4.2.3. Solid Electrolytes

ALD has showcased its versatility in the deposition of potential solid electrolytes for lithium-ion batteries (Table 6). Many materials, both crystalline and amorphous, have been deposited and ionic conductivities of the order of 10−<sup>7</sup> S/cm have been obtained with various materials. Of the traditional solid electrolytes, LiPON has been deposited using both ALD and PEALD [108,109].

**Table 6.** Examples of potential solid lithium-ion battery electrolyte materials deposited by ALD. Abbreviations used: LiHMDS = lithium hexamethyldisilazide, LiTMSO = lithium trimethyl silanolate, Ot Bu = *tert*-butoxide, TEOS = tetraethyl orthosilicate, TMPO = trimethyl phosphate, DEPA = diethyl phosphoramidate, OEt = ethoxide, thd = 2,2,6,6-tetramethyl-3,5-heptanedionato, TMA = trimethyl aluminum, La(FMAD)3 = lanthanum tris(N,N-di-*iso*-propylformamidinate), TDMAZ = tetrakis(dimethylamido)zirconium , TDMA-Al = tris(dimethylamido) aluminum, HF-py = mixture of HF and pyridine.


Lithium silicates can have reasonably high lithium-ion conductivities, especially in the amorphous state [125–127]. The silylamide precursor LiHMDS (lithium hexamethyldisilazide) provides a convenient route to lithium silicate deposition when combined with ozone [110,111]. The process exhibited good ALD behaviour at 250 ◦C, with saturation of both precursors seen and the film thickness increased linearly with the number of cycles. However, no ALD window was present and instead the growth rate increased from approximately 0.3 Å/cycle at 150 ◦C to 1.7 Å/cycle at 400 ◦C. This increase was explained with subsequent reaction mechanism studies [111]. The HMDS-ligand of the metal precursor reacts with surface hydroxyl groups, decomposing to different side products. Some of these side products are unreactive in the process but can still block active sites from the desired –Si(CH3)3 groups. At higher temperatures, the decomposition of the ligand to –Si(CH3)3 is enhanced and in addition desorption of unreactive products is faster. The deposited films were amorphous below 400 ◦C and showed only small amounts of carbon and hydrogen impurities, as determined by ERDA [110]. Notably, no nitrogen was detected in the film despite the lithium precursor being a silylamide. The Li:Si and Si:O ratios changed with deposition temperature but at 250 ◦C the film composition was Li2SiO2.9 which is very close to the lithium metasilicate Li2SiO3. The ionic conductivity of these films was not measured.

Recently, lithium silicates have been deposited also using LiTMSO (lithium trimethyl silanolate) [112] and lithium *tert*-butoxide [113]. With LiTMSO, to obtain good quality films both an ozone and a water pulse were needed after the metal precursor pulse [112]. It was postulated that the water generates hydroxyl groups on the surface of the silicate film, which are beneficial for the adsorption of LiTMSO. All the films were amorphous and the growth rate remained constant at 1.5 Å/cycle between 200 and 300 ◦C. Films deposited at low temperatures had significant hydrogen contents of 14 at.%. The amounts of both lithium and silicon increased with increasing deposition temperatures while the levels of impurities decreased but the Li:Si ratio remained at 2:1. No ionic conductivity information is available for these films.

Lithium *tert*-butoxide could deposit lithium silicates in combination with TEOS (tetraethyl orthosilicate) and water [113]. For these films, a Li:Si ratio close to Li4SiO4 was obtained at all deposition temperatures. The ionic conductivity of these films was quite low, reaching a maximum of <sup>5</sup> × <sup>10</sup>−<sup>9</sup> S/cm in films deposited at 250 ◦C.

LiPON, currently the most often used solid lithium-ion electrolyte material, was undoubtedly the stimulus for the ALD studies on lithium phosphate films. Li3PO4 can be deposited using either LiO<sup>t</sup> Bu or LiHMDS as the lithium source and TMPO (trimethyl phosphate, Figure 9) as the phosphate precursor [114,115]. The LiO<sup>t</sup> Bu + TMPO process showed a constant growth rate of approximately 0.7 Å/cycle between 225 and 275 ◦C [114]. However, no complete saturation was observed. The films were slightly crystalline and showed decreasing impurity levels at higher deposition temperatures in ERDA measurements. At 300 ◦C, the film composition was Li2.6PO3.7. The process utilizing LiHMDS is less than ideal, as the film growth rate varies strongly with deposition temperature, being 0.4 Å/cycle at 275 ◦C and 1.3 Å/cycle at 350 ◦C. At 300 ◦C, these films were close to stoichiometric lithium phosphate, being Li2.8PO3.9 as determined by ERDA. However, using LiHMDS as the lithium precursor led to higher carbon and hydrogen impurities than the LiOt Bu process. Regardless of the lithium precursor, the phosphate films crystallized into the orthorhombic Li3PO4 phase during HTXRD measurements.

**Figure 9.** The structures of three phosphorus precursors used to deposit lithium phosphate and LiPON. (**a**) TMPO, or trimethyl phosphate; (**b**) DEPA, or diethyl phosphoramidate; (**c**) TDMAP, or *tris*(dimethylamino)phosphine.

Wang et al. [115] and Létiche et al. [116] have studied the Li3PO4 process using LiO<sup>t</sup> Bu and TMPO in an effort to measure the lithium-ion conductivity of these films. Wang et al. reported an increasing growth rate as a function of the deposition temperature at 250–325 ◦C, which might be a result of the somewhat unsaturative behaviour of the process [114]. Electrochemical impedance spectroscopy showed that the films had rather good conductivities when deposited at 300 ◦C: 3.3 × <sup>10</sup>−<sup>8</sup> S/cm was extrapolated for a film with a composition of Li2.8POz (as determined by XPS) [115]. Similarly, Létiche et al. reported conductivities as high as 4.3 × <sup>10</sup>−<sup>7</sup> S/cm for Li3PO4 deposited at 300 ◦C [116]. These results are rather surprising, as it is common knowledge that lithium phosphate is generally no match for its nitrogen-doped counterpart LiPON and its conductivities of 10−8–10−<sup>6</sup> S/cm [32]. It appears that small film thicknesses can play a role in these high ionic conductivities [116]. Li3PO4 layers have been studied in contact with electrode materials [128,129] and it has been found that although the phosphate layer can decrease the electrode capacity, capacity retention is improved due to decreased transition metal dissolution and more stable SEI formation [128].

Recently, the deposition of LiPON was achieved both by thermal and plasma-enhanced ALD [108,109,130]. In the PEALD process, LiO<sup>t</sup> Bu was used as the lithium source combined with a pulsing sequence of water, TMPO and nitrogen plasma [109]. Deposition of Li2O/LiOH before exposure to TMPO resulted in less carbon impurities as compared to the process used by Hämäläinen et al. for Li3PO4 [114]. By using nitrogen plasma after the TMPO pulse, nitrogen could be incorporated into the films, causing the amorphization of the crystalline Li3PO4. In the thermal ALD process, the problems of nitrogen incorporation and nitrogen-phosphorous bond formation were resolved by using diethyl phosphoramidate, DEPA, a phosphate precursor with an amine group (Figure 9) [108]. By using DEPA with LiO<sup>t</sup> Bu, nitrogen contents as high as 9.7 at.% were achieved. However, the thermal ALD process led to high carbon impurities from 9.9 to 13.3 at.% compared to virtually none in the PEALD process [109]. Both processes deposited conformal coatings on demanding substrates as required from a potential solid electrolyte material. In addition, very good electrochemical properties were realized with ionic conductivities of 1.45 × <sup>10</sup>−<sup>7</sup> S/cm for the PEALD process (5 at.% nitrogen) and 6.6 × <sup>10</sup>−<sup>7</sup> S/cm for the thermal ALD process (9.7 at.% nitrogen) [108,109]. The plasma-deposited LiPON has already been studied as a protecting layer for a conversion lithium-ion battery electrode. It was found that the LiPON layer enhanced the capacity retention of the electrode by providing both a high lithium-ion conductivity and mechanical support during cycling [131].

The most recent addition to the ALD LiPON processes was reported by Shibata, using TDMAP or *tris*(dimethylamino)phosphine (Figure 9) as the phosphorous source, LiO<sup>t</sup> Bu as the lithium source and O2 and NH3 for oxidation and nitrification [130]. The high process temperatures of over 400 ◦C raise concerns of more CVD- than ALD-type film growth, especially combined with the changing growth rate as a function of cycles. However, no carbon impurities were found in the films with XPS. The N contents varied between 2 and 6 at.% and an ionic conductivity of 3.2 × <sup>10</sup>−<sup>7</sup> at 25 ◦C was obtained for these films.

Lithium tantalate, similarly to lithium niobate, is an interesting ferroelectric material [132,133]. Its amorphous form has also been suggested as a possible solid electrolyte material for lithium ions [117,134,135]. The material has been deposited by ALD using LiO<sup>t</sup> Bu, Ta(OEt)5 and water as precursors at 225 ◦C [117]. The film growth rate changed depending on the cycle ratio of the two binary processes, being 0.74 Å/binary cycle with a 1Li2O + 6Ta2O5 pulsing sequence. Similarly, the lithium contents of the films changed drastically with pulsing ratio (Figure 10). Both XANES (X-ray absorption near edge structure) and XPS measurements revealed that the chemical environment of tantalum in the films was similar to that of tantalum in stoichiometric LiTaO3. However, in the films deposited with the highest tantalum oxide pulsing ratios there were also some indications of a Ta2O5 phase. XPS also revealed some carbonate formation on the film surface. Less carbonate was formed on the surface of films deposited with high numbers of tantalum oxide subcycles, indicating that Ta2O5 was offering some protection for the lithium in the film against reactions with carbon dioxide in air. A lithium tantalate film with a composition of Li5.1TaO*<sup>x</sup>* was studied with electrochemical impedance spectroscopy (EIS) [117]. The film showed a room temperature lithium-ion conductivity of 1.2 × <sup>10</sup>−<sup>8</sup> S/cm, which increased to 9.0 × <sup>10</sup>−<sup>7</sup> S/cm at 100 ◦C. The material has later been used as a protective layer on lithium nickel cobalt manganese oxide cathodes [135]. With 5 supercycles of LiTaO3 (metal oxide pulsing ratio Li:Ta = 1:6), enhancements in both electrode capacity and cycling ability were obtained. Recently, LiTaO3 films were also made using ALD and solid state reactions: Li2CO3 was deposited from Lithd and O3 onto amorphous ALD-Ta2O5 and upon annealing at 750 ◦C in air crystalline LiTaO3 was formed with low impurity levels and a Li:Ta ratio of 1.5:1 [104]. The ionic conductivity of this film has not been measured but it can be expected to remain small due to both the crystalline structure of the film and its close-to-stoichiometric content of lithium.

**Figure 10.** The amount of lithium cations deposited into lithium tantalate and lithium niobate films as a function of the amount of lithium containing subcycles. Data points obtained from [117–119]. The solid line represents the stoichiometry obtained for the lithium tantalate deposited by using 3000 cycles of Lithd and O3 on 50 nm of Ta2O5 in Ref. [104]. The dashed line represents the stoichiometry obtained for the lithium niobate deposited by using same conditions but on Nb2O5 film in Ref. [104]. For stoichiometric LiTaO3 and LiNbO3, the lithium content is 50%.

Similar to lithium tantalate, lithium niobate thin films have also been deposited by ALD [118,119]. In the first paper, LiNbO3 was deposited using Nb(OEt)5, LiHMDS and water as precursors [118]. This work focused on the evolution of the lithium content in the films (Figure 10), on the epitaxial growth of the film on various surfaces and on the ferroelectric properties of the films. Later, the material was also deposited with LiOt Bu as the lithium source [119]. It is interesting to note that while the two processes use different lithium precursors, they produce films with the same stoichiometry when the pulsing sequence is Li:Nb = 1:1 (Figure 10). For other pulsing ratios, the LiO<sup>t</sup> Bu process seems to produce much higher lithium contents but this might be an artefact caused by the differing analysis methods used in these reports, XPS [119] and ToF-ERDA [118]. For the LiO<sup>t</sup> Bu process, the films with the lowest lithium contents showed the highest ionic conductivity of 6.4 × <sup>10</sup>−<sup>8</sup> S/cm at 30 ◦C [119]. In addition to these reports, LiNbO3 has also been made with the same combination of ALD and solid state reactions as was mentioned for LiTaO3 [104]. In this case also the post-deposition annealing of a bilayer of Li2CO3 and Nb2O5 produced a film with low impurity content. However, compared to the LiTaO3 case, here the films were slightly lithium deficient (Figure 10).

The first truly quaternary lithium material deposited by ALD was lithium lanthanum titanate (LLT), reported by Aaltonen et al. in 2010 [120]. Thin films were deposited by combining binary ALD processes for TiO2, La2O3 and Li2O/LiOH and were amorphous as deposited. In that work TiCl4 was used as the titanium precursor and it was found that applying the Li2O/LiOH subcycle after the TiO2 cycle resulted in rougher and less uniform films than when lithium was pulsed after the La2O3 subcycle. This is a clear indication that the pulsing order can have a large effect on the deposition of quaternary materials. As reactivity problems were also observed in the deposition of lithium titanate using TiCl4 [99], the chloride precursor might be playing a large, thus far unknown role in these processes. These problems could be related to LiCl formation, for example. For the LLT deposition, a pulse sequence where 3 cycles of La2O3 were applied after one TiO2 cycle and the number of lithium subcycles was varied, was used. The content of lithium in the film did not linearly follow the number of Li2O/LiOH subcycles. This could mean that the reactivity of LiOt Bu is lower on a Li2O/LiOH surface compared to its reactivity on a La2O3 surface, a somewhat similar conclusion as was made in the deposition experiments on lithium aluminate [136]. Nevertheless, saturation as a function of the LiO<sup>t</sup> Bu pulse length was observed [120]. The maximum lithium content reached with this pulsing scheme was approximately 20 at.%. The lanthanum content stayed constant in all experiments but the content of titanium decreased as a function of the increased lithium content. Under saturative conditions the film composition, as determined by ToF-ERDA, was Li0.32La0.30TiOz. Interestingly, SIMS (secondary ion mass spectrometry) depth profiling seemed to indicate that lithium was somewhat concentrated onto the film-substrate interface, whereas in many cases lithium has been reported to preferably reside on the outer surface of the film [99,118]. However, this observation could be an artefact caused by sputtering during SIMS. The films could be crystallized by annealing in oxygen. The XRD diffractograms matched well with the reported peak positions of Li0.33La0.557TiO3, however four peaks could not be identified [120].

Li*x*Al*y*Si*z*O, another amorphous solid electrolyte, has been studied by Perng et al. [121]. The material belongs to the lithium aluminosilicate family, which includes materials with high lithium-ion conductivities with various metal ratios [30]. Lithium aluminosilicate was deposited by ALD using a pulsing sequence of Al2O3 from TMA and water, Li2O/LiOH from LiOt Bu and water and SiO2 from TEOS and water [121]. With a pulsing sequence of Al:Li:Si = 10:6:4 the film thickness increased linearly with the number of supercycles. The lithium contents of the films, as determined by synchrotron ultraviolet photoemission spectroscopy (UPS), increased with increasing lithium oxide pulsing ratio but showed quite a lot of scattering (Figure 11). The deposited films were shown to be pinhole free and had ionic conductivities between 10−<sup>9</sup> and 10−<sup>7</sup> S/cm at room temperature, depending on the lithium content. Higher lithium contents led to higher conductivities but also increased the activation energy. It should be noted that the film thicknesses used in these experiments were very small, 10 nm and below. Larger thicknesses led to a lower ionic conductivity.

**Figure 11.** The lithium content in lithium aluminosilicate films as a function of the Li2O cycle fraction *a*/(*a* + *b* + *c*), as determined from syncrotron UPS spectra. Adapted with permission from [121]. Copyright 2014 The Royal Society of Chemistry.

Kazyak et al. have taken on the impressive task of depositing the garnet oxide Li7La3Zr2O12 by ALD [122]. This crystalline material is known to have a lithium-ion conductivity close to 10−<sup>3</sup> S/cm at room temperature [137]. In order to stabilize the desired cubic phase at room temperature, the material was doped with alumina [122]. This resulted in an ALD process combining 8 subcycles of Li2O, 28 subcycles of La2O3, 12 subcycles of ZrO2 and 1 Al2O3 subcycle at 225 ◦C to obtain an amorphous film with metal ratios Li:La:Zr:Al = 52:27:19:2 (ideal composition 54:26:17:2). Interestingly, despite using ozone as the oxygen source for all subcycles, no Li2CO3 or La2(CO3)3 formation was evident

from XPS results. The thickness of the films increased linearly with the number of supercycles and good conformality was also obtained. The ionic conductivity of the as-deposited, amorphous film was 1.2 × <sup>10</sup>−<sup>6</sup> S/cm at 100 ◦C and did not differ between in-plane and through-plane measurements. By extrapolation, the conductivity was 10−<sup>8</sup> S/cm at RT. The films could be crystallized to the cubic Li7La3Zr2O12 phase with annealing at 555 ◦C in inert atmosphere. A lithium-excess in the film and an extra lithium source were needed during the annealing due to lithium loss from the film. The annealed films had an island morphology, which prevented reliable conductivity measurements.

Although the majority of published ternary lithium ALD processes are for oxide materials, sulphides and fluorides have been studied as well [51,123]. Lithium aluminum sulphide Li*x*Al*y*S has been deposited using subcycles of Li2S [91] (LiOt Bu + H2S) and Al2S3 [138] (tris(dimethylamido)aluminum(III) + H2S). Using a 1:1 subcycle ratio resulted in a Li:Al ratio of 2.9:1 in films deposited at 150 ◦C, as determined with ICP-MS (inductively coupled plasma mass spectrometry) [123]. Evaluation of the metal ratio from QCM (quartz crystal microbalance) data, assuming stoichiometric growth, resulted in a metal ratio of 3.5:1 which is reasonably close to the value from ICP-MS. The ternary sulphide growth was linear as a function of cycles, with a reported growth rate of 0.50 Å/cycle. The growth rate during the Li2S subcycle seemed somewhat lower in the ternary process than the one reported for the binary process [91]. This difference was not commented on in ref. [123] but it most likely originates from different starting surfaces. A 50 nm Li*x*Al*y*S film was measured to have a room temperature ionic conductivity of 2.5 × <sup>10</sup>−<sup>7</sup> S/cm, which is among the best quoted conductivities of ALD-made films [108,109,117,121,123]. The sulphide was studied as an artificial SEI-layer on metallic Li and it was found to effectively stabilize the interface between the metal anode and an organic liquid electrolyte [123]. In addition, the coating decreased lithium metal dendrite formation during cycling, which considerably improves the safety of lithium metal anodes.

Multicomponent, lithium-containing fluorides have not been studied extensively by ALD [51,124,139]. Still, Li3AlF6, has been deposited using multiple processes. This material will be discussed in the next section on fluoride deposition using ALD.

#### **5. Atomic Layer Deposition of Metal Fluorides**

Metal fluorides have been of interest to ALD chemists since the beginning of the 1990s. In the very beginning, doping of electroluminescent materials with fluorine was studied [140] and soon after the first report on depositing CaF2, ZnF2 and SrF2 was published [141]. For the first two decades, metal fluorides were studied mainly because of their optical properties, namely low refractive indices and low absorption in the UV range [142]. However, with the rise of lithium-ion battery related ALD research, the potential of ALD metal fluorides in batteries has also been recognized [51,143–145]. Still, very few results on using atomic layer deposited metal fluoride thin films in lithium-ion batteries is available at this time.

Table 7 summarizes all reported ALD fluoride processes. Electrochemical analysis results are reported when available. The processes are divided into sections based on the fluorine precursor used. The materials are listed in the order of main groups followed by transition metals and lanthanides in the order of atomic numbers. For discussion purposes, a somewhat historical approach has been taken in the following subchapters. Besides this review, ALD of metal fluorides has been discussed in the academic dissertations of Pilvi [142], Lee [146] and Mäntymäki [79].

**Table 7.** ALD processes reported for fluoride materials. Abbreviations used: LiHMDS = lithium hexamethyldisilazide, HF-py = mixture of HF and pyridine, EtCp = ethylcyclopentadienyl, thd = 2,2,6,6-tetramethyl-3,5-heptanedionato, TMA = trimethyl aluminum, Ac = acetate, DEZ = diethylzinc, TEMAZ = tetrakis(ethylmethylamido) zirconium, Ot Bu = *tert*-butoxide, TDMAH = tetrakis(dimethylamido) hafnium, hfac = 1,1,1,5,5,5-hexafluoro-2,4-pentanedionato.

