As wafer circuit widths shrink down, wafer fabrication processes require stringent quality control. Therefore, fabs recently tend to clean a chamber after processing each wafer, in order to remove chemical residuals within the chamber. Such chamber cleaning, called purge operation, increases scheduling complexity in robotized cluster tools. In this paper, we examine scheduling problems of single-armed cluster tools with purge operations for series-parallel chambers. By extending the well-known backward sequence, we propose a backward(z) sequence that allows partial loading for parallel chambers, where vector z specifies how many chambers zi of each process step i are kept empty for cleaning. We then propose a way of finding optimal vector z* and identify when backward(z*) achieves the minimum cycle time among all possible sequences. We present experimental results on the accuracy of backward(z*).