Large field of view aberrations correction with deformable lenses and multi conjugate adaptive optics

Optical microscopes can have limited resolution due to aberrations caused by samples and sample holders. Using deformable mirrors and wavefront sensorless optimization algorithms can correct these aberrations, but the correction is limited to a small area of the field of view. This study presents an adaptive optics method that uses a series of plug-and-play deformable lenses for large field of view wavefront correction. A direct wavefront measurement method using the spinning sub-pupil aberration measurement technique is combined with correction based on the deformable lenses. Experimental results using fluorescence microscopy with a wide field and a light sheet fluorescence microscope show that the proposed method can achieve detection and correction over an extended field of view with a compact transmissive module placed in the detection path of the microscope. This method could improve the resolution and accuracy of imaging in a variety of fields, including biology and materials science.

Efficient and self-adaptive in-situ learning in multilayer memristor neural networks

Memristors with tunable resistance states are emerging building blocks of artificial neural networks. However, in situ learning on a large-scale multiple-layer memristor network has yet to be demonstrated because of challenges in device property engineering and circuit integration. Here we monolithically integrate hafnium oxide-based memristors with a foundry-made transistor array into a multiple-layer neural network. We experimentally demonstrate in situ learning capability and achieve competitive classification accuracy on a standard machine learning dataset, which further confirms that the training algorithm allows the network to adapt to hardware imperfections. Our simulation using the experimental parameters suggests that a larger network would further increase the classification accuracy. The memristor neural network is a promising hardware platform for artificial intelligence with high speed-energy efficiency.

Efficient and self-adaptive in-situ learning in multilayer memristor neural networks

Memristors with tunable resistance states are emerging building blocks of artificial neural networks. However, in situ learning on a large-scale multiple-layer memristor network has yet to be demonstrated because of challenges in device property engineering and circuit integration. Here we monolithically integrate hafnium oxide-based memristors with a foundry-made transistor array into a multiple-layer neural network. We experimentally demonstrate in situ learning capability and achieve competitive classification accuracy on a standard machine learning dataset, which further confirms that the training algorithm allows the network to adapt to hardware imperfections. Our simulation using the experimental parameters suggests that a larger network would further increase the classification accuracy. The memristor neural network is a promising hardware platform for artificial intelligence with high speed-energy efficiency.

Covalent-supramolecular hybrid polymers as muscle-inspired anisotropic actuators

Skeletal muscle provides inspiration on how to achieve reversible, macroscopic, anisotropic motion in soft materials. Here we report on the bottom-up design of macroscopic tubes that exhibit anisotropic actuation driven by a thermal stimulus. The tube is built from a hydrogel in which extremely long supramolecular nanofibers are aligned using weak shear forces, followed by radial growth of thermoresponsive polymers from their surfaces. The hierarchically ordered tube exhibits reversible anisotropic actuation with changes in temperature, with much greater contraction perpendicular to the direction of nanofiber alignment. We identify two critical factors for the anisotropic actuation, macroscopic alignment of the supramolecular scaffold and its covalent bonding to polymer chains. Using finite element analysis and molecular calculations, we conclude polymer chain confinement and mechanical reinforcement by rigid supramolecular nanofibers are responsible for the anisotropic actuation. The work reported suggests strategies to create soft active matter with molecularly encoded capacity to perform complex tasks.