The Rayleigh Criterion: How Wavelength and Numerical Aperture Actually Constrain Chip Scaling

A standard optical microscope resolves microscopic objects by gathering light through a lens, a process limited fundamentally by the wavelength of that light and the lens's collection angle. A semiconductor lithography scanner operates on the exact same physical principle, but with one critical difference: instead of projecting an image for a human eye to distinguish, it projects an aerial image onto a chemically reactive photoresist that must cross a strict solubility threshold.[1]

This projection process is the manufacturing heartbeat of the $600 billion global semiconductor industry. Every modern processor is built by shining light through a blueprint—a photomask—and shrinking that pattern onto a silicon wafer to create billions of transistors.[3]

The industry's relentless drive to pack more computational power into the same area is often obscured by marketing terminology. Foundries frequently announce "3-nanometer" or "18-angstrom" nodes, but these labels no longer correspond to any physical dimension on the chip; they are commercial shorthand for a generation of performance improvements.

Instead, the true physical minimum feature size—known as the Critical Dimension—is dictated by the Rayleigh criterion, an optical principle formulated by Lord Rayleigh in 1896.[4]

The equation governing this limit is elegantly simple: the Critical Dimension equals a process-dependent factor multiplied by the wavelength of the light, divided by the numerical aperture of the optical system.[1]

To make transistors physically smaller, chipmakers have exactly three mathematical levers to pull. They can decrease the wavelength, increase the numerical aperture, or lower the process factor.[1][2]

Historically, the most straightforward lever was wavelength. As documented in a 2006 historical perspective published in the Comptes Rendus de l'Académie des Sciences, the industry migrated from visible light to ultraviolet, moving from 436-nanometer g-line down to 365-nanometer i-line, and eventually to deep ultraviolet at 248 and 193 nanometers.[5]

The transition to Extreme Ultraviolet (EUV) lithography represented a massive engineering leap, dropping the wavelength to 13.5 nanometers. This required entirely new light sources—firing lasers at microscopic tin droplets—and reflective mirrors instead of refractive glass lenses, because 13.5-nanometer light is absorbed by almost all matter, including air.[6]

With wavelength currently stalled at 13.5 nanometers, the focus has shifted to the denominator of the Rayleigh equation: Numerical Aperture. This metric defines the ability of the optical system to collect and focus light.[1][3]

Standard EUV systems operate with a Numerical Aperture of 0.33. The next generation of scanners, known as High-NA EUV, pushes this figure to 0.55, allowing for a tighter focus and smaller features without changing the light source.[6]

Increasing this aperture requires vastly larger and more complex mirror systems. Ansys Optics methodology documentation highlights that higher Numerical Aperture increases the angles at which light rays converge on the wafer, which introduces severe new challenges with depth of focus.[3]

The depth of focus scales inversely with the square of the Numerical Aperture. This means that as the aperture increases from 0.33 to 0.55, the plane of perfect focus becomes razor-thin, requiring the silicon wafer to be flatter than ever before to prevent the image from blurring.[1][3]

The final lever is the process factor, which encapsulates the complexity of the manufacturing process, including the photoresist chemistry and computational lithography techniques.[2]

The theoretical physical limit for this factor in a single exposure is 0.25. ASML's documentation on pushing this boundary explains that reaching this limit requires aggressive resolution enhancement techniques, such as optical proximity correction and phase-shift masks, which pre-distort the blueprint so the light naturally bends into the correct shape.[2]

When the process factor cannot be pushed any lower, manufacturers resort to multiple patterning—splitting a dense pattern across two or more masks. This effectively circumvents the single-exposure Rayleigh limit but drastically increases manufacturing time, complexity, and cost.[4]

Furthermore, the photoresist itself becomes a bottleneck at these microscopic scales. A comparative study published in PMC on efficient photoresists beyond EUV lithography notes that at 13.5 nanometers, the number of photons available is so low that statistical fluctuations cause the edges of the printed lines to become rough.[7]

This phenomenon, known as shot noise, creates line edge roughness that can cause transistors to short-circuit or fail entirely. To smooth the lines, manufacturers must increase the radiation dose, which slows down the scanner and reduces the throughput of the multi-million-dollar machines.[7]

Because the primary literature from institutions like Ansys and the Institute for Microelectronics relies entirely on mathematical proofs and performance data, direct quotations from engineers are absent from these foundational texts. The physics, however, provides a rigid framework that cannot be negotiated.[3][4]

The semiconductor industry's roadmap for the next decade relies entirely on balancing these three variables. High-NA EUV provides a temporary reprieve by increasing the denominator, but the physics of light collection are nearing their practical limits.[6]

The next frontier, Hyper-NA systems targeting 0.75 or higher, or a theoretical shift to even shorter wavelengths, remains firmly in the research phase. The Rayleigh criterion will continue to dictate the pace of computing advancement, regardless of the marketing names applied to future nodes.[1][6]