Crystallite Size Control of Prussian White Analogues for Nonaqueous Potassium-Ion Batteries

Nonaqueous potassium-ion batteries have emerged as possible lowcost alternatives to Li-ion batteries for large-scale energy storage, owing to their ability to use graphitic carbon as the negative electrode. Positive electrode materials remain a challenge. Here, we report control of the crystal dimensions of the Prussian white hexacyanoferrate (HCF), K1.7Fe[Fe(CN)6]0.9, using solution chemistry to obtain either nano, submicron, or micron crystallites. We observe a very strong effect of crystallite size on electrochemical behavior. The optimal cathode material comprised of 20 nm crystallites delivers a close-to-theoretical reversible capacity of 140 mAh g−1 with two well-defined plateaus at 4.0 and 3.2 V vs K/K upon discharge. Slightly inferior electrochemical behavior is observed for crystallites up to ∼160−200 nm in diameter, but unlike the analogous Na HCFs, micron-sized crystals show very limited capacity. For the nanosized crystallites, however, the energy density of ∼500 Wh kg−1 is comparable to that of the best Na HCF cathode materials. At a relatively high current density of 100 mA g−1, half-cells cycled with ethylene carbonate/diethyl carbonate (EC/DEC) and 5% fluoroethylene carbonate (FEC) demonstrate an initial discharge capacity of 120 mAh g−1 with a capacity retention of 85% after 100 cycles and 65% after 300 cycles. Much interest in alternatives to lithium-ion batteries has emerged recently, focused on potentially lower cost systems based on alkali and alkaline earth-ion batteries based on Na, K, Mg, and Zn. Much of the work has concerned sodium-ion batteries (NIBs), which exhibit advantages such as the low cost of sodium-based electrodes and electrolyte salts and good mobility of Na in solid-state lattices. While potassium-ion batteries (KIBs) have been actively investigated with aqueous electrolytes, reports of potassium intercalation batteries utilizing organic electrolytes are still rare. This is due to the low capacities and poor K-ion mobility of many candidate positive electrodes. New motivation for studying such materials arises from recent exciting advances at the negative electrode. In the last two years, highly reversible potassium (de)intercalation in graphite electrodes was achieved in nonaqueous electrochemical cells. This is in sharp contrast to sodium, which does not intercalate into graphitic carbon. Discharge results in the product KC8, corresponding to a capacity on the order of 300 mAh g−1 below 0.25 V vs K/K, showing that the practical aspects of LIBs could potentially be extended to KIBs. Hard carbons are also suitable candidates for low-potential K insertion. KIBs have a further advantage because the K/K couple has a lower standard potential compared to Na/Na (SHE −2.94 vs −2.71 V), which can significantly offset the lower gravimetric capacity of the heavier alkali ion. Inspired by the above, a few studies on cathodes for nonaqueous KIBs have rapidly emerged. Ji et al. reported reversible capacities of 130−200 mAh g−1 with both 3,4,9,10perylene-tetracarboxylicacid−dianhydride (PTCDA) and poly(anthraquinonyl sulfide) (PAQS) between 3.5 and 1.5 V vs K/ K+. The overall energy density is <250 Wh kg−1 for either cathode material. Passerini et al. studied the layered oxide cathode K0.3MnO2, which exhibited a discharge capacity of <70 mAh g−1 in a similar window of 1.5−3.5 V vs K/K and complex multiple discharge steps. Other electrode materials include KTi2(PO4)3, which exhibits a reversible capacity of 80 mAh g−1 with a low potential of ∼1.7 V vs K/K+. The above work proves that reversible insertion/extraction of K-ions is possible in different frameworks. Prussian blue analogues (also called hexacyanoferrates or HCFs) have attracted tremendous interest as hosts for large alkali ions such as Na and K. HCF cathodes such as NaxMFe(CN)6·yH2O (M = Fe, Mn, Co, and Ni) have been reported for nonaqueous NIBs, and an interesting dual ion configuration achieved with a K−Na liquid alloy anode showed promising results. Potassium HCF analogues, however, have been mostly utilized for aqueous batteries. The first work Received: February 28, 2017 Accepted: April 4, 2017 Published: April 4, 2017 Leter http://pubs.acs.org/journal/aelccp © 2017 American Chemical Society 1122 DOI: 10.1021/acsenergylett.7b00179 ACS Energy Lett. 2017, 2, 1122−1127 on a nonaqueous potassium-based HCF cathode was published in 2004 by Eftekhari et al., who studied potassium intercalation in a KFeFe(CN)6 thin-film electrode. Surprisingly, this work was not followed up upon until extremely recently. During the course of finishing this Letter, preliminary work appeared on K-batteries based on HCF materials. Zhang et al. reported a vacancy-ridden, hydrated HCF cathode with a very low potassium content (K0.22Fe[Fe(CN)6]0.805□0.195·4H2O) that requires prepotassiation; the full cell exhibited a capacity of 70 mAh g−1 with low initial Coulombic efficiency that increased to 90% over cycling. Even more recently, it was communicated that two manganese HCFsK1.9Mn[Fe(CN)6]0.92·0.75H2O and K1.7Mn[Fe(CN)6]0.9·1.1 H2Oexhibited excellent reversible capacities, after initial activation cycling, of about 100 mAh g−1 vs a Na−K alloy in KClO4/PC electrolyte with an average potential of 3.6 V. In the seminal report of a full cell (vs a graphitic anode), K1.75Mn[Fe(CN)6]0.93·0.16H2O has been reported to deliver a discharge capacity of 60 mAh g−1 over 60 cycles. A communication that surveyed the electrochemical behavior of fully hydrated HCF materials with approximate compositions K1.5M[Fe(CN)6]0.92· nH2O (M = Mn, Fe, Ni, Cu; n ≈ 2−3) shows that these Prussian white HCFs exhibit very different cycling performance as a function of M. The difference was tentatively attributed to many factors such as water content, crystal defects, and crystal integrity. The above materials were obtained by simple precipitation of particles, either from aqueous chlorides, from KCl solutions, or by a Na-ion-assisted method. Herein, we report for the first time the crystallization of K1.69Fe[Fe(CN)6]0.90·0.4H2O by a novel citrate chelation route, which provides excellent control of crystallize size and morphology. The factors that dictate good cell performance in nonaqueous KIBs were evaluated to gain insight into the correlation between particle size and electrochemical behavior. We find that KHCFs exhibit inferior electrochemical performance to their sodium NaHCFs with a similar crystallite size; nonetheless, a high energy density of 500 Wh kg−1 was obtained with an electrode comprised of HCF nanocrystallites. Long-term cycling was realized with a capacity retention of 65% over 300 cycles, with 98% Coulombic efficiency. This performance is equal to the best cathodes for both KIBs and NIBs. Our preliminary results on HCF cathodes also show that the reactivity of potassium metal with alkylcarbonates such as PC is an obstacle to the studies of K half-cells (even in the presence of fluoroethylene carbonate (FEC) as an electrolyte additive). Prussian white samples with either an ultrasmall (∼20 nm) or intermediate (∼170−200 nm) crystallite size (denoted as Figure 1. Physical (a−c) and electrochemical (d−f) characterization of nano-sized KFeHCF-S: (a) Local environments of monoclinic K2FeFe(CN)6 show the connectivity of FeCN6 (brown) and FeNC6 (gray) octahedra. The silver, brown, and purple balls represent nitrogen, carbon, and potassium, respectively. (b) XRD pattern and accompanying Rietveld refinement of KFeHCF-S. (c) HR-TEM images of KFeHCF-S; the average particle size is ∼20 nm. (d) Cyclic voltammogram (at 0.05 mV s−1). (e) Initial discharge/charge profile; the inset shows the dQ/dV plot derived from the profile. ACS Energy Letters Letter DOI: 10.1021/acsenergylett.7b00179 ACS Energy Lett. 2017, 2, 1122−1127 1123 KFeHCF-S and KFeHCF-M, respectively) were synthesized by a modified citrate-assisted coprecipitation method. The crystallite size was tuned by varying the amount of citrate, whereas micron-sized crystals (KFeHCF-L) were synthesized using K4Fe(CN)4 as both the Fe and K source (see the Supporting Information for details). On the basis of inductively coupled plasma (ICP) analysis coupled with thermogravimetric analysis (TGA) below 180 °C to determine water content (Figure S1, Table S1), the compositions of the three materials are defined as K1.69Fe[Fe(CN)6]0.90·0.4H2O (KFeHCF-S), K1.78Fe[Fe(CN)6]0.92·0.4H2O (KFeHCF-M), and K0.68Fe[Fe(CN)6]0.89·0.7H2O (KFeHCF-L). All samples showed a small weight loss of 2−3 wt % between 150 and 300 °C, indicating low levels of interstitial water in the lattice after drying under vacuum. K2FeFe(CN)6 is isostructural to monoclinic K2MnMn(CN)6. 40 Figure 1a illustrates that the iron is either bound to six carbon atoms of the CN ligands to form FeC6 octahedra, or bound to six nitrogen atoms of the CN ligands to form FeN6 octahedra. Potassium is located in the void space between the octahedra. The X-ray diffraction (XRD) pattern of the KFeHCF-S showing a pure phase is obtained (Figure 1b), indexed in the space group P21/n. The crystallite size is ∼25 nm as calculated from the Scherrer equation, consistent with the estimation of 20 nm (Figure 1c) from high-resolution transmission electron microscopy (HR-TEM) where the welldefined lattice fringes confirm that the nanoparticles are highly crystalline. The fundamentals of potassium insertion/extraction in KFeHCF-S were evaluated by cyclic voltammetry (CV) and galvanostatic charge−discharge protocols. The CV curves demonstrate two pairs of peaks at 3.6/3.1 and 4.1/3.7 V vs K/K, which correlate with the redox reactions of Fe/Fe couples in different coordination environments (Figure 1d). The low-potential peaks are derived from redox on the Fe site with a high-spin configuration that is coordinated to nitrogen, while the high-potential peaks are from the Fe site with a lowspin configuration that coordinates to carbon. The initial charge and discharge capacities are 132 and 140 mAh g−1 at 10 mA g−1, corresponding to 1.7 and 1.8 potassium ions per formula, respectively (i.e, theoretical values). The voltage− capacity profile demonstrates two distinct plateaus; the precise potentials obtained from the dQ/dV curve (Figure 1e inset) are 3.28 and 3.93 V vs K/K upon discharge. By comparison, an optimal Na2FeFe(CN)6 cathode shows a high capacity in a Naion cell of 160 mAh g−1 and similar two-step potential profiles at 3.00 and 3.30 V. While the sodium electrode has a higher capacity, the potassium electrode is superior in potential. As a result, the two electrodes actually offer a similar energy density of 500 Wh kg−1. Long-term cycling performance is discussed below. Ex situ XRD patterns of KFeHCF-S after charge (Figure 2a) and subsequent discharge (Figure 2b) show the expected structural transformation upon cycling based on the Na analogue. After full extraction of K from KFeHCF-S (“K2Fe [Fe(CN)6]”) to form Fe [Fe(CN)6], the lattice transforms to cubic. Upon reinsertion of K-ions, the cubic symmetry is reduced to monoclinic, and this process is highly reversible. The bulky K-ion has long been a concern for the development of KIBs. Its radius of 1.40 Å is 84 and 40% larger than that for Li (0.76 Å) and Na (1.00 Å), respectively. This does not only impact the volumetric energy of KIBs but also hinders the kinetics. To illustrate the latter point, two materials, KFeHCF-M and KFeHCF-L, with larger crystallite sizes were fabricated for comparison with the KFeHCF-S electrodes. KFeHCF-M was synthesized with the same coprecipitation strategy as KFeHCF-S but using five-fold more potassium citrate. Citrate is well-known to chelate to transition metal ions such as iron and thus slows the crystallization of K2FeFe(CN)6. The KFeHCF-M material exhibits a similar XRD pattern as KFeHCF-S (Figure S2), but the crystallites exhibit a more pronounced cubic morphology and a much larger size of ∼170−200 nm (Figure 3a,b). The reaction to produce the KFeHCF-L was fairly slow even at 60 °C, which allowed the cubes to grow to micron dimensions (Figure 3c). The KFeHCF-M electrode demonstrates a similar two-step voltage profile to KFeHCF-S upon discharge, but the capacity is slightly reduced to 125 mAh/g (Figure 3e), only 8% lower than that for the KFeHCF-S electrode (Figure 3d) at the same current density. Further increasing the particle size to the submicron scale, however, results in a drastically decreased capacity of 10 mAh g−1. The lower K-ion content of the KFeHCF-L sample may affect the performance of the electrode, at least on the first cycle. Nonetheless, typical sodium HCF cathodes with micron-scale particle size and similar Na-ion content to our material offer much higher capacities of >120 mAh g−1. The difference in performance shows that HCF cathodes suffer from more severe kinetic barriers in KIBs. Nonetheless, the high energy density obtained with the KFeHCF-S cathode is comparable with that of the best electrodes in NIBs, and it is higher than that of other metal ion batteries. The high reactivity between potassium metal and some alkyl carbonate electrolytes is another challenge in KIBs in half-cell configurations. Simple experiments showed that contact of potassium metal with ethylene carbonate/diethyl carbonate (EC/DEC) resulted in no change overnight (although a film Figure 2. (a) XRD patterns and full profile refinement of the KFeHCF-S electrodes. (b) Schematic illustration demonstrates the reversible structural transition from monoclinic K2FeFe(CN)6 to cubic FeFe(CN)6 upon potassium extraction/insertion. The purple, yellow, green, silver, and brown colors represent potassium, Fe (low-spin), Fe (high-spin), carbon, and nitrogen, respectively. ACS Energy Letters Letter DOI: 10.1021/acsenergylett.7b00179 ACS Energy Lett. 2017, 2, 1122−1127 1124 appeared after several days); however, electrolytes such as propylene carbonate (PC) or EC/DMC reacted rapidly over several hours to produce a distinct opaque film on the surface of the potassium. Figure 4a shows the voltage profiles of a KFeHCF-S half-cell run in EC/DEC with 0.5 M KPF6 on the 1st, 10th, 50th, and 80th cycles at 100 mA g−1. Even with this “best” carbonate solvent, the cell exhibits regular charge− discharge behavior during the first cycles, but the charge process becomes unstable over cycling and the cell exhibits a low average Coulombic efficiency of <90%. The orange/yellow color of the glass fiber separator shows evidence of electrolyte reactivity (Figure S3). The addition of 2 or 5% FEC (as commonly used for sodium metal anode cells) remarkably improved the Coulombic efficiencies of the cells to 98% (Figure 4b). The cell with 2% FEC still suffered from sudden capacity fluctuation after 170 cycles, while the cell with 5% FEC cycled up to 300 cycles (Figure 4c). Nevertheless, the latter shows a capacity drop at about the 30th and 210th cycles, and electrolyte decomposition was suggested by the orange color of the separator after cycling. Thus, FEC enhances the Coulombic efficiency and long-term cycling but it cannot completely prevent electrolyte side reactions as it is successively consumed upon cycling. Full cells (or different electrolytes) are necessary to explore long-term cycling to screen new potassium cathodes. A comparison between the KFeHCF-S electrode and three “optimal” sodium HCF cathodes is presented in Table 1. Among the four materials, the potassium cathode has a lower capacity, but the energy density is offset by the higher cell voltage. As a result, the cells are comparable to each other in terms of energy density. The sodium HCF cathodes still exhibit superiority in rate and cycling performance (Figures S4 and Figure 4c), which highlights the importance of structural optimization for potassium cathode materials. In summary, three KFeHCF materials with different crystallite sizes were synthesized by controlled chemistry routes and examined in potassium cells. While the electrochemical performance of these materials is much more sensitive to crystallite size than that of their sodium HCF counterparts (we estimate an upper limit for optimal performance on the order of 200−300 nm), a high energy density of 500 Wh kg−1 was realized with the electrode composed of nanocrystallites. In half-cell configurations, PC is reactive with potassium, but even with the less reactive EC/DEC electrolyte, the addition of FEC Figure 3. Correlation between particle size and electrochemical performance at a current density of 20 mA g−1. (a−c) TEM images and (d−f) galvanostatic discharge/charge with KFeHCF-S, KFeHCF-M, and KFeHCF-L, respectively; note the difference in the y-axis scale. HR-TEM confirms that the average KFeHCF-S and KFeHCF-M cubes are ∼20 and <200 nm, respectively, while the KFeHCF-L cubes are >1.5 μm. The specific capacity was calculated using the actual composition of each sample. ACS Energy Letters Letter DOI: 10.1021/acsenergylett.7b00179 ACS Energy Lett. 2017, 2, 1122−1127 1125 was crucial to enhance the Coulombic efficiency and long-term cycling capability. Our findings reveal that nanocrystallite particle dimensions for active potassium cathode materials are of utmost importance, even for relatively flexible solids such as the HCF. Thus, searches for future novel cathode materials, such as rigid oxides, are unlikely to be fruitful unless crystallites are controlled to small dimensions. ■ ASSOCIATED CONTENT *S Supporting Information The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsenergylett.7b00179. Experimental methods, thermogravimetric analysis of the electrode active materials, XRD patterns of KFeHCF-M and KFeHCF-L materials, electrochemical cycling in the absence of FEC, and rate performance data (PDF) ■ AUTHOR INFORMATION Corresponding Author *E-mail: [email protected].