Sample dependence of the Casimir force

We have analysed available optical data for Au in the mid-infrared range which is important for a precise prediction of the Casimir force. Significant variation of the data demonstrates genuine sample dependence of the dielectric function. We demonstrate that the Casimir force is largely determined by the material properties in the low frequency domain and argue that therefore the precise values of the Drude parameters are crucial for an accurate evaluation of the force. These parameters can be estimated by two different methods, either by fitting real and imaginary parts of the dielectric function at low frequencies, or via a Kramers-Kronig analysis based on the imaginary part of the dielectric function in the extended frequency range. Both methods lead to very similar results. We show that the variation of the Casimir force calculated with the use of different optical data can be as large as 5% and at any rate cannot be ignored. To have a reliable prediction of the force with a precision of 1%, one has to measure the optical properties of metallic films used for the force measurement.