Solid-state nuclear magnetic resonance for garnet-type based solid lithium electrolytes

INTRODUCTION

With the rapid growth in demand for mobile electronics, large-scale energy storage, etc., exploring battery systems with high energy density and enhanced safety has become a key focus. Lithium-ion batteries (LIBs) are popular and widely used rechargeable batteries among various energy storage technologies. Solid-state batteries with solid electrolytes (SEs) can effectively achieve high energy density while improving safety. Since 1960s, several solid lithium-ion electrolytes have been reported, including Li-β-alumina^[1], Li₃N^[2], LISICON-type (Li₁₄ZnGe₄O₁₆, etc.)^[3,4], NASICON-type [Li_1.3Ti_1.7Al_0.3(PO₄)₃, etc.]^[5], garnet-type [Li₇La₃Zr₂O₁₂ (LLZO), etc.]^[6], perovskite-type [(Li, La)TiO₃, etc.]^[7], LiPON-type [Li_2.88PO_3.86N_0.14, etc.]^[8], Li₇P₃S₁₁^[9], argyrodite Li₆PS₅X (X = Cl, Br, I)^[10] and thio-LISICON (Li_4-xGe_1-xP_xS₄)^[11], etc. Among these solid ion conductors, garnet-type electrolytes are the most sought-after, because of their high ionic conductivity, high chemical stability and wide electrochemical window^[12,13].

The general structure of various oxide-based garnets is A₃B₂C₃O₁₂, which was first reported by Menzer in 1928^[14]. Kasper reported the first lithium-based garnet-type solids Li₃Ln₃M₂O₁₂ (M = Te, W) in 1968^[15], which were not further employed as SEs because of low ion conductivity. The first garnet-type lithium SE (Garnet-SEs) was claimed in 2003 by Werner Weppner, with the composition Li₅La₃M₂O₁₂ (M = Nb, Ta)^[16]. Some Garnet-SEs were investigated by the replacement of La by alkaline earth ions in Li₅La₃M₂O₁₂, such as Li₆ALa₂Ta₂O₁₂ (A = Sr, Ba)^[17], to achieve higher conductivity. Meanwhile, researchers reported Garnet-SEs by replacing M-ion by Bi^5+[18], Zr^4+[6], Te^6+[19], etc. Among these Garnet-SEs, Zr⁴⁺ replaced Li₅La₃Ta₂O₁₂ to form LLZO, offering the highest total conductivity. Figure 1 shows the molecular structures and corresponding ion conductivities of Garnet-SEs^[20]. Investigation shows that pure LLZO is tetragonal (I41/acd) at room temperature (RT) with an ionic conductivity in an order of 10^-6 S cm^-1, while the high-temperature cubic phase (Ia-3d) shows an ionic conductivity ranging from 10^-4 to 10^-3 S cm^-1 at RT. The cubic phase can be easily obtained by super valent dopants Al³⁺, which is likely from the ceramic crucible^[21]. Ion doping was introduced to improve the conductivity, for example, doping ion at the Li sites (such as Al^3+[22-26], Ga^3+[27,28]), La site (such as Ce^3+[29,30]), Zr site (such as Ta^5+[31,32], Nb^5+[33], Te^6+[34]) and O site (such as F^-[35], Cl^-[36]) or co-doping in these sites^[37-39]. Among them, the highest conductivity was achieved as 7.81 × 10^-3 s cm^-1 at 30 °C for Ga-doped LLZO^[40,41].

Figure 1. A schematic comparing the ionic conductivity and crystal structure of Garnet-SEs. Copyright 2021 Springer^[20].

However, the application of Garnet-SEs still faces many challenges, such as low ionic conductivity at RT, the formation of dendrites, etc. Improving room-temperature conductivity is a critical goal for practical battery applications. The growth of dendrites may destroy the interface between the electrolyte and the electrode, resulting in interface instability, which in turn affects the cycle life and safety of the battery. The formation of dendrites will lead to a short circuit phenomenon, which may cause thermal runaway in serious cases, affecting the safety and performance of the battery. Therefore, controlling dendrite growth is the key to improving the performance of garnet SEs.

Understanding the structure-activity relationship of ion conductivity is very important for the design and optimization of electrolytes. X-ray diffraction (XRD)^[42] and neutron diffraction (ND)^[43] methods provide average structural information from a long-range perspective. The ionic conductive behavior from electrochemical impedance spectroscopy (EIS)^[44] mainly focuses on bulk materials and their interface. X-ray photoelectron spectroscopy (XPS)^[45] can be used to characterize surface elements and their chemical states, and can disintegrate bulk phase information if combined with in-depth analysis, but it is limited in the detection of structural and dynamic information inside the material. However, the local structure plays a vital role for determining the ionic diffusion/transport and coordination stabilization, together with a significant influence on ion migration pathways. Therefore, it is crucial to probe the local structure, ion migration pathway, ion dynamic and microstructure formation in Garnet-SEs at the atomic scale for an accurate understanding of the ion transport mechanism. A high-resolution transmission electron microscopy (TEM)^[46] is a tool for the local structure, but it is not sensitive to light elements, such as Li atoms in the electrolytes. Instead, solid-state nuclear magnetic resonance (NMR) offers access to the investigation of local environments and dynamics on a molecular level^[47-50], with the details about local symmetry, coordination numbers, and the dynamics of Li-ion mobility. The information provided by NMR can effectively analyze the local structure and dynamics of Garnet-SEs, with possible insights into the key limiting factors for lithium migration, which would, in turn, optimize Garnet-SEs design from an atomic or molecular view to improve their conductivity and stability. The NMR signal comes from the magnetic moment of atomic nuclei with a non-zero spin. When placed in a strong external magnetic field (B₀), these nuclear spins align either parallel or antiparallel to the field, creating a net magnetization moment along the B₀ direction. When a radiofrequency (RF) pulse, which equals the resonant frequency of the nuclei, is applied, the net magnetization moment is deflected so that it is no longer in the B₀ direction. When the RF pulse ends, the nuclear spins will precess around B₀, while the net magnetic moment deflects back to the B₀ direction. The precessing spins generate a detectable signal, the free induction decay (FID), which contains information about the local environment of the nuclei. The basic principle of NMR could be referred to our previous book chapter^[51]. With the help of ultrafast-spinning, high field and different pulse techniques, such as cross-polarization (CP) under magic angle spinning (CPMAS), two-dimensional (2D) nuclear overhauser effect spectroscopy (NOESY), heteronuclear correlation (HETCOR), exchange spectroscopy (EXSY), rotary echo double resonance (REDOR), multiple-quantum magic-angle spinning (MQMAS) and temperature-dependent relaxation, etc., the resolution of the spectra can be effectively improved for reliable component assignment, together with dynamics and migration behavior. Furthermore, NMR and imaging [magnetic resonance imaging (MRI)] with the aid of pulsed-field gradient (PFG) are non-deconstructive tools, which could be extended for in-situ/in-operando studies to track the evolution and lithium distribution inside electrolytes.

LOCAL STRUCTURE DETERMINATION

The atomic environment of Garnet-SEs and phase transformation were determined by the annealing process. Van Wüllen et al.^[55] investigated the local structure of Li₅La₃Nb₂O₁₂ from NMR perspective, with the distribution of octahedrally-coordinated (Oct-Li) and tetrahedrally-coordinated Li (Tet-Li) cations sensitively depended on the annealing temperature. The electrolyte obtained at higher temperature (900 °C compared to 850 °C, called GR-900 and GR-850) displays higher concentrations of Oct-Li cations (84% than 60%) and higher lithium mobility, as shown in Figure 2A and B. The Oct-Li cations were considered as mobile species, whereas the Tet-Li cations seem to act as traps for Li cations. Nyman et al.^[56] and Truong et al.^[57] reported that decreasing the total Li⁺ population in Li₅La₃M₂O₁₂ (M = Ta, Nb) electrolyte by taking advantage of the instability of lithium oxide or ion exchange for protons results in a decrease in the population of immobile Li⁺ (Tet-Li cations) population, which is quantified by well-resolved^[6,7] Li NMR spectra.

Figure 2. (A) ⁶Li-MAS-NMR spectra for GR-900, taken at T = 200 K and; (B) GR-850, taken at room temperature. Copyright 2007, Royal Society of Chemistry^[55]; (C) Experimental ²⁷Al MAS NMR spectra of cubic Al-containing LLZO at 14.1 T. Copyright 2011, American Chemical Society^[21]; (D) ⁷¹Ga MAS NMR spectra of Ga-doped LLZO. Copyright 2014, American Chemical Society^[62]; (E) ⁷¹Ga (black), ²⁷Al (blue) MAS NMR spectra and; (F) ²⁷Al 3QMAS NMR spectra of Al, Ga-doped LLZO recorded at 21.1 T. Copyright 2015, American Chemical Society^[64]. MAS: Magic angle spinning; NMR: nuclear magnetic resonance; LLZO: Li₇La₃Zr₂O₁₂.

In the subsequent studies, Geiger et al.^[21] found that the use of a ceramic crucible in initial synthesis steps introduces Al³⁺ into LLZO, which can obtain a stable cubic phase with higher ionic conductivity. Solid-state^[27]27Al magic angle spinning (MAS) NMR in Figure 2C show the octahedrally-coordinated Al cations, being assigned to LaAlO₃, and two tetrahedrally-coordinated Al cations at 68 and 81 ppm. Besides, the content of LaAlO₃ will decrease after the sintering step, indicating that LaAlO₃ appears to be a precursor for the incorporation of Al into LLZO^[58].

Manipulating Li ion sites and occupancy is vital to improve the conductivity of Garnet-SEs. Many works actively doped external elements into Garnet-SEs to regulate the occupancy of lithium. Kuhn et al.^[59] and Düvelet al.^[60] prepared cubic LLZO electrolytes with different Al contents. XRD and NMR results show that Al doping would form LaAlO₃, γ-LiAlO₂ and other Al environments in LLZO, indicating that Al atoms replaced Li, La, and Zr atoms. Rettenwander et al.^[61] calculated ²⁷Al NMR chemical shifts using density functional theory and compared them to the experimental spectra, with the result showing that Al³⁺ prefers the regular (24d) and distorted (96h) tetrahedrally coordinated sites.

Rettenwander et al.^[62] and Bernuy-Lopez et al.^[63] prepared cubic Ga-bearing LLZO, and the ⁷¹Ga NMR spectra plotted in Figure 2D show that Ga³⁺ mainly occurs in a distorted tetrahedrally coordinated site (96h). The site occupations of Ga and Al in garnets were probed with ²⁷Al and ⁷¹Ga MAS NMR obtained from ultrahigh magnetic fields as well^[64,65]. As shown in Figure 2E, ⁷¹Ga MAS NMR spectra present two ⁷¹Ga NMR resonances, corresponding to the 24d and 96h sites, respectively. The ²⁷Al signals can also be further assigned to the Al ions at the 24d and 96h sites, respectively. The second-order quadrupole interaction always depresses the NMR spectral resolution, bringing difficulties for accurate assignments. The MQMAS technique was usually employed to improve spectral resolution for these materials with intensive quadrupole interaction. Figure 2F shows the ²⁷Al 3QMAS NMR spectra recorded at 21.1 T with two ²⁷Al environments, indicating that the line broadening is not caused by distinct overlapping sites.

Previous studies have generally assumed that the two resonances occurring in the ²⁷Al and ⁷¹Ga NMR spectra correspond to the tetrahedrally coordinated 24d sites and distorted tetrahedrally coordinated 96h sites, respectively. Karasulu et al.^[66] combined ¹⁷O, ²⁷Al, and ⁷¹Ga MAS NMR spectra and density-functional theory (DFT) calculations [Figure 3A-D]. In their opinion, both of the resonances observed in the ²⁷Al and ⁷¹Ga NMR spectra are 24d (Li1) sites [rather than doping at other 96h (Li2) or La sites], resulting from deviations in the symmetry of the polyhedral coordination/site and changes in the number of occupied adjacent Li2 sites that share oxygen atoms with these dopant sites [Figure 3E]. The ¹⁷O MAS spectra [Figure 3D] were used for the first time for the Garnet-SEs system, showing distinct and well-resolved resonances that are due to oxygen bound to the dopants (Al or Ga).

Figure 3. (A) Crystal structure of cubic-LLZO. DFT/GIPAW; (B) ²⁷Al and; (C) ⁷¹Ga MAS NMR spectra of cubic Al_0.25Li_6.25LZO models and cubic Ga_0.25Li_6.25LZO models compared to the experimental spectrum of the Ga_0.2Al_0.2LLZO, Al_0.3LLZO and Ga_0.3Al_0.1LLZO samples; (D) DFT/GIPAW ¹⁷O NMR spectra of cubic LLZO, X_0.25Li_6.25LZO and X_0.5Li_5.5LZO models compared to the experimental spectrum of the Ga_0.2Al_0.2LLZO sample; (E) Local surroundings of the (left) Al and (right) Ga-doped sites in the selected single-doped defect configurations. Copyright 2020, American Chemical Society^[66]. LLZO: Li₇La₃Zr₂O₁₂; DFT: density-functional theory; GIPAW: gauge-including projector augmented wave; NMR: nuclear magnetic resonance.

In the follow-up research, Vema et al.^[67] demonstrated that Al and Ga dopants occupy only the 24d site in the LLZO lattice; the additional higher frequency resonance (~ 80 ppm for ²⁷Al and ~ 230 ppm for ⁷¹Ga) in ²⁷Al and ⁷¹Ga NMR spectra has been identified as being due to the side products, γ-LiAlO₂/LiGaO₂. As shown in Figure 4A and B, it can be seen that the signal of γ-LiAlO₂ matches the resonance in the LLZO sample well. The DQ-SQ spectrum of LLZO [Figure 4C] contains signals at ~ 10.5 ppm and ~ 79 ppm, indicating that these two sites are in Al-rich local environments, which can be assigned as LaAlO₃ and γ-LiAlO₂. The presence of these side-products can considerably affect the ionic conductivity of LLZO. The high-conductivity LLZO sample with a total conductivity of up to 1.75 mS cm^-1 was prepared by adjusting the synthesis process.

Figure 4. (A) ²⁷Al MAS NMR spectrum of the same Al-LLZO sample; (B) An enlargement of the same spectrum to show the resonances corresponding to Al in tetrahedral environments; (C) DQ-SQ 2D spectrum for the same Al-LLZO sample. Copyright 2023, American Chemical Society^[67]; (D) ⁶Li MAS NMR spectra of Li_7-2x-3yAl_yLa₃Zr_2-xW_xO₁₂ (0 ≤ x ≤ 1). Copyright 2015, American Chemical Society^[68]. MAS: Magic angle spinning; NMR: nuclear magnetic resonance; LLZO: Li₇La₃Zr₂O₁₂; DQ: Double quantum; SQ: single quantum.

Compared with doping elements, the sites and environment of lithium ions will directly determine the migration of lithium ions. Wang et al.^[13] synthesized and investigated Te-doped LLZO with and without Al doping. The ⁶Li NMR spectra show that the bulk conductivity of garnet-SEs is mainly determined by the mobility of Li ions at 24d sites. The resolution of the one-dimensional ⁷Li or ⁶Li spectra is still insufficient to resolve more specific Li sites in Garnet-SEs. Therefore, multiple dimensions or MQMAS NMR spectra were usually employed to gain deeper insights into the site assignment and correlation. Alternately, the saturation recovery method was used to probe the spinl-attice relaxation time (T₁) of Al and W co-doped Garnet-SEs^[68], from which two different T₁ values, 1.0-5.0 s and 14-28 s, were obtained, respectively. The difference in T₁ was used to suppress the strong signals from the octahedral sites. The additional signal at 1.6 ppm was assigned as distorted octahedral sites (96h). With the help of this relaxation filter, 24d, 96h, and 48g sites can be well-resolved for the first time; the ⁶Li MAS NMR spectra and simulation are shown in Figure 4D.

LI⁺ MIGRATION PATHWAY PROBED BY NMR

Insight into ion migration pathways can enable researchers to target specific structural or compositional modifications that can enhance ionic conductivity and battery performance. Since Li⁺ ions are light and small in radii, it is very challenging to experimentally probe their migration pathway by conventional methods, such as electron microscopy [transmission electron microscopy (TEM), scanning electron microscopy (SEM)]. Fortunately, NMR is isotope-specific and sensitive enough to detect the ^6,7Li signal. Therefore, EXSY NMR spectroscopy and element replacement from electrochemical cycling are popularly developed to unveil the Li transport pathway.

Wang et al.^[68] applied the 2D ⁶Li EXSY spectrum and found a cross-peak exists at the resonance of the 96h site (octahedral) and 24d site (tetrahedral) [Figure 5A], indicating that the Li⁺ ions can jump between 24d and 96h sites, resulting in the total Li⁺ pathway of 24d-96h-48g-96h-24d. Garnet-SEs are also introduced into polymers to yield composite electrolytes, which occupy the merits of the two with improved ionic conductivity and enhanced mechanical properties. Ranque et al.^[69], Ghorbanzade et al.^[70], and Ghorbanzade et al.^[71] applied 2D ⁷Li EXSY spectrum in the LLZO-poly(ethylene oxide) (PEO) system, showing an additional interface phase (such as LiOH) between the LLZO and PEO phases. The study reflected that the interface is involved in ion transport and has a positive effect.

Figure 5. (A) ⁶Li EXSY spectra of Li_7-2x-3yAl_yLa₃Zr_2-xW_xO₁₂ (0 ≤ x ≤ 1). Copyright 2015, American Chemical Society^[68]; (B) ⁶Li NMR of LiClO₄ in PEO, pure LLZO, and LLZO-PEO composite; (C) ⁶Li CP NMR spectrum and quantification for the LLZO-PEO electrolyte; (D) Illustration of the symmetric ⁶Li/composite electrolyte/⁶Li battery and possible Li⁺ transport pathways; (E) Comparison of the ⁶Li NMR spectra of the LLZO-PEO composite electrolytes before (pristine) and after (cycled) cycling; (F) Quantitative analysis of ⁶Li amount in LiClO₄, interface, and LLZO before and after cycling. Copyright 2016, Wiley-VCH^[72]. EXSY: Exchange spectroscopy; LLZO: Li₇La₃Zr₂O₁₂; PEO: poly(ethylene oxide); NMR: nuclear magnetic resonance.

Zheng et al.^[72] creatively used isotope ion tracer exchange to determine the ion pathways in LLZO-PEO solid composite electrolyte. The signals of LiClO₄ in PEO, ceramic LLZO and interface can be clearly distinguished by highly resolved ⁶Li NMR spectra in Figure 5B. The ¹H-⁶Li CP NMR spectrum is plotted and quantified in Figure 5C, from which the LLZO takes up 71% and the concentration of interface is up to 19%. Figure 5D shows the symmetric battery, in which LLZO-PEO (LiClO₄) was used as an electrolyte and ⁶Li metal was used as electrodes. ⁶Li ions from the electrodes will replace ⁷Li ions on their pathway inside the composite electrolyte during cycling. The ⁶Li NMR difference between the pristine and the cycled composite electrolyte shows significantly increased and slightly increased intensity for the LLZO and interfacial Li, respectively. In contrast, the signal for the polymer LiClO₄ remains almost the same. The change in ⁶Li signal indicates that Li ions tend to pass through the LLZO ceramic phase rather than the interface of a PEO-LLZO or PEO phase [Figure 5E and F]. Subsequently, this method has been adopted to many similar studies for the migration path determination^[73,74]. Yang et al.^[75] reported that Li⁺ preferentially migrated within the LLZO/polymer interface in LLZO-PAN composite electrolyte. Zheng et al.^[76] reported that Li ions are mainly transported via tetraethylene glycol dimethyl ether (TEGDME)-associated phases, not the LLZO or PEO when TEGDME is added to LLZO-PEO composite electrolyte as a small molecule additive. This method was also used by He et al.^[77] to explore the influence of interfacial ion transport barriers in composite electrolytes under different current densities on the ion transport pathways. Wu et al.^[78] showed there is an extra high-speed Li-ion pathway between poly (vinyl ethylene carbonate) polymer phase and Li_6.75La₃Zr_1.75Ta_0.25O₁₂ by the aid of the formed chemical bonds and hydrogen bonds, resulting in high conductivity. Zhang et al.^[79] conducted the Li-ion conduction mechanism of individual components within the LLZTO-PEO composite electrolyte, as shown in Figure 6. In organic-inorganic composite SE (OISE), polymer, LLZTO and Li-deficient layer three-phase synergistically promotes ion conduction, while lithium ions in OISE-L (the composite electrolyte containing Li₂CO₃ in LLZTO) are mainly conducted through the Li-deficient layer and the polymers, and Li₂CO₃ is involved in ionic conduction. Besides, at higher currents, lithium ions tend to pass through the polymer phase. In short, the isotope replacement from NMR perspective would offer a direct and reliable probe for the pathway of ion transport, which will in turn inform the design and optimization of ion conductors.

Figure 6. (A and B) ⁶Li NMR spectra of OISE-L after ⁶Li → ⁷Li tracer exchange at 10 and 30 µA currents, respectively; (C) quantitative analysis of ⁶Li amount in LLZTO, Li-deficient layer, Li₂CO₃ and LiClO₄ in PEO; (D and E) ⁶Li NMR spectra of OISE after ⁶Li → ⁷Li tracer exchange at 10 and 30 µA currents, respectively; (F) quantitative analysis of ⁶Li amount in LLZTO, Li-deficient layer, and LiClO₄ in PEO. Copyright 2024, Wiley-VCH^[79]. NMR: nuclear magnetic resonance; OISE: Organic-inorganic composite solid-state electrolyte; PEO: poly(ethylene oxide).

LI⁺ DYNAMICS FROM NMR

While ion transport properties in garnet-SEs are usually studied by EIS, NMR study will provide comprehensive time-domain dynamic information. Van Wüllen et al.^[55] probed the dynamics of lithium ions in Garnet-SEs for the first time using temperature-dependent ⁷Li NMR spectra in 2007, which was extended into a series of ^[80-82].

At low temperatures, the ⁷Li NMR central line is Gaussian-like and broad due to the dipole-dipole interactions. The full width at half maximum (FWHM) at this temperature is wide (rigid lattice line width), and the jump rates of Li⁺ are much smaller than 10³ s^-1. As the temperature rises, the FWHM narrows due to the increasing averaged dipole-dipole interactions, which is called motional narrowing (MN), while line shape changes from Gaussian to Lorentzian. At high temperatures, dipole-dipole interactions are completely averaged due to sufficiently fast Li diffusion.

Narayanan et al. applied temperature-variable ⁷Li NMR spectra of tetragonal LLZO, as shown in Figure 7A, showing typical MN. The empirical expression of Waugh and Fedin was used to obtain the activation energy (E_a), which is 0.47 eV, as given in

Figure 7. (A) ⁷Li NMR line widths δ of the central line of tetragonal LLZO versus temperature. The dashed line represents a fit according to Equation (2) with no restrictions for the adjustable parameters. The solid line shows the same fit, however, with E_a being fixed at 0.5 eV. ⁷Li NMR; (B) SLR rates and; (C) SSR rates of tetragonal LLZO recorded in both the laboratory (•) and the rotating frame of reference (◦). Copyright 2011, American Physical Society^[82]. NMR: Nuclear magnetic resonance; LLZO: Li₇La₃Zr₂O₁₂; SLR: spin-lattice relaxation; SSR: spin-spin relaxation.

where T_onset denotes the onset temperature of MN^[82].

The expression proposed by Abragam was also used to fit MN line (the dashed line in Figure 7A) and shows E_a = 0.43 eV, as given by

The diffusion-induced spin-lattice relaxation (SLR) NMR rate R₁ changes linearly with temperature, as follows:

⁷Li NMR SLR rates of LLZO [Figure 7B] were recorded and analyzed. The SLR rates reach their maximum at approximately T_max = 325 K and 530 K, respectively. The mean Li jump rates at these two temperatures can be calculated to be 1.9 × 10⁵ s^-1 and 4.9 × 10⁸ s^-1, respectively. The self-diffusion coefficient D_sd can be obtained using the Einstein-Smoluchowski equation D_sd = a²/(6τ) (3D diffusion), which is 1.3 × 10^-17 m² s^-1 (325 K) and 3.3 × 10^-14 m² s^-1 (530 K).

As shown in Figure 7B, there is an asymmetric rate peak R_1ρ, showing the slope of the low-temperature flank (T < 330 K) is lower than that of the high-temperature flank, and leading to E_{a low} = 0.32(3) eV < E_{a high} = 0.48(3) eV. In general, only a small number of Li⁺ jumps are sampled on the low-temperature flank of R_1ρ, indicating SLR was mainly caused by short-range Li motions in this temperature regime, while long-range Li diffusion is probed on the high-temperature flank. Meanwhile, spin-spin relaxation (SSR) NMR can also be used to extract the ion dynamic, and a similar E_a (0.50 eV) was obtained, as shown in Figure 7C.

To probe the Li⁺ dynamics in different phases with various doped ions, Buschmann et al. and Bottke et al. applied ⁷Li NMR spectroscopy to analyze Li-ion dynamics in Al-doped and Mo-doped cubic LLZO on different range length scales^[58,83]. The central transition and the satellite transitions of ⁷Li NMR lines of Mo-doped LLZO are shown in Figure 8A. Figure 8B and C shows the line width of the central transition of Al-doped and Mo-doped LLZO, respectively. On the whole, the rigid lattice regime of cubic LLZO appears at lower temperatures, indicating higher Li⁺ jump rate and fast Li-ion hopping processes than tetragonal LLZO [Figure 7A]. Interestingly, in contrast to Al-doped LLZO, the ⁷Li NMR line shapes of Mo-doped LLZO undergo a relatively homogeneous MN. The smooth transition from the Gaussian shape to the Lorentz line at high temperatures suggests that most Li ions are involved in the diffusion process even at lower temperatures.

Figure 8. (A) Variable temperature ⁷Li NMR spectra of Mo-doped LLZO. Copyright 2015, American Chemical Society^[83]; The ⁷Li NMR line width of the central transition (FWHM) of (B) Al-doped and; (C) Mo-doped LLZO as a function of temperature. Copyright 2011, Royal Society of Chemistry^[58]. ⁶Li NMR of LiClO₄ in PEO, pure LLZO, and LLZO-PEO composite; (D) Arrhenius plot of the ⁷Li SLR NMR rates T₁^-1 and T_1ρ^-1 of (D) Al-doped and; (E) Mo-doped LLZO. Copyright 2015, American Chemical Society^[83]. NMR: Nuclear magnetic resonance; PEO: poly(ethylene oxide); LLZO: Li₇La₃Zr₂O₁₂; FWHM: full width at half maximum; PEO: poly(ethylene oxide); SLR: spin-lattice relaxation.

The rigid lattice line width of the central line and satellite intensity [Figure 8A] are 9 kHz and 150 kHz, respectively. At 273 K, the intensity of the satellite dropped drastically and completely evaporated at 446 K. At 396 K, the mean jump rate has already increased to the order faster than 10⁶ s^-1. This indicates that averaging of quadrupole satellites occurs at relatively low temperatures, accompanied by the narrowing of the central line. This is sensitive to slow Li-ion diffusion, points to a subensemble of Li ions that are quite mobile on the time scale set by 150 kHz. All these indicate several diffusion processes are involved in Mo-doped LLZO.

⁷Li NMR SLR rates (T₁^-1) are also investigated. The E_a of Al and Mo-doped LLZO acquired by R₁ is 0.12 and 0.15 eV [Figure 8D and E], respectively. These are lower than the E_a of pure tetragonal LLZO. However, the E_a is smaller than that obtained by EIS as expected since E_a from NMR relaxation is likely induced by short-range motions of the Li ions and possible unsuccessful Li jumps raised by correlation effects.

The broad shape of rate peak T_1ρ^-1 has been interpreted as a large distribution of jump rates and activation energies, resulting in the superposition of many SLR rate peaks at different temperatures. For Mo-doped LLZO, it can be well-reproduced with a combination of two individual rate peaks (dashed lines in Figure 8E), reflecting two possible different, fast ion diffusion processes in Mo-doped LLZO.

Other NMR techniques, such as stimulated echo correlation spectroscopy, could be developed to achieve E_a of SEs. Stimulated-echo experiments can powerfully analyze molecular dynamics with correlation times in the range of ms to s^[80,84,85]. Via SAE NMR, it is likely to access long-range ion transport parameters with possible diffusion pathways. SAE was employed to probe the Li⁺ dynamic in Mo-doped LLZO^[83]. As shown in Figure 9, the ⁷Li SAE NMR curves show there is one-step echo decay at low temperatures, and a second decay process gains in intensity at long mixing times with increasing temperatures. An E_a, E_a,SAE = 0.29 eV, was deduced as shown in Figure 8E, which is in good agreement with those obtained by T_1ρ. The two-step delay of SAE NMR reveals heterogeneous dynamics, indicating that both 24d and 96h sites are involved in Li⁺ diffusion. SAE decay curves recorded at long preparation times were also applied to accurately analyze dynamic processes, which shows a two-site jump process, with 96h-24d-96h’ present in Mo-doped LLZO.

Figure 9. (A) Fourier transforms (173.15 and 345.65 K) of the ⁷Li spin-alignment echoes of Mo-doped LLZO that were recorded at t_p = 10 μs and short t_m = 10 μs; (B) Stacked plot of the ⁷Li NMR SAE single-spin (sin-sin) two-time correlation functions recorded at t_p = 10 μs. Copyright 2015, American Chemical Society^[83]. LLZO: Li₇La₃Zr₂O₁₂; NMR: nuclear magnetic resonance; SAE: spin-alignment echoes.

Since SEs are used mainly in micrometers or larger sizes for electrochemical devices, the migration in µm space can be provided with the help of the PGSE NMR method. Hayamizu et al.^[86-91] applied the PGSE NMR method to measure the lithium-ion diffusion in Ta-doped LLZO. The observation time and PFG strength were varied to observe the lithium diffusion behavior, with lithium migration in Ta-doped LLZO distributed widely over a 10-100 ms timescale and a distance around 10^-6 m. Wang et al.^[68] obtained the diffusion coefficients of Al,W co-doped LLZO at different temperatures with the aid of PFG. The diffusion rates presented here are for those Li at the octahedral (48g) sites. Wang et al. probed the dynamics information of lithium-ion migration in LLZO and Al-doped LLZO^{[58-60,64,92,93]}, and Ga-doped LLZO^[94] as well.

In short, NMR offers a wide range of techniques to investigate ion dynamics on different length scales and time scales in SEs^[95-97], which has guiding implications for understanding the mechanism of ion migration within electrolytes and designing SEs with better performance. While T₁ relaxation times provide valuable information about local ion dynamics, they do not necessarily imply that these dynamics reflect the long-range transport behavior of charge carriers in operational cells.

DISTRIBUTION FROM MAGNETIC RESONANCE IMAGING

The growth of dendrites and the presence of dead lithium onto or inside SEs can seriously affect the life of solid-state batteries. The formation and growth of dendrites in garnet electrolytes are influenced by microstructure (i.e., SE density, grain size, etc.)^[98], so it is important to map the dendrite distribution upon electrochemical cycling and feedback to the underlying origins. Various methods, such as DFT^[99] and synchrotron X-ray tomography^[100], have been used to study the microstructure formation in Garnet-SEs. Compared with these characterization methods, MRI shows the advantages of non-destructive monitoring and observation in real time. ⁷Li MRI is widely used to detect the microstructural growth Li before observing visible symptoms in the voltage profiles during cycling. Marbella et al.^[101] applied ⁷Li NMR chemical shift imaging (CSI) to study the chemical and spatial information on Li microstructural growth in garnet-SEs. As shown in Figure 10A, LLZTO pellets with a diameter of ~ 5 mm and thickness of ~ 2.5 mm, and lithium metal disks in a diameter of 1.7 mm and thickness of 40 μm were used to assemble the symmetric battery. The ⁷Li CSI of a pristine and short-circuit Li-LLZTO-Li cell are shown in Figure 10B and C, respectively. Due to the difference in chemical shift between LLZTO, lithium metal and lithium dendrites, their resonances can be well distinguished. LLZTO spans the length of the pellet, with lithium metal electrodes capping either side. The signal for Li microstructure varies according to the size and the direction relative to the magnetic field raised by the magnetic susceptibility effects^[102]. Comparison of pristine, ⁷Li CSI of short-circuited cells shows the emergence of a new signal at ~ 267 ppm, corresponding to microstructural Li. Besides, the presence of Li microstructure spans the entire electrolyte region caused by short-circuiting.

Figure 10. (A) Schematic of a symmetrical Li-LLZTO-Li cell depicting the orientation with respect to the external magnetic field, B₀, during the ⁷Li CSI experiment; ⁷Li chemical shift images of (B) pristine and; (C) cycled Li-LLZTO-Li symmetrical cells; (D) Galvanostatic cycling at 0.5 mA cm^-2 for a maximum of 2 h per discharge/charge (plating/stripping) of Li-LLZTO-Li symmetrical cells stopped for ex-situ analysis with; (E) ⁷Li CSI for pristine and different cycling time. Copyright 2019, American Chemical Society^[101]. CSI: Chemical shift imaging.

To probe the formation process of Li microstructures, cells were stopped at different galvanostatic cycling times; the discharge/charge (plating/stripping) profile and ex-situ analysis with ⁷Li CSI of symmetric cells are shown in Figure 10D and E, respectively. Early during the cycling, the stable voltage plateaus and the CSIs which are similar to those of the pristine cell indicate almost no microstructural Li is formed. After one complete cycle, a slight sloping is observed in the voltage profile that coincides with the appearance of microstructural Li in the ⁷Li CSI, indicating the Li microstructures are localized close to the electrode surface and are not yet detrimental to the cell. During charge at t_cyc = 480 min, soft shorts (voltage spikes) are observed, indicating that the microstructures are fine Li dendrites that fuse due to the melting lithium metal, which is consistent with the assignment of these voltage spikes. Finally, at t_cyc = 720 min, a sharp voltage drop is observed, which is consistent with the ⁷Li resonance of microstructures that completely penetrates the SE and connects the two electrodes, indicating short-circuiting.

Chang et al.^[103] combine NMR, MRI and operando acoustic techniques to study the effects of different stack pressures (SP) and galvanostatic charge currents on the microstructure formation at mesoscale and microscopic scales as demonstrated in Figure 11. At 0.2 mA cm^-2 or higher current, a SP of 2 MPa is insufficient to maintain the conformal interface between Li and LLZO. A SP of 7 MPa or higher can induce creep-driven interfacial dynamics, but it cannot eliminate the growth of microstructures indefinitely. A high SP of 13 MPa increases the local yield stress which leads to rupture and short circuit in the battery. Considering the inherent degradation patterns observed under high SPs and long durations of charge passed, low SP operation is required.

Figure 11. Lithium metal–LLZO interface at (A) and (B) 2 MPa. (A) 1D slices of the ex-situ⁷Li CSI experiment in the z-direction highlighting the lithium metal region between 200 and 280 ppm at (A) 2 MPa; (C) 7.4 MPa and; (E) 13 MPa, as indicated by the dashed region in (B), (D) and (E). Contour plot of ex-situ⁷Li CSI depicting a strong lithium metal resonance (~ 250 ppm) extending into the electrolyte region (0 ppm) in the z-direction at (B) 2 MPa; (D) 7.4 MPa and; (F) 13 MPa; (G) Simplified cartoon illustration of interface behavior. Copyright 2021, Springer Nature^[103]. LLZO: Li₇La₃Zr₂O₁₂; CSI: chemical shift imaging.

SUMMARY, CHALLENGE AND PROSPECT

In this review, we summarized the applications of solid-state NMR techniques in garnet-type-based SEs. It illustrates that NMR could access to determine the local atomic structure, ion pathway, ion dynamic and ion distribution of garnet-type SEs.

(1) The atomic structure is usually assigned by fundamental one-dimensional and advanced multiple dimensional NMR spectra, such as CPMAS, DQ-SQ spectra, and MQMAS, which can provide the proximity between homonuclei or improve the resolution of half-integer quadrupole nuclear spectra.

(2) In addition, advanced techniques such as EXSY, isotope ion tracer exchange and variable temperature relaxation rates, SAE, and PGSE can determine the ion dynamics and migration pathway in SEs.

(3) Furthermore, MRI could track and map the formation and growth process of lithium microstructure, and the lithium evolution/distribution inside electrolytes upon electrochemical cycling. These obtained insights from NMR and MRI would help to understand atomic structure, dynamics and the limiting factors, which will, in turn, facilitate the design of SEs with enhanced performances.

However, several challenges hinder the widespread application of NMR in SEs.

(1) Traditional challenges. Sample preparation, cost and accessibility. Preparing suitable samples for solid-state NMR could be challenging, requiring careful consideration of factors such as particle size and packing density inside rotors, and external water and air need to be avoided at most of the time. High-field NMR spectrometers are expensive and require specialized expertise for operation and data analysis, limiting accessibility for many researchers.

(2) Low signal sensitivity. The inherently low sensitivity of NMR due to small Boltzmann distribution, together with the nuclei having large quadrupolar moments commonly used in SEs, requires long acquisition times and large sample quantities, which is impractical for scarce or expensive materials.

(3) Low spectral resolution. The rigid structure and anisotropic environments in SEs lead to broad NMR signals, reducing resolution and making spectral interpretation difficult, which will be exacerbated by the presence of paramagnetic impurities or defects. Complex SE compositions with multiple components can result in significant spectral overlap, complicating the assignment of individual signals and the quantification of different species.

(4) Interpretation of dynamic processes. Extracting quantitative information about ion transport mechanisms from NMR data can be challenging, requiring sophisticated experimental designs and data analysis techniques, which puts forward higher requirements for the measurement and data analysis of researchers.

(5) Design and assembly of battery cell. Battery cells used for solid-state NMR, especially in-situ solid-state NMR, require specialized cell designs that are non-magnetic and compatible with the NMR probe. These cells must allow for controlled electrochemical cycling while maintaining good NMR signal quality. This often involves compromises between electrochemical performance and NMR sensitivity, requiring careful optimization of cell design and materials^[104,105].

Meanwhile, researchers need to take the following aspects into account.

(1) Develop forward-looking NMR techniques. Solid-state NMR suffers from low sensitivity due to small Boltzmann distribution. High magnetic field and cryoprobe at low temperatures can generate high Boltzmann distribution to improve the sensitivity. Ultrafast MAS for echoing signals from spinning sidebands to the isotropic resonances will also enhance sensitivity^[106]. Dynamic nuclear polarization (DNP) can significantly increase sensitivity by transferring the high polarization of electrons to the nucleus under the condition of double resonance^[107]. DNP has been applied to composite SEs^[108]; Overhauser Effect DNP facilitates effective polarization transfer in the SEI for ^6,7Li isotopes via Li/Li⁺ exchange as the key mechanism. Ultrafast MAS also can be used to enhance the resolution because it averages out the anisotropy of nuclei to narrow the signals. Temperature-dependent NMR spectra and relaxation can easily probe the dynamic in SEs^[109], which depends on the development of high and low temperature probes. Additionally, 2D and 3D MRI and in-situ techniques can provide the distribution heterogeneity of specific elements in SEs; for example, ⁷Li MRI reveals Li depletion from the electrode-electrolyte interfaces and increased heterogeneity of Li distribution upon electrochemical cycling^[110].

(2) Expand the application of solid-state NMR to other systems to effectively study structural stability, interfacial reactions and products. Solid-state NMR is useful for probing the interfaces between electrodes and electrolytes, such as SE interphase (SEI) in solid batteries, which are critical for battery performance. Understanding the interfacial chemistry and stability can help in mitigating issues such as interfacial resistance and degradation. Additionally, solid-state NMR can help identify phase separation or segregation phenomena in composite materials or layered structures, which can affect the overall performance and stability of the battery. It can also monitor chemical reactions and degradation processes occurring during battery operation, providing insights into failure mechanisms and guiding the development of more stable materials. Furthermore, solid-state NMR can provide information about the mechanical properties of solid battery materials, helping to assess their brittleness, elasticity, and overall stability under operational conditions. The proposed NMR methods could also be adopted to other storage materials or systems, such as sodium-ion and lithium-air batteries.

With garnet-type SEs as the representative of SEs, this review summarizes what solid-state NMR technology can observe in solid batteries, which will enable researchers to have a more intuitive understanding of the role of solid-state NMR, and prospects of the challenges solid-state NMR faces in solid-state battery research and the future research directions, as shown in Table 1. In short, with the efforts of researchers, more cutting-edge NMR methods will be developed and applied to solve key problems in solid-state batteries for achieving higher performances.