INTRODUCTION
Tin rich lead-free solders are currently the predominant interconnection material for the microelectronics industry. The resulting interconnects provides both electrical and mechanical connection between the integrated circuit devices and their substrate [ref]. The emerging trends in miniaturization of electronic devices demands significant reduction in the size (pitch) of interconnects. As a consequence the solder’s reliability is of paramount importance that decides the integrity of the device. In the last decade, researchers have demonstrated various approaches by which the properties of the solder can be improved [ref]. Among them the composite approach where in suitable particles are introduced into the solders matrix seem to be one of the potential solution to engineer a stabilized microstructure with improved mechanical and thermomechanical properties.
The mechanical behavior of the solder joint is influenced by factors like geometry and plastic constraints [ref]. Therefore constraining effects on the mechanical response of the solder joint has to be removed so as to determine the constitutive properties of the solder material in most realistic conditions. Additionally these constitutive properties are of utmost importance for realisticmodeling/simulation of advanced technology like System-On-Chip or stacked 3D packaging of electronics devices [ref].
The overall objective of the current study is to investigate and understand the effect of addingmicrometer sized metallic particles to the solder alloy on its subsequent microstructural evolution and constitutive response of the solder. To address current objective composite solder prepared with Ni particles were compared with the baseline un-reinforced solder alloy.
2 MATERIALS AND EXPERIMENTAL PROCEDURE
Two bulk solder specimens were prepared for the current study: one with a reference and one with composite solder. The lead-free solder alloy Sn-4.0 wt% Ag-0.5 wt% Cu (SAC405-Alpha metals) in the form of solder paste was used for the reference specimen and as the matrix material for the composite solder. The composite solder as prepared by adding 4.2 vol% micrometer scale commercial Ni particles (3-7 µm, 99.9% purity, Leico industries) to the solder paste with a simple mechanical mixing method [ref]. Bulk specimens were fabricated by reflowing the prepared solder in a cylindrical aluminium nitride crucible. The amount of solder paste used for each specimen was fixed. Reflow soldering was conducted in a Zelflow RO4 oven (LPKF Electronics) with a temperature profile consisting of 170 °C preheating stage followed by peak temperature of 240 °C. The bulk specimens were reflowed individually and removed from the oven to cool in air. The reflowed specimens were degreased with ethanol, rinsed in deionised water, blown dry and finally cold embedded in epoxy resin for further metallographic analysis. The embedded specimens were polished to a 0.1 µm finish with suspension of alumina and etched with 5 % HCl in methanol to highlight the intermetallic phases. Optical and scanning electron microscopy (SEM) with energy dispersive spectroscopy (EDS) was used to characterize the intermetallic phases and provide micrographs for subsequent image analysis.
The micrographs of the polished and etched bulk specimens were analyzed to estimate the volume fraction of Ni and Ni containing intermetallic phases. An average volume fraction was calculated from 9 measurements conducted on 3 identical composite specimens.
The mechanical test specimens were produced in the following way. Cu plates were machined to produce symmetrical substrates of dimensions 60 x 20 x 2 mm. A single shear lap joint (Fig.1.) was then produced by affixing the substrates in a custom soldering jig that fixes the gap width to 1 mm. The solder paste prepared as previously described was used as filler material between the fixed substrates. The solder was reflowed, and joints were produced, by placing the entire soldering assembly (jig with affixed substrates and solder) in the reflow oven and subjecting it to an identical temperature profile as the bulk specimens prepared for microstructural analysis. Once removed from the jig, the test specimens were ground carefully to remove any excess solder. The nominal thickness of the joining layer was measured to be ~ 1.98 mm. A hole of 2 mm diameter was drilled at both the edges of joining layer. The curvature produced at the edges of the joining layer prevents premature failure at the interface (substrate/solder) during loading. The chosen geometry for the joining layer was found optimized by Finite element modeling, for stress uniformity ensuring uniform plastic strain at the center of the joining layer, Fig.1. A total of 22 test specimens were produced 12 with reference and 10 with composite solder.
Prior testing, microstructural analysis on joint specimens similar to that conducted on bulk specimens were performed to analyze for intermetallic layer thickness at the substrate/solder interface. At least two measurements were made for each specimen and condition.
All testing was performed with an Instron 5848 Micro Tester with a loading profile consisting ( ). A sequence of high resolution images of the joints was recorded during the test through an optical microscope at magnifications of 24 to 48 X. The recorded sequences were then processed in custom developed Digital Image Correlation (DIC) software to calculate the evolution of the average strain in the solder joint during the test [ref]. This technique provides for strain resolution in the joint of approximately 0.02 %. The fractured surface of the tested joint specimens were analyzed with SEM to provide micrographs for fractographic analysis.
3 RESULTS
3.1 Microstructural analysis
Representative microstructures from the bulk solder specimens are as presented in the Fig. 2-4 respectively. Here after the two specimens are referred to as follows: unreinforced reference alloy (SAC405) and composite solder prepared with Ni particles (SAC405+Ni). The SAC405 specimen contained three phases upon solidification: a matrix and two second phase particles, Fig.2. EDS analysis of these phases yielded compositions in weight percent of Sn-4.0 Ag, Sn-73 Ag and Sn-40 Cu surmised to be (Sn), Ag3Sn and Cu6Sn5, respectively [ref]. The Ag3Sn was organized as both network of fine needle-like phase and as few larger platelets extending upto tens of micrometers within the (Sn) matrix. The Cu6Sn5 was observed as a small fraction of fine particles within the matrix
The bulk microstructure of a SAC405+Ni composite specimen is presented in Fig. 3. The mechanically introduced Ni particles demonstrate a reasonably uniform distribution in the solder matrix. Further, the Ni particles have reacted with the solder alloy to form a surrounding reaction layer, Fig. 4. Results of a typical EDS line scan performed, as indicated by the dotted line in Fig.4 was plotted between elemental composition (weight %) versus scan distance (µm), Fig.5. The EDS line scan plot (Fig.5.) indicate an unreacted core of Ni particle surrounded by a reaction layer. Further the phase composition of the reaction layer corresponds closely to Ni3Sn4 intermetallic phase occurring at 73 wt % Sn [ref], considering that the composition of Cu innate in the alloy is considerably low (0.5 wt %). The average intermetallic layer thickness around the Ni particles was roughly 3.5 µm and is characterized by a “sunburst” morphology, an observation concurrent with similar studies [ref]. The randomly distributed Ag3Sn particles were found to be relatively coarser in size as compared to that observed for the SAC405 specimen. A few larger Ag3Sn platelets were also observed, very similar to that seen for the SAC405 specimen. Micro pores up to 5 µm in diameter were observed and unique to this specimen.
Average volume fraction of the reinforcing phases produced, due to 4.2 vol % Ni particle addition was calculated to be 13 vol %. The calculated value for volume fraction is in the expected range dictated by mass balance equation: ~ 16 vol % (for Ni3Sn4 phase occurring at 73 wt % Sn).
3.2 Mechanical characterization of the solder joint
The stress-strain responses of the shear joint specimens are presented in Fig. 6. The grey solid curve and the black bulleted curve represent the average stress-strain response of joint specimens produced with SAC405 and SAC405+Ni solder respectively. Average stress-strain curve was calculated from atleast 8 identical joint specimens for each condition. The gray shading represents the standard deviation in the experimental data. Considering an experimental error of (± 5 %) due to scatter in the test data, the ultimate shear stress of joints produced with SAC405+Ni composite solder (32.5 MPa) had a distinct increase of ~25 % over the joints produced with reference SAC405 solder (25.3 MPa).
3.3 Solder joint deformation
4 ANALYSIS – MODELING
4.1 Identification of constitutive properties of the solder
The measured load-displacement response of the joint specimens was utilized as input for the custom developed inverse identification procedure [ref]. The inverse identification procedure comprises of an identification loop which determines the constitutive properties of the solder material itself. The identification loop iteratively compares the simulated results obtained from the parametric Finite element (FE) model of solder joint with the apparent engineering stress-strain response of the joint (Fig.6.) obtained experimentally. The copper substrates have been considered as purely elastic and the solder has been modeled as an elasto-plastic material. The hardening law corresponding to the FE model is given by relation: σy = σy0 + Q∞(1-e-bε) where σy,σy0, Q∞,brepresents yield stress, intial yield stress,the asymptotic exponential hardening stress and exponential hardening rate respectively. The constitutive stress-strain response of the solder specimen (SAC405 and SAC405+Ni) plotted between equivalent stress versus equivalent strain (considering Von Mises yield criterion) is presented in Fig. 7. The grey (SAC405) and black (SAC405+Ni) solid curves represent the constitutive stress-strain response obtained from experimental results presented in Fig.6 through inverse identification procedure. Further, Table 1 summarizes the identified constitutive properties for the solder specimens. The identified ultimate stress of the SAC405+Ni composite specimen was calculated to be 25 % higher than the reference SAC405 specimen.
4.2 Modeling the constitutive response of the composite solder
A novel mechanical model based on Homogenization technique was employed [ref] to understand the influence of adding Ni particles on the mechanical response of the composite solder. The majority of the models reported in the literature are planar [2-4]. Though they capture some physical aspects of the material properties in regard to composites with randomly distributed particles, they are not able to predict the effective behavior of the composite [5, 6]. Plane stress models tend to underestimate the strengthening effect of the particles whereas plane strain models tend to overestimate it. In the present work a three dimensional unit cell homogenization model is therefore adopted [7]. This model will help to simulate and understand the constitutive stress-strain response of the composite solder based on the volume fraction of additional phases caused from particle addition.
In straightforward homogenization problems, the properties of the constituents are given and the unknowns are the composite properties. The Ni particles are assigned a linear elastic behavior while the matrix (SAC405) is assumed to be elastoplastic and to follow the von Mises yield criterion. The inputs for the model are: (i) the constitutive stress-strain curve obtained for the (SAC405) matrix presented in Fig. 7. (ii) the calculated average volume fraction of the reinforcing phases produced due to Ni particle addition (13 vol. %) and (iii) the mechanical properties, Young’s Modulus (E) and Poisson ratio (ע)of the reinforcing particles. EDS analysis revealed that these particles are constituted partly of Ni3Sn4 intermetallics and some unreacted Ni particles (Fig. 5.). Therefore two extreme cases for particle composition were considered: (i) when 100% Ni (E=214 GPa, ע=0.3) and (ii) 100% NiSn (E=133.3 GPa, ע=0.33) intermetallic particles [ref].
4.3 Generation of the Model
The cubic domain is discretized by the Constrained Delaunay Tetrahedralization (CDT), a variation of the Delaunay tetrahedralization which respects the domain boundary (the CDT is carried out by the software TetGen [8] which ensures quality mesh generation). The particles are then generated by assigning the inclusion material properties to the tetrahedra sharing a randomly chosen node. Typically a particle consists of about 24-26 tetrahedra (12-13, in case of particles generated from nodes lying on the cube faces) and exhibits an almost convex shape. The mesh can be further refined in a second step by carrying out a CDT which respects the particle boundary.
To assess the size of the Representative Volume Element (RVE) for the elastoplastic behavior the criterion proposed in [7] is adopted. Due to the presence of particles the stress-strain curve of the composite is expected to present a larger hardening than that of the matrix material. This effect can be described by following the approach proposed in [9], in which the local stress field is split into two contributions: the stress which would occur if the constituents were elastic and a self equilibrated residual stress field due to the plastic strain accumulated in the material:
(1)
Where is the elastic stress localization tensor, the macroscopic stress and the residual stress.
The average elastic energy in the material can be expressed as the sum of i the average macroscopic elastic energy and the average micro-stored elastic energy due to the residual stresses
(2)
According to equation (2) the unit cell is a RVE when the steadiness of the average elastic energy density with respect to the cell size is achieved (in particular the steadiness of is critical since it converges slower). Further, two contributions to the average plastic work, at the macroscopic level are identified: the average dissipated energy and the average elastic energy due to residual stresses , which is stored in the material and contributes to the macroscopic hardening [ref].
(3)
The application of the criterion to the present case leads to a unit cell containing 121 particles. Perfect adhesion between particles and matrix is assumed. Formation of Ni3Sn4 intermetallics around Ni particles supporting the previous assumption (Fig. 4). The finite element model is solved with ABAQUS® software package [10] and modified 10-node tetrahedral elements are employed. The mesh, (Fig. 7.) consists of 255802 nodes and 167818 elements. Pure tension simulations were carried out under mixed static-kinematic boundary conditions: the displacements are imposed at the nodes on the tensile faces and no load is imposed on the unit cell faces parallel to the tensile direction. The model predicted stress-strain curve for the composite (SAC405+Ni) almost coincide irrespective of the two extreme cases considered. Therefore the model predicted stress-strain curve for 100% Ni particles is reported (Fig. 7.). The three bulleted curves (Fig.7.) represent the constitutive stress-strain response of the SAC405+Ni solder by pure tension simulations of the model in X, Y, Z directions. Further themodel predicted stress-strain response of SAC405+Ni are compared with experimentally identified constitutive stress-strain curves (Fig. 7.).
5 DISCUSSION
The increase in the ultimate stress for the composite SAC405+Ni (~ 22 %) over the reference (SAC405) specimen may be rationalized by considering their microstructure (Fig. 2-5). The presence of additional Ni and Ni3Sn4 intermetallic phases unlike in SAC405 specimen can be attributed for this strengthening effect. In addition the reasonably uniform distribution of the Ni particles (Fig.3.) in the solder matrix could possibly add to further strengthening by distributing the stress in a more uniform manner resulting in homogenous deformation of the solder joint. Concurrent studies have also observed increase in strength upon adding metallic/ceramicreinforcements [ref]. The strengthening effect by particle addition is governed by two main mechanisms as suggested in the literature: (i) restriction of dislocation movement (ii) pinning of grain boundaries [ref]. In the current study an increase in the apparent hardening for the composite SAC405+Ni over the un-reinforced counterpart is discernable (Table 1). The relatively strong interfacial bonding between Ni3Sn4 particles [ref] and the solder matrix, compared to Cu [ref] or select inert reinforcements,can relatively withstandhigher stresses. This stress gathering capability of hard brittle Ni/Ni3Sn4reinforcement restricts the plastic deformation of the ductile solder matrix, consequently resulting in loss of ductility accompanied by increase in overall strength [ref]. The fractographic observations (Fig) ascertaining the previous suggested mechanism for the strengthening effect; absence of dimples revealing a more brittle fracture behavior for the composite SAC405+Ni joint specimens in contrast to a ductile fracture observed for the SAC405 joint specimens. Further, the failure mechanism of the solders’ deformation can be correlated with the images taken during DIC strain measurement (Fig.). Initial porosity % seems to play a crucial role on the deformation of the solder joint. The scatter in test data (5 %) can be associated with the porosity in the joint specimens. Failure mechanism for the reference solder seems to be primarily plastic deformation accompanied by micro voidal growth. In the case of the composite solder more hardening of the solder matrix is prominent as the reinforcements act as stress concentrations leading to more brittle fracture.
The adopted model (Section 4.2) is utilized as a tool to understand and predict the elasto-plastic response of the composite solder as a function of % of reinforcing phases with its associated mechanical properties. The model predictions are in very good agreement with experiments for values of the effective total strain larger than 1 %, with the model curve exhibiting an ultimate stress of about 51 MPa (Fig.7.). The experimental setup provides reliable results for strain values above 1 %, and this explains the reason for experimental stress values in the region of the onset of plasticity are lower than those predicted by the model. In the plastic regime, however the model predicted curve is slightly lower than experimental curve particularly in regard to the ultimate stress. This could be accounted for the error in the experimental test data. The model predicted curves for the composite solder (DX, DY, and DZ) are nearly superimposed exhibiting cubic symmetry of the model with respect to the loading direction.
6 CONCLUSIONS
FIGURE CAPTIONS
Figure 1. Schematic of mechanical test specimen and loading direction
Figure 2. Back scatter SEM image of bulk SAC405 specimen
Figure 3. Back scatter SEM image of bulk (SAC405+Ni) specimen at low magnification
Figure 4. Back scatter SEM image of bulk (SAC405+Ni) specimen at high magnification