Introduction

Solid electrolytes have become central to advancing all-solid-state battery technology1,2,3,4, as they enable safer and more compact designs by eliminating the liquid electrolytes traditionally used in lithium-ion batteries5. Liquid electrolytes are associated with numerous safety risks, including leakage, flammability, and instability in high-energy environments. Therefore, solid electrolytes4,6 offer a promising alternative, particularly when combined with a lithium metal anode7,8,9, which offers a higher theoretical energy density than conventional graphite anodes. The advancement of all-solid-state batteries4 requires, however, solid electrolytes that provide ionic conductivities on the order of 1 mS cm−1 or higher at room temperature. Achieving these conductivity levels10,11 is crucial to support the energy and power densities necessary for next-generation energy storage applications, particularly in electric vehicles and portable electronics4.

Among various types of solid electrolytes, ceramic solid electrolytes1,2,3,11,12 have attracted considerable attention due to their high ionic conductivity, mechanical stability, and thermal resistance. As examples, these ceramics typically comprise oxides (garnets)2,3,13, sulfides1,14,15, (thio-)phosphates14,16,17,18, and halides19,20,21,22, each offering specific benefits for ionic conduction and structural integrity. They also differ in electrochemical stability against lithium metal23,24,25, which is essential because direct contact with metallic lithium can lead to undesirable reactions, potentially degrading the electrolyte or causing the growth of lithium dendrites26. These reactions threaten both the safety and longevity of a given lithium metal battery4.

Achieving this interfacial stability9,25,26,27,28,29 is a complex and challenging aspect because lithium metal is highly reactive and tends to form undesirable interphases with many materials. Ceramic electrolytes that maintain a stable interface with lithium metal are therefore indispensable. If the ceramic electrolytes tend to form a mixed-conducting interphase (MCI)30 with the metallic anode, it is important to study the chemical, electrochemical and dynamic properties of such an interfacially formed layer. Since the operando formed interfacial layer is rather thin, some spectroscopic methods suffer from their poor sensitivity to capture the dynamics of the few Li ions in such a layer. Such a limitation in sensitivity is, for example, given in the case of nuclear magnetic resonance (NMR)31,32 that requires, if carried out at moderate magnetic fields, a sufficiently high number density of spins to produce a Li NMR signal with a reasonable signal-to-noise ratio. To overcome this limitation and to take advantage of the whole power of NMR, we mimic the operando lithiation of a solid electrolyte by treating the polycrystalline powder sample in n-butyllithium to prepare larger quantities of the lithiated electrolyte ex situ. As lithium aluminum titanium phosphate (LATP) is known to easily form an MCI30,33, it served as a suitable model substance for our investigation.

In this sense, the present study complements a recent one34 that used impedance spectroscopy to study macroscopic Li+ transport parameters and electronic conductivities in lithiated LATP. In that earlier study it turned out that bulk Li+ ion dynamics do not suffer from lithiation. In contrast, impedance spectroscopy even revealed a slight increase in Li+ conductivity at a low lithiation level. Information on (bulk) Li+ ion dynamics in heavily lithiated LATP (Li3Al0.3Ti1.7(PO4)3), which shows a crystal structure (space group R3) being different from that of unlithiated LATP (R3c, see below), is, however, still missing. By using 7Li NMR relaxation methods35, we were able to answer the important question of how the fast bulk ion dynamics present in the source material is affected by lithiation.

Results and Discussion

NMR spin-lattice relaxation rates

In Fig. 1a, c the 7Li NMR spin-lattice relaxation rates of unlithiated LATP Li1.5Al0.5Ti1.7(PO4)3 and those of lithiated LATP (Li1.3+xAl0.3Ti1.7(PO4)3) are shown using an Arrhenius representation, that is, log10R1 is plotted vs. the inverse temperature 1000/T. In all three cases we recorded not only the laboratory-frame R1 rates but also the diffusion-induced rates that correspond to spin-lattice relaxation (R1ρ) in the presence of a B1 magnetic field that is orders of magnitude lower than the B0 field being relevant for R1 relaxation36,37. While B0 corresponds to a Larmor frequency of ω0/2π = 116 MHz, B1 is determined by a locking-frequency of ω1/2π = 20 kHz ignoring any local fields that would increase ω1. In general, the rates R1 and R1ρ are expected to pass through characteristic diffusion-induced rate peaks whose flanks and position on the temperature scale contain valuable information on 7Li diffusion parameters such as activation energies, Arrhenius pre-factors and jump rates31,38. The analysis of such peaks has been documented elsewhere35. Here, data for unlithiated LATP were taken from an earlier work35 that also includes a detailed discussion of these rates. Though the Li and Al contents of this reference sample slightly differ from those of the lithiated sample with x = 0.2 (see Fig. 1a), we do not find any significant differences in spin-lattice relaxation behavior. The R1 rates of the two samples do almost coincide except for a shallow deviation at low temperatures. For the NMR R1ρ rates of the sample with x = 0.2 we cannot resolve the two-peak feature any longer that is seen for the reference sample. The rates R1ρ peaks in Fig. 1 have been analyzed with a proper NMR relaxation model (see the solid lines in Fig. 1a–c, see Supplementary Note 1 and Supplementary Table 1).

Fig. 1: NMR relaxation rates and spectra.
figure 1

Temperature evolution of the 7Li NMR spin-lattice relaxation rates (116 MHz, 20 kHz locking frequency) of (a) non-lithiated Li1.5Al0.5Ti1.7(PO4)3 and lithiated Li1.3+xAl0.3Ti1.7(PO4)3 [(a) x = 0.2, (b) x = 0.6, and (c) x = 1.3)]. The rates have been recorded in both the laboratory (R1) and rotating (R1ρ) frame of reference. In the latter case, we observe the transversal adjustment of the magnetization that is locked through a weak B1 field. As in the case of R1, this adjustment is driven through diffusive processes but sensitive to motional correlation rates in the kHz rather than the MHz regime. Upon lithiation, we observe a slight shift of the R1ρ(1/T) peaks towards lower temperature that is accompanied with a reduction in activation energy determining the low-T flank. For R1, an additional shallow peak emerges when x takes values equal or larger than 0.6 (b). The latter points to rapid localized Li+ exchange processes that are hardly seen in unlithiated LATP. Lines are to guide the eye. Values in eV refer to the activation energies corresponding to the linear parts of the curves. d to f; Variable-temperature 7Li (I = 3/2) NMR line shapes (116 MHz) of the same samples recorded under static conditions. For the samples with (d) x = 0.2 and (e) x = 0.6, central and satellite intensities, that is, the 90° singularities of the powder patterns, are seen (note the different scaling of the abscissae). f Increasing x to 1.3, causes the lines to be strongly affected by dipolar paramagnetic broadening because of the increasing number of Ti3+ centers formed.

Full size image

However, we notice an overall slight shift of the total R1ρ(1/T) rate peak towards lower temperature that is accompanied with a reduction in energy of both flanks, that is, the high-T (0.13 eV) and the low-T flank (0.10 eV). These changes reveal slightly faster Li+ exchange processes in the lithiated material as the rate peak appears at exactly the temperature where the mean correlation rate τc−1, being very similar to the jump rate τ−1, equals the Larmor or spin-lock frequency, τc−1 ≈ ω0(1). The same shift of the peak position is seen when increasing x to 0.6 (see Fig. 1b). While the shape of the R1ρ(1/T) peak remains untouched, it is slightly shifted to even lower temperatures, as is indicated by the vertically drawn dashed lines in Fig. 1a, b. This change reveals a further increase in (localized) Li+ hopping processes. In addition to this shift, we recognize a much stronger influence of lithiation on the R1 rates. The coupling of the 7Li spins with the increasing number density of paramagnetic Ti3+ centers might be at the origin here. Most importantly, a local maximum in R1(1/T) is seen at the temperature where the R1ρ(1/T) peak appears (Fig. 1b). It points to extremely fast, most likely localized, Li+ exchange processes including even forward-backward jumps, the ions are involved in.

This peak is also detectable for the highest lithiated sample characterized by x = 1.3, whose spin-lattice relaxation NMR rates are shown in Fig. 1c. In agreement with the trend seen for x = 0.2 and x = 0.6, the spin-lock rate peak shifts to even lower temperatures and appears only as a shallow one with quite flat slopes on either side ( < 100 meV). This flattening of the peak is most likely affected by the fact that the R1 magnetization transients cannot longer be approximated with a single (stretched) exponential function. Instead, they contain, besides the contribution R1slow that is similar to R1 of the other samples, a fast-relaxing component labeled R1fast (Fig. 1c) showing a rather weak temperature dependence. The bi-exponential feature of the transients is illustrated in Fig. 2a for selected curves. The relative amplitudes of the two contributions is shown in Fig. 2b as a function of the inverse temperature. With increasing temperature, the amplitude of the fast relaxing component increases and dominates the overall relaxation at temperatures higher than approximately 300 K. Likely, this component reflects the 7Li+ spins in the Ti3+ rich regions in heavily lithiated Li1.3+xAl0.3Ti1.7(PO4)3 with x = 1.3. As has been shown earlier, lithiation of LATP causes a phase transformation leading to a two-component system consisting of nonlithiated or less lithiated inner regions and highly lithiated regimes towards the surface of each particle34. For x = 1.3, SLR NMR can distinguish these two regions of a core–shell morphology. Evidence for Ti3+-rich surface regions and Ti3+-poor inner regions of a given LATP particle has directly been found by room-temperature X-ray photoelectron spectroscopy in our recent work34. We cannot exclude that homogenization of the Li+ distribution takes place with increasing temperature, which might lead to a change of the amplitudes for T > 300 K.

Fig. 2: NMR magnetization curves and amplitudes.
figure 2

a Magnetization Mz(tdelay) transients of the 1.3Li-LATP sample that show bi-exponential behavior leading to the rates R1 fast and R1 slow, see Fig. 1c. The solid lines show fits with a sum of two stretched exponential functions. b The relative amplitudes of the two contributions that control overall relaxation as a function of the reciprocal temperature.

Full size image

To sum up our findings up to here, we do not find any indications that lithiation at low levels leads to a decrease in overall Li+ diffusivity in LATP. Even at intermediate (x = 0.6) and high levels (x = 1.3) Li+ exchange seems to be slightly enhanced. These findings excellently agree with those extracted from bulk ion conductivity measurements, presented earlier34 and mentioned above. Hence, electrochemical degradation of LATP through the reaction with metallic lithium, which is here mimicked by the treatment with n-Buli, does not entail any disadvantages for Li+ hopping.

NMR line shape measurements

To refine the above-mentioned core-shell picture34 that evolves through lithiation of the LATP particles starting at the surface regions, we recorded 7Li (spin-quantum number I = 3/2) NMR line shapes as a function of temperature, see Fig. 1d–f. For the sample with x = 0.2 we find the typical motional line narrowing behavior39, which is driven by Li+ hopping processes with rates in the kHz range. In the absence of sufficiently fast Li+ motions (T < 140 K) the magnetic dipolar and electric quadrupolar interactions govern the NMR line leading to a broadened central line and almost invisible satellite intensities40,41. Motion-induced averaging causes the line to significantly narrow assuming the shape of a sharp Lorentzian at high T (493 K). Distinct 90° quadrupole satellites of the powder pattern appear first at T > 230 K. Their distance on the frequency scale increases with temperature as it might be expected for (anisotropic) lattice expansion. As an example, from 253 K to 453 K, the quadrupole coupling constant δq increases from 40 to 55.2 kHz. These values have been determined by assuming an axially symmetric (averaged) electric field gradient at the 7Li sites.

For the sample with x = 0.6 the same single satellite pair is seen; at 293 K we obtain δq = 42.4 kHz, which is slightly higher as for the sample with lesser amount of lithium. At 453 K the coupling constant turned out to be 53.6 kHz. These changes reveal a slight increase in the magnitude of the average electric field gradient sensed by the Li+ ions as soon as additional charge carriers enter the structure and Ti3+ centers are generated. More importantly, lithiation up to x = 0.6 leads to an increase in the dipolar broadening of the static lines. Considering the spectra recorded at 493 K, we do not see the extremely narrow central line that is detected for the sample with x = 0.2. Obviously, in terms of a two-phase core-shell picture, Ti3+ centers are not only restricted to an outer area of the crystallites. Instead, they are also found, to a lesser extent, in the interior of the LATP particles. Although diluted in these inner regions, the number of Ti3+ centers is sufficient to broaden the whole NMR signal. The finding that Li+/e− pairs diffuse into the inner regions of the LATP particles that still crystallize with R3c symmetry, is illustrated in Fig. 3a–d. Here, we assume a highly correlated if not coupled transport34 of Li+ and e− toward the inner regions of the particles (see Fig. 3d). The electron exchange between Ti3+ and Ti4+ is accompanied by Li+ hopping in the direct vicinity.

Fig. 3: Lithiation of LATP.
figure 3

Illustration of the surface degradation reaction, see (a) to (d), that takes place if LATP is either in contact with Li metal or any other Li source as n-BuLi, see the reaction shown at the bottom line. b The particles are covered with a mixed-conducting layer of the Li3-phase that grows towards the inner regions (c). d In addition, coupled Li+ and e− transport generates Ti3+ centers in the regions farther away from the surface; consequently, a broad 7Li NMR resonance is detected at sufficiently high lithiation levels x, see (c) and (d).

Full size image

In addition to the broadening of the original 7Li NMR line, we see that the 7Li spins directly residing in the Li-rich regions, that is, in the Li3-phase Li3Al0.3Ti1.7(PO4)3 (see also Fig. 3c, d) appear as a separate resonance line. This NMR line is indicated in Fig. 1e by the Lorentzian-shaped resonance shifted to the right on the frequency scale. The main resonance, consisting of a central line with a quadrupolar powder pattern, has been considered to isolate this new line, which reflects Li in the Li3-phase; for further details on the deconvolution using an appropriate spectrum simulation, see Supplementary Note 2 and Supplementary Fig. 1. The shift and the additional dipolar broadening of this line is due to Fermi-contact interactions of the Li spins with unpaired electrons42,43. Hence the Li line unequivocally indicates a fraction of LATP that has been transformed into the Li-rich, electronically conducting Li3-phase. The same spectral feature has been detected by high-resolution magic angle spinning 7Li NMR even at room temperature34. Narrowing of this broadened line is seen at higher temperatures indicating Li+ jump rates in the order of several kHz, see Fig. 1e and Supplementary Fig. 1.

Increasing x to 1.3, rather broad 7Li NMR lines are obtained, see Fig. 1f. As an example, at 191 K and below, we see that the lines are composed of two contributions, which we can attribute to Li spins in the Li3-phase (as indicated) and to even faster Li+ ions in the less lithiated inner regions. The latter manifest themselves as a narrowed line (12 kHz) superimposing the broader resonance with a full width at half maximum of 47 kHz, see the spectrum recorded at 191 K (Fig. 1f). Such a separation is possible until a temperature of approximately 230 K is reached. At higher T, a deconvolution of the whole signal might require more than two contributions. At 393 K, the main motional process is finished and at least three spectral features are seen with the main resonance showing, due to coupling of the 7Li spins with the Ti3+ centers, still a full width at half maximum of 6 kHz. Figure 4 compares the so-called motional narrowing curves that were constructed by plotting the full width at half maximum (FWHM) of the overall and deconvoluted NMR lines as a function of temperature to show the temperature mediated averaging processes. For 1.3-LATP we used, for the sake of simplicity, a combination of a Lorentzian and a Gaussian line to approximate the lines up to a temperature of 333 K.

Fig. 4: NMR line widths.
figure 4

Change of the 7Li NMR line widths (116 MHz) of the three LATP samples investigated. While for (b) 0.2-LATP and 0.6-LATP we simply used the full width at half maximum of the central line to construct the curves (see also (c), for (a) 1.3-LATP, we deconvoluted the line with appropriate (a) Gaussian and (c) Lorentzian functions up to T = 333 K. Up to 200 K, the combination of two lines reflect the overall NMR line shape quite well. Between 200 and 333 K, the data points shall be regarded as estimations as also other lines with lower intensity contribute to the overall shape.

Full size image

In agreement with 7Li NMR spin-lattice relaxation, for all samples we see a sharp decrease of the 7Li NMR central lines at T < 200 K indicating that the fast Li+ hopping process already seen in LATP is preserved. Considering that in the higher lithiated samples strong paramagnetic broadening is present, rather fast and efficient Li+ exchange processes are needed to average these increased interactions in these samples. This observation supports our earlier findings that also lithiated LATP represents a fast Li+ ionic conductor. Activation energies are lower than 0.2 eV, see Fig. 1, and underpin this interpretation. We believe that the very low activation energies of approximately 0.09 eV seen in 7Li NMR point to facile localized Li+ hopping processes in lithiated (and non-lithiated) LATP. The appearance of a local maximum in R1 NMR relaxation rates, which is seen for the samples with x = 0.6 and x = 1.3, points to even faster Li+ motional processes in the lithiated samples than in non-treated LATP. Hence, from a point of view of electrochemical degradation through lithiation, mixed-conducting LATP does not lose its “superconducting” properties. As has been shown earlier34, the electronic contribution to the overall ionic conductivity remains at a rather low level, that is, not reaching alarming values that might immediately cause electrical shorts or lead to further degradation.

Studying this system in real-world or practical Li-metal battery applications is extremely challenging, as the number of spins is crucial for generating an NMR signal. This is why we mimicked the process as explained above. We believe that lithiation from Li metal leads to the formation of a Li3-interphase covering the crystallites, which, depending on crystallite size and battery operating conditions, can propagate throughout the entire crystallites. Since Li+ diffusivity is high in the lithiated phase, Li uptake should not be detrimental to battery performance, provided that electronic conductivity does not reach harmful levels. Fortunately, for a system with µm-sized crystallites as studied here, our previous study has shown that electronic conductivity remains orders of magnitude lower than ionic conductivity34.

Conclusions

The fast ion-conductor Li1.3Al0.3Ti1.7(PO4)3 is known to rigorously accept Li+/e− if in contact with sources such as lithium metal or n-BuLi. Here, we prepared Li1.3+xAl0.3Ti1.7(PO4)3 (x = 0.2, 0.6, 1,3) by treatment of polycrystalline samples in a n-BuLi/hexane solution and mimicked the formation of degraded LATP in a lithium-metal battery where the solid electrolyte is in direct or indirect contact with the Li source. Lithiation of Li1.3+xAl0.3Ti1.7(PO4)3 leads to a mixed-conducting ceramic that shows fast Li+ diffusive processes as evidenced by 7Li NMR spin-lattice relaxation and line shape measurements. Activation energies as low as 0.09 eV and the appearance of diffusion-induced 7Li NMR maxima in R1 point to extremely fast (localized) Li+ exchange processes present in the lithiated samples with x = 0.6 and x = 1.3.

By considering earlier results from XPS, XRD and high-resolution 7Li MAS NMR, our findings point to the formation of an interfacial Li-rich and, thus, Ti3+-rich Li3-phase (Li3Al0.3Ti1.7(PO4)3) that propagates towards the inner regions of the crystallites. Broad 7Li NMR resonance lines seen for the sample with x = 0.6, which otherwise resemble those of unlithiated LATP, show that paramagnetic Ti3+ centers are also found in the inner regions of the crystallites. Hence, to a certain degree, e− diffuses into the inner regions. We believe that broadened 7Li NMR lines are caused by highly correlated if not coupled ion-electron (Li+/e−) transport.

Our study focuses on a microcrystalline powder with µm-sized crystallites, and size-dependent measurements are beyond the scope of this paper. However, we speculate that reducing the grain size could significantly enhance the kinetics of the lithiation process, potentially even leading to substantial structural changes.

In summary, electrochemical degradation of LATP-based electrolytes leads to lithiated compounds with very fast Li+ diffusion processes while earlier conductivity measurements revealed low electronic conductivities. Our work highlights that the intrinsic ion/electron transport properties of LATP remain resilient during the electrochemical degradation process, making it a robust candidate for advanced battery applications.

Methods

Preparation and Characterization

Synthesis and characterization of the samples used to record the 7Li NMR relaxation rates are described elsewhere34. For the present study, we used samples of exactly the same batch that have been recently investigated by X-ray powder diffraction, high-resolution 31P, 7Li magic angle spinning NMR, X-ray photoelectron spectroscopy, conductivity and electric modulus spectroscopy, as well as chronoamperometric measurements to study structural features, dynamic properties and electronic conductivities34. We strictly handled the lithiated powder samples under argon atmosphere and sealed them in glass ampules to protect them permanently from any contact with air or traces of moisture.

NMR measurements

Relaxation rates

Motion-induced 7Li NMR spin-lattice relaxation rates were recorded using a Bruker NMR connected to a shimmed cryomagnet with a nominal magnetic field of 7 T. This field corresponds to a Larmor frequency of ω0/2π = 116 MHz. A commercial high-temperature ceramic probe (Bruker) was utilized for experiments at temperatures ranging from 178 K to 493 K. For low-temperature measurements down to 112 K, a Bruker cryo-probe combined with a LakeShore 331 unit with two Cernox sensors for precise temperature control and monitoring was employed. The sample chamber was cooled with freshly evaporated nitrogen.

7Li NMR spin-lattice relaxation rates R1 in the laboratory frame were acquired by applying the saturation recovery pulse sequence consisting of a comb of ten π/2 pulses separated by 80 µs to destroy any longitudinal magnetization Mz37. The subsequent recovery of Mz was recorded with a π/2 reading pulse sent after a variable delay time td. Quadrature detection was used to sample the free induction decays. The π/2 pulse lengths ranged from 2.25 to 4.05 µs, depending on the sample and temperature. The resulting magnetization transients Mz(td), obtained by integrating the area under the FIDs, were analyzed with stretched single or double exponential functions to determine the relaxation rates.

Relaxation rates R1ρ in the rotating frame of reference were recorded with the spin-lock pulse sequence that consists of an initial π/2 pulse and a locking pulse of variable duration tlock37. Between each scan a sufficiently long recycle delay was ensured. During the spin-lock experiment the transversal magnetization M(xy)′ = Mρ(tlock) adapts to the weak locking field. In the present case, the decrease of Mρ(tlock) is induced by Li+ hopping processes and followed stretched exponentials that contain the rates R1ρ(T). Here, we used a locking field ω1/2π of 20 kHz to record the rates as a function of temperature. To generate the same locking field at each temperature the pulse power used was varied from 2.9 to 17 W.

Line shapes

The 7Li NMR spectra at 116 MHz were acquired using single-pulse excitation with a π/2 pulse. The recycle delay was chosen to be sufficiently long to allow complete longitudinal relaxation between the scans. For each spectrum, 16 to 32 scans were accumulated. Following zero filling, the free induction decays were directly Fourier transformed without additional signal processing. To ensure pure absorption-mode signals, zero-order and first-order phase corrections were applied.