Partially premixed combustion (PPC) is an advanced combustion strategy that has been proposed to provide higher efficiency and lower emissions than conventional compression ignition, as well as greater controllability than homogeneous charge compression ignition (HCCI). Stratification of the fuel-air mixture is the key to achieving these benefits. The injection strategy, injector-piston geometry design and fuel properties are factors commonly manipulated to adjust the stratification level. In the authors' previous research, the effects of injection strategy and fuel properties on the stratification formation process were investigated. The results revealed that, for a direct-injection compression ignition engine, by sweeping the injection timing from -180° aTDC (after top dead center) to -20° aTDC, the sweep could be divided into three different regimes: an HCCI regime, a Transition regime and a PPC regime, based on the changing of mixture stratification conditions. When running in the Transition regime, the engine's efficiency and emissions were poor. Hence, it is optimal to minimize the length of the Transition regime. At the same time, it would be very beneficial to expand the PPC regime as this allows greater tolerance between the stratification level and injection timing control, thus improving controllability. Accordingly, a method was proposed for lengthening the PPC regime and shortening the Transition regime by using a small spray angle injector or a wider bowl piston. In this paper, two piston designs with different bowl profiles were tested to observe the effect of piston bowl geometry on the stratification formation process. The results show that with a wider piston bowl, the early half of the PPC regime was lengthened by approximately 50% and the Transition regime was shortened. However, an unexpected "bump" in the required intake temperature was observed within the PPC regime with the wider bowl piston, which was assumed to be caused by a "hill" on the combustion chamber wall. Simulation work based on the experimental data was conducted using the KIVA-3v code. The results were used to analyze the fuel spray development processes and equivalence ratio distributions.