Ultrasonic Flip Chip Bonding
Modern electronic devices are required to be thin, lightweight and functionally sophisticated. To this end, small form factor and high density printed circuit board (PCB) assemblies have been realized using flip chip (FC) bonding technology.
However, the requirement for finer pitch PCBs continues to increase as integrated semiconductor devices with more signal pins require narrower intervals of input and output pads.
Recently large scale integration (LSI) assemblies with 40-micrometer pitches have appeared in devices such as liquid crystal display drivers. Additionally, the demand for lead-free products is increasing. Conventional FC bonding methods have difficulties adapting to these requirements. Inevitably, conventional FC bonding technology will not provide sufficient capability or productivity to meet future demands.
To address this problem, we recently studied flip chip bonding using ultrasonic vibration on flexible printed circuits (FPC) like those required in mobile phone applications. Focusing on large-size dies with multiple pads, we evaluated various bonding parameters and bond reliability using several kinds of copper-clad laminate (CCL). In doing so, we showed the correlation between formation of microscopic metallic bonds and actual chip on flex (COF) module performance. Ultimately, we succeeded in using ultrasonics to create a 35-micrometer-pitch COF assembly.
In response to higher density mounting of electronic components, Fujikura has established COF technology, mounting bare dies onto an FPC board. The wire bonding (WB) method and FC bonding have already been applied to mass production. The anisotropic conductive film (ACF-a lead-free, epoxy-based system) method and solder bump connection method have also been realized.
However, each of these methods has its drawbacks.
Compared with the ACF method, for example, ultrasonic bonding offers the advantages of lower electrical resistance and stronger mechanical retention. This is because of the metallic bond created between the gold (Au) bumps and the Au-plated pads on the FPC. Ultrasonic bonding also requires less down force. This reduces deformation and cracking of FPC pads.
With ultrasonic bonding it is not necessary to control solder wetting or wash away any kind of soldering flux. It is also superior to the ACF and solder bump methods in that it requires lower temperatures. Ultrasonic bonds can be created at room temperature, an important consideration given that FPCs can be harmed by excessive heat. The small dimensional deviation of FPCs at lower temperatures also helps realize fine pitch bonding with high accuracy.
In the past, the ultrasonic flip chip bonding method has been only applied to small dies with a few pins. This is because of the equipment used-a modified wire bonding machine that lacked effective vibration energy and caused harmful vertical vibrations. For our study we used a new concept ultrasonic bonding machine adequate to the large size dies and without these problems.
The ultrasonic flip chip bonding process is in principle a fairly straightforward one. First, the Au bumps on the bare die and the Au-plated FPC pads are brought into correct alignment. Then, the bumps contact the pads with a low force. The Au-Au metallic bond is generated through a combination of force and ultrasonic vibration.
In our study, two kinds of test elements groups (TEGs) with different types of CCL and bump shapes were used to evaluate the ultrasonic bonding method. The first, TEG 1, had no adhesive layer and plated bumps, while TEG 2 had a 10-micrometer adhesive layer between the copper foil and base polyimide film, and stud bumps. TEG 1 was intended to compare with the ACF method, TEG 2 with the solder bump method. Each FPC pattern was made from 18-micrometer-thick copper foil, plated with Ni-Au. The gap between the bare die and the FPC was reinforced with underfill resin and heated to 150 C for 30 minutes.
We considered a number of variables, including bonding force, temperature, amplitude of ultrasonic vibration and vibrating time. Changing these conditions, we then observed the cross sectional shape of each bond, its electric resistance and the die shear strength.
We also performed a thermal cycle test to evaluate the reliability of the bonding. The lower temperature was -40 C and the higher temperature was 125 C. Each extreme was maintained for 30 minutes. Electric resistance was measured over the whole test cycle. The threshold was a 20 percent rise of electric resistance to the initial value.
Evaluation of Bonds
In looking at the cross sectional views of the bonding points for the two test groups, we could not find any voids between the bump and pattern. Because the cross sectional shapes of all bumps are almost the same across the entire die, we concluded that the same power of ultrasonic vibration is distributed to all bumps.
It was observed that the bumps were well collapsed by the bonding force and ultrasonic vibration, although the collapsed part of the plated bump is much more than that of the stud bumps. Where the bonding points were intentionally broken, many dents were observed. We consider these dented sections to have been caused by the Au-Au bond between the bumps and the patterns.
New pure Au surfaces appeared at these sections while collapsing. When a bump is wider than a pattern, the Au-Au bond is formed mainly at the edge of the pattern. When a bump is narrower than a pattern, it is formed at the edge of the bump.
The bonding process itself takes less than one second and does not depend on the number of pins, bump shapes or the type of CCL. Compared with the 10 to 20 seconds of thermocompression required in the ACF method and the 3 to 5 minutes of reflow soldering in the solder bump method, the process time is greatly shortened.
The result of the die shear strength measurement showed that a breaking force of more than 0.2 newton per bump was required in either bump shape. In some cases, when the bumps were broken, the Au parts of the bumps remained on the patterns. This means the strength of the bonding is as strong as the original metal strength.
Our test showed that more bonding force, higher temperature, more amplitude of ultrasonic vibration and more vibrating time made for more collapsed bump area and higher strength. However, when increasing these parameters, it is important to recognize the risk of an electric short between adjacent bumps as well as possible breakage of the silicon die.
We consider the root cause of open circuit failures that occur during the underfill cure process to be shear stress between the silicon die and the polyimide base film. This stress results from different coefficients of thermal expansion. When an assembly is subject to high cure temperatures, stress concentrates on the bonding points and breaks the metallic bond. It is therefore very important to adjust bonding conditions to get enough bonding strength.
Thermal cycle testing of TEG 1 and TEG 2 showed that resistances did not rise during 1,000 cycles. This is because once the underfill cure process is completed successfully, sufficient reliability is obtained by the stress relaxation of the underfill.
Liquid crystal polymer (LCP) based CCL is regarded as superior to polyimide-based CCL in moisture absorption, dimensional stability and dielectric constant. Because these properties are useful in controlling impedance characteristics, we think LCP-based CCL is suitable for fine-pitch LSIs. Thus, we evaluated the bondability of the ultrasonic flip chip method using LCP-based CCL.
Again, looking at cross sectional views of the bonding points, we could find no voids between the bumps and the patterns. We also observed well-collapsed bumps and low initial electric resistances. Therefore, we conclude that LCP-based CCLs are compatible with the ultrasonic flip chip bonding method.
For comparison, we also tried the ACF method using an LCP-based CCL. Cross sectional examination of these bonding points showed the patterns had sunk down into the LCP base film as a result of the higher bonding temperatures and longer process time. The silicon die also moved down until the edge of the die would touch the copper patterns. This offered yet more evidence of the advantage of the ultrasonic method in terms of force, temperature and shorter process time.
35-micrometer Pitch Bonding
Having tested a number of different parameters, we next tried to assemble a 35-micrometer pad pitch COF module that we named TEG 3. TEG 3 was made of a polyimide-based CCL with 5-micrometer-thick copper foil.
The result of our efforts was an assembly in which the bumps and patterns were well positioned without misalignment by ultrasonic vibration. All the bumps collapsed in the same shape with no voids or no short-circuit failures. Breaking of the thin copper patterns by ultrasonic vibration energy did not occur, and we found acceptable values of resistance.
Therefore, we believe the ultrasonic flip chip bonding method has enough performance to assemble fine pitch COF modules with less than 35 mictrometers pin pitch. The next step will involve applying the method to actual modules and mass production.
This article is based on a paper presented at the Electronic Circuit World Conference, part of APEX/IPC Printed Circuits Expo 2005.