Lithography Theory Documentation

Comprehensive guide to the fundamental principles of optical lithography

1. Wave Optics Fundamentals

Huygens Principle

Huygens principle, formulated in 1678, states that every point on a wavefront acts as a source of secondary spherical wavelets. The new wavefront at any later time is the envelope of these secondary wavelets. This principle provides the foundation for understanding diffraction phenomena critical to lithography.

U(P) = (1/iλ) ∫∫ U(Q) [exp(ikr)/r] cos(n,r) dS

where:
U(P) = field at observation point P
U(Q) = field at source point Q on wavefront
k = 2π/λ (wave number)
r = distance from Q to P
cos(n,r) = obliquity factor

Maxwell Equations in Dielectrics

Electromagnetic wave propagation in lithographic materials is governed by Maxwell's equations. In a linear, isotropic, non-magnetic dielectric with permittivity ε and permeability μ₀:

∇ × E = -∂B/∂t
∇ × H = ∂D/∂t
∇ · D = 0
∇ · B = 0

Wave equation: ∇²E - (n²/c²)(∂²E/∂t²) = 0
where n = √(εᵣ) is the refractive index

Wavefront Propagation and Diffraction

When a wavefront encounters an aperture or obstacle, diffraction occurs. The Kirchhoff diffraction formula describes field propagation from an aperture:

Plane Wave Spectrum

Any optical field can be decomposed into a superposition of plane waves propagating at different angles. This angular spectrum representation is fundamental to understanding imaging systems:

U(x,y,z) = ∫∫ A(fx,fy) exp[i2π(fxx + fyy + fzz)] dfx dfy

fz = √[(1/λ)² - fx² - fy²] (propagating waves)
fz = i√[fx² + fy² - (1/λ)²] (evanescent waves)

2. Fourier Optics

Spatial Frequency Domain

Fourier optics treats optical systems in the spatial frequency domain, where patterns are decomposed into sinusoidal components. A lens performs an optical Fourier transform, mapping spatial frequencies to angles and positions in the pupil plane.

F{f(x,y)} = ∫∫ f(x,y) exp[-i2π(fxx + fyy)] dx dy

Spatial frequency: f = sin(θ)/λ
Maximum spatial frequency captured by lens: fmax = NA/λ

Fourier Transform Properties

Key properties relevant to lithographic imaging:

Property Spatial Domain Frequency Domain
Linearity a·f(x) + b·g(x) a·F(fx) + b·G(fx)
Shift f(x - x₀) F(fx)·exp(-i2πfxx₀)
Scaling f(ax) (1/|a|)·F(fx/a)
Convolution f(x) ⊗ g(x) F(fx)·G(fx)
Multiplication f(x)·g(x) F(fx) ⊗ G(fx)

Transfer Functions and Point Spread Function

The imaging system is characterized by its transfer function (frequency domain) and point spread function (spatial domain):

Coherent Transfer Function (CTF): H(fx,fy) = P(fx,fy)
Optical Transfer Function (OTF): H(fx,fy) = P(fx,fy) ⊗ P(fx,fy)

PSF = |F⁻¹{H(fx,fy)}|²
Image = Object ⊗ PSF

Convolution and Imaging

In the spatial domain, image formation is described by convolution of the object with the point spread function. For an incoherent system, intensity distributions convolve. For coherent systems, complex amplitudes convolve and the intensity is computed afterward.

Historical Note: Joseph Fourier developed his transform theory in 1822 for heat conduction. Ernst Abbe applied Fourier concepts to microscopy in 1873, establishing the foundation for modern Fourier optics formalized by Duffieux in 1946.

3. Hopkins Formulation

Partially Coherent Imaging Theory

H.H. Hopkins developed the rigorous mathematical framework for partially coherent imaging in 1953. This formulation is essential for lithography scanners, which operate with partial coherence (σ = 0.3 to 0.95).

I(x,y) = ∫∫∫∫ TCC(fx,fy; fx',fy') · M(fx,fy) · M*(fx',fy') ×
exp[i2π(fx-fx')x + i2π(fy-fy')y] dfx dfy dfx' dfy'

TCC = Transmission Cross Coefficient
M(fx,fy) = Fourier transform of mask pattern
* = complex conjugate

Transmission Cross Coefficient (TCC)

The TCC encodes all optical information about the imaging system: pupil function, illumination, aberrations, and defocus. It is a four-dimensional function in spatial frequency space.

TCC(fx,fy; fx',fy') = ∫∫ S(fx",fy") · P(fx-fx",fy-fy") ·
P*(fx'-fx",fy'-fy") dfx" dfy"

S(fx,fy) = source intensity distribution (normalized)
P(fx,fy) = pupil function including aberrations

Coherent vs Incoherent Limits

Hopkins formulation unifies coherent and incoherent imaging as limiting cases:

Practical Implementation: Modern lithography simulators compute the TCC matrix for representative frequency pairs, then perform matrix multiplication with mask spectrum vectors. This reduces computation from O(N⁴) to O(N²M) where M is the number of TCC eigenvalues retained.

4. Abbe Theory of Image Formation

Two-Step Diffraction and Interference Model

Ernst Abbe proposed in 1873 that image formation occurs in two steps: (1) the mask diffracts incident light into discrete orders at the pupil plane, and (2) these orders interfere at the image plane to reconstruct the pattern.

For a periodic mask with pitch p:

Diffraction orders at angles: sin(θₘ) = m·λ/p (m = 0, ±1, ±2, ...)
Orders captured by lens: |m| ≤ NA·p/λ

Minimum pitch requiring two orders (0th and ±1st):
p_min = λ/NA

Resolution Limit Derivation

For a line/space pattern to be resolved, at least the 0th and one ±1st order must pass through the lens. This leads to Abbe's resolution criterion:

Minimum resolvable period: p = λ/NA
Minimum resolvable linewidth: w = λ/(2·NA) = p/2

With partial coherence factor k₁:
Resolution: R = k₁·λ/NA
where k₁ ≈ 0.25 (aggressive RET) to 0.5 (conventional)

0th and ±1st Order Interference

The aerial image formed by interference of 0th and ±1st orders creates a sinusoidal intensity distribution. The image contrast depends on the relative amplitudes and phase of the interfering orders:

I(x) = |A₀ + A₊₁·exp(i2πx/p) + A₋₁·exp(-i2πx/p)|²
I(x) = |A₀|² + |A₊₁|² + |A₋₁|² + 2Re[A₀*A₊₁·exp(i2πx/p)] + ...

Image Modulation Transfer (IMT) = (Imax - Imin)/(Imax + Imin)

Higher-Order Effects

When higher diffraction orders (±2nd, ±3rd, etc.) pass through the lens, they contribute harmonics that sharpen edges and improve pattern fidelity. However, at small pitches approaching the resolution limit, only 0th and ±1st orders are captured.

Abbe's Experiment: Abbe demonstrated his theory using a microscope with an adjustable iris at the back focal plane (pupil). By blocking specific diffraction orders, he showed how different combinations create distinct images, proving that image formation is a diffraction-interference phenomenon.

5. Rayleigh Criterion

Resolution Definition

Lord Rayleigh (1896) defined the resolution limit as the minimum separation between two point sources where they can just be distinguished. The Rayleigh criterion states this occurs when the principal maximum of one Airy disk coincides with the first minimum of the other.

Rayleigh Resolution (circular aperture):
R = 0.61·λ/NA

Generalized for lithography:
R = k₁·λ/NA

where k₁ depends on:
- Illumination coherence (σ)
- Pattern type (lines, contacts, isolated vs dense)
- Acceptable image quality criteria

Depth of Focus Derivation

Depth of focus (DOF) quantifies the range of defocus over which the image remains acceptably sharp. It is derived from the acceptable wavefront error at the edge of the pupil:

Wavefront error at pupil edge due to defocus Δz:
W = (NA²/2n) · Δz

Setting W = λ/4 (Rayleigh quarter-wave criterion):
DOF = ±k₂·λ/NA²

Typical values: k₂ ≈ 0.5 to 1.0

k₁ and k₂ Process Factors

The factors k₁ and k₂ are not physical constants but depend on process conditions, pattern characteristics, and acceptable tolerances:

Condition k₁ Range k₂ Range
Conventional (no RET) 0.50 - 0.80 0.8 - 1.2
With OPC + OAI 0.35 - 0.50 0.5 - 0.8
With attPSM 0.30 - 0.40 0.4 - 0.6
Aggressive RET 0.25 - 0.30 0.3 - 0.5

Trade-off Between Resolution and DOF

There exists a fundamental trade-off in optical lithography: increasing NA improves resolution (R ∝ 1/NA) but degrades depth of focus (DOF ∝ 1/NA²). This inverse quadratic relationship motivates immersion lithography and EUV technology.

Example: For ArF lithography (λ = 193 nm), NA = 1.35 (immersion), k₁ = 0.28, k₂ = 0.45:
Resolution = 0.28 × 193nm / 1.35 = 40 nm
DOF = 0.45 × 193nm / 1.35² = 48 nm

6. Coherence Theory

Van Cittert-Zernike Theorem

The Van Cittert-Zernike theorem establishes the relationship between source size and spatial coherence at the mask plane. It states that the complex degree of coherence is the Fourier transform of the source intensity distribution.

Mutual coherence function:
J(x₁,x₂) = ∫∫ I_source(xs,ys) · exp[ik(xs·Δx + ys·Δy)/d] dxs dys

where Δx = x₂ - x₁, d = source-to-mask distance

Complex degree of coherence:
γ(x₁,x₂) = J(x₁,x₂) / √[J(x₁,x₁)·J(x₂,x₂)]

Spatial Coherence Parameter σ

In lithography, partial coherence is characterized by the σ parameter, defined as the ratio of illumination numerical aperture to imaging numerical aperture:

σ = NA_illumination / NA_imaging

σ → 0: Coherent illumination (point source)
σ = 0.3-0.7: Typical lithography range
σ → ∞: Incoherent illumination (Köhler)

Temporal Coherence and Laser Bandwidth

Temporal coherence, related to source bandwidth, affects imaging through chromatic aberration and interference contrast. Excimer lasers used in DUV lithography must have tightly controlled bandwidth:

Coherence length: Lc = λ²/Δλ

For ArF laser: λ = 193 nm, Δλ = 0.3 pm (FWHM)
Lc = 193² / 0.0003 ≈ 124 mm

This exceeds optical path differences in scanner, ensuring
temporal coherence effects are negligible.

Mutual Coherence and Imaging

The mutual coherence function describes correlations between field points. For lithography, spatial coherence is far more important than temporal coherence. The Hopkins formulation incorporates spatial coherence through the source distribution in the TCC.

Practical Impact: Adjusting σ (via illumination aperture) changes the balance between resolution and depth of focus. Lower σ gives better depth of focus but reduces resolution and throughput. Modern scanners use optimized off-axis illumination (annular, dipole, quadrupole) to maximize k₁ while maintaining acceptable k₂.

7. Aberration Theory

Seidel Aberrations

Seidel aberrations are the five primary third-order monochromatic aberrations that degrade image quality in optical systems:

Zernike Polynomials

Zernike polynomials form an orthogonal basis for describing wavefront aberrations over a circular pupil. They are extensively used in lithography for aberration characterization and correction.

Wavefront: W(ρ,φ) = Σ Cₙₘ · Zₙₘ(ρ,φ)

Zₙₘ(ρ,φ) = Rₙₘ(ρ) · cos(mφ) or sin(mφ)

Common Zernikes:
Z₁: Piston (constant offset)
Z₂, Z₃: Tilt (x and y)
Z₄: Defocus
Z₅, Z₆: Astigmatism (0° and 45°)
Z₇, Z₈: Coma (x and y)
Z₉: Spherical aberration
Z₁₀, Z₁₁: Trefoil

Wavefront Error and Image Degradation

Aberrations introduce phase errors in the pupil, degrading image quality. The relationship between wavefront error and image quality is quantified by the Strehl ratio:

Strehl ratio: S = (Peak intensity with aberration) /
(Peak intensity without aberration)

Maréchal approximation (small aberrations):
S ≈ 1 - (2π·RMS_wavefront/λ)²

Diffraction-limited: S > 0.8 → RMS < λ/14

Aberration Specifications for Lithography

Modern lithography lenses maintain extremely tight aberration budgets:

Scanner Generation RMS Wavefront Error Equivalent (nm @ 193nm)
ArF Dry (NA 0.75) λ/50 ~4 nm RMS
ArF Immersion (NA 1.35) λ/80 ~2.5 nm RMS
EUV (NA 0.33) λ/100 ~0.13 nm RMS

Measurement and Compensation: Scanner aberrations are measured using interferometry or image-based methods. Low-order aberrations can be corrected via lens element manipulators, while higher-order aberrations require compensation in mask design (OPC) or offline monitoring and metrology matching.

8. Resolution Enhancement Techniques

Optical Proximity Correction (OPC) Principles

OPC compensates for optical proximity effects by modifying mask geometries. As feature sizes approach and exceed the resolution limit, proximity effects become severe, requiring sophisticated correction strategies.

Model-based OPC iteration:
Edge Position Error: EPE_i = CD_target - CD_simulated(mask_i)
Mask update: mask_{i+1} = mask_i + α · EPE_i

Convergence criterion: max(|EPE|) < tolerance (typically 1-2 nm)

OPC techniques include:

Phase Shift Mask (PSM) Theory

PSM exploits destructive interference to enhance resolution and contrast. Two main types exist:

Attenuated PSM (AttPSM, 6% or HTPSM)

Transmission: t ≈ 6% (amplitude) = 0.245
Phase: φ = 180° relative to clear areas

Complex transmission: T_attPSM = -0.245 + i·0
This creates destructive interference at edges, sharpening the aerial image.

Alternating PSM (AltPSM or Levenson PSM)

Adjacent clear areas have 0° and 180° phase, creating destructive interference (dark line) between them. This doubles the effective cutoff frequency of the lens.

Minimum pitch with AltPSM: p = λ/(2·NA)
(compared to λ/NA without PSM)

Achieves k₁ as low as 0.25

Off-Axis Illumination (OAI)

OAI uses non-conventional source shapes to optimize diffraction order capture and improve process window:

Source-Mask Optimization (SMO)

SMO co-optimizes source shape and mask pattern to maximize process window for a given design. This is a high-dimensional optimization problem typically solved using gradient-based or genetic algorithms.

Objective function:
max {PW(source, mask)}
subject to: manufacturability constraints on source and mask

PW = Process Window (area in dose-focus space meeting specs)
Industry Trend: Modern 7nm and below nodes use full-chip computational lithography with OPC, ILT, complex freeform sources, and often multiple patterning. Mask data preparation can require weeks of computation on large clusters for a single layer.

9. Photoresist Physics

Chemical Amplification

Chemically amplified resists (CAR), invented by Willson and Ito in 1982, use photoacid generators (PAG) that create acid upon exposure. During post-exposure bake (PEB), each acid molecule catalyzes many deprotection reactions, providing high sensitivity.

Exposure: PAG + hν → H⁺ + products
PEB catalytic cycle: H⁺ + protected_polymer → deprotected + H⁺

Amplification factor: 100-1000×
(one photon generates one acid; one acid deprotects 100-1000 sites)

Dill Parameters A, B, C

The Dill exposure model describes resist bleaching during exposure. Resist contains photoactive compound (PAC or PAG) with concentration M and bleached product with concentration 1-M.

Absorption coefficients:
A = absorption of PAC (unbleached)
B = absorption of bleached product
C = quantum efficiency (PAC molecules destroyed per photon absorbed)

Rate equations:
dM/dz = -A·M (Beer's law for light intensity)
dI/dz = -[A·M + B·(1-M)]·I (absorption)
dM/dt = -C·A·M·I (bleaching kinetics)

Exposure Kinetics and Standing Waves

During exposure, light intensity varies with depth due to reflection from substrate and absorption in resist, creating standing wave patterns. These cause CD ripples that degrade sidewall quality.

I(z) = I₀·[(1+R)·exp(-αz) + R·(1-R)·exp(-α(2d-z))]

where:
R = substrate reflectivity
α = absorption coefficient
d = resist thickness

Standing wave period: λ/(2n_resist) ≈ 65 nm for ArF

PEB Acid Diffusion

During PEB (typically 90-130°C for 60-90 seconds), photogenerated acid diffuses laterally and vertically in the resist film. This diffusion smooths the latent image but also causes CD loss and line edge roughness.

Fick's second law:
∂C/∂t = D·∇²C

Diffusion length: L_diff = √(2·D·t)

Typical values: D = 1-10 nm²/s, t = 60-90 s
L_diff = 10-30 nm

Development Models

Development dissolves resist in aqueous base (typically 0.26N TMAH). The dissolution rate depends on local degree of deprotection, described by various models:

Mack model (empirical):
R = R_max / [1 + exp((T_sol - C_deprotect)/β)]

R_max = maximum dissolution rate
T_sol = solubility threshold
β = slope parameter
C_deprotect = local deprotection concentration

Development continues until cleared areas reach substrate. Final resist profile depends on the dissolution front propagation, which is determined by the 3D latent image distribution.

LER Origins: Line edge roughness arises from multiple sources: photon shot noise, acid diffusion stochastics, polymer molecular weight distribution, developer kinetics, and base contamination. Advanced resists and processes target LER < 2 nm (3σ) for sub-5nm nodes.

10. Metrology Fundamentals

Critical Dimension (CD) Measurement Principles

CD metrology quantifies feature dimensions. The primary techniques are CD-SEM (scanning electron microscopy) and scatterometry (optical CD):

CD-SEM Principles

Scatterometry (OCD) Principles

Measured: Diffraction spectrum S(λ, θ, polarization)
Model: Rigorous Coupled Wave Analysis (RCWA) or FEM
Fit: min ||S_measured - S_simulated(CD, height, SWA, ...)||²

Advantages: Non-destructive, measures buried structures, high throughput
Challenges: Model dependence, correlation between parameters

Overlay Metrology Theory

Overlay measures alignment between successive lithography layers. Modern overlay metrology uses diffraction-based imaging (DBI) or scatterometry techniques:

Box-in-box marks: Measure centroid offset between two nested squares
Precision: 0.1-0.3 nm (3σ) for advanced tools

Overlay error budget:
OVL_total² = OVL_scanner² + OVL_process² + OVL_metrology²

Target (7nm node): < 2 nm (mean + 3σ)

Focus Monitoring

Focus metrology uses special test structures designed to be sensitive to defocus:

Focus-sensitive CD:
CD(focus) = CD₀ + k·(focus - best_focus)²

Typical sensitivity: k = 0.5-2 nm per 100 nm² of defocus

Dose Control and Monitoring

Dose control maintains exposure energy within tight tolerances. Dose monitors are resist features with CD highly sensitive to dose variations:

Dose sensitivity:
∂CD/∂Dose = CD₀ · γ / Dose₀

where γ = resist contrast (typically 5-15)

For γ=10, CD₀=40nm, Dose₀=25 mJ/cm²:
1% dose change → 1.6 nm CD change

Total Measurement Uncertainty (TMU)

Metrology uncertainty must be much smaller than process tolerance. The TMU budget includes precision, tool-to-tool matching, and accuracy components:

TMU² = Precision² + Matching² + Accuracy²

Gauge R&R requirement: TMU < 10% of process tolerance

Example (7nm node, CD tolerance ±3nm):
Required TMU < 0.3 nm → Precision < 0.2 nm (3σ)
Metrology Challenges: As features shrink, metrology becomes increasingly difficult. 3nm node and below require sub-angstrom precision, multiple techniques (CD-SEM, scatterometry, AFM, TEM cross-sections), and extensive process-metrology interaction (PMI) studies to establish correlations between inline measurements and final electrical performance.