Translate

Monday, February 18, 2013

Zplasma to Attend SPIE Advanced Lithography 2013 Seeking Xenon Z-pinch EUV Source Funding


In the previous weeks I've written two blog articles which speak to the engineering complexities and capital intensity associated with the current development of EUVL (Extreme Ultra Violet Lithography) technology. An intensive effort is under way to provide a next generation 13.5 nanometer EUV light source required to print ever smaller nanoscale transistors and computer chips. Moore's law is marching on to again double the number of transistors on a chip and is being fueled by massive investments by major players in the semiconductor industry. Recently Intel, TSMC and Samsung invested over $6 Billion in ASML, a large well capitalized lithography equipment industry leader based in the Netherlands. ASML in turn purchased Cymer a critical supplier of laser technology providing key components in ASML's product line. The major investment in ASML was to ensure timely development and availability of HVM (High Volume Manufacturing) certified EUV lithography systems for shipment in late 2013 and projected 450 mm HVM production two or three years hence. EUV source development has been problematic and presents many challenges. Xtreme Technologies/Ushio and Gigaphoton and others are competing in the EUV source market with well funded programs. These efforts are being coordinated with imec, a major research and development center in Belgium. While billions are being spent on developing laser based, Sn plasma source technologies, innovative and cost effective engineering solutions are being researched with out the capital infusion and celebrity afforded some of the world's largest semiconductor equipment manufacturers.

Enter Zplasma. Funded by the U.S. Department of Energy in 1998, the University of Washington in Seattle, U.S.A., began construction of a ZaP Flow Z-Pinch Experiment. Henry Berg, an entrepreneur-in-residence at the university, started working with Nelson and Shumlak in 2011 to design and create a 200 watt commercial light source based on Sheared Flow Stabilization (SFS) technology, and then designed a proof of concept prototype to prove that the physics worked at the reduced size necessary for optical compatibility with steppers. Realizing that with xenon gas SFS produces a long pulse of EUV light at a wavelength of 13.5 nanometers, and that the semiconductor industry needed light sources of this type, the university filed for and was granted two patents on the invention. Zplasma was formed with Henry Berg as its CEO. The university and Zplasma executed an exclusive license agreement covering the two patents and the transfer of the commercial source prototype to Zplasma. More information on Zplasma may be found here at the company's web site.

During my most recent visit to the Zplasma website I sent a request to info@Zplasma.com for more detailed information on the company and its Z-pinch EUV source technology. Expecting a response from a MARCOM staffer I received a reply directly from Henry Berg, Zplasma's CEO. Having many questions about Zplasma and its Xe based EUV source technology, Henry and I exchanged questions and answers via email which allowed me to gain additional insight into Zplasma's feature/benefits and their significance given current concerns with impediments to HVM (High Volume Manufacturing) EUV source development. Much of the content of this blog entry is verbatim response to my questions from Henry Berg discussing the salient points of Zplasma technology and its differentiation from competitive EUV source technology. After discussion with Henry I thought this format might be the best way to convey a thorough discussion on a variation of Z-pinch ion source design in advance of SPIE Extreme EUV Lithography IV later this month.

I should point out that during the course of my blogging activities I have never held any financial interest in any of the companies or organizations I have mentioned. I've ceased my market trading activities over a year ago and currently do not hold positions in any stock or security (this disclosure does not preclude possible future investment or interest activities). My immediate interest is in rejoining active participation in the semiconductor industry which has been my career, and contributing meaningful discussion to critical issues facing our industry. In my opinion, Zplasma's patented technology could be of great interest to the EUV lithography community and the larger semiconductor industry. SPIE Advanced Lithography 2013 is an excellent opportunity for the EUV lithography community to initiate a thorough evaluation of Zplasma.  What follows below is a brief introductory description of Zplasma's current research and development effort as conveyed by Henry Berg. The questions related to the following discussion are mine followed by Henry's well articulated answers.



Henry Berg, CEO Zplasma, Inc.
“When we started the commercialization effort two years ago, I did not anticipate that it would be so incredibly difficult to get industry support to develop this technology. The primary barrier has been that since nobody has ever stabilized an EUV-emitting plasma before, the industry has no understanding of the benefits of a stable plasma. Every previous source of EUV light has been based on unstable plasma, where you have a very brief time to produce as much EUV light as possible, hence the focus on CE, the switch to molten tin and industrial lasers, and all of the problems with debris damaging the collector optics.We are very excited about the potential our source has to enable high volume EUV lithography. Our source technology is so much simpler than LPP or LDP that it will not only be much less costly to build and to operate, but it will also be durable and reliable. We need financial and development support to take the prototype we have and develop it into the HVM source it needs to become.”



Discussion

Zplasma's Xenon Plasma EUV Source Technology

Questions by
Thomas D. Jay, Semiconductor Industry Consultant

Answers provided by
Henry Berg, CEO Zplasma, Inc.



Could you explain Zplasma's Z-pinch function, its vacuum pressure regime, and plasma formation process?
"The vacuum chamber is pumped down to a few microtorr. A small amount of xenon gas is then injected. Power is applied to ionize the gas into plasma. The plasma is accelerated. The Z-pinch forms from the flowing plasma, pinches down and the central portion of the pinch emits a pulse of EUV light. The available plasma is consumed, the pinch dims, and the plasma exhaust exits through a small hole in the endwall of the outer electrode. There is a long time between pulses, our target operation at 7 kHz means there is a pulse every 143 microseconds. Since pulses last on the order of 10 microseconds, there is plenty of time for the gas plenum to recharge the acceleration region with gas for the next pulse."



In addition to the Z-pinch phenomenon are external magnets utilized to confine and/or control the plasma?
"No. Gas is injected into the annular space between an inner electrode and an outer electrode. The gas is ionized into plasma and accelerated axially. The inner electrode then has a rounded tip that allows the Z-pinch to form between the tip of the inner electrode and the outer electrode end wall. The outer electrode has openings to allow side-on optical collection of the light. The outer electrode also has a small hole at the end that allows the plasma exhaust to exit and be collected away from the optics."



What differentiates Zplasma's Z-pinch technology from competing systems?
"There are two things going on here. All DPP sources, including ours, use a Z-pinch. Gas is injected between two electrodes, ionized into plasma, and a pinch forms. Cymer and Xtreme both built DPP sources that worked this way. The problem was that the pinch that formed was highly unstable, so it blew apart. This meant that they couldn’t get meaningful high power levels without melting their electrodes, and even at low power levels debris was produced and they couldn’t collect enough light. Our core innovation is Sheared Flow Stabilization (SFS). With SFS, you inject the gas upstream of the pinch assembly point. The gas is ionized into plasma, and the plasma is accelerated axially and only then allowed to come together. The result is a Z-pinch which is stabilized by the moving plasma. The duration of the pinch can be adjusted by changing the amount of gas injected or adjusting the input power waveform."


Is there a specific plasma pressure regime which optimizes the Sheared Flow Stabilization effect?

"Sheared Flow Stabilization requires the gas to be injected into the acceleration region in a certain way and with a certain uniformity to create the proper sheared flows in the plasma. Unlike unstable discharges, our input waveform can significantly affect the duration of the pulse, we will optimize our output pulses in this manner for maximum EUV output. We have deliberately not factored in any such optimizations that we are not sure about. We do not have a power supply capable of outputting an adjustable waveform."



Your literature comments, “The duration of the pinch lasts ten to a hundred times longer than other Z-pinches (quoted at 100 nanoseconds), then has a controlled end.” Please describe all of the parameters acting on a “controlled end” to the pinch.
"The length of the pinch can be stretched by adding more gas and stretching the input power waveform. The pinch has a natural end because the available gas is consumed, but it fades quietly rather than ending explosively like the other DPP sources. You can also end the pinch early by ramping down your input power waveform. There are six critical advantages to SFS:

a) Longer Light Pulse: SFS pulses are 10 to 100 times longer due to their stable nature, allowing for more light collection.

b) High Power without High CE: Long SFS pulses and side-on optical collection access increase throughput and lower required CE, enabling HVM operation with xenon and eliminating the need for molten tin.

c) No Debris: SFS ends the plasma pinch without explosive termination, eliminating high-energy debris.

d) Low Instantaneous Power: SFS pulses allow for EUV light production without the high instantaneous power levels that cause electrode thermal stress and ablation.

e) Dose Uniformity: SFS allows the length of each EUV pulse to be adjusted under control of the power supply, allowing for extremely accurate dose uniformity.

f) Adjustable Geometry: SFS makes pinch geometry adjustable for optical matching to stepper IF."



What is the current Xe Z-pinch EUV source CE (energy Conversion Efficiency)?
"Our prototype CE is limited because we are powering it with a capacitor bank that rings. A proper pulsed power supply that can shape the output waveform will solve that problem. We have 2% BW CE of 0.5% in the lab now, that will go to 1.5% with the right power supply."



Given sufficient funding, what is the projected CE on a larger, scaled for power Z-pinch EUV source?
"Stable plasma means we do not need to focus on improving CE. Our 200W target assumes we never get more than 1.5%. Of course, we will be optimizing CE as well to increase beyond that. The theoretical CE limit with xenon is 2-4%."



Does the Zplasma ion source provide increased Xe CE when compared to competing Xe EUV source designs? If so what is the percentage in CE gain?
"All discharge-produced-plasma sources (Cymer DPF, Xtreme DPP, Zplasma Stable DPP) use a Z-pinch as the core EUV-producing mechanism. SFS is what has driven our current CE so high despite the ringing capacitor bank, we are already where Cymer and Xtreme were with optimized power supplies. Our HVM 200W target assumes we only get to 1.5% CE, but we will optimize for higher CE as part of the development process, and I would love to see us exceed 2%. But we didn’t assume ANY new inventions or innovations to hit 200W, so improvements would just drive us over 200W."


What components must be scaled to achieve desired HVM power levels?

"We need a high voltage pulsed power supply that can run at 7 kHz and we need to couple the electrodes to a forced-flow chiller system for cooling."


Can you describe an optimal 7kHz power supply design specification or is there a commercially available model you've selected?

"We cannot identify a commercially available model, though I would love to use something off the shelf if possible. Other discharge sources have all used custom-built magnetically switched supplies, and our source would work with any of those supplies. 7 kHz was chosen so as not to push the limits of previously developed pulsed power supplies. Our plan is to build an IGBT-switched supply that is fully adjustable so we can stretch our pulses out even further and produce more EUV light. We have identified a partner who can produce the power supply we need."


What variable parameters must this source power supply provide (example: pulse width, amplitude, wave form shape, pulse rise time ramp)?

"The supply we would have made would allow all of these to be adjusted. There is a lot of exciting work still to be done extending SFS Z-pinches and increasing the amount of EUV light that is produced from each pinch."


Would a new “scaled for power” HVM design be price and performance competitive with other EUV systems of similar HVM performance?
"Our source technology is much cheaper and simpler than other ways of generating EUV light. There are no other systems of similar HVM performance, but I estimate that our source will be about 10x cheaper to make and 5x cheaper to operate than the present HVM EUV sources."



What is the peak resolution at 13.5 nm for Xe EUV spectra? Is there comparable (or better) spectral resolution and out of band performance for Xe Z-pinch as compared to a Sn source?
"EUV source power is measured in a 2% bandwidth around 13.5 nm, thus 2% BW EUV. Radiation outside of this region is called out-of-band (OOB) radiation. So our goal is to deliver 200 W of 2% BW EUV to the stepper intermediate focus (IF). We produce the same light as previous DPP sources, but do it from a much smaller EUV emission volume and in a stable manner. I haven’t talked about it much, but our OOB results should be outstanding. We do not have the OOB light from a laser to contend with. Our source will need to be coupled with an ellipsoidal multilayer mirror to couple the light to the IF. One cool thing about how that works is Bragg layer reflection will reduce the transmission of any OOB light that is present. But it is not having to worry about light from a laser pulse that is the main OOB differentiator."



Given SFS and your comments on dose error, does Zplasma's Z-pinch produce fewer neutrals than Sn based plasma systems?
"One key advantage of SFS is that you can shorten or lengthen the duration of the pinch under control of the power supply. So the source can adjust based on light being delivered to the IF to keep the EUV light energy dose within very tight bounds. There will be no significant debris from the source, as the root cause of high-energy debris is the violent instabilities in the plasma. Present LPP and LDP sources have an exploding Sn plasma droplet, so high energy ions and fast neutrals are both produced. Previous DPP sources all had the xenon Z-pinch that blew apart, also showering out debris. Our core focus has been eliminating the root cause of the DPP debris. Once we have our source operating, we will place optical collector coupon material in the right places and conduct durability tests. In theory, fast neutrals are produced by collisions with fast ions, so the dramatic reduction in fast ions should reduce fast neutrals. Our plasma is also moving orthogonally to the collector optics, so no debris mitigation is needed."



With a reduced debris profile utilizing Xe instead of Sn source feed material, will the system accommodate an H2 plasma cycle for mitigation of residual Xe induced particulates or surface contamination of the source/optics?
"Our system would indeed accommodate an H2 plasma cycle if needed. We have done a whole bunch of runs with hydrogen as part of our lab work testing the prototype. Our goal is to operate without generating any high-energy debris that hits the optics."



Given the reduction in high energy ions and fast neutrals is it possible to minimize or eliminate shot noise by ramping and shaping the power supply output and resulting EUV pulse characteristics?
"We do not know the answer to this question. It is possible that there will be shot noise variation compared to LPP/LDP based on how our source produces the photons, but we do not expect any differences, and I am not sure I see a mechanism that could result in differences. This is something that we will look at when we have the ability to do so."


If sufficiently funded, could an HVM, high power, Zplasma Z-pinch EUV source be delivered on time and in quantity to accommodate requirements at ASML and other stepper vendors?

"Yes, if ASML and/or Nikon fully supported the stepper integration process. Our focus is on the core technology to produce the light. We would need Carl Zeiss or Media Lario to make the mirror and Fraunhofer to make the mirror coating. Development of this technology with the right funding and support could be very rapid – no new innovations are required."


What is the estimated cost to scale components to desired HVM power levels?

"$6-8M to get to a fully functional 200W HVM prototype. This is for the core EUV-producing unit only. Stepper integration will require adding an ellipsoidal multilayer mirror."


Recent major R&D investments have been made in EUV source technology by Intel, TSMC and Samsung. Zplasma's EUV source design with all its potential features and benefits seems worthy of major funding in the “best of breed” competition currently in progress. With the approach of SPIE Extreme EUV Lithography IV in February 25 – 28, have you attracted new interest in funding?

“We are talking to several companies about financial backing and partnering with us to develop our source. Recently, we have been talking to IMEC, and are hoping to have further discussions with Global Foundries, Intel, TSMC, SK Hynix and Samsung at the upcoming SPIE conference in San Jose.”

If Zplasma's Xe Z-pinch EUV source can be scaled to power levels required for HVM at existing or further optimized CE levels, significant improvements in MTBF and debris mitigation might be achieved. A more simplified EUV source design with stable plasma operation could enhance dose uniformity and device product yields offsetting further diversification of investment in the search for optimal EUVL for HVM.

I'd like to thank Henry Berg for his personal response to my inquiries concerning Zplasma and his enthusiastic assistance in providing engineering data affording an insightful overview of his company. 

Henry Berg, CEO of Zplasma will be attending SPIE Advanced Lithography, February 24 - 28, 2013 in San Jose, CA.

Thomas D. Jay
Semiconductor Industry Consultant
Thomas.Dale.Jay@gmail.com
For more information on Zplasma visit www.Zplasma.com

For information on the SPIE Advanced Lithography 2013 Extreme Ultraviolet Lithography IV program click on the link below:
http://spie.org/app/program/index.cfm?fuseaction=conferencedetail&export_id=x12540&ID=x10947&redir=x10947.xml&conference_id=1039349&event_id=996835



Friday, February 1, 2013

The 13.5 nanometer Physical Cliff?


As SPIE Extreme Lithography IV approaches will the EUV lithography program fall off a physical cliff?

SPIE is the International Society of Optics and Photonics. Its membership is comprised of engineers who research the scientific manipulation and applications of light. On February 24-28 in San Jose, CA, a large contingent of SPIE membership will meet to discuss current progress on an exotic EUV (Extreme Ultra Violet) light source scheduled for implementation in next generation, nanoscale computer chip manufacturing. For those outside the semiconductor industry, EUVL (Extreme Ultraviolet Lithography) is a next generation, extremely short wavelength light source (13.5 nanometers) providing improved lithographic capability to print ever smaller, nanometer scale transistor circuit patterns on computer chips. The time and expense invested in the development of EUV lithography spans many years and totals billions of dollars.   Recently, a few individuals (very few) have suggested to me that the physics challenges of 13.5 nanometer EUV lithography might be insurmountable and the continuing escalating expenditures to resolve EUV source power, uptime and mask issues (to name a few), will further delay the implementation of a work around strategy to preserve Moore's law.   Moore politics in the semiconductor industry? The recently celebrated investments in ASML by Intel, TSMC and Samsung collectively approximate $6 Billion, the price of a new state of the art wafer fab.  One might ask why not use these funds to build another foundry and utilize existing 193 nanometer manufacturing technology to creatively double or triple pattern DSA (Directed Self Assembly) device designs. This work around scenario might be an alternative in the shorter term but the economics and physics for this argument are not sustainable. At the 2011 EUVL Symposium, Rudy Peeters of ASML presented a compelling illustration (page 5 of his presentation) of the cost reductions attainable with EUV over 193 nanometer lithography. Given the same product (in one of his examples), a 193 nanometer process would entail as many as 5X the number of process steps with a >50% increase in cost.   EUV's superior image resolution and higher k1 value  at 13.5 nanometers extends lithography performance and ultimately reduces cost over time (k1 is a process evaluation coefficient that encapsulates process-related factors). These cost savings estimates are well within the ball park so long as critical EUV performance issues are resolved satisfactorily. Intel, TSMC and Samsung have invested heavily to ensure EUV performs on time. With additional time and expensive fine tuning, ASML will ramp production and Moore's law will again enable a new generation of semiconductor products, funding further R&D.

Is there an impending physical cliff for 13.5 nanometer EUV technology and beyond? Will complex physics issues limit EUV viability?   The semiconductor industry confidently says no and is also in concurrent pursuit of BEUV (Beyond Extreme Ultra Violet) lithography as a follow on evolutionary path.   BEUV 6.7 nanometer technology development will require additional time and investment and will no doubt foment additional engineering debate. Moore's law will be continually pushed to its limits but the current critical focus is on the timely delivery of HVM (High Volume Manufacturing) EUV lithography, and 450mm process/metrology tools. As the EUV program evolves, source designs will undergo modification and upgrades to reach required performance specifications but the over all program is moving forward. Semiconductor front end equipment manufacturers who are not EUV/450mm capable in a timely fashion risk the eventual loss of market share and possible forfeiture of future viability in the semiconductor manufacturing industry.

The key to success in the development of EUVL/BEUVL and related semiconductor technologies is the pooling of knowledge and distribution of R&D investment costs. The semiconductor foundries and consortiums have the capital resource to pursue technology development that can be cost prohibitive to a self funded corporate R&D program. However, collaboration on advanced R&D can be a delicate balancing act between managing intellectual property concerns and promoting the general welfare of a capital intensive industry. An excellent recent example of this concern is the protracted dispute between Apple and Samsung over intellectual property related to smart phone software. In spite of the on-going litigation, Apple A5 and A6 processors are being manufactured in an Austin, Texas wafer fab build by Samsung. Both companies benefit from the arrangement and share a major portion of the smart phone market place while making financial news headlines in the process.

Equally important to the pooling of financial resources is the cross linking of engineering groups collaborating on R&D programs. This interaction reduces development time by eliminating concurrent, redundant development programs and inefficient rediscovery of existing knowledge. As an example, I often recount an experience in which I visited a customer's corporate R&D facility to discuss a deep UV photostabilization application for his process. We began our discussion in the hallway outside his lab.   After a few minutes our discussion attracted the attention of another resident researcher who happened by. Without introduction he stopped and silently listened in on our conversation.  Our discussion began at 320 nanometers, a popular wavelength for photostabilization. We soon realized that a newly proposed process material would better stabilize at a higher wavelength in the 340 nanometer range. We wondered out loud where we might find a 340 nanometer range UV light source. Hearing this, our silent companion beamed a broad smile and blurted out, “I have what you're looking for.  I fabricated a cadmium vapor lamp for an experiment years ago and haven't used it since then. I thought someone might need it one day. It's in my desk, I'll go get it.”   We all laughed to celebrate a very brief but successful collaboration in which my customer discovered the answer to his question was a few doors down the hall from his lab. I didn't sell anything that day but planted the seeds for future collaboration and sales activity.   I often wondered what the collaborative mean free path might have been in that laboratory, and how long it might have taken for my two friends to discover their in house problem and solution without my presence as a catalyst. A good semiconductor industry statistician can probably provide an answer, but that's another story.


Thomas D. Jay
Semiconductor Industry Consultant
ThomasDaleJay@gmail.com
www.linkedin.com/pub/thomas-d-jay/26/aa3/499
www.ThomasDaleJay.blogspot.com

The Technology High Ground


For information on the SPIE Advanced Lithography 2013 Extreme Ultraviolet Lithography IV program click on the link below:



For additional information on the recent Intel, TSMC, Samsung investment in ASML, click on one of the referenced Bloomberg New links below:
http://www.bloomberg.com/news/2012-07-09/intel-agrees-to-buy-10-stake-in-asml-for-about-2-1-billion.html

http://www.bloomberg.com/news/2012-08-05/taiwan-semiconductor-agrees-to-invest-1-38-billion-in-asml.html

http://www.bloomberg.com/news/2012-08-27/samsung-to-buy-3-stake-in-asml-for-503-million-euros.html

http://www.bloomberg.com/news/2012-10-17/asml-to-buy-cymer-for-2-55-billion-to-speed-up-euv-development.html


For streaming updated technology news from Google, scroll to the very bottom of this page.