Micro 3D printing of a functional MEMS accelerometer

The study addresses the challenge of producing specialized MEMS sensors in low- to medium-volume markets where conventional semiconductor manufacturing is not cost-effective due to high fixed start-up costs. The research explores micro 3D printing—specifically two-photon polymerization—as a pathway to rapidly prototype and manufacture custom MEMS devices, focusing on a functional accelerometer. The purpose is to demonstrate feasibility and characterize performance (responsivity, resonance frequency, stability), thereby highlighting the potential of micro 3D printing to fill gaps in applications such as robotics, aerospace, and medicine where complex geometries and small batches are needed. The importance lies in enabling custom MEMS with design freedom and high resolution (<1 µm), scalable to a few hundred to a few thousand units.

Macro- and mesoscale 3D printing methods (fused filament fabrication, laser powder bed fusion, stereolithography) have produced inertial sensors with footprints from several mm² to cm², but these techniques are limited to features in the tens to hundreds of micrometers, constraining miniaturization and sensor bandwidth. Two-photon polymerization (also called multiphoton polymerization or direct laser writing) achieves sub-micrometer resolution in 3D, comparable to cleanroom lithography, and has been used for microfluidics, optics, and tissue engineering scaffolds. Prior functional 3D-printed micro-actuators using two-photon polymerization with metal sputtering (thermomechanical and electrostatic) have been reported, but microscale inertial sensors (accelerometers/gyroscopes) had not been realized before this work. This study fills that gap by demonstrating a functional 3D-printed MEMS accelerometer.

Design and fabrication: The accelerometer comprises a supporting pillar, two single-sided clamped horizontal cantilevers, and a proof mass at the cantilever ends. Structures were 3D printed on a glass substrate using a Nanoscribe Photonic Professional GT2 with commercial IP-S resin via two-photon polymerization. T-shaped shadow-masking features were integrated atop the cantilevers and pillar to enable selective metallization. Directional metal evaporation deposited 10 nm Ti and 30 nm Au, forming electrically isolated thin-film strain gauge resistors, interconnects, and probing electrodes via the shadow-masking geometry. The top surfaces of pillar and cantilevers were leveled to simplify electrical routing; the proof mass bottom surface was leveled with the cantilevers to avoid unsupported printing blocks. Mechanical/electrical transduction principle: Out-of-plane acceleration of the proof mass bends the cantilevers, inducing strain in the metal gauges and corresponding resistance changes proportional to applied acceleration. Design optimization and modeling: A parametrized COMSOL finite-element model was used to target a measurement range of 1–10 g. To increase sensitivity, cantilever thickness and width were minimized (both set to 20 µm) and length/proof mass size maximized within printability constraints. A parametric sweep yielded target dimensions: cantilever design length 500 µm (measured ≈480 µm), proof mass 350 µm × 300 µm × 210 µm (L×W×H). The COMSOL model was updated with the measured length for accurate comparison. Additional analytical models based on Euler–Bernoulli beam theory were developed (Supplementary S5) to estimate resonance frequencies. Characterization setup: Three nominally identical devices were fabricated. A piezoshaker provided controlled sinusoidal excitation; a laser Doppler vibrometer (LDV) measured proof mass oscillation amplitude. A lock-in amplifier drove the shaker (1.4–2.0 kHz, 1–7 Vrms) and demodulated LDV and electrical signals. The piezoshaker was calibrated (Materials and methods). Resonance was identified by sweeping frequency and fitting the amplitude response to a Lorentzian to extract f0 and Q. Mechanical displacement was measured at the proof mass surface (spot optimized for reflectivity). Electrical responsivity (ΔR/R vs acceleration) was measured using the lock-in demodulated signal at resonance and off-resonance; gauge factor was computed from responsivity and measured displacement. Stability tests: continuous repeated sweeps over 10 h at 5 Vrms while tracking resonance frequency and maximum ΔR/R.

- Resonance frequency: All three devices exhibited f0 ≈ 1.775 kHz ± 5 Hz. With probes attached, Q ranged 31–36 (without probes: 41.5 ± 1.4). Probe contact altered Q by up to ±30% but did not significantly affect f0. - Mechanical linearity: At resonance, displacement amplitude scaled linearly with applied acceleration up to ≈1.9 µm proof mass displacement. Simulations matched measured behavior when using an effective Young’s modulus of 6.5 GPa (vs 5.1 GPa from datasheet), consistent with the observed f0 (measured 1.775 kHz vs simulated 1.58 kHz with 5.1 GPa). - Responsivity: At resonance, ΔR/R responsivity across devices was 322–420 ppm/g. Normalized to off-resonance operation (100–160 Hz) by dividing by each device’s Q, the responsivity was 11 ± 0.7 ppm/g for all devices. - Gauge factor: Computed thin-film gauge factor ≈ 3.5 ± 0.6, higher than the COMSOL-based estimate (~2) expected for bulk Au, consistent with thin-film effects. - Stability over time (10 h at 5 Vrms): Resonance frequencies remained within ±3.8–4 Hz of their mean values, indicating stable mechanical properties. ΔR/R exhibited small drift: two devices showed ≈20 ppm variation in a 20 min rolling average from an initial ~205 ppm; one device was stable within ±4–5 ppm around ~193 ppm. - Additional observations: Resonance peak shifts at large oscillation amplitudes were attributed to polymer temperature increases. Simulated displacement amplitudes (with average Q ≈ 34) lay within the measured range.

The work demonstrates that two-photon polymerization, combined with directional metal evaporation and integrated shadow-masking, can produce electrically functional MEMS accelerometers with sub-micrometer feature control and true 3D design freedom. The accelerometers show predictable, linear mechanical response and measurable piezoresistive output with repeatable resonance characteristics. Agreement between measurement and simulation after updating the polymer’s effective Young’s modulus (6.5 GPa) validates the modeling approach and underscores the importance of accurate material parameter extraction for printed polymers. The observed gauge factor exceeding bulk Au expectations reflects thin-film metal behavior, supporting the efficacy of the deposited strain gauges. Q-factor sensitivity to probe contact emphasizes the need for optimized electrical interfacing or packaging, but resonance frequency robustness suggests that the core mechanical design is resilient. Short-term (10 h) stability in both f0 and responsivity indicates suitability for practical sensing tasks. Overall, the findings support micro 3D printing as a viable route to custom, low-/medium-volume MEMS sensors that are difficult to realize economically with conventional semiconductor processes.

This study reports, to the authors’ knowledge, the first functional 3D-printed MEMS accelerometer fabricated via two-photon polymerization with post-print directional Ti/Au evaporation forming integrated strain gauges and interconnects. The devices exhibit resonance near 1.775 kHz with Q ≈ 31–36 (up to ~41 without probes), linear mechanical response, responsivity of 322–420 ppm/g at resonance (≈11 ± 0.7 ppm/g off-resonance), gauge factor ≈3.5, and stable operation over 10 hours. Modeling and measurements align when accounting for an effective polymer Young’s modulus of 6.5 GPa. These results highlight micro 3D printing as an enabling technology for rapid prototyping and cost-effective production of specialized MEMS at low to medium volumes. Future directions include: optimizing polymer formulations and curing protocols to tune and stabilize mechanical properties; refined strain gauge materials and thicknesses to increase gauge factor and reduce temperature dependence; integrating full Wheatstone bridges and on-chip routing to improve sensitivity and reduce probe-loading effects; closed-loop readout and vacuum/packaged operation to enhance Q and bandwidth; comprehensive noise, bias stability, and environmental (temperature/humidity) characterization; and expanding to other inertial and physical sensors leveraging complex 3D geometries.

- Limited sample size (n=3) constrains statistical generalization. - Probe contact altered Q by up to ±30%, indicating measurement-induced loading; in-use packaging/readout could differ. - Resonance shifts at large amplitudes suggest thermal sensitivity of the polymer and thin-film gauges; detailed thermal characterization is not provided. - Material property discrepancies (effective Young’s modulus 6.5 GPa vs datasheet 5.1 GPa) highlight variability in printed polymer properties; broader material calibration is needed. - Long-term stability beyond 10 h, environmental robustness (temperature, humidity), and noise/bias drift metrics are not reported. - Affixed thin-film metal gauges formed by directional evaporation rely on shadow-masking geometries; scalability and robustness under shock/vibration were not evaluated.