What it is
Lift-off underlayers such as LOR and PMGI are non-photoimageable polymer coatings spun beneath a conventional imaging resist; they are never exposed themselves, but develop faster and sideways than the resist above them, carving a re-entrant undercut that breaks an evaporated metal film cleanly for lift-off.
How it patterns
- Substrate
- Underlayer
- Resist
- Exposed
A non-photoimageable layer beneath the pattern
LOR and PMGI are polydimethylglutarimide-based polymer coatings, spin-coated onto the substrate and soft-baked first. A conventional imaging resist is then spin-coated, exposed and developed on top of them. Neither LOR nor PMGI is itself exposed in the standard bilayer lift-off flow — the underlayer plays no photochemical role; it exists only to sit beneath the pattern and control what happens to it during develop.
Isotropic dissolution builds the undercut
The same develop step that defines the top resist's pattern also reaches the underlayer beneath it and dissolves it isotropically — sideways as well as downward — at a rate governed chiefly by the underlayer's own soft-bake temperature, not by the top resist's exposure dose; develop time then sets how far that sideways dissolution reaches. Because the underlayer clears laterally faster than the imaging resist's edge retreats, development leaves a re-entrant opening: wider at its base than at its top.
Why the overhang enables lift-off
When metal is evaporated over this profile, the overhanging lip of imaging resist breaks the film's continuity: metal lands on top of the resist stack and, separately, on the substrate inside the opening, with no continuous sidewall coating connecting the two. Soaking the stack in a dedicated stripper then dissolves the underlayer and imaging resist together and lifts away the unwanted metal cleanly, leaving only the metal that landed directly on the substrate.
Bake temperature, not exposure, is the process lever
Because the underlayer is never exposed, the undercut is tuned by its soft-bake temperature and time, which set how fast it dissolves, and by develop time, which sets how far the undercut reaches. The imaging resist's dose does not affect it.
What each step does
- 01 Coat the underlayer
LOR or PMGI is spun first. Its thickness sets how tall the gap under the overhang will be, and that gap has to be taller than the metal you deposit so the metal on the substrate never joins the metal on top.
Adjusting it: Choose the grade and speed on the spin curve to make the underlayer clearly thicker than the planned metal. Remove the edge bead as you would for any resist.
- 02 Bake the underlayer
Drives out solvent and also sets how fast the underlayer will dissolve in developer. A hotter bake gives a slower dissolution rate.
Adjusting it: This is the main control over undercut. A cooler bake gives a faster, larger undercut; a hotter one a smaller undercut. Because hotplate drift shows up as undercut drift, hold the temperature and time tightly once the process is tuned.
- 03 Coat the imaging resist
An ordinary imaging resist, usually a DNQ-novolak positive resist, is spun on the baked underlayer and soft-baked on its own schedule.
Adjusting it: Coat and bake it exactly as its own recipe says. Its thickness and resolution are those of the resist on its own.
- 04 Expose the imaging resist
Only the top layer is imaged. The datasheet says LOR and PMGI need no exposure step in the standard bilayer process.
Adjusting it: Dose it as you would the imaging resist alone. The top opening it defines is the size of the deposited feature. It has no effect on the undercut.
- 05 Develop
The developer clears the exposed top resist, reaches the underlayer and keeps dissolving it sideways as well as down, which forms the overhang.
Adjusting it: Once the bake is fixed, develop time sets how far the undercut reaches, and a longer develop gives a wider ledge. Stop before narrow resist lines lose support and lift.
- 06 Deposit and lift off
Evaporated metal lands on top of the stack and, separately, on the substrate inside the opening. A remover then dissolves both layers and takes the unwanted metal with them.
Adjusting it: Use a directional deposition, such as evaporation, and keep the metal thinner than the underlayer. Sputtered or over-thick metal joins across the gap and tears at the edge on lift-off.
When it goes wrong
Why does lift-off fail when the underlayer is thin?
Cause: If the underlayer isn't clearly thicker than the metal, the metal on the substrate joins the metal on top of the stack across the sidewall, and the break the overhang is meant to create never forms.
Fix: Coat the underlayer comfortably thicker than the planned metal, and re-check the margin whenever the metal thickness changes.
Sources: Kayaku, LOR and PMGI resists datasheet, Coating process; UCSB Nanofab wiki, lift-off with a PMGI underlayer
Why does LOR trap bubbles or coat unevenly over trenches?
Cause: The standard LOR A and B grades aren't formulated for coating into trenches or over deep topography.
Fix: Use an LOR C grade, which the manufacturer recommends for coating into trenches and over topography without trapping bubbles.
Source: Kayaku, LOR and PMGI resists datasheet, Coating process
Why does acetone edge-bead removal damage the underlayer edge?
Cause: Acetone and ordinary resist edge-bead removers aren't compatible with LOR and PMGI and aren't recommended for clearing their edge bead.
Fix: Use the edge-bead remover formulated for LOR and PMGI (EBR PG). LOR C grades also leave fewer whiskers on thick coats.
Source: Kayaku, LOR and PMGI resists datasheet, Edge bead removal
Why does PMGI lose adhesion on a reworked substrate?
Cause: A substrate that was coated and stripped before behaves differently: the manufacturer notes that PMGI soft-baked below the temperature its selection guide names for rework can lose adhesion there, where a fresh substrate would be fine.
Fix: On rework, soft-bake PMGI at or above the minimum temperature the selection guide gives for reworked substrates.
Source: Kayaku, LOR and PMGI resists datasheet, Product selection guide, advice note 2
Why is the undercut different at the wafer edge and the centre?
Cause: The bake temperature has the largest effect on undercut rate, so any temperature variation across the hotplate shows up directly as uneven undercut.
Fix: Check and calibrate the hotplate's uniformity, and make sure the wafer sits in even contact with it, since contact matters as much as the setpoint.
Source: Kayaku, LOR and PMGI resists datasheet, Soft-bake / prebake process
Why does the imaging resist mix with the underlayer?
Cause: Imaging resists cast in ethyl lactate can mix with an underlayer that was baked cool. The manufacturer names this solvent as the exception to the no-intermixing rule.
Fix: Bake the underlayer toward the hot end of its range before coating an ethyl-lactate resist, or use a PGMEA-based imaging resist, which carries no such caveat.
Source: Kayaku, LOR and PMGI resists datasheet, Product selection guide, advice note 3
Why does the imaging resist lift, flap or collapse after develop?
Cause: The undercut went too far, and the underlayer has dissolved away from under narrow or isolated features that no longer have anything holding them up.
Fix: Reduce the undercut: bake the underlayer hotter, which is the main control, or shorten the develop. For fine, closely spaced features, choose a slower-dissolving grade.
Sources: Kayaku, LOR and PMGI resists datasheet, Soft-bake / prebake process; Harvard CNS, SOP 112 lift-off resist processing, Fig. 2
Why is the undercut hard to control with my developer?
Cause: A full-strength TMAH developer dissolves some grades fast enough that a small, repeatable undercut is hard to hold.
Fix: Move to a weaker metal-ion-free developer, or to a metal-ion developer where compatible, to slow the undercut and make it steadier.
Source: Kayaku, LOR and PMGI resists datasheet, Product selection guide, advice note 1; development process
Why is there a hazy residue after developing the underlayer?
Cause: The underlayer leaves no scum only when it is baked properly and doesn't intermix. A short or cool bake, or a developer strength or develop mode that doesn't suit the grade, leaves residue around the undercut.
Fix: Check that the underlayer bake was complete and that the developer strength and immersion or spray mode match the manufacturer's recommendation for that grade.
Source: Kayaku, LOR and PMGI resists datasheet, Features; substrate preparation
Why are there metal ears or fences at the edges after lift-off?
Cause: The undercut was too small, so the deposited metal stayed continuous down the sidewall instead of breaking cleanly, and it tore when the stack lifted.
Fix: Increase the undercut by baking the underlayer cooler or developing longer, so the gap under the overhang guarantees a clean break.
Source: Kayaku, LOR and PMGI resists datasheet, Process flow, fig. 4
Why won't the underlayer strip in Remover PG after deposition?
Cause: The underlayer strips more slowly the hotter it was baked, and heat picked up during deposition hardens it further.
Fix: Run the remover warmer and soak longer for underlayers that were baked hot or saw a lot of process heat, and add ultrasonic agitation to help release the stack.
Source: Kayaku, LOR and PMGI resists datasheet, Removal rate of PMGI in Remover PG
Why does lifted metal settle back onto the wafer?
Cause: A single stripper bath fills with dissolved resist and loose metal flakes, which resettle on the substrate as the bath ages or the soak runs long.
Fix: Strip in two tanks, a bulk tank followed by a clean one, and demote the first tank once it is loaded.
Sources: Kayaku, LOR and PMGI resists datasheet, Lift-off process; Kayaku, Remover PG datasheet
Where it fits
- For lift-off patterning of evaporated metal, an LOR/PMGI underlayer beneath a conventional imaging resist builds the re-entrant undercut that a straight-walled single-layer resist profile does not reliably give.
- For process simplicity, these underlayers develop in the same TMAH-based developer as the imaging resist coated on top, so the bilayer clears in one develop step rather than a separate chemistry.
- For tuning the undercut, the underlayer's own soft-bake temperature is a dedicated process lever that sets its dissolution rate, independent of the imaging resist's exposure dose; develop time then sets how far the undercut reaches.
- For direct patterning, LOR and PMGI are not designed to be exposed and developed on their own — they carry no photospeed or resolution spec, since the pattern is defined entirely by the imaging resist coated above them.
- For a single-layer process, adding an underlayer means an extra coat, an extra bake, and a dedicated stripper step versus a resist that lifts off on its own.
- For dialing in a specific undercut geometry, published bake-temperature-vs-undercut-rate curves are not available for every LOR/PMGI grade, so datasheets recommend a small bake matrix per grade rather than trusting one stated condition.
Datasheet-cited recipes
Every value on these pages is cited to the manufacturer’s datasheet — spin curve, bake schedule, exposure dose and developer.
Further reading
- B. Cord, C. Dames, K. K. Berggren, J. Aumentado. Robust shadow-mask evaporation via lithographically controlled undercut. Journal of Vacuum Science & Technology B: Microelectronics and Nanometer Structures Processing, Measurement, and Phenomena (2006). doi:10.1116/1.2375090
- Y. Chen. A lift-off process for high resolution patterns using PMMA/LOR resist stack. Microelectronic Engineering (2004). doi:10.1016/s0167-9317(04)00111-x
Common questions
Why do I need an underlayer for lift-off?
A single, straight-walled imaging resist coats the sidewall of a deposited film continuously, leaving no break for a stripper to attack. An underlayer beneath the imaging resist develops faster and isotropically, carving an overhanging profile that breaks the metal film's continuity so lift-off works cleanly instead of tearing.
LOR vs PMGI — what's the difference?
Both are polydimethylglutarimide-based underlayers from the same manufacturer datasheet family, used the same way: spin-coated beneath an imaging resist, never exposed in the standard bilayer flow, and dissolved isotropically in develop at a rate set mainly by soft-bake temperature. They differ mainly in product-line grading.
What controls the size of the undercut?
Two things: the underlayer's own soft-bake temperature and time, which set how fast it dissolves, and the develop time, which sets how far the undercut reaches. The top resist's exposure dose does not affect it, since the underlayer is never itself exposed. Because bake-vs-undercut-rate curves aren't published for every LOR/PMGI grade, a small bake matrix per grade is the recommended way to dial in a target profile.
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.
General photolithography reference material, not a specification of any particular NANYTE BEAM configuration. Product names and trademarks belong to their respective owners; NANYTE is not affiliated with the manufacturers mentioned.