Thermomechanical properties of two-dimensional covalent organic frameworks studied by machine-learning molecular dynamics

Two-dimensional (2D) covalent organic frameworks (COFs) are promising emerging 2D polymeric materials with broad applications such as adsorption, energy storage, and catalysis. However, their mechanical properties have not been systematically studied yet. Here, the machine-learned neuroevolution potential (NEP) was developed to study the elastic properties of representative 2D COFs, e.g., COF-1 and COF-5. The trained NEP enables one to study the mechanical properties of 2D COFs in realistic situations (e.g., finite size and temperature) and enjoys greatly improved computational efficiency as compared with density functional theory calculations. With the aid of the obtained NEP, molecular dynamics (MD) simulations together with the strain fluctuation method were employed to evaluate the elastic constants of the considered 2D COFs at different temperatures. The elastic constants of COF-1 and COF-5 monolayers are found to decrease with growing temperature, though they are almost isotropic, irrespective of the temperature. The thermally induced softening of 2D COFs below a critical temperature is majorly attributed to their inherent ripple configurations at finite temperatures, while, above the critical temperature, the damping effect of anharmonic vibrations becomes the dominant factor. Based on the proposed mechanisms, analytical models were developed for capturing the temperature dependence of the elastic constants, which were found to agree with MD simulation results well. This work provides an in-depth insight into the thermomechanical properties of 2D COFs, which guides the development of COF materials with tailored mechanical behaviors for enhancing their performance in various applications.