An official website of the United States government
Ru(0001) and Co(0001) films with thickness d ranging from 5 to 300 nm are sputter deposited onto Al2O3(0001) substrates in order to quantify and compare the resistivity size effect. Both metals form epitaxial single crystal layers with their basal planes parallel to the substrate surface and exhibit a root-mean-square roughness < 0.4 nm for Ru and < 0.9 nm for Co. Transport measurements on these layers have negligible resistance contributions from roughness and grain boundary scattering which allows direct quantification of electron surface scattering. The measured resistivity ρ vs d is well described by the classical Fuchs-Sondheimer model, indicating a mean free path for transport within the basal plane of λ = 6.7 ± 0.3 nm for Ru and λ = 19.5 ± 1.0 nm for Co. Bulk Ru is 36% more resistive than Co; in contrast, Ru(0001) layers with d ≤ 25 nm are more conductive than Co(0001) layers, which is attributed to the shorter λ for Ru. The determined λ-values are utilized in combination with the Fuchs-Sondheimer and Mayadas-Shatzkes models to predict and compare the resistance of polycrystalline interconnect lines, assuming a grain boundary reflection coefficient R = 0.4 and accounting for the thinner barrier/adhesion layers available to Ru and Co metallizations. This results in predicted 10 nm half-pitch line resistances for Ru, Co, and Cu of 1.0, 2.2, and 2.1 kΩ/µm, respectively.
more »
« less
First-principles prediction of electron grain boundary scattering in fcc metals
The electron reflection probability r at symmetric twin boundaries Σ3, Σ5, Σ9, and Σ11 is predicted from first principles for the eight most conductive face-centered cubic (fcc) metals. r increases with decreasing interplanar distance of atomic planes parallel to the boundary. This provides the basis for an extrapolation scheme to estimate the reflection probability r r at random grain boundaries, which is relatively small, r r = 0.28–0.39, for Cu, Ag, and Au due to their nearly spherical Fermi surfaces, but approximately two times higher for Al, Ca, Ni, Rh, and Ir with a predicted r r = 0.61–0.72. The metal resistivity in the limit of small randomly oriented grains with fixed average size is expected to be proportional to the materials benchmark quantity ρ o λ × r r /(1 − r r ), where ρ o and λ are the bulk resistivity and bulk electron mean free path, respectively. Cu has the lowest value for this quantity, indicating that all other fcc metals have a higher resistivity in the limit of small randomly oriented grains. Thus, the conductivity benefit of replacement metals for narrow Cu interconnect lines can only be realized if the grains are larger than the linewidth or exhibit symmetric orientation relationships where r < r r .
more »
« less
- Award ID(s):
- 1712752
- PAR ID:
- 10349176
- Date Published:
- Journal Name:
- Applied Physics Letters
- Volume:
- 120
- Issue:
- 24
- ISSN:
- 0003-6951
- Page Range / eLocation ID:
- 241603
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Epitaxial W–Mo multilayers are employed as a model system to demonstrate how resistivity measurements parallel to metal–metal interfaces can be used to quantify the specific interface resistance without sub-10-nm patterning that would be required for direct transport measurements across the interface. 50-nm-thick epitaxial multilayer stacks containing 2–60 individual W(001) and Mo(001) layers are deposited on MgO(001) substrates and their resistivity ρ measured as a function of superlattice period Λ at 293 and 77 K. The measured room temperature ρ increases from 7.10 to 8.62 μΩ cm with decreasing Λ = 50–1.7 nm, which is attributed to the increasing electron-interface scattering. The semiclassical Fuchs–Sondheimer model for surface scattering dramatically overestimates the resistivity, which is attributed to coherent electron transmission across multiple interfaces. A new Boltzmann transport model treats each interface as a boundary condition where electrons either scatter diffusely or traverse without momentum loss with a probability T for the first encountered interface and with 100% transmission at subsequent interfaces until they are relaxed by a bulk scattering event. This model has a single unknown parameter T, which becomes the fitting parameter for experimental data analysis, yielding a temperature-independent T = 0.8 ± 0.1 and a corresponding contact resistance at the W(001)–Mo(001) interface of 2.6 × 10−16 Ω m2.more » « less
-
Abstract Owing to its superlative carrier mobility and atomic thinness, graphene exhibits great promise for interconnects in future nanoelectronic integrated circuits. Chemical vapor deposition (CVD), the most popular method for wafer-scale growth of graphene, produces monolayers that are polycrystalline, where misoriented grains are separated by extended grain boundaries (GBs). Theoretical models of GB resistivity focused on small sections of an extended GB, assuming it to be a straight line, and predicted a strong dependence of resistivity on misorientation angle. In contrast, measurements produced values in a much narrower range and without a pronounced angle dependence. Here we study electron transport across rough GBs, which are composed of short straight segments connected together into an extended GB. We found that, due to the zig-zag nature of rough GBs, there always exist a few segments that divide the crystallographic angle between two grains symmetrically and provide a highly conductive path for the current to flow across the GBs. The presence of highly conductive segments produces resistivity between 10 2 to 10 4 Ω μ m regardless of misorientation angle. An extended GB with large roughness and small correlation length has small resistivity on the order of 10 3 Ω μ m, even for highly mismatched asymmetric GBs. The effective slope of the GB, given by the ratio of roughness and lateral correlation length, is an effective universal quantifier for GB resistivity. Our results demonstrate that the probability of finding conductive segments diminishes in short GBs, which could cause a large variation in the resistivity of narrow ribbons etched from polycrystalline graphene. We also uncover spreading resistance due to the current bending in the grains to flow through the conductive segments of the GB and show that it scales linearly with the grain resistance. Our results will be crucial for designing graphene-based interconnects for future integrated circuits.more » « less
-
We present an ab initio evaluation of electron scattering mechanisms in Al interconnects from a back-end-of-line (BEOL) perspective. We consider the ballistic conductance as a function of nanowire size, as well as the impact of surface oxidation on electron transport. We also consider several representative twin grain boundaries and calculate the specific resistivity and reflection coefficients for each case. Lastly, we calculate the vertical resistance across the Al/Ta(N)/Al and Cu/Ta(N)/Cu interfaces, which are representative of typical vertical interconnect structures with diffusion barriers. Despite a high ballistic conductance, the calculated specific resistivities at grain boundaries are 70-100% higher in Al than in Cu, and the vertical resistance across Ta(N) diffusion barriers are 60-100% larger for Al than for Cu. These results suggest that in addition to the well-known electromigration limitations in Al interconnects, electron scattering represents a major problem in achieving low interconnect line resistance at fine dimensions.more » « less
-
Abstract Community detection is considered for a stochastic block model graph of n vertices, with K vertices in the planted community, edge probability p for pairs of vertices both in the community, and edge probability q for other pairs of vertices. The main focus of the paper is on weak recovery of the community based on the graph G , with o ( K ) misclassified vertices on average, in the sublinear regime n 1- o (1) ≤ K ≤ o ( n ). A critical parameter is the effective signal-to-noise ratio λ = K 2 ( p - q ) 2 / (( n - K ) q ), with λ = 1 corresponding to the Kesten–Stigum threshold. We show that a belief propagation (BP) algorithm achieves weak recovery if λ > 1 / e, beyond the Kesten–Stigum threshold by a factor of 1 / e. The BP algorithm only needs to run for log * n + O (1) iterations, with the total time complexity O (| E |log * n ), where log * n is the iterated logarithm of n . Conversely, if λ ≤ 1 / e, no local algorithm can asymptotically outperform trivial random guessing. Furthermore, a linear message-passing algorithm that corresponds to applying a power iteration to the nonbacktracking matrix of the graph is shown to attain weak recovery if and only if λ > 1. In addition, the BP algorithm can be combined with a linear-time voting procedure to achieve the information limit of exact recovery (correctly classify all vertices with high probability) for all K ≥ ( n / log n ) (ρ BP + o (1)), where ρ BP is a function of p / q .more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Journal Article:
- https://doi.org/10.1063/5.0098822
