancements, increasing $Q_{int}$ by up to 50% (specifically 45–50% for the 100 MPa sample and up to 45% for the 200 MPa sample).
   • Oxidation Stability: XPS and FTIR data indicated that silanization did not permanently inhibit surface oxidation; the O/Si ratio increased over time (up to 6 months) even in silanized samples, suggesting the oxide layer reforms or grows slowly.
3. Correlation Analysis:
   • Oxide Thickness: No direct correlation was found between oxide thickness and mechanical dissipation. HF etching thinned the oxide but yielded lower $Q$ gains than TMCS treatments, which left the oxide largely intact.
   • Surface Energy: Surface energy changes (indicated by contact angle) did not correlate with $Q_{int}$ improvements. HF etching increased surface energy, while TMCS decreased it, yet both improved $Q_{int}$.
   • Surface Hydroxyl Groups: The most consistent correlation was between the reduction of surface hydroxyl (−OH) groups and the increase in $Q_{int}$. HF etching partially removed −OH groups (replacing them with Si−F), while TMCS treatments chemically bound to and eliminated these groups (replacing them with Si−(CH$_3$)$_3$).

Significance and Claims
The paper claims that the primary driver of surface dissipation in low-stress SiNx resonators is not the native oxide layer itself, nor the surface energy or adsorbate uptake, but rather the presence of surface hydroxyl groups.

The authors conclude that:

• Surface hydroxyl groups are a key contributor to energy loss, likely through hydrogen bonding or dipole coupling to the oscillating strain field, or via interactions with water molecules.
• Chemical functionalization that removes or passivates these hydroxyl groups (via HF etching or, more effectively, TMCS silanization) substantially suppresses surface dissipation.
• The superior performance of TMCS treatments over HF etching alone suggests that complete elimination of hydroxyl groups is more effective than partial removal.

These findings provide a targeted pathway for improving the performance of SiNx resonators in high-sensitivity applications by focusing on surface chemistry modification rather than solely on geometric optimization or oxide removal.