On 300 mm silicon substrates coated with positive-tone DNQ/novolak photoresist, the spin-coating operation creates an edge bead that can be 20–60% thicker than the target film within an annular band extending 3–5 mm inward from the wafer bevel. Frontside edge bead removal on integrated coat/develop tracks is performed with a solvent such as PGMEA or ethyl lactate, dispensed through a nozzle positioned 1–2 mm from the substrate edge while the wafer rotates at 1000–3000 rpm. The lower control limit for critical-layer thickness uniformity is frequently set at ±3% of the target value, and the edge exclusion zone is typically 2 mm or 3 mm depending on the lithography node and device yield structure. Thinning beyond process limits refers to a condition in which the transition between fully removed edge bead and intact full-thickness resist extends inward more than 0.8–1.5 mm beyond the programmed EBR boundary, or the resist film thickness in the exclusion-adjacent region falls below the lower specification limit. Film thickness maps are acquired by multi-angle spectroscopic ellipsometry using 49-point radial mapping calibrated with NIST-traceable 25 nm silicon dioxide reference wafers. On a 300 mm track operating at 120 wafers per hour, batch-to-batch variation in the post-EBR edge profile can exceed ±5 nm at the 2.5 mm radial coordinate, which is sufficient to shift the post-etch critical dimension outside the allowed process corridor when exposure dose offsets are not adjusted. The production problem is therefore not merely incomplete edge bead removal, but the creation of a solvent-affected zone that violates thickness uniformity after the EBR sequence and before softbake.

How Does Solvent Retention After Edge Bead Removal Drive Capillary Thinning Toward the Wafer Center?

When the EBR solvent stream contacts the edge bead, it dissolves resist and creates a low-viscosity saturated solution whose local surface tension and viscosity are significantly lower than the bulk resist film. The surface tension difference between PGMEA at approximately 28.5 mN/m and the resist solution at 32–35 mN/m at 25 °C generates a Marangoni stress that pulls solvent-laden resist inward from the edge. Capillary pressure at the receding contact line scales inversely with the radius of curvature of the solvent front; with a contact line radius below 10 µm, the pressure gradient can exceed 1 kPa, which is sufficient to move diluted resist over the non-removed film. If the wafer deceleration phase following EBR dispense is longer than 2–3 s, residual solvent has time to diffuse into the film and reduce interfacial viscosity, allowing centrifugal and capillary forces to extend the thinned zone. This mechanism becomes more severe when the EBR solvent has a lower surface tension than the photoresist solvent mixture and when the ambient exhaust creates a net inward airflow across the wafer edge. The thin solvent layer does not simply evaporate from the edge; it interacts with the underlying resist as a swelling front, partially dissolving the topmost resist and leaving a wedge-shaped thickness transition that can extend 0.5–1.5 mm beyond the intended EBR boundary. On production-scale EBR stations, the resulting radial thickness profile often shows a negative slope from the full-thickness region toward the edge, with the steepest gradient located between 2.0 mm and 3.5 mm from the wafer bevel. Published data for specific resist-solvent combinations in this exact geometry is limited, but the general capillary-flow mechanism is consistent with thin-film hydrodynamics and solvent penetration models used in antifoaming and solvent-based edge-cleaning operations.

Measured by cone-and-plate rheometry calibrated to ISO 3219, a typical novolak resist has a viscosity of 8–12 mPa·s at 25 °C and a shear rate of 20 s⁻¹. After mixing with 0.5–1.0 parts of PGMEA by volume, the viscosity of the edge solution can drop to 2–5 mPa·s, depending on the resist solids content and the solvent affinity for the novolak matrix. The lower viscosity increases the Reynolds number of the radially flowing film and reduces the shear stress required to entrain the partially dissolved interface. On a 300 mm wafer spinning at 1500 rpm, the centrifugal acceleration at the 150 mm radius is approximately 37,000 m/s²; at the 3 mm edge exclusion zone, the local acceleration is slightly lower but still sufficient to produce drainage velocities that thin the solvent-swollen layer. The boundary layer thickness for the film under these conditions is typically on the order of 30–100 µm, and solvent diffusion into the underlying resist can extend to 0.2–0.5 µm within 2 s. This depth of solvent penetration is sufficient to soften the resist enough that subsequent spin-off removes more material than the intended EBR width, creating a wedge-shaped thickness profile that violates the process limit. The post-EBR thinning is not corrected by the subsequent spin-off step because the thinned region is no longer the original edge bead but a solvent-softened portion of the functional resist film. The result is a local reduction in resist thickness that can exceed 10–30 nm for a 1.0 µm target film, which is beyond the ±3% process limit and can propagate into linewidth variation after exposure.

When Exhaust Flow and Nozzle Positioning Deviate from Edge Exclusion Specifications

Production-scale EBR stations on integrated coat/develop tracks typically position the EBR nozzle at a radial distance of 0.8–2.5 mm from the wafer edge, with a dispense volume of 0.5–3.0 mL/wafer and a spin speed of 1000–3000 rpm. The exhaust airflow across the wafer edge is maintained within 0.5–1.5 m/s in the EBR cup, measured by hot-wire anemometer and verified against cleanroom airflow requirements under ISO 14644-1:2015. When the exhaust flow is below 0.5 m/s, solvent vapor accumulates around the wafer bevel, reducing the evaporation rate and extending the solvent-resist interaction time. When the exhaust flow exceeds 1.5 m/s, the enhanced evaporation produces localized cooling and increases the surface tension gradient, which can intensify Marangoni-driven inward flow. Nozzle positioning errors greater than 0.3 mm from the tool recipe value can shift the EBR boundary inward by an equivalent amount, but the thinning zone can extend an additional 0.5–1.0 mm due to solvent creep. Production logs from 300 mm tracks show that a 0.5 mm inward nozzle shift combined with a 0.25 mL increase in dispense volume can create a broad edge transition zone that fails the radial thickness uniformity specification. Backside EBR rinses that overlap the frontside EBR zone by more than 0.4 mm also generate a solvent reservoir at the bevel, which can wick onto the frontside and thin the resist during the spin-dry phase. The table below summarizes the process parameters, their typical boundaries, and the associated edge-thinning risk.

The table values are representative ranges from production track configuration checks; specific edge-thinning magnitudes depend on the photoresist chemistry, wafer type, and EBR solvent formulation. Published data for the exact combination of novolak resist and PGMEA-based EBR on 300 mm wafers with measured post-EBR thickness profiles is limited, and fabs typically generate internal process capability data rather than public test reports. The use of standard measurement methods such as ISO 3219 for viscosity and ASTM D3539-11 for solvent evaporation rate provides a common basis for comparing EBR solvent behavior, but the final thickness deviation must be verified by ellipsometry on each production reticle set.

Vapour Pressure Gradients and Solvent–Resist Interdiffusion Limits

The vapour pressure of the EBR solvent, determined by Shell thin-film evaporation per ASTM D3539-11, controls the rate at which the solvent leaves the edge zone after dispense. PGMEA has a vapour pressure of approximately 3.1 mmHg at 20 °C, while ethyl lactate is near 2.3 mmHg and cyclohexanone is near 3.4 mmHg. A higher-vapour-pressure solvent can evaporate more quickly, reducing the time for solvent diffusion into the resist, but rapid evaporation also cools the edge and may induce water condensation at high relative humidity. Water droplets on the edge of the resist create additional capillary pressure and can locally dissolve or swell the resist, leading to a thinning pattern that is irregular rather than smoothly radial. Solvent–resist interdiffusion in the EBR-affected zone is governed by the Flory-Huggins interaction parameter and the effective diffusion coefficient of the solvent in the novolak matrix. For PGMEA in novolak at 25 °C, the diffusion coefficient is on the order of 10⁻¹²–10⁻¹¹ m²/s, which corresponds to a diffusion length of 0.5–2.0 µm over a 1–2 s contact time. This diffusion length overlaps with the top portion of a 1.0 µm resist film, meaning that even brief EBR exposure can soften a significant fraction of the edge film thickness. Residual solvent content after EBR but before softbake is typically 2–5% of film mass in the center of the wafer, but the EBR-affected edge can retain 8–12% if the spin-dry step is too short or the exhaust is suboptimal. The higher residual solvent content lowers the local glass transition temperature of the novolak film by 10–30 °C, which permits additional film shrinkage during the subsequent softbake and increases the measured thickness loss at the edge. This shrinkage is not recoverable by extending the softbake time because the polymer matrix has already consolidated and the remaining solvent is trapped in the film. The process limit for post-softbake thickness uniformity therefore includes both the immediate EBR erosion and the later solvent-loss shrinkage, which together can produce a total edge-thinning deviation of 15–40 nm for a 1.0 µm target film.

Thermal Annealing Cannot Fully Recover Edge Film Thickness Lost to EBR Solvent Erosion

Post-application softbake on a hotplate at 90–110 °C for 60–90 s removes much of the residual casting solvent and consolidates the photoresist film, but it does not restore thickness lost through EBR solvent erosion. The thinned edge region consists of a polymer film that has already been dissolved or partially removed from the substrate, and the remaining polymer has a lower molecular weight distribution due to solvent-induced chain scission in the novolak resin. Differential scanning calorimetry of resist films exposed to EBR solvent shows a broadened glass transition and a lower transition midpoint compared to unexposed resist, consistent with plasticization and reduced chain entanglement. The viscosity of the thinned edge region at the softbake temperature is insufficient to allow reflow back to the original thickness; the film yields locally rather than flowing radially. Production-scale hotplate baking with a ramp rate of 15–25 °C/min cannot generate the thermal gradient needed to drive polymer transport over the 0.5–1.5 mm distance of the thinned zone. The final edge thickness after softbake therefore remains below the lower specification limit, and the defect is transferred into the exposure step as a local dose-to-clear shift. The thinned edge region clears at a lower exposure dose than the bulk film, leading to linewidth narrowing or resist scumming in the edge exclusion area. The only effective correction is to prevent the EBR solvent from extending inward beyond the programmed boundary, which requires tight control of dispense volume, nozzle position, spin speed, and exhaust flow. Published data on thermal recovery of EBR-thinned positive photoresist films is limited, and the available physical characterization indicates that post-softbake reflow is negligible for novolak systems with a molecular weight above 10,000 g/mol.

Inline spectroscopic ellipsometry on a 300 mm production track, using 49-point radial mapping and 25 nm NIST-traceable reference calibration, reveals that the edge-thinning defect is most severe when the EBR solvent dispense volume is increased to compensate for incomplete edge bead removal on wafers with higher edge bead thickness. Edge bead thickness itself varies with resist viscosity, dispense volume during spin coating, and wafer backside contamination, with a batch-to-batch standard deviation of 3–7% of target thickness on production tracks. When the EBR recipe is fixed but the incoming edge bead thickness increases by 10%, a single EBR pass may leave a residual bead that prompts the operator or automated process control to add a second EBR pass. The second pass delivers additional solvent to the already-thinned transition zone, widening the edge deviation and pushing the thinned region further inward. On a 300 mm track with a 120 wafer per hour throughput, the delay between EBR and softbake is typically less than 60 s, but even this short interval allows the solvent to continue diffusing into the edge film. Wafers that wait longer than 120 s before softbake show a measurable increase in edge-thinning depth of 3–8 nm compared to wafers processed within the standard queue time. The interaction between EBR solvent, resist chemistry, and queue time is a production control variable that is not always captured in standard process qualification, and it represents an operational boundary that must be monitored for stable edge thickness performance.