The Use of Solder as an Area-Array Package Interconnect

This tutorial details the role of solder as the primary interconnect material for area-array packages, such as BGAs and CSPs. This article details the nature, options and limitations of solders, solder bumping and soldering categories which are necessary to successfully implement area-array packaging.

- By Dr. Ning-Cheng Lee, Indium Corporation of America, Clinton, N.Y.

Throughout the history of electronics manufacturing, packaging trends have moved progressively toward being smaller, faster, lighter and cheaper. From surface mount technology (SMT), packages evolved further to the peripheral fine-pitch lead approach.

This development ran into limitations quickly when employed at 12-16-mil pitch. To address this challenge, area-array packaging (AAP) technology emerged, offering nearly a quantum leap over peripheral packaging technology. From BGA to CSP to flip chip, AAP now provides great benefits at both the IC and component levels. CSPs are particularly attractive AAPs, because they deliver a nice balance between miniaturization and manufacturability.

The interconnect technology has also been evolving over time. Although many methods have been developed, they can be roughly categorized into three groups: (1) liquid-metal bonding (soldering), (2) solid-metal bonding (wire bonding, TAB, etc.) and (3) metal-filled polymer bonding (anisotropic conductive adhesive, isotropic conductive adhesive, etc.).

Table 1 lists the performance of each of those three groups. Of the categories listed, solid-metal bonding cannot fill the role of interconnect for AAPs. Metal-filled polymer bonding, although somewhat promising for AAPs, also falls short in most of the needed properties when compared to liquid-metal bonding and can only be regarded as a supplementary interconnect technology.

Soldering is by far the favorite approach for providing the AAP interconnect. This is especially true for the second-level assembly stage for BGAs and CSPs. Consequently, it is important to understand the nature, options and limitation of solders, solder bumping and soldering categories to successfully implement

The choice of solder alloys is determined by the requirements of both process and reliability. Initially, besides meeting the solder wetting requirement, the solder chosen should be able to maintain its physical and mechanical integrity during subsequent processing.

In this manner, at the end of the packaging and assembly processes, the solder joints formed initially will not be altered or ruined. For area-array packages, interconnecting solder materials are usually introduced at two stages. The first stage is the predeposit of solder onto the packaging, usually accomplished through solder bumping.

The solder bumped package is then mounted onto the next level of packaging through soldering. The soldering process at this point may not need the introduction of additional solder. The additional solder materials may or may not be the same solder alloy as the solder bump on the packaging. When needed, additional solder materials are often introduced through either solder coating onto the next level of packaging or solder paste deposition as a bonding medium.

Solders can be generally separated into the following groups:

• Alloys used in flip-chip solder bumping and soldering:

  For FC in component (FCIP), the solders utilized for FC solder bumping and joining, such as 97Pb3Sn or 95Pb5Sn, must normally have high melting points. For direct chip attachment or FC on board applications, the solders utilized for FC bumping, as well as for solder coating on the next-level packaging, are often eutectic or near-eutectic tin-lead solders.

• Alloys used in BGA and CSP solder bumping and soldering:

  For heavy components such as CCGA or CBGA devices, the solder used for either column or ball is typically 90Pb10Sn. The column is mounted onto the package via either casting or 63Sn37Pb solder joining. For CBGA, the 90Pb10Sn solder ball is typically mounted with 63Sn37Pb solder paste.

  The high melting point of 90Pb10Sn solder ensures the required standoff of CCGA or CBGA on PC boards during board level soldering assembly using eutectic 63Sn37Pb or 62Sn36Pb2Ag solders. For light components, such as PBGA devices, the components are bumped with 63Sn37Pb or 62Sn36Pb2Ag and soldered onto the PC board with flux alone, or with solder pastes using similar alloy systems. The alloys employed with CSPs are similar to those used with PBGAs. However, the use of solder paste, rather than flux alone, is recommended for board level assembly.

• Lead-Free Solders

  Due to the toxicity of lead, there has been an effort to eliminate it from solder. The favorable Pb-free solder systems primarily comprise alloys of Sn with Ag, Bi, Cu, Sb, In or Zn (as shown in Table 2).

  These alloys may serve as substitutes for eutectic Sn-Pb solders in area-array packages. Nothing, however, has been developed yet as a substitute for high-melting-temperature solders.

Possible methods for solder bumping for CSPs, including the first-level interconnection, can be categorized as follows: (1) build-up process, (2) liquid solder transfer, (3) solid solder transfer and (4) solder paste bumping, as detailed below. For the CSP attachment process, depending on the type of packaging, either flux, fluxless or solder paste may be used as the bonding medium. CSPs generally offer a robust process, but attention should be given to reducing defects, such as delamination, misalignment, elongated joint, voiding, bridging, open, cracking, poor wetting and various attachment interactions.

The solder bump is built up by depositing solder gradually through a dry process, such as evaporation, or with a wet process, such as electroplating.

This is a dry solder build-up process, typically used for wafer bumping. The evaporated solder is deposited through a molybdenum mask onto the UBM surface. The metal mask is then removed, and the solder bump formed is often reflowed to fuse the solder. In general, the evaporation process is adequate for coarse pitch and low I/O devices, due to the constraints of metal mask technology.

Electroplating bumping can be regarded as a wet solder build-up process. At this stage, electroplating may be the most commonly used process for wafer bumping. Again, the solder alloys deposited are typically Sn-Pb systems. However, entrapment of impurities during the plating process may cause spattering during fusing, resulting in uneven bump size.

The solder bump is formed by transferring liquid solder onto the wafer metal base, either by solder dipping, such as meniscus bumping, or by liquid solder dispensing, such as solder jetting.

The wafer level bumping is based on the deposition of electroless Ni as a wettable UBM. Then, a solder layer, such as 80Au20Sn, is chosen and applied by a well-controlled dipping technique.

Solder jetting is a process whereby a molten solder droplet is ejected from an orifice with the use of a driving force. The most commonly used, and also the most successful, driving mechanism is piezoelectric force. Solder jetting used for BGA solder bumping has been demonstrated by Sandia National Laboratories, where the bumping of 400 pads can be accomplished with a single shot (see Figure 1).

Other mechanisms have also been attempted, such as electromagnetic driving force, reported by IBM. For a drop-on-demand wafer bumping process developed by MPM and Microfab, the molten solder droplet is directly ejected onto a wafer pad with an Au surface finish. The solder jetting may have the greatest potential as a low-cost bumping process. The throughput of this process, however, is still fairly low. In addition, the consistency of bump size also appears to be an issue.

The transferring of a solid solder mass to the pad area forms the defined solder bumps. This transfer process can be wire bumping, sphere welding, decal solder transfer, tacky dot solder transfer, pick-and-place solder transfer or integrated preform, as discussed below:

Wire bumping is similar to wire bonding. A solder wire, such as 97.5Sn2.5Ag, can be bonded directly onto the aluminum bond pad using thermosonic energy. The solder stud formed can then be reflowed to form a solder ball. The ball size and pitch limitations are determined by the diameter of solder wire and the thermosonic bonder's capability.

IBM Endicott has developed a solder bumping process for the TBGA package using a fluxless welding approach, as shown in Figure 2. During the welding process, the device-holding stage moves around while the welding tip fixture remains stationary. The welding process throughput is approximately 7.5 bumps/s.

In this process (developed by IBM-Endicott), solder is plated onto a non-wettable decal substrate such as aluminum and forms solder studs with a pattern matching that of a FC or CSP footprint pattern. This decal substrate, loaded with solder studs, is then placed on a fluxed wafer or CSP substrate, with each solder stud registered onto a metallized pad such as copper. This sandwiched assembly is then reflowed, followed by removal of the decal. The solder studs wet to the pad metallization and are detached from the decal substrate.

DuPont developed an approach using tacky dots on polyimide film to transfer solder spheres for bumping CSP and FC. The process involves using polyimide with tacky dots as a temporary sphere carrier. The sphere is loaded onto the tacky dots, followed by area-array package substrate placement, with spheres registered to the fluxed pads. The sphere is then transferred to the substrate through a reflow process.

For CSP solder bumping, the most common process involves using a pick-and-place machine to transfer solder spheres to a CSP substrate predeposited with flux or solder paste, followed by reflow. The pick-and-place mechanisms utilized include (1) vacuum pick-and-place and (2) gravity pick-and-place.

In the gravity pick-and-place approach, no vacuum is involved. A revolving process is used to first load spheres onto a sphere tray and subsequently to transfer the spheres onto a CSP substrate predeposited with either tacky flux or solder paste, as shown in Figure 3. All sphere transferring processes rely on gravity alone. This is possible through the proper positioning of both tray and CSP substrate via the revolving process. The populated CSP substrate is then reflowed to complete the solder bumping process.

Both pick-and-place designs for CSP solder bumping involve rolling of solder spheres back and forth between the sphere tray and the reservoir. This inevitably oxidizes the solder sphere and may pose a soldering quality issue at a later stage.

The integrated preform is a patterned solder preform. The patterned preform has a sub-preform corresponding to each pad of the CSP land pattern, and all neighboring subpreforms are interconnected with a thin solder link. Bumping with integrated preform can be achieved by placing the integrated preform on top of either flux or solder paste printed onto the CSP substrates. This approach has been reported by Indium Corporation of America to be promising.

Solder bumping can also be accomplished with the use of solder paste alone. This approach becomes more and more attractive when the pitch of area-array packages becomes smaller and smaller; therefore the solder bumps become smaller and smaller.

With the use of the sphere transfer approach, the cost of the sphere remains the same, regardless of sphere size, and the cost per bump is accordingly steady as well, regardless of bump size. Conversely, the cost of solder paste is determined by its volume. Therefore, with decreasing bump size, the cost per solder bump will be reduced significantly when employing the solder paste bumping approach (see Figure 4).

For sphere sizes smaller than 30 mils, bumping using solder paste alone becomes cheaper than using spheres, in terms of solder materials cost. If the equipment cost is also considered, the analysis should favor the paste bumping process even more. Processes of solder paste bumping include print-detach-reflow, print-reflow-detach and dispense.

The most desirable procedure is similar to the conventional surface mount process: print, detach the stencil, then reflow. This method offers the greatest potential for markedly reducing the bumping cost.

The second alternative involves printing the paste onto the area-array package with the use of a metal stencil; then reflowing the solder paste with the stencil left on and afterward removing the stencil; followed by cleaning.

CSP solder bumping may also be achieved with a solder paste dispensing approach. Although non-slump performance is still a required paste property, there is no issue related to an aperture nonclogging requirement. At this time the dispensing approach remains of interest, but its feasibility must still be proved.

Soldering is the primary interconnection technology for CSP. Methods for solder bumping for CSP, including for interconnects, can be categorized as follows: (1) build-up process, (2) liquid solder transfer, (3) solid solder transfer and (4) solder paste bumping. The first group includes both evaporation and electroplating processes, while the second group includes meniscus bumping and solder jetting. The third group includes wire bumping, sphere welding, decal solder transfer, tacky dot solder transfer, integrated preform and pick-and-place solder transfer processes, with the last one (pick-and-place solder transfer) being the current prevailing option.

Solder paste bumping exhibits great potential to reduce bumping costs dramatically and includes the print-detach-reflow, print-reflow-detach and dispense approaches. For CSP attachment processes, depending on the type of packaging, either flux, fluxless soldering or solder paste printing may be used as the attachment medium.