Gas excitation of post-starburst galaxies at 0.6 < z < 1.3

TL;DR

The paper investigates how quenching operates in post-starburst galaxies at $0.6 < z < 1.3$ by directly probing molecular gas excitation with CO(5-4) and comparing it to lower-$J$ transitions. Using ALMA data for eight targets and archival CO(2-1), CO(3-2), and CO(4-3) measurements, the authors quantify the excitation via $R_{52}=L'_{CO(5-4)}/L'_{CO(2-1)}$ and construct CO SLEDs to infer ISM conditions. They find an average $R_{52}\approx0.31$, similar to high-$z$ main-sequence galaxies, but non-detections yield $R_{52}<0.11$, indicating a suppressed dense/warm gas fraction in many systems; three mergers show higher excitation with $R_{52}\approx0.49$ and SLEDs peaking at $J\gtrsim4-5$, pointing to merger-driven heating or shocks. The CO SLEDs mostly peak at $J=3$ (Milky Way-like), suggesting low gas densities and temperatures dominate quenching, while a subset shows elevated excitation likely tied to mergers or AGN activity. The study concludes that gas stabilization, feedback, or stripping likely maintain quiescence, and residual star formation alone is insufficient to exhaust the remaining molecular gas; larger multi-transition samples are needed to generalize these findings and to explore variations in $\alpha_{CO}$ for quiescent systems.

Abstract

Molecular gas traces both the fuel for star formation and the processes that regulate it. Observing its physical state (e.g., excitation) reveals when and why galaxies stop forming stars. We observed the CO(5-4) emission of 8 post-starburst galaxies at z ~ 0.6-1.3. To our knowledge, this is the first time that high-J transitions are probed for quiescent galaxies beyond the local Universe. All targets are detected in CO(2-1) or CO(3-2) and have molecular gas fractions up to 20%. Using the ratio R52=L'CO(5-4)/L'CO(2-1) as a proxy for gas excitation, we test how quenching occurs. Low R52 values would indicate suppressed fractions of dense/warm gas relative to cold and diffuse gas, while ratios typical of main-sequence galaxies would imply that quenching is still ongoing and that star formation may exhaust the remaining gas. On average our post-starbursts have R52 = 0.31, comparable to high-redshift galaxies. However, CO(5-4) non-detections, corresponding to galaxies without signs of interaction, yield R52<0.11, 2 times lower than local star-forming galaxies. The average CO Spectral Line Energy Distribution (SLED) peaks at J = 3, similar to the Milky Way. Three galaxies show signs of ongoing mergers and have R52 = 0.49 and CO SLEDs peaking at J > 4-5, similar to high-redshift galaxies. At least one requires additional mechanisms (AGN, shocks) to explain the rise of the SLED up to J=5. CO excitation helps distinguishing among mechanisms driving the low star formation efficiency (SFE) of post-starburst galaxies. The low SFE might be due to high kinetic temperatures and low gas densities yielding high excitation, or due to low gas densities implying low excitation. On average, our sample favours the latter scenario, suggesting that gas stabilization, feedback, or stripping are needed to keep galaxies quiescent, and that residual star formation alone cannot deplete the remaining molecular gas.

Figures (6)

• Figure 1: ALMA data of our sample galaxies. First column: ALMA 2D maps of the continuum- CO(5-4) line. The cyan solid and dashed contours indicate respectively positive and negative levels of (2.5, 3.5, 4.5, 5.5) rms. The beam is reported in the bottom left corner as the white filled ellipse. Each stamp has a size of 10" $\times$ 10". The cyan cross indicates the center of our post-SB targets as estimated from the optical emission. Second column: 1D spectra of sources extracted using a PSF to maximize the S/N. The pink shaded areas indicate the 1$\sigma$ velocity range over which we measured the CO(5-4) line flux or its upper limit. For detections we also report the Gaussian fit of the emissions (Section \ref{['subsec:co54']}). The vertical dashed lines indicate the observed frequency corresponding to the redshift of the low-$J$ CO emission. Third column: rest-frame optical grz images from the DESI Legacy Survey DR9 Dey2019. The cyan contours show the CO(5-4) emission, while the orange and pink indicate the low-$J$ (CO(2-1) or CO(3-2)) or CO(4-3) emission respectively. The ellipses indicate the beam size. Each stamp has a size of $20\arcsec \times 20\arcsec$. Contour levels are the same as in the first panels. (Continues on next page)
• Figure 2: ALMA data of our sample galaxies (continued). As reported in Bezanson2021, the $\sim 1\arcsec$ offset of the CO(2-1) emission of J2202 from the optical centroid is not significant given the resolution and S/N of the data. However, it is interesting to note that the optical image of this galaxy appears to be slightly asymmetric.
• Figure 3: Ratio of CO(5-4) to CO(2-1), a proxy for molecular gas excitation, as a function of specific star formation rate (sSFR). Our sample is shown with red circles (non mergers) and red stars (mergers) and is compared to literature samples of local star-forming galaxies (blue pentagons), local (U)LIRGs (blue squares), high-redshift main-sequence galaxies (cyan circles), and high-redshift starbursts (cyan squares; Liu2021). The two star-forming companions of ID83492 are shown as brown circles. For post-starbursts where CO(2-1) is not observed, empty symbols indicate a pseudo CO(2-1) estimated from the CO(3-2) measurement assuming the Milky Way $R_{32} = 0.54$ ratio Carilli2013. Large symbols indicate the population averages. The average for post-SBs without CO(5-4) detection is estimated from stacking (large red circle). The average of the whole sample of post-SBs (large red diamond) including both mergers and non mergers is derived with the Kaplan–Meier estimator from survival analysis (see text). A fit to the literature samples (gray solid line) with its scatter (gray shaded area) from Valentino2020b is shown and extrapolated toward lower sSFR (gray dashed line).
• Figure 4: CO SLED of our post-starburst galaxies. Individual galaxies are shown with small colored circles (measurements) and down-pointing triangles (3$\sigma$ upper limits), while the average SLED is indicated by red empty circles. For comparison, we also show the Milky Way SLED (black squares; Carilli2013), the average SLED of main-sequence galaxies (black triangles; Valentino2020b), and the SLED predicted by the model of Narayanan2014 assuming a SFR surface density of $\Sigma_\mathrm{SFR} = 0.05\, \mathrm{M_\odot\, yr^{-1}\, kpc^{-2}}$ (blue diamonds) and $\Sigma_\mathrm{SFR} = 500\, \mathrm{M_\odot\, yr^{-1}\, kpc^{-2}}$ (pink diamonds).
• Figure 5: CO(5-4) luminosity of our sample galaxies compared with the maximum IR luminosity allowed by the continuum non detections. We computed the maximum L$_\mathrm{IR}$ by considering the average SED template from Magdis2021 and rescaling it to match the observations. We considered both a template with dust temperature of 20 K (red circles) and 30 K (blue squares). We refer to Section \ref{['subsec:residual_sf']} for details regarding the estimate of the IR luminosity. We overplot (black line) the CO(5-4) - L$_\mathrm{IR}$ relation estimated by Valentino2020b for star-forming galaxies that seems not to hold for our sample of post-starbursts. The top axis indicates the SFR estimated from the maximum $\mathrm{L_{IR}}$ with the Kennicutt2021 conversion.