ICMCTF1998 Session H1-2: Advanced Metalization: Materials and Processes (2)

Though the benefits of copper as an interconnect material for high-density, high-performance devices have long been recognized, its adoption has been gated by a number of architecture and fabrication challenges. Over the course of the past half decade, these hurdles have been steadily yielding to intensive engineering efforts, and the industry appears poised for mass implementation of copper metallization. Two of the most important innovations which have opened the way to copper integration include the identification of effective conductive barriers to limit copper diffusion, and adoption of the damascene process flow which sidesteps difficulties of plasma/RIE etching imposed by the limited volatility of copper halides.

For reliable interconnect performance, the copper deposition process must be capable of filling the deep, narrow recesses that are intrinsic to the damascene flow without voids or seams. Though conventional metal deposition techniques have proven inadequate for the challenges posed by damascene topography, a number of candidate processes are under intense evaluation at research and manufacturing facilities worldwide, and demonstrating varying degrees of success. One simple and cost-effective method is electrolytic deposition of the copper film in a solid/liquid biphase reaction, which represents a departure from conventional vacuum deposition techniques (PVD, CVD, evaporation). Furthermore, because copper electrochemical deposition (ECD) can be performed selectively on unmasked regions of a substrate, it introduces the option for a variety of interconnect architectures which share the benefits of damascene patterning while limiting the requirement for metal CMP. This talk introduces some of the fundamental capabilities of copper electrochemical deposition and includes an overview of the process, along with a discussion of the reactor architecture. A system which integrates these reactors for fully automated, dry-to-dry processing in high-volume manufacturing environments is described.

A process of ion implantation seeding of electroplated thin films on silicon substrates has been developed in this laboratory recently¹ . Electroplating of permalloy thin films on silicon substrate seeded with Pd+ ion implantation was reported in that paper. In the present paper, we describe our results on the use antimony ion implantation seeding for electroformed gold thin films. One of the major advantages with ion implantation seeding is that it obviates the need for the deposition for a conducting metallic seed layer on silicon. More importantly, it reduces the number of lithographic processing steps such as photoresist patterning and lift off at different stages especially when selective electro plating is to be performed in particular regions of the wafer. This process will be very useful especially in the fabrication of high aspect ratio and multi level structures, where removal of excess seed layer can at times be cumbersome.

Silicon(100) susbtrates have been implanted with Sb+ ions extracted at 10kV at several doses in the range 1e14 to 1e17 using a MEVVA implanter. Ion implantation has been carried out at room temperature. During implantation certain portion of the wafer is masked to have both implanted and unimplanted regions on the same wafer. These are electroplated under similar bath conditions using gold plating solution for the same duration. Electroplating is performed using back plating method, having a 2000 + aluminium film on the back side of the wafer as contact layer. There appear to be a threshold dose below which it was not possible to plate. In this particular case, 7e14 has been found to be the threshold dose. Adhesion is better with the films plated on substrates implanted at a higher dose (1e16 ions/cm2) than at the lower doses. Film thickness has been measured using profilometer whereas the adhesion has been evaluated using scotch tape test. SEM and RBS techniques were used to analyse the surface topography of the plated film and the interface between the implanted wafer and gold film. These results showed that the thickness and grain size of the plated film increases with the implantation dose. Typical grain size of the film plated on to the wafer implanted at 7e14 dose is approximately 2 micron. In order to demonstrate the usefulness of this technique, a gold coil has been plated using this selective seeding process. The detailed results of our investigations to understand the mechanism of seeding process will be discussed.

¹ G.K.Muralidhar and D.K.Sood, A novel method for electroplating on silicon without the need of a continuous plating base film, paper to be presented in the SPIE Far East Pacific Rim Symposium on Smart materials, Structures and MEMS, during 10-13th December, 1997 in Adelaide, AUSTRALIA

TiN is an extremely important barrier material in present day integrated circuits. The most common deposition method is reactive sputter deposition from a titanium target in a argon-nitrogen gas mixture. In reactive sputter deposition of TiN a fraction of the nitrogen feed gas is incorporated in the film. This consumption might lead to non-uniform deposition for large wafersizes and/or low pumping speeds. For relatively low pumping speeds, or a large fraction of incorporated nitrogen, it was expected that the film uniformity over the wafer would deteriorate. TiN was deposited at fixed partial pressures of argon and nitrogen. The pumping speed was varied over one order of magnitude. The process proved to be insensitive to the pumping speed. Even at the lowest pumping speed tested, i.e. at 84% consumption of the supplied nitrogen, the uniformity did not detoriate.

We propose that the explanation of this insensitivity is due to the high reactivity of Ti atoms on the target surface to react with N atoms, while the chance for additional nitrogen atoms to be adsorbed at the nitrided target is very small. The target is operating in the poisoned mode which means that the target surface consists of TiN. Calculations showed that the number of N atoms striking the target is about 80 times larger than the flux of Ti atoms sputtered from the target. The sticking coefficient of N₂ on Ti is high, therefore the chance for the target to recover its TiN surface after a nitrogen atom or a TiN fragment is sputtered off is high. On the other hand the sticking coefficient of nitrogen on TiN is very low. An increase in partial pressure of nitrogen will not result in extra "adsorbed nitrogen" on the target. Therefore the ratio of titanium to nitrogen atoms removed from the target by the ion bombardment is insensitive to the partial pressure of the nitrogen and hence to the puping speed.

We present the results of a project in which throughput for a selected deposition process is maximized via a programmed rate protocol. While doing this maximization, we also maintain acceptable film properties. In previous work, we proposed ¹ and verified ² the programmed rate chemical vapor deposition (PRCVD) process. Using this protocol, poor step coverage usually associated with high deposition rates is overcome by ramping a process parameter. In our experimental verification, we ramp temperature in a cold wall, lamp-heated single wafer LPCVD reactor to verify the procedure. Previous work reported ^{3, 4} the results of the PRCVD protocol. In this work we report the step coverage, stress, and resistivities that result from using an optimally based trajectory (as described in ⁴) that maximizes step coverage and minimizes time.

¹ T.S. Cale, M.K. Jain and G.B. Raupp, J. Electrochemical Society, 137(5), 1990, p. 1526.

² K. M. Tracy, Master's Thesis, Arizona State University, 1996.

³ J. J. Kristof, L. J. Song, K. S. Tsakalis, T. S. Cale, 1997 Proceedings of the 14th International VLSI Multilevel Interconnection Conference, p. 207.

⁴ J. J. Kristof, L. J. Song, K. S. Tsakalis, T. S. Cale, Proceedings of the Electrochemical Society 1997 Joint International Symposium on Chemical Vapor Deposition: CVD XIV and EUROCVD 11, p. 1566.

This presentation describes a novel powder metallurgy method of mechanical alloying to synthesize the tetragonal CuIn_0.7Ga_0.3Se₂ phase in a quaternary Cu-In-Ga-Se sputtering target. Mechanical alloying is a powder metallurgy process involving repeated welding, fracturing, and rewelding of powder particles in a high-energy ball mill. In this process, the required amounts of the constituent metal powders and a hard grinding medium are loaded into a container and subjected to intense agitation. Alloying occurs due to the heavy deformation leading to refinement of powder particles (and consequent short diffusion distances) and increased defect density.

A homogeneous alloy powder was synthesized by mechanical alloying of blended elemental powders in a planetary ball mill, and characterized by x-ray diffraction, differential thermal analysis, and electron microscopy techniques. Depending on the milling conditions, either a metastable cubic or a stable tetragonal phase was obtained in the as-milled powder. The mechanically alloyed powder was consolidated by hot isostatic pressing at 750^oC and 100 MPa for 2 hours when a well-recrystallized alloy with full density was obtained. The presentation will deal with a detailed characterization of the powder, the compact, and its sputtering behavior.

The effective heat of formation (EHF) model has been used to determine the reaction sequences of three of the binary thin film couples comprising the Cu(In,Ga)Se₂ system. Investigation of the reaction sequences of the films was performed by differential thermal analysis (DTA) and X-ray diffraction (XRD). The effective activation energy of each reaction has been determined using the Kissinger technique. The binary systems that have been examined are Cu-Se, In-Se, and Ga-Se. Each of the films were deposited by means of thermal evaporation.

To predict formation of a binary system using the EHF model, it is necessary to consider not only the thermodynamics involved, but also the effective concentration of the two reacting species and the growth interface. The effective concentration is assumed to be the concentration which leads to the highest mobility and is usually given by the composition of the liquidus minimum of the binary system under consideration. After first phase formation, the EHF model predicts that the next phase to form is that which is richer in the unreacted element and which has the most negative EHF. When a reaction occurs during DTA, the corresponding change in heat content is represented by peak on the DTA curve. The activation energy of the reaction can be determined by the change in peak temperature as the heating rate is varied.

The three films that were analyzed were approximately fifty percent each element comprising the binary film. They were multilayer films which were deposited onto 12.7 micron molybdenum foil on soda lime glass. For the binary systems of In-Se, Ga-Se, and Cu-Se, both the first phases to form and the subsequent phase formation sequences of these thin film systems were correctly predicted by the EHF model.

Novel Pd/Sn and Pd/Sn/Au Ohmic contacts have already been developed for n-GaAs [1]. The feasibility of these metallizations on practical devices is not reported yet. In this paper, GaAs MESFETs have been fabricated using Pd/Sn and Pd/Sn/Au Ohmic contacts and Al Schottky contact. MESFETs fabricated with Pd/Sn/Au metallizations show improved characteristics compared to Pd/Sn contacts. The edge uniformity/definition of the Pd/Sn/Au metallization is suitable for VLSI GaAs devices. MESFETs fabricated with Pd/Sn/Au Ohmic contacts exhibit a maximum transconductance of ~120 mS/mm for a gate length of 2 micrometre.

[1] M.S. Islam and P.J. McNally, "Thermally Stable Pd/Sn and Pd/Sn/Au Ohmic Contacts to n-type GaAs," Thin Solid Films, 1997 (in press).