Physics informed neural networks speed up solar spectral analysis by up to 60 times
Physics-Informed Neural Networks for Fast Multilayer Spectral Inversion of Hα 6562.8 A and Ca II 8542.1 A Spectra
Artificial IntelligenceMachine Learning
Summary
Analyzing detailed solar light spectra helps scientists understand the Sun’s atmosphere, but the usual methods are slow and compute-heavy. The authors developed a special kind of neural network, guided by physics rules, to quickly predict key solar atmospheric properties from observed light data. Their method is much faster than traditional techniques and still produces results that closely match the slower, standard approaches. This makes it easier to analyze large amounts of solar data in less time.
What this means in practice
- •For solar data analysts: Process large solar chromospheric spectral imaging datasets faster while keeping physical interpretability intact.
- •For astronomical observatories: Enable real-time or near-real-time analysis of solar active regions by accelerating spectral inversion routines with PINNs.
Authors
Ziyang Zhang, Qin Li, Vasyl B. Yurchyshyn, Kangwoo Yi, Haimin Wang, Wenda Cao, Bo Shen
Abstract
Strong chromospheric absorption lines such as H$α$ 6562.8 A and Ca II 8542.1 A provide vital diagnostics of plasma dynamics and thermal structure in the solar chromosphere. Multilayer spectral inversion (MLSI) offers a physically interpretable framework for modeling these lines using a finite number of radiative-transfer layers, but conventional MLSI relies on pixel-by-pixel nonlinear least-squares fitting, making it computationally expensive for large imaging spectroscopic data sets. Here, we introduce a physics-informed neural-network (PINN) framework to accelerate MLSI while preserving its analytic radiative-transfer formulation. The network predicts MLSI parameters directly from observed line profiles and passes them through a differentiable MLSI forward model to synthesize spectra. Training follows a two-stage approach: an initial stage optimized solely via spectral reconstruction loss, followed by fine-tuning that combines spectral consistency with parameter-space supervision from conventional MLSI results on a single reference image. This strategy eliminates the need for large precomputed training sets while maintaining physical interpretability. Applied to Fast Imaging Solar Spectrograph (FISS) observations from the Goode Solar Telescope (GST) targeting both quiet-Sun and active-region regions, MLSI-PINN parameter maps reproduce the primary spatial structures of direct inversions, achieving an arithmetic mean pixel-wise Pearson correlation coefficient of 0.933 across all evaluated parameters. The reconstructed spectra closely match both observed profiles and conventional MLSI fits. Post-training, MLSI-PINN processes a raster in approximately 5-15 seconds compared to 3-5 minutes for conventional MLSI, delivering an inference speedup of about 12-60 times without substantial loss in reconstruction quality, enabling efficient MLSI analysis on large chromospheric data sets.