Carl Zeiss SMT

We review the development of photolithography for semiconductor manufacturing from early 1960s to the present. From the 1960s to early 1970s, contact and proximity printing were the only methods used in manufacturing, achieving a minimum critical dimension of about 4 μm. Projection lithography was adopted in manufacturing with the advent of Perkin-Elmer’s unit-magnification wafer scanners in the 1970s, and was advanced with the arrival of wafer steppers from GCA, Canon, Nikon, ASML, and others, employing mercury (Hg) arc lamps as the light source. These Hg g-line (436 nm in wavelength) machines provided about 0.8-μm practical resolution in the fab. Wafer steppers using the Hg i-line (365 nm in wavelength) came to wide use starting about 1990. Combined with off-axis illumination, i-line lithography was able to reach below 0.35 μm in resolution. Deep ultraviolet lithography using the KrF excimer laser (248 nm in wavelength) came next, entering into production at about 1997 for the 0.25-μm logic node (with a minimum half-pitch of 0.25 μm), followed by the adoption of ArF-excimer-laser lithography (193 nm in wavelength), whose initial use in manufacturing, for the 0.13-?m node of logic integrated circuits, took place in 2001. Massive transition to 193-nm lithography happened in the 90-nm node. 193-nm lithography employing water immersion entered production at Taiwan Semiconductor Manufacturing Company (TSMC) in 2007 at the 40-nm node and enabled continued geometrical scaling for more than a decade with multiple-patterning techniques. Extreme ultraviolet (EUV) lithography entered into manufacturing in 2019 and printed dense lines of 20-nm half-pitch that year. With the arrival of the 0.55 NA EUV scanner in 2024, dense lines of less than 10-nm half-pitch can now be printed. In the course of sixty-five years, photolithography achieved a greater than 400X reduction in linear dimension and enabled a 2×105 increase in areal density.

This paper presents a design methodology for a multi-directional quasi-zero-stiffness (QZS) system that can achieve vibration isolation of large bodies with high masses across a broad frequency range in all six degrees of freedom in space. The system is based on a QZS design combined with an elastic bedding approach utilizing helical compression springs. Unlike many existing QZS approaches, the proposed system satisfies practical application constraints as well as requirements common in ultra-precision manufacturing. Specifically, it eliminates conventional mechanical joints to prevent particle abrasion and avoids critical high natural frequencies within the isolation system itself. In a two-stage design process, the system is first designed for vibration isolation in all three translational directions and then extended to all six directions through the spatial arrangement of the springs. By analytically deriving and evaluating the system’s stiffness matrix, the mutual dependencies between the positions of the springs and the resulting natural behavior are investigated to determine a suitable system configuration. The findings demonstrate the feasibility of designing a QZS system capable of stable vibration isolation for a body with a mass of 2500 kg that can be tuned to achieve target natural frequencies of about 2 Hz in all six degrees of freedom, while providing internal natural frequencies of the isolation system above a target value of 1000 Hz. In comparison to a conventional vibration isolation system, reductions in the natural frequency of up to 80% were achieved.

BackgroundContinued device scaling beyond high numerical aperture (NA) EUV (NA=0.55) resolution motivates the exploration of “hyper-NA” EUV lithography systems (NA≥0.75), which promise improved contrast and resolution but introduce stronger mask 3D effects, increased polarization sensitivity, and reduced depth of focus (DoF).AimWe assess whether 0.75 NA EUV can provide manufacturable imaging for post-2030 logic interconnect patterning and identify the principal ecosystem constraints on masks, depth of focus, and photoresists.ApproachWe analyze the consequences of NA scaling on the EUV scanner architecture, including mask-side angle distributions, and we model multilayer and absorber effects. Imaging performance is quantified through rigorous simulations with source mask optimization (SMO), comparing 0.55 and 0.75 NA under unpolarized illumination and central-obscuration constraints. Two representative use cases are studied: (i) a CFET Metal0 (M0) layer with 16-nm pitch lines/spaces and line-end structures, and (ii) 2D interconnect blocks down to 16-nm pitch. Photoresist effects are included via a simplified blur and absorption model, and local EPE tail behavior is assessed using ±20% dose contours with NA-scaled defocus.ResultsModest multilayer period adjustments are sufficient to support the larger mask-side angles at 0.75 NA with limited average reflectivity impact, while low-n/high-k mask absorbers improve contrast and reduce mask-3D penalties relative to standard Ta-based absorbers. Hyper NA consistently increases the image log slope (ILS)—with the largest gains observed for dark-field masks on the most critical CFET interconnect features—and reduces local EPE extremes compared with 0.55 NA in the evaluated patterns. The Rayleigh DoF at 0.75 NA decreases to ∼20 nm, indicating the need for thinner resist films and improved focus/wafer topography control.ConclusionsUnder the modeled assumptions, 0.75 NA hyper-NA EUV offers significant contrast and EPE-tail improvements for CFET-relevant interconnect layers without revealing fundamental imaging limitations, provided that mask multilayer/absorber optimization, scanner focus control, and thin-film resist performance advance in a coordinated manner.

BackgroundEUV scanners with a numerical aperture (NA) of 0.55 have recently been introduced into the market and are now on their way to high-volume production. Given the long product development cycles for EUV scanners, the question of a successor to the NA 0.55 scanners already arises.AimWe present the concept of an EUV scanner with an NA of at least 0.75, named hyper NA.ApproachHyper NA should enable reuse of as much of the present lithography infrastructure as possible, such as masks, sources, and production equipment for manufacturing those scanners.ResultsHyper NA can support an NA of 0.75 within practically the same volume claim as the current NA 0.55 scanners. The current light source can be reused as-is, and modern mask absorbers such as low-n or high-k are also sufficient for hyper NA.ConclusionsAn increase in NA, hence the introduction of hyper NA, is the preferable successor to the current NA 0.55 scanner as it offers an infrastructure-compatible path.