Researchers Detect Thermal Dust Echoes Revealing Missed Supernovae

Researchers release an all-sky, untargeted search for thermal dust echoes that can reveal historical, optically missed Galactic supernovae using 12 years of time-resolved NEOWISE co-adds. The team uses difference imaging to produce transient candidate maps, trains a convolutional neural network on echo patterns from Cas A plus archival catalogs, and classifies millions of variable sources. The search recovers Cas A as the only region with echoes at WISE sensitivity and produces a catalog of 20,477 echo positions now used for JWST studies.

The pipeline combines standard infrared imaging techniques with supervised deep learning. Key steps include:

• •difference imaging on decade-scale NEOWISE co-adds to isolate variable mid-infrared emission
• •morphological and temporal feature extraction to separate extended echoes from point-like variables
• •training a convolutional neural network on labeled examples drawn from Cas A echo morphology and archival variable-source catalogs
• •statistical searches for spatial over-densities across the Galactic plane

The classifier separates dust echoes, point sources, imaging artifacts, and high proper motion stars with high accuracy. The final output is a large, time-resolved echo position catalog optimized for targeted spectroscopic and imaging follow-up with JWST.

Thermal dust echoes are a rare but powerful tracer of luminous transients that occurred behind heavy extinction and therefore escaped optical discovery. Demonstrating an untargeted, scalable search across the entire sky establishes a repeatable methodology for cross-disciplinary teams. For ML practitioners, the work is an instructive case of applying CNN-based morphology classification to low signal-to-noise, extended infrared features on decade-long baselines. For observational astrophysicists, the non-detection of additional echoes at current WISE sensitivity quantifies the observational limits and highlights the value of higher-sensitivity, higher-resolution surveys. The catalog of 20,477 echo positions around Cas A is a practical resource for ISM tomography and shock physics with JWST.

Deeper, higher-resolution datasets from the Roman space telescope and continued NEOWISE operations will determine whether more historical Galactic transients can be recovered. Expect refinements in the classifier and feature engineering as teams adapt the pipeline to different wavelength bands and spatial resolutions.