macroscopic scales. This calls for a datadriven approach implementing efficient and accurate multi-scale modelling techniques to inform and interpret laboratory experiments. These include the lateralforce atomic force microscopy (AFM), which is the key experimental tool used for quantifying the nanoscale friction processes. [4] In particular, AFM is able to record the mechanical force exerted by a crystalline surface onto a nano-scale asperity dragged on top of it. Common computational models frequently used to study the microscopic laws of friction include quantum-mechanical first principles calculations, atomistic models based on molecular dynamics, nonlinear Prandtl-Tomlinson (PT) or Frenkel-Kontrova models, agentbased earthquake models, and models based on continuum mechanics applicable at macroscopic scales. The multi-scale modeling approach requires interfacing typically two or more of these levels of modeling into a systematic framework. [5] So far, the most refined multi-scale modelling of atomic friction has combined first principles calculations with molecular dynamics methods. [6,7] While highly accurate, this methodology requires significant computational resources, which restricts its applicability to relatively short time-and length-scales. More importantly, the lack of reliable force fields for the majority of materials is a major limitation for the transferability needed for new-materials screenings.The mesoscopic to macroscopic scale range of the friction processes has been studied by bridging atomistic molecular dynamics, linear response theory, and continuum mechanics into a unified multi-scale approach. [8] However, continuum theories inherently exclude the possibility of thermal and structural fluctuations and their applicability to the nanoscale friction range becomes problematic, as it is inherently a farfrom-equilibrium phenomenon dominated by such fluctuations and size effects. [9] Instead, it is often more fruitful to employ non-equilibrium statistical mechanics combined with transition state theory or stochastic Langevin dynamics. [10] This approach has been successful in generalizing, for example, the classical PT model to describe the thermally activated nano-scale friction in AFM experiments, [11][12][13][14][15] which qualitatively captured the velocity, load, and temperature dependencies observed in experiments. [13,16,17] To advance this statistical level of modeling requires incorporating the fundamental ability to describe the nanoscale frictional behavior of specific materials, which is the main objective in this work.The main result of this work is a fully consistent thermally activated thermodynamic model, which combines mesoscopic A multi-scale computational framework suitable for designing solid lubricant interfaces fully in silico is presented. The approach is based on stochastic thermodynamics founded on the classical thermally activated 2D Prandtl-Tomlinson model, linked with first principles methods to accurately capture the properties of real materials. It allows inves...