Fundamental Research and the Future of Semiconductors

Fundamental Research and the Future of Semiconductors

Show Video

Robert Margetta: So I want to thank everyone for joining us today. Semiconductors have been an issue that have been in the news a lot lately. The White House Congress Robert Margetta: Industry groups have all identified these as a top priority for us innovation. They've also identified this as an issue where manufacturing capacity within the US. Robert Margetta: Is a is a bit of an issue right now that we need to address, there's a shortage of chips as we see increase production of devices that require them and there there are questions about supply chain security.

Robert Margetta: So this is something that NSF National Science Foundation has looked at for years and we've supported research that has Robert Margetta: Examined semiconductors that every step of the way from manufacturing supply chain security to the innovations that are going to build Robert Margetta: The foundation for semiconductor science in the coming 10 2030 years it's talking about that and to end to give the reporters on the call some some good background. Hopefully some good sourcing for future stories we have an absolutely great panel. And we're going to Robert Margetta: I'm going to introduce them. Now we're going to go through some discussion with them and then we're going to open it up to questions. So on our panel.

Robert Margetta: And this list is in everybody's invitation. If you need name spellings and affiliations, we have Dawn Tilbury she is assessed assistant director for our Engineering Directorate. Robert Margetta: Margaret Martino see she's NSF assistant director for our Computer and Information Science and Engineering Directorate.

Robert Margetta: And Thomas coach with these a program manager who focuses on semiconductors and micro electronics, Tom, you know, I should have double check your, your last name pronunciation. So please correct me later if I got that wrong. Robert Margetta: We also have April brown Duke University. Daniel Holcomb from UMass Amherst Melinda Jeffries L from Boston University. Robert Margetta: Joan Red Wing from Pennsylvania State and soup, Ashish Mitra from Stanford University. And so with that I'm going to turn off.

Robert Margetta: I'm going to turn our attention to our NSF panelists. First, and I'm just going to ask them if dawn, Margaret, Tom. I was somebody could frame this issue for us. I'm going to start with dawn.

And could you talk about the sort of broad manufacturing in fabrication picture here. Dawn Tilbury - NSF: Oh. Thanks, Rob. So I like to think about semiconductors as being the foundation for all of the industries of tomorrow from artificial intelligence.

Dawn Tilbury - NSF: Algorithms that run on semiconductors through advanced manufacturing of all processes as well as the manufacturing of semiconductors. Dawn Tilbury - NSF: Also wireless biotechnology and quantum information science. All of these industries rely on semiconductor based devices platforms and systems. Dawn Tilbury - NSF: So NSF supports basic research in future semiconductor technology from advanced materials and materials processing methods to to to new Nano manufacturing and chip manufacturing methods.

Dawn Tilbury - NSF: Designs architectures and techniques for integrating semiconductors into systems, including systems on a chip and photonic and quantum devices and system. Dawn Tilbury - NSF: And assess takes them all of science approach, including basic materials advanced and future manufacturing user needs and Dawn Tilbury - NSF: Interfaces, as well as education and workforce development, all the way from lab technicians up to PhDs. Dawn Tilbury - NSF: So some of the key aspects that NSF is investing in for semiconductors includes energy efficiency. Dawn Tilbury - NSF: Security and heterogeneous integration, including micro electromechanical system nano electromechanical systems microfluidic communications and bio sensing Dawn Tilbury - NSF: The challenges with integrating all of these things into a chip can lead to more efficiency, but there are significant research challenges to make those integrations, especially in the manufacturing processes. Dawn Tilbury - NSF: So the investments that on assesses making in semiconductors are really leading to the next generation of electronics for both computing and communications enabling us to build the industries of tomorrow. Dawn Tilbury - NSF: Thank you.

Dawn Tilbury - NSF: Margaret. Margaret Martonosi NSF: Thanks, Don, and good morning everyone. It's wonderful to be here to be able to talk about such an important topic. So I'm Martin. Martin OC, and I lead, as was said. Margaret Martonosi NSF: The Computer and Information Science and engineering or science director here at NSF Margaret Martonosi NSF: And I serve on this role, while being on loan from Princeton University where I'm a computer science faculty member Margaret Martonosi NSF: And in fact, my own research as long focused on computer system design issues, including architected computing to be more secure, reliable and power efficient.

Margaret Martonosi NSF: And so I speak for my field for all of NSF and you know from my own personal experience, talking about the way Margaret Martonosi NSF: The ability to consider long term semiconductor technology trends in such research has always been centrally important Margaret Martonosi NSF: As as the ability to test ones research ideas through access to semiconductor fabrication facilities, by which to make prototypes. Margaret Martonosi NSF: When we incised speak about the future of our field. We talked about it in terms of three organizational themes Margaret Martonosi NSF: And one of those themes is very central to what we're talking about here today, namely the seismic shift.

We're seeing as Moore's Law slow Moore's Law scaling slows and ends. Margaret Martonosi NSF: So Moore's Law was first articulated, as many of you know by Gordon Moore and Intel executive in the 1960s. Margaret Martonosi NSF: To predict that transistor counts on chip could be cost effectively doubled every 18 months or so. And what's remarkable is that from the 1960s.

Margaret Martonosi NSF: To nearly the present. The industry has sustained that exponential doubling for over 50 years. But as many of you know Moore's Law scaling now is slowing and with each year becomes harder. Margaret Martonosi NSF: More expensive and more technically challenging to achieve that kind of scaling and as that happens, it's reshaping Margaret Martonosi NSF: The layers that sit on top of that, it's reshaping all aspects of computer systems design processes. Margaret Martonosi NSF: We're also seeing related challenges around enabling researchers to gain access to cutting edge fabrication facilities for their work. Margaret Martonosi NSF: So building advanced semiconductor and micro electronic systems today is definitely a technology and a circus challenge.

But I also want to stress that beyond that. Margaret Martonosi NSF: This is really an all hands on deck moment and opportunity for computing as a role. Margaret Martonosi NSF: As the slowing of horizontal scaling challenges are computing designs.

We're seeing big opportunities big changes already and how our phones and laptops are implemented. Margaret Martonosi NSF: And this dramatically impact software layers and sit on top of it within NSF, we're happy to be funding, where we can Margaret Martonosi NSF: Cross layer research that looks at how software researchers and harder researchers can take on these problems together. Margaret Martonosi NSF: And we're also looking at large scale research in which these challenges impact AI advanced wireless and so forth. I'm really grateful to the NSF funded researchers who come here today to discuss more ideas and opportunities as we look forward on these issues. Thanks very much. Robert Margetta: Tom, can you take us through some of the activities that that we've we've looked at so far, what we're looking at in the future.

Tom Kuech, NSF: Certainly, and Tom Kuech, NSF: We'd like to Tom Kuech, NSF: highlight a few things. First of all, you know, in I'm in the advanced manufacturing area of civil men civil mechanical manufacturing innovations. Tom Kuech, NSF: Division of NSF.

Our, our cluster our area really looks at new advances, a new science that goes into processing new materials. Tom Kuech, NSF: semiconductor technology and history has been built on several things. One, it's the innovations and new materials that lead to innovations and new ways to process and manufacturing of these materials that lead to new ideas and devices that lead to Tom Kuech, NSF: Innovations and advances sensing computing biomedical applications communications.

So there is a vertical structure to this ecosystem. Tom Kuech, NSF: Where we start with materials. Learn how to manufacture them at cost and with reliability and high efficiency and then Tom Kuech, NSF: Use these materials in a wide variety of ways they're building blocks to become bricks that are used to make all sorts of new new devices and I think you'll hear about new materials and new devices today. Tom Kuech, NSF: These are area then really looks at how we take these processes and push them forward as we move into the next generation or next year off of semiconductors and electronics Tom Kuech, NSF: Manufacturing is going to be key the integration of of artificial intelligence, machine learning and other types of new processing innovations are going to be crucial and NSF is funding research in these areas along with the new materials.

Tom Kuech, NSF: One of the things that we do directly, is that we solicit input. Tom Kuech, NSF: From the stakeholders you Stakeholders are people who are researchers at at universities, small companies large companies government labs. And we do this through a process of Tom Kuech, NSF: Usually typically holding workshops and these workshops have been going on. We have just finished a series of future of semi conductor workshops that will have reports that will be available. Eventually that look at the challenges in both Tom Kuech, NSF: You know, in the entire ecosystem, starting from materials up through, how do we actually get access to for researchers to actually use these materials make devices and at the highest level, as it was just mentioned.

And so these workshops are have been going on and will be ongoing. Okay. Robert Margetta: Great. And so I'm going to turn to our guests right now and I'm going to introduce them and ask them to describe their work. Just a little bit. Robert Margetta: One thing I should note is we're putting this panel together at NSF, we asked our staff to send us some great researchers and we had Robert Margetta: Way too many for one panel so report is logging in, will also be sending you as an information sheet with with a whole whole dozens of researchers Robert Margetta: Who is working may be interested in.

But I'm going to start us off right now with April Brown from Duke University. Can you talk to us a little bit about your, your work and the the big issues you're focusing on right now. April Brown: Sure. And thank you. Good morning.

I'm happy to be here. I hope you all are doing well. April Brown: It's an exciting time for people who have been engaged in research in semiconductor materials and devices and it's great that Rob pointed out the difference April Brown: I would like to point out one thing is the need to have convergence in that research context for me or background is I've been involved April Brown: In device and materials research for over 30 years at universities government and industry. What's exciting about April Brown: Right now, what's happening right now. And my focus through NSF funding.

April Brown: Is understanding how to advanced manufacturing of new materials and devices for quantum applications that's quantum computing sensing and communications. So there's lots of opportunity. April Brown: In a build someone our expertise and it also I think engages the enthusiasm of students.

So I'm going to stop right there. I look forward to questions and April Brown: Thank you. Robert Margetta: Well before I move on, I just want to Robert Margetta: You know your work focuses on quantum information devices and, you know, Robert Margetta: Preparing for this panel, as I've talked to experts in quantum computing they've talked about this idea that if you want to build effective quantum systems and networks. Robert Margetta: You need to go beyond the computers themselves and and sort of rethink and redesign entire systems. There's the work you're doing right now I'm potentially going to plug into that. April Brown: Absolutely.

And what you're saying is true. What's important to understand is that the devices and the integration of devices in quantum systems is fundamentally different. That said, April Brown: The integration of these with the strong silicon infrastructure in the US is a key important part of this. So we have a lot to build on, but it provides April Brown: Really a framework for great fundamental research that needs to be translated April Brown: To manufacture wearable new materials and devices and that's happening. So one of the things that I've been engaged with with NSF is a workshop. I'll hold at the end of May.

April Brown: To focus on the key challenges in manufacturing these new devices, how to integrate them with silicon our strong infrastructure and build upon that. Robert Margetta: And speaking of some of those manufacturing challenges. I'm going to go now to Daniel Holcomb from me from UMass who's doing some really interesting work in manufacturing and manufacturing security Daniel over to you. Can you talk about your work a little bit Daniel Holcomb -- UMass: Sure. Thanks, Rob.

Good morning around so my work in computer engineering here at UMass focuses on design techniques for securing semiconductor hardware or physical level. Daniel Holcomb -- UMass: So this means that we're dealing with all kinds of supply channel issues, right. So things like keeping kind of reports out of systems. Daniel Holcomb -- UMass: Ensuring the chips don't leak information through side channels to the outside world, preventing untrusted foundries from being able to manipulate your parts in a targeted way and so forth. Daniel Holcomb -- UMass: All of these issues.

I think really relate to just the immense complexity of modern chips and the processes that create them. Daniel Holcomb -- UMass: So these chips, you know, have billions of transistors, they're created by large design teams split across locations and maybe across multiple companies. Daniel Holcomb -- UMass: They're using design automation tools produced by other companies and, you know, after all, the design work is done and sent to a potentially offshore foundry to be fabricated Daniel Holcomb -- UMass: So every step in this process may have a different level of trust and it's really just a huge attack surface that we have to defend. So our NSF funded research in my group looks at the science of how to make secure chips in the presence of this huge and challenging attack surface. Robert Margetta: Well, and I'd like to focus on that for just a second, because you know everyone I've spoken to preparing for this.

Everybody probably on the line has read about this issue of Robert Margetta: Domestic fabrication versus foreign fabrication and the idea that more domestic fabrication be better. Robert Margetta: Your work seems to not necessarily taking if if domestic fabrication doesn't increase your work seems to look at this idea of like the if the present situation goes on as it is, how can we make it safer. Is that a fair way to characterize it.

Daniel Holcomb -- UMass: Yeah, I think it's it's fair. We sort of play both sides of the issue. I think we certainly look at what can be done today to fix problems that exist. Daniel Holcomb -- UMass: But this you know this complicated process of creating chips you know if fabrication capacity comes on shore and become more available to us.

Daniel Holcomb -- UMass: It certainly takes out one really challenging step. And what we're trying to secure. You know, there's still work to do. We have, you know, other things that we have to defend against but it would certainly help Robert Margetta: Thank you. We're going to turn out to Melinda Jefferies L, who, you know, when we spoke you introduce yourself as being at the fringe of semiconductors and I was hoping you can talk to us about that and and what you're working on now. Malika Jeffries-El: Certainly.

Good morning. Thank you for having me here. Robert, yes, as I, as I mentioned, my work is that the friends are. We're kind of the stepchildren to the field. Malika Jeffries-El: I'm really good. Jeff result, Department of Chemistry and division of material science from Boston University and I produce these problems from a chemistry perspective in that we are making organic semiconductors.

So these are Malika Jeffries-El: Where we look at making carbon frameworks and to see if we can inquire and develop materials that will have semi conducting properties. Malika Jeffries-El: And the biggest benefit of doing it this way is the tune ability Malika Jeffries-El: Of the material when you get into something that you can manipulate the chemical structures of you can kind of do this atomic level engineering to build in the different types of properties, you may need. And so we Malika Jeffries-El: were excited to be in that area. The area is evolving.

Malika Jeffries-El: I think you know back when they got started, people thought it was just going to be an academic curiosity. But now there's becoming Malika Jeffries-El: real interest in developing technologies for commercial applications and solar energy conversion in lighting and display technologies and transistors and so Malika Jeffries-El: Yeah, we're just doing our part my NSF funded project is on developing new materials for use in our gamma organic light emitting diodes and we're happy to contribute to the science that way. Robert Margetta: So your work focuses on these carbon based materials and as I understand it, there was a there was a previous push to commercialize that and, you know, it seemed to fall off of it. So there's renewed interest. What was pushing it right now. Malika Jeffries-El: Well, I think the previous push was kind of one of these, you know, these situations when the money people get involved and you know the venture capitalists and people who are funding it for commercialization kind of set this death.

Malika Jeffries-El: You guys will get efficiency level still a con by this deadline or this is dead. It's not going to happen. And when the science didn't get there by this pre determined Malika Jeffries-El: Financial deadline, people kind of gave up on it, but the research just kept going in the labs and then it's like, oh wait a Malika Jeffries-El: Few years later, all of a sudden these great results are coming out of academic labs and academic Malika Jeffries-El: Industry partnerships are starting to push to convert the the products for it and people are realizing the way we gave up. Malika Jeffries-El: You know, I think they kind of gave up on it a little bit to the peak too late and I think people gave up on it a little bit too soon, but Malika Jeffries-El: You know, one example on the solar energy conversion side, you know when when organic solar cells got started, we were getting fractions of percentages, you know, a percent of Malika Jeffries-El: The light that was isn't it to the device was being converted and now you know double digits was the goal.

And now you know people saying couldn't get to 20% you know it was a dream wants to get to Malika Jeffries-El: 10% and so people starting to think maybe there's something to this. And one of the other panelists mentioned this earlier, we can we can also ignore the supply and demand issue. Malika Jeffries-El: As there is increasing demand for semiconductors, we have to think about our supplies in terms of materials, the ability to solution process. Malika Jeffries-El: The material is a huge boon in terms of making large areas and large amounts of displays. It may also be the trick to getting around some of these limitations with Moore's Law and also. Lastly, because Malika Jeffries-El: These films are we make organic devices, based on thin films of materials, the amount of material that you actually need to cover a lot of surface you're talking about, you know, a meter of coverage with like Malika Jeffries-El: A gram of material.

And so there's there's there's some some real benefits there. And I think people are circling back around to take a to take another look. Malika Jeffries-El: The other side, you know, look at the US picture.

It was kind of where I said people gave up. It was kind of here the other countries kept going. And so we've got some catching up catching up to do Robert Margetta: Great. So I'm going to turn to sue.

But she's a teacher who is looking at, as I understand it, sort of a systems approach to to semiconductors. Can you tell us a little bit. A little bit about that.

Subhasish Mitra: Here. Thank you Rob, and good morning everybody. It's an honor to be here before even I get into what I work on, let me think the NSF, especially the size division of NSF Subhasish Mitra: For sponsoring my research and obviously other sponsors and, you know, students and collaborators worldwide, without whom I wouldn't have been able to do this work. Subhasish Mitra: So I can afford along two directions are the forceful what I call nano systems as dr Tilbury and Dr martino's you mentioned Subhasish Mitra: That progress in hardware technologies is absolutely essential, moving forward, but at the same time. Subhasish Mitra: conventional ways of advancing hardware technology that we pursued for the past 5060 years of have begun to stall at this exact moment when applications such as AI and Subhasish Mitra: You know, communications and so on are demanding the largest improvements.

So in that domain of nano systems. My research is making progress on two essential access Subhasish Mitra: One is beyond silicon nanotechnologies. And the second is, how do we build new three dimensional LAN assistance. Subhasish Mitra: That explored this unit properties of this nanotechnologies for significant performance benefits at the scale of 1000 x. So that's like one of the areas of my research. Subhasish Mitra: And the other aspect of my research is how do you ensure robustness of future systems as Daniel mentioned, you know, when he was talking about his research.

Subhasish Mitra: That robustness is a major challenge as systems become more complex more interconnected and more pervasive. Just think about a future chip that's implanted in somebody's brain and that she is crashing. Subhasish Mitra: So hardware failures are especially a growing concern because chips are so complicated that existing ways of screening for design bugs and manufacturing defects hardly cope with existing complexity. Subhasish Mitra: And in addition, as we move to smaller feature sizes and more advanced technologies, you know, somebody mentioned heterogeneous integration. Subhasish Mitra: Several reliability failure mechanisms that will largely benign in the past are becoming visible at the system level. Subhasish Mitra: And these barriers happier, at a time when robust computing is more essential than ever think of from self driving cars all the way to cloud data centers and supercomputers.

Subhasish Mitra: So in this area. My research is creating new solutions that are already impacting actual industrial products in major ways Robert Margetta: Great. And so you actually mentioned NSF workshops you recently organized. One of them, and I know the report is is forthcoming, but can you tell us a bit you know when you when you get together with all the stakeholders and experts. What a bit of what you took from that workshop Subhasish Mitra: Thank you very much. Rob, first of all I have to say I call organize this marker of in other colleagues, I was not the only organizer.

Subhasish Mitra: And of course in us have sponsored the workshop. The workshop actually focused on three major areas. The first one wasn't foundational nano technologies. And as I said nano systems. Subhasish Mitra: The second aspect was design enablement and electronic design automation for future generations of nano systems.

Subhasish Mitra: And the third area was how do we get access to founders and hardware prototyping facilities to actually build and demonstrate this nano systems and their unique capabilities and actual hardware. Subhasish Mitra: So those are the focus areas and the key learnings where I could just fairly quickly, you know, tell you a few Subhasish Mitra: Of, you know, of course, you know, recognition of the fact that, you know, future advances in nano systems are absolutely essential for you and communications and quantum and so on. Subhasish Mitra: But here's a really important aspect, which is that the field of nano systems is very rich and many promising research ideas can advance computing significantly Subhasish Mitra: Up from devices, all the way to systems.

So this perception that some people have the because our traditional miniaturization of transistors that's slowing down. Subhasish Mitra: That does not necessarily mean that hardware technologies are slowing down. There are several other nodes, for example, three dimensional integration and others that one can play to advance hardware technology is moving forward and another really important aspect that was Subhasish Mitra: Emphasized was this notion of cross layer going from nanotechnology devices for circuits and architecture to software and applications. Subhasish Mitra: Are there are a few other points that were made that led to kind of research advances where, for example, in a sub size, given their Subhasish Mitra: Interests in software and hardware foundations can play a very important role in kind of connecting Subhasish Mitra: These towards the walls of nanotechnologies with the world of actual systems and large scale applications such as AI and others and very importantly, education and workforce. Subhasish Mitra: That's super important because we need to attract the high school and the undergraduate students critically to this field.

Subhasish Mitra: To make real progress and Lord the entry barrier and shorten their learning curves. For example, you know, to access, you know, advanced technologies, their export control concerns. How do you get access to prototyping and so on. I can get into some of those details later. Robert Margetta: Great.

So I'm going to forget to our final panelist. I'm going to remind remind reporters. Robert Margetta: If you'd like to ask a question, please just type that question out in the Q AMP a box at the bottom of your zoom screen. Also, if you just prefer to ask a question. Robert Margetta: Verbally just note that and will will unmute you to to speak. We now have Joan Red Wing and honestly I love the, the group that you're representing the two dimensional crystal consortium, which I feel like should get a prize for naming.

Can you talk to us about the work you're doing. Joan Redwing - Penn State: Sure, yeah. Thank you Rob, and good morning everybody. Joan Redwing - Penn State: So I'm delighted to be part of this panel.

And I also want to thank the reporters who are joining us today and taking your time to cover this important project. Joan Redwing - Penn State: So my name is john Red Wing. I'm a faculty member in material science and engineering at Penn State University, and as Rob mentioned I serve as director of a user facility that we have. It's called the two dimensional crystal consortium.

Joan Redwing - Penn State: So it's funded by the National Science Foundation, it's part of the NSF materials innovation platform program. Joan Redwing - Penn State: So the focus of our facility is on advancing the development of two dimensional materials and devices. Joan Redwing - Penn State: So two dimensional materials, which includes two dimensional semiconductors really represent the ultimate scaling limit in layer thickness for devices. So they're only maybe an atom thick or a few atoms thick. Joan Redwing - Penn State: And and they crystallized in sheets, and the sheets are transparent their air stable.

Joan Redwing - Penn State: They're flexible and so it's possible to design new types of devices and new device architectures. Joan Redwing - Penn State: By stacking the sheets one on top of another, or maybe integrating them with existing silicon devices and circuits or by even twisting them to get new functionalities. Joan Redwing - Penn State: So our facility is focused on advancing the synthesis and processing of these two dimensional materials. So we're focused on being able to fabricate two dimensional materials that are Joan Redwing - Penn State: That have very high crystal and perfection overlarge areas. Joan Redwing - Penn State: That would be suitable for integrating into devices and systems and these materials can be used in integrated circuits.

So Joan Redwing - Penn State: More than more type applications where you're now stacking an additional layer of devices on top of your integrated circuit also flexible electronics Joan Redwing - Penn State: Application applications in Internet of Things just connecting up you know the the devices that we use in our daily lives. And also there are applications in quantum technologies. Robert Margetta: Thank you. And I was just actually going to ask you about the we've talked about the focus on manufacturing and Robert Margetta: How important, your, your, your work actually involves a sort of like large sheet manufacturing. That's, that's one of the challenges is to figure out how to get there.

Robert Margetta: Um, you know, when you're at the research stage in this scenario semiconductors. How important is it to start integrating the idea of how can we manufacture this early into the process. Joan Redwing - Penn State: Yeah, I think with any new technology, it's, it's really important to assess manufacturer ability Joan Redwing - Penn State: Early in the development stage, even when people are doing research. If something looks particularly promising you want to start thinking about Joan Redwing - Penn State: You know the cost of manufacturing potential roadblocks that may exist that's particularly important if you want to accelerate the transition of technology from research to commercial applications.

Robert Margetta: Um, so we actually have a couple of questions in the Q AMP a Q before we get there, just one thing I want to follow up. Robert Margetta: On a point that was made earlier. So we have people on this panel representing different areas of potential development in semiconductors.

Robert Margetta: And you know, when you when you read sort of the, the press release about this. It's always sort of like this could be the next big thing. And so I'd like to put to the group, including our NSF panelists, um, you know, Robert Margetta: How did these next big things. How can they interact.

Are we saying it's an all or nothing game or, you know, are these potentially technologies that could be integrated into current silicon systems. Robert Margetta: You know, different technologies may benefit in different areas of computing or semiconductor manufacturing how need them, you know, how do you see this working in the next several decades as far as emerging technology goes and open to anyone. Subhasish Mitra: I can take it because you know we are doing it, Rob. So for example, you know, our group has been working on, you know, carbon nanotubes transistors Subhasish Mitra: And new memory technologies such as resistive RAM and their dance model, a three dimensional integration.

Subhasish Mitra: And what is really important are two aspects, first of all, that these things are manufactured in an actual silicon found race. Subhasish Mitra: Right, so that you don't have to throw away existing infrastructure to be able to do this. And the second point is Subhasish Mitra: It can integrate with silicon itself because silicon is like what I call is rice or pasta and on top of that you put the sauce. Basically, right. Subhasish Mitra: And that's what's going to happen.

Moving forward, and, you know, we have been very fortunate that, first of all, NSF, you know, funded a lot of this research when we started Subhasish Mitra: And then DARPA picked it up and this is called a 3D Associate Program. I think it's a major DARPA program like Subhasish Mitra: Over $60 million six zero and sky water technology foundry, which is us foundry, you know, in Minnesota. And that's where the carbon nanotube transistors, there is a super am and monolithic 3D Subhasish Mitra: Has been implemented already and it's working in actual silicon foundry. So your point is spot on. Robert Margetta: Anyone else I know some of our researchers sort of have notes about how their technology would be sorry.

April. April Brown: Just to add to that, because I think it's important to realize that the infrastructure we have now is something to be built upon. So if you look at my areas of focus when I'm talking about quantum systems. April Brown: Which are really quite different. There's a lot of work to build hybrid systems and to understand how the system's level.

April Brown: Approaches architectures are built together all the way back to devices and materials integration with silicon and what we have in April Brown: Everything we've advanced in that area is crucial. So this is not an abrupt shift, although it is if you look at the end goal, but there are a lot of steps to build upon what we have achieved basically Robert Margetta: Anyone else from the panel before we move on to our next question. Tom Kuech, NSF: I can just add a little bit for that rob the, you know, we, one of the things that we fund at NSF is the integration of new materials into existing platforms. And it's not just in the semiconductor area, but it's cross platforms and all sorts of Tom Kuech, NSF: Industries.

So this is a is a common theme. You don't want as I think was previously just mentioned, you don't want to, you want to build upon the huge infrastructure. Tom Kuech, NSF: That you've built up and the capitalization that's already been put into into there. Tom Kuech, NSF: I think a lot of the research.

Certainly things that john talked about could be integrated into existing platforms, building on the silicon technologies or carbon nanotubes. Tom Kuech, NSF: And so one of the things that we would like to understand more is what is the processing that's involved. What is the new science that we have to involve to make this at cost and make it efficient and to get the best utilization of all these materials in these devices and systems. Robert Margetta: Great. Well, so I'm going to turn to our first report a question.

It's from Sam Wharf my Tripoli. The question is how important our back end of line active devices to future semiconductors. Anyone want to be the first to take a crack at this Subhasish Mitra: You know, I didn't want to grab their time. But, you know, but I can go if you want Robert Margetta: Please go ahead. That Subhasish Mitra: Hi, Sam.

By the way, so you know this this 3D technology that I was talking about, you know, like using carbon nanotubes and resistive RAM. Subhasish Mitra: And monolithic three dimensional integration, which is actually working in an actual silicon foundry. That's an example of back end of line transistors Subhasish Mitra: The carbon nanotube transistors. You can be at a very low temperature and that's where you can put it on the upper layers of a stack of a silicon chip. Subhasish Mitra: You know, and that's so important because when you put memory or other computer elements of the upper layers.

You have to be able to connect with them. Subhasish Mitra: And you want to be able to go and address the memory, memory controller. So you need a fork on a chip and so on and so forth. Subhasish Mitra: So that's going to be super important and carbon nanotubes, only one example of that, you know, one could do that, the 2D materials and other stuff.

Subhasish Mitra: But moving forward. That would be extremely important. But as somebody else mentioned Subhasish Mitra: All that has to be guided by new architectures. It's not just about and I have a bunch of devices. And I just, you know, spray them on the upper layers.

It's about what new architectures. You can build and what benefits they would bring Robert Margetta: April. Robert Margetta: What do you, what would you have to add here. April Brown: Absolutely. I agree.

I mean, and I think the most important thing I heard there was, again, this April Brown: Tomorrow convergence over cooperative research where if you think about quantum you're talking about advances and new materials and devices, but that's going to be most efficient if we can do that with new architectures and what's required overall. So a great Robert Margetta: Great, well. Our next question is from Pete singer from semiconductor digest. Robert Margetta: He's saying he's hearing that design technology co optimization.

A DTC O is critical to semiconductor manufacturing success moving forward, where the design and manufacturing concerns are balance. Robert Margetta: Is there work at NSF that's looking at this and so I'll first kick that to our NSF panelists and see programmatically, if you have anything and then then open it to the group. Tom Kuech, NSF: So, Tom Kuech, NSF: Donna. I didn't know Donna was going to jump in there, but the so you know there, there's a lot of interest in this encode design. Tom Kuech, NSF: Of manufacturing process with with the product in mind. And this isn't just specific to semiconductors and a lot of this has Tom Kuech, NSF: Been brought about by the integration of new data analytics into into the system as well to optimize the manufacturing process.

Tom Kuech, NSF: Without having to do a lot, lot of the trial and error that's been or physical modeling that's been done done before. So we have programs. We have a cyber manufacturing program and we have other programs that look at up optimization in manufacturing Tom Kuech, NSF: We have, we get, but this is open to all types of technology, including semiconductor technologies. Robert Margetta: Margaret.

Margaret Martonosi NSF: Yeah, sure. I would echo what's already been said about the fact that there's different opportunities for Co design at different levels. So one level is between the device technologies and the manufacturing that that fits with them. Margaret Martonosi NSF: And then the other levels are between the device technologies and the computational architectures and software that rests on them. And so those kinds of tight inter meetings are important.

Margaret Martonosi NSF: We have a range of cross directorate efforts related to this. For example, in the area of advanced manufacturing and future manufacturing, that's a place where we've seen size an inch of working together across that kind of space. Thanks.

Robert Margetta: And they don't Dawn Tilbury - NSF: Know i think i think it's been well said NSF absolutely support work in that area. Dawn Tilbury - NSF: You know, we could do a portfolio search and see what has been done, but this is a hot area because Dawn Tilbury - NSF: As we optimize each thing it doesn't necessarily optimize the whole stack or the interior layers so there's Dawn Tilbury - NSF: huge opportunities to develop methods to optimize up and down the stack or within the manufacturing process. So I think it's a hot area. There's a lot of people working on it and we're funding out of the best proposal that we have the money to fund in those areas so Robert Margetta: To if you want more details. Please follow up with us we'll be able to get it will be sure to get you more information anyone on the panel want to comment on that before we move on to the next question.

Subhasish Mitra: Just wanted to say, you know, even the world of duty CEO is moving to what is called St SEO. Subhasish Mitra: And that's like really hard, you know, which is system technological optimization. So, you know, you take your end system and then the technology and how you can optimize and NSF is playing a big role there you know I know have multiple funding, they're getting Robert Margetta: Okay. And so at first. Any anyone else I think aim.

Robert Margetta: Okay, so we're going to move on to Mitch. Mitch Ambrose from a IP how does NSF funded research in this area compliment DARPA's electronics resurgence initiative. So I'm going to Robert Margetta: Do the same thing.

I'm going to start with Margaret on the NSF side and then see if anybody else in the panel has anything to say. Margaret Martonosi NSF: Thanks, Mitch. That's a great question. So we partner extensively with DARPA actually in a range of different ways. Margaret Martonosi NSF: So we have truly joint program. So, for example, real time machine learning as an example of a program where DARPA has funded P is NSF has funded P eyes the solicitation.

Margaret Martonosi NSF: Was a joint process and the P eyes meet together. So that's an example. We're also in ongoing discussions with DARPA and to it. Oh, and elsewhere to sort of advance on these techniques we see our sort of positions being related but distinct Margaret Martonosi NSF: empirically speaking what we often see happen is that NSF give someone their first funding in a topic area. Margaret Martonosi NSF: Often when it seems a bit out there and it's a bit visionary and then once they get a few steps further into it to to prove the viability. Margaret Martonosi NSF: will sometimes see the darker programs that stand up at that point to really push it to the next step.

And so we work with them in a range of ways and we partner with them directly when we can. Thanks. Robert Margetta: It's an interesting note about how when one agency x, it has ripples through the other research funding agencies anyone on the panel have anything to add on that.

Subhasish Mitra: And I don't want to, you know, take up all the air time but we actually do have to concrete examples. So one is, you know, again, you know, when I started my career as a faculty Subhasish Mitra: NSF grant funded me on you know this nano systems for, you know, this was in collaboration with Philip long at Stanford. Subhasish Mitra: And we started working on this carbon nanotubes and then this 3D and all that. And, you know, thanks to NSF funding, we actually demonstrated for some of the forest carbon and the computer and so on. Subhasish Mitra: But then DARPA came around 2015 2016 timeframe and they started working with us and they started this major 3D SoC program, which was about Subhasish Mitra: Not religious demonstrations, but, you know, how do you implement in an actual commercial silicon foundry. Subhasish Mitra: And you know my former student.

Make sure locker at MIT. Now, you know, he got funded by NSF now through his students even now, you know, he's also he's actually the PR, you have the 3DS or see Subhasish Mitra: Similarly, on my robust systems work. You know, the work on design bugs. It actually kind of started with an NSF exclusions in computing program.

Subhasish Mitra: And then the darker posh, you know, which is part of the 3D, which is part of PRI, you know, picked up on that. So I'm very thankful and exactly what Margaret said I would. This kind of it start, you know, with the NSF and then kind of go to a doctor that I have experienced that in a big way. Robert Margetta: Great.

So, uh, this next question. I'm going to open up to the entire panel please jump in much of the focus of today's manufacturing push is on computing, but is there enough emphasis on wide band gap semiconductors and that's coming from Sam Moore and I Tripoli. Robert Margetta: Anyone can jump in. April Brown: Well, this is April, I can say something about that. Tom mentioned, the recent workshop that I participated in future of semiconductors that NSF failed.

April Brown: Which has been incredibly interesting but a key focus was on continued work on wiping gap semiconductors. So if you look at I think across applications and into April Brown: My area of focus in this conversation. And for example, quantum devices. There is a real focus and then understanding that we have a lot of great work going on and our understanding of where by band gaps to make continued. If the answers and applications is growing.

So, yes. Robert Margetta: So for move on to the next question. I just want to ask a follow up to that we can bring some of our other panelists and Robert Margetta: You know, right now we're talking about the issues directly in front of the manufacturing capability, some of the, the research questions about what can be done right now. Robert Margetta: I'd like to know from you know from a fundamental research perspective, you know, how far is the work you're focusing on looking at, you know, are you looking at stuff that Robert Margetta: You're looking to revolutionize semiconductors right now or are we also focused on 10 2030 years down the road and I open up to any of our panelists to see how long term is you're thinking right now. Joan Redwing - Penn State: I guess I can chime in. So for for two dimensional materials, you know, these are definitely materials that could impact applications 10 to 20 years down the road.

Joan Redwing - Penn State: There's still a lot of just basic research that needs to be done in order to identify you know applications where where they're going to have an impact, but then also just to develop the materials to the point where they could be used in manufacturing Robert Margetta: No, no. Malika Jeffries-El: I'll chime in. I mean, our field is in its infancy. But if you look at the progression over over just the short time period of the last 30 years, you know, again, I mentioned, you know, Malika Jeffries-El: 30 years ago it was all a dream and you look at you know where things are improving with the development of new organic semiconductor materials and their utility and the soul application.

It's really Malika Jeffries-El: Had a nice steady uptick over the years. And you know, I think I'd love to see what happens. You know, I'm pretty young.

Malika Jeffries-El: So I plan to be around and see what happens when we extrapolate this out for the next 20 years but to also look at ways in which we can get into, into newer applications. Malika Jeffries-El: You know, listening to you. I'll talk about manufacturing. I think there's some opportunities. Malika Jeffries-El: You know where we can get involved there.

And also on the somebody asked a question about why band gaps. We can't get to very wide band gap materials that is one of the Malika Jeffries-El: Challenges with organic materials. It's very difficult to make an organic material with a very wide intrinsic band gap that stable, so we can't really do the UV as well for the Malika Jeffries-El: LEDs, but we can do the blue, which is a hot topic area as well. So I think there's, I think we're just getting started. I think the outlook is bright for the next 20 years Robert Margetta: No notes, Daniel.

Would you have any any perspective on that. Daniel Holcomb -- UMass: Yeah, so you know as the security person here. I'm probably don't look as far out as some of the other panelists. Daniel Holcomb -- UMass: You know, I think that technology has to mature, a little bit to the point where, you know, the processes are known and understood and that we can find the vulnerabilities and then look at how to fix them.

So if we look too far ahead, and we're too speculative for security. Robert Margetta: A superstitious Subhasish Mitra: Sure. You know, I can you know share a little perspective and then I'll make like a funny comment, but it's not funny. So, you know, like when you look at you know kind of time horizons. I think we are always innovating. Subhasish Mitra: So even though we target something for they're out Subhasish Mitra: There are immediate advances that also happen and in I'll just give you, you know, three examples that three times skills.

You know, I was talking about robustness. You know, when I was at Intel Subhasish Mitra: That time there was a major challenge with manufacturing the cost of, you know, testing a transistor could be bigger than the cost of manufacturing it, for example. Subhasish Mitra: And I did something they are called X compact and that was adopted within three years and now you know like it's on every electronic system.

Subhasish Mitra: If you take another extreme right which was manana systems work. It started in as I said 2006 2007 timeframe. And now after 15 years you kind of see that, you know, it's kind of getting there. Subhasish Mitra: Are like a middle ground example is my work on reliability, you know, again, kind of in a 10 year time frame.

Subhasish Mitra: You see things happening. You know, in industrial products. But here's a really interesting one.

You know, via professors right at us. I'm a professor Subhasish Mitra: So my products also include students right and our students, their timeframe, sir, like three to five years, depending on whether they're a postdoc or their PhD students Subhasish Mitra: And, you know, they keep continuing. So I think our impact of our work enough keeps going through our students and postdocs that was going to be my kind of mediacom Robert Margetta: April, I don't want to leave you up before we move on. April Brown: Yeah, three nothing much, Dad.

Okay. Robert Margetta: So next question is from Pete singer is their work underway and neuro morphic computing. And I'm going to turn to Margaret for first comment on this. Margaret Martonosi NSF: Thanks, Pete.

Yeah, that's an area of widespread interest. So we see it come up in different places across the NSF portfolio. Margaret Martonosi NSF: Within science in our core programs we have work on neuro morphic that shows up in the software, hardware foundations area architecture and the software side of it.

Margaret Martonosi NSF: And also in our foundations of emerging technology area inside the core. The other program that I want to note here. Margaret Martonosi NSF: Because it's really sort of a fascinating example of different types of partnerships is the computational research.

Margaret Martonosi NSF: Collaborative research in computational neuroscience or CRC and s program. Margaret Martonosi NSF: In which multiple directorates that NSF partner with multiple federal agencies do you just joined us this year, specifically looking at at Nora morphic in that space. And so Margaret Martonosi NSF: CRC ins started out with sort of putting computational ways of doing research into the neuroscience community. Margaret Martonosi NSF: But as it has matured as a program. What we've seen is this really bi directional Margaret Martonosi NSF: flow of ideas back and forth from neuroscience into computing and vice versa.

It's fascinating. The other thing about CRC ins, is that this is an example of an international collaboration, there's on the order of eight different countries that collaborate on CRC ins proposals. Thanks. Robert Margetta: And what else want to jump in on neuro morphic computing Dawn Tilbury - NSF: I'll just jump in quickly because one of the opportunities that we see in your market computing. Is this a brain uses much less energy for computation than Dawn Tilbury - NSF: Silicon this current devices that we have.

So we have a program in engineering called etc VA or energy efficient computing from devices to architecture and several of the Dawn Tilbury - NSF: The projects within that program are looking at neuro morphic as a way to save energy in computation. Dawn Tilbury - NSF: And of course there are many other examples across NSF, but I think it is an ideal, right, the brain is incredibly energy efficient if we could an engineer devices that were as energy efficient as the brain, perhaps using their market computing, we would save a lot of energy. Robert Margetta: And I think that's also one of the points allows for our panelists have made is is the energy consumption issue, as this is another big, big question out there in semiconductors. Robert Margetta: So I just want to remind our reporters on the, on the, on the line that we're we're coming up on just nine minutes left on the panel. Robert Margetta: We have a few questions in the queue. If you haven't gotten a chance to ask a question, please get in now because we're probably only going to have time for a couple of more Robert Margetta: Right now I'm going to turn to Mitch Ambrose, who is asking, to what extent does NSF fun joint semiconductor research projects between universities and private companies.

Dawn Tilbury - NSF: So I'll start and say yes, all the time. There are many different avenues that we fund joint research between universities and industry, we have Dawn Tilbury - NSF: A goalie program within NSF where the PCI is a university and the KPIs and industry and they submitted joint research project to do the research together and they need to have an IP agreements. Dawn Tilbury - NSF: We also have several programs that foster collaboration between University and industry Engineering Research Centers several which are focused on semiconductors are semiconductor manufacturing Dawn Tilbury - NSF: Industry University cooperative research centers that again bring together industries and government agencies and universities do pre competitive basic research in the center environment. Dawn Tilbury - NSF: And of course, we have the SP IR program they fund small companies, many of whom spin out of universities. Dawn Tilbury - NSF: To do trance cancellation. I'll research.

It looks like Margaret, just want to chime in and add some more because we absolutely fun joint research with industry and universities, all the time, especially in this space. It's important. Margaret, please. Margaret Martonosi NSF: I'm sure Don gave a great list. I'll just add to it the sizes done set of direct partnerships with different industry entities in the semiconductor space.

Margaret Martonosi NSF: So we've had several program partnerships with Intel over the years, including in this kind of area. Margaret Martonosi NSF: We also have done partnerships with SRC, the semiconductor Research Corporation Margaret Martonosi NSF: And most recently, we have a program that's a bit more multi layer, not just semiconductor focused, but across the layers, looking at sustainable computing as a partnership with VMware Margaret Martonosi NSF: So that's a real interesting example of how these semiconductor issues are, as I mentioned, creating a seismic shift that goes across layers. And so, yes. Semiconductor vendors are interested in it, but other sort of cloud based vendors are interested in partnering on this as well. Thanks.

Robert Margetta: Anyone else on the question. Subhasish Mitra: I just wanted to add a slightly different perspective. You talked about the workshop that I organized, Rob. Subhasish Mitra: And one of the recommendations for the workshop and our NSF colleagues haven't seen the final report yet. Subhasish Mitra: Is that, you know, in addition to what NSF is doing, for example, CO sponsoring research projects. The idea was that moving forward for translational kind of projects, maybe NSF could think about actually even finding Subhasish Mitra: Some industrial projects in collaboration with the clause that they would collaborate with certain academics and certain, you know, research topics.

Subhasish Mitra: And that could actually create very bi directional flow of information. And this was something that was recommended, and it's a different twist to the answers that we heard today. Robert Margetta: And next up, and this is probably going to be our last question. I'm just going to tell reporters that you know we're looking at this as sort of being an opening this is obviously a very big topic.

Robert Margetta: If you have any further questions, please contact NSF, you're welcome to contact any of our panelists and so same or saying chip design, especially using advanced processes is is very expensive. Robert Margetta: And I'm so sorry. I lost the end of that there. Let me just a second. Sorry chip design, especially using advanced processes is very expensive will making design chain more secure add expense. Robert Margetta: Is the research in this area, aimed at keeping that cost down to, and I'm just going to add my own fault to that after we addressed security.

Robert Margetta: If any of our panelists, the work you're doing could address cost of manufacturing, I'd love to hear from you. After you've addressed the security issue so dance. It's just a security focused, let's start with you. Daniel Holcomb -- UMass: Sure, you know, short answer is yes.

Daniel Holcomb -- UMass: You know, anything that we're going to do that's going to make chips more secure is going to have some cost and, you know, certainly want to keep that down as much as possible. Daniel Holcomb -- UMass: You know, we're talking about getting your chips fabricated by an entity that you don't have trust that can be really expensive. Daniel Holcomb -- UMass: You know, to make this work. You have to withhold some information from the entity that's fabricating your chips. Daniel Holcomb -- UMass: Which means either that they're fabricating generic structures that you're going to the program and generic structures are always more expensive than Daniel Holcomb -- UMass: custom built or, it means that they're making you know say half your chip and then you're finishing the upper middle layers at another foundry you trust.

Daniel Holcomb -- UMass: So, you know, we do want to keep costs down. But yes, there will be costs if we can do more of our fabrication on shore. I think those costs can be brought down further. Robert Margetta: To bash. Yeah. Subhasish Mitra: I could say a few words, you know.

So when it comes to security, for example, but in general design bugs. You know, when you talk about complexity in advanced processes. Subhasish Mitra: Or even you know site channel, you know, security issues, and so on. So, there is a major need for new verification techniques because you could leak information through IP intellectual property, for example. Subhasish Mitra: And some of our work in that space, you know, verification is a 34 year old problem, but we haven't able to make some inroads where instead of taking Subhasish Mitra: 12 person months to verify stuff. We can do it in two person days.

So, you know, new ideas are possible. And here you know I don't want to put Subhasish Mitra: Margaret of the spot you know she has done really great work. You know, as part of her recent wearing her research had on new ways of security verification, for example. So a lot of really cool stuff is happening in that space to to bring the cost of design now. Robert Margetta: Bargain everything that other Margaret Martonosi NSF: Well, well thanks for the plug see Russia with my research hat on switching back to my NSF, Pat.

Margaret Martonosi NSF: Again, it gets to this notion of a seismic shift to that, if we can think about these things earlier and not wait until we're at the fabrication point. We do have some opportunity and some leverage. So, for example, Margaret Martonosi NSF: We have a program called formal methods in the field that seeks to take advances in formal methods, including verification advances. Margaret Martonosi NSF: And see how to sort of push them out into real usages. And so some of that can be about security verification that crosses the hardware, software interface, for example, thanks. Robert Margetta: Well, and again I just know we have we have short time left but I'd like to turn that to the panel, you know, are there any cost considerations that you're looking at in your research, particularly Robert Margetta: You know I know retooling manufacturing is always expensive to start.

But, you know, other other long term cost issues that any of your research might might address Robert Margetta: No. Okay. Well, in that case, we are just about a minute out. And so I'd like to thank all of our panelists for joining us. This has been great.

Thank you for taking the time. Robert Margetta: Thank you for joining us virtually I noticed as a new world of media briefings. I'd like to actually turn this over to our NSF panelists to see if they have any, any closing remarks to our guests. Robert Margetta: Because I know these are these are people that deal with often Margaret Martonosi NSF: Hey, Dawn Tilbury - NSF: Oh, go ahead. Dawn Tilbury - NSF: Excellent echo the thank you for joining us.

I think this has been a great panel and hopefully it's just the start of the conversation to help us convey Dawn Tilbury - NSF: The exciting results that are coming out of NSF funded research in collaboration with industry and collaboration with other government agencies for advancing the frontiers in this very important area of semiconductors. Robert Margetta: Great. Well, the market. Margaret Martonosi NSF: Just a big plus one, thanks to the panelists. Margaret Martonosi NSF: And and to the many folks that aren't on this panel who could have also spoken we we'd be happy to have the journalists follow up for more names as well.

Thank you. Robert Margetta: And reporters on the line will be sending you more names. Robert Margetta: This meeting has been recorded.

This will be available to you, please let us know if you have any more questions about semiconductor voice devices semiconductor materials and we will close out there. Thank you all for your time. Subhasish Mitra: Thank you for the opportunity. Thank you.

2021-04-01 03:33

Show Video

Other news