Differential interference contrast-photothermal microscopy (DIC-PTM), as a promising tool for trace analysis of nonfluorescent compounds, suffered low sensitivity in nanospace especially for aqueous samples, due to the poor thermophysical property of water and the unoptimised configuration. To improve its performance, a five-layer DIC-PTM model is built and influences of different parameters on the photothermal signal are investigated. The initial phase shift φ between two branches of the probe beam is found to be a key factor determining the detection sensitivity and response linearity: at a large φ (≤π/2) both a high sensitivity and a good linearity can be achieved, while a high signal-to-noise ratio occurs at a small φ . The steady-state photothermal phase shift φ has little impact on the linearity, which, however, is greatly influenced by the range of periodic photothermal phase shift φ . By introducing two coatings into a nanospace to confine the photothermal effect within and around the sample, the sensitivity can be enhanced from a few times to over 100 times. On an optimised DIC-PTM configuration and chip structure, detection limit down to 10 cm (or 40 molecules in a detection volume of 0.2 fL) was achieved in a 300-nm-thick nanospace. This work paves a way for optimising the DIC-PTM and chip structure for sensitive detection of analytes in nanospaces.