Mixed Polyanion Glass Cathodes: Mixed Alkali Effect

In lithium-ion batteries, mixed polyanion glass cathodes have demonstrated high capacities (200–500 mAh/g) by undergoing conversion and intercalation reactions. Mixed polyanion glasses typically have the same fundamental issues as other conversion cathodes, i.e.: large hysteresis, capacity fade, and 1 st -cycle irreversible loss. A key advantage of glass cathodes is the ability to tailor their composition to optimize the desired physical properties and electrochemical performance. The strong dependence of glass physical properties (e.g., ionic diffusivity, electrical conductivity, and chemical durability) on the composition of alkali mixtures in a glass is well known and has been named the mixed alkali effect. The mixed alkali effect on battery electrochemical properties is reported here for the ﬁrst time. Depending on glass composition, the mixed alkali effect is shown to improve capacity retention during cycling (from 39% to 50% after 50 cycle test), to reduce the 1 st -cycle irreversible loss (from 41% to 22%), and improve the high power (500 mA/g) capacity (from 50% to 67% of slow discharge capacity). of alkali and mixed alkali content on glass cathodes have not been previously reported to our knowledge. Single and mixed alkali copper phosphate vanadate (CuPV) glasses were tested as Li-ion battery cathodes to separately determine the electrochemical effects of alkali content and the mixed alkali content. The alkali effects on glass cathode performance were:

Most researched cathode materials for Li ion batteries are crystalline. Glass (amorphous) cathode materials have received relatively little attention. 1 The most explored class of polyanion glass cathodes is vanadate glasses, which typically demonstrate 100-300 mAh/g capacities when not discharged below 1V (vs. Li). [2][3][4][5][6] Phosphate glass cathodes have also been demonstrated in thin film batteries with a LiPON electrolyte. 7 Typically, polyanion glass cathodes have substantial capacity fade during tens of cycles.
Mixed polyanion glasses, which contain mixtures of polyanions such as phosphate, borate, vanadate, and molybdate, are high-capacity cathode materials (200-500 mAh/g) that can undergo conversion and intercalation reactions. 8 Phosphovanadate glasses with Ag, Cu, Co, Fe, and Ni cations have previously been demonstrated to be electrochemically active. 9 Mixed polyanion glasses, however, typically have the same fundamental electrochemical problems as other conversion cathodes i.e.: large hysteresis (typical ∼55% cycling efficiency), capacity fade during 10's of cycles, and a large 1 st -cycle irreversible loss (∼25%). Nevertheless, a key advantage of glass cathodes is the ability to tailor the composition and thereby change the physical properties and electrochemical performance. For example, previous work has demonstrated the ability to improve the cycling efficiency (as high as 67% efficiency), capacity retention, and/or 1 st -cycle irreversible loss (as low as 6% loss) by adding lithium and borate to copper phosphovanadate glasses. 10 Aoyagi,et al. 11 demonstrated high capacity (∼300 mAh/g) and reasonably good capacity retention (66% after 100 cycles) from Li-Fe-P-V-O glass and glass-ceramic cathodes. Yamauchi, et al. 12 demonstrated a tin-based mixed polyanion anode, which cycles well for tens of cycles. Additionally, Suman, et al. 13 demonstrated cycling of a mixed polyanion glass-ceramic in a Na-ion battery.
In polyanion glasses, a mixture of alkali cations is well known to have a particularly strong effect on the physical properties of a glass, which is commonly referred to as the mixed alkali effect. For a fixed total alkali content, many glass physical properties often exhibit a maxima or minima as a function of an alkali fraction (e.g., Na fraction for a Na x -Li 1-x containing glass). 14 The mixed alkali effect is stronger when the size difference between the alkali cations is larger. The mixed alkali effect can be particularly strong for the case of electrical conductivity (often reduced by 2-6 orders of magnitude), the alkali diffusion coefficient (often reduced by 2-4 orders of magnitude), and chemical durability (alkali extraction reduced by 4-6 times). 14 Transition metal cations in a glass can affect the intensity of a mixed alkali affect; Na/K iron phosphate glasses have shown a negligible mixed alkali effect on conductivity, density, and dissolution * Electrochemical Society Member. z E-mail: kercherak@ornl.gov rate (<10% property change), 15 while Na/K phosphate glasses with zinc 16 or vanadium 17 have shown a significant mixed alkali effects on conductivity. As discussed here, we found that the mixed alkali effect also has a significant influence on the electrochemical properties of glass cathodes, i.e., an influence that is likely due to these key physical property changes. The mixed alkali effect on glass cathodes in lithium ion batteries would be perforce complex, because the alkali content of the glass changes during the charge and discharge processes. For example, a Na-Li glass would have a higher alkali content and a lower Na fraction during the discharge process. As the alkali mixture and total alkali content change, the physical properties of the glass would change dramatically during electrochemical cycling. This manuscript represents an initial examination of the electrochemical impact of a mixed alkali effect for Na, K, Rb, Cs, Na-Li, and K-Li bearing glass cathodes. All compositions studied herein were based on copper metaphosphate metavanadate (CuPV) glasses. CuPV glass compositions were chosen, because their conversion reactions occur at relatively high voltages (2.3-2.5 V discharge plateau) and our previous work has studied the effect of polyanion content in similar compositions. 10

Experimental
Glass processing.-The CuPV glasses characterized herein had different alkali contents where the formula can be written as A x B y Cu ( 1 2 PO 3 • 1 2 VO 3 ) 2+x+y where A & B are alkali cations. In the following text, the names of these glasses will be written in the form A x CuPV (if only one alkali is present) or as A x B y CuPV. A list of all of the characterized glasses is shown in Table I (compositions based on dry precursor weights). The glasses were produced by melting the glass precursors in air at 1225-1250 • C (except for Li Cu ( 1 2 PO 3 • 1 2 VO 3 ) 3 that was melted at 1050 • C) and then splat quenching the glass melt between cold copper plates. The copper plates were 1.5 cm thick and 15 cm in diameter and were chilled in liquid nitrogen before use.
Glass splats were typically about 5-8 cm diameter and approximately 1 mm thick. The precursors were: copper (II) pyrophosphate hydrate, copper (II) oxide, vanadium (V) oxide, ammonium dihydrogen phosphate, lithium dihydrogen phosphate, sodium carbonate, potassium carbonate, rubidium carbonate, and cesium carbonate. Glass powders were produced by crushing the glass splats with a Sepor jaw crusher, dry milling the crushed pieces using a Spex 8000M mixer mill with 10 mm zirconia media for 10 minutes, and further milling in ethanol using 3 mm zirconia media for 30 minutes. Powder X-ray diffraction (Cu-Kα source, 10-70 • 2θ) was used to characterize powder from each glass splat; all reported materials exhibited XRD patterns with broad humps and without any observed crystalline peaks (representative XRD patterns of glass powders shown in Figure 1).

Theo. Cu
Theo. V single valence Composition capacity (mAh/g) capacity (mAh/g) While this type of XRD pattern is consistent with a pure glass phase, XRD analysis could not detect the presence of small quantities (<5 wt%) of a crystalline phase.
Cell fabrication & testing.-Cathodes tapes were produced by casting slurries of glass powder, graphite (Timcal C-Nergy KS 6L), carbon black (Timcal Super C65), and polyvinylidene fluoride (Kynar Powerflex LBG) (6/2/1/1 by weight, respectively) in n-methyl-2pyrrolidone using a square-frame tape caster (178 or 203 μm gap) on a 23 μm depassivated aluminum current collector using an MTI Corporation automatic thick-film coater (MSK-AFA-III). The cathode tapes were vacuum dried overnight at ∼80 • C and calendered to a 50 μm gap with an MTI Corporation MR-100A electric precision width rolling press. Cathode disks were then produced with a 1.11 cm hole punch.
Coin cells (CR2032, 316L stainless steel, polypropylene gasket) were assembled inside an argon-atmosphere glove box using cathode disks, an electrolyte (1.2 M LiPF 6 in ethylene carbonate/dimethyl car-  Coin cells were electrochemically evaluated on a Maccor series 4000 battery tester. Cycle testing was performed on the coin cells from 4.5 to 1 V at 20 mA/g and 100 mA/g. All voltages in this study were measured vs. Li/Li + . During charging, the cells were charged to 4.5 V and then were held potentiostatically until the current dropped below 10 μA. Specific capacity, 1 st -cycle irreversible loss, and cycling efficiency are all reported based on 20 mA/g testing. Capacity retention is reported based on 100 mA/g testing. Because of the 1 st -cycle irreversible loss, the cycling efficiency was measured between the 2 nd discharge and 1 st recharge (discharge energy/recharge energy).

Results and Discussion
Single alkali CuPV glasses.-The lithium, sodium, and potassium contents of CuPV glasses were varied from zero to 1 alkali per Cu (i.e., A x Cu ( 1 2 PO 3 • 1 2 VO 3 ) 2+x ). Also, rubidium and cesium CuPV glasses (Rb 0.05 CuPV & Cs 0.05 CuPV) were produced to further explore the effect of alkali cation size. Alternatively, these alkali-bearing glasses can be considered as cation mixtures with a fixed polyanion content: (A x Cu 1-0.5x ) PO 3 VO 3 . The Li x CuPV glasses do not have a mixed alkali content during charge & discharge processes, but the other single alkali CuPV glasses do. Comparing Li x CuPV with the other single alkali CuPV glasses will differentiate the effect of alkali content from a mixed alkali effect.
The first discharge, first recharge, & second discharge results at 20 mA/g (∼C/20) and cycling results at 100 mA/g (∼C/4) for single alkali glasses are shown in Figure 2. The high-capacity electrochemical reactions below ∼2.5 V are a Cu-based conversion reaction and a vanadium-based intercalation reaction (occurring concurrently or sequentially). 9 The plateau voltages of the first discharge curves were similar for all Li x CuPV glasses (Figure 2a); the Li 1 CuPV glass only had a slightly lower plateau voltage (2.3 V versus 2.4 V). However, Na x CuPV & K x CuPV with x ≥ 0.5 had significantly lower plateau voltages (Na 1 CuPV 2.0 V, K 1 CuPV 1.75 V) for the first cycle. This lowered plateau voltage during discharge was likely due to a lower ionic diffusivity and/or lower electronic conductivity in the discharging glass phase associated with a mixed alkali effect. Consistent with these results, the diffusivity/conductivity reduction of a mixed alkali effect would be expected to be more pronounced for increasing amounts of non-Li cations and for larger non-Li cations. The charging curves showed a distinctly different trend with alkali content. The CuPV glass showed a pronounced peak in the charge curve. Additions of small amounts of alkali (x = 0.05) to CuPV greatly reduced or eliminated the peak in the charge curve. Additions of large amounts of alkali (x = 1) appeared to increase the voltage of the charge peak above 4.5 V, which greatly increased the 1 st -cycle irreversible loss.
Some electrochemical metrics were primarily affected by the alkali content. Small alkali additions (A 0.05 ) improved the capacity retention of CuPV glasses, but only caused slight improvements in the 1 st -cycle irreversible loss and cycling efficiency (Figure 3). Large alkali additions to CuPV glasses caused higher 1 st -cycle irreversible loss (as high as 44% for K 1 CuPV) and higher cycling efficiency after the 1 st cycle (above 60% for all A 1 CuPV glasses). Increasing the alkali content does decrease the specific capacity, but not beyond what would be expected based solely on composition. If the composition is written in terms of constant polyanion content ((A x Cu 1-0.5x ) PO 3 VO 3 ), then it is readily apparent that alkali addition decreases the Cu available for a conversion reaction. If a full theoretical capacity of the Cu conversion reaction is assumed, then the vanadium-based intercalation reactions in these single alkali glasses would have capacities equivalent to 1.50-1.79 vanadium valence changes (based on Table I). A distinct mixed alkali effect could also be observed for the capacity retention of glass cathodes. For A 0.05 CuPV, the capacity retentions for the non-Li CuPV glasses (45% Na 0.05 CuPV, 50% K 0.05 CuPV, 46% Rb 0.05 CuPV, and 49% Cs 0.05 CuPV) exceeds the Li 0.05 CuPV value (43%) at 50 cycles. For A 1 CuPV, the capacity retention for the non-Li CuPV glasses (25% for K 1 CuPV) were lower than for Li 1 CuPV (41%).
Higher power performance also exhibited evidence of a mixed alkali effect (see Figure 4a). Glass cathodes were cycled at variable rates to determine their rate performance, starting at 20 mA/g (∼C/20) and progressing to higher currents (after test completion, a final 20 mA/g discharge was compared to the 2 nd 20 mA/g cycle to confirm nearly full capacity retention). The second 20mA/g cycle was compared to the first subsequent 500 mA/g cycle (∼1.3C), in order to eliminate the effect of 1 st -cycle irreversible loss. Li x CuPV, Rb 0.05 CuPV, and Cs 0.05 CuPV discharged at 500 mA/g all had capacities of 147-153 mAh/g (50% of their 20 mA/g capacities). In general, A 1 CuPV glasses had the worst high power capacities (e.g., 19-31% of their 20 mA/g capacities). For Na x CuPV & K x CuPV, the measured optimum high rate performances (63% & 67% of 20 mA/g values, respectively) occurred at different non-Li contents (50% Na & 5% K). The reason for different optimum compositions for best rate performance is unknown, but the alkali composition necessary to achieve a specific physical property value (such as ionic diffusivity) would be expected to be lower for larger alkali cation sizes based on the mixed alkali effect.
Mixing two alkali cations.-A series of CuPV glasses was made with a constant alkali content (Alkali-copper ratio = 1), but with mixtures of alkalis (i.e., A x Li 1-x Cu ( 1 2 PO 3 • 1 2 VO 3 ) 3 ). These mixed alkali glasses exhibited a mixed alkali effect "as made", but the alkali mixture would change during the charge and discharge processes. The first discharge, first recharge, & second discharge results at 20 mA/g and cycling results at 100 mA/g for mixed alkali glasses are shown in Figure 5. Note that all A x Li 1-x CuPV mixed alkali glasses (0 < x < 1) lacked the distinctive charge curve peak found in CuPV glass, but the 1 st -cycle irreversible loss was lower than Li 1 CuPV glass.
For the mixed alkali glasses, the most pronounced mixed alkali effects were observed for 1 st -cycle irreversible loss, the measured voltage for the conversion reaction, and high power performance. Mixtures of alkali cations consistently reduced the 1 st -cycle irreversible loss ( Figure 6), but the lowest irreversible loss was observed for Na 0.05 Li 0.95 CuPV (22%). The measured voltages for the conver-sion reaction in Na 0.05 Li 0.95 CuPV, K 0.05 Li 0.95 CuPV and Li 1 CuPV were very similar ( Figure 5), but glasses with a higher fraction of a non-Li alkali had progressively lower voltages in the first cycle (as low as 1.76 V for K 1 CuPV). The high power performance improved with 5% non-Li substitution (Figure 4), with the greatest effect observed for Na 0.05 Li 0.95 CuPV (its 500 mA/g capacity is 42% of its 20 mA/g capacity).

Conclusions
The mixed alkali content of a glass is known to strongly affect its physical properties, such as the electrical conductivity, alkali diffusion coefficient, and chemical durability. The electrochemical effects of alkali and mixed alkali content on glass cathodes have not been previously reported to our knowledge. Single and mixed alkali copper phosphate vanadate (CuPV) glasses were tested as Li-ion battery cathodes to separately determine the electrochemical effects of alkali content and the mixed alkali content.
The alkali effects on glass cathode performance were: r Increasing alkali content decreased the 1 st -cycle specific capacity due to a reduced amount of Cu available for a conversion reaction (e.g., 396 mAh/g for glass with 0 Li per Cu and 356 mAh/g for glass with 1 Li per Cu).
r Large alkali additions (1 alkali per Cu) caused a higher 1 st -cycle irreversible loss (up to 44% irr. loss) and a higher cycling efficiency (> 60% cycling efficiency).
The mixed alkali effects on glass cathode performance were: r Higher capacity retention after 50 cycles (up to 50%) occurred in glasses containing 0.05 non-Li alkali per Cu (compared to 39% for glass with no alkali). r Optimum non-Li alkali contents improved the high rate (500 mA/g) capacity from 50% to 67% of slow discharge capacity.