AVS1996 Session TF-ThM: Semiconductor Metallization

Thursday, October 17, 1996 8:20 AM in Room 107B

Thursday Morning

Time Period ThM Sessions | Abstract Timeline | Topic TF Sessions | Time Periods | Topics | AVS1996 Schedule

Start Invited? Item
8:20 AM TF-ThM-1 Microstructure Control in Semiconductor Metallization
J. Harper, K. Rodbell (IBM T.J. Watson Research Center)
The microstructure of thin film interconnections on semiconductor devices is becoming increasingly important as linewidths decrease below 0.5 \mu\m. At these dimensions, the reliability and performance are influenced by specific microstructural features, rather than by average material properties. Recent studies of crystallographic orientation distributions in device metallization have shown a wide variety of microstructures which depend on film composition, substrate conditions and deposition parameters. Examples include aluminum alloy metallization, copper alloy metallization and titanium silicide contact metallurgy, which demonstrate the effects of geometrical confinement, substrate roughness, second-phase precipitation and polymorphic transformation. Further complexity is found in films deposited with off-normal incidence atom flux and energetic particle bombardment, which have shown dramatic effects in molybdenum, niobium and transition metal oxide films. The mechanisms which control the microstructure will be reviewed for these examples, and the prospects for full control of the microstructure of semiconductor metallization will be discussed.
9:00 AM TF-ThM-3 Copper Diffusion into Aluminum Metallizations Determined by Auger Electron Spectroscopy Elemental Imaging
G. Ramseyer, L. Walsh, J. Beasock, H. Helbig (Rome Laboratory); R. Lacoe (The Aerospace Corporation)
Copper concentrations of up to a few percent in aluminum semiconductor metallizations increase interconnect reliability as evidenced by longer lifetimes reported during accelerated testing. Copper diffuses in the direction of the electrons along an interconnect, reducing aluminum migration, and leaving behind regions depleted in copper. Once copper has migrated from the aluminum matrix, aluminum rich regions have reduced lifetimes similar to that of pure aluminum. If the diffusing copper is replenished from a reservoir of copper, then lifetimes will increase. Colgan [1] and Barkshire [2] explored the distribution of 2-4% copper in aluminum and its influence on electromigration. Walsh [3] concentrated on an evaluation of the thermal aspects of accelerated testing of aluminum and copper with no applied current. Copper was found to have diffused vertically after annealing. The experiments reported here were designed to further elucidate the interactions of aluminum and copper. A series of blanket films of 930 nm of aluminum over 147 nm of copper were annealed at up to 400 C for 1 hr. The diffusion of copper into the aluminum was determined by Auger Electron Spectroscopy (AES) depth profiling with Zolar rotation. Also, AES elemental maps were made of aluminum and copper at various times during the depth profiles, which showed that the diffused copper was present as CuAl\sub 2\ precipitates. These maps showed not only vertical, but also horizontal diffusion of copper in the form of this precipitate. Unpassivated test structures were fabricated from the aluminum/copper blanket films for electrical testing. Horizontal and vertical copper diffusion into aluminum was determined as a function of both accelerated temperature and electrical stressing by a combination of AES depth profiling and elemental mapping. The contributions of electrical stressing were then separated from those of thermal stressing by subtraction the thermal results. The AES mapping of aluminum and copper during depth profiling resulted in a more complete understanding of the interactions of aluminum and copper. The CuAl\sub 2\ precipitates, which were determined by imaging, are critical to understanding the interactions of aluminum and copper when copper is present at greater than 4% in semiconductor metallizations. 1. E.G. Colgan and K.P. Robdell, J. Appl. Phys., 75, p. 3,423 (1994). 2. I.R. Barkshire and M. Prutton, J. Appl. Phys., 77, p. 1,082 (1995). 3. L.H. Walsh, G.O. Ramseyer, J.V. Beasock, H.F. Helbig and K.P. = MacWilliams, Polycrystalline Thin Films II - Structure, Texture, = Properties, and Applications, Mat. Res. Soc. Proc., 403, in press.
9:20 AM TF-ThM-4 Copper Deposition into Contact Holes using Self-sustained Magnetron Sputtering
Z. Radzimski (North Carolina State University); W. Posadowski (Technical University of Wroclaw, Poland); S. Rossnagel (IBM T.J. Watson Research Center); S. Shingubara (Hiroshima University, Japan)
As integrated circuits geometries reach deep submicron dimensions, the interconnect system becomes a limiting factor. Relatively high resistance of Al connections which limits the device speed and Al electromigration problems force researchers to look for other alternatives. One of the strong candidates is copper. One of the tasks to be addressed before copper implementing in ULSI technology will be to establish deposition techniques which yield a copper film of high quality, low resistance and good contact filling. Sputtering is a very attractive alternative because it enables one to control the microstructure and step coverage of the film. However, such effects as the kinetic energy of deposited species, the scattered or isotropic nature of deposition process as a function of pressure, the deposition rate and the arrival rate of impurity have to be considered. In this work we investigate copper deposition using an ETERNA 100 DC magnetron source operating in a self-sputtering mode without argon. This mode of magnetron operation offers several unique features which needs to be addressed in terms of contact hole coverage, such as: a reduced presence of inert gas particles in the deposited films, a possibility of deposition on substrates placed in a distance from the target much greater than the mean free path, a decrease of the plasma discharge influence on the substrate. We will illustrate various conditions of magnetron sputtering (substrate-target distance, substrate bias) where the contact hole filling is promoted and enhanced by the self-sputtering process when compared to standard sputtering in the presence of argon.
9:40 AM TF-ThM-5 Growth and Carbon Incorporation Chemistry of Dimethylaluminum Hydride
B. Willis, K. Jensen (Massachusetts Institute of Technology); A. Jones (Epichem)
For the next generation of submicron devices, there is interest in replacing tungsten chemical vapor deposition (CVD) with aluminum CVD for via filling in multi-level metallization. Dimethylaluminum hydride (DMAH) is a potential precursor for use in a manufacturable aluminum CVD process. We present temperature programmed desorption (TPD), molecular scattering, and high resolution electron energy loss spectroscopy (HREELS) studies of the surface reactions involved in the growth of aluminum films and the mechanisms for carbon incorporation when using DMAH. Previous surface studies of DMAH found carbon residue on the surface after the reaction in contrast to reported growth experiments showing no significant carbon contamination.\super 1\ \super 2\ We present evidence for a clean, carbon-free reaction on the Al(100) surface and propose an explanation consistent with the previous results, reported growth behavior, and our spectroscopy data. The results further suggest that a hydrogen terminated surface is not needed for clean growth of aluminum, contrary to previously proposed mechanisms for growth.\super 3\ Trimethylaluminum (TMA) is shown to be a product of the growth reactions and to provide a route for methyl elimination from the surface. If the reaction is conducted at high temperatures (>600K ), the reaction pathways which produce TMA compete with other pathways which lead to carbon contamination. The production of TMA from DMAH also has implications for the effectiveness of the precursor in terms of aluminum incorporated relative to the amount of precursor fed to the reactor. \super 1\D.R. Strongin and P.B. Comita, J. Phys. Chem, 95, 1329 (1991). \super 2\ N. Zhu, T. Cacouris, R. Scarmozzino, and R.M. Osgood Jr., J.Vac. Sci. Technol. B10 1167 (1992)
10:00 AM TF-ThM-6 Interfacial Reaction Pathways and Kinetics in Polycrystalline Al/W and Al/TiW Bilayers
D. Bergstrom, J. Chun, I. Petrov, J. Greene (University of Illinois, Urbana)
Polycrystalline bcc W and Ti\sub 0.25\W\sub 0.75\ films, 140 nm thick, were grown on oxidized Si substrates by ultra-high-vacuum (UHV) magnetron sputtering at T\sub s\ = 600 C. 190-nm-thick Al overlayers were then deposited at T\sub s\ = 100 C without breaking vacuum. Changes in bilayer resistivity R\sub s\ during UHV annealing were monitored continuously as a function of temperature T\sub a\ during thermal-ramping and as a function of time t during isothermal annealing. In addition, Rutherford backscattering spectroscopy, x-ray diffraction, transmission electron microscopy (TEM), and scanning TEM (in which cross-sectional specimens were analyzed by energy-dispersive x-ray analysis with a 1 nm resolution) were used to follow area-averaged and local interfacial reaction paths as well as microstructural changes as a function of annealing conditions. Information from microchemical and microstructural analyses was used to model the R\sub s\(T\sub a\,t) results and determine reaction kinetics and activation energies. For Al/W bilayers, the results show that the reaction begins at T\sub a\ = 450 C with the formation of both WAl\sub 4\ and WAl\sub 12\. The WAl\sub 4\ layer is discontinuous, and the formation of the dominant reaction product,WAl\sub 12\, is limited by W diffusion between the WAl\sub 4\ grains and into the Al layer with an activation energy of 2.7 eV. In the Al/TiW bilayers, however, Ti enhances the nucleation of the WAl\sub 4\ and a thin (~20 nm) but continuous layer of WAl\sub 4\ forms at the Al/TiW interface at T\sub a\ = 450 C resulting in a structure which is stable to annealing temperatures above 500 C where the formation of both WAl\sub 12\ and Al\sub 3\Ti are limited by diffusion through WAl\sub 4\ layer occuring with a combined effective activation energy of 4.6 eV.
10:20 AM TF-ThM-7 Crystallographic Texture of TiSi\sub 2\ as a Function of Deep Submicron Structure Geometry and Processing Conditions
V. Svilan, K. Rodbell, L. Clevenger, C. Cabral, Jr., R. Roy, J. Harper (IBM T.J. Watson Research Center)
Detailed analysis of the crystallographic texture of C54 TiSi\sub 2\ was performed and showed strong influence of structure geometry and processing conditions on the preferential crystallographic orientation. The study was undertaken on blanket and patterned TiSi\sub 2\ films formed in the reaction between 32 nm of Ti and undoped polycrystalline silicon using both in-situ x-ray diffraction during heating and post-anneal four-circle pole figure reflection geometry. Full pole figures were taken to determine the C54 TiSi\sub 2\ grain orientation distribution in narrow (0.2 to 1.1 \mu\m) lines which was compared with similar results obtained from the unpatterned (blanket) films. While in blanket films we found the presence of weak <311> C54 TiSi\sub 2\ crystallographic orientation perpendicular to the sample surface, the <040> preferential orientation dominated in patterned submicron line structures and increased dramatically with decreasing linewidth. Using pole figure analysis we observed strong <040> fiber texture in narrow lines with a slight variation in the tilt of the (040) plane normal in the direction perpendicular to the line (FWHM=6 deg.), and a smaller variation in the direction along the length of the line (FWHM=2 deg.). In addition, a preferred in-plane (azimuthal) orientation of <040> crystals was found which showed that most of the <040> grains had their (004) plane normal oriented parallel with the line direction. Finally, it is also shown that lower temperature isothermal holds favor the formation of <311> orientation, while higher temperature isothermals and 3 C/s heating rates favor the formation of <040> orientation.
10:40 AM TF-ThM-8 Development of Microstructure in Titanium Silicide CVD
R. Southwell, E. Seebauer (University of Illinois, Urbana)
# Titanium disilicide (TiSi\sub 2\ finds widespread application as a contact material for metallization in ULSI integrated circuits. As device dimensions shrink, CVD has become an increasingly attractive alternative to the conventional salicide process because of the potential to avoid problems with substrate consumption. A crucial issue in both salicide and CVD processes is the transformation of the growing film from the C49 phase to the desired low-resistivity C54 phase. This transformation occurs with great difficulty for salicide processing in confined geometries. The present work examines the same transformation in CVD through a combination of in situ real-time kinetic measurements and ex situ X-ray and resistivity measurements. At 700-750\super o\C, the films nucleate as C49, but transform very rapidly (seconds) to C54 as the separate grains coalesce into a continuous film at 400-700\Ao\ film thickness. Even in confined geometries, the phase change occurs with greater ease than in the salicide process. The relative roles of Si surface diffusion and SiH\sub 4\ adsorption in determining substrate consumption during this nucleation and coalescence have been determined, with very practical implications for source gas flow control in manufacturing processes.
11:00 AM TF-ThM-9 Thin Film Markers for Monitoring Vacancy Diffusion Outside the Growth Zone during Alloy Phase Formation
E. Zingu (University of the Western Cape, South Africa)
Interdiffusion of bilayers often results in the growth of one or more alloy phases. The planar growth of such an alloy phase requires at least one of the constituents to diffuse through the alloy layer towards the growth interface. The most common transport mechanism during the interdiffusion is vacancy diffusion. A vacancy flux is therefore associated with the flux of atoms through the alloy layer. These vacancies are generated at the growth interface and diffuse through the growing alloy layer in the opposite direction to the flux of atoms. The ultimate fate of the vacancies is of great interest in the development of alloys since the accumulation of vacancies could lead to the formation of voids. Silicide formation (alloying of metal and silicon) in various metal/silicon systems have been investigated. In the case where the metal atom is the dominant moving species, the metal vacancies are generated at the silicon/silicide interface, diffuse through the silicide layer, and arrive at the silicide/metal interface. It is of great importance to establish whether the vacancies are annihilated at the interface or whether they diffuse through the metal layer and become annihilated at the free metal surface. Very thin layers of inert material is usually used to monitor the flux of atoms through the growing alloy layer in order to identify the dominant diffusing species. In this work thin layers of Ta and Pt have been used as "inert" markers to monitor the diffusion of vacancies during the metal silicide formation. Thin films of metal (Co, Ni, Pd, Cr) were deposited on Si substrates. A thin layer of marker atoms was initially embedded in the metal layer, but far from the silicon/metal interface. The change in the position of the marker layer is a measure for the flux of metal atoms diffusing past the marker layer. Since the vacancy and atom fluxes are coupled, the movement of vacancies is also deduced from the repositioning of the marker atoms. The paper will present experimental results in support of the technique to monitor vacancies in metal layers.
Time Period ThM Sessions | Abstract Timeline | Topic TF Sessions | Time Periods | Topics | AVS1996 Schedule