Transition Metals Show Promise as Copper Barriers

Ultra-thin films in CMOS technology with shrinking via dimensions create new challenges for PVD at ≤65 nm. Conformal films with excellent step coverage are extremely important for copper barriers and seed layers at high aspect ratios. Traditional PVD has severe limitations due to line-of-sight deposition at ≤45 nm. Compared with PVD and CVD processes, atomic layer deposition (ALD) processes have demonstrated higher conformality, precise thickness control and improved film properties.

ALD is based on chemically saturated surface reactions between two reactants that are introduced sequentially. Repeated cycles of these reactants allow monolayer growth with each cycle or in a few cycles. The self-limiting nature of this chemisorption process gives rise to a very uniform and conformal growth behavior. ALD seems to be gaining acceptance in the semiconductor industry by enabling complete coverage on the sidewalls and bottoms of vias. Ultrathin (≤50 Å) barriers that reduce via resistance are therefore especially relevant for adherence to the International Technology Roadmap for Semiconductors (ITRS).

There is growing interest in barrier ALD applications for the 65 nm node and beyond. Although ALD is expected to become a mainstream approach at ≤45 nm, superior film properties from ALD are attractive for barrier applications even at 65 nm. Copper integration schemes at ≤130 nm use a low-k IMD/PVD Ta/TaN barrier/PVD copper seed/ECD copper material stack. Tantalum is chosen for better adhesion on low-k films and Ta/TaN films are used as copper diffusion barriers. The copper seed provides a conductive pre-layer for the copper electrochemical deposition (ECD) process. Thinner copper barriers could reduce via resistance as long as the barrier films do not impact electromigration of copper.

Ruthenium thin films deposited by CVD or ALD can potentially be used as seedless diffusion barriers between low-k IMD and copper interconnect at ≤45 nm. Ruthenium is a silvery white precious metal with low electrical resistivity and high thermal stability. It is relatively stable even in the presence of oxygen and water at ambient temperatures. It is conductive (resistivity ~7 µΩ-cm) and has a high melting point (~2300°C).

Thin transition metal films are known effective barriers against copper. Some of these metals such as chromium, tantalum, molybdenum and tungsten have been well studied, and exhibit very low solubility with copper. Other refractory metals such as ruthenium, rhodium and rhenium are expected to behave similarly. An important reason why alternate barrier layers attract interest is that a two-step Ta/TaN process may be reduced to a single ruthenium layer with adhesive and barrier properties. Copper films on ruthenium are reported to be continuous,¹ and there is ongoing work supporting ECD of copper directly on ruthenium. Ruthenium has negligible solubility in copper even at 900°C, as indicated in binary phase diagrams.² Consequently, ultrathin layers of ruthenium are being investigated not just as a replacement adhesive diffusion barrier layer, but also as a seed layer promoting ECD of copper in an IMD/ALD Ru/ECD copper stack.³

One of the most widely researched electronic applications of ruthenium is as a capacitor electrode in memory applications such as gigabit DRAMs and FRAMs, in which MIM capacitors have tantalum pentoxide- or perovskite-based capacitor dielectric films. The electrodes in these MIM capacitors require high comformality and low resistivity. Ruthenium thin films deposited using appropriate process conditions meet both requirements. Ruthenium can also be easily etched and patterned,⁴ and fluorine-based plasmas may be used for etching because ruthenium fluoride is a low melting and boiling point species.⁹

Yet another potential application of ruthenium is as a metal gate electrode in conjunction with high-k gate dielectric films. The increasing use of high-k gate dielectrics requires a switch from polysilicon gates to metal gates being used for PMOS and NMOS gate electrodes. Metal gates are expected to provide a range of benefits for gate-stack scaling such as reducing the high-k depletion region, preventing boron penetration and lowering gate resistance. Metal gates could be made of refractory metals, silicides and platinum group metals. Metals such as ruthenium, platinum, rhodium and iridium are being evaluated for this purpose as their work functions lie near the desired value for PMOS gate electrodes (~5 eV).

Most of the studies have experimented with CVD of metal electrodes, but reports of successful metal ALD are also emerging. Metal gates must be thermally stable up to dopant activation temperatures of 1000°C.⁵ This requires that there be no inter-diffusion between electrode and gate oxide, no chemical reactions at electrode/oxide interfaces, and no microstructure changes in the electrodes. Iridium is reported to have concerns with diffusion,⁵ while platinum may have issues with diffusion and etching.⁶ Ruthenium has been shown to be an effective candidate for PMOS gate electrodes.⁷ Additionally, RuO₂, the most thermodynamically stable oxide of ruthenium formed at temperatures above 200°C, has nearly as low a resistivity as ruthenium and a conductivity as good as that of some silicides used for metallization.

Precursor selection for CVD and ALD of ruthenium has been primarily based on thermal stability, reactivity, ease of reduction of the transition metal, and method of precursor delivery. Precursors that have been investigated are listed in Table 1 .

The most commonly studied precursors are Ru(Cp)₂, Ru(EtCp)₂ and Ru₃(CO)₁₂.^8-24 Other ruthenium precursors are b-diketonate complexes. Inorganic ruthenium halides such as ruthenium chloride are less interesting because of their low volatility and the potential concern of residual chlorine in the films. However, RuCl₃ is often the starting material for making other ruthenium complexes. Ru(EtCp)₂ is a liquid at room temperature, and is thermally stable even at higher temperatures. Ru(Cp)₂ is a solid whose melting point is ~200°C, and therefore must be heated for generating precursor flow. Ru₃(CO)₁₂ is also a solid and can be easily sublimed.

Since ALD operates at lower temperatures and can withstand lower mass flux than thermal CVD processes, high vapor pressure of precursors is not as stringent a requirement for ALD. Therefore, a wider selection of precursors can be used for ALD provided other requirements are fulfilled.

Ruthenium precursors are priced relatively higher than conventional CVD chemicals because ruthenium is a precious metal. However, precursor usage in ALD processes is also expected to be lower than in CVD processes, significantly reducing the impact of high precursor costs.

Ruthenium precursors used in CVD processes can be classified into two distinct groups: those that contain oxygen and those that are organometallic (direct M-CH_x bonds).

Organometallic precursors — CVD: The organometallic precursors of ruthenium such as Ru(Cp)₂ and Ru(EtCp)₂ have been widely used in CVD.^8-12 They react with O₂ to deposit ruthenium films at low temperatures (~250-500°C) and RuO₂ films at higher deposition temperatures.^7-8

Typically, there is a carrier gas such as argon in the process gas mix. RuO₂ formation can be limited by optimizing the partial pressure of oxygen and the deposition temperature. A process gas mix (O₂+Ar) with less than 80% O₂ resulted in mostly pure ruthenium films at 400°C.⁸RuO₂ films were predominant at temperatures higher than 500°C.⁹ Similar dependencies have been established for Ru(EtCp)₂. A 7% O₂ in the process gas mix at 230°C saturated the growth rate of ruthenium films while a 25% O₂ at 270°C still yielded a reaction-limited process.¹⁰ The formation of RuO₂ was observed at 300°C and higher temperatures.

Both precursors have also been compared under similar process conditions. The resistivity of ruthenium films deposited from these materials was as low as 11 µΩ-cm from Ru(Cp)₂ and 20 µΩ-cm from Ru(EtCp)₂ in the 240-260°C deposition temperature range.¹² Higher resistivity in the deposited films is typically caused by the incorporation of residual carbon, hydrogen and oxygen. The lower resistivities from Ru(Cp)₂ are likely attributable to the lower carbon and hydrogen content in the precursor. Controlled oxidation of the precursor by optimizing the process parameters has been demonstrated to reduce impurities and resistivities of the deposited films.¹¹ Even with Ru(EtCp)₂ and O₂ at 33% of the process gas mix, a film resistivity of 12 µV-cm was obtained at 300°C.

RuCp(i-PrCp), another purely organometallic precursor, was used to deposit ruthenium by CVD at 300-375°C.¹³ The thermal decomposition of this precursor and the oxidation behavior of the resulting ruthenium films were similar to Ru(EtCp)₂. The films had a resistivity of 13 µΩ-cm irrespective of deposition temperature. Single-phase ruthenium films were also deposited at 300-500°C from Ru(EtCp)₂ dissolved in a tetrahydrofuran (THF) solvent.¹⁴ The lowest resistivity for this combination was 25 µΩ-cm. Introduction of excess carbon and hydrogen from the solvent into the process is presumed to require more O₂ to deposit metallic ruthenium. Residual carbon and hydrogen in ruthenium films may have contributed to the higher resistivity. Additional reactions may also occur because of the solvent.

Film roughness was reported to be independent of O₂ flow rates, but dependent on deposition temperatures and type of substrate.¹³ The grain size increased with increasing deposition temperature from 325 to 400°C for 1000 Å films.

Film roughness may be attributed to nucleation difficulty on the substrate, which was ranked in the following descending order: SiO₂>Si>TiN>Ta₂O₅. The Gibbs function free energy per oxygen bond for metals decreases as Ti>Al>Ta>Cr>Si>Mo>W>Co>Re>Cu>Ru,¹⁵ which explains the difficulty of depositing ruthenium on oxides compared with other metals such as tantalum. Methods for overcoming ruthenium nucleation issues on dielectrics are being investigated.

A long incubation time was reported using Ru(EtCp)₂ at the initial stage of ruthenium growth on SiO₂ at 220°C.¹⁰ However, no incubation time was noted when the same ruthenium films were grown on a sputter-deposited ruthenium surface. Increasing the deposition temperature to 270°C reduced the incubation time to zero.

Ruthenium films have intrinsic tensile stress, which is reduced at higher deposition temperatures and growth rates. However, it may be desirable to deposit ruthenium at lower temperatures and produce low stress films that will not migrate, show good adhesion, demonstrate resistance to cracking, and reduce electromigration of copper.

Oxygen containing precursors — CVD: Of the oxygen-containing precursors, Ru₃(CO)₁₂ contains the metal atom in the Ru(0) state. The ligand here can cleanly and easily dissociate from the ruthenium metal.^9,17 Films deposited from Ru₃(CO)₁₂ by CVD at 250-400°C consisted entirely of ruthenium.⁹RuO₂ formation at temperatures <575°C was deemed unlikely. Carbonyl complexes deposit ruthenium films even in the absence of O₂, and excess oxygen favors the growth of RuO₂.¹⁶ Ruthenium film deposition from Ru₃(CO)₁₂ can be realized at as low a temperature as 150°C.¹⁷

Ruthenium films were also deposited using Ru(OD)₃ in methanol solvent at 300-400°C. The resistivity of ruthenium films was 20 µV-cm for 5-10% O₂ in the process gas mix.⁶ Pure metallic ruthenium films were also deposited on SiO₂ from a solution of Ru(THD)₃ in THF at 275-450°C without a distinct incubation time.¹⁸ It was proposed that the solvent reacted with the excess oxygen and thus prevented RuO₂ formation. The film thickness was proportional to the deposition time at all O₂ concentrations. This was compared with the case of Ru(EtCp)₂ dissolved in THF, where a distinct incubation time was noted. Ruthenium films with good step coverage on TiN/Si substrates (1:1 aspect ratios) were deposited at 250°C from Ru(THD)₂ COD.¹⁹ Deposition rates of 2.2 nm/min were realized on HfO₂/SiO₂/Si substrates at 250-270°C. The deposition process was shown to be surface reaction limited at these temperatures. The process became mass-transfer limited at 290°C and above.

All the prior work suggests that the solvent- and oxygen-containing precursors help in accelerated nucleation. But they may also contribute more residual carbon, hydrogen and oxygen impurities, where the purely organometallic precursors may have an advantage of being optimally oxidized. Typically, thicker films have shown lower resistivities (<25 µΩ-cm), but thinner films with resistivities close to the bulk value of 7 µΩ-cm are desired for barrier applications. The final precursor selection will depend on a balance of all these characteristics.

ALD of transition metals such as copper, cobalt, iron, nickel, tungsten, ruthenium and platinum has been demonstrated.^20-23 ALD ruthenium films from alternating pulses of Ru(Cp)₂ and air were deposited on Al₂O₃ and TiO₂ films on glass substrates at 350°C.²² The vaporization of Ru(Cp)₂ at 50°C was sufficient for realizing the desired growth rates, which were on the order of 0.45 Å/cycle. Air (20% O₂ in pulse mix) was used as the oxygen precursor. A period of low growth rate at the beginning of film deposition was reported, although film thickness could be controlled by the number of cycles. All films had resistivities <20 µV-cm, impurities were less than 0.4 at%, and no RuO₂ was formed. The carbon content in the films was always low, but the oxygen and hydrogen content increased (still <0.4 at%) as the deposition temperature was dropped from 400 to 300°C. ALD ruthenium films were deposited from Ru(EtCp)₂ on TiN/SiO₂/Si substrates at 270°C.²⁴ The O₂ ratio in the oxygen pulse mix was 44% (rest argon). Impurities were <2 at%, and resistivity was 15 µΩ-cm. Growth rates were 0.15 nm/cycle.

The precursors, process parameters and deposition methods described in the previous sections are summarized in Table 2 .

Transition metal complexes that can be used in ALD applications pose new challenges, especially with chemical stability and high precursor delivery temperatures. Ru(EtCp)₂, for instance, is one of the very few known liquid precursors of ruthenium. Many precursor candidates for ALD applications are solids (Table 3 ), and ruthenium ALD from solid sources has been reported.²⁵

It is more difficult to vaporize the solids in a controlled manner into a process chamber. This may require continuous heating of an organometallic solid or inorganic compound. Continuous heating causes the precursors to decompose and the equipment to endure harsh conditions. This may result in process fluctuations and film composition changes, and may contribute to particle contamination. Solid precursors have not been used very much because of insufficient vapor pressure and mass flux generation. A novel delivery system enabling the use of solid precursors consistently and reliably at low temperatures has been developed.²⁶ Such solid delivery systems will be required in high-volume manufacturing.