SemiWiki.com - Podcast EP300: Next Generation Metalization Innovations with Lam’s Kaihan Ashtiani

Episode Date: July 30, 2025

Dan is joined by Kaihan Ashtiani, Corporate Vice President and General Manager of atomic layer deposition and chemical vapor deposition metals in Lam’s Deposition Business Unit. Kaihan has more than 30 years of experience in technical and management roles, working on a variety of semiconductor tools and processes. Dan explores… Read More

Hello, my name is Daniel Neni, founder of Semiwiki, the Open Forum for Semiconductor professionals. Welcome to the Semiconductor Insiders podcast series. My guest today is Kehan Ashtiani. He is corporate vice president and general manager of atomic layer deposition and chemical vapor deposition metals in Lambs deposition business unit. Kayon has more than 30 years of experience in technical and management roles working on a variety of semiconductor tools and processes. Welcome to the podcast, Kayon. Hi, happy to be here. So first, to get started, let us know how you got interested in the semiconductor industry. You know, how did you make your first venture into semiconductors?

Yeah, I started my graduate school at University of Wisconsin, Madison at the time, and at a National Science Foundation had these excellent centers around the country that each would do a certain area of research, and they would get funding from NSF, and the industrial partners would match those funds. And as it turned out, University of Wisconsin-Madison had the center called plasma-aided manufacturing, and this was, as I said, it was NSF-funded center. that their area of research was to apply plasmas to applications in semiconductor devices. And I started my grad school at Wisconsin and became part of the center, and my PhD thesis was centered around doing research of plasmas used in manufacturing of semiconductor devices. So that's how it started. and I'm graduating from UW, I started working in industry, and I've been here for, as you said, 27 years. Great. So what challenges do semiconductor manufacturers face in the metallization phase of fabrication?

You know, do NAND, DRM, and Logic each have unique challenges? Yeah, each device has their own challenges in terms of metalization and how metals get integrated in these devices. But in general, the idea of metalization and semiconductor devices provide the interconnect where we connect millions and billions of transistors on a device together such that they can communicate and the signals traveling through these interconnects and essentially extract information from the devices. So the role of metals, there's memory or logic, is essentially in making that interconnect such that we can communicate with the transistors, you know, in the order of billions, on these devices. So the challenge obviously, when it comes to metalization, you know, as the

term is familiar with everybody metals, it conducts electricity. The challenge comes how to conduct electricity in the most efficient manner without much resistivity or generation of heat and essentially generating, moving the signals along the interconnect at the lowest resistance possible. That's the general challenge for all devices, be it memory, nan, DRAM, or in logic and foundry for metalization. And that challenge continues because as we put more and more devices in semiconductor chips, that means the interconnect becomes more and more intricate and smaller in dimensions such that we can connect all these devices together. So in that dimensionality, decrease in dimensionality, provides challenges for metals to conduct

electricity without generating too much heat or creating issues for the signal transfer. So that's really the metalization challenge is about this dimensional scaling that happens in semiconductor devices and how we cope with that such that we continue scaling and put more and more capability in semiconductor chips. Right. So Mali has emerged as the metal to replace tungsten in certain applications. What benefits could this bring? And why is a transition from tungsten needed now? Melitinum in general is a metal that provides better conductance or lower resistivity for

electricity to travel to it. That's a very general statement about molybdenum versus tungsten. That comes from two factors, what we call bulk resistivity of the material. This is something that's measured at large dimensions when you essentially measure resistivity of the metal. and you compare, let's say, tungsten and malady, malibdenum has lower resistivity, therefore it can conduct electricity better. So that's a bulk resistivity concept. As I said, dimensionally, these devices are getting smaller and smaller. So comes into play this concept of what we call thin film resistivity,

because as these bulk metals become thinner and thinner, such that we can scale them to these devices that are extremely small, There's another factor that comes into play that resistivity of the metal compared to its bulk resistivity increases as it becomes thinner and thinner. And that's mainly because the scattering of charge carriers, in this case, electrons in the metal. It's similar to thinking of a highway where you're, you know, a lot of cars are through traffic driving through it. But if you make that highway narrower and narrower, then the, you know, the flow of traffic is important. competed because you don't have as much room, I guess, to have enough cars going through. If you think about electrons as those cars in a highway, that the metals provide that highway,

as it becomes smaller and smaller resistance of that metal goes up, the electrons can't travel as fast. So that becomes an issue. Because of those two issues, molybdenum becomes very attractive compared to tungsten. In applications where dimensions are small enough that, you know, these dimensionality plays a role, making the thin film resistivity going high. So that's why molybidum comes into play, replacing tungsten in certain applications, obviously not in all applications, because dimensionality of all features are not the same

in a semiconductor chip. Great. Can you talk about some of the innovation required to make it viable for high-volume manufacturing? I mean, why was this difficult? As dimensions in semiconductor devices become smaller and smaller, the traditional methods of deposition of metal changes. It's, if I look 20 years ago, physical vapor deposition was a very common technique to deposit metals in semiconductor chips.

And then chemical vapor deposition came along as the dimension started shrinking, that gives the ability to do better filling of features with metals inside these semiconductor chips. If you think about features as these trenches or holes to be filled, as they get smaller, things become more difficult. In the past 15 to 20 years, the industry has moved towards a method of deposition, we call atomic layer deposition. And as the name suggests, we deposit these metals, one atomic layer at a time essentially in these lines or contact holes to ensure that these lines and contacts are filled with metal.

and no areas are left of void, because obviously voids don't conduct electricity. So atomic layer deposition has become a general technique, and with it, if you think about the term, atomic layer deposition of metals, one atomic layer at a time, you can automatically think about the challenges that brings along. One is in terms of efficiency of this process,

how good it is to fill these features, and second, as soon as we say, Atomic layer at a time, you start thinking, well, this process is going to be probably slower than other methods to reach a certain thickness of metal that you're depositing. So those two challenges are something that we always address than we do atomic layer deposition. I have to mention that Lam pioneered the atomic layer deposition of metals in the industry in high volume manufacturing 20 years ago. So we started at tungsten, ALD, about 20 years ago. And that evolved into about 10 years ago into what we call LFW tungsten. And then past that point, just about five or six years ago,

we started doing AOD atomic layer deposition on Mavi. And with that knowledge, there are certain techniques that comes into design of the tool that deposits the film, the architecture of the tool from the hardware perspective. And then there are challenges that come in terms of processing of the of the metal to deposit the films on the on the substrates both of those challenges are things that we've worked on for the past five to six years to bring a lD molly into hpm manufacturing okay lamb's altus halo tool is central to enabling the molly based metalization at scale what makes this tool unique you know how does it support the demands of advanced semiconductor manufacturing

The Altus Halo, similar to many of our deposition tools, is what we call an architecture that's called QSM, clot station module. That means within each of these modules or chambers, we have four stations that individually can process waferes. The ability to have four waivers processing at the same time, and if you think about what I mentioned about anatomically deposition being in general, process that gives us the ability to actually accelerate the process having four stations within one chamber. So that inherent architecture design, QSM, quad station module, is something that Nivellis came up with more than 25 years ago. Novellis and Lam merged about 10 or 11 years ago, and Lamb is essentially chamber.

So that architecture. So that architecture enables us to do a fast deposition of ALD for Mali. But there are specific technologies that we've developed for this tool that are specific to molyptenum or Mali. One is that all these deposition techniques or whatever we're depositing in metals via tungsten or Mali, the source of the metal itself comes from a precursor, the chemical precursor. In the case of Mali, the precursor, the chemical precursor comes in the in a solid form.

And a lot of innovation was required for us to enable solid precursors in a HVM environment, high volume manufacturing environment, such that we can run these processes day and day out and deliver reliably the amount of precursor needed to the chambers such that we can deposit the film. So that was a big innovation in terms of bringing a solid precursor at the volumes that are necessary,

HVM with militant. The second part that comes is that that solid precursor needs to become a change, transforming, to become a gas such that we can deliver that. So there is a delivery technique where we have to heat the solid to sublimate it, essentially provide the gaseous form in the chamber, which requires that delivery to be sophisticated enough to again reliably deliver the precursor to the chamber. And then finally, as part of the hardware innovation, comes the ability to deposit the film via AOD

and does require techniques of AOD in terms of the hardware itself, such that we can repeatedly wafer to wafer, chamber to chamber, deliver this material and deposit the films. So those are from a hardware perspective. A lot of innovation has gone into it. Partially those innovations came from our knowledge of many years of doing ALD with tungsten.

So, you know, those techniques were implemented for Mali as well. And then some new innovations because the material is different. The third challenge of this material is that the precursor itself is a very corrosive chemistry. And if enough precautions are not taken, that that becomes an issue in terms of contamination for depositing your pure molly film to get the lowest resistivity possible without damaging the rest of the semiconductor chip. And that's a third part of the innovations

that has gone into developing Altus Halen. And then finally, when it comes to the process itself, the steps that we take during the ALD process, there is a lot of innovation involved there. And that to some degree is specific, whether we do memory, DRAM, NAN, or logic, Each area requires innovation in its own to ensure that not only the film is deposited properly, but also is integrated with the rest of the processes in semiconductor chip manufacturing without issues.

So all of those over many years that we've been working on Mali and we pioneered Mali and the industry has gone into design of Altus Halo creating a differentiated product for MAP. No, interesting. So final question, Kahan, is how How is customer adoption shaping up? Will the transition to Mali be limited to 3D NAND or will it touch all the device types you work with over time? All device types will be impacted by adoption of Mali.

From a timing perspective, looking at these various applications in NAND, DRAM, and Logic, it appears the first applications in NAND are starting earlier than others. we are at a point that we've been evaluating our Mali process and our Altos Halo tool with our customers. It's at a point that the NAN manufacturers are in various stages of beyond CNF, what we call concept and feasibility. Beyond that stage, they're either in pilot qualification or in ramp for high volume manufacturing or in high volume manufacturing. So NAND has become the first adopter of Mali, and that's sort of understandable in terms of requirements of the device for Nant. Next is logic in certain areas of logic, what we call middle of the line applications in logic, where there is this transition from Tungsten to Mali that's happening. And again, that's because of the requirements of logic to look for the lowest specificity possible.

Just as a figure of merit to, you know, put audience understand what the impact of Mali is, if I compare a system of tungsten that deposit tungsten in a certain device or a certain contact and compared with Mali, the resistance of that contact, you're able to reduce that by about 50% or higher compared to tungsten. That's very, very significant in terms of device performance. When we think of metalization, you're always thinking you can speed. And speed translates how fast we can run these devices and, you know, be it artificial intelligence applications, be it high speed I.O., be it cloud, even in areas of streaming, all of these

devices with semiconductor devices, we're always looking for faster and faster in performance. And therefore, a contact that lowers the resistance by 50% is a significant achievement to improve the speed of the device. So it appears that logic is following very closely in terms of adoption of Mali into the devices and then finally the DRAM will follow later on as the DRAM transitions from what we call the 6F squared device to 4F squared device. That refers to how the integration of the device is done and the size of the chip in DRAM. As that transition, transition happens from six F squared to four F squared we expect Mali to be adopted in DRM as well. So as you can see all devices memory and logic are impacted or will be impacted by Mali and the

benefits that Mali brings in terms of lowering the resistance. Great conversation. Thank you Kian. Hopefully we can have you back again on a different topic. Well thank you for having me. I appreciate participation looking forward to it. our podcast. Thank you all for listening and have a great day.