Building a Photoresist Spin Curve: Thickness vs Spin Speed

A photoresist spin curve shows how coated resist thickness varies with spin speed. It is built by coating identical substrates at different controlled speeds, measuring each film thickness by stylus profilometry or ellipsometry, and fitting the thickness-versus-speed data. A validated curve lets process engineers pick spin speeds that hit target thickness and qualify photolithography processes.

Principle & mechanism

Spin coating turns a liquid resist into a uniform thin film through two coupled actions: centrifugal flow and solvent evaporation. A measured volume of resist is dispensed onto a stationary or slowly rotating substrate. When the chuck accelerates to the set spin speed, viscous shear spreads the liquid outward. Surface tension resists film thinning. At the same time, solvent evaporation raises the resist viscosity, which slows further flow. After a few seconds the film becomes effectively solid and its thickness stabilizes.

The final thickness depends on spin speed, spin time, resist viscosity and solids content, solvent volatility, and ambient conditions. In practice, thickness decreases as spin speed increases. Centrifugal and evaporative thinning both intensify at higher angular velocity. The spin curve records this relationship quantitatively. The mechanism also matters for troubleshooting: radial streaks usually indicate improper acceleration, while thickness drift across a batch often traces back to solvent evaporation from an open resist bottle or uncontrolled humidity.

Sample

preparation and substrate cleaning
Substrate condition directly governs adhesion and film uniformity, so cleaning precedes any coating trial. Silicon wafers, glass, or quartz substrates are first inspected for particles and scratches under bright illumination. A standard cleaning sequence uses solvent rinsing, for example acetone followed by isopropanol. Then comes deionized water rinsing and drying with filtered nitrogen. An oxygen plasma treatment or dehydration bake removes residual organic contamination and adsorbed moisture. These residues otherwise cause dewetting and pinholes.

Where the resist datasheet specifies an adhesion promoter, it is applied immediately before coating in the same controlled manner for every sample. For curve construction, all substrates must be identical in material, size, and surface finish, because edge behavior and wetting differ between wafer diameters. Each substrate is labeled with its intended spin speed before coating so the measurement data stay traceable. Skipping or varying the cleaning step introduces scatter. This appears later as false non-repeatability in the fitted curve, and the error is hard to isolate once coating is complete.

Spin-coating

procedure and parameter control
The coating run follows a fixed sequence: static dispense, spread step, high-speed thinning, and final constant-speed hold. A typical protocol dispenses a defined resist volume at the wafer center. It spreads the resist at a low intermediate speed, ramps to the target speed within a set acceleration, and holds it for a defined duration before deceleration. For a spin curve, every parameter except spin speed is held constant across the sample set. Only the target speed is varied, in discrete steps over the intended working range.

Discipline in the smaller variables determines curve quality. Resist is filtered and allowed to reach ambient temperature before dispense. Bottle opening time is minimized to limit solvent loss. Exhaust flow over the bowl, ambient temperature, and relative humidity are recorded, because evaporation-driven thinning responds to all three. Edge exclusion and spin time stay identical for each sample. After coating, a soft bake on a hotplate at the resist-recommended temperature and time drives off residual solvent before measurement. A well-documented recipe sheet listing every setting allows any operator to reproduce the run.

Thickness

measurement — stylus profilometry and ellipsometry
Two complementary techniques cover thickness determination.

Stylus profilometry requires a patterned step. A portion of the film is removed with a solvent-swabbed or plasma-etched window, and the stylus scans across the step edge. The vertical displacement between the film surface and the bare substrate gives the thickness directly. The method is destructive at the step location and needs careful leveling. Its accuracy depends on stylus force and step cleanliness, but it applies to most film–substrate combinations without optical modeling.

Ellipsometry is non-contact and model-based. Polarized light reflects from the film. The measured change in polarization is fitted to an optical model containing refractive index and thickness. Single-wavelength instruments suit single-layer resists of known optical constants. Spectroscopic ellipsometry resolves multilayer stacks and refines the model at the same time. Both techniques benefit from calibration on certified step-height or thickness standards before use. Measuring several points per wafer, including center and mid-radius positions, characterizes within-wafer uniformity. The mean value at each spin speed becomes one data point on the spin curve.

Spin

curve fitting and repeatability assessment
Thickness versus spin speed is conventionally fitted to a power law. Thickness is proportional to spin speed raised to a negative exponent. The exponent and a proportionality constant are determined by least-squares regression on log-transformed data. The exponent characterizes the resist–process combination. Deviations from a straight line on a log-log plot often indicate that the tested speed range crosses a transition between flow-dominated and evaporation-dominated behavior.

Repeatability is assessed before the curve is accepted for use. Each spin speed is replicated on separate substrates, ideally across different operators or days. The standard deviation of the replicate thickness values is compared against the process tolerance. Within-wafer uniformity and wafer-to-wafer variation are reported alongside the fitted curve. Points lying outside the fitted confidence band are investigated for cause rather than discarded silently. The accepted output is a curve equation with its valid spin-speed range, stated measurement uncertainty, and a repeatability statement. Together these define the accuracy of any thickness target selected from the curve.

Application

scenarios and process acceptance criteria
A validated spin curve is used whenever a photolithography step requires a specified resist thickness. Examples include wafer patterning, thin-film device fabrication, MEMS structuring, and packaging processes such as plating molds or lift-off layers. The curve lets an engineer read off the spin speed for a target thickness, or interpolate when the target falls between measured points. It also supports incoming inspection of new resist lots, where a reference coating run verifies that the delivered material matches the qualified thickness behavior.

Acceptance criteria follow the process requirement rather than a universal number. Typical criteria include agreement between measured thickness and target within a stated percentage tolerance. Also typical are within-wafer uniformity within a defined limit and fitted-curve residuals within measurement uncertainty. Documentation retained for audit includes substrate cleaning records, coating recipes, raw thickness data, calibration status of the measurement instruments, and the fitted curve with its valid range. Periodic requalification is scheduled after resist lot changes, equipment maintenance, or any drift observed in routine thickness monitoring. This keeps the curve representative of the actual process state.