Phase identification of magnetic islands based on runaway electron losses on EAST

This dataset originates from experimental discharge data of the EAST (Experimental Advanced Superconducting Tokamak) device, primarily including raw and processed signal data from two typical discharges: discharge #63862 (tearing mode experiment) and discharge #113599 (RMP experiment). The experimental data were acquired through the EAST device's MDSplus data system, involving various diagnostic data such as fast electron loss signals, magnetic perturbation signals, runaway electron signals, and auxiliary heating power.During the data acquisition process, the FILD (Fast-Ion Loss Detector) probe was used to measure local fast electron loss signals, with a maximum sampling rate of 2 MHz for the PMT channels and a spatial resolution of 1280 × 800 pixels for the CCD imaging system. Mirnov magnetic probes, with a sampling frequency of 1 MHz, were used to record magnetic perturbation signals to identify magnetic island phases. The RA (Runaway Electron Avalanche) and NF (Neutron Fluctuation) diagnostic systems were employed to monitor signals related to runaway electrons and neutrons, respectively. The ECRH power signal unit is MW, and the RMP current unit is kA. The time variable unit is seconds (s) or milliseconds (ms).The data content covers the complete discharge time series and the specific time windows highlighted in the paper. For discharge #63862, the key analysis interval is 0.84–0.88 s; for discharge #113599, the key analysis interval is 3.5–6.5 s. The time resolution depends on the sampling frequency of the specific diagnostic system (1 MHz or 2 MHz). Spatial information includes the position parameters of the diagnostics in the cylindrical coordinate system (R, Φ, Z), with units of meters (m), radians (rad), or degrees (°). The magnetic topology reconstruction data are based on Poincaré section data generated from MARS-F plasma response calculations and the TOP2D field line tracing code, with spatial resolution determined by the field line tracing step size.The raw signals underwent the following processing steps: First, baseline correction and high-pass filtering were applied to the magnetic probe signals to remove low-frequency drift and electronic noise. Baseline subtraction was performed on FILD and RA signals before analysis. Absolute flux calibration was not applied to any signals during phase analysis; they are presented in arbitrary units (a.u.) and are only used to reflect relative trends. The time axes were uniformly aligned to ensure temporal correspondence between multiple diagnostics.Data files are provided in MATLAB (.fig) format. Each file contains a time column and corresponding signal columns, with variable names consistent with the physical quantities used in the paper, such as t (time), FILD, KHP6T, LHP6T, RA, NF3, Ip, PECRH, IRMP, etc. The table data explicitly provide row labels (time series) and column labels (physical quantity names and units), including units such as kA, MW, MeV, m, rad, a.u., etc.In some individual discharge phases, due to diagnostic saturation or signals falling below the noise threshold, data may be missing or have a low signal-to-noise ratio for certain time periods. Sources of error primarily include electronic noise from the diagnostic systems, calibration errors in the effective area of the magnetic probe coils, time synchronization errors, and grid discretization errors in the numerical simulation of magnetic topology. For time measurements, errors are mainly limited by the sampling frequency (on the order of ~1 μs). For phase mapping, error sources include toroidal geometry positioning errors and approximation errors in the magnetic field model.All data files can be read directly using MATLAB software.This dataset supports the reproduction and validation of the main conclusions of this paper and does not contain restricted or classified information. For further clarification, the corresponding author can be contacted to obtain supplementary information.