arXiv subjects

Katherine Su

Publications and source records attributed to Katherine Su.

5 recordsLinked to original sources

arXiv2026

Strain patterning of flexomagnetism

Flexomagnetism, the coupling of magnetic ordering to strain gradients, provides access to novel symmetry-broken magnetic phases that cannot be accessed via uniform strain. However, flexomagnetism is hard to understand because it is extremely difficult to control a spatially varying strain. Here, we develop a top-down strategy to pattern transverse strain gradients using helium ion implantation through a lithographically defined mask. Using epitaxial films of the antiferromagnetic nodal line semimetal GdAuGe, we demonstrate that transverse strain gradients $\partial \varepsilon_{zz}/\partial x$ induce near-room-temperature ferromagnetic response, compared to the retained para or antiferromagnetism for homogeneously strained GdAuGe. We spatially correlate the magnetic response with the regions of largest strain gradient, via magnetic force microscopy and nanobeam x-ray diffraction, respectively, to confirm the flexomagnetic response. Our approach opens new avenues for the precise control of magnetic phases in thin films of quantum materials via a patterned strain gradient.

cond-mat.mtrl-sci↗

arXiv2024

Yi-Lightning Technical Report

This technical report presents Yi-Lightning, our latest flagship large language model (LLM). It achieves exceptional performance, ranking 6th overall on Chatbot Arena, with particularly strong results (2nd to 4th place) in specialized categories including Chinese, Math, Coding, and Hard Prompts. Yi-Lightning leverages an enhanced Mixture-of-Experts (MoE) architecture, featuring advanced expert segmentation and routing mechanisms coupled with optimized KV-caching techniques. Our development process encompasses comprehensive pre-training, supervised fine-tuning (SFT), and reinforcement learning from human feedback (RLHF), where we devise deliberate strategies for multi-stage training, synthetic data construction, and reward modeling. Furthermore, we implement RAISE (Responsible AI Safety Engine), a four-component framework to address safety issues across pre-training, post-training, and serving phases. Empowered by our scalable super-computing infrastructure, all these innovations substantially reduce training, deployment and inference costs while maintaining high-performance standards. With further evaluations on public academic benchmarks, Yi-Lightning demonstrates competitive performance against top-tier LLMs, while we observe a notable disparity between traditional, static benchmark results and real-world, dynamic human preferences. This observation prompts a critical reassessment of conventional benchmarks' utility in guiding the development of more intelligent and powerful AI systems for practical applications. Yi-Lightning is now available through our developer platform at https://platform.lingyiwanwu.com.

cs.CL↗

arXiv2023

Control of ternary alloy composition during remote epitaxy on graphene

Understanding the sticking coefficient $\sigma$, i.e., the probability of an adatom sticking to a surface, is essential for controlling the stoichiometry during epitaxial film growth. However, $\sigma$ on monolayer graphene-covered surfaces and its impact on remote epitaxy are not understood. Here, using molecular-beam epitaxial (MBE) growth of the magnetic shape memory alloy Ni$_2$MnGa, we show that the sticking coefficients for metals on graphene-covered MgO (001) are less than one and are temperature and element dependent, as revealed by ion backscattering spectrometry (IBS) and energy dispersive x-ray spectroscopy (EDS). This lies in stark contrast with most transition metals sticking on semiconductor and oxide substrates, for which $\sigma$ is near unity at typical growth temperatures ($T<800\degree$C). By initiating growth below $400 \degree$ C, where the sticking coefficients are closer to unity and wetting on the graphene surface is improved, we demonstrate epitaxy of Ni$_2$MnGa films with controlled stoichiometry that can be exfoliated to produce freestanding membranes. Straining these membranes tunes the magnetic coercive field. Our results provide a route to synthesize membranes with complex stoichiometries whose properties can be manipulated via strain.

cond-mat.mtrl-sci↗

arXiv2022

Controlling the balance between remote, pinhole, and van der Waals epitaxy of Heusler films on graphene/sapphire

Remote epitaxy on monolayer graphene is promising for synthesis of highly lattice mismatched materials, exfoliation of free-standing membranes, and re-use of expensive substrates. However, clear experimental evidence of a remote mechanism remains elusive. In many cases, due to contaminants at the transferred graphene/substrate interface, alternative mechanisms such as pinhole-seeded lateral epitaxy or van der Waals epitaxy can explain the resulting exfoliatable single-crystalline films. Here, we find that growth of the Heusler compound GdPtSb on clean graphene on sapphire substrates produces a 30 degree rotated epitaxial superstructure that cannot be explained by pinhole or van der Waals epitaxy. With decreasing growth temperature the volume fraction of this 30 degree domain increases compared to the direct epitaxial 0 degree domain, which we attribute to slower surface diffusion at low temperature that favors remote epitaxy, compared to faster surface diffusion at high temperature that favors pinhole epitaxy. We further show that careful graphene/substrate annealing ($T\sim 700 ^\circ C$) and consideration of the film/substrate vs film/graphene lattice mismatch are required to obtain epitaxy to the underlying substrate for a variety of other Heusler films, including LaPtSb and GdAuGe. The 30 degree rotated superstructure provides a possible experimental fingerprint of remote epitaxy since it is inconsistent with the leading alternative mechanisms.

cond-mat.mtrl-sci↗

arXiv2021

Quantifying Mn diffusion through transferred versus directly-grown graphene barriers

We quantify the mechanisms for manganese (Mn) diffusion through graphene in Mn/graphene/Ge (001) and Mn/graphene/GaAs (001) heterostructures for samples prepared by graphene layer transfer versus graphene growth directly on the semiconductor substrate. These heterostructures are important for applications in spintronics; however, challenges in synthesizing graphene directly on technologically important substrates such as GaAs necessitate layer transfer and anneal steps, which introduce defects into the graphene. \textit{In-situ} photoemission spectroscopy measurements reveal that Mn diffusion through graphene grown directly on a Ge (001) substrate is 1000 times lower than Mn diffusion into samples without graphene ($D_{gr,direct} \sim 4\times10^{-18}$cm$^2$/s, $D_{no-gr} \sim 5 \times 10^{-15}$ cm$^2$/s at 500$^\circ$C). Transferred graphene on Ge suppresses the Mn in Ge diffusion by a factor of 10 compared to no graphene ($D_{gr,transfer} \sim 4\times10^{-16}cm^2/s$). For both transferred and directly-grown graphene, the low activation energy ($E_a \sim 0.1-0.5$ eV) suggests that Mn diffusion through graphene occurs primarily at graphene defects. This is further confirmed as the diffusivity prefactor, $D_0$, scales with the defect density of the graphene sheet. Similar diffusion barrier performance is found on GaAs substrates; however, it is not currently possible to grow graphene directly on GaAs. Our results highlight the importance of developing graphene growth directly on functional substrates, to avoid the damage induced by layer transfer and annealing.

cond-mat.mtrl-sci↗