Optica Online Industry Meeting: Heterogeneous Integration on Silicon Photonics

Full Transcript

https://www.youtube.com/watch?v=rzgQMuFsltU

[00:00] Some very important in-person summits are being prepared right now.
[00:05] Take advantage of early bird registration rates, which are even lower if you're an Optica corporate member.
[00:13] Here's an overview of what's to come.
[00:16] Optica will bring together the brightest and best in Los Angeles.
[00:20] Corner New York, Brussels, Glasgow, Sunnyale, Paris, and Malaga.
[00:38] [music]
[00:38] [music]
[00:44] Got my shades in the lab at a quart 10.
[00:46] White coke cuff like a greaser again.
[00:50] Spinning that prism like a hub cap crown.
[00:53] Turning [music and singing] every photon in a fuel line.
[00:59] It's phonics [music]
[01:02] It's phonics baby.
[01:05] 2026 baby.
[01:05] 2026 riding a light wave doing new tricks.
[01:11] riding a light wave doing new tricks.
[01:11] From the fiber in the ground to the chip in my hand.
[01:15] We make that sunshine jump.
[01:19] Yeah.
[01:19] Photonics baby 26.
[01:44] Tiny little wave got on a thumbnail chip.
[01:48] chip.
[01:50] makes your wise look slow and steady.
[01:54] makes your wise look slow and steady.
[01:55] Lasers in the toolbox, markers on the floor, drawing up a future through a tiny door.
[01:57] Lasers in the toolbox, markers on the floor, drawing up a future through a tiny door.
[01:59] floor, drawing up a future through a tiny door.
[02:05] It's for there for a ton of shakes.
[02:11] Riding in a lightweight doing new tricks.
[02:16] From the fiber [music] in the ground to the chip in my hand.
[02:20] We make that sunshine jump on command. [music] Yeah.
[02:24] Photonics baby.
[02:33] >>> [music]
[02:44] From a single red dot on a dusted [music] bench to a whole wide [singing] world in a data trench.
[02:50] Flip that switch, see the spectrum spin [music] like a jukebox jumping when the coins rolling.
[02:59] Oh [screaming] yeah, baby. [music]
[03:11] It all will happen here at the ex campus's castle of Sanjen Ole in Paris.
[03:18] See you there.
[03:24] We're crossing boundaries layer by layer.
[03:31] What stood apart now works as one.
[03:33] Built with precision.
[03:35] Did we false quantum dark place right on time?
[03:49] We may obey force five quantum dot light placed right on time.
[04:09] Heterogeneous integration.
[04:13] integration not theory
[04:15] not theory not abstract.
[04:19] not abstract.
[04:19] This is how photonics moves
[04:21] This is how photonics moves forward.
[04:25] forward.
[04:25] Find out Tuesday, January,
[04:27] January, 27.
[04:31] 27.
[04:31] Time to look ahead to the next Optica
[04:33] online industry meeting.
[04:35] On Tuesday, January 27th, we bring together
[04:38] industry leaders to discuss challenges
[04:41] ahead for heterogeneous integration.
[04:44] This is no longer an academic
[04:45] discussion.
[04:45] It's about what can actually
[04:48] be built to a deadline, tested and
[04:50] manufactured at scale.
[04:56] We'll include perspectives from Lummech,
[04:59] Photon Bridge, Open Light, Fine, Cintel,
[05:03] Phetonics, Accelerint, Ligantech,
[05:06] Polariton, and Flex Compute.
[05:16] placed right on time.
[05:21] We look at some of the real bottlenecks
[05:23] We look at some of the real bottlenecks like alignment, thermal management,
[05:25] like alignment, thermal management, process integration, yield, and reliability.
[05:29] reliability. You may have seen some of the speakers on stage at Pixummit USA [music] on stage at Pixummit USA [music] on January 19th.
[05:36] January 19th. Now that deep dive conversation continues online for 90 minutes, and it matters more than ever.
[05:42] minutes, and it matters more than ever. Global tariffs and trade uncertainty are pushing companies to rethink supply chains from just in time to just in case.
[05:45] Global tariffs and trade uncertainty are pushing companies to rethink supply chains from just in time to just in case.
[05:47] chains from just in time to just in time to just in case.
[05:50] case. And thanks to the support from our sponsors, your attendance is free, but you do need to register in advance to actively participate in the discussion.
[05:52] And thanks to the support from our sponsors, your attendance is free, but you do need to register in advance to actively participate in the discussion.
[05:55] sponsors, your attendance is free, but you do need to register in advance to actively participate in the discussion.
[05:58] you do need to register in advance to actively participate in the discussion.
[06:02] actively participate in the discussion. Now you know who's here. Let's start.
[06:10] Hello everyone and welcome welcome to the online industry meeting on
[06:13] Hello everyone and welcome welcome to the online industry meeting on
[06:15] the online industry meeting on heterogeneous integration.
[06:17] I see many heterogeneous integration.
[06:20] I see many people who were with me last week in San Francisco.
[06:23] Safe travels everyone.
[06:26] I hope you all made it home safe and sound.
[06:28] Today is a very important day.
[06:30] Today we are talking about perhaps the big the most important challenge for us to match the demands of AIdriven companies.
[06:37] We are talking about bringing the best of all the worlds in photonix.
[06:42] We are talking about combining different materials.
[06:46] This meeting is free of charge for all of you because of the great support of our sponsors today.
[06:50] Lentech, open light, polar, vanguard automation, micron, photonbridge, iMac, fix and oe waves.
[06:55] Thank you very much for a great support as corporate member and today for sponsoring this event and facilitating this very important discussion for many of you.
[07:05] You know how it works.
[07:07] You know how this meeting runs is two groups of companies.
[07:12] There's a group of companies who are here manufacturing heterogeneously integrated
[07:17] manufacturing heterogeneously integrated photonics and there are companies here.
[07:19] photonics and there are companies here who have part of the solution and can actually make the work of the integrators even better than already is.
[07:26] integrators even better than already is.
[07:27] That's why what we're going to do in the next 90 minutes is to match the demands of the integrators with the technology of the enablers and all the people with me in the room.
[07:36] The integrators today are the siliconator platform of Lent Tech.
[07:39] The silicon photon is a silicon platform of IMC.
[07:41] The devices the photonic devices of luminum with the Indian for signing manufacturing open light offering the entry point of many foundaries including of course Tower gas and the Indian phosphon silicon platform unique silicon platform from Sintil photonics and the enablers are acceler and polaron.
[08:03] and polaron.
[08:07] This is a meeting that is made to give value to our corporate members.
[08:11] If you are not yet a corporate member of Optica, join corporate optica.org/join corporate.
[08:15] Go there right now, fill the
[08:17] corporate. Go there right now, fill the form and join us because we need you.
[08:20] We need you to continue doing your job.
[08:21] Also very important, this meeting is part of a long series of meetings, online industry meeting.
[08:25] You can find all of them in YouTube.
[08:27] The next meeting is laser beam shaping on the 10th of February.
[08:32] If you're making optics, micro optics and you can actually be in shape anything stay there on the 10th of February.
[08:38] But today, today is all about heterogeneous integration and I would like to remind everyone this meeting is live streaming YouTube.
[08:46] So let me say hello to YouTubers of the world.
[08:47] Hello YouTubers, if you have any question write them in the chat and I will read them in the room and this of course so valid for the people here with me in the zoom room.
[08:57] This meeting is as good as you want it to be.
[08:59] You are here to connect with each other.
[09:01] There's 120 people with you in the Zoom room.
[09:03] Use the chat. Write in the chat what you do and what you can do for others and connect on the next one hour and a half.
[09:11] Connect with your particular partners, suppliers, and customers.
[09:13] They are all already here waiting to do business with all of you.
[09:17] waiting to do business with all of you.
[09:19] And now I would like to give the floor to the person who made this fantastic agenda.
[09:25] He's director of optics, photonics, integrated circuits, and quantum technologies.
[09:30] Coming all the way from Bristol, UK, Dr. John Pew.
[09:33] The floor and the attention of the heterogeneous integrated photonic community is yours.
[09:39] Thank you so much, Jose, and hello everyone.
[09:43] Um, I'm of course delighted to be co-odderating today's very exciting Optica online industry meeting on heterogeneous integration in silicon phatonics.
[09:52] And heterogeneous integration is no longer a buzzword, but something people are actually trying to manufacture, qualify, and scale.
[09:58] So we've structured this session to start with a real integration partnership.
[10:02] Then we widen out to materials, devices, tooling, applications, and finally a deliberately different approach.
[10:10] The aim is simple.
[10:13] What works?
[10:13] What doesn't?
[10:13] And what's still missing?
[10:15] What partnerships can be formed to overcome these challenges?
[10:16] I'd like to thank in advance
[10:19] challenges? I'd like to thank in advance all our speakers for joining this meeting today and being prepared to answer all your questions.
[10:26] So please keep them coming. Uh if you type them in the chat as Jose said we'll welcome you to unmute and join the live conversation.
[10:33] So let's jump in with a concrete example. Your own guy and from Lijent and Alex Chakawi from uh Excelrint are working together on 35 in integration using microtransfer printing.
[10:47] Under the collaboration, Accelerint brings its microtransfer printing MTP expertise, a scalable way to place high performance active components onto diverse substrates, while Lentech brings its lowloss silicon nitride platform.
[11:03] Alex, your own. Please share your slides. The floor and the attention of everyone is yours. Thank you.
[11:15] Hi John. Hi everybody. Thank you for inviting us for this fantastic webinar
[11:20] inviting us for this fantastic webinar and I'm really uh looking forward for the discussion from now on.
[11:24] So my name is Alexandria Kawi.
[11:26] I'm business development director at Excel print.
[11:28] I'm going to update on MTP side and it will be followed by Jeron at Legend Tech who will update on uh the progress uh on their side as well.
[11:39] So MTP Excel print I mean I only have two minutes I could talk an hour on that.
[11:44] So um I decided to be short.
[11:47] So Excel print is the pioneer of microtransfer printing technology.
[11:50] [snorts] What is MTP?
[11:53] Again I'll try to do it in 2 minutes but I thought uh it would be uh easier for me to just focus on the benefits of the technology for this particular industry of silicon photonics.
[12:03] So the first thing that MTP brings on the table is that it enables the high volume dataccom market.
[12:09] When we're looking at the forcecoming forecast and number of wafers that we're going to need to deliver in the few coming years, what I can tell you for
[12:20] coming years, what I can tell you for sure is that at least flip chip technology is never going to make it.
[12:25] So that's one aspect.
[12:27] Um the other thing that MTP brings on the table is that it enables above 400G per lane transceivers.
[12:34] This is exactly I mean if you want to reach high bandwidths what you need is small thin components small pads few parasitics uh really optimized interconnects.
[12:46] that's what you need and this is also what MTP can bring on the table here and it's also the best material savings technology as we know it's all about cost at the end and MTP is saving and is 80% better than wafer to wafer bonding.
[13:03] If you think about the savings for 35, it's 60% better than diet to wafer.
[13:09] And there's another thing which many few people talk about, but where our industry partners think it's a very important parameter.
[13:17] You can reuse the 35 substrates.
[13:20] And as you know, there is a big shortage of 35 substrates.
[13:22] a big shortage of 35 substrates.
[13:22] With MTP, you can reuse and reuse and reuse MTP, you can reuse and reuse and reuse it.
[13:27] So you're not impacted by that.
[13:27] So this is a very good um and important parameter.
[13:29] And of course it enables low power systems uh because the placement accuracy today of 300 nanometer enables uh light low light coupling losses of much below.5 dB and we have a road map down to 100 nanometer.
[13:44] So um so in recent years right we have a huge amount of development and investment made by our RTO partners like Tindle uh institute where Excel print is actually located the headquarters are within the Tindle Institute but also uh other uh RTO's like IMC and uh millions and millions of dollars have been spent to validate MTP technology in the applications.
[14:15] because our partners in RTO's they have application experts and thank you so much for bringing that because without
[14:23] much for bringing that because without this our technology wouldn't be where it is today.
[14:29] Um at Excel print we made the conscious decision to focus on bringing our industry partner to production because our people are experts in micr fabrication.
[14:39] And so when you uh hear about our announcements, it's mainly focused on production on uh everything which has to do with supply chain.
[14:51] And so we have announced back in uh September 2024 the world first integrated value chain for HI integration in photonix and a partnership with XFAB and with Legend.
[15:03] And I'm not going to talk more about that and Geron will explain what's the status on that and the progress we have made in the integration of components.
[15:10] We have also just before um ECO uh last year announced that Intel has signed a letter of intent and together with Accelerint we're going to see how we can leverage uh MTP within Intel silicon.
[15:26] leverage uh MTP within Intel silicon photonics manufacturing site.
[15:29] Again this photonics manufacturing site.
[15:32] Again this is about production and this is really our focus.
[15:35] Uh you will not be successful I think in this market if you don't have some footprint in Asia and therefore we are also uh announced that we are joining force with Enostar uh into the silicon photonix integration.
[15:49] They have high capacity.
[15:51] They already have MTP printers.
[15:54] And so again the only discussion we have with them is to reach production.
[15:57] And uh last um Christmas we have announced uh that Excel print is acquiring the tool capabilities back from Xis company.
[16:10] So uh actually it's a knowhow and capabilities which is back to excel print and this is always with one vision and uh and target is to start production as soon as possible and uh today I'm uh really happy to talk about a specific um use case
[16:29] about a specific um use case uh which is uh the activity we have together with Seagate technologies.
[16:35] together with Seagate technologies.
[16:35] Uh so Seagate is an excellent customer.
[16:39] Uh so Seagate is an excellent customer.
[16:39] They are uh making and producing hard disk drive using hammer technology heat assisted magnetic recording.
[16:48] assisted magnetic recording and uh this technology is using a laser and uh I'm really happy to say that uh this is using an MTP laser.
[16:58] You can see here uh on uh the little cartoon uh description that you would need basically a laser coupled to a wave guide and this is you know how the system looks like and the print accuracy and the print requirement needed to to build that system are very equivalent to what is needed for any PC integration that will go next in transceivers.
[17:21] that will go next in transceivers.
[17:21] So and this high yield and high volume production is achieved using an excellent industrial printer.
[17:29] excellent industrial printer.
[17:29] So uh so we are really happy uh to
[17:31] So uh so we are really happy uh to really say that uh uh lasers um are now in mass production with MTP and a lot of people were saying that it's still years away but I would say it was a year be before that everything was prepared and it's ready today.
[17:52] So uh and um Seagate won the SPI photonix prism award uh just last week uh on the optical and material components uh with this technology.
[18:00] So with with all the learnings that we have made uh with uh this release to production experience that we have gained we want to apply that to any other type of material sets photo diodes microoptics any other components like synfim lithium nobate and this is what we're doing and uh I would leave the floor now to Jerome who is going to explain you what is being done and what's the progress of the 35 integration onto the silicon nitrite platform.
[18:31] Uh hello everyone uh I'm Yun Ko I'm a
[18:34] Uh hello everyone uh I'm Yun Ko I'm a technical lead for 35 integration at technical lead for 35 integration at Lentech.
[18:38] Lentech um you can press a few more times.
[18:40] um you can press a few more times Alexander.
[18:41] Alexander so legend is known for the silicon.
[18:43] so legend is known for the silicon nitrite wavekai platform u but then the topic of today is about uh specifically.
[18:46] topic of today is about uh specifically our uh photo detector integration where we use a micro transfer print technology.
[18:52] we use a micro transfer print technology offered by uh or developed by Excel print.
[18:57] offered by uh or developed by Excel print. Uh here you can see just an example of a demo circuit where we integrate a few of these photo detectors.
[18:59] example of a demo circuit where we integrate a few of these photo detectors and uh the snapshot is made right after printing.
[19:02] and uh the snapshot is made right after printing. Uh the next slide please.
[19:04] Uh the next slide please. Um here so you can see on the left hand side uh a source substrate.
[19:09] here so you can see on the left hand side uh a source substrate. So it's indium phosphide substrate full of photo detectors.
[19:11] So it's indium phosphide substrate full of photo detectors. Um so as what Alexander was hinting towards is that we can have a very high density integration on the source substrate.
[19:13] Um so as what Alexander was hinting towards is that we can have a very high density integration on the source substrate uh making for an efficient material usage and um specifically what's relevant for photo detectors is that uh for specific applications you can have uh a few photo.
[19:15] uh making for an efficient material usage and um specifically what's relevant for photo detectors is that uh for specific applications you can have uh a few photo.
[19:35] applications you can have uh a few photo detectors needed or you can have a lot.
[19:37] detectors needed or you can have a lot of photo detectors needed on your uh target pick application and that puts uh different constraints on the overall yield.
[19:47] So uh what we discussed together with the uh component provider is that we want to have this uh on PD testing capability so we can uh uh similar to flip chip assess known good type principle for the photo detectors and uh so the component provider u supplied this at and provided these measurements and uh they achieved above 99.5% yield and then we can kind of uh select the bad performingvices devices and filter them out.
[20:17] So on the right hand side you can just see a few of the statistics on uh uh junction capacitors and uh dark current or leakage current from the devices and uh these are very high performing photo detectors and performing up to spec as what the component provider uh stated for the next slide please.
[20:37] next slide please. So then after um you know receiving the
[20:40] So then after um you know receiving the devices there's some development effort
[20:42] devices there's some development effort needed in order to make the the devices
[20:44] needed in order to make the the devices essentially uh pick and print ready. Uh
[20:47] essentially uh pick and print ready. Uh so this is what we've been working on in
[20:49] so this is what we've been working on in the last year. Here you can see in the
[20:51] the last year. Here you can see in the bottom image an example of such a
[20:53] bottom image an example of such a component picked and print on top of a
[20:56] component picked and print on top of a wavegu uh a silicon nitrite a800
[20:58] wavegu uh a silicon nitrite a800 waveguides and we print the photo
[21:01] waveguides and we print the photo detectors locally in a cavity and then
[21:03] detectors locally in a cavity and then the light is coupled eancently uh into
[21:06] the light is coupled eancently uh into the absorption region. So on the top uh
[21:08] the absorption region. So on the top uh left hand graph you can see that the uh
[21:11] left hand graph you can see that the uh leakage current performs similar as
[21:13] leakage current performs similar as before. There's some slight variation
[21:15] before. There's some slight variation that we identified and we're working on
[21:17] that we identified and we're working on on improving that. Um and then on the
[21:19] on improving that. Um and then on the right hand side you can see one of the
[21:21] right hand side you can see one of the benefits of having these eancently
[21:22] benefits of having these eancently coupled detectors is that we can have a
[21:25] coupled detectors is that we can have a broadband response of the detectors uh
[21:27] broadband response of the detectors uh covering a large wavelength range with
[21:29] covering a large wavelength range with uh with good responsivity values. Um
[21:32] uh with good responsivity values. Um what is not shown here in the bottom
[21:34] what is not shown here in the bottom image is essentially the final
[21:35] image is essentially the final interconnect stage where we uh connect
[21:38] interconnect stage where we uh connect the printed photo detector to the
[21:40] the printed photo detector to the backend stack of the big platform. Um
[21:43] backend stack of the big platform. Um but that's just a very brief summary of
[21:45] but that's just a very brief summary of what we've been doing essentially the
[21:46] what we've been doing essentially the last year year and a half. Uh maybe if
[21:49] last year year and a half. Uh maybe if you can go to the next slide we can just
[21:51] you can go to the next slide we can just share a little bit on the road map. So
[21:53] share a little bit on the road map. So for the the detectors aside as mentioned
[21:55] for the the detectors aside as mentioned before we started a little while ago um
[21:59] before we started a little while ago um scoping the project and negotiating with
[22:01] scoping the project and negotiating with the suppliers in order to get to the
[22:03] the suppliers in order to get to the towards these transfer print ready
[22:05] towards these transfer print ready components. And then over the last year
[22:07] components. And then over the last year we essentially received the detectors,
[22:09] we essentially received the detectors, made them print ready and then trial out
[22:11] made them print ready and then trial out did some development and had some early
[22:13] did some development and had some early access runs. And then for this year uh
[22:16] access runs. And then for this year uh what's essentially on the road map is uh
[22:19] what's essentially on the road map is uh to transfer this from the prototyping
[22:21] to transfer this from the prototyping stage towards the volume commercial
[22:23] stage towards the volume commercial offering. uh actually uh within a few
[22:26] offering. uh actually uh within a few weeks uh the first lot will start and we
[22:29] weeks uh the first lot will start and we expect to to finish the end of the first
[22:32] expect to to finish the end of the first quarter where we combine all the
[22:34] quarter where we combine all the different separate modules that we've
[22:36] different separate modules that we've been developing on the 200 mm offering
[22:39] been developing on the 200 mm offering over the last 6 months and then put them
[22:41] over the last 6 months and then put them all together. Uh and then for the
[22:43] all together. Uh and then for the detectors essentially moving forward uh
[22:46] detectors essentially moving forward uh uh the tail end of 2026 into 2027 we'll
[22:49] uh the tail end of 2026 into 2027 we'll expand our offering with additional
[22:51] expand our offering with additional suppliers and also expand the product
[22:53] suppliers and also expand the product offering in terms of higher speed for
[22:55] offering in terms of higher speed for the detectors
[22:56] the detectors >> and then uh uh at later stage also
[22:59] >> and then uh uh at later stage also expand expand towards different uh
[23:01] expand expand towards different uh cross-sections to offer uh you know for
[23:04] cross-sections to offer uh you know for different uh application cases on the uh
[23:08] different uh application cases on the uh S SOAS and laser sides given the time I
[23:11] S SOAS and laser sides given the time I did not really go into that but
[23:13] did not really go into that but essentially we started the similar
[23:15] essentially we started the similar activity last year so here we're still
[23:17] activity last year so here we're still in the prototyping phase at the moment
[23:19] in the prototyping phase at the moment uh and here we're targeting the the the
[23:22] uh and here we're targeting the the the scale up of that activity uh beginning
[23:25] scale up of that activity uh beginning of next year and then uh towards the
[23:28] of next year and then uh towards the next slide this is it from my part back
[23:30] next slide this is it from my part back to Alexander
[23:32] to Alexander >> yes
[23:34] >> yes I mean this is uh the uh famous what can
[23:38] I mean this is uh the uh famous what can you do for us what can you do for you
[23:39] you do for us what can you do for you right So uh so what you can do for us
[23:42] right So uh so what you can do for us really is for component makers lasers,
[23:45] really is for component makers lasers, PDS, SOAs, modulators, microoptics,
[23:49] PDS, SOAs, modulators, microoptics, please I mean if you embrace our
[23:50] please I mean if you embrace our technology that will be great. Help us
[23:52] technology that will be great. Help us grow the supply chain trust in MTP in
[23:55] grow the supply chain trust in MTP in developing MTP compatible components. P
[23:58] developing MTP compatible components. P foundar is help us democratize and
[24:01] foundar is help us democratize and integrate MTP into your PDKs and ADKs.
[24:03] integrate MTP into your PDKs and ADKs. Right. Right now we have a lot of
[24:04] Right. Right now we have a lot of vertically integrated customers and it's
[24:06] vertically integrated customers and it's not really necessary for the success and
[24:09] not really necessary for the success and ramp in volume but to address the broad
[24:11] ramp in volume but to address the broad market we need this uh and please
[24:14] market we need this uh and please promote our technology there uh into
[24:16] promote our technology there uh into your PDK and ADK and in terms of what we
[24:20] your PDK and ADK and in terms of what we can do for you that's what we call the
[24:22] can do for you that's what we call the MTP alpha platform technology uh those
[24:25] MTP alpha platform technology uh those are the services we provide we can bring
[24:27] are the services we provide we can bring your proof of concept to life we can
[24:29] your proof of concept to life we can provide custom training on MTP if you
[24:31] provide custom training on MTP if you need MTP training you can contact
[24:34] need MTP training you can contact We support your manufacturing partners
[24:36] We support your manufacturing partners and your factory to reach production
[24:38] and your factory to reach production ready stage like we did with Seagate. We
[24:41] ready stage like we did with Seagate. We provide print services. We provide now
[24:45] provide print services. We provide now since Christmas uh last Christmas uh the
[24:47] since Christmas uh last Christmas uh the best-in-class production grade MTP
[24:50] best-in-class production grade MTP printers and we you have access
[24:52] printers and we you have access basically to the largest IP portfolio
[24:54] basically to the largest IP portfolio worldwide in MTP.
[24:57] worldwide in MTP. And I think that finishes the the talk
[24:59] And I think that finishes the the talk and looking forward for the questions.
[25:02] and looking forward for the questions. Alex Ziron, thank you so much and thank
[25:04] Alex Ziron, thank you so much and thank you for preempting the optica question
[25:06] you for preempting the optica question of what can you do for others and what
[25:08] of what can you do for others and what can others do for you. Uh for those
[25:10] can others do for you. Uh for those watching, if you'd like to answer any of
[25:11] watching, if you'd like to answer any of these calls, please get in touch
[25:12] these calls, please get in touch directly. Uh or I'll be happy to provide
[25:15] directly. Uh or I'll be happy to provide an introduction. So we have a couple of
[25:17] an introduction. So we have a couple of questions in the the chat. Um Punam,
[25:21] questions in the the chat. Um Punam, would you like to unmute and ask you a
[25:23] would you like to unmute and ask you a question, please?
[25:24] question, please? >> Yeah, sure. So I have a question like uh
[25:26] >> Yeah, sure. So I have a question like uh you fabric like uh first fabricated
[25:29] you fabric like uh first fabricated device on like India's uh on Indian
[25:32] device on like India's uh on Indian first and then transfer it on silicon
[25:34] first and then transfer it on silicon nide. So I have a question uh does your
[25:38] nide. So I have a question uh does your devices are passivated
[25:40] devices are passivated uh uh and how did you pass it? Did you
[25:43] uh uh and how did you pass it? Did you pass it before transferring or after
[25:45] pass it before transferring or after transferring on silicon nitrate?
[25:48] transferring on silicon nitrate? >> Um thank you for your question. Uh so to
[25:51] >> Um thank you for your question. Uh so to answer we yeah we use a custom um
[25:55] answer we yeah we use a custom um tethering mechanism as um as part of the
[25:58] tethering mechanism as um as part of the accelerant technology portfolio in order
[26:00] accelerant technology portfolio in order to encapsulate the photo detector and
[26:02] to encapsulate the photo detector and and make it pick and print ready. Uh and
[26:05] and make it pick and print ready. Uh and then on the uh silicon nitrate pig site
[26:09] then on the uh silicon nitrate pig site we essentially have a standard
[26:10] we essentially have a standard cross-section but uh we use what we call
[26:13] cross-section but uh we use what we call a loca module which is a local cladding
[26:15] a loca module which is a local cladding aperture opening uh which essentially
[26:18] aperture opening uh which essentially makes a cavity right above the main
[26:20] makes a cavity right above the main wavegu. So that opens up this main
[26:23] wavegu. So that opens up this main waveguide for eancent coupling. Um and
[26:26] waveguide for eancent coupling. Um and then we use a thin adhesion layer in
[26:28] then we use a thin adhesion layer in between as is typical for this
[26:30] between as is typical for this technology. And then afterwards indeed
[26:32] technology. And then afterwards indeed we pivate with um yeah within a final
[26:35] we pivate with um yeah within a final layer to make sure that the photo
[26:37] layer to make sure that the photo detectors are sealed uh and do not move
[26:41] detectors are sealed uh and do not move in any circumstance afterwards.
[26:43] in any circumstance afterwards. >> Okay. Yeah. Thank you so much.
[26:46] >> Okay. Yeah. Thank you so much. >> Right. And Stefan Heinman from FEX all
[26:48] >> Right. And Stefan Heinman from FEX all the way from Bosezeman, Montana.
[26:51] the way from Bosezeman, Montana. um what's on your mind?
[26:53] um what's on your mind? >> Yeah, thank you for the great
[26:54] >> Yeah, thank you for the great presentation. I would be interested to
[26:56] presentation. I would be interested to just get a rough understanding on what
[26:58] just get a rough understanding on what this trend what the bring up of such a
[27:00] this trend what the bring up of such a transfer process, you know, entails in
[27:03] transfer process, you know, entails in terms of timeline and partner
[27:04] terms of timeline and partner commitments.
[27:09] >> Well, I mean that's a complicated,
[27:12] >> Well, I mean that's a complicated, >> right?
[27:16] >> Yeah. I mean,
[27:22] And then I
[27:24] And then I know why.
[27:30] >> So um I guess I I would assume that you
[27:34] >> So um I guess I I would assume that you need special as as legend just
[27:38] need special as as legend just explained you need special interface
[27:40] explained you need special interface layers and then a special run on on the
[27:43] layers and then a special run on on the p side.
[27:44] p side. >> How about the active component? Uh how
[27:47] >> How about the active component? Uh how do they need to be specialized?
[27:49] do they need to be specialized? uh how do you qualify
[27:52] uh how do you qualify lasers/phor diets?
[27:54] lasers/phor diets? >> Okay, so uh thank you for specifying the
[27:57] >> Okay, so uh thank you for specifying the question a bit more. Uh indeed uh you're
[28:00] question a bit more. Uh indeed uh you're right. So I I will not say it's not a
[28:02] right. So I I will not say it's not a challenge to bring these component
[28:04] challenge to bring these component providers on board. It's a bit of a
[28:05] providers on board. It's a bit of a challenge and requires some time and
[28:07] challenge and requires some time and motivation uh to see the the benefits of
[28:11] motivation uh to see the the benefits of the technology. Um but you also have to
[28:13] the technology. Um but you also have to keep in mind that a lot of these
[28:15] keep in mind that a lot of these component providers have already a
[28:17] component providers have already a technology portfolio and a lot of it can
[28:20] technology portfolio and a lot of it can essentially be reused uh for a transfer
[28:23] essentially be reused uh for a transfer print offering like for instance for
[28:24] print offering like for instance for photo detectors uh it can be uh very
[28:28] photo detectors uh it can be uh very similar to already known offerings. Uh I
[28:32] similar to already known offerings. Uh I think you know the biggest difference
[28:33] think you know the biggest difference but it could also be almost the only
[28:35] but it could also be almost the only difference is to introduce this
[28:37] difference is to introduce this sacrificial release layer underneath the
[28:39] sacrificial release layer underneath the structure. All the other parts uh are
[28:42] structure. All the other parts uh are fairly uh I mean are almost always a bit
[28:46] fairly uh I mean are almost always a bit plugandplay from different uh technology
[28:48] plugandplay from different uh technology modules that these component providers
[28:50] modules that these component providers have. Um then in terms of on the P side
[28:54] have. Um then in terms of on the P side um you know eancent coupling grading
[28:56] um you know eancent coupling grading coupling different types of coupling
[28:58] coupling different types of coupling these are not unique to transfer
[29:00] these are not unique to transfer printing. Uh for instance eancent
[29:02] printing. Uh for instance eancent coupling this is also being used for
[29:04] coupling this is also being used for wafer to wafer approach or die to wafer
[29:07] wafer to wafer approach or die to wafer approach. So also in that sense for us
[29:09] approach. So also in that sense for us it's not really um that we have to put a
[29:13] it's not really um that we have to put a major investment towards developing the
[29:15] major investment towards developing the platform specifically for this one
[29:17] platform specifically for this one direction. I think these technology
[29:19] direction. I think these technology modules can be reused across different
[29:21] modules can be reused across different integration methods but it just happens
[29:23] integration methods but it just happens to be the case that transfer printing
[29:25] to be the case that transfer printing offers some nice benefits uh towards
[29:27] offers some nice benefits uh towards scaling and and uh plugandplay features
[29:30] scaling and and uh plugandplay features for the different material systems that
[29:32] for the different material systems that can be integrated on silicon nitrite.
[29:36] can be integrated on silicon nitrite. Thanks.
[29:38] Thanks. >> Lots of excitement to come with transfer
[29:40] >> Lots of excitement to come with transfer printing with the partnership with Intel
[29:43] printing with the partnership with Intel with uh you know the work with Seagate.
[29:45] with uh you know the work with Seagate. So uh looking forward to following to
[29:48] So uh looking forward to following to see where that goes. Thank you very much
[29:50] see where that goes. Thank you very much for joining. So let's now zoom out uh
[29:53] for joining. So let's now zoom out uh from this concrete example and see how
[29:55] from this concrete example and see how others in the ecosystem are tackling
[29:57] others in the ecosystem are tackling similar challenges. I'm delighted to
[29:59] similar challenges. I'm delighted to welcome Philip Suzanne from IMC. Philipe
[30:02] welcome Philip Suzanne from IMC. Philipe leads strategic development efforts that
[30:04] leads strategic development efforts that explore heterogeneous integration op
[30:06] explore heterogeneous integration op options particularly how diverse
[30:08] options particularly how diverse materials and photonic building blocks
[30:11] materials and photonic building blocks can be combined with silicon platforms
[30:13] can be combined with silicon platforms to unlock new levels of performance and
[30:16] to unlock new levels of performance and manufacturability.
[30:17] manufacturability. His insights will ground the session in
[30:19] His insights will ground the session in both cutting edge research and practical
[30:21] both cutting edge research and practical integration pathways that are shaping
[30:24] integration pathways that are shaping the future of integrated batonics. Phipe
[30:27] the future of integrated batonics. Phipe please share your slides. The floor and
[30:29] please share your slides. The floor and the attention of everyone is yours.
[30:31] the attention of everyone is yours. Thank you very much uh John for the kind
[30:33] Thank you very much uh John for the kind introduction.
[30:35] introduction. Um let's get there.
[30:38] Um let's get there. Yeah.
[30:49] All right. So um I'll be going through
[30:52] All right. So um I'll be going through actually a set of different integration
[30:55] actually a set of different integration technique for it integration. And you
[30:57] technique for it integration. And you might see I have been taking uh the
[31:00] might see I have been taking uh the eerogenous integration in a broad sense
[31:03] eerogenous integration in a broad sense and try to see how this apply to uh
[31:05] and try to see how this apply to uh integrated photonics. So you might
[31:07] integrated photonics. So you might encounter maybe different kind of
[31:10] encounter maybe different kind of techniques and we can discuss later on
[31:12] techniques and we can discuss later on uh about the pros and contrs of the
[31:15] uh about the pros and contrs of the different approaches. So uh when we look
[31:18] different approaches. So uh when we look into what's going on um and I think that
[31:21] into what's going on um and I think that was partially depicted in the by the
[31:23] was partially depicted in the by the previous speaker we have the dataccom
[31:25] previous speaker we have the dataccom ecosystem we have the emerging and what
[31:27] ecosystem we have the emerging and what we do see as you move towards co package
[31:30] we do see as you move towards co package optics and there's a density uh of
[31:33] optics and there's a density uh of modulators that is increasing there is a
[31:35] modulators that is increasing there is a larger bandwidth so basically silicon
[31:37] larger bandwidth so basically silicon might not be able to cut it and we have
[31:40] might not be able to cut it and we have on the emerging type of things more the
[31:43] on the emerging type of things more the light sources that have actually a
[31:44] light sources that have actually a variety of wavelengths so these
[31:46] variety of wavelengths so these different kind of APS sometime is
[31:48] different kind of APS sometime is amplification and the price as well. We
[31:51] amplification and the price as well. We have this good old debate that is never
[31:53] have this good old debate that is never resolved about being on chip or off chip
[31:56] resolved about being on chip or off chip for the lasers. Um today in dataccom
[31:59] for the lasers. Um today in dataccom they are faborizing the offchip because
[32:02] they are faborizing the offchip because it's more serviceable but is maybe a
[32:04] it's more serviceable but is maybe a sweet spot as well to move the item on
[32:06] sweet spot as well to move the item on the chip. So talking about integration
[32:10] the chip. So talking about integration processes uh in the next slide you will
[32:12] processes uh in the next slide you will see first etero epitexi we have a
[32:14] see first etero epitexi we have a tendency to forget that but why would
[32:16] tendency to forget that but why would you like to pick up the tree five from a
[32:18] you like to pick up the tree five from a different substrate why don't you just
[32:20] different substrate why don't you just grow it directly on a substrate so
[32:22] grow it directly on a substrate so that's the first things I would like to
[32:23] that's the first things I would like to uh depict and then you have the
[32:26] uh depict and then you have the different item where we are bringing
[32:28] different item where we are bringing actually in that case I take uh the 35
[32:31] actually in that case I take uh the 35 by different technique uh to the p
[32:34] by different technique uh to the p substrate and you will end up with
[32:36] substrate and you will end up with different kind of configuration which I
[32:38] different kind of configuration which I grossly depicted here but I wanted to
[32:40] grossly depicted here but I wanted to discuss with you what is maybe important
[32:43] discuss with you what is maybe important uh when it comes to the manufacturing is
[32:45] uh when it comes to the manufacturing is that on the first side we don't need a
[32:47] that on the first side we don't need a foreign substrate we directly grown
[32:49] foreign substrate we directly grown substrates very researchy today but
[32:51] substrates very researchy today but there's some good sign that it might
[32:53] there's some good sign that it might work and then there's a question whether
[32:55] work and then there's a question whether your ep is coming from an AP or let's
[32:57] your ep is coming from an AP or let's say the etherogenous by extension the
[32:59] say the etherogenous by extension the non-silicon material is coming from a
[33:01] non-silicon material is coming from a different substrate and you're going to
[33:03] different substrate and you're going to contact it after you have transferred it
[33:05] contact it after you have transferred it or whether it's already pre-contacted
[33:07] or whether it's already pre-contacted when you bringing to the substrate. As
[33:09] when you bringing to the substrate. As you understood, moving from left to
[33:11] you understood, moving from left to right is more or less the technology
[33:14] right is more or less the technology readiness level that you're seeing uh
[33:17] readiness level that you're seeing uh but it's it's something you need to keep
[33:19] but it's it's something you need to keep in mind when it comes to manufacturing.
[33:21] in mind when it comes to manufacturing. So let's dive into uh the first one uh
[33:25] So let's dive into uh the first one uh which is the most extreme uh I would say
[33:28] which is the most extreme uh I would say which is this epipexi. IMAC has
[33:31] which is this epipexi. IMAC has demonstrated with the right reactor that
[33:33] demonstrated with the right reactor that you can actually grow marinide
[33:37] you can actually grow marinide directly onto a 300 mm substrate and we
[33:40] directly onto a 300 mm substrate and we have been demonstrating lazing so
[33:42] have been demonstrating lazing so there's still a lot of challenge when it
[33:44] there's still a lot of challenge when it comes to reliability and so forth but
[33:46] comes to reliability and so forth but it's a very promising approach pretty
[33:48] it's a very promising approach pretty much forward looking uh but since I
[33:51] much forward looking uh but since I tried to be comprehensive today I just
[33:53] tried to be comprehensive today I just wanted to show it to you moving next and
[33:56] wanted to show it to you moving next and I think that was extensively discussed
[33:58] I think that was extensively discussed in a previous uh talk is about tium
[34:00] in a previous uh talk is about tium phosphide and t infinitium bitates by
[34:03] phosphide and t infinitium bitates by micro transfer printing. We have
[34:05] micro transfer printing. We have actually been demonstrating that we can
[34:08] actually been demonstrating that we can use that technique for tinfium niobates.
[34:11] use that technique for tinfium niobates. We are actually reaching 2.85 vial right
[34:15] We are actually reaching 2.85 vial right now for this kind of device. We expect
[34:16] now for this kind of device. We expect we will be announcing that. So what is
[34:18] we will be announcing that. So what is very interesting for us is that we do
[34:20] very interesting for us is that we do already have 100 gigahertz uh PDS
[34:24] already have 100 gigahertz uh PDS actually in the silicon photonics
[34:26] actually in the silicon photonics platform. So we were only missing a
[34:28] platform. So we were only missing a modulator. That's very interesting. By
[34:30] modulator. That's very interesting. By extension you can on the bottom left you
[34:34] extension you can on the bottom left you can actually see that we can also bring
[34:35] can actually see that we can also bring S SOAS on LPCVD nitride and I don't need
[34:39] S SOAS on LPCVD nitride and I don't need to repeat uh the previous uh discussion
[34:42] to repeat uh the previous uh discussion for us. It's very interesting approach
[34:44] for us. It's very interesting approach because it's adapted to scale up
[34:46] because it's adapted to scale up manufacturing. It's already demonstrated
[34:48] manufacturing. It's already demonstrated on 200 mm and that's a very interesting
[34:51] on 200 mm and that's a very interesting uh approach.
[34:54] uh approach. Now I'll be diverting a bit. So we
[34:57] Now I'll be diverting a bit. So we looked into die bonding. Uh it's very
[34:59] looked into die bonding. Uh it's very popular as well. There is a lot of
[35:01] popular as well. There is a lot of enhancement being done today and this is
[35:04] enhancement being done today and this is actually wafer to wafer chip to wafer
[35:06] actually wafer to wafer chip to wafer bonding molecular bonding is very
[35:09] bonding molecular bonding is very mainstream on simosc majors on DAM and
[35:12] mainstream on simosc majors on DAM and it's also what's going on for the coup
[35:13] it's also what's going on for the coup processes for bringing the EIC on the
[35:16] processes for bringing the EIC on the peak from TSMC and we are actually
[35:19] peak from TSMC and we are actually facing different challenges to uh to get
[35:22] facing different challenges to uh to get there as well if we want to use such
[35:24] there as well if we want to use such approach. first of all is that we are
[35:26] approach. first of all is that we are not bonding directly silicon to silicon.
[35:29] not bonding directly silicon to silicon. Uh so there is actually a first
[35:31] Uh so there is actually a first challenge there which is mainly about
[35:33] challenge there which is mainly about the cleanliness of the singulation as
[35:36] the cleanliness of the singulation as well as uh the defectivity and
[35:38] well as uh the defectivity and reliability that you get from this
[35:40] reliability that you get from this approach. Still at IMC we are looking
[35:42] approach. Still at IMC we are looking into this kind of approach as well and
[35:44] into this kind of approach as well and that's what I want to show on the next
[35:46] that's what I want to show on the next slide. So this is an absolutely non
[35:48] slide. So this is an absolutely non photonics topic but I think it's
[35:50] photonics topic but I think it's important that you uh look what this
[35:52] important that you uh look what this could mean for photonics is that we have
[35:55] could mean for photonics is that we have been demonstrating uh using uh the soy
[35:57] been demonstrating uh using uh the soy tech approach the imposi substrate so
[35:59] tech approach the imposi substrate so basically it's an Indian phosphide on
[36:01] basically it's an Indian phosphide on silicon uh and this Indian phosfi itself
[36:05] silicon uh and this Indian phosfi itself actually is under is going through the
[36:09] actually is under is going through the smart technology and we have been
[36:12] smart technology and we have been showing that we can manipulate Indian
[36:14] showing that we can manipulate Indian phosphi grown on silicon like a normal
[36:16] phosphi grown on silicon like a normal silicon die and we have shown that we
[36:19] silicon die and we have shown that we could use that in that case for a high
[36:20] could use that in that case for a high frequency device by extension you can
[36:23] frequency device by extension you can actually see that this would be very
[36:25] actually see that this would be very interesting as well for other kind of
[36:28] interesting as well for other kind of indium phosphide uh type of technology
[36:31] indium phosphide uh type of technology now I'll go into another item which we
[36:35] now I'll go into another item which we have been using more with a focal plane
[36:37] have been using more with a focal plane array focus so we can also perfectly
[36:40] array focus so we can also perfectly take the 35 itself and bond it here you
[36:44] take the 35 itself and bond it here you can see it's a very large square So very
[36:46] can see it's a very large square So very different use case for instance that MTP
[36:48] different use case for instance that MTP and I make we're pursuing both but it's
[36:50] and I make we're pursuing both but it's interesting to see that such approach
[36:52] interesting to see that such approach actually has its own merits and when we
[36:55] actually has its own merits and when we talk about integrated photonics in a
[36:57] talk about integrated photonics in a broad sense you can understand that if
[36:59] broad sense you can understand that if you want to make an array of emitters uh
[37:01] you want to make an array of emitters uh that is actually going to be extremely
[37:03] that is actually going to be extremely uh interesting as well. Here in this
[37:06] uh interesting as well. Here in this example we actually are showing that
[37:08] example we actually are showing that whether uh the device it's in that case
[37:11] whether uh the device it's in that case it's a p pin photo diet is behaving in
[37:14] it's a p pin photo diet is behaving in the same way before or after transfer.
[37:17] the same way before or after transfer. Finally and I think that was uh now
[37:20] Finally and I think that was uh now known for some time and it's already in
[37:23] known for some time and it's already in production uh you can use as well chip
[37:26] production uh you can use as well chip uh to bring your in that case a 35
[37:29] uh to bring your in that case a 35 directly on the substrate. It's it's
[37:32] directly on the substrate. It's it's actually very interesting because the
[37:33] actually very interesting because the existing OSAT infrastructure when it
[37:35] existing OSAT infrastructure when it comes to plating and D bonding is
[37:37] comes to plating and D bonding is already existing and there are use cases
[37:40] already existing and there are use cases uh where such an approach can be used
[37:42] uh where such an approach can be used because it does not require any
[37:44] because it does not require any magnification actually of the starting
[37:46] magnification actually of the starting substrate you can reuse the lasers the
[37:48] substrate you can reuse the lasers the DFB lasers and so forth that are
[37:50] DFB lasers and so forth that are actually uh made directly by the
[37:52] actually uh made directly by the compound foundaries. So the question is
[37:55] compound foundaries. So the question is ah Philip you're showing a lot of things
[37:57] ah Philip you're showing a lot of things uh but what does that mean? So what is
[37:59] uh but what does that mean? So what is very important is that each technique
[38:01] very important is that each technique has its merits when it comes to the
[38:04] has its merits when it comes to the relative area. So what is going to be
[38:05] relative area. So what is going to be the ratio that you of that material
[38:08] the ratio that you of that material you're going to have on the final
[38:09] you're going to have on the final substrate and the die that you need to
[38:11] substrate and the die that you need to pick up and and this is actually here
[38:14] pick up and and this is actually here the message is that you you do not have
[38:16] the message is that you you do not have an let's say an absolute answer to which
[38:19] an let's say an absolute answer to which intergenous integration is the best. It
[38:21] intergenous integration is the best. It actually depends on this ratio. Moving
[38:25] actually depends on this ratio. Moving further, if you now look at the
[38:27] further, if you now look at the industrial reality, the industriality on
[38:29] industrial reality, the industriality on the right side is that everything that
[38:30] the right side is that everything that has to do with silicon bonding is very
[38:32] has to do with silicon bonding is very mainstream in the silicon world is not
[38:34] mainstream in the silicon world is not actually in integrated photonics, but we
[38:37] actually in integrated photonics, but we do see layer transfer and chip bonding
[38:39] do see layer transfer and chip bonding has been uh you know quite well adopted
[38:42] has been uh you know quite well adopted so far. So that's interesting. And
[38:45] so far. So that's interesting. And finally and that's going to be a non uh
[38:48] finally and that's going to be a non uh decisive let's say conclusion from my
[38:50] decisive let's say conclusion from my side is people are asking what actually
[38:53] side is people are asking what actually is the interesting where is the optimum
[38:55] is the interesting where is the optimum well you know actually depends on the
[38:56] well you know actually depends on the use case uh usually what we do in iMake
[39:00] use case uh usually what we do in iMake we look at the throughput so that's
[39:01] we look at the throughput so that's something we don't mention but
[39:02] something we don't mention but throughput is very important for the
[39:04] throughput is very important for the cost of ownership of the technology we
[39:06] cost of ownership of the technology we look also what is the material that you
[39:08] look also what is the material that you are transferring you can understand that
[39:10] are transferring you can understand that transferring gas or Indian pulfide or
[39:13] transferring gas or Indian pulfide or very precious
[39:14] very precious lithium nubates is not the same. So
[39:16] lithium nubates is not the same. So you're going to end up with a cost and
[39:18] you're going to end up with a cost and as was mentioned previously by Alexon
[39:20] as was mentioned previously by Alexon the donor substrate reusability is also
[39:22] the donor substrate reusability is also very important. So on this I finish and
[39:26] very important. So on this I finish and with a nonconclusion just to tell you
[39:28] with a nonconclusion just to tell you that depending on the use case whether
[39:30] that depending on the use case whether you have a modulators again medium and
[39:32] you have a modulators again medium and the amount of material that you're going
[39:34] the amount of material that you're going to put on your target substrate you're
[39:36] to put on your target substrate you're going to pick up the best technique.
[39:37] going to pick up the best technique. Thank you for your attention.
[39:39] Thank you for your attention. >> Thank you so much Philip. Um you
[39:42] >> Thank you so much Philip. Um you mentioned TFLN that's been very very
[39:44] mentioned TFLN that's been very very popular um over recent times. So what
[39:47] popular um over recent times. So what are the real bottlenecks for scaling
[39:49] are the real bottlenecks for scaling TFLN adoption right now? Is it wafer
[39:53] TFLN adoption right now? Is it wafer size yield uh integration complex
[39:56] size yield uh integration complex complexity or is it packaging would you
[39:58] complexity or is it packaging would you say?
[40:00] say? >> Um I think that um the cost of TFLN is
[40:05] >> Um I think that um the cost of TFLN is still a major detractor to the adoption
[40:07] still a major detractor to the adoption of the technology. Um and indeed um
[40:12] of the technology. Um and indeed um there is going to be different use case.
[40:14] there is going to be different use case. Uh for instance in the field of
[40:16] Uh for instance in the field of pluggables you only have a few
[40:17] pluggables you only have a few components. So here in that case for
[40:19] components. So here in that case for chasing 400 gig per lane uh that might
[40:22] chasing 400 gig per lane uh that might be a very good use case for micro
[40:24] be a very good use case for micro transfer printing. Now whether people
[40:27] transfer printing. Now whether people had moved to CPU and have way more
[40:30] had moved to CPU and have way more modulators we'll have to see actually.
[40:32] modulators we'll have to see actually. So it's really a an economical question.
[40:35] So it's really a an economical question. In term of manufacturing readiness, I
[40:38] In term of manufacturing readiness, I think the eterogeneous integration part
[40:41] think the eterogeneous integration part is fairly mature because even in in labs
[40:44] is fairly mature because even in in labs or in small entities like like Gent, you
[40:48] or in small entities like like Gent, you can actually see quite good yield
[40:49] can actually see quite good yield already. Um and there is probably
[40:53] already. Um and there is probably another potential challenge is the so
[40:57] another potential challenge is the so lithium is a contaminating material. So
[40:59] lithium is a contaminating material. So you need specific infrastructure. If you
[41:01] you need specific infrastructure. If you want to move that to the place that is
[41:03] want to move that to the place that is manufacturing the peak then you have
[41:05] manufacturing the peak then you have some specific contamination specifically
[41:08] some specific contamination specifically if photonix is made into in semos fab.
[41:11] if photonix is made into in semos fab. If you now are a bit more open and can
[41:13] If you now are a bit more open and can tolerate that things that that is
[41:15] tolerate that things that that is something but you need volumes and it's
[41:16] something but you need volumes and it's a bit of a chicken and egg right now. Do
[41:18] a bit of a chicken and egg right now. Do you need specific infrastructure for
[41:21] you need specific infrastructure for contaminated you know corridors? Uh
[41:23] contaminated you know corridors? Uh that's actually the the big question
[41:25] that's actually the the big question right now
[41:27] right now >> and indeed and today is actually a very
[41:30] >> and indeed and today is actually a very timely following the press release uh of
[41:33] timely following the press release uh of your collaboration uh between IMC and
[41:36] your collaboration uh between IMC and VCO developing a 300 nanometer platform
[41:39] VCO developing a 300 nanometer platform for the epitexi of BTO um something I
[41:42] for the epitexi of BTO um something I think only quantum could boast up to
[41:45] think only quantum could boast up to now. So what do you expect BTO on
[41:47] now. So what do you expect BTO on silicon to unlock?
[41:50] silicon to unlock? So um from what we know um team
[41:53] So um from what we know um team [clears throat]
[41:53] [clears throat] lithium na beta as a material has the
[41:55] lithium na beta as a material has the lowest loss has the highest speed uh I
[41:59] lowest loss has the highest speed uh I think BTO will not be competing that
[42:01] think BTO will not be competing that even though in iMake we will see what we
[42:03] even though in iMake we will see what we can do but what we were particularly
[42:06] can do but what we were particularly chasing uh with the 300 mm work on BTO
[42:10] chasing uh with the 300 mm work on BTO is first to lower the cost of ownership
[42:14] is first to lower the cost of ownership because indeed there was for those who
[42:16] because indeed there was for those who don't know here if you Google you will
[42:17] don't know here if you Google you will see there is a prince announcement
[42:18] see there is a prince announcement between IMC
[42:20] between IMC and vehicle we are actually chasing to
[42:22] and vehicle we are actually chasing to move away from MBE as a technique. I
[42:26] move away from MBE as a technique. I cannot say more about it publicly and to
[42:30] cannot say more about it publicly and to show that it can be a process that can
[42:32] show that it can be a process that can be scalable. Now the integration of the
[42:34] be scalable. Now the integration of the BTO on the P is a subject of research in
[42:37] BTO on the P is a subject of research in IMC and it will depend on this relative
[42:39] IMC and it will depend on this relative area that I have mentioned. So whether
[42:42] area that I have mentioned. So whether we need to transfer centime size coupon
[42:45] we need to transfer centime size coupon or much smaller uh you know component is
[42:48] or much smaller uh you know component is going to be making the decision. Uh it
[42:51] going to be making the decision. Uh it could even be that it's wafer to wer
[42:52] could even be that it's wafer to wer that is being picked up at the end of
[42:53] that is being picked up at the end of the day.
[42:55] the day. >> Very exciting
[42:56] >> Very exciting >> if the cost of if the cost is low enough
[42:58] >> if the cost of if the cost is low enough right that's always the same question
[43:01] right that's always the same question >> indeed. Well congratulations on the the
[43:03] >> indeed. Well congratulations on the the collaboration. Very excited to see where
[43:05] collaboration. Very excited to see where it goes. Um so thank you Philip for
[43:07] it goes. Um so thank you Philip for joining. Uh we're very excited now to
[43:09] joining. Uh we're very excited now to welcome Matt Guzzi from Lmentum uh today
[43:13] welcome Matt Guzzi from Lmentum uh today today's meeting. Matt brings deep and
[43:15] today's meeting. Matt brings deep and broad experience in semiconductor
[43:17] broad experience in semiconductor packaging and assembly shaped by more
[43:19] packaging and assembly shaped by more than 21 years of leading complex
[43:21] than 21 years of leading complex integration initi initiatives at scale
[43:24] integration initi initiatives at scale at Intel. Now at Lamentum, he has been
[43:27] at Intel. Now at Lamentum, he has been central to aligning photonic component
[43:29] central to aligning photonic component development with real world
[43:30] development with real world manufacturing and packaging
[43:32] manufacturing and packaging requirements, translating high
[43:34] requirements, translating high performance designs into products that
[43:36] performance designs into products that can be reliably produced and deployed.
[43:39] can be reliably produced and deployed. Matt, we look forward to your insights
[43:40] Matt, we look forward to your insights on how packaging expertise is powering
[43:43] on how packaging expertise is powering the next wave of silicon photonics
[43:45] the next wave of silicon photonics innovation. The floor and the attention
[43:47] innovation. The floor and the attention of everyone is yours.
[43:49] of everyone is yours. >> Cool. Uh thanks for the uh introduction,
[43:51] >> Cool. Uh thanks for the uh introduction, John. Um yeah uh like John said I uh am
[43:55] John. Um yeah uh like John said I uh am in the CTO office at Lumenum. I've been
[43:58] in the CTO office at Lumenum. I've been there uh a bit under a year after a uh
[44:01] there uh a bit under a year after a uh pretty lengthy career at Intel. Um you
[44:04] pretty lengthy career at Intel. Um you know I spent that entire time developing
[44:07] know I spent that entire time developing advanced packaging flows and working on
[44:09] advanced packaging flows and working on heterogeneous integration. Of course
[44:11] heterogeneous integration. Of course advanced packaging 20 years ago isn't
[44:13] advanced packaging 20 years ago isn't quite what it is today. Um, so I've had
[44:15] quite what it is today. Um, so I've had a a really uh interesting history and I
[44:18] a a really uh interesting history and I think um my background is really uh why
[44:21] think um my background is really uh why I find the this this topic of today's
[44:24] I find the this this topic of today's session so powerful. Um I've seen the
[44:26] session so powerful. Um I've seen the effect that heterogeneous integration
[44:28] effect that heterogeneous integration has had in the um you know in the
[44:32] has had in the um you know in the semiconductor space and I see you know
[44:35] semiconductor space and I see you know the the photonics and optics industry
[44:37] the the photonics and optics industry ready to take advantage of those similar
[44:40] ready to take advantage of those similar techniques and get a lot of the same
[44:42] techniques and get a lot of the same benefits.
[44:44] benefits. Um so
[44:47] Um so at Lamentum um you know we are um
[44:51] at Lamentum um you know we are um experts across
[44:53] experts across >> [snorts]
[44:53] >> [snorts] >> um interconnect technologies um we both
[44:56] >> um interconnect technologies um we both provide 35 uh components. So individual
[45:00] provide 35 uh components. So individual uh lasers and our Vixel um platform is
[45:05] uh lasers and our Vixel um platform is uh widely known. So if you uh if you
[45:07] uh widely known. So if you uh if you have an iPhone and you unlock your phone
[45:09] have an iPhone and you unlock your phone with your face, you're using Vixels. Um
[45:12] with your face, you're using Vixels. Um but not only do we bring individual 35
[45:15] but not only do we bring individual 35 components to market um we also do offer
[45:18] components to market um we also do offer entire um package solutions um such as
[45:22] entire um package solutions um such as our 300x300
[45:24] our 300x300 MEMSbased uh optical switch. Um so we
[45:27] MEMSbased uh optical switch. Um so we really sit in a unique place where um
[45:30] really sit in a unique place where um not only do we understand the device and
[45:33] not only do we understand the device and the device physics but we also
[45:34] the device physics but we also understand the integrations and taking
[45:36] understand the integrations and taking something all the way to a high volume
[45:38] something all the way to a high volume product. Um so it's really an
[45:40] product. Um so it's really an interesting place um to sit now with
[45:43] interesting place um to sit now with where the industry is with the um you
[45:46] where the industry is with the um you know the AI revolution if you will. Um,
[45:49] know the AI revolution if you will. Um, and what's really great about um, you
[45:52] and what's really great about um, you know, AI is the inflect the the the
[45:54] know, AI is the inflect the the the influence it's having on the market,
[45:56] influence it's having on the market, right? Um, if you're getting a larger
[45:58] right? Um, if you're getting a larger model, right, if you're you're on the
[46:01] model, right, if you're you're on the latest version of um, of Chat GPT or if
[46:04] latest version of um, of Chat GPT or if you're training a model, um, you're
[46:06] you're training a model, um, you're going to be pushing your hardware
[46:07] going to be pushing your hardware resources uh to the limit. So, you ended
[46:09] resources uh to the limit. So, you ended up needing to buy more and more stuff.
[46:12] up needing to buy more and more stuff. Um, you know, so you've got all this
[46:13] Um, you know, so you've got all this capital investment that's going on. um
[46:16] capital investment that's going on. um you know at some point someone you know
[46:17] you know at some point someone you know someone's going to look at the books and
[46:18] someone's going to look at the books and be like well we need to find a better
[46:21] be like well we need to find a better lower cost faster solution to do that
[46:23] lower cost faster solution to do that which then drives um technology
[46:26] which then drives um technology innovation which then brings those
[46:27] innovation which then brings those efficiencies and brings those costs down
[46:29] efficiencies and brings those costs down but once you have the opportunity in
[46:31] but once you have the opportunity in operating a lower cost you go ahead and
[46:33] operating a lower cost you go ahead and bring in the next model on which is
[46:34] bring in the next model on which is going to be even larger so we're in this
[46:36] going to be even larger so we're in this cycle where we continually go on where
[46:38] cycle where we continually go on where the AI scaling that's occurring between
[46:41] the AI scaling that's occurring between both the teaching the modules and the uh
[46:45] both the teaching the modules and the uh the inference that's going on is really
[46:47] the inference that's going on is really driving a lot of innovation and really
[46:49] driving a lot of innovation and really pushing the industry forward. Um it's
[46:51] pushing the industry forward. Um it's really an exciting time to uh you know
[46:54] really an exciting time to uh you know for me it was really an exciting time to
[46:56] for me it was really an exciting time to enter um this space to really see how
[47:00] enter um this space to really see how those packaging techniques can apply um
[47:04] those packaging techniques can apply um at the most straightforward manner. You
[47:06] at the most straightforward manner. You know, I I think heterogeneous
[47:08] know, I I think heterogeneous integration and uh advanced packaging
[47:11] integration and uh advanced packaging techniques are really necessary um to
[47:15] techniques are really necessary um to allow co-packaged optics um to really
[47:18] allow co-packaged optics um to really take uh the the primary um the primary
[47:23] take uh the the primary um the primary seat of being the electrical to optic
[47:25] seat of being the electrical to optic solution. Um what we're seeing is that
[47:28] solution. Um what we're seeing is that the pluggable optics um you know that
[47:30] the pluggable optics um you know that have been uh common are slowly being
[47:33] have been uh common are slowly being pushed to NPO near packaged optics with
[47:36] pushed to NPO near packaged optics with the eventual target of having a
[47:38] the eventual target of having a co-ackaged optics solution. The benefit
[47:40] co-ackaged optics solution. The benefit is here as you bring the your um as you
[47:43] is here as you bring the your um as you bring your optic connection closer to
[47:45] bring your optic connection closer to your electronics uh you get faster
[47:48] your electronics uh you get faster bandwidth, lower power, um you know, a
[47:50] bandwidth, lower power, um you know, a higher degree of efficiency, but you are
[47:53] higher degree of efficiency, but you are trading it off um for uh more
[47:57] trading it off um for uh more complicated, you know, essentially if if
[47:59] complicated, you know, essentially if if you burn out a laser in your co-acked
[48:01] you burn out a laser in your co-acked solution, you can't really replace it
[48:03] solution, you can't really replace it easily. Um you're replacing the entire
[48:05] easily. Um you're replacing the entire chip. Um and also just the integration
[48:08] chip. Um and also just the integration uh of these devices is much more complex
[48:11] uh of these devices is much more complex than simply plugging a board into your
[48:13] than simply plugging a board into your um into your PCB. But this translation
[48:16] um into your PCB. But this translation and this this evolution of your um optic
[48:19] and this this evolution of your um optic solution from being something that
[48:21] solution from being something that exists uh on the board far away from
[48:24] exists uh on the board far away from your um AS6 uh to something that
[48:26] your um AS6 uh to something that coexists on the same substrate or even
[48:28] coexists on the same substrate or even on the same die um as your ASIC. um it's
[48:32] on the same die um as your ASIC. um it's really enabled by you know heterogeneous
[48:35] really enabled by you know heterogeneous integration and advanced packaging
[48:36] integration and advanced packaging because you're not going to use the same
[48:38] because you're not going to use the same silicon nodes um that you're using for
[48:41] silicon nodes um that you're using for your high performance compute. You're
[48:43] your high performance compute. You're not going to be using that same silicon
[48:45] not going to be using that same silicon node for you know your light source or
[48:48] node for you know your light source or your photo detectors. It you know you're
[48:50] your photo detectors. It you know you're just not going to do that. That's not
[48:52] just not going to do that. That's not going to be a uh a efficient solution
[48:55] going to be a uh a efficient solution for anything. So really I see advanced
[48:58] for anything. So really I see advanced packaging um really essential here but
[49:02] packaging um really essential here but I'm actually more excited about the
[49:03] I'm actually more excited about the other benefits um that these techniques
[49:06] other benefits um that these techniques bring to the industry because this is
[49:09] bring to the industry because this is this is some amazing enabling but uh the
[49:12] this is some amazing enabling but uh the power of these techniques and processes
[49:14] power of these techniques and processes go far beyond this. Um and I I sort of
[49:18] go far beyond this. Um and I I sort of see you know having done seen this
[49:21] see you know having done seen this evolve at Intel I think a lot of the
[49:24] evolve at Intel I think a lot of the same benefits can hear the first is um
[49:28] same benefits can hear the first is um it gives us an alternate pathway to
[49:31] it gives us an alternate pathway to extend a Moore's law type scaling um
[49:34] extend a Moore's law type scaling um it's not simply you know I don't need to
[49:37] it's not simply you know I don't need to push a single laser um faster and faster
[49:40] push a single laser um faster and faster and faster I can you know I can
[49:43] and faster I can you know I can integrate multiple laser sources run
[49:45] integrate multiple laser sources run multiple channels and get the bandwidth
[49:47] multiple channels and get the bandwidth that way. I don't need to be driving a
[49:49] that way. I don't need to be driving a single channel at 400G or 800G. I can
[49:52] single channel at 400G or 800G. I can find techniques of using several 200 or
[49:55] find techniques of using several 200 or 100G lasers and get that same scaling um
[50:00] 100G lasers and get that same scaling um but not need to extend my power envelope
[50:03] but not need to extend my power envelope or push my fab process so hard. Um and
[50:06] or push my fab process so hard. Um and we're even seeing this today with the
[50:07] we're even seeing this today with the aggressive growth of optical signaling
[50:09] aggressive growth of optical signaling bandwidth uh generation over generation.
[50:12] bandwidth uh generation over generation. um if we were simply being limited by
[50:14] um if we were simply being limited by the device physics, we wouldn't be able
[50:15] the device physics, we wouldn't be able to keep up with the bandwidth
[50:16] to keep up with the bandwidth requirements. So, we're needing to find
[50:18] requirements. So, we're needing to find alternate ways to satisfy that demand.
[50:21] alternate ways to satisfy that demand. Also, like I I said on on the previous
[50:23] Also, like I I said on on the previous layout about co-ackage optics, but this
[50:25] layout about co-ackage optics, but this um heterogeneous integration really
[50:27] um heterogeneous integration really allows us to bring the best solution for
[50:33] allows us to bring the best solution for each individual component. So, uh you
[50:36] each individual component. So, uh you can mix multiple optic technologies,
[50:38] can mix multiple optic technologies, right? So if you need a uh high quality
[50:42] right? So if you need a uh high quality laser source, you might look at a 35
[50:45] laser source, you might look at a 35 solution. You might be looking at GIM
[50:47] solution. You might be looking at GIM Mars 9 laser or a Vixel or um a pump
[50:51] Mars 9 laser or a Vixel or um a pump laser. Um for modulation, you might use
[50:54] laser. Um for modulation, you might use a TFLN, right? If you want that
[50:56] a TFLN, right? If you want that high-speed modulation, you're not going
[50:58] high-speed modulation, you're not going to do better than TFLN. Um but TFN can't
[51:01] to do better than TFLN. Um but TFN can't on its own. It doesn't generate light
[51:03] on its own. It doesn't generate light and it doesn't have a photo detector. So
[51:05] and it doesn't have a photo detector. So that in itself is not a complete
[51:06] that in itself is not a complete solution. um silicon phetonics uh you
[51:10] solution. um silicon phetonics uh you need to get the light from a location A
[51:12] need to get the light from a location A to B. Silicon Phetonics has uh amazing
[51:14] to B. Silicon Phetonics has uh amazing capabilities with waveguides and the
[51:17] capabilities with waveguides and the mucks and demucks and all the integrated
[51:19] mucks and demucks and all the integrated solutions that you can have within a
[51:20] solutions that you can have within a silicon phetonix um platform. But again,
[51:23] silicon phetonix um platform. But again, silicon phetonics lasers, their
[51:25] silicon phetonics lasers, their performance really isn't always there.
[51:27] performance really isn't always there. They are they show weird reliability um
[51:31] They are they show weird reliability um impacts and it's not going to
[51:33] impacts and it's not going to necessarily be your best solution. But
[51:36] necessarily be your best solution. But with heterogeneous integration is you
[51:37] with heterogeneous integration is you can take these components sort of almost
[51:39] can take these components sort of almost as a alocart menu right like you're at a
[51:42] as a alocart menu right like you're at a tapas restaurant and you can grab uh the
[51:45] tapas restaurant and you can grab uh the components that give you the best
[51:47] components that give you the best overall solution and the strongest
[51:50] overall solution and the strongest performance and also the best cost story
[51:53] performance and also the best cost story and put all of these together within a
[51:56] and put all of these together within a package solution and really give you the
[51:59] package solution and really give you the uh best-in-class
[52:02] uh best-in-class overall product by mixing these
[52:04] overall product by mixing these different technology elements. Um it's
[52:06] different technology elements. Um it's you know very similar to uh you would
[52:09] you know very similar to uh you would use one you know if you're putting
[52:11] use one you know if you're putting memory in a package uh with a compute
[52:14] memory in a package uh with a compute core. You're going to use your cutting
[52:17] core. You're going to use your cutting edge technology node for your compute
[52:19] edge technology node for your compute cores right your smallest possible
[52:21] cores right your smallest possible transistors you're going to want there.
[52:23] transistors you're going to want there. Your memory though your memory doesn't
[52:24] Your memory though your memory doesn't need to exist on that. So you can use an
[52:26] need to exist on that. So you can use an older cheaper node and integrate your
[52:28] older cheaper node and integrate your memory with your compute that way. It's
[52:30] memory with your compute that way. It's the exact same thought process here. Um,
[52:33] the exact same thought process here. Um, also I think one place, and I didn't put
[52:35] also I think one place, and I didn't put this on on on the slide, but yield
[52:39] this on on on the slide, but yield improvements. Um, if you're looking to
[52:41] improvements. Um, if you're looking to make an array of lasers, um, you know,
[52:44] make an array of lasers, um, you know, 35, the yields aren't always great, and
[52:47] 35, the yields aren't always great, and you're going to, you know, if you have a
[52:48] you're going to, you know, if you have a 90% device yield for a single laser
[52:51] 90% device yield for a single laser device, and you're looking to integrate
[52:53] device, and you're looking to integrate two of those, your yield will drop to
[52:55] two of those, your yield will drop to about 80%. If you're going down to, you
[52:58] about 80%. If you're going down to, you know, if you're going to four, you'll be
[52:59] know, if you're going to four, you'll be somewhere at 60%. So you're taking a
[53:01] somewhere at 60%. So you're taking a cost tip by providing an array of
[53:03] cost tip by providing an array of lasers. However, if you can integrate
[53:05] lasers. However, if you can integrate those lasers separately and
[53:07] those lasers separately and individually, you can maintain your fab
[53:09] individually, you can maintain your fab yield and keep that um economy of scale
[53:13] yield and keep that um economy of scale and that um cost story and integrate
[53:17] and that um cost story and integrate them together in a packaging solution
[53:18] them together in a packaging solution which will be lower cost than what you
[53:20] which will be lower cost than what you would pay on the fabit. So you can get
[53:21] would pay on the fabit. So you can get your four laser array um but at a lower
[53:25] your four laser array um but at a lower cost than a native fab solution. um it's
[53:28] cost than a native fab solution. um it's very similar by putting two 64 kit two
[53:31] very similar by putting two 64 kit two 64 core chips together and getting a 128
[53:34] 64 core chips together and getting a 128 core versus doing a native 128 core and
[53:37] core versus doing a native 128 core and getting a fraction of the yield. Um and
[53:40] getting a fraction of the yield. Um and then also what's great especially with
[53:42] then also what's great especially with how fast the cycle is moving is by using
[53:45] how fast the cycle is moving is by using a chiplet based architecture or um by
[53:48] a chiplet based architecture or um by reusing specific elements that have
[53:51] reusing specific elements that have previously be been validated. You can
[53:54] previously be been validated. You can reuse that IP in the next package
[53:57] reuse that IP in the next package forward and you don't necessarily need
[53:59] forward and you don't necessarily need to revalidate it or you don't need to
[54:00] to revalidate it or you don't need to redo your reliability data collection.
[54:03] redo your reliability data collection. you only need to worry about validating
[54:05] you only need to worry about validating the integration of that with the system.
[54:07] the integration of that with the system. And the individual components you
[54:09] And the individual components you already know are certified and you
[54:11] already know are certified and you already know that they work. So this can
[54:12] already know that they work. So this can both reduce your design timeline um
[54:14] both reduce your design timeline um because you're not needing to do a full
[54:16] because you're not needing to do a full chip design. You're simply using this
[54:18] chip design. You're simply using this module um but it also can reduce your
[54:20] module um but it also can reduce your validation timelines. So you can get a
[54:22] validation timelines. So you can get a faster time to market um and sort of get
[54:25] faster time to market um and sort of get ahead on that performance core and get
[54:27] ahead on that performance core and get your parts out to market um faster. So I
[54:31] your parts out to market um faster. So I find all of these sorts of advantage to
[54:35] find all of these sorts of advantage to be almost as powerful as simply enabling
[54:38] be almost as powerful as simply enabling co-ackaged optics. Yes, a co-acked
[54:40] co-ackaged optics. Yes, a co-acked optics solution, something that you can
[54:42] optics solution, something that you can slap into a coupe platform. Uh that's
[54:44] slap into a coupe platform. Uh that's kind of the holy grail right now. And to
[54:46] kind of the holy grail right now. And to find that way to mix these perfect
[54:48] find that way to mix these perfect ingredients into coupe, that is
[54:51] ingredients into coupe, that is incredibly powerful. But even at a more
[54:53] incredibly powerful. But even at a more basic level, all of these advanced
[54:55] basic level, all of these advanced packaging um can impact your um cost
[54:59] packaging um can impact your um cost story and your product offerings across
[55:02] story and your product offerings across the board and they really open up things
[55:04] the board and they really open up things that were extremely challenging to do
[55:06] that were extremely challenging to do previously. Um there's a lot of
[55:08] previously. Um there's a lot of excitement here. This is um this is
[55:10] excitement here. This is um this is actually where a fair amount of my
[55:12] actually where a fair amount of my personal work at Momentum um is based is
[55:15] personal work at Momentum um is based is trying to apply these techniques to
[55:17] trying to apply these techniques to solve some of these sorts of problems.
[55:21] And you know, at Lmentum, I
[55:24] And you know, at Lmentum, I unfortunately I'm not quite ready to to
[55:27] unfortunately I'm not quite ready to to um bring what we're doing to prime time
[55:29] um bring what we're doing to prime time yet. Um it's not quite uh ready to be on
[55:32] yet. Um it's not quite uh ready to be on stage, but we do have one very clear
[55:34] stage, but we do have one very clear example of some of our 2D 2.5D packaging
[55:38] example of some of our 2D 2.5D packaging and that is our Vixelon driver. Um so
[55:42] and that is our Vixelon driver. Um so our vixels are our vertically our
[55:43] our vixels are our vertically our vertical cavity emitting lasers. Um, so
[55:46] vertical cavity emitting lasers. Um, so basically this is a um laser chip and
[55:50] basically this is a um laser chip and this is stacked directly on the driver
[55:52] this is stacked directly on the driver which then would be wire bonded to a
[55:54] which then would be wire bonded to a PCB. Um but here you're getting the
[55:57] PCB. Um but here you're getting the fastest possible connection between your
[55:59] fastest possible connection between your your driver and your laser. um in a very
[56:02] your driver and your laser. um in a very small form factor um lower power um
[56:06] small form factor um lower power um better performance um and this sorts of
[56:10] better performance um and this sorts of module has is in high volume production
[56:14] module has is in high volume production and uh I'd say specifically vixels um
[56:18] and uh I'd say specifically vixels um you know vixels are an amazing optical
[56:20] you know vixels are an amazing optical source technology and we are
[56:23] source technology and we are specifically looking at a lot of ways to
[56:25] specifically looking at a lot of ways to use heterogeneous integration um to
[56:28] use heterogeneous integration um to bring vixels into different platforms
[56:30] bring vixels into different platforms and how to combine vixels with different
[56:33] and how to combine vixels with different sorts of chips and different um and
[56:36] sorts of chips and different um and different architectures using these
[56:38] different architectures using these sorts of um packaging. So, while I can't
[56:43] sorts of um packaging. So, while I can't quite show some of our killer apps that
[56:45] quite show some of our killer apps that we have in planned for applying these
[56:47] we have in planned for applying these techniques, um I'm really excited for
[56:50] techniques, um I'm really excited for for where it's going and really the
[56:52] for where it's going and really the opportunities that that this technology
[56:55] opportunities that that this technology is opening up. And um hopefully in the
[56:57] is opening up. And um hopefully in the the coming months you'll see some pretty
[56:59] the coming months you'll see some pretty cool packages coming out of Momentum
[57:01] cool packages coming out of Momentum that that leverage these techniques.
[57:07] >> Thanks so much Matt for for joining and
[57:09] >> Thanks so much Matt for for joining and I I personally cannot wait for um prime
[57:12] I I personally cannot wait for um prime time and for the for the big reveals. Um
[57:15] time and for the for the big reveals. Um I'm assuming something big at OFC. Um
[57:18] I'm assuming something big at OFC. Um but maybe you can't commit to that. I
[57:21] but maybe you can't commit to that. I mean, I'd say that I mean, I I think
[57:23] mean, I'd say that I mean, I I think that would be a great opportunity to
[57:25] that would be a great opportunity to share something.
[57:27] share something. >> Fantastic. So, I uh in the interest of
[57:30] >> Fantastic. So, I uh in the interest of time, I'm going to move on. I want to
[57:31] time, I'm going to move on. I want to didn't want to stop you is great
[57:33] didn't want to stop you is great insights, but um so next up is Steve
[57:36] insights, but um so next up is Steve Aliston from Open Light Phutonics. Steve
[57:39] Aliston from Open Light Phutonics. Steve brings a platform level perspective on
[57:41] brings a platform level perspective on how silicon phutonics design, boundary
[57:44] how silicon phutonics design, boundary access, and heterogeneous integration
[57:46] access, and heterogeneous integration come together in practice. Steve, the
[57:49] come together in practice. Steve, the floor and the attention of everyone is
[57:50] floor and the attention of everyone is yours. Thank you.
[57:52] yours. Thank you. >> So, thank you John. Um, so yeah, it's
[57:54] >> So, thank you John. Um, so yeah, it's great to be with everyone this morning.
[57:56] great to be with everyone this morning. Um, I just want to share a couple of
[57:57] Um, I just want to share a couple of slides uh in the session on the work
[57:59] slides uh in the session on the work that open light has done in terms of
[58:01] that open light has done in terms of bringing hetroenious integration to the
[58:02] bringing hetroenious integration to the market and explain a bit about some of
[58:04] market and explain a bit about some of the applications uh that we're seeing uh
[58:07] the applications uh that we're seeing uh uh this technology used in. [gasps]
[58:10] uh this technology used in. [gasps] Uh my slide removed. Thank you. Um so uh
[58:14] Uh my slide removed. Thank you. Um so uh in terms of you know who open light is
[58:16] in terms of you know who open light is and what we do. So open light is a
[58:17] and what we do. So open light is a company uh that was formed uh in 2022.
[58:21] company uh that was formed uh in 2022. We spun out from Juniper Networks at
[58:22] We spun out from Juniper Networks at that point. And our mission in life
[58:24] that point. And our mission in life really is to bring heterogeneous
[58:25] really is to bring heterogeneous integration which to our eyes is the
[58:28] integration which to our eyes is the integration of 35 materials and indeed
[58:30] integration of 35 materials and indeed others onto a silicon photonix platform
[58:32] others onto a silicon photonix platform uh to the outside world from there as
[58:34] uh to the outside world from there as well. And we've done that in conjunction
[58:36] well. And we've done that in conjunction with our partner Tower Semiconductor by
[58:38] with our partner Tower Semiconductor by creating a PDK that customers can design
[58:41] creating a PDK that customers can design into with a library of different
[58:43] into with a library of different components for different applications.
[58:45] components for different applications. A lot of our PDK uh is very much focused
[58:48] A lot of our PDK uh is very much focused towards dataccom, but we do have other
[58:50] towards dataccom, but we do have other devices in different bands uh for
[58:52] devices in different bands uh for different applications from there. So
[58:54] different applications from there. So probably pretty probably about 50% of
[58:55] probably pretty probably about 50% of our engineering time um is done to
[58:57] our engineering time um is done to develop components that are added into
[58:59] develop components that are added into that PDK that customers can design into
[59:02] that PDK that customers can design into on their own um or in fact uh with the
[59:04] on their own um or in fact uh with the assistance of ourselves or some of our
[59:07] assistance of ourselves or some of our design partners from there as well. But
[59:09] design partners from there as well. But on top of what is in the PDK, we do have
[59:12] on top of what is in the PDK, we do have a library of devices that are either
[59:13] a library of devices that are either coming into it and aren't fully mature
[59:15] coming into it and aren't fully mature ready for for for true public
[59:17] ready for for for true public consumption or some things that we could
[59:19] consumption or some things that we could consider uh to be licensed designs as
[59:22] consider uh to be licensed designs as well. And one thing that we've done over
[59:23] well. And one thing that we've done over the last couple of years as we've worked
[59:25] the last couple of years as we've worked with the companies, we've come to
[59:26] with the companies, we've come to understand that the ecosystem is
[59:28] understand that the ecosystem is incredibly important from this as well
[59:30] incredibly important from this as well that our ability to uh generate a PDK is
[59:33] that our ability to uh generate a PDK is not really uh uh uh a successful market
[59:36] not really uh uh uh a successful market solution unless we have the great
[59:38] solution unless we have the great support of not only you know the Foundry
[59:40] support of not only you know the Foundry partners uh but the infeed material uh
[59:43] partners uh but the infeed material uh from the Indian phosphide fabs uh
[59:45] from the Indian phosphide fabs uh support in the EDA tools where we have
[59:47] support in the EDA tools where we have great support from the community
[59:48] great support from the community including Lucida, GDS factory and
[59:50] including Lucida, GDS factory and synopsis But one thing we've also
[59:52] synopsis But one thing we've also started to do is increase the uh
[59:54] started to do is increase the uh engineering offering to enable things
[59:56] engineering offering to enable things like wafer level test and backend
[59:58] like wafer level test and backend services as well. Um so either in
[01:00:01] services as well. Um so either in conjunction with open light or or
[01:00:03] conjunction with open light or or independently uh customers are able to
[01:00:05] independently uh customers are able to use this technology to create their
[01:00:07] use this technology to create their designs uh as is needed
[01:00:11] designs uh as is needed in terms of what we believe
[01:00:12] in terms of what we believe heterogeneous integration is for from
[01:00:13] heterogeneous integration is for from our perspective. It really is as I
[01:00:15] our perspective. It really is as I mentioned wafer scale integration of 35
[01:00:17] mentioned wafer scale integration of 35 devices on two indium phospho silicon
[01:00:20] devices on two indium phospho silicon platforms from there as well. Uh the way
[01:00:22] platforms from there as well. Uh the way it works is that we take wafers uh at
[01:00:25] it works is that we take wafers uh at the tower fab uh we pattern silicon and
[01:00:28] the tower fab uh we pattern silicon and silicon nitride waveguides and then we
[01:00:30] silicon nitride waveguides and then we selectively prep areas for bonding um of
[01:00:33] selectively prep areas for bonding um of bulk epimeaterial from there as well.
[01:00:36] bulk epimeaterial from there as well. Now it's important to note that these
[01:00:37] Now it's important to note that these are not devices at these stage. This is
[01:00:39] are not devices at these stage. This is not a very accurate placement and we
[01:00:42] not a very accurate placement and we would typically do one placement of a
[01:00:43] would typically do one placement of a different type of material which will
[01:00:45] different type of material which will eventually become multiple devices as it
[01:00:47] eventually become multiple devices as it passes through the process chain. What
[01:00:49] passes through the process chain. What this enables is basically a very high
[01:00:51] this enables is basically a very high throughput of bonding and very high
[01:00:52] throughput of bonding and very high yield as as as mentioned by the previous
[01:00:55] yield as as as mentioned by the previous speaker. Um if your yield is not high uh
[01:00:58] speaker. Um if your yield is not high uh then you really the economics of of this
[01:01:00] then you really the economics of of this technology you know do do not work and a
[01:01:03] technology you know do do not work and a lot of the uh the effort the tower
[01:01:05] lot of the uh the effort the tower especially and our bonding partners have
[01:01:06] especially and our bonding partners have made is to make sure that this step is
[01:01:08] made is to make sure that this step is is very very high 90s in terms of uh a
[01:01:13] is very very high 90s in terms of uh a yield from there as well. Once those
[01:01:16] yield from there as well. Once those triplets of different types of Indian
[01:01:18] triplets of different types of Indian phosphide material are bonded onto the
[01:01:19] phosphide material are bonded onto the silicon uh the substrate is removed from
[01:01:22] silicon uh the substrate is removed from that to make it easier to process in a
[01:01:23] that to make it easier to process in a foundry environment. And then you we are
[01:01:27] foundry environment. And then you we are processing both the Indian phosphide uh
[01:01:29] processing both the Indian phosphide uh to put waveguides uh tapers electrical
[01:01:32] to put waveguides uh tapers electrical contacts onto those devices uh without
[01:01:35] contacts onto those devices uh without impeding the and destroying the
[01:01:37] impeding the and destroying the waveguides underneath that from there as
[01:01:38] waveguides underneath that from there as well. And this is really when they start
[01:01:40] well. And this is really when they start to become devices. And because this is
[01:01:42] to become devices. And because this is done at wafer scale using
[01:01:44] done at wafer scale using photoiththography
[01:01:45] photoiththography uh the actual alignment is is basically
[01:01:47] uh the actual alignment is is basically as tight as the photo lithography allows
[01:01:50] as tight as the photo lithography allows you from there as well. And everything
[01:01:51] you from there as well. And everything is done at the wafer scale as well. So
[01:01:53] is done at the wafer scale as well. So it typically will be creating thousands
[01:01:55] it typically will be creating thousands of devices uh in a couple of different
[01:01:57] of devices uh in a couple of different steps.
[01:01:59] steps. Once those devices are created on top of
[01:02:01] Once those devices are created on top of the silicon, they're then encapsulated
[01:02:02] the silicon, they're then encapsulated in a thick layer of silicon oxide with a
[01:02:04] in a thick layer of silicon oxide with a metallization on top layer on top uh for
[01:02:07] metallization on top layer on top uh for either contacting out to wire bond pads
[01:02:09] either contacting out to wire bond pads um or for
[01:02:12] um or for copper pillars for flip chip assembly
[01:02:13] copper pillars for flip chip assembly from there as well. And what you get at
[01:02:15] from there as well. And what you get at the end of the process is a wafer that
[01:02:17] the end of the process is a wafer that looks and smells and tastes like any
[01:02:18] looks and smells and tastes like any other silicon ponic wafer. So it can go
[01:02:21] other silicon ponic wafer. So it can go into a standard OSAP flow from there as
[01:02:23] into a standard OSAP flow from there as well for test, for burn-in, for
[01:02:25] well for test, for burn-in, for singulation, dicing, backgrounding and
[01:02:27] singulation, dicing, backgrounding and polishing without any need for special
[01:02:29] polishing without any need for special handling just because it's got Indian
[01:02:30] handling just because it's got Indian phosphate material in there. Um, so why
[01:02:33] phosphate material in there. Um, so why do we do this? Well, I mean obviously
[01:02:35] do we do this? Well, I mean obviously there is an economic argument for doing
[01:02:37] there is an economic argument for doing this as well. It allows customers to
[01:02:39] this as well. It allows customers to become truly vertically integrated and
[01:02:41] become truly vertically integrated and to buy a single skew direct from the
[01:02:43] to buy a single skew direct from the foundry gates with no markup uh to get
[01:02:45] foundry gates with no markup uh to get the lowest possible cost from there. But
[01:02:47] the lowest possible cost from there. But technically as as previous mentioned
[01:02:49] technically as as previous mentioned different material systems have
[01:02:50] different material systems have different advantages and by pushing
[01:02:52] different advantages and by pushing those together with very high coupling
[01:02:54] those together with very high coupling efficiency between the layers uh that
[01:02:56] efficiency between the layers uh that allows us to get to a point where we can
[01:02:58] allows us to get to a point where we can have very low laser output power uh very
[01:03:00] have very low laser output power uh very low current uh and we can have a
[01:03:03] low current uh and we can have a different classes of modulators uh uh
[01:03:05] different classes of modulators uh uh that compared to a pure silicon type
[01:03:07] that compared to a pure silicon type processors. Um it is very scalable in
[01:03:10] processors. Um it is very scalable in terms of production in terms of test
[01:03:12] terms of production in terms of test from there as well as well. And just as
[01:03:14] from there as well as well. And just as an example of some of the devices that
[01:03:15] an example of some of the devices that you can make on this in the bottom left
[01:03:17] you can make on this in the bottom left hand side is a DFB laser. Uh this is
[01:03:19] hand side is a DFB laser. Uh this is done by writing the grating into the
[01:03:21] done by writing the grating into the silicon layer and then bonding an Indian
[01:03:23] silicon layer and then bonding an Indian phosphide gate medium on top of that. Um
[01:03:25] phosphide gate medium on top of that. Um other laser structures are possible. Uh
[01:03:27] other laser structures are possible. Uh the one in the middle is a fullband
[01:03:28] the one in the middle is a fullband tunable laser that we offer primarily
[01:03:30] tunable laser that we offer primarily now for sensing applications and telecom
[01:03:33] now for sensing applications and telecom in both O and C band. Uh and you can
[01:03:35] in both O and C band. Uh and you can also make some very high power SOAs from
[01:03:37] also make some very high power SOAs from this as well. We have a road map up to
[01:03:39] this as well. We have a road map up to 20 dBm in both O and C band to support
[01:03:42] 20 dBm in both O and C band to support different applications.
[01:03:45] In terms of the contents of the design
[01:03:47] In terms of the contents of the design kit, it is not just active devices. It
[01:03:49] kit, it is not just active devices. It does include all of the passive devices
[01:03:51] does include all of the passive devices as well. And that includes ways of
[01:03:52] as well. And that includes ways of getting light on and off chip using
[01:03:54] getting light on and off chip using either grating couplers or spot size
[01:03:56] either grating couplers or spot size converters with different mode field
[01:03:57] converters with different mode field diameters. uh waveguides uh in both
[01:04:00] diameters. uh waveguides uh in both silicon and silicon nitride and
[01:04:01] silicon and silicon nitride and transitions between that uh and passive
[01:04:04] transitions between that uh and passive components including directional
[01:04:05] components including directional couplers, MMIs, BMIs, polarization
[01:04:08] couplers, MMIs, BMIs, polarization components uh control components uh and
[01:04:11] components uh control components uh and resonance structures based on ring
[01:04:13] resonance structures based on ring maxender and AWG and there's also design
[01:04:15] maxender and AWG and there's also design guides as well so that customers can
[01:04:17] guides as well so that customers can build their own passive devices but on
[01:04:20] build their own passive devices but on top of that there is a library of active
[01:04:21] top of that there is a library of active devices that are even represented as
[01:04:23] devices that are even represented as cells or if your EDA tool can handle it
[01:04:26] cells or if your EDA tool can handle it in some instances P cells as Well, so by
[01:04:29] in some instances P cells as Well, so by changing the physical properties of the
[01:04:30] changing the physical properties of the device, you can change for instance the
[01:04:32] device, you can change for instance the gain or the output power of an S SOA. Uh
[01:04:34] gain or the output power of an S SOA. Uh and that's reflected in the model that
[01:04:36] and that's reflected in the model that goes into those tools from there as
[01:04:37] goes into those tools from there as well. Uh the current list of devices
[01:04:40] well. Uh the current list of devices that we support in terms of laser
[01:04:42] that we support in terms of laser sources as I mentioned is the tunable
[01:04:43] sources as I mentioned is the tunable and DFB lasers from there as well. Uh
[01:04:46] and DFB lasers from there as well. Uh DFB lasers are currently targeted for
[01:04:48] DFB lasers are currently targeted for dataccom applications. Uh so are
[01:04:50] dataccom applications. Uh so are primarily available in Oband and CWDM.
[01:04:52] primarily available in Oband and CWDM. But we are starting to do custom designs
[01:04:54] But we are starting to do custom designs now uh with either DWDM spacings or
[01:04:56] now uh with either DWDM spacings or different wavelength bands uh depending
[01:04:58] different wavelength bands uh depending on a particular application needs. Both
[01:05:02] on a particular application needs. Both of those use lasers use an S SOA is the
[01:05:04] of those use lasers use an S SOA is the heart of the device as well and we offer
[01:05:06] heart of the device as well and we offer that as a standalone device with a
[01:05:08] that as a standalone device with a parameterized cell uh again in both O
[01:05:11] parameterized cell uh again in both O and C band for different applications uh
[01:05:13] and C band for different applications uh with different output powers and
[01:05:14] with different output powers and different gains uh for depending on on
[01:05:17] different gains uh for depending on on the circuit needs. And then for
[01:05:19] the circuit needs. And then for modulation, we have very much
[01:05:20] modulation, we have very much concentrated on electroabsorption
[01:05:22] concentrated on electroabsorption modulator rather than a maxender. Uh we
[01:05:24] modulator rather than a maxender. Uh we like the electroabsorption modulator at
[01:05:26] like the electroabsorption modulator at lower speeds because it could be driven
[01:05:28] lower speeds because it could be driven single-ended and look to the outside
[01:05:29] single-ended and look to the outside electronic ecosystem as as an EML. And
[01:05:32] electronic ecosystem as as an EML. And now we're getting to higher speeds uh by
[01:05:34] now we're getting to higher speeds uh by making the device shorter uh and uh
[01:05:37] making the device shorter uh and uh reducing the electrical parasitics.
[01:05:39] reducing the electrical parasitics. We're unable to extend the bandwidth of
[01:05:41] We're unable to extend the bandwidth of that into the 100 GHz range, making it
[01:05:44] that into the 100 GHz range, making it suitable for 400 gig PAM 4. On the
[01:05:47] suitable for 400 gig PAM 4. On the receive side, uh again we have photo
[01:05:49] receive side, uh again we have photo diodes based on Indian phosphide uh
[01:05:51] diodes based on Indian phosphide uh highspeed uh using a dedicated EP uh
[01:05:54] highspeed uh using a dedicated EP uh again with similar sort of bandwidths to
[01:05:56] again with similar sort of bandwidths to the modulator. Uh and then we monitor
[01:05:59] the modulator. Uh and then we monitor photo diodes uh using uh uh side corners
[01:06:02] photo diodes uh using uh uh side corners of gain EP primarily uh to prevent you
[01:06:04] of gain EP primarily uh to prevent you needing to put a specific uh uh
[01:06:06] needing to put a specific uh uh deposition down uh just to create a
[01:06:08] deposition down uh just to create a monitor photo diode from there as well.
[01:06:13] So here's a couple of examples of
[01:06:15] So here's a couple of examples of different devices that you can make from
[01:06:16] different devices that you can make from this. Uh these are what we call
[01:06:18] this. Uh these are what we call reference designs. So these are things
[01:06:19] reference designs. So these are things that we have done using our own
[01:06:21] that we have done using our own technology to either prove it out uh and
[01:06:23] technology to either prove it out uh and to show that it works uh but also to be
[01:06:25] to show that it works uh but also to be able to supply this as either samples or
[01:06:27] able to supply this as either samples or GDS examples uh to customers in the
[01:06:30] GDS examples uh to customers in the pluggable module type environment. Um uh
[01:06:33] pluggable module type environment. Um uh we're currently sampling uh to select
[01:06:35] we're currently sampling uh to select customers uh the uh DR8 version uh the
[01:06:38] customers uh the uh DR8 version uh the 2x FR4 version of that using the CWDM
[01:06:41] 2x FR4 version of that using the CWDM laser is in FAB. We also have some early
[01:06:43] laser is in FAB. We also have some early samples uh of a 3.2 TB version uh which
[01:06:47] samples uh of a 3.2 TB version uh which is very similar device from there as
[01:06:49] is very similar device from there as well uh just taking advantage of our
[01:06:51] well uh just taking advantage of our ability to increase the modulator
[01:06:53] ability to increase the modulator bandwidth uh without changing any of the
[01:06:55] bandwidth uh without changing any of the other designs within the chips from
[01:06:57] other designs within the chips from there as well. Um all of these devices
[01:07:00] there as well. Um all of these devices use a number of lasers typically four uh
[01:07:03] use a number of lasers typically four uh as EAMS and S SOAS uh in the instance
[01:07:07] as EAMS and S SOAS uh in the instance when the path loss uh is too high to
[01:07:10] when the path loss uh is too high to meet RMA targets from there. They're all
[01:07:12] meet RMA targets from there. They're all broadly the same size. They're about 5x7
[01:07:14] broadly the same size. They're about 5x7 mm in dimension and typically consume
[01:07:17] mm in dimension and typically consume between two or three watts total power.
[01:07:19] between two or three watts total power. And this is a device that can be flip
[01:07:21] And this is a device that can be flip chipped onto either a uh a substrate or
[01:07:24] chipped onto either a uh a substrate or a substrate like PCBA as you can see in
[01:07:26] a substrate like PCBA as you can see in the top right uh illustration which is
[01:07:29] the top right uh illustration which is an EVB that we put together for for
[01:07:31] an EVB that we put together for for demonstrations in conjunction with a
[01:07:32] demonstrations in conjunction with a driver from from Raycom sorry SEC in
[01:07:35] driver from from Raycom sorry SEC in this instance um and and that's a
[01:07:38] this instance um and and that's a complete uh integrated transmit solution
[01:07:41] complete uh integrated transmit solution from there as well and you can see that
[01:07:43] from there as well and you can see that the quality of the eye and this is an
[01:07:44] the quality of the eye and this is an LPO type configuration in conjunction
[01:07:46] LPO type configuration in conjunction with a max linear DSB board as you can
[01:07:48] with a max linear DSB board as you can see the setup in in the middle uh is
[01:07:50] see the setup in in the middle uh is nice and clean and open and we can
[01:07:51] nice and clean and open and we can happily exceed our OMA uh targets uh
[01:07:55] happily exceed our OMA uh targets uh with this technology from there as well.
[01:07:59] with this technology from there as well. And just to focus in really one of the
[01:08:00] And just to focus in really one of the advantages now as we moved from as the
[01:08:02] advantages now as we moved from as the previous speaker said from uh uh
[01:08:05] previous speaker said from uh uh pluggable to the CPO type format um and
[01:08:07] pluggable to the CPO type format um and that's you know the ability of the
[01:08:09] that's you know the ability of the modulator espec with uh two scaled up to
[01:08:11] modulator espec with uh two scaled up to the 400 gig demonstration. Um, as of
[01:08:14] the 400 gig demonstration. Um, as of last year, we demonstrated the first 400
[01:08:16] last year, we demonstrated the first 400 gig eye, which you can see on the right
[01:08:18] gig eye, which you can see on the right with a modulator bandwidth that had just
[01:08:20] with a modulator bandwidth that had just around 85 gigahertz S21 parameter. Uh,
[01:08:23] around 85 gigahertz S21 parameter. Uh, we're now working on devices uh, and we
[01:08:26] we're now working on devices uh, and we have samples uh, in excess of 100 GHz.
[01:08:29] have samples uh, in excess of 100 GHz. Uh, and we have a road map to increase
[01:08:31] Uh, and we have a road map to increase that uh, up to 110 and this is a very
[01:08:34] that uh, up to 110 and this is a very very short device. The intrinsic device
[01:08:35] very short device. The intrinsic device length is less than 100 microns long
[01:08:38] length is less than 100 microns long from there as well. which means some of
[01:08:40] from there as well. which means some of the challenges that we're seeing in CPO
[01:08:42] the challenges that we're seeing in CPO for instance in terms of just power
[01:08:43] for instance in terms of just power consumption and physical size of devices
[01:08:45] consumption and physical size of devices with other material structures don't
[01:08:47] with other material structures don't really apply to Indian phosphide from
[01:08:49] really apply to Indian phosphide from there as well
[01:08:51] there as well and then finally although AI is of
[01:08:53] and then finally although AI is of course the hot topic there are other
[01:08:54] course the hot topic there are other ways you things you can build out of
[01:08:56] ways you things you can build out of silicon and the other uh uh area that
[01:08:58] silicon and the other uh uh area that we've looked at uh and have partners
[01:09:00] we've looked at uh and have partners working on um is FMCW um so FMCW lidar
[01:09:04] working on um is FMCW um so FMCW lidar is a great technical application for
[01:09:05] is a great technical application for both automotive and other uh
[01:09:07] both automotive and other uh applications it has great range
[01:09:09] applications it has great range uh the tolerance to interference uh is
[01:09:12] uh the tolerance to interference uh is uh uh compared to time of flight or
[01:09:14] uh uh compared to time of flight or flash solutions uh is much greater. Uh
[01:09:17] flash solutions uh is much greater. Uh and uh one of the side benefits is is
[01:09:19] and uh one of the side benefits is is with FMCW because you can use the phase
[01:09:21] with FMCW because you can use the phase properties of the light you can
[01:09:22] properties of the light you can significantly reduce the amount of of
[01:09:24] significantly reduce the amount of of digital processing that you can do after
[01:09:26] digital processing that you can do after that. Uh but FMCW requires things like
[01:09:29] that. Uh but FMCW requires things like very narrow line width tunable fast
[01:09:31] very narrow line width tunable fast tuned lasers from there as well and
[01:09:33] tuned lasers from there as well and that's the devices that we can build in
[01:09:35] that's the devices that we can build in silicon from there as well. And then by
[01:09:37] silicon from there as well. And then by combining that laser either with or
[01:09:39] combining that laser either with or without an isolator and then a splisher
[01:09:41] without an isolator and then a splisher array uh and an array of S SOAS with the
[01:09:44] array uh and an array of S SOAS with the PBSRS you can create something like a
[01:09:47] PBSRS you can create something like a single chip FMCW type sensor. So the
[01:09:50] single chip FMCW type sensor. So the example here in the top right is
[01:09:51] example here in the top right is something that a partner has built. Uh
[01:09:53] something that a partner has built. Uh it's a 10x10 complete uh FMCW lidar
[01:09:56] it's a 10x10 complete uh FMCW lidar chip. Uh this actually has a laser
[01:09:57] chip. Uh this actually has a laser onboard as well from there as well or
[01:09:59] onboard as well from there as well or although other uh uh uh solutions use
[01:10:02] although other uh uh uh solutions use offboard lasers from there as well. And
[01:10:04] offboard lasers from there as well. And similar to the data compix, this is
[01:10:06] similar to the data compix, this is something that you can be flip chipped
[01:10:07] something that you can be flip chipped or wire bonded into a package as well.
[01:10:09] or wire bonded into a package as well. And it's a very compact uh uh thermally
[01:10:12] And it's a very compact uh uh thermally stable uh platform uh for a high volume
[01:10:15] stable uh platform uh for a high volume application.
[01:10:17] application. Um so hopefully that's given you an idea
[01:10:19] Um so hopefully that's given you an idea of what the offering and some of the
[01:10:20] of what the offering and some of the applications we are and
[01:10:22] applications we are and >> thank you so much. That was a great
[01:10:23] >> thank you so much. That was a great presentation. Very well summary. You are
[01:10:25] presentation. Very well summary. You are you actually kept yourself in time for
[01:10:27] you actually kept yourself in time for which I appreciate very much. But Steve
[01:10:30] which I appreciate very much. But Steve here there is a question for for you and
[01:10:32] here there is a question for for you and for Matt H which is very clear. All of
[01:10:35] for Matt H which is very clear. All of you presented the integration of light
[01:10:37] you presented the integration of light sources into silicon. Steve you
[01:10:39] sources into silicon. Steve you presented the DFV lasers and Matt
[01:10:42] presented the DFV lasers and Matt presented vixels. Uh so can we clearly
[01:10:45] presented vixels. Uh so can we clearly tell the customer what is the short and
[01:10:48] tell the customer what is the short and turn solution for bringing heterogeneous
[01:10:50] turn solution for bringing heterogeneous integration of 35 into silicon for light
[01:10:53] integration of 35 into silicon for light sources.
[01:10:56] sources. Steve go ahead.
[01:10:58] Steve go ahead. I'm sorry, Jose. Could you repeat the
[01:11:00] I'm sorry, Jose. Could you repeat the question, please?
[01:11:01] question, please? >> DFVS or Vixels, what is the best
[01:11:03] >> DFVS or Vixels, what is the best solution?
[01:11:04] solution? >> Uh, it really depends on the
[01:11:05] >> Uh, it really depends on the application. Uh, uh, I mean, as Matt
[01:11:08] application. Uh, uh, I mean, as Matt said, I mean I mean for for for for
[01:11:10] said, I mean I mean for for for for facial recognition, vixels are a great
[01:11:12] facial recognition, vixels are a great solution. Uh, not so great for 2 km
[01:11:14] solution. Uh, not so great for 2 km dataccom. Um, vixels of course work very
[01:11:16] dataccom. Um, vixels of course work very well for short reach. Uh, DFB lasers are
[01:11:19] well for short reach. Uh, DFB lasers are for longer reach and then for other
[01:11:20] for longer reach and then for other applications, narrowand high quality
[01:11:22] applications, narrowand high quality tunables are there as well. So the
[01:11:24] tunables are there as well. So the intent of the technology platform uh is
[01:11:27] intent of the technology platform uh is to offer a platform where different
[01:11:28] to offer a platform where different devices can be integrated into that.
[01:11:31] devices can be integrated into that. >> Yeah, I I agree completely with Steve.
[01:11:33] >> Yeah, I I agree completely with Steve. Uh different applications have different
[01:11:35] Uh different applications have different requirements and different laser
[01:11:36] requirements and different laser technologies suit those suit those best.
[01:11:40] technologies suit those suit those best. Um so there's no uh onesizefits-all
[01:11:42] Um so there's no uh onesizefits-all answer. Um it's about having a platform
[01:11:44] answer. Um it's about having a platform that's flexible that allows you to
[01:11:46] that's flexible that allows you to integrate the ingredients um as you need
[01:11:48] integrate the ingredients um as you need to to get a product that meets your
[01:11:50] to to get a product that meets your performance requirements.
[01:11:52] performance requirements. We all want to have uh key applications
[01:11:55] We all want to have uh key applications for each technology. But in my opinion,
[01:11:57] for each technology. But in my opinion, right now we have a huge challenge to
[01:11:59] right now we have a huge challenge to del the industry which is can we remove
[01:12:02] del the industry which is can we remove the external laser for CPO modules or do
[01:12:06] the external laser for CPO modules or do we have to? I would like to ask Doron
[01:12:08] we have to? I would like to ask Doron Tal from Neopotonics. Doron we are
[01:12:11] Tal from Neopotonics. Doron we are talking now you are pushing also near
[01:12:13] talking now you are pushing also near package optics NPO with external laser.
[01:12:16] package optics NPO with external laser. In your opinion, is this a shortterm
[01:12:19] In your opinion, is this a shortterm intermediary solution until companies
[01:12:22] intermediary solution until companies can provide a vixelbased CPO or
[01:12:24] can provide a vixelbased CPO or heterogeneous integration of a laser
[01:12:26] heterogeneous integration of a laser power fing for CPO or is the medium or
[01:12:29] power fing for CPO or is the medium or long-term solution already NPO?
[01:12:32] long-term solution already NPO? >> Yeah, I think NPO is starting now. Um,
[01:12:35] >> Yeah, I think NPO is starting now. Um, it's just a more pragmatic solution
[01:12:37] it's just a more pragmatic solution because people are looking for
[01:12:39] because people are looking for serviceability around uh laser
[01:12:41] serviceability around uh laser integration.
[01:12:42] integration. Um I think that um heterogeneously
[01:12:46] Um I think that um heterogeneously integrated
[01:12:47] integrated uh integrated circuits are ideal for NPO
[01:12:51] uh integrated circuits are ideal for NPO because they offer that serviceability
[01:12:53] because they offer that serviceability angle. um maybe not at the density at
[01:12:56] angle. um maybe not at the density at the high density that uh and uh CPO's
[01:13:00] the high density that uh and uh CPO's demanding but definitely uh near-term
[01:13:03] demanding but definitely uh near-term and midterm and perhaps even longer term
[01:13:06] and midterm and perhaps even longer term because we can evolve the um
[01:13:09] because we can evolve the um heterogeneous integration which is you
[01:13:12] heterogeneous integration which is you know all all the speakers here offered
[01:13:14] know all all the speakers here offered fantastic solutions um and this is
[01:13:18] fantastic solutions um and this is really the beginning of the cycle. So I
[01:13:20] really the beginning of the cycle. So I imagine with innovation and tighter
[01:13:24] imagine with innovation and tighter integration we can make near package
[01:13:26] integration we can make near package optics um last probably a lot longer
[01:13:29] optics um last probably a lot longer than we envision.
[01:13:31] than we envision. Congratulations on what you are
[01:13:33] Congratulations on what you are achieving with new photonics. I can't
[01:13:35] achieving with new photonics. I can't wait your next milestone at OFC. But
[01:13:37] wait your next milestone at OFC. But stay tuned because the next speaker
[01:13:39] stay tuned because the next speaker provides a technology that is very
[01:13:41] provides a technology that is very relevant to what was just presented by
[01:13:44] relevant to what was just presented by the previous speakers. We are talking
[01:13:46] the previous speakers. We are talking about photon bridge and the integration
[01:13:48] about photon bridge and the integration of fing for silicon in a very different
[01:13:51] of fing for silicon in a very different manner with passive alignment. The
[01:13:52] manner with passive alignment. The definition of pure semic oriented
[01:13:54] definition of pure semic oriented hetrogenous integration. Thank you very
[01:13:56] hetrogenous integration. Thank you very much John And here all the beautiful.
[01:13:59] much John And here all the beautiful. The floor is yours.
[01:14:04] We cannot hear very well.
[01:14:06] We cannot hear very well. >> I was too slow with the mute button.
[01:14:08] >> I was too slow with the mute button. Thank you very much for the wonderful
[01:14:09] Thank you very much for the wonderful introduction Jose and John. Thanks for
[01:14:11] introduction Jose and John. Thanks for arranging and inviting us. So very happy
[01:14:14] arranging and inviting us. So very happy to be here share an update on what's
[01:14:16] to be here share an update on what's happening with Photon Bridge. Maybe if
[01:14:18] happening with Photon Bridge. Maybe if you've not heard from Photon Bridge
[01:14:19] you've not heard from Photon Bridge before or know who we are, just let you
[01:14:21] before or know who we are, just let you know who we are. It's a relatively short
[01:14:23] know who we are. It's a relatively short presentation today. Um so this is going
[01:14:25] presentation today. Um so this is going to be the 30,000 foot view. Hopefully
[01:14:28] to be the 30,000 foot view. Hopefully there's lots of questions. Hopefully
[01:14:29] there's lots of questions. Hopefully there's not enough time, but my contact
[01:14:31] there's not enough time, but my contact details are there at the end. Please
[01:14:33] details are there at the end. Please reach out, challenge, ask questions, and
[01:14:35] reach out, challenge, ask questions, and I'll I'll give a shout out at the end
[01:14:37] I'll I'll give a shout out at the end for where we would love to hear help and
[01:14:38] for where we would love to hear help and support from the from the Optica
[01:14:41] support from the from the Optica community. So what am I going to talk
[01:14:42] community. So what am I going to talk about in the couple of minutes that I
[01:14:44] about in the couple of minutes that I have? Bigger is better for the AI
[01:14:45] have? Bigger is better for the AI nervous system or why Photon Bridge is
[01:14:47] nervous system or why Photon Bridge is the heterogeneous platform built for
[01:14:49] the heterogeneous platform built for manufacturer. That speaks to where we
[01:14:51] manufacturer. That speaks to where we see our value in the market, how we see
[01:14:53] see our value in the market, how we see the market evolving, where our focus
[01:14:55] the market evolving, where our focus areas are and what our heterogeneous
[01:14:57] areas are and what our heterogeneous platform is all about. Hopefully I can
[01:14:59] platform is all about. Hopefully I can do this in the time available.
[01:15:02] do this in the time available. So AI drives bifocation. This is kind of
[01:15:06] So AI drives bifocation. This is kind of how we see how the market evolving. We
[01:15:08] how we see how the market evolving. We think CPO is happening
[01:15:10] think CPO is happening and we think CPO is really about
[01:15:12] and we think CPO is really about connecting the brain of AI
[01:15:15] connecting the brain of AI and this is all about density as several
[01:15:17] and this is all about density as several people in the preceding presentations
[01:15:20] people in the preceding presentations have talked to so I won't go on this
[01:15:21] have talked to so I won't go on this again it's all about density and high
[01:15:23] again it's all about density and high bandwidth density but you need the
[01:15:25] bandwidth density but you need the nervous system you need to connect to
[01:15:27] nervous system you need to connect to the physical world and this requires
[01:15:29] the physical world and this requires very different things from integrated
[01:15:30] very different things from integrated photonics over the short term over the
[01:15:33] photonics over the short term over the next three to five years we believe it's
[01:15:35] next three to five years we believe it's going to be all about external light
[01:15:36] going to be all about external light sources Our platform scales to
[01:15:38] sources Our platform scales to integrated light sources next to CPO.
[01:15:40] integrated light sources next to CPO. But in the short term, three to five
[01:15:42] But in the short term, three to five years, this is all about connecting to
[01:15:44] years, this is all about connecting to the physical providing light sources for
[01:15:47] the physical providing light sources for infrastructure for powering the
[01:15:49] infrastructure for powering the connected brains, the CPO engines. But
[01:15:52] connected brains, the CPO engines. But it also scales to FMCW, LAR, and all
[01:15:54] it also scales to FMCW, LAR, and all sorts of things. But this is more about
[01:15:56] sorts of things. But this is more about density versus power. And where we focus
[01:15:58] density versus power. And where we focus and where our platform really shines is
[01:16:01] and where our platform really shines is on the right hand side.
[01:16:04] on the right hand side. And as mentioned, AI will be powered by
[01:16:06] And as mentioned, AI will be powered by light. We see the data rate going up. We
[01:16:08] light. We see the data rate going up. We see the transition to co-ackaged optics.
[01:16:10] see the transition to co-ackaged optics. And as Phippe from IMC mentioned, we
[01:16:12] And as Phippe from IMC mentioned, we believe there's going to be a transition
[01:16:13] believe there's going to be a transition to multiolor external light light
[01:16:16] to multiolor external light light sources. This is happening over the next
[01:16:17] sources. This is happening over the next two, three, four years, something like
[01:16:19] two, three, four years, something like that. We believe that the value for
[01:16:21] that. We believe that the value for heterogeneous integration is really how
[01:16:24] heterogeneous integration is really how you integrate all these lasers,
[01:16:25] you integrate all these lasers, integrate all the multiplexing of those
[01:16:27] integrate all the multiplexing of those lasers on silicon to provide it in a
[01:16:29] lasers on silicon to provide it in a scalable manufacturable way within the
[01:16:32] scalable manufacturable way within the same pluggable device or within the same
[01:16:34] same pluggable device or within the same uh serviceable and removable device.
[01:16:41] And to do this, what do you need? You
[01:16:43] And to do this, what do you need? You need a platform that's designed for high
[01:16:44] need a platform that's designed for high power, high yield, and low cost. And it
[01:16:49] power, high yield, and low cost. And it needs to be OSAC compatible
[01:16:50] needs to be OSAC compatible heterogeneous integration. Our platform
[01:16:53] heterogeneous integration. Our platform does this. It enables any device to be
[01:16:56] does this. It enables any device to be connected heterogeneously onto a piece
[01:16:58] connected heterogeneously onto a piece of silicon. So our process is material
[01:17:01] of silicon. So our process is material independent. We can support Indian
[01:17:02] independent. We can support Indian phosphide, thin film, lithium nabate.
[01:17:05] phosphide, thin film, lithium nabate. We're doing quantum well lasers today,
[01:17:06] We're doing quantum well lasers today, but it scales in our road map to quantum
[01:17:08] but it scales in our road map to quantum dots. It includes high-speed modulators,
[01:17:10] dots. It includes high-speed modulators, laser switches, isolators, everything.
[01:17:13] laser switches, isolators, everything. All can be combined like a Lego. All
[01:17:15] All can be combined like a Lego. All these things need to make manufacturable
[01:17:17] these things need to make manufacturable and to handle power. So if you look
[01:17:19] and to handle power. So if you look specifically at the multi
[01:17:21] specifically at the multi multi-wavelength light sources, one of
[01:17:23] multi-wavelength light sources, one of the things about our platform is that we
[01:17:25] the things about our platform is that we use large waveguides. They're two micron
[01:17:27] use large waveguides. They're two micron waveguides. This means they handle the
[01:17:28] waveguides. This means they handle the power and they are robust to process
[01:17:31] power and they are robust to process variation and tolerances. That means you
[01:17:33] variation and tolerances. That means you can build very high yielding
[01:17:34] can build very high yielding manufacturable AWGs and MXes. We of
[01:17:37] manufacturable AWGs and MXes. We of course support wafer level packaging.
[01:17:38] course support wafer level packaging. It's Dion wafer. The big thing about our
[01:17:41] It's Dion wafer. The big thing about our process is that it's OSAT compatible.
[01:17:45] process is that it's OSAT compatible. All the placement of the activives for
[01:17:47] All the placement of the activives for example the lasers are done passively
[01:17:50] example the lasers are done passively aligned. I'll come back to the next
[01:17:52] aligned. I'll come back to the next slide what that means but essentially we
[01:17:54] slide what that means but essentially we can resolve two to three micron
[01:17:56] can resolve two to three micron placement accuracy at a flip chip bonder
[01:17:59] placement accuracy at a flip chip bonder in an OSAT. So we don't believe flip
[01:18:01] in an OSAT. So we don't believe flip chip bonding is dead by the way. All of
[01:18:03] chip bonding is dead by the way. All of this is resolved to effectively 100
[01:18:05] this is resolved to effectively 100 nanometer optical alignment
[01:18:10] because it's wafer level packaging. We
[01:18:12] because it's wafer level packaging. We also have in our road map a path to
[01:18:14] also have in our road map a path to local hermiticity using no gold boxes.
[01:18:16] local hermiticity using no gold boxes. Now one of the things that we're acutely
[01:18:18] Now one of the things that we're acutely aware of any heterogeneous platform is
[01:18:21] aware of any heterogeneous platform is only valuable if you provide the
[01:18:23] only valuable if you provide the capability to get the light on and off
[01:18:25] capability to get the light on and off that platform. High throughput fiber
[01:18:28] that platform. High throughput fiber assembly is as important as the ability
[01:18:31] assembly is as important as the ability to combine Indian phosphide on silicon
[01:18:33] to combine Indian phosphide on silicon in a manufacturable scalable way. So as
[01:18:36] in a manufacturable scalable way. So as part of our roadmap we're developing a
[01:18:37] part of our roadmap we're developing a passive efficiently coupled to SMF uh
[01:18:40] passive efficiently coupled to SMF uh with integrated spot size conversion
[01:18:42] with integrated spot size conversion fiber attach capability and anticipating
[01:18:45] fiber attach capability and anticipating the question at the end John the optic
[01:18:48] the question at the end John the optic commun community optica community
[01:18:50] commun community optica community community how can we work together how
[01:18:52] community how can we work together how can we get the best value from everyone
[01:18:55] can we get the best value from everyone if you're working on fiber attach if
[01:18:56] if you're working on fiber attach if you're working on advanced packaging and
[01:18:58] you're working on advanced packaging and want to talk to us about how we can
[01:19:00] want to talk to us about how we can partner together to accelerate the
[01:19:02] partner together to accelerate the roadmap in terms of fiber attach we'd
[01:19:04] roadmap in terms of fiber attach we'd love to off to you.
[01:19:08] love to off to you. >> Thank you very much. And coming all the
[01:19:10] >> Thank you very much. And coming all the way from they provide with passive
[01:19:12] way from they provide with passive alignment they can integrate um
[01:19:14] alignment they can integrate um phosphide on silicon and that is great
[01:19:16] phosphide on silicon and that is great John but you for that you
[01:19:18] John but you for that you >> a trick for how it works but I see how
[01:19:20] >> a trick for how it works but I see how it works. Thanks for keeping me honest
[01:19:21] it works. Thanks for keeping me honest Jose.
[01:19:22] Jose. >> Thank you very much and that is exactly
[01:19:24] >> Thank you very much and that is exactly the secret here. We want to pro provide
[01:19:27] the secret here. We want to pro provide a passive alignment solution for indium
[01:19:29] a passive alignment solution for indium phosphate on silicon and look how the
[01:19:31] phosphate on silicon and look how the chip comes into the wake guide. To do
[01:19:33] chip comes into the wake guide. To do this, we need equipment. To do this, we
[01:19:35] this, we need equipment. To do this, we need equipment for passive alignment.
[01:19:38] need equipment for passive alignment. And here, John Andon, we brought you the
[01:19:40] And here, John Andon, we brought you the best one in the world. Let's go to
[01:19:42] best one in the world. Let's go to Berlin and meet Travis Scott from Fine
[01:19:44] Berlin and meet Travis Scott from Fine Techch. Tell us how your equipment can
[01:19:46] Techch. Tell us how your equipment can help Photon Bridge.
[01:19:50] >> Hello, Jose. Um,
[01:19:53] >> Hello, Jose. Um, sorry, I'm just trying to get to the
[01:19:55] sorry, I'm just trying to get to the screen sharing. Um, yeah, I'd just like
[01:19:57] screen sharing. Um, yeah, I'd just like to thank um Optica, Jose, and John for
[01:20:00] to thank um Optica, Jose, and John for putting this together. It's always an
[01:20:01] putting this together. It's always an extremely valuable meeting. it takes we
[01:20:03] extremely valuable meeting. it takes we take a lot of away from it and thank you
[01:20:06] take a lot of away from it and thank you for the invitation to speak. Um so at
[01:20:08] for the invitation to speak. Um so at fun at finec we manufacture flip chip uh
[01:20:12] fun at finec we manufacture flip chip uh die bonding equipment. We also offer
[01:20:14] die bonding equipment. We also offer tailored solutions for getting your
[01:20:16] tailored solutions for getting your prototype into production. Um, so we go
[01:20:20] prototype into production. Um, so we go all the way through technical technical
[01:20:22] all the way through technical technical consultancy, getting your packages built
[01:20:24] consultancy, getting your packages built in, uh, a demo lab or getting it into
[01:20:28] in, uh, a demo lab or getting it into systems to get it built um, uh, in the
[01:20:31] systems to get it built um, uh, in the equipment that we sell. And then we also
[01:20:33] equipment that we sell. And then we also ramp that up into production at OSATs at
[01:20:36] ramp that up into production at OSATs at offshore manufacturing uh, CMS and EMS
[01:20:38] offshore manufacturing uh, CMS and EMS facilities for example. Um, so our
[01:20:42] facilities for example. Um, so our philosophy at FineCH is to go from
[01:20:44] philosophy at FineCH is to go from prototype to production. So, our
[01:20:46] prototype to production. So, our equipment starts at tabletop flip chip
[01:20:48] equipment starts at tabletop flip chip die bonders all the way to fully
[01:20:50] die bonders all the way to fully automated uh production equipment.
[01:20:53] automated uh production equipment. Um, currently our most accurate system
[01:20:55] Um, currently our most accurate system is qualified for 0.3 micron at 3 sigma.
[01:20:58] is qualified for 0.3 micron at 3 sigma. Um, but we're trying to improve on this
[01:20:59] Um, but we're trying to improve on this and push it down. Like you said, we need
[01:21:01] and push it down. Like you said, we need to get to about sub 100 nanometer for a
[01:21:04] to get to about sub 100 nanometer for a lot of this um wafer level packaging
[01:21:06] lot of this um wafer level packaging equipment. We're also trying to bring a
[01:21:09] equipment. We're also trying to bring a system to market within the next one to
[01:21:11] system to market within the next one to one and a half years that's focused for
[01:21:13] one and a half years that's focused for chip to wafer bonding. Um, and we're
[01:21:17] chip to wafer bonding. Um, and we're trying to maintain this very modular
[01:21:20] trying to maintain this very modular um, die packaging approach that we've
[01:21:22] um, die packaging approach that we've always maintained at Fine Techch. And
[01:21:24] always maintained at Fine Techch. And this is where you can pack various
[01:21:26] this is where you can pack various different bonding technologies into a
[01:21:28] different bonding technologies into a single die assembly device. Um, here are
[01:21:31] single die assembly device. Um, here are some examples for some heterogeneous
[01:21:33] some examples for some heterogeneous applications we've seen over the past
[01:21:35] applications we've seen over the past few years.
[01:21:36] few years. um because of the difficulties with
[01:21:38] um because of the difficulties with heterogeneous integration 3D packaging
[01:21:40] heterogeneous integration 3D packaging vertical integration uh you're always
[01:21:43] vertical integration uh you're always fighting you know the refflow
[01:21:45] fighting you know the refflow temperature or the bonding temperature
[01:21:46] temperature or the bonding temperature of your previous bond layer uh so if
[01:21:49] of your previous bond layer uh so if you're stacking you know indium on gold
[01:21:52] you're stacking you know indium on gold uh gold on ultraviolet curing to get
[01:21:55] uh gold on ultraviolet curing to get various different functionalities from
[01:21:56] various different functionalities from these various different uh chips for the
[01:21:59] these various different uh chips for the heterogeneous integration uh we can
[01:22:01] heterogeneous integration uh we can offer all of these bonding technologies
[01:22:04] offer all of these bonding technologies inside of a extremely accurate It's
[01:22:06] inside of a extremely accurate It's passive flip chip die bonder. Um
[01:22:10] passive flip chip die bonder. Um yeah, so if you have any questions, if
[01:22:12] yeah, so if you have any questions, if you have any uh challenges with your
[01:22:14] you have any uh challenges with your packaging or your die bonding processes,
[01:22:17] packaging or your die bonding processes, please come and speak to us at Fine
[01:22:18] please come and speak to us at Fine Techch. So the question from Optica is
[01:22:20] Techch. So the question from Optica is always what can we do for you and what
[01:22:22] always what can we do for you and what can you do for us? Uh what you do for us
[01:22:25] can you do for us? Uh what you do for us is by coming to speak to us, showing us
[01:22:27] is by coming to speak to us, showing us these challenges. We take so much away
[01:22:28] these challenges. We take so much away from these technical discussions with
[01:22:30] from these technical discussions with customers because it shows us, you know,
[01:22:32] customers because it shows us, you know, the challenges that the market is
[01:22:34] the challenges that the market is facing. what we need to develop to get
[01:22:37] facing. what we need to develop to get uh you know in front of the market to
[01:22:39] uh you know in front of the market to keep up with what's happening uh in
[01:22:41] keep up with what's happening uh in production. Um and what we can do for
[01:22:43] production. Um and what we can do for you is offer 30 plus years of experience
[01:22:45] you is offer 30 plus years of experience in die bonding and die packaging um and
[01:22:48] in die bonding and die packaging um and a fully tailored custom solution for
[01:22:50] a fully tailored custom solution for whatever challenges you might be facing
[01:22:52] whatever challenges you might be facing with your packaging. Thank you very
[01:22:54] with your packaging. Thank you very much.
[01:22:56] much. >> Thank you so much Travis. Uh some great
[01:22:58] >> Thank you so much Travis. Uh some great work as usual. So this is a meeting
[01:23:01] work as usual. So this is a meeting about heterogeneous integration. So as
[01:23:03] about heterogeneous integration. So as more materials are introduced, we've
[01:23:05] more materials are introduced, we've heard about 35s, TFLN,
[01:23:08] heard about 35s, TFLN, BTO, and we will hear about metals. And
[01:23:10] BTO, and we will hear about metals. And in your experience, where's a thermal
[01:23:13] in your experience, where's a thermal mismatch and mechanical stress show up
[01:23:16] mismatch and mechanical stress show up first in the assembly process?
[01:23:19] first in the assembly process? >> Um, this is quite a difficult question
[01:23:21] >> Um, this is quite a difficult question with packaging because it bridges so
[01:23:25] with packaging because it bridges so many different parts of the the assembly
[01:23:26] many different parts of the the assembly process. like if you're bonding a laser
[01:23:29] process. like if you're bonding a laser to a carrier, if you're bonding a you
[01:23:31] to a carrier, if you're bonding a you know a large chip to an interposer to a
[01:23:34] know a large chip to an interposer to a PCB, um if this is done with indium or
[01:23:37] PCB, um if this is done with indium or gold, you know there's so many different
[01:23:39] gold, you know there's so many different variables that can come into it. So it's
[01:23:40] variables that can come into it. So it's a very difficult question to answer. Um
[01:23:43] a very difficult question to answer. Um but what we try to do is offer as many
[01:23:46] but what we try to do is offer as many different uh paths to fighting those
[01:23:49] different uh paths to fighting those challenges as possible. Uh so we offer
[01:23:51] challenges as possible. Uh so we offer for example like heating from the bottom
[01:23:53] for example like heating from the bottom as well as from the top. uh this heating
[01:23:55] as well as from the top. uh this heating from the top and the bottom can be done
[01:23:56] from the top and the bottom can be done at different temperatures, different
[01:23:58] at different temperatures, different ramp rates uh to try and you know match
[01:24:01] ramp rates uh to try and you know match those CTE mismatches in your bond
[01:24:04] those CTE mismatches in your bond layers. Um we've also integrated laser
[01:24:07] layers. Um we've also integrated laser bottom heating to get extremely fast
[01:24:09] bottom heating to get extremely fast ramp rates uh extremely high
[01:24:11] ramp rates uh extremely high temperatures um to move away from like
[01:24:13] temperatures um to move away from like traditional um resistance heaters that
[01:24:16] traditional um resistance heaters that take a very long time to heat up and
[01:24:18] take a very long time to heat up and cool down. Uh and this keeps your
[01:24:20] cool down. Uh and this keeps your thermal stress, you know, extremely
[01:24:21] thermal stress, you know, extremely localized around small chips.
[01:24:24] localized around small chips. Um so yeah it's again it's a very
[01:24:26] Um so yeah it's again it's a very difficult question to answer because the
[01:24:28] difficult question to answer because the the challenges could be infinite within
[01:24:30] the challenges could be infinite within the industry.
[01:24:32] the industry. >> Great. Well if you think uh fine tech
[01:24:35] >> Great. Well if you think uh fine tech can help you please get in touch with
[01:24:36] can help you please get in touch with me. I'll be very very happy to provide
[01:24:38] me. I'll be very very happy to provide an introduction to Travis. So uh thank
[01:24:41] an introduction to Travis. So uh thank you Travis. To bring this discussion up
[01:24:43] you Travis. To bring this discussion up up the stack we're pleased to welcome
[01:24:45] up the stack we're pleased to welcome our next speaker Matt Matt Crowley CEO
[01:24:48] our next speaker Matt Matt Crowley CEO of Cintel Phetonics. Matt will share the
[01:24:50] of Cintel Phetonics. Matt will share the application and product perspective
[01:24:53] application and product perspective drawing on Cintel's work translating
[01:24:55] drawing on Cintel's work translating advanced silicon phetonics and
[01:24:57] advanced silicon phetonics and heterogeneous integration into
[01:24:58] heterogeneous integration into deployable systems where performance
[01:25:01] deployable systems where performance packaging and manufacturability all have
[01:25:04] packaging and manufacturability all have to come together. Matt, the floor and
[01:25:06] to come together. Matt, the floor and the attention of everyone is yours.
[01:25:08] the attention of everyone is yours. Thank you.
[01:25:22] We don't hear you, Matt. Uh, are you
[01:25:24] We don't hear you, Matt. Uh, are you muted?
[01:25:36] Sorry about that. Okay.
[01:26:14] that
[01:26:15] that um
[01:26:18] um so thanks everybody. Uh heterogeneous
[01:26:20] so thanks everybody. Uh heterogeneous integration is a great topic for me. Uh
[01:26:23] integration is a great topic for me. Uh it's clearly core of what Cintel does.
[01:26:25] it's clearly core of what Cintel does. Uh so we're going to talk a little bit
[01:26:27] Uh so we're going to talk a little bit about the platform, a little bit about
[01:26:28] about the platform, a little bit about the product, and then a lot about the
[01:26:30] the product, and then a lot about the application and some of the view we
[01:26:32] application and some of the view we have. Uh we're a pretty focused company.
[01:26:34] have. Uh we're a pretty focused company. We like to joke we're laser focused. Um
[01:26:37] We like to joke we're laser focused. Um and our first product is targeting uh
[01:26:39] and our first product is targeting uh multi-wavelength uh external light
[01:26:41] multi-wavelength uh external light sources. We call them smart external
[01:26:43] sources. We call them smart external light sources for AI data centers with
[01:26:46] light sources for AI data centers with an even more focused target on scaleup
[01:26:48] an even more focused target on scaleup networks.
[01:26:50] networks. Um so what is SHIP? Uh ship is our cool
[01:26:53] Um so what is SHIP? Uh ship is our cool little acronym for heterogeneous
[01:26:55] little acronym for heterogeneous integrated photonics. Uh we don't have a
[01:26:57] integrated photonics. Uh we don't have a theme song yet, but I was inspired by
[01:26:59] theme song yet, but I was inspired by the optica one, so maybe we'll we'll
[01:27:00] the optica one, so maybe we'll we'll build one. Um really what we bring is a
[01:27:05] build one. Um really what we bring is a uh heterogeneous integration process
[01:27:06] uh heterogeneous integration process that uniquely leverages standard silicon
[01:27:09] that uniquely leverages standard silicon photonics. So working with our partner
[01:27:11] photonics. So working with our partner at Tower Semiconductor, we don't
[01:27:13] at Tower Semiconductor, we don't actually have to modify their base
[01:27:15] actually have to modify their base silicon photonix flow at all. we simply
[01:27:17] silicon photonix flow at all. we simply post-process 35 into that base flow
[01:27:20] post-process 35 into that base flow which has a lot of advantages for supply
[01:27:23] which has a lot of advantages for supply chain scaling and things like that. Uh
[01:27:26] chain scaling and things like that. Uh heterogeneous integration is great for
[01:27:28] heterogeneous integration is great for performance for density for reliability.
[01:27:30] performance for density for reliability. Uh it eliminates a lot of failure modes
[01:27:32] Uh it eliminates a lot of failure modes and of course it's scalable to huge
[01:27:34] and of course it's scalable to huge volume.
[01:27:37] volume. Um so talking a little bit about the
[01:27:39] Um so talking a little bit about the process itself.
[01:27:41] process itself. um we start with standard silicon
[01:27:43] um we start with standard silicon photonics and so you know block number
[01:27:45] photonics and so you know block number one is what we get from our foundry
[01:27:47] one is what we get from our foundry partner um what's nice is from this we
[01:27:49] partner um what's nice is from this we have a lot of IP blocks available that
[01:27:51] have a lot of IP blocks available that we don't have to redevelop our partner
[01:27:54] we don't have to redevelop our partner can know that any capex they do into
[01:27:56] can know that any capex they do into standard cipho uh for a normal
[01:27:59] standard cipho uh for a normal transceiver for example can also be used
[01:28:00] transceiver for example can also be used for our process so it's a great business
[01:28:02] for our process so it's a great business model for them and allows them to invest
[01:28:04] model for them and allows them to invest confidently in their own capacity
[01:28:07] confidently in their own capacity um and from now we have a lot of
[01:28:09] um and from now we have a lot of standard things we have this you know
[01:28:10] standard things we have this you know wave guides, photo detectors, uh
[01:28:13] wave guides, photo detectors, uh external couplers and things like that
[01:28:15] external couplers and things like that uh that we kind of get off the shelf
[01:28:17] uh that we kind of get off the shelf from our partner. The last part of the
[01:28:19] from our partner. The last part of the process is what is unique and this is
[01:28:21] process is what is unique and this is technology that originated at CATI with
[01:28:23] technology that originated at CATI with our founder um Solumno and has actually
[01:28:27] our founder um Solumno and has actually really been developed inside Cintil uh
[01:28:29] really been developed inside Cintil uh into a commercial process flow. I you
[01:28:32] into a commercial process flow. I you know didn't really put time frames but
[01:28:33] know didn't really put time frames but right now we are running hundreds of
[01:28:35] right now we are running hundreds of waiverss. So we did a fundra at the end
[01:28:37] waiverss. So we did a fundra at the end of last year and um we're kind of
[01:28:40] of last year and um we're kind of entering uh low initial low volume
[01:28:42] entering uh low initial low volume production sort of more focused on
[01:28:44] production sort of more focused on optimizing yields and such rather than
[01:28:47] optimizing yields and such rather than fundamentally changing the design or
[01:28:48] fundamentally changing the design or process flow. Um so the first step is we
[01:28:51] process flow. Um so the first step is we call back backside on box is we take the
[01:28:54] call back backside on box is we take the silicon photonix wafer we flip it upside
[01:28:57] silicon photonix wafer we flip it upside down we remove the handle and we bond it
[01:28:59] down we remove the handle and we bond it to a new handle. So we've effectively
[01:29:01] to a new handle. So we've effectively flipped the waf for upside down and
[01:29:03] flipped the waf for upside down and expose the oxide layer for bonding. In
[01:29:06] expose the oxide layer for bonding. In the next step, we do a molecular bond of
[01:29:08] the next step, we do a molecular bond of indium phosphide dye. These are not yet
[01:29:10] indium phosphide dye. These are not yet processed into lasers. It's our own
[01:29:12] processed into lasers. It's our own epistack that we purchased from third
[01:29:14] epistack that we purchased from third parties. And we bond it over the
[01:29:17] parties. And we bond it over the waveguide where it's going to eancently
[01:29:19] waveguide where it's going to eancently couple into the silicon photonics. Uh
[01:29:22] couple into the silicon photonics. Uh what's nice about this is um a couple of
[01:29:24] what's nice about this is um a couple of things. One is that the laser frequency
[01:29:26] things. One is that the laser frequency is determined by the grading that's made
[01:29:29] is determined by the grading that's made using advanced photoiththography. So we
[01:29:31] using advanced photoiththography. So we have extremely uh high level of
[01:29:33] have extremely uh high level of precision and control over the geometry
[01:29:35] precision and control over the geometry of the laser.
[01:29:37] of the laser. We also have high repeatability. So
[01:29:39] We also have high repeatability. So we're basically doing lithography. So we
[01:29:40] we're basically doing lithography. So we basically, you know, every laser on
[01:29:42] basically, you know, every laser on every reticle is going to be the same.
[01:29:43] every reticle is going to be the same. It's not written, nothing's moving. It's
[01:29:46] It's not written, nothing's moving. It's quite consistent. Uh in general,
[01:29:48] quite consistent. Uh in general, bringing let's say advanced
[01:29:49] bringing let's say advanced semiconductor process capabilities into
[01:29:51] semiconductor process capabilities into this has overall been very helpful to
[01:29:53] this has overall been very helpful to the process.
[01:29:54] the process. Uh critically, we also uh don't have to
[01:29:57] Uh critically, we also uh don't have to worry so much about alignment tolerances
[01:29:59] worry so much about alignment tolerances with this particular process flow
[01:30:00] with this particular process flow because we're doing the definition of
[01:30:02] because we're doing the definition of the laser post bond. And so even if the
[01:30:05] the laser post bond. And so even if the alignment's off at this phase, which
[01:30:06] alignment's off at this phase, which it's actually pretty good, it doesn't
[01:30:09] it's actually pretty good, it doesn't actually matter very much. And then in
[01:30:11] actually matter very much. And then in the fourth step, we actually fabricate
[01:30:12] the fourth step, we actually fabricate the laser. So we basically take that
[01:30:14] the laser. So we basically take that bonded uh dye that's made out of an
[01:30:17] bonded uh dye that's made out of an Indian phosphide stack and we etch and
[01:30:20] Indian phosphide stack and we etch and lithography and define the laser. Then
[01:30:21] lithography and define the laser. Then we passivate everything. We connect it
[01:30:23] we passivate everything. We connect it and voila, you have your laser die.
[01:30:28] and voila, you have your laser die. Our first product uh which we've talked
[01:30:30] Our first product uh which we've talked about is uh I think we really focused on
[01:30:32] about is uh I think we really focused on the external laser source as our first
[01:30:34] the external laser source as our first market. I think as a startup you have to
[01:30:37] market. I think as a startup you have to focus on a first market and if you're a
[01:30:39] focus on a first market and if you're a startup developing both a novel process
[01:30:40] startup developing both a novel process flow and a novel design, you have to
[01:30:43] flow and a novel design, you have to push your first process flow through
[01:30:44] push your first process flow through full qualification and your first design
[01:30:46] full qualification and your first design through full qualification so that you
[01:30:48] through full qualification so that you now have a platform on which you can
[01:30:50] now have a platform on which you can launch many many products. So our first
[01:30:52] launch many many products. So our first product is actually going to be an
[01:30:54] product is actually going to be an eightwavelength
[01:30:56] eightwavelength external laser source. Our customers,
[01:30:57] external laser source. Our customers, you know, were interested in eight and
[01:30:59] you know, were interested in eight and 16. Now they're thinking about eight as
[01:31:01] 16. Now they're thinking about eight as the first step, 16 is the second step.
[01:31:02] the first step, 16 is the second step. We've built both. Um the next spec is
[01:31:06] We've built both. Um the next spec is extremely important, which is you have
[01:31:08] extremely important, which is you have to have very precise frequency spacing.
[01:31:10] to have very precise frequency spacing. So it's not just the absolute distance,
[01:31:12] So it's not just the absolute distance, whether it be 100 GHz or 200 GHz or 400
[01:31:14] whether it be 100 GHz or 200 GHz or 400 GHz between channels. What's actually
[01:31:16] GHz between channels. What's actually critical to be able to work with the
[01:31:19] critical to be able to work with the micro ring resonators that our company
[01:31:21] micro ring resonators that our company our partners companies are using is that
[01:31:24] our partners companies are using is that the laser can't be too far off. And I
[01:31:25] the laser can't be too far off. And I like to think of like if I'm trying to
[01:31:27] like to think of like if I'm trying to shoot a target, the micro ring can pull
[01:31:30] shoot a target, the micro ring can pull a little bit, but if if your laser is
[01:31:32] a little bit, but if if your laser is too far off, it's not going to hit the
[01:31:34] too far off, it's not going to hit the target. And you think about over time,
[01:31:36] target. And you think about over time, temperature, and millions of units, you
[01:31:37] temperature, and millions of units, you need to have that precision. I can say
[01:31:39] need to have that precision. I can say not only do we have great uh accuracy
[01:31:41] not only do we have great uh accuracy off the wafer in this spec, but we also
[01:31:43] off the wafer in this spec, but we also have a built-in feedback system and a
[01:31:46] have a built-in feedback system and a custom ASIC that we developed in 65
[01:31:48] custom ASIC that we developed in 65 nanometer TSMC to have a complete system
[01:31:51] nanometer TSMC to have a complete system that not only has great process control,
[01:31:53] that not only has great process control, but also better than that. Um
[01:31:59] but also better than that. Um and so no AR codings. Uh so yeah, TW
[01:32:01] and so no AR codings. Uh so yeah, TW sorry it's a um yeah, that's correct. Um
[01:32:04] sorry it's a um yeah, that's correct. Um so no AR codings, no current through the
[01:32:06] so no AR codings, no current through the grading. um high power. Right now we're
[01:32:09] grading. um high power. Right now we're shipping 10 millatts per f per fiber per
[01:32:12] shipping 10 millatts per f per fiber per wavelength. We have our 20 millatt
[01:32:14] wavelength. We have our 20 millatt design in fab um and it's pretty
[01:32:17] design in fab um and it's pretty efficient. So that's you know something
[01:32:19] efficient. So that's you know something we have to measure over time and we have
[01:32:20] we have to measure over time and we have multiple fiber outputs per module. So we
[01:32:23] multiple fiber outputs per module. So we are actually also making a module where
[01:32:25] are actually also making a module where we can put you know potentially multiple
[01:32:26] we can put you know potentially multiple picks and our own custom ASIC and have a
[01:32:30] picks and our own custom ASIC and have a complete system that we can provide.
[01:32:31] complete system that we can provide. >> Matt, allow me to interrupt you because
[01:32:32] >> Matt, allow me to interrupt you because I cannot. This is this is a huge result
[01:32:35] I cannot. This is this is a huge result 20 mill per wavelength. This is this is
[01:32:37] 20 mill per wavelength. This is this is fantastic. Which application which is
[01:32:40] fantastic. Which application which is the key application that this is a
[01:32:42] the key application that this is a paradigm shift for?
[01:32:44] paradigm shift for? >> Uh well the amount of power per
[01:32:48] >> Uh well the amount of power per wavelength
[01:32:49] wavelength is really just um I think we're kind of
[01:32:52] is really just um I think we're kind of hitting some hard limits because we even
[01:32:54] hitting some hard limits because we even had to custom design our ASIC to even
[01:32:56] had to custom design our ASIC to even pump that much power. Um and I should
[01:32:58] pump that much power. Um and I should say that's not yet shipping. That's a
[01:33:00] say that's not yet shipping. That's a design infab. So, um, the reason people
[01:33:02] design infab. So, um, the reason people want more power is because they want to
[01:33:05] want more power is because they want to be able to use splits to basically run
[01:33:06] be able to use splits to basically run more channels. So, there's two
[01:33:09] more channels. So, there's two characteristics. One is, um, it's not
[01:33:12] characteristics. One is, um, it's not that our laser is necessarily higher
[01:33:14] that our laser is necessarily higher power than other people. It's that we
[01:33:16] power than other people. It's that we have less attenuation getting from the
[01:33:18] have less attenuation getting from the laser to the fiber, I think, is the key
[01:33:21] laser to the fiber, I think, is the key thing. So, there's other lasers out
[01:33:23] thing. So, there's other lasers out there that are higher. how we could make
[01:33:24] there that are higher. how we could make a design for that particular platform
[01:33:28] a design for that particular platform and let's see which partners they could
[01:33:29] and let's see which partners they could actually enable that allow me to give
[01:33:32] actually enable that allow me to give now the floor to Philip code Philip
[01:33:34] now the floor to Philip code Philip thank you very much for joining today
[01:33:36] thank you very much for joining today from flex compute I am really amazed
[01:33:38] from flex compute I am really amazed with what cintil photonix is doing I
[01:33:40] with what cintil photonix is doing I really am for for many many years Philip
[01:33:42] really am for for many many years Philip can you tell us a bit of flex compute
[01:33:44] can you tell us a bit of flex compute and how can you help companies like like
[01:33:46] and how can you help companies like like cintil and all the previous ones by
[01:33:48] cintil and all the previous ones by offering your platform to them the floor
[01:33:50] offering your platform to them the floor is yours
[01:33:52] is yours >> yeah absolutely thank you that just
[01:33:53] >> yeah absolutely thank you that just share my my screen.
[01:33:56] share my my screen. Okay, great.
[01:34:00] Okay, great. Yes. So, uh it's a pleasure to be here.
[01:34:02] Yes. So, uh it's a pleasure to be here. So, here at Flex Compute, what we
[01:34:04] So, here at Flex Compute, what we provide is uh tools for simulating and
[01:34:07] provide is uh tools for simulating and design photonic circuits and components.
[01:34:09] design photonic circuits and components. So, I believe you are in the a little
[01:34:10] So, I believe you are in the a little bit in the other end after you address
[01:34:12] bit in the other end after you address the process of integrating the the the
[01:34:15] the process of integrating the the the main materials, we are in the process of
[01:34:17] main materials, we are in the process of understand how our uh the device will
[01:34:20] understand how our uh the device will behave and providing tools for that. And
[01:34:22] behave and providing tools for that. And I believe a great advantage of flex
[01:34:23] I believe a great advantage of flex compute is that we actually have modern
[01:34:26] compute is that we actually have modern softwares and modern algorithms that
[01:34:28] softwares and modern algorithms that work native with GPUs. So we can scale
[01:34:30] work native with GPUs. So we can scale our simulations very much. So we can run
[01:34:33] our simulations very much. So we can run pretty large models that could not be
[01:34:35] pretty large models that could not be run before in CPU workstations. So it's
[01:34:37] run before in CPU workstations. So it's very fast and very scalable. Uh but for
[01:34:40] very fast and very scalable. Uh but for the integrated world when we need to put
[01:34:43] the integrated world when we need to put many place many things together work
[01:34:44] many place many things together work with many physics speed and scalability
[01:34:46] with many physics speed and scalability is just one part of the process. we need
[01:34:48] is just one part of the process. we need actually a full solution integrated to
[01:34:51] actually a full solution integrated to deal with complex problems. So I would
[01:34:54] deal with complex problems. So I would like to showcase here this interesting
[01:34:56] like to showcase here this interesting example. Uh this is inspired in a paper
[01:34:59] example. Uh this is inspired in a paper where the authors they created a
[01:35:01] where the authors they created a electroic modulator based on a silicon
[01:35:03] electroic modulator based on a silicon waveguide. So the idea is that we have a
[01:35:06] waveguide. So the idea is that we have a slot silicon waveguide and you fill the
[01:35:08] slot silicon waveguide and you fill the gap with a polymer and you create some
[01:35:11] gap with a polymer and you create some process in the polymer to induce the
[01:35:12] process in the polymer to induce the electrooptical effect in the polymer. So
[01:35:14] electrooptical effect in the polymer. So you can uh modulate the the optical
[01:35:17] you can uh modulate the the optical field in the silicon wave guide. And to
[01:35:20] field in the silicon wave guide. And to design and understand this device
[01:35:22] design and understand this device there's many steps. You need to carry
[01:35:24] there's many steps. You need to carry out the chart simulation to sulfur the
[01:35:26] out the chart simulation to sulfur the DC fields into the polymer. Uh you need
[01:35:28] DC fields into the polymer. Uh you need to solve for the RF fields that will
[01:35:30] to solve for the RF fields that will modulate the your optical signal. You
[01:35:32] modulate the your optical signal. You need to couple the RF fields into uh
[01:35:36] need to couple the RF fields into uh your optical wave guide. you need to run
[01:35:38] your optical wave guide. you need to run optical uh mode simulations considering
[01:35:41] optical uh mode simulations considering the both the RF simulation the RF fields
[01:35:44] the both the RF simulation the RF fields and also the doping in the silicon. So
[01:35:48] and also the doping in the silicon. So you need to put everything together in
[01:35:50] you need to put everything together in order to understand this uh this device
[01:35:53] order to understand this uh this device and this is what we offer. So we have
[01:35:54] and this is what we offer. So we have many tools that working together uh
[01:35:59] many tools that working together uh to fully characterize this device uh
[01:36:03] to fully characterize this device uh this design in a single workflow. So the
[01:36:05] this design in a single workflow. So the idea is that you have a single workflow
[01:36:07] idea is that you have a single workflow and the engineer just need to take care
[01:36:08] and the engineer just need to take care about the problem at hand and don't need
[01:36:10] about the problem at hand and don't need to wait a lot for for run this model or
[01:36:14] to wait a lot for for run this model or waste too much time integrating with
[01:36:16] waste too much time integrating with different tools and once you have your
[01:36:18] different tools and once you have your component uh after some point you need
[01:36:20] component uh after some point you need to put everything together in a circuit
[01:36:22] to put everything together in a circuit and for that we also offer photon fors a
[01:36:25] and for that we also offer photon fors a very powerful tool for photonics uh
[01:36:28] very powerful tool for photonics uh circuit design layout and the idea is
[01:36:31] circuit design layout and the idea is that you can uh integrate with a given
[01:36:33] that you can uh integrate with a given founder technology
[01:36:34] founder technology which allows us to after create our 2D
[01:36:37] which allows us to after create our 2D layout, we can automatically create 3D
[01:36:39] layout, we can automatically create 3D models from our structure and carry out
[01:36:42] models from our structure and carry out circuit level simulation and fully
[01:36:43] circuit level simulation and fully characterize our ship and at the same
[01:36:46] characterize our ship and at the same time exploit the integration of our
[01:36:48] time exploit the integration of our solvers and the speed and scalability.
[01:36:50] solvers and the speed and scalability. Well, uh I believe that's it for for
[01:36:52] Well, uh I believe that's it for for now. If you have any questions, I'm
[01:36:54] now. If you have any questions, I'm happy to answer. Thank you.
[01:36:56] happy to answer. Thank you. >> Thank you so much for that. It is so
[01:36:58] >> Thank you so much for that. It is so great to have the con support of flex
[01:36:59] great to have the con support of flex compute and how you are helping so many
[01:37:01] compute and how you are helping so many foundaries. Philip, I stay tuned because
[01:37:04] foundaries. Philip, I stay tuned because the next one is perhaps what I am most
[01:37:07] the next one is perhaps what I am most excited about this meeting. The company
[01:37:10] excited about this meeting. The company Polaron keeps breaking records when it
[01:37:12] Polaron keeps breaking records when it comes to high speeded modulation year
[01:37:14] comes to high speeded modulation year after year. Stefan Kosh from Polaron,
[01:37:17] after year. Stefan Kosh from Polaron, thank you for being here with us. You
[01:37:19] thank you for being here with us. You had the honor for being the closing
[01:37:21] had the honor for being the closing speech. Tell us what Polaron can do. And
[01:37:23] speech. Tell us what Polaron can do. And then in one or two sentences, tell us
[01:37:25] then in one or two sentences, tell us which partners you can work with after
[01:37:27] which partners you can work with after this meeting. The floor and the
[01:37:28] this meeting. The floor and the attention of everyone goes to Polaron.
[01:37:31] attention of everyone goes to Polaron. Thank you, Jose. Hello everyone. That's
[01:37:35] Thank you, Jose. Hello everyone. That's my name is Stefan. I'm from Polaron.
[01:37:38] my name is Stefan. I'm from Polaron. The best for last. Well, two minute
[01:37:40] The best for last. Well, two minute pitch.
[01:37:41] pitch. >> Two bit pitch from myself. I will be
[01:37:43] >> Two bit pitch from myself. I will be very tight. Um, so high-speed modulation
[01:37:47] very tight. Um, so high-speed modulation is essentially a a race of the materials
[01:37:51] is essentially a a race of the materials and this is data from light counting
[01:37:53] and this is data from light counting that says by 2029 all platforms will
[01:37:57] that says by 2029 all platforms will will be will be there in the market.
[01:37:59] will be will be there in the market. Well, silicon photonics will be well
[01:38:01] Well, silicon photonics will be well positioned in uh in in various forms,
[01:38:04] positioned in uh in in various forms, monolithic and in heterogeneous forms.
[01:38:07] monolithic and in heterogeneous forms. No, no, no doubt. But in a way, it's
[01:38:10] No, no, no doubt. But in a way, it's integration of materials. And the
[01:38:12] integration of materials. And the material we integrate is um is polymers.
[01:38:15] material we integrate is um is polymers. Um and this we call plasmonics. This is
[01:38:19] Um and this we call plasmonics. This is a post-processing step done on silicon
[01:38:22] a post-processing step done on silicon photonics. And uh the big advantage or
[01:38:26] photonics. And uh the big advantage or the big takeaway from this way of doing
[01:38:29] the big takeaway from this way of doing it because this is a very simple process
[01:38:31] it because this is a very simple process step. We have no complex layer stacks.
[01:38:35] step. We have no complex layer stacks. We have one redistribution layer. We
[01:38:37] We have one redistribution layer. We have one material iteration step and
[01:38:39] have one material iteration step and then the passivation
[01:38:41] then the passivation and the proof of concept of that is with
[01:38:43] and the proof of concept of that is with a silicon 200 mm silicon done from IMAC
[01:38:47] a silicon 200 mm silicon done from IMAC which we've made a press release towards
[01:38:50] which we've made a press release towards the end of last year about the
[01:38:52] the end of last year about the performance
[01:38:54] performance and the performance is high at any will.
[01:38:58] and the performance is high at any will. Maybe there are other fastest modulators
[01:39:01] Maybe there are other fastest modulators in the world and the frequency response
[01:39:04] in the world and the frequency response of a modulator uh in this specific uh
[01:39:07] of a modulator uh in this specific uh plot is flat until 100 gigahertz and
[01:39:12] plot is flat until 100 gigahertz and what you see is the impact of the test
[01:39:13] what you see is the impact of the test and measurement instruments the AWGs and
[01:39:16] and measurement instruments the AWGs and maybe and that make the entire system
[01:39:19] maybe and that make the entire system performance and the effect that uh we
[01:39:22] performance and the effect that uh we are dependent at the end on faster
[01:39:24] are dependent at the end on faster electronics or driving circuitry makes
[01:39:27] electronics or driving circuitry makes that the current um published eye
[01:39:30] that the current um published eye diagrams are maybe not better than the
[01:39:33] diagrams are maybe not better than the one from Steven. H nevertheless we know
[01:39:36] one from Steven. H nevertheless we know how to make the improvement improve them
[01:39:38] how to make the improvement improve them better and because of the driving
[01:39:40] better and because of the driving electronics and so this is h possible
[01:39:45] electronics and so this is h possible because our devices are in the microbit
[01:39:48] because our devices are in the microbit dimensions and what can be made small
[01:39:52] dimensions and what can be made small can as well be made power efficient
[01:39:54] can as well be made power efficient because you have to switch less energy
[01:39:56] because you have to switch less energy and this becomes a fantastic value
[01:39:59] and this becomes a fantastic value proposition it's not about being going w
[01:40:03] proposition it's not about being going w as as narrow or a gap about going slow.
[01:40:06] as as narrow or a gap about going slow. We can do fast and wide at the same
[01:40:07] We can do fast and wide at the same time. And this the value proposition for
[01:40:10] time. And this the value proposition for onboard optics, co- packaged optics and
[01:40:13] onboard optics, co- packaged optics and any other variants of heterogeneous
[01:40:15] any other variants of heterogeneous integration. So what can you do for us?
[01:40:21] integration. So what can you do for us? If there is anyone that can have help us
[01:40:23] If there is anyone that can have help us in doing encapsulation technology is
[01:40:27] in doing encapsulation technology is something that I would be willing to
[01:40:28] something that I would be willing to have a talk at any point in time. Why is
[01:40:31] have a talk at any point in time. Why is it important? It's one of the processing
[01:40:33] it important? It's one of the processing step. Does polarity not have that one?
[01:40:36] step. Does polarity not have that one? Of course, we have that one. But better
[01:40:38] Of course, we have that one. But better um encapsulation will permit you to um
[01:40:42] um encapsulation will permit you to um to to follow cost curves and all what is
[01:40:46] to to follow cost curves and all what is needed for being in production. Thank
[01:40:49] needed for being in production. Thank you so much.
[01:40:49] you so much. >> You you finished the game. You defeated
[01:40:51] >> You you finished the game. You defeated the final boss because you don't go
[01:40:53] the final boss because you don't go narrow and fast. You don't go slow and
[01:40:56] narrow and fast. You don't go slow and wide. You actually actually go fast and
[01:40:58] wide. You actually actually go fast and wide. But you just said something very
[01:41:00] wide. But you just said something very interesting. You are making devices now.
[01:41:04] interesting. You are making devices now. Is that the new the new business plan
[01:41:06] Is that the new the new business plan for for Polaron to to be a device
[01:41:09] for for Polaron to to be a device manufacturer?
[01:41:12] manufacturer? >> Well, whether it's new or not, h
[01:41:14] >> Well, whether it's new or not, h Plasmonics is it's a device. You're
[01:41:17] Plasmonics is it's a device. You're right. It's a it's a pick. Now we can do
[01:41:19] right. It's a it's a pick. Now we can do a pick with uh Baba choosing our own
[01:41:23] a pick with uh Baba choosing our own silicon photonics vendors but actually
[01:41:25] silicon photonics vendors but actually this is integrable into into any uh
[01:41:29] this is integrable into into any uh silicon photonix uh process baseline
[01:41:32] silicon photonix uh process baseline process in that sense it's a
[01:41:34] process in that sense it's a post-processing and yes well what you
[01:41:39] post-processing and yes well what you can do for you we can provide you faster
[01:41:42] can do for you we can provide you faster moderators
[01:41:44] moderators >> for all the packaging companies in the
[01:41:46] >> for all the packaging companies in the room if you are brave enough to package
[01:41:48] room if you are brave enough to package an 800 gig per lane chip. Polariton is a
[01:41:52] an 800 gig per lane chip. Polariton is a fantastic partner. As I said, this is
[01:41:54] fantastic partner. As I said, this is the one to watch. I would like to thank
[01:41:56] the one to watch. I would like to thank everyone in the room. Allow me now to
[01:41:58] everyone in the room. Allow me now to close the meeting for the YouTubers and
[01:42:00] close the meeting for the YouTubers and then we can have a private discussion to
[01:42:02] then we can have a private discussion to follow up on the main leads of the
[01:42:04] follow up on the main leads of the event.
[01:42:16] This is optical.
[01:42:24] >> [music]
[01:42:26] >> [music] >> This is lighting redefined.
[01:42:29] >> This is lighting redefined. [music]
[01:42:41] Okay. So first of all allow me to ask my
[01:42:43] Okay. So first of all allow me to ask my AT department first for
[01:42:46] AT department first for great cut the live stream

https://www.youtube.com/watch?v=rzgQMuFsltU

Summary

TL;DR — Optica's online industry meeting focused on heterogeneous integration in silicon photonics, bringing together industry leaders to discuss practical challenges and manufacturing scalability. The discussion highlighted advancements in integrating diverse materials like Indium Phosphide, thin-film Lithium Niobate, and polymers onto silicon platforms, moving beyond academic theory to real-world applications. Key themes included overcoming bottlenecks in alignment, thermal management, process integration, yield, and reliability to meet the growing demands of AI-driven industries.

Key points

• Heterogeneous integration is crucial for advancing silicon photonics, enabling the combination of different materials and devices to achieve higher performance and functionality.

• Microtransfer printing (MTP) technology offers a scalable solution for placing active components, potentially revolutionizing high-volume data communications markets and enabling higher bandwidth transceivers.

• Various integration techniques, including epitaxy, microtransfer printing, die bonding, and wafer-to-wafer bonding, are being explored and refined to combine different photonic materials with silicon platforms.

• Packaging expertise is vital for enabling co-packaged optics (CPO), bringing optical connections closer to electronics for faster bandwidth, lower power, and higher efficiency, though it introduces complexity in repair and integration.

• Companies are developing specialized platforms and PDKs to facilitate the integration of 35 materials onto silicon photonics, aiming for cost-effectiveness, high yield, and manufacturability at scale.

• Advancements in high-speed modulation, particularly with polymers and advanced silicon photonics, are pushing the boundaries of data rates, with companies demonstrating capabilities beyond 100 GHz.

Takeaway — Heterogeneous integration is rapidly maturing from a research concept to a practical manufacturing approach, essential for scaling silicon photonics to meet the demands of AI and high-bandwidth applications.