Ultrafast machine vision with 2D material neural network image sensors.

Machine vision technology has taken huge leaps in recent years, and is now becoming an integral part of various intelligent systems, including autonomous vehicles and robotics. Usually, visual information is captured by a frame-based camera, converted into a digital format and processed afterwards using a machine-learning algorithm such as an artificial neural network (ANN)1. The large amount of (mostly redundant) data passed through the entire signal chain, however, results in low frame rates and high power consumption. Various visual data preprocessing techniques have thus been developed2-7 to increase the efficiency of the subsequent signal processing in an ANN. Here we demonstrate that an image sensor can itself constitute an ANN that can simultaneously sense and process optical images without latency. Our device is based on a reconfigurable two-dimensional (2D) semiconductor8,9 photodiode10-12 array, and the synaptic weights of the network are stored in a continuously tunable photoresponsivity matrix. We demonstrate both supervised and unsupervised learning and train the sensor to classify and encode images that are optically projected onto the chip with a throughput of 20 million bins per second.

Band Nesting in Two-Dimensional Crystals: An Exceptionally Sensitive Probe of Strain.

Band nesting occurs when conduction and valence bands are approximately equispaced over regions in the Brillouin zone. In two-dimensional materials, band nesting results in singularities of the joint density of states and thus in a strongly enhanced optical response at resonant frequencies. We exploit the high sensitivity of such resonances to small changes in the band structure to sensitively probe strain in semiconducting transition metal dichalcogenides (TMDs). We measure and calculate the polarization-resolved optical second harmonic generation (SHG) at the band nesting energies and present the first measurements of the energy-dependent nonlinear photoelastic effect in atomically thin TMDs (MoS2, MoSe2, WS2, and WSe2) combined with a theoretical analysis of the underlying processes. Experiment and theory are found to be in good qualitative agreement displaying a strong energy dependence of the SHG, which can be exploited to achieve exceptionally strong modulation of the SHG under strain. We attribute this sensitivity to a redistribution of the joint density of states for the optical response in the band nesting region. We predict that this exceptional strain sensitivity is a general property of all 2D materials with band nesting.

Sparse pixel image sensor.

As conventional frame-based cameras suffer from high energy consumption and latency, several new types of image sensors have been devised, with some of them exploiting the sparsity of natural images in some transform domain. Instead of sampling the full image, those devices capture only the coefficients of the most relevant spatial frequencies. The number of samples can be even sparser if a signal only needs to be classified rather than being fully reconstructed. Based on the corresponding mathematical framework, we developed an image sensor that can be trained to classify optically projected images by reading out the few most relevant pixels. The device is based on a two-dimensional array of metal-semiconductor-metal photodetectors with individually tunable photoresponsivity values. We demonstrate its use for the classification of handwritten digits with an accuracy comparable to that achieved by readout of the full image, but with lower delay and energy consumption.

A photosensor employing data-driven binning for ultrafast image recognition.

Pixel binning is a technique, widely used in optical image acquisition and spectroscopy, in which adjacent detector elements of an image sensor are combined into larger pixels. This reduces the amount of data to be processed as well as the impact of noise, but comes at the cost of a loss of information. Here, we push the concept of binning to its limit by combining a large fraction of the sensor elements into a single "superpixel" that extends over the whole face of the chip. For a given pattern recognition task, its optimal shape is determined from training data using a machine learning algorithm. We demonstrate the classification of optically projected images from the MNIST dataset on a nanosecond timescale, with enhanced dynamic range and without loss of classification accuracy. Our concept is not limited to imaging alone but can also be applied in optical spectroscopy or other sensing applications.

Intrinsic donor-bound excitons in ultraclean monolayer semiconductors.

The monolayer transition metal dichalcogenides are an emergent semiconductor platform exhibiting rich excitonic physics with coupled spin-valley degree of freedom and optical addressability. Here, we report a new series of low energy excitonic emission lines in the photoluminescence spectrum of ultraclean monolayer WSe2. These excitonic satellites are composed of three major peaks with energy separations matching known phonons, and appear only with electron doping. They possess homogenous spatial and spectral distribution, strong power saturation, and anomalously long population (>6 µs) and polarization lifetimes (>100 ns). Resonant excitation of the free inter- and intravalley bright trions leads to opposite optical orientation of the satellites, while excitation of the free dark trion resonance suppresses the satellites' photoluminescence. Defect-controlled crystal synthesis and scanning tunneling microscopy measurements provide corroboration that these features are dark excitons bound to dilute donors, along with associated phonon replicas. Our work opens opportunities to engineer homogenous single emitters and explore collective quantum optical phenomena using intrinsic donor-bound excitons in ultraclean 2D semiconductors.

Optical imaging of strain in two-dimensional crystals.

Strain engineering is widely used in material science to tune the (opto-)electronic properties of materials and enhance the performance of devices. Two-dimensional atomic crystals are a versatile playground to study the influence of strain, as they can sustain very large deformations without breaking. Various optical techniques have been employed to probe strain in two-dimensional materials, including micro-Raman and photoluminescence spectroscopy. Here we demonstrate that optical second harmonic generation constitutes an even more powerful technique, as it allows extraction of the full strain tensor with a spatial resolution below the optical diffraction limit. Our method is based on the strain-induced modification of the nonlinear susceptibility tensor due to a photoelastic effect. Using a two-point bending technique, we determine the photoelastic tensor elements of molybdenum disulfide. Once identified, these parameters allow us to spatially image the two-dimensional strain field in an inhomogeneously strained sample.