Impact Factors on Low Temperature Cu-Cu Wafer Bonding

Metal thermo-compression wafer bonding is a promising candidate for three-dimensional integration of Integrated Circuits (3D ICs). 3D ICs are very interesting as they allow for higher performance due to shorter interconnect paths  fabricated in a wafer-level integration approach, even for wafers processed in different technologies. For example, Complementary Metal Oxide Semiconductors (CMOS) and Micro- Electro- Mechanical Systems (MEMS) may be integrated this way. Due to its high electrical and thermal conductivity, low cost, and current availability in numerous IC fabrication facilities Cu is a very attractive bonding material for 3D integration. During the past decade processes were developed for Cu-Cu thermo-compression bonding using about 1 h at 400°C under high contact pressure. A substantial bonding temperature reduction is required in order to optimize post-bond wafer-to-wafer alignment accuracy, to minimize thermo-mechanical stresses, to prevent device degradation (as it is the case for Dynamic Random Access Memories - DRAMs), and to increase the throughput of the bonding process.

The aim of this work was to study the physical mechanisms of Cu-Cu bonding, develop a deep knowledge of the effects of these mechanisms, and then apply this knowledge to optimize the bonding process for low temperatures. This was achieved by preparing 150 mm silicon sample wafers by two alternative methods (Cu sputtering and electroplating) either with or without Chemical Mechanical Polishing (CMP), running a set of experiments covering combinations of wet chemical ex-situ and dry in-situ oxide reducing pre-treatments and then evaluating the results using analytical techniques, such as Atomic Force Microscopy (AFM), Electron Back- Scatter Diffraction (EBSD), Transmission Electron Microscopy (TEM), C-mode Scanning Acoustic Microscopy (C-SAM), several elemental analysis techniques and a qualitative version of the Maszara crack test. Here the used wet chemistry was diluted citric acid (1wt%) and the reducing gas was forming gas (96% Ar + 4% H₂). Subsequent to the wafer bonding process, the post-bond annealing was performed in order to increase the bond strength while keeping the retention time in the wafer bonding system as short as possible.

It was determined that for successful low temperature Cu-Cu thermo-compression wafer bonding the following classes of variables have to be considered (i) the contact of the two Cu surfaces, even at atomic scale, (ii) the cleanliness of the Cu surfaces, (iii) the diffusion properties of the two Cu layers and (iv) process conditions. In this paper the measured experimental results are compared with current knowledge and theory. In addition, key variables for establishing a successful bonding process were discussed. Using the fundamental information learned in these experiments it was possible to perform Cu-Cu thermo-compression wafer bonding at and below 200°C (see Figure 1); even room temperature direct bonding is achievable.