AVS 71 Session AP+EM+PS+TF-FrM: Area Selective Processing and Patterning

Self-assembled monolayer (SAMs) and small molecule inhibitors (SMIs) are two types of inhibitors for area-selective atomic layer deposition (AS-ALD). One critical requirement for inhibitors is to form densely packed adsorbate layers on the non-growth area. SMIs can be delivered in the vapor phase, which makes them compatible with existing industrial processes. However, the random sequential adsorption of SMIs tend to leave gaps in between the adsorbates (1). Intermolecular interactions could improve SMI packing and coverage, potentially leading to a higher selectivity.

π-π interaction is an important class of intermolecular interactions that has been employed in different fields (2). To exploit π–π interactions in SMIs, one can introduce aromatic backbones to the molecule. In this work, 3-phenyl-2,4-pentanedione (Hppa (3)) is used as the structural variant of acetylacetone (Hacac) for inhibiting ALD on Al₂O₃ as the non-growth area over the growth area of SiO₂.

Using in–situ reflection–absorption infrared spectroscopy, we examined saturation behaviour, precursor blocking, and thermal desorption of Hppa. The results indicate that Hppa has similar adsorption behaviour as compared to Hacac, in terms of saturation dosages and bonding configurations. Still, various important differences are found: Firstly, the Hppa adsorbates exhibit a transition from a flat-lying configuration at low coverage, to a standing-up configuration at high coverage, suggested by the emergence of the sp² v(CH) peak. This indicates that the adsorbates are in orientations that could contribute to π-π interaction. Secondly, the results of blocking tests suggested that Hppa adsorbates are not displaced from the surface by bis(diethylamino)silane (BDEAS) precursor molecules, in contrast to the Hacac case in which ca. 5% of acac adsorbates are removed from the surface by BDEAS. Thirdly, Hppa adsorbates does not show thermal desorption at 150 °C for 10 hours. Instead, the increase in peak area suggest that the adsorbates might undergo rearrangement to other orientations. In summary, our results suggest that the Hppa can be an effective SMI for AS-ALD.

Reference

1. J. Li et al., J. Vac. Sci. Technol. A. 40, 062409 (2022).
2. R. Thakuria, N. K. Nath, B. K. Saha, Crystal Growth & Design. 19, 523–528 (2019).
3. O. A. Serra, E. J. Nassar, P. S. Calefi, I. L. V. Rosa, Journal of Alloys and Compounds. 275–277, 838–840 (1998).

In recent years, using small molecule inhibitors (SMIs) has become one of the main approaches to achieve area-selective atomic layer deposition (AS-ALD). The main challenge for area-selective deposition (ASD) using SMIs is to obtain a high packing of inhibitor molecules on the surface. Due to the vapor phase dosing, molecules arrive one-by-one at random locations, which leaves gaps in between the inhibitors where potentially a precursor can adsorb, causing loss of selectivity.[1]

In this work we explore the use of attractive intermolecular interactions for achieving higher packing of SMIs via simulations. When employing small molecules, it is expected that van der Waals interactions are too weak to contribute to packing. Consequently, one of the main questions of this work is how strong the interaction has to be to facilitate ordering of molecules on the surface. In addition, the synergy of intermolecular interaction and diffusion of SMIs over the surface is investigated.

The simulation method used for studying the packing of molecules with intermolecular interaction and diffusion is an augmented random sequential adsorption (RSA) model.[1] During every loop of the RSA model, either adsorption or diffusion takes place, based on the adsorption and diffusion rates. The intermolecular interaction energy is implemented in the form of a Metropolis-Hastings algorithm.[2] In the initial work, the molecule is a simple disk on a hexagonal grid with periodic boundary conditions, considering nearest neighboring blocking.

It was found that the coverage of the SMI improves significantly for low diffusion barriers and attractive intermolecular interaction. For interactions energies stronger than approximately -0.15 eV, the formation of domains of higher density on the surface is observed. The results suggest that coverage can be improved significantly by selecting SMIs with sufficient interaction energy and a low diffusion barrier.

1. J. Li; et. al., JVST A 2022, 40 (6), 062409
2. M. Kalos; et. al., The Annals of Statistics 1986, 22 (4), 1701-1762

In area-selective deposition, selectivity of surfaces could be manipulated to either suppress or promote surface reactivity with respect to the target reactants. Using model ALD processes with TiO₂ (TDMAT/water) or Al₂O₃ (TMA/water), the deposition onto semiconductor surfaces modified with small fluorine-containing molecules is analyzed by spectroscopic and microscopic techniques, including depth profiling with ToF-SIMS, supplemented by computational DFT modeling. The fluorinated functional groups are designed for easy spectroscopic characterization to analyze the potential AS-ALD schemes on silicon, as well as on oxide materials, including TiO₂, MgO, and Al₂O₃. The initial deposition steps are analyzed by comparing the behavior of modified surfaces with that of pristine substrates, and the distribution of the fluorine and fluorine-containing fragments within the ALD-deposited layers is followed by ToF-SIMS depth profiling once these F-containing functionalities are buried under the overgrown layers. This approach allows for identification of the deposition processes for both small molecule inhibitors (SMIs) and small molecule promoters (SMPs).

Complex 3D device architectures are proposed as the solution to make devices more energy efficient.¹ These architectures require many lithographic steps where the high costs of EUV lithography limits device throughput. As such, there is a need for process augmentation to reduce the EUV burdening. A potential solution is area-selective deposition(ASD), where film deposition occurs on a “growth” surface while it is inhibited on an adjacent “non-growth” surface allowing for bottom-up processing.

While ASD conveys the notion of selective deposition between two surfaces, however, in fabrication more might be exposed. Therefore, it is crucial to look at multiple surfaces, a “multi-color system.” The goal being to have the versatility of depositing the desired material only on the desired location(s). It is then important to know processes that activate or deactivate specific surfaces in a multi-color system.

Poly(3,4-ethylenedioxythiophene)(PEDOT) was deposited by oxidative chemical vapor deposition(oCVD) using 3,4-ethylenedioxythiophene(EDOT) monomer and antimony pentachloride(SbCl₅) as reactants to analyze how various pre-treatment strategies can tune the ASD between Si-H, SiN, and SiO₂ surfaces. As a control, single-material coupons were treated with a diluted hydrofluoric acid(DHF) wet etch prior to PEDOT oCVD. Results showed ~30nm of ASD on SiN and SiO₂ vs. Si-H. Three different pre-treatment strategies were evaluated: 1 cycle of molybdenum hexafluoride(MoF₆)/N_2, 7 cycles of N,Ndimethylamino-trimethylsilane(DMATMS)/N₂, and 7 cycles of DMATMS/N₂ followed by a water soak. Ellipsometer, water contact angle, XPS, and SEM results show that MoF₆ served to simultaneously activate and deactivate the Si-H and SiO₂, respectively, allowing for PEDOT ASD on SiN and Si-H vs. SiO₂. DMATMS exposure deactivated only the SiO₂ showing PEDOT ASD on SiN vs. SiO₂ and Si-H. Including a water soak after the DMATMS activated the Si-H to PEDOT deposition resulting in a deposition configuration of Si-H and SiN vs. SiO₂.

Overall, the concept of tunable selectivity for a three-color system is demonstrated by these results. These pre-treatment strategies providing a better understanding into controlling selectivity.

1.Datta, S.; Chakraborty, W.; Radosavljevic, M. Toward. Science 2022, 378 (6621), 733–740.

The selective etching of Si_1-xGe_x over Si enables the fabrication of the gate-all-around field-effect transistors. Thermal etching of Si/Si_1-xGe_x at near room temperature features high selectivity, exhibiting a non-linear relationship between etch rate and Ge% (Fig. 1(a)). There are no reported reaction mechanisms explaining this unique Ge%-dependent phenomenon.

In this work, thin films of Si_1-xGe_x of varying Ge content (Ge% = 0 to 1) were etched thermally by molecular F₂ gas at near room temperature under different F₂ partial pressures (0.5 to 10 mTorr) in Ar. The etch rates were quantified by ellipsometry measurement and the relationship between etch rate and Ge% resembled those shown in Fig. 1(a). Reported Si_1-xGe_x etch selectivity ranges from 100 to 1000 and the unpublished maximum etch selectivity is from 200 to 250. The unpublished experimental data is being reviewed for public release and will be presented at the conference. A kinetics model was established in this work to elucidate the reaction pathways in thermal etching of Si_1-xGe_x by F₂, considering reactions between atomic fluorine and various surface species and the interplay between reaction products involving Si and Ge. The model result (Fig. 1(b)) yielded the unique volcano-shaped relationship between etch rate and Ge%, validating the reactions considered in the model captured the main kinetics during F₂ etching of Si_1-xGe_x.

At advanced nodes targeting 10 nm feature size and below, lithography starts to dominate costs (EUV, multiple mask passes per layer, pattern placement error,...). Complementary techniques and materials are needed to continue 2D scaling and extend the Moore’s law. On the other hand, 2D scaling is reaching its limitations driving the transition to 3D and vertical integration schemes (such as 3DNAND, 3DDRAM, CFET…), which result in higher devices density per unit area and lower production cost. Area-selective atomic layer deposition (AS-ALD) is rapidly gaining interest because of its potential application in self-aligned fabrication schemes for next-generation nanoelectronics. In addition, ASD allows coping with high aspect ratio and complex 3D architectures. The strong sensitivity of ALD to surface chemistry and its self-limiting nature are particularly appealing for ASD.

In this talk I will illustrate a variety of ASD processes and applications spanning from nano-interconnects, logic and memories to patterning.

Achieving area selective deposition requires preventing growth on the non-growth surface (NGS), which often involves the use of molecules to block growth on those surfaces. Careful choice of the ALD process chemistry, thin film precursor and co-reactant, as well as the blocking molecule and how it is administered, is important. We report here a systematic examination of the effects of the precursor, co-reactant and co-adsorbate/blocking molecule on preventing growth of Al₂O₃ on SiO₂. We also consider the effects of temperature, and the dosing sequence employed for the blocking species. Concerning the precursor we compare trimethylaluminum (TMA) to a non-pyrophoric precursor containing only Al-N bonds and no Al-C bonds, i.e., BDMADA-Al [1]. For co-reactants we compare H₂O to t-BuOH. Finally, we consider two blocking species: octadecyl trichlorosilane (ODTS), and dimethylamine trimethylsilane (DMATMS). In this study we employ a quartz-crystal microbalance to monitor ALD in situ and in real-time, and the deposited thin films have been characterized ex situ using X-ray photoelectron spectroscopy, and a variety of techniques. Concerning the “pristine” processes, i.e., ALD in the absence of a blocking molecule, the properties of the films (density, C incorporation, stoichiometry, growth rates) are comparable using either BDMADA-Al or TMA as the precursor under similar reaction conditions. These species also react similarly with H₂O and t-BuOH as the co-reactant, where steady growth with the latter is only observed at sufficiently high temperatures. Concerning blocking growth, we have observed a number of identifiable trends. First, employing the same ALD process chemistry, ODTS produces better blocking in comparison to DMATMS in cases involving a single dose of the blocking molecule. When comparing TMA and BDMADA-Al, we observe that for both blocking molecules that the latter is more efficiently blocked. These two observations demonstrate the importance of molecular size as the larger BDMADA-Al is more efficiently blocked, and the larger ODTS is better for preventing growth. Temperature has a definitive effect on the efficiency of preventing growth where we find that higher temperatures lead to more effective blocking of growth. The dosing sequence employed for the blocking species also plays an important role. Repetitive dosing of DMATMS in an “ABC” process provides superior blocking with respect to a single pre-exposure, and these results exceed those produced by ODTS.

[1]J. V. Swarup, H.-R. Chuang, J. T. Jensen, J. Gao, A. L. You and J. R. Engstrom, J. Vac. Sci. Technol. A 43, 022404 (2025).

Molybdenum (Mo) halides and oxyhalides comprise the sole class of precursors that can deposit Mo metal films by ALD/CVD with sufficient purity for applications in advanced microelectronic devices. Unfortunately, solid, low vapor pressure Mo chloride precursors have challenges in flux stability and low vapor pressure. Metalorganic (MO) precursors are commonly used to address some of these challenges and are useful alternatives to halides for many non-metal films such as SiO₂, SiN, TiN, Al₂O₃, etc. Despite much effort, MO-precursors have not met the performance of chloride precursors for depositing pure Mo.

This presentation will summarize key findings of MO-Mo process development by the Lam ALD/CVD Metals concept and feasibility (C&F) group. Included will be typical precursor tests using coupon process modules, 300mm C&F chambers, fundamental mechanistic investigations of Mo surface reactivity, and novel deposition pathways such as conversion-reduction (Figure 1) and alloy formation (Figure 2).

As DRAM technology continues to scale to meet the increasing demands of high-performance computing, artificial intelligence, and data-intensive applications, maintaining capacitance while reducing cell dimensions presents a significant challenge. The aggressive scaling of DRAM unit cells necessitates innovations in high-k dielectric materials and electrode films to ensure high charge storage capability, low leakage currents, and excellent step coverage in high-aspect-ratio structures. Atomic Layer Deposition (ALD) has emerged as the key enabler for advanced DRAM capacitor and electrode fabrication, offering precise thickness control, excellent conformality, and superior material quality.

Eugenus, a leader in ALD technology, has developed next-generation solutions for DRAM capacitor and electrode deposition, leveraging its Sierra, Lassen, and Whitney ALD platforms. These systems enable the deposition of advanced materials such as ZrO₂, HfO₂, Al₂O₃, and ferroelectric HfZrOₓ for high-k capacitor stacks, as well as TiN, VN, and TSN for bottom and top electrodes. The Sierra ALD system, optimized for high-step coverage metal deposition, provides excellent process uniformity and throughput for TiN/VN electrodes, ensuring low resistance and high reliability in ultra-scaled DRAM architectures. The Lassen ALD system delivers high-quality dielectric films with superior conformality and electrical performance, enabling next-generation capacitor structures with minimal leakage and high breakdown strength. Additionally, the Whitney ALD system supports MoN electrode formation and gapfill applications, further enhancing DRAM performance by reducing resistance and improving integration flexibility.

This technical presentation will be focused on film properties, characterization, and device results. Key advancements in these ALD platforms include optimized precursor delivery, multi-station process modules, and improved reactor designs for high-aspect-ratio structures exceeding 50:1. These innovations enable step coverage exceeding 95% while maintaining excellent film quality, meeting the stringent demands of advanced DRAM fabrication. This presentation will discuss the technical advancements of Eugenus ALD solutions and their impact on future DRAM scaling, providing insights into high-volume manufacturing strategies for next-generation memory devices.