ALD2019 Session ALE1-TuM: ALE: Gas-phase and/or Thermal ALE

Transition metals are known as hard-to-etch materials for reactive ion etching (RIE) processes. Although Ar milling processes are now widely used for such metal etching, physical sputtering processes with high energy ions have problems of low selectivity and poorly controlled etched profiles. Surface damages induced by energetic ion bombardment may also cause degradation of the material surface properties. Therefore alternative transition-metal etching processes with high selectivity and low damage, rather than physical sputtering, have been sought after in the industry. In recent years, atomic layer etching (ALE) by the formations of volatile organic metal complexes has attracted much attention as a means to achieve atomically controlled and low damage etching. ALE reactions may be considered as reverse reactions of atomic layer deposition (ALD). For example, metal beta-ketoenolate complexes are often used as precursors for ALD, so stable adsorption of beta-ketones on a metal surface and the formation of metal beta-ketoenolates thereon are crucial steps for the development of corresponding ALE processes. In this study, we have examined surface reactions of transition metals or their oxides with hexafluoroacetylacetone (hfac) [1] and demonstrated that such reactions can be used to develop ALE processes for some transition metals. Experiments were performed in what we call the “Atomic-Layer-Process (ALP) Surface Analysis System,” which consists of ALP reaction chambers and an in-situ high-resolution X-ray photoelectron spectroscopy (XPS) system. In the reaction chamber, the substrate temperature can be controlled by a ceramic heater, which is installed on the back side of the sample. After exposure to reactive gases, the sample can be transferred from the reaction chamber to the XPS chamber without being exposed to ambient air for in- situ surface chemical analysis. Ion irradiation effects on an hfac adsorbed metal surface may be studied with the use of low energy Ar⁺ ion beam of the XPS system. We used Ni and Co substrates in this study. It has been found that hfac molecules adsorbed on a metal oxide surface are less likely to be decomposed at room temperature than those on a metal surface and, at an elevated temperature, a metal oxide surface is preferentially etched by hfac than a metal surface.

[1] H. L. Nigg and R. I. Masel, J. Vac. Sci. Technol. A, 17, 3477 (1999).

The thermal atomic layer etching (ALE) of silicon nitride was demonstrated using an oxidation and “conversion-etch” mechanism (see Supplemental Figure 1). In this process, the silicon nitride surface was oxidized to a silicon oxide layer using O₂ or ozone. The silicon oxide layer was converted to an Al₂O₃ layer using trimethylaluminum (TMA). The Al₂O₃ layer was fluorinated by HF to an AlF₃ layer prior to the removal of the AlF₃ layer by ligand-exchange using TMA. Silicon nitride ALE was studied using Si₃N₄ films deposited using low pressure chemical vapor deposition (LPCVD). In situ spectroscopic ellipsometry was employed to monitor the thickness of both the Si₃N₄ and the silicon oxide layer during ALE. These studies observed that the silicon nitride film thickness decreased linearly with number of reaction cycles while the silicon oxide thickness remained constant.

Using an O₂-HF-TMA reaction sequence, the Si₃N₄ ALE etch rate was 0.26 Å/cycle at 290°C. This etch rate was obtained using static reactant pressures of 250, 0.65 and 1.0 Torr, and exposure times of 10, 5 and 5 s, for O₂, HF and TMA, respectively. Employing similar dosing conditions, the process using O₃ yielded a higher Si₃N₄ etch rate of 0.47 Å/cycle (see Supplemental Figure 2). The Si₃N₄ etch rates remained the same for O₃ pressures from 30-250 Torr. The order of the reactant sequence affected the Si₃N₄ etch rate. Changing the reactant sequence from O₃-HF-TMA to O₃-TMA-HF reduced the Si₃N₄ etch rate from 0.47 to 0.20 Å/cycle at 290°C. The Si₃N₄ ALE etch rate was also reduced at lower temperatures. Using the O₃-HF-TMA reaction sequence, the Si₃N₄ etch rate was reduced from 0.47 Å/cycle at 290°C to 0.07 Å/cycle at 210°C.

Si₃N₄ ALE also decreased the roughness of the Si₃N₄ surface. The RMS roughness of the initial Si₃N₄ films was 4.7 Å measured using atomic force microscopy (AFM). The RMS roughness decreased to 3.1 Å after 80 ALE cycles. An SiO₂ oxide thickness of ∼10-15 Å remained after Si₃N₄ ALE at 290°C. This oxide could be removed by 15 sequential TMA and HF exposures after the Si₃N₄ ALE. Thermal Si₃N₄ ALE should be useful in advanced semiconductor fabrication. Thermal Si₃N₄ ALE could also find applications in optoelectronics, photonics and MEMS fabrication.

Thermal atomic layer etching is typically based on two sequential surface reactions. The first reaction activates the surface layer and the second reaction leads to material removal by the desorption of volatile etch products. The surface activation can be halogenation, conversion to a different material, or oxidation of the initial material. For example, BCl₃ is able to convert the WO₃ surface to a B₂O₃ surface layer during WO₃ ALE. The B₂O₃ surface layer is then spontaneously removed by etching using HF. In another example, O₃ is able to oxidize the TiN surface to a TiO₂ surface layer during TiN ALE. The TiO₂ surface layer is then spontaneously removed by etching using HF.

This study explored the spontaneous etching of B₂O₃ and TiO₂ with HF using Fourier Transform Infrared (FTIR) spectroscopy and quadrupole mass spectrometry (QMS) analysis. The initial B₂O₃ films were grown using B₂O₃ ALD with BCl₃ and H₂O as the reactants. The initial TiO₂ films were grown using TiO₂ ALD with TiCl₄ and H₂O as the reactants. FTIR measurements observed the growth of the B₂O₃ films and TiO₂ films by monitoring the absorbance of the B-O and Ti-O stretching vibrations, respectively, versus number of ALD cycles. FTIR experiments also observed the spontaneous etching of B₂O₃ and TiO₂ with HF by measuring the loss of the absorbance of the B-O and Ti-O stretching vibrations, respectively (See Supplemental Figure 1).

QMS studies were also able to monitor the volatile etch products during the spontaneous etching of B₂O₃ nanopowder with HF. The expected reaction products are BF₃ and H₂O based on the reaction B₂O₃ + 6HF → 2BF₃ + 3H₂O. In comparison, the QMS detected B(OH)F₂, BF₃ and H₂O as the main etch products (See Supplemental Figure 2). In addition, the QMS also revealed species at higher masses that were consistent with six-member ring species, such as B₃O₃F₃.

The reaction of HF with B₂O₃ and TiO₂ was also examined using a density functional theory (DFT) based computational approach. By comparing the thermodynamic free energy profiles of competing self-limiting surface and bulk reactions, the DFT calculations predicted the spontaneous etching of B₂O₃ by HF above -160°C and of TiO₂ (but not TiN) above 90°C, in agreement with the experimental findings.

Thermal based Atomic Layer Etching (th-ALEt) has opened a new horizon and triggered an increased interest in the Semiconductor Industry for the fabrication of sub-10 nm as well as complex 3D nano-devices. In the th-ALEt technique, a material is chemically etched from thermally activated surface by sequence of one or more reactants each separated by purge steps. Unlike the conventional anisotropic plasma etching, th-ALEt is isotropic, selective and its slow etch rate may possess excellent atomic-scale control. Most of the reported th-ALEt chemistries utilize hydrogen fluoride (HF from HF-pyridine) as one of the reactants. However, due to some safety concerns associated with the use and handling of HF, we have considered other fluorine donating compounds. In this work, we report niobium pentafluoride (NbF₅) as an alternative to HF. Carbon tetrachloride (CCl₄) is used as a co-reactant with NbF₅ to etch aluminum oxide (Al₂O₃) as well as titanium nitride (TiN). The various attributes of the etching process like etch rates, selectivity and post-etch surface roughness were studied. It was found that NbF₅ promotes the fluorination of Al₂O₃ and the fluorinated Al₂O₃ surface can be etched away by a subsequent exposure of CCl₄ gas. TiN can be etched in continuous pulsed mode just by CCl₄, while adding NbF₅ to the process enables etch-rate control. The etch results and proposed reaction pathway for the etching of Al₂O₃ and TiN will be discussed in the presentation.

Thermal atomic layer etching (ALE) can be achieved with sequential surface reactions using the fluorination and ligand-exchange mechanism. For metal oxide ALE, fluorination converts the metal oxide to a metal fluoride. The ligand-exchange reaction then removes the metal fluoride by forming volatile products. Previous studies have demonstrated the thermal ALE of amorphous HfO₂ and ZrO₂ ALD films. No previous investigations have explored the differences between the thermal ALE of amorphous and crystalline films. The thermal ALE of crystalline films is important because amorphous films may not crystallize easily when they are too thin. Consequently, amorphous films may have to be grown thicker, crystallized, and then etched back to obtain the desired ultrathin crystalline film thickness.

This study explored the thermal ALE of amorphous and polycrystalline films of hafnium oxide, zirconium oxide, and hafnium zirconium oxide. HF was used as the fluorination reactant. Dimethylaluminum chloride (DMAC) or titanium tetrachloride was employed as the metal precursor for ligand-exchange. The amorphous films had a much higher etch rate per cycle than the crystalline films. The differences were most pronounced for hafnium oxide. At 250 °C with HF and DMAC as the reactants, the etch rate was 0.03-0.08 Å/cycle for crystalline HfO₂ and 0.68 Å/cycle for amorphous HfO₂ (See Supplemental Figure 1).

Under the same conditions at 250 °C with HF and DMAC as the reactants, the etch rate was 0.60-0.82 Å/cycle for crystalline ZrO₂ and 1.11 Å/cycle for amorphous ZrO₂. In comparison, the etch rate was 0.16-0.26 Å/cycle for crystalline HfZrO₄ and 0.69 Å/cycle for amorphous HfZrO₄. The etch rates for HfZrO₄ were between HfO₂ and ZrO₂ for both the amorphous and crystalline films. When HF and TiCl₄ were used as the reactants at 250 °C, the etch rates were smaller than the etch rates with HF and DMAC as the reactants for every material. The etch rates also increased with temperature for both the amorphous and crystalline films. The differences between amorphous and crystalline HfO₂ are sufficient to obtain selective thermal ALE of amorphous HfO₂.

The experimental apparatus used in this study is 300-mm ALE tool equipped with inductively-coupled plasma source and infrared lamps. The cyclic ALE process is composed of three step repetitions: oxidation of cobalt surface with oxygen plasma, organometallization of the cobalt oxide with a low acidity ligand vapor, and sublimation of the organemetallic cobalt by thermal annealing.

The etching depth of cobalt increased with increasing the number of repetitions of the cycle. For one cycle of etching, it was 1 nm high. The root-mean-square (RMS) roughness of etched cobalt surface was estimated to be 0.8 nm. It was found that formation of homogeneous CoO (II) was important because compound oxide such as Co₃O₄ (II & III) resulted in rough etched surfaces with columnar morphology. Furthermore, controllability of etching amount was substantially improved by using low reactivity ligands compared to high-reactivity ligands. These results implied that the combination of homogenious CoO formation, low-reactivity ligands, and sublimation was essential for achieving smooth etched surfaces and excellent controllability of etching amount.

In conclusion, we have obtained well-controlled etch front roughness and etching depth of cobalt, which can be applied to semiconductor process, by controlling reactions of both oxidation and metal complex formation using the 300-mm ALE apparatus.

[1] K. Shinoda et al., J. Phys. D: Appl. Phys. 50, 194001 (2017).

[2] C. Winter, AVS 65th, PS+EM+TF-ThM5 (2018).