Understanding and modelling of moisture transport in cellulose-based materials is essential for biological processes in plants as well as industrial applications. At the continuum level, this transport is often described by Fick’s first law, where the moisture flux is proportional to a concentration gradient, with the transport diffusion coefficient playing a crucial role. However, determining this coefficient for cellulose structures remains a challenge due to the intricate interactions between water and cellulose. 
Cellulose materials are inherently porous and water confined between the cellulose has been found to be more “stubborn” than bulk water [1]. In fact, a thin layer of water between the cellulose nanofibrils is needed for thermodynamic equilibrium [2]. Inside the fiber wall of wood and paper, the pores can be on the nanoscale [3], where interactions between the water and cellulose in the pore become increasingly significant. To capture these effects, a molecular dynamics (MD) approach is motivated, which offers atomic-level resolution and means to quantify interactions. 
In the present study, a novel non-equilibrium MD simulation method, first introduced by H. Frentrup et al. [4] for a general nanopore system, was employed to determine the transport diffusivities of moisture in cellulose nanopores of varying shapes and sizes. By applying an external potential to a small region in the simulation box, a concentration gradient can be introduced, effectively mimicking the conditions of Fick’s first law. Previously applied to a variety of pores and fluids [5,6,7], the method is tested here for the first time in the context of cellulose and water. 
The results show the impact of pore morphology on the diffusivity and quantify the specific interactions between water and cellulose during concentration-driven transport. The determined coefficients provide initial values to be used in a continuum model, bridging the gap between molecular interactions and macroscopic transport.