The paper reports a new method for three-dimensional observation of the location of focused particle streams along both the depth and width of the channel cross-section in spiral inertial microfluidic systems. The results confirm that particles are focused near the top and bottom walls of the microchannel cross-section, revealing clear insights on the focusing and separation mechanism. Based on this detailed understanding of the force balance, we introduce a novel spiral microchannel with a trapezoidal cross-section that generates stronger Dean vortices at the outer half of the channel. Experiments show that particles focusing in such device are sensitive to particle size and flow rate, and exhibits a sharp transition from the inner half to the outer half equilibrium positions at a size-dependent critical flow rate. As particle equilibration positions are well segregated based on different focusing mechanisms, a higher separation resolution is achieved over conventional spiral microchannels with rectangular cross-section.

We present a systematic, multiscale, fully detailed numerical modeling for dynamics of fluid flow and ion transport covering Ohmic, limiting, and overlimiting current regimes in conductance of ion-selective membrane. By numerically solving the Poisson-Nernst-Planck-Navier-Stokes equations, it is demonstrated that the electroconvective instability, arising from the electric field acting upon the extended space charge layer, and the induced strong vortical fluid flow are the dominant factors of the overlimiting current in the planar membrane system. More importantly, at the transition between the limiting and the overlimiting current regimes, hysteresis of electric current is identified. The hysteresis demonstrates the important role of the electroconvective flow in enhancing of current in electrolyte systems with ion-selective membrane.

Cross-lingual transfer, where a high-resource transfer language is used to improve the accuracy of a low-resource task language, is now an invaluable tool for improving performance of natural language processing (NLP) on lowresource languages. However, given a particular task language, it is not clear which language to transfer from, and the standard strategy is to select languages based on ad hoc criteria, usually the intuition of the experimenter. Since a large number of features contribute to the success of cross-lingual transfer (including phylogenetic similarity, typological properties, lexical overlap, or size of available data), even the most enlightened experimenter rarely considers all these factors for the particular task at hand. In this paper, we consider this task of automatically selecting optimal transfer languages as a ranking problem, and build models that consider the aforementioned features to perform this prediction. In experiments on representative NLP tasks, we demonstrate that our model predicts good transfer languages much better than ad hoc baselines considering single features in isolation, and glean insights on what features are most informative for each different NLP tasks, which may inform future ad hoc selection even without use of our method. 1 * Equal contribution 1 Code, data, and pre-trained models are available at