1. Introduction
Because of their high energy performance, lithium-ion batteries are nowadays widely used for storage power generation and uninterruptible power supply (UPS) systems [
1,
2]. The use of Li-conducting organic liquid and/or polymer electrolytes, in the present battery technology, prevents the fabrication of completely safe devices as a result of low thermal stability. The use of an inorganic solid instead of a liquid/polymer electrolyte will significantly improve the safety of the lithium-ion battery, also extending its life by reducing the degradation processes [
3,
4,
5]. The rather poor temperature stability of Li-batteries based on organic electrolytes prevents the use of these devices in harsh environments where temperatures can rise above 100 °C. This limitation can be overcome by using ionic liquids, but the reduced ionic conductivity is deleterious for the output power of the battery. The preparation of all solid-state batteries (ASSBs) can be of interest for their use in a wider temperature and pressure range that the current liquid electrolyte-based batteries (LEEBs) cannot cover.
Another interesting niche of application of ASSBs is the result of the progressive miniaturization of electronic components. An increasing demand for micro-sized power sources incites the research of thin or thick films, so as to avoid the use of liquid electrolytes and to confine the batteries in a defined space (integrated batteries). Smart cards, implanted medical devices, micro electro-mechanical systems, memory blocks, sensors, transducers, and especially military equipment are potential consumers of ASSBs film batteries [
6,
7,
8]. Besides the implementation of ASSBs, the use of full dense solid electrolytes ceramic membranes as separators in liquid electrolyte-based batteries to prevent any short circuit caused by the growth of dendrites is an important application [
9]. These dense ceramic electrolyte membranes can be also used as separators in metal–air batteries, protecting the anodic compartment from the ambient [
10]. However, there are many drawbacks that limit their implementation. The first one is the relatively low ionic conductivity of the solid, which is usually one order of magnitude lower than that of their liquid electrolyte counterparts. But the main challenge to be faced is the procurement of electrochemically active solid electrodes/solid electrolyte interfaces. To accomplish this challenge, inter-diffusion must be avoided as much as possible. Another drawback is the mechanical fatigue produced on the interfaces during insertion de-insertion processes.
In ASSBs solid electrode/electrolyte, the interfacial area is small, as contacts are limited. In comparison with conventional secondary batteries, the diffusion path lengths for the electron draining at the current collectors, and for the Li insertion, are much larger. The length reduction of the diffusion path is fundamental for fast electrochemical kinetics and high power [
11]. The thin film battery ASSBs concept was proposed to solve this problem. Different designs have been made to increase the interfacial area. ASSBs based on thin film materials presented high charge and discharge rates, comparable to liquid electrolyte Li batteries, but the total energy stored remains poor [
12,
13]. The active mass of the electrodes must be increased, as this is proportional to the stored energy. Thick films with thicknesses in the range of 10–100 μm can be a solution. They will provide increased energy and enough power. Novel ASSB designs are needed in order to reduce the diffusion path length of the ions and electrons. Moreover, the strain effects due to the insertion/de-insertion of Li must be minimized to increase battery durability. This minimization will require important changes in battery designs [
14]. Electrolyte supported large area ASSB architecture can be a solution. It allows for flexible electrode design to minimize the strain effects, reduce the diffusion path, and balance the electrode mass between the anode and cathode. Two main different configurations of supported ASSBs can be foreseen. The first one with the anode and cathode at different sides of the electrolyte, and the second one with electrodes disposed at the same side of the electrolyte (inter-digital electrode configuration). A self-supported electrolyte thick-film with a high enough mechanical strength is mandatory for the first configuration. In the second one, a supported solid electrolyte layer on a refractory mechanically resistant substrate can be used. The procurement of planar large area electrolyte thick films is of main importance for both ASSB configurations. Among the different families of high conductivity Li solid electrolytes, Li
1.3Al
0.3Ti
1.7(PO
4)
3 with a NASICON structure is a good candidate for ASSBs, because of its high lithium conductivity (σ
bulk ∼10
−3 S·cm
−1 at RT), chemical resistance, and mechanical stability [
15,
16,
17,
18,
19]. These materials are widely studied in the form of ceramics and glass ceramics. Studies on thick-films ceramics are still scarce. The sulfide-based solid electrolytes presented larger Li conductivities, but are difficult to handle in a regular atmosphere compared to the same, but oxide-based.
For the procurement of solid electrolyte thick films with a thickness larger than 10 µm, tape casting or screen-printing methods are more appropriate [
20]. For the application, the thick film must be dense. The presence of pores and cracks in the ceramic films produce a negative impact on the electrical and mechanical properties of the films. It is necessary to find ways to produce full dense thick films [
21,
22].
The aim of this work is the preparation of self-standing thick-films of Li1.3Al0.3Ti1.7(PO4)3 with a high Li conductivity, to be used as support for all solid-state Li batteries. The structure, phase homogeneity, and microstructure of prepared thick films have been investigated by XRD, NMR, confocal micro-Raman, and SEM techniques. Impedance spectroscopy (IS) in a wide temperature and frequency range was used to extract the “total” and “bulk” direct current (DC) conductivity of the thick film samples prepared. A mechano-elastic characterization of the thick films was also performed, as high mechanical properties are needed to be used as support. The results obtained have been discussed and compared with those reported in other samples of a similar composition, especially the OHARA Li-NASICON glass-ceramic (OHGC), which is a commercial large area Li solid electrolyte that is available.
3. Discussion
Large area self-supported Li-NASICON thick films can be prepared by tape casting the slurries prepared with Li-NASICON powder obtained from the sol–gel powders. The crystallization temperature of the powder was 800 °C and the sintering was done at 900 °C in order to produce materials with a very small quantity of crystalline secondary phases. A small quantity of LiTiPO5 was found by XRD. In the commercial OHGC sample, a large quantity of AlPO4 was detected, which was not present in the LATP thick films. The NMR characterization of the samples, found other small quantities of secondary phases like LiAlP2O7 and LiTiPO5 in the LATP thick film samples.
On the other hand, the microstructure of our samples is similar to that of bulk ceramics, with a typical bi-modal grain distribution [
30], but with a larger porosity. The average grain size is much larger than the one found in the commercial sample, where the sizes are sub-micrometric [
31]. Another important difference in the microstructure, especially in the LATPT2 one, is that the facets of the large Li-NASICON crystals are clearly visible, in both the surface and in the fresh fracture micrographs. This feature is related with the easy breaking of the ceramic through the grain boundaries instead of through the intra-grain fracture. The larger crystals grow, but do not sinter with the neighbor ones, producing a poor attached interface in these samples at temperature and time-used in treatments. The poor sintering among the large grains joined to the large porosity can be responsible of the low Young’s modulus of the prepared thick films, as well as their brittle behavior. From the observed microstructure large grain boundary, impedance can be expected for the prepared thick films. Moreover, the grain size growth can be deleterious for the conductivity property, as it surpasses the critical size for the micro-cracks formation as a result of the strong dilatation anisotropy of the Li-NASICON [
32]. The absence of AlPO
4 as a secondary phase justifies large grain growths detected in the thick films samples from the sol–gel powders, as this phase prevents grain growth in LATP NASICON [
33].
The NMR characterization allowed for the quantification of the Al and Li incorporation in the LATP NASICON phase. The LATP
x = 0.3 phase is almost stoichiometric (
x = 0.34), showing a good quality in the prepared LATP crystallites of thick-films. The information about the Li
+ dynamics in the samples indicates enhanced dynamics for the OHGC sample, as it presented a narrower central line and reduced quadrupolar interaction. This result is not in agreement with the conductivity results, which indicate a better conductivity in the LATPT2 sample. The complexity of the Li
+ conductivity mechanism in Li-NASICON that implies a diffusion path along M1–M3–M1 sites with different residence times at different sites means that this simplified analysis of the line is not strictly valid. The changes on the residence time of two sites explain the increment of quadrupolar constants with temperature [
34].
The mechanical properties of an electrolyte thick-film must be high for the development of large-area ASSBs or Li–air batteries. The results of the Young’s modulus of the prepared thick-films are quite poor, compared with the OHGC commercial film. In the best case, LATPT1 is three times lower, increasing losses and dissipation effects at grain boundaries, making the material more viscous-elastic. LATPT2 presented the poorest mechanical properties. This drop of the mechanical properties can be related to the microstructure and the nature of the grain boundaries. Large grains imply that samples are prone to develop micro-cracking [
32] and poor sintering. This effect, joined with a rather large porosity of samples, can be the reason for the low Young’s modulus and the important brittleness of the sample providing a poor mechanical performance of prepared thick-films.
The ionic conductivity is also affected by the sample’s microstructure. The LATPT2 samples showed the highest “bulk” ionic conductivity, resulting in a good crystallization and stoichiometry, with a rather small quantity of secondary phases. The obtained value, 7 × 10
−3 Ω
−1·cm
−1 (measured at RT), is remarkable, displaying a very low activation energy of 0.22 eV. The conductivity measured in the micro-crystals of the LATP of the same composition is 5 × 10
−3 Ω
−1·cm
−1 at RT, and no activation energy value is given. [
35] Poor sintering, large porosity, and the presence of micro-cracks strongly reduces the “overall” DC Li conductivity. At 195 K, the difference between the “bulk” and the grain boundary values is three orders of magnitude. The LATPT1 shows the same problems, but in a lower degree.
The existence of secondary phases can also be behind the increase in the ceramic total resistivity. In the phase analysis of our samples, a secondary phase of LiTiPO
5 was found, as well as small quantities of LiAlP
2O
7 and rutile. This main secondary phase is much less Li conducting (1.1 × 10
−5 Ω
−1·cm
−1 at 400 °C, E
a = 1 eV) [
36] than LATP. If this secondary phase is accumulated at the grain boundaries, it can be, to some extent, responsible for their strong blocking behavior. To observe the distribution of the secondary phases in the sample, the confocal Raman technique is a powerful technique. From the compositional map of the LATPT1 thick-film sample (
Figure 4c), an even distribution of this and the other secondary phases is observed. The Raman spectrum of the unknown phase does not match with that of the LiAlP
2O
7 one (the main Raman bands of this phase are around 669 nm
−1, and 713 nm
−1, because of the υ P–O–P and vs P–O–P modes) [
37]. This secondary phase can be sub-micrometer in size, and thus out of the confocal Raman sensitivity.
As the secondary phases observed in the confocal RAMAN sweep are evenly distributed in the thick film, their Li
+ and electronic conductivities are well below those of the LATP, and the quantity is not enough to give a percolated path through the sample; the influence of them in the overall Li- conductivity is to reduce the volume fraction of the LATP phase, and reduce the grain boundary area. Both effects produce a reduction in the “bulk” conductivity as well as in the total DC conductivity, but in a lower extent than if they would be arranged in series with the LATP grains. Maybe the LiAlP
2O
7 secondary phase, which is not observed in the RAMAN spectroscopy, could be accumulated at the grain boundaries. The conductivity of this phase is 8.8 × 10
−6 Ω
−1·cm
−1 at 330 °C, just three orders of magnitude lower than the “bulk” one [
37]. With this conductivity, the distribution of this phase at the grain boundaries cannot explain the increase in the blocking behavior.
As the main difference between the LATPT1 and LATPT2 samples is the change of grain size, the quantity of the grain boundaries must be smaller for the LATPT2 sample, and the reduction in the total DC conductivity, which is as a result of the increase in the blocking effect of the grain boundaries, must be because of the worse grain boundaries in this sample. This statement is also supported by the reduction of the Young’s modulus of this sample in comparison with the LATPT1 one. A larger sintering time increases the grain size, but produces worse grain boundaries. This effect can be related to some thermodynamic instability of the Li-NASICON phase at these temperatures, added to the presence of microcracking at the grain boundaries. The in depth explanation of the origin of this effect is not in the scope of this work.
From our results, it can be concluded that, in pure phase LATP thick-films. the increase in grain size is deleterious for the functional properties. The large porosity joined with the bad quality of the interfaces between the large grains produced this effect. In fact, the OHGC thick film sample was designed to avoid the grain growth and the formation of LATP/LATP grain boundaries. The fabrication of the OHGC sample implies high temperatures to melt the mixture, and then the re-crystallization step (700 °C), to a specified extent. It should be kept in mind that heating the OHGC sample at temperatures larger than 700 °C increased re-crystallization and reduced the total DC Li conductivity; this is a limitation for its application in ASSBs.
In order to produce large area ceramic LATP thick films, the use of sintering aids is mandatory, as well as some additives that could reduce the grain growth. For this last aim, the use of a small volume of AlPO
4 to reduce grain size, as proposed in the literature [
33], could be a solution. Another way, which is totally different, is the use of spark plasma sintering (SPS), as it gives a higher density and smaller grain size; however, this is not a good choice for large area samples.
4. Materials and Methods
4.1. LATP Powder Preparation
The Li1.3Al0.3Ti1.7(PO4)3 solid-state electrolyte powder was synthesized by a sol–gel method. The raw materials used were LiNO3 (purity >97.0%; J.T. Baker Chemical Co, Phillipsburg, NJ, USA) City, Country), Al(NO3)3 9H2O (98%; Alfa Aesar, Ward Hill, MA, USA), NH4H2PO4 (purity >98%; Sigma-Aldrich, St. Louis, MO, USA) and Titanium(IV) isopropoxide (98%; Sigma-Aldrich). Titanium (IV) isopropoxide was added to the de-ionized water while stirring. Titanium hydroxide precipitate was formed immediately. The precipitate was filtered and washed. The washed precipitate was transferred to an empty beaker, and de-ionized water was added into the beaker. Nitric acid (65%, Aldrich) was then added to the titanium hydroxide precipitation with water. When a clear TiO2 and nitrate solution was formed, citric acid monohydrate (99%, Merck, Darmstadt, Germany) was added into the solution to stabilize it. LiNO3 and Al(NO3)3·9H2O added into the TiO2 and nitrate solution while stirring. After the salts were dissolved, NH4H2PO4 was then added to the solution and a sol was immediately formed. The gel was dried at 80 °C for 24 h and calcined at 600 °C for 4 h. The calcined powder was milled in ethanol with zirconia balls for 24 h, and pressing into pellets. The pressed pellets were sintered at 800 °C. The pellets were milled again and sieved through a 200 mesh.
4.2. Slurry and Thick Film Preparation
The LATP powder and binder (B73305 Ferro Corp., San Marcos, CA, USA; solute PVB, solvent Ethanol + Toluene) in 62.5:37.5 wt % proportions were mixed by ball milling in a Teflon bottle with zircona balls for 24 h. The mix was de-foamed in an oven at 70 °C for one hour. Then, it was ball milled again for 12–16 h. The recipient with the slurry was introduced into a vacuum chamber for 4–6 min for degassing. The slurry was tape casted using a commercial lab-size tape casting machine (uni99 from Huan-Yang Co., Ltd.) using a blade separation of 1 mm at a speed of 2 cm/s. The formed tape was dried at 40 °C and cut into pieces of 60 × 60 mm, and the final sintering was performed at 900 °C with a LATP powder bed during 4 h LATPT1 and 12 h LATPT2.
4.3. Samples Characterization
X-ray diffraction (XRD) patterns were recorded at room temperature using a Bruker D8 Advance diffractometer (40 kV and 30 mA, Bruker AXS, Karlsruhe, Germany), using the Cu Kα radiation. The XRD patterns were recorded in the 10–70° range. The angular step was 0.04°, and the counting time/step was 0.5 s. The crystalline phases were identified using the data of the International Centre for Diffraction Data (ICDD).
The thick films compositional homogeneity was also tested by using the confocal Raman microscope (Witec alpha-300R, WITec GmbH, Ulm, Germany), using a 532 nm excitation laser and a 100× objective lens (NA = 0.9). The incident laser power was 0.5 mW. The optical diffraction resolution in the confocal Raman apparatus microscope was limited to ~200 nm laterally and ~500 nm vertically. The resolution achieved in the 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. The collected spectra were analyzed by using the Witec Control Plus Software.
The 27Al, 7Li, and 31P MAS-NMR spectra were recorded with an AVANCE-400 Bruker spectrometer (9.4 T magnetic field). The frequencies used were 104.26, 155.45, and 161.97 MHz, respectively. The spectra were recorded after a π/2 irradiation (2.0 μs pulses for aluminum and lithium, and 4.0 μs for phosphorus) with the MAS technique (rotation of samples at 10 kHz around an axis inclined 54°44′, with respect to the external magnetic field). The number of scans was chosen in the range of 100–800. The position of the NMR components was referred to as 1 M AlCl3, 1 M LiCl, and 85% H3PO4 aqueous solutions. The spectral deconvolutions (position, linewidths, and intensities of components) were obtained with the commercial DMFIT (D. Massiot software, NMR @ CEMHTI CNRS UPR3079, Orléans, France).
The cross-section and plan-view micrographs of the crystalline oxide films were obtained using Scanning Electron Microscopy SEM (Phenom™ G2 pro SEM 5 kV; Eindhoven, Holland). The porosity of the thick films was calculated from the binarization of the plane view SEM images using the image processing software MIP 45 (Digital Imaging Systems Ltd., Buckinghamshire, UK).
The electrical properties of the thick films were measured on planar capacitors by the de-metallization of the opposite face of the samples with Au paste (Dupont 8216), and sintered at 800 °C 2 h. The metalized capacitors were mounted in a broad temperature cryostat Janis VPF-700 to perform impedance spectroscopy (IS) measurements as a function of the temperature (77–575 K) in the frequency range of 20 Hz–1 MHz, using an Agilent LCR E4192A apparatus (Santa Clara, CA, USA). Higher frequency measurements were performed in an ultra- broad frequency range from 1 Hz to 10 GHz, and temperatures up to 800 K [
38]. Two methods were combined, namely: a conventional two-electrode method [
39] and a coaxial line method [
40]. A sample voltage of 100 mV at a low frequency range and incident wave power of −10 to 0 dBm in the coaxial line was applied. All of the measurements provided up to 16 frequency points per octave, and were carried out in atmospheric air. The temperature stability during the measurements was around 0.4 K, and the speed of change was 2–3 deg/min. The films were measured in in-plane configuration, the samples of 1.5 mm length were prepared, and Pt electrodes were fired at 800 °C.
For the mechano-elastic characterization of the prepared samples at room temperature, samples with dimensions 6 × 2.5 mm2 were prepared from the large thick film samples, and were mounted in a triple point bending sample holder coupled to a dynamic mechanical analyzer DMA 7 (Perkin Elmer, Norwalk, CT, USA). The measurement conditions were as follows: static stress 6 × 106 Pa, dynamic stress 5 × 106 Pa, strain 0.01%, and frequency 10 Hz.