Protecting the EUV Collector

Abstract

The collector mirror sits closest to the tin plasma and carries the first severe lifetime trade in the EUV optical chain. It must accept radiation over a large solid angle, reflect the narrow 13.5-nanometer band toward the intermediate focus, survive heat and plasma exposure, and preserve figure and multilayer response while tin debris is generated nearby. I classify debris as fast ions, neutral vapor, clusters, droplets, and redeposited material, then match controls to each transport mechanism. Better droplet coupling reduces residue at the source. Magnetic fields influence charged species. Gas flow and geometric traps change neutral and particulate transport. Hydrogen chemistry can remove tin under controlled conditions. Sacrificial or replaceable elements protect selected surfaces. The performance metric is not initial reflectivity. It is useful intermediate-focus energy integrated between recoverable maintenance events, with optical quality and contamination state known well enough to schedule intervention before wafer throughput collapses.

Collection places the optic inside the experiment

The plasma emits radiation into many directions. The collector surrounds part of that emission and directs accepted light toward the scanner. Larger accepted solid angle increases useful power, but places more optical area in line of sight of the source.

If plasma radiates in-band energy $E_b$ with angular distribution $L(\theta,\phi)$, the collected energy can be screened as

$$E_c=\int_{\Omega_c}L(\theta,\phi)R(\theta,\lambda)\,d\Omega,$$

where $R$ is collector reflectivity over angle and wavelength. Contamination changes $R$ spatially, so the far-field distribution and intermediate-focus uniformity can degrade before average power alone reveals the problem.

The collector’s multilayer coating is tuned to EUV wavelength and incidence. Figure error shifts rays. Roughness scatters light. Deposits absorb or alter phase. Plasma heat and particles can damage the surface or coating interfaces. A few nanometers can matter, as ASML’s materials discussion makes clear.

Debris is not one species

Debris class	Transport	Likely consequence	Control family
Fast ions	Ballistic, charge-sensitive	Sputter, implant, roughen, damage coating	Magnetic fields, plasma optimization, gas
Neutral atoms and vapor	Ballistic and gas-collision transport	Deposit tin films	Gas flow, chemistry, source coupling
Clusters	Inertial and partly collisional	Deposit or impact	Traps, geometry, gas, target control
Droplets and fragments	High inertia	Local coating and surface damage	Target coupling, shields, geometric traps
Out-of-band radiation	Optical and thermal	Heat, unwanted exposure	Spectral filtering and coating response

Controls cannot be evaluated with one debris counter. A method that deflects charged ions may leave neutrals unchanged. A gas pressure that slows particles can absorb EUV or increase vacuum load. Hydrogen that supports cleaning can change surface chemistry elsewhere.

I therefore want species-resolved diagnostics and witness surfaces positioned through the vessel. Post-service analysis should return composition, thickness, roughness, morphology, and damage to the source model.

Prevent residue before intercepting it

The least expensive particle is the one never produced. Droplet size, pre-pulse shape, main-pulse coupling, alignment, and plasma condition determine how completely tin is converted and how matter leaves the interaction.

Source optimization should include collector cost:

$$J=\frac{\int P_{IF}(t)dt} {T_{maintenance}+C_{collector}+C_{consumables}+C_{downtime}}.$$

The units require normalization for a real economic decision, but the form makes the point. Maximum instantaneous conversion efficiency can be inferior if it accelerates irreversible damage.

Missed droplets deserve special treatment. Firing the main laser into an incorrectly shaped or positioned target can create unfavorable debris. Event-level detection may permit pulse suppression or a controlled recovery response, trading one missing photon event for optical protection.

PreventEfficient target coupling

Reduces large residual fragments and unfavorable plasma states.

DeflectMagnetic and electrostatic response

Redirects charged species before they reach critical optical area.

InterceptGas, shields, and geometric traps

Slows or captures neutral, cluster, and droplet transport.

RecoverHydrogen chemistry and cleaning

Removes reversible deposits before they become permanent loss.

Magnetic and gas controls have side effects

Charged particles follow trajectories influenced by magnetic fields. Their gyroradius is

$$r_L=\frac{mv_\perp}{|q|B}.$$

Slow, highly charged ions can be deflected more strongly than fast, massive, weakly charged species. Field geometry, plasma behavior, source components, and diagnostics must be included. A field that protects one collector zone can steer particles toward another surface.

Background gas creates collisions and changes particle energy or direction. It can also absorb EUV, conduct heat, load pumps, influence plasma, and alter chemistry. Flow topology may matter more than chamber-average pressure. Computational fluid and particle models need validation with local measurements.

Hydrogen is important because it can participate in tin-removal chemistry and surface conditioning. It also creates radical species and potential material compatibility concerns. Cleaning rate, selectivity, multilayer stability, redeposition, and byproducts must be understood across repeated cycles.

Cleaning must preserve the optic

A cleaning process can restore reflectivity and gradually damage the substrate or multilayer. The correct result is net recovered photon-hours, not the reflectivity immediately after one cycle.

I divide collector condition into reversible deposit, potentially reversible compound, roughness change, coating intermixing, erosion, and figure change. Each has a different disposition. Metrology should estimate which state dominates before applying an aggressive cleaning recipe.

BaselineMap reflectivity and scatter OperateAccumulate photon and debris history DiagnoseSeparate deposit from damage RecoverClean or condition selectively DispositionReturn, derate, or exchange

Spatial mapping is important. The collector does not receive uniform plasma, debris, gas, temperature, or cleaning exposure. A local damaged sector can change illumination distribution and imaging even if integrated power remains adequate.

Intermediate-focus power is necessary but incomplete

The fab observes source power delivered at an interface. That is the right operational quantity, but it cannot uniquely diagnose the collector. Source emission may rise while reflectivity falls. Beam pointing can shift. A sensor can drift. Control may increase laser demand to compensate, hiding degradation while energy and debris rise.

I want redundant evidence: calibrated in-band sensors, collector reflectometry or proxies, spatial far-field measurement, laser-to-EUV conversion estimate, contamination witnesses, cleaning response, and source history. State estimation should carry uncertainty.

Predictive maintenance can then schedule cleaning or exchange based on expected usable photon delivery and risk of irreversible damage. Waiting for a hard throughput threshold may sacrifice collector lifetime. Cleaning too often consumes availability and may erode the optic.

Design the collector for a service chain

Collector handling, transport, storage, installation, alignment, vacuum compatibility, contamination control, and disposal or refurbishment are part of the architecture. A replacement procedure needs protected surfaces, metrology references, tool access, verification, and a controlled chamber recovery.

The collector is a precision optic and a consumable exposed to plasma. Treating it as only one or the other produces a poor system. Its frontier is an observable, recoverable lifecycle in which photon production and material damage are measured together.

When that lifecycle is optimized, source power stops being a race against hidden optical decay. It becomes a managed flow of usable photons through a component whose condition, value, and next intervention are known.