Background

Nanotwinned copper (nt-Cu) has been regarded as a promising candidate for interconnection materials in integrated circuit (IC) since it possesses various superior properties such as high strength, low resistivity, high electromigration resistance and thermal stability. The performance, especially the mechanical properties of nt-Cu is strongly related to its twin density. Electrodeposition is so far the most efficient technique to fabricate nt-Cu interconnects, and the relationship between electrodeposition condition and twin density has been extensively studied. However, these investigations were mostly empirical, and the formation mechanism of copper nanotwins remains unclear. Although researchers have put forward some possible mechanisms, including strain energy theory and Winand theory, they still lack direct evidence. To efficiently improve the microstructural tunability of nt-Cu during DC electrodeposition, a clear understanding of the formation of nanotwins is of the essence.

Introduction

Researchers in school of materials science and engineering, Shanghai Jiao Tong University, established a two-dimensional nucleation and growth model of Cu nanotwins in DC electrodeposition. They evaluated the critical nucleation size and average twin spacing theoretically and unveiled a high probability of twin nucleation through random stacking with an average twin spacing derived to be ~1 nm. Therefore, nanotwins may form by random stacking without extra driving force. The gap between theoretical and observed twinning probability can be ascribed to the surface migration-induced dissipation of nuclei. By molecular dynamics simulation, the formation and subsequent dissipation of nuclei during deposition of Cu was confirmed. Two possible modes of nuclei dissipation, by migration of the whole nucleus and by domain boundary movement, were observed in the simulated system. Simulation of Cu deposition under different deposition rates and different temperatures well coincided with the experimental results. By providing a clear illustration of nanotwin formation process, this study contributes to the understanding of nt-Cu electrodeposition. 

Paper link: https://doi.org/10.1016/j.actamat.2023.119468

Research content

Figure 1. (a-d) Schematic illustration of periodic nucleation and growth of lamellae on (111) plane during electrodeposition of Cu; (e-f) surface morphology of an electrodeposited nt-Cu film showing terrace-like structure.

Table 1. MD Calculation results of total boundary energy difference (ΔG_t^b - ΔG₀^b), nucleation rate ratio (I₀ / I_t), and average twin spacing (t) for nuclei with different numbers of atoms (N_atom). Data for both unrelaxed and relaxed systems is presented.

Figure 2. Results of the MD simulation involving Cu deposition and surface atom migration. (a)The substrate used for simulation; (b)The atomic system after deposition of atoms as many as 12 monolayers. Common neighbor analysis was conducted with twin boundaries colored in black; (c1-c3) Motion of an adatom to be absorbed by the nucleus below; (d1-d6) The dissipation of a twinned domain by domain boundary movement, which is sketched by the red dotted line. (d)The interpretation of different colors in (c) and (d), which aims to manifest the difference between perfect and twinned domains.

Figure 3. (a-c) Top-view of systems after deposition of two monolayers of Cu atoms with deposition rates of 0.25 ps/atom, 5 ps/atom, and 50 ps/atom. Common neighbor analysis was conducted to reveal the twinned stacking with black color. (d) Content of hcp-structured atoms in the deposited atoms under different deposition rate, exhibiting the amount of twinned stacking. (e-g) Side-view of systems after deposition of 12 monolayers of Cu atoms at temperatures of 298K, 498K and 698K. (h) Content of hcp-structured atoms in the deposited atoms at different temperatures.

Conclusions

In summary, a two-dimensional nucleation and growth model for nt-Cu during DC electrodeposition was established. The calculated average twin spacing is much smaller than the experimental value, indicating a relatively high probability of twinned nucleation through random stacking. Dissipation of nuclei in the wake of frequent hopping of adatoms is a possible explanation for the gap between theoretical and experimental twinning probabilities. Such surface migration-induced dissipation of nuclei was confirmed by MD simulation, which takes place by migration of the entire nucleus or by domain boundary movement. This study gives insights into the formation mechanism of nt-Cu in DC electrodeposition and contributes to the advancement of nt-Cu processing.

Corresponding authors

Yunwen Wu: associate professor, institute of electronic materials and technology, school of materials science and engineering, Shanghai Jiao Tong University.

Tao Hang: professor, institute of electronic materials and technology, school of materials science and engineering, Shanghai Jiao Tong University.