ALD2019 Session AA2-WeM: ALD for ULSI Applications I

In the early days of the search to find a replacement for SiO₂-based gate oxides the goal was to find a material with a very high k value which could be incorporated into CMOS production for multiple technology nodes. A historical overview of the many promising high k materials considered for SiO₂ replacement leading to the selection of ALD HfO₂ as “the winner” will be presented. ALD HfO₂ has successfully been implemented in CMOS production for over a decade, starting at the 45nm node. There are two general integration approaches for implementing ALD High-k/Metal Gate stacks (HKMG) in production: gate first and gate last. Challenges with each integration approach, leading to the wider adoption of gate last will be discussed. Furthermore, as the dielectric constant of HfO₂ is only ~20 and a thin SiO₂-base interface was still required to maintain mobility and reliability, HfO₂ provided essentially a one-time scaling benefit. Further thinning of HfO₂ resulted in unacceptable leakage and thus to continue transistor scaling fully depleted devices such as FINFET and Ultra Thin Planar SOI (FDSOI) were pursued. High volume manufacturing flows for FINFET (with gate last integration) and FDSOI (with gate first integration) come with their own unique challenges. For example, with FINFET maintaining gate height uniformity is crucial for Vt targeting and control. With FDSOI, maintaining gatestack stability at high temperature is key. To continue future scaling, new device architectures (e.g. GAA, Vertical FETs, etc.) will pose further challenges for gate stack integration. Recent and historical progress in HfO₂ growth, interface control, selective deposition, morphology and etching will be discussed with respect to the possibility for future gate stack engineering.

References:

R.D. Clark, Materials, 7(4), 2913-2944, https://doi:10.3390/ma7042913 (2014).

D.H. Triyoso et al., ECS Transactions, 69 (5) 103-110 (2015).

R.J. Carter et al., ECS Transactions, 85(6) 3-10 (2018).

R.D. Clark et al., ECSarXiv. October 3. https://doi:10.1149/osf.io/qtxnd (2018).

R.D. Clark et al., APL Materials 6, 058203; https://doi.org/10.1063/1.5026805 (2018).

Gate dielectric materials with high-k are required for further scaling down in future years. As an alternative of conventional high-k materials such as HfO₂, the addition of elements to host high-k materials has attracted attention. Among various elements, rare-earth elements, such as Y, La, Dy, or Er has been known to transform the crystal structure of HfO₂ from the first-principles study. The theoretical study showed that the doping into HfO₂ can energetically stabilize the cubic or tetragonal phase at lower temperature than thermodynamic conditions of pure HfO₂. Since cubic (k~29) or tetragonal (k~70) HfO₂ has much higher dielectric constant than that of amorphous (k~16-19) and monoclinic (k~20-25) phases, it is noteworthy that the structural modulation by doping of rare-earth elements can enhance the electrical properties of HfO₂.

With accelerated scaling down and three-dimensional structuring of integrated circuits, it becomes very challenging to fabricate metal-insulator-metal (MIM) capacitors with low leakage current and high capacitance density. Specifically, the introduction of high-k dielectrics in conjunction with TiN electrodes has improved electrical properties in sub-100 nm processes. Various high-k dielectrics layers combined with TiN electrodes in MIM capacitors were studied for further improvement of MIM capacitors. Controlling an interfacial layer formation between dielectric layer and metal electrode is essential for depositing high-k dielectric thin film on a TiN electrode. When high-k dielectric films were placed on the TiN, interfacial layer was formed due to high reactivity of TiN. The interfacial layer acts as charge traps causing degradation of electrical properties. Surface treatment like plasma treatment on the TiN has been known to help suppress formation of an interfacial layer, but it would be hard to apply for mass-production of DRAM process due to difficulty of uniform treatment without damage caused by energetic species such as ions and radicals on the devices formed inside deep trenches with high aspect ratio.

Alternatively, selection of stable metal electrodes with high work function is required to improve electrical properties. Among several metals, Ru electrode can be appropriate option due to its good thermal and chemical stability, low resistivity, high work function. In this paper, we investigated effects of bottom electrodes on the thin film properties of atomic layer deposited (ALD) ZrO₂, concentrating on correlation between interfacial layer formation and electrical properties. Transmission electron microscopy (TEM) showed thinner thickness of the interfacial layer on the Ru electrode than TiN electrode. Chemical composition of the interfacial layer was analyzed by X-ray photoelectron spectroscopy (XPS) analysis, and ZrO₂ on Ru was less intermixed with bottom electrode due to good thermal and chemical stability of Ru electrode. Introducing Ru electrode improved symmetry of the normalized C-V characteristics. Simultaneously, the introduction of Ru electrode affects decrease of leakage current density from ~10^-5 A/cm² to ~10^-7 A/cm² in I-V characteristics. These results are very meaningful capacitor with Ru electrode can be a very promising device for MIM capacitor in DRAM production.

High-performance metal-oxide-semiconductor field-effect transistors (MOSFETs) require the semiconductor/high-κ interface with high-temperature thermal stability and a low interfacial trap density (D_it). Previously, in-situ atomic layer deposition (ALD) or molecular beam epitaxy (MBE) Y₂O₃ has effectively passivated GaAs(001) surface.^1,2 The growth was achieved in an integrated ALD/MBE ultra-high vacuum (UHV) system. Despite the difference in deposition, both Y₂O₃/GaAs interfaces withstand 900 ^oC annealing, and the D_it’s lie below 5×10¹¹ eV^-1cm^-2. MOS capacitors (MOSCAPs) with such interface outperform those with ex-situ deposited Al₂O₃.³ By in-situ synchrotron radiation photoemission study on ALD-Y₂O₃/GaAs(001)-4×6, we found that the faulted surface As atoms were removed and lines of Ga-O-Y bonds stabilized the interface.⁴ The interfacial Ga₂O (Ga⁺)-like state explains the low D_it.

In this work, we have improved the electrical characteristics in ALD-Y₂O₃/GaAs by in-situ UHV annealing the initial 1-nm Y₂O₃. The idea is motivated by removing the freed As atoms and hydrocarbons remained in the ALD layer. Note that, an amount of hydrocarbons at such critical interface can degrade the device performances. ALD-Y₂O₃ was grown by thermal ALD with sequential Y(EtCp)₃ and H₂O pulses, and in-situ UHV annealing up to 600 ^oC was conducted in another chamber in our system right after the ALD growth. MBE-Y₂O₃ is relatively pure and employed as a reference.

Fig. 1 shows the capacitance-voltage (CV) and quasi-static CV (QSCV) curves for MOSCAPs. The UHV-annealed ALD-Y₂O₃/GaAs was improved with a reduced frequency dispersion (F.D.) in the accumulation/depletion region, and a lowered trap-induced hump in the inversion region. Fig. 2 presents the D_it spectra extracted from QSCVs. The UHV-annealed ALD-Y₂O₃/GaAs shows a reduced D_it, also hinted by the sharp transition of QSCVs and the narrow gap between QSCVs and CVs. Fig. 3 shows the O 1s core-level spectra, where O-Y is from the stoichiometric Y₂O₃ and O* is from the interfacial Ga-O-Y and the surface Y-O-H.⁴ Note that the ratios of O-Y of our 1-nm Y₂O₃ films are significantly higher than the one reported.⁵ Upon UHV annealing, the residue O* may be attributed to the interfacial Ga-O-Y with the surface Y-O-H mostly removed. This UHV annealing approach is significant in perfecting ALD-Y₂O₃/GaAs and is applicable to many other material systems.

References:

¹ Y. H. Lin et al., Appl. Phys. Express 9, 081501 (2016).

² H. W. Wan et al., J. Cryst. Growth 447, 179 (2017).

³ T. Yang et al., Appl. Phys. Lett. 91, 142122 (2007).

⁴ C. P. Cheng et al., ACS Omega 3, 2111 (2018).

⁵ P. de Rouffignac et al., Chem. Mater. 17, 4808 (2005).