Issue



Flip Chip Interconnection Using Copper Wire Bumps


03/01/2006







Various flip chip systems have been used for advanced packages, including chip scale packages (CSPs) and ball grid arrays (BGAs). Wire-bumped flip chip systems offer the application flexibility, because direct bumping on an aluminum (Al) pad is possible without depositing an under-bump metallization (UBM) film and bump mask. Although gold (Au) wire is usually used for wire bumping on chips, flip chip systems based on Au stud bumps introduce several challenges. The Au stud easily forms brittle, intermetallic compounds, such as AnSn4, with most tin-based solder alloys. When conductive paste is used instead of solder, high contact resistance is caused by thermal strain or high temperature storage. To address these problems, copper (Cu) wire-bumped flip chip connection was evaluated.

Ultrasonic Ramp Control

Cu-wire stud bumps were formed using a conventional wire-bonding machine. A 25-µm-diameter, Cu wire of 99.99% was used for Cu bumping. The Cu ball was created by liquefying the end of the wire in a reducing gas atmosphere of 5% H2 in N2 to prevent it from oxidizing. SEM micrographs of the Cu balls show no oxidation after ball formation (Figure 1a). The Cu wire was bumped under ramp rate control of the applied ultrasonic power. Because Cu is not as malleable as Au, high ultrasonic power may cause chip pad cracks to occur. To evaluate the ramp effect, wire stud bumps were formed on the test chip using various ramp rates, and chip pads were examined after shearing off the bumps. When the ramp rate was decreased, the number of cracks decreased, and were finally eliminated at the ramp rate of 129 V/sec. Figure 1b shows the SEM micrograph of a typical Cu-wire stud bump.


Figure 1. SEM micrographs of a Cu ball and a Cu wire stud.
Click here to enlarge image

null

Flip Chip Assembly Using Cu-wire Bumps

Cu bumps were formed on daisy-chained test chips under optimized, ultrasonic ramp control. Au-wire stud bumps were also formed on test chips for comparison. The bumped chips were coined to a uniform bump height, and were mounted on FR4 substrates. Conventional 63Sn37Pb (SnPb) alloy solder paste and lead-free 96Sn3.5Ag0.5Cu (SnAgCu) alloy solder paste were deposited on the substrate by stencil printing. Interconnects were formed by reflowing the solder paste. A snap-cure epoxy underfill was applied to the flip chips. It was found that Au diffused into SnPb and SnAgCu and made a thick, intermetallic compound layer. Cu diffusion into the Sn-based solders was significantly less, as indicated by the ~5-µm-thick layer of intermetallic compound surrounding the bump.

Liquid-liquid thermal shock testing was conducted over a temperature range of -55° to 125°C with 10-minute cycles. Electrical probing and CSAM analysis was done every 100 cycles to identify failures. The SnAgCu test vehicles failed earlier than the other solder test vehicles. Half of the Au-SnAgCu failed after 200 cycles, and all of them failed after 600 cycles. All of Cu-SnAgCu failed after 100 cycles. Stress damage during the assembly process probably led to the poor reliability performance. The SnPb test vehicles also exhibited failure early in thermal shock testing, with first failure after 100 cycles in both Au- and Cu-stud test vehicles.

To improve the reliability of the Cu stud bump flip chip, a flip chip compression bonding process was also evaluated for the lead free SnAgCu test vehicles. Flip chip test vehicles were produced using B-stage compression bonding or high-temperature bonding. Reliability in thermal shock testing significantly improved, compared to the basic reflow assembly. The Cu-SnAgCu system showed a long mean time to failure; 2,269 cycles for the B-stage process, and 3,237 cycles for the high-temperature bonding process. This study indicates that Cu studs allow for an acceptable solder connection with the conventional SnPb alloy and the lead-free SnAgCu alloy.

Acknowledgments

The author would like to thank Satoru Zama, Hideaki Murata, and Toshiya Hikami of the Furukawa Electric Co., Ltd. for their collaboration on this study.

Click here to enlarge image

DANIEL F. BALDWIN, Ph.D., may be contacted at Engent, Inc., 3140 Northwoods Parkway, Suite 300A, Norcross, GA 30071; E-mail: [email protected].