page 1 of 27 Laser-produced plasma for EUV lithography UCLA MAE Department Thermo/Fluids Research Seminar Series 29 June 2007 M. S. Tillack

page 2 of 27 The end is in sight for semiconductor lithography based on direct laser exposure Efforts to drop the wavelength to 157 nm ended unsuccessfully Innovations such as immersion (high n) or phase-shift masks (interference) have pushed the limits of feature size at a given wavelength ( Current technology

page 3 of 27 EUV lithography has become the frontrunner “next generation” technology Discharge (DPP) and laser (LPP) alpha tools have been built LPP has several advantages: higher collection efficiency, more manageable thermal loads and debris, more scalable to HVM wafer mask EUV source laser

page 4 of 27 At UCSD we are developing technologies for a LPP light source using Sn targets Why 13.5 nm? Why Sn? Key challenges: Maximize in-band emissions, Minimize debris damage Research at UCSD: * Low-density and mass-limited targets * Wavelength and pulse length optimization * Double-pulse irradiation * Gas and magnetic mitigation UCSD Laser Plasma & Laser-Matter Interactions Lab

page 5 of nm is a large, but credible next step Mo/Si, 6.9 nm period 32 cm diameter ~70% reflectivity Yulin et al, Microelectronic Engineering, vol. 83, Issue 4-9, (2006) Transition to reflective system results in smaller NA (0.15 vs. 1), requiring much smaller wavelength for increased resolution (NA~1/2f) Multilayer mirrors as low as 13.5 nm are commercially available

page 6 of 27 The UTA of Sn is an efficient source at 13.5 nm Konstantin Koshelev, Troitsk Institute of Spectroscopy Sn +6 Sn +7 Sn +8 Gerry O’Sullivan, University College Dublin Sn Xe Sb Te I Light comes from transitions in Sn +6 to Sn+ 14, between 4p 6 4d n and 4p 5 4d n+1 or 4d n-1 (4f,5p) Lighting up these transitions, and only these transitions requires exquisite control of laser plasma

page 7 of 27 Optimum conditions for EUV light generation: a plasma temperature of ~35 eV, … … and laser heating where EUV light can escape Region of EUV emission Calculation of non-LTE ionization balance using Cretin Interferometry data 840 ns after pre-pulse Good News: These conditions can be achieved using relatively “ordinary” Joule-class pulsed (10 ns) lasers

page 8 of 27 Laser plasma is notoriously difficult to control  Steep spatial gradients  Strong time dependence  Intimate dependence on temperature  Non-LTE behavior (rate-dependent atomic populations) t (ns) Hyades: double-sided illumination of foam T (eV) n e (10 20 /cm 3 ) R (cm) t (ns)

page 9 of 27  Nomarski interferometer  Transmission grating EUV spectrometer  In-band energy monitor  Faraday cup  2 ns visible imaging  Spatially-resolved visible spectroscopy (2 ns)  EUV imaging at 13.5 nm … and difficult to measure

page 10 of 27 Maximum in-band emission with minimum out-of-band saves $$ on the laser and cost of ownership, and reduces optic damage. Intel’s EUV MicroExposure Tool Cymer’s HVM EUVL source concept Challenge #1: Maximize conversion of laser light to in-band EUV output  Techniques to narrow the UTA  Optimized pulse length & wavelength  Avoidance of reabsorption Our work:

page 11 of 27 Low density targets produce a narrower UTA  We attempted to reduce debris by reducing the target density.  The optical depth at 13.5 nm is only ~7 nm of full density Sn. Beyond that, light is reabsorbed.  An unexpected advantage is a narrower spectrum. Targets provided by Reny Paguio, General Atomics 100 mg/cc RF foam 0.1-1% solid density Sn e.g., 0.5%Sn = Sn 1.8 O 17.2 C 27 H 54

page 12 of 27 Conversion efficiency can be optimized by choosing the laser pulse length & wavelength Conversion efficiency is higher at longer wavelengths Laser absorption occurs at a lower density, allowing the EUV light to escape more efficiently. 2 ns gives better CE, but longer pulses may be more cost-effective

page 13 of 27 A spectral dip can occur due to reabsorption, degrading the conversion efficiency e.g., spot size is one of several factors that contribute to opacity control  CE depends on a balance between emissivity and opacity  In a smaller spot:  Lateral expansion wastes laser energy  less emissivity  Lateral expansion reduces plasma scale length  less opacity CE vs. laser intensity (I=2  W/cm 2 ) Emission spectrum

page 14 of 27 Challenge #2: Minimize optic damage from high-energy particles Laser-produced plasma generates debris and energetic ions  Optic lifetime must be >30,000 hours  Debris can be cleaned, but…  Energetic ions damage multilayer mirrors  Gas stopping  Magnetic stopping  Gas plus magnets  Double-pulsing  Full density  Mass-limited  Mass-limited plus gas Our work:

page 15 of 27 Gas stops ions, but also stops EUV photons Calculated Conversion efficiency at 45˚, 78 cm E-mon vs. attenuation calculation Ion yield at 10˚, 15 cm Faraday cup vs. SRIM estimate  Hydrogen is best, but not good enough (and nasty)  Collisionality in plumes also affects their range

page 16 of 27 Magnetic diversion is partially effective, but not sufficient 5 GW/cm 2 free expansion velocity v=6x10 6 cm/s aluminum

page 17 of 27 A magnetic field produces a synergistic effect in combination with background gas photoionization peak appears when gas is present Faraday cup time-of-flight measurements at 10˚, 15 cm from target 100% dense

page 18 of 27 We recently discovered a technique that dramatically reduces the ion emission energy Camera gate Pre-pulse: Wavelength532, 1064 nm Pulse length130 ps Energy2 mJ Intensity~10 10 W/cm 2 Spot size300 µm Main pulse: Wavelength1064 nm Pulse length7 ns Energy150 mJ Intensity2  W/cm 2 Spot size100 µm Low-energy short pre-pulse forms target; main pulse interacts with pre-plasma. Degrees of freedom to control performance. 2 mm Sn slab

page 19 of 27 Ion energy was reduced by a factor of 30! v=L/t, E= 1/2 mv keV Energy spectra of ions vs. time delay

page 20 of 27 The main pulse interacts with a pre-formed gas target on a gentle density gradient Pre-plume density profile 840 ns after the pre-pulse 130 ps pre-pulse, 2 mJ, 532 nm 840 ns1840 ns 440 ns10 ns Thermal plasma cold plume 500  m

page 21 of 27 The location of absorption of the main pulse is clearly displaced away from the surface Region of EUV generation Nomarski interferometer lasers single pulsedouble pulse pre-plasmamain plasma

page 22 of 27 At the optimum delay time, no loss of conversion efficiency is observed CE Particle energy reduction factor Delay (ns)

page 23 of mTorr Hydrogen TOF spectrum 10 mTorr Argon TOF spectrum Gas is more effective at stopping ions that already have their energy degraded In both cases, the predicted transmission of 13.5 nm light is >95%

page 24 of 27 “Mass-limited” targets should reduce the debris loading, without loss of light output  Thin films were fabricated using e-beam evaporative coating of Sn on plastic and glass  Film thicknesses from 20 to 100 nm, as well as foils from 1 to 10 µm were tested

page 25 of 27 Unlike single-pulse results, ion energy was reduced using a pre-pulse on thin coatings  Acceleration with single pulse likely due to low-Z substrate  Double-pulse pump beam never reaches the substrate

page 26 of 27 MCP Pre-pulse + gas + mass-limited target could satisfy requirements of a practical EUVL source We need better diagnostics to measure vanishingly small yields

page 27 of 27 Acknowledgements Contributors: Yezheng Tao, S. S. Harilal, Kevin Sequoia Financial support: General Atomics, William J. von Liebig Foundation, Cymer Inc., LLNL and the US Department of Energy

page 28 of 27 Center for Energy Research Laser-matter interactions at UC San Diego Thermal, mechanical and phase change behavior Relativistic laser plasma (fast ignition) optics damage Laser plasmas: EUV lithography HED studies (XUV, electron transport) Laser ablation plume dynamics, LIBS, micromachining