Permanent photoresist as a MEMS structural material

1 · Two ways a resist enters a MEMS process

A photoresist can play two completely different roles in a MEMS process, and the choice decides everything downstream. In the sacrificial role the resist is temporary — it masks an etch, shapes a lift-off, or holds a released part until a final step dissolves it away, and nothing of the resist remains on the finished device. In the structural role the resist *is* the device, or part of it: the patterned, cross-linked film stays put as a mechanical member — a microfluidic wall, a membrane, a spacer, a moulded post — and must survive every later step.

Most resists are designed for the first role. Only a few — chiefly the epoxy negative resists such as SU-8, and the thick dry-film epoxies — are formulated to become a permanent structure. Deciding the role first is the whole point: a resist chosen as a mask can be stripped and reworked freely, while a resist chosen as structure commits you to its geometry and its chemistry for good.

2 · What “photoimageable structural material” means

Photoimageable structural material is shorthand for a material you can pattern with light and then keep as a load-bearing part — collapsing two process steps into one. A conventional structure is built by depositing a material and then patterning it with a separate resist mask and an etch; a photoimageable structural resist is deposited, exposed and developed, and the developed film is already the structure. There is no transfer etch and no second material.

This is why SU-8 became the canonical MEMS structural resist: Lorenz et al. established it as a low-cost negative resist that produces thick, high-aspect-ratio structures directly by near-UV lithography, with the developed epoxy serving as the mechanical part rather than a stencil. The trade you accept in return is that the structure now has the mechanical and chemical properties of a cross-linked organic film — not those of silicon, metal or glass.

Sources: Lorenz et al., J. Micromech. Microeng. 1997

3 · Permanence is a design constraint, not a nuisance

Permanence is the defining property of a structural resist, and it is best treated as a design constraint rather than an inconvenience. A fully cross-linked epoxy resist such as SU-8 is effectively unstrippable — the cross-linked network that gives it its mechanical strength and chemical resistance is exactly what resists every practical stripper. Plan the geometry so you never need to remove it, because in practice you cannot without aggressive plasma ashing or laser ablation that also attacks everything around it.

At the conceptual level that permanence buys real robustness: a cross-linked structural resist holds its shape through wet chemistry, moderate heat and mechanical handling, which is why it works as a channel wall or a released beam. But the properties are those of an organic polymer, not a ceramic or a metal — it carries a much larger thermal-expansion mismatch to a silicon substrate than an inorganic film such as silicon dioxide, and it softens well below the temperatures silicon tolerates. Match the material to the mechanical and thermal envelope the part will actually see.

If you need the strength of a permanent film but want the option to remove it, a strippable epoxy is the alternative: KMPR 1050 is a thick negative resist that Lee et al. used as an electroforming mould and that can be stripped after use, unlike SU-8. It buys back reworkability at the cost of some of SU-8's ultimate resistance.

Sources: Lee et al., J. Micromech. Microeng. 2008

4 · Multi-layer builds and sacrificial pairings

Real MEMS parts are rarely a single developed layer. A structural resist can be built up in stages, and it is frequently paired with a sacrificial layer that is removed at the end to release a moving or suspended element.

Two sacrificial pairings recur in the literature. Tatikonda et al. use AZ 15nXT as a sacrificial layer that stays chemically stable through the full SU-8 process, then dissolve it to free metallized multilayer SU-8 devices such as an electrospray-ionization chip — the sacrificial resist survives structural processing and only lets go on demand. Wang et al. spin a thin AZ 4533 layer, about 4 µm, beneath an SU-8 cap and dissolve it in acetone to release optically driven micro-gears. In both, a positive DNQ–novolak resist is the sacrificial partner and the epoxy negative is the structure.

Structure can also be stacked directly. Working with dry-film epoxy, Farjana et al. laminated SUEX sheets — a 40 µm base plus 200, 50 and 20 µm sheets — to reach a target height for a millimetre-wave waveguide, a worked example of combining several sheet thicknesses into one part without spin-coating. Dry-film lamination is often easier than repeated spin-coats when the build is very thick or spans open cavities.

Sources: Tatikonda et al., Micromachines 2018; Wang et al., Nature Communications 2025; Farjana et al., Micromachines 2021

5 · When not to use resist as the structure

Resist-as-structure is the right call when you want the geometry a lithography step gives you and the part can live with an organic polymer's properties. It is the wrong call when the structure must have the mechanical, electrical or thermal properties of the substrate itself — then the resist should go back to being a mask, and the structure should be transferred into silicon, oxide or metal.

The tell is in the device papers. Haus et al. use ma-P 1215 as the lithographic mask through which hydrofluoric acid opens an oxide to define doping windows in an SOI pressure sensor — the resist is stripped and the silicon carries the structure. Lale et al. pattern AZ ECI 3012 to define suspended silicon nanowires that are then transferred into silicon by reactive-ion etching for gate-all-around ChemFETs — again the resist is a temporary mask and the released silicon is the mechanical member. If your part needs single-crystal stiffness, a defined electrical behaviour or high-temperature stability, pattern the resist and transfer; don't keep it.

See high-aspect-ratio patterning for the deep-feature version of this problem, and the etch-mask and lift-off tutorials for the sacrificial-role flows.

Sources: Haus et al., Sensors 2021; Lale et al., Proceedings 2017

Common questions

Where to next

1. Lorenz et al.. SU-8: a low-cost negative resist for MEMS. Journal of Micromechanics and Microengineering (1997). doi:10.1088/0960-1317/7/3/010
2. Lee et al.. Fabrication of thick electroforming micro mould using a KMPR negative tone photoresist. Journal of Micromechanics and Microengineering (2008). doi:10.1088/0960-1317/18/5/055032
3. Tatikonda et al.. Sacrificial Layer Technique for Releasing Metallized Multilayer SU-8 Devices. Micromachines (2018). doi:10.3390/mi9120673
4. Wang et al.. Microscopic geared metamachines. Nature Communications (2025). doi:10.1038/s41467-025-62869-6
5. Farjana et al.. Dry Film Photoresist-Based Microfabrication: A New Method to Fabricate Millimeter-Wave Waveguide Components. Micromachines (2021). doi:10.3390/mi12030260
6. Haus et al.. Robust Pressure Sensor in SOI Technology with Butterfly Wiring for Airfoil Integration. Sensors (2021). doi:10.3390/s21186140
7. Lale et al.. Development of All-Around SiO2/Al2O3 Gate, Suspended Silicon Nanowire Chemical Field Effect Transistors Si-nw-ChemFET. Proceedings (2017). doi:10.3390/proceedings1040419