arXiv subjects

Jiyong Zhou

Publications and source records attributed to Jiyong Zhou.

3 recordsLinked to original sources

arXiv2023

Melting of electronic and excitonic crystals in 2D semiconductor moiré patterns: a perspective from the Lindemann criterion

Using the Lindemann criterion, we analyzed the quantum and thermal melting of electronic/excitonic crystals recently discovered in two-dimensional (2D) semiconductor moiré patterns. We show that the finite 2D screening of the atomically thin material can suppress (enhance) the inter-site Coulomb (dipolar) interaction strength, thus inhibits (facilitates) the formation of the electronic (excitonic) crystal. Meanwhile, a strong enough moiré confinement is found to be essential for realizing the crystal phase with a wavelength near 10 nm or shorter. From the calculated Lindemann ratio which quantifies the fluctuation of the site displacement, we estimate that the crystal will melt into a liquid above a critical temperature ranging from several tens Kelvin to above 100 K (depending on the system parameters).

cond-mat.mes-hall↗

arXiv2023

Making Small Language Models Better Multi-task Learners with Mixture-of-Task-Adapters

Recently, Large Language Models (LLMs) have achieved amazing zero-shot learning performance over a variety of Natural Language Processing (NLP) tasks, especially for text generative tasks. Yet, the large size of LLMs often leads to the high computational cost of model training and online deployment. In our work, we present ALTER, a system that effectively builds the multi-tAsk Learners with mixTure-of-task-adaptERs upon small language models (with <1B parameters) to address multiple NLP tasks simultaneously, capturing the commonalities and differences between tasks, in order to support domain-specific applications. Specifically, in ALTER, we propose the Mixture-of-Task-Adapters (MTA) module as an extension to the transformer architecture for the underlying model to capture the intra-task and inter-task knowledge. A two-stage training method is further proposed to optimize the collaboration between adapters at a small computational cost. Experimental results over a mixture of NLP tasks show that our proposed MTA architecture and the two-stage training method achieve good performance. Based on ALTER, we have also produced MTA-equipped language models for various domains.

cs.CL↗

arXiv2023

Phonons of electronic crystals in two-dimensional semiconductor moiré patterns

We theoretically studied the phonon properties of the triangular-, stripe- and honeycomb-type electronic crystals recently found in two-dimensional semiconductor moiré patterns. By analyzing the phonon dispersions, we found the interaction induced lattice deformation in the zigzag-stripe crystal results in a much higher dynamical stability than in the linear-stripe crystal. Moreover, chiral phonons with finite magnetizations and large Berry curvatures can emerge in triangular and honeycomb crystals under time-reversal or inversion symmetry breaking. The small effective mass of the electrons allows the selective and efficient generation of chiral phonons from the optical activity of zone-center phonons combined with the anharmonicity, facilitating the realization of the phonon Hall effect. These findings point to an exciting new platform for exploring chiral phonons and the related topological phononic devices.

cond-mat.mes-hall↗