ZnGeN₂ and GaN are almost lattice matched and both have band gaps of approximately 3.4 eV. A large conduction band offset of ~ 1.4 eV results in a type II band alignment that has great potential for novel device structures. [1,2] For the 50-50 alloy, a slightly positive mixing energy, indicating a tendency toward phase separation, has been predicted. [3] For this mixture the lowest energy configuration is predicted to be an octet-rule-preserving orthorhombic Pmn2₁ phase. Other compositions may in principle be made in octet-rule-preserving (and thus lower energy) phases, compared to those that break the octet rule, by random stacking of ZnGeN₂ and GaN layers along the orthorhombic b axis. [3] Whether random stacking, phase separation, or octet rule violations occur will determine whether, and by how much, the band gap may be tuned with composition, and whether the transport properties are isotropic or anisotropic. The only other work on this alloy reported to date employed a gas reduction nitridation method for synthesis of powders of different compositions, from pure ZnGeN₂ to a 50-50 mixture, for photocatalytic applications [4].

Here we report the results of MOCVD growth of this alloy on c-, r-, and a-plane sapphire and c-GaN/sapphire substrates, at temperatures varying from 550 ⁰C to 700 ⁰C. Films at the 50-50 composition exhibit better surface morphologies when grown on r-sapphire substrates. Zn incorporation increases with the increase of Ga. The highest growth rate, 3.46 µm/hr, was obtained for a film grown on r-sapphire at 670 ⁰C and 550 torr, for which a 2θ-ω XRD measurement yielded a wurtzite (110) diffraction peak at 2Ɵ = 57.70⁰ with FWHM of 0.76⁰ and an RMS surface roughness of ~ 10 nm by AFM. The Hall mobility is 8.19 cm²/v-s with an n-type carrier concentration of 8.5 x 10¹⁸ cm^-3. Atomic-resolution HAADF-STEM revealed the atomic arrangement of the film near the substrate interface. Introduction of a low-temperature-grown ZnGeN₂ buffer layer (480 ⁰C at 600 torr with low injection of precursors compared to the film growth conditions) led to improved surface morphology and crystal quality, and yielded a room temperature photoluminescence spectrum indicating a band edge at approximately 3.5 eV, close to that predicted for the Pmn2₁ phase. [3]

The authors acknowledge support from the National Science Foundation DMREF: SusChEM: grant 1533957.

[1] L. Han, K. Kash, H. Zhao, J Appl Phys 120, 103102 (2016)

[2] M. R. Karim, H. Zhao, J Appl Phys 124, 034303 (2018)

[3] B.H.D. Jayatunga, S. Lyu, S. Kumar, K. Kash, W. R. L. Lambrecht, Phys Rev Mat 2 (2018)

[4] T. Suehiro, M. Tansho, T. Shimizu, J Phys Chem C 121, 27590 (2017)

We report on the growth of device quality β-Ga₂O₃ and related alloys using MOCVD method. β-Ga₂O₃ thin films are grown using Ga(DPM)₃, TEGa and TMGa as Ga sources, and molecular O₂, H₂O vapor, and N₂O as an oxidizer. Films grown from each Ga source had high growth rates with up to 10 m/hr achieved using TMGa [1]. The effect of the oxidizer identity on the growth rate, electron mobility (µ_e), background carrier concentration, surface and crystalline quality of the films will be discussed. Using pure O₂ as an oxygen source, optimal growth conditions have led to the growth of smooth epitaxial UID Ga₂O₃ thin films with a RT µ_e of 176 cm²/Vs at n~7x10¹⁵ 1/cm³ [2]. The highest µ_e of ~3500 cm²/Vs has been measured at 54 K. C and H impurities have been shown to be below the SIMS detection limit for a wide range of process conditions, whereby films with n~ 2x10¹⁴ cm^-3 were demonstrated. We will also present the growth of device quality β-Ga₂O₃ layers doped with Si, Fe, and N impurities. Critical growth conditions influencing the incorporation of these dopants will be discussed. Using optimum growth conditions, controllable doping with a concentration between 10¹⁵ and 10²⁰ 1/cm³ were obtained for each dopant. In this work, we will also present on the MOCVD growth of (Al_x, Ga_1-x)₂O₃ alloys. The MOCVD process enables the growth of AlGaO at a temperature >800 ^oC, improving the solubility of Al₂O₃ in β-Ga₂O₃ by preventing the formation of volatile suboxides. The MOCVD reactor used in this work has a unique feature that enables it to minimize premature reaction between the species, thereby improving the Al incorporation. AlGaO alloys with Al content of up to 43 % was obtained. The epitaxial growth of high quality strained β-(Al, Ga)₂O₃/Ga₂O₃ heterostructures and superlattices will be discussed. The composition homogeneity, structural quality, surface morphology and electrical properties of the heterostructures will be discussed as a function of growth conditions. Finally, the growth of alloys including β-(In, Ga)₂O₃ and ZnGaO using MOCVD and their application for photodetection purposes will be discussed.

[1] F. Alema et al., J. Cryst. Growth, 475, 77(2017).

[2] Y. Zhang et al., APL Materials, 7, 022506 (2019).

Phase-Change Memory (PCM) is today the most mature among innovative back-end non-volatile memory technologies, thanks to a wide set of interesting features making PCM technology enough versatile to meet different applications’ requirements [1]. A PCM device experiences a physical change of a chalcogenide material sandwiched between two electrodes made possible by the current induced Joule heating flowing through the cell. To achieve the amorphous phase, the PCM in the crystalline phase has to be melted, then rapidly quenched (RESET operation). Thanks to the switching phenomenon, the material in the amorphous phase changes abruptly its conductivity starting to be highly conductive, and can recover the crystalline phase thanks to a specific thermal profile during the pulse application, that provides the energy necessary to the atomic reorganization (SET operation). Thereby, PCM thermal stability relies on the magnitude of the activation energy of the crystallization that results from the combination of crystals nucleation and growth phenomena, on which also the device programming speed relies. Thus, a general trade‑off exists between the time required for the SET operation and the device data retention performance [2] (Figure 1). Sb-rich GeSbTe compounds are suitable for high-speed performances with a programming time down to tens of ns still ensuring high endurance and scalability, promising for Storage Class Memory applications (SCM) [3]. Reliability at high temperature is the main requirement to target automotive embedded applications. Ge-rich compositions revealed an endurance of 10⁷ cycles up to 175 °C and high temperature data retention compatible with embedded standards. We present here the device performance tuning thanks to the phase‑change material stoichiometry engineering (Figure 2). Moreover, we highlight the possibility to boost the PCM performances, such as SET speed and Multi Level Cell capability, thanks to dedicated programming strategies [4].

REFERENCES

[1] F. Arnaud et al, “Truly Innovative 28nm FDSOI Technology”, IEDM 2018.

[2] G. Navarro et al, “Non-Volatile Resistive Memory”, ECS 2016.

[3] V. Sousa et al, “Phase Change Memory”, Chapter 7, Springer 2018.

[4] J. Kluge et al, “High Operating Temperature Reliability”, IMW 2016.

The ability to dynamically learn and adapt to changes in the environment is one of the hallmarks of biological systems. In the last years there has been a lot of research focused on exploring novel materials, such as those exhibiting memristive behavior, that could enable this type of systems. However, there are comparatively fewer studies focusing on understanding which are the ideal properties that memristive materials should have in order to optimize the performance of architectures capable of dynamic learning. This type of information is crucial to provide design targets for new materials and accelerate the integration of novel devices into architectures optimized for specific applications.

In this work, we identify the subset of the design space of memristive materials that is optimal for dynamic learning applications: in this type of application, a system, in this case a neural network, evolves dynamically and learns as it processes information in real time. This type of behavior is highly desirable for smart sensors or edge processing applications. We have implemented a benchmark architecture consisting of a discrete implementation of spiking neurons where dynamic learning takes place on a set of plastic synapses formed by memristor pairs in a crossbar array. This architecture, which is inspired on the learning center of the insect brain, is capable of dynamically learning standard machine learning datasets such as MNIST and Fashion-MNIST. We have used this model to identify the key properties that memristive materials should have to be optimal dynamic learners, exploring the impact of the kinetics of the memristor's internal state on the system's learning ability, as well as the impact that materials and device variability and errors in tuning the memristor's internal state have on the system's performance.

The results obtained show that a fine degree of control of the memristor internal state is key to achieve high classification accuracy during dynamic learning, but that, within this optimal region, learning is extremely robust to both device variability and to errors in the writing of the internal state, in all cases allowing for 2σ variations greater than 40% without significant loss of accuracy. Moreover, the dynamics of the internal state can show distinct kinetics depending on the polarity, something that is critical for bipolar memristors. These criteria are significantly different from those required for ReRAM applications or even for neuromorphic applications based on offchip training, where the robustness of reading and writing operations are critical.

VO₂ is a phase change material that undergoes a first order crystalline phase transition at a critical temperature (T_c = 68^oC), resulting in significant changes in intrinsic electrical and optical properties, especially in the infrared. Optical changes with this phase transition are of particular interest as passive and active components of optoelectronic devices, specifically for thermal regulation and modulated signaling. Realizing this type of device often requires the integration of thin, conformal VO₂ films with complex, non-planar structures (like metamaterials). Thus, atomic layer deposition (ALD) is the ideal deposition method in these cases.

Traditional metal-based plasmonic materials suffer from high optical losses, which has promoted research towards alternative low-loss materials that can support plasmonic-like effects. One such approach employs phonon-mediated collective-charge oscillations (surface phonon polaritons, SPhPs) that are supported by nanostructured polar dielectric materials (SiC, AlN, etc), which inherently are low-loss. Geometric design of the nanostructures enables spectral tuning of resonant features between the longitudinal and tranverse optical phonons of the polar material, typically in the infrared regions. However, the spectral position and amplitude of these resonances remain fixed after fabrication. Integrating phase change materials with these structures provides a way to achieve active modulation of resonances.

Phase change materials (PCM) are attractive due their tunability, wide range of applications, and quasi-passive actuation as compared to traditional active integrated circuits. Vanadium dioxide (VO₂) is particularly appealing because of its high contrast between dielectric and conductive states and the relatively low temperature (68 °C) at which its metal-insulator transition (MIT) occurs. This work details a process for depositing high quality VO₂ thin films on C-plane sapphire wafers as well as alumina (Al₂O₃) buffer layers using atomic layer deposition (ALD) on silicon substrates. We compare resistivity vs. temperature measurements of VO₂ on sapphire with VO₂ on the Al₂O₃ buffer layers both as-deposited and after rapid thermal annealing (RTA) at temperatures ranging from 950 to 1150°C. The VO₂ thin films yielded ratios of resistivity between conductor (heated) and dielectric (room temperature) states of 9.8×10⁴ Ω-cm, 5.2×10³ Ω-cm, and 1.5×10⁴ Ω-cm when deposited on crystalline sapphire, amorphous Al₂O₃ buffer layers, and annealed Al₂O₃ buffer layers, respectively. This corresponds to an improvement by a factor of 2.9 in the annealed buffer layers over the amorphous buffer layers. Moreover, we studied various VO₂ thin films using X-ray diffraction, which showed clear indications that the films are highly pure and have a preferred crystal orientation. The deposition process we have developed will allow us to use high quality VO₂ thin films on silicon substrates, especially millimeter-wave devices such as reconfigurable antennas, sensors, and meta-surfaces.

The information age has unleashed an exponential growth in access and quality of life for human beings, but sustaining this growth is key to our future. Currently, 10% of the world’s energy is consumed for processing, communication or storage of digital media and this share is constantly increasing, which is a slowly unfolding energy crisis.¹ With the growing need for artificial intelligence, efficient use of energy for computing warrants a departure from digital logic. Brain-inspired neuromorphic computing is envisioned to be several times more efficient for associative learning problems. Phase change materials demonstrating metal to insulator transitions are suitable to realize solid state neurons in this neuromorphic framework. VO₂ undergoes a 5 order-of-magnitude change in resistivity near room temperature with minimal structural change.² We designed highly energy efficient, low power oscillators based on VO₂/SrRuO₃ heterostructures in the light of thermal engineering. Early simulation work has demonstrated low power consumption down to 45 µW with 3 GHz driving frequency.³ High-quality epitaxial VO₂ thin films have been deposited on various 3m surface symmetry substrate, SrTiO₃ and [LaAlO₃]_0.3[Sr₂AlTaO₆]_0.7.⁴ VO₂ thin films have shown sharp 4 orders of magnitude change in resistivity around the transition temperature. The thermal boundary conductance is measured by time-domain thermoreflectance. We also present our experimental effort to achieve energy efficient and low power oscillators for neuromorphic computing.

¹J. Nicola, Nat. Mag. 163 (2018).

²H.-T. Kim, B.-J. Kim, S. Choi, B.-G. Chae, Y.W. Lee, T. Driscoll, M.M. Qazilbash, and D.N. Basov, J. Appl. Phys. 107, 023702 (2010).

³B. Zhao and J. Ravichandran, Phys. Rev. Appl. 11, 014020 (2019).

⁴Y. Liu, S. Niu, T. Orvis, H. Zhang, H. Zhao, H. Wang, and J. Ravichandran, J. Vac. Sci. Technol. A 36, 061506 (2018).