Design for reliability of BEoL and 3 D TSV structures A joint effort of FEA and innovative experimental techniques

Transcription

1 March 17, 2011 Santa Clara, California Design for reliability of BEoL and 3 D TSV structures A joint effort of FEA and innovative experimental techniques Jürgen Auersperg 1, D. Vogel 1, M.U. Lehr 3, M. Grillberger 3, J. Oswald 3, S. Rzepka 1, B. Michel 1,2 1 Fraunhofer Institute for Electronic Nano Systems ENAS, Germany 2 Micro Materials Center at Fraunhofer Institute for Reliability bl and Microintegration IZM, Germany 4 GLOBALFOUNDRIES Dresden Module One LLC & Co. KG, Dresden, Germany

2 Outline Motivation Multi level FE modeling of CPI Multi failure evaluation for Near Chip Edge and Near Bump Cracks in BEoL Role of Initial Stresses TSV damage and delamination investigation Required experiments Summary and Outlook

3 Task Heterogeneous Integration Packaging Bridging the gap between chip and application Bump side <10 µm Transistor side nm, µm Nanostructures cm, m Applications

4 Multiscale why? Geometry Packaged microprocessor (eight symmetric model Packaged microprocessor (10 40 mm size) BEoL stack region Ni layer PI layer Bump side HMT solder Underfill BEoL stack (5 10 µm size) Feature sizes: some nm! Transistor side <10 µm gap LJT solder Solder interconnect with Boardpad BEoL stack (50 80 µm size) Soldermask

5 Multiscale why? Geometry Bridge 40 mm to sub nm for Chip Packaging Interaction investigations of a BEoL structure Structural details: sub nm

6 Multiscale why? Structure Nano lawn for interconnect formation Can these materials longer be modeled as homogeneous materials? Atomistic level modeling Molecular dynamics Low k dielectrics (nano porous) Polymer: 23 H2O for 98 C/ 100 RH

7 Multiscale Molecular Dynamics Two Ways used Structure The way I: 1. Modeling of the molecular structure 2. Homogenization in a unit cell 3. Use it directly inside a Macro model, a FE Model for instance The way II: 1. Modeling of the molecular structure 2. Simulations towards extraction of key properties FiniteElement (YOUNG s modulus, Region CTE, diffusion coefficients) Crack Tip Molecular 3. Use these properties Dynamics Region inside a Macrodla FE Model lfor instance model, Polymer: 23 H2O for 98 C/ 100 RH

8 Moisture Diffusion in Epoxy by Molecular Dynamics Motivation: i Understand diffusion phenomena as structureproperty correlation for reliability prediction Structure Method Materials: Epoxy with sev. chem. composition Exp: measure diffusion D & saturation S Sim: molecular dynamics Result: D and S are strong functions of (as tested) density, polarity, chain length, stoichiometry, temperature Sim & Exp show correct tendencies Quantitative Results need correct 3D network representation of epoxy resins E.D. M. Dermitzaki, Grillberger, J. Bauer, J. Oswald, B. Wunderle, S. Rzepka, B. B. Michel Michel m^2/s) oeff. Diff D co (cm 2.50E E E E E E+00000E 000 Polymer: 23 H2O for 98 C/ 100 RH Varied parameters = Arrow: increasing polarity and chain lenght Exp Experiment:98 C/100RH Diff co oeff. 2/s) D (cm^ 3.0x x x x x x C/ 100 rh Sim Molecular Dynamics 98 C/100RH

9 Multiscale Finite Element Techniques Substructuring Continuum Multi level Substructuring 256x

10 Multiscale Finite Element Techniques Submodeling Continuum Multi level Submodeling 256x

11 Multi failure Evaluation

12 2 nd Failure Mode Risk for Die Cracking, Cracking of Substrates, Smooth bending w/o singularities in Stress field with well known stress stresses/strains singularity 1 r Stress field with stress singularity 1 r Well known angle from anisotropic etching, for instance Classical l strength thhypotheses Generalized stress intensity it Fracture mechanics: max. surface stress evaluation factor (gsif) A factor K factor (SIF), J integral, energy Weibull plots release rate (ERR)

13 3 rd Failure Mode Delamination at Materials Interfaces 20µm ( i ) 0 xx yy K r i / 2 r Hutchinson et al r 0 2a exp 2 for yy and r * 0 a 2 exp for xy Mixed Mode Situation

14 Cohesive Zone Modeling Drawbacks and Challenges CZM DCB specimen with crack propagation under continuous displacement controlled opening! CZM Models have to handle and deliver damage parameters (damage progress per cycle/time) CZM with damage options Subjects of current research!

15 XFEM Mesh Independent Crack Propagation XFEM FEM shape functions FEM unknowns Discontinuous enrichment functions XFEM unknowns FEM displacement all nodes Crack crossing an element XFEM Heaviside enrichment subset of nodes Crack tip in an element

16 BEoL Stack of an IC Cracking/Delamination Locations and Impacts Near-chip-edge cracks under Chip Package Interaction (CPI) Crack stop structure Near-bump cracking during reflow (NBC) Leadfree Copper Pillars

17 Chip Package Interaction (CPI) Near-chip-edge cracks

18 BEoL Stack of an IC Near Bump Cracks Region of Interest NBC Packaged FE model (eight symmetric model) LID adhesive BEoL stack region Ni layer PI layer TIM1 Die Underfill Soldermask HMT solder Underfill Board gap LJT solder Soldermask Boardpad

19 3D BEoL Stack Part model Submodeling with Substructures NBC Substructure Finite elelement characteristic size in BEoL stack < nm Substructure region Replacement of the finest (1x) structures by superelements (repeatedly used) leads to a dramatically reduced model size (by factor 3), complete model (with and without element edges) vias and metal traces

20 3d Local Model of a BEoL Part Crack Introduced SubModeling NBC Global model Submodel Initial crack

21 3d Local Model of a BEoL Part SubModel Placing and Orientation NBC Submodel driving nodes

22 Multiscale Substructure & Submodeling of a BEoL stack of an IC NBC Mode II (In Plane Shear) Mode III (Out of Plane Shear)

23 Multiscale Substructure & Submodeling of a BEoL stack of an IC NBC Crack kdii driving force along the crack front Results: Find most critical places with Material interface delamination and/or Cohesive fracture of material

24 BEoL Stack of an IC Die Edge Cracking Interface Fracture Mechanics CPI 1,40 max. norm. ERR vs Crack Length orm. G n 1,20 1,00 0, ,60 Chip corner 0,40 margin 51 µm to crack-stop-structure 0,20 0,00 margin 51 µm to crack-stop-structure w/o crack-stop-structure structure crack length [µm] Crack stop structure t Margin

25 BEoL Stack of an IC Die Edge Cracking Interface Fracture Mechanics CPI 10 bimaterial interface paths + 4 cohesive crack paths in ULK 3 Max. ERR in diff. Crack Paths, 40 µm Crack Length marg 100 2,9 2,8 ge en. G 2,7 2,6 Chip corner 2,5 Path 24 2,4 Crack stop structure t 2, path 10 Margin

26 NBC Reflow Process Dependent Near Bump Cracks in BEoL Leadfree

27 3d Global Local Model Simulations Stress States NBC Peeling Stresses at 0 C (under CPI) Peeling Stresses at RT (end of reflow) Chip center Chip center Chip center Avoid worst case: white spots in SAM

28 Cooling Down from Soldering Reflow Profiles NBC Highlead Leadfree Aggressive cooling slope Some studies have shown that an aggressive cooling slope can provide optimal diffusion of material and fine eutectic grain structures Mohanty, R., Apell, M. C., Burke, R.: Reflow Process Control Monitoring, and Data Logging, URL: es/white%20papers/speedreflo wprocesscontrol.pdf

29 FE Modeling of the Soldering Process Reflow Profile Modification NBC Highest G for Leadfree Soldering Stress relaxation of the solder joints Temperature vs. Time 300 Leadfree profile 250 Leadfree profile 02 Highlead profile 200 Highled profile 02 Highled profile Cooling down period discussed here

30 Crack and Damage Initiation/Propagation Investigation NBC Damage Problems: Material/Interface properties Integration Stability/Controlability Damage propagation investigation in order to find highest damage risk locations XFEM XFEM in order to find crack starting locations and paths

31 Chip Package Interaction and Initial Stresses From Manufacturing Processes

32 Local Residual Stress Measurement Approaches Electron Back Scattered Diffraction (EBSD) based method deformation of Kikuchi diffraction pattern as measure of stress determines stress tensor comp. Residual Stresses Stress Release Technique (fibdac) utilizing deformation field after trench milling by FIB Raman Spectroscopy uses shift of Raman bands due to stress TERS effect with potential to obtain resolution beyond 100 nm D. Jürgen Vogel, A. Auersperg, Gollhardt D. Vogel, M.U. Lehr,

33 Local stress measurement concept stress relief by FIB milling (fibdac) Material removal by FIB feature milling Stress relief, i.e. deformation nearby milling pattern Solution of the Quantifying mechanical stress relief Displacement df deformation field problem by field fitting by Digital Image Finite Element Analysis Correlation (DIC) (FEA) Stress values D. Jürgen Vogel, A. Auersperg, Gollhardt D. Vogel, M.U. Lehr,

34 Near Bump Cracks in BEoL Taking Initial Stresses into Account Crack Paths 6 and 14 with ULK vs. Gen3 During whole processing w/o and with initial stresses 1,2 1 max. ERR for all Steps at the Crack Front With initial stresses 1,00000 G 0,8 0,66285 gen. 0,6 0, , ,4 0,2 0, , , , Path 6 Path 6 Path 14 Path 14 Path 6 Path 6 Path 14 Path 14 ULK Gen3 ULK Gen3 ULK is Gen3 is ULK is Gen3 is

35 3D Integration ti TSV and BEoL Cracking/Delamination risks during BEoL manufacturing on top of TSVs?

36 TSV and BEoL structure Protrusion

37 TSV and BEoL structure Possible Failure Mode Delamination Protrusion

38 TSV and BEoL structure Possible Failure Mode Delamination Protrusion Assumtions: Initial delamination between TSV and barrier Initial stresses in TSV and introduced step by step during BEoL processing Partly delaminated BEoL Cracks in BEoL

39 TSV and BEoL structure Delamination Investigation Changing stress traction direction Strong dependence on stress free assumptions Crack flanks partly in contact negative Jint

40 Sequential Build Up Cohesive Zone Modeling BEoL to MOL Delamination Cohesive zone elements introduced between MOL and BEoL damaging during thermal loading delamination of initially undamaged material interface d Measure for interface loading (BEoL removed for better demonstration Interface totally damaged at distance d TSV Silicon Define CZM properties: Tn, Ts, Tt, GI, GII, GIII, En, Es, Et (all are estimated here!)

41 and BEoL structure Possible Failure Mode Damaging of Copper Define damage properties: pl, p/q, rate, G c, )all are estimated here!)

42 Material lproperty Dt Determination ti Size Dependent Material Behavior

43 Material Behavior Measure Required Properties Dynamical Mechanical Analysis (DMA) Thermomechanical Analysis (TMA) Creep data eval. (Solder) Temp.dep. creep data (Epoxy) Stress relaxation data (Epoxy) Poisson s s ratio eval. H. Walter, M. Grillberger, J. Auersperg J. Oswald, S. Rzepka, B. Michel

44 Nanoindentation in low K Dielectrics Bump side Thin layers BEoL stack 7 µm F, u Intenter Thin Layer Substrate t For TSV: Initial yield stress Hardening ade gbehavior N] Force [µ Chip side Sim Measurement Experiment E=145, sy= Distance [nm] Experiment Simulation Agreement Parameter extraction only possible by coupled Sim & Exp Raul Jürgen Mrosko, Auersperg, Saskia Huber, D. Vogel, O. Wittler M.U. Lehr,

45 TSV Copper YOUNGs Modulus by nano Indentation P13 Messung 3 P13 Messung 3 46,405 demo demo demo demo demo Y(mm) 46,400 demo demo demo demo demo demo demo demo demo demo demo demo demo demo demo Er(GPa) ,395 demo demo demo demo demo 40-82, , , , X(mm) demo demo demo demo demo 46,390-82,275-82,270-82,265-82,260 X(mm)

46 Copper Material Behavior Properties Necessary Necessary to know: Young's modulus dep. On Temperature Initial yield stress Hardening vs. plastic strains Regarding TSV: initial stress state Xi Liu, Qiao Chen, Pradeep Dixit, Ritwik Chatterjee, Rao R. Tummala, and Suresh K. Sitaraman, Failure Mechanisms and Optimum Design for Electroplated Copper Through Silicon Vias (TSV), 2009 Electronic Components and Technology Conference, pp

47 Size Effects and Damage in Thin Metallic Films during Nano Indentation Meso mechanics Homogeneous Strain gradient plasticity Higher order stress theories Dislocation movement restrictions inside boundaries D. Gross, A. Trondl: Numerical Simulation of Size Effects and Damage in Thin Metallic Films during Nano Indentation, Proc. ICF 12, Ottawa, CA, 2009, Paper on CD: fin00457.pdf

48 Damage and Fracture Toughness Characterization ti Size Dependent Material Behavior

49 Interface Fracture Toughness Tests P Pull off test schematic of substrate, interface and pull stud Modified CT specimen Peel test Pull out test Brazilian disk specimen Superlayer adhesion test A.A. Volinsky et al. / Acta Materialia 50 (2002)

50 Interface Fracture Toughness Tests Bending Asymmetric Double Cantilever Beam Brazil Nut Sandwich Single Leg Bending Symmetric Double Cantilever Beam 4 point bending Asymmetric DCB Asymmetric End Notched Flexure End Notched Flexure Sundaraman, Sitaraman 1997 Center Cracked Beam Yeung, Lam, Yuen 2000 Double Cantilever Beam Miller, Ho 2000

51 Mode Separation/Evaluation Critical ERR and Phase Angle Measured fracture toughness failure region Calculated crack loading

52 Mode Separation/Evaluation Critical ERR and Phase Angle 1 1 G c G IC sin 2 With (a further material/interface parameter) is a function of temperature and moisture concentration G c failure region G( ) initiationi i i G( ) no initiation Hutchinson and Suo, "Mixed Mode Cracking in Layered Materials", Advances in Applied Mechanics, Vol. 29, pp K. Liechti, Adhesion Measurements Relative to Electronic Packaging, Proc. 4th Annual Topical Conference On Reliability, October 30 - November 1, 2000

53 Nano Indentation for Interface Fracture Toughness Evaluation Nanoindentation for measuring Young s modulus and hardness Nanoindentation for measuring fracture toughness Buckling Nanoindentation for measuring interfacial fracture toughness (adhesion) Conical and wedge indenters For instance: E.E. Gdoutos, A. Volinsky, Interfacial Fracture Toughness of Thin Films, Tutorial AC Conf., June 6, 2005 M. R. Begley, D. R. Mumm, A. G. Evan, J. W. Hutchinson, Analysis of a Wedge Impression Test for Measuring the Interface Toughness Between Films/Coatings and Ductile Substrates, Acta mater. 48 (2000)

54 Stress State t Determination/Verification ti ifi ti Fitting modeling assumption with experimental results

55 TSV Stresses near the Surface vs. Raman Spectroscopy zz rz rr M. Hecker, GlobalFoundries, Dresden

56 Surface vs. Midd Stress rr Surface - rr Middle - rr

57 Summary Multilevel lfe modelling of cracks in BEoL Struktures For Chip Package Interaction and Reflow leadfree bumping Goal: Reliability enhancement Utilizing fracture mechanics concepts (cohesive and adhesive) and/or Damage mechanics approaches

58 What we need Intensive experimental work necessary for Evaluation of residual stresses from manufacturing, Evaluation of the material behavior (constitutive, size dependent) Characterization of the damage behavior of materials Characterization of strength/toughness properties (fracture mechanics characterization) of materials and materials interfaces Verification of simulation results Extensive simulation work necessary Huge simulation models and computing resources Theoretical work delamination und fracture (CZM, XFEM), materials behavior (SGPM)

59 What we Need for Helpful Simulation Results Residual stress characterization Damage and fracture characterization Material behavior and property characterization Simulation results verification

60 Thank You for Your Attention! ti