MEMS design is the engineering discipline of creating micro-electro-mechanical systems, devices with moving mechanical parts built at the scale of micrometers, roughly the width of a human hair or smaller. At this scale, familiar physics starts behaving differently: surface forces overwhelm gravity, materials flex in ways bulk objects never would, and air itself becomes a sticky drag on moving parts. Designing a working MEMS device means navigating all of these quirks while choosing the right materials, fabrication methods, and actuation strategies for a given application.
Why Microscale Physics Changes the Design Rules
The most fundamental thing that makes MEMS design different from designing, say, a wristwatch mechanism is that the physics at micrometer scales do not behave the way you would expect from everyday experience. When you shrink a mechanical part down to a few micrometers, surface area becomes enormous relative to volume. That shift means forces you would never worry about in a large machine, like electrostatic attraction or molecular-level stickiness between surfaces, suddenly dominate everything. Gravity, by contrast, becomes almost irrelevant.
This creates a major headache called stiction, where two tiny surfaces that come into contact can become permanently stuck together because adhesive forces at the microscale are so strong. Research on MEMS surface interactions has shown that adhesive force is higher at the microscale, even though at even smaller (nanoscale) dimensions, forces like electrostatic attraction that are negligible at the microscale can become dominant in their own right.1Nanotechnology. Scale dependence of micro/nano-friction and adhesion of MEMS/NEMS materials, coatings and lubricants For designers, this means every surface interaction has to be carefully managed. Anti-stiction coatings are a standard part of the toolkit. Studies have found that perfluoroalkylsilane coatings applied to silicon surfaces produce the highest water contact angle and lowest surface energy compared to other common coatings, making them particularly effective at preventing unwanted adhesion.2Sensors and Actuators A: Physical. AFM study of perfluoroalkylsilane and alkylsilane self-assembled monolayers for anti-stiction in MEMS/NEMS
Air damping is another microscale nuisance. A MEMS resonator vibrating in open air loses energy to the surrounding gas far more readily than a larger structure would, because the thin air film between closely spaced surfaces acts almost like a viscous fluid. Designers counter this by operating devices in vacuum packages or by engineering comb-like geometries that reduce the squeezing effect of trapped air between moving parts.
Choosing the Right Material
Silicon dominates MEMS design for the same reason it dominates the semiconductor industry: decades of manufacturing infrastructure already exist for working with it, and its mechanical properties are excellent. Single-crystal silicon is stiff, strong, and does not creep or fatigue as easily as metals at the microscale. Its stiffness varies by crystal orientation, ranging from about 130 to 188 GPa depending on which direction you measure, while polycrystalline silicon (polysilicon) sits around 170 GPa.3Sensors and Actuators A: Physical. Review Materials: Silicon and beyond That directional variation matters in design because a pressure sensor’s diaphragm or an accelerometer’s proof mass will flex differently depending on how it is oriented relative to the crystal lattice.
Polysilicon is the workhorse for many surface-micromachined MEMS because it can be deposited as thin films at relatively low temperatures, making it compatible with building structures layer by layer on top of a wafer. Single-crystal silicon tends to be used in bulk-micromachined devices where the wafer itself is carved into the structure.
Beyond silicon, two-dimensional materials like graphene and molybdenum disulfide have attracted serious research interest. Suspended membranes made from these atomically thin sheets have demonstrated higher sensitivity in mass sensors, gas sensors, accelerometers, pressure sensors, and microphones compared to their silicon-based counterparts.4PubMed Central. Nanomaterials Based Micro/Nanoelectromechanical System (MEMS and NEMS) Devices These materials represent the ultimate limit of thickness, and their nanoscale dimensions, high mechanical strength, and easy electronic integration open up capabilities that bulk silicon cannot match.5Applied Physics Reviews. Nanoelectromechanical systems from two-dimensional materials That said, manufacturing processes for 2D materials are still far less mature than for silicon, so most commercial MEMS devices remain silicon-based for now.
How MEMS Structures Are Made
MEMS fabrication borrows heavily from semiconductor chip manufacturing but adds its own set of processes for creating three-dimensional mechanical structures. The two broad approaches are surface micromachining (building up thin-film layers and selectively removing sacrificial layers to free moving parts) and bulk micromachining (etching directly into the silicon wafer to carve out structures like cantilevers, diaphragms, and channels).
Deep reactive ion etching, commonly known as DRIE, is one of the most important tools in the MEMS fabrication toolbox. Using a technique called the Bosch process, DRIE alternates between etching and passivation steps to cut deep, nearly vertical trenches into silicon. This enables high-aspect-ratio structures and gives designers a lot of geometric flexibility. The trade-off is that DRIE suffers from a phenomenon called etch lag, where narrow trenches etch more slowly than wide ones, creating depth inconsistencies across a device.6PubMed Central. Reduced Etch Lag and High Aspect Ratios by Deep Reactive Ion Etching (DRIE) Getting around etch lag often requires complex oxide masks and careful process tuning, which adds time and cost to development.
For biomedical applications, fabrication gets even more interesting. Both hard micromachining (etching silicon and glass) and soft micromachining (molding polymers like PDMS) are used, sometimes on the same device. This combination allows controllable microfluidic systems that can handle biological fluids, deliver drugs, or perform chemical analyses on tiny sample volumes.7PubMed. Hard and soft micromachining for BioMEMS: review of techniques and examples of applications in microfluidics and drug delivery
Making Things Move at the Microscale
A MEMS device with no moving part is just a chip. The moving part needs something to drive it, and at the microscale, the options differ from what you would find in a conventional machine shop. The main actuation mechanisms designers choose from are electrostatic, piezoelectric, thermal, and electromagnetic, each with different strengths and drawbacks.
Electrostatic actuation is the most common. It works by applying a voltage between two closely spaced conductive surfaces, generating an attractive force between them. The classic implementation is the comb-drive actuator, where two sets of interleaved fingers slide past each other like the teeth of two combs. Comb drives are elegant because the force they generate is nearly linear with voltage and because they can be fabricated entirely in polysilicon using standard surface micromachining. The catch is that those slender comb fingers can suffer from side pull-in instability: if two adjacent fingers bend too much, they snap together and short out the device. Research has modeled this problem and shown that it is driven by the mechanical compliance of the fingers relative to the electrostatic force pulling them sideways.8Sensors and Actuators A: Physical. On the side instability of comb-fingers in MEMS electrostatic devices Designers counter this by making fingers stiffer, shortening them, or adding mechanical stops.
Piezoelectric actuation uses materials that deform when voltage is applied to them, or conversely, generate voltage when deformed. This makes piezoelectric MEMS popular for applications that go both ways, like microphones and speakers. A piezoelectric MEMS loudspeaker, for instance, can be modeled using lumped element and finite element approaches to predict the sound pressure it generates in open air or inside an ear canal.9Microelectronics Reliability. A piezoelectric MEMS loudspeaker for in-ear and free field applications lumped and finite element models Piezoelectric devices tend to produce larger displacements than electrostatic ones for a given voltage, but the piezoelectric thin films add complexity to fabrication.
Thermal actuators work by heating a beam and exploiting its thermal expansion. They can generate large forces and displacements but are slow and power-hungry by MEMS standards. Electromagnetic actuators use coils and magnetic fields, offering strong forces but requiring external magnets or on-chip coils that consume significant die area.
Integrating MEMS with Electronics
A MEMS sensor or actuator rarely works alone. It almost always needs companion electronics to read signals, apply drive voltages, or process data. How you combine the MEMS structure with its electronics is one of the biggest architectural decisions in a MEMS design project, and the industry has settled into two main approaches.
The hybrid approach puts the MEMS die and the electronics die in the same package but on separate chips. They connect through wire bonds or through-silicon vias. Roughly half of the MEMS market uses this method because it is modular: you can optimize the MEMS fabrication process and the circuit fabrication process independently, and development time is shorter.10Sensors and Actuators A: Physical. Viewpoint Paper CMOS–MEMS integration today and tomorrow If your MEMS process needs a high-temperature anneal that would fry transistors, no problem: the transistors live on a different chip.
The monolithic approach builds MEMS structures and transistors on the same chip. This can yield smaller packages, lower parasitic capacitance (which helps signal quality), and potentially lower unit cost at high volumes. But the development effort is much heavier because the MEMS and CMOS processes have to be made compatible, and any change to one can break the other. Monolithic integration tends to pay off for very high-volume consumer products like smartphone accelerometers, where the extra engineering investment is amortized across billions of units.
Simulation Before Silicon
Because MEMS fabrication is expensive and slow, with each design iteration taking weeks to months from mask layout to tested die, getting the design as close to right as possible before committing to silicon is critical. MEMS simulation spans multiple physical domains simultaneously. A single device might involve structural mechanics, electrostatics, fluid dynamics, and thermal transport, all coupled together.
Finite element analysis is the standard tool for detailed device-level simulation. It discretizes the structure into tiny elements and solves the governing equations over each one, capturing complex effects like stress concentrations, mode shapes, and resonant frequencies. Lumped element models, by contrast, represent the device as a network of idealized components (springs, masses, dampers, capacitors) and solve much faster. The trade-off is fidelity: lumped models work well for predicting overall behavior but miss spatial details that finite element models capture. In practice, designers use both. Lumped models handle early-stage parameter sweeps to narrow the design space, and finite element models verify the final design in detail.
For a piezoelectric MEMS loudspeaker, for example, researchers have used lumped models to quickly estimate acoustic output, then confirmed the predictions with full finite element simulations that account for the actual geometry of the vibrating membrane and the acoustic coupling to the ear canal or open air.11Microelectronics Reliability. A piezoelectric MEMS loudspeaker for in-ear and free field applications lumped and finite element models
Designing for Reliability
MEMS devices in cars, phones, and medical implants need to survive years of continuous use, which means reliability is not an afterthought. The failure modes differ from those of purely electronic chips because MEMS have moving parts subject to mechanical wear, fatigue, and the stiction problem mentioned earlier.
Fatigue is a particular concern for devices that vibrate or cycle millions of times, like inertial sensors and resonators. In polysilicon MEMS, failure under repeated loading can occur within the moving structure itself or at the anchor points where the polysilicon film bonds to the underlying silicon dioxide. Research has explored debonding mechanisms at this interface and proposed geometry optimization strategies for on-chip test structures that help characterize anchor strength before devices go into production.12PubMed Central. Fatigue-Induced Failure of Polysilicon MEMS: Nonlinear Reduced-Order Modeling and Geometry Optimization of On-Chip Testing Device The idea is to build dedicated test devices right on the wafer alongside the product devices, subject them to accelerated cycling, and use the results to validate the main design’s lifetime predictions.
Packaging is another reliability battleground. A MEMS accelerometer needs a hermetic cavity to keep out moisture and particles, but a MEMS pressure sensor needs an opening to the outside world. A MEMS microphone needs an acoustic port but not a leak path for water. Each application demands a different packaging strategy, and package-related failures (seal leaks, wire bond fatigue, thermal mismatch stresses) account for a large share of field returns.
Applications That Drive Design Choices
The application dictates nearly every design trade-off. A MEMS accelerometer for a car airbag has very different requirements from a MEMS microphone in a hearing aid, even though both are sensors. Airbag accelerometers need to survive harsh shock environments and trigger reliably in a crash, while hearing-aid microphones need ultra-low noise and tiny footprints. The design philosophy, materials, actuation type, and packaging are all chosen to fit the use case.
Optical MEMS devices, like digital micromirror devices used in projectors, illustrate how specialized the design can get. Each micromirror is an individually addressable actuator that tilts to direct light toward or away from a projection lens. These devices have been shown to outperform liquid crystal spatial light modulators in speed, spectrum sensitivity, and polarization modulation, making them attractive for applications well beyond cinema projectors, including scientific instruments and adaptive optics.13Annalen der Physik. Tailoring light with a digital micromirror device
In radio-frequency communications, MEMS switches offer reconfigurability that solid-state switches struggle to match. Designers have used RF-MEMS switches to build tunable bandpass filters for 5G and satellite systems, where the switch transitions between states to shift the filter’s passband.14Engineering, Technology & Applied Science Research. A Low-Profile Reconfigurable Wide Band BPF with RF-MEMS Switches for 5G/ Satellite Applications The advantage of a MEMS switch over a transistor switch at RF frequencies is lower insertion loss and better isolation, though MEMS switches are slower to actuate and need higher driving voltages.
Harvesting Energy from Vibrations
One of the more ambitious MEMS design challenges is building devices that scavenge energy from ambient vibrations to power themselves, eliminating the need for batteries. These vibrational energy harvesters typically use a proof mass on a spring that oscillates in response to external motion, converting kinetic energy into electrical energy through piezoelectric, electrostatic, or electromagnetic transduction.
The design challenge is that real-world vibrations are often at low frequencies and low amplitudes, which makes power output painfully small. An electrostatic MEMS energy harvester with a comb structure designed to reduce air damping demonstrated a maximum power output of about 13 microwatts at its best operating point, and a maximum single-cycle energy of 0.45 microjoules.15Microsystems & Nanoengineering. A power supply module for autonomous portable electronics: ultralow-frequency MEMS electrostatic kinetic energy harvester with a comb structure reducing air damping That is enough to trickle-charge a sensor node but nowhere near enough to run a smartphone. To squeeze more power from weak vibrations, researchers have developed design strategies that deliberately increase the coupling between mechanical motion and electrical output, even at the cost of temporarily reducing output, in order to reach a higher-performance operating point under realistic excitation conditions.16PubMed Central. MEMS vibrational energy harvesters
The practical vision is a maintenance-free wireless sensor that harvests enough energy from machinery vibrations, bridge traffic, or even body motion to wake up periodically, take a measurement, transmit a data packet, and go back to sleep. Getting there requires co-designing the mechanical resonator, the transduction element, the power conditioning circuit, and the communication protocol as one integrated system rather than optimizing each in isolation.
Pushing Toward Nanoscale Resonators
While most commercial MEMS operate with feature sizes in the range of a few micrometers to tens of micrometers, research continues to push dimensions smaller. Nanoelectromechanical systems (NEMS) shrink critical dimensions below one micrometer, which brings higher resonant frequencies, lower masses, and extreme sensitivity to external forces. A demonstration of a silicon-based resonant body transistor achieved acoustic resonance at 11.7 GHz with a quality factor of 1,830, marking the highest-frequency acoustic resonance measured on a silicon wafer at the time of its publication.17Nano Letters. The resonant body transistor That device embedded a transistor directly into the resonating body itself, eliminating the need for a separate sense electrode and sidestepping signal losses that plague conventional capacitive detection at such high frequencies.
Devices like this point toward a future where MEMS and NEMS resonators could replace bulky quartz crystals and surface-acoustic-wave filters in communication systems, offering smaller footprints and direct integration with CMOS circuits. The design challenges at this scale are formidable: thermal noise becomes a larger fraction of the signal, fabrication tolerances tighten dramatically, and surface effects that were merely annoying at the microscale become performance-limiting at the nanoscale. But the payoff in sensitivity and frequency range keeps researchers pushing the boundaries of what can be built and measured on a chip.