The strength reduction finite element method （SRFEM） is an effective approach for evaluating the seismic safety factor of slopes. The key to this method lies in selecting a rational criterion for slope failure. At present, three types of criteria are commonly used, namely the penetration of plastic zones （equivalent plastic strain）, non-convergence of numerical simulation calculations, and abrupt displacement changes at characteristic points. Most scholars believe that the penetration of equivalent plastic strain is a sufficient but not a necessary condition for slope failure. Moreover, the selection standard for the magnitude of plastic strain has certain non-deterministic and other subjective factors. The problem with using non-convergence of finite element calculations as a criterion is that the convergence standard is limited by the software, and its physical meaning is not clear. The abrupt displacement change or non-convergence of characteristic points usually involves selecting characteristic points at the toe of the slope, the midpoint of the slope face, and the crest of the slope. This situation can only reflect the local state of the slope. The slope may be locally unstable rather than globally unstable. Therefore, these three types of criteria have certain differences and subjectivity in practical applications.

In this study, through the SRFEM, the concept of a rainbow-like displacement contour map （RLDCM） during slope instability was proposed. A rainbow-like displacement contour map was established as a criterion for the seismic stability of slopes. That is, when the slope fails globally, the displacement contour map should meet the following three conditions: ① The soil elements within the landslide mass undergo relative sliding along the circular arc slip surface compared with the stable area inside the slope, presenting a clear circular-arc-shaped boundary. ② The displacement of the landslide mass shows a circular-arc-shaped belt distribution along the slope direction, with distinct layers resembling a rainbow. ③ The displacement of each layer of the landslide mass first increases and then suddenly decreases from the slope surface to the inside. The layer with the maximum displacement is the slip surface of the slope.

The main conclusions drawn from this study are as follows:

1） The appearance of the RLDCM can serve as a criterion for the overall failure of soil slopes under seismic action. It has a clear physical meaning and is easy to operate. Moreover, it can also be combined with other criteria, such as the penetration of plastic zones, to confirm the accuracy of the safety factor.

2） The safety factors obtained using four criteria, i.e., penetration of plastic zones, abrupt displacement of characteristic points, displacement time-history curves, and RLDCM, are not significantly different. Among them, the safety factor obtained using the abrupt displacement criterion has subjective differences and a range of acceptable values.

3） When using the abrupt displacement criterion of characteristic points, several points should be selected on the slope surface. Monitoring points at the toe of the slope may yield inaccurate results because they may be located above the shear exit.