The stellar initial mass function ( IMF ) plays a crucial role in determining the number of surviving stars in galaxies , the chemical composition of the interstellar medium , and the distribution of light in galaxies .
A key unsolved question is whether the IMF is universal in time and space .
Here we use state-of-the-art results of stellar evolution to show that the IMF of our Galaxy made a transition from an IMF dominated by massive stars to the present-day IMF at an early phase of the Galaxy formation .
Updated results from stellar evolution in a wide range of metallicities have been implemented in a binary population synthesis code , and compared with the observations of carbon-enhanced metal-poor ( CEMP ) stars in our Galaxy .
We find that applying the present-day IMF to Galactic halo stars causes serious contradictions with four observable quantities connected with the evolution of AGB stars .
Furthermore , a comparison between our calculations and the observations of CEMP stars may help us to constrain the transition metallicity for the IMF which we tentatively set at \textrm { [ Fe / H ] } \approx - 2 .
A novelty of the current study is the inclusion of mass loss suppression in intermediate-mass AGB stars at low-metallicity .
This significantly reduces the overproduction of nitrogen-enhanced stars that was a major problem in using the high-mass star dominated IMF in previous studies .
Our results also demonstrate that the use of the present day IMF for all time in chemical evolution models results in the overproduction of Type I.5 supernovae .
More data on stellar abundances will help to understand how the IMF has changed and what caused such a transition .