Using photoresist as an etch mask

1 · What an etch mask actually does

An etch mask is a patterned, sacrificial layer: the resist stays where you want the substrate protected and is cleared where you want the substrate attacked, so the pattern in the resist transfers into the material below. A general-purpose positive resist is the usual starting point — for example the S1813 recipe is a classic DNQ–novolak etch-mask workhorse, and the ma-P 1215 recipe is a thin wet-etch mask whose bake and exposure are published outright in the literature.

The resist often masks an intermediate hard layer — Zhang et al. 2020 pattern ma-P 1215 to define a chromium electrode mask, then wet-etch the chromium through it — or opens windows straight through a stacked film, as when Arulanandam et al. 2022 use a published S1818 recipe to open vias through a dielectric rear-mirror stack. Either way the resist's only requirement is to still be there, with its edges intact, when the etch finishes.

Sources: Zhang et al. 2020; Arulanandam et al. 2022

2 · Wet versus dry etch: what each demands

A wet etch immerses the substrate in a liquid etchant that attacks the exposed material chemically. It is isotropic — it etches sideways as fast as it etches down — and it is unforgiving of poor adhesion, because the bath creeps under any edge the resist hasn't bonded to and lifts the mask. So a wet-etch mask lives or dies on adhesion: a clean, dehydrated surface and, on oxide, an HMDS prime (the adhesion section of the guidelines explains why).

A dry (plasma) etch removes material with reactive ions in vacuum and can be made highly anisotropic — etching down far faster than sideways — which is how deep, vertical features are cut. The demand it places on the mask is selectivity: the plasma erodes the resist too, so the mask has to etch slower than the substrate for the whole depth. That is why deep etches reach for thick or hardened films — Bernardin et al. 2018 coat AZ 12XT at 15–18 µm over HMDS as a deep-RIE mask, and Schermer et al. 2022 use thick AZ 10XT under projection lithography for MEMS etch masks.

Negative resists serve as plasma masks too: Gerlt et al. 2021 use a 7.2 µm AZ nLOF 2070 film as a DRIE mask. The AZ nLOF 2070 recipe publishes a 174–198 mJ/cm² exposure at 365 nm for its 5.3–11.8 µm range, and for the thickest, most etch-resistant masks the KMPR 1050 recipe coats 34–115 µm as a strippable, DRIE-compatible epoxy mask.

Sources: Bernardin et al. 2018; Schermer et al. 2022; Gerlt et al. 2021

3 · Sidewalls, undercut and hard bake

The resist edge shapes the etched edge. In an anisotropic dry etch a near-vertical resist sidewall reproduces as a near-vertical wall in the substrate, so the profile you develop is roughly the profile you get. In an isotropic wet etch the etchant undercuts the mask by design — it eats sideways beneath the resist edge — so the etched opening ends up wider than the patterned opening by roughly the etch depth. Plan for that undercut: a wet-etched feature must be drawn narrower than its target, and features spaced too tightly will merge as their undercuts meet.

Hard bake is the lever for a wet etch. After development, baking the patterned resist hotter than the soft bake drives off residual solvent and reflows the film slightly, hardening it and improving how tightly it grips the substrate so the bath can't lift it. It costs a little edge sharpness, which is why it is standard for a long wet etch and often skipped before a short plasma etch where edge fidelity matters more. The hard-bake section of the guidelines covers the trade-off.

4 · Sizing thickness to the etch budget

The mask has to be thicker than the etch will consume, with margin. Estimate the etch depth, divide by the resist-to-substrate selectivity for your chemistry, and add headroom for non-uniformity — that number is the minimum film you can coat. A thin, high-resolution film like the AZ ECI 3012 recipe (a chemically amplified etch/RIE-mask resist at 1.05–1.9 µm) is right for a shallow etch of fine features; a thick positive resist like the AZ 9260 recipe, which doubles as an etch mask and a plating mould at 7.9–11.4 µm, suits a long or deep etch where the mask has to outlast a lot of material.

Don't over-coat blindly: a thicker film resolves a coarser feature and needs more dose and longer bakes. Read the thickness for a given spin speed off the resist's spin curve on its recipe page, pick the thinnest film that still survives the etch, and browse the library filtered to etch mask to narrow the field.

5 · The process, step by step

The full flow is eight steps. Every value for your resist — spin speed, bake temperatures and times, dose, developer — lives on its recipe page; treat the list below as the skeleton the numbers hang on.

• —Clean and prime. Start from a clean, dry substrate; dehydration-bake it and apply an HMDS adhesion prime where the resist calls for one — non-negotiable before a wet etch.
• —Coat. Spin-coat to the film thickness your etch budget demands, read from the resist's spin curve.
• —Soft-bake. Bake on a hotplate to drive off the casting solvent so the film is dry and stable before exposure.
• —Expose. Deliver the resist's working dose at your wavelength — for a straightforward mask the S1813 recipe publishes a sizing energy of 150 mJ/cm² at the g-line (436 nm), 1.3× its 82 mJ/cm² dose-to-clear. On a maskless tool this is a direct write from your layout: Wilke et al. 2022 pattern an S1822 micropillar etch mask on an MLA150 maskless aligner.
• —Develop. Develop for the specified time to clear the pattern; under-develop and residual film blocks the etch, over-develop and features thin.
• —Hard-bake. For a wet etch, hard-bake to harden the film and lock down adhesion before the substrate meets the bath.
• —Etch. Run the wet or dry etch, watching that the mask isn't consumed faster than expected; it reaches target depth where the resist has cleared and stops where it hasn't.
• —Strip. Remove the remaining resist in solvent (or an oxygen plasma / resist stripper), leaving the etched substrate. The strip section of the guidelines covers the options.

Sources: Wilke et al. 2022

6 · Failure modes

Most etch-mask problems show up either at develop or partway through the etch, and each points back at a step above.

• —Mask lifts or undercuts badly in a wet etch → an adhesion failure: revisit the clean, dehydration bake, HMDS prime and hard bake. The bath is getting under the resist edge.
• —Mask erodes before the etch finishes (dry etch) → too little film for the selectivity, or an over-baked/under-cured resist. Coat thicker, or move to a more etch-resistant resist.
• —Etched features come out too wide → in a wet etch this is expected undercut — draw features narrower; in a dry etch it points to over-development or an over-dose widening the opening.
• —Substrate not fully cleared where it should etch → resist scum in the opening: raise dose or develop time within the recipe window, or refresh the developer.
• —Ragged, rounded edges after hard bake → the hard bake ran too hot or too long and reflowed the film; back it off, or skip it for a short plasma etch.

As with any first run, expose a short dose array — the same feature at stepped doses — to find the dose that clears cleanly on your own tool before committing the pattern to an etch.

Common questions

Where to next

1. Schermer et al.. High-resolution projection lithography for MEMS-applications using thick photoresist AZ 10XT. 2022 Smart Systems Integration (SSI) (2022). doi:10.1109/SSI56489.2022.9901437
2. Bernardin et al.. Demonstration of a Robust All-Silicon-Carbide Intracortical Neural Interface. Micromachines (2018). doi:10.3390/mi9080412
3. Gerlt et al.. Reduced Etch Lag and High Aspect Ratios by Deep Reactive Ion Etching (DRIE). Micromachines (2021). doi:10.3390/mi12050542
4. Zhang et al.. Linked optical and gene expression profiling of single cells at high-throughput. Genome Biology (2020). doi:10.1186/s13059-020-01958-9
5. Wilke et al.. Turning traditionally nonwetting surfaces wetting for even ultra-high surface energy liquids. Proceedings of the National Academy of Sciences (2022). doi:10.1073/pnas.2109052119
6. Arulanandam et al.. Fabrication methods for high reflectance dielectric-metal point contact rear mirror for optoelectronic devices. MethodsX (2022). doi:10.1016/j.mex.2022.101898