In-situ study of creep in Sn-3Ag-0.5Cu solder

Abstract

The creep behaviour and microstructural evolution of a Sn-3Ag-0.5Cu wt% sample with a columnar microstructure have been investigated through in-situ creep testing under constant stress of 30 MPa at ~298 K. This is important, as 298 K is high temperature within the solder system and in-situ observations of microstructure evolution confirm the mechanisms involved in deformation and ultimately failure of the material. The sample has been observed in-situ using repeat and automatic forescatter diode and auto electron backscatter diffraction imaging. During deformation, polygonisation and recrystallisation are observed heterogeneously with increasing strain, and these correlate with local lattice rotations near matrix-intermetallic compound interfaces. Recrystallised grains have either twin or special boundary relationships to their parent grains. The combination of these two imaging methods reveal grain 1 (loading direction, LD, 10.4° from [100]) deforms less than the neighbour grain 2 (LD 18.8° from [110]), with slip traces in the strain localised regions. In grain 2, (1$\overline{1}$0)[001] slip system is observed and in grain 1 (1$\overline{1}$0)[$\overline{1}\overline{1}$1]/2 and (110)[$\overline{1}$11]/2 slip systems are observed. Lattice orientation gradients build up with increasing plastic strain and near fracture recrystallisation is observed concurrent with fracture.

Introduction

For economical and sustainable success of electronic components, estimating the lifetime of solder components is critical. Solders are used to electronically connect components and their structural integrity is important in maintaining this connection. Due to the local geometry of the joint, solders can be subjected to creep deformation. Creep deformation is a time-dependent plastic deformation mode where the stress is below the yield point. Excessive creep can lead to component distortion and can lead to fracture of the solder joint, reducing service lifetime. It is known that Pb-free solders are hot working metals ($\frac{\mathrm{T}}{{\mathrm{T}}_{\mathrm{M}}}\ge$ 0.6 [1]) at room temperature, and therefore creep performance is an important material selection parameter for these high temperature applications. Sn-3Ag-0.5Cu wt.% (SAC305) has become the most widely used commercial Pb-free solder alloy in electronics such as servers and computers, and yet the mechanisms of creep failure in these alloys remain uncertain.

In SAC305, three stage creep has been reported [2], [3], [4], [5]–6] to which the slope at the steady stage (stage II) of the curve is referred to as the creep strain rate. In primary creep, the initial deformation takes place rapidly and the creep strain rate decreases because of strain hardening, which occurs due to dislocation multiplication during plastic deformation. Creep reaches the secondary stage with a nearly constant creep strain rate when the strain hardening and dynamic recovery compete and the stress exponent obtained from plotting secondary creep rate with stress indicates that dislocations can move through or around obstacles such as grain boundaries and intermetallic compounds (IMCs) by cross slip, climb and thermally activated slip [7]. Finally, in tertiary stage creep, the creep strain rate increases with significant accumulated damage. An effective reduction in cross-sectional area is often found due to necking and/or internal void formation [4,[8], [9]–10].

Plastic deformation can occur via different slip systems in the β-Sn matrix [11], [12], [13], [14], [15], [16], [17], [18], [19], [20], [21]–22], and the activation of β-Sn slip systems is typically related to the crystal orientation, local stress state, and the ratio of the critical resolved shear stress (τ_CRSS), which modifies the (effective) weighted Schmid factor (m) when determining the likely active slip systems. However, a large variation has been obtained due to the influence of temperature history, composition, and orientation of the solder samples [13], [14],16,23].

There are many families of slip systems that have been identified in Sn, but the relative activity of these different families of slip systems is only partially understood. Zhou et al. [12], Bieler and Telang [23], and Zuo et al. [17] showed that slip in the [001] and [111] directions on and {110} type planes are the most active and are frequently observed in thermomechanical deformation, while activation in the slip of [100] and [101] directions was relatively hard [12]. Bieler and Telang [23] carried out shear lap tests for 0.8 shear deformation at strain rate of 0.1 s⁻¹ using an eutectic Sn-3.5Ag/Cu solder to join two pieces of Cu half-dog-bone at 25°C. They concluded that the (010)[101] slip system was activated and the slip systems in {101}<10$\overline{1}$> and {121}<10$\overline{1}$> families were very difficult to be activated for a sample with its [001] direction near the loading direction. Zuo et al. [17] performed thermomechanical fatigue tests for SAC307 solder joints and discovered that the (110)[$1\overline{1}$1]/2 is the activated slip system. Slip traces were observed in the region with large Cu₆Sn₅, which resisted the development of low angle grain boundaries (LAGBs). In addition, it has been found that the τ_CRSS of the slip systems in the {110}<1$\overline{1}$1>/2 family decrease exponentially with increasing temperature [22] which means that high temperature was required to overcome a certain energy barrier so as to initiate these slip systems. However, Matin et al. [15,16] reported that the slip systems ($\overline{1}$10)[111]/2, (1$\overline{1}$0)[001], (121)[$\overline{1}01$] and (121)[$\overline{1}01$] were activated for Sn-3.8Ag-0.7Cu bulk samples under mechanical [16] and thermomechanical [15] fatigue deformation. Slip traces and microcracks were observed predominantly in the eutectic regions. It is noted that the applied τ_CRSS values for Matin et al. [15,16] are different to Zamiri et al. [12,17,22,23], who consider different crystal orientations for single-crystal samples under uniaxial tensile test from ref [12]. Matin et al. [15] selected ‘consistent’ values of τ_CRSS for families, assuming a constant strain rate at 293 K and for family 1, 2, 9 and 10, these were extracted from their experimental measurements from single-crystal experiments of pure β-Sn (99.99%) performed in ref [16] and family 3 - 5 are estimated values by correlating the slip systems with the atomic line density (Peierls model based). The CRSS ratios are given in Supplementary Table S1.

Furthermore, the dependency of orientation of the solder joint on activation of slip systems is reported in the literature [13,20,23,24]. Zhou et al. [13] stated that thermally cycled SAC305 solder joints with c-axis nearly parallel to the chip-side interface showed more slip activity than samples with c-axis perpendicular to the chip-side interface. They also reported that the recrystallisation generating new grains was caused by gradual lattice rotation, which was observed around the Sn <110> axes and formed as a result of the easier activation of slip systems in the {110}<001> family.

By carrying out a first principle atomistic study and uniaxial tensile tests, Kinoshita et al. [20] and Dong et al. [19] found that samples with an orientation close to [001] had slip in the (101) plane activated easily for pure β-Sn and SAC305. For the [100] and [110] orientations, the (110) and (100) + (010) planes are easier to slip [20].

However, there is not only a very limited number of works supporting the used τ_CRSS values, but also a lack of a systematic study on the τ_CRSS values for β-Sn. This is because most works consider the microstructure within these alloys as a homogenous phase [12,[15], [16], [17],22,23,[25], [26], [27]] and yet we know it contains regions of primary β-Sn and eutectic regions consisting of β-Sn with embedded IMCs.

Some works have studied microstructural evolution of the solder during in-situ creep deformation [2,3,28–32]. The main observations are an increase in surface roughness [2,28], the formation of shear bands [2,30], growth of subgrains [2,3,29–32] and dynamic recrystallisation [2,30] with continuous straining. Tian et al. [30] also reported that the creep deformation of each sample varies due to the different grain orientations (due to the anisotropic properties of β-Sn). Among them only Zhang et al. [3,31,32] has used in-situ EBSD to improve the understanding of dynamic deformation of Sn-Cu and Sn-Ag solder joints during creep tests. They found that at a low strain rate of 1 × 10⁻⁴ s⁻¹, a significant strain concentration was observed for Sn-Ag/Cu solder joints at the Sn-Cu₆Sn₅ interface with the formation of shear bands, while a relatively uniform deformation was introduced with wave-like deformation bands formed for Sn-4Ag/Cu solder joints. The grains deformed with lattice rotation, which depended on their orientation, leading to grain subdivision and grain boundary migration, however, for these tests recrystallisation was not observed [32]. Furthermore, the samples used in their work [3,31,32] have different microstructures, namely orientations and number of grains, which will have a significant effect on the creep behaviour.

In previous work the twin nucleated recrystallisation has been suggested [33–36], however there has been no direct evidence of this provided, as without in-situ observations the mechanisms are unclear. In the work of Arfaei et al. [33], the correlation of solder fatigue with both recrystallisation and intergranular crack growth was observed for SAC305 samples that had undergone in thermal cycling. They observed that recrystallisation started to form in the highly-strained regions such as in the bulk solder near the Si chip-side and close to coarsened precipitates, likely from the spread in orientation around LAGBs and twin boundaries. In the work of Telang et al. [34], they reported continuous recrystallisation in a Sn-3.5Ag joint during thermomechanical fatigue. Based on ex-situ EBSD observations, they found that some of the newly recrystallised orientations were twin orientations to the initial orientations and the {110} plane slip traces were generated simultaneously with increasing recrystallisation. Furthermore, Mattila et al. [35] tested Sn-Ag-Cu solder joints under thermomechanical loading and demonstrated that the twin nucleated recrystallisation can appear preferentially at grain or phase boundaries and near solder / substrate interface to achieve less energy consumption.

To address these complexities, here we perform an in-situ study of creep in controlled SAC305 microstructures, with a focus on when and where grain polygonisation and recrystallisation occur, with respect to grain boundaries and eutectic regions where the β-Sn contains embedded Ag₃Sn and Cu₆Sn₅ IMCs, which can deform differently during creep [37]. An in-situ study also removes ambiguities over features that could develop as the material relaxes after unloading due to room temperature creep and recovery.

In this work, electron microscopy is combined with in-situ deformation. This enables us to follow the evolution of slip and recrystallisation, via changes in lattice orientation. Fast imaging with electron channelling contrast imaging (ECCI) enabled us to confirm that recrystallisation occurred during the deformation process, and EBSD enabled us to understand the relationships of new grains with respect to the parent microstructure.