System performance can be improved by stacking and vertically interconnecting distinct device layers by replacing long 2D interconnects with shorter vertical interconnects. This reduces power loss across interconnects while allowing the design of high bandwidth communication between the two dies which is challenging in a side by side placement. In order to allow communications between stacked dies, metallic connections are required. Cu-Cu solid state direct bonding is ideal to form the required metallic connections if the bonding process is compatible to semiconductor process technologies. Solid state direct bonding allows reduction in pitch size required for the bonding of interconnects from non-solid state bonding methods ( 1). Surface activated bonding (SAB) method is an attractive option for Cu-Cu bonding as the force application duration is considerably short compared to thermocompression while multiple wafers or dies can be processed together in the annealing process at less than 400 °C ( 2). However SAB is difficult for Cu-Cu in atmospheric condition and even more difficult for die to die bonding. Die to die bonding will allow flexibility and optimization in the design of the individual die in the case of heterogeneous integration. In this report, blanket and patterned Cu-Cu die to die SAB have been demonstrated. Two different types of plasma were used for the surface activation of Cu surfaces. Both Ar and Ar / N2plasma activation were performed on the Cu surfaces enabling direct bonding at room temperature, ambient air, and atmospheric pressure without any wet chemical process. Cu-Cu SAB samples with 2mm by 2mm bonding area were prepared for shear strength test. Bonding process was carried out in ambient air, and atmospheric pressure and room temperature with a bonding force of 500N for each sample for 5 minutes. After the bonding process, force was removed and annealing was carried out after multiple purges to form a N2 environment in the chamber. Annealing was carried out at 300 °C for 1 hr with a slow ramp rate. After annealing, the bonded dies were shear strength tested by a shear test machine. The bonded samples are shown in Figure 1. The shear strength results for the SAB with Ar plasma and Ar/N2plasma are shown in Figure 2. Most samples failed at the substrate while some have a mixture of both substrate and interface failure. Patterned dies with cross bridge kelvin resistor (CBKR) structure were also bonded together successfully. Each pair of top and bottom dies was sent for plasma activation together. After the plasma activation, the dies were immediately aligned and bonded together using a Finetech die bonder. The bonding force used in the bonding process was 210N for 4 minutes. After bonding a few set of dies, the samples were transferred into the wafer bonder for the annealing process. Annealing was carried out in N2environment at 300 °C for 1 hour with a slow ramp rate. After annealing, the samples were assessed to be bonded by shaking the dies in a waffle tray. The CBKR structures are shown in Figure 3 and the bonded samples in Figure 4. The dies were then tested electrically to ascertain that all CBKR structures were bonded. Blanket Cu-Cu die to die SAB have been demonstrated with Ar plasma and Ar/N2 plasma. Good shear strength results have been achieved from the bonded samples. CBKR structures patterned dies were also successfully bonded together. This shows that both Ar and Ar/N2plasma activation enables patterned die to die direct bonding at room temperature, ambient air, and atmospheric pressure without any wet chemical process. Works Cited T.Suga. Feasibility of surface activated bonding for ultra-fine pitch interconnection-a new concept of bump-less direct bonding for system level packaging. Electronic Components and Technology Conference, Las Vegas, NV, 2000; pp 702 - 705. T.Suga. Cu-Cu Room Temperature Bonding - Current Status of Surface Activated Bonding(SAB). ECS Transactions 2006, 3 (6), 155-163. Figure 1
Read full abstract
Dec 28, 2016
Bo Zhang + 6
Open Access
