SearcharxivSearch

arXiv subjects

Yi-Cheng Lo

Publications and source records attributed to Yi-Cheng Lo.

2 recordsLinked to original sources

Band Gap Engineering of Nitrogen-Doped Monolayer WSe$_2$ Superlattice and its application to Field Effect Transistor

We systematically investigate the electronic structures of pristine monolayer WSe$_2$ and WSe$_2$ superlattices with periodic nitrogen substitution. Unlike random doping, which often introduces in-gap impurity states, periodic nitrogen doping primarily modulates the band gap, thereby facilitating effective band gap engineering for electronic and optoelectronic applications. The gap narrows monotonically with increasing dopant density (pristine $>$ 8-row $>$ 6-row $>$ 4-row), directly influencing device switching. We also evaluate the FET performance of nanojunctions created by these configurations by examining the contour plot of current density as a function of temperature and gate voltage, which quantifies how bandgap engineering affects switching characteristics. Our calculations clarify the classical-quantum crossover in sub-10 nm 2D FETs: as $T$ rises, $J$ approaches the thermionic current; as $T$ falls, quantum tunneling dominates, and the steep energy dependence of $\tau(E)$ may break the classical limit of subthreshold swing imposed by the Boltzmann tyranny. The optimal gating range ($V_g^\mathrm{ON}$, $V_g^\mathrm{OFF}$) is investigated for each temperature, insensitive to temperature in the high-temperature regime, confirming the good thermal stability of the FET devices. A comparison study demonstrates that the 4-row structure, with large $J_\mathrm{OFF}$ and restricted operation range, is inappropriate for realistic FET applications. The pristine structure has a high $V_g^\mathrm{OFF}$ ($\sim$1.1 V) makes it less practical, since such a large threshold voltage may promote time-dependent dielectric breakdown (TDDB) of the oxide layer, reducing device dependability. The 6-row and 8-row structures exhibit more favorable $V_g^\mathrm{OFF}$ values ($\sim$0.75 V), achieving compromise, making them more promising candidates for future FET integration.

cond-mat.mes-hall

Post-Training Quantization for Vision Mamba with k-Scaled Quantization and Reparameterization

The Mamba model, utilizing a structured state-space model (SSM), offers linear time complexity and demonstrates significant potential. Vision Mamba (ViM) extends this framework to vision tasks by incorporating a bidirectional SSM and patch embedding, surpassing Transformer-based models in performance. While model quantization is essential for efficient computing, existing works have focused solely on the original Mamba model and have not been applied to ViM. Additionally, they neglect quantizing the SSM layer, which is central to Mamba and can lead to substantial error propagation by naive quantization due to its inherent structure. In this paper, we focus on the post-training quantization (PTQ) of ViM. We address the issues with three core techniques: 1) a k-scaled token-wise quantization method for linear and convolutional layers, 2) a reparameterization technique to simplify hidden state quantization, and 3) a factor-determining method that reduces computational overhead by integrating operations. Through these methods, the error caused by PTQ can be mitigated. Experimental results on ImageNet-1k demonstrate only a 0.8-1.2\% accuracy degradation due to PTQ, highlighting the effectiveness of our approach.

eess.IV