Selected Success Stories from Twenty Years of High-k Gate Dielectric Research

Abstract

There are many success stories from the high-k gate dielectric research, carried out during the last two decades or so (say, 1996 to 2016), only a few of which could be accommodated in this paper. We plan to highlight four successful enterprises, namely, (1) Atomic Layer Deposition (ALD) of HfO2 (including HfON, HfSiON), (2) Enhancement of the dielectric constant k through structural transformation and/or compound formation, (3) Role of the SiO2/high-k Interface Dipole and the Cap Layer at the high-k/metal interface in controlling the flat-band voltage, (4) Passivation of the Ge surface and Ge MOSFET with high channel mobility. These topics will be analyzed critically and the underlying physics will be presented. Atomic Layer Deposition scores over competitors (MOCVD, Magnetron Sputtering) in terms of better film uniformity and conformity (against uneven surfaces) with less pinhole. The alkyl-amide precursors: (TDMAH - Hf[N(CH3)2]4), (TEMAH - Hf[N(C2H5) (CH3)]4), (TDEAH - Hf[N(C2H5)2]4) were found to be suitable for low temperature deposition of smooth (RMS roughness ca. 1% of film thickness), uniform (thickness and composition wise), dense (ρ = 9.23 gcm-3, i.e. 95% of the density of the monoclinic phase), low contamination (< 1% C and < 0.25% N), and conformal (100% site coverage on holes with > 35 aspect ration) films of HfO2 with a deposition rate of ca. 0.093 nm per cycle.ALD is claimed to be a self-limiting process, but it was found that self limiting surface reaction occurred at a particular precursor temperature corresponding to a specific precursor dose; these precursor temperatures for the saturating dose were 75, 115, and 130 oC for TDMAH, TEMAH, and TDEAH, respectively. High deposition temperatures culminated in precursor decomposition followed by non-uniform films or no deposition on the substrate; moreover, HfO2 film surface roughness increased due to growth of crystallites. One main concern is the contamination of the deposited high-k layer by the chemical reaction by-products – generally C, Cl, N, and H. Post-deposition O3 treatment can reduce C, H contamination by producing H2O, CO or CO2 followed by their out-diffusion. Higher deposition temperature results in lower C contamination. High temperature annealing between ALD stages has been found to lower the C contamination; likewise has been NH3/Ar plasma treatment. Typically, the value of the dielectric constant k for a high-k alloy varies in a monotonous (somewhat linear) fashion as a function of the mixing ratio between those of the two components of a binary alloy. Departure from the above trend can happen, as experiments suggest, if a major structural transformation of the alloy matrix takes place. HfO2 can occur in three crystal structures – monoclinic, cubic, and tetragonal – with different values of permittivity. The cubic and the tetragonal HfO2 have higher dielectric constant k (from theory – 29 and 70 respectively) than the monoclinic polymorph. Permittivity is determined by the ratio αm/Vm, where αm is the molar polarization and Vm is the molar volume. Analysis suggests that a reduced molar volume may be mainly responsible for the larger k value of the cubic and tetragonal HfO2. Unfortunately, at room temperature and up to about 1750 oC, the stable phase of HfO2 is monoclinic - one with the lowest value of k among the polymorphs. The option therefore lies in finding phase modifiers, which can transform the low-k to the high-k phase and stabilize that phase in the temperature range the MOSFET is subjected to during processing. Experiments have revealed the ability of some dopants, e.g. Y, Si, La, Gd, Er, Dy, Sc, Ce, to be effective phase modifiers of HfO2. An interface dipole refers to an electrical bilayer of equal and opposite charges located within atomic distance on each side of an interface; as the electric field in the interface dipole will have a Dirac delta form, it will be represented by an interface dipole potential, which has no dependence on the physical thickness. There exists strong evidence to indicate that a dipole is present across the IL/high-k interface, when the IL is an SiO2 layer. Elements of the SiO2/high-k interface dipole model have been affirmed by the results of many investigations. Perhaps, its strongest support comes from the wide technological use of a capping layer in high-k MOSFET manufacturing to control and adjust the threshold voltage. The experimental data reported by many investigators appear to suggest that the SiO2/high-k interface dipole is the dominant entity controlling the flat-band voltage VFB.