Almost everyone has had the experience of quitting a job.  It’s not that hard.  You just go to your boss and say “Take this job and … find someone who will better appreciate its benefits.”  But things are more complicated when that job is with the National Security Agency.

I joined NSA at Fort Meade, MD as my first job out of college, and worked there for nine years in the 80s and early 90s.  (These were the good old days of NSA, before they started spying on Americans.)  I was very proud of my work there, making semiconductor encryption chips.  Like most of the people there, I had the highest security clearance the government offered – Top Secret Special Intelligence.  We scoffed at those piddly Top Secret clearances that they gave away like candy over at the CIA – this was the real deal.  And when you agree to accept this clearance there are certain freedoms you voluntarily give up.

The most obvious freedom I gave up was being able to talk about my work.  I could only talk about work at work, and outside of work I would end up talking about the people at work, which is more interesting anyway.  But the strangest lifestyle difference was the inversion of what is private and what is public.  Work became private, and my personal life could only remain private if it was boring.  The security office at NSA worried about its employees being blackmailed, so the skeletons in each of our closets were actively exhumed through regular security reviews and polygraph tests.

In 1988 an organization called SEMATECH was started up in Austin, Texas.  For those of us working in chip manufacturing, this was a big deal – a consortium of the largest US semiconductor companies, Intel, IBM, Motorola, sending people and money to this new research lab, with the Department of Defense contributing 50% of the budget.  It was the cutting edge of my profession.  So I thought, maybe I could get assigned to SEMATECH for a year or two to work with these great people doing important research in the interest of US security.  And moving from Baltimore to Austin for a while could be fun.  Everyone at NSA agreed this would be a great idea, but there was one problem:  it had never been done before.  And the first rule of bureaucracy is “If it’s never been done before, it can’t be done.”

Bureaucrats have a bad reputation, mostly undeserved since they are the people that actually get things done in government.  After all, the other group of government employees is politicians, and we know how much good they do.  Government employees operate under an important constraint that most people don’t realize.  At your job, you have rules, and if you break those rules you get in trouble, unless breaking the rules makes your company a lot of money, and then you get a pat on the back and a promotion.  But in the government these rules go by a different name:  they’re called laws.  You don’t just break the rules, you break the law.  A guess what – government employees don’t like breaking the law, so they work very, very hard to make sure everything they do follows the huge and continuously growing set of regulations that apply to almost everything they do.

It was hard to figure out how to make the regulations work so that I could be temporarily assigned to SEMATECH, but I was determined, and I was a pretty good bureaucrat.  Through a lot of careful study of regulations and a lot of advocacy from my boss and my boss’s boss and my boss’s boss’s boss and a team of people from finance and legal at NSA, we created a plan and made it work. I left Maryland to come to Austin to work at SEMATECH.  It was great!  I did cool work with cool people, I wrote and published papers and presented at conferences, and we made a real difference.  And when my 16 month assignment was over, I decided I couldn’t leave.  I had fallen in love with Austin.  I had found my home.  So six weeks before I was supposed to go back, I told my boss I was going to resign.  Of course, she was disappointed, but she understood my decision.

For my last week of working for NSA, I flew back to Maryland and started the week-long debriefing process.  Quitting a job with a TSSI clearance is a lengthy process.  On Monday and Tuesday it was meetings with my boss and group about projects and hand-offs, cleaning out my desk, all the normal stuff.  On Wednesday I got the library to sign off that I had no overdue books.  On Thursday I went to finance to finish up with my final paycheck and expense reports and to get their sign-off before the final security debriefing on Friday.  And this is where things went horribly wrong.  When I handed finance my exit form for their signature, they said “Sorry, we can’t sign this because our records show you owe us for an expense overpayment we made.  Once you write us a check for the overpayment we can sign your form.”  Well that’s odd, I thought, but OK.  “How much do I owe?”  “$120,000.” 

What?!?!?  $120,000?  That was more than two years of my salary!

There must have been a mistake.  No, they assured me that they had been working on these calculations for over a week, all the numbers had been double checked and approved, and they had applied the rules pursuant to joint federal travel regulation §301-74.24 scrupulously.

Regulation 301 what?

They explained to me what this regulation meant.  Suppose you are on a TDY (that’s Temporary Duty, government speak for a business trip) and you fail to meet your obligations of the TDY.  For example, you are sent on a trip to attend a conference but instead you go to the beach all week.  The government can require reimbursement of their expenses in sending you on the trip, including your salary for the week.  I was on an extended TDY, and the claim was that my obligation for the 16 month trip was to bring everything I had learned at SEMATECH back to NSA for use there, and by not returning it was the same as if I had spent 16 months at the beach instead of working.

I knew this could not possibly be right, but I discovered that finance was just doing what they were told.  My boss’s boss, who had gone to the mat two years before to figure out how to send me to SEMATECH, was pissed!  He had written a memo to finance the week before invoking this regulation, and the bureaucratic wheels had begun to spin.  It didn’t matter that I was being crushed under them.  I tried to call him, my new nemesis, but he refused my calls.  I went to his office but a security guard was stationed there to keep me out.  That night I was angry, but also determined.  I would fight this, lawyer up if needed.  I was obviously in the right, and I was not about to let some petty paper pusher ruin my life.

On Friday I went to my all-day security debriefing.  At the end of the day I handed my exit form to the security officer for that final signature.  “I can’t sign this,” he said.  “It’s not signed by finance.  They have to sign it first before I can sign.”  OK.  But what does that mean?  “It means you can’t resign.”

Can’t.  “Can’t,” he said. I can’t quit my job.  That was a thought that had never occurred to me.  My first reaction was, “Oh yeah, watch me.”  I’ll just leave and never come back.  What’s that if not quitting?

But then the security officer said four words that froze me in my tracks:  “The 59 minute rule.”

I’m guessing none of you know what the 59 minute rule is, but I knew perfectly well what he meant.  In the 1960s two NSA mathematicians had gone on vacation together for a week.  On the following Monday they didn’t show up for work.  Their boss was surprised, but assumed they had just decided to extend their trip.  The next day goes by, and then the next.  They still don’t show up.  The supervisor was getting worried, but still didn’t do anything.  The following week the two mathematicians show up on Soviet television.  They had defected, taking everything they knew about NSA with them.

NSA reacted by improving their security processes, and the 59 minute rule was born.  This is it:  If you are going to be more than 59 minutes late for work, you must call your supervisor to inform them.  If you don’t, your supervisor must contact the security office.  A security officer will then try to contact you.  If they can’t get in touch with you, they will call the FBI, and the FBI will come looking for you.

The 59 minute rule.  Every NSA employee knew and lived by this rule every day.  I instantly grasped what this meant for me.  I could choose to never show up for work again.  But if I did not officially severe my employment with the NSA, then I would have to call in by 8:59am, Monday through Friday, every week, for the rest of my life.

I was trapped, and I knew it.  When the shock wore off, I took a deep breath.  Slowly I came to understood what must be done.  The only thing that can beat bad bureaucracy is better bureaucracy.  And I could do bureaucracy as well as anyone.  I would show up the next Monday, and work the system.  And that’s what I did.  I spent the weekend reading regulations, I set up meetings, provided documentation, wrote memos and got other people to write memos, and slowly the wheels of the system began spinning in my direction.  I met with finance, with legal, with a director who had 10,000 people working under him.  More memos were written.  Consensus was formed.  I never once saw or spoke to my nemesis, but his novel legal theory was discredited, and so was he.  NSA paid me for an extra week to quit.  By Friday I had my signature and became a free man.

It is easy to take for granted a freedom that has never been infringed.  After my experience at NSA, I have always appreciated the freedom to quit.

I’m sad to report that Vladimir Ukraintsev died last Saturday. He was 64. I knew Vladimir from his long involvement in the Metrology conference of SPIE’s Advanced Lithography Symposium. He was conference chair this year, but his illness kept him from attending.

I enjoyed his rigorous mind and kind demeanor. Here are the details from his SPIE profile:

Vladimir Ukraintsev has PhD in Solid State Physics. Before joining Qorvo, Inc. Vladimir founded Nanometrology International, Inc., directed Technical Marketing at Veeco Instruments and developed metrology solutions for 6 technologies at Texas Instruments. Vladimir published over 85 articles focusing on development of metrology and characterization solutions for industrial applications.

What a difference one year makes. 2017 was my first year with solar panels, and I was very proud that I generated 96% of the electricity I used for the year (including the charging of my electric car – driving on sunshine!). Last year was not as astonishing, though still good. I generated 81% of my total home electric consumption. What changed? Consumption was up 6% in 2018, due to a much hotter summer (in Texas, AC is the big electricity hog). But even bigger, generation was down 10% due to many more cloudy days. I suspect that these two results will be the extremes, and about 90% generated solar power will be about average. I’m still happy with that.

In the middle of the picture below is Bill Arnold, a familiar face to many lithographers. He was head of lithography at AMD for many years, become the head of a technology development group (TDC) at ASML, as well as Chief Scientist (I think I have the title right) at ASML. He always gives some of the best talks at the SPIE Advanced Lithography conference, and is a senior editor at JM3. In the field of lithography, he is one of the smartest guys around.

To Bill’s right is his sister, Frances Arnold, and to his left is Donna Strickland, former president of OSA who Bill got to know while he was president of SPIE. The occasion for the fancy attire? They are all in Stockholm last month at the Nobel Prize Award Ceremony, where Frances Arnold received the Nobel Prize in Chemistry, and Donna Strickland received the Nobel in Physics.

Smart definitely runs in the Arnold family.

A little over one year ago I turned on my solar panels at my house.  Now that a full year of use has passed, I can assess their effectiveness.  My goal was to generate from solar 100% of the electricity I use.  I didn’t quite make it – I generated 96% of the electricity my house consumed.  But that included charging my electric car for the year, which means I was driving on sunshine.

If you are anything like me, you want to geek out on all the numbers.  So here they are.

The Panels

• I have 30 panels, 320 W each, for a total capacity of 9.6 kW (LG320 NeON2 MonoX Plus panels)
• Installed cost:  $28,000
• Austin Energy Rebate:  $7,500
• Federal tax credit: $6,100 (30 percent of the cost of the system)
• Net cost:  $14,400

Energy Usage and Generation for Year One

• Consumed:  13.59 MW-h
• Generated: 12.75 MW-h
• Net Consumption:  540 kW-h

Electric Car Energy Consumption

• Nissan Leaf Miles Driven:  ~8,000/year
• Mileage:  3.3 miles/kW-h
• Approximate energy consumption:  ~2,400 kW-h/year (about 18% of my total house consumption)

It has been eight years since I have been to Monterey to attend the Photomask Technology conference, commonly known as Bacus.  The name Bacus comes from the group, the Bay Area Chrome User’s Society, that originally sponsored the conference before merging with SPIE.  Fall in Monterey is a beautiful time to think about masks!  And this year we have the added benefit of thinking about Extreme Ultraviolet Lithography (EUVL) as well.  The EUVL Symposium has also decided cooperated with SPIE to host their conference and this year the two meetings have been collocated for the first time.  The result is a great synergy that makes both conferences better and gave me the necessary excuse to come again.

The synergy seems to have worked, with 560 technical attendees and about 100 more attending just the equipment exhibit (significantly more people than the organizers had expected).  I missed most of the first day of the photomask conference, but managed to notice how trendy the topic of machine learning has become.  Unfortunately, this human didn’t learn enough to make much sense of the machine learning papers.

By Tuesday the EUVL conference had begun and I greatly enjoyed Greg McIntyre’s keynote talk on EUV readiness for manufacturing.  He pointed out how lithography scaling is slowing significantly, but that “device cleverness” is taking up the slack.  These one-time innovations (like reducing the number of wire tracks making up the height of a standard cell) are helping to keep density scaling on track.  The hope, then, is that EUV will arrive and get lithography pitch scaling back up to speed.  Greg also clearly identified stochastic yield loss (extremes of roughness) as the number one problem facing EUV lithography.  When printing lines and spaces (or contact holes) is there a process window that enables no bridges and no breaks (no “missing or kissing” contact holes) at the same time?  Some significant metrology innovations may be needed to answer this question.

Many speakers discussed the imminent availability of the 250W EUV light source from ASML.  Word on the street (or at least the conference halls) says Samsung is getting this first 250W source on their first NXE:3400B scanner.  It has already shipped is supposed by be up by the end of this year with first results in early 2018.  Everyone will be waiting anxiously for those results, I am sure.

It is now clear that ASML is positioning the 3400 as the first EUV high-volume manufacturing (HVM) scanner.  It looks like most of the NXE:3300 and 3350 tools will not be upgraded to higher source powers (many will stay at 80W) and will remain learning tools.  The transition to a high power source is not an easy one.

Besides the ability to handle higher source powers, the 3400 has other improvements, most particularly the ability to use extreme off-axis illumination (out to sigma of 1) with half the pupil fill ratio (20%) of previous generation tools.  This will allow k1 as low as 0.32 (down from 0.38 or so).  Zeiss has shipped 12 of the 3400 optical systems to ASML to date.

Progress on EUV lithography as reported at the conference continues to be good, but none of the major risk factors have yet to be retired:  sufficiently high source power to achieve good throughput, making and maintaining defect-free masks, and no yield loss due to stochastic effects in lithography.

Erik Hendrickx of imec gave an update on their efforts to identify new absorber materials for EUV masks.  The current tantalum absorber requires a thickness of 55nm, resulting in tall structures with undesirable shadowing effects for tilted illumination.  Absorbers like Nickle or Cobalt could shrink the thickness below 35nm.  The problem is etching these materials, and more work is still required.

Heebom Kim of Samsung was optimistic about making defect-free masks.  They have developed their own internal actinic mask inspection system that seems to have put Samsung ahead in EUV mask making.  He claimed a current mask yield of 90%.  They will also be getting the first EUV AIMS tool from Zeiss next year in order to qualify repairs with actinic light.  But the combination of expensive inspection plus expensive repair verification plus expensive blanks will make EUV masks cost 8 times more than a 193 mask.  Wow.  I remember 15 years ago hearing promoters of EUVL saying that one of the big benefits of EUV will be cheaper masks compared to 193.  Whenever I feel a need for a good laugh, I go back and look at those early cost projects (when an EUV scanner was going to cost $20M, for example).

The resist papers focused on understanding the exposure mechanisms of EUV resists, and on reducing roughness.  Bits of progress were made on both fronts, but not nearly enough to reduce the risk of stochastic problems delaying or stopping the use of EUV in manufacturing.  I’m not sure what a major breakthrough in roughness management will look like, but it won’t look like what we saw this week.

ASML’s one billion euro investment in Zeiss is showing tangible effects as construction has begun on new buildings for the manufacture of high-NA EUV optics.  The NA=0.55 tool will be bigger than a freight train locomotive, and is essentially a two-story building.  If you thought $140M was a lot for an NA=0.33 EUV scanner, imagine how expensive the new NA=0.55 tool will be.  Then imagine higher.

On Thursday I skipped over to the photomask conference to hear about progress on multibeam tools for mask making.  Both IMS and NuFlare are making 50KeV e-beam lithography tools with 512X512 beams.  IMS has hit the market first and has already shipped many tools targeting the N7 node.  NuFlare has a beta version of their tool installed at Samsung, and so is behind IMS.  Their tool is spec’d for a higher resolution, however, and is geared toward the N5 node.  The competition is encouraging and both companies are making great progress, enabling future mask making with improved specs and reasonable write times.

Vinayan Menon of imec gave an extremely refreshing talk – an unfiltered look at one year in the life of an EUV scanner.  It wasn’t pretty.  After installing an ~80W source on their 3300, imec faced a series of ugly trade-offs:  either operate the source at near its full power and live with extremely low tool availability, or operate at a much lower power (below 30W) in order to keep the tool running and available.  There was also a fairly significant reduction in source power over time that shows rate is not equal to actual power.  Tool upgrades could help things, but those upgrades often took one or more months, so imec chose availability over peak performance.  I found another point he made intriguing as well.  A focus difference between the two chucks of the scanner was first detected as a systematic difference in linewidth roughness, suggesting that LWR might be a good focus monitor.

The closing remarks of the EUVL Symposium always include the results of a survey of the conference steering committee.  The basic survey asks which potential roadblock for EUV success is most concerning.  This year’s survey asked two questions:  what is most concerning for initial HVM insertion, and what is most concerning for continued advances in EUV lithography beyond initial insertion.  Interestingly, while the availability of a high-power source was considered the most pressing issue for initial HVM insertion, stochastic-induced variation was considered the number one issue for continued advances in EUV lithography.

Unfortunately I missed quite a few good papers at the conference.  I was busy rehearsing.  For many of the EUV crowd new to the Bacus conference, it was a surprise to discover that the conference banquet on Wednesday night was followed by an entertainment show put on by members of our community (three of the six cast members had papers at the conference).  This 40 minute show is skit humor, replete with singing, dancing, and often a fair amount of silliness (if you can’t imagine what I’d look like in an elf costume, don’t try).  I was very glad to be a part of the show this year after a 12-year hiatus.  I hope the audience had half as much fun watching it as we did putting it on.  (Photo Credit:  Bernd Geh)

I write my posts the morning after that day of the symposium. And today definitely feels like a “morning after”.  Two days of late nights at the hospitality suits followed by far too little sleep are beginning to have their effects.  Let’s see if adrenaline and desire can carry me through the rest of the week…

For those reading this blog who do not attend the SPIE Advanced Lithography Symposium, let me explain that there are seven conferences as a part of the symposium, and there are always at least five sessions happening in parallel (Wednesday morning will see all seven).  There is almost always more than one paper at any given time that I want to see, but all of my attempts at quantum entanglement with a doppelgänger have led to decoherence.  (Yes, that is the ultimate in bad nerd humor.)  Be aware that my extremely limited sampling of the symposium does not begin to do it justice.

For me, the day started with ASML’s talk on their new NXE:3400 EUV scanner, soon to be released.  As a bit of history, the NXE platform was introduced to us at this symposium in 2010.  The NXE:3100 was a “pre-production” tool, described in this way:  “With an NA of 0.25 and a productivity of 60wph this tool is targeted for EUV process implementation and early volume production at the 27nm node.”  But the NXE:3300 was to be the true production tool, targeted at 125 wph and the 22nm node.  As we all know, the 3300 missed its window for use in production, but the much improved NXE:3350 soon become the target production tool.  Since there was an upgrade path from the NXE:3300 to the NXE:3350, there was still a chance for those first 3300s to be used in production.  But after listening to Intel’s Monday talk, I am getting the impression that all the existing tools in the field are playing the original role of the original 3100.  It is the NXE:3400 that is now the targeted tool for high volume manufacturing.  It has many improvements (such as the Flex-illuminator and a membrane just above the wafer that blocks unwanted out-of-band radiation), with throughput again targeted at 125 wph.

A quick word about throughput.  Since throughput is a function of the dose used to expose the resist, and this dose is decided by the customer, ASML must make some assumption about the dose in order to specify the throughput of their tool.  In the very early days of EUV development (15 years ago), many people hoped for a 5 mJ/cm2 sizing dose.  That dream quickly relaxed to the more realistic (but still unrealistic) 10 mJ/cm2.  The throughput specs for the NXE:3100 were based on this assumed dose.  But since pattern quality improves with higher dose, the production spec of 125 wph for the NXE:3300 was based on a dose of 15 mJ/cm2.  Since then, the unforgiving onslaught of stochastic randomness brought a concession by ASML to a dose of 20 mJ/cm2.  This is now the assumption used to predict a 125 wph throughput for the NXE:3400.  This dose is also a function of the mask level being printed, with contact holes, vias, and cut masks requiring more dose (maybe twice as much, possibly more).  Since I don’t think that a dose of 20 mJ/cm2 is remotely possible due to roughness effects, significant downward scaling of the true throughput from the specified value is inevitable.

I enjoyed Tim Brunner’s paper on how to intelligently determine roughness specifications (but as a co-author, I am certainly biased).  The old ITRS specifications for linewidth roughness, useful in their day, and now rightly ignored as both irrelevant and unachievable.  Tim’s results, though, are scary.

I know that I exhibit selection bias, since I seek out the papers that deal with roughness and stochastic effects, but is seems that stochastics are everywhere at the symposium this year.  From linewidth control specifications to edge placement error, stochastic effects are almost never ignored anymore and often are admitted to be the dominant source of error in the lithography process.  After years of complaining that roughness was not getting the attention it deserved, that no longer seems to be a problem.

At the resist conference (Advances in Patterning Materials), the theme was often better roughness through chemistry.  Or if we don’t have the chemistry ready, it is often better roughness through cartoons of the chemistry.  Let me explain a test that I use when examining proposed solutions to stochastic-induced roughness:  If I don’t understand how it works, I don’t believe it.  Granted, this convolves skepticism with my own quite considerable ignorance, so I have to continually try to find my own errors in thinking and be open to being convinced.  Some ideas that fall into the “don’t understand, so don’t believe” category include PSCAR and second-order deprotection kinetics.  I hope to be convinced (preferably with good LER data).

We are half way through the technical conferences.  I have two more papers to give, and many more to listen to.

The first day of the symposium began with the awards.  I was very happy to see a great group of new SPIE fellows from our community:  Emily Gallagher of Imec, Yuri Granik of Mentor Graphics, Qinghuang Lin of IMB, David Pan of the University of Texas at Austin, Mark Phillips of Intel, and James Thackeray of Dow.  Congratulations to each of you for this well-deserved recognition.  Donis Flagello, CEO of Nikon Research Corporation of America, won this year’s Frits Zernike award (full disclosure, I nominated him).  For a history of the Zernike award, see this brief article.

For a change, I enjoyed all three plenary speakers.  Usually, at least one is a dud, but not this year.  I have to admit that I didn’t care for JSR CEO Nobu Koshiba’s disciple-like references to Ray Kurzweil and his singularity predictions (I’m not a Kurzweil fan), but it was just one part of his overall optimism for Moore’s Law.  I don’t agree that Moore’s Law will continue to the 2-nm node, but I guess it’s important that sufficient optimism exists, otherwise we’ll never try.  And we should try.

The first two talks of the EUV session were keynote addresses.  Britt Turkot of Intel painted a fairly rosy picture of the progress of EUVL towards manufacturing readiness.  “It’s been a long and winding road,” and we still have a ways to go, but the eight NXE:3300s and six NXE:3350s in the field are giving semiconductor manufacturers opportunities to shake out enough of the reliability problems to enable process learning.  Tool availability continues to creep up (past the 70% mark), and mask making has progressed to the point where Intel has made “multiple” defect-free EUV masks.  Intel showed data on “adders” (defects that get added to the mask during use) and reiterated their message from last year that that production without a pellicle is not an option.  Thus, it makes sense that she listed the availability of a manufacturing-capable pellicle as the biggest risk.

She also mentioned stochastics, saying that “CD and edge placement variability is a deal breaker.”  But then her conclusion slide said that resist performance won’t gate the introduction of EUV.  I didn’t know what to make of these mixed messages, especially when she explained that the target dose for EUV manufacturing was 20 mJ/cm2.   At that dose, there will be plenty of CD and edge placement variability.

Seong-Sue Kim of Samsung was similarly encouraged by EUVL improvement.  He expressed amazement at the progress in mask blank defectivity saying it had reached the benchmark of 5 defects per blank that he thinks can enable manufacturing.  He also said that the mask blistering problems he mentioned last year have largely been solved.  For resists, he thinks that current performance is good enough for 7nm development, but sensitivity (at low roughness) needs to be improved for production.  Of course, everyone agrees with that statement.  The question is how to do it.

My favorite technical talk was Bill Hinsberg’s modeling of metal-oxide resists – a much needed start.  John Biafore gave a great paper modeling millions of contact holes at various EUV conditions and looking for stochastic-related failures.  He expressed skepticism at any possible breaking of the RLS trade-off (“resolution, LER, sensitivity – pick two”).

Finally, I was extremely gratified by the reception I received to my tutorial talk and was grateful for the many people willing to stay till 6:30pm to hear me speak.  Thanks to Eric Panning and Ken Goldberg and the EUV Lithography conference for giving me such a great opportunity to talk about stochastic-induced roughness.