Measurements of the scaling of \/f noise magnitude versus resistance were made in metal films as the metal was removed by sandblasting. This procedure gives an approximate experimental realization of a Swiss-cheese continuum-percolation model, for which theory indicates some scaling properties very different from lattice percolation. The ratio of the resistance and noise exponents was in strong disagreement with lattice-percolation predictions and agreed approximately with simple continuum predictions. PACS numbers: 72.70. + m, 05.40. +j, 05.70.Jk, 71.30. + h The critical scaling of various transport properties on percolating clusters provides a probe of the structure of those clusters. Recently, attention has been drawn to the possibility that the scaling of the mean square fluctuations S R in electrical resistance may provide information not obtainable from the scaling of R, the resistance itself. 1,2 In particular, since S R is sensitive to a higher moment of the current density distribution than is R it may be among those transport properties 3 " 5 for which the universality of scaling found in lattice models and many continuum models breaks down, according to theory. In this Letter we report measurements of the scaling of S R vs R in a simple experimental system which show unambiguously that the lattice models are inapplicable to continuum systems and which approximately confirm theoretical expectations for a simple continuum model.Monte Carlo simulations 1,6 of percolating clusters on a lattice (consisting of identical resistors with independent fluctuations) yield resistance and noise critical exponents K=1.12 and /3 L = -0.973, where the exponents are defined by R ~ £ L --A v L and S R /R 2 -A _K , with £ the percolation coherence length and p = p c + A the filling fraction. Halperin et al? have recently suggested that the permeability and elasticity exponents for the Swiss-cheese model, in which round holes with a fixed radius and randomly placed centers are removed from a material, are larger than those for the standard lattice model, while the resistance exponent vf3 L should not differ significantly. These Swiss-cheese calculations were made with a nodeslinks-blobs (NLB) model. 7 The agreement of the resistance exponent with the lattice value is also consistent with previous works, 4,5 which predict deviations only for singular distributions of single-link conductivities.Whether a quantity scales the same in a continuum model as in lattice models depends on how sensitive it is to the behavior of the weakest or narrowest links in the network. In other words, it depends on what moment of the current density, strain, etc., is probed by that quantity. Although the NLB model is not especially accurate for f3 L , which probes the second mo-ment of the current density and thus depends on blobs as well as links, it is expected to become increasingly accurate for higher moments, since the scaling of the number of links is known exactly. Recent experiments have shown that the conclusion that /3 L is unchang...

Low-frequency noise measurements were made in the temperature range of 77-340 K on submicrometer resistors made from GaAs gro~n by molecular-beam epitaxy. Two types of noise were found, depending on surface treatment. One type consisted of discrete spectral components which showed no anomalous statistical behavior. The other type was a small l/f component which sho~ed anomalously large variations in spectral density, with these variations themselves having a 1/f spectrum.

The dynamics of atomic motions in the amorphous conductors Si-Au and C-Cu were probed using conductance fluctuations in mesoscopic samples in the temperature range 4-300 K. Near 4 K the number of mobile sites was close to that expected from the anomalous heat capacity of amorphous materials. Although some sites showed two-state-switching kinetics, a variety of statistical indicators all indicated that intersite interactions were important. Amorphous materials show a surprisingly universal term in their low-temperature heat capacity, which is ordinarily attributed to two-level systems (TLS), largely on the basis of nonlinear acoustic properties. ' However, the extent to which it is appropriate to view such TLS as independent sites, or even whether the TLS picture might break down due to intersite interactions, remains uncertain. 3 In this paper we describe the use of conductance fluctuations in mesoscopic samples of amorphous materials to examine whether the mobile sites are generally independent two-state systems.The spectral densities of conductance noise in macroscopic samples of amorphous conductors has been shown to be consistent with a conventional picture of double-well systems (TLS at low temperatures) coupled to the conductance (below about 50 K) via universal conductance fluctuations (UCF), assuming a TLS density comparable to that expected from extrapolated heat-capacity measurements. According to standard UCF theory, 5 the conductance noise should be very nonselective, picking up all types of atomic motions, and thus is expected a priori to see the same mobile sites that appear in time-dependent heat capacity. Since the temperature range of our measurements is above 4 K, and particularly in view of the low conductivities of the materials, the relaxation of the TLS or other structural fluctuations should be dominated by the same phonon processes as in insulating amorphous materials. ' With mesoscopic samples we could directly measure the density offluctuating sites using the size of spectral features, and use statistical properties to test if the f1uctuators were actually independent TLS.The materials studied were Cl -"Cu"and Sii -"Au" (with x typically 0.15 to 0.20) prepared by rf sputtering of a composite target. Transmission-electron-microscopy analysis showed that the samples were almost entirely amorphous, although about 2% of crystalline inclusions occurred. Both samples of Si-Au and four of the five samples of C-Cu were in the metallic conduction regime (with resistivity in the 10 Acm range). One C-Cu sample was near the metal-insulator transition.A new double-step-edge technique, described in detail elsewhere, 6 allowed four-probe mesoscopic samples (necking down to small wires of 10 ' to 10 ' cm ) of continuous films to be prepared by ordinary photolithography. The approximate dimensions of the short, narrow necks in the samples (about 30 nm laterally by 150-nm length) were confirmed by scanning electron microscopy and by resistance measurements.Noise measurements were made with standard dc te...