Identifying lopsidedness in spiral galaxies using a deep convolutional neural network

Context. About 30% of disk galaxies exhibit lopsidedness in their stellar disk. Although such a large-scale asymmetry in the disk can be primarily considered as a long-lived mode (m = 1), the physical origin of the lopsidedness in the disk continues to present a puzzling issue. Aims. In this work, we employ a transfer-learning approach for the automated identification of lopsided galaxies, using SDSS DR18 imaging by fine-tuning a Zoobot model, a deep convolutional neural network (DCNN) package pre-trained on the Galaxy Zoo dataset. Methods. We obtained 7042 well-resolved, nearly face-on spiral galaxies from SDSS DR18 over the redshift range 0.01 ≤z ≤ 0.1, with an extinction-corrected g-band model magnitude < 16 and Petrosian radius (enclosing 90% of the flux) ≥ 3 arcsec. Of these, we visually identified 490 lopsided and 444 symmetric galaxy samples suitable for training. The trained model achieved a testing accuracy of (87 ± 0.02)%, averaged over ten independent trials. Results. Using the best-performing model, we identified 3679 lopsided and 2429 symmetric galaxies from the remaining sample. Of these, 2658 lopsided and 1455 symmetric galaxies were predicted with a high level of probability, namely, Ppred ≥ 0.85. The lopsided galaxies in our predicted samples are relatively high star-forming, bluer, low-concentration (late- type), and low-mass galaxies, as compared to the symmetric galaxies. Conclusions. Our study offers an usable catalogue of lopsided and symmetric galaxies, providing new insights into the formation of lopsidedness in disk galaxies.