AbstractThe thermodynamic data of iron oxides reduction reactions from the most recognized thermodynamic database (NIST‐JANAF) show remarkable difference between stoichiometric and nonstoichiometric iron oxides. Iron oxides reduction equilibria in both CO‐CO2 and H2‐H2O atmospheres are calculated with Dieckmann's and Weiss’ defect models. Relevant literatures are investigated and reduction experiments are carried out to complement and interpret the newly calculated equilibrium diagram. The results suggest a conjecture of two routes for hematite reduction reactions. With the ideal and widely accepted mechanism, hematite is reduced to magnetite and then to iron below 576 °C, while the reduction route follows the sequence of Fe2O3 → Fe3−δO4 → FexO → Fe above 576 °C. With the regular but always unrecognized mechanism, the reduction process of hematite experiences Fe3O4, FeO, and Fe step by step in all possible temperature above 156 °C. In the regular mechanism, sufficient scattered impurities occupy crystal interstices of magnetite and prevent the newly produced FeO unit cell from dissolving into the solid solution of magnetite, and then the FeO will accumulate above 156 °C. Actually the regular mechanism is hard to realize and usually confused with the ideal mechanism. The presence of the regular mechanism is proved by experimental phenomena of the drop of eutectoid temperature.