- Wang, Xin;
- Jones, Tucker A;
- Treu, Tommaso;
- Morishita, Takahiro;
- Abramson, Louis E;
- Brammer, Gabriel B;
- Huang, Kuang-Han;
- Malkan, Matthew A;
- Schmidt, Kasper B;
- Fontana, Adriano;
- Grillo, Claudio;
- Henry, Alaina L;
- Karman, Wouter;
- Kelly, Patrick L;
- Mason, Charlotte A;
- Mercurio, Amata;
- Rosati, Piero;
- Sharon, Keren;
- Trenti, Michele;
- Vulcani, Benedetta
We combine deep Hubble Space Telescope grism spectroscopy with a new Bayesian method to derive maps of gas-phase metallicity for 10 star-forming galaxies at high redshift (). Exploiting lensing magnification by the foreground cluster MACS1149.6+2223, we reach sub-kiloparsec spatial resolution and push the limit of stellar mass associated with such high-z spatially resolved measurements below for the first time. Our maps exhibit diverse morphologies, indicative of various effects such as efficient radial mixing from tidal torques, rapid accretion of low-metallicity gas, and other physical processes that can affect the gas and metallicity distributions in individual galaxies. Based upon an exhaustive sample of all existing sub-kiloparesec resolution metallicity gradient measurements at high z, we find that predictions given by analytical chemical evolution models assuming a relatively extended star-formation profile in the early disk-formation phase can explain the majority of observed metallicity gradients, without involving galactic feedback or radial outflows. We observe a tentative correlation between stellar mass and metallicity gradients, consistent with the "downsizing" galaxy formation picture that more massive galaxies are more evolved into a later phase of disk growth, where they experience more coherent mass assembly at all radii and thus show shallower metallicity gradients. In addition to the spatially resolved analysis, we compile a sample of homogeneously cross-calibrated integrated metallicity measurements spanning three orders of magnitude in stellar mass at z ∼ 1.8. We use this sample to study the mass-metallicity relation (MZR) and find that the slope of the observed MZR can rule out the momentum-driven wind model at a 3σ confidence level.