To reduce theRCinterconnect delays and cross‐talk noise associated with the sub‐130 nm technology nodes, copper interconnects must be combined with low‐kinterlayer dielectrics (ILDs) having dielectric constantsk⩽ 2. In order to obtain sufficiently low dielectric constants, pores are introduced into ILD materials thereby lowering the average density of the ILD. Information concerning porous, low‐kthin‐films can be obtained non‐destructively using grazing‐incidence X‐ray reflectivity (XRR) methods. Specular XRR provides information on the thickness, roughness and porosity, whilst diffuse (non‐specular) XRR yields valuable information about the average pore‐size, ⟨D⟩, and the pore‐size distribution. In this work, we present XRR data from a spin‐on, porous low‐kdielectric material (MesoELK) deposited on a Si substrate. The XRR data were analyzed by fitting to simulated intensity curves using a genetic algorithm (GA). The diffuse XRR intensity curve was calculated using the distorted‐wave Born approximation (DWBA) and two common models were used to describe the microstructure of porous materials,i.e.the polydispered sphere model and the random two‐phase model. The resulting pore‐size distribution corresponds well with that obtained from complementary methods. The assumptions and limitations of the XRR method for the non‐destructive characterization of porous thin‐films are discussed. © 2003 American Institute of Physics