470655 Mechanistic Insights into the Reactive Sorption of Sulfur Compounds Using Metal Oxides
Copper (II) oxide nanofibers were synthesized via electrospinning using copper nitrate trihydrate, copper acetate, or zinc acetate as precursors. Electrospinning solutions consisted of the metal salt precursor and polyvinylpyrrolidone (PVP) dissolved in ethanol/water mixtures. Electrospinning syntheses were conducted at 30-35 kV, 5 mL/hr extrusion rate, and 15-17 inch distance between the tip of the extrusion syringe needle and the collector plate. The nanofibers were subsequently treated in air at 625°C (at 0.7°C/min). CuO nanoparticles were prepared via precipitation using 0.1 M solutions of copper salts and 0.1 M NaOH solutions. The washed precipitates were dried in air at 80°C for 16 h and then treated in air at 500°C for 4 h. Breakthrough curves were collected in a tubular reactor with plug flow hydrodynamics. Adsorption models based on irreversible adsorption isotherms and linear driving force (LDF) approximations were used to analyze data from these fixed bed experiments. These models are appropriate for making quantitative assessments of reactive sorption rates because these reactions are essentially irreversible (i.e., equilibrium constants >1010). The resulting equations relate the properties of the sorbent material and the experimental conditions to the breakthrough curve of the contaminant from the bed using two parameters: the sorption rate parameter (k) and the saturation concentration of the contaminant on the solid phase (qmax). The rate parameter can be de-convoluted into three different resistances to H2S removal: (i) bulk phase mass transfer of H2S, (ii) intraparticle diffusion of H2S, and (iii) reaction/diffusion phenomena occurring at the surface of the reactive phase. These model parameters were regressed from the breakthrough curve data, while qmax was also determined from a mass balance of the contaminant in the vapor phase.
Fresh and spent nanostructures were analyzed via XRD and SEM to identify composition, crystallite agglomerate size, and crystallite size (using the Sherrer Equation). XRD patterns of fresh CuO nanostructures showed the characteristic peaks of tenorite (2q = 32.3°, 35.4°,38.5°, 48.5°, 53.1°, 57.9°, 61.1°, 67.5°), the monoclinic structure of CuO, for both nanofibers and nanoparticles regardless of precursor. Furthermore, nanofibers and nanoparticles consisted of CuO crystallites of similar size (16-18nm). SEM images revealed that the nanofibers exist as approximately rectangular solid shaped agglomerates of crystallites that are 250-300 nm is size and arranged in a one-dimensional stacking pattern. Nanoparticles also consist of 250-300 nm sized crystallite agglomerates, however, these agglomerates are packed into sphere like structures that are 5-10 microns in size.
The mechanism of H2S removal was probed by observing the effects of various process parameters (i.e., pressure, H2S concentration, contact time, and particle diameter) on H2S breakthrough curves from fixed beds of a commercially produced Cu-Zn-Al mixed metal oxide. Breakthrough curves that exhibit symmetrical S-shaped patterns indicate sorption rates that are modeled by an LDF equation that is first order in both gas phase H2S concentration and residual concentration of CuO in the solid phase. The rate parameters derived from this model exhibit contributions from fluid phase mass transfer and intraparticle diffusion. Sickle shaped curves result from LDF equations that are zero order in H2S but still first order in solid phase CuO concentration. LDF equations that are zero order in gas phase H2S concentrations indicate a sorption mechanism controlled by diffusion of S-atoms through CuS layers around unreacted CuO crystallites and agglomerates. For large granule sizes (550 micron average), the H2S removal was described by an equation that is first order in H2S at low linear velocities. As linear velocity was increased from 5.3 to 9.1 cm/s (holding contact time constant at 0.72 s), the removal of H2S became governed by a zero order (in H2S) LDF equation. This change in mechanism indicates that external and intraparticle mass transfer contribute to the overall rate of H2S sorption for large granules, likely because mass transfer contributions are proportional to the square of the granule radius. The higher linear velocities associated with the shorter contact times reduce external and internal mass transfer effects. Thus, the reaction/diffusion phenomena at the reactive surface becomes rate controlling. Under conditions where reaction/diffusion at the crystallite-agglomerate scale dominate (i.e., 150 micron granules), rate parameters were invariant with changing total pressure (8.6 x 10-5 1/s for P = 1-2 atm), however, rate parameters increased from 3.2 x 10-5 to 1.1 x 10-4 1/s as the partial pressure of H2S in the feed to the reactor increased from 620 to 2200 ppm-mol. This increase in rate parameter suggests that it is a lumped parameter that can be further de-convoluted into components from different elementary steps that are dependent upon H2S concentration.
To further probe the reaction between H2S and CuO at the agglomerate/crystallite level, H2S removal using CuO nanofibers and nanoparticles was compared. These nanostructures exhibited breakthrough curves indicative of LDF equations that are zero order in H2S. Nanofibers and nanoparticles exhibited similar maximum capacities (2.6 g H2S per 100 g CuO), however, sorption rate parameters for nanofibers (3.0 x 10-3 1/s) were 1.7 times larger than those for nanoparticles (1.7 x 10-3 1/s). This faster sorption rate is likely the result of the higher intra-agglomerate void fraction in the nanofibers compared to the nanoparticles. The one-dimensional arrangement of the nanofibers provides smaller and more easily accessible reactive domains compared to the nanoparticles.
These studies have confirmed that linear driving force models can quantitatively describe the kinetics and mechanisms for the removal of H2S from gaseous streams. Furthermore, these studies have shown that the rate parameters in these models consist of contributions from mass transfer as well as reaction/diffusion at the sub-micron scale, with the critical reactive domain at the level of the arrangement of agglomerates of crystallites. The rate parameters have been shown to be lumped parameters that are dependent upon H2S concentration, and thus, these parameters must contain contributions from various elementary steps that are not zero order in H2S. This knowledge can be further exploited to develop detailed microkinetic models which can be used to improve the efficiency of existing sorbents and to provide a rational framework for designing new materials for reactive sorption.