Paper SE-TuP3
Investigations on Physical Processes for Low Temperature Plasma Activated Wafer Bonding

Tuesday, November 1, 2011, 6:00 pm, Room East Exhibit Hall

Direct wafer bonding is a straightforward method of directly connecting wafers, with suitable (in terms of micro-roughness, flatness and cleanliness) surfaces, permanently to each other, by bringing them into contact and subsequently annealing them or simply storing them. The conventional process for hydrophilic oxidized Silicon surfaces (native as well as thermal oxide) is well understood, and explained the following way [1]:

Up to 100°C the substrate surfaces are held together via van der Waals interaction which is mediated by a few monolayers of water. In the range of 100-200°C the water diffuses away from the interface both along the interface and through the oxide into the crystalline bulk, where it reacts with the silicon and forms oxide. The increase of the bond strength from 50% to 100% of Si bulk strength is usually attributed to a closing of gaps at the interface, which starts at the softening temperature of the thermal oxide at around 900-1000°C, depending on whether dry or wet oxide was used.

Low temperature plasma activated direct wafer bonding is a process that lowers the required annealing temperatures necessary for reaching high bond strength. One example for such an improvement is a pair of native oxide – thermal oxide wafers, where bulk strength can be realized by plasma activation with subsequent annealing at temperatures below 200°C. At this temperature conventional wafer bonding reaches half of Si bulk strength, and is limited by gaps at the bonding interface. The mechanism behind this improvement compared to the non activated process is still under discussion.

To clarify the mechanism for this commercially available process, different bonding experiments were performed to evaluate the lifetime of the surface activation and the achievable bond strength when varying some of the boundary conditions of the process (substrates with different orientations, various plasmas, and lowering the annealing temperature).

By partly covering wafers during plasma activation, comparisons between the activated and non-activated regions could be made on single wafers. Therefore the influence of the slightly different substrates could be eliminated. Such wafers were then analyzed by atomic force microscopy, by spectroscopic ellipsometry, by Auger analysis and by X-ray photoelectron spectroscopy.

Finally a model for the mechanism, which was derived from the model for the conventional bonding process, and which explains the experimental results will be presented.

[1] Q.-Y. Tong, U. Gösele, Semiconductor Wafer Bonding: Science and Technology, Wiley, (1998)