Figures 1 (a) and (b) show the XRD patterns of (a) NH4(Mg1-xLix)F3-x and (b) (NH4)2(Mg1-xLix)F4-x. The most of XRD peaks could be indexed with the cubic (Pm\(\stackrel{-}{3}\)m) symmetry for NH4(Mg1-xLix)F3-x and the tetragonal symmetry (I4/mmm) for (NH4)2(Mg1-xLix)F4-x. In NH4(Mg1-xLix)F3-x, the diffraction peaks of the cubic phase gradually shifted to lower angle with increasing the Li content. This indicated that larger Li+ (0.76 Å) was substituted into the smaller Mg2+ (0.72 Å) sites. The lattice parameters of NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x were calculated from the diffraction angles and were plotted in Figs. 1 (c) and (d) as a function of the Li content. Except for the c-axis of (NH4)2(Mg1-xLix)F4-x, the lattice parameters changed monotonically with the Li content in NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x phases, suggesting that solid solution is formed at least within the compositional range of 0 < x < 0.3 in NH4(Mg1-xLix)F3-x and 0 < x < 0.2 in (NH4)2(Mg1-xLix)F4-x and the solubility limit of Li is higher than 30 mol% in NH4(Mg1-xLix)F3-x and 20 mol% in (NH4)2(Mg1-xLix)F4-x. Small diffraction peaks of NH4NO3 could be found in some compositions, especially in NH4(Mg0.8Li0.2)F2.8. In order to investigate the state and location of the impurity, the SEM observation and EPMA analysis were carried out. The results for NH4(Mg0.8Li0.2)F2.8 were presented in Figs. S1. The impurity, possibly NH4NO3, was observed as indicated by the yellow circles in Fig. S1. However, since the impurity particles seemed to exist sparsely from the main compound and their amount was not significant, the influences of the impurity on the observed ionic conductivities were supposed as negligibly small.
The SEM images of the cross sections of the pressed compacts of NH4(Mg0.8Li0.2)F2.8 and (NH4)2(Mg0.85Li0.15)F3.85 were shown in Fig. S2. The compacts seemed dense as just pressed compact, and the relative densities of all the compacts were approximately 75 %.
Figure S3 (a) and (b) show the results of TG measurement. NH4MgF3 and (NH4)2(Mg0.8Li0.2)F3.8 were stable below approximately 443 and 413 K, respectively. As shown in Figs. S3 (c) amd (d), XRD analysis indicated that NH4MgF3 was decomposed to MgF2 at around 443 K and (NH4)2MgF4 was decomposed into NH4MgF3 and MgF2 near 413 K forming NH4F gas.
Figures 2 show Nyquist plots observed with (a) NH4(Mg1-xLix)F3-x and (b) (NH4)2(Mg1-xLix)F4-x at 323 K. Although the results are not given in Figs. 2, only scattered signals were observed in EIS measurements with non-doped NH4MgF3, indicating its extremely low electrical conductivity. On the other hand, the Li-doped samples showed typical impedance responses of an ionic conductor with blocking electrodes, e.g. a semicircle in the high frequency region and a sharp spike in the low frequency region. These impedance behaviours suggested ionic conductivity in these samples. The resistance of the sample was determined from the semicircle in high frequency region. Figure 3 shows temperature dependences of the electrical conductivities of NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x. The conductivities were enhanced by Li-doping, but they showed the maximum and decreased with further increasing the Li content. At 323 K, the maximum conductivity was observed at x = 0.1 for NH4(Mg1-xLix)F3-x (8.4×10-6 S cm-1) and at x = 0.15 for (NH4)2(Mg1-xLix)F4-x (4.8×10-5 S cm-1). The decrease in electrical conductivity in highly doped samples is considered to be caused by cluster formation or ordering of fluoride ions and vacancies, and etc..25,26
In order to confirm dominant fluoride ion conduction in the investigated materials, we prepared a blocking cell consisting of Pb/PbSnF4/sample/PbSnF4/Pb. Since PbSnF4 is an almost pure fluoride ion conductor, this cell conducts only fluoride ion under steady-state DC bias, while the AC conductivity of the cell includes the contribution of all mobile carriers in the sample. Thus, if the conductivities measured by AC EIS and DC polarization methods are comparable, it can be concluded the dominant carrier is fluoride ion. The voltage transient curves observed in DC polarization measurements with a Pb/PbSnF4/samples/PbSnF4/Pb at various temperatures are shown in Figs. S4 (b)-(e) and S5 (b)-(h). The measured voltages were considerably increased immediately after the DC polarization and then gradually increased with time. From the impedance spectra shown in Figs. S4(a) and S5(a), the relaxation times for electrical conduction in NH4(Mg0.9Li0.1)F2.9 and (NH4)2(Mg0.95Li0.05)F3.95 were faster than 10-1 s. Thus, the gradual increase of the voltage might be mainly caused by the formation of resistive interphases by the decomposition of PbSnF4 at the PbSnF4/current-corrector interface. Therefore, the DC conductivity of the blocking cell was evaluated from the current and the voltage drop observed at 1 second after applying DC current. Figure 4 shows temperature dependence of conductivities of NH4(Mg0.9Li0.1)F2.9 and (NH4)2(Mg0.95Li0.05)F3.95 measured by AC EIS and DC polarization methods with a Pb/PbSnF4/sample/PbSnF4/Pb cell. The conductivities by AC EIS and DC polarization methods were comparable both NH4(Mg0.9Li0.1)F2.9 and (NH4)2(Mg0.95Li0.05)F3.95. Thus, it can be concluded that the dominant carrier was fluoride ion both in NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x.
In Fig. 3, the conductivities of conventional fluoride ion conductors are shown by dash-dotted lines3, 27–29. The fluorides investigated in this work exhibited relatively high ionic conductivity, although not as high as that of the best fluoride ion conductor, PbSnF4. It is also noteworthy that pressed compacts of NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x showed relatively high conductivities without sintering. This can be a great advantage for the fabrication of all-solid-state batteries. The activation energies of NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x were approximately 1.0 eV, as summarized in Table S1. The activation energies of NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x were slightly higher than those of the reported typical fluoride ion conductors.
In the case of the layered perovskite structure, interstitial anions sometimes can be mobile, as interstitial oxygens in Ln2NiO4+d (Ln = rare earth).30 Based on this idea, the introduction of interstitial fluoride ions was tried for the layered perovskite (NH4)2MgF4 by partially substituting trivalent cation Sc3+ for Mg2+. However, as shown in Fig. S6, this trial was not effective for improving the ionic conductivity of (NH4)2MgF4.
In order to demonstrate the influence of the molecular cations on the anionic conductivity, K(Mg0.9Li0.1)F2.9 having the same crystal structures was prepared. The lattice constant of K(Mg0.9Li0.1)F2.9 was 3.989 Å which was comparable with NH4(Mg0.9Li0.1)F2.9, 4.072 Å. The electrical conductivities of K(Mg0.9Li0.1)F2.9 and K2(Mg0.9Li0.1)F3.9 were considerably low, 5.2×10-6 S cm-1 at 789 K and 7.3×10-5 S cm-1 at 717 K, respectively (Fig. S7). Although the reason for the conductivity enhancement by the substitution of K+ for NH4+ is not clear at this moment, extension of the bottleneck for ion conduction, reduction of the interaction between the host and carrier ions, or assistance of the ion conduction by the rotation of the molecular ions might occur, as in fact cation conductors containing molecular anions. In this work, we succeeded to achieve relatively high fluoride ion conductivity in compounds containing molecular cations, NH4(Mg1-xLix)F3-x and (NH4)2(Mg1-xLix)F4-x, by introducing fluoride ion vacancies. The findings of this works suggested that compounds containing molecular cations can be new host materials for fast anion conductors.