In atomic layer deposition (ALD), film thickness control by counting the number of deposition sequences is poor in the initial, nonlinear growth region. We studied the growth of TiN films formed by sequentially controlled reaction of TiCl4 and NH3 on thermal SiO2 during the transient, nonlinear period. Using low-energy ion scattering and Rutherford backscattering spectroscopy analysis, we have found that a three-dimensional growth of islands characterizes the ALD TiN growth on SiO2. Growth at different temperatures (350 °C and 400 °C) affects the extent of the transient region and the rapid closure of the film. At 400 °C, a reduced growth inhibition and an earlier start of three-dimensional growth of islands results in film closure at about 100 cycles, corresponding to a TiN thickness of 24±3 Å. At 350 °C the minimum thickness at which the TiN layer becomes continuous is 34±3 Å, deposited with 150 cycles.

Copper electrodeposition in fine recesses requires the use of several additives to induce a strongly accelerated deposition rate inside recesses. This leads to so-called bottom-up filling for fine recesses, induced by curvature-enhanced accelerator coverage.Here we show that more of the additives are incorporated in the deposit when this mechanism is active by studying the impurity concentrations with time-of-flight scanning ion mass spectroscopy in specially prepared samples with a high density of recesses. The undesired higher copper resistivity for small dimensions, induced by this incorporation, results in a reduction of the expected surface-scattering induced deviation from Matthiessen's rule.The resistivity of electroplated copper is of critical importance for interconnects between active areas in an integrated circuit. When the dimensions of such interconnects are reduced to close to the intrinsic electron mean free path, , of copper ͑39 nm at room temperature͒, there are several effects that play into the strong resistivity increase. 1-3 Intuitively it is expected that surface scattering will become increasingly important. However, grain growth is also inhibited by reduced dimensions and in fact the obtainable grain size during thermal annealing has been shown to scale with size. This results in the interesting complication that both surface and grainboundary scattering are dependent on the physical size in a similar manner, 4 rendering it difficult to separate the influence of these effects. Measuring the resistivity vs. temperature down to 4 K is seen as a reliable approach. As pointed out by van Attekum et al., 5 the temperature-dependent part of the resistivity predicted by the Mayadas-Shatzkes ͑MS͒ grain-boundary scattering model is almost identical to that of bulk material and is independent of film thickness. In contrast, surface scattering in the Fuchs-Sondheimer ͑FS͒ model leads to a nonlinear temperature-dependent contribution at cryogenic temperatures where is much larger than the sample size. Hence it results in the so-called surface-scattering induced deviation from Matthiessen's rule ͑SSIDMR͒. 6 A more complicated effect is that of impurity incorporation during electrochemical deposition. First, this can lead to an increase of resistivity simply by increasing the number of scattering sites. Second, as grain growth has been shown to push some impurities into the grain boundaries during recrystallization, thereby purifying the material, the grain-boundary reflection coefficient, R, can be modified. Also, impurities in a grain boundary can inhibit grain boundary motion, and thus reduce the obtainable grain size. All of these phenomena can have an influence on the background scattering length, , which enters into the FS model.The importance of impurities for the resistivity of electroplated copper has regained attention in light of the use of several additives in the electroplating chemistry for filling high aspect ratio recesses. These additives are needed to induce so-called bottom-up filling,...

Geometric linewidth and the impact of thermal processing on the stress regimes induced by electroless copper metallization for Si integrated circuit interconnect technology The dimensions of transistor gates are scaling to small dimensions and the relative variation of the edge, referred to as line edge roughness ͑LER͒, is addressed as one of the critical issues for front-end technology. These local variations deteriorate the device performance ͓Croon et al., Tech. Dig. -Int. Electron Devices Meet. 2002, 307; Croon et al., Proceedings of the ESSDERC, 2003͑unpublished͒, p. 22͔. However, also for backend, interconnects wires are scaling down and the relative contribution of the LER is increasing. The impact of LER for interconnects has rarely been studied yet. In this article we address the electrical performance of the resulting copper interconnects with additional LER generated with e-beam lithography. The added LER will have spatial frequencies and amplitudes that are similar to the current resists used.

Articles you may be interested inProcess variation-aware three-dimensional proximity effect correction for electron beam direct writing at 45 nm node and beyondIn situ transmission electron microscopy observations of 1.8 μ m and 180 nm Cu interconnects under thermal stresses Appl.

A metal-insulator-semiconductor (MIS) capacitor with La2O3 dielectric is proposedin this work as a sensor for measuring CO2 in air. In this device, a 10 nm thick La2O3 dielectriclayer, which serves as a CO2 sensitive material, was atomic-layer-deposited (ALD) on p-typesilicon. Change in the at band voltage (VFB) of the MIS capacitor due to the reactionbetween CO2 and oxide layer and its interfaces, is used as the gas sensitive parameter of thesensor. The deposition temperature for the La2O3 layer has been optimized for maximizingCO2 sensitivity. The process ow including post annealing (rapid thermal annealing) has beenoptimized to allow further possibility to integrate the sensor with CMOS read-out circuitries. The sensor shows a sensitivity of 84 mV per decade to CO2 in air in a concentration rangefrom 300-5000 ppm at ambient temperature with a response time (t90) of 34 minutes.