λPIC：一个以回调为中心的粒子模拟框架
$λ$PIC: A callback-centric particle-in-cell framework
AI总结：
介绍基于Python的电磁粒子模拟框架λPIC，以回调为中心架构打破现有代码局限，能实现自定义算法与原位分析，关键内核用C扩展等编写，支持动态负载平衡，特别关注强激光与等离子体相互作用，未来将扩展到GPU加速等。
AI中文摘要：
我们展示了λPIC，这是一个基于Python的电磁粒子模拟框架，围绕以回调为中心的架构构建。现有的粒子模拟代码通常将高性能与静态、预编译的时间步长循环绑定，阻碍了自定义物理、诊断或输出逻辑的实现。λPIC通过将循环的每个阶段作为命名阶段（钩子）公开来打破这种耦合，允许附加对完整模拟状态进行操作的任意Python函数，从而在不修改核心算法的情况下实现自定义算法和原位分析。在这个灵活的框架下，性能关键内核用C扩展和Numba编写，场和粒子存储在NumPy数组中，MPI并行与图分区相结合以支持动态负载平衡。虽然λPIC设计为通用框架，但特别关注强激光与等离子体相互作用。未来工作将把框架扩展到GPU加速以及包括隐式求解器和核物理在内的其他物理模块。
英文摘要：
We present $λ$PIC, a Python-based electromagnetic particle-in-cell framework built around a callback-centric architecture. Existing PIC codes typically tie high performance to static, pre-compiled timestep loops, hindering implementation of custom physics, diagnostics, or output logic. $λ$PIC breaks this coupling by exposing every stage of the loop as a named stage (hook), permitting attaching arbitrary Python functions that operate on the full simulation state, enabling custom algorithms and in-situ analysis without modifying the core algorithms. Under this flexible framework, performance-critical kernels are written in C extensions and Numba, fields and particles are stored in NumPy arrays, and MPI parallelism is paired with graph partitioning to support dynamic load balancing and non-rectangular domains. Although $λ$PIC is designed as general-purpose, it has special focus on intense laser-plasma interactions. Future work will extend the framework to GPU acceleration and additional physics modules including implicit solvers and nuclear physics.
