產業_SEMICON2026矽光子論壇_LamResearch逐字稿_20260831

場次：2026 SEMICON Taiwan 矽光子國際論壇（2026-08-31） 講者：Dr. David Haynes（Managing Director, Specialty Technologies, Lam Research） 題目：The Path from Electrons to Photons — Enabling Silicon Photonics with Advanced Fabrication Solutions

[00:00:00.00] David Haynes, and there's a domain for our research, and I look after specialty technology. I will explain what that means a little bit more in a couple of slides, but it very much focuses on our electronics and photonics today. [00:00:15.92] I'm here to kind of, I guess, represent the tool vendor's perspective a little bit on how we fabricate these jets that are incredibly enabling, but also a little bit crazy in terms of some of the designs. [00:00:28.76] And why do I say that? Well, you know, if you take a silicon photonics chip today, it has [00:00:35.76] critical-electric processes that are as challenging as some of the processes at the very leading [00:00:40.76] end of technology. At the same time, with the back end of the line, we have processes that [00:00:44.76] look a lot more like MEMS. Deep TSVs, big cavities, crystallographic wet edges, all sorts of things. [00:00:49.76] So it's really the convergence of a lot of technologies. That's why I call that a little bit crazy. [00:00:55.76] But at the end of the day, you know, these are very, very important in terms of the future of AI and semi-seamfter systems. [00:01:03.24] So I'll give you what I promise will just be a one-slide introduction for Lam Research, [00:01:08.10] and I'll explain what I mean by specialty technologies and where we fit within the Lam Research business. [00:01:14.62] And then I'm going to focus on why is so much of photonics important. [00:01:17.74] And so many industry experts today have told you why it's so important, [00:01:22.94] but I think I can manage that section quite quickly. [00:01:25.76] But I'm then going to move on and talk about something that hasn't been talked about too much today, which is really the fabrication challenges associated with silicon photonics and how they're being addressed by companies like LAM. [00:01:37.56] At the very end, if we've got some time, I'm also going to talk a little bit about packaging and CPO, because there is a lot about the co-design and engineering of silicon photonics and packaging that's very important. [00:01:52.94] and LAM is working very, very hard in both fields, [00:01:56.14] so I think it will help bring the discussion together. [00:02:00.50] So for those of you who may not be familiar with LAM, [00:02:03.10] we're a $23 billion company. [00:02:05.60] We employ about 22,000 people worldwide. [00:02:08.44] This is a little bit, a few months out of date, [00:02:10.14] that could be 23,000 by now. [00:02:12.44] The rate of the industry is growing, and we're now new people. [00:02:15.64] But I'll actually raise your attention to the kind of middle number. [00:02:18.48] We're a very innovative company. [00:02:19.74] We invested $2.4 billion in R&D last year. [00:02:23.74] And a non-trivial part of that actually goes on non-leading-edge technologies that support areas like silicon photonics. [00:02:31.74] We are better known for definition, etch, and script and clean technologies, and all of these are absolutely critical to silicon photonics. [00:02:40.74] And of course we need to support these applications as they emerge, and we have an incredible team here in Taiwan [00:02:46.74] with great depth of experience in silicon photonics from a unit process and integration perspective. [00:02:55.14] So when we spoke to speak about specialty technologies for many people in [00:02:59.06] Taiwan would not be an unfamiliar term but you know lab today is probably best known for [00:03:06.10] leading end logic and memory applications and about packaging applications and of course we [00:03:11.62] and we are a technology leader in those spaces. [00:03:14.70] But we're also very, very influential in this space [00:03:17.54] of silicon photonics, and that's all part [00:03:21.36] of our specialty technologies. [00:03:23.00] And specialty technologies more broadly [00:03:24.50] include CMOS image sensing, MEM sensing, [00:03:27.80] power applications, new materials like wide bandgap [00:03:30.62] materials, ferroelectric materials, [00:03:32.42] which we'll talk about later, [00:03:33.62] but we'll go back to silicon for time. [00:03:35.94] RF and other applications. [00:03:37.62] But by far the fastest growing part of this space today [00:03:41.62] is photonics and photonics, [00:03:43.62] both for AR applications in AI classes, [00:03:47.22] but more importantly, photonics and photonics [00:03:52.02] for photonic IC applications to support data centers. [00:03:57.62] And if you looked at a data center rack, [00:03:59.82] everybody's mind immediately goes to the advanced logic, [00:04:02.72] the GPUs, the high bandwidth memory, [00:04:05.02] and of course, LAM again is working in supporting [00:04:07.58] and enabling every generation of those new technologies. [00:04:10.58] But if you looked at all the orange words on this slide, [00:04:13.86] and there's a lot of them, I'm not gonna go through them, [00:04:16.06] but sensing, power management, MCUs, [00:04:20.36] and of course, silicon photonics are all really key parts [00:04:24.12] of what go into a data center today. [00:04:26.38] So specialty technologies plays a really important role [00:04:30.50] in enabling data center architectures. [00:04:33.66] And it's growing rapidly. [00:04:35.14] And the one thing that I will be absolutely right about today [00:04:39.20] is any forecast I show you on silicon photonics will be wrong. [00:04:42.70] And it will almost certainly be very conservative, [00:04:46.08] and it will be higher again [00:04:48.24] when we look back in a year's time. [00:04:50.28] But actually from this, you can see [00:04:51.96] that one of the most important parts [00:04:54.20] of the specialty technology business in data centers, [00:04:57.64] along with sensors, along with power and analog SEs, [00:05:01.00] is silicon photonics, and it's the fastest growing part of it, [00:05:03.88] which is why it's in critical of the world. [00:05:07.04] Why is it so important? [00:05:08.20] Well, AI is of course AI hardware, [00:05:11.24] and enabling AI hardware is of course very difficult. [00:05:14.14] We've seen this exponentially increase in compute power. [00:05:17.12] It's changed all of our lives. [00:05:19.04] But we still have bandwidth limitations [00:05:21.82] in terms of memory bandwidth [00:05:23.44] and in terms of internet bandwidth. [00:05:25.04] And again, many industry experts have spoken to that today, [00:05:27.84] so I'm not going to go into the detail of it. [00:05:29.94] Similarly, they've spoken about the power consumption issues. [00:05:32.50] If we don't fix this, [00:05:33.90] it's going to become self-limited technology. [00:05:36.18] So, bandwidth and power need to be dealt with. [00:05:39.14] And there are other challenges such as [00:05:40.94] thermal management and yield control. [00:05:43.08] This is becoming a high volume business. [00:05:45.56] And yield is a big thing. [00:05:46.66] And I'll touch on one of my slides later, [00:05:49.50] one of the very enabling things that LAM is doing [00:05:51.62] to focus on improving yield [00:05:53.50] in silicon's atomic manufacturing. [00:05:57.82] It's already well established. [00:05:59.22] And again, we've seen similar, [00:06:02.36] similar information in prior presentations today. [00:06:05.42] but what are also already a key part [00:06:07.56] of what being the center infrastructure is. [00:06:09.80] We've seen the emergence of MPO. [00:06:12.12] I think the jury's somewhat out on how much of that [00:06:14.82] will be displaced by the next generation [00:06:16.82] of CPO technologies. [00:06:18.62] And there's been a lot of discussion today. [00:06:20.16] I think there was some great commentary [00:06:21.86] in the imec presentation about optical interconnect [00:06:25.54] technologies as well. [00:06:26.58] So when you bring all these things together, [00:06:28.52] this is a very exciting and innovative space. [00:06:32.36] I'll go back to my comment [00:06:33.52] about every market forecast being wrong. [00:06:35.40] I submitted this presentation to the deadline to Semi a couple of weeks ago, and this was accurate at that point in time. [00:06:43.28] Since then, there's another one's gone up by Semi-X, and I think somebody commented on this morning about the latest Yole. [00:06:49.36] So either we need to make the presentation submission a bit later, or we need to control the market forecast. [00:06:56.70] But I'm going to leave the forecast to the analysts for the time being, and I'm going to go back to the technology, which is really what we're here to talk about. [00:07:05.40] I'm going to focus today on what's in the kind of purple what, the silicon photonics engine. [00:07:10.98] But at the end, we'll talk, just in a few slides, about advanced packaging. [00:07:15.16] We have another presentation from one of my colleagues, who's been reading, [00:07:19.66] on Friday, I believe, from the Heterogenes Integration Bookshop, [00:07:22.18] that will talk more about the packaging part of it, and particularly hybrid bonding applications. [00:07:29.62] There's a lot of words on this slide, but the one thing I'll say is silicon photonics [00:07:33.74] is very technically challenging but I would sum up those challenges in kind of three or four areas [00:07:40.94] and I think they've all been addressed to some extent by the people today. Number one, you've [00:07:46.06] got to make the waveguides better. You've got to get the losses done, particularly to enable [00:07:50.38] backside waveguides, long-range waveguides. Improving waveguide loss is really really important. [00:07:56.94] That also goes for performance whether it's silicon, waveguides, silicon, nitride waveguides, [00:08:01.82] and what kind of architecture is, [00:08:03.36] we also have a name for the same category. [00:08:05.94] We're making a way to expect it. [00:08:07.18] Number two, what do we do for modulators beyond 200g? [00:08:11.00] Again, it's been a huge topic of discussion today, [00:08:13.10] so you can see how important it is [00:08:15.10] to the silicon photonics community. [00:08:17.30] So how do we enable modulation beyond 200g [00:08:21.06] in new materials? [00:08:22.30] That's number two. [00:08:23.58] Number three, and a lot of people think [00:08:25.16] it's really easy, is fiber coupling. [00:08:27.18] How do we get the fibers interconnected [00:08:29.18] with silicon photonics? [00:08:30.28] It's really not easy. [00:08:31.76] It's quite challenging, in fact, [00:08:33.12] that it's where some of these more men's-like processes come into play. [00:08:37.70] So we're going to talk about that a bit. [00:08:39.04] And related to that, but we'll call it number four, [00:08:42.00] is really the co-design and co-engineering of the silicon photonics in the package. [00:08:46.68] Having the most discrete units doesn't work. [00:08:48.80] You really need to think about how the whole system and the packaging comes together [00:08:52.92] as you develop the technology. [00:08:57.42] So let's start with how do you make weight goes better? [00:09:00.00] Well, the first thing you do is how do you make high-quality silicon nitride? [00:09:03.86] Well, if you're making silicon nitride wave-dy-generated in the process [00:09:07.02] and you have a lot of thermal headroom, [00:09:09.22] you can use LPCBD at a high temperature and the other. [00:09:12.26] That's great. [00:09:13.42] But if you can't, you have to go to other processes, [00:09:16.86] such as plasma-based processes and PECVD. [00:09:19.80] And we know that hydrogen in PECVD-positive silicon nitride can cause issues. [00:09:25.02] Certainly cause issues in C-band, but it can even cause losses in O-band. [00:09:28.74] it's quite challenging to get that hydrogen out of a conventional PCVD deposited silicon [00:09:34.56] nitrate film. People who have tried to use DVD or sputtering do the same thing, but the [00:09:39.30] quality of the films tends to be very good, even though you can get zero hydrogen films [00:09:46.06] effectively. One of the things we have at LAM is a technology called Pulse Laser Deposition, [00:09:50.60] which we've been developing over the last few years. It's a type, really, of physical [00:09:54.46] vapor deposition I'll explain it a little bit later and it has the potential to [00:09:59.38] produce hydrogen free films with really really low losses it's a very interesting [00:10:04.84] technology but today it's still somewhat in CNN so today we're focused on new [00:10:10.24] BKM's low temperature processes low by vision actually we've been delivering [00:10:14.32] these PCV capabilities to our customers now for a few months multiple customers [00:10:20.44] in the leading place. [00:10:23.84] We hope that PLD is currently in concept of feasibility. [00:10:27.18] We're looking at that as well as some other approaches. [00:10:29.82] And we believe those will be ready for dead-run [00:10:32.30] probably early next year. [00:10:34.28] But again, they give the potential [00:10:36.40] to give very high quality costs. [00:10:38.20] And here's some data on the PECVD films, [00:10:41.38] just really showing how we can reduce [00:10:43.26] the hydrogen concentration [00:10:45.58] and improve our radiation process temperature. [00:10:47.80] So again, it's promising for low temperature [00:10:50.48] silicon nitride films, these are good. [00:10:52.30] Work is ongoing to look at new chemistries, [00:10:54.74] new process capabilities, et cetera. [00:10:56.94] But in the future, we think we can also use PLD. [00:11:01.58] And PLD is interesting for many photonics applications, [00:11:04.50] and the way it works is you use a nanosecond [00:11:07.88] of pulsed simulasia to basically remove the material [00:11:11.80] from a small target. [00:11:12.90] It's basically a 100 millimeter target, [00:11:14.66] so it's small, it's small, [00:11:15.96] which is usually a very big target. [00:11:18.92] You basically, whatever's in the target [00:11:20.74] ends up on your wafer, [00:11:22.18] and you can rotate and move your wafer around [00:11:24.52] to control the stoichiometry [00:11:26.38] and thickness of the films you're depositing. [00:11:28.88] The advantage is, if you think about a waveguide layer, [00:11:31.92] number one priority is your thickness uniformity [00:11:35.04] in terms of your RI uniformity has to be excellent. [00:11:37.94] And this gives you the ability to control very carefully [00:11:40.84] both the RI uniformity and the thickness uniformity. [00:11:43.36] So number one, you can deposit film. [00:11:45.24] but a very highly uniform film. [00:11:48.36] But if you have a hydrogen in your pocket, [00:11:50.30] you won't get any hydrogen on your wafer. [00:11:52.42] So therefore, if you have a computer target [00:11:54.74] and we take the word of producing those, [00:11:58.16] then in theory, we can produce zero hydrogen films [00:12:02.42] with exceptional uniformity. [00:12:04.38] Now, again, this is still in contact with feasibility today. [00:12:06.90] There are still challenges around it. [00:12:08.78] We're working on it. [00:12:09.62] We hope to be able to demonstrate it, [00:12:11.68] or run demos for customers later. [00:12:15.24] I will say, while that's on the screen, that's actually a time-lapse video of the real plasma and what happens. [00:12:22.58] It's taken through the side corner of the chamber so you can see what the plasma actually looks like. [00:12:27.32] It's about the size of an apple and you move it around the wafer. [00:12:30.98] But the interesting thing is the target doesn't have to be silicon nitride. [00:12:34.54] It could be BTO. [00:12:36.50] So if you want to do BTO (barium titanate), next you need some very clever steps to be able to deposit BTO (barium titanate) on a silicon wafer. [00:12:45.24] with the right crystallography, [00:12:47.58] but that's something we're also working on. [00:12:49.82] So this could be a very enabling technique for clients. [00:12:55.88] Once you've deposited your waveguide material, [00:12:58.28] you need to etch it. [00:12:59.40] And this is where these critical etch steps [00:13:02.30] really come into play. [00:13:03.38] It's really, really hard to etch these waveguides. [00:13:05.66] I mean, state of the art today, [00:13:07.52] maybe sidewall roughness of two nanometers or one nanometers, [00:13:10.54] there's definitely a goal to get to the long one nanometer [00:13:13.56] in terms of sidewall roughness. [00:13:15.16] With silicon you can process the films, maybe at high temperature, to improve the sidewall roughness. [00:13:21.16] With nitride waveguards built later on you don't have that thermal budget. [00:13:25.16] So you really need to optimize the edge process. [00:13:28.16] It's challenging because you need no sidewall roughness, but you also have a high degree of iso-dense loading. [00:13:34.16] I go back to my comment about some silicon photonics designs look a bit crazy. [00:13:39.16] So a semiconductor engineer, these iso-dense loading loads on your hands, which can never be something you would do. [00:13:44.16] You have these, I mean the picture of the micro [00:13:46.16] that resonated there, you have these huge open areas [00:13:48.74] with very small gaps, and the gaps, you know, [00:13:52.30] inform the performance of the robust. [00:13:54.64] So these things are very, very challenging [00:13:56.46] in terms of aspect ratio. [00:13:57.54] Now you may be stopping on oxide, [00:13:59.46] so depth is no problem, but CB can prove up, [00:14:03.20] are a problem, even if you're stopping on oxide. [00:14:06.10] If you're actually a grating, [00:14:07.30] for the grating coupler, you have a blind edge, [00:14:09.80] you don't even have a stop layer, [00:14:11.18] and then the aspect ratio dependent [00:14:12.68] for different features sizes and integrating new and new are the challenges. [00:14:16.68] So this is where technologies like high voltage bias pulsing and Lam Research's advanced mixed motor pulsing technologies [00:14:22.68] and these are available on our Lam Research water bank edge tools, they were developed for leading edge applications [00:14:27.68] for enabling silicon photonics. Without these type of capabilities you wouldn't be able to address the roadmap [00:14:33.68] of silicon photonics moving forward. It's very advanced capabilities. [00:14:37.68] People think it's 45 nanometer technology, it's 65 nanometer technology. [00:14:42.68] CD node does not make any difference. [00:14:45.68] The edge challenges are shown to be in some of these processes are as challenging as anything that we've seen on any application. [00:14:53.68] Next thing is modulated materials. [00:14:56.68] So modulated materials, and these have all been mentioned today, beyond silicon, [00:15:01.68] I think on lithium niobate, BTO, [00:15:03.68] electropyl polymers, 2D materials, [00:15:05.68] such as that. [00:15:06.68] The plan is working on all of these, [00:15:08.68] to some extent, to support our customers. [00:15:13.68] I've mentioned that PLD could be used to deposit BTO, [00:15:17.68] and there are some tricks needed, [00:15:19.68] and template materials to be able to do that in silicon. [00:15:22.68] But you also need to be able to etch it. [00:15:24.68] If you have a material like BTO, lithium niobate, [00:15:26.68] you can actually develop a batch process [00:15:28.68] that's quite well understood [00:15:29.68] that it's been around a long time. [00:15:32.14] That's fine. [00:15:32.98] But if you go to BTO, how do you etch it? [00:15:35.28] There's no real conventional etch process for BTO. [00:15:39.72] There's no easy way of cleaning the etch-byte products [00:15:41.88] off the sidewalk after you etched it. [00:15:44.32] So you go to a purely physical etch process using IDE [00:15:47.14] and we have a proven process already [00:15:49.42] that's being used to make BTO-based switches, [00:15:52.84] optical switches, and it's well established, works well, [00:15:56.96] and it's been used in a large, fragile way. [00:16:00.72] So we have a solution to meet these material requirements [00:16:04.88] of the new materials coming through in the next few years. [00:16:08.24] Fibre coupling, again, I mentioned, [00:16:10.04] it's already a significant challenge. [00:16:12.94] We've talked, I think other speakers have talked about [00:16:16.00] edge coupling versus grating coupling. [00:16:18.52] The grating edge on these grating couplers [00:16:20.80] is one of the hardest things you do [00:16:22.32] in silicon photonics. [00:16:23.64] But there's also a lot of other features, [00:16:27.36] more men's light features, big features, [00:16:29.44] cavity etches, deep-range etches, [00:16:32.44] TSVs, even metal lenses to replace conventional lenses [00:16:36.74] for multi-wave-like fiber coupling. [00:16:39.56] Well, in this case, we use our Syndion tools. [00:16:42.58] This is a deep-requivite edge tool. [00:16:44.72] It uses LAM's rapid voltage, or rapid process, [00:16:49.66] which allows us to etch at high edge rate, [00:16:51.80] but still get very smooth surfaces. [00:16:53.56] Again, in this case, it's why it's a market-leading tool in TSV, but for optical applications, [00:16:58.44] it's equally important. [00:16:59.44] We need smooth sidewalls, we need high X-rate patrupa on deep cavities, for example. [00:17:05.52] And similarly, you may want to do a fiber isolation for an edge cuppler, where you want [00:17:11.56] to do a deep cavity edge. [00:17:13.56] Here you can see you can use an isotropic edge process to edge silica under a nitrate [00:17:19.56] wave of nitrate. [00:17:20.56] The good thing is it's very flexible. All these things can be done in the same time. [00:17:24.56] So the same process capabilities can enable isotropic etch, highly isotropic deep silicon etch, [00:17:30.56] shadow etches, even to some extent dielectric etches. [00:17:35.56] I mentioned the yield earlier, and this is another big problem, where you etch all these really critical etches [00:17:41.56] and then back in the line of your process you do some really deep etches, maybe through 20 microns of oxide, [00:17:47.56] maybe through deep silica. [00:17:50.02] And you can cause a lot of damage to the bevel [00:17:52.24] of your wastebreder, [00:17:52.94] and these final processes. [00:17:54.82] So, Lam Research's Coronus tool, [00:17:56.32] which is basically developed for barrel engineering, [00:17:58.98] can make a huge difference to yield. [00:18:00.94] Because now you can use bevel deposition [00:18:03.10] and barrel-ex processes to either reconstitute [00:18:05.94] or protect the bevel. [00:18:07.32] So when you get this damage in these processes [00:18:09.90] and the far back end of the silica process, [00:18:13.10] you can use Coronus to basically eliminate them [00:18:15.90] from the yield issue. [00:18:16.84] and that makes a big difference. [00:18:18.16] These have been quite challenging problems [00:18:20.62] in silicon photonics. [00:18:22.52] So we're well positioned from a silicon photonics perspective [00:18:26.96] for a wide range of applications, [00:18:28.58] it's not a product presentation as such, [00:18:30.46] I'm not going to get into the different products, [00:18:32.52] but we can meet the needs in the silicon photonics space. [00:18:35.66] But that's only part of the challenge, [00:18:38.32] so what about packaging? [00:18:39.80] As I said earlier, we need to co-engineer, [00:18:42.54] co-design and co-engineer the silicon photonics [00:18:46.02] and the packaging. [00:18:46.86] So these two things need to come together. [00:18:48.98] So Lam Research also a leader in advanced packaging capabilities. [00:18:52.12] When we think about CPO, [00:18:53.98] we think about two packaging routes. [00:18:56.60] The first is high density fanout. [00:18:58.76] The second is 3D stacking using high-density. [00:19:02.90] And Lam Research working both of these areas. [00:19:05.06] I don't have time to go into a huge amount of detail today. [00:19:08.32] I just wanted to give you a taste. [00:19:10.40] We have some other presentations. [00:19:12.68] So let's start with fanout, racial level package. [00:19:15.72] We have our wet products that are most used in this space. [00:19:20.74] So we have both our clean tools [00:19:22.74] and also our Sabre electric and deposition tools, [00:19:25.96] our Sabre 3D tool in particular. [00:19:28.02] So if you think about that, [00:19:29.10] some of the challenges of high density packaging, [00:19:32.94] there's a wide distribution of patterns [00:19:35.22] that you need to be able to fill using these tools. [00:19:37.94] There's fine line RVLs, five by five, [00:19:40.24] Michaels for example, [00:19:41.24] right up to mega color structures [00:19:42.82] that are 150 by 150 c. [00:19:45.40] These are huge features that you need to build copper. [00:19:50.34] The turbo cell technology, [00:19:52.06] grain engineering capabilities we have [00:19:54.48] and optimized chemistry can address all of these. [00:19:57.86] When we think about hybrid bonding for 3D stacking, [00:20:01.70] it really brings into play pretty much all of our equipment [00:20:04.94] across our wet water portfolio, [00:20:06.74] our deposition portfolio and our X portfolio. [00:20:10.26] Just one example of where that's really important [00:20:14.10] is to start thinking about interface fire. [00:20:17.30] So, it might be some of the carbon, [00:20:19.00] it was mentioned on the other information, [00:20:21.54] but it's become a standard material to be, [00:20:24.08] you know, even the bonding material, [00:20:25.62] produces very high strength, very good performance. [00:20:28.68] It's got, it's a great copper diffusion barrier, [00:20:31.82] has very specificity, but also has superior bonding strength. [00:20:35.66] Similarly, copper grain engineering [00:20:38.26] to produce small grains makes a big difference. [00:20:41.60] So, the combination of ABC [00:20:43.66] and fine-brain copper can really improve [00:20:46.36] the bottom quality of the water. [00:20:48.52] So that's addressing the packaging side of activities. [00:20:52.50] Last slide, to conclude, [00:20:55.64] I think Lamb is very, very focused on this area. [00:20:57.98] Soil-comitronics is a really big area of interest for us. [00:21:01.28] We are doing a lot of development around this space. [00:21:04.30] New materials integration, improving processes [00:21:06.80] of developing technologies for fiber integration. [00:21:10.36] And from a packaging perspective, [00:21:12.74] Again, it's compacted optics enabling [00:21:15.86] around packaging solutions to bring all that together. [00:21:19.14] But this is an ecosystem. [00:21:20.40] It's gonna take a village to enable silicon photonics. [00:21:24.12] I'm kind of very pleased that the consortiums [00:21:26.56] that have been set up for me to address this. [00:21:28.42] I think it's really bringing people together. [00:21:30.40] It's a huge group of experts today. [00:21:32.36] It's really good. [00:21:33.62] So ecosystem collaboration is really important to us. [00:21:36.36] We can't do this unless we collaborate with our customers, [00:21:39.56] particularly in photonics. [00:21:40.90] You have to measure stuff that is really quite difficult to measure when that gets routines [00:21:46.26] that our customers have. [00:21:47.62] So partnerships are really important. [00:21:49.72] So if we can help you, please reach out to myself. [00:21:53.14] And either myself or somebody from my team will definitely meet at the end of the day [00:21:58.14] because it's about working to generate the community to make this the technology flexion [00:22:02.58] that it's going to be. [00:22:03.58] So thank you very much. [00:22:04.58] Thank you. [00:22:05.58] Thank you. [00:22:06.58] Thank you. [00:22:07.58] Thank you. [00:22:08.58] Thank you, David. [00:22:09.58] Thank you David. We have time for a couple of questions. [00:22:19.58] I'm really excited about the equipment vendors embracing photonics now. I spent myself a couple of years in a large company. [00:22:28.58] I wanted to ask about the silicon nitride temperatures maybe a little bit. [00:22:35.58] What, for your laser assistant, what kind of temperatures are you getting there? [00:22:42.30] Well, certainly less than 200 degrees, but in theory we can deposit silicon nitrate at room temperature. [00:22:49.24] I mean, in fact, because PLD has no self-heating on the way for, like, you get it sputtering. [00:22:55.36] You can't really do room temperature definition using PLD. [00:22:58.72] Whether that's truly necessary or not, I don't know. [00:23:01.72] But certainly below 200 degrees centigrade is what the target is. [00:23:05.58] And I would add to the comment you made about equipment [00:23:08.86] vendors becoming involved in this. [00:23:10.34] I think somebody said it earlier. [00:23:12.14] You need some very advanced equipment [00:23:14.10] to really meet these roadmap challenges. [00:23:16.10] Because honestly, the capabilities and requirements [00:23:20.46] are becoming incredibly strange in terms of performance. [00:23:24.10] Absolutely. [00:23:25.22] Have you, so for all these light trends, [00:23:27.94] have you looked at the losses? [00:23:29.94] Have you characterized them? [00:23:31.74] So the early days that we have [00:23:35.06] is very very compelling. It's very low-cost. But we haven't yet got data of customers' test [00:23:43.94] structures that's just basically doing public transmission loss data. But it looks very good. [00:23:48.42] It's the best data we've got from any form of micro-deposition. So it's compelling but it's early days yet. [00:23:55.06] I think all my questions answered. Any other questions? [00:23:59.70] If not, we're on time for our breaks, so please be back at 325.