arXiv subjects

Qile Zhang

Publications and source records attributed to Qile Zhang.

At least 19 recordsLinked to original sources

arXiv2026

Resonant Pitch-Angle Scattering Of Runaway-Electrons by Externally-launched Helicon Waves in the DIII-D Tokamak

Resonant wave-particle interactions between externally launched helicon waves (also known as whistler waves) and runaway electrons (REs) have been demonstrated on the DIII-D tokamak. In this work we extend the initial results reported in Choudhury, H. et al. Phys. Rev. Lett. 136, 025101 (2026) by exploring the effects of antenna alignment with the edge magnetic field, toroidal wave propagation direction, and coupled power on RE scattering in the quiescent RE experimental scenario. Two distinct experimental configurations have been investigated: one in which the antenna aligns well with the edge background magnetic field, known as the ideal antenna configuration, and one with misalignment, known as the non-ideal case. Previously, it had been found that helicon power in the ideal antenna configuration prevented RE growth despite the normalized toroidal electric field remaining high enough to drive exponential RE growth in the absence of helicon power. In this paper, we show that scattering via the normal Doppler resonance (n=1) effectively limits the growth of the RE population in both the ideal and non-ideal antenna configurations, with evidence of a power threshold in the latter case. In contrast, launching waves that favour the anomalous Doppler resonance (n=-1) is observed to enhance rather than reduce the RE population. In addition, fast magnetic measurements reveal rising-tones in the 30-60 MHz range during helicon-off periods, which are not observed prior to helicon power. Finally, the challenges of using launched helicon waves to scatter post-disruption RE beams are discussed. Collisional damping and a large vacuum gap between the plasma and antenna on the outboard side present significant obstacles to helicon waves propagating into the plasma core.

physics.plasm-ph↗

arXiv2026

Farewell to Item IDs: Unlocking the Scaling Potential of Large Ranking Models via Semantic Tokens

Recent studies on scaling up ranking models have achieved substantial improvement for recommendation systems and search engines. However, most large-scale ranking systems rely on item IDs, where each item is treated as an independent categorical symbol and mapped to a learned embedding. As items rapidly appear and disappear, these embeddings become difficult to train and maintain. This instability impedes effective learning of neural network parameters and limits the scalability of ranking models. In this paper, we show that semantic tokens possess greater scaling potential compared to item IDs. Our proposed framework TRM improves the token generation and application pipeline, leading to 33% reduction in sparse storage while achieving 0.85% AUC increase. Extensive experiments further show that TRM could consistently outperform state-of-the-art models when model capacity scales. Finally, TRM has been successfully deployed on large-scale personalized search engines, yielding 0.26% and 0.75% improvement on user active days and change query ratio respectively through A/B test.

cs.IR↗

arXiv2026

TAP-ViTs: Task-Adaptive Pruning for On-Device Deployment of Vision Transformers

Vision Transformers (ViTs) have demonstrated strong performance across a wide range of vision tasks, yet their substantial computational and memory demands hinder efficient deployment on resource-constrained mobile and edge devices. Pruning has emerged as a promising direction for reducing ViT complexity. However, existing approaches either (i) produce a single pruned model shared across all devices, ignoring device heterogeneity, or (ii) rely on fine-tuning with device-local data, which is often infeasible due to limited on-device resources and strict privacy constraints. As a result, current methods fall short of enabling task-customized ViT pruning in privacy-preserving mobile computing settings. This paper introduces TAP-ViTs, a novel task-adaptive pruning framework that generates device-specific pruned ViT models without requiring access to any raw local data. Specifically, to infer device-level task characteristics under privacy constraints, we propose a Gaussian Mixture Model (GMM)-based metric dataset construction mechanism. Each device fits a lightweight GMM to approximate its private data distribution and uploads only the GMM parameters. Using these parameters, the cloud selects distribution-consistent samples from public data to construct a task-representative metric dataset for each device. Based on this proxy dataset, we further develop a dual-granularity importance evaluation-based pruning strategy that jointly measures composite neuron importance and adaptive layer importance, enabling fine-grained, task-aware pruning tailored to each device's computational budget. Extensive experiments across multiple ViT backbones and datasets demonstrate that TAP-ViTs consistently outperforms state-of-the-art pruning methods under comparable compression ratios.

cs.CV↗

arXiv2025

Toward a unified framework for data-efficient evaluation of large language models

Evaluating large language models (LLMs) on comprehensive benchmarks is a cornerstone of their development, yet it's often computationally and financially prohibitive. While Item Response Theory (IRT) offers a promising path toward data-efficient evaluation by disentangling model capability from item difficulty, existing IRT-based methods are hampered by significant limitations. They are typically restricted to binary correctness metrics, failing to natively handle the continuous scores used in generative tasks, and they operate on single benchmarks, ignoring valuable structural knowledge like correlations across different metrics or benchmarks. To overcome these challenges, we introduce LEGO-IRT, a unified and flexible framework for data-efficient LLM evaluation. LEGO-IRT's novel design natively supports both binary and continuous evaluation metrics. Moreover, it introduces a factorized architecture to explicitly model and leverage structural knowledge, decomposing model ability estimates into a general component and structure-specific (e.g., per-metric or per-benchmark) components. Through extensive experiments involving $70$ LLMs across $5$ benchmarks, we show that LEGO-IRT achieves stable capability estimates using just $3\%$ of the total evaluation items. We demonstrate that incorporating structural knowledge reduces estimation error by up to $10\%$ and reveal that the latent abilities estimated by our framework may align more closely with human preferences.

cs.AI↗

arXiv2025

Particle Injection Problem in Magnetic Reconnection and Turbulence

Magnetic reconnection and turbulence in magnetically-dominated environments have been proposed as important nonthermal particle acceleration mechanisms that generate high energy particles and associated emissions. While the acceleration to high energy that produces the power-law energy distribution has drawn strong interest, recent studies actively discuss pre-acceleration, or injection, to a sufficient energy for a sustained and prolonged Fermi-like acceleration. The injection process is important for determining the fraction of nonthermal particles and energy partition between thermal and nonthermal particles. We review recent advances in understanding the injection mechanisms responsible for populating these nonthermal power-law spectra, and conclude with an outlook for studies and applications of injection models.

physics.plasm-ph↗

arXiv2025

Excitation of whistler and slow-X waves by runaway electrons in a collisional plasma

Runaway electrons are known to provide robust ideal or collisionless kinetic drive for plasma wave instabilities in both the whistler and slow-X branches, via the anomalous Doppler-shifted cyclotron resonances. In a cold and dense post-thermal-quench plasma, collisional damping of the plasma waves can be competitive with the collisionless drive. Previous studies have found that for its higher wavelength and frequency, slow-X waves suffer stronger collisional damping than the whistlers, while the ideal growth rate of slow-X modes is higher. Here we study runaway avalanche distributions that maintain the same eigen distribution and increase only in magnitude over time. The distributions are computed from the relativistic Fokker-Planck-Boltzmann solver, upon which a linear dispersion analysis is performed to search for the most unstable or least damped slow-X and whistler modes. Taking into account the effect of plasma density, plasma temperature, and effective charge number, we find that the slow-X modes tend to be excited before the whistlers in a runaway current ramp-up. Furthermore, even when the runaway current density is sufficiently high that both branches are excited, the most unstable slow-X mode has much higher growth rate than the most unstable whistler mode. The qualitative and quantitative trends uncovered in current study indicate that even though past experiments and modeling efforts have concentrated on whistler modes, there's a compelling case that slow-X modes should also be a key area of focus.

physics.plasm-ph↗

arXiv2024

Self-mediation of runaway electrons via self-excited wave-wave and wave-particle interactions

Nonlinear dynamics of runaway electron induced wave instabilities can significantly modify the runaway distribution critical to tokamak operations. Here we present the first-ever fully kinetic simulations of runaway-driven instabilities towards nonlinear saturation in a warm plasma where collisional damping is subdominant. It is found that the slow-X modes grow an order of magnitude faster than the whistler modes, and they parametrically decay to produce whistlers much faster than those directly driven by runaways. These parent-daughter waves, as well as secondary and tertiary wave instabilities, initiate a chain of wave-particle resonances that strongly diffuse runaways to the backward direction. This reduces almost half of the current carried by high-energy runaways, over a time scale orders of magnitude faster than experimental shot duration. These results beyond quasilinear analysis may impact anisotropic energetic electrons broadly in laboratory, space and astrophysics.

physics.plasm-ph↗

arXiv2024

Constraints on Primordial Magnetic Fields from High Redshift Stellar Mass Density

Primordial magnetic fields (PMFs) play a pivotal role in influencing small-scale fluctuations within the primordial density field, thereby enhancing the matter power spectrum within the context of the $\Lambda$CDM model at small scales. These amplified fluctuations accelerate the early formation of galactic halos and stars, which can be observed through advanced high-redshift observational techniques. Therefore, Stellar Mass Density (SMD) observations, which provide significant opportunities for detailed studies of galaxies at small scales and high redshifts, offer a novel perspective on small-scale cosmic phenomena and constrain the characteristics of PMFs. In this study, we compile 14 SMD data points at redshifts $z > 6$ and derive stringent constraints on the parameters of PMFs, which include the amplitude of the magnetic field at a characteristic scale of $\lambda=1\,{\rm Mpc}$, denoted as $B_0$, and the spectral index of the magnetic field power spectrum, $n_{\rm B}$. At 95\% confidence level, we establish upper limits of $B_0 < 4.44$ nG and $n_{\rm B} < -2.24$, along with a star formation efficiency of approximately $f_*^0 \sim 0.1$. If we fix $n_{\rm B}$ at specific values, such as $-2.85$, $-2.9$, and $-2.95$, the 95\% upper limits for the amplitude of the magnetic field can be constrained to 1.33 nG, 2.21 nG, and 3.90 nG, respectively. Finally, we attempt to interpret recent early observations provided by James Webb Space Telescope (JWST) using the theory of PMFs, and find that by selecting appropriate PMF parameters, it is possible to explain these results without significantly increasing the star formation efficiency.

astro-ph.CO↗

arXiv2024

Plasma Dynamics and Nonthermal Particle Acceleration in 3D Nonrelativistic Magnetic Reconnection

Understanding plasma dynamics and nonthermal particle acceleration in 3D magnetic reconnection has been a long-standing challenge. In this paper, we explore these problems by performing large-scale fully kinetic simulations of multi-xline plasmoid reconnection with various parameters in both the weak and strong guide field regimes. In each regime, we have identified its unique 3D dynamics that leads to field-line chaos and efficient acceleration, and we have achieved nonthermal acceleration of both electrons and protons into power-law spectra. The spectral indices agree well with a simple Fermi acceleration theory that includes guide field dependence. In the low-guide-field regime, the flux-rope kink instability governs the 3D dynamics for efficient acceleration. The weak dependence of the spectra on the ion-to-electron mass ratio and $β$ ($\ll1$) implies that the particles are sufficiently magnetized for Fermi acceleration in our simulations. While both electrons and protons are injected at reconnection exhausts, protons are primarily injected by perpendicular electric fields through Fermi reflections and electrons are injected by a combination of perpendicular and parallel electric fields. The magnetic power spectra agree with in-situ magnetotail observations, and the spectral index may reflect a reconnection-driven size distribution of plasmoids instead of Goldreich-Sridhar vortex cascade. As the guide field becomes stronger, the oblique flux ropes of large sizes capture the main 3D dynamics for efficient acceleration. Intriguingly, the oblique flux ropes can also run into flux-rope kink instability to drive extra 3D dynamics. This work has broad implications for 3D reconnection dynamics and particle acceleration in heliophysics and astrophysics.

astro-ph.SR↗

arXiv2024

Multi-species Ion Acceleration in 3D Magnetic Reconnection with Hybrid-kinetic Simulations

Magnetic reconnection drives multi-species particle acceleration broadly in space and astrophysics. We perform the first 3D hybrid simulations (fluid electrons, kinetic ions) that contain sufficient scale separation to produce nonthermal heavy-ion acceleration, with fragmented flux ropes critical for accelerating all species. We demonstrate the acceleration of all ion species (up to Fe) into power-law spectra with similar indices, by a common Fermi acceleration mechanism. The upstream ion velocities influence the first Fermi reflection for injection. The subsequent onsets of Fermi acceleration are delayed for ions with lower charge-mass ratios (Q/M), until growing flux ropes magnetize them. This leads to a species-dependent maximum energy/nucleon $\propto(Q/M)^α$. These findings are consistent with in-situ observations in reconnection regions, suggesting Fermi acceleration as the dominant multi-species ion acceleration mechanism.

astro-ph.SR↗

arXiv2023

Particle Injection and Nonthermal Particle Acceleration in Relativistic Magnetic Reconnection

Magnetic reconnection in the relativistic regime has been proposed as an important process for the efficient production of nonthermal particles and high-energy emissions. Using fully kinetic particle-in-cell simulations, we investigate how guide-field strength and domain size affect characteristic spectral features and acceleration processes. We study two stages of acceleration: energization up until the injection energy $γ_{\rm inj}$ and further acceleration that generates a power-law spectrum. Stronger guide fields increase the power-law index and $γ_{\rm inj}$, which suppresses acceleration efficiency. These quantities seemingly converge with increasing domain size, suggesting that our findings can be extended to large-scale systems. We find that three distinct mechanisms contribute to acceleration during injection: particle streaming along the parallel electric field, Fermi reflection, and the pickup process. Fermi and pickup processes, related to the electric field perpendicular to the magnetic field, govern the injection for weak guide fields and larger domains. Meanwhile, parallel electric fields are important for injection in the strong guide field regime. In the post-injection stage, we find that perpendicular electric fields dominate particle acceleration in the weak guide field regime, whereas parallel electric fields control acceleration for strong guide fields. These findings will help explain the nonthermal acceleration and emissions in high-energy astrophysics, including black hole jets and pulsar wind nebulae.

astro-ph.HE↗

arXiv2022

Comment on "Nonideal Fields Solve the Injection Problem in Relativistic Reconnection"

Recently, Sironi (PRL, 128, 145102; S22) reported the correlation between particles accelerated into high energy and their crossings of regions with electric field larger than magnetic field (E>B regions) in kinetic simulations of relativistic magnetic reconnection. They claim that electric fields in E>B regions (for a vanishing guide field) dominate in accelerating particles to the injection energy. S22 presented test-particle simulations showing that if particle energies are reset to low energies in E>B regions, efficient injection is suppressed. This Comment re-examines these claims by analyzing a simulation resembling the reference case in S22. We show that during crossings E>B acceleration only contributes a small fraction to the injection energy as E>B regions only host particles for a short duration. The energization before any E>B crossings has a comparable contribution, indicating E>B regions are not unique in pre-accelerating particles. A new test-particle simulation shows that zero-outing electric fields in E>B regions does not strongly influence the injection. We suggest that the procedure used in S22 to exclude E>B acceleration partly removes acceleration outside E>B regions, leading to a false conclusion.

astro-ph.HE↗

arXiv2022

Dynamical Stability of the Power Law K-essence Dark Energy Model with a New Interaction

We investigate the cosmological evolution of the power law K-essence dark energy (DE) model $F(X)= -\sqrt{X} + X$ with a new interaction $Q = αρ_mρ_{ϕ}H^{-1}$ in FRWL spacetime. The evolution behavior of dark energy under this interaction is analyzed by using dynamical systems method, and ten critical points are obtained. Among those critical points, a new stable point, which we called Scaling-like dark energy(DE) solution, is very important and interesting. The cosmological meaning of this attractor is different from the Scaling solution and dark energy dominated solution. For some value of model parameters, the universe will evolve to the attractor solution with the dark energy density parameter $Ω_ϕ=0.682946$ and the the equation of state $w_ϕ=-0.99$, which can be in good agreement with the observed data, and alleviate the Coincidence Problem.

gr-qc↗

arXiv2021

Slow Shock Formation Upstream of Reconnecting Current Sheets

The formation, development and impact of slow shocks in the upstream region of reconnecting current layers are explored. Slow shocks have been documented in the upstream region of magnetohydrodynamic (MHD) simulations of magnetic reconnection as well as in similar simulations with the {\it kglobal} kinetic macroscale simulation model. They are therefore a candidate mechanism for preheating the plasma that is injected into the current layers that facilitate magnetic energy release in solar flares. Of particular interest is their potential role in producing the hot thermal component of electrons in flares. During multi-island reconnection, the formation and merging of flux ropes in the reconnecting current layer drives plasma flows and pressure disturbances in the upstream region. These pressure disturbances steepen into slow shocks that propagate along the reconnecting component of the magnetic field and satisfy the expected Rankine-Hugoniot jump conditions. Plasma heating arises from both compression across the shock and the parallel electric field that develops to maintain charge neutrality in a kinetic system. Shocks are weaker at lower plasma $β$, where shock steepening is slow. While these upstream slow shocks are intrinsic to the dynamics of multi-island reconnection, their contribution to electron heating remains relatively minor compared with that from Fermi reflection and the parallel electric fields that bound the reconnection outflow.

physics.plasm-ph↗

arXiv2021

Variable Ion Compositions of Solar Energetic Particle Events in the Inner Heliosphere: A Field-line Braiding Model with Compound Injections

We propose a model for interpreting highly variable ion composition ratios in solar energetic particles (SEP) events recently observed by Parker Solar Probe (PSP) at $0.3 - 0.45$ astronomical unit. We use numerical simulations to calculate SEP propagation in a turbulent interplanetary magnetic field with a Kolmogorov power spectrum from large scale down to the gyration scale of energetic particles. We show that when the source regions of different species are offset by a distance comparable to the size of the source regions, the observed energetic particle composition He/H can be strongly variable over more than two orders of magnitude, even if the source ratio is at the nominal value. Assuming a $^3$He/$^4$He source ratio of $10 \%$ in impulsive $^3$He-rich events and the same spatial offset of the source regions, the $^3$He/$^4$He ratio at observation sites also vary considerably. The variability of the ion composition ratios depends on the radial distance, which can be tested by observations made at different radial locations. We discuss the implication of these results on the variability of ion composition of impulsive events and on further PSP and Solar Orbiter observations close to the Sun.

astro-ph.SR↗

arXiv2021

Efficient Nonthermal Ion and Electron Acceleration Enabled by the Flux-Rope Kink Instability in 3D Nonrelativistic Magnetic Reconnection

The relaxation of field-line tension during magnetic reconnection gives rise to a universal Fermi acceleration process involving the curvature drift of particles. However, the efficiency of this mechanism is limited by the trapping of energetic particles within flux-ropes. Using 3D fully kinetic simulations, we demonstrate that the flux-rope kink instability leads to strong field-line chaos in weak-guide-field regimes where the Fermi mechanism is most efficient, thus allowing particles to transport out of flux-ropes and undergo further acceleration. As a consequence, both ions and electrons develop clear power-law energy spectra which contain a significant fraction of the released energy. The low-energy bounds are determined by the injection physics, while the high-energy cutoffs are limited only by the system size. These results have strong relevance to observations of nonthermal particle acceleration in space and astrophysics.

astro-ph.SR↗

arXiv2021

Magnetic Energy Release, Plasma Dynamics, and Particle Acceleration during Relativistic Turbulent Magnetic Reconnection

In strongly magnetized astrophysical plasma systems, magnetic reconnection is believed to be a primary process during which explosive energy release and particle acceleration occur, leading to significant high-energy emission. Past years have witnessed active development of kinetic modeling of relativistic magnetic reconnection, supporting this magnetically dominated scenario. A much less explored issue is the consequence of 3D dynamics, where turbulent structures are naturally generated as various types of instabilities develop. This paper presents a series of 3D, fully-kinetic simulations of relativistic turbulent magnetic reconnection (RTMR) in positron-electron plasmas with system domains much larger than kinetic scales. Our simulations start from a force-free current sheet with several different modes of long wavelength magnetic field perturbations, which drive additional turbulence in the reconnection region. Because of this, the current layer breaks up and the reconnection region quickly evolves into a turbulent layer filled with coherent structures such as flux ropes and current sheets. We find that plasma dynamics in RTMR is vastly different from their 2D counterparts in many aspects. The flux ropes evolve rapidly after their generation, and can be completely disrupted due to the secondary kink instability. This turbulent evolution leads to superdiffusion behavior of magnetic field lines as seen in MHD studies of turbulent reconnection. Meanwhile, nonthermal particle acceleration and energy-release time scale can be very fast and do not strongly depend on the turbulence amplitude. The main acceleration mechanism is a Fermi-like acceleration process supported by the motional electric field, whereas the non-ideal electric field acceleration plays a subdominant role. We discuss possible observational implications of 3D RTMR in high-energy astrophysics.

astro-ph.HE↗

arXiv2019

Instabilities and Turbulence in Low-$β$ Guide Field Reconnection Exhausts with Kinetic Riemann Simulations

The role of turbulence in low-$β$, guide-field reconnection exhausts is explored in 2D reconnection and 2D and 3D Riemann simulations. The structure of the exhaust and associated turbulence is controlled by a pair of rotational discontinuities (RDs) at the exhaust boundary and a pair of slow shocks (SSs) that are generated by counterstreaming ions beams. In 2D the exhaust develops large-amplitude striations at the ion Larmor radius scale that are produced by electron-beam-driven ion cyclotron waves. The electron beams driving the instability are injected into the exhaust from one of the RDs. However, in 3D Riemann simulations, the additional dimension (in the out-of-plane direction) results in strong Buneman and electron-electron streaming instabilities at the RD which suppress electron beam formation and therefore the striations in the exhaust. The strength of the streaming instabilities at the RD are controlled by the ratio of the electron thermal speed to Alfvén speed, lower thermal speed being more unstable. In the 3D simulations an ion-ion streaming instability acts to partially thermalize the counterstreaming ion beams at the SSs. This instability is controlled by the ratio of the sound speed to Alfvén speed and is expected to be stable in the low $β$ solar corona. The results suggest that in a guide field reconnection exhaust with $1 \gg β> m_e/m_i$, the kinetic-scale turbulence that develops will be too weak to play a significant role in energy conversion and particle acceleration. Therefore, the energy conversion will be mostly controlled by laminar physics or multi-x-line reconnection.

physics.plasm-ph↗