Germany-based IP licensing company ALLOS Semiconductors sees itself as the leading provider of GaN-on-Si wafer technology for LED, power and RF applications and claims to have the best performance in the market with respect to essential properties like crystal quality, uniformity, and low wafer bow.
ledcax had a chance to sit down face to face with Burkhard Slischka, CEO of ALLOS, to talk about the company’s proprietary technology and his vision for the future of micro LED.(Image: ALLOS Semiconductors)
Q: Burkhard, glad to have you here. Let’s start with the question many readers have: Why GaN-on-Si? Almost all conventional LED production is on sapphire. Why should that be different for micro LEDs?
A: For a successful production of micro LEDs fundamental changes are needed. New device designs and novel manufacturing techniques are to be used and – most importantly – production yield has to be brought to a level far better than anything the LED industry has seen. For all these changes GaN-on-Si is a key enabler and saves cost.
Q: So moving to GaN-on-Si is all about better yield and cost in your opinion?
A: Yes, exactly. And to understand this we need to look at the entire value chain from growing the epiwafer until the micro LED chips are successfully build into the display.
Q: On this value chain, where do you see the biggest impact by using GaN-on-Si?
A: We see three factors along the entire chain: the cost advantages of bigger wafer diameter, the advantages of silicon for thin film processing and the huge influence of wafer uniformity on yield.
Q: Let’s start with uniformity. End of 2017 Veeco and ALLOS made a joint press release claiming that your technology has achieved the best uniformity published for 200 mm GaN-on-Si. How does this compares with GaN-on-sapphire?
A: We showed a standard deviation for emission wavelength of below 1 nm. Other uniformity values like thickness are even better. These values for 200 mm GaN-on-Si compare very well with 150 mm GaN-on-sapphire. Especially as our technology has very good wafer-to-wafer reproducibility.
“The next target is to have 99 % within a +/- 1 nm bin”
Q: Do you see room for further improvement?
A: Yes, we certainly do. We achieved these results after only four weeks project work on Veeco’s Propel reactor. So there is still a lot of room for improvement. The next target is to have 99 % of the wafer area within a +/- 1 nm bin.
Q: Are you saying that you can get the same or even a much better uniformity on 200 mm GaN-on-Si compared to what others can achieve on 150 mm GaN-on-sapphire?
A: Correct, same uniformity, but on a wafer area almost twice as big.
Q: It seems having excellent uniformity is of much higher importance than for conventional LEDs. Why is that?
A: Uniformity is the essential enabler for yield. For every display you need millions of micro LED chips with precisely the same wavelength. Sorting LED chips before assembly or any pixel repairing afterwards is driving up the cost of the display. Therefore, this issue must be solved right at the beginning with very uniform epiwafer material, otherwise micro LED displays will not be cost competitive.
Q: In this context, what is more important in your opinion, having the best wafer uniformity or the lowest epi cost per wafer area?
A: Good question. Looking at one of the two items alone is short-sighted. A wafer with better uniformity drives up yield in all following manufacturing steps. So even if the epi cost would be higher, it can easily be over-compensated by the cost savings resulting from the better yield.
“Using 200 mm CMOS saves 45 %”
Q: Which leads us to the second factor you mentioned before, the cost advantages of bigger wafer diameter. I understand that moving to larger wafer diameters is driving down cost per chip. But wouldn’t that be the same for 200 mm GaN-on-sapphire?
A: Yes, in principle one should think so. Except that there are three problems for sapphire: 1) The high cost of 200 mm sapphire substrates, 2) It might be impossible to achieve the needed wafer uniformity and 3) to get to sufficient manufacturing yield due to strong wafer bow. But even more important is another advantage of GaN-on-Si: You can use existing 200 mm CMOS processing capacity and expect a cost saving per area of 45 %. At the same time, it offers the high processing yield needed for micro LEDs which conventional LED sites could match only after massive CAPEX and skill upgrades.
Q: And the 45 % you mentioned is only from chip processing?
A: Right, there are more contributions coming from higher yield at substrate removal and bonding and better area economics at mass transfer.
Q: There is one more technical argument I would like to understand. How can you achieve the uniformity target on 200 mm on silicon while it is so difficult on sapphire as you are saying?
A: Our GaN-on-Si technology comes with a unique strain engineering technology. This allows us to achieve two things at the same time: We can have very good temperature homogeneity during growth – enabling the uniformity we talked about before – and a flat and crack-free wafer after cooling down. Usually these two items are seen as trade-offs, but we can provide both. And by the way, we are very confident to scale this even to 300 mm.
“Flat and crack-free 300 mm is possible”
Q: 300 mm GaN-on-Si?
A: Yes, 300 mm is possible. Unfortunately, there is no industry-grade MOCVD equipment for 300 mm available yet. But we are optimistic that this will change and we can demonstrate the first flat and crack-free 300 mm GaN-on-Si epiwafer soon.
Q: Very interesting! I got the cost advantages of GaN-on-Si for micro LEDs you just described. But what about performance differences to GaN-on-sapphire? From what we hear there are concerns in the industry about brightness and light efficiency on GaN-on-Si.
A: In fact, it is very difficult to achieve good enough crystal quality with GaN-on-Si and this damages the LED performance as you mentioned. We overcame this limitation and have the best crystal quality on GaN-on-Si in the industry. Actually, our top layer defect density is only 2 x 108 cm-² TDD which is on pair with the value on sapphire.
Q: What does that mean in terms of LED performance?
A: It means that once we have licensed and transferred our technology to a customer their existing light emitting layer as they grow them today on sapphire and our GaN buffer on silicon will be integrated into one recipe. This results in a GaN-on-Si wafer with all the superior properties we have already talked about and also with the same light emitting structure and performance our customers have today on sapphire.
“Same performance as on sapphire”
Q: Let’s move on to the business side. How will the supply chain look like for micro LEDs?
A: The target applications of the large device makers will determine development and business decisions. Either the device makers for smartphones, TVs or other kinds of gadgets will follow Apple’s example and built up micro LED capabilities in-house. Or they incentivize strategic partners to provide solutions for parts along the value chain. But in either case they need to coordinate technology, yield and cost along the value chain and finance to a certain degree the development. Just to wait for some supplier to provide the end-to-end solution for micro LEDs does not look like a viable strategy to me.
Q: How will that impact ALLOS’ future business strategy?
A: ALLOS has strong GaN-on-Si technologies for three markets, LED, power and RF. We see that these three markets require different technology and business set-ups and that we also need to apply different strategies and strategic partnerships in these three markets.
Q: What are your objectives in micro LEDs in the next two years?
A: In micro LEDs we are engaged with big corporations and start-ups alike to further develop the technology. We see that we are in a strong position based on the excellent performance we have already today and our solid IP portfolio. Now it is equally important to have the other right partners along the value chain and the needed funds to develop micro LEDs into production readiness.
Q: If I were one of the large device makers you mentioned before and would lead and sponsor such development I would demand to get exclusive rights on the developed technology too…
A: Yes, we hear this argument a lot these days and with the right set-up this might be a win-win solution.
Q: Thanks for talking to us, Burkhard. See you in July at our micro LED forum!
A: I am looking forward to it!总部设在德国的IP授权公司Aules半导体公司将自己视为GaN领先的LED晶片、电源和射频应用技术供应商，并声称其在市场上具有最佳性能，如晶体质量、均匀性和低晶片弓。.莱德内尔有机会与OLOS首席执行官Burkhard Slischka面对面讨论公司的专有技术和他对未来微型LED的展望。.（图像：半导体）Q：伯克哈德，很高兴你在这里。.让我们从许多读者的问题开始：为什么在赣？几乎所有传统LED生产都是蓝宝石。.为什么微型LED会有所不同呢？答：为了成功生产微型LED，需要进行根本性的改变。.新的设备设计和新的造技术将被使用，最重要的是，生产收益必须达到比LED行业所看到的水平好得多的水平。.对于所有这些变化，GaN on Si是一个关键因素，节约了成本。.问：那么在你看来，移动到GaN上的产量和成本都更好吗？答：是的，确切地说。.为了理解这一点，我们需要从增长外延晶片看整个价值，直到微LED芯片成功地建成显示器。.问：在这个价值上，你看到在GaN上使用GaN的最大影响是什么？答：我们看到了三个因素：整个晶圆直径的成本优势、硅薄膜加工的优势以及晶片均匀性对产量的巨大影响。.问：让我们从均匀性开始。.2017年底，VECO和OLAS联合发布了一份新闻稿，声称你的技术已经达到了200 mm GaN的最佳一致性。.这如何与蓝宝石上的GaN进行比较？A:我们对1 nm以下的发射波长进行了标准偏差。.其他均匀值如厚度甚至更好。.Si上的200 mm GaN的这些值与蓝宝石上的150 mm GaN相比非常好。.特别是由于我们的技术具有很好的晶片到晶片的再现性。.“下一个目标是在+/- 1纳米箱中有99%个”Q：你看到进一步改进的空间了吗？是的，我们当然愿意。.我们在VEECO推进反应器仅四周的项目后就取得了这些成果。.所以还有很大的改进空间。.下一个目标是在+/- 1纳米仓内有99%的晶片面积。.问：你是说你可以在200 mm GaN上得到相同甚至更好的均匀性，相比之下其他150个蓝宝石上的mm GaN能达到什么样的效果呢？答：正确，同样的一致性，但在晶片面积几乎两倍大。.问：它似乎具有优异的均匀性比传统LED具有更高的重要性。.为什么会这样？A:均匀性是产量的基本推动力。.对于每一个显示器，您需要数百万个具有相同波长的微型LED芯片。.在组装之前对LED芯片进行分类或任何像素修（后来）都会抬高显示器的成本。.因此，这个问题必须首先用非常均匀的外延片材料来解决，否则微型LED显示器将不具有成本竞争力。.问：在这种情况下，你认为什么是更重要的，晶圆均匀性最好还是每个晶圆面积的EPI成本最低？好问题.只看这两个项目中的一个是短视的。.在所有后续造步骤中，具有更好均匀性的晶片提高了成品率。.因此，即使EPI成本会更高，它也很容易被过度补偿所带来的成本节约。.“使用200毫米CMOS节省45%”Q：这导致我们提到的第二个因素，更大的晶片直径的成本优势。.我知道，移动到更大的晶片直径正在降低每芯片的成本。.但是对于蓝宝石上的200个mm GaN来说，这不一样吗？答：是的，原则上我们应该这样想。.除了蓝宝石存在三个问题：1）200毫米蓝宝石衬底的高成本，2）可能无法实现所需的晶片均匀性和3）由于晶圆的强弓而获得足够的造收益率。.但更重要的是GaN在Si上的另一个优点：您可以使用现有的200毫米CMOS处理能力，并预期每45%的面积节省成本。.同时，它提供了微型LED所需的高加工率，传统LED站点只能在大规模CAPEX和技能升级之后才能匹配。.问：你提到的45%只来自芯片处理？A:是的，在基质去除和粘方面有较高的贡献，在传质方面具有较好的区域经济性。.问：还有一个我想理解的技术论点。.如何在硅上实现200毫米的均匀性目标，而在蓝宝石上这么难？答：我们的GaN Si技术具有独特的应变工程技术。.这允许我们同时实现两件事：在生长过程中，我们可以有很好的温度均匀性，使我们以前所说的均匀性和冷却后的扁平无裂纹晶片。.通常这两个项目被看作是取舍，但我们可以同时提供这两个项目。.顺便说一下，我们很有信心把这个比例扩大到300毫米。.“平坦和无裂纹300毫米是可能的”Q：300 mm GaN在Si上？答：是的，300毫米是可能的。.不幸的是，没有工业级MOCVD设备300毫米可用。.但我们乐观地认为，这将发生变化，我们可以证第一个平板和无裂纹的300 mm GaN在硅外延片上。.问：非常有趣！我刚才在GaN描述了微LED的成本优势。.但是蓝宝石上的GaN性能差异如何呢？从我们所听到的，业界对GaN上的亮度和光效率的关注。.答：事实上，在Si上用GaN获得足够好的晶体质量是非常困难的，这损害了您提到的LED性能。.我们克服了这种局限性，在工业硅上具有最好的晶体质量。.事实上，我们的顶层缺陷密度仅为2×108×cm-TDD，其值与蓝宝石上的一对值有关。.问：在LED性能方面，这意味着什么？答：这意味着一旦我们将技术授权给客户，将他们现有的发光层在蓝宝石上生长，我们在硅上的GaN缓冲区将被整合成一个配方。.这导致了GaN在Si晶片上具有我们已经谈论过的所有优异性能，并且我们的客户在蓝宝石上具有相同的发光结构和性能。.“与蓝宝石一样的性能”Q：让我们走向商业面.供应如何看起来像微型LED？答：大型设备造商的目标应用将决定发展和业务决策。.无论是智能手机、电视机还是其他类型的设备造商，都将仿效苹果的例子，在内部建立微型LED功能。.或者激励战略合作伙伴为价值上的零部件提供解决方案。.但无论在哪种情况下，它们都需要在价值上协调技术、收益和成本，并在一定程度上促进金融的发展。.仅仅等待一些供应商为微型LED提供端到端的解决方案对我来说不是一个可行的策略。.问：这将如何影响未来的商业战略？A:在GaN的SI技术上拥有强大的三个市场，LED，功率和射频.我们看到，这三个市场需要不同的技术和业务设置，我们还需要在这三个市场应用不同的战略和战略伙伴关系。.问：在未来两年内，微型LED的目标是什么？答：在微型LED中，我们与大公司和初创公司一样，进一步发展这项技术。.我们看到，我们在强大的地位，基于我们今天的出色表现和我们的坚实的知识产权投资组合。.同样重要的是，在价值上还有其他合适的伙伴和需要开发微型LED的资金。.问：如果我是你以前提到过的大型设备造商之一，我将领导和赞助这样的发展，我也会要求在开发的技术上获得专有权……A：是的，我们最近听到了很多关于这个问题的观点，如果正确的话，这可能是一个双赢的解决方案。.问：谢谢你和我们谈话，Burkhard。.七月在我们的微型LED论坛上见！我很期待！来自百度翻译