AVS 71 Session PS1-TuA: Plasmas in Advanced Packaging 

Die-to-wafer (D2W) and wafer-to-wafer (W2W) hybrid and fusion bonding are integral to advanced packaging applications. Prior to bonding, for both D2W and W2W approaches, surface preparation is performed to facilitate the bonding process. Surface preparation consists of plasma activation and wet cleaning and hydration process steps. These steps are critical to obtain good interface quality and in turn high yield bonding that is void-free with high bond strength. Plasma surface activation is typically a relatively short, low power process. Nonetheless, the plasma source hardware and process conditions must be optimized to provide sufficient activation without roughening the dielectric layer (fusion and hybrid bonding) or sputtering and/or heavily oxidizing the bond pad Cu (hybrid bonding).

D2W bonding is required for chiplet heterogenous integration, and also offers the potential for yield improvement by the use of known good die for high bandwidth memory (HBM) and CMOS image sensor (CIS) applications. For D2W bonding, singulated die are bonded directly onto the target wafers. However, the quality of the die singulation process directly impacts the bonding yield. Defectivity levels for traditional saw dicing are too high for high volume D2W manufacturing. As such, advanced singulation techniques, such as plasma dicing are an essential part of the D2W ecosystem. For plasma dicing, there are two approaches, referred to as dice before grind (DBG) and dice after grind (DAG). For DBG, the etching process is performed before wafer thinning. The etching process trenches or grooves the full thickness wafers, and the dies are then singulated during the backgrind thinning process. For DAG, the etching process directly singulates or dices the thinned wafer, landing on a carrier. Consequently, the DBG and DAG have different process requirements and integration challenges.

In this paper, an overview of the fundamental mechanisms, chamber hardware factors, key process parameters, and process integration considerations for surface activation and plasma dicing steps will be presented. In addition, implementation of surface activation plasma for onto bonding cluster tools will be discussed.

Bonding plays an important role in advanced microelectronics integration and packaging by bringing together components and devices fabricated separately.In addition to other important factors, surface-pretreatments on materials including metals, dielectrics and III-V’s have been consistently found to be crucial to achieving a high-quality bond, despite incomplete understanding of why certain treatments have greater success than others.Our project aims to improve understanding of some of these bonding pre-treatment effects to help enable more efficient development of bonding protocols, focusing on materials relevant for photonic and electronic devices.

Bonding of III-V semiconductors onto silicon enables fabrication of heterogeneous integrated photonic devices featuring low-loss waveguides and optical gain.[1]Al₂O₃ grown by atomic layer deposition (ALD) constitutes an effective intermediate layer between III-V’s and SiO₂, as Al₂O₃ films can achieve a higher interface energy even with low temperature annealing.[1,2]We have used x-ray photoelectron spectroscopy (XPS) to characterize GaAs surfaces prior to and post common ‘preps’ as well as after plasma or thermal ALD of sub-6 nm Al₂O₃ layers.Such thin films allow observation of both the 3d and 2p regions for Ga and As, giving insight into chemistry at different depths of the material, including the interface.Results provide insight into surface prep effects and support previous reports that thermal ALD removes native GaAs oxides.

Regarding electronic devices, plasma-pretreatments can promote better bonding of important contact metals such as In, which is used in e.g. cryo-cooled detectors.Preliminary results show He/H₂/N₂ atmospheric plasma treatments increase InO_x with little change to the metal signal and suggest that plasma-treated In displays a lower surface work function as compared to a control.We plan to extend our studies to include Cu, a critical material for hybrid bonding applications, and also to investigate plasma surface treatments via an in-situ remote plasma source on the XPS without atmospheric exposure.We anticipate that these results will add valuable insight into various bonding chemistries and their relation to bond strength or efficacy to help improve bonding process development in the future.

Acknowledgement: This work was performed with funding from the CHIPS Metrology Program, part of CHIPS for America, National Institute of Standards and Technology, U.S. Department of Commerce.

[1] H.K. Sahoo, L. Ottaviano, Y. Zheng, O. Hansen, K. Yvind, J. Vac. Sci. Technol. B 36(1) (2018) 011202-1 to 6.

[2] T. Gougousi, Prog. in Crystal Growth and Char. of Mat. 62 (2016) 1-21.