Abstract
Through-Silicon-Vias (TSV) for 3D integration are susceptible to mechanical failures such as TSV extrusion, chip cracking, and carrier mobility change, mainly due to significant thermal mismatch between electroplating copper and silicon wafer. Assessing TSV reliability requires accurate mechanical properties of electroplating Cu, which may not be achieved by traditional testing methods for large-size samples. Alternatively, nanoindentation can be used as a versatile and non-destructive technique to obtain material properties at micro-scale or nan-scale. The technique, however, is sensitive to experimental data variation and may yield non-unique properties, which is not yet fully understood for TSV-Cu and other packaging materials.
This paper reports the results and analysis of in-situ nanoindentation for TSV-Cu specimens. Each specimen has a diameter of 20 um and a depth of 180 um, fabricated on a 200-mm silicon wafer. The relationship of power-law hardening was used to characterize the mechanical behaviors of TSV-Cu with Young's modulus (E), yielding strength (Y), and strain hardening exponent (n). To extract these properties, two methods of so-called reverse analysis were applied to experimental test data. Finite element models were also built to reproduce the displacementloading curve based on the extracted material properties. The following results were observed: 1) with very different yield strength and hardening coefficient, identical load-displacement curves were produced. This indicates that single indentation testing cannot provide unique material properties for TSV-Cu; 2) the loading-displacement curves vary with the locations of the specimens, possibly due to the effect of grain size on nanoindentation, or experimental "errors"; 3) the plastic properties can vary dramatically with a slight change in the loading-displacement curve. This implies the sensitivity of extracted properties on experimental data variations. Generally, the non-uniqueness and sensitivity due to experimental data variations of nanoindentation testing are not independent. This study concludes that single nanoindentation testing itself may not be able to adequately and uniquely determine the material properties of TSV-Cu. To make the best use of the nanoindentation technique, a robust methodology is needed to minimize the sensitivity to experimental errors and analysis.