Atomic Interdiffusion and Diffusive Stabilization of Cobalt by Copper

Mar 13, 2014 - Electromigration of copper in integrated circuits leads to device failure. Potential solutions involve capping the copper with ultrathi...
0 downloads 0 Views 3MB Size
Letter pubs.acs.org/JPCL

Atomic Interdiffusion and Diffusive Stabilization of Cobalt by Copper During Atomic Layer Deposition from Bis(N-tertbutyl‑N′‑ethylpropionamidinato) Cobalt(II) Tyler D.-M. Elko-Hansen,† Andrei Dolocan,‡ and John G. Ekerdt*,† †

Department of Chemical Engineering, and ‡Texas Materials Institute, University of Texas at Austin, Austin, Texas 78712-1589, United States S Supporting Information *

ABSTRACT: Electromigration of copper in integrated circuits leads to device failure. Potential solutions involve capping the copper with ultrathin cobalt films. We report the properties of cobalt films after deposition on polycrystalline Cu at 265 °C by atomic layer deposition from H2 and bis(N-tert-butyl-N′-ethylpropionamidinato) cobalt(II) (CoAMD). We find intermixing of Co and Cu producing a transition layer on the Cu nearly as thick as the Co-rich overlayer. X-ray photoelectron spectroscopy and time-of-flight secondary ion mass spectrometry depth profiling reveal that a finite amount of Cu continuously segregates to the progressing Co surface, minimizing the free surface energy, throughout deposition up to at least 16 nm. The Cu-stabilized Co film initially follows 2D growth and strain-relieving 3D crystal formation is apparent beyond 2 nm of film growth. Depth profiling indicates that Cu likely diffuses within the Co film and along the polycrystalline Co grain boundaries. SECTION: Surfaces, Interfaces, Porous Materials, and Catalysis

K

much interest exists to develop alternative deposition methods.10 Chemical vapor deposition (CVD) and atomic layer deposition (ALD) processes are of particular interest due to their ability to deposit ultrathin and conformal films and their potential application for surface selective deposition of barriers.11,12 Previous work from Lim et al. demonstrated the deposition of Cu, Ni, and Co transition metals on a variety of substrates from a series of chelating amidinate precursors under ALD and CVD conditions.13 Bis(N-tert-butyl-N′-ethylpropionamidinato) cobalt(II) (CoAMD) is a proven ALD precursor for Co deposition that is amenable to carrier-gas-based vapor delivery and can be deposited using reducing agents like H2 or NH3 rather than O2.14 Avoiding oxidation is important for BEOL components, especially the Cu metallization lines. We have separately demonstrated that CoAMD exhibits self-limiting adsorption on Cu and a preference to deposit on Cu rather than on SiO2 or carbon-doped oxides that might comprise the BEOL interlayer dielectric materials (unpublished work). The inherent selectivity of CoAMD for Cu over Si-based dielectric materials makes it an interesting candidate for selective-ALD processes. In this study, we deposit sub-20 nm Co films by ALD from CoAMD on Cu substrates to better understand the Co/Cu interface and Co film properties. Understanding the properties

nown generally to exhibit good adhesion on Cu surfaces, Co and Co alloy films have recently been applied as Cu capping layers to mitigate premature microelectronic device failure due to Cu interconnect electromigration (EM).1−3 Decreasing microelectronic device dimensions continue to exacerbate the EM-induced self-diffusion of Cu. Therefore, reducing Cu EM-induced failures in back end of line (BEOL) interconnects without increasing resistance-capacitive delay is an ongoing concern.4 Cu alloys and EM-resistant metal capping layers are possible solutions to the EM challenge; however, capping is preferred to alloying due to resistivity increases associated with alloying Cu interconnects.5 Co/Cu multilayers have been of great interest for many years for their applications in magnetic and microelectronic materials. The Co/Cu interface and structure has garnered particular interest, and efforts have demonstrated lattice matching of Co on Cu (111) surfaces. Further, it was suggested that Co overlayers may be stabilized by monolayer-thick Cu segregating at the surface and by alloying at the interface.6,7 While alloying is generally unexpected in Co/Cu (111) systems given their low bulk solubility, it is energetically favorable for monolayers up to ∼50/50 mixtures.8 Further, Co capping layers have been demonstrated to reduce Cu EM more effectively than silicidation of the Cu surface. Co caps result in less resistance-capacitance increase in the metallization structure than SiCN caps and adhere better to Cu.1−3,5,9 Currently selective, electrolessly deposited CoWP is a benchmark Cu EM barrier.1,3,5 Nevertheless, contamination of the adjacent dielectric materials from the plating bath is a concern, and © 2014 American Chemical Society

Received: February 10, 2014 Accepted: March 13, 2014 Published: March 13, 2014 1091

dx.doi.org/10.1021/jz500281k | J. Phys. Chem. Lett. 2014, 5, 1091−1095

The Journal of Physical Chemistry Letters

Letter

of ultrathin films is important for meeting the ITRS roadmap goal of 0.5 nm barrier films by 2025.5 In situ X-ray photoelectron spectroscopy (XPS) and ex situ depth profiling time-of-flight secondary ion mass spectrometry (ToF-SIMS) measurements reveal the films’ chemical composition and atomic mixing. Ex situ characterization by scanning electron microscopy (SEM), cross-sectional transmission electron microscopy (TEM), and atomic force microscopy (AFM) provide structure and surface morphology of the films. Figure 1 depicts plan-view SEM images of Cu surfaces after 250 to 3000 ALD cycles (∼1 to 16 nm films, respectively,

doubles from 500 to 1000 ALD cycles and again from 1000 to 3000 cycles giving a final Co:Cu atomic ratio of 49.8:1. Although the surface faceting is indistinguishable between 500 and 1000 cycles, the increasing Co:Cu ratio suggests that the film thickens or increases in density with little surface modification. By 3000 cycles, the Co film surface appears polygranular and multifaceted (Figure 1d). Figure 2a,b presents AFM and cross-sectional TEM data from 1000- and 3000-cycle Co films, respectively. The TEM images indicate that these Co films are continuous and polygranular. The 1000- and 3000-cycle films, 4.5 and 16 nm thick, exhibit root mean square (RMS) roughnesses of 1.9 and 4.8 nm, respectively. By comparison, the underlying Cu substrate has an RMS roughness of 1.86 ± 0.05 nm after cleaning with acetic acid. Figure 3a,b presents normalized ToF-SIMS depth profiles of the 1000-cycle film, and Figure 3c,d presents normalized ToFSIMS depth profiles of the 3000-cycle film. As suggested by similar Co2− and Cu2− secondary-ion yields for both samples in the Co and Cu regions, respectively, the two films comprise roughly the same Co/Cu matrix; thus, a particular measured species, e.g., CuO−, may be compared directly between samples. Consequently, the depth profiles for a certain species of interest were normalized to their maximum between the two samples to give a relative quantitative comparison. For converting the sputtering time into depth, a sputtering-rate model assuming the instantaneous sputtering rate at the interface of two films as a linear combination of the individual sputtering rates was used.15 The sputtering rates for Co and Cu were calculated as 1.2 Å/s and 2 Å/s, respectively, based on the TEM-determined thicknesses (see Figure 2) and the corresponding time to sputter through the respective film layers. Simplified schematics of the relevant species in order of their appearance during depth profiling are depicted in the insets. The insets describe the sample as measured after storage in air before ex situ analysis by ToF-SIMS. Oxides of the film elements are due to ex situ storage before ToF-SIMS depth profiling. In situ XP spectra of the as deposited films show no O contamination of the Co or Cu (Supporting Information).

Figure 1. Plan-view SEM images of (a) 250, (b) 500, (c) 1000, and (d) 3000 cycle ALD depositions of CoAMD on Cu at 265 °C. The calibrated Co:Cu atomic ratio detected in XPS is listed in the upper right-hand corner of each image.

assuming conformal coverage and extrapolating from crosssectional TEM images of thicker 1000- and 3000-cycle films in Figure 2). The SEM images illustrate that minimal 3D faceting is evident after 250 cycles (Figure 1a). By 500 cycles (2.3 nm Co), 3D faceting of the Co surface is apparent in SEM and is effectively unchanged up to 1000 cycles (4.5 nm, Figure 1b,c, respectively). From 250 to 1000 cycles, the relative Co:Cu atomic ratio detected in XPS increases from 3.3:1 to 26.5:1 and

Figure 2. Cross-sectional TEM images and surface topography renderings from AFM of (a) 1000 cycle and (b) 3000 cycle ALD of CoAMD on Cu at 265 °C. 1092

dx.doi.org/10.1021/jz500281k | J. Phys. Chem. Lett. 2014, 5, 1091−1095

The Journal of Physical Chemistry Letters

Letter

Figure 3. Normalized ToF-SIMS depth profiles of 1000-cycle Co film (a,b) and 3000-cycle Co film (c,d). “Advent. Matl.” refers to adventitious material accumulated on the substrate during sample storage and after air transfer for analysis and was not present in situ following deposition.

The Co2− and Cu2− secondary ions were chosen as markers for the Co and Cu films, respectively, to avoid the intrinsic artifacts that Co− and Cu− signals have due to residuals originating from CoO− and CuO−, respectively. The Co/Cu interfaces for the 1000-cycle and 3000-cyle films are shown in Figures 3a and 3c, respectively, where the measured interface lengths are indicated as the distance between the 10% level of the markers forming the interface, i.e., the normalized Co2− and Cu2− signals, and read ∼3.6 nm and ∼10.2 nm, respectively.16 Interfacial atomic mixing lengths of ∼3 nm and ∼9.6 nm between the Cu substrate and the Co overlayer for the 1000cycle and 3000-cycle samples, respectively, were extracted from their corresponding measured interface lengths by applying the so-called mixing-roughness-information (MRI) model (details in the Supporting Information).17 Figure 3b,d contains more comprehensive ToF-SIMS profiles. Both samples present a finite amount of Cu, about 2 nm thick on the 1000-cycle and a few monolayers thick on the 3000-cycle film, mostly oxidized, segregated to the films’ free surfaces, as indicated by the black CuO− traces. A Co-rich layer is observed preceding the Co/Cu mixing region after which only bulk Cu is detected. For the 3000-cycle film, the double peak feature exhibited by the depth profile of CuO− located closely around the Co2− peak (Figure 3d) suggests that Co continues to burrow under the surface Cu as the film grows in order to reduce its surface free energy and that Cu diffuses along the Co grain boundaries at the base of the film. Moreover, the local minimum in the CuO− signal indicates that the surface Cu is a result of the initial Co burrowing and not

from grain boundary diffusion of Cu, which would result in a constantly declining Cu concentration from the substrate to the Co surface. The lower CuO− secondary-ion signal at the surface of the 3000-cycle film when compared to the 1000-cycle film suggests that the Cu, initially segregated at the Co surface, is pushed both toward the substrate and the surface by the progressing Co layer. Additionally, partial loss of CuO− signal at the surface may be attributed to an increasing roughness following the expansion of the Co capping layer from 3.6 to 16 nm. An increased roughness translates into a larger surface area (i.e., lower Cu surface density) and stronger shadowing effects during sputtering, both leading to a reduced and slower decaying secondary-ion yield of Cu-related species. Figure 4 is a schematic representation of Co film growth from CoAMD on Cu. The as-deposited Co films generally comprise a Co-rich, polycrystalline layer with a thin Cu surface coating and are free from oxidation. Co film growth on Cu proceeds first with the formation of a 2D wetting film and an intermixing layer of Co and Cu. Energy dispersive X-ray spectroscopy (EDX) analysis of the region beneath the Co/Cu interface in Figure 2b reveals Co has diffused into the bulk of the Cu substrate, providing additional evidence of intermixing (Supporting Information). Cu is observed to segregate to the surface of the growing Co film in order to minimize the free surface energy.18 The surfactant-like behavior that maintains Cu concentration at the uppermost layer of the Co film as it deposits suggests that Cu may stabilize the first several monolayers of Co leading to a wetting film.3 Venables previously described a phenomenon by which higher energy 1093

dx.doi.org/10.1021/jz500281k | J. Phys. Chem. Lett. 2014, 5, 1091−1095

The Journal of Physical Chemistry Letters

Letter

indicate that Cu migrates both toward the surface and Co/Cu interface following the minimization of the Co crystal lattice energy. Finally, highly mobile Cu and Co at the film interface leads to significant intermixing of the species during Co deposition even at 265 °C as well as grain boundary diffusion of Cu through the Co film. These phenomena suggest that pure Co EM barrier layers deposited by ALD from CoAMD will likely suffer greater resistance-capacitive losses than their electrolessly deposited CoWP counterparts due to the large degree of intermixing at the interface. Incorporating P or W or both in the ALD process could significantly reduce Cu migration through the film by removing the accelerated diffusion pathway of Co grain boundaries. Whether W or P would help mitigate atomic mixing at the Co/Cu interface is an important question for the viability of 0.5 nm ALD Co EM barriers from CoAMD.

Figure 4. The growth sequence of Co films on Cu by ALD from CoAMD.

overlayers burrow into lower energy surfaces and become coated in a thin layer of the lower energy substrate.19 This occurs especially during the deposition of magnetic materials on noble metal substrates.8,19 Moreover, Li and Tonner have demonstrated that annealing Co face-centered cubic (fcc) films grown epitaxially on Cu (001) to 400 °C induces the migration of a thin layer of Cu to the Co free surface.20 They suggested that Cu atoms undergo an inversion process by which Cu substrate atoms segregated to the surface of the Co film to act as a stabilizing layer for the mismatched Co film. Once the Co-rich layer is thick enough, 3D faceting of the Co begins. The apparent transition from 2D to 3D growth may reflect preferential deposition at grains of a specific crystallographic orientation or reconciliation of heteroepitaxial strain between the Co, for which bulk films prepared below 420−450 °C are generally hexagonal close packed (hcp), and Cu for which an fcc structure is more stable.6,10,21,22 The crystalline structure of the Co is likely a mixture of fcc and hcp Co. X-ray diffraction studies revealed strong fcc (111) signals for blank Cu wafers and 4.5 nm Co/Cu film samples (Supporting Information). However, distinct hcp signals were indiscernible from the signal noise if present, and the Co contribution to the fcc signal is inseparable from the Cu contribution due to the magnitude of the signal. A thicker (∼25 nm) Co film was deposited by molecular beam epitaxy but revealed only a very weak fcc (200) signal in addition to those observed on blank Cu. Efforts by Deo et al. suggest that even very thick Co films (∼ 200 nm) exhibit small XRD signals and that data from Co films less than 40 nm thick are obscured by noise.23 It is, therefore, reasonable to expect that hcp signals from the 4.5 and 25 nm films are too small to observe. Nevertheless, evidence of pseudoepitaxial Co/Cu layers in superlattices formed by fcc/hcp stacking faults fit well with our Co film behavior.6,7 In summary, Co deposition by ALD from CoAMD proceeds first by a metastable 2D Co-rich layer that grows on the Cu substrate until enough Co is deposited that the film begins to roughen and grow three dimensionally. The Co-rich layer remains while Co continues to accumulate as indicated by ToFSIMS depth profiling. Further, a surfactant-like ∼2 nm Cu layer is apparent on the surface of the Co up to at least 16 nm film thicknesses, suggesting that the Cu plays a role in the Co deposition and may stabilize the Co film especially at the early stages of film deposition. The Cu surface layer may improve Co lattice matching as pseudoepitaxial fcc and hcp Co grains can be expected on the Cu surface. Additionally, TOF-SIMS data



EXPERIMENTAL METHODS



ASSOCIATED CONTENT

CoAMD was acquired from Dow Chemical and used without further modification. Cu substrates (300 nm physical vapor deposited, polycrystalline Cu, predominantly (111), on TaN on SiO2) were provided by Intel. Co films were deposited by ALD in an in-house constructed vacuum system with in situ XPS. The vacuum system comprises a load lock and an ALD chamber with a base pressure of ∼1 × 10−7 Torr and in situ transfer to a PHI model 1600 XPS for chemical analysis (base pressure ∼1 × 10−9 Torr). System gases include Ar and H2 (99.999%, Matheson) each used with 50 sccm flow rate. Ar is used as the precursor carrier gas and inert sweep. Prior to deposition, the Cu substrates were cleaned by rinsing in order with acetone, ethanol, and deionized water (18 MΩ) and then rinsed in a 35 °C bath of glacial acetic acid (99.9%, Fisher Scientific) for 1 min to remove surface oxide. The wafers were blown dry with pressurized Ar and loaded immediately into the vacuum system load lock. A typical ALD cycle comprises a 2 s CoAMD exposure and 15 s H2 exposure separated by 15 s purges with Ar. Films are deposited at 265 °C, a temperature consistent with technical data provided by the precursor supplier. The precursor temperature was maintained at 80 °C during growth. ToF-SIMS data were collected using a TOF.SIMS 5 instrument (ION-TOF GmbH, 2010). For depth profiling of the Co films and Co/Cu interfaces, a short-pulsed (18 ns) primary ion beam (Bi1+, 30 keV energy, ∼3.1 pA measured sample current) was typically raster-scanned over a 100 × 100 μm2 area centered within a 250 × 250 μm2 regressing area that was previously sputtered by a secondary ion beam (Cs+, 500 eV energy, ∼53 nA measured sample current). The depth profiles were acquired at a base pressure of about 7.5 × 10−10 Torr in noninterlaced mode (i.e., sequential data acquisition and sputtering) and with the primary ion beam set in high current bunched mode. All detected secondary ions had negative polarity and mass resolution >8000 (m/δm). A detailed explanation of the ToF-SIMS data acquisition and the methods used to convert sputter time to depth and estimate the Co/Cu atomic mixing length at the interface are presented in the Supporting Information.

S Supporting Information *

Explanation of ToF-SIMS methods and MRI model, asdeposited XPS data, EDX, and XRD results. This material is available free of charge via the Internet at http://pubs.acs.org/. 1094

dx.doi.org/10.1021/jz500281k | J. Phys. Chem. Lett. 2014, 5, 1091−1095

The Journal of Physical Chemistry Letters



Letter

(17) Hofmann, S. Profile Reconstruction in Sputter Depth Profiling. Thin Solid Films 2001, 398−399, 336−342. (18) Vitos, L.; Ruban, A. V; Skriver, H. L.; Kolla, J. The Surface Energy of Metals. Surf. Sci. 1998, 411, 186−202. (19) Venables, J. A. Introduction to Surface and Thin Film Processes; Cambridge University Press: New York, 2000. (20) Li, H.; Tonner, B. P. Structure and Growth Mode of Metastable FCC Cobalt Ultrathin Films on Cu (001) as Determined by AngleResolved X-Ray Photoemission Scattering. Surf. Sci. 1990, 237, 141− 152. (21) Nilsen, O.; Karlsen, O. B.; Kjekshus, A.; Fjellvåg, H. Simulation of Growth Dynamics for Nearly Epitaxial Films. J. Cryst. Growth 2007, 308, 366−375. (22) Nilsen, O.; Karlsen, O. B.; Kjekshus, A.; Fjellvåg, H. Simulation of Growth Dynamics in Atomic Layer Deposition. Part I. Amorphous Films. Thin Solid Films 2007, 515, 4527−4537. (23) Deo, N.; Bain, M. F.; Montgomery, J. H.; Gamble, H. S. Study of Magnetic Properties of Thin Cobalt Films Deposited by Chemical Vapour Deposition. J. Mater. Sci. Mater. Electron. 2005, 16, 387−392.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was sponsored by Intel Corporation. Additionally, we acknowledge the National Science Foundation grant DMR0923096 used to purchase the TOF-SIMS instrument of the Texas Materials Institute and the Welch Foundation for support of facilities used in this work.



REFERENCES

(1) Moon, P.; Dubin, V.; Johnston, S.; Leu, J.; Raol, K.; Wu, C. Process Roadmap and Challenges for Metal Barriers. IEEE Int. Electron Devices Meet. 2003, 841−844. (2) Gordon, R. G.; Kim, H.; Bhandari, H. Cobalt Nitride Layers for Copper Interconnects and Methods for Forming Them. US Patent 7,973,789 B2, 2011. (3) Kohn, A.; Eizenberg, M.; Shacham-Diamand, Y. Copper Grain Boundary Diffusion in Electroless Deposited Cobalt Based Films and Its Influence on Diffusion Barrier Integrity for Copper Metallization. J. Appl. Phys. 2003, 94, 3015−3024. (4) Hau-Riege, C. S. An Introduction to Cu Electromigration. Microelectron. Reliab. 2004, 44, 195−205. (5) International Technology Roadmap for Semiconductors, 2011 ed.; Sematech International. (6) Tonner, B. P.; Han, Z.-L.; Zhang, J. Structure of Co Films Grown on Cu(111) Studied by Photoelectron Diffraction. Phys. Rev. B 1993, 47, 9723−9732. (7) Lamelas, F. J.; Lee, C. H.; He, H.; Vavra, W.; Clarke, R. Coherent Fcc Stacking in Epitaxial Co/Cu Superlattices. Phys. Rev. B 1989, 40, 5837−5840. (8) Christensen, A.; Ruban, A. V; Stoltze, P.; Jacobsen, K. W.; Skriver, H. L.; No, J. K.; Besenbacher, F. Phase Diagrams for Surface Alloys. Phys. Rev. B 1997, 56, 5822−5834. (9) Zhang, L.; Kraatz, M.; Aubel, O.; Hennesthal, C.; Im, J.; Zschech, E.; Ho, P. S. Cap Layer and Grain Size Effects on Electromigration Reliability in Cu/Low-k Interconnects. IEEE Int. Interconnect Technol. Conf. 2010, 1−3. (10) Shimizu, H.; Sakoda, K.; Momose, T.; Shimogaki, Y. Atomic Layer Deposited Co(W) Film as a Single-Layered Barrier/Liner for Next-Generation Cu-Interconnects. Jpn. J. Appl. Phys. 2012, 51, 1−7. (11) Yang, C.-C.; Flaitz, P.; Wang, P.-C.; Chen, F.; Edelstein, D. Characterization of Selectively Deposited Cobalt Capping Layers: Selectivity and Electromigration Resistance. IEEE Electron Device Lett. 2010, 31, 728−730. (12) Chioncel, M. F.; Haycock, P. W. Structural Characterization of Cobalt Thin Films Grown by Metal-Organic CVD. Chem. Vap. Deposition 2005, 11, 235−243. (13) Lim, B. S.; Rahtu, A.; Gordon, R. G. Atomic Layer Deposition of Transition Metals. Nat. Mater. 2003, 2, 749−754. (14) Bhandari, H. B.; Yang, J.; Kim, H.; Lin, Y.; Gordon, R. G.; Wang, Q. M.; Lehn, J.-S. M.; Li, H.; Shenai, D. Chemical Vapor Deposition of Cobalt Nitride and its Application as an AdhesionEnhancing Layer for Advanced Copper Interconnects. ECS J. Solid State Sci. Technol. 2012, 1, N79−N84. (15) Zimmerman, J. D.; Lassiter, B. E.; Xiao, X.; Sun, K.; Dolocan, A.; Gearba, R.; Vanden Bout, D. a; Stevenson, K. J.; Wickramasinghe, P.; Thompson, M. E.; et al. Control of Interface Order by Inverse QuasiEpitaxial Growth of Squaraine/Fullerene Thin Film Photovoltaics. ACS Nano 2013, 7, 9268−9275. (16) Berglund, S. P.; He, H.; Chemelewski, W. D.; Celio, H.; Dolocan, A.; Mullins, C. B. p-Si/W2C and p-Si/W2C/Pt Photocathodes for the Hydrogen Evolution Reaction. J. Am. Chem. Soc. 2014, 136, 1535−1544. 1095

dx.doi.org/10.1021/jz500281k | J. Phys. Chem. Lett. 2014, 5, 1091−1095