发表机构
Università degli Studi di Brescia（布雷西亚大学）
机构由 AI 辅助整理，请以论文原文为准。AI 中文总结
研究利用H-MOG变分方法，基于喷流宽度和线性极化两个投影可观测量重构相对论性喷流三维磁场，测试三种打破符号简并性途径，虽能恢复无符号场方向，但极向场方向无法恢复，打破简并性的途径因物理原因失败，恢复方向需宇称奇可观测量。
AI 中文摘要
相对论性喷流的每一张解析图像都编码了其磁场，但没有图像记录磁场沿轴的指向方向。同步辐射强度和线性极化对此并不敏感：它们仅通过磁场分量的偶数组合来探测磁场。我们引入了H-MOG，这是一种变分方法，可根据两个投影可观测量，即喷流宽度W(z)和线性极化p(z)，在晶格上重构三维磁场，并通过哈密顿先验进行正则化，利用自动微分进行优化。我们将H-MOG应用于十个跨越MAD和SANE状态以及五个黑洞自旋（a* = -0.94, -0.5, 0, +0.5, +0.94）的GRMHD模拟中，并测试了三种打破重构固有符号简并性的途径：法拉第旋转测量项、全三维采样和布兰福德-兹纳耶克自旋先验。无符号场方向的恢复率在<|cos|> ~ 0.95 - 0.98之间，远高于随机预期的0.5。然而，极向场的方向无法恢复，我们证明了这是由于W和p在B -> -B下是不变的。所有三种打破这种简并性的途径都因不同的物理原因而失败；这些湍流喷流中的自旋-方向关系不是单调的，在MAD和SANE之间有很大差异。恢复方向需要一个宇称奇的可观测量：法拉第层析成像或圆极化。
英文摘要
Every resolved image of a relativistic jet encodes its magnetic field, yet no image records which way the field points along the axis. Synchrotron intensity and linear polarization are blind to this: both probe the field only through even combinations of its components. We introduce H-MOG, a variational method that reconstructs the 3D field on a lattice from two projected observables, the jet width W(z) and the linear polarization p(z), regularized by a Hamiltonian prior and optimized with automatic differentiation. We apply H-MOG to ten GRMHD simulations spanning MAD and SANE states and five black-hole spins (a* = -0.94, -0.5, 0, +0.5, +0.94), and test three routes to break the intrinsic sign degeneracy of the reconstruction: a Faraday rotation-measure term, full 3D sampling, and a Blandford-Znajek spin prior. The unsigned field orientation is recovered at <|cos|> ~ 0.95-0.98, far above the random expectation of 0.5. The sense of the poloidal field, however, is not recovered, and we prove it cannot be: W and p are invariant under B -> -B. All three routes to break this degeneracy fail for distinct physical reasons; the spin-sense relation in these turbulent jets is not monotonic, differing sharply between MAD and SANE. Recovering the sense requires a parity-odd observable: Faraday tomography or circular polarization.
Comments14 pages, 13 figures, 4 tables
