Search

Atomically thin semiconductor heterostructures provide a two-dimensional (2D) device platform for creating high densities of cold, controllable excitons. Interlayer excitons (IEs), bound electrons and holes localized to separate 2D quantum well layers, have permanent out-of-plane dipole moments and long lifetimes, allowing their spatial distribution to be tuned on demand. Here, we employ electrostatic gates to trap IEs and control their density. By electrically modulating the IE Stark shift, electron-hole pair concentrations above $2\times10^{12}$ cm$^{-2}$ can be achieved. At this high IE density, we observe an exponentially increasing linewidth broadening indicative of an IE ionization transition, independent of the trap depth. This runaway threshold remains constant at low temperatures, but increases above 20 K, consistent with the quantum dissociation of a degenerate IE gas. Our demonstration of the IE ionization in a tunable electrostatic trap represents an important step towards the realization of dipolar exciton condensates in solid-state optoelectronic devices., Comment: 14 pages, 4 main figures, 1 extended data figure

Employing flux-grown single crystal WSe$_2$, we report charge carrier scattering behaviors measured in $h$-BN encapsulated monolayer field effect transistors. We perform quantum transport measurements across various hole densities and temperatures and observe a non-monotonic change of transport mobility $\mu$ as a function of hole density in the degenerately doped sample. This unusual behavior can be explained by energy dependent scattering amplitude of strong defects calculated using the T-matrix approximation. Utilizing long mean-free path ($>$500 nm), we demonstrate the high quality of our electronic devices by showing quantized conductance steps from an electrostatically-defined quantum point contact. Our results show the potential for creating ultra-high quality quantum optoelectronic devices based on atomically thin semiconductors., Comment: 6 pages, 4 figures

Deep learning in computational biochemistry has traditionally focused on molecular graphs neural representations; however, recent advances in language models highlight how much scientific knowledge is encoded in text. To bridge these two modalities, we investigate how molecular property information can be transferred from natural language to graph representations. We study property prediction performance gains after using contrastive learning to align neural graph representations with representations of textual descriptions of their characteristics. We implement neural relevance scoring strategies to improve text retrieval, introduce a novel chemically-valid molecular graph augmentation strategy inspired by organic reactions, and demonstrate improved performance on downstream MoleculeNet property classification tasks. We achieve a +4.26% AUROC gain versus models pre-trained on the graph modality alone, and a +1.54% gain compared to recently proposed molecular graph/text contrastively trained MoMu model (Su et al. 2022)., Comment: 2023 ICML Workshop on Computational Biology

Language models can be prompted to reason through problems in a manner that significantly improves performance. However, \textit{why} such prompting improves performance is unclear. Recent work showed that using logically \textit{invalid} Chain-of-Thought (CoT) prompting improves performance almost as much as logically \textit{valid} CoT prompting, and that editing CoT prompts to replace problem-specific information with abstract information or out-of-distribution information typically doesn't harm performance. Critics have responded that these findings are based on too few and too easily solved tasks to draw meaningful conclusions. To resolve this dispute, we test whether logically invalid CoT prompts offer the same level of performance gains as logically valid prompts on the hardest tasks in the BIG-Bench benchmark, termed BIG-Bench Hard (BBH). We find that the logically \textit{invalid} reasoning prompts do indeed achieve similar performance gains on BBH tasks as logically valid reasoning prompts. We also discover that some CoT prompts used by previous works contain logical errors. This suggests that covariates beyond logically valid reasoning are responsible for performance improvements., Comment: ICML 2023 Workshop: Knowledge and Logical Reasoning in the Era of Data-driven Learning

Double descent is a surprising phenomenon in machine learning, in which as the number of model parameters grows relative to the number of data, test error drops as models grow ever larger into the highly overparameterized (data undersampled) regime. This drop in test error flies against classical learning theory on overfitting and has arguably underpinned the success of large models in machine learning. This non-monotonic behavior of test loss depends on the number of data, the dimensionality of the data and the number of model parameters. Here, we briefly describe double descent, then provide an explanation of why double descent occurs in an informal and approachable manner, requiring only familiarity with linear algebra and introductory probability. We provide visual intuition using polynomial regression, then mathematically analyze double descent with ordinary linear regression and identify three interpretable factors that, when simultaneously all present, together create double descent. We demonstrate that double descent occurs on real data when using ordinary linear regression, then demonstrate that double descent does not occur when any of the three factors are ablated. We use this understanding to shed light on recent observations in nonlinear models concerning superposition and double descent. Code is publicly available.

Moir\'e materials provide fertile ground for the correlated and topological quantum phenomena. Among them, the quantum anomalous Hall (QAH) effect, in which the Hall resistance is quantized even under zero magnetic field, is a direct manifestation of the intrinsic topological properties of a material and an appealing attribute for low-power electronics applications. The QAH effect has been observed in both graphene and transition metal dichalcogenide (TMD) moir\'e materials. It is thought to arise from the interaction-driven valley polarization of the narrow moir\'e bands. Here, we show surprisingly that the newly discovered QAH state in AB-stacked MoTe2/WSe2 moir\'e bilayers is not valley-polarized but valley-coherent. The layer- and helicity-resolved optical spectroscopy measurement reveals that the QAH ground state possesses spontaneous spin (valley) polarization aligned (anti-aligned) in two TMD layers. In addition, saturation of the out-of-plane spin polarization in both layers occurs only under high magnetic fields, supporting a canted spin texture. Our results call for a new mechanism for the QAH effect and highlight the potential of TMD moir\'e materials with strong electronic correlations and spin-orbit interactions for exotic topological states.

Excitons are composite bosons that can feature spin singlet and triplet states. In usual semiconductors, without an additional spin-flip mechanism, triplet excitons are extremely inefficient optical emitters. Transition metal dichalcogenides (TMDs), with their large spin-orbit coupling, have been of special interest for valleytronic applications for their coupling of circularly polarized light to excitons with selective valley and spin$^{1-4}$. In atomically thin MoSe$_2$/WSe$_2$ TMD van der Waals (vdW) heterostructures, the unique atomic registry of vdW layers provides a quasi-angular momentum to interlayer excitons$^{5,6}$, enabling emission from otherwise dark spin triplet excitons. Here, we report electrically tunable spin singlet and triplet exciton emission from atomically aligned TMD heterostructures. We confirm the spin configurations of the light-emitting excitons employing magnetic fields to measure effective exciton g-factors. The interlayer tunneling current across the TMD vdW heterostructure enables the electrical generation of singlet and triplet exciton emission in this atomically thin PN junction. We demonstrate electrically tunability between the singlet and triplet excitons that are generated by charge injection. Atomically thin TMD heterostructure light emitting diodes thus enables a route for optoelectronic devices that can configure spin and valley quantum states independently by controlling the atomic stacking registry., Comment: Extended work of previous study: arXiv:1912.07678

Excitons are composite bosons that can feature spin singlet and triplet states. In usual semiconductors, without an additional spin-flip mechanism, triplet excitons are extremely inefficient optical emitters. Large spin-orbit coupling in transition metal dichalcogenides (TMDs) couples circularly polarized light to excitons with selective valley and spin. Here, we demonstrate electrically controlled brightening of spin-triplet interlayer excitons in a MoSe$_2$/WSe$_2$ TMD van der Waals (vdW) heterostructure. The atomic registry of vdW layers in TMD heterostructures provides a quasi-angular momentum to interlayer excitons, enabling emission from otherwise dark spin-triplet excitons. Employing magnetic field, we show that photons emitted by triplet and singlet excitons in the same valley have opposite chirality. We also measure effective exciton g-factors, presenting direct and quantitative evidence of triplet interlayer excitons. We further demonstrate gate tuning of the relative photoluminescence intensity between singlet and triplet charged excitons. Electrically controlled emission between singlet and triplet excitons enables a route for optoelectronic devices that can configure excitonic chiral, spin, and valley quantum states.

Excitons in semiconductors, bound pairs of excited electrons and holes, can form the basis for new classes of quantum optoelectronic devices. A van der Waals heterostructure built from atomically thin semiconducting transition metal dichalcogenides (TMDs) enables the formation of excitons from electrons and holes in distinct layers, producing interlayer excitons with large binding energy and a long lifetime. Employing heterostructures of monolayer TMDs, we realize optical and electrical generation of long-lived neutral and charged interlayer excitons. We demonstrate the transport of neutral interlayer excitons across the whole sample that can be controlled by excitation power and gate electrodes. We also realize the drift motion of charged interlayer excitons using Ohmic-contacted devices. The electrical generation and control of excitons provides a new route for realizing quantum manipulation of bosonic composite particles with complete electrical tunability.

Guided modes in anisotropic two-dimensional van der Waals materials are experimentally investigated and their refractive indices in visible wavelengths are extracted. Our method involves near-field scanning optical microscopy of waveguide (transverse electric) and surface plasmon polariton (transverse magnetic) modes in h-BN/SiO2/Si and Ag/h-BN stacks, respectively. We determine the dispersion of these modes and use this relationship to extract anisotropic refractive indices of h-BN flakes. In the wavelength interval 550-700 nm, the in-plane and out-of-plane refractive indices are in the range 1.98-2.12 and 1.45-2.12, respectively. Our approach of using near-field scanning optical microscopy allows for direct study of interaction between light and two-dimensional van der Waals materials and heterostructures.

Transition metal dichalcogenide (TMD) monolayers are direct bandgap semiconductors that feature tightly bound excitons, strong spin-orbit coupling, and spin-valley degrees of freedom. Depending on the spin configuration of the electron-hole pairs, intra-valley excitons of TMD monolayers can be either optically bright or dark. Dark excitons involve nominally spin-forbidden optical transitions with zero in-plane transition dipole moment, making their detection with conventional far-field optical techniques challenging. Here, we introduce a new method for probing the optical properties of two-dimensional (2D) materials via near-field coupling to surface plasmon polaritons (SPPs), which selectively enhances optical transitions with dipole moments normal to the 2D plane. We utilize this method to directly detect dark excitons in monolayer TMDs. When a WSe2 monolayer is placed on top of a single-crystal silver film, its emission into near-field-coupled SPPs displays new spectral features whose energies and dipole orientations are consistent with dark neutral and charged excitons. The SPP-based near-field spectroscopy significantly enhances experimental capabilities for probing and manipulating exciton dynamics of atomically thin materials., Comment: 4 figures