Ferroelectrics for GaN based High Power Devices
Moritz Nunnenkamp is a PhD student in the department Inorganic Materials Science. Promotors are prof.dr.ir. G. Koster and prof.dr.ing. A.J.H.M. Rijnders from the faculty Science & Technology.
Power electronics is a rapidly growing sector with diverse applications. The market for power electronics, valued at $43.3 billion in 2022, remains highly attractive in the semiconductor industry. The increasing use of integrated circuits in ICT systems has led to a rise in global power consumption, emphasizing the need for power efficiency in various applications. Power field-effect transistors (FETs) play a vital role in power conversion modules, but achieving a balance between breakdown voltage and specific on-resistance is challenging.
Extensive research is currently focused on gallium-nitride (GaN), a member of the tri-nitride (III-N) mineral class, due to its potential for high-power applications. The desirable 3C-GaN, or GaN in Wurtzite phase, exhibits high polarization, which is crucial for the formation of a two-dimensional electron gas (2DEG) in heterointerfaces containing aluminum-gallium-nitride (AlGaN) in high-electron mobility transistors (HEMTs). GaN's broad bandgap (3.4 eV) and large breakdown field make it attractive for high-power and high-frequency applications, surpassing the figure of merits of Si power devices. Encouraged by previous studies, this research focuses on innovative dielectric RESURF III-N HEMTs that incorporate ferroelectric materials into GaN-based HEMTs, aiming to achieve superior performance compared to Si and GaN devices. The choice of a suitable buffer layer is crucial for the growth of ferroelectric materials on GaN, as it should withstand a strong electric field while allowing for well controlled and crystalline growth.
The epitaxial growth of lead zirconate titanate (PZT) on gallium nitride (GaN) with and without buffer layer is investigated. Without a buffer layer, the PZT film on GaN exhibits a rough surface with small triangular grains and larger hexagonal agglomerates. X-ray diffraction (XRD) analysis shows a higher proportion of the pyrochlore phase in the film. To promote perovskite growth, a magnesium oxide (MgO) buffer layer is integrated, resulting in improved surface morphology with hexagonal grains and reduced roughness. Single phase PZT in (111) direction was observed. Another approach of integrating PZT on GaN involves using oxide nanosheets, such as titanium oxide (TiO) and cerium oxide (CNO), as buffer layers. PZT grown on TiO nanosheets shows mixed phase formation, while PZT on CNO nanosheets exhibits single-phase (001) growth. Surface morphology analysis reveals pronounced nanosheet imprints on PZT/CNO and less distinct imprints on PZT/TiO. The lattice mismatch between nanosheets and PZT influences the growth behavior. Lead-free perovskites, such as barium titanate (BTO) and strontium titanate (STO), are also grown on Ca2Nb3O10 nanosheets. BTO shows single-phase (001) growth, while STO exhibits square features corresponding to the cubic unit cell lattice. The electrical properties of these films are further investigated.
Following from the results earlier, different oxide buffer layers were introduced between Ca2Nb3O10 nanosheets and PZT. Topological, structural and ferroelectric behavior of poly-epitaxial Ca2Nb3O10 nanosheet buffered PZT, utilizing SrTiO3, LSAT and LaMnO3 as additional oxide buffer layers was investigated. An increase in crystalline quality and ferroelectric response in the case of SrTiO3 buffered PZT was attributed to the good lattice matching between Ca2Nb3O10 and SrTiO3 , which promote well aligned PZT growth. LSAT showed improved structural and electrical properties as well with a less strong increase mainly attributed to the presence of a second PZT orientation within the film. Bare Ca2Nb3O10 nanosheets as well as LaMnO3 as buffers resulted in columnar and therefore less crystalline PZT films also showing less strong ferroelectric polarization. LaMnO3 showed a decrease in polarization compared to bare Ca2Nb3O10 nanosheets. This was attributed to the large mismatch in lattice constants between Ca2Nb3O10 and LaMnO3 which promoted rough and strained LaMnO3 buffer films, leading to rough and less aligned PZT growth. SrTiO3 with a lattice mismatch of -1,29% to Ca2Nb3O10 and -3,58% to PZT is relaxing the accompanied strain in this material system well with a 10nm buffer. XRD as well as electrical measurements show well improved properties due to the dense and well aligned growth of the SrTiO3 buffered PZT compared to growth of PZT without buffer or utilizing LaMnO3 and LSAT. Multilayer nanosheet films as growth templates were tested in order to eliminate possible ‘open’ areas in the CNO nanosheet films to possibly improve device performance. It was found that the roughness in the nanosheet films as well as the later deposited PZT on top increased with the number of nanosheet layers. This roughness increase was majorly driven by a larger misalignment in the grains within the PZT films. Following from this, the polarization was also found to decrease for multilayer buffered PZT compared to monolayer buffered PZT.
Interfacial effects and their influence on PZT utilizing SrTiO3 as additional oxide buffer layers was investigated in order to understand the mechanism behind the significant structural changes as discovered earlier. The analyzed films could be grouped in two major types of films. PZT with buffer layers thinner than 2.5nm and PZT films with STO buffers thicker than 2.5nm. A major structural change was observed with a change in growth behavior going from columnar growth for direct growth of PZT on Ca2Nb3O10 compared to dense film growth when introducing a SrTiO3 buffer layer thicker than 2.5nm. This also reflected in the crystallinity of the PZT films. PZT was found to grow in the (001) direction regardless of the SrTiO3 buffer, however the crystallinity was significantly improved when utilizing 2.5nm or more of SrTiO3 buffer. The polarization of PZT changed towards a squarer loop with buffer compared to a more diagonally stretched P-E loop without buffer. More in-plane polarization for SrTiO3 was attributed to that behavior. Interdiffusion of Pb at the interface between PZT and CNO nanosheets was shown by cross-section TEM on individual nanosheets. SrTiO3 showed to prevent this Pb diffusion at the interfacial region acting as a diffusion barrier.
To make the step towards the actual investigation on breakdown voltage and contact resistance of ferroelectric GaN HEMTs through the utilization of dedicated test structures, various ohmic contacts with different Al/Ti ratios were examined. It was found that a contact ratio of 170/55nm Al/Ti, annealed at 850°C, exhibited superior performance compared to other ratios. Finally, we fabricated and experimentally studied breakdown structures comprising high-k gate oxides (PZT, BTO, and STO) on buffer layers of epitaxial MgO, STO buffered CNO nanosheets, as well as bare CNO nanosheets, all of which were characterized at earlier stages. Our findings reveal that the average breakdown field in the PZT/MgO GaN structure has increased by a factor of 1.6 when compared to the conventional GaN counterpart. However, nanosheet buffered high k dielectric gate oxides did not enhance the breakdown performance of the devices. STO and BTO, as well as STO-buffered PZT, exhibited breakdown at significantly lower fields, with factors ranging between 0.2 and 0.4 compared to the reference GaN device without any gate oxide. Although CNO buffered PZT demonstrated breakdown fields with a factor of 0.6, showcasing more comparable breakdown behavior compared to the reference device, it still fell short of improving the breakdown characteristics adequately. It can therefore be concluded from the obtained results that only fully epitaxial buffered PZT has the potential to enhance the breakdown characteristics of HEMT devices.