星系内禀 alignments（IA）是宇宙剪切信号（GG）中的重要污染源。对于交叉关联红移子样本而言，星系本征椭率-引力剪切（IG）相关性通常是这种污染的主要组成部分。自校准（SC）方法是减轻GG信号中此类污染最有效的方法之一。在光学巡天中，SC方法首先利用透镜-源对红移依赖关系从观测到的星系-星系透镜相关性中提取星系数密度-星系本征椭率（gI）相关性。通过gI相关性和其他透镜观测值的尺度关系计算IG相关性。我们通过对尺度关系进行修改，将SC方法扩展到非线性区域，使其能够考虑非线性星系偏倚模型以及各种IA模型。本研究为鲁宾天文台时域巡天（LSST Y1）提议的第一年源星系红移子样本提供了检测IG相关的框架。我们针对潮汐排列和潮汐转矩（TATT）模型进行了IA测试，并发现对于交叉关联和自关联红移子样本，尺度关系的精度分别在10%和20%以内。因此，IG污染的抑制因子分别为10和5。我们还测试了该方法的稳健性，并发现即使在IA和偏倚参数存在适度不确定性的情况下，对于所有交叉关联红移子样本组合，IG污染的抑制因子仍可达到5。[摘要缩短版]

Intrinsic alignment (IA) of galaxies is a challenging source of contamination in the Cosmic shear (GG) signals. The galaxy intrinsic ellipticity-gravitational shear (IG) correlation is generally the most dominant component of such contamination for cross-correlating redshift bins. The self-calibration (SC) method is one of the most effective techniques to mitigate such contamination from the GG signal. In a photometric survey, the SC method first extracts the galaxy number density-galaxy intrinsic ellipticity (gI) correlation from the observed galaxy-galaxy lensing correlation using the redshift dependence of lens-source pairs. The IG correlation is computed through a scaling relation using the gI correlation and other lensing observables. We extend the SC method beyond the linear regime by modifying its scaling relation which can account for the non-linear galaxy bias model and various IA models. In this study, we provide a framework to detect the IG correlation for the redshift bins for source galaxies for the proposed year 1 survey of the Rubin Legacy Survey of Space and Time (LSST Y1). We tested the method for the tidal alignment and tidal torquing (TATT) model of IA and we found that the scaling relation is accurate within 10$\%$ and 20$\%$ for cross-correlating and auto-correlating redshift bins, respectively. Hence the suppression of IG contamination can be accomplished with a factor of 10 and 5, for cross-correlating and auto-correlating redshift bins, respectively. We tested the method's robustness and found that the suppression of IG contamination by a factor of 5 is still achievable for all combinations of cross-correlating bins even with the inclusion of a moderate amount of uncertainties on IA and bias parameters, respectively. [Abridged]