(First Best Paper Award) Vacuum Quality Impact on Covalent Bonding

Abstract

As already described in the literature [1, 2, 3, 4], covalent bonding is based on direct bonding process. The first step of this process consists in creating dangling bonds at the wafer surface using an Ar+ ion bombardment. Then, the two activated surfaces are brought in contact, resulting in covalent bonds between them. Using silicon surfaces, no further annealing is required to enhance the bonding strength, as opposed to direct hydrophilic bonding. The adherence energy (Gc=2γc) of such Si/Si bond pairs reaches the silicon fracture energy (5J/m²). Indeed, all bonded samples break during double cantilever beam (DCB) measurements of it.Dangling bonds are highly reactive. Queue time between activation and bonding is crucial, as dangling bonds must be preserved. One way, commonly used, is to process activated materials while always staying under Ultra High Vacuum (UHV). Indeed, reactive dangling bonds should not get in contact with others reactive molecules, like H2O, O2 or N2, for instance. However, some molecules always remain, for any given vacuum value. Depending on their nature, the UHV quality for bonding purpose is affected. This paper aims at studying the impact of the UHV quality on the dangling bonds stability and bonding quality.In order to study the UHV quality, a mass spectrometer has been installed inside the bonding chamber. It has been shown that among all the reactive potential molecules, water is the most harmful one. Indeed, its removal enhances the vacuum quality. Baking vacuum systems is a well-known technic to remove adsorbed water molecules from metallic surfaces [5]. Although background water concentration in UHV is hardly detectable at room temperature, tracking its partial pressure during bakeout cycles enables to define a water threshold. Indeed, during a bakeout cycle, a water concentration peak is clearly seen by the mass spectrometer, as shown in figure 1. The impact of several system bakeouts at 100°C was also evaluated. After each bakeout, a Si/Si bonding energy was measured. After the first bakeout, the adherence energy was only around 2.5J/m². However, after the second bakeout, it reached the silicon fracture energy of 5J/m². More bakeout cycles were also tested. Adherence energy results will be discussed as well as partial pressures of other reactive molecules.The impact of queue times between wafer surfaces activations and bonding was also studied. For a standard process, this waiting time represents the handling time to activate both substrates (in a system having only one activation chamber), and to move them from the activation chamber to the bond chamber. We showed that the impact of queue time depended on the number of bakeout cycles and that the adherence energy decreasing rate was strongly affected by the UHV quality. With our best tool conditioning, an adherence energy higher than 3,4J/m² can be maintained up to 5mn added to the minimum queue time.Thanks to our EVG®Combond® system, covalent bonding in temperature can be performed using heated electrostatic chucks (ESC)[6]. However, even after two or more system bakeouts, 200°C covalent bonding failed to exhibit high adherence energy and it was not possible to reach the silicon fracture energy. Tracking the pressure level inside the bonding chamber during a high temperature bakeout of the ESC (>300°C) showed that water still desorbed during this process. Optimized tool preconditioning thus have to be implemented for such specific hot bonding, with a need for global baking (even of internal chamber pieces).In conclusion, we will describe the experimental setup and the impact of tool preconditioning (i.e. bakeout sequences) on manufacturing. Mass spectrometers and bond strength measurements will be presented, as well as curves of bond strength against the waiting time between the activation and the bonding, for bond pairs processed at room temperature or higher (200°C). Suga T, Takahashi Y. STRUCTURE OF A1-A1 A N D A1-Si3N4 INTERFACES BONDED AT ROOM TEMPERATURE BY MEANS OF THE SURFACE ACTIVATION METHOD. :5.Takagi H, Kikuchi K, Maeda R, Chung TR, Suga T. Surface activated bonding of silicon wafers at room temperature. Appl Phys Lett. 15 avr 1996;68(16):2222-4.Taniyama S, Wang YH, Fujino M, Suga T. Room temperature wafer bonding using surface activated bonding method. In: 2008 IEEE 9th VLSI Packaging Workshop of Japan. 2008. p. 141-4.Flötgen C, Razek N, Dragoi V, Wimplinger M. Novel Surface Preparation Methods for Covalent and Conductive Bonded Interfaces Fabrication. ECS Trans. 14 août 2014;64(5):103.Berman A. Water vapor in vacuum systems. Vacuum. avr 1996;47(4):327-32.Lomonaco Q, et al. Stress Engineering in Germanium-Silicon Heterostructure Using Surface Activated Hot Bonding. ECS Trans. 30 sept 2022;109(4):277-87. Figure 1