Effects of Copper Pattern Density and Orientation on the Modulus of BGA Substrates

Abstract

Substrate material properties directly impact package mechanical performance. Estimation of substrate mechanical properties based on each component layer provides maximum flexibility and better accuracy. The component materials of a BGA substrate are the fiberglass-reinforced epoxy layers, copper and soldermask. Even when the material properties (modulus and CTE) of each component are known, the behavior of the final substrate is not easily determined because factors such as copper density and pattern have significant effects. For example, the copper pattern may be oriented such that conductor traces run in one direction for a considerable length. The modulus in the direction parallel to the traces will be different from the modulus in the direction perpendicular to the traces. Package mechanical simulations typically do not mesh the details of the copper patterns, but rather average within each layer, thus neglecting the orientation effects. Therefore FEM (Finite Element Model) simulations may yield inaccurate stress and warpage results. This study utilized a test vehicle (TV) to study the copper pattern density and orientation effect on substrate modulus. Each TV sample was a two-metal-layer copper-clad laminate designed with one of eight simple geometric patterns. One set of patterns had Cu traces running parallel in one direction, with varied metal densities of 0%, 25%, 50%, 75% and 100% respectively. Measurements were made both parallel and perpendicular to the trace orientation. Also studied were a pattern of holes (continuous metal plane with unconnected holes) and a dot pattern (unconnected circular metal pads). Substrate flexural modulus measurements were made using two types of instruments: DMA and Instron. Data was collected from −65 to 260°C. FEM simulations which meshed the details of copper patterns were used to calculate the overall substrate modulus. These correlated better with the empirical data than simulations utilizing traditional layer averaging. The copper pattern had significant effects; the substrate modulus was always much lower in the direction perpendicular to the traces than the modulus in the direction parallel to the traces. The amount varied depending on temperature, but could be as high as 35%. When measured parallel to the copper traces, the substrate modulus was observed to increase with copper pattern density as predicted by theoretical mixture rules. However, the substrate modulus in the direction perpendicular to the traces was not easily predicted theoretically. Therefore, more sophisticated averaging techniques and simulations are needed.