Large Scale Quantum Chemistry with Tensor Processing Units.

We demonstrate the use of Googles cloud-based Tensor Processing Units (TPUs) to accelerate and scale up conventional (cubic-scaling) density functional theory (DFT) calculations. Utilizing 512 TPU cores, we accomplish the largest such DFT computation to date, with 247848 orbitals, corresponding to a cluster of 10327 water molecules with 103270 electrons, all treated explicitly. Our work thus paves the way toward accessible and systematic use of conventional DFT, free of any system-specific constraints, at unprecedented scales.

Large-scale distributed linear algebra with tensor processing units.

We have repurposed Google tensor processing units (TPUs), application-specific chips developed for machine learning, into large-scale dense linear algebra supercomputers. The TPUs' fast intercore interconnects (ICIs), physically two-dimensional network topology, and high-bandwidth memory (HBM) permit distributed matrix multiplication algorithms to rapidly become computationally bound. In this regime, the matrix-multiply units (MXUs) dominate the runtime, yielding impressive scaling, performance, and raw size: Operating in float32 precision, a full 2,048-core pod of third-generation TPUs can multiply two matrices with linear size [Formula: see text] in about 2 min. Via curated algorithms emphasizing large, single-core matrix multiplications, other tasks in dense linear algebra can similarly scale. As examples, we present 1) QR decomposition; 2) resolution of linear systems; and 3) the computation of matrix functions by polynomial iteration, demonstrated by the matrix polar factorization.

Optimal light trapping in ultra-thin photonic crystal crystalline silicon solar cells.

Crystalline silicon is an attractive photovoltaic material because of its natural abundance, accumulated materials and process knowledge, and its appropriate band gap. To reduce cost, thin films of crystalline silicon can be used. This reduces the amount of material needed and allows material with shorter carrier diffusion lengths to be used. However, the indirect band gap of silicon requires that a light trapping approach be used to maximize optical absorption. Here, a photonic crystal (PC) based approach is used to maximize solar light harvesting in a 400 nm-thick silicon layer by tuning the coupling strength of incident radiation to quasiguided modes over a broad spectral range. The structure consists of a double layer PC with the upper layer having holes which have a smaller radius compared to the holes in the lower layer. We show that the spectrally averaged fraction of photons absorbed is increased 8-fold compared to a planar cell with equivalent volume of active material. This results in an enhancement of maximum achievable photocurrent density from 7.1 mA/cm(2) for an unstructured film to 21.8 mA/cm(2) for a film structured as the double layer photonic crystal. This photocurrent density value approaches the limit of 26.5 mA/cm(2), obtained using the Yablonovitch light trapping limit for the same volume of active material.

Characterization of collagen fibers in Bruch's membrane using chemical force microscopy.

Bruch's membrane is a layer composed of collagen fibers located just beneath the retina. This study validates a strategy used to map the morphological and adhesion characteristics of collagen fibers in Bruch's membrane. Atomic force microscopy tips were functionalized with different chemical groups and used to map the hydrophilic and hydrophobic regions on the surface of the eye tissue. The largest adhesion forces were observed when tips functionalized with NH(2) groups were used. The trend in the adhesion forces was rationalized based on the distribution of different functional groups in the triple-helical structure of the collagen fibers. The results of this study can be used to design more effective strategies to treat eye diseases such as age-related macular degeneration.

Study of the morphological and adhesion properties of collagen fibers in the Bruch's membrane.

The Bruch's membrane is located beneath the retina in vertebrate eyes. We have used atomic force microscopy to examine the morphological and adhesion properties of collagen fibers located in different portions of the membrane. The D-periodicity of the fibers was 62.54 +/- 4.25 nm and 63.78 +/- 4.14 nm for regions away from the optic nerve and close to it, respectively. The adhesion properties of the collagen fibers were evaluated using force volume imaging on a number of different eye samples. The adhesion force we recorded in regions away from the optic nerve was different compared to regions close to the optic nerve. The reported results allow us to understand the nanoscopic properties of connective tissues in the eye and are important for the design of new and improved biomaterials.