Low temperature inorganic chemical vapor deposition of Ti-Si-N diffusion barrier liners for gigascale copper interconnect applications

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be important differences between the submitted version and the official published version of record. People interested in the research are advised to contact the author for the final version of the publication, or visit the DOI to the publisher's website. • The final author version and the galley proof are versions of the publication after peer review. • The final published version features the final layout of the paper including the volume, issue and page numbers.

A new low temperature inorganic thermal chemical vapor deposition process has been developed for the growth of titanium-silicon-nitride ͑Ti-Si-N͒ liners for diffusion barrier applications in ultralarge scale integration copper interconnect schemes. This process employs the thermal reaction of tetraiodotitanium (TiI 4 ), tetraiodosilane (SiI 4 ), and ammonia (NH 3 ) as, respectively, the individual Ti, Si, and N sources. Ti-Si-N films were successfully grown over a broad range of deposition conditions, including wafer temperature, process pressure, and TiI 4 , SiI 4 , and NH 3 flows ranging, respectively, from 350 to 430°C, 0.1-1 Torr, and 2.5-8.0, 2.5-12.5, and 100-250 sccm. Film stoichiometry was tightly tailored through independent control of the Ti, Si, and N source flows. Film properties were characterized by x-ray photoelectron spectroscopy, Rutherford backscattering spectrometry, transmission electron microscopy, scanning electron microscopy, x-ray diffraction, and four-point resistivity probe. Resulting findings indicated that the texture and resistivity of the Ti-Si-N system were dependent on composition. In particular, films with a Ti 33 Si 15 N 51 stoichiometry exhibited a nanocrystalline TiN phase within an amorphous SiN matrix, highly dense morphology, resistivity of ϳ800 ⍀ cm for 25 nm thick films, and step coverage of ϳ50% in 130 nm wide, 10:1 aspect ratio trenches. Oxygen and iodine contaminant levels were below, respectively, 3 and 1.4 at. % each. Preliminary copper diffusion-barrier studies indicated that barrier failure for 25 nm thick Ti 34 Si 23 N 43 films did not occur until after annealing for 30 min at 700°C. © 2000 American Vacuum Society. ͓S0734-211X͑00͒07904-X͔ Copper based interconnects have almost universally replaced aluminum in high-performance integrated circuitry applications. This transition was driven by copper's lower resistivity and improved electromigration resistance, which allow faster signal propagation speed and higher performance characteristics. However, the successful incorporation of copper into subquarter-micron device generations requires effective chemical, structural, mechanical, and electrical compatibility with the surrounding low dielectric constant insulators. Most of the resulting target specifications could be achieved through the identification of appropriate liners that prevent copper diffusion into the dielectric, and promote viable copper-dielectric interlayer adhesion. 1 The need for a diffusion barrier and adhesion promoter is further mandated by the high diffusivity of Cu into silicon and its poor adhesion to silicon dioxide and low dielectric constant materials. The presence of copper in silicon results in the formation of deep level traps that cause device degradation and failure. Liner materials are also required to be thermodynamically stable with respect to the copper and dielectric layers, and preferably exhibit an amorphous structure to eliminate the high diffusion pathways typically provided by grain boundaries. More important, they must sustain their desirable properties at extremely reduced thicknesses to ensure that most of the effective volume of the trench and via structures is occupied by the actual copper conductor. 1 In this respect, ternary refractory metal liners such as the titanium-silicon-nitrogen ͑Ti-Si-N͒, tantalum-siliconnitrogen ͑Ta-Si-N͒, and tungsten-boron-nitrogen ͑W-B-N͒ systems could act as viable diffusion barriers in copper metallization due to their favorable chemical, structural, and thermal properties. In particular, the Ti-Si-N phase represents a highly desirable option due to the fact that Ti-based liners have already gained wide acceptance in semiconductor fabrication flows. In addition, the availability of Ti-Si-N in amorphous form provides an added incentive, in view of the a͒ Author to whom correspondence should be addressed; electronic mail: ee@cnsvax.albany.edu absence of grain boundaries that tend to act as fast diffusion paths for copper migration. In this respect, the amorphous Ti-Si-N phase has been shown to be stable against crystallization at temperatures as high as 1000°C, with the latter being strongly dependent on film stoichiometry. 2 As a result, various research groups have investigated the formation of Ti-Si-N films by a variety of physical vapor deposition ͑PVD͒ and metalorganic chemical vapor deposition ͑MOCVD͒ techniques, and documented their resulting performance as copper diffusion barriers.
In the PVD case, most of the diffusion barrier studies employed liners with thicknesses larger than 100 nm, which made the resulting conclusions less applicable to subquartermicron device structures. [3][4][5] The MOCVD route, on the other hand, exploited the reaction of tetrakis diethylamido titanium ͑TDEAT͒, silane (SiH 4 ), and NH 3 to deposit Ti-Si-N films over the temperature range from 300 to 450°C. 6 Barrier thermal stress ͑BTS͒ testing was subsequently performed on 10 nm thick Ti 23 Si 14 N 45 O 3 C 3 H 12 samples that were MOCVD grown at 400°C. The samples were shown to possess a mean time to failure ͑MTTF͒ approximately 10-100 times that of PVD TiN. 7 More recently, metalorganic atomic layer deposition ͑MOALD͒ was employed to deposit conformal Ti-Si-N layers at low substrate temperatures (180°C). However, no information was available with regard to film purity, resistivity, or barrier properties. 8 The work presented herein has focused on the development of a low temperature thermal CVD process that employs the thermal reaction of tetraiodotitanium (TiI 4 ), tet-raiodosilane (SiI 4 ), and ammonia (NH 3 ) as, respectively, the individual Ti, Si, and N sources. These source precursors are attractive because of their chemical simplicity, low dissociation energy, and reduced decomposition temperature as compared with other inorganic chemistries ͑such as TiCl 4 ͒. Additionally, SiI 4 was selected over SiH 4 in order to prevent the possible incorporation of hydrogen into the film as a result of incomplete decomposition of the SiH 4 molecule. The strategy espoused was to grow the Ti-Si-N phase in situ through the simultaneous reaction of appropriate mixtures of the Ti, Si, and N sources in the presence of hydrogen. Film stoichiometry was tightly tailored through independent control of the TiI 4 , SiI 4 , and NH 3 gaseous flows, with the reaction proceeding schematically as follows: Hydrogen was employed to ensure complete passivation and elimination of any free iodide reaction by-products from the reaction zone.
All Ti-SiN films were deposited in a customized, stainless steel, 200 mm wafer capable, warm-wall CVD system equipped with a high vacuum loadlock for wafer transport and handling without exposing the deposition chamber to air. This approach allowed tight control over the process stability and reproducibility. The chamber was also equipped with a parallel-plate-type, radio-frequency ͑rf͒ plasma capability for in situ wafer plasma cleaning. A Roots blower stack was used for process pumping. The TiI 4 and SiI 4 source precursors, which are solid at room temperature, were delivered to   This delivery approach did not require carrier gas, and allowed repeatable and controllable control over reactant flows to the process chamber. NH 3 and hydrogen (H 2 ) were used as coreactants. Undesirable gas phase reactions were eliminated through the use of a ''no-mix'' showerhead architecture, where the ammonia line was isolated from all other reactants until introduction into the reaction zone. For all depositions, three types of substrates were employed. Si͑100͒ wafers were employed for thermal diffusion barrier testing, while 500 nm thick thermally grown SiO 2 on Si was applied for composition, resis-tivity, and texture measurements, and patterned oxide structures were used for assessment of film conformality. Table I summarizes pertinent deposition parameters and corresponding ranges explored in this work.
The composition, microstructure, surface morphology, conformality, and electrical properties of the CVD Ti-Si-N films were analyzed by x-ray photoelectron spectroscopy ͑XPS͒, Rutherford backscattering spectrometry ͑RBS͒, x-ray diffraction ͑XRD͒, transmission electron microscopy ͑TEM͒, and four-point resistivity probe. Relevant information regarding these characterization techniques is given elsewhere. 9 In this respect, selected film properties are summarized in Table II, and the corresponding optimized processing conditions are listed in Table III.
Additionally, a preliminary evaluation was carried out of the performance of the CVD Ti-Si-N as a diffusion barrier in copper metallization. For this, 100 nm thick Cu films were ex situ sputter deposited onto 25 nm thick CVD Ti 33 Si 15 N 51 films grown on Si. The resulting stacks were annealed in 1 atm argon at 450, 515, 600, and 700°C for 30 min, along with sputter-deposited Cu/CVD TiN/Si stacks of identical thickness. The latter provided a comparative assessment of barrier performance. After annealing, the copper was stripped off in a diluted nitric acid solution, and the resulting samples were then analyzed by RBS to detect the presence of Cu, either in the liner material or in the underlying Si substrate. Figure 1 presents a typical RBS spectrum of an 82 nm thick CVD Ti-Si-N film deposited on Si at a wafer temperature of 430°C. RBS analysis indicated that the films were free of any heavy elemental contaminants, with the exception of ϳ1.4 at. % I. In this respect, higher NH 3 and SiI 4 flows at constant TiI 4 flow yielded, respectively, increased nitrogen and silicon incorporation into the resulting Ti-Si-N phase. For illustration purposes, Fig. 2 displays the XPS depth profile for the same type sample as that shown in Fig. 1. The XPS depth profile yielded a film composition of Ti 33 Si 15 N 51 , and indicated the presence of oxygen levels of ϳ3 at. %, which was within the typical background levels of the XPS system. Figure 3 presents a typical XRD spectrum of a 25 nm thick CVD TiSiN film grown on SiO 2 . The only XRD peaks detected were ascribed to TiN x phases. It is believed that this finding indicates that film microstructure consisted of a nanocrystalline TiN phase within an amorphous SiN matrix, in agreement with prior results in the literature on sputterdeposited Ti-Si-N films. 10 It was suggested that this Ti-Si-N microstructure is desirable from a diffusion barrier performance perspective, especially in view of the absence of grain boundaries. 11 The latter tend to act as fast diffusion paths for copper migration.
TEM was used to determine the film conformality in aggressive device structures. In this respect, Fig. 4͑a͒ exhibits the bright field TEM micrograph of 10 nm thick CVD-grown Ti-Si-N in a nominally 130 nm wide, 10:1 aspect ratio, trench structure. In addition, Figs. 4͑b͒ and 4͑c͒ display higher magnification TEM images of film profiles at, respectively, the top and bottom of the trench structure. TEM analysis yielded a film conformality of ϳ50%, and indicated the absence of thinning or loafing effects at the top and bottom corners of the trench structure. TEM imaging also confirmed the XRD results with respect to the existence of a nanocrystalline structural phase.
In terms of diffusion barrier performance, RBS results indicated the absence of any diffused copper in the CVD Ti-Si-N liner or underlying Si substrates after annealing at 600°C, as shown in Fig. 5͑a͒. In contrast, RBS detected the onset of copper diffusion for the same film after annealing at 700°C, as shown in Fig. 5͑b͒.
In summary, a new low temperature inorganic CVD Ti-Si-N process has been developed for applications as a diffusion barrier in copper metallization schemes. This process employs the thermal reaction of TiI 4 , SiI 4 , and NH 3 as, respectively, the individual Ti, Si, and N sources. Ti-Si-N films were successfully grown over a broad range of deposition conditions, including wafer temperature, process pressure, and TiI 4 , SiI 4 , and NH 3 flows ranging, respectively, from 350 to 430°C, 0.1-1 Torr, and 2.5-8.0, 2.5-12.5, and 100-250 sccm. In particular, films with a Ti 33 Si 15 N 51 stoichiometry exhibited a nanocrystalline TiN phase within an amorphous SiN matrix, highly dense morphology, resistivity of ϳ800 ⍀ cm for 25 nm thick films, and step coverage ϳ50% in 130 nm wide, 10:1 aspect ratio trenches. The oxygen and iodine contaminant levels were below, respectively, 3 and 1.4 at. % each. Preliminary copper diffusion-barrier studies indicated that barrier failure for 25 nm thick Ti 33 Si 15 N 51 films did not occur until after annealing for 30 min at 700°C. Additional studies are presently underway to further optimize the film conformality through modeling of film profiles in aggressive device topographies using fast simulators and to fully document corresponding diffusionbarrier characteristics by employing barrier thermal stress measurements.
The work presented herein was supported in part by SE-MATECH under Contract No. MLMA006, by MKS Instruments, and by the New York State Center for Advanced Thin Film Technology. This support is gratefully acknowledged.