Copyright 2004 by Marcel Dekker, Inc. All Rights Reserved.

shows the dephasing times T 2 derived from the plasmon bandwidth as afunction of the particle radius [15]. The dephasing times are extremely short(sub-10-fs scale). Because the dephasing time is inversely proportional to thelinewidth, the broader the line is, the faster the dephasing is. For smallparticles, the faster dephasing time can be attributed to the enhanced electronsurfacescattering leading to the loss of coherence. On the other hand, forlarge nanoparticles, retardation effects due to inhomogeneous polarization ofthe nanoparticle determine the dephasing time.The above analysis assumes a monodispersity of metal nanoparticles,whereas real samples of metal colloids are always polydisperse (for goldcolloids, the size dispersion is typically f10%). However, from the obviousshift of the plasmon band maximum for spherical gold nanoparticles withaverage sizes that only differ by a few nanometers, one can conclude that thelinewidth is mainly determined by dephasing processes in the individualparticles despite the sample-size polydispersity. Single-particle studies[14,16,49,50] are well suited to measure the homogeneous linewidth and,hence, T 2 , without the problems associated with the inhomogeneous sizedistribution. In Ref. 14, a conventional microscope with illumination from ahalogen lamp was used to measure homogeneous linewidths for individualgold nanodots and nanorods [14] (these experiments took advantage of largescattering cross sections of metal nanoparticles). Dephasing times of 1–5 fs forspherical gold nanoparticles were found, in agreement with results in Fig. 4b.On the other hand, much longer dephasing times (up to 18 fs) for goldnanorods were derived from the homogeneous width of the longitudinalsurface plasmon resonance. These results were explained by the decreasedspectral overlap between the longitudinal plasmon resonance at lowerenergies and the interband transition in gold. The results of single-particlestudies also indicated that pure dephasing (time T 2*) was negligible and theplasmon dephasing (time T 2 ) was dominated by nonradiative decay (time T 1 )into single-particle excitations [14].In the other single-particle study [49], 40-nm gold nanoparticles embeddedin polymer films of various thicknesses showed a strong dependence ofthe plasmon resonance frequency on the particle environment. In particular,the plasmon absorption red-shifted with increasing film thickness. In thethinner films, the particles were assumed to be partially surrounded by air,which accounted for the change in the local environment. Using a scanningnear-field optical microscope (SNOM), the near-field transmission spectra ofindividual gold nanoparticles with an average size of 40 nm were measured[50]. The dephasing times extracted from homogeneous linewidths werearound 8 fs. Deviations in linewidths observed for individual particles wereattributed to variations in the local environment. These examples clearlyCopyright 2004 by Marcel Dekker, Inc. All Rights Reserved.