Packaging Effect on Reliability for Cu/Low k Damascene Structures* Guotao Wang and Paul S. Ho Laboratory of Interconnect & Packaging, TX 78712 * Work supported by SRC through the CAIST Program TRC 2003
1. Introduction Packaging requirements and processes Packaging impact on reliability for Cu interconnect 2. Multilevel sub-modeling technique 3D Modeling with Modified Virtual Crack Closure (MVCC) method 3. Simulation results Results for TEOS and low k structures Line dimension scaling effect Parametric study for package level components - Material and size effect for underfilled packages - Effect of solder reflow without underfill in plastic packages 4. Recent fracture studies on low k structures
Single-chip Packaging Technology Requirements Year of Production 2001 2002 2003 2004 2005 2006 2007 Chip Size (mm 2 ) Hand-held 57 59 61 63 65 65 65 High-performance 310 310 310 310 310 310 310 Power: Single Chip Package (Watts) Hand-held 2.4 2.6 2.8 3.2 3.2 3.5 3.5 High-performance 130 140 150 160 170 180 190 Package Pin Count Maximum Hand-held 100-420 112-464 122-508 134-560 144-616 160-680 176-748 High-performance 1700 1870 2057 2263 2489 2738 3012 Sub-threshold Leakage Current, I sd,leak (@25C) Nominal low-power NMOS (pa/μm) 100 100 100 300 300 300 700 Nominal high-performance NMOS (μa/μm) 0.01 0.03 0.07 0.1 0.3 0.7 1 Int. Semicond. Tech. Roadmap Update 2002
Flip-Chip Packaging Die Solder bumping Substrate Solder reflow 183ºC 37Pb/63Sn High stress in the package during cooling down to room temp. Underfilling 125º to 180ºC Thermal cycling -55 o C-125 o C High stress can be introduced during thermal cycling.
Wire-bond Packaging Die Substrate Die attach 150º to 300ºC High stress at bond pad during wire bonding. Wire bonding Encapsulation 140º to 160ºC
Oxide low k Low k dielectrics have Weak mechanical strength Large coefficient of thermal expansion Poor adhesion to cap and barrier layers Interfacial delamination is a serious reliability concern for Cu/low k interconnects ~10mm Active Side Silicon Chip Passivation Metal Dielectric Device Level Interfacial delamination
Interfacial Delamination Interfacial delamination The crack driving force can be evaluated from the energy release rate (ERR) using finite element analysis. For a stand-alone Cu/low k interconnects, ERR has been calculated to be about 1 J/m 2 during cooling from 400 o C to room temperature. The fracture energy, or the critical ERR, is usually about 4-5 J/m 2 for low k interfaces (lower for porous ILD) and considerably higher for TEOS interfaces. (Y. Du et al., Proceedings of ECTC, 2002) Sub-critical crack growth instead of critical fracture will be more of concern for stand-alone Cu/low k structures
Solder bump Die BT substrate Underfill Solder bump Underfill BT substrate Through hole via Die High density signal layer Material mismatch and processing can induce thermal deformation and stress concentration in flip-chip packages: Packaging effect on low k interconnect reliability?
Verification with Moiré Interferometry 0 Package cross-section U Field Displacement (um) -1-2 -3-4 -5-6 Moire result FEA result 0 0.5 1 1.5 2 2.5 3 3.5 4 Distance from neutral point (mm) Package warpage V Field High resolution moiré interferometry was used to measure the thermal deformation in the flip-chip package and verified the modeling results at the package level.
High resolution U phase map (208nm per fringe) U Thermal load 20 C to 102 C
High resolution V phase map (208nm per fringe) V Thermal load 20 C to 102 C
Displacement distribution (U field, 52nm per contour) U Thermal load 20 C to 102 C
Displacement distribution (V field, 52nm per contour) V Thermal load 20 C to 102 C
Shear strain γ xy distribution at the solder bump A B C
Normal strain ε y distribution at the solder bump A B C
Impact from Packaging For a chip attached to a board, Large thermal stresses are induced due to thermal mismatch between package level components. In a plastic underfilled flip-chip package, Large peeling stress at chip/solder bump or chip/underfill interface due to thermal mismatch between solder and underfill. Large shear and tensile stress at die corner due to thermal mismatch between die and PCB. (M.R. Miller et al., Proceedings ECTC, p.979, 1999) Effect of thermal stress induced by packaging on structural reliability of Cu/low k interconnects?
Finite Element Analysis To study packaging impact on interconnect reliability In FEA model, details for both packaging and interconnect levels have to be considered. Modeling challenges Maximum dimension at package level: 10 to 30mm Minimum dimension at interconnect level: 20nm or less for barrier layer The ratio can be as high as 10 6. It is impossible to consider the details of interconnect structures when modeling a whole package. Solution: multilevel sub-modeling technique
Multilevel Sub-model A 4-level 3D sub-model was developed to analyze Cu interconnects in flip-chip packages Starting from the packaging level, sub-modeling was conducted one level of detail at a time to reach the interconnect level. ANSYS built-in cut boundary technique was used at each sub-modeling level. At the final interconnect level, a crack with fixed length was introduced at various relevant interfaces. A Modified Virtual Crack Closure (MVCC) method was used to calculate the crack driving force (energy release rate).
Hierarchical Levels of Submodeling Level 1: Package level Level 2: Critical Solder Region Level 3: Die-Solder Interface Level 4: Detailed Interconnect
Level 1: Flip-chip Package Die Underfill PCB PCB Die At the package level, a quarter section of the package was modeled based on symmetry. Details of the interconnect structure was not considered because of its small dimension.
Verification with Moiré Interferometry 0 Package cross-section U Field Displacement (um) -1-2 -3-4 -5-6 Moire result FEA result 0 0.5 1 1.5 2 2.5 3 3.5 4 Distance from neutral point (mm) Package warpage V Field High resolution moiré interferometry was used to measure the thermal deformation in the flip-chip package and verified the modeling results at the package level.
Level 2: Critical Solder Region Die Critical solder bump Underfill PCB With underfill shown Without underfill shown The sub-model focused on the critical solder bump region with a uniform ILD layer at the die surface but no detailed interconnect structure included.
Level 3: Die-Solder Interface Die (Si) BPS G ILD PASS Solder pad This sub-model focused on the die-solder interface region containing a portion of die, ILD layer and a portion of solder bump but included only a uniform ILD layer for the interconnect.
Level 4: Detailed Interconnects Si Metal Line BPSG Metal 1 ILD PASS Solder pad Metal 2 This sub-model focused on the die-solder interface taking into account the detailed interconnect structure. A crack with fixed length was introduced along various interfaces to calculate the crack driving force.
z y x MVCC Technique (Modified Virtual Crack Closure) A 2 1 3 A FEA elements and nodes near crack tip (2) δ z 2 G 3 (1) F z 1 (1) F z Mode 1 component I = F δ (1) z (2) z /(2 A) (2) δ x 2 3 (1) F x 1 (1) F x Mode 2 component G II = F δ (1) x (2) x /(2 A) G (2) δ y Total energy release rate: 2 (1) F y 1 (1) F y 3 Mode 3 component III = F δ (1) z (2) z /(2 A) G = G + G + I II F X, F y and F z are nodal forces at node 1 along x,y and z direction, respectively. δ X, δ y and δ z are relative displacements between node 2 and 3 along x,y and z direction, respectively. G III
Interconnect Interfaces BPSG M1 Via ILD Crack 6 Crack 5 Crack 1 Crack 4 PASS M2 Solder pad Crack 4, 5 and 6 are at the horizontal cap and barrier layer interfaces. Crack width is taken to be the line width.
Interconnect Interfaces (cont d) BPSG ILD ILD Metal 1 Crack 3 Crack 2 Barrier TiN Crack 2 and 3 are at the vertical barrier interfaces.
Material Properties Materials Al Cu TEOS SiLK Underfill PCB E ( GPa ) 72 122 66 2.45 6.23 Anisotropic ν 0.36 0.35 0.18 0.35 0.40 elastic property CTE ( 10-6 /C) 24 17 0.57 66 40.6 16 (in plane) 84 (out of plane) Thermal loading: for stand-alone wafer structure from 400 o C to 25 o C and for packaging from 55 o C to 125 o C. All materials are taken to be linear elastic.
ERR (J/m^2) 1.2 1 0.8 0.6 0.4 ERR for Stand alone Wafer Structures (from 400 o C to 25 o C) Al/TEOS Structure Cu/TEOS Structure Cu/SiLK Structure 0.2 0 crack 1 crack 2 crack 3 crack 4 crack 5 crack 6 The SiLK/barrier interface in Cu/SiLK structure has the highest energy release rate (about 1.16 J/m 2 ). Fracture mode is primarily mode I driven by the high CTE of SiLK.
Packaging effect (-55 o C to 125 o C) ERR (J/m^2) 18 15 12 9 6 Al/TEOS Structure Cu/TEOS Structure Cu/SiLK Structure 3 0 crack 1 crack 2 crack 3 crack 4 crack 5 crack 6 Packaging has little effect on energy release rate for Al/TEOS or Cu/TEOS structure, but is significant for Cu/SiLK structure. Mixedmode delamination with both peeling and shear stresses contributing.
Why energy release rate is much higher in Cu/low k structure than in Cu/TEOS structure? Assuming that the thermal stress induced from package level deformation is σ, the strains for low k ILD and TEOS will be σ σ ε SiLK =, εteos = for simple 1-D case E E SiLK TEOS The strain energy densities for SiLK and TEOS are ξ ξ SiLK TEOS 1 = σε 2 1 = σε 2 SiLK TEOS = = 1 2 1 2 σ E 2 SiLK σ E 2 TEOS E SiLK is about 30 times lower than E TEOS, hence the strain energy density in SiLK will be about 30 times higher, leading to a much higher energy release rate in the Cu/SiLK structure. Confinement effect due to the damascene structure also affects the stress driving force and deformation behavior of the interconnect.
Scaling Effect Cu/SiLK Structure, SiLK/PASS Interface 1.2 1 Normalized ERR 0.8 0.6 0.4 0.2 wafer level only (400--25oC) Packaging effect (-55--125oC) 0 0.1 0.2 0.3 0.4 0.5 Line width (um) The driving force for interface fracture increases slightly with decreasing line width at Cu SiLK/PASS interface.
Effect of Die Attach Process A critical process step in flip-chip packaging is the solder reflow step before underfilling the package. Without underfill serves as a stress buffer, thermal mismatch between the die and substrate can generate large thermal stress at the solder/die interface near the die corner. Solder reflow temperatures are different for Pb-based and Pb-free solders. Thermal loads used in our simulation: High lead solder: 300 o C-25 o C Eutectic solder: 160 o C-25 o C Lead free solder: 250 o C-25 o C
Parametric Study of Die Attach Process 1. Substrate effect Plastic vs. ceramic substrate 2. Die size: 7x8mm vs. 13.4x14.4mm die 3. Solder materials: High lead solder Eutectic solder Lead free solder
Material Properties E(GPa) v CTE(x10-6 ) Die 162 0.28 2.6 Plastic substrate Anisotropic elastic property 16(in plane) 84(out of plane) Ceramic substrate 300 0.3 5.0 High lead solder 50.81-0.102*T 0.35 29.7 Eutectic solder 75.84-0.152*T 0.35 24.5 Lead free solder 88.53-0.142*T 0.40 16.5 Underfill 1 8.40 0.40 28.0 Underfill 2 7.10 0.40 34.0 Underfill 3 6.23 0.40 40.6
ERR(J/m^2) 36 30 24 18 12 Solder Materials Effect (Plastic substrate, 7x8mm die) High lead solder package Eutectic solder package Lead-free solder package 6 0 crack 1 crack 2 crack 3 crack 4 crack 5 crack 6 Solder reflow before underfill increases the driving force for interfacial delamination in Cu/SiLK structures, particularly for lead-free solders.
Substrate Effect (Eutectic solder, 7x8mm die) 8 Plastic substrate 6 Ceramic substrate ERR (J/m^2) 4 2 0 crack 1 crack 2 crack 3 crack 4 crack 5 crack 6 Solder reflow reduces significantly the driving force for interfacial delamination in Cu/SiLK structure for packages with ceramic substrate compared to plastic substrate.
Die Size Effect (Plastic substrate, Eutectic solder) 10 8 8x7mm Die 14.4x13.4mm Die ERR (J/m^2) 6 4 2 0 crack 1 crack 2 crack 3 crack 4 crack 5 crack 6 During reflow, the driving force for interface fracture increases with increasing die size. The effect is larger for high lead solders due to a higher reflow temperature
Summary Maximum energy release rate (J/m 2 ) for 0.5 mm line width Interfaces ILD/PASS ILD/BARR Metal/PAS S Metal/BARR Wafer level 0.0076 0.3854 0.2044 0.4375 Al/TEOS Packaging 0.2086 0.0845 0.0630 0.0716 Wafer level 0.0072 0.3669 0.2079 0.4287 Cu/TEOS Packaging 0.2095 0.0865 0.0617 0.0702 Wafer level 0.2908 1.1556 0.1718 1.0476 Cu/SiLK Packaging 8.3392 11.1109 16.7080 7.5216
Conclusions 3D multilevel sub-modeling technique was developed to investigate the packaging effect on interfacial fracture for TEOS and low k interconnect structures. For stand-alone interconnects, the crack driving force for interfacial delamination is usually lower than the critical fracture energy, so subcritical crack growth and fatigue crack growth are important in controlling structural reliability. Packaging effect can significantly increase the energy release rate to cause critical crack growth in Cu/low k structures, particularly at the interfaces parallel to the chip surface. Interfacial chemical bonds are important in controlling interfacial adhesion. Residual stress can enforce thermal stress to drive crack growth, particularly in a humid environment. The effect on low k interconnect reliability has to be investigated.