This market research report was originally published at the Yole Group’s website. It is reprinted here with the permission of the Yole Group.
The VCSEL market for consumer applications took off in 2017 with the introduction of the 3D facial recognition feature in smartphones. Since that time, the consumer market for VCSELs has kept growing and is expected to reach $1.7 billion in 2027, according to the VCSELs market and technology trends 2022 report published by the Yole Group. The 940 nm GaAs-based VCSELs and the sensors were integrated on top of the display, usually in a notch bar.
This implementation created a frustration as the notch took some space, split the display and was not compatible with the trend of having an OLED display occupying the full size of the smartphone. However, some recent advancements have been achieved to reduce the notch size in the Apple’s iPhone 14 pro family. The proximity sensor has been placed under the display thanks to the use of the InP-based long-wavelength emitter and receiver at 1,380 nm, thus being in the optical transmission window of OLED screens. The Yole Group’s compound semiconductor team is expecting the InP market to reach $5.7 billion in 2027 (Source: InP 2022 report – Compound Semiconductor Market Monitor) where the InP consumer sensing applications could reach a $356 million market.
Founded in 1923 as a series of mechanical workshops, TRUMPF has since developed into one of the world’s leading companies for machine tools, laser technology, and electronics for industrial applications. A few months ago, TRUMPF announced that it had already mastered an innovative manufacturing process for long-wavelength VCSELs at 1380 nm on 4” wafers with more than 85% area yield at die level. These VCSELs are based on Indium Phosphide material and could pave the way for future applications in consumer, automotive and medical photonic integrated circuits (PICs), not to mention in quantum photonic integrated circuits (QPICs) and optical communication applications.
It is in this context that Ezgi Dogmus, Pierrick Boulay, and Ali Jaffal from Yole Intelligence, part of the Yole Group, had the chance to discuss with Berthold Schmidt, CEO at TRUMPF Photonic Components, to learn more about this innovation and its future applications.
Pierrick Boulay (PB): Berthold, could you introduce yourself to our readers?
Berthold Schmidt (BS): I am the CEO of TRUMPF Photonic Components, which is a subsidiary of the TRUMPF Group. I have been with Photonic Components for almost three years and with the TRUMPF Group for ten years now.
Before my time as CEO, I was head of the Corporate Research Group, always trying to drive innovation from within. Additional steps have been that of CEO in the US for the semiconductor laser division, where we produce our EELs and that of CTO of the laser division back at the headquarters in Germany. This gives me a broader view within the organization. Before joining TRUMPF, I worked for Uniphase / Nortel Networks which is now owned by Coherent. So, my experience with the photonics industry goes back more than 25 years.
PB. So, let’s get started and dig a little bit more into TRUMPF’s activities. We understand that TRUMPF is a major player in the [Gallium Arsenide] GaAs photonic business and that it has recently expanded into the InP photonics business.
What was the reason behind this expansion?
BS: At TRUMPF, we consider ourselves as being a leader in GaAs, and we continue to drive all the innovation there. We have experience in GaAs, not only with VCSELs but also with edge emitter laser devices. The original motivation for the latter was to have a vertically integrated supply chain for pumps and for our disk and fiber lasers. This is important for our independence and market leadership, so we continue to drive this.
It follows that we have a similar approach to VCSELs. We have observed that there is an increase in demand in the market for long wavelengths and a mix of materials. There are obstacles: It is always a challenge when there is no volume manufacturing process for a technology. Because of this, we didn’t reinvent long wavelength VCSELs, but it was important to establish a volume-capable, high yielding process. We started with wavelengths of around 1380 nm, because there is a lot of interest for them in the market, but we have started addressing 1310 nm, 1450 nm and other wavelengths as well. The capability for the InP technology has gone up to 2µm. But initially, the idea was to demonstrate a high-volume-capable, industrial-grade process.
Ali Jaffal (AJ): We see decreases in the notch size in smartphones. Is this integration of sensors under displays a driver for InP use in smartphones?
BS: There has been a drive in the development of OLED displays for the smartphone and handheld market for some time now. It is a fundamental aim to have under-display sensors. There are naturally multiple paths to that; for example, we have just released a VCSEL that is highly polarised to serve as a sub-OLED display sensor, although this technology operates at a 940 nm wavelength. However, the transmission window of OLED displays in the longer wavelength regime is wide – 1380 nm-1450 nm – and this 70 nm window provides a lot of opportunities in multiplexing light sources simply and with lower-cost technologies. There is a drive to have components at this longer wavelength available and this is what we are trying to bring to the market. Due to their better eye safety, you can also allow higher output powers at this wavelength, which might expand the use of these devices for longer distance applications.
AJ: This year we have seen [InP] indium phosphide materials used in the iPhone 14 Pro with its proximity sensors.
Do you see applications other than under-display 3D sensing? Do you see it, for example, being used in rear facing lidar in smartphones for AR, VR applications?
BS: If you develop devices at such wavelengths with the right characteristics, I don’t see any obstacles for this technology from the emitter perspective. I think the challenge would be, and this is something that is often not addressed, on the sensor side. The market also needs SPAD arrays and other sensors with the right quantum efficiency at the right cost levels. That is one of the trickier technologies which need to be provided as well. We are not working on both fronts in parallel; we supply VCSEL chips and VCSEL arrays. This must be a joint effort, where TRUMPF plays its part. But the development takes time – it took more than a decade to develop them using short-wavelength technology.
AJ: I have a question about front-end equipment.
You have demonstrated InP VCSELs on 4″ InP wafers, is there any plan for expansion to 6″ InP in the long term? And what could drive this expansion?
GaAs and InP share the same processing tools, so will TRUMPF need to invest more with InP VCSELs in terms of front-end equipment? And is it possible to do InP epitaxy internally at TRUMPF?
BS: It was always the aim at TRUMPF to run this process on standard processing equipment that is, for example, installed in our German line. There is a new technology platform that operates at 4-inch today but has also 6-inch capability.
To be frank, with the slight slowdown in the market, jumping directly to 6-inch is probably questionable, at least in the next one to two years. We are quite comfortable with this because it gives us flexibility and fast throughput for our development at the 4-inch level. Furthermore, InP material at good wafer quality is widely available only at 4-inch at the moment.
To your question on the front-end equipment, we naturally need to be thorough in running GaAs and InP in parallel on certain equipment due to cross contamination. We don’t do that; certain things are totally separate. Other technologies, however, like passivation and metallization, can have dual source equipment. With regards to the MOCVD or other types of epitaxial reactors, it was very important to us that we only use processes that can run on industrial-grade equipment. Some long-wavelength specialists have special reactors, but we wanted to run the whole thing on industrial-grade MOCVD reactors. We therefore have flexibility to do that in-house as well as using external partners. This is important when you aim to go into high-volume production because you don’t want to depend on special-purpose equipment. You want to have multiple sources to meet a potential volume demand.
AJ: You have demonstrated in your first VCSELs with 85% yield, which was a really good result for the first attempt. However, higher yields are needed to address the consumer, high volume markets.
What are the challenges involved in exceeding 90% yields with InP?
BS: In the first phase of product introduction, with yields of 85%, everybody will tell you that the next 5% to 10% is just continuous optimization and engineering work. It has a lot to do with the agreement on optical specs and characteristics with a customer, the binning, and the exclusion zones. These are discussions you would normally have when you are introducing a product into the market, anyway. Briefly put, I see that 90% yield is absolutely achievable with this technology.
AJ: That’s great to hear. At Yole Intelligence, we know the bottleneck in the indium phosphide edge emitters is the low yield, which drives up the cost. How can we compare the InP VCSELs cost with InP EEL cost for this specific application?
BS: The only InP-based EEL I ran was in my PhD thesis. So, I can’t say too much about that. But if I refer to the VCSEL and EEL story from the GaAs world. The last 20 years have proven VCSELs to be superior when it comes to miniaturised laser devices at moderate output powers, at a good cost level and therefore at high yield. VCSEL technology has proven this repeatedly. So, why would the story be different at a slightly different wavelength? Besides, we have seen all the material costs come down in the long wavelength range, which makes this absolutely feasible.
AJ: You said earlier that with the [economic] slowdown, 6-inch InP will take one or two more years to get into the market.
Now that it has been demonstrated on 4-inch InP VCSELs, do you plan in the long term to go to 6-inches and what could drive this expansion in InP?
BS: We are already looking into 6-inch solutions, but I am being a little cautious to offer this to the industry right now, knowing the material qualities and cost of substrates. However, we are already exploring this expansion. Previously InP was based on a 3-inch solution, so scaling up the process to 4-inch, with good stability and an industrial-grade process, was a significant step forward.
What can drive the expansion? Maybe aiming for larger arrays. If customers want sensors, not just in handheld and electronics, but in LiDAR and automotive applications, it requires a larger volume of devices with a larger chip size, so scaling up to 6-inch is a good strategy. We have already demonstrated that we can easily expand from a single emitter to arrays, with the same power levels per emitter on a chip. Expanding long-wavelength technology over VCSEL arrays may also be the way for some industries, in which case we will need to expand the wafer space, and scale up the process to 6-inch.
AJ: That is interesting. We always thought it would be the smartphone business, but indeed going from single-die size to arrays justifies the move. There will also be InP photonic integrated circuits (PICs) for datacom-telecom applications, which could justify going to 6-inch InP.
Even though InP VCSELs at TRUMPF are only in the demonstration phase, may I also ask what your target revenue or market share for the next five years from InP will be?
BS: We are not communicating specific financial figures from TRUMPF’s subsidiaries, but it is true that it is our aim to establish this technology. Having more than two decades in this industry, I know that some technologies need time to evolve. If I had an industrial-grade 6-inch process tomorrow, I would also need the right partners and demand levels to fill the fab. We have to make sure that the sensors, the packaging, and the driver side – in other words, “the whole industry” – is ready to integrate long-wavelength VCSEL technology.
TRUMPF is an innovation-driven company and we continue to drive this long-wavelength sector. We have the InP building block, which, within the right environment, will develop to be the same size business as our GaAs short-wavelength one in the next three to five years.
We want to have our first products in 2025. It takes a long time for qualification and alignment with customers, but we see this as feasible. There is also, as you said, datacom applications for long-wavelength VCSELs and requests for PICs with detailed specifications. We have two building blocks now for photonic components, consumer sensing and datacom on GaAs where TRUMPF is already established as one of the key suppliers, with 25G, 50G and soon 100Gbit VCSELs, and with 1310 nm on InP we see a lot of opportunities as well.
PB: In this discussion, we are speaking about long-wavelength VCSELs. Some companies are working on dilute nitride to make long-wavelength VCSELs.
Why have you chosen to work on InP-based VCSELs and not dilute-nitride VCSELs?
BS: We have been supporting dilute-nitride VCSELs for around 15 years and we have learned the hard way that, when going significantly beyond 1280nm, it becomes more complicated to establish a stable, repeatable process. Going into the even longer wavelength range, we looked at both process options and decided that we would rather accept the engineering challenge of scaling up an InP-based process to 4-inch than spending another four or five years on material research.
Ezgi Dogmus (ED): How do you see the competitiveness in the supply chain for the consumer market?
BS: The supply chain is stuck at a mid-point. Trying an entry level solution has yield challenges and process complexities, while companies working with special equipment and special technologies are not capable of reaching higher volumes on industrial-grade equipment. And then there is the material challenge for dilute nitride. Taking all that into account, we took the decision to offer an alternative path based on InP technology. We know all the material characteristics and were able to establish it from the beginning at 4-inch with a good yield.
ED: Do you think we will see others entering the market, too?
BS: Being currently a key supplier helps, and we are filing IP on special steps of course, but even if we had other participants in the market, it would not be harmful because it develops more momentum establishing the technology. We are also filing a lot of adjacent functionalities and technologies, so it is a very interesting time right now for us to drive this technology forward.
ED: Still with the consumer market, you said in the next three to five years that InP revenues could be like those of GaAs.
Is this only in the consumer market, or do you think InP will replace GaAs, or could both technologies coexist?
BS: I think most functionalities that are currently realized with GaAs-based devices can be replicated with new InP technology. Having a broader wavelength regime for smartphones, that require a whole variety of sensors means it is feasible that for the next 10 to 15 years we will have both. The coexistence of both, GaAs and InP technologies allows us to separate functionalities and make sensor technologies more cost-efficient in smartphones.
ED: You also spoke of the slowdown in the consumer market. We are already seeing the effects of Covid on the supply chain.
Do you think we can recover on a global level?
BS: There are always cycles in the consumer market. We still have Covid restrictions in Asia and a stock increase in the supply chain from the last two to three years, but I think things will slowly level off.
We are naturally sensitive at TRUMPF Photonic Components, but as a TRUMPF organization we see and believe in the value of VCSEL and photonic sensor technology. We therefore see it as an opportunity to establish ourselves and to win more market share, not only with new technologies, but also with more established technologies with new, possibly updated products due to the change in the supplier environment.
AJ: Just one last thing, would you like to add any closing thoughts for our readers?
BS: Sure. I think you can see my fascination for photonics. I have been in the industry now for almost 30 years and I’ve always been committed and believed in photonic components, whether for industrial, telecom or datacom applications.
TRUMPF really believes that innovation will lead to a continuous increase in our business, so we are fully focused on that. We are not doing everything in the VCSEL world, we’re rather trying to be distinctive in our efforts. Therefore, we have intensive communication with our customers, and we are going to be consistent in our development effort. There are a number of opportunities and it’s really exciting to be in this industry right now. Overall we are driving our roadmaps to make the VCSEL components smarter by adding additional features such as integrated optics, multi-junction technology and polarization locking.
Think about it – we are now making VCSELs that will drive a quantum sensor to control a satellite with a launch date in 2027! How cool is that?
I have a childlike excitement for this technology and I enjoy going to the office every morning to drive this technology forward.
Pierrick Boulay
Senior Technology and Market Analyst, Photonics and Sensing Division, Yole Intelligence (part of Yole Group)
Ezgi Dogmus, PhD.
Team Lead Analyst, Compound Semiconductor and Emerging Substrates, Power and Wireless Division, Yole Intelligence (part of Yole Group)
Ali Jaffal, PhD.
Technology and Market Analyst, Compound Semiconductor and Emerging Substrates, Power and Wireless Division, Yole Intelligence (part of Yole Group)
