We investigate the differential effects of metal cooling and galactic stellar winds on the cosmological formation of individual galaxies with three sets of cosmological , hydrodynamical zoom simulations of 45 halos in the mass range 10 ^ { 11 } < M _ { \mathrm { halo } } < 10 ^ { 13 } M _ { \odot } .
Models including both galactic winds and metal cooling ( i ) suppress early star formation at z \gtrsim 1 and predict reasonable star formation histories for galaxies in present day halos of \lesssim 10 ^ { 12 } M _ { \odot } , ( ii ) produce galaxies with high cold gas fractions ( 30 - 60 per cent ) at high redshift , ( iii ) significantly reduce the galaxy formation efficiencies for halos ( M _ { \mathrm { halo } } \lesssim 10 ^ { 12 } M _ { \odot } ) at all redshifts in overall good agreement with recent observational data and constraints from abundance matching , ( iv ) result in high-redshift galaxies with reduced circular velocities in agreement with the observed Tully-Fisher relation at z \sim 2 and ( v ) significantly increase the sizes of low-mass galaxies ( M _ { \mathrm { stellar } } \lesssim 3 \times 10 ^ { 10 } M _ { \odot } ) at high redshift resulting in a weak size evolution – a trend in agreement with observations .
However , the low redshift ( z < 0.5 ) star formation rates of massive galaxies are higher than observed ( up to ten times ) .
No tested model predicts the observed size evolution for low-mass and high-mass galaxies simultaneously .
Without winds the sizes of low-mass galaxies evolve to rapidly , with winds the size evolution of massive galaxies is too weak .
Due to the delayed onset of star formation in the wind models , the metal enrichment of gas and stars is delayed and agrees well with observational constraints .
Metal cooling and stellar winds are both found to increase the ratio of in situ formed to accreted stars - the relative importance of dissipative vs. dissipationless assembly .
For halo masses below \sim 10 ^ { 12 } M _ { \odot } , this is mainly caused by less stellar accretion and compares well to predictions from semi-analytical models , but differs from abundance matching models as the in situ formed fractions of stellar mass are still too low in the simulations .
For higher masses , however , the fraction of in situ stars is over-predicted due to the unrealistically high star formation rates at low redshifts .