Sputter-Deposited Amorphous Li3PO4 Solid Electrolyte Films

This paper reports the thin-film synthesis of Li3PO4 solid electrolytes by RF magnetron sputtering. A relatively high ionic conductivity of more than 1 × 10–6 S cm–1 is achieved. It is revealed that the crystallization of Li3PO4 impedes ionic conduction, and a moderate amount of O2 addition to Ar suppresses the crystallization and guarantees long-term deposition. Another important finding in this study is that when Li3PO4 is deposited on a LiCoO2 film to construct a thin-film battery, the LiCoO2 film can be damaged depending on the substrate bias potential relative to the cathode potential propagated through the sputtering plasma. Active control of the bias potential to avoid the damage realizes negligible interface resistance in the thin-film battery.


■ INTRODUCTION
Solid-state Li-ion batteries are promising next-generation power supplies to replace current Li-ion batteries because of their superior features such as high energy density, long cycle life, and safety. However, it is reported that the interface between a cathode material and a solid electrolyte in a solidstate battery can be rate-determining and thus governs the power density. 1 Converting solid-state batteries into the thinfilm form is an effective way to investigate the interface properties since it simplifies the geometry and provides important information about the interfaces. 2−5 Lithium phosphorus oxynitride (LiPON) is widely used as a solid electrolyte layer in thin-film batteries because of its relatively high ionic conductivity (∼3 × 10 −6 S cm −1 ). LiPON was first developed by Bates et al. 6 by sputtering a Li 3 PO 4 target in pure N 2 . Their Li/LiPON/LiCoO 2 thin-film batteries operate for more than 30,000 cycles with a capacity fading of less than 5%. 2 Since then, a number of thin-film batteries with LiPON have been reported. 3,7,8 However, recent firstprinciples calculations indicate that LiPON is not thermodynamically stable, but kinetically stabilized, upon contact with Li metal and LiCoO 2 , 9 and the calculation results are consistent with experimental results. 10,11 Although partial replacement of O with N (and Li uptake) improves the ionic conductivity, and the conductivity reaches 6.4 × 10 −6 S cm −1 with simultaneous Li enrichment in the target, 12 the incorporation of N into Li 3 PO 4 narrows its electrochemical stability window according to the aforementioned calculations. 9 Li 3 PO 4 itself is also used as a solid electrolyte layer in thinfilm batteries. Bates et al. examined it along with LiPON by sputtering a Li 3 PO 4 target with 40% O 2 in Ar. However, the conductivities of their Li 3 PO 4 films were as low as 7 × 10 −8 S cm −1 ; 6 another group also reported similar values, 7 and in both of these studies, Li 3 PO 4 films were deposited by radiofrequency (RF) magnetron sputtering. Meanwhile, Li 3 PO 4 films prepared by pulsed laser deposition (PLD) using a high-photon-energy ArF excimer laser showed a relatively higher ionic conductivity of ∼5 × 10 −7 S cm −1 , 13,14 and thinfilm batteries made with the PLD Li 3 PO 4 operate rather well. 14−16 Here, we report the Li 3 PO 4 solid electrolyte film synthesis by RF magnetron sputtering with a much improved ionic conductivity. Although there are difficulties in LiPON synthesis in terms of controlling the amount of N incorporated and the simultaneous Li addition to achieve charge neutrality, Li 3 PO 4 synthesis is much simpler. We also report thin-film batteries constructed by depositing Li anodes and Li 3 PO 4 on LiCoO 2 epitaxial thin films.

■ RESULTS AND DISCUSSION
A schematic configuration of our specially designed RF magnetron sputtering system is shown in Figure 1. Since multiple sputter cathodes with 2″-diameter targets are equipped, each cathode is oriented to the center of a 2″diameter substrate holder with a 60°incident angle. The substrate holder is continuously rotated during deposition, and 10 mm square or 10 mm diameter substrates are located around the middle radius position of the 2″ inconel holder for simultaneous multiple deposition. Ar as well as O 2 gases can be introduced through mass flow controllers. The chamber is evacuated by a turbo molecular pump, and a conductancecontrollable gate valve is equipped between the chamber and the pump to adjust the chamber pressure independent of the gas flow rate. The substrate holder potential can be adjusted by a bipolar direct current (DC) power supply. Stainless steel and 0.5 wt % Nb-doped SrTiO 3 substrates were used, and they were electronically connected to the substrate holder during Li 3 PO 4 deposition. For the Nb:SrTiO 3 substrates, a piece of metal Mg was bridged to make a low-resistance connection with the inconel holder. The sputtering plasma potential around the substrate position can be measured by a substrate shutter. Besides, the cathode DC potential can be measured through a low-pass filter during RF sputtering. Improvement of Li 3 PO 4 Conductivity. It is obvious from Figure 2 that the crystallization of Li 3 PO 4 film drastically decreases the ionic conductivity. The figures show the substrate temperature (T sub ) dependence of the film deposition rate, ionic conductivity, and X-ray diffraction (XRD) patterns of 2 h-deposited Li 3 PO 4 films on mirror-polished stainless steel substrates. The 10 mm square and 0.5 mm-thick stainless steel substrate, which works as the bottom blocking electrode, was vacuum-annealed before use to remove the insulative oxidation layer on the surface. The deposition rate was evaluated by the film thickness measured with X-ray reflectance measurement, and ionic conductivity was estimated by alternating-current (AC) impedance measurements with 2 mm diameter Pt blocking electrodes deposited by DC magnetron sputtering. The AC impedance data were obtained in the frequency range of 5 × 10 5 −0.01 Hz with an AC amplitude of 20 mV, and ionic conductivity was estimated from the diameter of the semicircle at a higher-frequency region by fitting. The XRD patterns were measured by the surface-sensitive grazing-incidence method (GIXRD), where the X-ray incident angle is fixed at 0.25°and the intensity is recorded with 2θ scanning. Ar and O 2 gas flow rates were 20 and 5 sccm, respectively, and the total gas pressure was controlled to be 0.6 Pa during deposition. A RF power of 150 W was used, and 200 W data are also plotted in the left panel of Figure 2 for a comparison.
The deposition rate, i.e., the film thickness, is almost constant and independent of T sub , but the ionic conductivity is sensitive to T sub ; a higher T sub results in lower conductivity. According to the GIXRD results, when T sub is lower than 150°C , the Li 3 PO 4 film is in an amorphous state, showing a halo centered at 2θ = 23°. On the other hand, when T sub is higher, sharp diffraction peaks appear, which correspond to the Li 3 PO 4 crystal phase, and at 300°C, additional peaks at 2θ ≈ 14 and 28°appear, which are attributable to the Li 4 P 2 O 7 crystal phase. Because the T sub at the starting of crystallization coincides well with that during the conductivity drop, it is concluded that the crystallization of Li 3 PO 4 impedes ionic conduction, and an amorphous state is essential for high ionic conductivity. The activation energy estimated from the temperature dependence of ionic conductivity of the amorphous films in the range of 200−350 K was 0.53−0.55 eV. The obtained activation energy and frequency dependence of the impedance were similar to those reported for a PLD-deposited film under an O 2 atmosphere (0.58 eV), 15 but quite different from that of a RF-sputtered film deposited under pure Ar (0.38 eV), 7  suggesting the importance of O 2 introduction. Although low T sub is preferable to make the films amorphous, the substrate is heated up by sputtering plasma during much longer deposition processes, resulting in an unstable T sub in our deposition system, because it does not have a substrate cooler. Heating at moderate temperatures between 50 and 150°C is reliable to keep T sub constant throughout the deposition and to deposit amorphous films. Although the 200 W data show higher deposition rates (almost double), the higher cathode power tends to damage the target surface severely and quickly (e.g., by causing cracking and color change); thus, a lower RF power is preferred for long-term deposition. Figure 3 shows the T sub dependence of the mixing ratio of O 2 and Ar gases, in the same manner as in Figure 2, under a total pressure of 0.6 Pa, which was controlled by the conductance valve. According to the results of Figure 2, a T sub of 100°C is selected, and the room-temperature deposition is also examined without O 2 introduction. It is obvious that the deposition rate is higher when none or a small amount of O 2 is introduced. However, the ionic conductivity is low (less than 10 −6 S cm −1 ) when no O 2 is introduced. GIXRD results indicate that Li 3 PO 4 is crystallized clearly when the O 2 ratio is 1% or less and only slightly when it is 50%. The latter conditions seem similar to those examined by Bates et al.
reporting low conductivity; 6 thus, it can be concluded that O 2 is necessary to avoid crystallization; however, excess of O 2 also results in crystallization and decreases the conductivity. In addition, a thin film deposited at room temperature reveals the importance of O 2 introduction. The Bragg peaks indicating crystallization are observed for the thin film deposited without O 2 introduction, even though the film is deposited without substrate heating, and the film shows low ionic conductivity. It means that O 2 gas is anyway needed to suppress the crystallization of Li 3 PO 4 . Because sputtering is a vacuum process and the deposited film is an oxide, the film tends to become oxygen-deficient. In the field of thin-film growth of high T c superconducting and other functional oxides, it is wellknown that the melting point (T melt ) and crystallization temperature of oxide materials tend to be lowered when oxygen is deficient, and thus, high-crystallinity thin films of oxide materials with high T melt can be obtained via a vacuum process, 17 e.g., molecular beam epitaxy and PLD, under much lower T sub relative to their T melt values. Besides, oxygen deficiency is introduced not only in the deposited films but also in the sputtering target, resulting in serious target damage. O 2 gas introduction is therefore necessary to suppress film crystallization and to avoid target damage in the long-term deposition for a thick solid electrolyte layer. O 2 introduction  also makes the deposited films stable in air; otherwise, the transparent films devitrify after long-term storage in air, probably because of the humidity. Figure 4 shows the film thickness dependence of the deposition rate, ionic conductivity, and AC impedance Nyquist plots. Deposition conditions are as follows: T sub , 100°C; RF power, 150 W; Ar, 20 sccm; O 2 , 1 sccm; and total pressure, 0.6 Pa. When the thickness fringes were unclear in X-ray reflectance curves for films thicker than the typical 200 nm, the thickness was evaluated with a stylus profilometer, i.e., measuring the step height made by peeling off masking tapes. It appears that films with a thickness of 50 nm or more are needed to guarantee ionic conduction because all six Pt pads prepared on a 44 nm or thicker Li 3 PO 4 film were electronically open and showed similar conductivities, but on a 25 nm or thinner Li 3 PO 4 film, the conductivity was low and some Pt pads were short-circuited already. The constant deposition rate and ionic conductivity observed in the thickness range beyond 50 nm demonstrate that long-term and stable deposition is possible under the current conditions. For the preceding experiments, including this thickness dependence, the substrate potential was kept at +0.5 V by the DC power supply, and the target−substrate distance was fixed to be 150 mm during deposition.
Damage to the Underlying LiCoO 2 Film in Battery Devices. Li 3 PO 4 films were deposited on PLD-grown epitaxial LiCoO 2 thin films. The substrates were 0.5 wt % Nb-doped SrTiO 3 (111) single crystals with a 10 mm square or 10 mm diameter and 0.5 mm thickness, and LiCoO 2 grew in the c-axis orientation with a thickness of 100−200 nm. Details of the LiCoO 2 thin-film synthesis are described elsewhere. 18,19 Figure   5 shows the 2θ−ω scan XRD patterns of 15 h Li 3 PO 4deposited (2−2.5 μm thick) LiCoO 2 thin films under different substrate bias potentials during sputtering. Diffraction patterns before Li 3 PO 4 deposition are also shown in blue curves. Deposition conditions were as follows: T sub , 100°C; RF power, 120 W; Ar, 20 sccm; O 2 , 7 sccm; and total pressure, 0.6 Pa. Because the deposition rate of Li 3 PO 4 was stable under certain conditions, the deposited thickness was controlled by the deposition time. It is obvious that there is a clear substrate bias potential dependence of LiCoO 2 crystallinity after Li 3 PO 4 deposition. When the potential is lower than −3 V or higher than +0.5 V, LiCoO 2 diffraction peaks disappear or the intensity decreases drastically, i.e., the LiCoO 2 crystal lattice is destroyed. Besides, it seems that there is an optimal substrate potential, and −2.0 V is close to the optimal value in these depositions as the intensity decrease of LiCoO 2 diffraction is minimal. It should be noted here that the conductivity of Li 3 PO 4 films deposited on stainless steel substrates within this potential range is almost constant (≈1.2 × 10 −6 S cm −1 ), independent of the substrate bias potential.
It appears that an optimal substrate bias potential exists; however, it is not constant but varies gradually with each deposition. Figure 6 shows a plot of the cathode potential when the RF power is set at 100 W before each 5 h deposition. Other deposition conditions were as follows: Ar, 20 sccm; O 2 , 10 sccm; total pressure, 0.8 or 1.0 Pa; and the target−substrate distance was shortened to 95 mm from and after this experiment to approximately double the deposition rate. The target was changed between depositions #525 and #526. The target change alters the various deposition conditions, but the biggest change is in the DC cathode potential during RF sputtering. When the target is being worn out, the cathode potential increases and reaches around −250 V under these conditions, and finally, the rear indium bond and Cu backing plate appear around the eroding part of the Li 3 PO 4 target. In the case of a new target, the cathode potential is as low as around −400 V. Scattering of the cathode potential is mainly caused by the total pressure change. It should also be noted that the total pressure is related to the deposition rate as well as the shutter potential. When the pressure is increased in the range of 0.6−1.0 Pa, the deposition rate slightly decreases mainly due to stronger gas scattering, and the shutter potential increases and approaches 0 V. It should be noted that the cathode potential has a strong correlation to the shutter potential around the substrate position as shown in Figure 6 (top). Although the shutter potential is always negative, it changes and correlates to the cathode potential. It can be assumed that the optimal substrate bias potential that preserves the LiCoO 2 crystal quality after the Li 3 PO 4 deposition is related to the cathode potential, and it changes gradually with the change in the target surface state.
The change in the optimum bias potential can be seen in Figure 7. It shows 2θ−ω scan XRD patterns of two 15 h Li 3 PO 4 -deposited LiCoO 2 thin films under the same substrate bias potential of −0.5 V, but before and after the target change. Other deposition conditions were as follows: T sub , 100°C; RF power, 100 W; Ar, 20 sccm; O 2 , 10 sccm; and total pressure, 0.6 Pa. Between the two depositions, the Li 3 PO 4 target was changed for a new one because the target was worn out. The LiCoO 2 crystallinity of the top one is almost preserved, whereas that of the bottom one is degraded severely. It means that the optimal substrate bias potential was around −0.5 V, but it has shifted after the target change. Thin-film batteries made by ∼500 nm-thick Li anode deposition with vacuum thermal evaporation on Li 3 PO 4 films support the tendency; an open-circuit voltage (ocv) of the thin-film battery with the top sample just after cell construction was 3.9 V, whereas that of the bottom one was 4.3 V.
Here, we discuss what happens when the substrate bias potential is not optimal. Because Li 3 PO 4 and LiCoO 2 do not react with each other at substrate temperatures as low as 100°C , there must be other reasons that relate to the potential. In a battery, the LiCoO 2 cathode can be damaged by over-charging and over-discharging, with too high and too low cutoff voltages, respectively. In the sputtering process, the cathode DC potential always becomes negative to sputter the target material by positively ionized Ar gases. Even though it depends on the total pressure, the target−substrate distance, and the on-axis/off-axis geometry, 7 the target and substrate are connected to each other by plasma, which is an electronconductive gas; thus, the substrate holder is subjected to the cathode potential to some extent, as shown in Figure 6. Since LiCoO 2 is underneath the previously deposited Li 3 PO 4 , which is connected to the plasma, LiCoO 2 can be charged or discharged depending on the substrate bias potential relative to the plasma potential. In Figure 7 (bottom), the LiCoO 2 film appears to be overcharged as the X-ray diffraction intensity is considerably decreased, and additional reflections appearing at 2θ ≈ 21.6 and 44.0°are attributable to the 001 and 002 diffractions of the O1 phase, respectively, which is an overcharged phase of LiCoO 2 . 20,21 Besides, the ocv of the asconstructed battery was 4.3 V, which is higher than the standard charging cutoff voltage of 4.2 V. In Figure 5, the top film is also overcharged, but the bottom one seems overdischarged as LiCoO 2 cannot take up extra Li, 22    plasma, discharging (Li-ion infusion) takes place, probably due to the charge build-up of negative plasma by an electronically isolated substrate holder. The shutter potential can be a good reference to determine the optimal substrate bias potential; however, it is not stable enough for long-term Li 3 PO 4 sputtering, e.g., 10 h deposition; it can shift due to changes in the target surface state, e.g., sudden target cracking, oxygen deficiency introduction, and so on. Therefore, even if the substrate bias potential is determined once from the shutter potential before the deposition, the shift of the plasma potential deviates the substrate bias potential from the optimum value during deposition, which results in low experimental reproducibility. Here, we introduce a strategy to solve the reproducibility problem rather easily. During Li 3 PO 4 deposition, the substrate bias potential is adjusted in real time so that the current meter between the substrate holder and DC power supply (Figure 1) shows zero current. Ideally, no current flows when the substrate bias potential and potential induced from the plasma are balanced. With this active control, we successfully fabricated 10 mm-scale Li/ Li 3 PO 4 /LiCoO 2 thin-film batteries exhibiting high performance; some results have been published already, 21,27 and another example is described below. Figure 8 shows our thin-film battery structure and roomtemperature performances of the battery in which Li 3 PO 4 (ca. 1.1 μm thick) was deposited under the consideration of the above-described over-charging/discharging processes. Li 3 PO 4 deposition conditions were as follows: T sub , 50°C; RF power, 150 W; Ar, 20 sccm; O 2 , 10 sccm; total pressure, 0.8 Pa; and 5 h deposition. The increased total pressure was to reduce the absolute value of the plasma potential. The RF magnetron sputtering-grown LiCoO 2 was a (104)-oriented epitaxial film on a 10 mm square, 0.5 mm-thick 0.5 wt % Nb-doped SrTiO 3 (100) substrate, with a film weight of 561 μg (ca. 1.4 μm thick), which was measured with an electronic balance. Details of LiCoO 2 sputtering growth are reported elsewhere. Before the Li 3 PO 4 deposition, an ∼100 nm-thick Pt current collector was deposited on a LiCoO 2 film by DC magnetron sputtering, which is a 10 mm square film with a circular opening with a diameter of 10 mm, 21 and connected to the substrate holder electronically during Li 3 PO 4 deposition. Meanwhile, all of our LiCoO 2 films were air-exposed after synthesis for all purposes including weighing, XRD measurement, and Pt deposition. After Li 3 PO 4 deposition, a circular Li anode with a diameter of 8.5 mm was formed above the opening of the Pt current collector. 21 Although the LiCoO 2 cathode layer was relatively thick, the battery showed a rather high rate capability. Here, a charging/discharging rate of 1 C is defined as 137 mA g −1 , which is based on the expected capacity when LiCoO 2 is charged up to Li 0.5 CoO 2 at 4.2 V. 28 The thin-film battery was charged at a 1 C constant current up to a cutoff voltage of 4.2 V, followed by a constant-voltage charging at 4.2 V for 1 h before each discharge to guarantee a fully charged state. The capacity of low-rate discharge with 7.69 μA (0.1 C) was 108 mAh g −1 , whereas the capacity of high-rate discharge with 7.69 mA (100 C) was 60 mAh g −1 , maintaining >55% of the lowrate discharge. Figure 8 (bottom) shows a Nyquist plot at the 4.2 V charged state. The semicircle at the higher-frequency region originates from the resistance of Li 3 PO 4 , and its ionic conductivity calculated from the x-intercept of the semicircle at a lower frequency is 1.1 × 10 −6 S cm −1 . It should be noted that other resistances, e.g., the interface resistance between the Li anode and Li 3 PO 4 , or Li 3 PO 4 and the LiCoO 2 cathode, are not clearly observed, which is different from previously reported results. 7,14 Most recently, the reduced interface resistance was reported to be 10.3 Ω cm 2 between LiCoO 2 and RF-sputtered Li 3 PO 4 , 29 and also a further reduced resistance of 8.6 Ω cm 2 was reported between LiCoO 2 and LiPON. 7 As the active area of our battery was 0.567 cm 2 (8.5 mm-diameter Li anode), at least 15 Ω of interface resistance should appear in the Nyquist plots, if such a resistance exists; however, it cannot be seen in Figure 8 (bottom). The IR drop at 7.69 mA discharge is 1.37 V (Figure 8 middle) and thus R becomes 178 Ω. The resistance of Li 3 PO 4 in Figure 8 (bottom) is 179 Ω, which is almost the same value, i.e., no additional resistance exists, even though the LiCoO 2 film is air-exposed. It strongly suggests that the intrinsic interface resistance of Li/Li 3 PO 4 or Li 3 PO 4 /LiCoO 2 is much smaller than that reported and is almost negligible, and this can be achieved by the active control of the substrate potential during Li 3 PO 4 deposition. In other words, sputtering 7 Figure 8. Room-temperature performances of a Li (∼500 nm)/ Li 3 PO 4 (1.1 μm)/LiCoO 2 (561 μg) thin-film battery. Top: the battery device structure; middle: discharge curves for different C-rates; bottom: AC impedance Nyquist plot at 4.2 V charged state, where the top x-axis is areal resistance calculated with an 8.5 mm-diameter Li anode. AC impedance data were obtained in the frequency range of 5 × 10 5 −0.01 Hz, with an AC amplitude of 20 mV. Li 3 PO 4 deposition conditions were as follows: T sub , 50°C; RF power, 150 W; Ar, 20 sccm; O 2 , 10 sccm; total pressure, 0.8 Pa; and target−substrate distance, 95 mm. or bombardment of high-energy ablated species 14 during the Li 3 PO 4 deposition has been proposed to cause damages to the LiCoO 2 layer, which results in a large interface resistance, and eliminating these causes lowered the interface resistance to ca. 10 Ω cm 2 . However, the interface resistance has not reached the intrinsic value yet due to remaining damages under uncontrolled bias potential. In fact, we also observed such an interfacial resistance (≈10 Ω cm 2 ) when the bias potential control was not adequate, and the details are reported in a separate paper.

■ CONCLUSIONS
This study revealed that RF magnetron sputtering-deposited Li 3 PO 4 can have a relatively high ionic conductivity of more than 1 × 10 −6 S cm −1 , close to that of LiPON, by avoiding Li 3 PO 4 crystallization; the addition of a certain amount of O 2 gas into Ar as well as a low substrate temperature are effective to suppress the crystallization. Besides, this study points out the importance of substrate potential control during Li 3 PO 4 deposition on LiCoO 2 films, which has been a hidden, but predominant, process parameter for achieving high performance in thin-film batteries. This manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript and contributed equally.