Multifunctional Polymer Restraint of the Agglomeration of SnO2 Nanocrystals for Efficient and Stable Planar Perovskite Solar Cells.

The aggregation of SnO2 nanocrystals due to van der Waals interactions is not conducive to the realization of a compact and pinhole-free electron transport layer (ETL). Herein, we have utilized potassium alginate (PA) to self-assemble SnO2 nanocrystals, forming a PA-SnO2 ETL for perovskite solar cells (PSCs). Through density functional theory (DFT) calculations, PA can be effectively absorbed onto the surface of SnO2. This inhibits the agglomeration of SnO2 nanocrystals in solution, forming a smoother pinhole-free film. This also changes the surface contact potential (CPD) of the SnO2 film, which leads to a reduction in the energy barrier between the ETL and the perovskite layers, promotes effective charge transfer, and reduces trap density. Consequently, the power conversion efficiency (PCE) of PSCs with a PA-SnO2 ETL increased from 19.24% to 22.16%, and the short-circuit current (JSC) was enhanced from 23.52 to 25.21 mA cm-2. Furthermore, the PA-modified unpackaged device demonstrates better humidity stability compared to the original device.

Acoustic quasi-periodic bioassembly based diverse stem cell arrangements for differentiation guidance.

Arrangement patterns and geometric cues have been demonstrated to influence cell function and fate, which calls for efficient and versatile cell patterning techniques. Despite constant achievements that mainly focus on individual cells and uniform cell patterns, simultaneously constructing cellular arrangements with diverse patterns and positional relationships in a flexible and contact-free manner remains a challenge. Here, stem cell arrangements possessing multiple geometries and structures are proposed based on powerful and diverse pattern-building capabilities of quasi-periodic acoustic fields, with advantages of rich patterns and structures and flexibility in structure modulation. Eight-fold waves' interference produces regular potentials that result in higher rotational symmetry and more complex arrangement of geometric units. Moreover, through flexible modulation of the phase relations among these wave vectors, a wide variety of cellular pattern units are arranged in this potential, such as circular-, triangular- and square-shape, simultaneously. It is proved that these diverse cellular patterns conveniently build human mesenchymal stem cell (hMSC) models, for research on the effect of cellular arrangement on stem cell differentiation. This work fills the gap of acoustic cell patterning in quasi-periodic patterns and shows promising potential in tissue engineering and regenerative medicine.

Heterogeneous tissue construction by on-demand bubble-assisted acoustic patterning.

Highly heterogeneous structures are closely related to the realization of the tissue functions of living organisms. However, precisely controlling the assembly of heterogeneous structures is still a crucial challenge. This work presents an on-demand bubble-assisted acoustic method for active cell patterning to achieve high-precision heterogeneous structures. Active cell patterning is achieved by the combined effect of acoustic radiation forces and microstreaming around oscillating bubble arrays. On-demand bubble arrays allow flexible construction of cell patterns with a precision of up to 45 µm. As a typical example, the in vitro model of hepatic lobules, composed of patterned endothelial cells and hepatic parenchymal cells, was constructed and cultured for 5 days. The good performance of urea and albumin secretion, enzymatic activity and good proliferation of both cells prove the feasibility of this technique. Overall, this bubble-assisted acoustic approach provides a simple and efficient strategy for on-demand large-area tissue construction, with considerable potential for different tissue model fabrication.

ECBC: Efficient Convolution via Blocked Columnizing.

Direct convolution methods are now drawing increasing attention as they eliminate the additional storage demand required by indirect convolution algorithms (i.e., the transformed matrix generated by the im2col convolution algorithm). Nevertheless, the direct methods require special input-output tensor formatting, leading to extra time and memory consumption to get the desired data layout. In this article, we show that indirect convolution, if implemented properly, is able to achieve high computation performance with the help of highly optimized subroutines in matrix multiplication while avoid incurring substantial memory overhead. The proposed algorithm is called efficient convolution via blocked columnizing (ECBC). Inspired by the im2col convolution algorithm and the block algorithm of general matrix-to-matrix multiplication, we propose to conduct the convolution computation blockwisely. As a result, the tensor-to-matrix transformation process (e.g., the im2col operation) can also be done in a blockwise manner so that it only requires a small block of memory as small as the data block. Extensive experiments on various platforms and networks validate the effectiveness of ECBC, as well as the superiority of our proposed method against a set of widely used industrial-level convolution algorithms.

Tunable and Dynamic Optofluidic Microlens Arrays Based on Droplets.

Microlens arrays (MLAs) are acquiring a key role in the micro-optical system, which have been widely applied in the fields of imaging processing, light extraction, biochemical sensing, and display technology. Compared with solid MLAs, liquid MLAs have received extensive attention due to their natural smooth interface and adjustability. However, manufacturing tunable liquid MLAs with ideal structures is still a key challenge for current technologies. In this paper, a novel and simple optofluidic method is demonstrated, enabling the tunable focusing and high-quality imaging of liquid MLAs. Tunable droplets are fabricated and self-assembled into arrays as the MLAs, which can be easily adjusted to focus, form images, and display different focal lengths. Tuning of MLAs' focusing properties (range from 550 to 5370 µm) is demonstrated by changing the refractive index (RI) of the droplets with a fixed size of 200 µm, which can be changed by adjusting the flow rates of the two branch streams. Also, the corresponding numerical apertures of the MLAs range from 0.026 to 0.26. Furthermore, the MLAs' functionality for microparticle imaging applications is also illustrated. Combining the MLAs with a 4× objective, microparticle imaging is magnified two times, and the resolution has also been improved on the original basis. Besides, both the size and RI of the MLAs in an optofluidic chip can be further adjusted to detect samples at different positions. These MLAs have the merits of high optical performance, a simple fabrication procedure, easy integration, and good tunability. Thus, it shows promising opportunities for many applications, such as adaptive imaging and sensing.

On-demand liquid microlens arrays by non-contact relocation of inhomogeneous fluids in acoustic fields.

Microlens arrays (MLAs) are key micro-optical components that possess a high degree of parallelism and ease of integration. However, rapid and low-cost fabrication of MLAs with flexible focusing remains a challenge. Herein, liquid MLAs with dynamic tunability are presented using non-contact acoustic relocation of inhomogeneous fluids. By designing ring-shaped acoustic pressure node (PN) arrays, the denser fluid of miscible liquids is relocated to PNs, and liquid MLAs with ideal morphology are obtained. The experimental results demonstrate that the liquid MLAs possess a powerful reconfigurability with long-term stability and sharp imaging that can conveniently switch between the on and off state and can dynamically magnify by simply adjusting the acoustic amplitude. Moreover, the high biocompatibility inherited from liquids accompanied by the acoustic treatment allows cells to be within working distance of the MLAs without immersion, as would be required for a solid lens. This innovative liquid MLA is inexpensive to manufacture and possesses continuous focus, fast response, and satisfactory bioaffinity, and thus offers promising potential for microfluidic adaptive imaging and biomedical sensing, especially for live cell imaging.

Smart acoustic 3D cell construct assembly with high-resolution.

Precise and flexible three-dimensional (3D) cell construct assembly using external forces or fields can produce micro-scale cellular architectures with intercellular connections, which is an important prerequisite to reproducing the structures and functions of biological systems. Currently, it is also a substantial challenge in the bioengineering field. Here, we propose a smart acoustic 3D cell assembly strategy that utilizes a 3D printed module and hydrogel sheets. Digitally controlled six wave beams offer a high degree of freedom (including wave vector combination, frequency, phase, and amplitude) that enables versatile biomimetic micro cellular patterns in hydrogel sheets. Further, replaceable frames can be used to fix the acoustic-built micro-scale cellular structures in these sheets, enabling user-defined hierarchical or heterogeneous constructs through layer-by-layer assembly. This strategy can be employed to construct vasculature with different diameters and lengths, composed of human umbilical vein endothelial cells and smooth muscle cells. These constructs can also induce controllable vascular network formation. Overall, the findings of this work extend the capabilities of acoustic cell assembly into 3D space, offering advantages including innovative, flexible, and precise patterning, and displaying great potential for the manufacture of various artificial tissue structures that duplicatein vivofunctions.

Digital optofluidic compound eyes with natural structures and zooming capability for large-area fluorescence sensing.

Compound eyes are ubiquitous natural biosensors that possess high temporal resolution and large fields of view (FOVs). While for solid materials based artificial imaging systems, flexible zooming ability while keeping the constant FOV is still challenging, as well as the low-cost fabrication. Herein, liquid compound eyes with natural structures are presented that synthesize optofluidics and bionics in a non-trivial manner, which enables the deformation-free zooming and flexible cell fluorescence sensing. Experimental results indicate that the innovatively manufactured bionic template possesses low roughness and uniform lens configuration with more than two thousands units, which endows the eyes with high-quality and low aberration imaging ability. Besides, digital controlled miscible liquids switching enables the focus of ommatidia simultaneously be adjusted from 150 µm to 5 mm with 100° view angle, and without bending the microlens curvature, to avoid FOV changing and image aberration. Due to large FOV and tunable ability, large-area cell fluorescence signal arrays and dynamic cell motion are imaged using this liquid compound eyes. This work presents novel strategy for compound lens manufacture at low-cost, and proposes deformation-free and continuous focus-tuning strategy, offering potentials for numerous applications, including biomedical sensing and adaptive imaging with large FOV.

Adversarial Binary Mutual Learning for Semi-Supervised Deep Hashing.

Hashing is a popular search algorithm for its compact binary representation and efficient Hamming distance calculation. Benefited from the advance of deep learning, deep hashing methods have achieved promising performance. However, those methods usually learn with expensive labeled data but fail to utilize unlabeled data. Furthermore, the traditional pairwise loss used by those methods cannot explicitly force similar/dissimilar pairs to small/large distances. Both weaknesses limit existing methods' performance. To solve the first problem, we propose a novel semi-supervised deep hashing model named adversarial binary mutual learning (ABML). Specifically, our ABML consists of a generative model GH and a discriminative model DH , where DH learns labeled data in a supervised way and GH learns unlabeled data by synthesizing real images. We adopt an adversarial learning (AL) strategy to transfer the knowledge of unlabeled data to DH by making GH and DH mutually learn from each other. To solve the second problem, we propose a novel Weibull cross-entropy loss (WCE) by using the Weibull distribution, which can distinguish tiny differences of distances and explicitly force similar/dissimilar distances as small/large as possible. Thus, the learned features are more discriminative. Finally, by incorporating ABML with WCE loss, our model can acquire more semantic and discriminative features. Extensive experiments on four common data sets (CIFAR-10, large database of handwritten digits (MNIST), ImageNet-10, and NUS-WIDE) and a large-scale data set ImageNet demonstrate that our approach successfully overcomes the two difficulties above and significantly outperforms state-of-the-art hashing methods.

On-chip rapid drug screening of leukemia cells by acoustic streaming.

Rapid and personalized single-cell drug screening testing plays an essential role in acute myeloid leukemia drug combination chemotherapy. Conventional chemotherapeutic drug screening is a time-consuming process because of the natural resistance of cell membranes to drugs, and there are still great challenges related to using technologies that change membrane permeability such as sonoporation in high-throughput and precise single-cell drug screening with minimal damage. In this study, we proposed an acoustic streaming-based non-invasive single-cell drug screening acceleration method, using high-frequency acoustic waves (>10 MHz) in a concentration gradient microfluidic device. High-frequency acoustics leads to increased difficulties in inducing cavitation and generates acoustic streaming around each single cell. Therefore, single-cell membrane permeability is non-invasively increased by the acoustic pressure and acoustic streaming-induced shear force, which significantly improves the drug uptake process. In the experiment, single human myeloid leukemia mononuclear (THP-1) cells were trapped by triangle cell traps in concentration gradient chips with different cytarabine (Ara-C) drug concentrations. Due to this dual acoustic effect, the drugs affect cell viability in less than 30 min, which is faster than traditional methods (usually more than 24 h). This dual acoustic effect-based drug delivery strategy has the potential to save time and reduce the cost of drug screening, when combined with microfluidic technology for multi-concentration drug screening. This strategy offers enormous potential for use in multiple drug screening or efficient drug combination screening in individualized/personalized treatments, which can greatly improve efficiency and reduce costs.

Quantized CNN: A Unified Approach to Accelerate and Compress Convolutional Networks.

We are witnessing an explosive development and widespread application of deep neural networks (DNNs) in various fields. However, DNN models, especially a convolutional neural network (CNN), usually involve massive parameters and are computationally expensive, making them extremely dependent on high-performance hardware. This prohibits their further extensions, e.g., applications on mobile devices. In this paper, we present a quantized CNN, a unified approach to accelerate and compress convolutional networks. Guided by minimizing the approximation error of individual layer's response, both fully connected and convolutional layers are carefully quantized. The inference computation can be effectively carried out on the quantized network, with much lower memory and storage consumption. Quantitative evaluation on two publicly available benchmarks demonstrates the promising performance of our approach: with comparable classification accuracy, it achieves 4 to $6 \times $ acceleration and 15 to $20\times $ compression. With our method, accurate image classification can even be directly carried out on mobile devices within 1 s.

[Bioavailability of nickel and its complexes during anaerobic digestion].

The biouptake of nickel and its complexes for methanogenic enrichment in the presence of different chelators during batch methane fermentation were investigated in this paper. The results showed that the chelators had obvious effects on anaerobic digestion. At sodium acetate concentration of 85 mmol/L, sulfides concentration of 1 mmol/L, nickel concentration of 200 micromol/L and temperature was 35 degrees C, methane production in the NTA added system were 15% and 9% which was higher than that in CA and EDTA amended ones. While nickel concentration was 100 micromol/L, methane production in NTA added system were 43% and 57% which was higher than that in CA and EDTA amended ones. The biouptake of nickel for methanogenic enrichment related to the species of nickel complexes. NTA was the best chelator for stimulating nickel biouptake in the anaerobic reactors, and EDTA was the better one. The biouptake of Ni-CA complexes was the minimum for the methanogenic enrichment.

[Effect of metal chelating agent on anaerobic digestion under different substrates].

Aided by GC analysis, the effect of metal chelating agent on anaerobic digestion under different substrates was investigated. The results showed that the addition of chelating agent A could greatly promote the methane production under different substrates and speed the conversion of organic acid to methane, especially for propionic acid. When the dosage of chelating agent A was 10 micromol/L, the methane production was, respectively, 19.8%, 133.0% and 45.2% for butyric acid, propionic acid and acetic acid. Also, the arrearage time of methane production was shortened and the production rate increased by 4-fold. During 4-8 d operation, the degradation rate of butyric acid reached 56%. The accumulation of acetic acid was not detected by GC analysis, which demonstrated acetic acid converted from butyric acid was quickly utilized by methane producing bacteria. The increase in methane production due to the addition of chelating agent A could be supported by enzymology. Taking acetic acid as an example, the content of coenzyme F420 in sludge increased to 1.52 micromol/g from 1.20 micromol/g due to the addition of 10 micromol/L chelating agent A.