Two rest-frame wavelength measurements of galaxy sizes at $z<1$: the evolutionary effects of emerging bulges and quenched newcomers
Authors: Angelo George, Ivana Damjanov, Marcin Sawicki, Stéphane Arnouts, Guillaume Desprez, Stephen Gwyn, Vincent Picouet, Simon Birrer, John Silverman
Abstract: We analyze the size evolution of $16000$ star-forming galaxies (SFGs) and $5000$ quiescent galaxies (QGs) with mass $M_*>10^{9.5}M_\odot$ at $0.1<z<0.9$ from the COSMOS field using deep CLAUDS+HSC imaging in two rest-frame wavelengths, $3000$\r{A} (UV light) and $5000$\r{A} (visible light). With half-light radius ($R_e$) as proxy for size, SFGs at characteristic mass $M_0 = 5\times10^{10}M_\odot$ grow by $20\%$ ($30\%$) in UV (visible) light since $z\sim1$ and the strength of their size evolution increases with stellar mass. After accounting for mass growth due to star formation, we estimate that SFGs grow by $75\%$ in all stellar mass bins and in both rest-frame wavelengths. Redder SFGs are more massive, smaller and more concentrated than bluer SFGs and the fraction of red SFGs increases with time. These results point to the emergence of bulges as the dominant mechanism for the average size growth of SFGs. We find two threshold values for the stellar mass density within central $1$kpc (${\Sigma}_1$): all SFGs with $\log{\Sigma}_1 > 9$ are red and only QGs have $\log{\Sigma}_1>9.7$. The size of $M_*=M_0$ QGs grows by $50\%$ ($110\%$) in the UV (visible) light. Up to $\sim20\%$ of this increase in size of massive QGs is due to newcomers (recently quenched galaxies). However, newcomers cannot explain the observed pace in the size growth of QGs; that trend has to be dominated by processes affecting individual galaxies, such as minor mergers and accretion.
Ask questions about this paper to our AI assistant
You can also chat with multiple papers at once here.
The previous summary was created more than a year ago and can be re-run (if necessary) by clicking on the Run button below.
Look for similar papers (in beta version)
By clicking on the button above, our algorithm compares the title and abstract of this paper with all the papers of the same arXiv category that we have summarized, and finds the closest ones. You can rerun it from time to time: the results improve as our database grows.
Disclaimer: The AI-based summarization tool and virtual assistant provided on this website may not always provide accurate and complete summaries or responses. We encourage you to carefully review and evaluate the generated content to ensure its quality and relevance to your needs.