Current modeling and verification approaches for self-adaptive software rarely consider temporal properties. However, in time-critical application domain, the correct operation of self-adaptive software depends on the correctness of self-adaptive logic as well as temporal properties of self-adaptive software dynamic processes. For this end, temporal properties for self-adaptive software were explicitly defined, such as, monitoring period, delay trigger time, deadline of self-adaptive process, self-adaptive adjusting time and self-adaptive steady time. Then, a Timed Automata Network (TAN) based modeling templates for temporal properties of self-adaptive software dynamic processes were constructed. Finally, the temporal properties were formally described with Timed Computation Tree Logic (TCTL), and then were analyzed and verified. Combining with a self-adaptive example, this paper has validated the proposed approach. The results show that the proposed approach can explicitly depict temporal properties of self-adaptive software, and can reduce its formal modeling complexity.