Stimulated by the growing manifestation in advanced semiconductor device development, an extensive multifrequency electron spin resonance (ESR) study has been carried out on the thermal (110)Si/SiO${}_{2}$ interface in terms of occurring paramagnetic point defects as a function of oxidation temperature ${T}_{\mathit{ox}}$ (200--1125 \ifmmode^\circ\else\textdegree\fi{}C), with seclusion of the H-passivation factor. The main type of defect observed is a P${}_{\mathrm{b}}$-type interface center closely related to the P${}_{\mathrm{b}}$${}^{(111)}$ and P${}_{\mathrm{b}0}$${}^{(100)}$ variants (Si${}_{3}$ \ensuremath{\equiv} Si${}^{\ifmmode\bullet\else\textbullet\fi{}}$) characteristic for the (111) and (100)Si faces, respectively. The inferred principal g matrix values (${g}_{//}$ = 2.0018 and ${g}_{\ensuremath{\perp}}$ = 2.0082 for ${T}_{\mathit{ox}}$ = 800 \ifmmode^\circ\else\textdegree\fi{}C), splitting parameters of the resolved ${}^{29}$Si hyperfine doublet, and line width behavior closely resemble those of P${}_{\mathrm{b}0}$${}^{(100)}$, from which the defect is typified as P${}_{\mathrm{b}0}$${}^{(110)}$. For low ${T}_{\mathit{ox}}$, an unexpectedly high density of P${}_{\mathrm{b}0}$${}^{(110)}$ defects (\ensuremath{\sim}7 \ifmmode\times\else\texttimes\fi{} 10${}^{12}$ cm${}^{\ensuremath{-}2}$) is observed, which gradually dwindles for ${T}_{\mathit{ox}}$ increasing above \ensuremath{\sim}700 \ifmmode^\circ\else\textdegree\fi{}C to approach \ensuremath{\sim}4 \ifmmode\times\else\texttimes\fi{} 10${}^{12}$ cm${}^{\ensuremath{-}2}$ for ${T}_{\mathit{ox}}$ \ensuremath{\rightarrow} 1125 \ifmmode^\circ\else\textdegree\fi{}C. The behavior is related to interfacial stress release as a result of global structural relaxation of the top SiO${}_{2}$ layer, an effect also signaled by attendant alterations in ESR parameters, including a drop in ESR line width and a change in line shape symmetry and ${g}_{\ensuremath{\perp}}$. Comparison with previous ESR data on (111)Si/SiO${}_{2}$ and (100)Si/SiO${}_{2}$ interfaces indicates that, in terms of P${}_{\mathrm{b}}$ type, the (110) face is the worst of all three low-index Si interfaces, i.e., [P${}_{\mathrm{b}0}$${}^{(100)}$] [P${}_{\mathrm{b}}$${}^{(111)}$] [P${}_{\mathrm{b}0}$${}^{(110)}$], in contrast with the common electrically inferred interface trap density order; only for ${T}_{\mathit{ox}}$ \ensuremath{\geqslant} 900 \ifmmode^\circ\else\textdegree\fi{}C does the (110) face slightly improve on the (111)Si one, raising caution with the application of (110)Si/SiO${}_{2}$ in terms of vulnerability during device operation. The comparison further shows that, unlike a textbook quote, the density of occurring P${}_{\mathrm{b}(0)}$ centers is not found to be proportional to Si surface areal atom density or available Si bond density. Instead, an empirically inferred matching criterion appears to be the surface areal Si atom density scaled by the number of bonds per atom directed into the oxide. Besides P${}_{\mathrm{b}0}$${}^{(110)}$, an apparently isotropic second type of interface center is revealed, baptized ${I}_{\mathrm{x}}$, at $g$ = 2.0048 with a density of \ensuremath{\sim}1 \ifmmode\times\else\texttimes\fi{} 10${}^{12}$ cm${}^{\ensuremath{-}2}$ that is rather independent of ${T}_{\mathit{ox}}$. Showing a similar passivation behavior in H${}_{2}$ as the P${}_{\mathrm{b}0}$${}^{(110)}$ center, it is also interpreted as a Si dangling-bond-type defect, now residing in an interfacial randomized Si environment---a variant of the D center.
Read full abstract