Faceted Anomalous Roughening

• Faceted anomalous roughening is a kinetic regime marked by the coexistence of facet-dominated morphology with separated global, local, and spectral scaling exponents.
• The analysis employs Fourier space methods, correlated-noise models, and nonlinear growth equations to distinguish faceted behavior from ordinary self-affine scaling.
• Practical insights include using structure factor diagnostics and applications to systems like epitaxial films, radial imbibition, and critical Ising interfaces.

Searching arXiv for relevant papers on faceted anomalous roughening and related anomalous scaling. Faceted anomalous roughening is a kinetic-roughening regime in which an interface is not described by a single self-affine exponent set, but instead develops facet-dominated morphology together with a separation between global, local, and spectral scaling exponents. In the generic dynamical scaling description, the distinctive faceted case occurs when the spectral roughness exponent exceeds unity and differs from the global roughness exponent, while related super-rough cases display a local exponent locked to unity and global roughness beyond the ordinary self-affine range. This phenomenology has been identified in correlated-noise growth equations, anharmonic and quenched-disorder elastic interfaces, epitaxial semiconductor films, radial imbibition fronts, and critical Ising-derived interfaces (Nascimento et al., 2011, Alés et al., 2019, Alés et al., 2019).

1. Scaling framework and classification

The modern description of anomalous roughening is formulated most cleanly in Fourier space. The structure factor is written as

$S(k,t)=k^{-(2\alpha+d)}\,s(kt^{1/z}),$

with asymptotic behavior

$$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$

Here $\alpha$ is the global roughness exponent, $z$ the dynamic exponent, and $\alpha_s$ the spectral roughness exponent. In real space, local height fluctuations are commonly expressed as

$$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$

A positive $\kappa$ signals anomalous roughening, because local fluctuations continue to evolve after the global rescaling has been taken into account (Nascimento et al., 2011, Alés et al., 2021, Pino et al., 11 Jul 2025).

The GDST-based classification used across the literature distinguishes ordinary self-affine roughening from anomalous variants by the relations among $\alpha$, $\alpha_s$, and $\alpha_{\mathrm{loc}}$.

Regime	Exponent condition	Morphological implication
Family–Vicsek	$$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$0	Ordinary self-affine scaling
Intrinsic anomalous	$$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$1, $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$2	Anomalous scaling without faceting
Super-rough	$$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$3, $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$4	$$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$5
Faceted anomalous	$$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$6, $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$7	Facet-dominated morphology

Within this scheme, super-roughness and faceted anomalous roughening are closely related but not identical. Several works use the real-space criterion $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$8 together with $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$9 to identify super-rough behavior; faceted anomalous roughening is the more restrictive case in which the structure factor reveals $\alpha$0 in addition to $\alpha$1. The practical consequence is that width-based diagnostics alone are generally insufficient for identifying faceting.

2. Correlated-noise routes in linear theories

A central theoretical result is that anomalous roughening does not require a nonlinear growth term. For the $\alpha$2-dimensional Edwards–Wilkinson equation

$\alpha$3

driven by spatiotemporally correlated Gaussian noise

$\alpha$4

the exact global exponents are

$\alpha$5

The interface becomes super-rough when

$\alpha$6

so that $\alpha$7. The same threshold appears in the local slope field $\alpha$8, whose roughness exponent is

$\alpha$9

Thus, super-roughening begins exactly when the slope field itself becomes rough. Above threshold,

$z$0

and local fluctuations acquire the slope-dominated form associated with faceted geometry. In the temporally correlated case $z$1, the threshold is exact at $z$2 (Alés et al., 2019).

This linear result is important because it relocates the origin of some apparently nonlinear anomalous roughening phenomena to the statistics of the driving itself. The correlated-noise mechanism already produces exponent splitting, rough slope fields, and a local exponent fixed at unity. The resulting picture is reinforced by the generalized elastic model, which identifies a broader morphological transition at

$z$3

For $z$4, local and global scaling coincide; for $z$5, the local roughness exponent is pinned to $z$6, yielding superrough morphology under both thermal and non-thermal conditions. In that framework, anomalous roughening is not restricted to standard surface growth, but extends to polymers, membranes, and fluctuating interfaces with long-ranged hydrodynamics (Taloni et al., 2012).

3. Nonlinear, localized, and disorder-driven faceting

In the Kardar–Parisi–Zhang equation with temporally correlated noise,

$z$7

extensive simulations of ballistic deposition and a discretized KPZ equation show a threshold-like transition from ordinary self-affine roughness to macroscopic faceting as the temporal correlation exponent increases. Above approximately $z$8 for ballistic deposition and $z$9 for the discretized KPZ equation, the height profile develops system-spanning linear pieces with sharp cusps, and the structure factor requires distinct exponents. At $\alpha_s$0 in ballistic deposition, the reported values are $\alpha_s$1 and $\alpha_s$2, with $\alpha_s$3. The proposed mechanism is localization of the auxiliary field $\alpha_s$4: in the columnar-noise limit, the multiplicative-noise equation for $\alpha_s$5 produces exponentially localized states, which map back to piecewise linear cusped profiles in $\alpha_s$6. The same work argues that the singularities found in perturbative dynamic renormalisation group calculations at $\alpha_s$7 are signatures of a genuine physical crossover rather than purely technical pathologies (Alés et al., 2019).

An anharmonic elastic interface in a temporally correlated random medium provides a parallel route to faceted anomalous scaling. For

$\alpha_s$8

the $\alpha_s$9 interface becomes faceted for

$$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$0

In that regime, the local roughness exponent saturates to

$$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$1

and the anomalous time exponent is

$$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$2

The same study concludes that anomalous roughening cannot exist in this model for $$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$3, because the Laplacian dominates asymptotically and restores single-exponent scaling (Alés et al., 2021).

A closely related quenched-disorder example is the anharmonic Larkin model, in which a nonlinear elastic term $$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$4 is added to the Larkin random-force problem. In $$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$5, numerical results show

$$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$6

for any finite $$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$7. The interface is therefore faceted, standard single-exponent two-point scaling fails, and the small-gradient approximation of the elastic energy density breaks down in the thermodynamic limit. This model also connects the static endpoint fluctuations to a family of Brownian functionals, interpolating between the random-acceleration process at $$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$8 and the Lévy arcsine-law problem at $$C(\ell,t)\sim \ell^{\alpha_{\mathrm{loc}}} t^\kappa, \qquad \kappa=\frac{\alpha-\alpha_{\mathrm{loc}}}{z}.$$9 (Purrello et al., 2018).

A distinct but related facet-controlled setting is the smooth $\kappa$0He crystal surface below the roughening transition. There, the nonanalytic cusp in the orientation-dependent surface tension,

$\kappa$1

forces crystallization waves to propagate as trains of flat facet segments connected by macroscopic steps rather than as ordinary harmonic modes. In the zero-width-step limit, the spectrum becomes amplitude dependent,

$\kappa$2

so that $\kappa$3. This is not a standard kinetic-roughening result, but it is a clear example of cusp-controlled facet dynamics (Burmistrov, 2010).

4. Geometric and critical realizations

Experimental radial imbibition in a Hele–Shaw cell gives a geometry-driven realization of anomalous roughening. In forced radial imbibition of water into a porous medium, the mean front radius obeys Washburn scaling,

$\kappa$4

while the rough front displays a strong separation between global and local exponents. The measured global growth exponents increase from $\kappa$5 at $\kappa$6 to $\kappa$7 at $\kappa$8, whereas the global roughness exponents decrease from $\kappa$9 to $\alpha$0. At $\alpha$1, $\alpha$2, and $\alpha$3, the interface is explicitly super-rough, with $\alpha$4 and $\alpha$5. The authors interpret the flow-rate dependence as competition between pinning by quenched disorder and imposed advective forcing, and they argue that radial geometry itself challenges geometry-independent notions of universality (Chen et al., 2015).

A complementary geometric mechanism appears in curvature-driven roughening of a radially expanding disk domain with a variable interface window. There the reported exponents are $\alpha$6, $\alpha$7, $\alpha$8, $\alpha$9, and $\alpha_s$0, together with the explicit inequalities

$\alpha_s$1

The key claim is that the time-dependent circumference acts as a variable observation window, producing a new anomalous roughening class distinct from fixed-window planar growth (Chen et al., 2012).

Critical Ising interfaces show that anomalous roughening is not confined to growth equations. In a random-interface representation of the three-dimensional Ising model, the full 3D membrane has a size-independent cusp in the interfacial width at $\alpha_s$2, while the 2D cross section becomes super-rough with

$\alpha_s$3

This is intrinsic anomalous scaling rather than faceted anomalous roughening, because the local and spectral exponents coincide below unity even though the global exponent is about one. The same work reports that these geometric exponents coincide with those of the pure 2D Ising model at criticality (Dashti-Naserabadi et al., 2019).

The reinterpretation of model-A critical dynamics of the two-dimensional Ising model as a surface roughening process sharpens the role of preparation history. For the original TDGL or Glauber dynamics, a critical quench from the ordered phase follows Family–Vicsek scaling, whereas a quench from the disordered phase shows an initial overgrowth regime with intrinsic anomalous roughening. For the related integral GL model, the ordered quench again gives FV scaling, but the disordered quench yields faceted anomalous roughening with a measured spectral exponent $\alpha_s$4. This difference is traced to a linear instability of the critical stochastic Ginzburg–Landau equation, followed by nonlinear stabilization (Pino et al., 11 Jul 2025).

5. Experimental diagnostics and measurement issues

The decisive observable for faceted anomalous roughening is usually the structure factor, not the global width. The KPZ study with temporally correlated noise explicitly shows that the width $\alpha_s$5 obeys a standard finite-size scaling form and does not reveal $\alpha_s$6; the spectral anomaly is encoded in $\alpha_s$7. This is why the high-$\alpha_s$8 sector, and specifically the separation between the envelope exponent and the tail exponent, is central to diagnosis (Alés et al., 2019).

The epitaxial growth of CdTe polycrystalline films on glass covered with SnO$\alpha_s$9:F provides a clear experimental demonstration. GDST analysis of stylus-profiler power spectra at $\alpha_{\mathrm{loc}}$0C gives

$\alpha_{\mathrm{loc}}$1

Since $\alpha_{\mathrm{loc}}$2, the surface falls in the faceted anomalous class. At $\alpha_{\mathrm{loc}}$3C, the spectral exponent is even larger, $\alpha_{\mathrm{loc}}$4. Yet a naive real-space fit to the profiler correlation function yields $\alpha_{\mathrm{loc}}$5, which would suggest a self-affine surface. The discrepancy is attributed to convolution effects caused by the finite probe tip, whose effective lateral resolution is estimated as $\alpha_{\mathrm{loc}}$6. High-resolution AFM images then directly confirm the faceted grain morphology predicted by the power-spectrum analysis (Nascimento et al., 2011).

This measurement problem is conceptually important. Faceted anomalous roughening can be missed if one probes only coarse real-space observables or only the global width. Conversely, apparent local self-affinity does not exclude faceting when instrumental smoothing suppresses short-scale slopes.

6. Universality, misconceptions, and current interpretation

A recurring issue is whether faceted anomalous roughening defines a single universality class. Several studies argue against such a strong formulation. Radial imbibition and curvature-driven growth both report that geometry and a variable interface window modify the measured exponents and can even challenge the notion that roughening universality is geometry independent (Chen et al., 2015, Chen et al., 2012). The Ising model A study adds that the same critical dynamics can show FV, intrinsic anomalous, or faceted anomalous scaling depending on the initial condition and on whether the field or its spatial integral is analyzed (Pino et al., 11 Jul 2025).

Another common misconception is that faceting necessarily requires nonlinear growth. The exact Edwards–Wilkinson analysis with spatiotemporally correlated noise shows that a linear diffusion equation already generates super-roughening, rough slope fields, and $\alpha_{\mathrm{loc}}$7 when the noise correlations exceed the threshold $\alpha_{\mathrm{loc}}$8. This suggests that at least some nonlinear faceted regimes inherit their scaling structure from correlated-noise mechanisms already present in linear theory (Alés et al., 2019).

A further clarification concerns the relation between anomalous roughening and faceting. They are not synonymous. Intrinsic anomalous roughening corresponds to $\alpha_{\mathrm{loc}}$9 with $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$00, as in the 2D Ising cross section and the disordered TDGL quench. The two-species sandpile model in $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$01 dimensions likewise emphasizes anomalous crossover from logarithmic smoothing to asymptotic power-law roughening, but does not formally classify the long-scale regime as a faceted universality class, even when some exponents exceed $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$02 (Dashti-Naserabadi et al., 2019, Chakrabortty et al., 2012). By contrast, faceted anomalous roughening requires direct evidence that the spectral exponent exceeds unity and that the morphology is controlled by facets, cusps, or piecewise linear segments.

Taken together, these results suggest that faceted anomalous roughening is best viewed as a morphology-and-scaling regime rather than as a single microscopic model class. Its defining content is the coexistence of facet-like geometry with exponent splitting—typically $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$03, $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$04, and in many super-rough realizations $$s(u)\sim \begin{cases} u^{2(\alpha-\alpha_s)}, & u\gg 1,\ u^{2\alpha+d}, & u\ll 1. \end{cases}$$05—but the mechanisms capable of producing that regime include correlated noise, localization, anharmonic elasticity, quenched disorder, variable geometry, and critical fluctuations.