In this paper, hierarchical model predictive control is applied to mild HEVs in order to improve fuel economy and reduce NOx emissions. There is a difference in the timescale between the temperature control (high-level) and the torque control (low-level), which are connected problems via the battery. In high-level of the hierarchical control, the optimal amount of heating from the battery to the catalyst in the exhaust gas after-treatment system is determined. The amount of heating must be determined in such a way that the catalyst can optimally purify NOx emitted from the engine. At the low-level of the hierarchical control, the optimal torque distribution for the engine and motor is determined. The controller at both levels then applies model predictive control to account for the future behavior of the HEV's components. Fuel consumption, battery SoC and NOx emissions from the engine are taken into account. Finally, the effectiveness of our proposed method is demonstrated by simulations.