Power Spectral Density of Rough Surfaces

power spectral density of a rough surface

Roughness is known to be an important factor in friction, wear and related performance of the machine parts. It was shown recently that Power Spectrum Density (PSD) can be used as a metric to describe the surface from tribological point of view. Besides the field of tribology, roughness determines surface properties, such as hydrophulicity, adhesion, fatigue resistance, etc. Some of the analytical models of contact mechanics predict contact area, stiffness and other important parameters based on the PSD. Therefore, knowledge of PSD of a surface can bring valuable information about its (tribological) properties.

Power spectral density reveals presence and contribution of different spatial frequencies (wavelengths, wavevectors) in the surface. For example, if the surface roughness is composed of two sine waves of different wavelengths, its PSD will have only two peaks (see the figure below). These peaks indicate presence of two wavelengths in the surface.

Power Spectral Density
Power Spectral Density of a surface composed of two sine waves of different wavelength

Power spectral density of a randomly rough surfaces was shown to frequently look like in the figure below:

PSD of a rough surface
Power spectral density of randomly rough surface

In this case, the PSD reveals a self-affine surface, which is not always the case (if not self-affine, it may look curved and not straight line).

There are various definitions of the power spectral density, which differ typically with a constant. Here we follow [1] and define the PSD as follows:


 C^{2D}_{q_x,q_y} = \frac{1}{A}*\abs{\widetilde{h}_{q_x,q_y}}^2 ,

where  C^{2D}_{q_x,q_y} is the 2 dimensional PSD, A is the area of the surface,  q_x, q_y are wavevectors in x and y directions. Fourier transform of the surface is given as follows:

 \widetilde{h}_{q_x,q_y} = \int_{A} h_{x,y} e^{-i(q_x+q_y)} dxdy.

Equivalent definition of the PSD is used by B. Persson [2]:

 C^{2D}_{q_x,q_y} = \int_{A}  <h_{x,y} h(0,0)> e^{-i(q_x+q_y)} dxdy ,

where the <..> denotes average of the ensemble. These definitions are equivalent due to Wiener-Khintchine theorem [2].


The definition of the power spectral density in terms of Fourier transform seems to be more convenient, since it also shows a direct way of calculating it for a given surface.


  1. [1] Quantitative characterization of surface topography using spectral analysis, Tevis Jacobs, Till Junge and Lars Pastewka.
  2. [2] On the nature of surface roughness with application to contact mechanics, sealing, rubber friction and adhesion, Bo Persson, O Albohr, U Tartaglino, A I Volokitin and E Tosatti

Administration of the project

1 Comment

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.