GaN HEMT has attracted a lot of attention in recent years owing to its wide applications from the high-frequency power amplifier to the high voltage devices used in power electronic systems. Development of GaN HEMT on Si-based substrate is currently the main focus of the industry to reduce the cost as well as to integrate GaN with Si-based components. However, the direct growth of GaN on Si has the challenge of high defect density that compromises the performance, reliability, and yield. Defects are typically nucleated at the GaN/Si heterointerface due to both lattice and thermal mismatches between GaN and Si. In this article, we will review the current status of GaN on Si in terms of epitaxy and device performances in high frequency and high-power applications. Recently, different substrate structures including silicon-on-insulator (SOI) and engineered poly-AlN (QST^®) are introduced to enhance the epitaxy quality by reducing the mismatches. We will discuss the development and potential benefit of these novel substrates. Moreover, SOI may provide a path to enable the integration of GaN with Si CMOS. Finally, the recent development of 3D hetero-integration technology to combine GaN technology and CMOS is also illustrated.

In this paper, we will discuss the rapid progress of third-generation semiconductors with wide bandgap, with a special focus on the gallium nitride (GaN) on silicon (Si). This architecture has high mass-production potential due to its low cost, larger size, and compatibility with CMOS-fab processes. As a result, several improvements have been proposed in terms of epitaxy structure and high electron mobility transistor (HEMT) process, particularly in the enhancement mode (E-mode). IMEC has made significant strides using a 200 mm 8-inch Qromis Substrate Technology (QST^®) substrate for breakdown voltage to achieve 650 V in 2020, which was further improved to 1200 V by superlattice and carbon-doped in 2022. In 2016, IMEC adopted VEECO metal-organic chemical vapor deposition (MOCVD) for GaN on Si HEMT epitaxy structure and the process by implementing a three-layer field plate to improve dynamic on-resistance (R_ON). In 2019, Panasonic HD-GITs plus field version was utilized to effectively improve dynamic R_ON. Both reliability and dynamic RON have been enhanced by these improvements.

A typical method for normally-off operation, the metal–insulator–semiconductor-high electron mobility transistor (MIS-HEMT) has been investigated. Among various approaches, gate recessed MIS-HEMT have demonstrated a high gate voltage sweep and low leakage current characteristics. Despite their high performance, obtaining low-damage techniques in gate recess processing has so far proven too challenging. In this letter, we demonstrate a high current density and high breakdown down voltage of a MIS-HEMT with a recessed gate by the low damage gate recessed etching of atomic layer etching (ALE) technology. After the remaining 3.7 nm of the AlGaN recessed gate was formed, the surface roughness (Ra of 0.40 nm) was almost the same as the surface without ALE (no etching) as measured by atomic force microscopy (AFM). Furthermore, the devices demonstrate state-of-the-art characteristics with a competitive maximum drain current of 608 mA/mm at a V_G of 6 V and a threshold voltage of +2.0 V. The devices also show an on/off current ratio of 10⁹ and an off-state hard breakdown voltage of 1190 V. The low damage of ALE technology was introduced into the MIS-HEMT with the recessed gate, which effectively reduced trapping states at the interface to obtain the low on-resistance (R_on) of 6.8 Ω·mm and high breakdown voltage performance.

Non-radiative energy transfer (NRET) from quantum dots (QDs) to monolayer MoS₂ has been shown to greatly enhance the photoresponsivity of the MoS₂ photodetector, lifting the limitations imposed by monolayer absorption thickness. Studies were often performed on a photodetector with a channel length of only a few μm and an active area of a few μm². Here, we demonstrate a QD sensitized monolayer MoS₂ photodetector with a large channel length of 40 μm and an active area of 0.13 mm². The QD sensitizing coating greatly enhances photoresponsivity by 14-fold at 1.3 μW illumination power, as compared with a plain monolayer MoS₂ photodetector without QD coating. The photoresponsivity enhancement increases as QD coating density increases. However, QD coating also causes dark current to increase due to charge doping from QD on MoS₂. At low QD density, the increase of photocurrent is much larger than the increase of dark current, resulting in a significant enhancement of the signal on/off ratio. As QD density increases, the increase of photocurrent becomes slower than the increase of dark current. As a result, photoresponsivity increases, but the on/off ratio decreases. This inverse dependence on QD density is an important factor to consider in the QD sensitized photodetector design.