Current Issue An Independent Journal Dedicated to the Advancement of Chip - Scale Electronics October 2001 Email the editor

By Mike Fenner, Indium Corp. Europe, London

Figure 1. Flip chip, a technology with critical solder-bumping requirements, has become pervasive in consumer electronics, such as this Handspring Visor. The wafer with the devices is also shown (inset). (Flip Chip Division of Kulicke & Soffa Industries)

Components were first mass soldered to PWBs over 40 years ago, and soldering is still the preferred means for component attach.

Soldering has retained its popularity for component attachment because it is a versatile process that uses relatively cheap materials-materials that are suited to mass production methods, and which don't require expensive plants or processes.

Finally, soldering is relatively easily reversible, making soldered assemblies repairable. This article looks at solders and fluxes, and how they are specified for flip-chip and Tessera's µBGA packages (Figures 1 and 2).

Special Requirements

At first sight, the requirements for soldering flip chip and Tessera's µBGA packages are the same as for regular SMT components.

However, there is more to soldering miniature components than simply scaling an existing process. A further evolution of materials is required to account for the different surfaces and materials used and to recognize the increased structural role of the solder in leadless devices.

Bumping

Flip chip and µBGA devices do not have leaded terminations. In these devices, the leads are replaced by a metal bump that also makes the physical bond between the device and the circuit.

The two basic methods of applying metal bumps are co-deposition and soldering.

In co-deposition, a mixture of tin lead is applied to selected areas by sputtering, evaporation or plating via a mask. The mask is removed and the mixture of tin and lead is melted to form a solder alloy, simultaneously taking on the bumped profile.

The size of the bump is controlled by the coating thickness. Alloy choice is generally restricted to binary compositions because of the difficulty of depositing more than two constituent metals simultaneously.

In soldering, pre-alloyed solder, together with a flux, is applied to selected areas and then reflowed. The use of pre-alloyed solder provides a much wider alloy choice, since alloying is done "off-chip" before deposition.

Figure 2. Tessera's µBGA packages are leadless, with special structural bumping requirements. (Tessera)

Solder Application

There are two methods of solder application. In the first, flux and solder are applied one after another; in the second, they are applied concurrently. In either case, solder is first formed into spheres.

In the sequential method, the spheres are applied individually to each pad or land area of the fluxed chip or µBGA package. In the concurrent method, the solder is formed into many tiny spheres, which are blended into a paste with the flux, and then printed like ink and reflowed.

If the soldering process is to be successful, the quality and quantity of both the flux and the solder has to be carefully considered and controlled.

Alloy Selection and Control

Solder bumps are made from a mixture of tin and lead. The proportions determine the melting point of the alloy, and the most widely used is 63% tin, 37% lead (Sn63).

The use of pre-alloyed solder provides very accurate alloy control, increases conventional alloy choice and enables the use of ternary (three constituent) or even quaternary (four constituent) alloys.

Accuracy is as important as the melting. This is important for device users in setting their own consistent process parameters. The use of ternary (and above) alloys will become increasingly important with the move to Pb-free.

Most Pb-free alloys are composed of three or more constituents, and many are patented. Some of the patents for the Sn/Ag/Cu alloys are for very similar compositions, so that a slight change (fractions of a percent) in a minor alloy constituent can mean an inadvertent switch from one patented alloy to another.

Technical Challenges

Setting those aside for the moment, the major technical challenge posed by the move to Pb-free is that virtually all the proposed alternatives offer both a much higher tin content (50% increase) and a raised (+20°C) liquidus.

For the µBGA package, the challenge relates to the higher processing temperature requirements, which push construction materials to their limits.

At the same time, the new alloys require modified flux formulations to work at the same levels of efficiency as the common Sn63. These issues, however, are common to all polymer-bodied devices and are being addressed on an industry-wide basis.

With respect to flip chips, the greater tin content, together with the higher required temperatures, constitute a major problem. Together, these two factors mean that these proposed Pb-free alternatives can dissolve the under-bump metallization (UBM).

Most of the industry's investment in developing Pb-free materials has gone into the mainstream area of PWB assembly, with specialized areas, such as pad array devices, left behind.

Fortunately, certain companies with a greater awareness of solder applications in component manufacturing have developed Pb-free solders for specialized applications, such as balling pad array devices.

These specialty solders offer a smaller tin increase and lower operating temperatures, and can be employed as a drop-in for Sn63, with only minor changes in process set-up, as shown in Table 1.

The basic material cost is higher for these alloys, and this expense can be significant in volume use. In the highly manufactured forms used in pad arrays, however, this cost is much less significant. At present, there is no economic lead-free alternative to high lead (5Sn/95Pb) bumps.

Flux Selection

In a "classical" soldering operation, the primary functions of the flux are chemical tarnish removal and physical protection of the surface until solder wetting is accomplished. A well-designed flux will assist the spreading process, although this is not necessarily the case when soldering pad arrays.

A production process will have to consider not only the theoretical functions of the flux, but the practical considerations as well, including application methods, safety of residues in service and methods of removal.

There are three separate areas of interest, each with slightly different requirements. The first is ball attach to the device and the second is attaching the balled device to the circuit substrate. The third is removal and replacement, should this prove necessary.

Each process has a slightly different emphasis. For example, tarnish removal is a lower order of priority in device manufacture and repair situations, since the surfaces have already been cleaned and wetted during manufacture. Requirements now mostly center on physical properties to optimize the applications.

Concurrent or Printed Solder

Solder paste is printed through a stencil, with the stencil openings corresponding to the pad areas on the device (Figure 3).

The maximum height that can be printed is limited, so the area is increased to achieve the required volume of solder. The mechanical properties of the flux are more important than oxide reduction, as its primary role is to ensure consistent print characteristics (and therefore solder volume) and also to ensure that all the solder spheres coalescence into a single bump.

Sequential Processing

Flux is applied to the device by a wide variety of methods. The simplest method employs a brush, while the most complex probably uses an X/Y computer-controlled, high-speed precision dispensing machine. Other methods in common use include direct and indirect transfer printing, needle dispensing and spraying.

Essentially, the choice is between "whole device" and "bumps/pads only," and, of course, there is the usual balancing of speed against cost, and volume against flexibility.

There is no right or wrong way and no industry consensus either. Consequently, flux suppliers offer both liquid and paste materials to accommodate these methods. Liquid fluxes are available in a range of viscosities to suit the chosen method of application. (A representative selection of liquid fluxes is shown in Table 2).

After flux application, the spheres are placed and then reflowed to form bumps (Figure 4).

Figure 3. Sn63 solder paste, as printed for a 50-mil pitch device.ÊThe particle size is 25-45 microns.	Figure 4. Column Ball Grid Array package bumped with preformed spheres (IBM Corp.)

Solder Spheres or Pastes?

The price of solder spheres does not relate much to size; the manufacturing cost is much greater than the material cost.

The manufacturing cost of solder paste is also independent of intended sphere size, but smaller bumps need less paste than bigger bumps. This means that the material cost of smaller bumps is less with paste, and-with larger bumps-is more.

The crossover point is around 0.75 mm (0.030"). The final cost, of course, and the decision on which process to use, will depend on a variety of other factors.

For example, at very fine pitches, paste becomes unreliable-and eventually impossible to employ-due to the oversize printing requirements. At some point, the overprint from one bump will meet the overprint from another.

Machine Compatibility

Naturally, the most important requirement is that packaging shape be compatible with the machine. This also requires selecting a packaging size commensurate with the rate at which the product will be consumed.

Solder pastes, in particular, are time limited once opened. The rule is to select the container size that can be used completely in one work period per machine.