Elsevier

Acta Materialia

Volume 67, April 2014, Pages 297-307
Acta Materialia

Damage evolution during cyclic tension–tension loading of micron-sized Cu lines

https://doi.org/10.1016/j.actamat.2013.12.006Get rights and content

Abstract

In this study, the low-cycle fatigue properties (1–15,000 cycles) of electrodeposited Cu, which is frequently used as metallization in the semiconductor industry, is analyzed with respect to its microstructure. Freestanding Cu tensile samples 20 μm × 20 μm × 130 μm were fabricated by a lithographic process. The grain size of the samples was modified by using three different process conditions for electrochemical Cu deposition. All samples were subjected to cyclic tension–tension testing performed with a miniaturized stress-controlled stage in situ in a scanning electron microscope until failure occurred. The number of cycles sustained prior to failure depends on the accumulated creep strain and can be related to the failure strain in a tensile test. It will be shown that the microstructure influences the number of cycles to failure and the failure mode.

Introduction

Cu in microelectronic devices, such as light-emitting diode submounts and microelectromechanical systems, has to fulfill two important requirements. It must provide a low electrical resistance even at high current densities and simultaneously provide excellent thermal conductivity to transfer dissipated energy. Furthermore, the occurrence of thermal stresses due to the difference of the coefficient of thermal expansion (CTE) of Si and Cu requires optimization of the mechanical properties [1], [2]. This includes the mechanical fatigue behavior as in many cases more than 108 load cycles must be endured during the lifetime of semiconductor devices [3]. If the Cu line is damaged, the electrical resistance increases, leading to additional Joule heating, and resulting in a local hot spot and subsequent failure of the device. Thus, compromises in the grain size must be made to fulfill opposing trends. To prevent electromigration and to provide excellent electrical conductivity a large grain size is favored [4], [5], while a fine grain size is best to achieve high mechanical strength and a low susceptibility to mechanical failure under cyclic loading [6]. Since as-grown electrodeposited Cu films undergo uncontrolled grain growth at room temperature [7] an annealing treatment is made after the film deposition with the intention to stabilize the microstructure at a specific grain size.

The strength of Cu films could also be increased by alloying. However, alloying reduces the electrical and thermal conductivity and, as a consequence, the temperature of the metallization would increase, leading to higher thermal stresses. The potential of alloying Cu to optimize the strength while maintaining sufficient electrical and thermal conductivity is limited. Alternatively, controlling the texture may open routes to tailor the mechanical performance of Cu due to its high elastic anisotropy while maintaining excellent electrical conductivity. Possible methods to influence the texture are ion bombardment as reported in Ref. [8] or abnormal grain growth where the competition of surface energy and strain energy permit the alteration of the texture from 〈1 1 1〉 to 〈1 0 0〉 with increasing film thickness [9]. Both methods of tailoring the texture are inconvenient for industrial device processing as abnormal grain growth is undesirable because of decreasing strength, and ion bombardment is only applicable for films thinner than 5 μm. A simpler approach to improve the strength without adversely affecting the thermal and electrical conductivity is to control the grain size. It is well established that smaller grains lead to higher strength (Hall–Petch effect [10], [11]). Furthermore, recent publications have shown that the number of grains in a cross-section has a significant influence on the mechanical behavior [12], [13], [14]. For Cu samples with less than about 25 grains in a cross-sectional area the strength decreases significantly compared to the bulk material. The strength decreases because of the high percentage of surface-intersecting grains, where dislocations can easily escape and thus do not contribute to hardening. Similar findings were made for Cu [12], Pt [13] and Ni [14]. This approach was taken in the present study where the median grain size, d, of the Cu lines (line width w) was varied between 2.7 μm (w/d = 7.5) and 13.5 μm (w/d = 1.5) by using different methods for electrochemical Cu deposition and analyzed with respect to their fatigue properties.

Up to now low-cycle fatigue (LCF) experiments were performed for similar grain sizes on Cu films on a substrate [15], Cu foils [16], [17], [18], microsized Cu wires [19], [20] and ultrafine-grained (ufg) bulk Cu [21]. Additionally, high-cycle fatigue (HCF) experiments were performed for similar grain sizes on thin Cu films on a substrate [6], [22], [23], [24] and ufg bulk Cu [25], [26], [27]. In several of these publications the size effect of different sample dimensions with different grain sizes was analyzed. During fatigue testing, transgranular extrusions and intergranular cracks are formed in large grains accompanied by surface roughening and damage formation [6], [22]. For thin films and other small-scale materials the fatigue lifetime significantly exceeds that of bulk material [23]. Increasing the loading frequency reduces the number of cycles to failure [24]. In the LCF regime the deformation behavior of ufg Cu in fatigue experiments is different to that of conventional bulk Cu. The static tensile strength of ufg Cu is 1.8 times higher than that of conventional bulk Cu, contrary to the HCF properties that show no significant difference [25]. Up to 107 cycles ufg Cu shows a much higher strength compared to conventional bulk Cu; above 107 cycles grain coarsening starts, resulting in a much lower strength in this area [26]. Contrary to this, in other fatigue experiments with ufg material a change in microstructure was not found until >109 cycles [27].

In this paper detailed investigations on the influence of grain size for free-standing micron-sized samples are reported. A further target of this work was to compare the cyclic tension–tension behavior (R  0, R = σmin/σmax) with cyclic compression–tension experiments (R = −1) from bulk material. The main outcome is that with changing R values different deformation mechanisms are expected. Furthermore, macroscopic samples with a microcrystalline grain structure usually possess >1010 grains in the sample volume, while for microscopic samples with microcrystalline grain structure the number of grains per volume typically ranges between a few and 103 grains.

Performing micromechanical experiments on small samples adds an additional complication to fatigue testing. It is challenging to accurately measure the strain between subsequent loading cycles for micromechanical experiments. In this study it will be shown that accurate strain measurements can be achieved by in situ scanning electron microscopy (SEM) fatigue experiments and the role of grain size and number of grains per sample volume will be analyzed.

Section snippets

Film deposition and microstructure analysis

The Cu samples were produced by photolithographic process and electrolytic deposition from liquid solution; for details see Ref. [28]. Three different electrolyte compositions were used for Cu deposition with the aim of modifying the grain size. All Cu films were annealed at 400 °C for 30 min directly after deposition to stabilize the microstructure at specific grain sizes. The microstructure of all samples was characterized by SEM and electron backscatter diffraction (EBSD). No surface treatment

Results and interpretation

In this section the microstructural findings are reported first before presenting the results concerning the static tensile deformation including the Young’s modulus. Subsequently, the dynamic tension–tension data and the in situ SEM observations are provided.

Discussion

In this section the main results concerning the static and dynamic tension tests will be discussed. We will first focus on the influence of grain size on yield stress, UTS and strain to failure before comparing the cyclic creep experiments with R  0 to the static tension test and R = −1 fatigue tests reported in literature. Differences in the deformation mechanism between experiments with R = 0 and R = −1 will be analyzed followed by the application of the Basquin equation. An equation will be

Summary and conclusions

Tensile samples with a cross-section of 20 × 20 μm2 and a length of 130 μm were produced by electrolytic deposition from aquatic solutions with different process conditions to produce tensile samples with four different grain sizes. The samples were cyclically loaded in a tension–tension LCF experiment with the number of cycles ranging between 1 (tensile test) and 15,000. When the grain size was much smaller than the sample width, the samples showed polycrystalline behavior with multiple slip. When

Acknowledgments

Part of this work was jointly funded by the Austrian Research Promotion Agency (FFG, Project No. 831163) and the Carinthian Economic Promotion Fund (KWF, Contract KWF-1521|22741|34186). The authors want to thank M. Cordill, J. Fugger and M. Smolka. Thanks for technical assistance is expressed to M. Augustin, H. Felber, R. Grilz, H. Groß, F. Hubner, M. Krug, R. Leuschner, C. Lindner, K. Matoy, S. Modritsch, G. Moser, T. Ostermann, G. Reiter, H. Schönherr and K. Schrettlinger.

References (55)

  • S. de Filippis et al.

    Microelectron Reliab

    (2011)
  • O. Kraft et al.

    Mater Sci Eng A

    (2001)
  • M. Seita

    Acta Mater

    (2010)
  • P. Sonnweber-Ribic et al.

    Acta Mater

    (2006)
  • D.C. Jang et al.

    Scripta Mater

    (2011)
  • G.P. Zhang et al.

    Acta Mater

    (2006)
  • D.T. Read

    Int J Fatigue

    (1998)
  • S. Hong et al.

    Thin Solid Films

    (1996)
  • R. Hofbeck et al.

    Scripta Metall

    (1986)
  • B. Yang et al.

    Mater Sci Eng A

    (2009)
  • H.J. Maier et al.

    Int J Fatigue

    (2006)
  • Y.-B. Park et al.

    Thin Solid Films

    (2006)
  • G.P. Zhang et al.

    Microelectron Reliab

    (2007)
  • Y.-B. Park et al.

    Thin Solid Films

    (2007)
  • S.Z. Han et al.

    J Alloys Compd

    (2007)
  • G. Khatibi et al.

    Int J Fatigue

    (2010)
  • L. Kunz et al.

    Proc Eng

    (2011)
  • P.G. Sanders et al.

    Acta Mater

    (1997)
  • A.W. Thompson et al.

    Acta Metall

    (1971)
  • A. Vinogradov et al.

    Scripta Mater

    (2009)
  • L. Kunz et al.

    Mater Sci Eng A

    (2006)
  • H. Mughrabi et al.

    Scripta Mater

    (2004)
  • A. Zhao et al.

    Int J Fatigue

    (2012)
  • J.H. Ryu et al.

    Int J Fatigue

    (1989)
  • D.-Y. Lee et al.

    Int J Fatigue

    (2012)
  • X. Hou et al.

    Acta Mater

    (2012)
  • D. Kiener et al.

    Acta Mater

    (2008)
  • Cited by (33)

    • Cyclic response of electrodeposited copper films. Experiments and elastic–viscoplastic mean-field modeling

      2021, Mechanics of Materials
      Citation Excerpt :

      It has been also observed that microstructural features influence the mechanical response of copper film. Film thickness, thermal treatments or grain size, shape and orientations, manufacturing process generate different copper mechanical behaviors, as pointed out experimentally by Schwaiger et al. (2003) or Wimmer et al. (2014) for PCB applications or by Adzima et al. (2017) for forming process of copper thin sheet. To account for all expected dependencies on microstructural features, elastic–viscoplastic multiscale approaches for polycrystalline materials were proposed in the literature, usually as counterparts of formulations developed earlier for composite materials.

    View all citing articles on Scopus
    View full text