Processing and Reliability of Corner-bonded CSPs
BY BRIAN J. TOLENO AND JOSEF SCHNEIDER
Underfill materials are used to increase the reliability of devices such as flip-chips, chip scale packages (CSPs), ball grid arrays (BGAs), micro BGAs and other components. The underfilling process can be costly in materials, capital equipment, and process time. Many manufacturers use underfills to compensate for the large coefficient of thermal expansion (CTE) mismatch between the silicon die and the FR4 substrates typically used on flip chip on board (FCOB) or flip chip in a package. Other packages may only be underfilled if there is a perceived risk such as a cell phone withstanding frequent drops, or for use in high reliability applications such as avionics.
The CTE of underfill materials is typically matched by adding silica-based fillers that lower the CTE to provide a gradient between the silicon chip and the substrate. Underfills that provide physical protection to packages typically have a higher glass transition temperature, Tg, with a modulus that is better suited for increasing reliability.
Today's most common underfill systems are low-viscosity liquids that flow under a component by capillary action, wetting to the chip and substrate surfaces and encapsulating the solder joints. Underfills designed for assembly-level flip chip generally cure in five minutes or less at 165°C to form a hard seal with high adhesion to both the component and the substrate. Underfills designed for package-level assemblies offer improved assembly reliability, but require more time to flow under the die and cure. Since most capillary flow underfills are permanent, faulty or skewed components must be detected prior to underfill cure. Reworkable capillary underfills are available, but their use is not yet widespread.
Underfill is applied close to the edge of a flip chip or CSP to enable capillary forces to encapsulate the gap between the component and the board. Dispensing of capillary underfill materials requires specialized equipment to achieve the accuracy and precision required for high volume assembly. At a minimum, the dispenser must reproducibly position successive assemblies and apply a predetermined volume of underfill to the edge of the component. Heating the substrate is a common secondary requirement that accelerates capillary flow. Cure is usually accomplished in belt-style reflow or curing ovens.
Both dispensing and curing of underfills require extra capital equipment and add steps to the manufacturing process. Fluxing underfills have improved processing concerns, because they can be applied to the substrate prior to reflow. The component is then placed in the underfill, which acts as a flux. The underfill cures during the reflow process. Fluxing underfills typically are not reworkable.
This article presents an alternative to fully underfilling components to protect them from vibration, shock and dropping trauma. New corner-bonding underfill materials increase CSP reliability, reduce related manufacturing costs and allow for post-cure rework.
This study used an in-house designed test board and selected packages, and relied on a standard SMD assembly process: 1) no-clean solder paste printing, 2) corner bond underfill dispensing, 3) package placement, and 4) reflow soldering (Table 1).
Table 1. Package types tested.
Optimum process parameters were evaluated before the assemblies could be built for reliability testing. One key issue is a lack of space for the underfill material between the corner solder bump and the edge of the package. All the study packages have full array or perimeter solderball patterns; even in the package corners, there are solderballs. No corners are solderball-free for underfill application. Two packages from Asia are available without solderballs in the corners to allow the placement of underfill dots in these areas. These two exceptions were unique package designs not available when this study was conducted.
Table 2. Edge spacing by package type.
The available space between the package edge and the first solderball is quite small. Measured values for the packages used in this study are presented in Table 2.
Reflow Soldering Profile
A stepped reflow profile was used for reflow soldering and curing the underfill dots: 40-60 seconds above 200°C; peak temperature of 230°-245°C; and DSC analysis showed that the corner underfill was fully cured.
Dot Dispensing and Placement
A rotary pump dispensing system was used to dispense the corner bond material. To determine the dot diameter increase after placement, glass die replaced the actual packages. Figure 1 shows each assembly sequence using the LFBGA 160 package. The dispensed underfill dot contacted the printed solder paste deposit (shown in Figure 1 c. and d.) and badly affected the solder joint during the reflow soldering process. To avoid contaminating the solder paste deposits with underfill material and thereby producing bad solder joints (1e.), the right underfill diameter and height must be used for a specific package. This geometry depends on the gap between the substrate and the package, and the available space between the package's outside edge and the closest solderball row. The typical dot geometry is peaked. If the center of the adhesive dot is defined as the edge of the package, there is minimal interference with corner bump solder joint formation. This position also allows a fillet to be formed with the corner of the package.
Self-alignment Capability of Corner-bonded Area Array Packages
The following conditions were used to check the self-alignment capability of underfilled packages: application of one suitable underfill dot in each corner; four dots applied for each package; placement offset adjusted in x-direction to simulate the worst case; and the solderball center placed on the pad edge — an exaggerated offset half of the pad width (Figure 2).
Figure 2. Placement equipment image showing placement position (50 percent off related to pad width).
Three test runs were conducted to achieve optimized dispense settings for each package (Tables 3, 4, 5 and 6).
Table 3. Dispense parameters for the first test run.
Table 4. Contact resistance results after the first test run.
After soldering, the contact resistance of each daisy chain package was measured and the position of the solderballs was checked in relation to the substrate pads. Two packages created problems: the LFBGA 208 and the UCSP 50 both showed an offset after soldering. The UCSP 50 package exhibited a short as well.
Dispense volume was adjusted for the following packages: PBGA 313 (increase), LFBGA 208 (decrease), UCSP 50 (decrease). No other parameters were changed. The LFBGA 208 and the UCSP 50 still showed an offset after soldering.
Two packages that still showed an offset after reflow, LFBGA 208 and UCSP 50, were soldered without the corner underfill.
While self-alignment of the two critical packages (LFBGA 208 and UCSP 50) was much better without underfill, the components alone could not align without underfill with a 50 percent offset. All the other tested packages showed perfect self alignment with and without corner underfill. In a production environment with good controls, optimized parameter settings that ensure placement will be much lower than 50 percent offset. This study's 50 percent offset is a worst-case scenario and is used to demonstrate the effectiveness of the corner underfill in overcoming a worst-case process defect.
Figure 3 a.) PBGA 256 offset placement after reflow, b.) PBGA 256 after reflow (corner underfill on the right).
Package position after placement is shown in Figure 3a., while 3b. verifies the self-alignment capability of underfilled packages after the reflow soldering process — confirming that packages can still self-align with the corner underfill material.
Test vehicles were assembled using the dispense parameters from the second test run (Table 5). Devices were tested against thermal cycling and shock/drop. The thermal cycling testing conditions were –55° to +125°C air-to-air, up to 2,000 cycles. The shock/drop testing consisted of drops from 1.5 m onto the edge of a steel plate. The results are compared against devices with either full fast flow snap cure (FFSC) and reworkable underfill or no underfill that were produced in previous studies. The PBGA 313 fails early because of the package's increased weight, even without underfill.
Table 5. Dispense parameters for second test run.
The reworkable material was the only underfill that survived 50 drops on all packages without any defect, most likely due to its low modulus. All the other packages treated with the corner underfill passed the 50 drops without any disconnected solderballs. In all cases, the corner-bonded packages had increased reliability over non-underfilled packages. Only the larger packages showed any defects.
In addition to drop testing and thermal cycling, the material's adhesion strength was tested through two reflow cycles. Three different dot sizes were dispensed (dispense time: 500 msec/ 700 msec/ 1,600 msec) directly onto a soldermask in order to simulate underfill dots of adjacent packages during the package rework process. The dots were fully cured and removed with a Dage shear tester. Three values were measured in grams for each set of 10 dots: minimum, average (mean) and maximum force. Two different cure conditions were used: 1x or 2x reflow cycles.
Table 6. Contact resistance results after the second test run.
Cured twice, the dots require equal or even slightly more shear force for removal than underfill dots cured only once. Therefore, a second heating cycle should have no negative effect, for example, decreasing adhesion properties of the underfill to the soldermask layer of a laminate board.
Although never a planned outcome, defects occur even in controlled processes. The ability to rework an underfilled device is an advantage that can reduce scrap costs. Since this material is not completely underfilling the device, the reworkability of the underfill was investigated.
Two packages were used for rework trials: UCSP 50 and PBGA 256. Standard rework equipment removed the soldered package. The cured underfill dots were then removed from the soldermask.The board was heated from the bottom using IR, and from the top using hot air to remove the soldered and underfilled package. To remove the cured underfill residues from the soldermask layer, the board was again heated on the bottom to a temperature of 90°C, and the top to a temperature of 140°C. A spatula, followed by a flux pen, scraped off the underfill residues. The site, finally, was cleaned with isopropanol. After rework, little underfill residue remained on the soldermask. These remnants are not a problem for later underfill dispensing and package placement.
Plans are underway to develop materials that can be pre-applied, allowing packages to be supplied with the corner-bond underfill already attached. Adhesive suppliers also plan to develop materials that are suitable for a Pb-free process.
Please contact the authors for a complete list of references and tables.
The authors would like to thank Andreas Karch, of Henkel Germany.
Editor's Note: This article originally was presented as a paper at the IEEE/CPMT/SEMI STS: International Electronics Manufacturing Technology Symposium in 2003.
BRIAN J. TOLENO, senior chemist, and JOSEF SCHNEIDER, manager of Electronics Engineering, may be contacted at Henkel Corp., 15051 E. Don Julian Rd., City of Industry, CA; Gutenbergstrasse 3, D-85748 Garching-Hochbrück, Germany.