Electrochemical Stability of Garnet-Type Li₇La_2.75Ca_0.25Zr_1.75Nb_0.25O₁₂ with and without Atomic Layer Deposited-Al₂O₃ under CO₂ and Humidity

The lithium garnet Li₇La_2.75Ca_0.25Zr_1.75Nb_0.25O₁₂ (LLCZN) samples were prepared via two separate methods – solid-state method for samples without ALD treatment and combination sol-gel and solid-state method for the samples with ALD coating. The sample without the Al₂O₃ coating was synthesized at the University of Calgary via the solid-state method.^22,23 Stoichiometric amounts of LiNO₃ (99%, Alfa Aesar) (10 wt% excess), La₂O₃ (99.99%, Alfa Aesar) (dried at 900°C for 12 h), Ca(NO₃)₂·4H₂O (>99% Alfa Aesar), ZrO₂ (99% Alfa Aesar), and Nb₂O₅ (99.99%, Alfa Aesar) were thoroughly mixed by planetary ball mill (Pulverisette, Fritsch, Germany) at 200 rpm with zirconia balls for 6 h in isopropanol. The resulting powders were dried and fired at 700°C for 6 h to decompose the nitrates and were then ball-milled a second time similar to the first milling step. The powders were isostatically pressed in rubber molds in oil with a pressure of 150 MPa. Compressed powder rods were placed on a powder bed from the same batch inside a clean alumina crucible and covered with additional mother powder to reduce the lithium loss during high-temperature treatment. The rods were sintered via a two-step heating process. The samples were heated to 900°C for 24 h then directly to 1100°C for an additional 6 h for densification.

The sample with Al₂O₃ (ALD-LLCZN) coating was synthesized at the University of Maryland (UMD) via the sol-gel method, as previously reported.¹¹ Stoichiometric amounts of La(NO₃)₃ (99.9%, Alfa Aesar), ZrO(NO₃)₂ (99.9%, Alfa Aesar), LiNO₃ (99%, Alfa Aesar), NbCl₅ (99.99%, Alfa Aesar) and Ca(NO₃)₂ (99.9%, Sigma Aldrich) were dissolved in de-ionized water and 10% excess LiNO₃ was added to compensate for lithium volatilization during high-temperature synthesis. Citric acid and ethylene glycol in the molar ratio of 1:1 were consequently added to the solution. The solution was slowly evaporated on a hotplate to produce the precursor gel with stirring, which was then heated to 400°C for 10 h to burn out the organics. The obtained powder was ball-milled and pressed into pellets for calcination at 800°C for 10 h. The synthesized powders were then uniaxially pressed into pellets, which were sintered at 1050°C for 12 h in alumina boats covered with the same powder. ALD coating was also performed at UMD. Sintered pellets were coated with an ultra-thin Al₂O₃ layer using atomic layer deposition (Beneq TFS 500). The thickness of the Al₂O₃ layer was measured using a spectrophotometer (MET03 – N&K spectrophotometer) and was to found be 10 nm. The crystal phase of the sintered materials was analyzed by powder X-ray diffraction at room temperature using a Bruker D8 Advance (Cu K_α, 40 kV, 40 mA) over the 2θ range 10–70° with a rate of 3 sec/step and step size of 0.02°. Further characterization of the Al₂O₃ layer was reported in our previous work.¹¹

In situ electrochemical impedance spectroscopy (EIS) measurements were performed using Li-ion blocking gold electrodes in air using a Solartron 1260 impedance analyzer (0.1 Hz-1 MHz; 100 mV) to characterize the conductivity of the samples at different temperatures. The electrochemical stability of the samples upon exposure to CO₂ and humidity was investigated using a Parstat 4000 (Princeton Applied Research 0.1 Hz-1 MHz; 100 mV) coupled to an in-house environment chamber under flowing CO₂ and 3% H₂O in Ar, N₂ or CO₂. A schematic of the in situ EIS setup is shown in Fig. 1. The ALD-coated and uncoated samples were placed within a millimeter of each other inside the same tubular furnace during the measurements to minimize any temperature gradient effect, and to ensure that any change in impedance due to temperature fluctuation in the background is the same in both samples. Before the electrical conductivity measurements, pellets that are 1 to 1.5 mm thick and 10 mm in diameter (without Al₂O₃ coating) were polished using SiC polishing papers and isopropanol media, starting with #320 grit followed by #600 and finally #1000. The pellets were sonicated in isopropanol to remove any SiC particulate and dried in ambient at 80°C for 5 mins. The samples with Al₂O₃ coating were used as received. Both coated and uncoated pellets were then sputtered (Bal-Tec SCD 500) with gold on each side with a thickness of ∼30 nm/side to act as the current collector. Humid atmospheres were mimicked by passing Ar through a bubbler containing deionized water at room temperature to create a 3% humidity with Ar as the carrier gas. CO₂ conditions were regulated by diluting CO₂ with N₂; all gases were controlled with mass flow meters (Alicat Scientific).

Zoom In Zoom Out Reset image size

The electrochemical stability window was determined by cyclic voltammetry using a Solartron 1287 electrochemical interface with gold as the working electrode and lithium as the reference and counter electrode. The gold working electrode was sputtered on one side of the pellet. The samples were transferred to an Ar-filled glove box where lithium granules were rolled and punched into discs (∼0.2 cm²) and then melted onto the bare garnet surface (200°C, 1 h). The samples were sandwiched between stainless steel coin cell spacers to ensure uniform pressure. Cross-sectional microstructure imaging of each sample was done before and after CO₂ and humidity exposure using a Zeiss Sigma VP scanning electron microscope (SEM). The post-tested samples were fractured right before placing inside the SEM chamber to rule out any surface impurity formation caused by exposure to ambient air.

The changes on the surface of the ALD-coated and uncoated samples with time upon exposure to ambient air or humidity were monitored using atomic force microscopy (AFM) (Agilent 5500) with an environmental chamber in contact mode using diamond-like carbon-coated Si tips. Humid conditions were achieved by bubbling pure N₂ through water at room temperature, establishing a 3% H₂O/N₂. The uncoated samples were polished like those for electrical conductivity measurements, but with additional steps of polishing in cloth with 7 and 0.3 μm alumina powder to ensure that the roughness will not impede continuous scanning of the AFM tip. AFM was performed immediately after polishing. The ALD-coated samples which were stored in an Ar-filled glove box were characterized immediately after mounting to the AFM holder.

The X-ray diffraction patterns collected from the crushed powder of both samples prepared via solid-state and sol-gel methods align well with the isostructural parent phase, Li₅La₃Nb₂O₁₂ (powder diffraction file (PDF) No.: 45-0109), confirming the formation of garnet-type crystal structure (Fig. 2). The XRD data also indicates that the cubic phase was formed as opposed to the lower Li-ion conducting tetragonal garnet phase.^24,25 Although the bulk samples were coated with a thin (∼10 nm) Al₂O₃ layer, the Al₂O₃ signal was not detected, which is most likely explained by the low amount of the material that is below the resolution of XRD. The very weak impurity peak at ∼22° is attributed to LiAlO₂ phase (PDF No.: 18-0714). We have seen similar Al contamination in the earlier work.²¹ The effect of such impurity phases on the response of the garnet to CO₂ and humidity is not known at the moment, but since both the uncoated and ALD-coated samples contain such impurity phase, it is reasonable to compare the samples based on their overall behavior under exposure to CO₂ and humidity. A small peak at 21° from an unknown impurity is also found in the uncoated sample. The small amount of impurities did not significantly affect the conductivity of the samples as seen in the following discussion.

Zoom In Zoom Out Reset image size

We have confirmed that the LLCZN samples prepared by both methods result in pellets with comparable ionic conductivities and relative density (Fig. 3). The relative density of both the pellets was found to be about 80% based on their geometry and weight. The conductivity of both pellets are ∼ 10⁻⁴ S cm⁻¹ at room temperature, and both have activation energies of ∼ 0.37 eV. The conductivities were estimated from the low-frequency tail intercept in the Nyquist plots (Fig. 3a). The samples have similar microstructures with grain size of approximately 2–10 μm in diameter (Figs. 3b and 3c).

Zoom In Zoom Out Reset image size

Garnets exhibit wide electrochemical stability window (up to 9 V vs. Li/Li⁺).⁶ It has been reported that Nb-doped Li-garnets do not show the same stability due to the reduction of Nb⁵⁺ to Nb⁴⁺.^26,27 To investigate the electrochemical stability of our LLCZN samples, cyclic voltammetry was performed in both Al₂O₃-coated and uncoated samples. The potential was scanned between −0.5 to 6 V vs. Li/Li⁺ with a sweep rate of 1 mV/s. The cyclic voltammogram of both uncoated and ALD-coated (Figs. 4a and 4b, respectively) samples only show peaks centered around 0 V vs. Li/Li⁺ which are attributed to the Li deposition and dissolution, as seen in other studies.^6,28 The wide electrochemical stability window suggests that these materials can be used for high voltage battery.

Zoom In Zoom Out Reset image size

Many groups have suggested that CO₂ is the main culprit to the high interfacial resistance between Li electrode and lithium garnet electrolyte.^14,16,18 CO₂ is believed to react with the garnet through a two-step reaction (protonation then carbonation as in Equations 1 and 2). Direct carbonation is another possibility:¹⁷

Li₂CO₃ is formed on the surface of the sample that impedes the interfacial reactions. However, it is not known if the presence of CO₂ affects the bulk properties of garnets, most importantly their Li-ion conductivity. Here, we investigated the electrochemical stability of lithium garnets in CO₂ rich environments. In situ impedance measurements of LLCZN pellets were carried out under varying CO₂ concentrations (400 ppm CO₂/N₂ to pure CO₂) at 25°C. The lowest CO₂ concentration studied was 400 ppm as this is the approximate concentration in ambient air.²⁹ The samples were monitored for a minimum of 100 hours at 25°C under varying CO₂ concentrations to confirm stability. Before exposure to CO₂, the conductivity of the samples was monitored under dry N₂ for 100 h, and no significant change in impedance was observed in both coated and uncoated samples.

Figs. 5a and 5b show the Nyquist plots under dry N₂, 400 ppm CO₂/N₂ (100 h) and 3000 ppm CO₂/N₂ (100 h) at 25°C for the uncoated and ALD-coated samples, respectively. A low-frequency spike in the impedance spectra was observed in all cases due to the blocking nature of Au to Li⁺. The impedance spectra of the uncoated LLCZN sample do not show a clear grain boundary response (Fig. 3a) even though both samples were fabricated by the same procedure. We have tested several samples to check for similar features of grain boundary response and values of total conductivity. The impedance shown in Figs. 5a and 5b were typical behavior. The reason for the difference is not clear, but it is known that impedance may vary from sample to sample even with similar chemical composition.^{11,19–21,30} It was not possible to separate the bulk and grain boundary contribution. Only a total electrolyte resistance R_T [Ω], which accounts for the bulk (R_b) and grain boundary (R_gb) contributions, can be obtained from the real part of the impedance before the low-frequency spike (Fig. 5a). The data in Fig. 5a (also Fig. 5b) is expressed as resistivity, ρ [Ω cm], i.e., multiplied impedance by {a}/{l} where a and l are the area and thickness of the sample in cm² and cm, respectively.

Zoom In Zoom Out Reset image size

The ALD-coated sample's impedance spectra show a bend just before the low-frequency spike which could be due to a grain boundary response. The simplest equivalent circuit that could fit the data is (R_bC_b)(R_gbCPE_gb)(R_eCPE_e), where R_b, R_gb, and R_e are the resistance of the bulk, the grain boundary and the electrode due to possibly gas-related reaction, respectively. C_b, CPE_gb, and CPE_e are the capacitance effects from the bulk, grain boundary, and electrode, respectively (CPE refers to a constant phase element). Ideally, CPE_e is expected for a blocking electrode. The electrode impedance was found to be in the order of 10⁶ Ω, and was only added to provide a better fit to the data. The quasi-equivalent capacitance (QEC) associated with the fitted capacitance can be calculated using:³¹

where R is the resistance and n is the exponent. Based on this calculation, the (R_gbCPE_gb) semicircle has a QEC in the order of 10⁻¹⁰ F, which is in the range of grain boundary contribution.³² The equivalent circuit and the simulation of the arcs from the fitting are shown in Fig. 5b.

The total electrolyte resistance for the uncoated LLCZN sample was the real impedance value before the low-frequency spike, while the total electrolyte resistance value of the ALD-coated LLCZN sample is R_b + R_gb from the equivalent circuit fitting. A slight increase in resistivity was observed in the uncoated sample (6.9 to 7.6 kΩ cm) while no significant change in impedance was observed in the ALD-coated sample after exposure to 3000 ppm CO₂. The CO₂ was then removed and the samples were exposed to ambient air for the next 100 h to examine the effect of ambient humidity (Equations 1 and 2) on the impedance of the samples. CO₂ concentrations were then increased progressively up to 3000 ppm. The total experiment time was approximately 1000 h.

The total electrolyte resistivity change upon gas exposure (3% H₂O, Ar, N₂, or CO₂) in % was calculated relative to the initial total electrolyte resistivity before exposure (0 h in N₂ or Ar):

where ρ⁰ is the resistivity at 0 h in N₂ or Ar and ρ^x_gas is the resistivity upon exposure to a certain x concentration of gas. Figs. 5c and 5d summarize the effect of CO₂ content changes on the total electrolyte resistivity. Further increase in resistance was observed in the uncoated sample after exposure to 100 h in ambient air while the ALD-coated sample appears to show no change in impedance. The resistance of the uncoated sample had clearly increased by about 8% from its initial resistance before exposure to CO₂ while the ALD-coated sample showed no appreciable increase in resistance. Another sample of each of the ALD-coated and uncoated Li-garnets were exposed to accelerated CO₂ increase test as shown in Figs. 5e and 5f to confirm the reproducibility of the results. Note that at 2000 ppm CO₂, the increase in resistance for the uncoated sample is already 5.9% while the ALD-coated sample show improved conductivity after exposure to the same CO₂ content. After exposure to pure CO₂, the increase in resistance compared to dry N₂ for the uncoated sample was 7.6%, almost the same as the other uncoated sample exposed to 3000 ppm CO₂, while the ALD-coated sample appears to have decreased resistance. The results in Fig. 5 highlights an apparent protecting property of the ALD coating to the garnet from CO₂ exposure.

Wang and Lai³³ have also studied the stability of lithium garnet, Li₇La₃Zr₂O₁₂, under pure CO₂. However, at 350°C Wang and Lai noticed an initial decrease in impedance for the first 24 hours, followed by a small increase after 48 hours. The decrease in impedance was attributed to a phase transition from tetragonal to cubic structure.³³ The carbonation of the garnet sample leads to increased Li vacancies and ultimately a stabilized cubic garnet structure. The investigated samples have been stabilized in the cubic phase by Ca- and Nb-doping in Li₇La₃Zr₂O₁₂; hence, this decrease in impedance due to phase change is not expected.

We took a similar approach as the CO₂ stability tests for the humidity stability of LLCZN. The samples were transferred from an Ar-filled glove box to the impedance set-up with the environmental chamber. The chamber was purged with dry Ar for a minimum of 100 h before starting the humidity tests. Purging was performed to ensure no humidity or CO₂ was present before the measurements, and to confirm that the sample impedance does not change under an inert atmosphere. Upon introduction of 3% H₂O/Ar at 25°C, clear responses in the Nyquist plots were observed in both the uncoated and ALD-coated samples (Figs. 6a and 6b). Specifically, in the ALD-coated, the bend found in the impedance spectra before exposure to humidity has evolved into a clear semicircle upon exposure for about 100 h (Fig. 6b). Note that the semicircle becomes apparent even after 1 h of exposure to humidity.

Zoom In Zoom Out Reset image size

The semicircle in the impedance spectra before and after 95 h exposure to humidity in the ALD-coated sample was successfully fit with (R_b)(CPE_gbR_gb)(R_eCPE_e) equivalent circuit. The QEC of the additional semicircle is in the order of 10⁻¹⁰ F, lying in the range of grain boundary capacitance.³² A reasonable fit to an equivalent circuit accounting for the semicircle in the uncoated sample could not be established. However, we believe the response will also be from a grain boundary-like resistance increase. Similar grain boundary increases were observed by Shimonishi et al.³⁴ after treating bulk garnet samples with H₂O, 1 M LiOH or 0.1 M HCl for 1 week at 50°C.

Within 1 hour of removing the humidified Ar, the impedance began recovery, and the grain boundary contribution began to retreat. After prolonged exposure (93 h) to dry Ar, the impedance spectra would lose the grain boundary-like response and the total resistance is almost the same as the pristine sample for the uncoated sample, while the ALD-coated sample seems to have lowered total resistance. Figures 6c and 6d summarize the change in resistivity of the uncoated and ALD-coated samples, respectively, as defined by Equation 5, when cycling between 3% humidified Ar and dry Ar. Each cycle shows the appearance of grain boundary (under humidity) and recovery (under dry Ar) on both the uncoated and ALD-coated samples. Cycling was performed with numerous samples of both ALD-coated and uncoated LLCZN, confirming the excellent reproducibility of the stability tests.

The mechanism for the appearance of grain boundary-like response in the impedance spectra and the recovery upon removal of humidity is not clear. One possible explanation is that in humid environments, water may adsorb onto the grain boundaries of the polycrystalline material, impeding the Li-ion transport. When the humidity is removed, and dry gas flow is reintroduced, the adsorbed water likely removed and the impedance to return to that of initial values. It is also uncertain why Al₂O₃ did not protect the garnet pellet from the humid environment unlike in CO₂ environments.

After humidity exposure, the garnet sample become protonated followed by a LiOH coating that will further react with CO₂ forming a Li₂CO₃ passivation layer. The sample was exposed to 3% H₂O/Ar for 4 h to induce the increased impedance and then humidity was removed. Subsequently, 400 ppm CO₂ was introduced. As seen with dry Ar, the impedance was found to recover to pristine sample value under 400 ppm CO₂. The experiment was repeated with pure CO₂ and similar recovery was observed, as seen in Figs. 7a and 7b for both uncoated and ALD-coated, respectively. Recovery was not only limited to the continuous flow of dry gas. Leaving the sample in ambient air after removing the humidity causes the impedance to recover its original shape.

Zoom In Zoom Out Reset image size

The fracture cross-section SEM of CO₂ and humidity-treated samples are shown in Fig. 8. Figs. 8a and 8b compare the microstructure of the uncoated and ALD-coated samples, respectively, after 1000 h of testing and exposure to a maximum of 3000 ppm CO₂ (Figs. 5c and 5d), while Figs. 8c and 8d compare that of the uncoated and ALD-coated samples, respectively, after 700 h of alternating between dry Ar and 3% H₂O/Ar (Fig. 6). Except for the uncoated sample exposed in CO₂, all the samples did not show any evidence of significant change from the observed microstructure before the long-term tests (Figs. 3b and 3c). The uncoated sample exposed in CO₂ (Fig. 8a) exhibits grains that are covered or partially covered with a porous, fine-grained layer, which is most likely Li₂CO₃. The layer covers almost the entire cross-section of the pellet, which might be the reason for the increase in the total electrolyte resistance of the uncoated sample in Fig. 5c. The unchanged microstructure exhibited by the ALD-coated sample was consistent with the impedance characterization showing a negligible increase in total electrolyte resistance.

Zoom In Zoom Out Reset image size

In situ AFM was employed to study the surface features in both samples upon exposure to ambient and humidity. Fig. 9 shows the AFM images of coated and uncoated samples after continuous scanning of the surface that is exposed to ambient air. Even after leaving the samples overnight, no drastic change was observed in both the uncoated and ALD-coated samples. The findings are consistent with the negligible change in the impedance during in situ characterization of both samples upon exposure to 400 ppm CO₂ and ambient air for many hours as shown Figs. 5c and 5d.

Zoom In Zoom Out Reset image size

Figure 10 shows the surface of the samples after scanning every hour that the samples were exposed to humidity. A distinct droplet-like formation is observed in the uncoated sample while no noticeable change is observed in the ALD-coated sample. The AFM imaging also included a final step wherein the humidity was removed and a continuous flow of N₂ was held in the ALD-coated sample for 2.5 h to mimic the effect of recovery as seen in the impedance measurements. No noticeable change in the surface features was observed (the distinct droplet-like formation did not disappear). The droplet-like formation may not be directly related to the observed changes in impedance, as there was no noticeable change in the surface features upon removal of humidity and exposure to a continuous flow of N₂ for 2.5 h. Future in situ technique work is required to resolve grain boundary features during electrochemical tests upon exposure to humidity.

Zoom In Zoom Out Reset image size