The Lowdown On Substrates
Asif Anwar, Director of the GaAs and Compound Semiconductors Technologies (GaAs) Program at Strategy Analytics, is the author of several market reports that include: GaAs Industry Forecast 2008-2013 and InP Bulk Substrates Market 2007 2012. Here he provides an insight into the trends in the GaAs and InP substrate markets.
Q How has the market for semi-insulating GaAs substrates, which are the starting point for making power amplifiers for handsets, been affected by the current economic storm?
A In the final quarter of 2008 handset manufacturers stopped buying power amplifiers, and the power amplifier manufacturers stopped buying substrates. There was also very minimal business in Q1 of 2009, but there was an upturn in the second quarter, and the second half of the year was actually quite strong forsubstrate manufacturers.
In terms of overall volume – and we haven’t got actual survey results yet - our modelling suggests a 2 percent increase in volume year-on-year. So even in a down year, manufacturers still managed to hold their own, if nothing else.
Q How do you see revenue in this substrate sector evolving over the next five years?
A We are forecasting a 5 percent compound annual growth rate in terms of revenue. The key driver remains the multimode, multi-band handset. Over the next three to five years we will see the real emergence of 3G, 3.5G (HSPA) and LTE.
Q Who are the big substrate manufacturers in this market?
AThe four key manufacturers are Sumitomo Electric, Freiberger, AXT and Hitachi Cable.
Q Is AXT continuing to grab market share from its competitors?
A Yes. Over the past couple of years they have managed to come back from being a company that was out of the running, and re-establish themselves as a viable supplier. Our most recent numbers suggest that they have overtaken Hitachi Cable, in terms of commercial supply of substrates. Obviously the situation is cloudy if you take the overall market, because companies like Hitachi Cable also have captive demand. But in 2008 the rankings in the commercial market were Freiberger, Sumitomo Electric, AXT and then Hitachi Cable.
Q Different technologies are used for substrate manufacturing. Is a winning approach emerging?
A The industry has now settled on two main manufacturing technologies – LEC (liquid encapsulation Czochralski), and VGF (vertical gradient freeze), which is also referred to as VB (vertical Bridgman). Sumitomo Electric is almost exclusively using VB, AXT is exclusively using VGF, Freiberger has a mix of both, and Hitachi Cable is almost exclusively using LEC. So what has happened is that VGF or VB technologies have taken control of the market. This has been driven by the use of VGF for the production of HBT technologies, which form the bedrock for power amplifier production.
PHEMT manufacturers typically use LEC-based substrates. PHEMT is still a volume market, especially for the switch side. For higher frequency applications, for example, such as point-to-point mm-wave radios, PHEMT technology is used for power amplifier functions.
Feature-rich hand-held wireless devices, such as Research in Motion’s range of Blackberry products, have a high content of GaAs electronics. This is helping to spur the growth of GaAs substrate sales
Q What size of semi-insulating GaAs substrate is used for the majority of manufacturing?
A 6-inch is the dominant diameter now by far now – it started to take over the market in 2005. We’ve seen some of the larger players move to this size over the last few years. RFMD has converted to 6-inch and bought the Filtronic fab that has a 6-inch line, and Skyworks has done a partial conversion to 6-inch. You also have WIN Semiconductor, TriQuint, and Anadigics high-volumemanufacturing on 6-inch.
There are still players working on 4-inch platforms in Europe, Japan, and the US. But overall, 4 inch is only about 27 percent of the total market, and the bulk of the remainder is 6-inch.
Q Are substrate prices falling?
A There is pricing pressure across the whole supply chain, including substrate vendors. They are seeing average price declines of around 4 to 5 percent year-onyear for the pricing of their substrates. The actual pricing depends on individual relationships, the size of the contract, and how much volume companies are buying.
Q Do you think that the GaAs chipmakers will move to a 200 mm platform by the end of this decade?
If we see 8-inch manufacturing, it’s probably going to come in through the foundry route as opposed to the IDMs (independent device manufacturers). It makes sense for them – the larger the diameter that they can get and produce, the more cost effective it is for them, whereas for an IDM, they have to invest significantly in new equipment, new infrastructure, and then a 5-10 year cycle that equipment has to be amortized over and depreciated. I don’t see any of the major GaAs manufacturers, the IDMs, making that kind of investment.
Looking forward, the multi-mode, multi-band approach might be replaced by broadband power amplifiers. If that was the case the actual potential demand for power amplifiers, in terms of volume, would start to drop, and have a knock-on effect on how many substrates are needed.
Q How would you describe the current semiconducting GaAs substrate market that serves LED manufacturers?
A It is quite healthy. The LED market has seen resurgence, but the thing to remember is that LEDs aren’t exclusively based around GaAs technology. The primary technology coming into play right now is GaNbased LEDs. But having said that, we expect to see about a 4 percent compound annual growth rate for semiconducting GaAs substrates over the next four to five years. And LED manufacturing is going to be about 30-40 percent of that demand.
Q What applications will drive the semiconducting GaAs market over the next few years?
In terms of volume demand, the market is going to be driven by LEDs. And what’s going to drive that is the flat panel display market. Your mainstream TVs will have edgelit or back-lit white, nitride-based LEDs, but then you have premium models that have red, green and blue lighting. The red LEDs are grown on GaAs substrates.
We have also modeled in a growth market for solar. What you have to bare in mind is that the bulk of III-V concentrating photovoltaic production has typically been on germanium substrates. However, there are efforts to grow III-V solar cells on GaAs as well. There will be some impressive compound annual growth rates, but I would not expect solar to represent more than 10 percent of the market in the next four or five years, as far as volume is concerned.
Q Is the manufacturing leader board for semiconducting GaAs similar to that for semiinsulating GaAs?
A We see some of the same names, but the market tends to be dominated by the Japanese companies, such as Sumitomo Electric, Mitsubishi Chemical, and Hitachi Cable – those are the market leaders. Other Japanese companies include Dowa. AXT and Freiberger are behind the larger Japanese suppliers.
Q What is the dominant substrate size for sales of semiconducting GaAs?
A It is still 2-inch, but you are seeing larger diameters moving forwards. Right now we see 2-inch representing about 40 percent, but there has been a move towards larger diameter substrates over the past couple of years, and that growth will continue. We expect that by the 2013-2014 timeframe, 3-inch and 4-inch will represent over 70 percent of semiconducting substrate demand.
Q Moving on to InP, how would you describe this market?
A It is in a fairly healthy state. There is going to be strong growth for semiconducting InP bulk substrates, resulting from the fiber-optics market. On the optoelectronic side, InP demand is driven by photodetectors and transmission lasers.
On the semi-insulating side, it’s the same market driver – fiber optics. However, the overall volumes are a bit lower, because the devices manufactured on this substrate are analog ICs such as TIAs (trans-impedance amplifiers) and laser drivers. The penetration of that technology is dependent on 40 G and 100G growth moving forwards. And those are still emerging markets, so that is why the microelectronics market is a bit slower. But it is still going to see significant growth. We are forecasting for InP, in terms of dollar value, a compound annual growth rate of over 47 percent over the next four to five years.
Q Who are the leading manufacturers of InP substrates?
A The Japanese are dominating that segment. Companies such as Sumitomo Electric and Nikko Materials are in the lead, but AXT is also a significant supplier of InP. Other providers include the French company InPACT. There is also IQE subsidiary Wafer Technology, though they now seem to be focusing on more esoteric substrates.
Philippe Roussel from Yole Développement has been taking a close look at the SiC substrate market for several years. He has authored several reports, including SiC market 2010: a 10-year projection. He offers his take on the SiC substrate market.
Q How would you describe the state of the current SiC substrate market? Has it been rocked by recession?
A I would say that the market was quite flat in 2009 compared to 2008. In 2009 the market for all SiC substrates, including semi-insulating and all the R&D substrates, was $48 million, and in 2007 it was $47 million. In terms of volumes we saw an increase in sales, but there was also an erosion of prices.
Q The SiC market can be sub-divided into semiconducting and semi-insulating material, and 4H and 6H polytypes. What are the types of devices grown on each of these platforms?
A To be frank, we should no more speak of 4H and 6H polytypes, because most of the products today are made on 4H. Even Cree is manufacturing everything on 4H SiC.
So the most important divide is between semi-insulating verses semiconducting SiC. Semi-insulating SiC is only used for RF applications. Conductive substrates are used for LEDs, power electronics, and MEMs, but the later application is a very small market.
Q Some manufacturers of SiC substrates use their material for manufacturing their own chips. For example, Cree produces LEDs on its own substrates. In terms of volume, how does the total internal market compare in size to that of the merchant market?
A We have tried to simulate Cree’s captive SiC consumption and translate it into a value. In 2008 this was in the range of $235 million. The merchant market was only $47 million, so peanuts in comparison.
But it was very tricky to get a real number, because it is hard to estimate the internal costs of Cree wafers. Our actual figure is probably wrong, but Cree’s captive market will probably be somewhere within the range of two-tothree hundred million dollars.
The development of a reliable SiC transistor could lead to the deployment of SiC electronics in hybrid electric vehicles. If this happens, it will fuel an increase in SiC substrate sales
Q How quickly will SiC market revenues grow over the next few years, and what will drive this growth?
A The substrate business is linked to the device business, and it will be boosted by the launch of the transistor. We believe that the MOSFET will start volume consumption somewhere between 2013 and 2014.
This is the inflection point for us. If the transistor is introduced in the market place in 2010 it will take one to three years before the system makers will implement it, validate it and make all qualification processes. We expect the substrate business to reach $350 million by 2019.
Q Who are the leading manufacturers of SiC substrates?
A It’s no surprise that Cree is leading the business. But their market share is decreasing year after year. In 2008, Cree had a little bit more than half the market share of the merchant market.
The two big challengers were II-VI and SiCrystal. In 2010 we are expecting to see Dow Corning take-off, and this will further decrease Cree’s market share.
Q Are SiC substrate makers using different techniques to produce their material?
A Most of them are using sublimation from powder. Only Norstel is using high-temperature CVD. Sumitomo has an interesting liquid phase epitaxy technology, but it’s not in production.
If we consider the market situation, then 95 percent of material is made using sublimation. Most of the newcomers are starting with sublimation, so the dominance of this technique will probably continue.
Q Are there still issues of SiC material quality, or is this now behind us?
A I think that only Cree is able to produce zero micropipe material. Their challengers are working on that. There are very good results with dislocation densities of less than one per square centimeter. So the dislocation density is under control. The next challenge is the Basal plane dislocations. If these dislocations are in the raw material, then they can degrade the quality of the epilayers.
Q How do sales of 2-inch, 3-inch and 4-inch substrates compare?
AThe split is currently moving from 3-inch to 4-inch. In power electronics production, in 2009, it was 80 percent on 4-inch, versus 20 percent on 3-inch and 2- inch. And 2-inch is less than 2 percent now.
6-inch is expected by a lot of players. Its introduction will depend on Cree, because it has announced a 6-inch substrate for 2010. But nobody has seen any real 6-inch material yet. Due to the qualification time – one to oneand- a-half years – we don’t expect to see 6-inch in production until 2012. Its success depends on the price.
Q I’ve heard that it’s more expensive to make products on 4-inch substrates than 3-inch equivalents?
A That’s not true. It was true during the introduction of the first samples, but not any more. The cost-persquare- inch is now less with 4-inch material if you buy in volume - an order of 50 to 100 substrates or more. But an R&D lab will pay more by a factor of four or five.
Q What is the trend of the average selling price of SiC substrates for high volume orders?
A It’s a bit like in the PC world. You always pay the same price, but you get something better every year. The typical price for a volume order from a large substrate manufacturer is $1000-$1200 per substrate. This was the price for 2-inch SiC seven or eight years ago, for 3-inch four years ago, and now for 4-inch.
QDo you see any changes in the number of players in the SiC market over the next few years?
A We don’t see a lot of incomers. The market size is not big enough to sustain any new companies to enter. And I expect one or two established SiC substrate manufacturers to disappear, because they have to live until 2013 with quite low-volume products. I don’t think that there is enough room for all of them, so the weaker will disappear.
Strategies Unlimited is well known for it analysis of the LED market. In addition, it covers the GaN substrate market, with reports co-written by Bob Steele and consultant Hank Rodeen. These authors, which wrote GaN Substrates: Performance Comparisons and Market Assessment – 2009, offer a view on the GaN market.
Q How would you describe the state of the GaN substrate market?
A In 2008 it was just over $100 million. We haven’t tabulated the number for 2009, but we don’t think that market would have been much bigger. Demand is driven mainly by lasers for Blu-ray players and the Sony PlayStation 3. The Nintendo Wii has been so popular, and PlayStation 3 sales may have suffered from a combination of this and the recession.
GaN substrates are predominantly used to make 405 nm lasers for reading Blu-ray discs. This technology features in the Sony Plastation3
Q How will the market evolve over the next few years?
A We forecast a compound annual growth rate for sales of about 10 percent through to 2013, and by then the market will be worth about $190 million. The unit volume will increase by about 20 percent per year, but there will be attrition in price. The market will be driven mainly by lasers, but there might be some small demand for electronic devices. We don’t foresee any demand for LEDs. We’ve looked at this carefully, and talked to people, researchers are getting outstanding lab results on sapphire and SiC. Cree announced 186 lm/W [at 350 mA] from a lab device in the last 6 months. The price of GaN is higher than sapphire, or even SiC, and it would be very difficult to envisage a hike in LED performance that’s commensurate with that difference in price.
Q Who are the big players in the GaN substrate market?
A The leader for GaN sales is Sumitomo Electric, and they have about 80 percent of the business. They are the major exponent of HVPE, which involves the growth of thick layers of GaN onto another substrate. They were using GaAs to put GaN on, but we had heard that they were changing. They are shipping volume, and most of their volume is 2-inch.
Others producing HVPE-grown GaN substrates are: Kyma Technologies; TDI, which is now part of Oxford Instruments; Lumilog; Hitachi Cable; Mitsubishi Chemical; Samsung Corning; and TopGaN. A lot of GaN that’s produced is sold as rectangles, or squares that are typically 15 mm by 15 mm, or 25 mm by 25 mm in size. But these are difficult to process into volume, and anyone going into business should target a 2-inch disc.
Q The US and Europe are way behind Japan in terms of GaN producers. Will this change?
A We don’t have any reason to think it will. In Europe and the US there aren’t any large companies involved in GaN. When the big companies – Sumitomo, Mitsubishi, and Hitachi Cable – want to be serious in the business, they have the resources and the needs to do it. Most of the others are start-ups, or small operations. One exception is the Korean firm Samsung Corning, which has enough resources to strongly pursue this market if it wants to.
The other factor is who makes the lasers. It’s Japanese companies, mainly Sony, but Nichia and a few others. Historically speaking, Japanese companies have relied on other Japanese companies to be their suppliers. So our guess is that Japan will dominates GaN manufacturing for the foreseeable future, unless someone else comes up with something that is so much better. However, there are some larger companies in the US with substantial resources and infrastructure that are doing R&D on GaN substrates under the radar. They may well be able to compete with the major Japanese players in the future.
Q What other methods are being used to produce GaN substrates?
A In addition to HVPE, which can give defect densities close to 104 or 105 per square centimeter, you have ammonothermal growth. There are three suppliers – Ammono, Mitsubishi Chemical, and Sixpoint Materials, which is a spin-off out of the University of California, Santa Barbara - but none of them have very large volume. The defect densities depend on the type of seed, and they run in the range of 104 or 105 per square centimeter.
Then you have the high-temperature, high-pressure (HTHP) approach, which is the ultimate from the standpoint of low defect density. But it’s also the ultimate in terms of price. We don’t think any crystals are made, except by happenstance, that are any bigger than 0.6 or 0.7 inches in diameter.
The leading companies, as far as capability is concerned, are: TopGaN; TDI; and A. F. Ioffe Physical Technical Institute in Russia, which is a research organization that ships stuff out for evaluation.
If you use a HPHT approach with a HPHT seed that starts out with a defect density of 102 per square centimetre, then you can produce some pretty good material. It’s an approach that a lot of people would take. They would get the best material that they could and get the best price that they could on it, while realizing that it’s not practical for manufacture until the price came down.
But they could demonstrate their capability, carry out device characterization, and generate a lot of meaningful data from the standpoint of performance. They will say: here’s where we are going to be – all we need is a good price and availability.
Q Although GaN substrates are far more expensive than silicon, sapphire and even SiC, they do provide a superior platform for making low defect density epiwafers. Do they also improve device lifetimes?
A The reliability issues for devices on GaN substrates may well become apparent on devices operating at high power and high temperature. However, there has not been sufficient testing and evaluation at this stage to know just how big a difference in reliability that GaN substrates might make.
Cree is by far the biggest seller of SiC substrates, and it also uses this material as the basis for its own LED production
Q I’ve heard that a typical price for a piece of 2-inch GaN is $5000. Is that correct?
A$5000 is the number we used in 2008, but we’ve heard that the laser price has been pushed down, and this has had a knock-on effect on substrate prices. Lasers cost $15-20 in 2008, and we’ve heard that they have been as low as $6-8 in the past year. However, Sony is the dominant supplier of lasers, and it supplies itself. In Japan you have very vertically integrated companies, and money is made at the system level. So the component side gets squeezed to provide as low a price as possible for the system people. On the other hand, the component business has to make money too. I suspect that the prices have come down, because substrates are a huge percentage of the cost of the laser.
Q Are any companies developing 3-inch GaN substrates?
A We don’t believe so. At this point we don’t think that there is a huge driving force to go to 3-inch, because you can, in principle, get 5,000 lasers off a two-inch wafer. Laser volumes are just 15-20 million per year, which is very low to compared to LEDs, which are over 50 billion per year. However, even though there is not a large demand at the moment, Sumitomo and others could fairly readily scale up to 3-inch if necessary.