Tailored Synthesis of Hexagonal Boron Nitride: Chemical Vapor Deposition and Next Generation Devices

Abstract

The introduction of two dimensional (2D) materials in electronic device applications has provided a platform for several new device concepts and architectures. Hexagonal boron nitride (hBN) is the only insulator in the family of 2D materials. It has thus been widely regarded as the preferred substrate for 2D nanodevices and its heterostructures due to its electrically insulating nature, excellent thermal conductivity, suitable dielectric characteristics, and very high optical phonon energy. Owing to such characteristics, it also has several applications as a gate dielectric, tunnel barrier layer, and resistive switching memory, among others. To better exploit its capabilities, hBN must be synthesized over large areas with good crystalline quality and uniformity, along with control over the number of layers and grain size. In this work, by physico-chemical modelling of the ammonia borane (BH3NH3) route, currently the most popular chemistry for hBN deposition, a predictive CVD process is established to grow high crystalline quality uniform hBN over large areas, 6”x6”. Most other approaches have been empirical. Another key distinguishing factor in this work is the fact that the BH3NH3 precursor is placed outside the growth chamber which allows regulation of the precursor flux. This approach allows for control over the formation of the impurity known as nanocrystalline-BN (nano BN) which is observed with ammonia borane as the precursor. Eliminating nano BN is very important to obtain clean surfaces/interfaces and to integrate hBN with other materials for device implementation. The CVD process parameter window is identified that allows for uniform deposition of n-layered hBN, n = 1, 2… > 10 with grain sizes approaching 5 microns. To demonstrate that the hBN growth method reported in this paper yields layers of the required quality despite its polycrystalline nature, graphene on hBN FETs have been fabricated with record - when compared with existing literature reports for CVD graphene/CVD hBN devices - room temperature mobilities. Ammonia borane (BH3NH3) as a precursor contains both boron and nitrogen which therefore fixes the B/N ratio. In deposition of compound semiconductor films the ability to vary this ratio has been found to be crucial to exercise better control over the fundamental aspects of growth. Also, in the CVD hBN process using ammonia borane, due to the relatively larger solubility of B in Cu which then acts as a secondary source of B, lateral growth of the islands, is limited by availability of nitrogen on the surface. Hence, to further exercise control over the growth process, an ammonia-assisted chemical vapor deposition method is explored to synthesize larger grain sizes, in excess of 50 microns of hexagonal boron nitride (hBN) for the very first time to the best of our knowledge. The mechanistic understanding of the growth is established by correlating the effect of ammonia on the nucleation density and growth rates of the hBN grains. To validate the effect of larger grain sizes on device performance, resistive random-access memories (RRAMs) are fabricated based on hBN which exhibit non-volatile bipolar resistive switching (RS). The fabricated RRAMs establish improved switching performance due to larger grain sizes of the hBN. This result is due to the reduction of grain boundaries which facilitate lower cycle-to-cycle variability, better endurance and higher current on/off ratio as the grain boundaries are known to assist ion migration. A 4 order of magnitude improvement in retention capability of the memory and a 5-order improvement in on/off ratio is observed. The hBN deposited here is poly-domain with grain boundaries. It has been shown in graphene that by defect engineering of grain boundaries, properties of CVD graphene can approach that of single crystal material. It has been established that the grain boundaries or defect sites in polycrystalline CVD grown hBN provide leakage routes. These paths in turn result in local generation of percolation paths that lower the dielectric breakdown (BD) of multi-layered hBN. To verify if whether the same ammonia annealing route can also heal defects in BN as in graphene, a post growth annealing technique involving ammonia was employed. Structural and Raman characterization of the obtained films has been correlated to the partial pressure of ammonia during annealing to demonstrate reduced defect density. The post growth annealed hBN films exhibited a high breakdown field strength of ∼13.1 MV cm−1 , which is the highest reported breakdown fields of CVD grown hBN films. Using the understanding of resistive switching, neuromorphic behavior in a scalable two-dimensional material structure is demonstrated consisting of CVD-hBN grown on copper (Cu) and contacted with silver (Ag). In this system, avalanche dynamics is examined such as those seen in cortical tissue structures which exhibit critical neuromorphic network dynamics due to presence of atomic-scale networks which develop as a result of diffusion of Ag inside the hBN matrix. The development of Ag filaments by application of persistent I-V sweeps also gives rise to a resistive switching memory device that has two states, a low resistance (LRS) and a high resistance (HR) state. The avalanche dynamics are observed in the HRS due to the intercalation of Ag inside the hBN matrix which results in formation of a percolation network when Ag clusters are within the tunneling distance. In the LRS state, the filamentary networks of Ag are formed which exhibit avalanche behavior under the application of a constant electric field. Thus, a first of its kind brain-like avalanche behavior is reported in a 2D material system comprising of Ag-hBN. This kind of system can be scaled up to form large-area devices and with hBN being a 2D material, it allows for engineered heterostructures for future applications in neuromorphic computing. In summary, large area deposition of hexagonal boron nitride (hBN) is enabled by controlling the nucleation density, grain size, layer thickness and defect density. The quality of the films grown are demonstrated to be state of the art by various standards. The highest values of mobility of CVD Graphene on CVD hBN and highest breakdown field of CVD hBN are reported. The improved resistive switching characteristics and the enablement of novel neuromorphic architecture shows the potential of CVD grown hBN in next generation electronic applications