Temperature dependence of noise processes in metals

Here we consider the analysis of the temperature and frequency dependences of the power spectral density $S_V$(f,T) of excess low-frequency noise. Two limiting cases that arise naturally from a superposition of thermally activated relaxation processes are investigated. The Lorentzian spectra associated with various relaxation times τ=${\tau }_0$ $e^{E/\mathrm{kT}}$ are supposed to arise from either (i) a distribution D(E) of activation energies E (invoked in the Dutta-Horn model), or (ii) a distribution H(${\tau }_0$) of prefactors ${\tau }_0$ (as required for activated diffusion processes). Much can be learned from the analysis of noise spectra with distinctive spectral features or known temperature-dependent variance 〈δ$V^2$(T)〉, as the analysis of noise from ${\mathrm{H}}^{+}$ diffusion in Nb films has illustrated. In the absence of such distinguishing features or independent data, particularly with 1/f noise, we find that there is generally insufficient information to distinguish even these limiting models, much less to determine their parameters. We have analyzed the resistance fluctuations of small-metallic-film conductors. We find that existing 1/f noise data from noble metals can be fit by either model with a wide range of assumptions about the variance 〈δ$V^2$(T)〉. In contrast new 1/f noise data from Cr films are inconsistent with either model unless strong ad hoc temperature-dependent variance is postulated.