Chapter 4 Results & Discussion
4.3 Measurement results of high resolution Moiré interferometry
In U field fringe patterns by regular Moiré interferometry, there were only 1-2 fringes in the bump/underfill layer. We could not observe the thermo-mechanical deformation of bumps in detail. The V field fringe patterns by regular Moiré interferometry were similar. Thus, the resolution of regular Moiré interferometry was not enough to measure the thermo-mechanical deformation of solder bumps with 110 μm diameter because the displacement was too small to be resolved. Therefore, a high resolution Moiré interferometry was used. Its spatial resolution was enhanced to 26 nm by phase shifting technology [39]. Such sensitivity would be adequate for resolving the displacement of solder bumps.
Since delaminations or cracks often occurred near die edge, we employed high resolution Moiré interferometry to observe the thermo-mechanical deformation behaviors of the critical region. Figures 4.3 and 4.4 showed the continuous displacement images of U field and V field for the UF-1 package, respectively. The fringes were shifted precisely by the PZT controller. A program developed by Prof.
Paul S. Ho’s group at the University of Texas at Austin was employed to calculate and analysis the displacement and strains of the two assemblies [39]. The U and V field images would be transformed to a phase contour map by this program. The cross-sectional SEM image of the assembly was superimposed onto the phase contour maps with the help of the obvious turns at die/underfill interface of the contour maps for relating with their relative positions. Figures 4.5 (a)-(b) showed the phase contour maps for U field and V field, respectively. Each fringe space was equal to 208 nm for the contour maps. The contour resolution in Fig. 4.6 was enhanced to 52 nm to help us to study the distribution of thermal induced strain in bump/underfill layer. Figures
4.7 – 4.10 were relative images of the UF-2 assembly.
Figure 4.3 The U field continuous displacement images of the package with underfill-1
(a) Ix1 (b) Ix2 (c) Ix3 (d) Ix4
Figure 4.4 The V field continuous displacement images of the package with underfill-1
(a) Iy1 (b) Iy2 (c) Iy3 (d) Iy4
(a)
(b)
Figure 4.5 The phase contour maps of the package with UF-1 (each fringe spacing = 208nm)
(a)
(b)
Figure 4.6 The displacement contour maps of the package with underfill-1 (each contour spacing = 52 nm)
(a) U field (b) V field
Figure 4.7 The U field continuous displacement images of the package with UF-2 underfill material
(a) Ix1 (b) Ix2 (c) Ix3 (d) Ix4
Figure 4.8 The U field continuous displacement images of the package with UF-2 underfill material
(a) Iy1 (b) Iy2 (c) Iy3 (d) Iy4
(a)
(b)
Figure 4.9 The phase contour maps of the package with UF-2 underfill materials (each fringe spacing = 208nm)
(a)
(b)
Figure 4.10 The phase contour maps of the package with underfill-2 (each contour spacing = 52nm)
(a) U field (b) V field
From the U field phase contour map, the undulate contour lines could be observed in underfill layer due to different mechanical properties between bumps and underfill material. These contour maps could transform to the axial displacement data for the two assemblies by the Moiré analysis program. In order to study the thermal deformation at different position of bumps, we extracted displacement data alone 3 lines as: Top (line A), middle (line B) and bottom (line C) of bumps. Figures 4.11 (a) and (b) showed the X and Y direction displacement difference of UF-2 assembly between the 3 lines, respectively. From Fig. 4.11(a), the X displacement alone line A was observed as smallest. Since the silicon die (α=2.6 ppm/℃) had smaller CTE than Sn0.7Cu bumps (α=22 ppm/℃) and UF-2 material (α=27 ppm/℃), the X direction displacement alone line A in bump/underfill layer would be constrained by silicon chip. From Fig. 4.11(b), the Y displacement alone line A had largest Y direction displacement. Since the lateral multi-structure FC-BGA assembly was under bending, the underfill layer and BT substrate would deform easily due to their lower elastic modulus. Therefore, the Y displacement alone line A would be slightly larger than which alone the other two lines.
(a)
Displacement of X direction (μm)
Top (line A) Middle (line B) Bottom (line C)
0.0 0.2 0.4 0.6 0.8 1.0
Displacement of Y direction (μm)
Relative distance of X axis (μm) Top
Displacement of Y direction (μm)
Relative distance of X axis (μm) Top Middle Bottom
(b)
Figure 4.11 (a) X direction (b) Y direction displacement alone the three lines in bump/underfill layer
Figure 4.12 revealed the XY plane shear strain alone line B was higher than wich alone line A and C. The obtained XY plane shear strain data was solved by the Eq.
(3.11), ⎟⎟
γ 1 . Since the underfill/bump layer had higher CTE than
the substrate and silicon chip, the upper and lower side of bump would be confined by chip and substrate. Therefore, the XY plane shear strain alone line B (middle of bump) would produce a larger thermal induced strain while a thermal loading applied in the package.
Enlarged
Figure 4.12 The XY shear strain alone the three lines in bump/underfill layer
Figures 4.13(a) and 4.13(b) depicted the X and Y displacement along the center of bumps for UF-1 and UF-2 assemblies, respectively. The displacement of V field was larger than U field for both assemblies. This implied the assemblies were under bending deformation. From the U field, the displacement curves dropped sharply in the region which the underfill layer did not exist under the silicon chip because underfill materials had larger CTE than silicon chip. When the region was no longer confined by the silicon chip, the underfill would produce a larger thermal displacement in the region. The X direction displacement of UF-1 assembly was less than that of UF-2 assembly, while Y direction displacement showed the opposite trend.
This could be ascribed to UF-2 had higher modulus (969 kg/mm2) than UF-1 (826.5 kg/mm2).
Relative distance of X axis (μm) Top Middle Bottom
(a)
(b)
Figure 4.13 The displacement measurement by high resolution Moiré interferometry (a) X direction (b) Y direction
0.0 0.2 0.4 0.6 0.8 1.0
Displacement of X direction (μm)
Relative distance of X axis (μm)
UF-1 with SnCu
Displacement of Y direction (μm)
Relative distance of X axis (μm)
UF-1 with SnCu UF-2 with SnCu
0.0 0.2 0.4 0.6 0.8 1.0
Relative distance of X axis (μm)
UF-1 with SnCu UF-2 with SnCu
Figure 4.14 showed the shear strain data along the plane of bump center near die edge for the two different underfill materials based on high resolution Moiré measurement. The shear strain obtained from Moiré interferometry was the total strain, i.e., the sum of stress induced strain and thermal strain. The shear strain distributions revealed that the highest shear strain located near the bottom of die edge. This confirmed our previous inferences. The difference between solder bumps and underfill was readily observed. The shear strain increased in solder bump region and decreased in underfill region. This implicated the compliant underfill material mitigated the thermal induced shear stress. Moreover, the outmost solder bump had the highest shear strain compared to other ones. It implied that the outmost solder bump has highest risk of crack or delamination. Since the UF-1 possessed lower modulus than UF-2 material, the bump/underfill would sustain more thermal induced stress. From Fig. 4.14, the larger shear strain of UF-1 was readily observed. This may raise the bump crack potential.
Figure 4.14 The XY plane shear strain results by high resolution Moiré
4.4 Comparison between TCT results and simulation data
The accuracy of the simulation model had been validated by Moiré interferometry measurement results. However, previous simulation model was not a complete FC-BGA package because the assembly had been cut at plane about 12 mm from the edge of frame. This did not correspond with the actual assemblies. Therefore, a quarter symmetric model was created to predict the thermal induced stresses for different packaging samples, and compared simulation results with actual reliability tests.
Temperature cycling test (TCT) under 125 to -55 ℃ for 1000 cycles was carried out for the assemblies with UF-1, UF-2 and Sn0.7Cu solder. For comparative study, two more underfill materials (UF-3 and UF-4) with higher CTE but lower modulus and lower Tg temperature were introduced along with eutectic Sn37Pb, high-lead (Sn95Pb) and Sn0.7Cu solders. In total, 6 FC-PGA package assemblies with different underfill and solder bumps underwent the same TCT process. Table 4.2 summarized the temperature cycling test results for these various samples.
Table 4.2 TCT1000 results for various packaging samples with different underfill materials and solder bumps
Item underfill solder bump bump failure low-K failure
A UF-4 Sn 95Pb failed
Since the bump fracture was caused by shear force mode based on failure analysis, we investigated the failure potential by the XY plane shear stress, σxy, and von Mises stress, σe, at the outmost solder bump. Figure 4.15 showed the stresses distribution of the outmost solder bump in package D. The maximum stresses located at right bottom corner of the out most solder bump and the die edge corner of lay-k layer. This corresponded excellently with the actual crack position shown by SEM viewgraph in Fig. 4.16, kindly provided by UMC corp.
(a)
(b)
Figure 4.16 The SEM picture of sample D
Figures 4.17 and 4.18 showed the maximum XY plane shear stress and von Mises stress of the outmost solder bump for the six packaging samples from simulation results, respectively. Good correlation between simulation and TCT reliability test was found for bumping cracking in sample A (UF-4) with high-lead Sn95Pb solder.
The high-lead alloy (Sn95Pb) had lower elastic modulus (E=2388 kg/mm2) than conventional Sn37Pb eutectic alloy (E=2600 kg/mm2). Thus, high-lead solder bumps may deform easily during reflow process or TCT. Therefore, high-lead packaging assembly needed more rigid underfill material to provide enough protection for solder bumps. Since UF-4 material had lowest Tg temperature and elastic modulus than the other three underfill materials, solder bumps would experience higher stress than the other underfill materials. Based on FEA results, the von Mises stress of the outmost bump in sample A (60.3 kg/mm2) was 23.8% higher than sample B (48.7 kg/mm2).
Thus, bump crack may occur easily in sample A resulting from the higher stress induced during thermal fatigue test.
0
σ xy, XY plane shear stress (Kg/mm2 ) A: UF-4 with Sn95Pb
B: UF-3 with Sn95Pb C: UF-3 with Sn37Pb D: UF-3 with Sn0.7Cu E: UF-2 with Sn0.7Cu F: UF-1 with Sn0.7Cu
Figure 4.17 The max shear stress of the outmost bump
0
σ e, von Mises stress (Kg/mm2 )
A B C D E F
Among three kinds of underfill material (UF-1 to UF-3) under evaluation for lead-free solder bump, only the package with UF-3 underfill material showed failure with bump crack in Sample D. The von Mises stress of the outmost bump in the assembly with UF-3 underfill materials showed the highest stress (56.8 kg/mm2) than the other two underfill materials, because UF-3 material provided more compliant mechanical properties (low Tg and low modulus). However, the UF-3 material was also used in the packaging sample with eutectic solder (Sample C) and high-lead bumps (Sample D). The von Mises stress in the package with eutectic solder was slightly higher than lead-free Sn0.7Cu assembly, but it did not cause any bump failure.
This implied that weaker solder joints strength existed in the lead-free Sn0.7Cu solder bumps in the package even though the modulus of Sn0.7Cu was higher than lead-tin solder. Since lead-free solder introduced new fluxes in the process and high reflow temperature, it was suspected that the residues and byproducts may degrade the adhesion between the die and the underfill material [41]. Furthermore, Sn0.7Cu alloy may form a brittle Cu-Sn inter-metallic compound (IMC) layer at the solder/Cu pads interface. Moreover, the thickness of IMC layer was proportional to reflow temperature and time [42]. The sustained growth of Cu-Sn layer during reflow or reliability temperature cycling test may decrease the interface strength of solder joints.
If there was a crack occurring in this region, it would propagate easily along the IMC layer [43]. Therefore, if lead-free alloy was adopted as the solder bump, we must pay more attention to the bump cracking issue.
The UF-3 material had the lower Tg temperature and lower elastic modulus than UF-1 and UF-2 underfill materials. Thus, UF-3 material could not provide enough protection to solder bumps, especially when the environmental temperature exceeded its Tg temperature. Therefore, underfill material with higher elastic modulus and higher Tg temperature was highly desired for bump protection.
Although there was no low-K delamination observed after TCT 1000 cycles, low-K delamination risks in these 6 samples was still assessed by FEA. For the low-K delamination, σxy and σe did not play an import role. The low-K delamination potential index was changed to be the first principal stress, σ1 [44]. Figure 4.19 showed the maximum σ1 occurred at the corner of low-K layer. Figure 4.20 showed the maximum σ1 in low-K layer for samples A through F. Clearly, sample F (UF-1 with Sn0.7Cu) had the lowest stress than other samples in low-K layer. The σ1 value appeared to increase with decreasing Tg or increasing CTE among samples D, E and F. The effect of underfill materials’ properties such as Tg, CTE and E on the die/package reliability warranted our further examination in the following section.
Figure 4.19 The σ1 distribution of low-K layer for sample D
0 5 10 15 20 25 30 35 40 45
σ 1 , first principal stress (Kg/mm2 )
A B C D E F
A: UF-4 with Sn95Pb B: UF-3 with Sn95Pb C: UF-3 with Sn37Pb D: UF-3 with Sn0.7Cu E: UF-2 with Sn0.7Cu F: UF-1 with Sn0.7Cu
Figure 4.20 The max first principal stress of low-K layer
4.5 The effects of different mechanical properties of underfill
In the aforementioned study, stresses of bump and low-K layer were calculated by finite element method. Moreover, the simulation results correlated well with the thermal reliability test results (bump cracking) for high-lead solder with weaker strength. The higher bump crack risk in the lead-free (Sn0.7Cu) process was believed to be caused by the formation of IMC and poor adhesion due to flux residue.Nevertheless, it is of significant importance to further study how to modify the properties of underfill materials to protect both bump and fragile low-K.
The glass transition temperature (Tg), the coefficient of thermal expansion (CTE), and the elastic modulus (E) were the major thermo-mechanical properties of underfill material which directly influenced on thermal reliability. Most of the underfill materials were composed by epoxy resin and silica filler. Silica filler possessed low CTE and high modulus properties which were opposite to epoxy resin. The quantity and shape of filler determined the CTE and modulus of underfill materials. Therefore, the CTE and modulus were inversely related in the traditional formulation of underfill materials.
However, these two properties can be treated as independent parameters to examine the optimal properties of underfill materials for protecting both the lead-free bumps and low-k fragile layer. The UF-1 and UF-2 materials were a case in point, although the Tg temperatures were not the same. We chose Sample E as a reference to understand how CTE and E may affect the stress distribution in FC-BGA packages by using finite-element analysis. We modified the input data of UF-2 mechanical properties (E1 or CTE1) to study the stress variation of solder bumps and low-K layer.
Figures 4.21 and 4.22 showed the impacts of solder bump and low-K layer stresses
26
underfill material showed opposite trend for solder bump and low-K layer protection.
The modulus of underfill material with high modulus reduced bump stress but increased stress for fragile low-K layer. The CTE of underfill material should be as low as possible for reducing stresses at critical region. However, the minimum CTE of underfill was about 20 ppm/℃ due to more than 70% filler content would cause capillary flow issue [45].
The Tg temperature of underfill material was also a critical parameter for TCT reliability because CTE and modulus changed acutely for temperature above Tg. An optimal underfill material should provide enough protection to solder bumps at higher temperature region and mitigate the stress to fragile low-K layer at lower temperature during thermal cycling test. Based on FEA results, UF-1 with high Tg, moderate E and lower CTE, can simultaneously provide the lowest stress in both bump and low-K layer.
Figure 4.21 The max stress comparison of different E1 of underfill material for sample E
Figure 4.22 The max stress comparison of different CTE1 of underfill material for sample E
26 28 30 32 34 36 38 40 42 44
Stress value (Kg/mm2 )
CTE1= 20 ppm CTE1=27 ppm CTE1=30 ppm Low-k stress,σ
1
Bump stress,σe E1=969 ; E2=11 kg/mm2
4.6 Mechanical properties of underfill material modification
The underfill material can be seen as a two-phase composite. It is often composed by epoxy matrix and silica fillers as shown in Fig. 4.23. J. Qu proposed the more filler loading and smaller filler particle size would result in smaller CTE [46]. In addition, more filler loading also leaded higher effective elastic modulus of underfill material. Therefore, the effective E and CTE of underfill could be modified by adjusting the ratio of filler to epoxy resin.
Figure 4.23 Morphology of conventional underfill system
However, the filler loading has no influence on Tg temperature. Therefore, the chemical structure of epoxy resin needs modification in order to change the Tg temperature of underfill material. Figure 4.24 showed the main chain of epoxy resin chemical structure. The more phenyl groups and double bond attached in the main chain will raise Tg temperature due to their rigidity. Crosslink of epoxy resin can be
C CH3
CH3
OCH2CH CH2 O
HCH2CO H2C
O
also enhanced by addition of curing agent to the epoxy system. Typically, the aromatic and aliphatic curing agent system could be employed to increase degree of crosslink in epoxy system. E. Crawford proposed the Tg temperature will increase with higher crosslink functionality of curing agent and lower molecular weight between cross-links [47]. Therefore, we can create desired mechanical properties of underfill material by changing filler loading and epoxy system modification.
Figure 4.24 The chemical structure of epoxy resin
Chapter 5 Conclusions
As low-K dielectric material and lead-free solder alloy introduced in FC-BGA packages, the underfill material needs to re-design its mechanical properties in order to protect both low-K layer and solder bumps from delamination and crack issues. In this study, the thermo-mechanical deformation of two representative underfill materials were measured and compared by high resolution Moiré interferometry.
Based on the experimental results, underfill material with higher elastic modulus may induce larger die warpage and higher strain of solder bumps.
In addition, a simplified three-dimension model for finite element analysis by ANSYSTM was established. The difference of die warpage between FEA prediction and Moiré interferometry measurement were less than 5%. Moreover, thermal reliability of six kinds of FC-BGA samples by TCT 1000 cycles were studied and compared. The reliability test results illustrated that lead-free and high-lead solder bump had higher crack potential due to the higher reflow temperature and weaker mechanical strength of Sn95Pb alloy, respectively. The thermal reliability test results showed good correlation with simulation data. This finite element model can be used to predict the failure potential of new packages without lots of money and time.
Furthermore, the optimal mechanical properties (E, CTE, and Tg) of underfill material for low-K FC-BGA packages by FEA was evaluated. Based on the simulation results, the lower CTE of underfill can mitigate the thermal induced stress for both bumps and low-K layer. However, it showed opposite elastic modulus requests of underfill for bump and low-K layer protection. The underfill with higher modulus could provide more support to solder bumps, but induced higher stress to low-K layer. The Tg temperature played an critical role in underfill material property due to both the CTE and E would possess a abrupt change while the environmental
temperature exceeded the Tg point. Thus, higher Tg temperature of underfill could result in lower CTE and higher E to protect both low-K layer and bumps at high temperature. Therefore, the mechanical properties of novel underfill material for low-K FC-BGA packages should be moderate E, low CTE and high Tg temperature.
References
[1] B. Halg, IEEE Micro Electro mechanical Systems Workshop, 172 (1990).
[2] L. Shen and K. Zeng, Microelectronic Engineering, 71, 221 (2004).
[3] 鐘文仁, 陳佑任, IC 封裝製程與 CAE 應用, 全華科技圖書公司, 台北市, (2005).
[4] R. R. Tummala, E. J. Rymaszewski, andA. G. Klopfenstein, Microelectronics
[4] R. R. Tummala, E. J. Rymaszewski, andA. G. Klopfenstein, Microelectronics