ALD2023 Session AF1-TuM: Precursors and Processes I

Although AgCl, AgBr and AgI have unique attractive properties, our primary motivation for developing ALD processes for these materials is the deposition of silver halide perovskites. Silver halides are IR-transparent and antiseptic, which has enabled their use in niche optical and medical applications. Silver halides are also light-sensitive, which we believe can be exploited for patterning applications. Most importantly, silver halides are components of double perovskites, such as Cs₂AgBiBr₆.

Halide perovskites are a major topic in materials science and are associated with Pb compounds, photovoltaics, and challenges with scalability, stability, and toxicity. However, one-third of the publications on perovskites are unrelated to photovoltaics. The largest and fastest growing non-photovoltaic applications of perovskites are light-emitting diodes, sensors, and microelectronic components. In these applications, it is possible to use Pb-free alternatives like double perovskites. Silver and bismuth double perovskites, like Cs₂AgBiBr₆, are stable and nontoxic, eliminating two of the three challenges.¹ Depositing double perovskites with ALD could address the remaining scalability challenge.

The development of ALD processes for ternary and quaternary compounds begins with the processes for the corresponding binary compounds. From our previous work² we know how to deposit cesium halides, but no processes for silver and bismuth halides are known. This work focuses on silver halide processes using Ag(fod)(PEt₃), a silver precursor well established in the ALD and CVD of metallic silver.

Our metal-halide ALD processes employ volatile metal halides, like SnI₄, as halide precursors. In our previous work, we discovered that the choice of the volatile metal halide makes or breaks the process.³ The byproducts generated by the volatile metal halide can be benign or detrimental by being able to etch the film material or by being nonvolatile, which results in the incorporation of impurities and poor crystallinity. Therefore, we screened the candidates to identify suitable pairs. The candidates are the corresponding halides of titanium, gallium and tin.

For example, Ag(fod)(PEt₃) and SnI₄ are one such pair. These precursors produce crystalline β-AgI films in the 100 – 200 °C temperature range with the largest GPC of 0.9 Å at 140 °C. The films were smooth, uniform, and contained a negligible amount of impurities. We continue our process studies on suitable precursor pairs for the other silver halides.

[1] Lei et al., Adv. Funct. Mater. 2021, 31, 2105898.

[2] Weiß et al., Chem. Mater. 2022, 34, 6087.

[3] Popov et al., Dalt. Trans. 2022, 51, 15142.

The number of ALD processes for metal fluorides has been limited, especially when compared to ALD processes for metal oxides, nitrides, and sulfides. Recently, however, interest towards ALD of metal fluorides has increased. The applications for metal fluoride films range from optical coatings to lithium-ion batteries (LIB) and luminescence devices. This work summarizes recent studies on ALD of metal fluorides at University of Helsinki.

New ALD processes for rare earth and transition metal fluorides are presented. Of the rare earth metal fluorides, we have included an ALD process for ScF₃, which is a negative thermal expansion (NTE) material. To our knowledge, this is the first wide-temperature range NTE material deposited by ALD. The films are close to the stoichiometric, and, e.g., in films deposited at 300 °C the total impurity content (O, C, and H) is only ~2.6 at-% as measured by ToF-ERDA. In addition, an ALD process for GdF₃ and its in-situ conversion to NaGdF₄ by Nathd (thd=2,2,6,6-tetramethyl-3,5-heptanedione) are presented. GdF₃ is an important material for antireflection coatings, whereas NaGdF₄ is a potential host material for luminescence centers, especially for medical applications. Of the transition metal fluorides, an ALD process is presented for CoF₂ that is a potential LIB cathode material.

In addition to the new metal fluoride ALD processes, we introduce a new fluoride source, NbF₅, the use of which has been inspired by the successful use of TiF₄ and TaF₅ as fluoride sources in ALD. In this work NbF₅ was combined with Ho(thd)₃ to deposit HoF₃. In HoF₃ films, Nb impurity content as low as 0.2 at-% was obtained.

We aim to give a comprehensive overview of ALD of metal fluorides. Therefore, also future research directions will be discussed.

State-of-the-art devices require surrounding Cu interconnects with a layer of tantalum (Ta liner) and a layer of tantalum-nitride (TaN barrier) to prevent diffusion of Cu atoms into the surrounding dielectric, which is detrimental to the lifetime of the IC. Downscaling of integrated circuits (ICs) faces significant challenges because the resistivity of Cu features increases at smaller dimensions according to the product λ×ρ₀, and further thinning of the Ta/TaN layers would result in increased resistivity and poorer performance. Among metals with lower λ×ρ₀ values, cobalt (Co) requires a barrier film and ruthenium (Ru) is difficult to process during CMP. Molybdenum (Mo), which does not require a barrier and is CMP processable, is a suitable alternative for Co and Ru and the entire Cu/Ta/TaN interconnect. Current commercial processes use solid halide-based precursors (MoCl₅ and MoO₂Cl₂). As device integration becomes more complex with each node generation, halide-free deposition processes are necessary. We designed and successfully synthesized a halide free Mo precursor with small ligands, which is low melting and has high volatility. We studied its thermal stability, volatility and chemical properties. Our precursor consists of a) Mo at zero oxidation state, b) neutral halide-free ligands – easier to detach from metal thermally during deposition.

Transition metal carbides (TMCs) are widely used in catalytic and wear resistance applications. They exhibit excellent chemical and thermal stabilities, exceptional hardnesses, and low resistivities. Additionally, TMCs typically have good electromigration resistances. These properties make them relatively good conductors for metal wires when the dimensions shrink to the sub-10 nm range. Development of TMC ALD processes opens the possibility to use carbides in semiconductor applications. Molybdenum carbides (MoC_x) have the potential to improve the performance, efficiency, and reliability of semiconductor devices. Recently, they have emerged as potential candidates for diffusion barriers, interconnects, and gate electrodes.^1–3 The ALD of metal carbides is, however, still in its infancy, and current challenges include a lack of thermal ALD processes, high process temperatures, and low growth rates.

In this work, we report a novel thermal ALD process for MoC with MoCl₅ and bis(trimethylgermyl)-1‚4-dihydropyrazine ((Me₃Ge)₂DHP) as precursors. (Me₃Ge)₂DHP has previously been used as a reducing agent in ALD of nickel and gold.^4,5 In the current process, (Me₃Ge)₂DHP acts as both the reducing agent and carbon source for the first time. The process was investigated at temperatures between 200 and 300 °C. At 275 °C, high growth rates of 1 Å/cycle were observed. The films are very smooth with XRR roughnesses of approximately 0.25 nm. The growth rate is strongly affected by the MoCl₅ pulse length, and we observed a small etching component by MoCl₅ in all depositions. Remarkably, according to XPS the MoC films do not contain any metallic Mo. The resistivities of the moderately crystalline films are ~200 μΩcm at a film thickness of ~100 nm, which is slightly higher than those for bulk MoC_x. No change in the crystallinity was observed after annealing up to 1000 °C under N₂ atmosphere. The influence of process parameters on the MoC film properties as well as the mechanism of the process is discussed in detail.

[1] Tripathi, C. C., et al. Appl. Surf. Sci. 255, 3518–3522 (2009). https://doi.org/10.1016/j.apsusc.2008.09.076.

[2] Leroy, W. P., et al. J. Appl. Phys. 99, 063704 (2006). https://doi.org/10.1063/1.2180436.

[3] Ha, M.-J. et al. Chem. Mater. 34, 2576–2584 (2022). https://doi.org/10.1021/acs.chemmater.1c03607.

[4] Vihervaara, A. et al. Dalt. Trans. 51, 10898–10908 (2022). https://doi.org/10.1039/D2DT01347A.

[5] Vihervaara, A. et al. ACS Mater. Au (2023). https://doi.org/10.1021/acsmaterialsau.2c00075.

The element bismuth (Bi) is an important component of materials ranging from insulators to superconductors. For many applications, thin films containing Bi need to be grown in high aspect ratio features with perfect conformality and Angstrom-level thickness control. Atomic layer deposition (ALD) is a film growth method that can afford uniform thickness films, even in narrow and deep nanoscale features. While Bi ALD precursors have been reported for materials such as oxides,¹ there have been no reports to date of the ALD growth of Bi metal films. Herein, we will describe a family of thermal ALD processes for Bi metal thin films. Bi precursors used in this work include BiCl₃, BiPh₃, and Bi(NMe₂)₃. Reducing co-reactants fall into two general classes. Processes with BiCl₃ and 2-methyl-1,4-bis(trimethylsilyl)-2,5-cyclohexadiene (1) or 1,4-bis(trimethylsily)-1,4-dihydropyrazine (2)² afforded Bi metal films at substrate temperatures ranging from 50 to 175 ºC. Here, 1 and 2 serve as the reducing agents, with elimination of Me₃SiCl and toluene (1) or pyrazine (2). ALD processes were also developed using BiCl₃, BiPh₃, or Bi(NMe₂)₃ in combination with nitrogen sources such as ammonia, hydrazine, alkyl hydrazines, or alkyl amines. The processes with nitrogen-based co-reactants are proposed to proceed via the formation of unstable “BiN”, which decomposes to afford Bi metal films. All of these processes afforded crystalline Bi metal films, as determined by X-ray diffraction. X-ray photoelectron spectroscopy demonstrated that the films were >94% pure Bi metal after argon ion sputtering to remove adventitious surface impurities.

Transparent conductive oxide thin films, including SnO, SnO₂, In-Sn-O (ITO), Zn-Sn-O (ZTO), and In-Zn-Sn-O (IZTO) films, have recently attracted much attention for various applications such as flat-panel displays, gas sensors, and solar cells. We have reported the atomic layer deposition (ALD) of In₂O₃ and ZnO thin films using cyclopentadienyl-based precursors [1, 2]. In order to deposit such transparent conductive oxide thin films by using ALD, ALD-Sn precursor is essential. This time, we report ALD of tin oxide (SnO_x) thin films using a new liquid cyclopentadienyl-based precursor.

As a new ALD-Sn precursor, bis(ethylcyclopentadienyl) tin, Sn(EtCp)₂, was synthesized. Sn(EtCp)₂ is a liquid precursor at room temperature. Differential scanning calorimetry (DSC) was conducted to measure its thermal decomposition temperature. The decomposition temperature was estimated approximately 230 °C, so the deposition temperature was set to 200 °C, which is the same temperature in the case of ALD of In₂O₃ and ZnO thin films [1, 2]. The vapor pressure of Sn(EtCp)₂ was determined by directly measuring the equilibrium vapor pressures at several points. From the Clausius-Clapeyron plot for Sn(EtCp)₂, the precursor temperature was set to 70 °C, which corresponds to the vapor pressure of approximately 0.8 Torr.

SnO_x thin films were deposited on 150 mm Si wafers with native oxide films. ALD process was conducted by using Sn(EtCp)₂ as a precursor and O₂ plasma as an oxidant. Saturation of reaction was confirmed when 14 s of Sn(EtCp)₂ and 45 s of O₂ plasma pulse times were applied. At this condition, linear growth of SnO_x thin film was observed. The growth per cycle (GPC) was approximately 0.18 nm/cycle with this ALD condition. This growth rate was relatively fast compared with a previous experiment using tetrakis(dimethylamino)tin (TDMASn) as a precursor and O₂ plasma as an oxidant (~0.13 nm/cycle) [3]. On the contrary, by applying H₂O for 30 s in the place of O₂ plasma, the thickness of SnO_x films scarcely increased by increasing the number of ALD cycles.

SnO_x thin films were deposited by ALD using a new cyclopentadienyl-based precursor Sn(EtCp)₂, and linear growth of SnO_x thin film was confirmed.

References

[1] F. Mizutani, S. Higashi, M. Inoue, and T. Nabatame, AIP Advances 9, 045019 (2019).

[2] F. Mizutani, M. Mizui, N. Takahashi, M. Inoue, and T. Nabatame, ALD2021, AF1-10 (2021).

[3] M. A. Martínez-Puente, J. Tirado, F. Jaramillo, R. Garza-Hernández, P. Horley, L. G. S. Vidaurri, F. S. Aguirre-Tostado, and E. Martínez-Guerra, ACS Appl. Energy Mater. 4, 10896 (2021).