Storing and handling photoresist in a Bangladeshi cleanroom
Why a lithography recipe stops repeating, and how much of that traces back to the bottle rather than the aligner. Storage, spin curves, softbake, edge bead, developer control and the humidity problem that runs from June to September.
A student reproduces a lithography recipe from a published paper, gets features half a micron wider than expected, and spends three weeks on the aligner. The aligner is fine. The resist bottle has been on the bench under a window since March, the developer is being made up by eye, and the room was at one humidity in February and another in July.
Photoresist is the least respected consumable in a fabrication lab and the one that causes the most unexplained results. It is a chemical with a defined storage condition, a defined shelf life and a strong sensitivity to the environment it is processed in, and almost every reproducibility problem in a university cleanroom sits somewhere in that chain.
Temperature, light and the bottle itself
Store at the temperature on the datasheet, which for most resists means refrigerated and never frozen. A dedicated chemical refrigerator, not the one with the lunch in it, and not one that defrosts by cycling above the storage temperature.
Equilibrate before opening. A cold bottle opened in a Dhaka cleanroom condenses water on the inside of the neck and into the resist. Bring the sealed bottle to room temperature first, and for a full bottle that takes hours, not minutes. This single habit removes a whole class of coating defects.
Keep it in the dark. Amber bottles, closed cabinets, and yellow lighting in the coating and developing area. Check the filter sleeves on the light fittings after any maintenance visit: a maintenance team replacing a failed tube with a plain white one is a common and invisible cause of a facility suddenly losing contrast.
Never pour resist back into the bottle. Dispense what you need, and discard the excess. Returning it puts particles and absorbed water back into the stock.
Log the date opened, not just the date received. Shelf life on the label assumes an unopened container. Write the opening date on the bottle in marker, and treat an undated bottle as unknown stock.
Filter at the point of dispense and use a fresh pipette each time. Do not shake a bottle to mix it: shaking entrains air, and the bubbles come back as comets in the coated film.
Age shows up in specific ways rather than as general poor performance. Positive resists lose contrast and develop increased dark erosion, so unexposed areas thin during development. Adhesion gets worse. Small features foot or lift. If a process that used to work has slowly drifted over a year, the resist age is the first thing to check and the cheapest to eliminate.
Getting the spin curve to repeat
Film thickness falls roughly as the inverse square root of spin speed for a given resist at a given temperature, which is why every datasheet has a spin curve. Reproducing that curve on your own tool needs several things fixed that people leave floating.
Acceleration, not just final speed. The ramp determines how the resist spreads before it thins, and two recipes with the same final speed and different acceleration give different thickness and different edge behaviour.
Static or dynamic dispense, chosen and then kept. Dispensing onto a stationary substrate and dispensing onto a slowly turning one produce different results, and mixing the two between users is a common source of scatter in a shared facility.
Dispense volume and coverage. Too little resist gives radial streaks; too much is waste and can flood the chuck. Fix the volume in the written recipe.
Bowl extraction. Airflow over the spinning substrate controls solvent evaporation, which controls the final thickness. A spinner sharing an extract with something else, or with a damper that anybody can adjust, will not repeat. Set it, mark it, and check it.
Substrate temperature. A wafer taken straight off a hotplate or straight out of an air-conditioned store is not at the same temperature as the resist, and thickness follows.
Substrate preparation belongs in the same recipe. A dehydration bake followed by HMDS priming, preferably vapour priming rather than a puddle, is what makes resist stick to silicon and to oxide. When small features lift during development or undercut badly, adhesion is the usual culprit, and the fix is at the priming step rather than in the exposure dose.
Softbake, and why a hotplate beats an oven
The softbake drives solvent out of the film. A hotplate with vacuum contact heats from the substrate upwards, so solvent leaves through the free surface. A convection oven heats the top first and can skin the film, trapping solvent underneath. The trapped solvent then comes out during exposure or development and gives inconsistent results that look like a dose problem.
Two timings need to be fixed in the written procedure and then actually followed: the delay from spin to bake, and the delay from bake to exposure. Resist keeps losing solvent the whole time it sits on the bench, so a wafer exposed five minutes after bake and one exposed forty minutes after bake are not the same wafer. In a teaching lab where a group of students shares one aligner, that delay is often the largest uncontrolled variable in the whole process.
Edge bead, and what it does to contact printing
Surface tension leaves a raised rim of resist at the substrate edge. On a full wafer the bead can be several times the thickness of the film in the centre.
In contact or proximity printing that rim holds the mask off the substrate. The gap is largest in the middle, exactly where the pattern usually is, and the result is a wafer where the outer features print sharply and the centre is soft. Groups chase this for weeks as an exposure or a mask problem.
Edge bead removal with a solvent nozzle during the spin is the clean answer and is a specification point when buying a coater.
Manual removal with a solvent-wetted swab works on small substrates and in a teaching lab, provided it is done consistently and the swab does not shed.
Design around it by keeping the pattern away from the edge region, which is acceptable for research devices and free.
Check it rather than assume it. A quick measurement across the substrate on a profiler or a reflectometer shows the bead immediately, and it is worth doing once for every resist and thickness the facility uses.
Developer strength, temperature and agitation
Development is a chemical etch of the exposed resist, and etch rates depend on concentration, temperature and transport. All three get treated casually.
Use the concentration the resist datasheet specifies. Diluting developer to make it go further does not simply slow the process down, it changes the contrast and therefore the feature size. If a diluted developer is used deliberately, it is a different process and needs its own written recipe.
Control the temperature. Development rate is strongly temperature dependent. A beaker on a bench in a Dhaka February and the same beaker in June are two different processes. A temperature controlled bath is the proper fix; at minimum, measure and record the developer temperature with every batch.
Fix the agitation. Puddle development, immersion with agitation and immersion without agitation give different profiles. Choose one and write it down.
Decide the reuse policy. Developer is consumed as it works. Reusing a bath across a day drifts the process from the first wafer to the last. Fresh developer per run is the reproducible choice, and where that is too expensive, cap the number of substrates per bath and record it.
Rinse and dry consistently. A long rinse in DI water followed by a controlled nitrogen dry is part of the recipe, not an afterthought. Water spots and residual developer both show up as defects at inspection.
Humidity is the recurring problem here
From June to September the ambient dew point in Dhaka stays high for weeks at a time, and that is the environmental variable that most often defeats a Bangladeshi lithography process.
Air conditioning controls temperature, not moisture. A split unit or a chilled water coil sized for cooling will hold the room temperature and let the relative humidity wander. Controlling humidity needs either reheat after a deep cooling coil or a desiccant dehumidifier. If lithography is a core capability, that has to be in the air handling specification.
Primed surfaces lose their priming. HMDS makes a surface hydrophobic, and leaving a primed wafer in humid air undoes it. Prime and coat in the same session.
Chemically amplified resists are the most sensitive. They are affected by humidity and by airborne amines between exposure and post-exposure bake, and the classic symptom is a T-shaped profile at the top of the feature. Fixed, short delays and filtered air are the controls.
Water in the film changes thickness and adhesion. An unexplained shift in coated thickness between the dry season and the monsoon is normally moisture, not the spinner.
Log the room humidity with every batch. A logger costs very little and turns arguments into evidence. When a run fails, the humidity trace for that afternoon is often the answer.
A routine that survives student turnover
A research group loses its trained user every two years. The handling discipline has to live in the facility rather than in a person.
One written recipe per resist and thickness, stored on the tool and locked so that students cannot edit the master. Copies are for experiments; the master is not.
A resist log at the refrigerator: what came in, when, when it was opened, when it was discarded.
A control substrate at the start of each session, coated to a standard recipe and measured. An ellipsometer or a reflectometer gives a thickness in seconds, and a chart of that number over months detects a drifting process long before a project fails.
A named owner for the coater and the developer bench, with the authority to stop somebody using them.
An induction that includes the chemistry, not just the buttons. A user who knows why the bottle warms up before opening will do it when nobody is watching.