*5.2. ALD of Metal Fluorides Using Metal Fluorides as the Fluorine Source*

The deposition of many metal fluorides has been studied at the University of Helsinki using TiF4 as the fluorine source, including materials such as MgF2, LaF3 and CaF2 [153,155,157,158]. TiF4 is a relatively safe alternative to HF, since it is a solid at room temperature. It possesses relatively high volatility and thermal stability, combined with high reactivity, which are vital attributes for an effective ALD precursor. The use of TiF4 is made possible by a net ligand exchange reaction with a metal thd-complex (Equation (2)):

$$4\text{M(thd)}\_x\text{ (g)} + x\text{TiF}\_4\text{ (g)} \rightarrow 4\text{MF}\_x\text{ (s)} + x\text{Ti(thd)}\_4\text{ (g)}\tag{2}$$

Other volatile side products, such as TiF*x*(thd)4−*<sup>x</sup>* can form in addition. Recently this precursor was also demonstrated in an ALD process used in conjunction with AlCl3 [156].

Generally, metal fluorides deposited using TiF4 as the fluorine source show a decrease in growth rate as the deposition temperature is increased (Figure 12) [155]. This decrease has been proposed to be due to a decrease in the TiF*<sup>x</sup>* adsorption density but this has not been verified experimentally. Using TiF4 as the fluorine source leads to higher growth rates compared to the use of HF, possibly due to the formation of the fluoride in question during both precursor pulses [155,156]. When using TiF4 as the fluorine source, the film growth usually shows saturation with respect to the fluorine precursor (Figure 13b) but the metal precursor can show either poor [150,153,157] or good saturation [158], depending on the material deposited (Figure 13a). Pilvi et al. postulated that the reason for

the non-saturative behaviour might be either slow kinetics or an enhancement of metal precursor decomposition caused by the TiF*x*-surface groups [153]. Films deposited with TiF4 are generally very close to stoichiometric, as determined with ERDA (Figure 12d). Titanium is often found as an impurity in the deposited films but usually in only very small amounts [150,153,155]. Still, this impurity can limit the UV transmittance of these films when optical applications are the goal. The impurity level decreases as the deposition temperature is increased but at the same time film roughness increases resulting in more scattering of UV-light [153,157,158].

In an effort to obtain even purer metal fluoride films for optical applications, deposition of MgF2 has been studied using TaF5 as the fluorine source [154]. The growth process is very similar to that using TiF4 (Figure 12a) although with using TaF5 saturation with respect to the Mg(thd)2 pulse length is observed. The films contained lower metal impurity levels than those deposited with TiF4 and in addition the films were much smoother at high deposition temperatures [153,154]. This low roughness resulted in improved optical properties.

**Figure 13.** (**a**) Saturation curves for the metal precursor for CaF2, MgF2, YF3 and LiF deposition processes; (**b**) Saturation curves for the fluorine precursor TiF4 for CaF2, MgF2, YF3 and LiF processes. Data obtained from [150,153,155,157].

Following the work of Pilvi et al., a process for the deposition of LiF was developed [150]. Lithd and TiF4 were used as precursors and the deposition temperature was varied between 250 and 350 ◦C. Crystalline LiF films with only small proportions of impurities were obtained at all deposition temperatures. Saturation of the growth was not found for Lithd at 325 ◦C (Figure 13a). TiF4, on the other hand, showed saturation type behaviour between 0.5 and 2.0 s pulses. With longer pulse times, the growth rate increased linearly. Such an increase has not been observed with other processes utilizing TiF4, however in these studies the TiF4 pulse lengths have been limited to 2 s or less (Figure 13b) [153,155,157,158].

The pulsing sequence Mg(thd)2 + TiF4 + Lithd + TiF4 can produce LiF thin films between 300 and 350 ◦C [151]. Unlike the Lithd + TiF4 process of ref. [150], this sequence shows both an ALD window between 325 and 350 ◦C and saturation with respect to both Lithd and TiF4 (Figure 14). The growth rate at 325 ◦C was 1.4 Å/cycle, as opposed to 1.0 Å/cycle in the previous LiF process. All the films were again highly crystalline, with the film roughness being 19–20 nm for 70–80 nm films regardless of the deposition temperature. ToF-ERDA measurements showed the films to be very pure LiF, with only minute amounts of Mg and Ti impurities. C and H formed the largest part of impurities, however both were below 1 at.% in the deposition temperature range of 300–350 ◦C. It is surprising that despite the use of Mg(thd)2 in the pulsing sequence, no magnesium ended up in the LiF films. We have proposed a mechanism to explain the deposition process (Equations (3)–(6)). In Equation (3), Mg(thd)2 and TiF4 deposit MgF2 as has been previously reported [153]. In Equation (4), Li<sup>+</sup> from Lithd replaces Mg2+ in the fluoride film and forms LiF. Magnesium leaves the films as Mg(thd)2, because the low levels of O, C and H impurities imply that virtually no ligand decomposition is occurring during the growth. This type of fluoride-to-β-diketonate ligand exchange might at first seem unexpected, however it has been reported that metal oxides can be dry etched using β-diketone vapours to form volatile β-diketonato complexes of metal ions [174]. After the removal of magnesium, Lithd adsorbs onto the formed lithium fluoride (Equation (5)) and is converted to LiF during the last TiF4 pulse (Equation (6)).

$$2\text{Mg(thd)}\_{2}\text{ (ads)} + \text{TiF}\_{4}\text{ (g)} \rightarrow 2\text{MgF}\_{2}\text{ (s)} + \text{Ti(thd)}\_{4}\text{ (g)}\tag{3}$$

$$\text{MgF}\_2\text{ (s)} + 2\text{Litdd (g)} \rightarrow 2\text{LiF (s)} + \text{Mg(thd)}\_2\text{ (g)}\tag{4}$$

$$\text{LiF (s)} + \text{Lithd (g)} \rightarrow \text{Lithd (ads)}\tag{5}$$

$$4\text{Liftd (ads)} + \text{TiF}\_4\text{ (g)} \rightarrow 4\text{LiF (s)} + \text{Ti(thd)}\_4\text{ (g)}\tag{6}$$

**Figure 14.** (**a**) Growth rate of LiF films as a function of Lithd (black squares) and TiF4 (white squares) pulse lengths at 325 ◦C; (**b**) Growth rate of LiF films as a function of deposition temperature. Data from [151].

Xie et al. have studied the use of LiOt Bu instead of Lithd with TiF4 as the fluorine source [124]. This precursor combination led to the deposition of crystalline LiF between 200 and 300 ◦C. The maximum growth rate of 0.5 Å/cycle was achieved at 250 ◦C. The films had a refractive index close to 1.4 and a Li:F ratio of 1:0.97. Little else has been reported on the process thus far.

Similar to the methodology of using TiF4 or TaF5 as fluorine sources, Mane et al. have reported on the deposition of LiF using LiO<sup>t</sup> Bu and either WF6 or MoF6 as the fluorine source [152]. Film growth took place between 150 and 300 ◦C, with amorphous films being deposited at the lowest temperature. This is an interesting finding, as using LiHMDS and HF-py at 150 ◦C led to crystalline LiF films [147]. With MoF6 as the fluorine source the films had a growth rate of 2.6 Å/cycle, which is much higher than that obtained with the Lithd + TiF4 process [150]. The films showed a 1:1 ratio of Li and F, with very small amounts of oxygen and carbon impurities. Most importantly, metal impurities were not detected with XPS.

After the success of alkaline and alkaline earth metal fluoride deposition, our group studied Al(thd)3 and TiF4 as precursors for AlF3 [156]. However, this precursor combination led to no film growth on silicon and aluminum oxide. Our assumption is that this lack of reactivity has to do with the presence of aluminum-oxygen bonds in the Al(thd)3 complex. Therefore, another aluminum precursor was needed in combination with TiF4. AlCl3 is a widely used ALD precursor, with no oxygen present in the molecule. In addition, TiCl4 is a well-known, volatile ALD precursor [175,176], which is encouraging considering the expected ligand-exchange reaction taking place between AlCl3 and TiF4. Thus, this combination was studied for the deposition of AlF3 [156]. Film growth was observed between 160 and 340 ◦C (Figure 12c). The saturation of the growth rate was studied at 160, 200 and 240 ◦C (Figure 15a). TiF4 showed similar behaviour as in the LiF case [150], with an increased

growth rate with the longest pulse times. AlCl3, on the other hand, showed an opposite trend with growth rates decreasing as a function of pulse time. AlCl3 vapour has been reported to enhance the volatility of AlF3 [177], which might explain the etching-type behaviour seen in this process. As already mentioned, a similar decrease in the growth rate of AlF3 was also observed in the TMA + HF-py process. Despite these effects, the AlF3 growth rate remained constant with different cycle numbers at all temperatures studied (Figure 15b) [156]. ToF-ERDA measurements revealed that the films contained decreasing amounts of Cl and Ti impurities as the deposition temperature was increased, both being well below 1 at.% at 280 ◦C. However, the H and O impurities showed the opposite trend, with the films deposited at 280 ◦C containing up to 6 at% of oxygen.

**Figure 15.** (**a**) Growth rate of AlF3 films as a function of AlCl3 (black) and TiF4 (white) pulse lengths at 160, 200 and 240 ◦C; (**b**) Growth rate of AlF3 films as a function of deposition cycles at various deposition temperatures. The AlCl3 pulse time was 0.5 s and the TiF4 pulse time was 1 s. Data from [156].

Jackson et al. have attempted to combine the methods discussed above for the deposition of AlF3 by using TMA and TaF5 as precursors [145]. This approach is somewhat questionable, as it has been reported that combining TMA with metal halides generally leads to metal carbide deposition [178–180]. In addition, a similar ligand exchange reaction between TMA and TaF5, as was depicted in Equation (2), should produce pentamethyl tantalum which has been reported to be unstable [181,182]. Indeed, significant amounts of TaC*x* were deposited at elevated temperatures [145]. At 125 ◦C the content of tantalum impurity was decreased and the process showed ALD-like behaviour. The films contained approximately 20 at.% of oxygen, meaning that the process is unable to deposit good quality AlF3. Despite the large amounts of impurities, the deposition of this material onto a high-voltage lithium-ion battery cathode nickel-manganese-cobalt oxide (NMC) led to significant improvements in its rate performance (Figure 16).

Similar to Jackson et al., Park et al. used TMA with WF6 to deposit an amorphous composite fluoride composed of AlF3 and metallic W and WC*<sup>x</sup>* [144,159]. The material was studied as an artificial SEI layer for LiCoO2 cathodes and was found to improve the cycling properties of the material. It appeared that the composite nature increased the electron conductivity of the fluoride layer while still retaining its chemical inertness.

Our group has also studied ternary fluoride deposition using metal fluorides as precursors [139]. The deposition of Li3AlF6 proved complicated when using the binary processes of LiF and AlF3 described in Refs. [150,151,156]: uncontrollable conversion of AlF3 to LiF by Lithd was a threat to using the ALD subcycle approach. More importantly, exposing LiF to AlCl3 could result in undesirable LiCl formation. Thus, attempts at depositing Li3AlF6 were made using two processes (Figure 17) [139]. In Process 1, Al(thd)3 and TiF4 were pulsed sequentially onto LiF thin films and a conversion reaction

to the ternary Li3AlF6 took place during the deposition process. In Process 2, AlF3 films were exposed to Lithd vapour in an ALD reactor (described in more detail in section "5.3. Other Approaches to ALD of Metal Fluorides"). In Process 1, Al(thd)3 was used instead of AlCl3 to avoid Li+ contact with Cl− from AlCl3. Despite the prior knowledge that Al(thd)3 and TiF4 do not produce AlF3 on silicon substrates [156], a reaction did occur between these precursors when pulsed onto LiF films. The fluorides mixed together during the deposition process, resulting in crystalline Li3AlF6 with crystalline LiF residues. High deposition temperatures together with long Al(thd)3 pulses resulted in less LiF impurity in the film, as observed with GIXRD. However, these same conditions worsened the visual appearance of the films. ToF-ERDA revealed that even in the best samples, the content of Al was very low, although Li3AlF6 was clearly visible in the X-ray diffractograms. In addition, the level of titanium impurity was high. Thus, it was concluded that in the end Process 1 was not efficient in depositing Li3AlF6.

**Figure 16.** Gravimetric capacity of coin cells as a function of cycle number with different discharge rates. Black squares denote an NMC-cathode coated with TMA + TaF5. Red circles denote uncoated NMC. Reprinted with permission from [145]. Copyright 2016 American Vacuum Society.

**Figure 17.** Process 1 utilizes Al(thd)3 and TiF4 in a conversion reaction to form Li3AlF6 out of LiF thin films. Process 2 uses a conversion reaction between ALD-made AlF3 and the lithium precursor Lithd to deposit Li3AlF6. Reprinted with permission from [139]. Copyright 2017 Elsevier.

Recently, Xie et al. published their results on the deposition of LiAlF4 [124]. By combining the processes for LiF and AlF3 using LiOt Bu, AlCl3 and TiF4 as precursors they were able to deposit amorphous LiAlF4 with a Li:Al ratio of 1.2:1 and an ionic conductivity of 3.5 × <sup>10</sup>−<sup>8</sup> S/cm. The LiAlF4 film was used as a protective layer on top of a lithium nickel manganese cobalt oxide cathode and was found to improve the stability of the cathode without sacrificing its rate performance.
