**2. Theoretical Model for the Creep Size E**ff**ect**

To begin with, the creep process under indentation tests is considered to be accommodated with the evolution of dislocation microstructures beneath the indenter tip. At the onset of creep deformation, dislocation loops with Burgers vectors normal to the surface plane are generated to address the geometrical shape change at the contact surface [41]. Then, the existing GNDs are forced to move radially from the inner creep region (close to the indenter tip) to the boundary between the creep and

elastic regions. For simplicity, we assume that the creep deformation can be discretized into numbers of sequential activation events. During each activation event, the indenter tip moves downwards by a distance of *b*, and forces the *i*-th dislocation loop (1 ≤ *i* ≤ *N* with *N* being the number of dislocation loops) to sweep a distance of *s*. Here, *b* is the magnitude of the Burgers vector and *s* is the spacing between dislocation loops [41], as illustrated in Figure 1.

**Figure 1.** Schematic of indentation creep with the evolution of geometrically necessary dislocations (GNDs), which are performed by a conical indentation. When tan <sup>θ</sup> <sup>=</sup> <sup>√</sup> π/24.5 = 0.358 (θ is the angle between the indenter and sample), the model is also applicable to the Berkovich indentation following the self-similar principle [41]. The creep process is discretized into the expansion of circular dislocation loops. During each creep activation, the indenter tip moves forward by a length of *b*, and a new dislocation loop is generated from the indenter tip that forces existing dislocation loops to creep radially with a distance of *s*. Thereinto, *b* and *s* are, respectively, the magnitude of the Burgers vector and the spacing between individual dislocation loops.
