|
By Dr. Thorsten Teutsch, Pac Tech USA, Santa Clara, Calif. [pactech-usa.com]
|
| Many different methods have been proposed or developed for providing solderable bumps on wafers for flip-chip- or wafer-level package bumping. While much attention has been given to stencil printing, because of its attractive cost structure, solder ball placement methods are proving to be cost-effective alternatives to stencil printing in several areas. |
|
Soldering remains the most important interconnection method for assembling electronic devices to interconnection substrates.
Since the introduction of area-array packaging, soldering has advanced from art to true science. Over the years, many different methods have been developed to provide the bumps normally required to make assembly possible.
Solderable Interfaces
The use of preforms to create a solderable interface is one example. The pre-form method has largely been used for BGA balling in both automated and manual processes over the years.
However, because of the more significant challenges, the use of solder spheres for wafer bumping is not very common and has emerged only recently as a method that has found its greatest initial use in packaging wafer-level CSPs (WLCSPs).
For low-cost wafer bumping, solder ball placement methods offer certain advantages over other low-cost techniques such as solder paste printing.
Solder-ball placement also offers greater flexibility. For example, preformed solder balls are available in nearly all alloy compositions and in a broad range of diameters down to 60µm. Even so, when cost is the primary concern, stencil printing is a strong competitor.
While bump height may vary somewhat, the preformed solder balls offer one of the lowest production costs, especially when combined with electroless Ni/Au under-bump metallization (UBM). The technology, however, does impose limits at finer pad pitches (<200µm) and where overall interconnection performance is concerned.
Spheres Offer Tight Tolerances
In contrast, solder spheres offer very tight diameter tolerances and, therefore, height distributions of the final, reflowed solder bumps are very uniform.
This uniformity is achievable because, for example, in the lower diameter range, the spheres are usually available with tolerances of ±5µm absolute. Moreover, the generation of voids, which is definitely an issue for lead-free solder pastes, is much lower for solder spheres that are reflowed for attachment by a tacky flux.
Pre-fluxing of the of the UBM surface with a tacky flux prior to ball placement is necessary for almost all ball placement methods. The exceptions are solder jetting and laser reflow applications.
Solder Activation During Reflow
Flux not only provides the solder activation during reflow, it also holds the solder ball at the pad location after initial placement. Thus, solder ball placement usually requires a two-piece equipment solution: The first part is a flux system, that can utilize a needle transfer, spin coating or printing process; and the second component is a ball placement system.
The higher capital cost of a two-piece system implies a big disadvantage compared to stencil printing; however, there are a number of good reasons for considering the ball placement method.
The balance of this paper will discuss the principles and methods of ball placement for three product applications.
WLCSP Applications
To date, WLCSPs are the only embodiment of flip-chip technology that has kept the promise of low cost. With relaxed pitches of around 500µm and the use of small die sizes that enable underfill-free assembly, the devices have proven virtually 100 percent compatible with SMT.
Like everything in the electronics industry, semiconductors are very cost-sensitive and can only remain competitive if low-cost technologies for bumping and assembly are used.
Screen printing appears to be a good choice, however, since stencil printing is limited and cannot transfer enough solder volume to allow for underfill-free process flows during assembly. Gang-ball placement (Figure 1) of 300µm solder spheres is well established and has, thus far, proven to be the most suitable technology for those devices.
 |
 |
 |
| Figure 1. Micro-gang ball placement |
Figure 2. Ball transfer |
Figure 3. Ball drop attachment method |
Vacuum Process
There are a number of potential methods for accomplishing ball transfer and attachment (Figure 2). For example, different equipment and process concepts are currently available within the industry that utilize vacuum.
In the vacuum process, solder balls are sucked from a reservoir onto a metal template, which transfers the spheres using a multi-axis placement system. The UBM of the wafer, typically, is prepared in advance with a tacky flux that is commonly applied by stencil printing.
Another method for ball attachment is referred to as "ball drop" (Figure 3). This method is a modified printing process where solder spheres instead of paste is used to fill the openings of a metal stencil; each opening contains a single solder ball.
|
| Since the introduction of area-array packaging, soldering has advanced from art to true science. |
|
To ensure that the stencil is not contaminated with the previously applied flux, the stencil has to be elevated by organic columns functioning as spacers. The preparation process for this separation (spacer layer) on the underside of the stencil requires the use of a photolithography process and adds complications and cost to the stencil-making process.
Still, the solder ball transfer process provides specific advantages, especially in terms of placement precision, fine-pitch capability and solder ball diameter flexibility. In the final analysis, it appears that the ball-drop process offers a more economical alternative for solder ball sizes of 300µm and higher.
 |
| Figure 4. Replacement of missing solder balls |
Flip-Chip Applications
Once a seemingly overwhelming challenge, sandard stencil printing limitations, including fundamental pitch and pad density, can be overcome to a substantial degree by more complex and expensive process solutions.
For example, among the new solutions are those requiring the use of fine lithography and organic masks that serve as a stencil. Meanwhile, placement of micro- solder spheres, in sizes below 200µm, may prove to be a more cost-effective alternative.
In one transfer process, a reservoir of solder spheres, called a "jumper station," is ultrasonically agitated. A patented ball-vacuum unit, comprised of a metal template with holes, is then employed to create a very homogeneous vacuum distribution over the openings. This ball-vacuum unit is used to transfer solder spheres with diameters as small as 80µm.
This method may be used for assembly down to 120µm pitch and can transfer up to 400,000 balls to pads on a 200mm wafer, with process times ranging between 1.5 and 4 minutes/wafer-which is quite low.
Moreover, virtually 100 percent final assembly yield (relative to ball attach) can be reached when the technology is combined with a ball inspection and repair unit.
Misaligned solder balls are removed and replaced, and missing solder balls are simply replaced (Figure 4). The rework system is relatively simple, consisting only of a modified solder ball bumper and a 2D inspection system, which can be integrated in the micro-ball placer or implemented in a ball placement production line as a stand-alone unit.
Using laser heating, misaligned solder balls can be removed prior to or after the reflow process, and missing solder balls can be replaced.
|
| There are a number of potential methods for accomplishing ball transfer and attachment. For example, different equipment and process concepts are currently available within the industry that utilize vacuum. |
|
Other Applications
Another interesting process technology for new applications is the laser-assisted, solder-jetting method (Figure 5). In the jetting process, preformed solder spheres as small as 80µm in diameter are singulated by a rotating disk and injected into a ceramic capillary with an opening diameter smaller than the solder ball with nitrogen.
Clogged in the capillary, the solder ball is released by heating it to melting temperature with a millisecond laser pulse. The result is a liquid solder droplet of very controlled volume. This controlled volume (the volume amount of the initial solder ball), is jetted out from a distance of several 100µm to form a contact or bump when it makes contact with a solder-wettable surface.
 |
| Figure 5. Laser-assisted solder jetting |
MEMS Bumping
The contactless nature of this process makes the technology very well suited to the task of MEMS bumping and packaging.
The fluxless capabilities are also an advantage in the assembly of optoelectronic components and devices. However, one application for solder jetting that has found great favor is in hard disk drive (HDD) head-gimble assembly.
By forming a three-dimensional contact between a silicon device and the suspension of a hard disk, overall drive structure sizes can be reduced, providing advantages relative to both performance and final cost of the storage device.
As a result, many HDD manufacturers are beginning to switch to this technology in an effort to replace gold wire bonding and enable potentially much finer- pitch capability.
Summary
Solder ball placement methods have proven to be cost-effective and viable alternatives to stencil printing in a number of areas.
There are many methods of placing solder balls, making the process an attractive choice for a broad range of applications, especially where very fine-pitch applications (~120µm) or where very small solder balls (~60µm) are required.
The table summarizes the characteristics of the various processes and their relative benefits.
In summary, preformed solder ball technology offers the user the ability to accomplish substantial assembly volume in a relatively small space. It also offers freedom in the use of a wide variety of tools and processes, allowing for greater flexibility.
|
|
Dr. Teutsch is president and chief technical officer of Pac Tech USA, the U.S.-based subsidiary of Pac Tech GmbH. An eight-year veteran of the company, he received his doctorate in physical chemistry from the Fritz-Haber Institute of the Max-Planck Society. [teutsch@pactech-usa.com]
|
|
|
