(image) (image) [Previous] [Next]

Degradation of Electrical Parameters of Power Semiconductor Devices – Process Influences and Modeling

1.3 Semiconductor-insulator interface point defects

Electron spin resonance (ESR), which is also frequently referred as electron paramagnetic resonance (EPR), allows investigating the composition of point defects in semiconductors and insulators by analysis of the interaction of microwave absorption or carrier generation/recombination processes under the influence of a magnetic field [Lep72; WB06; LSS11]. In detail, the interaction occurs with unpaired electrons of unsaturated atomic bonds. For a semiconductor-insulator system these paramagnetic point defects may act as trapped charges which change the device parameters. Consequently, ESR investigations are very valuable for device degradation research.

In the following the most important results of ESR studies of the interface between silicon (Si) and silicon dioxide (SiO2 ) are reviewed. Subsequently, the manifold reactions of hydrogen with point defects near the Si-SiO2 interface are discussed to provide a basis for the discussions in the subsequent Chapters. To divide between the possible reactions, the two most prominent point defects, the dangling bond on a Si atom at the Si-SiO2 interface (Pb center) and the dangling bond on a Si atom bonded to three oxygen (O) atoms within the SiO2 (E′ center), are handled individually, while other reactions are only briefly summarized. A few of the conclusions of this review Section are also documented in [PobegenINBOOK13].

1.3.1 Pb center

The natural lattice mismatch between silicon (Si) and the amorphous layer of thermally grown silicon dioxide (SiO2 ) causes the existence of unsaturated bonds at the interface of these two materials [HP94]. These dangling bonds are visible in ESR and identified as a dangling bond on a Si atom at the Si-SiO2 interface (Pb center) because of their anisotropy and their Landé \( g \)-factor [Cap+79]. These point defects were then connected to the electrically measurable interface traps [LD82]. The orientation of the surface of Si has an impact on the type and properties of the Pb center. On (111)Si-SiO2 surfaces only one type of the Pb center exists. However, for (100)Si a first Pb center variant at the (100)Si-SiO2 interface (Pb0 center) and a second Pb center variant at the (100)Si-SiO2 interface (Pb1 center) exist. See Fig. 1.2 for a sketch of the structure of the two Pb center variants.


Fig. 1.2: Schematic drawing of a Pb0 center and Pb1 center at the (100)Si-SiO2 interface from [Cam+07].

To obtain an upper limit for the number of interfacial dangling bonds the atomic density of Si atoms at the Si surface can be calculated from the lattice constant of Si of 0.5431 nm [Lig61]. For the technologically relevant (100) and (111) surfaces of Si this equates to a density of 6.78 × 1014 cm−2 and 11.76 × 1014 cm−2 [Lig61; KA+02], respectively. Using ESR it was found that the interface to SiO2 exhibits a two decades smaller number of 1013/cm2 dangling bonds at the interface for both (111) and (100)Si [Ste93; SNA98].

Hydrogen interaction

Annealing in a neutral hydrogen atom (H) containing atmosphere at temperatures of 400 °C to 600 °C is widely used to reduce the number of interface traps or Pb centers [Kar+00]. Optimization of the annealing process in forming gas (mixture of nitrogen and hydrogen gas) can bring the remaining electrically active dangling bond density down to 109 cm−2. This results in an interface where approximately every hundredth atom at the Si side of the interface is passivated by an H atom and about every millionth of these could not find an H atom for passivation or the nearest neighbor of the SiO2 and are left unsaturated and thus electrically active. The mean distance between two hydrogen passivated Si dangling bonds is thereby in the nanometer range and two unsaturated bonds are separated by a few hundred nanometers, respectively. This points out that even devices with just a few tens of nanometer width and length will have a few interface traps [PobegenINBOOK13].

First investigations towards the understanding of the BTI of Si based MOSFETs [JS77] suspected the dissociation of hydrogen from H passivated interface traps as the root cause of the instability. However, later performed studies revealed that the values for the dissociation of a H passivated Pb center (Pb H complex) are greater than \( \SI {2.83}{\electronvolt } \). In detail, the passivation and dissociation energy values are normally distributed due to the amorphous nature of the oxide [Ste96b; Ste96a; Ste00; Sta95a; Sta95b] (and not single-valued [BM90; Bro88; Bro90]). See Table 1.1 for a summary.

Table 1.1: Dissociation and passivation energy values for the Pb center family of defects at the Si–SiO2 interface [Sta95a; Sta95b; Ste96b; Ste96a; Ste00]. All values are given in electron-volt.
\( P_{b}^{(111)} \) \( P_{b0}^{(100)} \) \( P_{b1}^{(100)} \)
Passivation mean 1.51 1.51 1.57
variance 0.06 0.14 0.15
Dissociation mean 2.83 2.86 2.91
variance 0.08 0.07 0.07

Also theoretical calculations for H passivated silicon dangling bonds (Si–Hs) in bulk silicon revealed rather large activation energies for H depassivation [PobegenINBOOK13; TW99]. Such rather large dissociation energies cannot normally be reached during operation of a device.

Consequently, in the context of negative BTS (NBTS) in Si based devices, the challenging question arises how the rather strong Si–H bond can be broken at typical BTS temperatures of 100 °C to 150 °C. Various possible explanations will be discussed in Section 3.1.

1.3.2 E′ center

Electrical measurements and ESR studies on the Si-SiO2 system revealed also several types of defects which are situated in the bulk of the oxide [Dea+67; Bun+00]. Among them, the dangling bond on a Si atom bonded to three O atoms within the SiO2 (E′ center), as sketched in Fig. 1.3, and its variants are most important [Wee56; Sil61; FFY74].


Fig. 1.3: Schematic drawing of an E′ center in SiO2 (right) and its precursor state, an oxygen vacancy.

Also for this family of defects possible interactions with hydrogen have been reported.

Hydrogen interaction

Transitions involving H and E′ centers have been studied widely by ESR measurements. It was shown that the interaction of the E′ center with hydrogen may either lead to paramagnetic variants of the E′ center [Vit78; Tak+87; TTS87; TG87; CL92; LJFC98] or to passivation of the dangling bond [AS97; BS99]. Only the former is important for argumentations conducted later in this thesis.

The E′ center may turn to a particular paramagnetic variant, the 74 G doublet, when one of the three oxygen atoms bonded to the silicon atom is exchanged with a hydrogen atom. The 10.4 G doublet is formed when one of the three oxygen atoms bonds to a hydrogen atom. These two variants were shown to form even at room temperature when the sample is exposed to molecular hydrogen gas [CL92]. Therefore these defect types could also be existent in the devices used in this thesis.

1.3.3 Proton

Even though not measurable with ESR, the positively charged hydrogen ion (H+ ) in SiO2 deserves a brief mention because it could be involved in defect creation during NBTS. Experimental studies [AS98; AS99; AS01; AAS01] and theoretical calculations [BS99; Bun+00; Pan+00] indicate that a proton may reside within the SiO2 and may give rise to a positive fixed charge. Especially, calculations suggest that H+ is the only stable state of H at the Si-SiO2 interface [Ras+01]. Experimental work using ESR provided evidence that the proton bonds to an intact Si–O–Si complex through rearrangement of the surrounding amorphous SiO2 lattice [AS98]. Theoretical calculations using density functional theory and supercells suspect the proton to bond to the oxygen vacancy [Pan+00] instead of an intact Si–O–Si complex. Thus, the H atom could become trapped in the close vicinity of the Pb center after Pb H complex dissociation. This idea has already been formulated in context of NBTI [Soo+03].