No abstract

Three types of iron oxide Janus particles are obtained by varying the deposition rate of iron in a 3:1 Ar/O(2) atmosphere during physical vapor deposition. Each type of iron oxide Janus particle shows a distinct assembly behavior when an external magnetic field is applied, i.e., formation of staggered chains, double chains, or no assembly. A detailed deposition rate diagram is obtained to identify the relationship between deposition rate and assembly behavior. The extent of iron oxidation is identified as the key parameter in determining the assembly behavior. In addition, the effects of particle volume fraction, thickness of the iron oxide cap, and assembly time on the final assembly behavior are studied. Cap thickness is shown not to influence the assembly behavior, while particle volume fraction and assembly time affect the chain growth rate and the chain length, but not the overall assembly behavior. The samples are characterized by optical, scanning electron, and atomic force microscopies.

With the emergence of a spectrum of high-end mobile devices, many applications that formerly required desktop-level computation capability are being transferred to these devices. However, executing Deep Neural Networks (DNNs) inference is still challenging considering the high computation and storage demands, specifically, if real-time performance with high accuracy is needed. Weight pruning of DNNs is proposed, but existing schemes represent two extremes in the design space: non-structured pruning is fine-grained, accurate, but not hardware friendly; structured pruning is coarse-grained, hardware-efficient, but with higher accuracy loss.In this paper, we advance the state-of-the-art by introducing a new dimension, fine-grained pruning patterns inside the coarse-grained structures, revealing a previously unknown point in the design space. With the higher accuracy enabled by fine-grained pruning patterns, the unique insight is to use the compiler to re-gain and guarantee high hardware efficiency. In other words, our method achieves the best of both worlds, and is desirable across theory/algorithm, compiler, and hardware levels. The proposed PatDNN is an endto-end framework to efficiently execute DNN on mobile devices with the help of a novel model compression techniquepattern-based pruning based on an extended ADMM solution framework-and a set of thorough architecture-aware compiler/code generation-based optimizations, i.e., filter kernel reordering, compressed weight storage, register load redundancy elimination, and parameter auto-tuning. Evaluation results demonstrate that PatDNN outperforms three state-ofthe-art end-to-end DNN frameworks, TensorFlow Lite, TVM, and Alibaba Mobile Neural Network with speedup up to 44.5×, 11.4×, and 7.1×, respectively, with no accuracy compromise. Real-time inference of representative large-scale DNNs (e.g., VGG-16, ResNet-50) can be achieved using mobile devices.

Manganese-doped magnetite nanoparticles as magnetic resonance imaging (MRI) contrast agents have been well developed in recent years due to their higher saturation magnetization and stronger transverse (T 2 ) contrast ability compared to parent magnetite. However, the underlying role that manganese doping plays in altering the contrast ability of magnetite is still not thoroughly understood. Herein, we investigate the effects of manganese doping on changes of ferrite crystal structures, magnetic properties, and contrast abilities. We developed a successful one-pot synthesis of uniform manganese-doped magnetite (Mn x Fe 3−x O 4 ) nanoparticles with different manganese contents (x = 0−1.06). The saturation magnetization and T 2 contrast ability of ferrite nanoparticles increase along with rising manganese proportion, peak when the doping level of Mn x Fe 3−x O 4 reaches x = 0.43, and decrease dramatically as the manganese percentage continues to augment. At high manganese doping level, the manganese ferrite nanoparticles may undergo lattice distortion according to analysis of XRD patterns and lattice distances, which may result in low saturation magnetization and eventually low T 2 contrast ability. The Mn x Fe 3−x O 4 nanoparticles (x = 0.43) with a diameter of ∼18.5 nm exhibit the highest T 2 relaxivity of 904.4 mM −1 s −1 at 7.0 T among all the samples and show a much stronger T 2 contrast effect for liver imaging than that of other iron oxide contrast agents. These results indicate that the optimized T 2 contrast ability of manganese ferrite nanoparticles could be achieved by tuning the manganese doping level. This work also opens a new field of vision for developing high-performance T 2 contrast agents by modulating the metal composition of nanoparticles.