You are here
Silicon is reaching its limit. What's next?
IF cities today are truly high-tech, why can't we drive electric cars up to Kuala Lumpur on a single charge? Why are we tethered by a cable to a wall socket each time our smartphones run out of juice? Why is information wireless, but not power? The problem, some say, lies with silicon, the raw material of modern power electronics. It's difficult to eke out further efficiency gains from silicon-based power systems, analysts say, because silicon is close to reaching the physical limits of Moore's Law.
In the semiconductor industry, Moore's Law guides innovation cycles. It refers to the expectation that a chip's computing power should double every 2.5 years or so. Advancements in silicon computer chips have kept pace with this rule for the last 55 years, but cost is also a limitation, and it's now more technically difficult to double the number of transistors on a given silicon chip or make its glass-like wafer any thinner.
To keep up with Moore's Law, many traditional silicon power semiconductor companies have been actively developing new materials. One of these materials is gallium nitride, or GaN, a compound semiconductor formed by combining gallium and nitrogen into crystals.
Ng Geok Ing, a professor at Nanyang Technological University (NTU) who conducts GaN research, says: "GaN is poised to revolutionise the semiconductor electronics industry. This is due to its attractive and unique material properties that make it well-suited for high-power, high-efficiency and high-temperature electronics applications."
GaN is more expensive than silicon, which is the most abundant element on earth after oxygen. But GaN more than makes up for this by reducing costs on a system level by up to 20 per cent, says Jim Witham, chief executive of GaN Systems, a Canadian maker of GaN power semiconductors.
Because GaN loses less energy as heat than silicon during power conversion, has higher switching frequencies and can operate at higher maximum temperatures than silicon, it allows engineers to build smaller, faster, more reliable devices without the need for fans or liquid cooling.
"So people can have better performance and lower costs, which is kind of unusual," Mr Witham explains, adding: "Sonnen, the German solar energy storage system company, said they saved 8 per cent on their bill of materials cost by using my parts. They increased their efficiency 4 per cent, so it's 4 per cent more revenue that they can sell out of their energy storage system. So, overall, a 12 per cent swing just by using my parts."
More customers are now exploring the use of GaN for battery charging, 5G telecommunications and automotive applications, says Paul Wiener, GaN Systems' vice president of strategic marketing: "It's forcing everybody to think about how they're going to use GaN in their next designs.
"In the power world, the fundamental building block when you design any one of these things is the power transistor (a switch that governs the flow of electrical current, physically larger than a normal transistor). So you need to have that in your portfolio. Every semiconductor company now needs to have a GaN strategy. It's at that maturity point now and we see that happening."
According to market research firm Yole, sales of GaN power transistors and semiconductors are expected to grow at a compound annual growth rate of 86 per cent between 2018 to 2025, to reach more than US$700 million in 2025.
The catch is that it's not easy to grow GaN crystals on silicon wafers, which is what you want to do if the so-called GaN-on-silicon chips are to be compatible with existing processing lines in standard silicon chip factories.
In fact, very few firms in the world can do it. Singapore's IGSS GaN, or IGaN, is one of the few that can, says Raj Kumar, the startup's chief executive: "Apart from TSMC (Taiwan Semiconductor Manufacturing Company), no other proven foundry has the know-how to build GaN-on-Si epiwafers."
IGaN's proprietary GaN-on-silicon growth recipe is the result of hundreds of millions of dollars of research by various groups in Singapore over a period of 14 years, he points out.
IGaN then spent another six years perfecting the technology it had licensed from the Agency for Science, Technology and Research, before finally bringing it to market last year.
Mr Kumar expects initial production revenue to start rolling in by the end of this year, and accelerate in 2021 and 2020: "The world has spent 20 years or more on GaN-on-silicon and today the traction is beginning to happen. We have more and more players asking to sample our wafers."
Some of IGaN's customers are multi-billion dollar entities, though most are startups. One of them makes quantum sensors that require GaN to be more sensitive.
IGaN is the only firm in Singapore on the GaN train right now, but Prof Ng hopes there will be more: "Singapore's semiconductor industry already has the needed infrastructure, supply chains, and know-how to make GaN technology more affordable in a large manufacturing scale. This is essential to enable high-volume low-cost GaN technology for the mass market."
Singapore's goal to achieve net zero greenhouse gas emissions will also be expedited by GaN's usefulness in clean energy and cutting energy expenses at data centres, one of the biggest big carbon emitters here.
Prof Ng adds: "It is essential that we put in the resources and efforts to enable this technology which can help to fight climate change, besides economic benefits."
Singapore no longer has any local semiconductor giants, and it's difficult for any company built tomorrow to become a leader in mainstream semiconductor technologies.
Mr Kumar says: "If you want to chase Moore's Law, that's going to cost you US$10 billion to put up a state-of-the-art (silicon) wafer fab. It would be too much effort."
Only the big foundries like TSMC and Samsung or state-led firms can afford to play on this bleeding edge.
But in the market for emerging technologies like GaN and other silicon alternatives, startups do have a good shot at being world champions, Mr Kumar believes. "Focus on emerging, next-generation technologies, not mainstream ones that follow Moore's Law (where your competitors advance two times every 18 months). Niche technologies do not follow Moore's Law. They take a much, much longer time to evolve to the next generation.
"Today, we (Singapore) still have a lot of strengths in semiconductors and we should not lose them. We can still use the ecosystem here for niche technology."
For instance, NTU researchers here are working on GaN-based micro-inverters for building-integrated photovoltaics or solar panels as well as a new generation of high-energy radiation sensors using GaN solid-state technology.
Prof Ng is also collaborating with IGaN and the Singapore-MIT Alliance for Research and Technology Low Energy Electronic Systems to develop a GaN manufacturing technology which can be adopted by existing silicon wafer fabs to produce high-volume and low-cost GaN products, he says.
Over the last decade or so, many GaN startups that sprung up around the world were spun out of academic institutions, notes Richard Eden, principal research analyst for power semiconductors at tech research house Omdia.
Transphorm, one of the pioneers of high voltage GaN semiconductors, came out of the University of California at Santa Barbara. It was funded by KKR and Google Ventures before heading for a US listing through a reverse merger.
Mr Eden adds: "If Singapore wants to establish itself as a key GaN semiconductor hub, it should take advantage of its technological research institutes and financial strength by investing in startups, in the same way that NTU's School of Electrical and Electronic Engineering collaborates with IGaN on research projects."
Besides, many traditional semiconductor companies were slow to start in-house GaN projects, and may catch up through acquisitions, like how industrial giant STMicroelectronics took over GaN startup Exagan last month, he points out.
The GaN adoption roadmap
Last year was a breakthrough year for GaN (gallium nitride) in consumer electronics, says Ezgi Dogmus, technology and market analyst at market research firm Yole.
"Before 2019, the GaN power market was tiny and devices were mainly deployed in some niche applications. Starting from 2019, we witnessed the ground-breaking entry of GaN devices in high-power in-the-box fast chargers, meaning chargers sold together with the smartphones, in several brands such as Oppo and Realme. We also see major original equipment manufacturers such as Samsung and Xiaomi adopting GaN in accessory fast chargers. This opens the doors for the high volume smartphone market for GaN."
Apple is widely expected to retail chargers powered by GaN this year too, together with its new iPhone.
GaN's smaller form factor, high efficiency and cost-competitiveness for high-power fast-charging beyond 60 Watts are currently the key parameters desired by laptop and smartphone makers, Ms Dogmus adds, so GaN adoption will continue to rise.
Kong Xin, deputy programme director of NTU's Energy Research Institute, adds: "It is amazing to note that these GaN-based ultra slim phone and PC chargers are so small and light that they can easily be placed into your pocket. It will not be a surprise to see the IT market possibly taken over by the GaN technology in the near future."
So faster chargers are now here. But Jim Witham, chief executive of GaN Systems, has bigger ambitions. He wants to cut the power cord entirely, the same way we cut out wired telephones and ethernet cables: "To me, the future is wireless."
The wireless charging market is expected to grow at a compound annual growth rate of 41 per cent between 2018 to 2023. So far, low-power applications for wearables and smartphones have dominated. These use silicon-based transistors to achieve power levels of less than 15 Watts.
GaN is necessary to enable efficient resonant wireless charging at power levels from 25 Watts (so you can charge multiple phones or tablets over a charging pad) to 2,500 Watts, GaN Systems says.
GaN is already being used by automakers for in-car wireless charging systems. Other firms are already selling full product lines of transmitter units, charger units and receiver coils to enable wireless drone and factory robot charging, as well as underwater robot charging.
GaN Systems has also demonstrated a power amplifier that can send 65 Watts of power through an eight-inch wall, with more than 80 per cent efficiency end-to-end. This solves the problem of how to power electronics outside of a building, like security cameras.
Consumer electronics markets will lead in GaN adoption, followed by automotives, says Yole.
Raj Kumar, chief executive of IGSS GaN, notes: "What is stopping electric vehicles (EVs) from being used by people like you and me is how far a distance you can cover on a single battery charge. Most of the EVs can go as far as 250-300 kilometres. Some people will say this is good enough for Singapore, but that's missing the point. That's a limiting factor. GaN and silicon carbide (another compound semiconductor) will double that distance for the same space and weight."
Mr Kumar has been in talks with a few system companies and big auto giants on GaN applications for two years, but change will take time because new technologies take a longer time to qualify due to safety requirements, he says.
Meanwhile at the 2019 Tokyo Motor Show last October, Toyota and the Nagoya University Institute for Future Materials and Systems showcased an "all-GaN" concept car.
It featured a GaN traction inverter that improved efficiency by 20 per cent, extending the driving range of the car on one battery charge, and a GaN electric power converter which reduced the size of the system by 75 per cent compared to current models.
Mr Witham says: "Making things smaller and lighter are especially important in things that move... It has (GaN parts have) allowed Toyota to make the electronics one third the size, one third the weight and that extends the distance that you can have the car go. It also means you need less batteries. Batteries are one of the most expensive things in an electric vehicle, so it helps with the overall cost of the system. So GaN electronics for vehicles is a super-hot topic."
By Marissa Lee
Will GaN displace silicon?
"As the incumbent technology, silicon offers good performance for a very low price. GaN (gallium nitride) is a disruptive technology that offers better performance, but at higher costs. GaN device costs will eventually approach those of silicon, as both use silicon wafers as the basic substrate on which semiconductors are formed. But almost all analog and digital/logic ICs (integrated circuits), microprocessors and microcontrollers will use silicon, so the huge economies of scale will keep silicon production costs down. Silicon will always dominate. In the long term, for power semiconductors, silicon will dominate in the low-voltage range (0-80 volts). GaN has benefits from 80-650 volts, and silicon carbide offers the best performance above 650 volts."
- Richard Eden, principal research analyst for power semiconductors at Omdia
"The rate of conversion (to GaN parts) depends a lot on the amount of power that each device is using.
The more power they use, the faster they will move (to GaN or silicon carbide) because the efficiency gains monetise. So, If you're only doing 5 Watts, you save 2 per cent, you're not saving a lot of dollars - the incentive to do that is less. But if you're doing a Kilowatt and you save 2 per cent, the dollars add up and you get a very fast return on investment. In fact, the world isn't doing this fast enough because data centres could be saving money today if they all switched over to GaN, it would pay for itself."
- Rick Reigel, vice president of sales for GaN Systems