Melanie
Rosen
ab,
Ruijie
Ye
ac,
Markus
Mann
a,
Sandra
Lobe
a,
Martin
Finsterbusch
*ad,
Olivier
Guillon
ad and
Dina
Fattakhova-Rohlfing
abd
aInstitute of Energy and Climate Research – Materials Synthesis and Processing (IEK-1), Forschungszentrum Jülich GmbH, 52425 Jülich, Germany. E-mail: m.finsterbusch@fz-juelich.de
bFaculty of Engineering, Center for Nanointegration Duisburg-Essen, Universität Duisburg-Essen, Lotharstr. 1, 47057 Duisburg, Germany
cInstitute for Power Electronics and Electrical Drives (ISEA), RWTH Aachen University, Jägerstr. 17-19, 52066 Aachen, Germany
dHelmholtz Institute Münster: Ionics in Energy Storage (IEK-12), Forschungszentrum Jülich GmbH, Corrensstr. 46, 48149 Münster, Germany
First published on 22nd January 2021
Ceramic solid state-electrolytes attract significant attention due to their intrinsic safety and, in the case of the garnet type Li6.45Al0.05La3Zr1.6Ta0.4O12 (LLZO), the possibility to use Li-metal anodes to provide high energy densities on a cell and battery level. However, one of the major obstacles hindering their wide-spread application is the translation and optimization of production processes from laboratory to industrial scale. Even though the plausibility of manufacturing components and cells via wet processing routes like tape casting and screen printing has been shown, the impact of the sensitivity of LLZO to air and protic solvents due to Li+/H+-exchange is not fully understood yet. An uncontrolled alteration of the powder surface results in poorly reproducible processing characteristics and electrochemical performance of the final battery components and full cells. This knowledge gap is the cause of the large performance variations reported across different research labs worldwide and is unacceptable for up-scaling to industrial level. To close this gap, the influence of the Li+/H+-exchange taking place at various steps in the manufacturing process was systematically investigated in this study. For the first time, this allowed a mechanistic understanding of its impact on the processability itself and on the resulting electrochemical performance of a free-standing LLZO separator. The importance of a close control of the pre-treatment and storage conditions of LLZO, as well as contact time with the solvent could be extracted for each step of the manufacturing process. As a result, we were able to optimize the processing of thin, dense, free standing LLZO separators and significantly improve the total Li-ion conductivity to 3.90 × 10−4 S cm−1 and the critical current density to over 300 μA cm−2 without making structural changes to separator or the starting material. These findings do not only enable a deeper understanding and control over the manufacturing process, but also show potential for further improvement of cell concepts already existing in literature.
In contrast to e.g. sulphide based solid electrolytes, LLZO can be handled in ambient air. However, in the presence of water the surface undergoes a very fast Li+/H+-exchange with formation of a poorly conductive LiOH and subsequently Li2CO3-layer on the surface. This process takes place when LLZO is exposed to the humidity in air16–19 and to all solvents commonly used for wet-processing of ceramic components.20 Although this Li+/H+-exchange on the LLZO surface during storage in air (or any atmosphere with traces of humidity) or exposure to solvents is practically unavoidable, little is known about the influence of this exchange on the sintering behaviour and processability of LLZO powders via wet-processing routes and on the resulting component properties.
Some encouraging studies already exist that show the possibility of large-scale synthesis of LLZO in air21,22 and component manufacturing via solvent based fabrication routes.4,23–27 Thus, for industrial application of LLZO, air and solvent contributions to the Li+/H+-exchange need to be considered. However, the treatment of the material prior to component fabrication, such as storage conditions or pre-treatment, are often not detailed. Also, while stating the introduction of Li-excess during the synthesis to accompany loss during processing, many studies refrain from reporting the actual chemical composition of their samples during or after processing in favour of reporting just the target composition. Furthermore, the contact time between LLZO and the solvent during the wet-processing is scarcely reported and varies greatly between individual works. Thus, a detailed understanding of the impact of air exposure and processing in solvents is hard to generate from the existing work.
To close this knowledge gap, the change of LLZO during air exposure and in the solvent based processing steps was investigated in this study, including the influence on the electrochemical performance of the resulting components. The LLZO powder obtained after lab-scale synthesis typically requires a wet-milling process to acquire a suitable particle size for tape-casting. However, on industrial scale, suitable particle size distributions might be acquired directly after an optimized synthesis route. To investigate to influence of storage conditions, the milled LLZO powder was compared to powders annealed in air and argon after wet-milling. Both annealed powders were also stored in air and solvent for increasing amounts of time to produce a variety of surfaces by the inevitable Li+/H+-exchange. Combined with surface analysis techniques, this enabled us to obtain a mechanistic understanding of the effect of storage and processing of LLZO in air and solvents. This knowledge was used to improve the processing parameters for tape casting of LLZO resulting in a large improvement of both total ionic conductivity and CCD and pushed it closer towards feasibility for industrial application.
To reduce the particle size for further processing, the obtained powder was milled in ethanol with ZrO2 jar and balls at 1000 rpm for 15 minutes using a planetary ball mill (Pulverisette 7 premium, Fritsch) and subsequently dried at 70 °C for 8 h. All processing steps up to this point have been carried out in ambient air.
Li6.4La3.2Zr1.6Ta0.4Al0.02O12 + xH2O → Li6.4−xHxLa3.2Zr1.6Ta0.4Al0.02O12 + xLiOH | (1) |
LiOH + 1/2CO2 → 1/2H2O + 1/2Li2CO3 | (2) |
These reactions can be reversed with a suitable heat treatment above 673 K (ref. 28) in both air and inert atmosphere. These phenomena have been investigated in great detail for sintered LLZO components. Nevertheless, a detailed analysis of the behaviour of LLZO powder in presence of excess lithium-sources is needed, to gain a better understanding of the influence on the particle surface and consecutively the grain boundaries.
Similarly to LiOH formed during the protonation of LLZO, the excess LiOH can also react to Li2CO3 in presence of CO2viaeqn (2) and will be labelled as Li2CO3(ex) in the following. To reverse this reaction and re-form the excess LiOH, ambient water needs to be present. However, in inert atmosphere no water is present and only a direct decomposition of Li2CO3 according to
Li2CO3 →Li2O + CO2 | (3) |
To investigate the impact of different surfaces produced by the Li+/H+-exchange, a sample matrix using the same starting material which is then exposed to different storage conditions in air and solvent was designed. As starting material, calcined and milled LLZO powder (see Experimental part for the details) was divided into three parts. The first part was directly used for fabricating tape cast samples and thus labelled as “milled”. From the other two part, one was annealed in air (assigned further as “air0”) and one in argon atmosphere (assigned further as “argon0”) for 10 h at 750 °C.
Furthermore, to increase the impact of storage in air and solvent, parts of each of the two annealed subsets were stored in air and immersed in ethanol for increasing amounts of time. This experimental matrix, compiled in Fig. 1, allows a systematic assessment of the impact of both ambient air and solvent onto the processability via tape casting and electrochemical performance of the final component.
To analyse the total Li2CO3 content and LLZO surface composition after milling and assess the impact of annealing in Ar and air, we used Raman spectroscopy (Fig. 2). In contrast to X-ray diffraction, which does not show significant difference between as-milled and annealed powder samples, Raman spectroscopy is very sensitive to Li2CO3 and various LLZO crystal phases (cubic vs. tetragonal) and probes only the surface of the particles as the information depth is only several nm. The as-milled LLZO powder (Fig. 2 blue) shows a significant Li2CO3 peak at 1090 cm−1. From previous studies,7 we expect LLZO to undergo a significant proton exchange during high-energy milling in ethanol, the formation of protonated LLZO and lithium hydroxide, as described in (1). Subsequently, during drying in air, lithium carbonate forms from the freshly produced LiOH as well as the excess LiOH, as described in (2). The spectrum of the sample annealed in Ar (Fig. 2 purple) shows good agreement with the high-temperature cubic garnet phase, whereas the milled sample (Fig. 2 blue) and the sample annealed in air (Fig. 2 green) show significant changes of the spectrum between 100 cm−1 and 800 cm−1, that cannot be attributed to the formation of Li2CO3. These spectra indicate the presence of the low-temperature cubic LLZO phase due to the protonation and additional adsorption of CO2 and subsequent incorporation of CO32− into the crystal lattice.29,30,35 Besides a much lower ionic conductivity than the high temperature cubic phase, it is also more ordered and could lead to differences in the interaction of solvents and dispersants in the subsequent wet-processing route.
![]() | ||
Fig. 2 Raman spectra of powders with different pre-treatments in comparison to sintered and polished cubic LLZO34 and commercially available Li2CO3. |
During the annealing in air (Fig. 2 green), the overall amount of Li2CO3 is greatly reduced, since Li2CO3(LLZO) can react back to LLZO. Li2CO3(ex) can also react back to LiOH, using atmospheric water as a reaction partner. CO32− is also removed from the LT-cubic phase of LLZO during the annealing. However, since the annealing takes place in a crucible in open but stagnant atmosphere, both the CO32−-stabilized low-temperature cubic phase of LLZO and some Li2CO3(LLZO) and Li2CO3(ex) can be formed again during cooling. However, as the intensities are much smaller compared to the milled powder, they reform to smaller degree as the reaction time is greatly decreased.
During the heating in argon atmosphere (Fig. 2 purple), the partial pressure of water and CO2 is negligible. Thus, Li2CO3(LLZO) can react back according to (2) and (1) to reform LLZO by removing protons from the structure and emitting CO2 and H2O. However, the Li2CO3(ex) cannot easily react back according to (2) to reform LiOH, as no water is available from the oven atmosphere and the direct thermal decomposition of Li2CO3(ex) according to (3) does not occur at the temperatures chosen in this study.
Overall, this leads to the much higher signal for Li2CO3 observed in the Raman spectrum of the Ar-annealed sample (Fig. 2 purple). Furthermore, CO32− is again removed from the LT-cubic phase of LLZO, but in contrast to the annealing in air, the partial pressures of CO2 and H2O remain low over the course of the treatment. Thus, during the cooling, CO2 preferably reacts with the excess LiOH, rather than reincorporating into the crystal structure of LLZO preserving the HT-cubic LLZO structure observed in Fig. 2 green. In summary, the Ar-annealed sample shows a higher amount of Li2CO3(ex), but the surface of the particles is very close to the high temperature cubic phase.
To obtain more detailed information about the surface composition of the particles, X-ray photoelectron spectroscopy (XPS) was employed on the same powders and the total amount of surface species extracted (Fig. 3). Verifying the Raman results, the comparison of the air-to the Ar-annealed powder (Fig. 3 green and purple, respectively) shows increasing amounts of C due to the higher amount of Li2CO3. Interestingly, for La and Zr the XPS measurements show a higher value for the milled powder, whereas the Ar and air annealed powders are about the same. Since La and Zr signal are proportional to the bulk LLZO measured via XPS, this indicates that the milled powder has a thinner or incomplete coating with LiOH and/or Li2CO3 compared to the air- or Ar-annealed ones. The most plausible explanation of this effect is, that the excess LiOH and Li2CO3 formed after milling does not homogeneously cover the surface but forms small particles itself (e.g. see schematic representation in Fig. 9, second column). During the high temperature treatment at 750 °C both LiOH (Tm = 426 °C) and Li2CO3 (Tm =720 °C) melt, coat and partially connect the particles (Fig. 9, third column). From the C signal in XPS and the Raman measurements we can now confirm that the composition of these coatings is LiOH-rich for air annealed powders and Li2CO3-rich for the Ar annealing. It can be expected that these coatings on the particle surface and their different compositions will have a significant effect on their behaviour in a subsequent wet-chemical processing step and final component performance.
![]() | ||
Fig. 3 Absolute atomic concentration of the major elements as determined via XPS for the milled powder (blue), the air annealed (green) and the Ar annealed (purple) sample. |
The rheological behaviour is shown in Fig. 4. All starting powders (milled (blue), air0 (green) and argon0 (red)) showing shear-thinning behaviour. When getting into contact with the solvent during preparation of the slurry, fast protonation of the surface is expected, as LiOH is formed and consecutively dissolved in the surrounding solvent as proposed by R. Kun et al.20 In our work, the effect of the excess LiOH and Li2CO3 on the particle surface (as determined by Raman and XPS) on the protonation in the solvent and the resulting rheological behaviour needs to be considered. On the one hand, the excess LiOH dissolves very well in ethanol, on the other hand, Li2CO3 has a very low solubility, which means it will remain in its original state and place upon contact with the solvent. From the variation of the viscosity in Fig. 4 in the shear rate range of 0.01 to 1 s−1 we can see there is clear correlation between the Li2CO3 content of the sample, as determined by Raman and XPS measurements, and the viscosity of the slurry. The as milled powder has the highest surface area uncovered by excess material, which has already been protonated by the previous exposure to the solvent and therefore allows fast attachment and full coverage of the dispersant, resulting in the lowest viscosity. The air-annealed sample (air0) is mainly coated by LiOH which can also be dissolved by the solvent. The subsequent protonation of the free particle surface allows for still reasonable coverage of the dispersant and results in medium high viscosities. The Ar-annealed powder (argon0) features a Li2CO3-rich coating of the particles, which has a much lower solubility than LiOH and thus only allows poor coverage with the dispersant (see Fig. 9 fourth column).
In addition to the strong impact of the surface pre-treatment of starting powders on their rheological behaviour, further significant changes in the rheological behaviour are observed when the LLZO powders are stored for different time both in air and a solvent.
When comparing the behaviour of the freshly annealed powders to the ones stored in air for prolonged times, a clear change can be observed for the air-annealed powders stored in air (Fig. 5a) and the Ar annealed powders stored in air (Fig. 5b). The air-annealed powder shows a decrease in viscosity, caused by the slow protonation of the LLZO surface when in contact with ambient air and the subsequently better coverage with dispersant in the slurry. After one day of exposure to ambient air, the obtained viscosity is close to the viscosity of the as-milled slurry, which means the protonation of the surface is almost complete.
The behaviour or the Ar-annealed powders is less linear. As the partial pressures of water and CO2 are greatly lowered in inert atmosphere, short exposure to air causes the unreacted species obtained after Ar annealing to react with ambient moisture and CO2 to reach equilibrium. As a result, the thicker, insoluble layer of Li2CO3 covering the particles allows for less protonation of LLZO. Therefore, these particles are less covered with dispersant, causing an increase in viscosity of the obtained slurry. With prolonged exposure to ambient air, protonation of the LLZO surface through the Li2CO3 layer can take place, allowing for a better coverage with dispersant in the slurry. Thus, the viscosity of the obtained slurry is vastly lowered (Fig. 5b light red). After 7 days in air, the viscosities of air-annealed and Ar-annealed powders are almost the same.
For the storage in solvent, similar trends can be observed. The LLZO annealed in air with prolonged storage time in the slurry (Fig. 5c) shows a significant decrease in viscosity already after only 1 day in ethanol (green), while a longer storage time does not change it much further (light green). Again, this can be explained by ethanol dissolving the LiOH predominately covering the particle surface, exposing fresh particle surface. This saturates the solvent with OH−, which is confirmed by the measured pH values of over 13 for all samples. The dispersant can then attach to the fresh particle surface, leading to the observed drop in viscosity. For Ar-annealed powder, the behaviour for prolonged storage in ethanol (Fig. 5d) is much different and resembles the storage in air (Fig. 5b). In contrast to the air annealed sample, the particles are covered by a much larger amount of Li2CO3, which will not dissolve as easily in ethanol. Thus, after the initial increase in viscosity, the dissolution of LiOH and Li2CO3 covering the surface and subsequent drop in viscosity due to improved attachment of the dispersant is much slower. After 7 days, the behaviour resembles the one for air annealed sample as after 1 day storage in air. In summary, the excess LiOH and Li2CO3, covering the surface after annealing, greatly affects the rheological behaviour of the slurry. Common variation in the recipes like homogenizing the slurry on a roller bench for 24–72 h can result in large deviations in castability, depending on the storage and pre-treatment of the original powder. This poses challenges regarding the variation and control of time scales in the manufacturing processes, demonstrated exemplary in Fig. 6. Fig. 6a shows the green tape cast with freshly Ar-annealed powder (argon0) having only minor defect on the edges of the tape due to surface tension based flow during drying. After 1 day of storage in ethanol (argon0,1) the casting results improve, and the green tape shows no defects. After 7 days of storage in ethanol (argon0,7) the green tape shows major defect on the edge due to surface tension based flow and severe coffee-staining. Defects of such severity do not allow for further processing of this tape.
![]() | ||
Fig. 6 Casting behaviour resulting from different rheological profiles (a) argon0 (b) argon0,1 (c) argon0,7. |
This difference in casting behaviour can of course be compensated (within limits) by changing the composition of the slurry. However, it clearly demonstrates the importance of a detailed understanding of the influence of the pre-processing and storage on the particle surface and its effect on the slurry viscosity to predict the resulting changes in casting behaviour. Additionally, the impact of these changes in particle surface on the properties of the sintered sample needs to be investigated.
![]() | ||
Fig. 7 Top: picture of successfully sintered LLZO components bottom: corresponding cross-section pictures by SEM. |
Sample | Pre-treatment | Storage | Relative density [%] | Conductivity 25 °C [S cm−1] | Critical current density 50 °C [μA cm−2] | Li2CO3 content |
---|---|---|---|---|---|---|
Argon0 | Argon | — | 91.2 | 1.89 × 10−4 | 5.1 | + |
Argon0,1 | Argon | 1 day ethanol | 88.1 | 2.20 × 10−4 | 35.7 | + |
Argon1 | Argon | 1 day air | 88.2 | 2.66 × 10−4 | <5.1 | ++ |
Argon7 | Argon | 7 days air | 88.8 | 2.75 × 10−4 | 114.6 | +++ |
Argon1,7 | Argon | 7 days ethanol | — | — | — | +++ |
Air0 | Air | — | 92.8 | 3.90 × 10 −4 | 318.5 | 0 |
Air0,1 | Air | 1 day ethanol | 91.5 | 2.08 × 10−4 | 35.7 | + |
Air1 | Air | 1 day air | 89.0 | 1.19 × 10−4 | 127.4 | ++ |
Air7 | Air | 7 days air | 79.5 | 7.87 × 10−6 | — | ++ |
Air1,7 | Air | 7 days ethanol | — | — | — | +++ |
Milled | As-milled | — | 89.4 | 1.09 × 10−4 | 127.4 | +++ |
Due to the low thickness and high conductivity of the samples, the resolution of bulk and grain boundary contributions was not possible at room temperature. Therefore only total ionic conductivities of the components can be reported, taken from the low frequency intercept of the semicircle corresponding to the onset of the blocking electrode behaviour of the sputtered gold electrodes (Fig. 8, left).
![]() | ||
Fig. 8 Left: impedance spectrum of air0 at room temperature right: CCD measurement of air0 at 50 °C. |
No direct correlation between the relative density and conductivity can be drawn from the measurements summarized in (Table 1). Samples obtained from freshly air-annealed LLZO powder show the highest total conductivity of 3.90 × 10−4 S cm−1 at 25 °C. With prolonged storage in air the amount of LiOH decreases in favour of Li2CO3 on the particle surface, resulting in a lower conductivity of the resulting sample. The negative effect of Li2CO3 is even more pronounced for the freshly Ar-annealed sample. Yet, prolonged storage in air improves the conductivity of Ar-annealed samples, as the onset of LLZO-protonation increases the LiOH content. Prolonged storage in the slurry reduces the conductivity of samples obtained from air-annealed powders. Nevertheless, protonation by the solvents seems to have a less disadvantageous influence on the conductivity than the reaction with ambient air, as the subsequent reaction to Li2CO3 is inhibited. The highest ionic conductivity obtained in this work of 3.90 × 10−4 S cm−1 at 25 °C for the components prepared from freshly air-annealed LLZO powder, without the use of sintering aids, is the highest value reported in literature so far.
These findings highlight the importance of a detailed understanding of the influence of the pre-processing and storage on the particle surface and its effect on the ionic conductivity of sintered components.
The critical current densities (CCD) that can be achieved with LLZO separators and Li anodes without failure caused by the development of a lithium dendrite are an important indicator of their performance. To determine CCDs for different LLZO separators prepared in this study, Li plating and stripping experiments were performed on symmetric Li‖LLZO‖Li cells. To improve the wetting of the LLZO with lithium and to minimize the Li/LLZO interfacial resistance,17 a sputtered Au interlayer was introduced to ensure comparable interface resistances amongst all samples. Each sample was characterized via impedance spectroscopy prior to the CCD measurements. Only in the case of argon1, extremely high values for the interface resistance clearly lead to a low CCD. All other samples show similar interfacial resistances (ESI Table 1†), which can be ruled out as the reason for their differences in CCD. Thus, the differences in their CCDs are mainly governed by the differences in the total ionic conductivity of the sample. At low current densities, most samples show flat voltage plateaus, indicating a uniform plating and stripping of the metallic lithium. With higher currents, the growth of a dendrite is marked by an increase in voltage, followed by a sharp decrease, marking the critical failure (Fig. 8, right).
No linear connection of the critical current density to the total ionic conductivity can be derived from our data (Table 1). Since the bulk conductivity should be similar for all samples, a detailed investigation of the dependency of the CCD on the grain boundary structure and composition of LLZO should be undertaken in the future. Still, the overall trends found this far are also valid for the critical current density measurements. Freshly-air annealed powder produces samples with the high CCDs of 0.32 mA cm−2 at 50 °C. Prolonged storage in air causes the protonation of the LLZO surface with subsequent reaction to Li2CO3, causing a decrease in the CCD of the sintered sample. The obtained values match those of the as-milled samples. Sintered samples obtained from freshly argon-annealed powder show very low CCDs, as again the coating of the particles with Li2CO3 negatively influences the grain boundary of the sintered component. Storage of the argon-annealed powder in air for a short time worsens the CCD to <5 μA cm−2, as LiOH remaining on the particle surface after annealing forms additional Li2CO3. Prolonged storage in air allows for the protonation of the LLZO surface, resulting in an improvement of the CCD. The CCD of samples from as-milled powders is similar to the values for both air-annealed and argon-annealed samples after prolonged storage in air.
Prolonged storage in the slurry of the air-annealed powder reduces the CCD of the sintered component vastly, as proton exchange is expected to be faster in the solvent than in air. Yet, prolonged storage in the slurry of the argon-annealed powder improves the CCD, as the protonation is faster in the solvent and the subsequent reaction with CO2 is inhibited.
The highest critical current density of 0.32 mA cm−2 was measured for the separator produced from freshly air annealed LLZO powder, which had the shortest exposure times with both air and solvents. Unfortunately, the prevailing majority of publications deal with pressed and sintered LLZO pellets, which are not suitable for the large-scale fabrication of solid state batteries. Most publications on tape-cast LLZO do not report plating-stripping results.22,24 Even so the critical current density obtained for our optimized tape cast membranes is among the highest reported for thin freestanding tape-cast LLZO separators. The only comparable work on tape cast LLZO films by Hitz et al.4 deals with 3D structured separators. Due to a higher contact area of their electrodes the higher total CCD value of the component is achieved. However, the equivalent CCD of their material estimated from the total current and the specific surface area of the porous electrodes is only around 0.25 mA cm−2, which is significantly lower than the value obtained in our work. It can be therefore expected that a combination of 3D structuring that was shown to be beneficial for the cell performance, with the optimized processing route developed in this work would further increase the critical current density of LLZO separators to reach values relevant for industrial application.
Footnote |
† Electronic supplementary information (ESI) available. See DOI: 10.1039/d0ta11096e |
This journal is © The Royal Society of Chemistry 2021 |