*5.1. ALD of Metal Fluorides Using HF as the Fluorine Source*

ZnF2, SrF2 and CaF2 films, reported in the first paper on ALD of fluorides in 1994, were deposited using HF as the fluorine source [141]. The HF gas was generated in the reactor in situ by thermal decomposition of ammonium fluoride, NH4F. Thus, there was no need to store and handle large amounts of gaseous HF. An added benefit of this method was that excess HF can be condensed again inside the reactor as ammonium fluoride, without the gas entering and damaging the vacuum pump. Metal thd-complexes were used as precursors for strontium and calcium and zinc fluoride was deposited using zinc acetate. All the films were close to stoichiometric and polycrystalline, with carbon impurities of the order of 0.5 at.%. For the calcium and strontium fluoride processes, the growth rates decreased with increasing deposition temperatures (Figure 12b).

The work on fluoride deposition using HF has been continued by many groups, including Hennessy et al., who have deposited magnesium and aluminum fluoride films using anhydrous HF with bis(ethylcyclopentadienyl)magnesium and TMA as metal precursors [148,149]. Magnesium fluoride is an interesting material due to its large band gap and low refractive index [142,153]. Aluminum fluoride is a material with a similar variety of possible optical applications [149,162–164]. In addition, as already mentioned, AlF3 is a potential artificial SEI-layer for protecting both cathodes and anodes [47,143,145,165–167]. Thus, AlF3 has become a much studied ALD material in the past few years [143,145,147,149,156]. Magnesium fluoride showed growth rates of 0.6 to 0.3 Å/cycle in the deposition temperature range of 100 to 250 ◦C (Figure 12a) [148]. AlF3 showed a similar decrease in growth rate, being 1.2 Å/cycle at 100 ◦C and 0.5 Å/cycle at 200 ◦C (Figure 12c) [149]. MgF2 films were crystalline and showed small amounts of carbon and oxygen impurities and a slight fluorine deficiency in XPS measurements (Figure 12d) [148]. AlF3 films were amorphous, with 1–2 at.% of oxygen [149]. The aluminum fluoride films were stoichiometric based on XPS measurements. The anhydrous HF required an unconventionally long purging time to obtain good film uniformity. It was speculated that multilayer physisorption might be the cause of this effect. However, it has been reported in another publication that MgF2 does not readily adsorb HF during the ALD growth process [147].

**Figure 12.** (**a**) Growth rates of MgF2 films as a function of deposition temperature using different precursor combinations; (**b**) Growth rates of CaF2 films as a function of deposition temperature using different precursor combinations; (**c**) Growth rates of AlF3 films as a function of deposition temperature using different precursor combinations; (**d**) F-Mg ratios of MgF2 films deposited at different temperatures and with different precursor combinations. The HF-process was analysed with XPS and the TiF4 and TaF5 processes with ERDA. Data obtained from References: (**a**) [147,148,153,154]; (**b**) [141,155,160]; (**c**) [143,145,149,156]; (**d**) [148,153,154].

A number of metal fluorides have recently been deposited by Lee et al. using HF, including AlF3 [143], LiF, ZrF4, ZnF2 and MgF2 [147]. Of these materials, lithium fluoride is of special interest because of its band gap of approximately 14 eV and low refractive index of 1.39 at 580 nm, much like AlF3 as discussed previously [168,169]. For the deposition of these fluorides, HF was generated from a solution containing 30 wt.% of pyridine and 70 wt.% HF ("*Olah's reagent*") to mitigate the safety concerns of anhydrous, gaseous HF. This solution is in equilibrium with gaseous HF, with no pyridine detected in the gas phase and provides a safer alternative to anhydrous HF [143,147]. Metal precursors used included a diethylcyclopentadienyl complex for magnesium, a silylamide for lithium and an alkylamide for zirconium. All processes resulted in saturation at 150 ◦C, with growth rates below 1 Å/cycle (Figure 12). All films, except AlF3 and ZnF2, were crystalline. Generally, the films contained less than 2 at.% of oxygen impurities, as determined with XPS. Only ZrF4 contained some carbon impurities in addition to the oxygen. The films appeared somewhat fluorine deficient, however this is speculated to be a result of preferential fluorine sputtering during the XPS measurement [143,147]. The AlF3 deposition from TMA and HF showed an interesting etching reaction: above 250 ◦C the precursor pulses etched the AlF3 film [143,149]. This reaction has later been exploited in developing atomic layer etching processes [170–172]. The AlF3 process has been successfully utilized in protecting freestanding LiCoO2/MWCNT (multi-walled carbon nanotube)/nanocellulose fibril electrodes, showing better protection at high potentials compared to the more common artificial SEI material Al2O3 [173].

Li3AlF6 has been deposited using subcycles of AlF3 and LiF using TMA and HF-pyridine and LiHMDS and HF-pyridine as precursors [51]. One subcycle of AlF3 and one subcycle of LiF were used at 150 ◦C and monoclinic Li3AlF6 was obtained with a growth rate of 0.9 Å/cycle [51,146]. The films had a Li:Al ratio of 2.7:1, as determined with ICP-MS and carbon, silicon and oxygen impurities were below the XPS detection limit. The films had an ionic conductivity of 7.5 × <sup>10</sup>−<sup>6</sup> S/cm at room temperature [146], which is similar to the first reports from the literature on thermally evaporated amorphous Li3AlF6 films [12,13]. Interestingly, changing the pulsing ratio to three subcycles of AlF3 and one LiF resulted in the same metal ratio in the as-deposited film as the pulsing ratio of 1:1, suggesting a similar conversion reaction as we have seen in our experiments on LiF [151] and Li3AlF6 [139].
