11 Fundamentals of Solder Alloys in 3D Packaging
339
layers were all consumed after bonding. The Pd layer retards the growth of Ni 3 Sn 4
layer on the substrate side while forms large (Pd, Ni)Sn 4 IMC in the joint [22].
11.4 Microstructure Variation of Microbump Under
Thermal Mechanical Conditions
The thermal mechanical conditions referred herein include the high temperature storage and thermal cycle treatment of the joint. The phase transformation induced by thermal ageing is influential on the final microstructure of the
microbump due to the small solder volume. The Cu/Sn/Cu microbump formed
Cu/Cu 3 Sn/Cu 6 Sn 5 /Sn/Cu 6 Sn 5 /Cu 3 Sn/Cu structure after reflow [3]. The residual Sn
is of limiting volume which may not exist for some joints. Thermal ageing of the
joint at 240 °C gradually converted the Cu 6 Sn 5 into Cu 3 Sn. The long time storage,
3 h, made the microbump a Cu 3 Sn only joint, Fig. 11.5b [3]. A Cu pillar bump with
Sn4Ag0.5Cu solder cap forms scallop Cu 6 Sn 5 IMC in the as fabricated microbump.
The Cu 6 Sn 5 IMC transform from scallop structure to planar structure after ageing
of the microbump at 180 °C. The discontinuous Cu 3 Sn IMC formed between the
Cu 6 Sn 5 and Cu pillar. Voids were formed between Cu 3 Sn and Cu pillar where the
fracture occurs after the shearing test [8].
The high volume fraction of the intermetallic compound in the microbump indicates that the intermetallic compound may be determinant for the performance of
the joint. The thermal fatigue life, assessed by thermal cycle testing, of a microbump
joint was found to be affected by the microstructure and types of the intermetallic
compound. A theoretical simulation indicates that the thermal fatigue life is disproportional to the Young’s modulus of the IMC [7]. However, a combinatory effect
of the modulus and the coefficient of thermal expansion (CTE) seems to actually
determine the fatigue life. The microbump of high volume of IMC may show fatigue
life in the order of Ni 3 Sn 4 (136 GPa, 13.7 ppm) > Cu 6 Sn 5 (124 GPa, 19.0 ppm) >
Cu 3 Sn (143 GPa, 18.2 ppm), the values in the parenthesis are (Young’s Modulus,
CTE) [7]. The volume fraction of the IMC affects the fatigue life of the micro joint.
An increasing IMC thickness or fraction in the micro joint will increase the thermal
fatigue life of the joint. The thermal fatigue life reaches the minimum value at 42.8%
volume fraction of the Ni 3 Sn 4 IMC [7].
The growth of intermetallic compound in the microbump is controlled by the
intrinsic diffusion coefficients of the elements. The diffusion coefficient of element in
dilute solution can be measured with tracer method. However, it is difficult to measure
the diffusion coefficient in concentrate volume such as intermetallic compound with
similar approach. A simulated annealing numerical method was developed for estimating the diffusion coefficients in intermetallic compounds. The intrinsic diffusion
coefficients of the constituent elements in the commonly encountered intermetallic
compounds and the pure metals are listed in Table 11.2. These values indicate that
diffusion of Cu is slightly faster than Sn in Cu 3 Sn, while of similar speed in Cu 6 Sn 5 .
339
layers were all consumed after bonding. The Pd layer retards the growth of Ni 3 Sn 4
layer on the substrate side while forms large (Pd, Ni)Sn 4 IMC in the joint [22].
11.4 Microstructure Variation of Microbump Under
Thermal Mechanical Conditions
The thermal mechanical conditions referred herein include the high temperature storage and thermal cycle treatment of the joint. The phase transformation induced by thermal ageing is influential on the final microstructure of the
microbump due to the small solder volume. The Cu/Sn/Cu microbump formed
Cu/Cu 3 Sn/Cu 6 Sn 5 /Sn/Cu 6 Sn 5 /Cu 3 Sn/Cu structure after reflow [3]. The residual Sn
is of limiting volume which may not exist for some joints. Thermal ageing of the
joint at 240 °C gradually converted the Cu 6 Sn 5 into Cu 3 Sn. The long time storage,
3 h, made the microbump a Cu 3 Sn only joint, Fig. 11.5b [3]. A Cu pillar bump with
Sn4Ag0.5Cu solder cap forms scallop Cu 6 Sn 5 IMC in the as fabricated microbump.
The Cu 6 Sn 5 IMC transform from scallop structure to planar structure after ageing
of the microbump at 180 °C. The discontinuous Cu 3 Sn IMC formed between the
Cu 6 Sn 5 and Cu pillar. Voids were formed between Cu 3 Sn and Cu pillar where the
fracture occurs after the shearing test [8].
The high volume fraction of the intermetallic compound in the microbump indicates that the intermetallic compound may be determinant for the performance of
the joint. The thermal fatigue life, assessed by thermal cycle testing, of a microbump
joint was found to be affected by the microstructure and types of the intermetallic
compound. A theoretical simulation indicates that the thermal fatigue life is disproportional to the Young’s modulus of the IMC [7]. However, a combinatory effect
of the modulus and the coefficient of thermal expansion (CTE) seems to actually
determine the fatigue life. The microbump of high volume of IMC may show fatigue
life in the order of Ni 3 Sn 4 (136 GPa, 13.7 ppm) > Cu 6 Sn 5 (124 GPa, 19.0 ppm) >
Cu 3 Sn (143 GPa, 18.2 ppm), the values in the parenthesis are (Young’s Modulus,
CTE) [7]. The volume fraction of the IMC affects the fatigue life of the micro joint.
An increasing IMC thickness or fraction in the micro joint will increase the thermal
fatigue life of the joint. The thermal fatigue life reaches the minimum value at 42.8%
volume fraction of the Ni 3 Sn 4 IMC [7].
The growth of intermetallic compound in the microbump is controlled by the
intrinsic diffusion coefficients of the elements. The diffusion coefficient of element in
dilute solution can be measured with tracer method. However, it is difficult to measure
the diffusion coefficient in concentrate volume such as intermetallic compound with
similar approach. A simulated annealing numerical method was developed for estimating the diffusion coefficients in intermetallic compounds. The intrinsic diffusion
coefficients of the constituent elements in the commonly encountered intermetallic
compounds and the pure metals are listed in Table 11.2. These values indicate that
diffusion of Cu is slightly faster than Sn in Cu 3 Sn, while of similar speed in Cu 6 Sn 5 .