Challenges and solutions for extreme ultraviolet lithography at 22nm

According to the International Technology Roadmap for Semiconductors, extreme ultraviolet lithography (EUVL) at 13.5nm is the most likely source for patterning the next generation of semiconductor chips at 22nm and smaller.¹ Finer and finer dimensions are needed every year to reduce costs and improve performance. Leading tool manufacturers are planning this integration in 2013 and have promised beta tools for evaluation within the next year. The throughput of wafers in prototype EUVL alpha-machines is 2.5 wafers per hour (wph), which is far below the 100wph required for high-volume manufacturing. In this brief review we describe the technical challenges ahead and highlight work at our Center for Plasma Material Interactions (CPMI).

The top critical issues facing EUVL² are the following: a reliable, high-power source with 100W at the intermediate focus (IF) and 5MJ/day to enable sufficient throughput to make the technology affordable; availability of defect-free masks; resist resolution, sensitivity, and low line-edge roughness (LER); protection of reflective masks during storage, handling, and use; and quality and lifetime of projection and illuminator optics. Each of these issues will require one to several orders of magnitude improvement before EUVL can be considered production-worthy. Missing a target specification by as little as 30% could mean EUVL would not print the first production 22nm chips. Source power has been increasing every year, but not at the rates projected. Likewise, resist sensitivity has improved, but slowly. Progress has been made on each of the critical issues—and there do appear to be viable ways to solve them—but there is still a considerable way to go. Targeted applied research is the key to overcoming these hurdles.

Figure 1. Measured ion and neutral particle debris at the intermediate focus (IF), not in the line of sight of the EUV pinch. The presence of these energetic neutral atoms was not expected.

At CPMI, the commercial EUV discharge-produced plasma (DPP) experimental setup called XCEED (XTREME commercial EUV experiment and diagnostic chamber) is equipped with tools³ to detect ions, neutral particles, and charged particles from the Z-pinch (a type of plasma device) and the secondary plasma formed during the laser pulse. Our DPP is almost identical to the ones used and being proposed to make chips at 22nm. By examining what comes out of it, we can work to devise methods for increasing power. XCEED also studies lifetime issues employing inductively coupled reactive ion etching (ICP-RIE) to remove contaminating tin from the collector optics. Figure 1 shows measured ion and neutral debris using an electrostatic spherical sector analyzer combined with a neutral detector at the output of the line of sight of the EUV pinch.

Figure 2. Presence of electrons measured using Faraday cup (FC) outside of the collector shell.

Figure 3. Schematic of a new combined detector to measure the debris and contamination at the IF. XCEED: XTREME commercial EUV experiment and diagnostic chamber.

Energetic neutral atoms were also found at the IF, which connects to the lithographic stepper. Faraday cup measurements show the presence of electrons outside of the collector shell, as illustrated in Figure 2. Moreover, x-ray photoelectron spectroscopic analysis of the silicon witness plates confirmed carbon and oxygen contamination at the IF. A triple Langmuir probe placed inside the chamber detected a very weak plasma in the vicinity of the probe surface. We are constructing a new combined detector that will measure the debris and contamination at the IF as shown schematically in Figure 3.

ICP-RIE tin-cleaning experiments using a mixture of argon and chlorine gases to etch the tin from the ruthenium mirrors in a small collector mock-up showed a significant tin removal rate of ~100±10nm/min.⁴ Experiments in a full-size collector mock-up showed that significant cleaning occurs even with only a 4cm gap between shells near the source. The CPMI has also developed a nanoparticle-cleaning system for masks and wafers called PACMAN (plasma-assisted cleaning by helium metastable atom neutralization).⁵ PACMAN works by using ions and the energy from the metastable helium atoms to volatilize the contaminants. The PACMAN cleaning process resulted in a 100% particle removal efficiency of polystyrene latex (PSL) test particles with sizes between 30 and 200nm. Figure 4 shows scanning electron microscope images before and after cleaning of 30nm PSL particles.

Figure 4. Scanning electron microscope (SEM) images before and after cleaning of 30nm PSL particles using the PACMAN technique.

Figure 5. SEM images of unprocessed and processed samples showing the significant reduction in LER.

Finally, an LER reduction technique developed at CPMI uses an ion beam at grazing incidence unidirectional with the features. LER reduction numbers are reported at both short and long spatial wavelengths using neon, helium, and argon ion beams. Tests have achieved a decrease in LER from 9.8±0.67 to 5.5±0.86nm for 45nm 1:1 lines using an argon beam at 500eV for 6s at an 85^° angle of incidence (see Figure 5).

CPMI intends to continue working in each of these areas: diagnostics of the source, cleaning of the optics, and particle removal for masks and wafers. Through targeting specific applied research to the most pressing problems facing the introduction of EUVL, we hope to enable this technology to be introduced to manufacturing in a timely manner.