When Machine Learning Meets 2D Materials: A Review.

The availability of an ever-expanding portfolio of 2D materials with rich internal degrees of freedom (spin, excitonic, valley, sublattice, and layer pseudospin) together with the unique ability to tailor heterostructures made layer by layer in a precisely chosen stacking sequence and relative crystallographic alignments, offers an unprecedented platform for realizing materials by design. However, the breadth of multi-dimensional parameter space and massive data sets involved is emblematic of complex, resource-intensive experimentation, which not only challenges the current state of the art but also renders exhaustive sampling untenable. To this end, machine learning, a very powerful data-driven approach and subset of artificial intelligence, is a potential game-changer, enabling a cheaper - yet more efficient - alternative to traditional computational strategies. It is also a new paradigm for autonomous experimentation for accelerated discovery and machine-assisted design of functional 2D materials and heterostructures. Here, the study reviews the recent progress and challenges of such endeavors, and highlight various emerging opportunities in this frontier research area.

Toward a formal theory for computing machines made out of whatever physics offers.

Approaching limitations of digital computing technologies have spurred research in neuromorphic and other unconventional approaches to computing. Here we argue that if we want to engineer unconventional computing systems in a systematic way, we need guidance from a formal theory that is different from the classical symbolic-algorithmic Turing machine theory. We propose a general strategy for developing such a theory, and within that general view, a specific approach that we call fluent computing. In contrast to Turing, who modeled computing processes from a top-down perspective as symbolic reasoning, we adopt the scientific paradigm of physics and model physical computing systems bottom-up by formalizing what can ultimately be measured in a physical computing system. This leads to an understanding of computing as the structuring of processes, while classical models of computing systems describe the processing of structures.

In-Materio Reservoir Computing in a Sulfonated Polyaniline Network.

A sulfonated polyaniline (SPAN) organic electrochemical network device (OEND) is fabricated using a simple drop-casting method on multiple Au electrodes for use in reservoir computing (RC). The SPAN network has humidity-dependent electrical properties. Under high humidity, the SPAN OEND exhibits mainly ionic conduction, including charging of an electric double layer and ionic diffusion. The nonlinearity and hysteresis of the current-voltage characteristics progressively increase with increasing humidity. The rich dynamic output behavior indicates wide variations for each electrode, which improves the RC performance because of the disordered network. For RC, waveform generation and short-term memory tasks are realized by a linear combination of outputs. The waveform task accuracy and memory capacity calculated from a short-term memory task reach 90% and 33.9, respectively. Improved spoken-digit classification is realized with 60% accuracy by only 12 outputs, demonstrating that the SPAN OEND can manage time series dynamic data operation in RC owing to a combination of rich dynamic and nonlinear electronic properties. The results suggest that SPAN-based electrochemical systems can be applied for material-based computing, by exploiting their intrinsic physicochemical behavior.

A deep-learning approach to realizing functionality in nanoelectronic devices.

Many nanoscale devices require precise optimization to function. Tuning them to the desired operation regime becomes increasingly difficult and time-consuming when the number of terminals and couplings grows. Imperfections and device-to-device variations hinder optimization that uses physics-based models. Deep neural networks (DNNs) can model various complex physical phenomena but, so far, are mainly used as predictive tools. Here, we propose a generic deep-learning approach to efficiently optimize complex, multi-terminal nanoelectronic devices for desired functionality. We demonstrate our approach for realizing functionality in a disordered network of dopant atoms in silicon. We model the input-output characteristics of the device with a DNN, and subsequently optimize control parameters in the DNN model through gradient descent to realize various classification tasks. When the corresponding control settings are applied to the physical device, the resulting functionality is as predicted by the DNN model. We expect our approach to contribute to fast, in situ optimization of complex (quantum) nanoelectronic devices.

Electron Trapping Mechanism in LaAlO_{3}/SrTiO_{3} Heterostructures.

In LaAlO_{3}/SrTiO_{3} heterostructures, a still poorly understood phenomenon is that of electron trapping in back-gating experiments. Here, by combining magnetotransport measurements and self-consistent Schrödinger-Poisson calculations, we obtain an empirical relation between the amount of trapped electrons and the gate voltage. The amount of trapped electrons decays exponentially away from the interface. However, contrary to earlier observations, we find that the Fermi level remains well within the quantum well. The enhanced trapping of electrons induced by the gate voltage can therefore not be explained by a thermal escape mechanism. Further gate sweeping experiments strengthen that conclusion. We propose a new mechanism which involves the electromigration and clustering of oxygen vacancies in SrTiO_{3} and argue that such electron trapping is a universal phenomenon in SrTiO_{3}-based two-dimensional electron systems.

Classification with a disordered dopant-atom network in silicon.

Classification is an important task at which both biological and artificial neural networks excel1,2. In machine learning, nonlinear projection into a high-dimensional feature space can make data linearly separable3,4, simplifying the classification of complex features. Such nonlinear projections are computationally expensive in conventional computers. A promising approach is to exploit physical materials systems that perform this nonlinear projection intrinsically, because of their high computational density5, inherent parallelism and energy efficiency6,7. However, existing approaches either rely on the systems' time dynamics, which requires sequential data processing and therefore hinders parallel computation5,6,8, or employ large materials systems that are difficult to scale up7. Here we use a parallel, nanoscale approach inspired by filters in the brain1 and artificial neural networks2 to perform nonlinear classification and feature extraction. We exploit the nonlinearity of hopping conduction9-11 through an electrically tunable network of boron dopant atoms in silicon, reconfiguring the network through artificial evolution to realize different computational functions. We first solve the canonical two-input binary classification problem, realizing all Boolean logic gates12 up to room temperature, demonstrating nonlinear classification with the nanomaterial system. We then evolve our dopant network to realize feature filters2 that can perform four-input binary classification on the Modified National Institute of Standards and Technology handwritten digit database. Implementation of our material-based filters substantially improves the classification accuracy over that of a linear classifier directly applied to the original data13. Our results establish a paradigm of silicon-based electronics for small-footprint and energy-efficient computation14.

Electrochemical Detection of Tumor-Derived Extracellular Vesicles on Nanointerdigitated Electrodes.

Tumor-derived extracellular vesicles (tdEVs) are attracting much attention due to their essential function in intercellular communication and their potential as cancer biomarkers. Although tdEVs are significantly more abundant in blood than other cancer biomarkers, their concentration compared to other blood components remains relatively low. Moreover, the presence of particles in blood with a similar size as that of tdEVs makes their selective and sensitive detection further challenging. Therefore, highly sensitive and specific biosensors are required for unambiguous tdEV detection in complex biological environments, especially for decentralized point-of-care analysis. Here, we report an electrochemical sensing scheme for tdEV detection, with two-level selectivity provided by a sandwich immunoassay and two-level amplification through the combination of an enzymatic assay and redox cycling on nanointerdigitated electrodes to respectively enhance the specificity and sensitivity of the assay. Analysis of prostate cancer cell line tdEV samples at various concentrations revealed an estimated limit of detection for our assay as low as 5 tdEVs/µL, as well as an excellent linear sensor response spreading over 6 orders of magnitude (10-106 tdEVs/µL), which importantly covers the clinically relevant range for tdEV detection in blood. This novel nanosensor and associated sensing scheme opens new opportunities to detect tdEVs at clinically relevant concentrations from a single blood finger prick.

Hard Superconducting Gap and Diffusion-Induced Superconductors in Ge-Si Nanowires.

We show a hard superconducting gap in a Ge-Si nanowire Josephson transistor up to in-plane magnetic fields of 250 mT, an important step toward creating and detecting Majorana zero modes in this system. A hard gap requires a highly homogeneous tunneling heterointerface between the superconducting contacts and the semiconducting nanowire. This is realized by annealing devices at 180 °C during which aluminum interdiffuses and replaces the germanium in a section of the nanowire. Next to Al, we find a superconductor with lower critical temperature (TC = 0.9 K) and a higher critical field (BC = 0.9-1.2 T). We can therefore selectively switch either superconductor to the normal state by tuning the temperature and the magnetic field and observe that the additional superconductor induces a proximity supercurrent in the semiconducting part of the nanowire even when the Al is in the normal state. In another device where the diffusion of Al rendered the nanowire completely metallic, a superconductor with a much higher critical temperature (TC = 2.9 K) and critical field (BC = 3.4 T) is found. The small size of these diffusion-induced superconductors inside nanowires may be of special interest for applications requiring high magnetic fields in arbitrary direction.

Versatile Micropatterns of N-Heterocyclic Carbenes on Gold Surfaces: Increased Thermal and Pattern Stability with Enhanced Conductivity.

Patterned monolayers of N-heterocyclic carbenes (NHCs) on gold surfaces were obtained by microcontact printing of NHC-CO2 adducts and NHC(H)[HCO3 ] salts. The NHC-modified areas showed an increased conductivity compared to unmodified gold surface areas. Furthermore, the remaining surface areas could be modified with a second, azide-functionalized carbene, facilitating further applications and post-printing modifications. Thorough elucidation by a variety of analytical methods offers comprehensive evidence for the viability of the methodology reported here. The protocol enables facile access to versatile, microstructured NHC-modified gold surfaces with highly stable patterns, enhanced conductivity, and the option for further modification.

Bottom-Up Single-Electron Transistors.

As the downscaling of conventional semiconductor electronics becomes more and more challenging, the interest in alternative material systems and fabrication methods is growing. A novel bottom-up approach for the fabrication of high-quality single-electron transistors (SETs) that can easily be contacted electrically in a controllable manner is developed. This approach employs the self-assembly of Au nanoparticles forming the SETs, and Au nanorods forming the leads to macroscopic electrodes, thus bridging the gap between the nano- and microscale. Low-temperature electron-transport measurements reveal exemplary single-electron tunneling characteristics. SET behavior can be significantly changed, post-fabrication, using molecular exchange of the tunnel barriers, demonstrating the tunability of the assemblies. These results form a promising proof of principle for the versatility of bottom-up nanoelectronics, and toward controlled fabrication of nanoelectronic devices.

Monolayer Contact Doping from a Silicon Oxide Source Substrate.

Monolayer contact doping (MLCD) is a modification of the monolayer doping (MLD) technique that involves monolayer formation of a dopant-containing adsorbate on a source substrate. This source substrate is subsequently brought into contact with the target substrate, upon which the dopant is driven into the target substrate by thermal annealing. Here, we report a modified MLCD process, in which we replace the commonly used Si source substrate by a thermally oxidized substrate with a 100 nm thick silicon oxide layer, functionalized with a monolayer of a dopant-containing silane. The thermal oxide potentially provides a better capping effect and effectively prevents the dopants from diffusing back into the source substrate. The use of easily accessible and processable silane monolayers provides access to a general and modifiable process for the introduction of dopants on the source substrate. As a proof of concept, a boron-rich carboranyl-alkoxysilane was used here to construct the monolayer that delivers the dopant, to boost the doping level in the target substrate. X-ray photoelectron spectroscopy (XPS) showed a successful grafting of the dopant adsorbate onto the SiO2 surface. The achieved doping levels after thermal annealing were similar to the doping levels acessible by MLD as demonstrated by secondary ion mass spectrometry measurements. The method shows good prospects, e.g. for use in the doping of Si nanostructures.

Highly doped silicon nanowires by monolayer doping.

Controlling the doping concentration of silicon nanostructures is challenging. Here, we investigated three different monolayer doping techniques to obtain silicon nanowires with a high doping dose. These routes were based on conventional monolayer doping, starting from covalently bound dopant-containing molecules, or on monolayer contact doping, in which a source substrate coated with a monolayer of a carborane silane was the dopant source. As a third route, both techniques were combined to retain the benefits of conformal monolayer formation and the use of an external capping layer. These routes were used for doping fragile porous nanowires fabricated by metal-assisted chemical etching. Differences in porosity were used to tune the total doping dose inside the nanowires, as measured by X-ray photoelectron spectroscopy and secondary ion mass spectrometry measurements. The higher the porosity, the higher was the surface available for dopant-containing molecules, which in turn led to a higher doping dose. Slightly porous nanowires could be doped via all three routes, which resulted in highly doped nanowires with (projected areal) doping doses of 1014-1015 boron atoms per cm2 compared to 1012 atoms per cm2 for a non-porous planar sample. Highly porous nanowires were not compatible with the conventional monolayer doping technique, but monolayer contact doping and the combined route resulted for these highly porous nanowires in tremendously high doping doses up to 1017 boron atoms per cm2.

Charge transport in nanoscale vertical organic semiconductor pillar devices.

We report charge transport measurements in nanoscale vertical pillar structures incorporating ultrathin layers of the organic semiconductor poly(3-hexylthiophene) (P3HT). P3HT layers with thickness down to 5 nm are gently top-contacted using wedging transfer, yielding highly reproducible, robust nanoscale junctions carrying high current densities (up to 106 A/m2). Current-voltage data modeling demonstrates excellent hole injection. This work opens up the pathway towards nanoscale, ultrashort-channel organic transistors for high-frequency and high-current-density operation.

Passivation and characterization of charge defects in ambipolar silicon quantum dots.

In this Report we show the role of charge defects in the context of the formation of electrostatically defined quantum dots. We introduce a barrier array structure to probe defects at multiple locations in a single device. We measure samples both before and after an annealing process which uses an Al2O3 overlayer, grown by atomic layer deposition. After passivation of the majority of charge defects with annealing we can electrostatically define hole quantum dots up to 180 nm in length. Our ambipolar structures reveal amphoteric charge defects that remain after annealing with charging energies of 10 meV in both the positive and negative charge state.

Effect of Orbital Hybridization on Spin-Polarized Tunneling across Co/C60 Interfaces.

The interaction between ferromagnetic surfaces and organic semiconductors leads to the formation of hybrid interfacial states. As a consequence, the local magnetic moment is altered, a hybrid interfacial density of states (DOS) is formed, and spin-dependent shifts of energy levels occur. Here, we show that this hybridization affects spin transport across the interface significantly. We report spin-dependent electronic transport measurements for tunnel junctions comprising C60 molecular thin films grown on top of face-centered-cubic (fcc) epitaxial Co electrodes, an AlOx tunnel barrier, and an Al counter electrode. Since only one ferromagnetic electrode (Co) is present, spin-polarized transport is due to tunneling anisotropic magnetoresistance (TAMR). An in-plane TAMR ratio of approximately 0.7% has been measured at 5 K under application of a magnetic field of 800 mT. The magnetic switching behavior shows some remarkable features, which are attributed to the rotation of interfacial magnetic moments. This behavior can be ascribed to the magnetic coupling between the Co thin films and the newly formed Co/C60 hybridized interfacial states. Using the Tedrow-Meservey technique, the tunnel spin polarization of the Co/C60 interface was found to be 43%.

Boosting the Boron Dopant Level in Monolayer Doping by Carboranes.

Monolayer doping (MLD) presents an alternative method to achieve silicon doping without causing crystal damage, and it has the capability of ultrashallow doping and the doping of nonplanar surfaces. MLD utilizes dopant-containing alkene molecules that form a monolayer on the silicon surface using the well-established hydrosilylation process. Here, we demonstrate that MLD can be extended to high doping levels by designing alkenes with a high content of dopant atoms. Concretely, carborane derivatives, which have 10 B atoms per molecule, were functionalized with an alkene group. MLD using a monolayer of such a derivative yielded up to ten times higher doping levels, as measured by X-ray photoelectron spectroscopy and dynamic secondary mass spectroscopy, compared to an alkene with a single B atom. Sheet resistance measurements showed comparably increased conductivities of the Si substrates. Thermal budget analyses indicate that the doping level can be further optimized by changing the annealing conditions.

Electron-Hole Confinement Symmetry in Silicon Quantum Dots.

We report electrical transport measurements on a gate-defined ambipolar quantum dot in intrinsic silicon. The ambipolarity allows its operation as either an electron or a hole quantum dot of which we change the dot occupancy by 20 charge carriers in each regime. Electron-hole confinement symmetry is evidenced by the extracted gate capacitances and charging energies. The results demonstrate that ambipolar quantum dots offer great potential for spin-based quantum information processing, since confined electrons and holes can be compared and manipulated in the same crystalline environment.

Controlling the dopant dose in silicon by mixed-monolayer doping.

Molecular monolayer doping (MLD) presents an alternative to achieve doping of silicon in a nondestructive way and holds potential for realizing ultrashallow junctions and doping of nonplanar surfaces. Here, we report the mixing of dopant-containing alkenes with alkenes that lack this functionality at various ratios to control the dopant concentration in the resulting monolayer and concomitantly the dopant dose in the silicon substrate. The mixed monolayers were grafted onto hydrogen-terminated silicon using well-established hydrosilylation chemistry. Contact angle measurements, X-ray photon spectroscopy (XPS) on the boron-containing monolayers, and Auger electron spectroscopy on the phosphorus-containing monolayers show clear trends as a function of the dopant-containing alkene concentration. Dynamic secondary-ion mass spectroscopy (D-SIMS) and Van der Pauw resistance measurements on the in-diffused samples show an effective tuning of the doping concentration in silicon.

Selective hydrothermal method to create patterned and photoelectrochemically effective Pt/WO3 interfaces.

A hydrothermal method based on the use of hydrogen peroxide is described to grow a homogeneous layer of tungsten oxide (WO3) on a platinum (Pt) film supported on a silicon wafer. WO3 growth is highly selective for Pt when present on silicon in a patterned arrangement, demonstrating that Pt catalyzes decomposition of the WO3 precursor in solution. The obtained Pt/WO3 interface yields high photocurrents of 1.1 mA/cm(2) in photoelectrochemical water splitting when illuminated by a solar simulator. The photocurrents are significantly higher than most previously reported values for hydrothermally grown layers on indium-tin oxide and fluorine-tin oxide glasses. The selective growth method thus provides new options to effectively implement WO3 in photoelectrochemical devices.