We have determined the optical constants in the energy range 0.1-6 eV for bulk Cu, Ag, and Au using Kramers-Kronig analysis of previously unpublished reflectance data. The results are compared to those commonly used from the literature. 2 Studies of the optical properties of the transition metals culminated in 1981 in a two volume set of figures and tables [1, 2]. These volumes have gone out of print but the data were abridged and published in the Handbook of Chemistry and Physics [3]. The originals have also been scanned and are available on one of the authors' websites [4]. Other authors also compiled data sets, including the well known volumes by Palik [5] which included the transition metal data. To our knowledge, there have been no more recent experimental studies, such work having gone out of vogue as investigations of the band structure through photon spectroscopies have yielded to photoemission and related spectroscopies that probe E(k) with exquisite precision. The optical constants of the noble metals Cu, Ag, and Au are important to those who use the data for such studies as photonics, plasmonics and nanoparticle arrays [6]. While the optical constants in Ref. 2 have stood the tests of time for Au, those for Cu and Ag were taken from tabulations by Hagemann, Gudat, and Kunz (HGK) [7, 8] who in turn relied on the results by Johnson and Christy (JC) [9]. To provide a critical reassessment of the optical properties of Cu and Ag, we returned to some largely unpublished absorptivity measurements by Weaver et al. While the reflectance results were shown in Ref. 2, the authors were interested at that time in structure in the dielectric constants and those structures were well described by JC [9] (see the comparison figures in Ref. 2). Here, we are interested in the magnitudes of  and N, the complex dielectric constant and complex index of refraction, and we performed Kramers-Kronig analyses to compare with prior results in the 0.1-6 eV range. Extensive comparisons that extend to 30 eV are available from the Weaver website [4]. The samples were mechanically polished single crystals that were chemically cleaned to remove work damage, as described previously [1, 2]. The measured quantity was the absorptivity, A = 1-R where R is the reflectivity. In the infrared, R is nearly 100%, and a measurement of A rather than R provides inherently greater accuracy. For Ag, to extend the spectral range for the KK analyses, we used the results from Leveque, Olson, and Lynch to 30 eV [10] For Cu, we tied our results to those of JC [9] and then HGK to 30 eV. Standard power law extrapolations of the form R = RoE-3.5 were used to 1000 eV.

We describe an atomic layer etching (ALE) method for copper that involves cyclic exposure to an oxidant and hexafluoroacetylacetone (Hhfac) at 275 • C. The process does not attack dielectrics such as SiO 2 or SiN x , and the surface reactions are kinetically self-limiting to afford a precise etch depth that is spatially uniform. Exposure of a copper surface to molecular oxygen, O 2 , a weak oxidant, forms a ∼0.3 nm thick layer of Cu 2 O, which is removed in a subsequent step by exposure to Hhfac. The etch reaction involves disproportionation of Cu(hfac) intermediates, such that ∼0.09 nm copper is removed per cycle. Exposure of copper to ozone, a stronger oxidant, affords ∼15 nm of CuO; when this oxidized surface is exposed to Hhfac, 8.4 nm of copper is removed per cycle. The etch products, Cu(hfac) 2 and H 2 O, are efficiently pumped away; H 2 O, a poor oxidant, does not attack the bare Cu surface. The roughness of the copper surface increases slowly over successive etch cycles. Thermochemical and bulk etching data indicate that this approach should work for a variety of other metals.

We describe an example of a new phenomenon: the use of a growth inhibitor to homogenize nucleation and improve the smoothness of a thin film deposited by chemical vapor deposition. For many film–substrate combinations, the rate of nucleation on the substrate is slow relative to the growth rate, a situation that produces a broad distribution of island sizes and a rough surface morphology. We show an example in which this outcome is avoided by directing a second component onto the substrate that has little effect on the nucleation rate but significantly retards the island growth rate. The case studied is the growth of HfB2 films on SiO2 substrates using the chemical vapor deposition precursor Hf(BH4)4 with NH3 as the inhibitor. The addition of the inhibitor increases the island density at coalescence by 50× and decreases the roughness by 10× to the subnm range. We suggest that the use of inhibitors to homogenize nucleation may be applicable to other film–substrate combinations.

We report a method to control the surface morphology of thin copper films during growth by chemical vapor deposition from the precursor Cu(hfac)VTMS. A molecular inhibitor -an additive that modifies the surface attachment kinetics but does not decompose and contribute impurity atoms to the film -is added during the nucleation and/or growth stages of the film. Here we show that the reaction by-product VTMS can serve as such an inhibitor. If the inhibitor is added during the nucleation stage, when bare substrate surface is still exposed, the inhibitor greatly reduces the rate of coalescence and promotes the formation of a large density of uniformly-sized copper islands. Alternatively, if the film is allowed to nucleate in the absence of the inhibitor, subsequent addition of the inhibitor leads to a continuous copper film that is remarkably smooth on the nm scale. © 2014 The Electrochemical Society. [DOI: 10.1149/2.009405jss] All rights reserved.Manuscript submitted January 14, 2014; revised manuscript received February 27, 2014. Published March 11, 2014 Copper is used in many advanced nanoscale technologies due to its high electrical and thermal conductivity, and its strong surface plasmon resonance when in the form of nanoparticles.1-5 For continuous films, such as those used as the seed layer for electrodeposition in integrated circuits, the film must be less than 10 nm thick, pinholefree, and extremely smooth, with an rms roughness of less than 1 nm. For optical devices based on copper nanoparticles, it is important to control the nanoparticle size and morphology. 4,6 Rigorous control of copper growth can be difficult: the surface energy of copper is high and the atomic diffusion rate is significant, so that dewetting often occurs during growth or subsequent annealing. 7-12Thin films of copper can be deposited by a wide variety of techniques including wet chemical growth, physical vapor deposition, chemical vapor deposition (CVD) and atomic layer deposition (ALD). To deposit copper conformally in substrate architectures such as trenches and vias that have re-entrant or high aspect ratio features, ALD and CVD are preferred techniques because of the ability of the precursor molecules to diffuse throughout the structure.13-17 A general difficulty arises when the substrate is relatively unreactive, such as an oxide surface: the resulting films tend to be rough owing to a combination of sparse nucleation and the tendency of the deposited material to agglomerate. 18 Once surface roughness on the length scale of the island separation is formed, it cannot be eliminated by the overgrowth of more material. 18The use of additives to enhance film smoothness is well established in the electrochemical deposition of copper 19,20 but is not common in CVD. For CVD, the morphology of copper films can sometimes be improved by adding a second component to the growth gas. For example, addition of H 2 O to a flux of Cu(hfac)VTMS (hfac = hexafluoroacetylacetonate and VTMS = vinyltrimethylsilane) enhances the wettability of the surfac...

The deposition of Au onto thin Xe films at a low temperature leads to cluster formation. The subsequent Xe sublimation results in cluster aggregation and delivery to the substrate in a process known as buffer-layerassisted growth. Previously, this process was described in terms of a diffusion-limited cluster-cluster aggregation process during layer-by-layer desorption of the buffer. Instead, significant diffusion, restructuring, and dewetting of the Xe occur prior to desorption, and this leads to cluster aggregation. Cluster motion and aggregation are driven by capillary forces as the dewetting film retreats and sublimes. Kinetic Monte Carlo simulations reproduce the experimentally observed particle shapes and size distributions, and they provide additional insight into the interaction of the particles with the dewetting front. The presence of nanoscale particles on the film inhibits dewetting and significantly alters the shape of the front.