Elsevier

Acta Materialia

Volume 58, Issue 18, October 2010, Pages 6035-6045
Acta Materialia

Microstructure evolution during dewetting in thin Au films

https://doi.org/10.1016/j.actamat.2010.07.021Get rights and content

Abstract

Thin metal films can degrade into particles in a process known as dewetting. Dewetting proceeds in several stages, including void initiation, void growth and void coalescence. Branched void growth in thin Au films was studied by means of electron backscatter diffraction (EBSD). The holes were found to protrude into the film predominantly at high angle grain boundaries and the branched shape of the holes can be explained by surface energy minimization of the grains at the void boundaries. (1 1 1) Texture sharpening during dewetting was observed and quantified by EBSD and in situ X-ray studies.

Introduction

Thin metal films can disintegrate into an array of particles upon annealing. This process is referred to as solid state dewetting; it is driven by surface and strain energy minimization [1]. Dewetting starts with the formation of voids, which can take place by various mechanisms, e.g. thermal grooving at grain or twin boundaries [2], [3], [4], [5], nucleation of vacancies at the film–substrate interface [6], [7], [8] or at grain boundaries and grain boundary triple junctions in areas of high local stress [9]. In the latter case the cavity is expected to grow along grain boundaries, since this is the fastest diffusion path for vacancies. Furthermore, void formation due to dust particles or heterogeneities in the substrate is often observed. It is assumed that hole growth, i.e. the removal of material from the hole border, takes place by film surface diffusion [10], [11]. During hole growth a rim of piled-up material is formed around the hole. This rim has to be moved for further void growth, thus a decrease in hole growth rate for increasing hole size is expected if the hole remains circular. Jiran and Thompson [12] observed branched hole growth at the edges of thin gold lines and proposed a corrected model for hole growth kinetics which is based on branched growth. The model states that a rim is formed at the sides of the finger-shaped hole while at the tip of the hole the material can diffuse away in several directions and no pile-up of material is observed. According to this model hole growth preferably takes place at the tips of holes and no decrease in the void growth rate is expected. Kan and Wong [13] have shown that an instability in the rim at the edge of a film exists that leads to the growth of such regularly spaced fingers. Jiran and Thompson [12], [14] also attributed branching to capillary effects and stated that branching is not affected by the presence of grain boundaries; they found that voids grew across grain boundaries and grains alike. In contrast to their observations, several experimental studies have suggested that hole growth is taking place preferably along grain boundaries [9], [15], [16]. The presented study provides insights into the mechanism of branched hole growth based on the analysis of film microstructure by diffraction methods.

Dewetting of thin metal films is usually studied by analysis of the post-annealed film surface by electron microscopy or atomic force microscopy (AFM), but in situ observation in a microscope during annealing, conductivity/resistivity measurements of the film during annealing or X-ray diffraction (XRD)-based methods can also be used. The advantage of the in situ techniques is that they allow the investigation of the dewetting process in real time. There are two well-developed in situ characterization techniques that have been employed by various groups. The first technique is to monitor the surface of the sample by light microscopy. This is done in either reflection [17] or transmission [12], [14]. The method is based on evaluation of the coverage parameter, i.e. the area still covered by the film divided by the total sample area. The main disadvantage of this approach is that it is not possible to detect changes at the film–substrate interface. The second well-developed technique is to measure the sample resistivity. A typical resistivity measurement shows a sharp increase in resistivity as soon as the film disintegrates. This is a well-known phenomenon [18] and can be explained by a percolation model [19]. The main disadvantage of the resistivity technique is that it can track hardly any changes after the film becomes discontinuous. Sieradzki et al. [20] investigated both the resistance and the coverage evolution from SEM pictures during isothermal annealing. They showed a linear relationship between the resistance normalized to its initial value and the area fraction still covered by film normalized to the coverage at electrical failure. Another possibility for the investigation of dewetting is XRD. It has the potential to become an important technique for the characterization of dewetting because it can also provide information about texture, phase composition and grain size – information that would prove especially helpful in studying the dewetting process of alloys. To the authors’ knowledge a paper by Mizsei and Lantto [21] constitutes the only report in the literature on in situ XRD measurements applied to study dewetting. Mizsei and Lantto [21] studied the dewetting of thin silver films and found an increase in the intensity of the (1 1 1) reflection during annealing. They attributed this to an increase in particle height, however, no conclusive evidence was given to support this assumption. The true nature of the increase in (1 1 1) reflection intensity thus remains to be investigated.

Section snippets

Film deposition

Three inch p-type 380 μm thick Si(1 0 0)/SiO2/SiNx wafer substrates (Crystec Kristalltechnologie) were used as received. The thickness of the SiO2 and SiNx barrier layers was 50 nm each. A 99.99% pure gold wire (Umicore) and a Bal-Tec Med 020 Coating System were used to deposit thin gold films by thermal evaporation. A quartz crystal monitor was used to measure film thickness and deposition rate in situ. The deposition rate was 0.3 nm s−1 for all films. Films of 10, 15, 20 and 30 nm thickness were

XRD measurements as a method to follow the dewetting process

An experiment analogous to the one reported by Mizsei and Lantto [21] was carried out using a 30 nm thick gold film. The film was annealed inside the XRD spectrometer at 600 °C in an N2 atmosphere and measurements were carried out every 10 min. Fig. 1a shows an excerpt of the spectra obtained. As can be seen from Fig. 1, the (1 1 1) reflection of gold became increasingly sharp during annealing. After 1 h annealing the (1 1 1) peak was split into two narrower peaks corresponding to the Kα1 and Kα2 lines

XRD as an alternative to coverage studies

It can be seen in Fig. 1b that (1 1 1) peak intensity variations can be used to follow the dewetting of thin gold films. Fig. 8a shows the (1 1 1) peak intensity and normalized hole area as a function of annealing time. The normalized hole area data was measured from separate ex situ experiments in an oven, not from the sample investigated in situ by XRD. It is thus possible that the observed shift between the (1 1 1) peak intensity and the normalized hole area was caused by differences in sample

Conclusions

The pioneering work of Mizsei and Lantto [21] on characterization of dewetting by XRD measurements has been continued in this work. It has been shown that the increase in (1 1 1) peak intensity during dewetting is caused by both a height increase of the features on the sample and an increase in texture. The dewetting of gold films of 15–30 nm thickness was found to occur via branched hole growth. Holes were found to protrude into high angle grain boundaries. EBSD measurements have shown that the

Acknowledgements

C.M. thanks D. Spori for support during film deposition and B. Grant for help with the thickness determinations by AFM. Special thanks go to M. Seita for carrying out the EBSD measurements, to F. Mornaghini for FIB operation and to S. Frank, C. Borgia and D. Koziej for assistance during XRD measurements. Support by the Nonmetallic Inorganic Materials Group (SEM), the Surface Science and Technology Group (thermal evaporation), the Multifunctional Materials Group (XRD) and Electron Microscopy at

References (27)

  • F.Y. Génin et al.

    Acta Metall Mater

    (1992)
  • D.C. Agrawal et al.

    Acta Metall

    (1989)
  • C.M. Kennefick et al.

    Acta Metall

    (1989)
  • E. Shaffir et al.

    Acta Mater

    (2009)
  • N. Kristensen et al.

    Thin Solid Films

    (1991)
  • A.E.B. Presland et al.

    Prog Surf Sci

    (1972)
  • E. Jiran et al.

    Thin Solid Films

    (1992)
  • A.E.B. Presland et al.

    Surf Sci

    (1972)
  • V.A. Mazur et al.

    Phys Lett A

    (1989)
  • D.J. Srolovitz et al.

    Thin Solid Films

    (1986)
  • W.L. Winterbottom

    Acta Metall

    (1967)
  • D.T. Danielson et al.

    J Appl Phys

    (2006)
  • W.W. Mullins

    J Appl Phys

    (1957)
  • Cited by (105)

    • Solid state dewetting of Ni-Co bilayers on sapphire during slow heating and cooling

      2022, Acta Materialia
      Citation Excerpt :

      For the lowest studied temperature (500 °C), uniformly distributed anisotropic holes with branches were observed in Ni (Fig. 2(a)). Such holes are typical for solid state dewetting of thin polycrystalline films and have been observed in several previous studies [18,34,35]. In the Ni-30Co film, only few isolated anisotropic holes with sections of rimless edges at very nascent stage of nucleation were present, as indicated by arrows in Fig. 2(b).

    View all citing articles on Scopus
    View full text