Separation of gases and liquids mixtures into their constituents is crucial for many industrial processes. Accordingly, development of high performance membranes with enhanced permeability, salt rejections and low energy consumption has drawn many researchers' attentions. High potential candidates are Mixed-Matrix Membranes (MMMs) combining the benefits of both polymer and carbon nano-tubes (CNTs) for high processability, selectivity and throughput. However, there are still challenging issues, such as, poor miscibility between the polymeric phase and the carbon nano-tubes and the ensuing phase separation problems. In this study, by combining experiments and multiscale modeling approaches (i.e., molecular dynamics and mesoscale simulation), we focus on (1) the functionalization of multi walled CNTs (MWCNTs) to improve the compatibility between MWCNTs and a co-polyimide (P84) and to obtain homogeneous dispersion without phase separation problems and on (2) the mechanism and validation of homogeneous dispersion of MWCNTs through P84. First, we perform molecular dynamics (MD) simulation to calculate the solubility parameters of various CNTs, pristine and functionalized with the polar groups -OH and -NH2, and polymer matrix used for CNT/polymer composite membrane fabrication. After then, mesoscale simulations are executed to understand the macroscopic CNT dispersion through the polymer matrix. Finally, modeling results are validated by experimental data obtained by the preparation of composite P84 membranes containing MWCNTs pristine or functionalized. The MWCNTs functionalization is carried out by oxidation or amination procedure in order to introduce oxygenated and amine species, respectively. A good agreement between theoretical calculation and experimental results is achieved confirming the necessity to functionalize the MWCNTs with polar groups in order to obtain a good dispersion in the hydrophilic P84 matrix and the prediction of an optimum range of functional groups in order to obtain well dispersed MMM. (C) 2016 Elsevier B.V. All rights reserved.