Impact of crystalline orientation on Cu–Cu solid-state bonding behavior by molecular dynamics simulations

概要

Title

Impact of crystalline orientation on Cu–Cu
solid-state bonding behavior by molecular
dynamics simulations

Author(s)

Tatsumi, Hiroaki; Kao, C. R.; Nishikawa, Hiroshi

Citation

Scientific Reports. 2023, 13(1), p. 23030

Version Type VoR
URL
rights

https://hdl.handle.net/11094/93537
This article is licensed under a Creative
Commons Attribution 4.0 International License.

Note

Osaka University Knowledge Archive : OUKA
https://ir.library.osaka-u.ac.jp/
Osaka University

www.nature.com/scientificreports

OPEN

Impact of crystalline orientation
on Cu–Cu solid‑state bonding
behavior by molecular dynamics
simulations
Hiroaki Tatsumi 1*, C. R. Kao 2 & Hiroshi Nishikawa 1
High-density electronics are hindered by the constraints of Sn-based solder joints, necessitating the
exploration of Cu–Cu solid-state bonding. However, current bonding methods are expensive and
time-consuming; therefore, understanding the Cu–Cu bonding mechanism is crucial for optimization.
This study utilizes molecular dynamics (MD) simulation to elucidate the Cu–Cu solid-state bonding
behavior, focusing on interfacial densification and diffusion phenomena. Furthermore, it highlights
the influence of crystal orientation on the interfacial bonding behavior. To analyze the impact
of crystal orientation, monocrystalline Cu slabs with a simplified periodic surface structure were
employed to replicate surface roughness and subsequently bonded at a specific temperature. The
results indicate the critical influence of crystalline orientations on the bonding process: identical
orientations result in slower densification at the interface, whereas misoriented orientations
significantly accelerate it. This effect, attributed to the grain boundary (GB) structures formed
owing to misorientation, suggests a central role for GB diffusion in bonding progression. Diffusion
coefficients calculated using the mean square displacement (MSD) confirmed these findings and
exhibited significantly larger values for misoriented joints. Additionally, the simulations reveal an
activation energy for GB diffusion that is lower than conventional values, highlighting the impact of
the crystallographic orientation and voids at the bonding interface. Our research elucidates the role of
crystalline orientation in diffusion phenomena at bonding interfaces, offering valuable implications for
optimizing bonding-based manufacturing processes.
In recent years, the rapid proliferation of highly integrated electronics and increased electrical current density
has generated a growing demand for improved bonding techniques. Sn-based solder joints commonly utilized
for electronic device bonding are rapidly approaching their theoretical limitations, particularly in the latest
three-dimensional integrated circuits (3D-ICs) where the input/output pitch is expected to decrease to 1 μm1.
Conventional Sn-based solder bumps with a minimum pitch of 20 μm2 are prone to electrical short-circuit
failures owing to intermicrobump contact during bonding. As the size of solder bumps decreases, joint characteristics are increasingly affected by the formation of Kirkendall voids and brittle intermetallic compounds
(IMCs) at the i­nterface3. Moreover, the reliability of solder joints is limited by the achievable current density
resulting from ­electromigration4. An alternative approach is transient liquid phase (TLP) bonding, commonly
referred to as solid–liquid interdiffusion (SLID) ­bonding5. High-melting-point materials such as Cu and lowmelting-point materials such as Sn are supplied as layers that solidify isothermally during the bonding process,
enabling a higher melting-point joint consisting of IMCs. Some studies have successfully fabricated Cu–Sn IMC
joints using the TLP bonding technique with Cu/Sn microbumps and demonstrated its versatility for 3D-ICs and
micro-electro-mechanical system (MEMS) p
­ ackaging6,7. However, TLP-bonded joints still face issues, including
not only low electrical and thermal conductivity but also poor mechanical properties derived from the IMCs.
Owing to their high electrical conductivity, high electromigration resistivity, excellent mechanical properties,
and the absence of concerns regarding Kirkendall voiding, Cu–Cu solid-state bonding technologies have attracted
considerable attention to accommodate the demands of high-end 3D-ICs5,8–11. This method can be employed
using room-temperature ­bonding12 or a combination of room-temperature bonding and post-annealing8,9.

1

Joining and Welding Research Institute, Osaka University, 11‑1 Mihogaoka, Ibaraki, Osaka  567‑0047,
Japan. 2Department of Materials Science and Engineering, National Taiwan University, No. 1, Sec. 4, Roosevelt
Road, Taipei 10617, Taiwan. *email: tatsumi.jwri@osaka-u.ac.jp

Scientific Reports |

(2023) 13:23030

| https://doi.org/10.1038/s41598-023-50427-3

1
Vol.:(0123456789)

www.nature.com/scientificreports/
Achieving activated atomically flat surfaces is crucial for optimal bonding contact and can be obtained using
chemical mechanical polishing (CMP) and complex surface activation processes. However, this approach is
expensive and time-consuming. An alternative method is the thermal compression p
­ rocess10,11, typically involving a bonding time of ~ 1 h at temperatures of 300–400 °C, with high contact pressure a­ pplied5. However, these
low-throughput processes pose limitations for industrial applications.
To improve the bondability at lower temperatures and for shorter times, understanding the bonding mechanism of Cu solid-state bonding is essential. First, the removal of impurities and oxides from the bonding surfaces
is crucial. Several surface-modification techniques have been proposed, including high-vacuum ­plasma13, reacting ­gases14, passivation c­ oatings15, electromagnetic i­ rradiation16, and Ag thin layer d
­ eposition17. After successfully removing oxides and impurities from the bonding surfaces, the key step is to eliminate gaps at the bonded
interface via deformation and diffusion. The process of traditional solid-state bonding, known as diffusion
bonding, has been theoretically explained in previous l­ iterature18,19. The significance of interfacial plastic deformation, creep deformation, and surface/grain boundary (GB) diffusion in the formation of neck surfaces has
been highlighted. Higher bonding temperatures, pressures, and times are known to improve bondability. For
semiconductor applications requiring bonding under relatively low-temperature and low-pressure conditions,
diffusion phenomena have a significant ­impact20,21. One dominant mechanism driving Cu–Cu solid-state bonding is GB diffusion. Rebhan and ­Hinger22 systematically evaluated the effects of possible dominant parameters
on bondability by comparing physical vapor deposited (PVD) and electrochemically deposited (ECD) Cu thin
films on Si wafers. The study demonstrated that PVD Cu can be bonded at a lower temperature compared to
ECD Cu, suggesting the improved bondability may be attributed to significant GB diffusion on the smaller
grain size and an increased concentration of random high-angle GBs. Another possible mechanism is surface
diffusion. Liu et al.23,24 reported that the utilization of ECD highly (111)-oriented Cu polycrystalline thin films,
known as nanotwin Cu, improved bondability. They suggested that the improved bondability can be primarily
attributed to surface diffusion, as the surface diffusion coefficient of Cu on the (111) plane is three to four orders
of magnitude larger than that on other surfaces. Additionally, Shie et al.21 attempted to provide a comprehensive
bonding mechanism considering both surface and GB diffusion. They assumed a bonding behavior consisting
of initial contact formation, GB formation on the contacted surfaces, migration of GBs at the bonding interface,
void ripening, and grain growth eliminating the bonding interface. The effect of crystallographic orientation
on surface diffusion has been extensively discussed; however, the impact of crystallographic orientation on GB
diffusion and its contribution to bondability remain unclear.
Molecular dynamics (MD) simulation, an atomistic-scale simulation method, can be a powerful tool for
understanding the bonding behavior and diffusion phenomena at the atomic level. Li et al.25 investigated the
diffusion phenomenon at the Al–Cu interface using MD simulations and discussed the dominant diffusion
mechanism from the perspective of activation energy. Xydou et al.26 investigated the void-closing behavior of
Cu–Cu GBs and highlighted the importance of GB diffusion. Long et al.27 reported the mechanism of microweld formation and the breakage of ultrasonically bonded wires using MD simulations. Thus, MD simulations
offer valuable insights into atomic migration behavior. Furthermore, our previous i­ nvestigation28 explored the
morphological evolution of the solid-state bonding interface, specifically the Au–Cu interface, by examining
atomic diffusion. Through atomic displacement analyses and diffusion coefficient estimations, this examination
successfully probed the behavior of interfacial atom penetration into interstices, which was described as interfacial densification behavior. Therefore, our previously established MD simulation and analysis techniques can
provide novel perspectives on the bonding behavior of Cu–Cu solid-state bonding, particularly the influence of
surface diffusion and GB diffusion on the bonding mechanism.
In this study, MD simulations were employed to elucidate the Cu–Cu solid-state bonding behavior, focusing
on interfacial densification and diffusion phenomena. This study aimed to investigate how the crystal orientation of the bonding surfaces affects the bonding behavior. A Cu slab was constructed with a simplified periodic
surface structure to simulate surface roughness. The Cu slabs were subsequently bonded at specific temperatures
and pressures. ...

この論文で使われている画像