The push of leading-edge manufacturing technologies toward sub-0.1μm feature sizes places extreme performance demands on manufacturing processes and equipment. Improved device performance requirements have led to the examination of various new electronic materials (such as high-k dielectrics and advanced gate materials) [1, 2]. The stringent requirements on the control of film properties drive the development of advanced chemical delivery techniques, control hardware, and choice of materials [1, 2].
The implementation of Al2O3, HfO2, or HfSiOx in advanced device generations drives development and characterization in deposition methods, precursor design, process development and integration, and improved mechanical and electrical properties of materials. Atomic-layer deposition (ALD) and metal-organic chemical-vapor deposition (MOCVD), using novel metal-organic precursors, are frequently used to deposit these new materials.
The implementation of exotic new precursors is complicated by their low vapor pressure and requires chemical delivery systems that accurately and reproducibly deliver the requested quantities. Furthermore, the non-steady-state nature of the ALD process also necessitates pulsing the chemical flow. Innovation in the vapor delivery system, including mass flow controllers, valves, sensors, and other hardware, must meet these challenges (e.g., reduce unnecessary waste of chemicals).
MOCVD vs. ALD
ALD is a surface-reaction controlled process that works by subsequent exposure of the substrate surface to the various precursor gases. MOCVD proceeds by thermally initiated reactions of one or more simultaneously injected precursors. MOCVD allows high throughput of uniform films. However, the development of MOCVD processes for advanced materials has been restricted by a lack of suitable precursors. Binary and tertiary films deposited from mixtures of conventionally available precursors can have poor film uniformity, composition control problems, and reduced deposition efficiency resulting from differences in precursor physical properties and decomposition characteristics.
MOCVD processes can result in excessive impurity incorporation, leading to inferior device performance characteristics such as high leakage current. For example, ALD TiN has shown a 100-fold improvement in leakage current as compared with TiN MOCVD. MOCVD films often have inferior step coverage to conformal ALD films. Gas phase reactions lead to particulate contamination problems not seen with ALD. ALD, however, has low throughput and thus is most appropriate for thin (<1 to 20nm) layers.
Single-wafer ALD vs. batch ALD
The use of single-wafer ALD systems for production is justified by sequential integrated processing (clustering of gate dielectric and electrode material and/or pre- and post-processing). For single-layer deposition (e.g., capacitor dielectrics), batch-based ALD is favorable from a cost-of-ownership point-of-view (Fig. 1). A reduction of 50% or more in per-wafer chemical consumption can be achieved by migrating to a batch tool.
Figure 1. Comparison of chemical consumption and throughput for ALD Al2O3 deposited in single-wafer cluster and batch systems
Plasma-enhanced single-wafer ALD
Plasma-enhanced ALD is a conventional ALD process enhanced by the presence of atomic radical species (such as atomic hydrogen, oxygen, or nitrogen) and other excited atomic and molecular species created by the plasma. This enables the deposition of high-quality thin films at lower temperatures and from a larger variety of chemical precursors than is possible without plasma. The concern of plasma damage, however, needs to be addressed.
New precursors and films
In addition to the tighter constraints on film thickness and uniformity, material properties, such as dielectric constant, change for advanced device geometries. In addition, typical ALD precursors are much more expensive than traditional MOCVD chemicals. A potential solution to this problem is the use of “single-source” precursors containing two or more of the elements required in the binary or ternary compound film in a single molecule.
The use of “virtual chemistry” in design and screening of potential monomolecular candidates for ALD is the evolutionary result of principles laid out earlier . A major advantage of this approach is that the element-to-element ratio in the precursor matches that required in the deposited film. Another, more flexible approach is the simultaneous co-injection of multiple, compatible precursors. For example, co-injection of compatible alkyl-amine Hf and Si precursors allows precise composition control of hafnium silicate film stoichiometry .
To achieve the uniformity necessary to meet the requirements of <90nm nodes, the process chamber volume of the ALD system used (in this case, a Verano 5000) must be minimized to decrease the time required to cycle purge between chemical pulses. Dual vertical injectors positioned at the wafer edge are used for rapid alternation between metal-organic precursor and oxidizer while establishing cross-flow gas dynamics, regardless of load size.
Figure 2. Step coverage of HfSiOx in a deep trench (AR 80:1) deposited by dual-DLI ALD in the Verano 5000 batch ALD system; thickness increased to emphasize step coverage.
High-k dielectric films including aluminum oxide, hafnium oxide, and hafnium silicate are deposited using liquid metal-organic precursors and ozone. For hafnium silicate, mean within-wafer thickness uniformity of 0.8% (1-σ/mean) and within batch wafer-to-wafer thickness repeatability of 0.8% (1-σ/mean) were achieved (Fig. 2). Run-to-run thickness uniformity of 2.5% and 100% conformal step coverage into high aspect-ratio trenches were also obtained.
Now that researchers have evaluated the films deposited by single-wafer ALD, the drive toward economic manufacturability accelerates. Thickness control, uniformity specifications, and economics of advanced materials deposition processes are pushing innovation in control systems, hardware, and design for chemical delivery. The development and implementation of exotic, low vapor-pressure precursors are posing new challenges as well. A batch ALD tool that meets the technical demands of future technology nodes for large-scale production was presented.
The authors would like to acknowledge the hard work, dedication, and support of the technical staff at Aviza Technology. Special thanks go to Carl Barelli, S.G. Park, Larry Bartholomew, and Yoshi Senzaki. Verano is a trademark of Aviza Technology Inc.
- H. Treichel, O. Spindler, T. Kruck, “Molecular Engineering in Semiconductor Technology: Borosilicate Glass by Decomposition of a Monomolecular Precursor,” Proc. 7th European Conf. on CVD, pp. 747-756, 1989.
- H. Treichel, A. Mitwalsky, G. Tempel, G. Zorn, D.A. Bohling, “Deposition, Annealing, and Characterization of High-dielectric Constant Metal Oxide Films,” Adv. Mat. for Optics and Electron., Vol. 5, pp. 163-175, 1995.
- Y. Senzaki, S. Park, H. Chatham, L. Bartholomew, W. Nieveen, “Atomic Layer Deposition of Hafnium Oxide and Hafnium Silicate Thin Films Using Liquid Precursors and Ozone,” J. Vac. Sci. Technol. A 22, pp. 1175-1181, July/Aug. 2004.
Helmuth Treichel is director of process technology at Aviza Technology Inc., 440 Kings Village Dr., Scotts Valley, CA 95066; e-mail firstname.lastname@example.org.
Hood Chatham is principal R&D engineer at Aviza Technology Inc.
Cole Porter is the applications lab manager at Aviza Technology Inc.
Yoshi Okuyama is a senior process engineer at Aviza Technology Inc.