Deposition and characterization of thin films for 3D lithium-ion micro-batteries

Electricity has since its first discovery evolved to become one of the most commonly used energy carriers. It is clean and versatile and enables a large part of the technology that is currently playing an important role in everyday life. Storage of this electrical energy has been of interest since its discovery. One of the main storage methods of electrical energy is the use of galvanic cells, which are to the general public better known as batteries or accumulators. These cells have made feasible that electrical energy is not only available in stationary applications, but also for mobile applications. The miniaturization of these mobile wireless applications has created a demand for miniaturized batteries. Especially now a trend is observed towards distributed sensor networks, which consist of individual miniature sensor devices that communicate wirelessly with a base station. The energy for these sensors can be supplied by an energy harvester, e.g. based on photovoltaic, thermoelectric or piezoelectric generators. This energy needs, however, to be stored to ensure a continuous energy supply to the sensors. For this purpose, all-solid-state micro-batteries are a suitable form of energy storage. The disadvantage of these micro-batteries is, however, that their energy density is relatively low. To increase the energy density, the concept of 3D integrated all-solid-state thin-film micro-batteries was proposed. This concept is described in Chapter 1. Several concepts and fabrication methods for 3D all-solid-state batteries have been proposed in recent literature, which are reviewed in Chapter 2. Key factors of 3D micro-batteries are the formation of a 3D geometry and, subsequently, the conformal deposition of battery materials. One method for the fabrication of 3D geometries is the use of anisotropic etching of silicon. This is a well-established method that is commonly applied in the semiconductor industry. The focus of the research in this dissertation lies on the deposition of thin-film battery materials using Low Pressure Chemical Vapour Deposition (LPCVD), and the characterization of the deposited films. In Chapter 3 the used LPCVD set-up and other applied deposition and characterization methods are described. Because LPCVD is not commonly applied as a deposition technique for battery materials, the deposition parameters were first systematically investigated for a standard cathode material in lithium-ion batteries. The results of these experiments are described in Chapter 4. It was concluded from these measurements that electrochemically active LiCoO2 can be deposited using LPCVD. The composition of the deposited film can be controlled using the composition of the precursor gas. A sufficiently high temperature is required for the deposition of crystalline electrochemically active LiCoO2. Finally, it was observed that the deposited LiCoO2 revealed a stable cycling stability when combined with a solid-state electrolyte. Li3PO4 and LiTaO3 were investigated as solid-state electrolyte candidates in Chapter 5. These layers both have a similar ionic conductivity, in the order of 5-6·10-8 S·cm-1. The deposited Li3PO4 was, however, more stable at low potentials, which opens more possibilities to use it in combination with various anode materials. Reduction reactions occurred for LiTaO3 films at 1.8 V vs. Li/Li+, while Li3PO4 was stable at substantially lower voltages. And even though the composition control of Li3PO4 films was less straightforward than in the case of LiTaO3, it was concluded that because of the larger voltage window Li3PO4 is a more suitable material for allsolid-state thin-film micro-batteries manufactured by LPCVD. Several anode materials are explored in Chapter 6. Co3O4, MoO3, Ru and RuO2 were deposited using LPCVD and their properties were examined. All these materials had their own advantages and disadvantages, which can play a role for the choice of the material for the battery to be deposited. Co3O4 is practically a convenient battery material, because the same cobalt precursor can be used as for LiCoO2. It showed, however, a large difference in lithiation and delithiation voltage (0.8 and 2.0 V, respectively) and the observed cycling stability was very limited. It should, however, be noted that these measurements were performed in liquid electrolyte, and that further stability testing should be performed in combination with solid-state electrolytes. MoO3 is an electrode material that shows electrochemical activity at around 2.5 V. It can therefore be applied as either anode or cathode, depending on the voltage level of the other electrode. The main disadvantage of this material is that it is complex to control the morphology in the condition range available in the used LPCVD set-up. The deposition of RuO2 is well established in LPCVD, and also in this case it was demonstrated successfully. However, also for this material the cycling stability in a liquid electrolyte is limited. The analyses of fully working all-solid-state batteries, using Neutron Depth Profiling (NDP) are reported in Chapter 7. This technique is capable of making depth profiles of thin-film batteries while these are being cycled. The measured profiles agree well with the measured electrochemical responses, and it can be applied to evaluate the degradation of thin-film micro-batteries uponcycling. A general capacity loss mechanism for all-solid-state lithium-ion batteries involves the loss of mobile lithium as a charge carrier. It was concluded that for the analysed planar all-solidstate micro-batteries, which was deposited using Physical Vapour Deposition (PVD), this was indeed the case. Lithium became in this battery immobilized at the anode side of the solid-state electrolyte. One possible explanation for this degradation mechanism is the formation of a lithium-silicon-phosphate. The average composition of this material was calculated to be Li3.3Si0.1PO4. The previous experiments were performed for thin films on planar substrates. Conformal deposition in 3D structures is more complex, as is described in Chapter 8. It is shown that the conformality of the deposition is strongly dependent on the geometry of the 3D-structure and the deposition temperature. The geometry is a variable that cannot be easily modified, as each new structure will involve the design of a new patterning mask and a new etching optimization process. Therefore, a single procedure was developed to produce a silicon substrate containing trenches with different widths. Trenches were chosen as the shape because these are most suitable for cross-section imaging and because this geometry can easily be applied to run 2D simulations. The investigated trenches were 30, 10, 3 and 1 µm wide and 30 µm deep. The second variable influencing the step conformality is the deposition temperature. Two different processes form the basis of LPCVD: transport of precursor molecules from the gas phase to the surface of the substrate, and the chemical reactions at the surface. Important for conformal 3D-deposition is that the deposition rate is controlled by the surface reaction. This chemical reaction rate is strongly dependent on the deposition temperature, while the temperature dependence of the transport processes is much weaker. Therefore, LPCVD processes generally have a low-temperature regime in which the deposition rate is reaction rate limited, and a higher temperature range where transport limits the deposition rate. For the deposition of Ta2O5, which is the basis of the LiTaO3 electrolyte, these two temperature ranges were clearly observed, and a good conformal deposition was obtained for all trench widths. For TiO2, a potential electrode material, the layer thickness, and thus the deposition rate, were more difficult to determine due to the polycrystalline appearance of the film at some temperatures. However, it was observed that at the lowest temperature at which deposition is possible the deposition is not step-conformal. LiCoO2 is the proposed cathode material. This is a more complex metal oxide, because both lithium and cobalt need to be deposited. With LiCoO2 it was also observed that at low temperature the best conformality was achieved. However, it was previously observed that deposition at high temperatures is required to obtain electrochemically active, high temperature, modification of LiCoO2. This electrochemically active LiCoO2 can, however, also be formed using deposition at low temperature followed by a high temperature anneal step. With this approach a film with 80 % step conformality could be achieved, of which it was found that the ratio of Li:Co was the same, at the top, in the centre and at the bottom of a 50 µm deep trench. Although TiO2 was not deposited conformally it may well serve as a model material to simulate the deposition process. The deposition of TiO2 using Titanium Tetra Iso Propoxide (TTIP) as precursor is most suitable to model the deposition process as the precursor already contains oxygen, and therefore no extra oxygen is required to deposit TiO2. The number of species involved in the modelling is therefore limited. Moreover, the resulting model can lead to an improvement of the 3D deposition with a reduced experimental effort. First, the deposition of TiO2 was investigated experimentally. From these measurements it was concluded that using LPCVD anatase TiO2 could be deposited and that these films were indeed lectrochemically active. Subsequently, two model approaches were used to simulate the deposition process: a multi-scale finite elements model and a Monte Carlo simulation. The multi-scale finite elements method uses three different length scales: the reactor scale (10-30 cm), the micro-scale of the trenches (10-30 µm) and a meso-scale, which is used as a transition scale between the macro and micro-scale. This model has the advantage that it is a model for the complete system, involving every aspect of the deposition process. That is, however, also its disadvantage: it is a complex model that needs a large number of input variables. It was observed that the agreement of the model and experimental results were only limited. Monte Carlo simulations, on the other hand, are more simple, and involves only a particle model on the micro-scale. This model managed to obtain a good agreement with experimental results, but can only be applied for the prediction of experimental results inside micro-trenches. Overall, it is concluded in Chapter 10 that LPCVD is a successful method to deposit electrode and electrolyte materials for thin-film all-solid-state micro-batteries. The conformal deposition was investigated both experimentally and using modelling approaches. Moreover, it was demonstrated that NDP can be applied as a technique to in situ study the lithium distribution inside thin-film batteries. Furthermore, several approaches for future research were suggested, e.g. the research to single-source precursors and growth inhibitors to tune the chemical reaction rate. Finally, because the properties of batteries can be tuned with the choice of materials, a well matched combination of a 3D battery and application will yield the most efficient device.