Nucleophilic substituted NH2-MIL-125 (Ti)/polyvinylidene fluoride hollow fiber mixed matrix membranes for CO2/CH4 separation and CO2 permeation prediction via theoretical models

Improper degree of nucleophilic substitution on PVDF may reduce the crystallinity of the membrane and weaken mechanical properties. Subsequently, the drop in membrane crystallinity might cause a decline in membrane performance, particularly in gas separation, which was unpropitious. Also, inappropriate models for predicting CO2 permeation of MMMs would lead to incompatibility between theoretical results and experimental CO2 permeation data, subsequently affecting the design of the membrane separation system. In the current work, ammonia solutions with 3 different concentrations (25, 30, and 35 v/v) were used to modify PVDF incorporated with 2 wt NH2-MIL-125 (Ti). The SEM, XPS, XRD, contact angle, and mechanical analyses were performed to investigate the structural changes, degree of substitution on PVDF, crystallinity, hydrophilicity, and changes in mechanical strength, respectively. In single and CO2/CH4 (50:50) mixed gas test, PVDF-4 showed the highest ideal and real selectivity, which also showed a good increment in CO2 permeance compared to pure PVDF and 2 wt NH2-MIL-125 (Ti)/PVDF HFMMM. Therefore, appropriate ammonia solution modification conditions are required to enhance the membrane properties without severely compromising the stability of the membrane and ensuring better CO2 gas separation. After aging, PVDF-4 demonstrated a 23 reduction in CO2 permeance, while also showing a 9 improvement in CO2/CH4 ideal selectivity. The optimized membranes were then coated with PDMS to enhance its CO2/CH4 (50:50) mixed gas separation performance. The basic Maxwell, Bruggeman, Pal, Lewis�Nielson, and Bottcher model, along with modified Pal model, were used to estimate the CO2 permeation of PVDF HFMMMS developed. Modified Pal model demonstrated �3.68 AARE towards PVDF-2 which has higher accuracy in the prediction of CO2 permeation compared to other models in this study. © 2023 Elsevier B.V.