Multilayer Mirrors and Atomic-Precision Optics

Abstract

At 13.5 nanometers, ordinary refractive optics absorb rather than transmit useful light. EUV systems therefore use mirrors whose reflectivity comes from constructive interference in repeated molybdenum and silicon layers. The mirror is simultaneously a substrate, nanometer-period coating, thermal body, mechanical component, contamination-sensitive surface, and phase element. I derive the Bragg condition that sets multilayer period, then trace how interface width, roughness, interdiffusion, thickness variation, capping layers, stress, substrate figure, mounting, and temperature affect amplitude and phase. A projection path contains several mirrors, so modest loss at each surface compounds. High NA requires larger, more strongly curved mirrors and metrology capable of validating them at operating scale. I organize fabrication around a traceable coordinate system that follows the optic from substrate polishing through coating, mounting, vacuum, thermal state, active correction, and scanner wavefront. Optical perfection is not a free-state surface measurement. It is the controlled in-system wavefront.

Reflection is built from interference

EUV wavelengths are absorbed within shallow depths of most solids. A single metallic surface does not provide the reflectivity needed for a multi-mirror projection system. Alternating layers create partial reflections whose phases add near a designed wavelength and angle.

For multilayer period $d$, grazing angle convention $\theta$, order $m$, and wavelength $\lambda$, a simplified Bragg relationship is

$$m\lambda=2d\sin\theta.$$

The actual design uses complex refractive indices, absorption, interface layers, polarization, angle distribution, and many periods. The relationship still reveals the scale: layer thickness control is tied directly to a 13.5-nanometer wavelength.

Molybdenum and silicon are widely used because their optical contrast and absorption balance can produce useful reflectivity. Barrier and capping layers may manage interdiffusion, oxidation, cleaning, or lifetime. Every added material changes optical phase, stress, and chemical behavior.

Interfaces determine the ideal that materials cannot reach

An ideal multilayer has abrupt, smooth, compositionally stable interfaces. Real deposition produces roughness, intermixing, thickness drift, particles, stress, and local defects. Heat and radiation can evolve the stack.

Interface roughness reduces coherent reflectivity and increases scatter. If roughness propagates from layer to layer, the stack can amplify substrate features. Interdiffusion changes the effective period and optical contrast. Thickness error shifts peak wavelength and phase. Spatial nonuniformity becomes a wavefront error, not only a power loss.

Multilayer variable	Optical consequence	Lifecycle consequence
Period thickness	Peak wavelength and phase	Sensitivity to deposition drift
Layer ratio	Bandwidth, reflectivity, absorption	Thermal and stress balance
Interface width	Lower contrast and reflectivity	Evolution under heat
Roughness	Scatter and phase noise	Cleaning sensitivity
Cap material	Surface chemistry and reflectivity	Oxidation and contamination response
Residual stress	Figure distortion	Mounting and thermal stability

The coating design should be optimized for the distribution of incidence angles seen by each mirror. High NA broadens angular demands. A uniform stack may not be ideal over the full surface, so graded multilayers can tune response spatially. That capability raises deposition and metrology difficulty.

Transmission compounds across mirrors

If mirror $i$ has reflectivity $R_i$, ideal optical transmission through $n$ mirrors is

$$T=\prod_{i=1}^{n}R_i.$$

Six mirrors at 70 percent reflectivity transmit about 11.8 percent before mask, collector, spectral, obscuration, and other losses. Raising each to 71 percent increases the product to about 12.8 percent, an 8.6 percent relative gain. Small improvements at each surface can materially affect wafer-plane power.

The same compounding applies to degradation. A one-percent relative loss on several mirrors raises exposure time and source demand. The optical column therefore needs contamination control and condition monitoring, not only excellent initial coatings.

SurfaceCap and contamination boundary

Controls chemistry, cleaning, oxidation, and first optical interface.

StackMo/Si interference multilayer

Creates wavelength- and angle-selective amplitude and phase response.

BodyUltra-low-error substrate

Provides figure, roughness, stiffness, and thermal stability.

SupportMount, frame, sensors, and actuators

Preserves or corrects operating figure inside the optical system.

Figure exists across spatial scales

Low-spatial-frequency figure error changes wavefront and aberration. Mid-spatial-frequency error can create flare, image-placement variation, and contrast loss. High-spatial-frequency roughness scatters light. A single rms surface value can hide which regime contains the error.

Metrology should report a power spectral description with instrument transfer functions and uncertainty. Interferometry, profilometry, scatter measurement, microscopy, and actinic methods cover different scales. Stitching data from several instruments requires common coordinates and overlap checks.

Surface error contributes approximately twice its height to reflected wavefront optical path at normal incidence, modified by geometry. At EUV wavelength, picometer and subnanometer scales become consequential across the pupil. ZEISS describes subnanometer metrology and the exceptional scale of High NA optical systems because larger, more curved mirrors still need atomic-level precision.

Substrate materials must combine polishability, stiffness, low thermal expansion, homogeneity, and stability. Internal material variation can distort the surface under temperature or load. Polishing must remove error without imprinting damaging mid-spatial signatures.

Coating stress and mounting rewrite the surface

A perfect substrate can deform after multilayer deposition. Film stress acts across the optic. Spatially varying stress produces figure change. Subsequent capping, cleaning, annealing, and aging can alter the balance.

Mounting introduces support forces, gravity orientation during assembly, contact behavior, preload, actuator forces, and thermal paths. In operation, vacuum removes convective heat transfer. Absorbed EUV and out-of-band energy create gradients. The mirror’s free-state room-temperature figure is therefore not the final optical state.

I want a digital thread:

PolishMap substrate figure and spectrum CoatMeasure amplitude, phase, and stress MountTrace support-induced distortion OperateEstimate thermal in-vacuum state CorrectClose scanner wavefront

All data should share fiducials and coordinates so a phase variation measured actinically can be associated with deposition, substrate, or mount features. Model uncertainty must accompany the transformation.

Actinic metrology tests the relevant response

Visible-light metrology can measure figure with extraordinary precision but does not directly observe the EUV multilayer’s wavelength-specific amplitude and phase. Actinic measurement uses the relevant wavelength and can reveal coating behavior invisible to non-actinic methods.

Actinic tools are expensive and photon-limited. They should be used where they add unique information, calibrated against other instruments, and connected to model-based prediction. The goal is not to replace all fabrication metrology with EUV light. It is to validate the transformations that infer operating optical performance from accessible measurements.

Precision needs production yield

An optic that takes a year to manufacture and fails at final coating or system integration represents enormous lost capacity. Process control must detect drift early through witness samples, in-situ deposition monitoring, substrate maps, predictive stress models, and intermediate acceptance gates.

Repair options are limited. Some figure can be corrected through further material removal, deposition, or active control. Particles and coating defects may be terminal for a critical surface. Prevention, clean handling, and transport are part of optical yield.

The mirror frontier is not an isolated record for smoothness. It is a reproducible chain that turns raw substrate into a known in-system wavefront, then preserves that state through years of exposure and maintenance. EUV optics succeed when atomic precision becomes an industrial population rather than one exceptional artifact.