Two-dimensional electron gas oxide remote doping of Si(001)

We demonstrate the integration of a two-dimensional electron gas (2DEG) oxide structure composed of $\mathrm{LaTi}{\mathrm{O}}_3/\mathrm{SrTi}{\mathrm{O}}_3$ (LTO/STO) on undoped Si(001) that possesses the attractive attributes of high charge density from the oxide and high mobility of silicon. Key to this approach is modification of the oxygen content at the STO-Si interface, which tunes the band alignment and induces electron carriers to move from the oxide 2DEG to form a 2DEG in the silicon substrate. As a consequence, the overall mobility of the heterostructure increases by two orders of magnitude compared to that in the oxide 2DEG to up to $100\mathrm{c}{\mathrm{m}}^2{\mathrm{V}}^{-1}{\mathrm{s}}^{-1}$ at room temperature, with a carrier density of $\sim 1\times {10}^{13}\mathrm{c}{\mathrm{m}}^{-2}$. This approach can be applied to technologies that require both high carrier and high carrier mobility for applications in plasmonics and high power electronics.

Figure 1

Schrödinger-Poisson simulations of oxide 2DEG heterostructures on Si with different STO-Si band alignments. (a) Simulation showing a −0.7 eV CBO between STO and Si. (b) Simulation showing a smaller CBO of −0.15 eV between STO and Si. The structure is 1.95 nm (5 uc) of STO on 0.79 nm (2 uc) of LTO on 4 nm of Si. Band offsets of 0.25 eV are used between the LTO and STO layers. The conduction band potential across the heterostructure is shown as a red curve, and the predicted electron distribution is shown as a blue curve.

Figure 2

XPS band alignment measurements of oxygen annealed STO/Si structures. (a) Spectra of Si $2p$ core levels. The energies of the Si $2p$ peaks are shifted to the as grown (control) sample. Raw data are shown as open symbols and lines are fits. “${\mathrm{O}}_2$” refers to samples annealed in molecular oxygen for 1800 s, “short” refers to a 30 s plasma anneal, and “long” refers to a 600 s plasma anneal. (b) Expanded scale showing the shoulder near the Si $2p$ peak, indicating the presence of interface $\mathrm{Si}{\mathrm{O}}_x$ formation due to the oxygen anneals. (c) Intensity differences from (b) between the annealed samples and the unannealed sample (black dashed line). (d) Ti $3p$ core level spectra. Gray arrows denote positions of peaks and show a shift to lower binding energy for the annealed samples. (e) Summary of calculated conduction band offsets between STO and Si (black squares) along with interface $\mathrm{Si}{\mathrm{O}}_x$ thickness (blue circles) from XPS data. The values are plotted so that the anneals with the highest oxygen activity and time increase from left to right, and lines drawn in between data points serve as guides to the eye.

Figure 3

Two-channel Hall fits to $n_{ox},n_{\mathrm{Si}},{\mu }_{ox}$, and ${\mu }_{\mathrm{Si}}$ for oxygen plasma annealed STO/LTO/STO/Si heterostructures. (a) Sheet carrier density as a function of measurement temperature. (b) Electron mobility as a function of measurement temperature. Filled shapes represent transport in the bulk Si channel, and open shapes represent transport through the oxide layer close to the STO/Si interface. Anneals are performed on the 4.5 uc STO/Si template layer for 10 minutes and range from room temperature (RT) to $600{}^{\circ }\mathrm{C}$. Unannealed control samples are also shown (black squares). Lines are meant as guides to the eye. The unannealed sample shows a linear Hall signal $<200\mathrm{K}$. The annealed samples show nonlinear Hall signals from room temperature down to $<100\mathrm{K}$. In the linear regime, transport is associated with a single channel in the oxide layer. Linear fits have errors smaller than 3%, whereas nonlinear fits have errors ranging from 15 to 30%.

Figure 4

Study of oxide and silicon channel (a) carrier densities and (b) carrier mobilities as a function of oxygen plasma anneal length. The heterostructures consist of 5 uc STO/2 uc LTO/4.5 uc STO deposited on undoped Si. The template 4.5 uc STO/Si is annealed at 300 °C in oxygen plasma for different durations ranging from 0 s (i.e., unannealed) to 600 s. Two-channel Hall fits to $n_{ox},n_{\mathrm{Si}},{\mu }_{ox}$, and ${\mu }_{\mathrm{Si}}$ are plotted, with Si channel data shown as filled shapes and oxide channel data shown as open shapes. Measurement temperatures of room temperature (300 K) and low temperature (100 K) are taken. Thick lines connecting data points are shown as guides to the eye. Error bars are shown in gray for the fits to the 300 K data and in light blue for fits to the low temperature data.