Abstract　 　　
Although the thick disk of the Milky Way has been observed for about 30 years, its origin is still unknown. Dozens of models of the formation of the thick disk has been proposed during the past two decades, and finally four of them turn out to be the most likely ones:

1) the accreted debris of the disrupted dwarf galaxies;

2) heating of the old thin disk by the merging events;

3) gas-rich mergers (form in situ);

4) radial migration induced by transient spiral arms.

These four scenarios are very hard to be distinguished only with observed kinematic data. Therefore, chemo-kinematic studies seem the more plausible way to test them. With about 13000 G dwarf disk stars selected from SDSS DR7/SEGUE survey, we found that the thick disk stars can be separated into two groups:

 i) the stars on near-circular orbits with enhanced alpha-abundance and high metallicity are smoothly transited from the thin disk, which are also metal-rich but with low-alpha abundances;

 ii) the stars on eccentric orbits have enhanced alpha abundances, low metallicity and distinguished kinematic properties with thin disk stars. Qualitatively comparing with simulations we believe that the former is likely related to the radial migration. And the later is consistent with neither the accreted debris nor the heating scenarios and hence may be originated from gas-rich mergers. Therefore, we claim that the thick disk stars have at least two distinct origins. A follow-up investigation of the elemental abundances of the thick disk stars compiled from the literature provides more chemo-kinematic evidence that the thick disk stars have two distinct chemical evolution channels. The correlation between the abundances, e.g. [Fe/H], [a/Fe], [Ba/Fe], and [Eu/Fe], and the orbital properties, e.g. eccentricity and angular momentum, are naturally interpreted by the gas-rich mergers