Although molecular beam epitaxy (MBE), one of three types of epitaxy equipment, has long been considered niche, it is poised to transition to volume applications.
Unprecedented demand for more powerful and energy-efficient devices has spurred the need for compound semiconductors such as gallium arsenide, gallium nitride, and silicon carbide. Such materials require ultra-pure thin films grown by epitaxy. Although molecular beam epitaxy (MBE), one of three types of epitaxy equipment, has long been considered niche, it is poised to transition to volume applications.
During a recent webinar, Riber, French provider of MBE equipment serving the semiconductor industry, and Yole Développement, French research and strategy consulting firm, presented the current status and outlook of the global MBE equipment market.
MOCVD, HTCVD, MBE
The epitaxy equipment market is segmented by technology. Metal-organic chemical-vapor deposition (MOCVD) epitaxy equipment covers most of the III–V epitaxy, while high-temperature CVD (HTCVD) is the most common deposition technique for silicon and SiC devices. As Yole explains in a recently published report, “Epitaxy Equipment for More Than Moore 2021,” MOCVD is required for commodity devices such as GaN-based traditional LEDs. However, a growing number of high-end applications, such as fast chargers, MicroLED displays, and vertical-cavity surface-emitting lasers (VCSELs) for 3D sensing, will drive demand in the coming years. For HTCVDs, the primary market is power applications, which are based on silicon and SiC epi material and are mainly deployed in market segments such as automotive and industrial.
MBE is based on the evaporation of complex materials with beams of molecules in an ultra-vacuum environment. Atoms are deposited from the evaporated materials to the substrate, where they form a crystalline layer. This technology makes it possible to achieve electronic components with remarkable physical characteristics. MBE is used in low-volume, high-performance–demanding applications.
According to Yole, the epitaxy equipment market, including MOCVD, HTCVD, and MBE, will grow at a compound annual growth rate (CAGR) of 8%, from US$692 million in 2020 to US$1.1 billion in 2026. More specifically, MOCVD, which represented more than 60% of the equipment market share in 2020 revenues, will grow at a CAGR of 7% over the forecast period, to US$630 million in 2026. The HTCVD market is set to grow at a 9% CAGR over the same period, to US$393 million in 2026. The MBE equipment market value has long remained limited, with revenue of US$45 million in 2020, but will grow at a healthy 7.1% CAGR to reach US$68 million in 2026. “This forecast is only for equipment and does not include services and spare parts, which represent a significant value of that market,” said Jean-Christophe Eloy, CEO of Yole.
MBE is often considered to be a niche technique and faces the unfortunate misconception that it is suitable only for research and pilot production. For example, while it is true that Riber began in 1964 by supplying research laboratories and universities, it now has 750 MBE machines in operation in the world, and some leading semiconductor foundries use MBE exclusively in their production environment.
Riber is at the start of the compound semiconductor value chain. It delivers MBE systems for both research on compound semiconductor materials and for volume production of epi wafers. The Bezons, France-based company also offers ultra-high–vacuum chemical-deposition machines for growing a material or crystal in successive layers of atoms and has diversified into evaporators for the organic LED (OLED) and photovoltaics industry.
Today, Riber serves the telecom and infrastructure (satellites, 4G/5G base stations, fiber optics, lasers), defense and aerospace (night vision, radar, infrared), and industry (photovoltaics, OLEDs, and ultraviolet) markets. Yet Yole is convinced that MBE offers many opportunities and that it is time to unleash its full potential.
“MOCVD has a very strong growth potential with multiple applications in multiple markets, but in terms of technology and scalability for large wafers, it’s quite complex and it takes time,” Eloy said. In contrast, he said, “MBE is an epitaxy equipment for extremely uniform layers on large dimensions and … any type of layers on top of wafers. MBE, with its extremely precise ability to achieve epitaxial layers, offers significant opportunities for devices, processes, and functionalities that cannot be achieved by MOCVD.
“In the medium and long term, there are potential game-changers that are now in development and could move into very high-volume production,” Eloy added.
This perspective applies to Riber and its direct and only competitor, Veeco Instruments.
Indeed, Riber and Veeco Instruments are the only two players offering high-capacity/high-throughput MBE production tools for volume manufacturing. Other MBE manufacturers such as DCA Instruments Oy, SVT Associates, Eiko, and VJ Technologies offer R&D or pilot production systems.
Yole’s MBE market growth forecast of approximately US$68 million by 2026 is encouraging, but the best is yet to come. Opportunities related to MicroLEDs, VCSELs, and quantum computing are indeed expected after 2026.
“MBE is well-entrenched in the RF and photonics business because the market started tens of years before,” Eloy told EE Times Europe during the Q&A session. “Photonics is really driven by data centers and high-performance computing, and it’s here to stay for the next 20 years. It’s very stable.”
He continued, “New growth opportunities are significant because MicroLEDs are about to turn the display industry upside down.” MicroLED displays are set to move into volume production in the 2025–2026 timeframe.
VCSEL is a more short-term application, while quantum computing is a more long-term volume application.
All in all, these different areas of development “can add a lot of value and a lot of equipment on top of what is already on the market,” said Eloy. The prospect is “very significant in terms of business upside.”
Quantum computing promises to be a game-changer for MBE, said Nicolas Grandjean, professor of physics at the École polytechnique fédérale de Lausanne (EPFL) and vice chairman of the supervisory board at Riber.
When asked about Riber’s current and future developments in quantum computing, Philippe Ley, Riber’s chairman of the executive board, told EE Times Europe that the group is developing a completely new machine to deposit materials at low temperature. “Currently, it’s about 800°C, and the target is to have some deposition at –100°C or –200°C,” said Ley. In that respect, Riber and the CNRS Laboratory for Analysis and Architecture of Systems in Toulouse, France, established a joint laboratory in June 2021. Dubbed Epicentre, the laboratory aims to develop a series of in situ measurement tools, as well as a dedicated solution for growing superconductor materials for quantum computing. Expected within the next two years, the co-developed solution will be made available to all research centers in the world, Ley said.
In parallel, Riber is conducting a project on the convergence of silicon and compound semiconductors to overcome the current limitations of silicon. Various materials are emerging because they offer extremely high-performance switching-speed properties with virtually zero optical losses, the company says. This is opening up possibilities for applications that have never been considered, such as light-beam shaping, holography, and the production of networks of programmable optical neurons (AI, quantum computing, recognition).
Dubbed Rosie, the project aims to develop the first 300-mm machine, making it possible to create thin films of these materials with an epitaxy process on silicon, Ley said.
“Arsenide, nitride, and blue LEDs are grown by MOCVD, but if you wish to grow oxide, then there is an issue with the precursors and the MOCVD by itself,” said Grandjean. “I do see [an opportunity for] MBE to really compete, maybe not on nitride, but more for functional oxides and superconductors where MOCVD simply can’t grow those materials, as far as I know. This is not a short-term business — maybe five years for functional oxide on silicon.”
This article was originally published on EE Times Europe.
Anne-Françoise Pelé is editor-in-chief of eetimes.eu and EE Times Europe.