Critical reviewStress and strain in polycrystalline thin films
Introduction
When today one searches for “stress AND (thin films OR coatings)” on the web of science, one gets over ten thousand hits. This huge amount of papers demonstrates the interest in stresses in thin films and coatings. These extensive search results do not include a reference to the pioneering work by Stoney, who in his 1909 paper derived the archetype of the equation that now goes by his name [1]. Some work on evaporated iron films was already done in the 1950s by Finegan and Hoffman [2] but the field really took off in the 1970s. In 1970, Francois d'Heurle suggested that the compressive stresses measured in sputter deposited Mo films might be due to a “shot-peening” action of the depositing atoms. This effect is now generally accepted and goes by the name “ion-peening” (Ref. [3]).
In 1968, Klokholm and Berry proposed constrained shrinkage as an explanation for the measured tensile stresses in evaporated metal films [4]. In 1972, Doljack and Hoffman published their paper on tensile stress in evaporated Ni films [5]. In that paper, they argue that tensile stress is generated at the grain boundaries. For Cr- and CrN-films, we now have evidence that the Doljack and Hoffman explanation is appropriate, see section “Stress in zone T films”. For WC–W multilayers, constrained shrinkage may very well be the appropriate description in view of the very high apparent interfacial stresses (see section “Stress in multilayers”).
In 1978, Abermann published the first in-situ stress measurements on a growing film [6]. These measurements, made during evaporation of Ag, show a complicated compressive–tensile–compressive–tensile (CTCT) behavior in the development of stress with thickness. In Fig. 1 a result by Abermann et al. is reproduced. This result, today, still lacks a satisfactory explanation.
Various aspects of the field have been reviewed by Hoffman in 1976, Doerner and Nix in 1988, Windischmann in 1992, Koch in 1994, and Spaepen in 2000 (Refs. [7], [8], [9], [10], [11]). In the present review, I shall limit myself to polycrystalline films that have no epitaxial relation to the underlying substrate.
Films or coatings on a substrate are usually in a stressed state. The film “wants” to be smaller or larger than the substrate allows it to be, hence the film is in tensile stress (film “wants” to shrink) or compressive stress (film “wants” to expand). The unit of stress is Pascal [Pa]. Stresses in thin films typically range from minus a few GPa to plus a few GPa. Tensile stresses are designated by a positive sign, compressive stresses by a negative sign.
The trivial reason for a film to be in a stressed state is deposition at a different (higher) temperature than the temperature at which the stress is measured. Due to a difference in thermal expansion coefficients between the film and substrate, the film is in a stressed state. This dependency in differential form is [8]:with σf the stress in the film, T the temperature, αs the coefficient of linear thermal expansion of the substrate, αf the coefficient of linear thermal expansion of the film, and Ef / (1 − νf) the biaxial modulus of the film. In addition to this thermal stress, an intrinsic stress may also be observed. The present review is entirely concerned with this intrinsic stress. The stresses observed during the thermal cycling of thin films on silicon wafers are not considered in this paper. These stresses are the combined effect of Eq. (1) and temperature dependent stress relaxation mechanisms. A starting point for the study of stress during thermal cycling is the review by Doerner and Nix [8].
Large stresses have to be avoided for many applications. Large tensile stresses may lead to cracking of the film. Large compressive stresses may lead to buckling [12], [13], [14]. Conventional wisdom is to strive for a compressive stress of a few hundred MPa. This can be understood based upon results from the widely applied “scratch test” for protective coatings. In this test, a diamond stylus is pushed into the coated material with a prescribed force and subsequently moved laterally [15]. In the wake of the moving indenter, a tensile stress occurs leading to film cracking. In the case for which the film was under compressive stress prior to the “scratch testing”, a higher load can be applied before the resulting stress becomes tensile and the film fails [16].
We have knowledge of stress in thin films in two ways: (1) by directly measuring the crystal lattice strain in the film using X-ray diffraction [17] and (2) by measuring the elastic deformation of the substrate, usually a single crystal silicon wafer. The wafer becomes curved due to the stress in the film [18]. Since we know the elastic properties of the wafer, we can calculate the stress in the film. Mechanical equilibrium requires the net force (F) and bending moment (M) equilibrium to vanish on any film/substrate cross section [15]. Thusandwith σ(z) the stress at height z in the film, z the moment lever arm, and A the sectional area through the film and substrate.
In this review, I will treat stress measurement by wafer curvature, both conceptual as well as practical. I shall touch only slightly on X-ray strain measurements, since an excellent review on that topic has been published recently [17]. The actual stress measurements, that are discussed, are ordered according to the observed microstructure, either zone T or zone II, (Refs. [19], [20], [21]). A special topic, with its own section is stress relaxation in c-BN.
Section snippets
Stress measurements by wafer curvature
Stress in a thin film attached to a flexible substrate will induce a bending of that substrate. Stoney used steel strips for his experiments. Today most substrate curvature stress measurements are performed on silicon wafers, hence the name “wafer curvature measurements”. Since the film exerts a force on one side of the substrate, the substrate bends with a curvature 1 / R ∝ σftf, where σf is the average stress in the film and tf is the thickness of the film.
In Eqs. (1), (2a), (2b), stress was
Practical wafer curvature measurements
In Eq. (6), the term 1 / R represents the curvature of the wafer. This curvature is proportional to the product of stress in the film and the thickness of the film. In most of our films, the stress is biaxial and rotationally symmetric, therefore 1 / R[110] = 1 / R[11¯0], but we record both curvatures routinely to check this equality. In this section, we will use 1 / R, assuming that indeed the film stress is biaxial and rotationally symmetric.
Wafer curvature measurements are carried out both in-situ and
Some remarks on X-ray diffraction strain measurements
In a recent review, Welzel [17] treats strain measurements by X-ray diffraction extensively. The reader is referred to that review. Here only the basic concept of the sin2Ψ method is treated together with a remark on absolute lattice spacing measurements.
In elastically strained crystalline material, the lattice spacing depends on the strain. In polycrystalline material, this dependence is used in the so called “sin2Ψ method” to determine the strain in the material. For a biaxial rotationally
Types of thin film microstructure
Thin films are far from thermodynamic equilibrium. Thermodynamically, the film material would “prefer” to be compact, bounded by low index planes. The same holds for the substrate. Together the film and substrate might form a joint structure in the case for which the total free energy is reduced by the formation of an interface. In all practical cases, however, we are far from this situation. The thin films we produce are a consequence of kinetic restraints. The possible range in
Stress in “zone T” films
In zone T films, the number of grain boundaries in the film is a function of the height in the film. In the analysis by Doljack and Hoffman [5], tensile stress is generated at the grain boundaries by shrinking the distance between adjacent grains due to interatomic forces across the grain boundary together with the adherence of the film to the substrate. Since in zone T films, the number of grain boundaries is a function of the height in the film, the stress in the film is also a function of
Stress relaxation in c-BN
c-BN is a very hard and wear resistant material. c-BN films have been studied since the late 1980s [55], [56]. The problem with this material is the metastable nature of the cubic phase. The material has to be forced into this phase. Deposition conditions leading to the desired cubic structure also give rise to a compressive stress of the order of 10 GPa in the films. Recently, it has been shown that the compressive stress in c-BN films can be relaxed by irradiation during deposition with ions
Growth stress in zone II films
The microstructure of “zone II films” (Tdep / Tmelt > 0.4 for evaporation, different temperature windows for other deposition methods) [20], [21] does not reflect the growth history. Grain growth of the larger grains at the expense of the smaller grains occurs during the deposition process. This development of the microstructure during deposition complicates finding the origin of the observed stresses in these films. Several researchers have worked on in-situ wafer curvature measurements during
Multilayers, interfacial and surface stress
Surface energy arises from the absence of binding on the vacuum side of the atoms in the outermost layer of the sample. Surface energy is always positive, otherwise the bulk phase would not be stable. Surface energies are of the order of 1–10 J/m2. Surface stress is related to, though not equal to, surface energy despite the fact that the units [J/m2] or [N/m] are identical. According to Sutton and Balluffi [79], the surface stress fijS is a symmetric tensor of rank two. The surface stress is
Conclusions and prospects
For some “zone T films” (Cr, CrN) a satisfactory description of the stress in the films has been reached based on the observed microstructure and the known ion bombardment during deposition [34], [44], [45]. For other “zone T films” (e.g. TiN) the observed stress is not yet correlated to the microstructure [52], [53]. For these films, ex-situ analyses of the films by TEM and X-ray diffraction offer a fair chance to elucidate the data.
For the “zone II films”, the need for in-situ structural
Acknowledgement
Th. H. de Keijser, of our laboratory is acknowledged for the stimulating discussions on an earlier version of this paper.
References (81)
Proc. R. Soc. Lond., A
(1909)- et al.
J. Appl. Phys.
(1959) Metall. Trans.
(1970)J. Electrochem. Soc.
(1968)- et al.
Thin Solid Films
(1972) - et al.
Thin Solid Films
(1978) - et al.
CRC Crit. Rev. Solid State Mater. Sci.
(1988) Crit. Rev. Solid State Mater. Sci.
(1992)J. Phys.: Condens. Matter.
(1994)