ArF Immersion Lithography: Physics, Process Principles, and Sub-10nm Scaling

Introduction

For decades, the semiconductor industry has relentlessly pursued scaling to maintain the performance gains and cost reductions predicted by Moore’s Law. The primary vehicle for this scaling has been optical projection lithography. As the industry pushed beyond mature planar nodes, conventional "dry" optical lithography utilizing argon fluoride (ArF) excimer lasers operating at a deep ultraviolet wavelength encountered a critical physical barrier. Optical lithography scaling faces physical constraints because lithography limitations arise partly from the limited depth of focus of submicron lithography instruments [T1]. The numerical aperture (NA) of dry systems, which is physically constrained by the refractive index of air, could not exceed unity, thereby limiting the minimum resolvable feature size and depth of focus (DOF).

To bypass this limitation, the industry evaluated transitioning to a shorter vacuum ultraviolet wavelength using fluorine ($F_2$) lasers. However, dry lithography at shorter wavelengths faced severe material bottlenecks, including the lack of suitable pellicle materials, high absorption in conventional lenses, and the optical degradation of calcium fluoride ($CaF_2$) optics under intense ultraviolet radiation. This paved the way for ArF immersion lithography (also known as 193i), which introduced a high-refractive-index liquid (ultra-pure water) between the final lens element of the projection system and the photoresist-coated wafer. By altering the optical medium, ArF immersion lithography bypassed the vacuum ultraviolet generation, enabling the industry to scale down through the 28nm process flow and into advanced multi-gate regimes.

Physics & Mechanism

The Refractive Index and Effective Wavelength

The fundamental physical mechanism of ArF immersion lithography relies on the modification of the optical environment through which light propagates. According to classical wave optics, when an electromagnetic wave transitions from air into a dense optical medium, its frequency remains constant while its phase velocity and wavelength decrease. The wavelength in the medium is defined as:

$$\lambda = \frac{\lambda_0}{n}$$

where $\lambda_0$ is the vacuum wavelength of the light source and $n$ is the refractive index of the medium. By inserting ultra-pure water (which exhibits a refractive index significantly higher than air at the exposure wavelength) into the optical path, the effective exposure wavelength within the liquid is proportionally reduced. This reduction in wavelength is achieved without altering the underlying excimer laser light source, thereby preserving the existing optical material platform and photoresist chemistry developed for deep ultraviolet radiation.

Numerical Aperture Expansion

The resolution ($W$) of an optical lithography system is governed by the classical Rayleigh resolution limit:

$$W = \frac{k_1 \lambda_0}{\text{NA}} = \frac{k_1 \lambda_0}{n \sin\theta}$$

where $k_1$ is a process-dependent factor determined by the photoresist performance and resolution enhancement techniques (RET), and $\theta$ is the half-angle of the maximum cone of light that can enter the objective lens.

In a dry system, the maximum NA is strictly limited by the refractive index of the medium between the lens and the wafer (air, $n_\text{air} \le 1.0$). In an immersion system, the refractive index of the immersion fluid ($n_\text{fluid}$) scales the system's maximum NA, allowing hyper-NA lithography systems with numerical apertures exceeding unity. The larger the numerical aperture, the higher the spatial frequencies of diffracted light collected by the objective lens, directly enabling the projection of denser patterns.

Depth of Focus and Polarization Effects

In addition to resolution enhancement, the introduction of an immersion fluid dramatically influences the DOF. The classical DOF formula for a projection system is given by:

$$\text{DOF} = \frac{k_2 \lambda_0}{\text{NA}^2}$$

At a constant NA, transitioning from air to water decreases the refraction angle at the wafer surface, effectively increasing the latitude of focus positioning and expanding the usable process window.

At large convergence angles, image formation becomes more sensitive to polarization. The orientation of the electric field affects how components from different rays overlap and interfere. In some imaging geometries, p-polarized illumination reduces contrast relative to s-polarized illumination. Appropriate polarization can improve pattern transfer, but the outcome depends on the illumination and feature orientation; it is not a universal guarantee of high contrast.

Process Principles

Source-Mask Optimization and Freeform Illumination

To maximize the capability of ArF immersion systems at their physical resolution limits, engineers leverage RET such as source-mask optimization (SMO). Traditional illumination modes (e.g., annular or dipole) are often insufficient to resolve complex, multi-directional features simultaneously. By implementing programmable illuminators consisting of micro-mirror arrays, systems can generate custom freeform illumination sources that redistribute diffraction orders across spatial frequencies to enhance contrast and stabilize the critical dimension (CD). Advanced algorithms compute symmetric or asymmetric freeform sources that ensure both horizontal and vertical features remain within the required after development inspection (ADI) CD tolerance.

Chemically Amplified Resists (CAR)

The chemical reaction principle of the photoresist under exposure is critical to achieving stable patterns. Deep ultraviolet resists that use chemical amplification were introduced in submicron technology generations and remain foundational for scaling [T1]. Upon absorption of deep ultraviolet photons, a photoacid generator (PAG) embedded within the polymer matrix undergoes photolysis to generate a strong acid molecule.

During the subsequent post-exposure bake (PEB) process, this acid acts as a catalyst, diffusing through the matrix and driving deprotection reactions that cleave acid-labile ester groups on the polymer backbone. This deprotection exposes hydrophilic carboxylic acid groups, creating a massive solubility contrast between exposed and unexposed regions in a basic developer (e.g., tetramethylammonium hydroxide, TMAH). The rate of acid diffusion during PEB must be carefully balanced; excessive diffusion degrades the line-edge roughness (LER) and line-width roughness (LWR), while insufficient diffusion limits sensitivity and contrast.

Challenges & Failure Modes

Fluid-Related Defects: Bubbles and Watermarks

The presence of water directly at the interface between the lens and the wafer introduces unique defect categories. The high-speed scanning motion of the wafer stage can drag air into the immersion hood, forming microscopic bubbles. These bubbles act as refractive index inhomogeneities, scattering the incident light and causing localized projection shadowing that leads to pattern bridging or micro-bridging.

Additionally, water droplets left behind on the photoresist surface during scanning can dissolve soluble components of the resist, such as PAGs or base quenchers. Upon evaporation, these localized residues form "watermarks" that locally alter the deprotection rate during PEB, leading to severe CD variation and local pattern failure.

Photoresist Leaching and Topcoats

Because the immersion water is in direct contact with the photoresist, low-molecular-weight additives within the resist can leach into the water column. Leached PAGs or organic bases can contaminate the final lens element of the projection optics, undergoing photodeposition under high-intensity UV exposure and degrading lens transmission.

To prevent this leaching, a protective, hydrophobic topcoat is often applied over the photoresist layer. This topcoat must be impermeable to water and photoresist components while remaining highly transparent to deep ultraviolet light and easily dissolvable during the development step.

Polarization-Induced Contrast Loss

Polarization-dependent imaging can contribute to differences in printed profiles and exposure latitude. Its influence depends on the pattern, pupil illumination and resist response. The relevant distinction is the interference of differently oriented field components, rather than an assertion that every p-polarized component always cancels throughout the resist.

Technology Node Evolution

28nm Node: Single-Exposure Limits

At the 28nm node, ArF immersion lithography reached its single-exposure limit for dense features in advanced photolithography. With high numerical aperture optics and process parameters near theoretical physical limits, a single-exposure ArF system reached a fundamental single-exposure pitch limit. To scale below this, advanced RET, complex optical proximity correction (OPC), and highly optimized phase-shifting masks were required.

14nm to 7nm Nodes: Multi-Patterning Schemes

To bridge the gap before extreme ultraviolet (EUV) lithography became commercially viable, ArF immersion was extended to the 14nm FinFET and 7nm FinFET nodes using multi-patterning schemes. By splitting a dense layout into multiple, coarser sub-patterns, the effective pitch was halved or quartered.

• Self-Aligned Double Patterning (SADP): A mandrel layer is printed using ArF immersion, followed by conformal atomic layer deposition of a spacer material and anisotropic dry etching to form sidewall spacers. The mandrel is then selectively stripped, leaving behind twice the density of features.
• Self-Aligned Quadruple Patterning (SAQP): By repeating the spacer deposition and etching sequence on the SADP-generated features, the spatial frequency is doubled again, allowing sub-resolution pitches to be printed with ArF immersion.
• Litho-Etch-Litho-Etch (LELE): Multiple separate exposure and dry etching cycles are performed sequentially, which requires tight overlay tolerances and places massive demands on stage alignment and metrology systems.

Related Processes

ArF immersion lithography does not operate in isolation; it is deeply coupled with several upstream and downstream process steps:

• Chemical Mechanical Planarization (CMP): Because hyper-NA immersion lithography has a narrow depth of focus, the incoming topography must be thoroughly planarized using chemical mechanical planarization to prevent focus-induced CD variations across the die.
• Dry Etching: Due to the thinness of the photoresists used to prevent aspect-ratio collapse in high-resolution features, the photoresist pattern alone cannot survive deep silicon or dielectric etching. Hard mask stacks—often comprising spin-on glass (SOG), spin-on carbon (SOC), or bottom anti-reflective coating layers—are patterned by the resist and subsequently used to transfer the layout into the device layers.
• Thin-Film Underlayers: Specialized underlayers are utilized beneath the photoresist stack. These underlayers (which may contain heavy elements like Sn) provide adhesion, minimize backscattered light reflection, and act as hard masks during downstream dry etching steps.

Future Outlook

While EUV lithography has taken over critical layer printing at sub-7nm nodes, ArF immersion lithography remains the workhorse of the modern semiconductor fab, patterning the vast majority of non-critical metal, via, and implant layers in advanced logic and memory architectures.

Furthermore, emerging applications are adapting ArF immersion tools to non-traditional fabrication fields. For instance, researchers have demonstrated the high-throughput, wafer-scale fabrication of optical metasurfaces on silicon platforms. By leveraging the mature, high-speed, and precise CD control of 193i scanners, the optical and photonics industries are transitioning away from slow electron-beam lithography to realize commercially viable flat-lens systems, highlighting the enduring legacy of this powerful optical technology.