Physica A: Statistical Mechanics and its Applications
Possible origin for the experimental scarcity of KPZ scaling in non-conserved surface growth☆
Introduction
The kinetic roughening of surfaces [1] is a subject of wide interest, due both to its implications for processes of technological relevance [2], [3], [4], and to the interesting instances that it offers of extended systems evolving in the presence of fluctuations [5]. A very successful theoretical framework for the study of rough interfaces has been the use of stochastic growth equations for the interface height. Among these, the one proposed by Kardar, Parisi and Zhang (KPZ) [6] has played a prominent role, since in particular it has allowed to make connections to other physical problems, like directed polymers in disordered media or randomly stirred fluids [1]. Denoting by the surface height at time t above position on a reference substrate plane, the KPZ equation is the simplest nonlinear coarse-grained description of a surface growing (or eroding) at a constant average rate v along the local normal direction, and readswhere ν is a positive constant, and is an uncorrelated Gaussian noise representing fluctuations, e.g., in a flux of depositing particles. Thus, the KPZ equation was postulated to describe generically the dynamics of surfaces growing in the absence of additional conservation laws, and is thus expected to be relevant to such diverse physical growth systems as erosion by low energy ion-beam sputtering (IBS) [2], electrochemical deposition (ECD) [4] or chemical vapor deposition (CVD) [3]. The generality of KPZ scaling would be a consequence of the phenomenon of universality observed for the scaling properties of rough surfaces.
However, despite some attempts at measuring KPZ scaling in, e.g., IBS [7] or ECD [8], to date very few experiments have been reported which are unambiguously described by the KPZ equation [9], [10], [11]. The identification of the physical mechanisms responsible for this paradoxical (termed “anomalous” in some early studies of kinetic roughening) non-KPZ scaling behavior has been complicated by two main reasons: on the one hand, while detailed derivations of the KPZ equation have actually been achieved, they apply to discrete or continuous theoretical models [12] which are indirectly related with experiments, or else the derivations themselves need resort to approximations which are not free from ambiguities.1 On the other hand, symmetry arguments such as those leading to the KPZ equation do not enable a detailed connection with phenomenological parameters describing specific experimental systems. These facts have led to invoking additional effects on a coarse-grained level, such as specific noise statistics, non-local effects, etc. [1], in order to account for the difference between the observed and the predicted scaling behaviors of rough surfaces. However, a wide range of scaling exponents ensued, there being no theoretical argument that could identify the correct exponents for a specific growth experiment.
Physically, many of the experimental growth systems expected to be in the KPZ universality class—and which seem to fail such expectations—feature dynamical instabilities leading to development of large slopes on the surface, and even to the production of characteristic surface features whose presence breaks scale invariance. A prototype example is IBS, where standard experimental conditions lead to the production of nanometric ripples [14] or dots [15] on the surface at short and intermediate times. Actually, it has been possible to derive [16] the relevant interface equation for these experiments, which for ions bombarding the target at perpendicular angles readsthe constants , KIBS and λIBS being in this case functions of phenomenological quantities, such as the angle of incidence, ion energy, ion penetration depth, etc. Eq. (2) is the noisy generalization of the Kuramoto–Sivashinsky (KS) equation [17]. The KS equation has been seen to describe other dynamic processes such as flame front propagation or chemical turbulence [17], or terrace growth in epitaxial systems [18]. Eq. (2) displays a linear instability at short times, there being a most unstable Fourier mode which induces the observed [14], [15] periodic array of ripples or cells. At very long times, scale invariance is restored, the values of the scaling exponents being—at least for a one-dimensional substrate—those of the KPZ universality class [19]. Hence, the specific example of IBS suggests transients due to instabilities as a physical reason for the difficulty in observing the asymptotic KPZ scaling, even if it should occur, since probing asymptotic times may well lie beyond experimental capabilities. For instance, in the CVD experiment in Ref. [11], deposition runs up to 2 days long had to be carried out in order to confirm KPZ behavior, which is both unusual and almost prohibitive. The IBS example illustrates another important fact: trying to assess the origin of non-KPZ behavior implies addressing pre-asymptotic features of the growth process under study. Hence, symmetry arguments do not suffice to derive the interface equation of motion, since contributions which are very important in the early dynamics can be overlooked. For instance, in the IBS system symmetry suggests a simple KPZ description [20], which misses the correct negative sign of the linear Laplacian term in (2) accounting for the physical ripple instability.
Section snippets
Unified equation for CVD and ECD: a stochastic moving boundary problem
Inspired by the example of IBS, one may wish to reconsider other growth techniques, such as CVD and ECD, for which KPZ scaling is expected to be relevant. These two types of experiments are conceptually similar, their simplest representations [21], [22] being essentially as processes in which particles from a vapor (CVD) or solution (ECD) diffuse until they react at an aggregate (CVD) or cathode (ECD) surface, onto which they stick leading to growth. This similarity can be made more precise, to
Conclusions and outlook
Summarizing, we have illustrated that instabilities appear rather generically in non-conserved surface growth, this occurring as long as the diffusion length lD is larger than the capilarity length d0, and independently of the sticking probability at the interface. Nevertheless, the mass transfer coefficient does control both the type of linear dispersion relation and the extent of the transients associated with the instabilities. In deriving a universal equation of motion it then seems that a
References (30)
- et al.
Phys. Rev. Lett.
(1982) - et al.
Fractal Concepts in Surface Growth
(1995)Adv. Phys.
(1997) - et al.
Ion–Solid Interactions: Fundamentals and Applications
(1996) - et al.
- et al.
Modern Electrochemistry
(1970) - et al.
Noise in Spatially Extended Systems
(1999) - et al.
Phys. Rev. Lett.
(1986) - et al.
Phys. Rev. Lett.
(1991) - et al.
Phys. Rev. Lett.
(1992) - et al.
Phys. Rev. B
(1997)et al.Phys. Rev. Lett.
(1997)
Phys. Rev. B
Phys. Rev. Lett.
Phys. Rev. A
Commun. Math. Phys.
Phys. Rev. E
Phys. Rev. Lett.
Cited by (11)
Dynamical renormalization group study for a class of non-local interface equations
2011, Journal of Statistical Mechanics: Theory and ExperimentKinetic roughening in a realistic model of non-conserved interface growth
2009, Journal of Statistical Mechanics: Theory and ExperimentThree-dimensional topography evolution algorithm for ion beam etching process
2014, International Journal for Multiscale Computational EngineeringModeling of feature profile evolution for ion etching
2013, Journal of Applied PhysicsDynamics of nanodots formation on irradiated materials
2012, Proceedings of the Institution of Mechanical Engineers, Part N: Journal of Nanoengineering and Nanosystems
- ☆
Work partially supported by DGES (Spain) Grant No. BFM2000-0006.