February_EDFA_Digital
edfas.org 21 ELECTRONIC DEVICE FAILURE ANALYSIS | VOLUME 20 NO. 1 pillar geometries. The very large decreases in solder volume realized in copper pillar solder joint geometries (Fig. 1) wouldbe expected to affect initial microstructures. Changes in solder joint composition and thermal history are correlatedwith themicrostructure, shear strength, and ATC lifetimes of solder joints. The precipitate morphol- ogy is compared with changes in solder composition and aging time at elevated temperature. All of these values were correlated to solder joint mechanical properties in an effort to better understand and control copper pillar solder joint reliability. Copper pillars are much stiffer than lead-free solder, resulting in higher stress during assembly processes and in use, with the possibility of reduced reliability because fracture or delaminationmay occur. Therefore, studies of themechanical and reliability properties of packaging interconnect copper pillar joint structures are critical. Thepresent investigationwas focusedon the reliability of silicon/micro-copper pillar/solder/glass assemblies and silicon/micro-copper pillar/solder/silicon assem- blies—fundamental building blocks of 2.5-D packaging. These studies should also be relevant to 3-D packaging, where copper pillars are also used. The inclusion of glass interposers in this study is important, because thesemate- rials are important new alternatives to through-silicon interposers; glass interposers are relatively stable, low cost, and afford great flexibility inmanufacturing arbitrary sizes. [7-9] The investigation used shear strength measure- ments and ATC tests to characterize the reliability of silicon/micro-copper pillar/solder/glass assemblies with awide range of parameter values, including a varied range of pitches. Previous investigations of the shear strength of copper pillar solder joints have revealed sensitivities to the thickness of intermetallic compounds in the joint, shear speed, and shear height. [11-30] This study incorporates these variables and extends such investigations to include careful correlations with solder composition, substrate chemistry, solder joint microstructure, and the results of ATC tests with assemblies with the same copper pillar solder joints. EXPERIMENTAL PROCEDURES FABRICATION OF SUBSTRATE AND CHIP (TOP DIE) The substrates were metallized to allow continuous monitoring of the electrical resistivity ofmost of the solder joints. The substratewiring extendedbeyond the chiparea to an array of test probe pads. For both glass and silicon interposer substrates, a0.8µmSiO 2 /0.8µmSiNbilayerwas deposited prior to thewiring deposition. A serpentine line was created by alternating lines between the chips and the substrates. The thin-film structure consisted of two levels of wiring. The first level, on top of the SiN layer, was fabricated by electroplating 3- to 4-µm-thick copper lines, either 15 or 20 µm wide, depending on the copper pillar pitch. Thiswas followedby a dual dielectric layer of 0.6 µm of SiN and 3 µmof photosensitive polybenzoxazole (PBO) dielectric. Then, the top layerwas depositedas either 7 µm of copper or 2 µmof nickel/0.08 µm gold on the substrate or copper pillar structures on the chips. Silicon wafers with copper pillar structures, and both glass and silicon interposer substrate, were fabricated at IBM. Three different interposer materials weremetallized at IBM: two Corning glass interposer substrates (one with a CTE of 3 ppm/K and one with a CTE of 8 ppm/K) and a silicon interposer substrate (wafer). Separate silicon wafers were fabricatedwith copper pillars and solder cap at IBM in a proven, standard lithography process. The pad diameters were 30, 50, or 100 µm, while the pitch was 90, 140, or 190 µm, respectively. Three different copper pillar heights were produced for each pad diameter (15, 25, or 35 µm). Solder compositions of Sn, Sn0.9%Ag, Sn1.8%Ag, and Sn2.4%Ag were deposited at layers of 10 or 20 µm thick (Table 1). Results were correlated with careful characteriza- tions of microstructure and a wide range of parameter values, including copper pillar height, aging time, solder and metallization compositions, and number of reflows. Different assemblies were fabricated, allowing variation of a number of different parameters, including interposer Fig. 1 Comparison of microbump (as background) and copper pillar 30 µm (inset) technologies. Presented on the same scale. The larger aspect ratio of copper pillar bumps allows finer pitches.
Made with FlippingBook
RkJQdWJsaXNoZXIy MjA4MTAy