The high abundance of disk galaxies without a large central bulge challengespredictions of current hydrodynamic simulations of galaxy formation. We aim toshed light on the formation of these objects by studying the redshift and massdependence of their 3D shape distribution in the COSMOS galaxy survey. Thisdistribution is inferred from the observed distribution of 2D shapes, using areconstruction method which we test using hydrodynamic simulations. We find amoderate bias for the inferred average disk circularity and relative thicknesswith respect to the disk radius, but a large bias on the dispersion of thesequantities. Applying the 3D shape reconstruction method on COSMOS data, we findno significant dependence of the inferred 3D shape distribution on redshift.The relative disk thickness shows a significant mass dependence which can beaccounted for by the scaling of disk radius with galaxy mass. We conclude thatthe shapes of disk dominated galaxies are overall not subject to disruptivemerging or feedback events below redshift $z=1.0$. This favours a scenariowhere these disks form early and subsequently undergo a tranquil evolution inisolation. In addition, our study shows that the observed 2D shapes of diskdominated galaxies can be well fitted using an ellipsoidal model for the galaxy3D morphology combined with a Gaussian model for the 3D axes ratiodistribution, confirming findings from similar work reported in the literature.Such an approach allows to build realistic mock catalogs with intrinsic galaxyshapes that will be essential for the study of intrinsic galaxy alignment as acontaminant of weak lensing surveys.

We use the Fundamental Plane (FP) to measure the redshift evolution of thedynamical mass-to-light ratio ($M_{\mathrm{dyn}}/L$) and thedynamical-to-stellar mass ratio ($M_{\mathrm{dyn}}/M_*$). Althoughconventionally used to study the properties of early-type galaxies, we hereobtain stellar kinematic measurements from the Large Early Galaxy AstrophysicsCensus (LEGA-C) Survey for a sample of $\sim1400$ massive ($\log( M_*/M_\odot)>10.5$) galaxies at $0.6<z<1.0$ that span a wide range in star formationactivity. In line with previous studies, we find a strong evolution in$M_{\mathrm{dyn}}/L_g$ with redshift. In contrast, we find only a weakdependence of the mean value of $M_{\mathrm{dyn}}/M_*$ on the specific starformation rate, and a redshift evolution that likely is explained bysystematics. Therefore, we demonstrate that star-forming and quiescent galaxieslie on the same, stable mass FP across $0<z<1$, and that the decrease in$M_{\mathrm{dyn}}/L_g$ toward high redshift can be attributed entirely toevolution of the stellar populations. Moreover, we show that the growth ofgalaxies in size and mass is constrained to occur within the mass FP. Ourresults imply either minimal structural evolution in massive galaxies since$z\sim1$, or a tight coupling in the evolution of their morphological anddynamical properties, and establish the mass FP as a tool for studying galaxyevolution with low impact from progenitor bias.