After several hours of discharge cleaning, it has been observed that the decay time of the water‐vapor partial pressure changes from one well defined value to another. A long decay time observed in the later phase can be interpreted as a result of a slow deoxidization rate of chromium oxide, which may dominate the cleaning process in this phase. Optimization of the electron density for the cleaning is discussed, comparing the experimental results on density dependence of the water‐vapor partial pressure with a result based on a zero‐dimensional calculation for particle balance. One of the essential points for effective cleaning is found to be raising the electron density of the plasma high enough that the dissociation loss rate of H2O is as large as the sticking loss rate. A density as high as 1011cm−3is required for a clean surface condition where sticking probability is presumed to be around 0.5.