Improved conductivity in tape casted Li-NASICON supported thick films: Effect of temperature treatments and lamination

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

Journal of the European Ceramic Society journal homepage: www.elsevier.com/locate/jeurceramsoc

Original Article

Improved conductivity in tape casted Li-NASICON supported thick films: Effect of temperature treatments and lamination ⁎

R. Jiméneza, , A. del Campoc, M.L. Calzadaa, J. Sanza, S.D. Kobylianskab, B.O. Liniovab, A.G. Belousb, А.V. Ragulyad a

Instituto de Ciencia de Materiales de Madrid, ICMM-CSIC. C / Sor Juana Inés de la Cruz 3, 28049, Cantoblanco, Madrid, Spain Vernadsky Institute of General and Inorganic Chemistry, National Academy of Sciences of Ukraine, Palladine Ave 32-34, 03680, Kyiv, 142, Ukraine c Instituto de Cerámica y Vidrio, ICV-CSIC. C / Kelsen 5, 28049, Cantoblanco, Madrid, Spain d Frantsevich Institute for Problems in Materials Science of National Academy of Sciences of Ukraine, Krzhizhanovsky Str 3, 03680, Kyiv, Ukraine b

A R T I C L E I N F O

A B S T R A C T

Keywords: Tape casting Sintering Li- NASICON thick films Confocal Raman Solid state Li batteries

In this work, the influence of different thermal sintering treatments on Li1.3Al0.3Ti1.7(PO4)3 NASICON thick films has been investigated. The isostatic lamination step performed before the thermal sintering of thick films has demonstrated to improve film density and grain size, increasing "bulk" and grain boundary Li-conductivities. The confocal Raman spectroscopy characterization allowed the observation of the connectivity of the particles present in the ceramic samples and so a deeper understanding of ionic conductivity. The dependence of total and "bulk" Li conductivity on the samples microstructure is discussed. The films sintered by slow heating sintering with a previous lamination step, displayed an overall Li- conductivity > 10−4 Ω-1 cm-1, that is superior to that reported in commercial OHARA Li- NASICON glass ceramics. The tape casting deposition method is scalable for preparation of large area thick supported electrolyte films with high conductivity for novel Li ion all solid state batteries (ASSB) architectures.

1. Introduction Nowadays lithium-ion batteries are widely used for storage power generation and uninterruptible power supply (UPS) systems due to its high energy performance [1–3]. However, using Li-conducting organic liquid or polymer electrolyte prevents the manufacture of completely safe devices. Replacement of the organic electrolyte by an inorganic solid will not only significantly improve the safety of lithium-ion battery, but extend its life by reducing the degradation processes [4–8]. The rather poor temperature stability of organic electrolyte based Libatteries makes that this storage technology cannot be applied in harsh environments where the temperatures can rise above of 100 °C. The use of ionic liquids attenuate this limitation but ionic conductivity decreases, reducing the power of the battery. The implementation of all solid state batteries (ASSBs) can be of interest for applications in relative high temperatures and wide range of pressures that current liquid electrolyte based batteries (LEBBs) cannot cover. As a result of progressive miniaturization of electronic components there is an increasing demand for micro-sized power sources that incite the research of thin or thick films for all-solid-state batteries. Smart cards, implanted medical devices, micro electro-mechanical systems,

⁎

memory blocks, sensors, transducers, and specially military equipment are potential consumers of ASSBs film batteries [9–11]. However, there are many drawbacks that limits their implementation, among them the relatively low ionic conductivity of the solid, the difficulty in obtaining coherent electrodes/solid electrolyte interfaces and mechanical fatigue produced on interfaces during insertion de-insertion process. In ASSBs solid electrode/electrolyte contacts are limited and the interfacial area is lower. The diffusion paths for the Li insertion and electron collection from the electrodes are much larger than in conventional secondary batteries (composite electrodes). One solution was the implementation of ASS thin films batteries. Many designs have been made in order to increase the interface area. The thin film concept has reached high specific numbers and high charge and discharge rates comparable to liquid electrolytes Li batteries but the total energy stored remains poor [12,13,14]. In order to increase the energy stored, the active mass (electrodes) must be increased, leading to the thick films technology with thicknesses in the range 10–100 μm, that provide higher specific energy and power than thin film electrodes. The need of reducing the diffusion path of the ions and electrons in ASSB and the minimization of the strain effects due to the insertion/de-insertion processes requires important changes in battery designs. Electrolyte supported ASSBs

Corresponding author. E-mail address: [email protected] (R. Jiménez).

https://doi.org/10.1016/j.jeurceramsoc.2017.12.017 Received 3 August 2017; Received in revised form 1 December 2017; Accepted 9 December 2017 0955-2219/ © 2017 Elsevier Ltd. All rights reserved.

Please cite this article as: Jiménez, R., Journal of the European Ceramic Society (2017), https://doi.org/10.1016/j.jeurceramsoc.2017.12.017

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

Prepared Li1.3Al0.3Ti1.7(PO4)3 nanoparticles were mixed with organic reagents to prepare the slurry for "tape casting" thick films deposition. The composition of the slurry was 22 wt.% acetylacetone, 2 wt.% isopropanol (solvents), 2 wt.% polymethyl-methacrylate (binder), 30 wt.% dibutyl phthalate (plasticizer), 9 wt.% dybutyl fosfate and 5 wt.% hallotannine (dispersants). The ratio of organic reagents to Li1.3Al0.3Ti1.7(PO4)3 nanoparticles was 70:30 wt.%. For preparing the slurry and to promote homogenization, a planetary mill Fritsch Pulverisette 7 was used for 2 h at a rotational speed of 300 rev/min. For deposition of thick films slurry mixtures, an Applicator and a Drying Time Recorder Coatmaster 510 (Erichsen, Germany) were used. The rate of deposition of the slurry on the substrate (polikor α-Al2O3) was 0.1 mm/s. After drying in air, the film thickness was 120 μm and the deposited area 15 × 15 mm2. Thick-film materials and substrate were subjected to different heat treatments, reaching the maximum temperatures of 1000 / 1100 °C. In some cases the Li1.3Al0.3Ti1.7(PO4)3 crude films were subjected to vacuum degasification under a pressure of 0.1 MPa for 60s in a Original HENKELMAN vacuum system (Jumbo 30). The evacuated films were introduced in a sealed flexible container, and then laminated in a IL4008PC isostatic laminator apparatus. Prepared films with and without isostatic lamination, were annealed under different conditions. In particular, five samples were produced:

allow specific electrode configurations to balance the electrode mass between the anode and cathode, minimize the strain effects and reduce the diffusion path. The electrolyte supported ASSBs can be foreseen in two different configurations with the anode and cathode at different sides of the electrolyte or disposed at the same side of the electrolyte. The first configuration implies a self-supported electrolyte thick-film with high enough mechanical strength, the second one allows the use of a supported solid electrolyte layer on a refractory mechanically resistant substrate. For both architectures the procurement of planar large area electrolyte thick films is of main importance in order to build large area electrolyte supported ASSBs devices with enough energy and power. The lithium Li1.3Al0.3Ti1.7(PO4)3 conductor, displaying the NASICON structure, is a promising material for producing solid electrolytes because of high lithium conductivity (σbulk ∼ 10−3 Ω−1 cm−1 at RT), chemical resistance and mechanical stability [15–19]. These materials are widely studied in the form of glass ceramics; however, studies on films are still scarce. The optimal methods to obtain thick films with a thickness of more than 10 μm are the "tape casting" or screen printing [20]. The presence of pores and cracks have a negative impact on the electrical properties of films; therefore, it is necessary to find ways to increase the density of thick films at preparation. By using appropriate heat treatments or by reducing the deformation of the surface (wet film lamination), density of films can be increased [20–22]. Isostatic lamination is one of the most important processing techniques for improving quality of multilayer ceramic three-dimensional structures. It is known that isostatic lamination, when pressure and temperature are applied to green tapes simultaneously, improves the particles packing and density of green tapes, providing uniform films with reduced internal stresses, and without cracks in laminated films. The aim of this work is the preparation of high Li-conducting thick films of Li1.3Al0.3Ti1.7(PO4)3 (σoverall ≥10−4 Ω-1 cm-1 at RT) by tape casting on insulator substrates. Due to the large area to volume ratio and to the rather large temperatures used in the preparation of the thick films, the microstructures of the solid electrolytes differ from that obtained in the “bulk” ceramics. The microstructure and phase homogeneity of prepared thick films has been investigated with SEM, and confocal micro-Raman techniques. This characterization will provide the spatial distribution of phases in prepared electrolytes. The connectivity between solid electrolyte particles and the distribution of secondary phases are fundamental to explain ionic conductivity in resulting thick films. Total and bulk conductivity will be compared with that reported in other samples of similar composition, specially the OHARA Li- NASICON (OH-NAS) glass-ceramic, that is the unique commercial large area Li solid electrolyte available.

- Sample 1 (TFS1).-"Thermal shock" (the film was preheated in a oven to 500 °C, then heated from 500 °C to 1000/1100 °C at the heating 60 °C/h rate. - Sample 2 (TFS2).-"Slow heating" (the film was heated at 20 °C/h to 500 °C, then from 500 °C to 1000/1100 °C at 60 °C/h). - Sample 3 (TFS3).-An "Isostatic lamination" step was applied before "Thermal shock". - Sample 4 (TFS4).-An "Isostatic lamination" step was applied before "Slow heating". - Sample 5 (TFS5).- "Constant speed heating" (the film was heated from room temperature to 1000/1100 °C at 60 °C/h). X-ray diffraction patterns of resulting compounds were recorded in a DRON-3M (CuKα-radiation; Ni-filter) apparatus, with a goniometer rotation speed 0.5 deg/min. XRD diffraction patterns were recorded in the range 2θ= 10-60°, with 0.01° steps of 1 s/per step. The thick films compositional homogeneity was tested by using the confocal Raman microscope (Witec alpha-300R, WITec GmbH; Ulm, Germany), using a 532 nm excitation laser and a 100x objective lens (NA = 0.9). The incident laser power was 0.5 mW. The optical diffraction resolution in confocal Raman apparatus microscope was limited to ∼200 nm laterally and ∼500 nm vertically. Resolution achieved in recorded Raman spectra was 0.02 cm−1. The sample was mounted on a piece-driven scan platform, displaying a 4 nm lateral and 0.5 nm vertical positional accuracy. Collected spectra were analyzed by using the Witec Control Plus Software. Cross-section and plan-view micrographs of the crystalline oxide films were obtained by field-emission gun scanning electron microscopy (FEG-SEM, Nova Nanosem 230 FEI Company equipment, Hillsboro, OR). In supported ceramic thick films, the volume fraction of functional phase cannot be determined using many of the methods applied in "bulk" ceramics. The α-Al2O3 support used is 1 mm thick, the weight of the thick films is around the 2% of the total preventing the use of Arquimedes method. Mercury intrusion porosimetry is a destructive technique so it is discarded. In this work digital image analysis of SEM micrographs of oxide films plan-view has been used to deduce porosity volume fractions in thick films. Apart of the intrinsic error due to the image conditions and the calculations, due to its local nature, this method can lead to larger error in the volume fraction estimation in heterogeneous samples. As the confocal Raman spectroscopy also gives information about the volume fraction of phases in the thick films, a

2. Experimental Starting Li1.3Al0.3Ti1.7(PO4)3 nanopowders used as precursor for films were prepared by sol-gel method. Li2CO3 (99,99% Merck), Al (NO3)3•9H2O (99,97%, Sigma-Aldrich), diisopropoxytitanium bis(acetylacetonate) solution (∼75% in isopropanol) (purum, Sigma-Aldrich), titanium diisopropoxide bis(acetylacetonate) C16H28O6Ti, phosphoric acid 85% H3PO4 (all of high purity grade), citric acid C6H8O7 and ethylene glycol C2H6O2 were used as starting reagents. To prepare aqueous solutions of salts of lithium nitrate LiNO3 and aluminum nitrate Al(NO3)3, Li2CO3 and Al(NO3)3•9H2O were dissolved in nitric acid. Citric acid was dissolved in ethylene glycol at a ratio of 1:4. Both solutions were mixed and stirred at 70 °C for 1 h. Then diisopropoxytitanium bis(acetylacetonate) solution with phosphoric acid were added. For homogenization and to promote the esterification and polymerization between ethylene glycol and citric acid, the solution was stirred for 12 h at 100 °C to produce gels. The polymer gel was heated on a sand bath at 350 ± 10 °C (5 h) and calcinated at 700 °C (2 h) to obtain Li-NASICON powder particles. 2

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

preparation method.

good agreement between both volume fraction and porosity estimations support the deduced values and the micro-structural homogeneity of the samples. Electrical properties of thick films as a function of temperature (77–575 K) were investigated by Impedance Spectroscopy (IS) in the frequency range 20Hz-1MHz, using an Agilent LCR E4192A apparatus. For low frequency measurements (1mHz–10 kHz), a Zhaner IM6ex apparatus was used. In-plane measurements were performed to characterize supported electrolyte samples. Two stripe electrical contacts were deposited on the surface of electrolytes supported thick films by Au sputtering through a shadow mask. For OH-NAS sample, a planar capacitor configuration was used for electric measurements. Samples of 5 mm x 0.5 mm x 250 μm were cut from a two inch wafer, then 4 mm diameter gold electrodes were deposited by d.c. sputtering through a shadow mask onto opposite sample faces.

3.2. Microstructural study TFS5 film heated with a "constant speed" of 60 °C/h displayed a large number of pores, very small volume fraction of particles (∼ 0.2), cracks and de-lamination defects, from this fact, this treatment has been discarded hereafter. Fig. 2 shows the surface of Li1.3Al0.3Ti1.7(PO4)3 films (12 × 12 μm2) and the cross-section micrographs (inset) obtained after various heating rates with and without previous lamination treatments. To estimate the volume fraction occupied by grains in thick films, digital image treatment of larger scanned area micrographs (40 × 40 μm2) were used. From the surface micrograph of TFS1 sample (Fig. 2a), the volume fraction calculated was 0.45 ± 0.04. The TFS2 sample (Fig. 2b) display a volume fraction of 0.68 ± 0.04, decreasing considerably the pores volume. Lamination treatments before sintering produce important changes on the films microstructure. Previous lamination before thermal shock increases in a significant way the volume fraction occupied by particles from 0.45 ± 0.04 to 0.61 ± 0.04 (Fig. 2c). The lamination of the sample before sintering by slow heating, increases the volume fraction to a value of 0.72 ± 0.04 in TFS4 sample (Fig. 2d). The thickness of the samples is illustrated as insets in surface micrographs. In all cases, film thickness are near 30 ± 5 μm except for the TFS1 that presents uneven thickness with an estimated average value of 70 ± 5 μm. In micrographs, it was observed an increment on the grain size of the Li-NASICON grains caused by the lamination process. The TFS1 presented grain sizes below 1 μm. The previous lamination increases the grain size between 1–2 μm and reduces the porosity in TFS3 sample. The TSF2 presented a larger grain size compared to the TSF1 and a lower porosity. For this sample the grain size was between 1–2 μm. The isostatic lamination produced an important increase of the grain size to values between 3–4 μm in the TFS4 sample. From the microstructural study it can be concluded that best performances are displayed by the TFS4 sample, submitted to slow heating for sintering after lamination.

3. Results 3.1. XRD study Diffraction patterns of Li1,3Al0,3Ti1,7(PO4)3 thick films are displayed in Figs. 1a and b. As shown in Fig. 1a, Li1,3Al0,3Ti1,7(PO4)3 thick-film (TFS5) patterns recorded at 1000 °C are characteristic of a single-phase, with a tiny contribution from the α-Al2O3 substrate. The increment of the sintering temperature to 1100 °C leads to the detection of an additional TiO2 peak, that can be caused by the partial evaporation of lithium from NASICON's phases. Based on this, the sintering temperature used in thick films manufacture was limited to 1000 °C. Fig. 1b shows Li1.3Al0.3Ti1.7(PO4)3 films sintered at 1000 °C with and without lamination step. All patterns seem to be single phase regardless of the

3.3. Electrical characterization Conductivity measurements were performed using two co-planar electrodes, following the procedure described in a previous paper [23]. Deduced results are given in pF/m and normalized to the intersection length of stripe electrodes. The conductivity of the samples was deduced from the immittance function that is directly related to the conductivity, the real part of the admittance. In Fig. 3, the frequency variation of the real part of the complex conductivity of four samples is given for different temperatures (170, 220, 300 K). In the last plot, the frequency dependence of OH-NAS film conductivity was included for comparison. High conductivity values of Li-NASICON films required the sample cooling to detect the "bulk" response in the frequency window used (20 Hz–2 MHz) in electrical measurements. The sample TFS4 exhibits the best conductivity, TFS2 and TFS3 displaying lower values, and TFS1 the worst conductivity. From collected data, bulk and the grain boundary conductivity were deduced using the derivative criterion [24]. Conductivity results are expressed as a function of the inverse temperature in Ahrrenius plots of Fig. 4. It was not possible to obtain reliable values of the bulk conductivity in sample TFS1 due to the strong convolution of "bulk" and grain boundary responses. In all cases, activation energies for grain boundary contributions are larger than the corresponding bulk ones. In Table 1, the samples conductivity at two representative temperatures, and activation energies for bulk and grain boundary responses are displayed.

Fig. 1. XRD patterns of Li1,3Al0,3Ti1,7(PO4)3 thick films: а) “heating with an average speed” and sintering at different temperatures. 1 - 1000 °C, 2–1100 °C; b) Different speed of heating of films: 1 – “thermal shock”, TFS1;2 – at 20 °C/h, “slow heating”, TFS2; 3 – lamination of film and “thermal shock”, TFS3; 4 – lamination of film and “slow heating”, TFS4. The main peaks of NASCON phase are indexed in both figures. * reflections form αAlumina substrate, + reflections from rutile (TiO2).

3

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

Fig. 2. The microstructure of the surface films and of the cross section (inset) obtained by treating different conditions: а) TFS1 (thermal shock), b) TFS2 (slow heating).c) TFS3 (lamination step and thermal shock). d) TFS4 lamination step and slow heating.

4. Discussion

evaporation processes and degradation of organic compounds, instantly producing the pyrolysis of organic components [26,27]. Owing to the rapid pyrolysis of organic components the inorganic carcass remain. It results into a porous structure of the TFS1 film, where particles are far from each other and where the mechanism of inter-bubble coalescence doesn’t work. Consequently, size of grains after sintering are between 0.6 – 1 μm (Fig. 2a). In the case of “slow heating” (20 °C/h to 500 °C) organic solvents (isopropanol and acetylacetone) evaporate gradually, and the polymer matrix degrades progressively, avoiding cracks in the surface of TFS2 films. The grains size obtained after sintering at 1000 °C increased up to 1.5 μm (Fig. 2b), as a result of the shrinkage of the film by capillary phenomena and the sintering of inorganic particles by inter-bubble coalescence [28]. The use of "isostatic lamination" on the microstructure of thick films decreased films porosity (Fig. 2c and d). This was the result of the residual compressive stresses applied at outer layers of films [29]. The heating of the polymer matrix where inorganic particles are incorporated, produces the partial destruction of the matrix and the sintering of inorganic particles. The analysis of the microstructure of laminated Li1.3Al0.3Ti1.7(PO4)3 films after "slow heating" TFS4 and "thermal shock" TFS3 (Fig. 2c and d) showed that the isostatic lamination promoted the increment of the grain size. The analysis of the grains morphology showed that the maximum grain size, 4 microns, was achieved after 20 °C/h heating till 1000ºC ("slow heating")

4.1. Thick film formation It is known, that thermal treatments used for preparation of films differ from those used for ceramic samples, in the first case thermal treatments must be carried out in several stages [25]. A first stage is required to remove organic components (annealing of film), and a second one for grains sintering (sintering of films). The first stage is particularly important because green tape contains 70% of organic compounds. Annealing of films is formed by two processes: 1) the solvent evaporation from the surface of the film; and 2) diffusion of the solvent from volume to surface. Here the limiting process is diffusion phenomena. Ideally, the rate of evaporation of solvents from the surface should be equal to the diffusion rate: vev. = vdiff.. When the rate of evaporation of solvents is faster than rate of diffusion (vev. > vdiff.), cracks are formed, and when rate of diffusion is faster (vev. < vdiff.) the surface will be covered by a thick crust that strongly affects total ion conductivity. Therefore it is very important to use annealing temperatures for which vev. = vdiff; as a result of the uniform evaporation of the solvent, small volume porosities are produced. In the case of films obtained after "heating with a constant speed" (TFS5) the resulting microstructure was crumbled making this film useless as solid electrolyte. In order to optimize annealing processes, “thermal shock” and “slow heating” treatments were used. The "thermal shock" treatment avoids 4

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

Fig. 4. D.C. Conductivity results. “Bulk” response (Hollow symbols). Grain boundary response (filled symbols). The lines are the corresponding linear fitting to the Ahrrenius law. The resulting activation energies are collected in Table 1. Large hollow triangles in the grain boundary response are the results of the grain boundary conductivity calculation as a function of the observed microstructures. The relation of the symbols with the different samples studied is depicted in the plot.

are related in a different way with the microstructure. The "bulk" response is related to the volume fraction and composition of NASICON grains. On the other hand, the grain boundary response (if not secondary phase are formed at grain/grain interfaces) is related to the contact area between particles and the number of grain boundaries along Li conductivity paths. The "bulk" response of four samples can be deduced from complex conductivity vs. frequency plots recorded at 170 K, Fig. 3a. Applying the derivative criterion, the "pseudo-plateau" minima associated to the bulk contribution were deduced for TFS1, TFS2, TFS3, TFS4 samples: 0.64, 0.25, 0.34, 0.20 respectively. The minimum value deduced in the ceramic Li1.3Al0.3Ti1.7(PO4)3 sample was close to 0.21 [30], being 0.23 in the OH-NAS sample (this work). This value is related to the bulk conductivity partially convoluted with the grain boundary response. The closer the conductivity relaxation time of "bulk" and grain boundary responses the higher derivatives result (bad resolution of "bulk" d.c. plateau). In the case of the grain boundary the conductivity time constant depends on the area and thickness of contacts between neighbor particles [31]. Then, large slopes measured in "bulk d.c. contributions" were derived from small contact areas or/and from bottle-necks between grains. Thick grain boundary layers due to depletion of charge carriers can also give raise to large slope values in d.c. conductivity "pseudo-plateaus". Following these arguments large slope values detected have been ascribed to poor inter-particle contacts. In ceramic samples with high porosity, the existence of small volume fractions of the conducting phase, near the percolation threshold, could also produce a similar effect on the shape of the a.c. conductivity "bulk" response [32]. The analysis of the Ahrrenius plot showed the largest "bulk" conductivity value for TFS4 and lower ones for TFS2 and TFS3 samples (Fig. 4). These results are related to changes produced in the volume fraction of the NASICON particles due to different preparation conditions. For example the effective "bulk" conductivity response for an isotropic ceramic pellet, with a volume fraction of 0.6 is 41% of the single-crystal conductivity value; the same estimation for a ceramic with a volume fraction of 0.8, was 71% of the single crystal conductivity [32]. In Fig. 5, the plot of the normalized "bulk" conductivity as a function of the estimated volume fraction of Li- NASICON grains in the thick film samples is included. The black triangles correspond to films studied here with their estimated volume fractions, supposing that the crystallite conductivity is the same despite the thermal annealing used.

Fig. 3. Variation of the a.c. conductivity as a function of frequency for the four samples studied in this work at three representative temperatures . a) 170 K b)220K. c) 300 K. In the later plot the a.c. conductivity of the OHARA Li- NASICON glass-ceramic (OH-NAS) is included for comparison as well as the Nyquist representation of the TFS4 impedance. In the impedance plot the low frequency spike typical of ion blocking at the Au electrodes is observed. At the highest frequencies the impedance does not go to zero indicating that the "bulk" response is out of the experimental frequency window.

(Fig. 2d). 4.2. Conductivity of thick films Different preparation treatments produced a strong effect on the microstructure and electrical properties of films. In TFS4, Li overall conductivity is 1.2 10−4 Scm-1 at 300 K, that is a good result, even superior to that reported on commercial OH-NAS samples (8.8 105 Scm-1 at RT), see Fig. 3c. Changes on the conductivity of the samples must be related to those produced in micro-structure of films. The analysis of conductivity has been performed at two levels: First, information deduced from the "bulk" conductivity response and second, information deduced from the grain boundary response. Both responses 5

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

Table 1 Li conductivity parameters of the thick film samples prepared in this work.

*

Sample

σ"bulk" @ 170K (Ω−1cm−1)

σ"total" @ 300K (Ω−1cm−1)

Ea "Bulk" (eV)

Ea"Grain Boundary" (eV)

TFS1 TFS2 TFS3 TFS4

8 ± (0.8) x10−8* 3 ± (0.1) x 10−7 1.8 ± (0.1)x 10−7 8.7 ± (0.2)x 10−7

1.4 ± (0.08)x 10−6 6.2 ± (0.1) x 10−5 2.8 ± (0.09) x 10−5 1.2 ± (0.08) x 10−4

——— 0.27 ± (0.003) 0.27 ± (0.003) 0.24 ± (0.002)

0.29 0.32 0.34 0.29

± ± ± ±

(0.004) (0.005) (0.005) (0.004)

TFS1 presented large convolution with grain boundary response.

possible to roughly explain the observed changes with geometrical features; but changes on grain boundary activation energy must be related to changes on the chemical nature of grain boundaries. These differences can be explained by different blocking factors produced by defects at grain boundary interfaces, as suggested by M. Dessemond et al [34] more interface blocking defects are created the larger activation energy result. Obtained results indicate that the TFS4 sample displays a larger conductivity than the OH-NAS sample. As conductivity measurements (two co-planar electrodes) can lead to large errors due to approximations made, it is interesting to compare complex a.c. conductivity and derivative results to highlight Li migration dynamics. The a.c. conductivity of three Li-NASICON samples, measured at 175 K, are included as well as the corresponding derivatives in Fig. 6. The TFS4 sample, displayed a "bulk" d.c. conductivity close to that of Li1.2Al0.2Ti1.8(PO4)3 (LATP) ceramic reported in reference [30], with a much better grain boundary contribution. The OH-NAS conductivity is below the conductivity value of other samples. The derivative indicates that the TFS4″bulk" plateau is produced at larger frequency than that of the OH-NAS sample, indicating a faster Li dynamics, slightly below that of the LATP ceramic. This result is in agreement with the better conductivity reported in Li-NASICONS with Al/Ti rather than Al/(Ti,Ge) substitutions. It is worth to stress the quite similar blocking behavior detected in TFS4 and OH-NAS grain boundaries contributions, despite the preparation route. In a previous paper, it was demonstrated that thick films are more heterogeneous than deduced from X-ray diffraction patterns [23]. All thick films prepared in this work are presumably formed by single LiNASICON phases. However, amorphous/glassy secondary phases, not detected by XRD, are also present in analyzed samples. As previously commented, changes detected in activation energies of the "bulk" response can be related to compositional deviations and to the presence of secondary phases. The confocal RAMAN spectroscopy is a useful tool to study the homogeneity of samples, due to its capability for identifying different amorphous or crystalline phases. Due to the sweeping procedure used, this technique gives also information, with sub-micrometer resolution, about distribution of phases that is the key to understand

Fig. 5. Normalized effective “bulk” conductivity as a function of the volume fraction of electrolyte particles. The line is the calculation made using the Percolation effective medium theory [32] assuming that the particles are spheres and a tridimensional percolation threshold of 0.33. Black triangles, expected effective ‘bulk’ conductivity taking into account porosity. White triangles, expected effective ‘bulk’ conductivity taking into account porosity and secondary phases as deduced from the confocal RAMAN spectroscopy.

Important changes can be expected if the volume fraction change due to the preparation procedure. On the other hand, the activation energy of samples should not change significantly with the volume fraction, when values are bigger than 0.60, as deduced from the composite model [32]. The activation energy of TFS2 and TFS3 samples are almost the same, but the activation energy for the TFS4 sample is lower. Changes in activation energy for the "bulk" response are often associated with changes in the grains composition. The lowest activation energy deduced for the TFS4 sample can be related to a closer Li/Al stoichiometry to that reported for the best LATP NASICON conductor (lower activation energy) [15]. However, a large scattering was reported in published works [33], preventing the use of the activation energy alone as a indicator of the sample quality. The grain boundary response of prepared thick films is often related to the samples microstructure. The grain boundary impedance is proportional to the number of grain boundaries and inversely proportional to the grain boundary area, assuming that the nature of contacts does not change with thermal treatments. As an approximation, the number of grain boundary interfaces can be deduced from the micrographs choosing a connected path among the grains. Using the SEM micrographs of Fig. 2, the number of grain boundaries along selected paths resulted between 33–36, 12–15, 16–18 and 8–9 in TFS1, TFS2, TFS3 and TFS4 samples. The approximate grain boundary areas were 0.49, 2.2, 1.6 and 4.8 μm2 in analyzed samples, assuming square contacts between adjacent grains and observed contact lengths. With these data a calculation of the grain boundary factor was performed, that after normalization to that of TSF4 sample, yielded the results included as large hollow triangles in Fig. 4. It can be observed a quite good agreement between the approximate variation of the grain boundary impedance, based on geometric factors, and that deduced from grain boundary impedance measurements. The agreement indicates that it is

Fig. 6. A.C Conductivity spectroscopic plot for three Li- NASICON samples at 175K. The inset corresponds to the corresponding derivatives.

6

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

Fig. 7. Confocal Raman study of the TFS4 sample. A) Color map of the surface scanning with the confocal RAMAN device, the color code is explained in figure E. B),C) in deep confocal RAMAN mapping in the X and in the Y axis respectively . D) optical micrograph showing approximately the sample area mapped. E) Different main RAMAN spectra found in the sample, red (Crystalline Li-NASICON), Cyan (Amorphous phase,). Green (Rutile).

technique. In Fig. 7a, a 35 × 35 μm2 micrograph of the TSF4 surface is displayed. Figs. 7a–c display the result of the confocal RAMAN analysis in surface and in deep along the major two axes. The color code used in micrograph is red for the Li-NASICON phase, green for the TiO2 rutile

macroscopic properties. This technique has been only applied to the TSF4 sample, the best sample analyzed here. In Fig. 7, the phase distribution between two gold stripes used for the electrical measurements was investigated with the Confocal Raman 7

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

secondary phases is already produced at 1000ºC as deduced from the Raman results. A possible solution for this problem could be the reduction of the sintering temperature, to get pure Li-NASICON phases. However, lowering the sintering temperature produces a reduction of the grain size, worse grain boundary quality and an increment of the film porosity and the number of grain boundaries interfaces. These effects will lead to a reduction of the total Li conductivity despite the better phase purity. From this fact, a compromise between purity and microstructure must be achieved to maximize overall Li-conductivity in thick films. From these results, it can be inferred that thermal treatments have an important effect on the microstructure and the conductivity of LiNASICON thick films. The isostatic lamination step, previous to sintering treatments, improves the microstructure by reducing the porosity, increasing the grain size, favoring orientation of crystallite faces and reducing the grain boundary impedance. All these effects enhance the total ionic conductivity in prepared thick films. At 1000 ºC, it is not possible to obtain pure Li-NASICON phases; however, the presence of the usually reported AlPO4 secondary phase was difficult to observe in X-ray diffraction patterns. In this sample (TFS4) confocal RAMAN spectroscopy reveals the existence of peaks ascribed to rutile and amorphous phases. The existence of secondary phases can be related to some decomposition of the NASICON phase. The formed amorphous Li titanates can retain Li ion conductivity at the particles surface producing a moderate conductivity blocking effect. High values obtained in ionic conductivity of films suggest that grain boundary conductivity has been considerably enhanced as a consequence of an improved interface coherence among crystallites. In these boundaries, the quantity of blocking defects decreases and ion conductivity is improved. Further work in the preparation conditions is on its way to reduce samples porosity and to further increase NASICON phase volume fraction in the thick film. An ideal full dense thick film with the same grain and grain boundaries conductivities of the TSF4 sample, would yield a much higher total Li conductivity than the OH-NAS sample. The study of the amorphous phases is also required. In this work, confocal RAMAN spectroscopy plays an important role in understanding electrical behavior of thick films.

phase and blue for an amorphous phase. The volume fractions of each phase were 0.56, 0.11, 0.02 for Li-NASICON, amorphous phase and rutile phases. The summation of different fractions (0.69) are in good agreement with volume fraction deduced from SEM micrographs. This result supports the volume fraction estimations made by the image analysis of the SEM micrographs of the thick film sample and good homogeneity. The Li-NASICON phase is the most abundant and presents good connectivity, but its volume fraction is (0.56) lower than deduced from the SEM micrograph, (0.72 ± 0.04), due to contribution of secondary phases. In this preparation, an improvement with respect to a previous work was detected (secondary phases fraction = 0.2) [23]. The secondary phases, still present in the prepared sample increases tortuousness in Li-paths, increasing the total impedance. In the present work, the sample is more dense, with a larger NASICON volume fraction than reported in previous works. The main secondary phase found out in the previous paper was an amorphous phase with a large band at 782 cm−1. A less abundant secondary phase, with a large band at 747 cm−1, was also detected [23]. These secondary glassy phases could correspond, according to literature, to lithium titanates, where tetrahedral Ti can be stabilized by the presence of lithium [35,36]. In the present study the large band is located at 747 cm−1, suggesting that composition of the main secondary phase corresponds to the less abundant secondary phase reported previously [23]. This result indicates the strong effect of the preparation method on the homogeneity of the NASICON films. The ratio between the NASICON phase and this secondary phase (cyan color) is 0.19 in the present sample larger than in the previous work 0.13. The total secondary phase/ NASICON ratio is 0.24 in TSF4 and 0.87 in the previous work, indicating a better homogeneity. It is important to remark that despite the large porosity and the presence of secondary phases, the total conductivity of the film remains quite high. This result is related to the high Li conductivity of the NASICON phase as well as the establishment of good inter-grain interfaces. The reduction of the NASICON volume fraction ascribed to the presence of secondary phases was taken into account in Fig. 5, where corrected volume fractions of the NASICON phase are considered. As the confocal Raman spectroscopy analysis has not been performed in other samples, corrected volume fractions were deduced from NASICON / secondary phase ratios obtained in the TFS4 sample. After the correction of the volume fraction, the TFS1 point shifts into the critical region of the curve, as a consequence of the 3D percolation threshold proximity (0.33). This proximity can also be responsible for the quite large value of the minimum slope in the ″bulk" d.c.″ pseudo-plateau" in addition to the poor grain boundaries. It should be noted that the commercial OH-NAS sample is not a pure phase but a composite including two Li-NASICON phases with different composition (main Li1+xAlxGeyTi2-x-yP3O12 and less abundant Li1+x+3zAlx(Ge,Ti)2-x(SizPO4)3 phases) and one secondary AlPO4 phase [37]. The calculation of the volume fraction of Li-NASICON based on the published OH-NAS, SEM micrographs is near 0.8 ± 0.02. The fraction of the minor Li- NASICON phase is below 0.02. This phase is evenly distributed and do not contribute much to the Li conductivity. The true conductivity of crystals in glass ceramics is 1.4 times the effective "bulk" response of the sample for the volume fraction considered. In Fig. 5, the position of the OH-NAS sample can be found in the right part of the normalized "bulk" d.c. conductivity plot (hollow triangle). The sample heterogeneity is emphasized in the published report of the company, where different compositions of phases are given as well as micrographs illustrating secondary phases. For the TFS4 sample, the crystallites conductivity is 2.9 times the measured effective "bulk" one, due to the lower volume fraction of Li-NASICON in the sample investigated here than in the OH-NAS sample. The results of this work indicate that the sintering temperature increases the quantity of secondary phases, Fig. 1a. The appearance of

5. Conclusions The preparation of Li-NASICON (Li0.3Al0.3Ti1.7(PO4)3) thick films on Al2O3 substrates is described, with Li conductivities slightly higher than reported in commercial OHARA glass ceramics (> 10−4 Ω-1 cm-1). The presence and distribution of secondary phases has been assessed by the combined use of X-ray, SEM and confocal Raman techniques. Different thermal treatments, affected microstructure of films, changing Li transport properties. The variation of the "bulk" and grain boundary activation energy of different samples cannot be explained by geometrical factors related to the microstructure, but to variations on NASICON grain compositions induced by thermal treatments. The compositional variation can be related to some decomposition of the NASICON phase. No AlPO4 phase is present and the presence of Ti rich phases is produced in prepared films. With the confocal Raman technique the presence of rutile and of amorphous lithium titanate phases was deduced. The use of an isostatic lamination step before sintering treatments enhanced density and grain size in thick films. The slow heating of the sample produced denser samples in comparison with thermal shock treatments. The relatively low grain boundary impedance suggest the presence of good inter-grain interfaces, improving electrical properties of prepared films. The preparation method used here can be scaled to produce large area electrolytes for new Li all-solid state batteries architectures.

8

Journal of the European Ceramic Society xxx (xxxx) xxx–xxx

R. Jiménez et al.

Aknowledgements [19]

The Spanish group has been funded by the regional MATERYENER3-CM (S2013/ MIT-2753), and national projects MINECOMAT2013-46452-C4-2R, MAT2016-76851-R and I-Link 1064. The Ukrainian group thanks the NAS Ukraine project 34/15-H funding ("Fundamental Problems of Nanostructured Systems, Nanomaterials, Nanotechnology" program). Both groups were also funded by the European NANOLICOM (FP7-PEOPLE-2009-IRSES) projects.

[20]

[21]

References

[22]

[1] J.-M. Tarascon, M. Armand, Issues and challenges facing rechargeable lithium batteries, Nature 414 (2001) 359–367, http://dx.doi.org/10.1038/35104644. [2] C. Brent, Melot, J.-M. Tarascon, Design and preparation of materials for advanced electrochemical storage, Acc. Chem. Res. 46 (2013) 1226–1238, http://dx.doi.org/ 10.1021/ar300088q. [3] S.J. Visco, E. Nimon, L.C. De Jonghe, Secondary batteries: metal-air systems lithium-air, in: G. Jürgen (Ed.), Encyclopedia of Electrochemical Power Sources, Elsevier, Amsterdam, 2009, pp. 376–383, , http://dx.doi.org/10.1016/b978044452745-5.00184-2. [4] D.X. Ji, L.K. Tae, L.F. Nazar, Challenges of lithium-sulfur and lithium-air cells: old chemistry, new advances, Scalable Energy Storage: Beyond Lithium Ion, Almaden Institute, San Jose, USA, 2009 Aug. 26. [5] P.G. Bruce, S.A. Freunberger, L.J. Hardwick, et al., Li-O2 and Li-S batteries with high energy storage, Nat Mater. 11 (2012) 19–29, http://dx.doi.org/10.1038/ nmat3191. [6] K.I. Cho, S.H. Lee, K.H. Cho, et al., Li2O-B2O3-P2O5 solid electrolyte for thin film batteries, J. Power Sources 163 (2006) 223–228. [7] M. Abdel-Baki, A.M. Salem, F.A. Abdel-Wahab, F. El- Diasty, Bond character, optical properties and ionic conductivity of Li2O/B2O3/SiO2/Al2O3 glass: effect of structural substitution of Li2O for LiCl, J. Non-Cryst. Solids 354 (2008) 4527–4533, http://dx.doi.org/10.1016/j.jnoncrysol.2008.07.003. [8] B.K. Money, K. Hariharan, Relation between structural and conductivity relaxation in PEO and PEO based electrolytes, Solid State Ionics 179 (2008) 1273–1277, http://dx.doi.org/10.1016/j.ssi.2013.09.033. [9] P. Knauth, Inorganic solid Li ion conductors: an overview, Solid State Ionics 180 (2009) 911–916, http://dx.doi.org/10.1016/j.ssi.2009.03.022. [10] K. Takada, T. Inada, A. Kajiyama, et al., Solid-state lithium battery with graphite anode, Solid State Ionics 158 (2003) 269–274, http://dx.doi.org/10.1016/S01672738(02)00823-8. [11] M. Kotobuki, Y. Isshiki, H. Munakata, K. Kanamura, All-solid-state lithium battery with a three-dimensionally ordered Li1.5Al0.5Ti1.5(PO4)3 electrode, Electrochim. Acta 55 (2010) 68–92. [12] S. Shiraki, H. Oki, Y. Takagi, et al., Fabrication of all-solid-state battery using epitaxial LiCoO2 thin films, J. Power Sources 267 (2014) 881–887, http://dx.doi. org/10.1016/j.jpowsour.2014.05.133. [13] M. Haruta, S. Shiraki, T. Suzuki, et al., Negligible “Negative space-charge effects" at oxide-electrolyte/electrode interfaces of thin-film batteries, Nano Lett. 15 (2015) 1498–1502, http://dx.doi.org/10.1021/nl5035896. [14] P.H.L. Notten, F. Roozeboom, R.A.H. Niessen, L. Baggetto, 3-D integrated all-solidstate rechargeable batteries, Adv. Mater. 19 (2007) 4564–4567, http://dx.doi.org/ 10.1002/adma.200702398. [15] K. Arbi, S. Mandal, J.M. Rojo, J. Sanz, Dependence of ionic conductivity of composition of fast ionic conductors Li1+xTi2−xAlx(PO4)3, 0≤x≤0.7. A parallel NMR and electric impedance study, Chem. Mater. 14 (3) (2002) 1091–1097, http://dx. doi.org/10.1021/cm010528i. [16] H. Aono, E. Sugimoto, Y. Sadaoka, et al., The electrical properties of ceramic electrolytes for Li1+xMxTi2-x(PO4)3 + yLi2O, M = Ge, Sn, Hf and Zr systems, J. Electrochem. Soc. 140 (1993) 1827–1833, http://dx.doi.org/10.1149/1.2220723. [17] G. Adachi, N. Imanaka, H. Aono, Fast Li+ conducting ceramic electrolytes, Adv. Mater. 8 (1996) 127–135, http://dx.doi.org/10.1021/ar00045a002. [18] J. Fu, Superionic conductivity of glass-ceramics in the system Li2O-Al2O3-TiO2-P2O5,

[23]

[24]

[25] [26]

[27]

[28]

[29]

[30]

[31]

[32]

[33]

[34]

[35]

[36]

[37]

9

Solid State Ionics 96 (1997) 195–200, http://dx.doi.org/10.1016/S0167-2738(97) 00018-0. K. Arbi, J.M. Rojo, J. Sanz, Lithium mobility in titanium based Nasicon Li1+xTi2−xAlx(PO4)3 and LiTi2−x Zrx(PO4)3 materials followed by NMR and impedance spectroscopy, J. Eur. Ceram. Soc. 27 (2007) 4215–4218, http://dx.doi.org/ 10.1016/j.jeurceramsoc.2007.02.118. D. Jurków, H. Roguszczak, L. Golonka, Cold chemical lamination of ceramic green tapes, J. European Ceramic Society 29 (2009) 703–709, http://dx.doi.org/10.1016/ j.jeurceramsoc.2008.07.035. Hae-Gu Park, Hwan Moon, Sung-Chul Park, et al., Performance improvement of anode-supported electrolytes for planar solid oxide fuel cells via a tape-casting/ lamination/co-firing technique, J. Power Sources. 195 (2010) 2463–2469, http:// dx.doi.org/10.1016/j.jpowsour.2009.11.086. D. Jurków, L. Golonka, Low-pressure thermo compressive lamination, J. Eur. Ceram. Soc. 32 (2012) 2431–2441, http://dx.doi.org/10.1016/j.jeurceramsoc. 2011.12.033. R. Jimenez, A. del Campo, M.L. Calzada, et al., Lithium La0,57Li0,33TiO3 perovskite and Li1,3Al0,3Ti1,7(PO4)3 Li-NASICON supported thick films electrolytes prepared by tape casting method, J. Electrochem. Soc. 163 (2016) A1653–A1659, http://dx.doi. org/10.1149/2.0881608jes. W. Bucheli, R. Jimenez, J. Sanz, A. Varez, The log(sigma) vs. log(omega) derivative plot used to analyze the ac conductivity. Application to fast Li+ ion conductors with perovskite structure, Solid State Ionics 227 (2012) 113–118, http://dx.doi.org/10. 1016/j.ssi.2012.09.018. Richard E. Mistler, Eric R. Twiname, Tape Casting: Theory and Practice, Wiley, 2000 298 P. J. Chen, K.R. Udayakumar, K.G. Brooks, L.E. Cross, Rapid thermal annealing of sol‐gel derived lead zirconatetitanate thin films, J. Appl. Phys. 71 (1992) 4465–4469, http://dx.doi.org/10.1063/1.350789. S.Ju Kang, Y.J. Park, J. Sung, et al., Spin cast ferroelectric beta poly(vinylidene fluoride) thin films via rapid thermal annealing, Appl. Phys. Lett. 92 (2008) 012921, http://dx.doi.org/10.1063/1.2830701. Ging H. Hsiue, L.W. Chu, I.N. Lin, Optimized phosphate ester structure for the dispersion of nano-sized barium titanate in proper non-aqueous media, Colloids Surf. A: Physicochem. Eng. Aspects 294 (2007) 212–220, http://dx.doi.org/10. 1016/j.colsurfa.2006.08.013. F. Doreau, G. Tari, M. Guedes, T. Chartier, C. Pagnoux, Mechanical and lamination properties of alumina green tapes obtained by aqueous tape-casting, J. Eur. Ceram. Soc. 19 (1999) 28672873. K. Arbi, W. Bucheli, R. Jimenez, J. Sanz, High lithium ion conducting solid electrolytes based on NASICON Li1+xAlxM2-x(PO4)3 materials (M = Ti, Ge and 0 < = x < = 0.5), J. Eur. Ceram. Soc. 35 (2015) 1477–1484, http://dx.doi.org/ 10.1016/j.jeurceramsoc.2014.11.023. P.G. Bruce, A.R. West, The AC conductivity of polycrystalline LISICON, Li2+2xZn1xGeO4, and a model for intergranular constriction resistances, J. Electrochem. Soc. 130 (1983) 662–669, http://dx.doi.org/10.1149/1.2119778. R. Jimenez, Application of the PEM theory to solid electrolyte pressed powder pellets: electrically isotropic solid electrolyte particles, Solid State Ionics 90 (1996) 233–244, http://dx.doi.org/10.1016/S0167-2738(96)00371-2. S. Breuer, D. Prutsh, Q. Ma, et al., Separating bulk from grain boundary li ion conductivity in the sol-gel prepared solid electrolyte Li1.5Al0.5Ti1.5(PO4)3, J. Mater. Chem. A 3 (2015) 21343–21350, http://dx.doi.org/10.1039/C5TA06379E. M. Kleitz, L. Dessemond, M.C. Steil, Model for ion-blocking at internal interfaces in zirconias, Solid State Ionics. 75 (1995) 107–115, http://dx.doi.org/10.1016/01672738(94)00143-G. R. Baddour-Hadjean, J.P. Pereira-Ramos, Raman microspectrometry applied to the study of electrode materials for lithium batteries, Chem. Rev. 110 (2010) 1278–1319, http://dx.doi.org/10.1021/cr800344k. K.J. Rao, K.C. Sobha, S. Kumar, Infrared and raman spectroscopic studies of glasses with NASICON-type chemistry, Proc. Indian Acad. Sci. 113 (2001) 497–514, http:// dx.doi.org/10.1007/BF02708786. K. Nakajima, T. Katoh, Y. Inda, B. Hoffman, lithium ion conductive glass ceramics: properties and application in lithium metal batteries, Symposium on Energy Storage Beyond Lithium Ion; Materials Perspective, October, Oak Ridge National Laboratory, 2010, pp. 7–8.