Date of Award


Document Type

Open Access Dissertation


Chemical Engineering


College of Engineering and Computing

First Advisor

Miao Yu


To develop ultrathin GO membranes with both high permeation rate and excellent selectivity, it is essential to understand the interlayer nanostructure and its influence on water purification performance. We purposely deposited GO at a fast rate and at a rate ~12 times slower to control the interlayer nanostructure of the resulting membranes. The d-spacing difference between proposed thermodynamically favored interlayer structure which formed at slow deposition rate and another relative randomly packed interlayer structure formed at fast deposition rate were corroborated by XRD, organic vapor deposition, and AFM. Molecular dynamics simulations further confirmed that, in type I structure, functionalized patches and pristine graphene patches on neighboring GO layers facing themselves, which not only leaded to smaller d-spacing but also facilitated fast water permeation; in type II structure, functionalized and pristine were mismatched, leading to larger d-spacing and drastically retarded water permeation. Our experimental results also showed that compared with type II structure, narrower hydrophobic nanochannels in type I structure lead to 2.5~4 times faster water permeation rate and 1.8~4 times higher salt rejection. We believed our finding, tuning the GO interlayer nanostructure by simply controlling GO flake deposition rate in solution phase deposition process, helped break the current trade-off between water flux and precise sieving performance of GO membranes, and may eventually bring about novel design of ultrathin GO-based membranes for high flux and high selectivity water purification.

After figured out the process-structure-performance relationship of GO membranes, a layer-by-layer deposition method was designed to prepare GO membranes, through this deposition technique, GO layers could have enough time to self-assemble and form the thermodynamically favored structure. To overcome GO’s inherent dispensability in the water environment and to lock the d-spacing at sub-nanometer scale, the as-prepared GO membranes were thermally reduced under vacuum. The 3 nm reduced graphene oxide (rGO) membrane exhibited no permeation even for water molecules and then O2 plasma was introduced to create extra defects on the membrane surface, which dramatically facilitated water permeation but still could block large molecules (such as methylene blue) in high efficiency. By tuning the plasma treatment time, the 3 nm rGO membranes achieved ~98% rejection for MB and pure water flux as high as about 44 Lˑh-1·bar-1ˑm-2. Moreover, the optimized 10s plasma etched 3 nm rGO also exhibited 100% rejection and good antifouling ability for humic acid.

As the flux cross the membrane decreases with membrane thickness, to balance this pay off between permeability and selectivity, membrane should have thin thickness to provide high flux and appropriate pores to allow the passage of water but block large solutes at the same time. Therefore, the “ultimate” target in membrane science is to fabricate a membrane in the form of only one atomic thickness and with suitable pores on its surface. The properties of GO just meet these two requirements. In this part of research, by clarifying two distinct water transportation mechanisms for membranes with sub-monolayer and multilayer GO coverage, we proposed a methodology to fabricate nominal single-layered GO membrane. While the calculated GO coverage increased from less than 100% to multilayers, the water flux exhibited a transition from two stages of linear decreases to exponential decrease, the condition to prepare nominal single-layered GO membrane was extrapolated from the turning point of the linear-to-exponential transition, the as-fabricated membrane with thickness closing to one-carbon-atom exhibited high water permeance around 64 L· h-1·bar-1 m-2. In this nominal single-layered GO membrane, defects on GO flakes provided major contribution for its sieving properties. By evaluating the separation performance of this membrane with rigid molecules, the effective defect size of GO was determined to be ~1.2-1.7 nm. This membrane with nominal single-layer GO cover also show great potential in protein separation.

movie.mpg (63082 kB)