TiO2 has been extensively studied due to the possible application in solar cells and photoelectrochemical (PEC) water-splitting. However, the energy conversion efficiency is rather low because of the large band gaps (larger than 3.0 eV) of rutile and anatase TiO2. Here we introduce the multiobjective differential evolution (MODE) method as a novel global optimization algorithm to predict new polymorphs of bulk TiO2 with better optical properties than rutile and anatase TiO2. The band gaps of the new PI (Pnma) and Cl (C2) phases are found to be 1.95 and 2.64 eV. The calculation of formation energy, phonon dispersions, and thermal stability shows that the two novel phases are dynamically and thermally stable. These new TiO2 polymorphs with better electronic and optical properties may pave a new way for high-efficiency solar energy conversion.