Electrospun Mn₂O₃ web electrodes: Influence of fabrication parameters on electrochemical performance

As alternative energy sources become more critical, improving energy storage is an essential step toward a widespread implementation of clean and renewable energy.^1,2 Because of their uniquely high power and energy densities, supercapacitors have the ability to bridge the gap between traditional capacitors and secondary batteries.^3,4 Their inherently long cycling life and fast charging and discharging capabilities are attractive for applications including hybrid and electric vehicles, portable electronic devices, and large-scale energy storage for load leveling.

Supercapacitors store charge as electrostatic interactions between ions in the electrolyte and the electrode surface, so maximizing the available surface area as sites for interactions is one route to increasing theoretical capacities. Nanostructured materials possess high surface-area-to-volume ratios, making them attractive as electrode materials.^5,6 Charge transfer pathways are also shortened in nanostructures, improving electrical characteristics and allowing materials with poor bulk electrical properties to be used as electrodes.¹ Traditional fabrication methods for nanofiber- or nanoparticle-based electrodes involve mixing the active material with conductive agents and binders and affixing the slurry to the charge-collecting substrates. However, electrospinning directly onto charge collectors is a facile and direct method for producing nanofiber-based ceramic electrodes. Polymer fibers containing ceramic precursors may be electrospun onto conductive substrates and calcined to remove the polymer and convert the precursors to oxide.^7–9 After calcination, electrodes are simply cut from the bulk mat and used directly. Eliminating the need for binders and conductive agents, as well as simplifying processing methods, has the potential to decrease overall device weight and fabrication costs of supercapacitors.

Manganese oxides are materials of interest for supercapacitor electrodes due to their low cost, abundance, and environmental benignity.^5,10 Although possessing relatively low conductivities and poor bulk performance, Mn₂O₃ has exhibited promising electrochemical capacitance and stability when processed into nanoparticles,^10,11 nanofibers,^9,12 and thin films.¹³ The weaker energy storage performance of Mn₂O₃ compared with other phases of manganese oxide is offset by the ease of preparation: it is readily produced by direct calcination of precursors in air at relatively low temperatures. Manganese oxide web electrodes with variable composition have previously been examined as potential structures for use as supercapacitor electrodes.⁹ However, the effects of fabrication parameters on the ultimate electrochemical performance of freestanding Mn₂O₃ web electrodes have not been well explored to date. The concentrations of polymer and oxide precursor in the electrospinning solution during initial fabrication, duration of fiber deposition during electrospinning, and finally, calcination duration are examined herein for their effects on electrode capacitance. Phase, morphology, and electrochemical capacitance of the electrospun electrodes are characterized, and the effects of each response variable are determined. A numerical model fit to the collected capacitance data is used to compare the relative impact of each factor on electrode performance.

To examine the effects of fabrication parameters on electrospun web electrodes, a design matrix was prepared using a full factorial design of experiments (DOE) with three midpoints to check for curvature in the data (Table I).

The low and high conditions for polyvinyl pyrrolidinone (PVP) and manganese (II) acetate [Mn(ac)₂] were set at 8 and 20 wt. %; the electrospinning deposition time was varied between 10 and 30 min, and the calcination time was tested at a low value of 120 min and a high value of 240 min. The 16 standards evaluating low and high conditions were fabricated and tested in a random order to eliminate temporal effects, while the three midpoints (std 17, 18, and 19) were evaluated as the first, tenth, and nineteenth samples.

Electrospinning solutions were prepared from PVP (MW 1 300 000, Sigma-Aldrich) and manganese (II) acetate tetrahydrate [Mn(ac)₂⋅4H₂O, ≥99%, Sigma-Aldrich] in a 1:1 volume ratio of deionized water:ethanol (≥99.8%, Sigma-Aldrich) solvent. All solution components were used as-received without further purification. Under stirring, PVP was added to the solvent solution and mixed at room temperature until fully dissolved, 2–3 h. Mn(ac)₂⋅4H₂O was then added to the solution and stirred for approximately 1 h, until the solution was fully homogeneous. All solutions were used the same day as prepared due to a rapid degradation during storage.

Fibers were electrospun in a Spraybase vertical electrospinner equipped with a 22-gauge needle at high voltage and a grounded collection plate. The solution flow rate was fixed at 0.20 ml/h, and the voltage was maintained at 6.5 kV with a separation distance of 4.5 cm between the needle and the collector. Fibers were electrospun onto 9-cm-diameter circular aluminum foil substrates. Manganese oxide films were also spin cast for comparison to the electrospun structures. To achieve similar areal masses to the electrospun electrodes, the solutions were dynamically deposited onto 9-cm aluminum foil substrates at 10 000 rpm and spun for 60 s. Deposition volumes were the same as electrospun deposition volumes based on the electrospinning time and solution flow rate—100 μl for 30 min deposition times and 33 μl for 10 min deposition times.

Electrospun fiber mats and spin cast films were calcined in a 2-in.-diameter tube furnace (MTI Corp, CA, USA) in air to remove the polymer from the structures and convert Mn(ac)₂ to manganese oxide. The furnace was ramped to 600 °C at a rate of 10 °C/min and maintained at temperature for 120, 180, or 240 min before cooling to room temperature.

Calcined samples were tested in a standard configuration three-electrode electrochemical cell with Ag/AgCl reference electrodes and Pt wire counter electrodes and 1.0M Na₂SO₄ electrolyte. A Keithley 2450 SourceMeter was used as the potentiostat. Circular samples were cut from the calcined fiber mats and placed into a circular foil holder with a defined area of 0.421 cm². Cyclic voltammograms were recorded between −0.3 and +0.3 V at scan rates of 25, 50, 100, and 200 mV/s. Three samples were tested for each standard, and the average of the calculated capacitances was taken as the areal capacitance of the standard.

To prepare samples for powder x-ray diffraction (XRD), PVP and Mn(ac)₂⋅4H₂O were mixed in a crucible in 1:1, 1:2.5, and 2.5:1 weight ratios to replicate the weight ratios present in the electrospun structures. The powders were calcined for 120 min under the same conditions as the electrospun fibers and crushed to remove agglomerations. Diffractograms were collected with a Rigaku Ultima IV x-ray diffractometer from 20° to 80° at a scan rate of 2°/min. A Cu-Kα source was used with an Ni filter at 40 kV and 40 mA.

Scanning electron micrographs of the calcined electrodes were collected with a Tescan Mira3 FE-SEM operated at 15.00 kV. Micrographs of the as-electrospun composite fibers were taken with a Hitachi S4500 SEM at 20.00 kV. Samples were cut from the as-spun and calcined fiber mats, mounted onto aluminum stubs, and lightly gold coated prior to imaging. Fiber diameters were measured in imagej as the average of 15 measurements.

Powder XRD was used to confirm the phase of the electrospun oxide structures and to ensure that interactions between the carrier polymer and oxide precursor during calcination did not affect the resulting phase for the electrodes (Fig. 1).

Individual powder samples reflecting the weight ratios of polymer and acetate resulting from each of the electrospinning solutions were produced and analyzed. Powders with 1:1 weight ratios of PVP and Mn(ac)₂ represented the 8%/8% and 20%/20% electrospinning solutions, while 1:2.5 PVP: Mn(ac)₂ (or vice versa) weight ratios represented 8%/20% solutions. For all weight ratios, all XRD peaks were indexed to Mn₂O₃. Lack of variation between conditions indicates that the relative amounts of PVP and Mn(ac)₂ in the fibers do not affect the oxidation state of manganese in the final structure; and therefore, variations in the areal capacitance of the web electrodes are not due to chemical differences.

Fibers were electrospun directly onto aluminum foil charge-collecting substrates from four solution compositions of 8%/8%, 8%/20%, 20%/8%, and 20%/20% PVP/ Mn(ac)₂ for 10 and 30 min. As-spun fibers morphologies were examined prior to calcination (Fig. 2).

All electrospinning conditions result in discrete fibers, without evidence of electrospraying or excessive fiber melting upon or after deposition. Fibers prepared from 8% PVP/8% Mn(ac)₂ exhibit beading, which is not observed with other solution compositions. While the 8%/8% fibers have a smaller average diameter than those from other solution compositions, a significant difference in fiber diameter was not observed between fibers prepared with 8% PVP/20% Mn(ac)₂ and 20% PVP/20% Mn(ac)₂, suggesting that the concentration of oxide precursor in the fibers has a strong effect on the electrospinning process. With a 20% PVP/8% Mn(ac)₂ electrospinning solution, the resultant fibers are approximately twice as large as those prepared from the other solutions. Solution conductivity is primarily owing to Mn(ac)₂ content, so increasing the concentration of the precursor elevates the conductivity of the electrospinning solution. With 20% PVP, the solution is highly viscous, and with low Mn(ac)₂concentration, the applied electric field has a weaker stretching effect on the fibers, resulting in higher average diameter.

Each solution from the experimental design was electrospun for both 10 and 30 min deposition times, and samples from each of those conditions were calcined for 120 and 240 min prior to testing. Electrospinning solution composition had the greatest effect on final electrode morphologies (Fig. 3).

The polymer concentration in the electrospinning solution had the most consistent and visible effect on fiber morphology, with low-PVP solutions exhibiting a nanowrinkle morphology and high-PVP solutions maintaining a fused-fiber structure. The nanowrinkle structures observed in 8% PVP fibers likely result from the small initial fiber diameter previously discussed. Previous work has shown that during heating, Mn(ac)₂ is converted to Mn₂O₃ before PVP removal,¹⁴ so the oxide structure is already formed upon polymer burnout. Small-diameter electrospun fibers from the 8% PVP solution contain relatively smaller amounts of Mn(ac)₂, inhibiting stable oxide network formation as the polymer softens, melts, and eventually burns off. The mechanical weakness of the fine oxide network formed causes the fibers to collapse and melt together into a filmlike structure with fibrous wrinkles at the surface. Conversely, the thicker fibers deposited from 20% PVP solutions form more mechanically stable oxide networks that are capable of supporting themselves once the polymer structures are removed.

The concentration of Mn(ac)₂ in the electrospinning solution did not have a visible effect on the nanowrinkle structures but affected the porosity of the fiber-based structures (Fig. 4).

All samples exhibited some porosity [Fig. 4(a)], but the number of large pores was greatly increased for fiber-based structures prepared with 8% Mn(ac)₂ [Fig. 4(b)]. High porosity results from greater dispersion of Mn(ac)₂ within the electrospun composite fibers. With high Mn(ac)₂ concentration, the oxide particles that form during the initial stages of calcination are in sufficiently close proximity that densification at higher temperatures nearly eliminates major porosity in the structure. However, with low Mn(ac)₂ concentration, oxide-rich regions in the fibers are initially spaced distantly enough that densification cannot effectively eliminate pores over the same heating duration. While low Mn(ac)₂ content results in higher porosity, accompanied by a higher specific surface area, the total mass of oxide deposited is lower, so improvements in capacitance from the higher specific surface area will be mitigated by a decrease in active mass.

Deposition time does not have a visible effect on the morphologies of the calcined electrodes, and calcination time has only a small effect. For the 20% PVP/20% Mn(ac)₂ electrospinning solution, fiber definition within the structure decreases slightly with increasing calcination times (Fig. 5), but the trend is not visible in the 20% PVP/8% Mn(ac)₂ condition, which also exhibits a fused-fiber structure.

Somewhat less distinct fiber definition is evident in samples that were calcined for 240 min than those treated for shorter time periods. During calcination, the fiber structures always exhibit some degree of intermelting, resulting from sintering at the fiber junctions and subsequent self-diffusion of the oxide toward the lowest surface energy structure of a smooth surface. With increased calcination time, more diffusion and bulk structure densification causes further intermelting and loss of discrete fiber structures. With the 8% Mn(ac)₂ electrospinning solution, the trend is not observable as the initial fiber structure is already less defined than that exhibited with the 20% Mn(ac)₂ solution. Loss of definition may, in fact, be present but is too subtle to detect by visual inspection.

Cyclic voltammetry was used to measure the areal capacitance of the freestanding Mn₂O₃ web electrodes. Voltammograms were collected for each fabrication condition at scan rates between 25 and 200 mV/s (Fig. 6).

The web electrodes exhibit electrochemical traces about an inclined axis. The incline of the voltammograms results from the conductive behavior of the aluminum foil charge-collecting substrate: aqueous electrolyte wets the electrodes and substrate thoroughly, so the direct contact between the electrolyte and charge collector results in the addition of a conductive I–V curve to the capacitive component of the measured current. Relatively low mass loading onto the substrates decreases the scale of the capacitive voltammogram and gives the current collector a significant impact on the shape of the scan, which is characteristic of resistive capacitors. No redox peaks are evident in the forward or reverse scans, and voltammetric curves broadened with increasing scan rate.

The areal capacitance of the electrospun web electrodes was calculated according to the equation

where C is the areal capacitance (F/cm²); i and v are current (A) and voltage (V), respectively, as recorded by the potentiostat; μ is the scan rate (V/s); A is the exposed area (cm²) of the electrode, defined by the electrode holder; and ΔV is the voltage window (V) of the scan. Tabulation of the calculated areal capacitance for each sample at all scan rates is presented in Table II.

Design Expert 12 was used to fit a statistical model to the data and determine the significance and relative impact of PVP and Mn(ac)₂ concentrations, deposition time, and calcination time on the areal capacitance. To visually demonstrate the effect of each parameter on capacitance, datasets were collapsed to the parameter of interest by taking the average of all measurements encompassing the data point (e.g., for the effect of calcination time, the data point for each scan rate in the 120 min dataset represents the average of all standards calcined for 120 min; standards 1–8) (Fig. 7). The midpoint conditions were not tested in combination with the high and low limits from the DOE, so the results for those three standards could not be treated with the same methodology and are therefore not included in the graphical representation. However, the statistical models indicate that curvature is not significant within the range studied, so their exclusion does not affect the trends described by the model.

Increasing PVP concentration in the electrospinning solution slightly increases the areal capacitance. Polymer concentration controls the morphology of calcined structures, with fibers resulting from a high polymer concentration and nanowrinkles from low concentration. The relatively greater specific surface area of fiber-based morphologies, compared to nanowrinkle structures, increases storage capacity directly due to a greater number of storage sites available. However, the nanowrinkle morphologies still have rough surfaces with high surface areas, so the capacitance increase is minor compared to the magnitude of measured capacitance. Increasing concentration of Mn(ac)₂ in the electrospinning solution decreases the areal capacitance of the web electrodes. Despite the reduced active mass in electrodes prepared from solutions with low Mn(ac)₂ concentration, the high porosity observed in the structures has a greater effect on the available surface area than total oxide mass. The lower porosity and specific surface area resulting from increased precursor concentration causes decreased overall performance due to fewer active sites for electrolyte interaction. At the slowest scan speed, Mn(ac)₂ concentration has only a moderate effect on performance, with a 27% decrease in capacitance between the low and high Mn(ac)₂ concentrations, but at faster scan speeds concentration becomes the most significant factor examined. The shift in significance suggests that at slow scan speeds, the nanopores observed in all the fabricated electrode structures are accessible to the electrolyte ions and massively increase accessible surface area, regardless of whether the electrode is fiber or film-based. At faster scan speeds, however, diffusion kinetics limit access to the nanopores, and the larger pores observed in structures fabricated from 8% Mn(ac)₂ solutions become the dominant source of accessible storage sites. The additional surface area presented by the larger pores present is accessible at fast scan rates, whereas the universally observed nanopores are not.

As anticipated, increasing deposition time improves electrochemical performance, as the amount of oxide present for energy storage increases with longer deposition times. With a greater active mass, especially for fibrous structures, more sites are available for energy storage. Although the total mass of Mn₂O₃ deposited is approximately three times greater in samples deposited for 30 min than those deposited for only 10 min, the observed increase in capacitance is only 34% at 25 mV/s and decreases at faster scan rates. With longer deposition times, fiber intermelting and loss of discrete fiber structure prevent the theoretical threefold increase in the available surface area that could be expected with a threefold increase in fiber mass. Fiber layers beneath the surface of the electrode may not be as easily accessible, so the increase in mass would not correlate directly to the increased surface area. Additionally, increasing the thickness of the electrode negatively affects electronic conductivity through the structure, so the oxide layers furthest from the charge collector cannot store charge as effectively as those closest to the substrate. As a result, the areal capacitance does not increase proportionally to increasing active mass via longer deposition time.

Lengthening calcination duration also elevates the areal capacitance of the electrospun web electrodes. With extended time held at high temperature, densification of the ceramic structures may proceed more completely, and interior porosity is minimized. The densified ceramic webs provide more direct conductive pathways than structures with significant interior porosity acting as insulating regions. Shorter charge transfer pathways allow for more effective electronic conduction between the charge collector and electrolyte/electrode heterojunction at the surface of the oxide webs. Without the ability to effectively charge the oxide surfaces, the electrodes cannot attract and repel electrolyte ions, so optimized electrical conductivity through the structure is essential to electrochemical performance.

In addition to decreasing interior porosity, longer calcination times encourage grain growth in the ceramic, thereby increasing conductivity in the structures via reduction in grain boundary scattering.^15,16 At 25 mV/s, calcination time has the strongest effect on capacitance of the parameters tested, with an average 62% capacitance increase observed between the 120 and 240 min conditions. At faster scan rates, the effects of deposition time and calcination duration have similar magnitudes, which are stronger than the effect of PVP concentration but weaker than the effect of Mn(ac)₂ concentration.

Two-factor interactions were also considered based on their statistical significance as determined by the Design Expert model. Interactions between electrospinning duration and other parameters were found to be significant in that short deposition times depress the effects of the other examined variables. For the samples prepared with 10 min electrospinning stages, little to no difference was observed in capacitance resulting from changes to calcination time or Mn(ac)₂ concentration.

For short electrospinning times, a thin fiber mat is deposited, with a shorter distance for charge transfer to occur between the charge collector and top surface of the electrode. Because the charge transfer pathways are inherently shorter for the thin fiber mats, the effects of calcination time on structure conductivity become less significant and more difficult to detect. Similarly, the smaller mass of oxide in electrodes produced with short electrospinning times leads to a less noticeable increase in the measurable surface area with high-porosity electrodes produced from 8% Mn(ac)₂. Because the surface area increase from the porous structure is necessarily proportionate to the initial mass of fibers deposited, a larger improvement in capacitance is observed from a larger starting mass of fibers. Longer deposition times result in a greater mass of fibers and therefore more detectable changes relating to the other experimental variables.

Additionally, the interaction between deposition time and Mn(ac)₂ concentration in the electrospinning solution suggests that the major limiting factor in performance increase of the high-porosity structures fabricated from 8% Mn(ac)₂ solutions is primarily limited conductivity through the structure, rather than poor accessibility of the electrode surfaces with increasing thickness. If fiber layers were inaccessible to the electrolyte due to the increased electrode thickness from longer deposition time, the effect of Mn(ac)₂ concentration on areal capacitance would become less significant with longer deposition, as the proportionate surface area increase from high porosity would only be observable for the small fraction of active material on the top layers of the electrodes. Because the opposite is observed [Mn(ac)₂ concentration has a larger effect with longer deposition], it is inferred that the thicker electrodes from long electrospinning stages retain excellent electrotype permeability but possess poor electrical conductivities.

Comparison to planar Mn₂O₃ electrodes was performed in order to confirm that electrospun structures have higher capacitances than planar structures that have otherwise been treated identically. Electrospun structures were determined to exhibit universally higher areal capacitance compared to spin cast films prepared from the same solutions. While the maximum performance of 560 μF/cm² was observed with an 8% Mn(ac)₂/20% PVP solution, the film electrode fabricated from the same solution had an average areal capacitance of only 49 μF/cm². The film electrode made from 8% Mn(ac)₂/8% PVP fared only slightly better, with a capacitance of 82 μF/cm², as compared to 389 μF/cm² for its analogous web electrode. The improved performance of the low-PVP film is due to a more uniform film quality during spin casting and calcination, owing to the relatively lower viscosity of the solution allowing for a more uniform dispersion on the substrate. Although the effects of solution composition are not consistent with those observed in the electrospun structures, when the method of deposition is the only variable changed, electrospinning results in a remarkable capacitance increase over spin cast films.

Average areal capacitance values for the 19 standards tested range from 55 to 560 μF/cm², which is several orders of magnitude lower than the literature values reported for Mn₂O₃ thin films and nanofiber-based electrodes.^12,13 The performance is attributed to a relatively poor charge transfer through the charge collecting substrate to the electrode surfaces as well as low active mass. Mass loading in the freestanding electrodes is much lower than that observed with traditionally processed nanofiber and nanoparticle-based electrodes, leading to decreased capacitance from fewer active sites. Thin film electrodes have low active mass but superior electrical properties; as increasing active mass via deposition time was not observed to have a strong effect on areal capacitance, improving electrical properties of the electrospun web electrodes may provide a more effective route to optimization.

The maximum performance of the web electrodes fabricated within the scope of the study was from 20% PVP/8% Mn(ac)₂, 30 min deposition, and 240 min calcination (standard 14). The average capacitance for the set of conditions measured was 560 μF/cm² at 25 mV/s, which was 44% greater than the next-highest value of 389 μF/cm², documented for 8% PVP/8% Mn(ac)₂, 30 min deposition, and 240 min calcination (standard 13). The conditions used to produce the highest performing electrodes are consistent with the statistical evidence that polymer concentration has the weakest effect on performance, as all resulting morphologies demonstrated high surface roughness, while calcination time has the strongest effect. Because electronic properties are one of the most significant limiting effects for manganese oxide electrodes, the improved crystalline and physical structure resulting from longer calcination translates to significantly better performance. Further testing is necessary to determine the optimal calcination time, past which detrimental effects from surface area loss via smoothing and intermelting negate the positive electronic effects.

Freestanding web electrodes were fabricated via a facile electrospinning and direct calcination procedure, and the effects of several fabrication parameters on electrochemical storage performance were examined. Electrospinning solution composition, deposition time, and calcination duration were varied between low and high conditions, and a statistical model was fit to the effects of each parameter on areal capacitance. Increasing polymer concentration was found to result in improved capacitance due to the high surface area of the fibrous morphologies. The precursor concentration has a negative correlation with capacitance due to decreasing structure porosity with increasing Mn(ac)₂ content. The high-concentration samples exhibit an average decrease in capacitance of 27% as compared to those prepared with low precursor concentrations. Increasing deposition time and calcination time both have consistent positive effects on capacitance (34% and 62% improvements between the low and high values) from increased active mass and improved charge transfer, respectively. Within the range of conditions studied, the performance was maximized at 20% PVP, 8% Mn(ac)₂, 30 min deposition, and 240 min calcination, which resulted in a highly porous fiber-based electrode with greater active mass and improved electrical qualities. While current capacitance values are lower than high-performance Mn₂O₃ electrodes reported in the literature, further work into improving the electrical properties of the calcined oxide structures via controlled processing has the potential to improve performance to levels comparable to current nanostructure-based electrodes, providing a route to low-cost, high-performance supercapacitors.

The research was sponsored by the Combat Capabilities Development Command Army Research Laboratory and was accomplished under Cooperative Agreement No. W911NF-15-2-0020. The views and conclusions contained in this document are those of the authors and should not be interpreted as representing the official policies, either expressed or implied, of the Combat Capabilities Development Command Army Research Laboratory or the U.S. Government. The U.S. Government is authorized to reproduce and distribute reprints for Government purposes notwithstanding any copyright notation herein.