Ion Jump Motion as the Background for Muon Diffusion in Battery Materials Research Using µ SR Ryosuke Kadono∗ arXiv:2511.21013v2 [cond-mat.mtrl-sci] 5 Feb 2026 Muon Science Laboratory, Institute of Materials Structure Science, High Energy Accelerator Research Organization (KEK), Tsukuba, Ibaraki 305-0801, Japan Numerical simulations of muon spin relaxation (µ SR) in ion diffusion were performed using the extended Kubo-Toyabe (KT) relaxation function GEA z (t) that incorporates an Edwards-Anderson type autocorrelation function for the jump motion of ions. The analysis of the generated µ SR spectra using the conventional KT function GKT z (t; ∆KT , νKT ) (mimicking the previous analysis procedure) suggest that the anomalous peak in the fluctuation rate νKT around a specific temperature T ∗ and associated decrease of the linewidth ∆KT above T ∗ , often observed in the previous µ SR studies on ion diffusion, originate from the sharp increase in the ion jump rate νi against that of the muon νµ with increasing temperature. This indicates that a more detailed reanalysis of the vintage data using GEA z (t) is useful for the proper evaluation of νi and νµ . Meanwhile, it also suggests that the µ SR results showing no such anomaly convey little information on ion diffusion. Muon spin rotation/relaxation (µ SR) has been used for investigating ion dynamics in various non-magnetic materials including those for batteries in the past fifteen years, where nuclear magnetic moments carried by ions served as a marker of their dynamics. Assuming that the muon itself remains immobile in the time scale of µ SR measurements (∼101 µ s), the magnitude and fluctuation frequency of random local fields B(t) from these magnetic moments are deduced by analyzing µ SR spectra under zero or longitudinal magnetic field (ZF/LF-µ SR) using the Kubo-Toyabe (KT) function, GKT z (t; ∆KT , νKT ) (with ∆KT and νKT being the static linewidth and its fluctuation rate, repectively) [1], and νKT is interpreted as the ion jump frequency νi [2–21]. However, a recently developed spin relaxation model based on the insight that the time correlation of B(t) associated with ion diffusion retains finite values even as t → ∞ has revealed that the relaxation function predicted in such cases exhibits behavior qualitatively different from that of the KT function [22]. Specifically, the situation is described by introducing an Edwards-Anderson type parameter Q to the autocorrelation function, C(t) = hB(0)B(t)i = (1 − Q) + Qe−νit , ∆2 /γµ2 (1) where h...i denotes the statistical average over the canonical ensemble, ∆2 = γµ2 |B(0)|2 is the static linewidth given by the nuclear magnetic moments of ions and their distances from a muon (with γµ = 2π × 135.53 MHz/T being the muon gyromagnetic ratio). Note that ion diffusion corresponds to Q < 1 as only a portion of the ions around muon undergo jumps, which is in contrast to the muon jump motion that causes the entire amplitude ∆ to fluctuate (Q = 1). While the conventional KT function (∆ = ∆KT , νi = νKT , Q = 1) exhibits the motional narrowing with increasing νi (i.e., GKT z (t; ∆KT , νKT ) ≃ 1 for νi ≫ ∆), the extended KT function GEA z (t; ∆, νi , Q) under Eq. (1) is predicted √ to approach a quasi-static limit with a reduced linewidth 1 − Q∆ for Q < 1 ∗ ryosuke.kadono@kek.jp [22]. This has raised questions about the validity of analyzing µ SR data on ion diffusion with the conventional KT function. Furthermore, it has been pointed out very recently that the activation energy Ei and prefactor νi0 of the Arrhenius equation, νi = νi0 exp(−Ei /kB T ), deduced from such analyses deviate significantly from the common values expected for ion diffusion in nearly all cases [23]. Under these circumstances, a detailed reexamination of the µ SR literature on ion diffusion has revealed striking anomalies: νKT exhibits the maximum at a specific temperature T ∗ , and this peak behavior is often accompanied by a significant reduction in ∆KT at higher temperatures (T & T ∗ ). Despite that such behaviors are clearly inconsistent with the monotonous increase of ion jumping frequency with temperature expected for the thermally activated diffusion, no explicit explanation for this discrepancy is found in the literature. On the contrary, it is argued in these studies that the activation energy of νKT for T < T ∗ corresponds to Ei . In this paper, we “simulate” the µ SR spectra for various νi and the muon jump frequency νµ using the extended Kubo-Toyabe function GEA z (t; ∆, νi , νµ , Q) over the temperature range 10–400 K, assuming the appropriate activation energies and prefactors for νi and νµ . Subsequently, by analyzing these simulated spectra via the χ 2 fitting with GKT z (t; ∆KT , νKT ) (that mimics the data analysis procedure in the previous literature), we demonstrate that the anomalous behavior of ∆KT and νKT can be semi-quantitatively reproduced. This indicates that such behavior stems from the sharp increase in νi around T ∗ , and that a more detailed reanalysis using GEA z (t) is useful to properly evaluate νi together with νµ . Conversely, it strongly suggests that µ SR results showing no such anomalies clearly fail to capture ion diffusion. Provided that both ions and muons exhibit diffusion by jump motions and that the correlation between them is negligible, the autocorrelation function may be given by the following equation [22, 24, 25], C(t) ≈ [(1 − Q) + Qe−νit ]e−νµ t , (2) where the term e−νµ t represents the fluctuation in the total amplitude of ∆ due to the muon jump motion, and Eq (2) is reduced to Eq (1) with νµ = 0. In Li/Na-ion battery materi- 2 als, which are of our particular interest, Q and 1 − Q are respectively presumed to represent √ the contribution from Li/Na nuclear magnetic moments √ ( Q∆) and that from other ions (e.g., Co ions in Lix CoO2 , 1 − Q∆). However, considering that Q is determined by ions involved in a single fluctuation event of B(t), it is the contribution of one Li/Na ion next to muon that dominates Q when νi is low: thus other Li/Na ions are likely to contribute to 1 − Q as well [23]. (This will be discussed later in more detail.) Based on Eq. (2), Monte Carlo simulations have been performed by Ito [26] to generate GEA z (t; ∆, νi , νµ , Q), where the local field upon ion jump event was given as a vector sum of two independent fields B1−Q and BQ (corresponding to the 1 − Q and Q components in Eq. (2)). These fields were randomly generated according to the Gaussian p probability distribution with the standard deviation of (1 − Q)∆/γµ and √ Q∆/γµ , respectively. The fluctuation due to ion jump motion was implemented in accordance with the strong-collision model by regenerating the BQ at the average rate of νi . For the muon jump event, both B1−Q and BQ components were regenerated at once with the average rate νµ . The time evolution of a muon spin under B(t) = B1−Q + BQ , initially oriented to the z direction, was calculated and its z-component was averaged for 106 muons to obtain GEA z (t) for ZF. For finite LFs, the calculation was performed for the vector sum of the local field and the LF along the z direction. Next, to simulate µ SR spectra presumedly observed in the previous literature, we calculated GEA z (t; ∆, νi , νµ , Q) up to t = 20 µ s at temperatures ranging from 10 to 400 K using νi and νµ given by the Arrhenius equations, νi = νi0 exp(−Ei /kB T ), νµ = νµ 0 exp(−Eµ /kB T ), (3) (4) with νi0 = 5 × 1012 s−1 , νµ 0 = 2 × 106 s−1 , Ei = 0.35 eV, and Eµ = 0.02 eV. Their temperature dependence in the relevant temperature region is shown in Fig. 1. Note that these parameter values are well within the ranges of those suggested by 6/7 Li-NMR and µ SR studies, and are also in line with the respective diffusion mechanisms expected in the relevant temperature range (T . ΘD , the Debye temperature): ion diffusion is dominated by the over-barrier jump motion with Ei ranging from 0.2 to 0.6 eV with νi0 determined by the Debye frequency kB ΘD ≈ 1012–1014 s−1 , whereas muon diffusion is dominated by the phonon-assisted quantum tunneling that requires Eµ and νµ 0 much smaller than those for ion diffusion [23]. For other parameters, ∆ was fixed at 0.3 MHz, close to the value known for the relevant materials, and Q was set to 0.8 so that the ratio of (1 − Q)∆2 to Q∆2 was 1:4. Treating these calculation results as simulated µ SR spectra, we performed χ 2 fitting analysis using GKT z (t) to derive ∆KT and νKT . Before comparing the above simulation with past experimental results, let us first examine the general characteristics of ∆KT and νKT observed for the case of νµ = 0. Figure 2(a) plots these parameters versus νi which were obtained by global fitting with GKT z (t) (assuming all parameters except magnetic field are common between ZF and LF) of the time spectra generated by GEA z (t) at ZF and LF = 1 mT for various FIG. 1. Temperature dependences of jump frequencies for ions (νi ) and muon (νµ ) assumed for the simulation of µ SR spectra. The main panel shows these in semi-log scale against inverse temperature, while inset shows a magnified view of the region relevant to µ SR in linear scale. νi values. As is generally expected, GEA z (t) exhibits decay of the 1/3 component (motional broadening) and subsequent decrease in linewidth (motional narrowing) with increasing νi . However, as νi further increases, the 1/3 component reappears, and the function converges toward a quasistatic KT function with a smaller linewidth [22]. Such behavior in the lineshape is quantitatively illustrated in Fig. 2(a). More specifically, ∆KT√decreases as νi increases beyond the region where √ νi > Q∆ ≈ 0.27 MHz, asymptotically approaching 1 − Q∆ = 0.134 MHz. Meanwhile, νKT increases with νi , exhibits a pronounced maximum at νi ∼ √ 3 Q∆ (which is slightly larger than where ∆KT begins to decay), and then decreases monotonically as νi increases. Namely, GEA z (t) is quasistatic at both ends of νi , exhibiting behavior significantly different from the KT function where spin relaxation is quenched by motional narrowing as νi increases. These characteristics of ∆KT and νKT provides a key distinction between ion and muon diffusion and becomes crucial when comparing the simulation with the experimental results below. Furthermore, as seen in Fig. 2(b), the χ 2 value obtained from the global fit is largely enhanced in the region of νi where ∆KT and νKT exhibit characteristic changes, confirming that the KT function fails to adequately reproduce the spin relaxation expected for ion diffusion. Some examples of global fitting results are shown in Figs. 2(d)–(f), where the fit becomes poor around the νi region where ∆KT exhibits a significant decay with increasing νi . In all µ SR studies of ion diffusion published to date, analysis has been routinely performed using the conventional KT function. Table I lists materials in which the maximum value of νKT is observed at a specific temperature T ∗ . The occurrence/absence of a relatively sharp decrease (kink) in ∆KT for T & T ∗ is also indicated. To the best of our knowledge, these 3 a) νi=0.12 MHz Q=0.8 a) d) b) νi=3.3 MHz Q = 0.8 e) c) νi=45 MHz νi=3.3 MHz b) f) FIG. 2. (a) Linewidth ∆KT and fluctuation frequency νKT , and (b) associated χ 2 versus ion jumping frequency νi (in logarithmic scale) obtained by fitting the time spectra (ZF and LF=1 mT) generated by GEA z (t) (with ∆ = 0.3 MHz, Q = 0.8) using the conventional KT function. (c) Q dependence of ∆KT and νKT at νi = 3.3 MHz. Examples of the simulated ZF and LF spectra (EA) and those obtained by curve fitting (KT) are shown for (d) νi = 0.12 MHz, (e) 3.3 MHz, and (f) 45 MHz. behaviors of νKT and ∆KT have not been addressed until now, despite contradicting the naive expectation that the jump rate of ions driven by over-barrier jump motion would keep increasing with temperature. Here, Fig. 3 shows the temperature dependence of ∆KT and νKT obtained by the χ 2 fits of the simulated µ SR spectra with νi and νµ given by Eqs. (3) and (4). The behavior of the two fitting parameters can be qualitatively understood as replacing Material T ∗ (K) Kink in ∆KT Ref. Li0.73 CoO2 270 Yes [2] Li0.53 CoO2 310 Yes Li0.98 Ni1.02 O2 310 Yes [3] Li1 FePO4 300 Yes Li0.9 FePO4 250 Yes [5] Li0.8 FePO4 250 Yes Lix FePO4 230–250 No [8] Li0.3 (Ni/Co/Mn)O2 430 – [9] Li4 Ti5 O12 400 No [10] LiC6 370 – [12] LiC12 300–350 Yes Li(Fe/Mn)PO4 260 Yes [13] NaCx 250 Yes [17] TABLE I. List of materials in which a peak νKT at a certain temperature T ∗ has been reported in µ SR studies of ion diffusion: also shown is the presence or absence of ∆KT reduction/kink for T > T ∗ (information not available for “–”). T* FIG. 3. (a) Linewidth ∆KT and (b) fluctuation frequency νKT versus temperature obtained by fitting the time spectra (ZF and LF=1 mT) generated by GEA z (t) (with ∆ = 0.3 MHz, Q = 0.8) using the conventional KT function. Dashed line in (a) is the best fit using Eq.(5), and those in (b) are νi and νµ given by Eqs. (3) and (4) (see the inset of Fig. 1). the horizontal axis in Fig. 2(a) from νi to temperature for 200– 400 K and adding the contribution of νµ . Thus the peak of νKT and √ associated decay in ∆KT is a hallmark of the crossover in the Q∆ sector from the broadening to the narrowing regime with increasing νi , as has been shown in Fig. 2(a). The decay in ∆KT√corresponds to the decrease in the mean linewidth from ∆ to 1 − Q∆ associated with the sharp increase in νi above T ∗. These two features are regarded as evidence for the occurrence of ion jumps, which become observable when the jump rate νi falls within the time window of the µ SR near T ∗ . In the materials listed in Table I, Lix CoO2 , Li0.98 Ni1.02 O2 , Lix FePO4 , Li(Fe/Mn)PO4, LiC12 , and NaCx exhibit these features simultaneously, suggesting for the contribution of ion diffusion around T ∗ . Conversely, in other cases where these two features are not observed, µ SR likely fails to capture ion diffusion and reflects only muon self-diffusion. According to the simulation, T ∗ is determined as the temperature satisfying √ νi ≈ 3 Q∆ (see Fig. 2(a)). Although it is difficult to estimate Ei or νi0 in Eq. (3) from this information alone, one may be able to have a rough estimate of νi at T = T ∗ . Meanwhile, it should be also emphasized that νKT is dominated by νµ except at temperatures around T ∗ , and that ion diffusion makes negligible contribution in the low temperature range. Considering that all results attributed to ion diffusion in 4 the literature listed in Table I are derived from νKT in the temperature region T . T ∗ , these are likely to correspond to νµ in reality. Considering the complex temperature dependence exhibited by νKT , it would be difficult to evaluate νi versus temperature from νKT , indicating the needs for reanalyses of data using GEA z (t). Concerning the temperature dependence p of ∆KT , it can be (1 − Q)∆ due to the regarded as the transition from ∆ to √ quenched contribution of the Q component by motional narrowing, and it may be approximated by the following twostate transition model [27],   Q  + (1 − Q) ∆2 ,  ∆2KT =  (5) 1 + N exp − kEB∆T where E∆ is the activation energy, and N is the density of states. As shown in Fig. 3(a), the curve fit using Eq. (5) reasonably reproduces the temperature dependence of ∆KT , yielding E∆ ≈ 0.4 eV (with Q ≈ 0.82 and N ≈ 1.2 × 108 ) which is in fair agreement with Ei = 0.35 eV. Thus, while it is difficult to evaluate Ei from νKT , it may be feasible to obtain a rough estimate of Ei from ∆KT versus temperature. It should be noted that within the scope of this simulation, ∆KT remains constant for T < T ∗ . Therefore, the decrease in ∆KT with increasing temperature in the corresponding temperature range reported in the previous literature is strongly suggested to be irrelevant with ion diffusion and instead a phenomenon due to other causes including muon diffusion (see below). The first example where the model based on Eq. (2) was applied is µ SR studies on the local ion dynamics of cation molecules in hybrid perovskites [22]. It should be recalled that in these materials too, the two aforementioned features in the temperature dependence of ∆KT and νKT were clearly observed [28–30], suggesting that these anomalies are hallmarks for the presence of ion dynamics in general. However, while the lack of such anomalies suggests insensitivity of µ SR to ion diffusion, whether Q ≈ 0 or νi ≈ 0 should be carefully distinguished in the interpretation of experimental results. As shown in Fig. 2(c), the extent of the increase in νKT near T ∗ is sharply suppressed as Q decreases. Similarly, √ the decrease of ∆KT with increasing νi is suppressed as 1 − Q approaches unity. Therefore, when Q is small, even when ion diffusion is actually occurring (νi > 0), it is expected that µ SR cannot detect the contribution from νi . Results from µ SR studies on Nax CoO2 [7, 14] have already demonstrated to be in such situation (Q < 0.25) [23, 31]. The same is likely true for other examples where no anomalies are observed in νKT and/or ∆KT . Now, let us discuss the magnitude of Q in the cases where the nuclear magnetic moments of ions other than those of Li/Na are negligibly small: this is presumed to be the case for Li0.98 Ni1.02 , Lix FePO4 , Li4 Ti5 O12 , and Li/NaCx in Table I. Even in such cases, the probability for a single fluctuation event of B(t) involving multiple ions around the muon is virtually negligible except at extremely high temperatures. Therefore, given that the magnitude of ∆ is dominated by nn n ions, its fluctuation amplitude associated with single ion jump is approximately evaluated as Q ≈ 1/n. (This is readily understood by considering that ∆2KT is determined by the sum of ∆2i for each magnetic dipolar field exerted from a single ion to yield ∆2KT ≈ n∆2i , where n = 4–6 in typical materials.) Moreover, provided that it takes m steps in ion jump motion to entirely reset B(t) exerted from the n ions (depending the details of jumping √ path, so that m ≥ n), the true decay rate of C(t) for the Q∆ component in Eq. (2) is reduced to νi /m [23]: this can be regarded as the increase of effective 1 − Q by e−νi Tµ /m − e−νi Tµ over the µ SR time window Tµ . These two factors serve to reduce the effective Q for the relatively slow ion jump (νi /m . ∆KT ), where the n − 1 ions contribute to the 1 − Q component. In fact, the observed value of νKT at T ∗ (which places the upper limit of νi ) is below ∼1 MHz in all materials listed in Table I. Finally, we discuss the effect of ion vacancies on muon diffusion. While muons are supposed to occupy both interstitial sites and vacancies randomly upon their implantation, they often exhibit diffusion-limited trapping (DLT) to vacancies because of a deeper potential well than that at the interstitial sites. Since ∆KT at the vacancies is considerably smaller than that at the interstitial site due to the increase in the distance rµ between the muon and the nn ions (∆KT ∝ 1/rµ3 ), ∆KT is expected to be reduced by the promotion of DLT with increasing νµ [22]. The gradual decrease of ∆KT below T ∗ observed in some battery materials [2, 3, 8, 10, 13, 17] can be explained by this scenario. Note that Eq. (2) is virtually unaffected by the DLT process, as it only serves to reduce ∆KT and νµ as mean values. To conclude, based on simulation data using the extended KT function GEA z (t; ∆, νi , νµ , Q) and the analysis of them by the conventional KT function GKT z (t; ∆KT , νKT ), it has been demonstrated that the two anomalous features, the peak of νKT around T ∗ and associated decrease of ∆KT , observed in the previous µ SR studies of ion diffusion can be explained by the sharp increase in νi against νµ . The characteristic temperature T ∗ corresponds to the boundary where the mode of √ relaxation for the Q∆ component (describing ion jump motion) exhibits a crossover from the motional broadening to the narrowing regime with increasing νi . The motional √ narrowing also leads to the decrease in ∆KT toward 1 − Q∆ for T > T ∗ . Therefore, when these two features are observed simultaneously in the behavior of ∆KT and νKT , νi may be properly evaluated by the reanalysis using GEA z (t). It has been also shown that, as a temporary approach, the activation energy for ion diffusion may be evaluated from the temperature dependence of ∆KT . On the other hand, it is highly likely that the previous results showing no such anomalies in νKT and/or ∆KT convey little information on ion diffusion. The development of the extended KT function has led to a deeper understanding of how ion dynamics are reflected in µ SR spectra. As a result, it has become strongly suggested that most µ SR results previously interpreted in terms of ion diffusion actually pertain to muon self-diffusion. However, this does not necessarily mean that previous efforts were made in vain. For example, oxides are important materials not only for battery applications but also in various fields such as semiconductors and dielectrics, where it is well known that trace 5 amount of hydrogen significantly influences the physical and functional properties of these materials [32]. Previous µ SR studies on battery materials may be valuable as they provide information on the role of such “hidden hydrogen” via muon as pseudo-hydrogen. It is strongly hoped that those involved will revisit their studies from this renewed perspective. The author thanks T. U. Ito and M. Hiraishi for helpful discussion. This work was partially supported by the MEXT Program: Data Creation and Utilization Type Material Research and Development Project (Grant No. JPMXP1122683430). [1] R. S. Hayano, Y. J. Uemura, J. Imazato, N. Nishida, T. Yamazaki, and R. Kubo, Phys. Rev. B 20, 850 (1979). [2] J. Sugiyama, K. Mukai, Y. Ikedo, H. Nozaki, M. Månsson, and I. Watanabe, Phys. Rev. Lett. 103, 147601 (2009). [3] J. Sugiyama, Y. Ikedo, K. Mukai, H. Nozaki, M. Månsson, O. Ofer, M. Harada, K. Kamazawa, Y. Miyake, J. H. Brewer, E. J. Ansaldo, K. H. Chow, I. Watanabe, and T. Ohzuku, Phys. Rev. B 82, 224412 (2010). [4] J. Sugiyama, H. Nozaki, M. Harada, K. Kamazawa, O. Ofer, M. Månsson, J. H. Brewer, E. J. Ansaldo, K. H. Chow, Y. Ikedo, Y. Miyake, K. Ohishi, I. Watanabe, G. Kobayashi, and R. Kanno, Phys. Rev. B 84, 054430 (2011). [5] P. J. Baker, I. Franke, F. L. Pratt, T. Lancaster, D. Prabhakaran, W. Hayes, and S. J. Blundell, Phys. Rev. B 84, 174403 (2011). [6] J. Sugiyama, H. Nozaki, M. Harada, K. Kamazawa, Y. Ikedo, Y. Miyake, O. Ofer, M. Månsson, E. J. Ansaldo, K. H. Chow, G. Kobayashi, and R. Kanno, Phys. Rev. B 85, 054111 (2012). [7] M. Månsson and J. Sugiyama, Phys. Scr. 88, 068509 (2013). [8] T. E. Ashton, J. V. Laveda, D. A. MacLaren, P. J. Baker, A. Porch, M. O. Jones, and S. A. Corr, J. Mater. Chem. A 2, 6238 (2014). [9] M. Månsson, H. Nozaki, J. M. Wikberg, K. Prša, Y. Sassa, M. Dahbi, K. Kamazawa, K. Sedlak, I. Watanabe, and J. Sugiyama, J. Phys.: Conf. Ser. 551, 012037 (2014). [10] J. Sugiyama, K. Mukai, H. Nozaki, M. Harada, and I. Umegaki, R & D Rev. Toyota CRDL 46, 29 (2015). [11] M. Amores, T. E. Ashton, P. J. Baker, E. J. Cussen, and S. A. Corr, J. Mater. Chem. A 4, 1729 (2016). [12] I. Umegaki, S. Kawauchi, H. Sawada, H. Nozaki, Y. Higuchi, K. Miwa, Y. Kondo, M. Månsson, M. Telling, F. C. Coomer, S. P. Cottrell, T. Sasaki, T. Kobayashi, and J. Sugiyama, Phys. Chem. Chem. Phys. 19, 19058 (2017). [13] J. V. Laveda, B. Johnston, G. W. Paterson, P. J. Baker, M. G. Tucker, H. Y. Playford, K. M. Ø. Jensen, S. J. L. Billinge, and S. A. Corr, J. Mater. Chem. A 6, 127 (2018). [14] J. Sugiyama, I. Umegaki, S. Takeshita, H. Sakurai, S. Nishimura, O. K. Forslund, E. Nocerino, N. Matsubara, M. Månsson, T. Nakano, I. Yamauchi, K. Ninomiya, M. K. Kubo, and K. Shimomura, Phys. Rev. B 102, 144431 (2020). [15] P. Benedek, O. K. Forslund, E. Nocerino, N. Yazdani, N. Matsubara, Y. Sassa, F. Jurànyi, M. Medarde, M. Telling, M. Månsson, and V. Wood, ACS Appl. Mater. Interfaces 12, 16243 (2020). [16] I. McClelland, B. Johnston, P. J. Baker, M. Amores, E. J. Cussen, and S. A. Corr, Annu. Rev. Mater. Res. 50, 371 (2020). [17] K. Ohishi, D. Igarashi, R. Tatara, S. Nishimura, A. Koda, S. Komaba, and J. Sugiyama, ACS Phys. Chem Au 2, 98 (2022). [18] K. Ohishi, D. Igarashi, R. Tatara, I. Umegaki, A. Koda, S. Komaba, and J. Sugiyama, ACS Appl. Energy Mater. 5, 12538 (2022). [19] I. Umegaki, K. Ohishi, T. Nakano, S. Nishimura, S. Takeshita, A. Koda, K. Ninomiya, M. K. Kubo, and J. Sugiyama, J. Phys. Chem. C 126, 10506 (2022). [20] K. Ohishi, D. Igarashi, R. Tatara, I. Umegaki, J. G. Nakamura, A. Koda, M. Månsson, S. Komaba, and J. Sugiyama, ACS Appl. Energy Mater. 6, 8111 (2023). [21] O. K. Forslund, C. Cavallo, J. Cedervall, J. Sugiyama, K. Ohishi, A. Koda, A. Latini, A. Matic, M. Månsson, and Y. Sassa, Carbon Energy , e70017 (2025). [22] T. U. Ito and R. Kadono, J. Phys. Soc. Jpn. 93, 044602 (2024). [23] T. U. Ito and R. Kadono, arXiv:2508.05968 (2025). [24] S. F. Edwards and P. W. Anderson, J. Phys. F: Metal Phys. 5, 965 (1975). [25] S. F. Edwards and P. W. Anderson, J. Phys. F: Metal Phys. 6, 1927 (1976). [26] https://github.com/tuito0/musrfitdynGssEALF/tree/main/dual_diffusion_model (accessed on November 2025). [27] S. F. J. Cox, J. L. Gavartin, J. S. Lord, S. P. Cottrell, J. M. Gil, H. V. Alberto, J. Piroto Duarte, R. C. Vilão, N. Ayres de Campos, D. J. Keeble, E. A. Davis, M. Charlton, and D. P. van der Werf, J. Phys.: Condens. Matt. 18, 1079 (2006). [28] A. Koda, H. Okabe, M. Hiraishi, R. Kadono, K. A. Dagnall, J. J. Choi, and S.-H. Lee, Proc. Nat. Acad. Sci. 119, e2115812119 (2022). [29] M. Hiraishi, A. Koda, H. Okabe, R. Kadono, K. A. Dagnall, J. J. Choi, and S.-H. Lee, J. Appl. Phys. 134, 055106 (2023). [30] K. Papadopoulos, O. K. Forslund, S. Cottrell, K. Yokoyama, P. K. Nayak, F. M. A. Noa, L. Öhrström, E. Nocerino, L. Börjesson, J. Sugiyama, M. Månsson, and Y. Sassa, Adv. Phys. Res. 3, 2300120 (2024). [31] T. U. Ito and R. Kadono, arXiv:2505.09147 (2025). [32] R. Kadono and H. Hosono, Adv. Phys. , DOI: 10.1080/00018732.2024.2413342 (2024).