Gold wire bonding on aluminium pads plays an important role in microelectronic packaging. The reliability and lifetime of the interconnection is determined by intermetallic phases formed at the Au/Al interface. Five phases are stable at room temperature. Some of them are supposed to undergo degradation processes due to mechanical stress or chemical reactions during IC lifetime. For microstructural characterization, investigations by EBSD are applied as an alternative to expensive TEM analyses.
In the manufacturing of microelectronic devices one of the common interconnection technologies is thermosonic wire bonding. Mostly gold wire is used to connect the microchip with its periphery. At first the wire, especially a small ball at its end, is pressed on an aluminum pad at the chip to form a ball bond contact. Then the wire is moved to the second contact pad where a so called wedge contact is formed. The use of temperature and a given bond force in combination with ultrasound movement of the bonding tool during the bond process ensures the generation of stable contacts.
In the contact area a microscopically thin layer of Au / Al compounds forms. On the one hand it is responsible for mechanical stability but on the other hand it is a weak point due to possible degradation processes. At room temperature five intermetallic phases (IMP) are stable (tab. 1). Elevated temperatures, originating e.g. from device operation, lead to an increased interdiffusion of gold and aluminum atoms. Thus the intermetallic region grows and different phases develop (fig. 2). Some of the IMP are more susceptible to chemical reactions activated by the ingress of moisture or corrosive substances from the encapsulation. Mechanical shocks as well as the formation of Kirkendall voids during diffusion lead to cracks propagating through the brittle IMP. The degradation of the contact region reduces the reliability of the bond contact as well as the entire unit. Thus, for a precise failure analysis it is essential to know what phases form and where they are located within the Au/Al interface.
Electron Backscatter Diffraction (EBSD)
To characterize the intermetallic phase region and for a qualitative phase determination on the basis of crystallographic characteristics mainly transmission electron microscopy (TEM) is used.
This method has a very high spatial resolution, but lacks of the fact that only small samples can be observed that may not always be representative for the whole specimen. By examination in the scanning electron microscope (SEM), the expensive and time consuming TEM investigations can be avoided. The sample preparation is limited to "only" the preparation of a nearly defect-free cross-section. The electron backscatter diffraction (EBSD) method used in the SEM is based on the crystallographic structure of the investigated materials as well. The electron beam strikes the sample tilted by an angle of 70°. So called Kossel cones of elastically scattered electrons are generated by the Bragg reflection of the electrons at the lattice planes. Conical sections appear on a fluorescent screen as Kikuchi bands, which are registered by a fast CCD camera. Each scanned point on the sample provides a crystallographic phase and its orientation by automatic comparison of measured patterns with calculated ones. A reliable identification of the intermetallic Au / Al phases is possible because of their different crystal structures.
Sample Preparation and Experiments
For our investigations ball bond contacts were made with 4N gold wire with a diameter of 25 µm bondet on commercially available Al metallization. Different reaction stages were achieved by heat treatment for various times at a temperature of 150°C. The temperature is far below the liquidus, so only solid-state reactions take place. The embedding of the specimen was done with cold-curing epoxy resin in order to avoid influences on the phase composition. The samples were ground and finally polished with ¼ µm diamond suspension. Furthermore argon-ion polishing was applied for the removal of the surface artifacts from the previous mechanical polishing step. For charge dissipation during SEM analysis, a thin carbon layer was deposited.
The EBSD measurements were performed at 20 keV and, for scans, with a step size of 50 nm, as small grains could be expected especially in the early stages of thermal treatment. Because of the very thin IMP layer formed directly after bonding, an EBSD phase map could not be performed in this stage. But it was still possible to detect the phase Al3Au8 by EBSD point analysis (fig. 3). A possible second phase could not be confirmed. After heat treatment for 6h/150°C the phases Al3Au8 and AlAu2 could be found. At the same time the aluminum of the bond pad is still present (fig. 4a). In the next phase evolution stage, the Aluminum pad is completely consumed and the formation of a new phase AlAu4 begins. At the same time AlAu2 is nearly totally converted into the now dominating Al3Au8. AlAu2 only exists in areas in contact with the aluminum. Figure 4b shows the corresponding EBSD phase map. Another measurement after 500 h (fig. 4c) now shows almost only AlAu4. Both Al3Au8, as well as AlAu2 exist in a small amount only at the periphery, i.e. in contact with the aluminum. This can be understood particularly well by planar measurements, where all three phases are located in a concentric arrangement with decreasing Au content in the direction to the edge.
Until now, only the gold-rich phases could be detected by means of EBSD, whereas in the literature sometimes all five phases are found in the connection area . It has to be discussed if the two phases AlAu and Al2Au cannot be resolved by the EBSD method in general. Also it has to be considered whether certain factors in these experiments (such as wire material) have led to the fact that these two phases do not arise or even more are not stable.
With regard to the latter assumption, another gold wire with a certain amount of dopants was investigated in comparison to study the influence of wire chemistry on the phase formation. The experimental parameters were not changed. EBSD phase maps show the same general growth sequence but a slower growth rate. Additionally it has to be noticed that the arrangement of single phases in the IMP region is different. In contrast to the common layered structure with decreasing gold content, now veritable block-like growth can be observed (fig. 5). The cause for the deviant behavior may be a diffusion barrier consisting of dopants that has formed between the phase region and the gold ball. Thus, the horizontal phase expansion can primarily explained by the existence of only a few descrete points where the unhindered contact between aluminum and gold is guaranteed and from which the phase formation starts. Though it was not possible to confirm the assumption that a different wire chemistry may lead to the growth of further IMPs, like the AlAu and AlAu2 phases. However another growth mechanism could be revealed.
With the help of the example described above it could be shown that EBSD is an applicable method for highly resolved investigations of crystalline materials. Due to faster and cheaper sample preparation, as well as shorter investigation periods, EBSD is an excellent alternative method compared to TEM.
 Büchler H. et al.: J. of the Less Common Metals 160, 143-152 (1990)
 Puselj M. et al.: J. of the Less-Common Metals, 35, 259-266 (1974)
 Range K.-J. et al.: J. of the Less Common Metals, 154, 251-260 (1989)
 Büchler H. et al.: J. of the Less-Common Metals, 161, 347-354 (1990)
 Noolu N. et al.: J. of Electronic Materials 33, 340-352 (2004)
Related Articles :
Keywords: Aluminium Benjamin März EBSD Electron Backscatter Diffraction Gold Intermetallic Phases Material Sciences Microelectronic Devices Microelectronic Packaging Scanning Electron Microscope Scanning Electron Microscopy Wire Bond Interconnects Wire BondingEmail requestCompany Homepage
Fraunhofer Institute for Mechanics of Materials