EUCLIDES: European EUV lithography milestones

09/01/1999

Abstract

In Europe, the EUCLIDES program has obtained preliminary throughput and cost-of-ownership results in its comparison of laser plasma and synchrotron sources. In addition, researchers have doubled the EUV flux at the wafer to increase throughput and have obtained impressive results with optical coatings and optical substrate fabrication. An EUV test bench is now under construction, moving this consortium toward its 2000 goal of beginning assembly of a beta system.

In this article:
EUVL system
The optics
Synchrotron source
Conclusion
About the authors

Jos P.H. Benschop, Anton J.J. van Dijsseldonk, ASM Lithography, Veldhoven, The Netherlands,
Winfried M. Kaiser, Carl Zeiss, Oberkochen, Germany
David C. Ockwell, Oxford Instruments, Oxford, England

As 2000 approaches, the European EUCLIDES program (Extreme UV Concept Lithography Development System) is nearing completion of its first phase. EUCLIDES began in 1998 with system integrator ASM Lithography (ASML), optics manufacturer Carl Zeiss, and synchrotron source supplier Oxford Instruments forming an industrial consortium for R&D to evaluate extreme ultraviolet lithography (EUVL) based on their strengths and expertise [1]. ASML coordinates the program, which also includes subcontractors Philips Research and CFT, TNO-TPD, FOM-Rijnhuizen, PTB, and FhG-IWS (Fig. 1).

Figure 1. EUCLIDES program organization. Participants include the Research Lab of Philips Electronics; TNO-TPD, a Dutch institute for applied scientific research; FOM-Rijnhuizen, a Dutch research institute for plasma physics; PTB, the German synchrotron facility; and FhG-IWS, a German institute for applied R&D. Click here to enlarge image

Other regional EUV programs include the US consortium EUV-LLC/VNL [2] and Japan's ASET program [3]. Recently, ASML announced its participation in the EUV-LLC program, linking it to EUCLIDES.

At inception, the partners decided to execute the EUCLIDES program in two phases:

• Phase 1, set for 1998 through 2000, targeted development of key technology needed to demonstrate technical solutions for the list of possible EUVL "showstoppers." Here, since 1998, EUCLIDES has focused on R&D for mirror substrates, high reflectivity multilayer coatings, and vacuum stages. Program researchers have compared plasma and synchrotron sources. In addition, the consortium has investigated total system architecture to ensure a viable system concept that meets the requirements of the semiconductor industry at an acceptable cost of ownership.
• Phase 2, slated to begin in 2000, will eventually lead to an EUVL "beta" system (timing is still being discussed) and, eventually, commercial systems.

Everyone looking at lithography's future knows there are enormous risks and uncertainties facing the various new next generation lithography technologies. For example, a defect-free mask is considered to be a potential showstopper for EUVL. Within EUCLIDES, however, the partners are concentrating on core competencies within the system. Here, high-risk items include a high power EUV source with acceptable cost of ownership, multilayer-coated optics, vacuum stages, and contamination control. Mask development, on the other hand, is a matter that will be done in cooperation with potential end-users.

The view of the EUCLIDES partners is that, compared to competitive methods, EUVL using 11-14nm radiation has a high probability to become the next generation lithography of choice for several reasons:

• Due to the absence of stochastic coulomb interaction that couples throughput with resolution, EUVL (as well as x-ray lithography) provides higher throughput potential (>40 300mm wafer/hr) down to the 35nm technology node compared to electron projection lithography and ion projection lithography.
• Compared to x-ray, EUVL is a reduction projection technology; this greatly reduces requirements for mask technology.
• EUV has the best fit with the current optical lithography technology base, including precision optics and scanning stages.

EUVL system

The work executed to date by ASML and subcontractors Philips Research & CFT and TNO-TPD has addressed system architecture, vacuum stages, and an EUVL test bench, specifically physical and technical limitations of source, optics, reticle, stage, wafer handling, and contamination control subsystems.

For example, engineers at TNO-TPD have invented a novel technique for contamination control that mitigates resist outgassing without the need for a window between the projection optics and the wafer. The principle consists of producing a gas flow toward the wafer in a tube surrounding the optical beam (Fig. 2). This gas flow limits debris flow toward the optics. Model calculations show that outgassing generated at the wafer that would otherwise enter the optics compartment is reduced by >6 orders of magnitude by using a gas flow of ~2 liter/sec at ~0.01 mbar.

Figure 2. EUCLIDES novel mechanism for mitigating optic-coating problems from resist outgassing.Click here to enlarge image

This purging technique eliminates any need for a silicon membrane to protect the optics, thereby enabling 90% energy transmission with a commensurate throughput advantage. In addition, the relative simplicity of the components used means that maintenance is less frequent, compared to a membrane subject to coating.

Early tests show that debris flow in a nonoptimized configuration is already reduced by a factor of 100. A design study is now in progress to integrate this component on an EUVL tool and to perform more detailed measurement of its advantages.

For all methods of lithography, stage accuracy is constantly increasing together with stage size and stage mass. Maintaining system throughput in the face of these developments presents a major challenge for mechanical design, servo control, and metrology. EUVL adds the need and complexity of a vacuum stage.

Currently, various vacuum stage concepts, which have patents pending, are being evaluated, including the experimental verification of their performance.

Figure 3. The water droplet source on EUCLIDES test bench.Click here to enlarge image

The first EUCLIDES EUVL test bench is under construction at Philips Research Labs. Already the energy source, a laser produced plasma based on a droplet-source concept pioneered by L. Rymell and H. Hertz [4], is operational; Fig. 3 shows a schematic of the source and some technical details. This source is clean and relatively simple, and, extended with a differential-pumping scheme, has been run continuously.

Figure 4. Measured and modeled spectrum from the water droplet laser.Click here to enlarge image

Simplified models have been used to predict the spectral distribution from this energy source [5], comparing closely to the measured spectrum (Fig. 4). In addition, the calculated plasma radius of approximately 57µm at 3.6ns agrees well with the measured plasma radius of 65µm. The intensity of this source at the 13nm line is approximately 2.7x10¹² photons/shot/sr, which corresponds to a conversion efficiency of 9x10^-5 sr^-1. This conversion efficiency is in accordance with that found by Hertz et al., who obtained 4x10¹² photons/shot/sr at laser energy of approximately 0.7 J/shot [6].

Currently for EUCLIDES, an at-wavelength interferometer is under construction that will be added to the test bench [7]. This extended-source interferometry scheme will allow efficient use of the spatially incoherent laser-plasma source. The challenge is to achieve subnanometer vibration and drift stability for the interferometer frame in vacuum, commensurate with the targeted 20 milliwaves RMS wavefront accuracy (i.e., 0.26nm at 13nm). Such mechanical requirements will be needed in future EUV lithography tools. Through analysis of at-wavelength interferograms, EUCLIDES researchers can evaluate the imaging performance of EUVL's multilayer mirrors.

The optics

To date, work done by Carl Zeiss and its subcontractors FOM and FhG-IWS (providing multilayer coatings) and PTB (providing multilayer coating characterization at EUV wavelengths using a synchrotron source) includes activities on optical and mechanical design, optics fabrication, multilayer coatings, and metrology.

Click here to enlarge image

EUVL is in principle the extension of optical lithography to 11-14nm wavelengths. Within this range, there are no transmissive materials suitable for lens fabrication and mirrors have maximum reflectivity of around 70%. Therefore, the EUVL optical subsystems - the illuminator and projection system - only work with a limited number of aspherical mirrors as optical elements. As with present day lithography systems, the quality of the optical elements, especially those of the projection systems, is crucial for the imaging performance of the whole system.

Process development at Carl Zeiss for the fabrication of mirror substrates has been performed using the "ELT2" secondary mirror from Sandia Lab's three-mirror EUV projection lithography system design. This mirror has a concave surface with ~2µm asphericity; to obtain good imaging performance aspheric mirrors are required, but the exact amount of asphericity depends on optics design.

The basic quality terms of these mirrors and their specific impact on imaging are "figure," which impacts wavefront aberrations; mid-spatial frequency roughness (MSFR), which impacts stray light and contrast loss; and high spatial frequency roughness (HSFR), which impacts reduced reflectivity and intensity loss.

With the ~2µm asphericity mirror, the priority in the first development phase is to improve figure and MSFR because both have strong impact on imaging performance.

Figure 5. Reflectivity curve of two different Mo/Si multilayer coatings.Click here to enlarge image

Three copies of mirror substrates have been manufactured in Zerodur (see table). Initially, all three substrates show a relatively large HSFR. (The larger HSFR is an effect of the polycrystalline structure of Zerodur at the actual state of development. On amorphous materials like fused silica, a HSFR of 0.07nm was achieved with a similar process.) However, Substrate 2 in the table, overall considered the best mirror, was reworked using an improved manufacturing process with the aim of achieving better HSFR. This work resulted in HSFR and MSFR values close to the final specification. While the figure of Substrate 2 was not determined, the figure obtained with Substrate 1 is the best figure achieved to date, worldwide, on an aspheric EUVL substrate.

Good process convergence is needed for future economic manufacturing of these optical elements. Therefore highly reproducible surface metrology is necessary. Based on a Carl Zeiss digital interferometry method [8], reproducibility of 0.08nm RMS can be achieved. Also, the convergence of the figuring process is impressive: all results above were achieved with <15 processing steps.

Researchers at FOM have been developing multilayer-coating technology based on e-beam evaporation. Specifically, they are investigating in detail the influence of a number of process and design parameters of Mo-Si multilayer coatings [9]. So far, optimization of the evaporation processes has produced a reflectivity of 68.6% at a wavelength of 12.75nm (Fig. 5). (The at-wavelength characterization of the coatings was done at the PTB laboratory at the BESSY synchrotron in Berlin.)

Synchrotron source

Oxford Instruments has been working on the design and evaluation of a synchrotron-based source for EUV lithography. The light output specification requires the maximum power of useful light within the bandwidth, set by the multilayer optics. The radiation must be homogeneous in the horizontal plane to illuminate a suitable ring field on the reticle with 1% homogeneity.

An initial analysis has indicated that multipole wigglers, which yield the required light output, have better throughput potential than radiation from bending magnets and undulators [10].

Synchrotron light is produced by the action of bending a high-energy electron with a magnetic field. On initial synchrotron light sources, the light was produced by the bending magnets used to form a closed path. Modern synchrotron sources are fitted with insertion devices. These are arrays of magnets that make multiple bends back and forth about the straight path between two bending magnets. Within boundaries set by the electron beam parameters, the x-ray light output can be selected by the insertion device design. "Wiggler" describes an insertion device that produces high broadband power. "Undulator" describes an insertion device that produces high power density, but low total powers, in a set of narrow spectral lines.

Figure 6. Wiggler power spectrum - W/horizontal mrad in 2% bandwidth integrated over all vertical angles.Click here to enlarge image

At present, multipole wiggler designs show that a 3.5m long wiggler can generate 8W of in-band power with a 1A, 500MeV electron beam. A wiggler has a broadband output with the wavelength of maximum power set by the magnetic field and electron energy (Fig. 6). Once the magnetic field is set, the output power is directly proportional to wiggler length and stored beam current. The period of the magnetic field within the device sets the horizontal extent of the output beam. For the wiggler design now being considered, the output beam horizontal is ±33mrad. The vertical opening angle is the same as for a bending magnet. The geometrical collection efficiency of this source can easily be made 100%.

The next stage is to design collimator optics suitable for this device and then optimize the optics and source together to provide the most suitable illumination of the reticle. A significant element of this design will be how to cope with the large amount of out-of-band power produced.

Oxford researchers have already explained the present design of a suitable storage ring (Fig. 7) for the multipole wiggler that yields a throughput of twenty-five 300mm wafers/hr for an EUV lithography system with a six-mirror projection system [10].

Figure 7. Storage rings for two-wiggler layout.Click here to enlarge image

Two designs are being pursued, one with two wigglers ("Eu-Two") and one with four ("Eu-Fouric"). In both designs, there are two additional straights to contain the RF and injection components required by a synchrotron. In addition, these designs have radiation shield walls that are required for personnel protection and scanners in a potential site layout. Engineers at Oxford Instruments envision that in a large IC fab several synchrotrons would be installed in the same shield enclosure.

In a storage ring, the electron beam current decays with time due to several mechanisms such as scattering off of gas molecules in the vacuum vessel and scattering of electrons within the beam. Consequently, the beam current must be periodically restored by injection of fresh electrons into the system. Injection can be done by one of two techniques:

• Ramp the synchrotron down to take low energy electrons directly from a pre-injector (microtron or linac) and then ramp up to full energy.
• Keep the storage ring at full energy and inject electrons that have been ramped up to full energy by an intermediate booster ring.

The second option costs more, but potentially allows higher stored beam currents, less injection time and higher throughput. Our selection will be based upon cost of ownership.

Certainly, advantages include the fact that a synchrotron produces no debris and that both synchrotron storage rings and multipole wigglers are established and presently achievable technology (no new ground needs to be broken). A wiggler produces a light beam of restricted solid angle compared to a point source.

EUCLIDES is evaluating the manufacturing costs of a synchrotron in detail. Initial estimates show the price/source for a "Eu-Two" with booster to be the same as the projections for the laser plasma source in series production. We expect, however, that the laser plasma source, in contrast to the synchrotron, will require periodic replacement of key components, therefore the operational cost is likely to be lower for the synchrotron source. Long-term reliability of synchrotrons in industrial use has been proven with HELIOS 1 at the IBM facility in East Fishkill [11]. While the synchrotron does require extra facilities, such as a shield wall and personnel protection, we believe these can be provided quite simply, at low cost and in a standard factory, as has been done at the HELIOS test facility at Oxford Instruments.

The excellent process extensibility, the worldwide development efforts, and a good fit to existing technology base are key factors that give EUVL a high probability to become the preferred lithographic technology for below 70nm technology nodes. The European EUCLIDES program, which is focusing on a number of the technical challenges to EUV lithography, has demonstrated some milestones. For example, Carl Zeiss has demonstrated that it is capable of producing optical substrates close to, or even exceeding, final specifications. Optics should therefore no longer be considered as a technology showstopper. Overall, the EUCLIDES partners are confident that EUV lithography will be a technological success and well positioned to commercialize, which is the next phase of the EUCLIDES program.

Acknowledgment

The authors thank numerous team members at ASML, Carl Zeiss, Oxford Instruments, Philips, TNO-TPD, FOM, PTB, and FhG-IWS for their contributions to this program. This work has been supported in part by the European Commission within the ESPRIT program (Project EP 28160). Zerodur is a registered trademark of Carl Zeiss.

References

1. Jos P. H. Benschop, Winfried M. Kaiser, David C. Ockwell. " EUCLIDES, the European EUVL program," to be published in the SPIE 3676 Symposium on Microlithography 1999.
2. R.H. Stulen et al., "Progress in the development of extreme ultraviolet lithography," to be published in the SPIE 3676 Symposium on Mircrolithography 1999.
3. S. Okazaki, "EUV program in Japan," to be published in the SPIE 3676 Symposium on Mircrolithography 1999.
4. L. Rymell, H.M.Hertz, "Droplet target for low-debris laser-plasma soft X-ray generation," Optics Communications, 103, pp. 105-110, 1993.
5. Jeroen Jonkers et al., "Laser produced oxygen plasmas," to be published in The Annals of the New York Academy of Sciences (Proceedings of 2nd International Symposium on Heat and Mass Transfer under Plasma Conditions, ed. by P. Fauchais), 1999.
6. H.M. Hertz et al., "Debris-free soft x-ray generation using a liquid droplet laser-plasma target," in Applications of laser plasma radiation II, SPIE, 2523:,88-93, 1995.
7. Matthieu Visser, Martijn Dekker, Petra Hegeman, Joseph Braat, "Extended-source interferometry for at-wavelength test of EUV-optics," to be published in the SPIE 3676 Symposium on Mircrolithography 1999.
8. European Patent EP 0455 218 B1.
9. Eric Louis et al., "Reflectivity of Mo/Si multilayer system for EUVL," to be published in the SPIE 3676 Symposium on Microlithography 1999.
10. D.C. Ockwell, N.C.E Crosland, V.C. Kempson, "A synchrotron light source for EUV lithography," to be published in these proceedings.
11. L.G. Lesoine, J. Leavey, "IBM advanced lithography facility: The first five years," Solid State Technology, July 1998.

About the authors

Jos P.H. Benschop received his MSc and PhD from Twente University, Netherlands. Benschop is heading the EUCLIDES program and managing research at ASML, de Run 1110, 5503 LA Veldhoven, The Netherlands; ph 31/40-230-3968, fax 31/40-230-4455, e-mail [email protected].

Anton J.J. van Dijsseldonk worked at Philips Research for 10 years and European Southern Observatory for 19 years. He is a senior scientist at ASML, Veldhoven, The Netherlands.

Winfried M. Kaiser received his diploma in physics from the University of Tübingen and a degree in optics from the Technical University of Stuttgart. Kaiser is responsible for product development and strategy of the Lithography Optics Division at Carl Zeiss, Postfach 1380, D-7082 Oberkochen, Germany; ph 49/7364-20-4690, fax 49/7364-20-4509, e-mail [email protected].

David C. Ockwell is a physics graduate from The University of London. He is engineering manager for synchrotron products at Oxford Instruments, Osney Mead, Oxford OX2 0DX, England; ph. 44/1223-427400, fax 44/1223-425050, e-mail [email protected].