| Summary: | 博士 === 國立清華大學 === 材料科學工程學系 === 87 === The development of the microstructure at the solder/metallization on solder/substrate interface significantly affects the mechanical properties of the joint assembly, and the degradation of the mechanical properties of the solder joint is caused by the formation of the intermetallic compound (IMC) during both the soldering and aging processes. Investigation concerning the growth kinetics, thermodynamics, and phase transformation related to the IMC, therefore, dominate the critical evaluation of the solder joint reliability in the microelectronic packaging.
In the first phase of this research, a joint assembly of lead-free solder/intermetallic layers/copper was prepared by hot-dipped solder coated on a copper substrate and then by thermal aging at 100, 125, 150, and 170℃ for 50, 100, 200, and 600 hours, respectively. Results of interfacial morphologies and concentration profiles on the solder/copper joint were presented. OM and SEM were used to measure the thickness of intermetallic layers and then to illucidate the development of microstructure at the joint assembly. The phases of intermetallic compound were identified to be Cu3Sn and Cu6Sn5 by both X-ray mapping in EPMA, and X-ray diffraction. The intermetallic layers, subtracted from the initial thickness formed by hot dipping, showed a linear dependence of square root of aging time at various aging temperature. The diffusion coefficients of intermetallic compounds are estimated by Arrhenius equation, and the pre-exponential terms of Cu3Sn layer and Cu6Sn5 layer are 7.10×10-7 cm2/sec and 6.1×10-3 cm2/sec, respectively. The associated activation energies of Cu3Sn layer and Cu6Sn5 layer are 57.03 KJ/mol and 83.76 KJ/mol, respectively. A model of diffusion-controlled mechanism is used to fit the concentration profiles of the joint assembly, and exhibits a fairly good quantitative agreement with the measured data. The initial thickness formed as soldering is also taken into account to evaluate the apparent thickness by introducing a term of correcting aging time.
The phase transformation of the joint assembly was compared to the phase diagram of binary alloy, Cu - Sn, and it showed an agreement with the resultant intermetallic phases formed between the pure tin and pure copper. Two theoretical models proposed by Gosele and Shatynski were developed and then employed to characterize the assembly. The Gosele''s model was used to predict whether the intermetallic layers grew or shrank during aging, while the Shatynski''s model was employed to estimate the related reactive thicknesses and hence the ratios of the interdiffusivities in the joint assembly. After a series of calculations, the Gosele''s model predicted that Cu6Sn5 and Cu3Sn intermetallic layers became thicker. Evaluation of intermetallic interdiffusivities was also proven to approach theoretical ones from the Shatynski''s model.
In addition, X-ray color mappings by electron probe microanalyzer (EPMA) of copper and tin were also applied to study the concentration variations near the interfaces in the joint assembly. According to the intensities of Cu and Sn, collected by color mapping, a developed software was employed to construct series of statistical graphs, and the detailed concentration profiles at the interfaces of the assembly were investigated from these graphs. Two important results are derived. The first is that analysis of interfacial concentration profile exhibits the phases of Cu3Sn-rich, Cu6Sn5-rich and tin-rich, which match with boundaries of solder/Cu6Sn5, Cu6Sn5/ Cu3Sn and Cu3Sn/copper, respectively. The second is that the semi-quantitative measurement with a peak-fitting model employed suffices to evaluate the interfacial concentration profiles with a statistical variation less than 5 mol %.
A MLC/Sn-Ag-Zn/Pt-Ag metallization joint assembly aged at 150 ℃ was then prepared and investigated with electron microscope. Both secondary electron image (SEI) and backscattered electron image (BEI) were employed to observe the intermetallic layers, Ag3Sn, which were formed between solder and metallization with an irregular shape. The thickness profile of intermetallic layer showed a typical normal distribution by the employment of Person''s goodness-of-fit. The mean of intermetallic thickness was obtained by transformation from one-dimensional linear length measurement to two-dimensional area measurement with an image processor. A criterion of which deviation of average thickness from the mean is less than 0.5% was established as the reference to estimate the relative error of other measurements with different measuring times. With the increasing measuring times, both the relative error to mean for average thickness and relative deviation to the standard deviation of the criterion were decreased with an exponential decay function. In addition, a quantitative measurement method to reduce the measuring error was demonstrated by the exponential function.
Finally, the kinetics of the IMC growth was investigated with the joint assemblies of Sn-Ag-1Zn/Pt-Ag, Sn-Ag/PtAg, Sn-Ag-1Zn/Cu, or Sn-Ag/Cu experienced thermal aging at 1500C. The morphologies of these joint assemblies were investigated by electron microscope. Related kinetics models were derived to cope with the microstructure development of the intermetallic compound (IMC) growth. After modified fitting with the derived model, the measured IMC thicknesses exhibited good regression with theoretically modeled curves. The derived kinetics for various joint assemblies were nonplanar-diffusion-controlled for Sn-Ag-1Zn/Pt-Ag, grain boundary diffusion controlled for Sn-Ag/PtAg, planar-diffusion-controlled for Sn-Ag-1Zn/Cu, and grain boundary diffusion controlled for both Sn-Ag/Pt-Ag and Sn-Ag/Cu, respectively. The obtained interdiffusivity of IMC for the joint assemblies of Sn-Ag-1Zn/Pt-Ag and Sn-Ag-1Zn/Cu were 2.76’10-12 cm2/sec for DAg3Sn and 1.12’10-12 cm2/sec for DCu6Sn5, respectively. Correlations of the derived model in various joint assembly with the morphological evolution in the microstructure investigation of the interfaces are discussed. Particularly the time dependence of IMC growth thickness is probed and compared with those in literature. In addition, a detailed precision measurement of IMC thickness is proposed.
|
|---|
