*5.3. Other Approaches to ALD of Metal Fluorides*

Putkonen et al. provided other interesting pathways for avoiding the use of HF in fluoride deposition by depositing metal fluorides through oxide chemistry [160]. They found that by using the fluorinated β-diketonate precursor Ca(hfac)2 and ozone as precursors CaF2 is deposited, instead of CaCO3 that is formed when the non-fluorinated β-diketonate Ca(thd)2 is used together with ozone [183]. The fluoride films had a growth rate of 0.3 Å/cycle at 300 ◦C and were close to stoichiometric, although approximately 5 at.% of oxygen was present in the films [160]. An even more interesting approach to CaF2 used the non-fluorinated metal precursor Ca(thd)2 in combination with ozone and the Hhfac molecule. First, Ca(thd)2 was pulsed onto the substrate followed by an ozone pulse, resulting in CaCO3 deposition. Then Hhfac, which is known to adsorb to surfaces, was pulsed followed again by an ozone pulse. Ozone breaks down the Hhfac on the surface, providing fluoride ions which react with the calcium ions, resulting in a conversion reaction from CaCO3 to CaF2. This process provided a growth rate of 0.4 Å/cycle between 250 and 350 ◦C, which is close to the rate of CaCO3 deposition from Ca(thd)2 and ozone in similar conditions. The CaF2 films were polycrystalline and close to stoichiometric and the amount of oxygen was below the detection limit of Rutherford backscattering spectroscopy (RBS). The same approach was reported to be successful also for MgF2 and LaF3 deposition.

Vos et al. recently reported on an interesting process for depositing AlF3 using TMA and SF6 plasma [161]. Plasma-enhanced ALD processes have not been previously available for fluoride deposition. However, F-containing plasmas, such as CF4, have been widely used for etching processes. In this work, SF6 was used because it is stable and non-toxic. Inductively coupled SF6 plasmas contain species such as F, F2, SF5+ and F<sup>−</sup> with S and S+ as minor components. The AlF3 deposition showed saturation with both TMA and the plasma with good film uniformity and conformality. As with all other AlF3 processes, the film growth rate decreased with increasing deposition temperature, being 0.85 Å/cycle at 200 ◦C. All the films were amorphous with very low S and O impurity levels and an Al:F ratio close to stoichiometric.

After noting that the pulsing sequence Mg(thd)2 + TiF4 + Lithd + TiF4 produced LiF with little to no Mg impurities, our group studied the conversion reaction of MgF2 films upon exposure to Lithd [151]. As it turned out, with high enough Lithd doses MgF2 films of 150 nm in thickness could be converted into LiF with no indication of MgF2 or Mg impurities in GIXRD, EDX or ToF-ERDA measurements. The lower the reaction temperature, the larger the Lithd dose needed to completely convert the MgF2 film into LiF. Although the resulting LiF films were again highly crystalline as determined with GIXRD, they showed much smaller grain sizes and thus lower roughness than the films deposited with either the two [150] or four [151] step LiF processes (Figure 18). In addition, the adhesion of the films to the silicon substrates was markedly improved. Our experiments demonstrated that the conversion reaction with Lithd is not limited to the surface regions of MgF2 films but can in fact proceed very deeply into the films. The high mobility and reactivity of Li+ seen in these experiments is likely to play a role in many processes used to deposit materials containing lithium, especially in the case of ternaries. For example, Miikkulainen et al. later reported similar results in their conversion experiments to form spinel LiMn2O4 using MnO2 films and Lithd, as was already discussed in a previous section [90]. However, their conversion reaction also led to significant amounts of hydrogen and carbon impurities when no ozone was used after the Lithd pulse, as opposed to our conversion reactions of MgF2 with Lithd which led to purer film products.

**Figure 18.** AFM images of LiF films deposited at 325 ◦C using three different processes: (**a**) Lithd + TiF4, thickness 73 nm, rms roughness 15.9 nm; (**b**) Mg(thd)2 + TiF4 + Lithd + TiF4, thickness 68 nm, rms roughness 20.1 nm; (**c**) conversion from a MgF2 film using Lithd, thickness 94 nm, rms roughness 4.8 nm.

Li3AlF6 was also deposited using a similar conversion reaction as with MgF2 and Lithd (Process 2, Figure 17) [139]. Approximately 50 and 100 nm thin films of AlF3 were exposed to the lithium precursor at different temperatures and for different Lithd pulse numbers to determine whether good quality Li3AlF6 could form from this conversion reaction. GIXRD analyses showed that amorphous AlF3 transformed into monoclinic Li3AlF6 with Lithd exposure. Choosing too large a number of Lithd pulses resulted in crystalline LiF formation. The mechanism of the conversion is most likely similar to the MgF2 conversion, as the oxygen, carbon and hydrogen impurity levels were very low in the films after the conversion, as determined with ToF-ERDA. Despite crystalline Li3AlF6 being visible in the X-ray diffractograms, obtaining the correct Li:Al ratio was challenging. ToF-ERDA revealed that doubling the Lithd exposure from 20 to 40 pulses increased the Li:Al ratio from 0.93:1 to 7.9:1 for approx. 50 nm AlF3 films (Figure 19). With 100 nm AlF3 films exposed to 40 Lithd pulses, the ratio varied between 1.33:1 and 1.49:1. Thus, thinner films were much faster to convert than thicker ones, as was to be expected. The exposure temperature also played a role in the conversion. Just a 25 ◦C increase from 250 to 275 ◦C increased the Li:Al ratio for a 40 pulse sample from 1.49:1 to 2:1. However, despite the lithium deficient metal ratio, the 275 ◦C exposure temperature sample already contained a prominent amount of crystalline LiF based on GIXRD. The converted films showed a porous structure in FESEM (field emission scanning electron microscopy) (Figure 20), preventing ionic conductivity measurements due to top electrodes short circuiting with the bottom electrode.

**Figure 19.** The content of lithium cations in converted AlF3 films as a function of the number of Lithd pulses. Black and white symbols denote samples prepared at different times but with same exposure parameters. Dotted and dashed lines illustrate that the content of lithium increases linearly with the number of Lithd pulses. The solid line illustrates the correct metal stoichiometry of Li3AlF6. Data from [139].

**Figure 20.** FESEM images of (**a**) ~90 nm crystalline Li3AlF6 film made by conversion from AlF3 and Lithd and (**b**) ~50 nm crystalline Li3AlF6 film made by conversion from AlF3 and Lithd and taken by tilting the sample.

## **6. Conclusions**

Metal fluorides could provide an interesting, high voltage and high capacity alternative to current oxide based lithium-ion battery cathode materials. However, because most metal fluorides act as conversion cathodes, special effort must be made to enhance their cycling ability and to alleviate problems related to electrode pulverization. For some fluorides, encouraging lithium-ion conductivities have been measured, meaning that they could act as highly stable solid electrolytes in all-solid-state Li-ion batteries. In addition, fluoride thin films have been shown to function as artificial solid-electrolyte-interface layers, protecting both cathodes and anodes from metal dissolution and other undesired side reactions during battery cycling. All these results show that there is a lot of untapped potential in this class of battery materials, worthy of more research. For future applications, methods producing high quality, conformal fluoride thin films with precise thickness control are needed, for example for depositing cathodes in intricate 3D structures and for depositing protecting films of 1 nm or less on current electrode materials. Atomic layer deposition can be the method for answering these demands. The lithium-ion exhibits unique ALD chemistries through its high mobility, which needs to be taken into account when designing processes for battery material deposition by ALD.

Many metal fluoride thin films have been deposited using ALD, however the focus of these studies has mainly been on optical applications. Only very recently studies on the deposition of materials such as AlF3 and L3AlF6 have emerged, with the main motivation of utilizing these materials in lithium-ion batteries. In the future, much more research effort should be put into depositing transition metal fluorides and multi-component fluorides for use in lithium-ion batteries. These materials have received next to no interest in the past 25 years of ALD fluoride studies, meaning there is much to be discovered in this area of materials science. These future studies will undoubtedly greatly benefit from recent studies utilizing new fluorine sources such as metal fluorides and SF6 plasma.

**Author Contributions:** Writing–Original Draft Preparation, M.M.; Writing–Review & Editing, M.M.; M.R.; M.L.; Visualization, M.M.; Supervision, M.R.; M.L.

**Funding:** This work was funded by the Finnish Centre of Excellence in Atomic Layer Deposition (284623).

**Acknowledgments:** Peter J. King is thanked for proof-reading this work.

**Conflicts of Interest:** The authors declare no conflict of interest.
