« PreviousUpNext »Contents
Previous: 4 Point defects and their correlation to hydrogen    Top: 4 Point defects and their correlation to hydrogen    Next: 5 Bias dependence of degradation and recovery

4.4 The \( \textrm {K}’ \) center

Due to process innovations going along with continuous device scaling, an alternative type of gate oxide insulator increasingly gained importance, namely silicon nitride (Si(math image)N(math image)). Silicon nitrides and nitrided oxides (NOs) have a number of advantages over pure silicon dioxide (SiO(math image)) insulators, for example, they have a higher dielectric constant (high–\( \kappa \)), a lower gate leakage current due to the larger oxide thickness, a denser structure and a better resistance to hydrofluoric (HF) acid than SiO(math image) [123]. The higher dielectric constant provides larger capacitances at lower oxide thicknesses, while the denser structure makes the oxide a better barrier against the diffusion of various impurities like doping atoms, mobile ions and moisture [124].

The nitridation process is realized in different ways depending on the oxide thickness. While thin oxide devices (<10\( \,\mathrm {nm} \)) undergo a so-called thermal nitridation where a pure silicon surface is exposed to N\( _\mathrm {2} \) or NH\( _\mathrm {3} \) at relatively high temperatures (950 °C – 1300 °C), thick oxide devices (>10\( \,\mathrm {nm} \)) are usually processed by plasma enhanced nitridation where the oxide is grown at lower temperatures by adding a plasma (photons, electrons or ions). A third method is realized by ion implantation where nitrogen ions are implanted directly beneath the surface of the silicon wafer followed by a high temperature anneal. The thickness of the nitride film is controlled by the implantation energy, the ion mass and the ion dose, thereby leading to a Gaussian-like implantation profile with poor interfacial properties.

By introducing oxygen either subsequently to the actual nitridation process (re-oxidized nitroxide (RONO) or rapid thermal oxynitridation (RTNO)), or directly during low pressure chemical vapor deposition (LPCVD) or plasma enhanced chemical vapor deposition (PECVD), some nitrogen atoms are replaced by oxygen, resulting in an SiON film (SiO\( _\mathrm {x} \)N\( _\mathrm {y} \)) which has a similar structure but much less tensile stress as opposed to the pure Si(math image)N(math image) layer. The bandgap energy of the SiO\( _\mathrm {x} \)N\( _\mathrm {y} \) layer can be adjusted between 5\( \,\mathrm {eV} \) and 9\( \,\mathrm {eV} \) depending on the [O]/[N] ratio [124].

Figure 4.6:  A schematic diagram of nitrogen related defects in silicon nitride and oxynitrides

As a consequence of nitrogen incorporation, the micro structure of the oxide is modified considerably which introduces some new defect types [125, 126] in plasma nitrided oxides (PNOs) and thermally nitrided oxides (TNOs). One of the new defect classes is labeled \( \mathrm {K}’ \) centers in pure Si(math image)N(math image) and \( \mathrm {K}’_\mathrm {N} \) centers in silicon oxynitrides [127, 128]. Following Fig. 4.6, both trapping centers are likely dangling bond defects in which the silicon is back-bonded to three nitrogen atoms, the dangling orbital electron being located at the central silicon atom [129, 130, 131, 132]. Indeed, there are only small discrepancies in the magnetic-resonance parameters between \( \mathrm {K}’ \) centers and \( \mathrm {K}’_\mathrm {N} \) centers which might arise from slightly different bonding environments in silicon nitride and SiON [133].

In SiON one or two of the second nearest neighbors of the central silicon atom may be oxygen atoms instead of silicon atoms. The \( \mathrm {K}’ \) center defects are assumed to be located in the near interfacial region [134] of the dielectric and have a narrow-peaked effective density of state profile near the middle of the silicon bandgap, hence, constituting very effective trapping centers in MOSFET devices. Besides \( \mathrm {K}’ \) centers, other defects like bridging nitrogen configurations have been identified in nitrided oxides as well, cf. Fig. 4.6. ESR spectroscopy revealed that one (paramagnetic) or two (diamagnetic) electrons can be captured in the N 2p\( _\pi \) non-bonding orbital of the brigding nitrogen defect which is oriented normally to the Si\( _\mathrm {2} \)N plane [135, 136].

The PNO process is known to enhance NBTI considerably because a lot of hydrogen is incorporated into the gate oxide during PECVD forming Si–H and N–H bonds within the gate oxide and at the interface [137]. Assuming that \( \mathrm {K}’ \) centers in nitrided oxides can be passivated by hydrogen in a similar way as P\( _\mathrm {b} \) centers in SiO(math image) oxides [138], oxynitrides provide a large number of hydrogen precursors which may be distorted during NBTI stress [139], explaining the similar degradation dynamics as observed in SiO(math image) devices [140]. Besides that \( \mathrm {K}’ \) centers and other nitrogen related defects have a much lower hole trapping barrier than \( \mathrm {E}’ \) centers, hence, positive oxide charge is assumed to be prevalent in nitrided oxides. It is also reported that \( \mathrm {K}’ \) centers represent effective hydrogen traps [141] and can therefore trigger hydrogen desorption from Si–H bonds at the interface and hence P\( _\mathrm {b} \) center creation [8]. This would explain the link between hole trapping and interface trap creation observed in nitrided oxides.

« PreviousUpNext »Contents
Previous: 4 Point defects and their correlation to hydrogen    Top: 4 Point defects and their correlation to hydrogen    Next: 5 Bias dependence of degradation and recovery
- b -->\( _\mathrm {b} \) centers become electrically reactivated as they become positively charged when the interface is flooded by holes during NBTS, cf. Fig. 1.2.

« PreviousUpNext »Contents
Previous: 4 Point defects and their correlation to hydrogen    Top: 4 Point defects and their correlation to hydrogen    Next: 5 Bias dependence of degradation and recovery
m " class="lateximage"> bandgap which may contribute to leakage currents and defect charge centers after electrical or ionizing-radiation stress.

Figure 4.5:  A schematic diagram of hydrogenated oxygen vacancies and their hydrogen related defect centers.

It has also been suggested that H\( _\mathrm {2} \) molecules can be cracked at \( \mathrm {E}’ \) centers, thereby creating various hydrogenated species of oxygen vacancies [93, 121, 122] and free hydrogen atoms that may interact with interface defects. Furthermore, Lenahan et al. [83] suggested that there may be hydrogen exchange between passivated P\( _\mathrm {b} \) centers and stress induced hydrogen-free \( \mathrm {E}’ \) centers. He argued that the presence of a large number of hydrogen passivated silicon dangling bond sites at the Si/SiO\( _\mathrm {2} \) interface and a large number of unpassivated silicon dangling bond sites in the adjacent oxide (\( \mathrm {E}’ \) centers) corresponds to a well-ordered configuration which cannot be favorable in thermal equilibrium since its Gibbs free energy would be very high. Thus, basic statistical thermodynamics arguments based on entropy considerations indicate that the exchange of a large number of hydrogen atoms from interface dangling bonds toward the \( \mathrm {E}’ \) centers is thermodynamically favorable, thereby creating a direct link between hydrogen release from the interface and the creation of oxide defects.

« PreviousUpNext »Contents
Previous: 4 Point defects and their correlation to hydrogen    Top: 4 Point defects and their correlation to hydrogen    Next: 5 Bias dependence of degradation and recovery