Natural gas feeds often contain contaminants such as CO₂, H₂S, H₂O, and small hydrocarbons. Carbon dioxide is a major contaminant reducing the heating value of the gas and causing pipeline corrosion, so CO₂ level should be lowered to below 2% to meet the United States pipeline specifications. Membrane separation technology can be advantageous over cryogenic distillation and amine adsorption in terms of cost and efficiency. The key hurdle to overcome in polymeric membrane separation technology is improvement in selectivity, productivity, and durability without introducing significant additional cost. The ultimate goal of this study is to analyze effects due to polymerization conditions and imidization methods on properties of 1,3-propanediol monoesterified crosslinkable polyimide (PDMC). Hillock, Omole, Ward, and Ma did work on PDMC synthesis; however, variability of polymer properties remains a challenge that must be overcome for industrial implementation of PDMC material. First, reaction temperature and reaction time of polymerization prior to imidization were considered as key conditions to affect molecular weight, crosslinkability and transport properties of polymer. Batches with controlled reaction temperature and time were prepared, and properties of each dense film were measured and optimized in terms of permeability, selectivity, and plasticization suppression. Second, imidization methods for PDMC were also studied. There are mainly two kinds of Imidization: chemical Imidization and thermal Imidization. Surprisingly, thermally imidized PDMC showed 70% higher permeability than chemically imidized samples with minimal acrifice in selectivity. At high reaction temperature during the thermal imidization, transamidation can occur. It is believed that the transamidation led to more randomized sequence distribution in the thermally imidized samples. We thus hypothesize that the higher permeability of the thermally imidized PDMC results from greater uniformity of the sequence distribution, as compared to the chemically imidized sample that does not experience high temperature during imidization. XRD, DSC, DMA, and permeation instruments checked and supported this hypothesis. FTIR, TGA, and NMR ruled out the possibility of an alternate hypothesis related to side reaction. Finally, effects of aggressive feed conditions on both chemically imidized PDMC and thermally imidized PDMC dense film were examined. The aggressive feed conditions include high CO₂ partial pressure, operating temperatures, and exposure to high feed pressure. Testing aggressive feed conditions for dense film should be pursued before pursuing hollow fiber applications, to decouple effects on the basic material from those on the more complex asymmetric morphology. This study enables understanding of the disparity between various previous researchers’ selectivity and permeability values. The work shows clearly that polymerization conditions and imidization methods must be specified and controlled to achieve consistently desirable polymer properties. In addition, for batch scale-up and development to a hollow fiber, this fundamental study should enable production of high molecular weight PDMC with good fiber spinnability and defect-free structure.

Mixed matrix nanocomposite membranes composed of a crosslinkable polyimide matrix and high-silica molecular sieve particles were developed for purifying natural gas. It was shown that ideal mixed matrix effects were not possible without sieve surface modification. A previously developed Grignard procedure was utilized to deposit magnesium hydroxide nanostructures on the sieve surface in order to enhance polymer adhesion. Analyses of Grignard-treated sieves pointed to the formation of non-selective voids within the surface deposited layer. These voids were suspected to lead to lower-than-expected membrane performance. In order to improve membrane transport, a reactive sizing procedure was developed to fill these voids with polyimide-miscible material. In a serendipitous discovery, as-received sieves--when treated with this reactive sizing procedure--resulted in nearly identical membrane performance as reactive-sized, Grignard-treated sieves. This observation lead to the speculation of a non-ideal transport mechanism in mixed matrix membranes.