Elsevier

Microelectronics Reliability

Volume 54, Issues 9–10, September–October 2014, Pages 1963-1968
Microelectronics Reliability

Stress analyses of high spatial resolution on TSV and BEoL structures

https://doi.org/10.1016/j.microrel.2014.07.098Get rights and content

Abstract

Knowledge and control of local stress development in Back-End-of-Line (BEoL) stacks and nearby Through Silicon Vias (TSVs) in advanced 3D integrated devices is a key to their thermo-mechanical reliability. The paper presents a combined simulation/measurement approach to evaluate stresses generated in the result of the TSV and BEoL stack manufacturing and 3D bonding processes. Stress measurement methods of high spatial resolution capability (microRaman and Focused Ion Beam (FIB) based stress release techniques) are used to obtain stress data from real components as manufactured. Finite Element Analysis (FEA) allows a more accurate interpretation of measurements results as well as a subsequent comprehensive analysis of failure behaviour. The paper gives an introduction to the applied local stress measurement on advanced multi-layer systems and 3D integration components referring to the state-of-art capabilities and limitations. The need of experimental stress data generation is illustrated on FEA examples. Illustration is given for FEA applications on 3D IC integration components currently lacking appropriate residual stress input for an assumed initial state.

Introduction

Advanced microprocessors utilizing low-K and ULK materials have led to new reliability concerns due to the tremendously changed material and interface properties in the BEoL stack. Chip Package Interaction (CPI) became an important issue for interfacial crack initiation and propagation in the stack. Ongoing 3D-IC-integration development introduced new critical problems corresponding to the used TSV technology. Because of the high CTE (Coefficient of Thermal Expansion) mismatch between the TSV metals and the surrounding silicon, significant stresses occur nearby TSVs. As a result, charge carrier mobility as well as mechanical device reliability is affected. So, reliable knowledge and control of local critical stress states is an important precondition of successful device design.

Numerical simulation and failure modelling on complex 3D-integrated packages still is a challenge. Although whole manufacturing process simulations are carried out, they can suffer from multisided necessary simplifications. Right identification of material laws and respective property determination is another difficulty. Hence, Finite Element Analysis (FEA) of sub-components and sub-process steps is a common approach. Already frozen intrinsic stresses for a particular starting point of an FEA is one of the essential input entries needed. Fig. 1 illustrates this fact for a case study of the authors on interface reliability in a BEoL stack. Although a trend estimation of crack propagation risk comparing different dielectrics materials/crack propagation paths is possible, no real failure risk can be estimated, because of missing reliable absolute values of the Energy Release Rates (ERR), as seen from Fig. 1.

Incorporating stress measurement methods to feed simulations with right initial stresses can be an appropriate solution. Basic requirements on such tools would be

  • sufficient spatial stress resolution laterally and into depth (some μm to a few 10 nm),

  • no or low degree of destructive measurement,

  • measurement feasibility on components and systems as manufactured,

  • at least low field measurement.

Among the high spatial resolution methods demonstrated on BEoL stacks, TSVs or similar small objects, we find X-ray and electron diffraction [1], [2], stress relief techniques [3], [4] and microRaman spectroscopy [5]. Diffraction based methods require sufficiently crystalline materials. Higher lateral spatial resolution is limited to electron diffraction (some 10 nm) or synchrotron sources with small spot sizes (some 100 nm). Electron diffraction allows relative measurements in single crystals only. Diffraction measurements on synchrotrons are expensive with limited availability.

At first glance microRaman is an attractive method with feasible and fast stress mapping, however restricted to Raman-active materials, like semiconductors. In the past many groups applied this method to access stresses around TSVs [5], [6]. However, at second glance stress extraction from Raman shifts is not as simple.

Section snippets

Applied stress measurement approaches

Two experimental stress measurement methods are reported in this work – a stress relief technique carried out in Focus Ion Beam equipment (FIB), and microRaman spectroscopy. Both of methods possess high lateral spatial resolution, even though the first one outmatches by an order of magnitude. Furthermore, they differ in their sensitivity and response to the particular stress tensor components relevant for the residual stress state nearby TSV structures or in BEoL stacks. I.e. they are

fibDAC stress relief measurements

Stress relief measurements have been performed in a ZEISS Auriga60 FIB using magnification typically between 10,000 and 100,000. In order to reduce processing time due to stage tilts and re-adjustments, ion milling was carried under 54° observing the object with the SEM under 0°. Low current deposition mode has been used for ion milling.

Fig. 3 shows a first example of stress measurement carried out on a TSV at room temperature, Cu filled and with 8 μm diameter. Obtained experimental stress data

Finite Element Analysis (FEA)

In order to avoid thermo-mechanical failures or charge carrier mobility changes around TSVs, extended FEA is being performed to simulate and understand the impact of TSVs. Fig. 10 shows a typical failure analysis performed by FEA for a TSV application. Accumulated plastic strain in Cu due to thermal cycling leads to Cu pumping and influences the failure behaviour. Bi-material fracture mechanics analyses e.g. by modified Virtual Crack Closure Technique (VCCT) are a common approach to evaluate

Conclusions

It has been shown how a combination of local stress measurement techniques, as microRaman and fibDAC stress relief, and Finite Element Analysis (FEA) can be combined to evaluate stresses caused by TSVs and BEoL stacks. Experimental results can be utilized to validate FE modelling and to make sure that simulations lead to trustworthy data. Moreover experimentally found intrinsic stresses can be used as input data for FEA modeling, if not the whole component manufacturing process is emulated by

Acknowledgements

Some of the referred to Raman measurements have been performed at Fraunhofer IZM. We acknowledge the assistance by A. Gollhardt. We also acknowledge some simulation work carried out by U. Hoelck, at that time being with Technical University Chemnitz. Part of the work reported has been funded by the BMBF Germany (Cluster of Excellence “MERGE”).

References (13)

  • A.J. Wilkinson et al.

    Ultramicroscopy

    (2006)
  • A.M. Korsunsky et al.

    Mater Lett

    (2009)
  • K.J. Kang et al.

    Thin Solid Films

    (2003)
  • Song Xu et al.

    Mater Sci Forum

    (2008)
  • Vogel D, Rzepka S, Michel B. In: Proc. of semiconductor conference, Dresden;...
  • de Wolf I, Simons V, Cherman V, Labie R, Vandevelde B, Beyne E. In: Proc.of 60th ECTC. San Diego; 2012. p. 331–337...
There are more references available in the full text version of this article.

Cited by (19)

  • Crack arrest in thin metallic film stacks due to material- and residual stress inhomogeneities

    2018, Thin Solid Films
    Citation Excerpt :

    Thin film components for microelectronic applications are most commonly subjected to residual stresses that emerge due to variations in the coefficient of thermal expansion or an atomic lattice spacing mismatch between the materials [36]. Several techniques that determine the residual stress state in thin films have been recently proposed: X-ray diffraction analysis [37–39], focused ion beam milling in combination with digital image correlation [40–43] and the Ion beam Layer Removal (ILR) method [44]. Experimental and computational results have shown that residual stresses can strongly influence the fracture behavior of materials [45–50].

  • High resolution determination of local residual stress gradients in single- and multilayer thin film systems

    2016, Acta Materialia
    Citation Excerpt :

    A third group of methods for the local determination of residual stresses comprises incremental focused ion beam (FIB) milling to determine material relaxation due to stress redistribution upon material removal. Different measurement geometries such as slits [4,14,15], pillars [16–19] and micro cantilevers [20,21] are available. These methods are generally applicable to crystalline and amorphous, as well as to textured materials, and provide data from specific positions of a film/sample or local features on a patterned structure.

View all citing articles on Scopus
View full text