光刻投影物镜小比例模型光学系统镜片数量较多，用每个补偿量去补偿对应失调量的传统装调方法对其不适用。利用灵敏度矩阵，并对其进行奇异值分解，对分解后的失调量向量的数值与符号进行分析，确定各失调量之间的相关性，提出了系统补偿量的选择方法，该方法仅需要分析少量向量即能选择补偿量的最佳组合。编写宏模块进行大量仿真计算，验证了补偿量敏感度与相关性分析的正确性。物镜系统初装完成后，对系统进行了仿真预装调，以确保补偿方案的可行性。实验结果表明，补偿后系统波像差均方根(RMS)值由70 nm降低到29.6 nm，优于技术指标要求，证明了提出的补偿量选择方法的正确性和有效性。

Because of the large number of the lenses in a small scale projected objective, it is unsuitable to use a traditional method that each compensator only compensates the corresponding misalignment. A sensitivity matrix is utilized to solve this problem. The sensitivity matrix is decomposed using singularly valuable decomposition method. The values and signs of displaced parameters vector are analyzed, and the correlations among each displaced parameter are obtained. A method of selecting a set of optimum combination of system compensators which only need to analyze a few displaced parameters vectors is proposed. The selected compensators are proved as correct by massive simulation calculations with the aiding of macro commands modes. After the original assembling, a pre-aligning procedure of optical system is carried out to ensure the feasibility of the aligning scheme. Test results show that the wave front errors root mean square (RMS) reduced to 29.6 nm from 70 nm after compensating, which is better than the requirements of the qualifications. The results prove the methods of selecting compensators are right and effective.