The temperature-dependent characteristic of band offsets at the heterojunction interface was studied by an internal photoemission (IPE) method. In contrast to the traditional Fowler method independent of the temperature (T), this method takes into account carrier thermalization and carrier/dopant-induced band-renormalization and band-tailing effects, and thus measures the band-offset parameter at different temperatures. Despite intensive studies in the past few decades, the T dependence of this key band parameter is still not well understood. Reexamining a p-type doped GaAs emitter/undoped Al x Ga 1−x As barrier heterojunction system disclosed its previously ignored T dependency in the valence-band offset, with a variation up to ∼−10 −4 eV/K in order to accommodate the difference in the T-dependent band gaps between GaAs and AlGaAs. Through determining the Fermi energy level (E f), IPE is able to distinguish the impurity (IB) and valence bands (VB) of extrinsic semiconductors. One important example is to determine E f of dilute magnetic semiconductors such as GaMnAs, and to understand whether it is in the IB or VB.