A chip-level MEMS short-period seismometer for distributed planetary seismic networks Authors: Xingyu Wei, Chenhao Du, Shaolin Zhang, Qi Wang, Shimin Jiao, Qiu Wang, Hao Ouyang, Huafeng Liu Published: 2026 ABSTRACT Precise measurement of seismic activity is fundamental to planetary seismology. However, existing planetary microelectromechanical system (MEMS) seismometers encounter a critical bottleneck in balancing compact chip footprint, a low noise floor, and a wide measurement range. This paper presents a miniaturized MEMS seismometer, which has a compact chip footprint of only 5.2 mm × 6.5 mm. The device utilizes corrugated-gap capacitive sensing comb fingers, which significantly reduce sensing nonlinearity while maintaining high displacement sensitivity, effectively extending the linear measurement range and elevating the pull-in acceleration threshold. Furthermore, the system’s effective resonant frequency is reduced by exploiting the electrostatic negative stiffness mechanism, which aims to suppress the seismometer’s noise floor. Experimental results validate the seismometer’s comprehensive performance, demonstrating a noise floor of 5.5 ng/Hz1/2 (closely approaching the mechanical thermal noise limit), a measurement range of ±34 mg, and a bandwidth of 70 Hz. Under continuous operation, the instrument successfully captured natural seismic signals from a magnitude 7.0 earthquake 3682 km away and a magnitude 5.5 event 1240 km away, achieving a 99.1% waveform correlation coefficient when compared against a commercial reference seismometer. Ultimately, this work presents a novel chip-level MEMS seismometer for distributed seismic networks and planetary seismic observation. Introduction The precise measurement of planetary seismic activity is not only fundamental to scientific research, such as revealing planetary internal structures1–4, clarifying the laws of planetary formation5,6, and inverting the thermal states of extraterrestrial bodies7, but also provides essential core data for engineering applications, for instance, extraterrestrial resource exploration8–13, extraterrestrial habitat safety assessment14, and planetary geological disaster early warning15–19. The Apollo Passive Seismic Experiment (APSE) pioneered this field. Deployed on the lunar surface between 1969 and 1972, APSE acquired continuous lunar seismic data until 1977. However, its Short Period (SP) seismometer relied on a traditional macroscopic moving-coil design. The sensitive element was massive (23 cm × 29 cm). The entire system weighed 11.5 kg and consumed up to 7.4 W of power. It even required an additional 2.5 W heater to maintain thermal stability20. Although groundbreaking, APSE’s performance was heavily reliant on this bulky, power-hungry architecture, which has driven the evolution of planetary seismometers toward miniaturized, highly integrated, and low-power silicon-based MEMS. In 2018, NASA’s InSight Mars lander deployed the Seismic Experiment for Interior Structure (SEIS). It integrated a MEMS SP seismometer developed by Imperial College London21. Operating in its high-gain mode, the noise floor of this instrument drops to an impressive 0.25 ng/Hz1/2. This currently represents the absolute peak precision of planetary MEMS seismometers. However, achieving this sensitivity comes with strict functional trade-offs. Its immense gain of 27,000 V/(m/s) strictly limits the output range to just 0.3 mm/s, making it susceptible to saturation during larger seismic events. Furthermore, the core sensor chip has a dimension of 25 mm × 25 mm and employs a highly complex electromagnetic force-balanced closed-loop architecture22,23. This design necessitates the installation of bulky permanent magnets on the housing frame to provide a constant magnetic field, which severely restricts further size and weight reductions. Building on this established foundation, the same research team developed a lunar MEMS SP seismometer capable of achieving a 0.5 ng/Hz1/2 noise floor and a 100 Hz bandwidth. The Farside Seismic Suite (FSS) will soon carry this advanced payload to the lunar Schrödinger basin24,25. Other institutions have also made significant strides. In 2020, wang et al. developed the MEMS-M³ microgravity module, operating in a low Earth orbit environment. This closed-loop device successfully suppresses noise to between 2 ng/Hz1/2 and 5 ng/Hz1/2 in the 0.1 Hz to 10 Hz band, but consumes 3 W of power26. In 2023, the Indian Space Research Organisation (ISRO) deployed the ILSA payload during the Chandrayaan-3 mission. This marked the first stable lunar operation of a fully MEMS-based scientific payload27. To accommodate different signal strengths, the instrument adopts a fine- and coarse-range dual-chip fusion strategy, integrating two distinct sets of MEMS sensing units. It utilizes an open-loop architecture with gap-varying capacitive plates to measure displacement. Specifically, its fine-range seismometer features a compact 20 mm chip, with noise floor of 30 ng/Hz1/2, making it exceptionally difficult to detect the weak, deep geological signals that are crucial for understanding planetary interiors28,29. This study introduces a novel miniaturized MEMS SP seismometer with an ultra-compact footprint of merely 5.2 mm × 6.5 mm. To decrease the mechanical thermal noise without increasing the chip area, an extra mass is appended vertically. Crucially, we propose a corrugated-gap sensing comb finger. This unique geometry reduces sensing nonlinearity by a factor of three while maintaining high displacement sensitivity, effectively extending the linear measurement range and elevating the pull-in threshold. Furthermore, by exploiting the electrostatic negative stiffness effect derived from the capacitive electrodes, adjusting the AC carrier voltage actively reduces the system’s effective resonant frequency. Reducing the effective resonant frequency suppresses the overall noise floor to 5.5 ng/Hz1/2, closely approaching the mechanical thermal noise limit. The proposed device is fabricated by a simple four-mask process, demonstrating excellent comprehensive performance and successfully capturing multiple natural earthquake signals with high fidelity during continuous operation. Subsequent sections discuss the structural design, microfabrication process, and test results of the MEMS seismometer in detail. Results Device design and working principles The mechanical structure of the designed MEMS seismometer (as shown in Fig. 1a) consists of a sensor outer frame, a spring system composed of four cantilever beams, fixed electrodes, a proof mass, and an extra mass. The fixed electrodes and the movable electrodes on the proof mass interleave to form six sets of capacitive comb electrodes, creating differential capacitors Cs1 and Cs2. The extra mass is connected to the proof mass via the silicon oxide layer, which can increase the effective mass without expanding the footprint area. The MEMS seismometer first realizes the mechanical conversion of input acceleration into displacement through its sensitive element. This displacement signal is then converted into a capacitance signal via the differential capacitive comb electrodes. With the subsequent detection and readout circuit, the capacitance signal is ultimately converted into a voltage signal. Fundamentally, the sensing element of the MEMS seismometer can be modeled as an equivalent mass-spring-damper system. Thus, the equation of motion can be derived by incorporating the damping factor γγ. The mechanical response of the sensing element is calculated as follows30:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$m\frac{{d}^{2}x\left(t\right)}{d{t}^{2}}+\gamma \frac{{dx}\left(t\right)}{{dt}}+{k}_{m}x\left(t\right)=m{a}_{{in}}\left(t\right)$$\end{document}md2xtdt2+γdxtdt+kmxt=maintwhere mm is the effective mass, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{m}$$\end{document}km is the mechanical stiffness, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{{in}}$$\end{document}ain is the input acceleration. Through the Laplace transform, the transfer function between displacement and acceleration can be expressed as:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X\left(s\right)=\frac{{A}_{{in}}\left(s\right)}{{s}^{2}+\frac{s{\omega }_{0}}{Q}+{\omega }_{0}^{2}}$$\end{document}Xs=Ainss2+sω0Q+ω02where ss is the Laplace variable, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$X\left(s\right)$$\end{document}Xs and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${A}_{{in}}\left(s\right)$$\end{document}Ains represent the Laplace transforms of the proof mass displacement xx and the external acceleration \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{{in}}$$\end{document}ain, respectively. Here, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\omega }_{0}=\sqrt{{k}_{m}/m}$$\end{document}ω0=km/m is the natural angular frequency and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$Q=m{\omega }_{0}/\gamma$$\end{document}Q=mω0/γ is the quality factor of the sensing element. When the angular frequency of the input acceleration signal is much smaller than the natural angular frequency \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\omega }_{0}$$\end{document}ω0, the system gain can be simplified to a linear gain, yielding \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{{in}}={{\omega }_{0}}^{2}x.$$\end{document}ain=ω02x.Fig. 1Design of the MEMS seismometer chip.a 3D schematic of the MEMS seismometer chip, showing core components including sensing comb fingers, the spring-mass system, and the substrate frame. b Finite element modal analysis of the sensitive element; the first primary working mode (in-plane translation) has a fundamental frequency of 156 Hz. c Magnified local schematic of the corrugated-gap sensing comb fingers. d Kinematic model of a single corrugated unit. e Simulation results of the small-displacement sweep (up to 1.6 µm), confirming consistent sensitivity across varying angles θθ. f Comparison between theoretical and simulated second-order and third-order nonlinearity coefficients across various corrugation angles θθ Design of the MEMS seismometer chip. a 3D schematic of the MEMS seismometer chip, showing core components including sensing comb fingers, the spring-mass system, and the substrate frame. b Finite element modal analysis of the sensitive element; the first primary working mode (in-plane translation) has a fundamental frequency of 156 Hz. c Magnified local schematic of the corrugated-gap sensing comb fingers. d Kinematic model of a single corrugated unit. e Simulation results of the small-displacement sweep (up to 1.6 µm), confirming consistent sensitivity across varying angles θθ. f Comparison between theoretical and simulated second-order and third-order nonlinearity coefficients across various corrugation angles θθ To verify the design parameters obtained through theoretical calculation and to determine the vibration mode frequencies and sensitivities, a finite element method (FEM) analysis was conducted using COMSOL Multiphysics. As illustrated in Fig. 1b, the working mode is the in-plane translational mode with a mechanical resonant frequency of 156 Hz. Higher-order modes, including torsional and out-of-plane vibrations, were also investigated, demonstrating that the sensing element possesses a high mode rejection ratio Table 1.Table 1The dimensions of the deviceNameParameterValueCantilever BeamLength1630 μmWidth9 μmThickness50 μmDifferential Comb FingerGap3 μmNumber of Pairs160Extra MassLength3900 μmWidth700 μmThickness400 μm Table 1The dimensions of the deviceNameParameterValueCantilever BeamLength1630 μmWidth9 μmThickness50 μmDifferential Comb FingerGap3 μmNumber of Pairs160Extra MassLength3900 μmWidth700 μmThickness400 μm The dimensions of the device Considering the displacement detection characteristics of the differential capacitors, the input acceleration induces a change in the gap between the electrodes, which in turn alters the capacitance value. The sensitivity of traditional parallel sensing gaps is inversely proportional to the square of the gap distance. While decreasing the gap distance can yield ultra-high sensitivity, it drastically deteriorates linearity, severely impacting the linear measurement range and inducing prominent pull-in effects. This work adopts corrugated-gap comb fingers as shown in Fig. 1c. By altering the electric field distribution along the sensitive axis, the sensing nonlinearity coefficient is reduced while maintaining constant sensitivity. The model of a single corrugated unit is depicted in Fig. 1d. Due to the introduction of the corrugation angle θθ, the maximum travel range of the movable electrode is dd1 = \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d/\sin \theta > d$$\end{document}d/sinθ>d, which expands the allowable travel distance of the sensing element and improves shock resistance. When the structure undergoes a displacement xx, the capacitance change of a single corrugated unit is:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta C = \frac{\varepsilon h\left(\frac{{l}_{1}}{\sin \theta }-\frac{d}{\tan \theta }+x\cos \theta \right)}{d-x\sin \theta }-\frac{\varepsilon h\left(\frac{{l}_{1}}{\sin \theta }-\frac{d}{\tan \theta }\right)}{d}\\ = \frac{\varepsilon {l}_{1}hx}{d\left(d-x\sin \theta \right)}$$\end{document}ΔC=εhl1sinθ−dtanθ+xcosθd−xsinθ−εhl1sinθ−dtanθd=εl1hxdd−xsinθWhere \varepsilonε is the permittivity, hh is the thickness of the comb-finger electrode, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${l}_{1}$$\end{document}l1 is the corrugation height and dd is the electrode gap. To further investigate the capacitive sensing characteristics, a Taylor expansion at xx = 0 yields:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\Delta C= \frac{\varepsilon {l}_{1}h}{{d}^{2}}x+\sin \theta \frac{\varepsilon {l}_{1}h}{{d}^{3}}{(x)}^{2}\\ +{\sin }^{2}\theta \frac{\varepsilon {l}_{1}h}{{d}^{4}}{\left(x\right)}^{3}+O\left({x}^{4}\right)$$\end{document}ΔC=εl1hd2x+sinθεl1hd3(x)2+sin2θεl1hd4x3+Ox4The first term of Eq. (4) represents sensitivity, while the second and third terms are higher-order nonlinearity coefficients. Notably, sensitivity is independent of the corrugation angle, whereas the nonlinearity coefficients decrease proportionally to their terms. Thus, optimizing effectively suppresses sensing nonlinearity without compromising sensitivity. Furthermore, the total capacitance change for a single electrode is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\varepsilon {hl}x/d(d-x\sin \theta )$$\end{document}εhlx/d(d−xsinθ), where ll is the overlapped length of the finger. Considering the comb finger width, the corrugation width ww was set to 10 μm. Meanwhile, the corrugation heights \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${h}_{1}$$\end{document}h1 were evaluated at 4 μm, 6 μm, 8 μm, 10 μm, and 12 μm, corresponding to the θθ angles of 22°, 31°, 38°, 45°, and 50° (90° corresponds to the conventional parallel-plate electrode). In this work, electromechanical simulations were conducted over comb fingers by using COMSOL MULTIPHYSICS. A displacement sweep (up to 1.6 µm) confirmed that sensitivities remain fundamentally identical across various θθ values (Fig. 1e). The simulated sensitivities and high-order nonlinearity coefficients for different angles are summarized in Table 2.Table 2Simulated sensitivity and nonlinear coefficients of a single comb fingerθθ(o)Sensitivity (fF/um)2nd-order nonlinearity (fF/um2)3rd-order nonlinearity (fF/um3)9018.76.12.15018.64.61.24518.64.31.13818.53.80.93118.53.20.72217.92.60.7 Table 2Simulated sensitivity and nonlinear coefficients of a single comb fingerθθ(o)Sensitivity (fF/um)2nd-order nonlinearity (fF/um2)3rd-order nonlinearity (fF/um3)9018.76.12.15018.64.61.24518.64.31.13818.53.80.93118.53.20.72217.92.60.7 Simulated sensitivity and nonlinear coefficients of a single comb finger To identify the optimal angle that yields the best linearity, the theoretical and simulated capacitance-displacement curves for various θθ were compared. (Fig. 1f). At θθ = 90°, nonlinearity peaked and was tightly aligned with theoretical values. As θθ decreases, the second-order coefficient reduces accordingly. For θθ < 30°, edge effects cause the simulation results to diverge from theoretical predictions. Similarly, the third-order coefficient aligned with theory at 50° and 45°, but increased to 1.2, 1.3, and 2.5 times the theoretical values at 38°, 31°, and 22°, respectively. Balancing these characteristics and laboratory fabrication capabilities, the corrugation angle was ultimately set to 31°. Pull-in effect and electrostatic negative stiffness effect The differential capacitance readout circuit of the proposed device requires a high-frequency AC carrier voltage \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}$$\end{document}Vp~ applied to the detection mass to accurately measure small capacitance changes. This carrier voltage modulates the capacitance into a high-frequency voltage signal, suppressing the impact of 1/f noise. Simultaneously, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}$$\end{document}Vp~ generates an electrostatic field, exerting an additional force on the sensitive mass. The stored electric field energy, WW, is expressed as:5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$W = \frac{1}{2}{C}_{s1}\left(x\right){\widetilde{{V}_{p}}}^{2}+\frac{1}{2}{C}_{s2}\left(x\right){\widetilde{{V}_{p}}}^{2}\\ = \frac{n\varepsilon h\left[d\left(l-{dcos}\theta \right)+{x}^{2}{\sin }^{2}\theta \cos \theta \right]}{s{in}\theta ({d}^{2}-{x}^{2}{\sin }^{2}\theta )}\widetilde{{V}_{p}}$$\end{document}W=12Cs1xVp~2+12Cs2xVp~2=nεhdl−dcosθ+x2sin2θcosθsinθ(d2−x2sin2θ)Vp~where nn denotes the number of pairs of differential comb electrodes. By taking the spatial partial derivative of WW the equivalent electrostatic force \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{e}$$\end{document}Fe along the sensitive axis is obtained:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{e}=\frac{\partial W}{\partial x}=\frac{2n\varepsilon {hld}x\sin \theta {\widetilde{{V}_{p}}}^{2}}{{\left({d}^{2}-{x}^{2}{\sin }^{2}\theta \right)}^{2}}$$\end{document}Fe=∂W∂x=2nεhldxsinθVp~2d2−x2sin2θ2Equation (6) shows that when the proof mass is at its initial equilibrium position, the electrostatic force is zero. However, external acceleration displaces the mass by xx. Consequently, the mass experiences three simultaneous forces: inertial force, elastic restoring force, and electrostatic force. Under steady-state conditions, mechanical equilibrium is established, which satisfies the following equation:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${ma}={k}_{m}x-{F}_{e}={k}_{m}x-{\widetilde{{V}_{p}}}^{2}\frac{2n\varepsilon {hld}x\sin \theta }{{\left({d}^{2}-{x}^{2}{\sin }^{2}\theta \right)}^{2}}$$\end{document}ma=kmx−Fe=kmx−Vp~22nεhldxsinθd2−x2sin2θ2For stable operation, the elastic restoring force must overcome the combined inertial and electrostatic forces, satisfying \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${ma}\le {k}_{m}x-{F}_{e}$$\end{document}ma≤kmx−Fe. Equation (6) reveals that as xx increases, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${F}_{e}$$\end{document}Fe becomes highly nonlinear. The total restoring force initially increases, reaches a peak, and then decreases. This peak defines the maximum tolerable inertial force. Exceeding this limit causes pull-in, forcefully pulling the mass onto the fixed electrodes. As shown in Fig. 2a, at |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}$$\end{document}Vp~ | = 2 V, decreasing the inclination angle θθ significantly broadens the linear motion range and raises the pull-in displacement threshold. Figure 2b further quantifies that smaller θθ angles increase the pull-in acceleration threshold. When θθ > 80°, the device undergoes pull-in directly without any acceleration input.Fig. 2Theoretical analysis of the pull-in effect and electrostatic negative stiffness.a Total restoring force versus displacement at various θθ under | \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}|$$\end{document}Vp~∣= 2 V, illustrating broadened linear motion ranges as θθ decreases. b Pull-in acceleration corresponding to varying θθ, at |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}|$$\end{document}Vp~∣= 2 V; when θθ > 79°, the device experiences pull-in without any acceleration input. c Pull-in voltage limits across different θθ, at |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}|$$\end{document}Vp~∣= 2 V, demonstrating higher voltage thresholds for smaller θθ, with a 2.75 V threshold at θθ = 31°. d Dynamic regulation of the system’s effective resonant frequency by adjusting the carrier voltage amplitude Theoretical analysis of the pull-in effect and electrostatic negative stiffness. a Total restoring force versus displacement at various θθ under | \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}|$$\end{document}Vp~∣= 2 V, illustrating broadened linear motion ranges as θθ decreases. b Pull-in acceleration corresponding to varying θθ, at |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}|$$\end{document}Vp~∣= 2 V; when θθ > 79°, the device experiences pull-in without any acceleration input. c Pull-in voltage limits across different θθ, at |\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}|$$\end{document}Vp~∣= 2 V, demonstrating higher voltage thresholds for smaller θθ, with a 2.75 V threshold at θθ = 31°. d Dynamic regulation of the system’s effective resonant frequency by adjusting the carrier voltage amplitude Equation (6) also reveals an attractive electrostatic force that opposes the mechanical springs, effectively driving the mass away from the center. The equivalent negative stiffness, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{e}$$\end{document}ke, is expressed as:8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{e}=-\frac{\partial {F}_{e}}{\partial x}=-\frac{2n\varepsilon {hldsin}\theta \left({d}^{2}+3{x}^{2}{\sin }^{2}\theta \right)}{{\left({d}^{2}-{x}^{2}{\sin }^{2}\theta \right)}^{3}}{\widetilde{{V}_{p}}}^{2}$$\end{document}ke=−∂Fe∂x=−2nεhldsinθd2+3x2sin2θd2−x2sin2θ3Vp~2Under a small displacement approximation (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$d\gg x\approx 0$$\end{document}d≫x≈0), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{e}\approx 2n\varepsilon {hlsin}\theta {\widetilde{{V}_{p}}}^{2}/{d}^{3}$$\end{document}ke≈2nεhlsinθVp~2/d3. System stability requires \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{m}$$\end{document}km > \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$|{k}_{e}|$$\end{document}∣ke∣, meaning | \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}$$\end{document}Vp~ | < \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\sqrt{{k}_{m}{d}^{3}/(2n\varepsilon {hlsin}\theta )}$$\end{document}kmd3/(2nεhlsinθ). Figure 2c shows pull-in voltages across different θθ angles. A smaller θθ yields a higher pull-in voltage; at θθ = 31°, it is roughly 1.4 times higher than that of parallel electrodes. Furthermore, this electrostatic negative stiffness corrects the system’s effective resonant frequency \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\omega }_{{eff}}$$\end{document}ωeff to:9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${\omega }_{{eff}}=\sqrt{\frac{{k}_{m}}{m}-\frac{2n\varepsilon {hlsin}\theta }{{d}^{3}m}{\widetilde{{V}_{p}}}^{2}}$$\end{document}ωeff=kmm−2nεhlsinθd3mVp~2Thus, adjusting \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}$$\end{document}Vp~ can regulate the device’s effective resonant frequency and sensitivity as shown in Fig. 2d. In summary, the corrugation angle θθ is a crucial parameter for tuning electrostatic nonlinearity and stability. Smaller angles expand the linear range and elevate pull-in thresholds, and excessive angles risk premature pull-in. Finally, exploiting electrostatic negative stiffness provides an adjustable method for optimizing sensor performance by simply tuning the carrier voltage31. Setup The fabricated MEMS chip is encapsulated in a ceramic package, as shown in Fig. 3a. The packaged chip was subsequently integrated onto a customized printed circuit board (PCB) and tested inside a vacuum chamber. The test instruments utilized in this study are shown in Fig. 3b. A power supply (KEYSIGHT E3631A) supplies power to the PCB, and the AC carrier voltage was supplied by a waveform generator (KEYSIGHT 33600A). The output of the device was recorded using a data acquisition system (NI PXIe-4464). A commercial reference seismometer (Streckeisen STS-2.5) was placed parallel to the sensitive axis of the proposed MEMS seismometer. Performance evaluations of the proposed MEMS seismometer were conducted under vacuum conditions. During the experiments, both the MEMS chip and the PCB were installed within the vacuum chamber.Fig. 3Fabricated MEMS seismometer and experimental setup.a The chip encapsulated in a ceramic package. b All experimental equipment used in this paper Fabricated MEMS seismometer and experimental setup. a The chip encapsulated in a ceramic package. b All experimental equipment used in this paper Calibration experiments To characterize the regulatory effect of the carrier voltage amplitude on the system’s effective resonant frequency, the device’s output signals under different carrier voltages were collected using random ambient noise excitation. These signals underwent a Fast Fourier Transform (FFT) to produce frequency-domain response curves, as shown in Fig. 4a. Because the device possesses an extremely high quality factor, according to Eq. (2), the output signals show a sharp resonance peak at the effective resonant frequency. As the electrostatic drive voltage increases, the position of the resonance peak shifts. The shifted resonance peak with drive voltage is presented in Fig. 4b, and the theoretical fitting using Eq. (9) determines the initial mechanical resonant frequency of the device to be 168 Hz. For subsequent testing, the amplitude of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\widetilde{{V}_{p}}$$\end{document}Vp~ was set to 2 V.Fig. 4Calibration experiments for the MEMS seismometer.a Amplitude-frequency response tested under random ambient noise excitation with varying carrier amplitudes. b Effective resonant frequency versus carrier voltage, yielding a fitted initial mechanical resonant frequency of 168 Hz. c Ring-down test, extracting a time constant of 2.9 s and an effective resonant frequency 122 Hz. d Gravity field calibration, deriving a scale factor of 227 V/g through linear fitting, with an R2 value of 0.9999 and a nonlinearity of 0.84% Calibration experiments for the MEMS seismometer. a Amplitude-frequency response tested under random ambient noise excitation with varying carrier amplitudes. b Effective resonant frequency versus carrier voltage, yielding a fitted initial mechanical resonant frequency of 168 Hz. c Ring-down test, extracting a time constant of 2.9 s and an effective resonant frequency 122 Hz. d Gravity field calibration, deriving a scale factor of 227 V/g through linear fitting, with an R2 value of 0.9999 and a nonlinearity of 0.84% Furthermore, the time-domain ring-down curve of the device was calibrated by applying impact vibrations, as shown in Fig. 4c. Based on the 2.9 s required for its amplitude to decay to its 1/e and the effective resonant frequency of 122 Hz derived from its oscillation period, a quality factor of 1112 was extracted; notably, this frequency is consistent with the aforementioned experiments. Moreover, the mechanical thermal noise \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{{th}}$$\end{document}ath induced by thermal Brownian motion is expressed as follows32:10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{th}=\sqrt{\frac{4{k}_{B}T{\omega }_{{eff}}}{{mQ}}}$$\end{document}ath=4kBTωeffmQwhere \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${k}_{B}$$\end{document}kB is the Boltzmann constant, and TT = 298.15 K denotes room temperature. The calculation yields \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${a}_{{th}}$$\end{document}ath = 4.9 ng/Hz1/2. The experiment adopts the gravity field calibration method to test the sensitivity of the device, controlling the tilt angle of the device via a high-precision dual-axis turntable. This applies a static acceleration of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$g\times \sin (\alpha )$$\end{document}g×sin(α) along the sensitive axis, where αα is the turntable tilt angle33. The linear fitting reveals a scale factor of 227 V/g and a full-scale nonlinearity of 0.84%, where an R2 value is 0.9999. Furthermore, the sensor features a wide linear measurement range of 34 mg. When the acceleration input exceeds 52 mg, the device experiences pull-in. Performance testing To characterize the device’s amplitude-frequency response, the waveform generator applied a 0.5–500 Hz sinusoidal drive voltage to fixed electrodes, electrostatically actuating the proof mass along its sensitive axis. Fitting this experimental data with Eq. (2) yielded an R2 value of 0.9962, establishing a +3 dB bandwidth of 70 Hz and a −3 dB bandwidth of 180 Hz . Next, we evaluated the device’s self-noise. The MEMS seismometer was placed adjacent to the commercial reference seismometer, and measurements were started simultaneously over the same time period. The raw measured signal represents a superposition of the sensor’s self-noise and Earth’s micro-tremors34. Power Spectral Density (PSD) analysis confirmed that both devices successfully captured these micro-tremors signals, as shown in Fig. 5b. Since the background noise behaves as a common-mode signal in both sensors, we adopt coherence analysis to remove the correlated ambient noise and extract the self-noise of the MEMS seismometer35. The evaluated noise floor at 1 Hz is 5.5 ng/Hz1/2, which closely approaches the mechanical thermal noise limit. Correspondingly, the device achieves a dynamic range of 136 dB at 1 Hz.Fig. 5Comprehensive performance evaluation and seismic observation.a Amplitude-frequency response testing, yielding a +3 dB bandwidth of 70 Hz. b PSD comparison with a commercial reference, extracting a self-noise floor of 5.5 ng/Hz1/2 at 1 Hz via coherence analysis. c Time-domain seismic records of two natural earthquakes captured during a 48-h continuous test, exhibiting high waveform consistency with the commercial seismometer, with a correlation coefficient reaching 99.1% Comprehensive performance evaluation and seismic observation. a Amplitude-frequency response testing, yielding a +3 dB bandwidth of 70 Hz. b PSD comparison with a commercial reference, extracting a self-noise floor of 5.5 ng/Hz1/2 at 1 Hz via coherence analysis. c Time-domain seismic records of two natural earthquakes captured during a 48-h continuous test, exhibiting high waveform consistency with the commercial seismometer, with a correlation coefficient reaching 99.1% Finally, a 48-hour continuous operational test verified the seismometer’s practical detection performance. The seismometer recorded two natural earthquakes. The first was an M7.0 event near Kalimantan Island (116.2°E, 6.95°S), located 3682 km away. The second was an M5.5 event in Yilan County, Taiwan (121.81°E, 24.7°N), located 1240 kilometers away. Our waveform data proved highly consistent with the commercial reference seismometer. Both seismometers exhibited clear, identifiable responses to P-waves, S-waves, and surface waves36. Ultimately, the waveform correlation coefficient reached an impressive 99.1%, demonstrating the high reliability and practical effectiveness of the miniaturized seismometer. Discussion To evaluate the overall performance of the seismometer, a figure of merit (FOM) is suggested to include three factors: range, bandwidth, chip size, and noise floor. Therefore, the overall FOM can be expressed as:11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$${FOM}=\frac{{Range}\times {Bandwidth}}{{Noise\; floor}\times \sqrt{C{hip}{size}}}$$\end{document}FOM=Range×BandwidthNoisefloor×ChipsizeTable 3 summarizes the core performance comparison between the MEMS seismometer developed in this work and existing planetary MEMS seismometers. Existing planetary MEMS seismometers generally suffer from excessively large dimensions. Solutions employing electromagnetic force-balanced closed-loop architectures can achieve extremely low noise floors (down to 0.25 ng/Hz1/2), but require additional magnets to provide electromagnetic force, which significantly increases chip volume and system complexity. Open-loop architecture MEMS solutions, while having accomplished extraterrestrial on-orbit deployments, still suffer from excessively large chip dimensions, relatively high noise floors, and insufficient linear measurement ranges.Table 3Performance comparison of existing planetary MEMS seismometersDeviceNoise floorBandwidthWork modeRangeChip sizeFOMNASA InSight SEIS230.25 ng/Hz1/250 HzClosed-loop2 mg*25*25 mm216NASA CLPS FSS24,250.5 ng/Hz1/2100 HzClosed-loop—25*25 mm2—HUST MEMS-M³ 262 ng/Hz1/250 HzClosed-loop2 mg25*25 mm22ISRO Chandrayaan-3 ILSA** 28,2930 ng/Hz1/240 HzOpen-loop15 mg17*20 mm21This work5.5 ng/Hz1/270 Hz(+3 dB)180 Hz(−3 dB)Open-loop34 mg5.2*6.5 mm274*The MEMS seismometer of NASA InSight SEIS possesses a noise floor of 0.25 ng/Hz1/2 in high-gain mode, corresponding to a velocity range of 0.3 mm/s. At a characteristic frequency of 10 Hz, this velocity range converts to an acceleration range of approximately 2 mg**The parameters reported for the ISRO Chandrayaan-3 ILSA correspond specifically to its fine-range sensing unit Table 3Performance comparison of existing planetary MEMS seismometersDeviceNoise floorBandwidthWork modeRangeChip sizeFOMNASA InSight SEIS230.25 ng/Hz1/250 HzClosed-loop2 mg*25*25 mm216NASA CLPS FSS24,250.5 ng/Hz1/2100 HzClosed-loop—25*25 mm2—HUST MEMS-M³ 262 ng/Hz1/250 HzClosed-loop2 mg25*25 mm22ISRO Chandrayaan-3 ILSA** 28,2930 ng/Hz1/240 HzOpen-loop15 mg17*20 mm21This work5.5 ng/Hz1/270 Hz(+3 dB)180 Hz(−3 dB)Open-loop34 mg5.2*6.5 mm274*The MEMS seismometer of NASA InSight SEIS possesses a noise floor of 0.25 ng/Hz1/2 in high-gain mode, corresponding to a velocity range of 0.3 mm/s. At a characteristic frequency of 10 Hz, this velocity range converts to an acceleration range of approximately 2 mg**The parameters reported for the ISRO Chandrayaan-3 ILSA correspond specifically to its fine-range sensing unit Performance comparison of existing planetary MEMS seismometers 70 Hz(+3 dB) 180 Hz(−3 dB) *The MEMS seismometer of NASA InSight SEIS possesses a noise floor of 0.25 ng/Hz1/2 in high-gain mode, corresponding to a velocity range of 0.3 mm/s. At a characteristic frequency of 10 Hz, this velocity range converts to an acceleration range of approximately 2 mg **The parameters reported for the ISRO Chandrayaan-3 ILSA correspond specifically to its fine-range sensing unit In contrast, the miniaturized MEMS seismometer proposed in this paper achieves a balanced, high-performance breakthrough on a miniaturized foundation. The device chip size is merely 5.2 mm × 6.5 mm, making it the smallest in area among all benchmarked solutions—only 24% of the SEIS, FSS, and MEMS-M³ footprints, and 30% of the ILSA footprint. Its noise floor is as low as 5.5 ng/Hz1/2, approaching the mechanical thermal noise limit of 4.9 ng/Hz1/2. Concurrently, it realizes a wide linear measurement range of ±34 mg and a high dynamic range of 136 dB, leading peer open-loop solutions by 22 dB and performing comparably to the closed-loop NASA InSight SEIS. The bandwidth of the device is 70 Hz, which is higher than that of currently on-orbit MEMS seismometers. In summary, compared to existing MEMS seismometers, the proposed seismometer achieves a 4.5-fold improvement in the FOM, demonstrating superior overall performance. Through the innovative design of corrugated-gap capacitive sensing comb fingers, this work reduces the second-order and third-order nonlinearity coefficients by 2-fold and 3-fold, respectively, greatly broadening the linear range, increasing the pull-in voltage by a factor of 1.4, and resolving the sensitivity-linearity contradiction in traditional architectures. By leveraging the electrostatic negative stiffness mechanism and adjusting the carrier voltage to regulate the effective resonant frequency, it suppresses the noise floor to the mechanical thermal noise limit. Future work will focus on novel packaging methods and improved interface circuits. Specifically, utilizing an Application-Specific Integrated Circuit (ASIC) for capacitive signal detection is expected to yield higher resolution and greater integration. The low manufacturing cost, compact size, and mass-production scalability of the proposed MEMS seismometer highlight its potential for broader applications, enhanced flexibility, and ease of integration. It is conceivable that these devices could be deployed as distributed planetary seismic networks in practical scenarios, serving purposes such as extraterrestrial resource exploration, safety assessments of extraterrestrial habitats, and infrastructure health monitoring. Materials and methods Fabrication of the MEMS chip The MEMS seismometer is fabricated on a standard silicon on insulator (SOI) wafer using an efficient four-mask process (Fig. 6). The specific workflow is as follows: (a) The wafer surface is cleaned with a buffered HF solution (1:5) to remove native oxide. (b) A 30 nm Cr and 200 nm Au layer are deposited on the handle layer. The first mask defines the deposited gold through lithography and etching. (c) Using the second mask, thick photoresist is patterned. Deep reactive ion etching (DRIE) on the handle layer forms the release cavity and extra mass. (d) The SOI is rinsed with buffered HF. A 350 nm Al layer is deposited on the device layer, where the third mask patterns the pads. (e) The fourth mask and DRIE etch on the device layer. This precisely defines the movable structures and fixed electrodes. (f) Finally, dry etching removes the buried oxide (BOX) sacrificial layer. This releses the micromechanical structures, finalizing the chip fabrication.Fig. 6Microfabrication process flow of the MEMS seismometer chip.a Cleaning the SOI wafer and removing the native surface oxide. b Depositing Cr/Au layers and patterning the substrate grounding electrodes. c Etching the handle layer via DRIE to form the release cavity and extra mass block. d Depositing an Al layer and patterning the interconnection pads on the device layer. e DRIE on the device layer to define the mass-spring structures and comb fingers. f Dry etching the buried oxide (BOX) sacrificial layer to release the micromechanical structures Microfabrication process flow of the MEMS seismometer chip. a Cleaning the SOI wafer and removing the native surface oxide. b Depositing Cr/Au layers and patterning the substrate grounding electrodes. c Etching the handle layer via DRIE to form the release cavity and extra mass block. d Depositing an Al layer and patterning the interconnection pads on the device layer. e DRIE on the device layer to define the mass-spring structures and comb fingers. f Dry etching the buried oxide (BOX) sacrificial layer to release the micromechanical structures Interface circuit The capacitance detection interface circuit relies on a differential transformer bridge and amplitude modulation to mitigate electronic 1/f noise, as shown in Fig. 7. Initially, minute low-frequency capacitance variations are modulated by a 50 kHz AC carrier. A transformer-based charge amplifier then converts the resulting current into an AC voltage. This signal passes through a band-pass filter to reject out-of-band noise, significantly improving the signal-to-noise ratio. Next, a lock-in amplifier performs coherent demodulation to down-convert the high-frequency signal back to the baseband. To ensure optimal phase coherence, a phase conditioning circuit precisely aligns the demodulation reference with the original carrier source. Finally, a low-pass filter eliminates residual high-frequency carrier and mixing components, delivering the extracted low-frequency voltage signal.Fig. 7Architecture of the capacitance readout system.Schematic diagram of the interface circuit based on a differential transformer bridge and modulation-demodulation technology Architecture of the capacitance readout system. Schematic diagram of the interface circuit based on a differential transformer bridge and modulation-demodulation technology