Blog
Thoughts on plasma physics, accelerators, and machine learning
-
理解 Transformer
介绍 Transformer 的基本计算过程：token embedding、注意力、多头结构、encoder–decoder，以及训练与推理的区别。
-
The Fourth Generation: Where the X-Ray Light Sources Stand in 2026
A survey of the synchrotrons and free-electron lasers that produce the world’s brightest X-rays — what is running, what is being rebuilt, what is only proposed — and an assessment of where plasma acceleration fits.
-
Laser-Driven Proton Therapy: Strong Physics, Hard Road to the Clinic
A follow-up to my post on VHEE electrons: why laser-driven proton acceleration is great physics but, in my view, unlikely to become a competitive medical product — and where its real strengths are.
-
The 10 TeV Horizon: A Perspective on the Future of High Energy Physics
As the community looks toward the 10 TeV energy frontier, a look at the strategic trade-offs between linear and circular approaches, and why the future of plasma acceleration may lie in high-brightness photon science.
-
A Verdict on the Annual Modulation: Settling a 20-Year Dark Matter Debate
A new combined analysis from the COSINE-100 and ANAIS-112 experiments provides the strongest evidence yet against the long-standing DAMA/LIBRA dark matter claim.
-
The Kernel Trick: A Guide to High-Dimensional Feature Spaces
A technical explanation of how the kernel trick allows linear models to solve non-linear problems by implicitly operating in high-dimensional feature spaces.
-
A Scalpel of Electrons: The Promise and Clinical Reality of Laser-Driven Radiotherapy
An analysis of laser-wakefield accelerators for cancer therapy, exploring how this remarkable technology has overcome physics barriers but now faces the challenge of clinical necessity.
-
The Next Light: Can Particle Accelerators Power the Future of Moore’s Law?
An overview of the push for next-generation EUV light sources, comparing the incumbent laser-produced plasma technology against emerging accelerator-based Free-Electron Laser concepts.
-
Reconstructing Electron Bunch Current Profiles with Conditional Diffusion Models
This work proposes using a conditional diffusion model to solve the ill-posed inverse problem of reconstructing electron bunch current profiles from downstream diagnostics in plasma wakefield accelerators.
-
WIMP detection reaches critical threshold
Four decades of WIMP searches reach 2.2×10^-48 cm² sensitivity—just one order of magnitude from the neutrino floor that will either reveal dark matter or redirect physics toward axions.