Numerical Modeling of Ion Chromatography in a Novel Microfluidic Device
COMSOL simulations were designed to replicate a novel 3D-printed microfluidic device (MFD) used for online monitoring of rare earth element (REE) separation via ion exchange. Combination of modeling and experimental efforts are critical to quickly and effectively produce efficient methods for separating REEs from existing waste streams to introduce alternative sources the elements. Multiple COMSOL modules were used to simulate the separation of neodymium (Nd) on cation exchange resin in acidic solutions. The CAD Import Module was used to import the entire MFD geometry. A simplified geometry of the ultraviolet-visible (UV-VIS) portion of the device was used to test the effects of air mirrors on experimental analysis via absorption spectroscopy using the Ray Optics Module. Compared to a geometry without air mirrors around the UV-VIS channel, the current chip design directed approximately 280% more light at the detector, increasing the sensitivity and lowering the detection limit of absorption analysis. Another, simplified geometry of the chromatographic column was used to investigate laminar flow around a static, spherical, and porous resin bead for the separation of Nd using the Transport of Diluted Species in Porous Media Module. A Langmuir-isotherm adsorption model was used to predict loading of Nd from the resin. The model was evaluated using time-resolved fluorescence emission spectroscopic measurements taken directly on the resin. The model was found to be highly non-linear and required extremely fine meshing of an axisymmetric 2D geometry to produce repeatable loading profiles. Usage of intraparticle and film diffusion models in the literature derived from large-scale, fully packed columns predicted significantly slower loading times than observed experimentally. Using the faster self-diffusion rate of Nd yielded much less discrepancy between experiment and model. Preliminary simulations failed to capture the initial loading profile up to 10% of resin capacity. The model, however, predicted loading very well between 10-90% capacity. These findings suggest that future work analyzing ion exchange on singular resin beads can be simulated with less experimentally derived parameters, further reducing experiments required to generate alternative sources of REEs. Future simulations discussed at the conference will address washing and elution of the resin. A ramping function to match experimental lag from tubing, which is not currently considered, will be added and is expected to increase prediction accuracy of the initial loading step. Future work will include simulations for the washing and elution profiles of the ion exchange process. These steps are expected to match experimental data well after including the resin affinity for H+ ions in solution. The completed model will be critical in future works to predict optimal experimental conditions used to evaluate ion exchange resins for large-scale separation of REEs.
