Skip to content
NANYTE

SU-8 2050 process recipe

SU-8 2050 is the mid-range thick-film workhorse of the SU-8 2000 family — the grade most MEMS and microfluidics processes default to when they need tall, robust structures without stepping up to the very thickest SU-8 2100/2150 coats. SU-8 2050 is a high-viscosity (12,900 cSt, 71.65% solids) grade in Kayaku Advanced Materials' SU-8 2000 epoxy photoresist series, formulated for thick, high-aspect-ratio, permanent MEMS and micromachining structures. It crosslinks via UV-generated photoacid followed by thermally-driven epoxy crosslinking during post-exposure bake, yielding a chemically and thermally stable, mechanically robust film that is normally left on the device rather than stripped.

https://nanyte.com/photoresists/su-8-2050 · last updated 2026-07-12

At a glance
Manufacturer
Kayaku Advanced Materials
Tone
negative
Chemistry
Epoxy (SU-8 type)
Thickness
40.1–171 µm
Developer
SU-8 Developer (Kayaku's proprietary PGMEA-based developer); other solvent developers such as ethyl lactate or diacetone alcohol are also stated to work
Applications
MEMS structural · High aspect ratio · Microfluidics · Electroplating / molding

Unverified — not yet human-checked; values transcribed from the datasheet, characterize on-tool.

Coat40.1–171 µmSoftbake95 °C · 7.5 min · hotplateDevelop6 min
—01 / Coating

Spin coating

SU-8 2050 is spin-coated to 40.1–171 µm. The curve below is redrawn from the manufacturer's published data — read your target thickness off the vertical axis and take the matching spin speed as a starting point.

Spin curve for SU-8 2050: film thickness in µm against spin speed in rpm.0.00501001502001k2k3k4kSPIN SPEED (rpm)THICKNESS (µm)
Data points
SU-8 2050 — film thickness (µm) by spin speed (rpm)
Seriesrpmµm
SU-8 20501000171
200076
300053
400040

Values are the manufacturer's starting points, not a guarantee; verify on your own tool. Characterize on-tool. Series digitized from a published figure were independently cross-checked by a second blind read; treat those values as approximate (±10 %).

read from Figure 1 'SU-8 2000 Thickness vs. Spin Speed', p.2 of SU-8 2000 (2025-2100) Technical Data Sheet, Kayaku Advanced Materials, April 2021 — the chart plots four family curves (2025/2035/2050/2075) together with axis gridlines every 20 µm (0-240) and every 500 rpm (500-4500); no numeric table is published for this figure. ADJUDICATED 2026-07-12 (third pixel read, WL directive: figure over table): this figure is VECTOR-DRAWN, not a raster image — re-extracted directly from the PDF's vector marker paths (PyMuPDF get_drawings(), the small filled+stroked rects at each data point) with axes calibrated from the y-axis and x-axis tick-LABEL TEXT coordinates (exact, not eyeballed): y=0 at PDF-space y=619.5, y=240 at y=464.95 (1.5529 µm/pt); x=500rpm at x=48.9, x=4500rpm at x=249.2 (0.02 rpm/pt-inverse, i.e. 20.03 pt per 1000 rpm). The chart plots exactly FOUR markers per series, at 1000/2000/3000/4000 rpm only — the crawl-era raw read's 1500/2500/3500 rpm points were fabricated (not present in the source figure) and are dropped. SU-8 2050 identified as the second-from-top series (triangle marker, filled+stroked rect) by y-order at each x-column, consistent with its rank between SU-8 2075 (top) and SU-8 2035 (third); vector-derived values (171.17, 75.49, 53.03, 40.08 µm, rounded here to 3 sig figs) match the independent cross-check pixel read (171, 75.5, 53.0, 40.1 µm) to better than 0.2% at every point, confirming the crawl-era raw read (165/90/58/44, plus the fabricated 115/72/50 points) erred and the cross-check extraction was correct. Uncertainty on the vector-derived read is well under ±2% (exact marker geometry, not an eyeball estimate).

Redrawn from the manufacturer's published data — hover to read values between points, click to pin.

Kayaku's stated 'Recommended Program' (family-wide, not thickness-specific): dispense 1 ml resist per inch (25 mm) of substrate diameter; spin at 500 rpm for 5-10 s at 100 rpm/s acceleration (spread step); then spin at 2000 rpm for 30 s at 300 rpm/s acceleration as a generic starting point — the actual final rpm/time should be taken from Figure 1 for the desired thickness. Edge-bead removal (EBR) with a solvent stream (Kayaku EBR PG) at the wafer edge is recommended so the mask can seat in close contact. Secondary/practical (Cornell Nanoscale Facility SOP, university source, not the vendor): pour SU-8 2050-or-thicker from stock into small working bottles at least 24 h before spinning to let entrained bubbles dissipate; for layers ≥150 µm, let the wafer rest on the spinner after spin-up so hanging edge resist can retract back onto the wafer before wiping the edge bead, since a fully removed edge bead will just reflow into the gap during softbake.

Adhesion
HMDS not required — Adhesion promoters are stated as 'typically not required' for standard use. Substrates should simply be clean and dry — piranha wet etch (H2SO4 + H2O2) + DI rinse, or RIE / O2 barrel-ashing, is recommended for best results. The one exception: for applications involving electroplating, a substrate pre-treatment with HMDS IS recommended.
—02 / Bake

Soft bake

Soft bake
95 °C · 7.5 min · hotplate
Notes
Datasheet times are given per THICKNESS bracket, not per grade. For a nominal SU-8 2050 film in the 45-80 µm range (matching ~3000 rpm per the spin curve above), Table 2 specifies 0-3 min at 65°C followed by 6-9 min at 95°C (450 s = midpoint of the 95°C range used here; the 65°C step is optional/short). A level hotplate is required — convection ovens are explicitly not recommended, as a skin can form on the resist and trap solvent. Vendor's own optimization method: after the prescribed bake, remove the wafer and let it cool to room temperature, then return it to the hotplate; if the film wrinkles, continue baking a few more minutes and repeat the cool/reheat cycle until wrinkles disappear.

SOURCE: Table 2 'Soft Bake Times', p.3 of SU-8 2000 (2025-2100) Technical Data Sheet, Kayaku Advanced Materials, April 2021

—03 / Exposure

Exposure dose

The manufacturer does not publish a clearing dose for SU-8 2050. Determine it with a dose array on your own tool.

As published
Table 3 'Exposure Dose' gives dose vs. FILM THICKNESS (not grade): 25-40 µm -> 150-160 mJ/cm2; 45-80 µm -> 150-215 mJ/cm2; 85-110 µm -> 215-240 mJ/cm2; 115-150 µm -> 240-260 mJ/cm2; 160-225 µm -> 260-350 mJ/cm2; 230-270 µm -> 350-370 mJ/cm2. For a SU-8 2050 film in its typical ~45-80 µm working range (per the spin curve above, roughly 2500-4000 rpm) the applicable dose is 150-215 mJ/cm2.
Post-exposure bake
95 °C · 6.5 min

Not published for this resist: Dose at 365 nm, Dose at 405 nm — characterize on-tool.

—04 / Development

Development

Developer
SU-8 Developer (Kayaku's proprietary PGMEA-based developer); other solvent developers such as ethyl lactate or diacetone alcohol are also stated to work
Dilution
undiluted
Time
6 min
Method
immersion (spray or spray-puddle also usable per datasheet)
Rinse
Fresh SU-8 developer spray/wash ~10 s, optionally repeated, then air/N2 dry with filtered pressurized gas. A white film during an IPA rinse indicates under-development of the unexposed resist -- treat with more SU-8 developer and repeat, rather than relying on IPA to finish the job. Ultrasonic/megasonic agitation is recommended for high-aspect-ratio or tight-pitch structures.
Developer family
Solvent

SOURCE: Table 6 'Development Times for SU-8 Developer' (45-75 µm bracket: 5-7 min, 360 s used as midpoint), p.4 of SU-8 2000 (2025-2100) Technical Data Sheet, Kayaku Advanced Materials, April 2021

—05 / Post-processing

Hard bake, etch & strip

Etch resistance
Described as chemically and thermally very stable once crosslinked: thermal stability to 315°C (5% wt. loss), and the Removal section states it is 'extremely difficult to remove... with conventional solvent based resist strippers.' This robustness makes cured SU-8 usable as a durable etch/plating mold in many wet-chemical processes, at the cost of being hard to strip afterward (see stripper field).
Stripper
Minimally-crosslinked SU-8: Kayaku Remover PG, 50-80°C bath, 30-90 min immersion (swells and lifts off partially-cured material; if OmniCoat 30-100 nm sacrificial layer was applied first, Remover PG gives a clean full lift-off). Fully cured/hard-baked SU-8 CANNOT be removed with Remover PG at all unless OmniCoat was used underneath. To rework fully crosslinked SU-8: oxidizing acid solutions (piranha etch), plasma ash, RIE (200 W, 80 sccm O2, 8 sccm CF4, 100 mTorr, 10°C), laser ablation, or pyrolysis.
Storage
Store upright in tightly closed containers, in a cool dry environment away from direct sunlight, light, acids, heat, and ignition sources, at 40-70°F (4-21°C). Shelf life is 13 months from date of manufacture.

Not published for this resist: Hard bake, Descum — characterize on-tool.

SOURCE: Hard Bake (cure) section, p.5 of SU-8 2000 (2025-2100) Technical Data Sheet, Kayaku Advanced Materials, April 2021

—06 / Applications

Where it's used

MEMS structuralHigh aspect ratioMicrofluidicsElectroplating / molding

SU-8 2050 sits in the middle of the SU-8 2000 family by viscosity (12,900 cSt vs. 4,500 for 2025 and 22,000 for 2075) and covers roughly 45-165 µm in a single coat over the datasheet's plotted 1000-4000 rpm range (the family description states thicknesses down to 0.5 µm and beyond 200 µm are achievable across the whole SU-8 2000 line, implying lower speeds than plotted would push a single 2050 coat higher still, but that isn't shown numerically for this grade). Choose 2050 over the thinner SU-8 2025 when a structure needs more height and mechanical robustness, and over the much thicker SU-8 2100/2150 grades when their hour-long bakes and very deep develops would otherwise dominate the process. PEB is the step where crosslinking actually completes (thermally-driven, acid-catalyzed epoxy reaction) -- a visible latent mask image appearing within 5-15 s of starting the 95°C PEB is the datasheet's own go/no-go check for adequate exposure and heating. Thick films are prone to cracking and delamination from thermal stress; besides the softbake wrinkle-check cycle, a University of Cornell Nanoscale Facility SOP (secondary source, not the vendor) recommends gradual heating/cooling for any layer >=50 µm, since silicon cools faster than SU-8, and a ~10 minute wait between exposure and PEB start for full latent-image (photoacid) formation. Because fully cross-linked SU-8 is essentially unstrippable with ordinary solvents, any process that will need to remove or release the structure later should plan for an OmniCoat sacrificial layer from the start, not as an afterthought.

—07 / Sources

Sources & disclaimer

  • Kayaku Advanced Materials — SU-8 2050 datasheet (SU-8 2000, 2025-2100, Technical Data Sheet, April 2021 (per footer on every page; fetched via a university-hosted mirror of the same document, https://nanofab.sites.olt.ubc.ca/files/2026/01/KAM-SU-8-2000-2025-2100-Datasheet-4.9.21-final-2.pdf, since kayakuam.com returned HTTP 403 to automated fetches during this session)) · accessed 2026-07-10
  • https://www.cnfusers.cornell.edu/sites/default/files/Equipment-Resources/SU8%20processing%20suggestions.pdf — Cornell Nanoscale Facility 'SU-8 Processing Suggestions' (v2, July 2013) -- used only for practical process tips not in the vendor datasheet: pre-spin bubble-dissipation wait time for SU-8 2050+, the ~10 min exposure-to-PEB delay for latent image formation, and thermal-stress/cracking guidance for thick and multi-layer coats. Not used for any headline numeric spec (spin curve, dose, bake time/temp all come from the Kayaku datasheet).
Research using this resist
  1. Shaw et al.. Negative photoresists for optical lithography. IBM Journal of Research and Development (1997). doi:10.1147/rd.411.0081
    The IBM origin paper of SU-8.
  2. 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
    15:1 aspect ratios in 1997 — the paper that made SU-8 a MEMS resist.
  3. del Campo et al.. SU-8: a photoresist for high-aspect-ratio and 3D submicron lithography. Journal of Micromechanics and Microengineering (2007). doi:10.1088/0960-1317/17/6/R01
    The canonical SU-8 review.
  4. Qasaimeh et al.. Microfluidic probes for use in life sciences and medicine. Lab Chip (2013). doi:10.1039/C2LC40898H
    Free-standing SU-8 microfluidic probes for live-cell chemistry.

Manufacturer datasheet values are starting points; optimal parameters depend on your substrate, equipment and environment. Product names and trademarks belong to their respective owners. NANYTE is not affiliated with the manufacturers listed. Last updated 2026-07-12.

Cite this recipe

NANYTE. "SU-8 2050 process recipe." NANYTE Photoresist Library. https://nanyte.com/photoresists/su-8-2050. Accessed 2026-07-12.

Ready to run this process on your own tool?Book a demo →

Expose it at 365 and 405 nm

NANYTE BEAM is a desktop maskless lithography system with software-selectable dual-wavelength exposure and 16-bit grayscale — no photomask, no mask cost, same-day iteration.

Book a demo

Improve this recipe

Run this resist in your lab? Send us what actually works, or flag a value that's wrong. No account, no email needed — a handle and affiliation are optional and become your credit line. Every contribution is checked against published sources before it appears.