-
Multi-objective Deep Data Generation with Correlated Property Control
Authors:
Shiyu Wang,
Xiaojie Guo,
Xuanyang Lin,
Bo Pan,
Yuanqi Du,
Yinkai Wang,
Yanfang Ye,
Ashley Ann Petersen,
Austin Leitgeb,
Saleh AlKhalifa,
Kevin Minbiole,
William Wuest,
Amarda Shehu,
Liang Zhao
Abstract:
Developing deep generative models has been an emerging field due to the ability to model and generate complex data for various purposes, such as image synthesis and molecular design. However, the advancement of deep generative models is limited by challenges to generate objects that possess multiple desired properties: 1) the existence of complex correlation among real-world properties is common b…
▽ More
Developing deep generative models has been an emerging field due to the ability to model and generate complex data for various purposes, such as image synthesis and molecular design. However, the advancement of deep generative models is limited by challenges to generate objects that possess multiple desired properties: 1) the existence of complex correlation among real-world properties is common but hard to identify; 2) controlling individual property enforces an implicit partially control of its correlated properties, which is difficult to model; 3) controlling multiple properties under various manners simultaneously is hard and under-explored. We address these challenges by proposing a novel deep generative framework that recovers semantics and the correlation of properties through disentangled latent vectors. The correlation is handled via an explainable mask pooling layer, and properties are precisely retained by generated objects via the mutual dependence between latent vectors and properties. Our generative model preserves properties of interest while handling correlation and conflicts of properties under a multi-objective optimization framework. The experiments demonstrate our model's superior performance in generating data with desired properties.
△ Less
Submitted 17 October, 2022; v1 submitted 30 September, 2022;
originally announced October 2022.
-
Strain-engineering of twist-angle in graphene/hBN superlattice devices
Authors:
Adolfo De Sanctis,
Jake D. Mehew,
Saad Alkhalifa,
Freddie Withers,
Monica F. Craciun,
Saverio Russo
Abstract:
The observation of novel physical phenomena such as Hofstadter's butterfly, topological currents and unconventional superconductivity in graphene have been enabled by the replacement of SiO$_2$ with hexagonal Boron Nitride (hBN) as a substrate and by the ability to form superlattices in graphene/hBN heterostructures. These devices are commonly made by etching the graphene into a Hall-bar shape wit…
▽ More
The observation of novel physical phenomena such as Hofstadter's butterfly, topological currents and unconventional superconductivity in graphene have been enabled by the replacement of SiO$_2$ with hexagonal Boron Nitride (hBN) as a substrate and by the ability to form superlattices in graphene/hBN heterostructures. These devices are commonly made by etching the graphene into a Hall-bar shape with metal contacts. The deposition of metal electrodes, the design and specific configuration of contacts can have profound effects on the electronic properties of the devices possibly even affecting the alignment of graphene/hBN superlattices. In this work we probe the strain configuration of graphene on hBN contacted with two types of metal contacts, two-dimensional (2D) top-contacts and one-dimensional (1D) edge-contacts. We show that top-contacts induce strain in the graphene layer along two opposing leads, leading to a complex strain pattern across the device channel. Edge-contacts, on the contrary, do not show such strain pattern. A finite-elements modelling simulation is used to confirm that the observed strain pattern is generated by the mechanical action of the metal contacts clamped to the graphene. Thermal annealing is shown to reduce the overall doping whilst increasing the overall strain, indicating and increased interaction between graphene and hBN. Surprisingly, we find that the two contacts configurations lead to different twist-angles in graphene/hBN superlattices, which converge to the same value after thermal annealing. This observation confirms the self-locking mechanism of graphene/hBN superlattices also in the presence of strain gradients. Our experiments may have profound implications in the development of future electronic devices based on heterostructures and provide a new mechanism to induce complex strain patterns in 2D materials.
△ Less
Submitted 4 December, 2018; v1 submitted 9 October, 2018;
originally announced October 2018.
-
Novel circuit design for high-impedance and non-local electrical measurements of two-dimensional materials
Authors:
Adolfo De Sanctis,
Jake D. Mehew,
Saad Alkhalifa,
Callum P. Tate,
Ashley White,
Adam R. Woodgate,
Monica F. Craciun,
Saverio Russo
Abstract:
Two-dimensional materials offer a novel platform for the development of future quantum technologies. However, the electrical characterisation of topological insulating states, non-local resistance and bandgap tuning in atomically-thin materials, can be strongly affected by spurious signals arising from the measuring electronics. Common-mode voltages, dielectric leakage in the coaxial cables and th…
▽ More
Two-dimensional materials offer a novel platform for the development of future quantum technologies. However, the electrical characterisation of topological insulating states, non-local resistance and bandgap tuning in atomically-thin materials, can be strongly affected by spurious signals arising from the measuring electronics. Common-mode voltages, dielectric leakage in the coaxial cables and the limited input impedance of alternate-current amplifiers can mask the true nature of such high-impedance states. Here, we present an optical isolator circuit which grants access to such states by electrically decoupling the current-injection from the voltage-sensing circuitry. We benchmark our apparatus against two state-of-the-art measurements: the non-local resistance of a graphene Hall bar and the transfer characteristic of a WS2 field-effect transistor. Our system allows the quick characterisation of novel insulating states in two-dimensional materials with potential applications in future quantum technologies.
△ Less
Submitted 30 January, 2018;
originally announced January 2018.
