Final Report: Metal Deposition for Microelectronics Using CO2 as a SolventEPA Contract Number: EPD04042
Title: Metal Deposition for Microelectronics Using CO2 as a Solvent
Investigators: DeYoung, James P. , Taylor, Doug
Small Business: MiCell Technologies Inc.
EPA Contact: Manager, SBIR Program
Project Period: March 1, 2004 through August 31, 2004
Project Amount: $70,000
RFA: Small Business Innovation Research (SBIR) - Phase I (2004) RFA Text | Recipients Lists
Research Category: Nanotechnology , SBIR - Nanotechnology , Small Business Innovation Research (SBIR)
The goal of this research project was to develop a new method for depositing metallic thin films of copper and other metals for use as barrier layers, seed layers, and copper interconnects. This process could replace the current electroplating approach used in filling deep trenches and forming thin films in microelectronic circuit manufacturing and physical vapor deposition, which is limited in barrier deposition application because it is a line-of-sight metal deposition method. The electroplating process generates large quantities of aqueous waste that must be treated in place. The process described here utilizes liquid or supercritical CO2 as the solvent. This research project is part of an overall strategy to replace all aqueous and organic solvents in microelectronics fabrication.
Fluid displacement deposition utilizes a two-step approach to form thin metallic films on substrates. In the first step, organometallic precursors are dissolved in either liquid or supercritical CO2. The wafer to be coated is immersed in either the liquid or supercritical solution, and the solution is displaced with either CO2 itself or another supercritical fluid such as helium. The displacement step causes the formation of a thin organometallic precursor film to be deposited on the wafer surface. Film thickness can be controlled by adjusting the precursor solution concentration and/or the displacement rate. Because of the low surface tension and viscosity of the CO2 phase, the precursor will penetrate uniformly into the high aspect ratio features on a patterned wafer substrate. After the displacement step, the system is heated and a reducing agent such as hydrogen is added to the deposition chamber to remove the ligands bound to the central metal atom of the deposited organometallic precursor film. The net reduction reaction leaves behind a metal film and hydrogenated ligand byproducts that are ideally displaced from the surface.
Experiments also were conducted with the organometallic precursor dissolved in supercritical CO2. The substrate was heated to a minimum temperature and a reducing agent was added. A metal film was observed to grow on the substrate.
Micell Technologies, Inc., successfully deposited Cu and Ru metal films onto the indicated substrates using fluid displacement deposition for barrier, seed layer, and interconnect applications. The films were smooth, with root mean square roughness values similar to chemical vapor deposition (CVD) deposited films depending on the specific metal. The films were resistive and contained low levels (less than 3 percent) of impurities as measured by x-ray photoelectron spectroscopy. The significant impurity was carbon, which is believed to result from inefficient removal of the organometallic precursor ligands from the growing metal film. During the reduction process to remove the ligands, significant pressure from the reducing gas impedes the efficient removal of the ligands as they are produced during the reduction reaction. The partial pressure of the reducing gas was not optimized in these experiments for contaminate reduction. Experiments in which CO2 was added to the system in addition to the reducing agent did not reduce the carbon contamination. It was not possible to add CO2 at pressures that would have created liquid CO2 because the liquid CO2 would have removed the adsorbed organometallic precursor film prior to the film having the opportunity to be reduced to metal. Scanning electron microscopy (SEM) micrographs revealed the films to have granular morphology similar to, but smoother than, CVD-deposited metal films. The film resistivity arises from the disconnected grains as revealed in SEM micrographs and the presence of impurities in the film.
Some films passed the pressure-sensitive adhesive peel test but failed scratch and peel tests. This suggests that the metal film is weakly adhered to the substrate. The adhesion properties of the film were dependent on the time permitted for the reduction reaction. Short reduction reaction times (typically 5 minutes) led to films that could be brushed off with finger pressure, while films produced from longer reduction reaction times (up to 1 hour) were moderately conductive but could be scratched with a fingernail. Images from SEM micrographs suggest that films produced at shorter reduction reaction times consists of particle-like grains appearing more spherical in nature than hemispherical. The spherical nature of the particle produces a higher contact angle with the substrate surface so that the surface is not well wetted by the metal film. Many experiments aimed at cleaning the wafer surface prior to deposition did not produce any improvement in film adhesion to the cleaned substrate. Experiments also were conducted in which the wafer chip was heated during the displacement step. It was thought that this would improve the film adhesion because of the nature of metal film formation kinetics. These experiments did not leave behind a visible organometallic precursor film as was seen when the displacement step was carried out at room temperature. X-ray photoelectron spectra, however, revealed the presence of metal atoms left behind from the reduction reaction. This suggests that a very thin film of the organometallic precursor was deposited on the substrate. Attempts to cyclically deposit visible organometallic precursor films by recycling the precursor solution and displacing the recycled solution still did not produce a visible precursor film despite multiple cycles.
By altering the conditions of the deposition and reduction process, it was possible to produce strongly adherent films that passed all adhesion tests if the reduction was carried out at the same time as deposition. SEM micrographs showed that the film consisted of hemispherical grains that wet the substrate surface. Films produced in this manner were conducting and had no detectable impurities as measured by x-ray photoelectron spectroscopy after initial surface sputtering. Copper metal films produced in this manner were conductive, with resistivity of 4 μΩ-cm for copper and 25 μΩ-cm for ruthenium as measured by 4-point probe methods). These films had low-root mean square surface roughness values of 11 and 5 nm for Cu and Ru, respectively. SEM micrographs revealed a granular structure similar to CVD-deposited films. X-ray diffraction showed that films produced in this manner were polycrystalline. Film thickness could be controlled and most films produced in these experiments were 17 ± 3 nm thick.
Micell Technologies, Inc., successfully demonstrated its capabilities to deposit metal films using supercritical CO2 in application areas potentially advantageous to the semiconductor industry. Depending on the specific method used for deposition, the films generally are of high quality and are suitable for use in electronic devices. Metal film adhesion remains problematic if the metal film is deposited in a two-step process, but it is not problematic when the metal film is deposited in a single-step process. This technology could find widespread use within the semiconductor industry at a sufficiently small technology node. The technology discussed here has commercialization potential within the semiconductor industry.