1/fnoise in mono- and polycrystalline aluminum

~Received 27 June 1997! The 1/f noise in three types of aluminum lines has been investigated in the temperature range 140-510 K. The types are one long single crystal, a chain of short single crystals ~''bamboo'' !, and a polycrystal. In the lines of the first two types the 1/f noise power is significantly lower than in the polycrystalline specimens. The temperature dependence of the noise power in the polycrystalline lines shows a plateau between 370 and 415 K, corresponding to activation energies 0.9-1.0 eV. Both types of monocrystalline lines have equal noise power with a peak around 340 K, corresponding to an activation energy of about 0.8 eV. In the polycrystalline lines the dominant contribution to 1/ f noise appears to be the thermally activated motion of atoms in grain boundaries. The measurements on the monocrystalline lines reveal the existence of at least one further contri- bution to 1/f noise in metals, presumably associated with the thermally activated diffusion of atoms along dislocations. @S0163-1829~98!06601-6# Resistance fluctuations in continuous metal films exhibit a power spectrum that is approximately inversely proportional to the frequency f . Mobile lattice defects are known to con- tribute considerably to these resistance fluctuations, 1 e.g., low-temperature irradiation with electrons, by which atomic defects are introduced, increases the 1/ f noise of metal films drastically. 2,3 Dutta, Dimon, and Horn ~DDH! demonstrated that a significant part of the 1/f noise in metal films can be attributed to thermally activated processes with activation energies and relaxation times characteristic for atomic diffusion. 4 In measurements of the 1/f noise versus tempera- ture for pure Al and for Al with Si and Cu additives, Koch and co-workers found that the additives increased the activa- tion energies. 5 A similar increase in the activation energy upon Si and Cu addition is generally observed for Al elec- tromigration ~atomic diffusion! along grain boundaries. 6 These experimental findings strongly indicate that a signifi- cant part of the 1/ f noise in polycrystalline metal films is due to diffusion along grain boundaries. Further support for the prevailing role of grain boundaries stems from the work of Verbruggen and co-workers. 7 In their comparison of several gold samples, the lowest 1/f noise intensity was observed in the samples with fewer grain boundaries. The question arises whether mechanisms not involving grain boundaries or atomic defects contribute to 1/ f noise in metal films. To answer this question, it is necessary to inves- tigate the relationship between 1/ f noise and the microstruc- ture of the films. Even though the importance of the micro- structure for 1/ f noise is commonly acknowledged, so far little work has been done to investigate this relationship in