Mode-dependent photothermal responsivity mapping of tensile-stressed nanomechanical resonators

Technical Summary: Mode-Dependent Photothermal Responsivity Mapping of Tensile-Stressed Nanomechanical Resonators

Problem Statement
Nanomechanical photothermal sensing relies on the principle that absorbed optical power raises a resonator's temperature, relaxing its tensile stress and shifting its resonance frequency. While the relative power responsivity (fractional frequency shift per absorbed power) is a central design quantity, existing understanding has been limited. Previous work established the mean temperature framework (MTF) for fundamental modes and demonstrated mode-dependence via simulations or specific defect-mode enhancements. However, a closed-form theory describing the combined position- and mode-dependence of the photothermal responsivity, validated by mode-by-mode experimental mapping, has been missing. This gap is critical for emerging applications such as dual-mode thermal sensing, where a mode pair with differential response to localized heating but common response to global drifts is required to suppress noise.

Methodology
The authors developed a closed-form theoretical model for square membranes and validated it through high-resolution scanning photothermal measurements.

• Theoretical Framework: The resonator is modeled as a square membrane of side length $L$ and thickness $h$ under biaxial tensile prestress $\sigma_0$. The core physical insight is a "fiber picture": the membrane behaves as a dense grid of independent tensile fibers along the $x$ and $y$ axes. In mechanical equilibrium, the tension change of each fiber is determined by its mean temperature, not its local temperature.

  • Using first-order perturbation theory and the steady-state heat equation, the fractional frequency shift $\delta\omega_{nm}/\omega_{nm}$ is derived as the product of a temperature responsivity constant and a dimensionless map $F_{nm}$.
  • The map $F_{nm}$ decomposes exactly into two terms: a mode-independent conduction envelope $A_{nm}$ (resembling the MTF) and a closed-form mode term $B_{nm}$.
  • Crucially, the model predicts that $B_{nm}$ peaks on the antinodal rows and columns of a mode, rather than at the displacement antinodes, because modal strain energy is carried by fibers (lines) rather than points.

• Experimental Setup:

  • Samples: A square silicon nitride (SiN) membrane ($L=500$ µm, $h=50$ nm) and a "trampoline" resonator.
  • Readout: An optical lever system with a 785 nm laser and a quadrant photodetector (QPD) tracks the resonance frequency using a self-sustaining oscillator (SSO) loop.
  • Excitation: A broadband light source filtered to 320 nm (to maximize SiN absorption) is modulated at 8 Hz and focused to a 30 µm spot.
  • Scanning: An xyz nanopositioner raster-scans the heating spot across the resonator surface on a 15 µm grid.
  • Analysis: For the square membrane, the measured maps for modes (1,1), (1,2), (2,1), and (2,2) were compared to the model. The conduction envelope was subtracted to isolate the mode term. For trampoline resonators, the tether-limited thermal conductance naturally flattens the envelope, allowing direct observation of the mode term.

Key Contributions and Results

1. Closed-Form Model Derivation: The paper provides an exact analytical decomposition of the photothermal responsivity map into a smooth conduction background and a specific mode-dependent term.
2. Experimental Validation: Scanning measurements on the SiN membrane confirm the model with a correlation coefficient $r \geq 0.998$ across all four modes, with residuals of only 1.0–1.6%. The only fitted parameter was the overall envelope amplitude.
3. Revealing the "Fiber" Effect: The experiments confirm the counterintuitive prediction that the responsivity peaks on antinodal rows and columns, not at the antinodes.
   • For the (2,2) mode, which has a nodal crossing at the center, the raw response is still maximal at the center due to the dominance of the conduction envelope. However, after subtracting this envelope, the mode term correctly shows zero response at the center and maxima along the antinodal lines.
   • Line cuts decisively distinguish the fiber model from local models (which predict peaks only at antinodes).
4. Trampoline Resonators: Measurements on trampoline resonators demonstrate that the tether-dominated thermal conductance renders the conduction envelope nearly position-independent. This allows the mode term to be directly visible in raw scans without envelope subtraction, clearly showing the distinct patterns of torsional and pad-localized modes.
5. Design Guidelines for Dual-Mode Sensing: The authors generated differential maps for potential mode pairs.
   • For square membranes, the (1,1)/(2,2) pair offers a maximal differential signal of ~20% at the center.
   • For trampolines, orthogonal torsional mode pairs show differential signals exceeding the fundamental mode reference.
   • These maps serve as design inputs for selecting mode pairs that combine large differential responsivity with a shared thermal and mechanical environment for common-mode drift suppression.

Significance
The paper claims to provide the first closed-form theory and comprehensive experimental validation of position- and mode-dependent photothermal responsivity. By establishing that the membrane acts as a grid of independent fibers responding to mean temperature, the work resolves the discrepancy between point-based intuition and observed mode shapes. The resulting maps are presented not as a final solution for noise suppression, but as a necessary design tool. They identify specific mode candidates (such as the (1,1)/(2,2) membrane pair or trampoline torsional pairs) that are suitable for dual-mode sensing schemes, shifting the focus from theoretical possibility to practical design parameters where the correlation of frequency fluctuations can be further investigated.