Tuesday, March 11, 2014

A Perspective on EUV Lithography Feb. 2014 The NIF Shot Heard Around the World Advanced Lithography V 2014 provided no encouraging news on further development of EUV power output for advanced semiconductor HVM. During the week of the conference it was announced that a recently shipped ASML NXE:3300B [A] stepper/scanner with a 30 Watt EUV source failed during its trial run at TSMC. Accidents happen. Over the years I have witnessed several spectacular meltdowns of high energy/high value wafer fab equipment. Recovery is rapid as wafer fab crash teams resolve such incidents in short order. 

The quest for higher power EUV has been a greater challenge than originally anticipated. Unfortunately this latest occurrence at TSMC punctuated a ten year continuance of forward looking statements in which ASML/Cymer repeatedly anticipated imminent arrival of EUV power levels of 100 watts or more.

The Ultimate Shot Noise

Interestingly, the search for advanced, future semiconductor EUV lithography technique has been an on-going effort that began many years ago. In 1994 the National Ignition Facility (NIF) at Lawrence Livermore National Laboratory began the Laser Science and Technology (LS&T) Program [1] whose research would chart the course for many future advanced technology projects. The National Ignition Facility  [1A] team is to be congratulated on their most recent August 13, 2013 experiment which produced output power greater than that of input levels. The NIF utilizes 192 individual high energy lasers focused on a small deuterium target with the goal of emulating the physics of our sun and unleashing large amounts of fusion energy.

See Wikipedia NIF photos:


A subset of the Laser Science and Technology program was AMP, the Advanced Microtechnology Program, providing research and development resources in semiconductor imaging and detection. AMP was considered a show case example of the U.S. Department of Energy's (DOE) efforts to transfer and commercialize newly developed technologies to U.S. commercial interests. The semiconductor industry now had the attention of world experts in plasma and light source technology.
The Birth of Laser Produced Plasma EUV 

The NIF began work on Laser Produced Plasma EUV. Plasma produced from Sn (tin) or Xe (xenon) enables the creation of a 13.5nm EUV light source, an item of key interest to next generation lithographers in the semiconductor industry. The NIF built a 13.5nm Laser Produced Plasma (LPP) test stand which successfully provided this desired wavelength of vacuum EUV. AMP and its associated EUV research and development would become LS&T's largest program. 

Later, three DOE laboratories; Lawrence Livermore, Lawrence Berkeley, and Sandia Laboratories in California went on to form the Virtual National Laboratory (VNL) to further research and develop extreme ultraviolet lithography (EUVL) technology. The VNL was funded by the Extreme Ultraviolet LLC, a consortium of Intel, Motorola, Advanced Micro Devices, and Micron Technology. Semiconductor industry heavy weights were now interacting commercially with the formidable technology base of the U.S. Department of Energy. The three year, $250 million venture was dedicated to developing EUVL for commercial manufacturing of computer chips and to foster migration of the technology to semiconductor production facilities by 2010. Each national laboratory contributed expertise to this effort; Lawrence Livermore (optics, precision engineering, and multilayer coatings), Sandia Labs (systems engineering, photoresists, and light source). Berkeley contributed its Advanced Light Source capability, generating EUV light to characterize optics and resists at the nanometer scale. SEMATECH now similarly sponsors and benefits from the development of actinic EUV metrology at the Lawrence Berkeley Center for X-ray Optics (CXRO). 

A Fifteen Year Chronology of EUV
Source Development 

- On May 6, 1998 Arthur W. Zafiropoulo, Chairman, CEO and president of Ultratech, formed United States Advanced Lithography LLC, and reached an agreement with EUV LLC (the consortium of Intel, Motorola, Advanced Micro Devices, and Micron Technology) in order to further develop and transfer EUV technology to American lithography manufacturers. Zafiropoulo wanted to ensure U.S. semiconductor equipment vendors remained competitive in the world economy by producing EUV lithography tools on American soil. [1B]

- On June 24, 1999 ASML of the Netherlands reached an agreement with EUV LLC, (the consortium of Intel, Motorola, Advanced Micro Devices, and Micron Technology) to participate in the further development and transfer of EUV technology to semiconductor lithography manufacturers. By participating in the EUV program facilitated by EUV LLC, ASML became a defacto beneficiary of the EUV research conducted by the U.S. DOE. Martin van den Brink, executive vice president of marketing and technology at ASML was later quoted as saying “While EUV is expected to have the highest throughput and most extendable resolution, the complexity of non-optical techniques requires the parallel evaluation of multiple options."  ASML moved rapidly to secure its position in the future lithography market place.

- In June, 2006 Cymer put its first LPP EUV source into operation.

- In November 2007 Cymer reported achieving 100 watts of EUV burst power on its LPP source. [2]

- On May 14, 2008, Cymer reported the achievement of continuous EUV source operation for over one hour at an average power level of 25 watts. [3]

- In July 2009 Cymer announced the shipment of an LPP source to ASML, claiming it had achieved 75 watts of “EUV exposure power” and anticipated 100 watt power levels within 90 days enabling 60 wafer/hour throughput on 300mm wafers. [4] 

- In 2010 Cymer reported achieving 100 watts of EUV peak power for brief periods but was only able to provide 10 watts of continuous EUV output. ASML began evaluating three potential suppliers of EUV sources; Cymer, Gigaphoton and Extreme Technologies. [5]

- In July 2011, at a company earnings conference call Bob Akins, then Cymer's Chairman and CEO reported “As a result of increased source availability and stability improvements, the eight (EUV) sources have cumulatively produced greater than 40 megajoules of EUV since March of this year and it is sufficient to expose greater than 3,000 wafers”. [6]

- In February 2012, Cymer reported shipping three 8 Watt EUV sources but 20 watt upgrade shipments for NXE-3100 systems were delayed. [7]

- In May 2013, Cymer's EUV source power output was still short of HVM targets. ASML completed the Acquisition of Cymer in a cash and stock transaction estimated to be $3.7 Billion. [8]

- As reported on February 24, 2014 during SPIE Advanced Lithography V, an NXE:3300B, was shipped to TSMC with an integrated 30 Watt EUV source from ASML/Cymer, failed during testing but was later repaired. [9]

A Perspective on Extreme Ultra Violet Lithography 
March 11, 2014  

In 2008 Arthur W. Zafiropoulo, Chairman, CEO and President of Ultratech, estimated that EUV lithography systems could be premium priced as high as $15 to 20 million each, affording a significant market opportunity. The interplay of the NIF and EUV programs promulgating the current lithography initiative has exposed two starkly differing cost center/ROI models. The National Ignition Facility took tewlve years to build and houses 192 high power laser bays 300 yards long, producing 500 Terawatt laser “shots” (500 Trillion watts) focused on a single deuterium pellet with the goal of replicating the fusion energy created in the core of our sun. Funded by the U.S. Government's Department of Energy, the NIF facility cost $3.5 Billion to construct. Given current ASML pricing at $120 Million each, a quantity of 25 ASML EUV stepper/scanners, each anticipated to produce 150 watts of front end LPP EUV illumination, are now estimated to cost $3 Billion. If we utilize Mr. Zafiropoulo's original high end estimate of $20 Million per stepper, the cost for the same 25 EUV steppers is reduced to $500 Million (the number to the right of the decimal point on the NIF's construction cost). The collective investments in ASML made thus far by Intel, Samsung and TSMC actually exceed the NIF's $3.5 Billion construction cost. Is an ASML equipped semiconductor front end EUV lithography fab (25 EUV steppers) really at cost parity with a U.S. government sponsored fusion energy project?  Given current wafer fab construction costs approximating $5 to 6 Billion, ASML's recently quoted EUV lithography pricing is unprecedented.  This singular discrepancy in the semiconductor industry's cost continuum has displaced Moore's Law as a viable operand.  EUV technology originally developed within the U.S. DOE/NIF program has been transferred to cooperative multinational interests outside any U.S. based cost control infrastructure.  Electron beam lithography as an alternate HVM solution was never funded on a large scale leaving ASML as a defacto sole source for nanometer scale HVM. This is why the EUV program is on hold. It's time to call the accountants, get costs under control, and restore U.S. based best of breed lithography competition to the semiconductor industry.  We all applaud the efforts of our friends at ASML who have made extraordinary strides in the development of EUV.  However, with ASML as the primary beneficiary of the NIF's EUV Laser Produced Plasma program, the U.S. based semiconductor equipment industry should be competing with them for both economic and strategic considerations.
The current and on going status of EUV endures as a great drama for those of us with keen interest in the semiconductor industry and the phenomenon of Moore's Law. Although ASML stock holders should continue to benefit from their dominate front end market share, it would appear ASML's customer/investors are getting less return on their subsidy of EUV as progress on HVM power output development has stalled. Although there are few remaining EUV players (none with ASML's front end market share), the current economic complexity of the EUV program compounded by the throttling of the 450mm initiative has quashed enthusiasm for large scale investment in new, competitive EUV and alternative lithography technologies targeting CDs <28nm. The current over capacity status at many fabs has also delayed further investment in tweaking strategic product positioning, best illustrated by Intel's idling of newly constructed fab 42. 

With regard to next generation semiconductor products, we might choose to continue optimizing cloud based CPUs and servers as another way of offsetting increasingly heavy processing demands until nanometer scaling enabled by restoration of the EUV initiative or SEMATECH's alternate choice, electron beam lithography enables us to re-institute the spirit of Moore's Law.

Please join me in supporting the National Photonics Initiative, SPIE and the United Nations proclaimed International Year of Light 2015.

Thomas D. Jay 
Semiconductor Industry Consultant
Thomas D. Jay YouTube Channel

Corporate or private entities mentioned in this article are the respective owners of their logos, trademarks, service marks and intellectual property. Unless otherwise disclosed, Thomas D. Jay has no financial interest in companies referenced in blog articles or other published media communications. No representation is made to either buy or sell securities. Opinions expressed by Thomas D. Jay are his own. Thomas D. Jay does not employ or otherwise utilize/authorize third party agents to express his opinions, represent his interests or conduct business on his behalf except where formally contractually designated.

Acknowledgements and Reference Links

 Photonics for a Better

National Photonics Initiative

 [A] ASML NXE:3300B, ASML Web Site

[A1] Ultratech

[1] Extreme Ultra Violet Lithography, Imaging the Future

[A-G] NIF Photos, Wikipedia 

[1A] National Ignition Facility, Lawrence Livermore National Laboratory  

[1B] May 6, 1998 Business Wire

[2] November 30, 2007 Business Wire 

[3] May 14, 2008 FABTECH

[4] EE Times July 13, 2009 

[5] August 19, 2011 Engineering and Technology Magazine, by Chris Edwards

[6] July 21, 2011, Morning Star 

[7] February 3, 2012 Semiconductor Engineering, by Mark LaPedus

[8] May 30 2013 UT San Diego, by Mike Freeman

[9] February 24, 2014 Semiconductor Engineering, by Mark LaPedus

Related blog articles of interest
by Thomas D. Jay

August 2013
The SCRUM of All Fears 

January 2013