# The FEMTUM PHOTONICS SUCCESS STORY FROM QUEBEC CITY

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

[00:09] Here I am in Quebec City at the headquarters of Femto the company that develops femtosecond lasers at three micrometers that is revolutionizing the photonic integrated circuit manufacturing.
[00:20] Luis Rafael I'm here I came all the way to Quebec City because I wanted to see with my own eyes what is going on with the fixing the trimming and the cleaning of silicon photonics.
[00:30] But before that, what is Femto these days?
[00:32] It's a amazing days.
[00:33] We have basically launched a trimming and cleaning and we are fixing pics.
[00:38] And now we have a huge demand for that.
[00:40] We have wafers coming from everywhere in the world from mini fabs, mini customers, third ones, third twos.
[00:45] It's very exciting days and now we are basically constrained on delivering to the customers.
[00:50] It's going to get a while since we made that interview at OFC.
[00:51] You launched a product and things have happened, right?
[00:56] Now the company is already desperate to use your machine.
[00:57] How does it work?
[00:59] Which are the customers that are more exciting for you?
[01:03] So the customers that are very
[01:05] excited about trimming it's it's really the ones that are trying to launch the next generation 3.2T uh transceivers.
[01:12] And right now with trimming what you can do is really solve the thermal phase shifter problem and get rid of most of the energy and fix fundamentally the design and bring huge amount of value.
[01:24] So this is for trimming.
[01:24] The second one for cleaning it's right now the co-package optics we know it's super hard to manufacture and only one dust can kill the full thing.
[01:32] So we are working exactly on that process.
[01:34] How can you using a femtosecond laser being able to trim silicon photonic waveguides?
[01:39] Am I missing something here?
[01:40] So the key IP on this one is uh very interesting.
[01:43] It's actually what we do is we are stressing so we are melting the clad with ultra fast pulse and the this melt basically stress the waveguide and change the refractive index.
[01:55] So it's kind of a laser welding, ultra-fast pulse process, but very similar to a piezo effect.
[02:02] Do you have the IP for that particular process?
[02:05] Absolutely.
[02:05] So, your customers can
[02:06] actually buy equipment from you and use it in their facilities.
[02:09] Exactly.
[02:09] And uh we are basically uh a laser solution, so we sell the laser source, laser head, the IP, the software.
[02:16] So, this is a subsystem, a bit like a Keysight or X-FAB subsystem, and we integrate into a machine like a ficonTEC and MPI, which are very close collaborator to us.
[02:24] We're open also for other machines.
[02:24] You have worked with ficonTEC and MPI at I saw in your in your demo at SiO collaboration with ficonTEC.
[02:31] How is that collaboration going?
[02:32] It's going very well.
[02:34] They're super supportive.
[02:34] And right now there's crazy demands around the world about Sick Photonics, as everybody knows.
[02:39] So, I mean it's a it's really timely with them.
[02:41] I called you last night and I told you that I was coming and you told me that you are preparing a demo for us.
[02:46] Thank you very much for working on it last night.
[02:47] What is the demo about?
[02:49] So, we're going to show just the simplified way because I mean, we we have to stay public.
[02:53] So, we have a wafer from AMF, and it's a with Well Picks, it's a very it's a new startup in Quebec City.
[02:59] They're developing a data com chips, and we have a wafer with them.
[03:02] We are close collaborators, and basically we're going to show how we trim one Mach-Zehnder in
[03:07] a live and in a live way, and so people can really capture the essence of trimming.
[03:11] So, I think we can go there.
[03:13] Let's go to see the wave the the the trimming of wave guides from AMF, now GlobalFoundries, by the way.
[03:17] Let's go and see how it works.
[03:18] I am really so looking forward to seeing this because one thing is to see the companies at an exhibition and the booth, and the other one is to to come here.
[03:25] How many employees do you have by now?
[03:27] So, since last year we grew from like I think we were 12, now we're 37.
[03:31] 37 people.
[03:35] Oh, fantastic.
[03:36] This is a big growth.
[03:37] And all of them working now on the silicon photonics side or are you still targeting other applications?
[03:42] So, the cleaning and trimming it's really Sick Photonics, and there's a new product upcoming next year for new process, which I cannot tell you.
[03:48] Okay, yes, that's a teaser, you know.
[03:50] We are now at the at the lab of Fhentum, at the laboratory, the application lab.
[03:54] Thank you very much Jaime here and Jose from Optica.
[03:56] It is great to meet you.
[03:58] I came all the way here because Luis Rafael, the CEO and founder of Fhentum, told me that last night you guys were working on
[04:07] a demo for me and you wanted to show it.
[04:10] So, let me explain to you to everybody how it works the laser trimming.
[04:12] So, here it's a very simple case of a non-balanced Mach-Zehnder that we're going to trim.
[04:14] So, what you see here is uh we're starting the process where the wing sum AOI, so Z scan to make really make sure that we are at the at the right layer on the stack.
[04:26] Now, we are looking at the grating couplers to make sure that we can couple light very efficiently into the the circuit.
[04:31] And now uh very shortly, what we're going to do is we are going to see the first light.
[04:35] So, now we are with the IFU, the fiber array unit, injecting into like this uh array of a grating couplers, making sure that we have a very nice signal into the the circuit.
[04:45] And once done, uh we will see uh we will measure the optical response.
[04:51] So, since it's an unbalanced Mach-Zehnder, uh the response is actually um very cyclical.
[04:59] And now we see the the the real natural response since it's an analog circuit.
[05:03] And uh very soon now, what you will see it's the laser
[05:08] trimming itself.
[05:11] So, now we're applying uh our pulses to the waveguide which stress on the sides of each and it was very, I mean, very fast.
[05:16] And those stress rods uh will actually change the refractive index in a permanent fashion.
[05:23] It's a set once and forget uh process step.
[05:28] And now, as you can see, uh the response and this is all in dB.
[05:30] Basically, you see that there's basic like by the eye, there's a very minimal uh background loss.
[05:37] Actually, this is a almost lossless uh process.
[05:40] Then we have applied a very specific amount of uh of phase shift done by a laser trimming step.
[05:46] You can do this on every resonator.
[05:48] You can do this on ring resonators, on mass tenders, any any periodic structure.
[05:53] the fundamental thing is that a wave guide is is a very fundamental structure and and it's all about phase and refractive index and effective refractive index.
[06:01] So, we can shoot on on multi-mode, we can shoot on single mode, on slabs, on ribs.
[06:06] I mean, there's many
[06:08] many things that we can do and that's the beauty, I think, of it.
[06:11] But Rafael, so everybody's wondering right now putting lots of resonating structures on silicon photonic chips.
[06:17] We are seeing companies like Light Matter, for example, making a huge huge impact on that.
[06:22] And we have the problem of putting heaters to tune them locally.
[06:26] You're actually solving that with set and forget settings.
[06:27] Exactly.
[06:30] So, this is the same function as a thermal phase shifter.
[06:32] Instead of applying 24/7 energy, so actually consuming power, and you know, AI the the key button on AI is power.
[06:39] Mhm.
[06:42] And actually, for making them work, the micro rings, you need to apply power.
[06:43] So, we are actually fixing this power issue.
[06:46] And it could sum up like for a a dense WDM low and wide and slow and wide like to up to 30% of the power consumption of a PIC.
[06:54] So, we can remove, I mean, a lot of that energy into a single circuit.
[06:58] The amazing thing that this is a process that is incorporated to the fiber contact machine, so you can do this at the same time as you do the packaging and testing.
[07:05] Exactly.
[07:07] So, it's an add-on, so
[07:09] you actually need to test everything.
[07:11] You need to test every chip, every circuits.
[07:13] So, what we're adding on is after that test, we add a trimming step and another final test.
[07:19] So, when you're doing the wafer level testing of the silicon photonics wafer, you can actually trim it at the same time?
[07:25] Exactly. That's the insertion.
[07:26] It's that exactly what we have here. So, this is a full 300 mm wafer level tester.
[07:31] We have east-west optical coupling with six axis on each side.
[07:35] We have the the femto head.
[07:37] We have also on the north side this electro electro test.
[07:42] So, right now we can do this at wafer scale and it's super efficient and fast.
[07:47] What is the the next big thing?
[07:48] So, right now you we have the machine, we can trim the wafers, you are getting wafers from all the silicon photonic foundries, so you can make sure that this works for everyone.
[07:57] What is the next thing that you want to do?
[07:59] So, I think the the the crazy thing that is happening with the AI is the the data com is going to change.
[08:05] So, the slow and wide is really a killer app. We
[08:09] want to win that killer app with laser trimming.
[08:12] And there's also many other killer apps for us, but of course data com and telecom are very a key focus and we really want all the customers bring the most value to their chip with this process.
[08:24] But you're seeing all these short reach interconnects using this large amount of ring resonators, this large amount of Mach-Zehnder interferometers.
[08:30] What What do you think they are not coming to you yet?
[08:33] Is because they don't know you or because they have different solutions that you think are perhaps competing?
[08:38] I think of course, I mean, this is a new process, they don't know us.
[08:42] We need to do our I mean, every foundry, every stack, every designs are unique.
[08:46] So, we need to like re-engineer the process for each of those combinations.
[08:49] Um so, I think that's part of the thing.
[08:52] Second thing, but at the end of the day, the the rings for instance, it's a very hard problem to solve because the thermo-optic shifter, if for instance, if you're right, it's like the process variations are just strong enough, so you're past the the the target.
[09:08] Actually, you need to do it like a full
[09:10] to buy a full cycle and the thermo shifter over time burn itself.
[09:15] So, it's a reliability problem.
[09:18] And you can also anneal the dopant inside the micro ring, which also can kill the micro ring.
[09:25] So, there are many many known issues with micro rings with thermo-optic shifters.
[09:28] So, it's very serious problem for the industry.
[09:32] And I think the trimming is is one way to really solve this.
[09:36] You work directly with the foundry, but I you also have to work directly with the customer of the foundry to make sure that they that this process satisfy their needs.
[09:43] So, where is the Femto machine going to be?
[09:45] At the foundry or at the customer location?
[09:47] So, we It's very good question.
[09:50] So, right now we work with the fabless companies.
[09:52] So, they send us their wafers.
[09:54] What is fully processed wafers.
[09:56] So, they do the their first test and it's basically the OSA.
[10:01] So, when you do the real functional test of these wafers, that's where you apply the trimming because you have to know the real response of the die.
[10:10] Uh remember, it's an analog chip.
[10:12] So, you have to measure the real response.
[10:17] Like not uh it's not a digital chip, not yet.
[10:19] So, you have showed us how the machine looks outside.
[10:22] But can actually can we actually see what's inside?
[10:24] Of course.
[10:24] Safety first.
[10:27] We'll put the the laser glasses.
[10:29] And let's open it.
[10:35] So, what are we looking at?
[10:36] So, it's a it's the complete system.
[10:39] So, first uh this machine is first of all it is a wafer level tester provided by Falcon Tech.
[10:47] So, this is their newest version for top side test.
[10:49] So, it's a 300 mm wafer chuck.
[10:52] You have on the east and on the west side two six optical alignment six axis optical aligners.
[10:57] You see that if a U is coming on the right side over here.
[10:59] On the north side, you have the electrical test.
[11:03] So, and we could also add if we want south uh south test for electrical, too.
[11:07] For this setup, we also have the instruments.
[11:09] So, the lasers, the detectors to really characterize the
[11:13] die.
[11:13] So, there's the Falcon Tech's and of course, there's the Femto product, the Femto laser solution.
[11:19] We have this laser head where basically you have this fiber delivery.
[11:21] It goes inside the system.
[11:24] You have a galvo scanner.
[11:26] You have optimized lenses really to focalize the most accurate fashion and also the largest field of view possible and we also have the feature of having inline vision.
[11:35] So, it's collinear.
[11:37] We can see through the same objective.
[11:39] All these optics was optimized by Femto M and at the very top we have the light source, the laser controllers and everything.
[11:45] So, it's a rack mount fiber laser at the outside.
[11:46] And of course, you have the wafer.
[11:49] So, this is a 8-in wafer, but this machine can do a 12-in wafer, too.
[11:53] Can do 12 in.
[11:56] So, here you have the the the laser from Femto M focus on the wafer and you're actually melting on both sides on the wave guide.
[12:03] How accurate is this?
[12:06] So, right now what we specify is a it's like a 500 nm accuracy from the wave guide and because we're shining these pulses
[12:14] very close, it's basically like a Gaussian and this size is really important,
[12:20] but basically this is the the the full accuracy of the full system with the vibration and everything included.
[12:26] Well, Rafael, of course, this is the FiconTEC machine, but what I can see as well here is that the presence of the Quebec ecosystem.
[12:29] I can see Exfo.
[12:33] How does it work here?
[12:35] How does the optronic community in Quebec work and interact with each other?
[12:39] To me, success is a collaborative effort.
[12:42] Especially for a startup like us.
[12:45] So, without the collaborators, we will not meet the speed and the quality of the customer that what they ask and we work with very nice customers.
[12:52] So, we need high quality ISP.
[12:55] So, we have to partner up.
[12:57] Our key IP and key value add is the is the light source, is the process, is the software.
[13:04] The rest we outsource and we collaborate and of course, having local presence that help so much because then we can fix the problem in a very timely fashion.
[13:14] Your customers are from all over
[13:16] world, of course, from Canada, from United States, from Asia.
[13:21] Is it a good advantage for you to be in Canada being able to work on both sides of the world?
[13:25] I I I I'm not like but Canada, right now, worldwide, it's a very nice position, geopolitically speaking.
[13:32] So, we're not messing with anybody.
[13:33] And uh basically, people are very free to work with us worldwide.
[13:38] We have seen, of course, that there is a race for who is going to make the first short-reach interconnect links on silicon photonics for AI.
[13:46] Being in Canada allows you to work on both sides.
[13:48] Allows you to work with the companies in Silicon Valley and also with the companies in Asia.
[13:53] Of course. And the fact that we're in Canada, we're very close to the US.
[13:56] And there there's basically two poles in photonics.
[13:58] There's the northeast, which is very close to us, where where we are.
[14:01] And there's also the the west coast.
[14:02] So, these two poles are well, we're very close to the first one.
[14:06] And And the other one is like 6 hours by flight, so it's pretty efficient.
[14:08] You already mentioned to us that the wafer is from AMF, now Global Foundries.
[14:12] And you already mentioned to us that you can work with different wafer sizes.
[14:16] you looking also at other material systems like silicon nitride or So.
[14:20] right now, we I just we just published at WEC a double ring trimming of silicon nitride.
[14:24] So, you can look it up.
[14:26] It was a very nice paper we presented.
[14:28] So, yeah, silicon silicon nitride, but we have had the wafers from Tower.
[14:32] We have upcoming wafers from TSMC.
[14:34] We have worked on CEA-Leti, IMEC, AMF.
[14:38] I mean, the the process itself needs to basically be fine-tuned for each of these stacks and the and and foundries and and unique designs.
[14:48] But that's our job, man.
[14:50] We love to do that.
[14:50] What I love the most about this is not only that you do magic by by trimming your waveguides, but that you actually do it at wafer level and at the same time as you're doing the wafer level testing.
[15:00] But this only trimming.
[15:02] You told me something else.
[15:03] You can do cleaning.
[15:03] Can you tell us a bit more about this?
[15:05] So, the cleaning is actually way like easier to integrate.
[15:09] So, we have another machine over there.
[15:11] So, we can look it up.
[15:11] It's basically a strip-down version.
[15:12] So, the we shine pulses.
[15:15] So, as we were developing this
[15:17] trimming process, we found a lot of way not to trim.
[15:18] Uh and we know really much how to not to interact.
[15:21] So and we also find ways how to clean.
[15:24] So we found like three new patterns on the cleaning.
[15:26] And basically on this side we can do the cleaning of the grating couplers of all these critical surface.
[15:31] Especially when you arrive with the fiber attached just before the final fiber attached.
[15:33] These critical surface like interface between the fiber and the silicon photonics bits this need to be pristine and without any particles.
[15:35] And right now the key problem is that people are doing that by hand.
[15:47] So they will wipe that off or use a dry ice which is really hard to integrate or use plasma clean which basically blow everything.
[15:57] So um laser laser cleaning is a widely used in other industries.
[15:59] Was not yet in use for silicon photonics and we're doing that with our very special lasers.
[16:02] So you have a stream down version of this machine only for laser cleaning.
[16:05] Yes.
[16:07] Can we look at it?
[16:09] Of course.
[16:11] Let's go.
[16:18] So here is the cleaning machine.
[16:18] Yes.
[16:21] So this is the new S laser cleaner.
[16:24] So it's a it's a cleaning machine that we work with.
[16:26] You can take you see the machine is is a smaller and basically we can put parts below and do this laser cleaning.
[16:32] So everything is integrated.
[16:34] It's the the same software environment and it's an integration of um of of this process.
[16:42] At the end of the day you're bringing the the wafer here and it gets clean.
[16:44] You have to make sure it doesn't get dirty afterwards.
[16:46] What is the process after getting the wafer clean?
[16:51] So these machines are basically in clean rooms like at the Ozarks or at the fab.
[16:56] So you have to handle properly the the parts.
[16:58] But they go basically to the next step in the production flow.
[17:01] So basically it could be a bonding step and assembly step, a test step because for test if it's not clean you could find a dirt that killed a chip just by they will find that the test is not good.
[17:14] So there's like for those critical process you need clean samples.
[17:19] It feels great to be in Quebec City seeing this, being able to touch machines that are real and that they are real and all the companies in the world can actually use them.
[17:26] Congratulations really Rafael on what you're doing.
[17:28] This time for me to go.
[17:30] Can we walk outside?
[17:32] Yes.
[17:33] You You keep defining Phantom as a startup.
[17:35] 35 people already working with different companies.
[17:38] How How do you consider yourself a startup if you already have a very well-defined product that is already commercially available?
[17:46] So we could say we are more a scale up now.
[17:48] I think we are the the I think it's a VC term more than a industrial commercial term.
[17:53] So yes, I mean we're a company.
[17:55] We deliver.
[17:57] We have a manufacturing.
[17:58] We make our stuff.
[18:01] We have you know, we sell, we produce, we have projects.
[18:03] So we are company and but it's just that we just raised a very nice series A.
[18:05] So a 16 million round.
[18:07] So I think that's that's why.
[18:12] In the next months we are going to meet again in Malaga at the Global Photonics Economic Forum and there we are talking about success stories on companies who
[18:20] are doing business globally.
[18:22] What is the your your message for those who are really struggling on on benefiting from the fact that we can do business in Asia in US at the same time?
[18:32] It's a very good question.
[18:34] I think relationship matters.
[18:36] You have to be face-to-face with the customer.
[18:39] You have to show real data.
[18:43] Show the real the the the real valuable data.
[18:44] Show the value.
[18:46] But face-to-face.
[18:46] So go out there and understand the problems and but also build a relationship specially with new countries like other countries than what you do.
[18:58] I think that's the that's key and having very strong partnerships like we do have a mouth like I can take an MPI but we're very welcome to other partnerships.
[19:07] We do have a IMEC say a lady.
[19:09] I know.
[19:09] I mean we collaborate a lot.
[19:11] So, those partnerships are key for success.
[19:14] Luis Rafael, it is time for me to go, but I will see you in Malaga for the Global Photonics Second Week Forum.
[19:19] Bye-bye.