Historical research guide · 1073 · 2003

Performance and scaling of a dense plasma focus light source for EUV lithography

Igor V. Fomenkov · Richard M. Ness · Ian R. Oliver · Stephan T. Melnychuk · Oleh V. Khodykin · Norbert R. Böwering · Curtis L. Rettig · Jerzy R. Hoffman

In one minute

This is the plasma focus doing a day job. Igor Fomenkov and his colleagues at Cymer report on a dense plasma focus — two coaxial electrodes that drive a current sheet down a barrel and collapse it into a pinch — built not to study plasma but to sell light. Every pinch of xenon radiates extreme ultraviolet at 13.5 nanometres, the wavelength the semiconductor industry needs to print the smallest circuits, and a factory tool needs that pinch to happen thousands of times a second, for years, without eroding its own electrodes into the optics. The team report about 60 millijoules of in-band light per pulse at a conversion efficiency near 0.45 percent, bursts at 4 kilohertz and continuous running at 1 kilohertz, 13 kilowatts of heat pulled out of the discharge region by water, anodes lasting beyond ten million pulses, and a pinch about 0.38 millimetres across whose brightness into different directions tracks almost perfectly pulse by pulse.

Connection

Chapter 9 treats the pinched plasmoid as a real, engineerable object rather than a curiosity, and this paper is the proof in the strongest possible form: a dense plasma focus running as a commercial product line, water-cooled, fired thousands of times a second, with electrode lifetimes and dose stability quoted like any other industrial component. Every number here — pinch size, shot-to-shot repeatability, erosion rate, heat removed — is an engineering figure someone had to earn.

From the source

Improvements to electrode geometry, the gas recipe, the gas delivery and pre-ionisation systems and the plasma driver raised the in-band conversion efficiency at 10 joules by about 85 percent over two years, giving a source delivering 60 millijoules per pulse at a conversion efficiency of 0.45 percent for about 12 joules of dissipated energy, with a pulse-to-pulse root-mean-square energy stability of about 7 percent at low repetition rate.Section 2, In-band EUV energy and conversion efficiency, page 808; Figures 1 and 2
The authors state the gap that remains: for the output power a high-volume manufacturing tool needs, the conversion efficiency must rise to at least 0.7 percent, and such a system could conceivably produce more than 100 watts of collectable in-band extreme ultraviolet power — but the present electrode configuration has reached a plateau, so further gain requires additional modification of the electrode structure.Section 2, page 808, closing paragraphs
High-repetition-rate operation was demonstrated with a 4 kilohertz resonant charging system: bursts of 1000 pulses at 2 kilohertz and 800 pulses at 4 kilohertz, with root-mean-square in-band energy variation of 7 and 9 percent respectively, while second-generation direct water-cooled electrodes and a water-cooled vacuum vessel raised the maximum steady-state repetition rate five-fold in a year from 200 hertz to 1000 hertz, at which 13 kilowatts of heat was removed from the discharge region with a coolant flow of about 27 litres per minute — roughly 50 percent of it from the anode, 40 percent from the cathode and 10 percent from the vessel.Section 3, High-repetition rate operation, page 809; Figures 3 and 4
A transmission grating spectrometer — zirconium filter, 50 micrometre slit, a gold grating of 5000 lines per millimetre and an extreme-ultraviolet-sensitive back-illuminated camera — recorded complete single-pulse spectra at a source-size-limited resolution of about 0.3 nanometres, and comparing the relative intensities of the xenon ion-stage line groups shows the average plasma temperature rising as the xenon flow rate falls, making the instrument a real-time tuning tool for the pinch.Section 4, Spectral characterization of the source, page 810; Figures 5, 6 and 7
The emission is highly repeatable in direction as well as in energy: off-axis in-band output measured with 45-degree multilayer mirrors is directly proportional to the on-axis output with almost perfect correlation, two off-axis monitors correlate at 0.95 through a 2 kilohertz burst with 7 percent root-mean-square energy stability, and at-wavelength one-to-one imaging with two spherical multilayer mirrors gives a typical source size of 0.38 millimetres full width at half maximum with a centroid fluctuation usually under 0.05 millimetres.Section 5, Angular stability, page 811, Figures 10 to 12; Section 6, Source imaging, Figures 13 and 14; Section 8, Summary
Anode erosion under pulsed heat load scales with a material figure of merit the authors call the impulsivity, the melting temperature times the square root of the product of thermal conductivity, density and specific heat, and this holds for tungsten, copper, aluminium, molybdenum and stainless steel but not for graphite; refractory anodes extended electrode life from a few million to more than ten million pulses, and with a multichannel debris shield stopping electrode material so effectively that surface analysis behind it finds only carbon, ultraviolet-catalysed carbon deposition is now the limit on producing a clean source.Section 7, Debris mitigation, pages 811 to 812, Equation 1 and Figures 15 to 17

The way in

https://doi.org/10.1117/12.505303

Historical source-access record. This guide links to the original; it does not reproduce the paper text.

SOURCE REACHED AND READ IN FULL, REPRODUCED IN PART. Page 807 of the paper carries the line Proceedings of SPIE Vol. 5037 (2003) © 2003 SPIE, so this is not a United States Government work and no Creative Commons statement exists anywhere on it; only the authors’ own abstract is reproduced here, and it is given verbatim, with the line-break in at-wavelength closed up and the greater-than sign in item two written out in words. The paper itself, printed as pages 807 to 813, was read end to end on 2026-09-08 from the authors’ own copy, which Cymer published on its technology page and which survives in the Internet Archive as the 22 March 2012 capture of cymer.com/files/pdfs/Technology/2003/Performance_and_Scaling_of_a_Dense_Plasma_Focus_Light_Source_for_EUV_Lithography.pdf; every claim locator below points to a numbered section, equation or figure of that paper. The publisher’s deposited metadata spells the sixth author Bowering; the paper itself prints Böwering, and the paper’s spelling is used here. The work was done at Cymer, Inc., 16750 Via Del Campo Court, San Diego, California, with plasma modelling in collaboration with the University of Nevada, Reno. Companion sheets: the same group’s earlier optimisation paper at /library/stm-deef169c45, the Lee model code for plasma focus radiative behaviour at /library/stm-a9cc428c6d, and foam-target experiments on the PF-1000 plasma focus at /library/stm-77cc1c549c.

How to cite it

Igor V. Fomenkov, Richard M. Ness, Ian R. Oliver, Stephan T. Melnychuk, Oleh V. Khodykin, Norbert R. Böwering, Curtis L. Rettig, Jerzy R. Hoffman (2003) Performance and scaling of a dense plasma focus light source for EUV lithography doi:10.1117/12.505303

Where it sits in the curriculum

Plasmoids, charge clusters and the orbs