The electrical and optical properties of wafer bonded unipolar silicon-silicon junctions were investigated. The interfaces, both n-n type and p-p type, were prepared using wafers with hydrophilic surfaces. The current versus voltage characteristics, the current transients following stepwise changes in the applied bias, and the capacitance versus voltage characteristics as well as the temperature dependence of the current and capacitance were experimentally obtained and theoretically modeled. The proposed model assumes two distributions of interface states, one of acceptors and one of donors, causing a potential barrier at the bonded interface. It is argued that the origins of the interface states are impurities and crystallographic defects in the interfacial region. The capacitance of the bonded structures includes contributions from the depletion regions as well as from minority carriers. When bonded n-n type samples were illuminated with light of photon energies larger than the silicon band gap the current across the junction increased. This is caused by the photogenerated increase in the minority carrier concentration in the interfacial region, which results in a lowering of the potential barrier. Illumination of n-n type structures with light of photon energies lower than the band gap caused a considerable photocurrent at low temperatures. In this case the observed behavior cannot be explained by interaction with the interface states. Instead, the mechanism is the change in the occupancy of deep electron traps caused by the illumination. These traps are located in the silicon in a small volume around the bonded interface with energies close to the center of the band gap and with a peak concentration of about 1013 cm−3. Impurities present on the silicon surfaces before bonding and impurities gettered to the bonded interface are possible reasons for the increased concentration of deep electron traps in the vicinity of the bonded interface.

No abstract

We have studied the rate at which positive charge is generated starting near the oxide-silicon interface when electrons are injected from the gate through the very thin oxide layer in metal-oxide-(p)silicon tunnel diodes. By varying the oxide thickness, we find that the charging rate is not strongly controlled by the flux of tunneling electrons over a five order of magnitude range in current density. This implies that if the tunneling electrons do participate, then the charge generation in these oxides is at least a two-step process. A comparison of charge generation in aluminum and polycrystalline silicon gate devices suggests that the process does not involve aluminum-related defects. Measurements of the charging rate versus temperature, T, show that it is weakly dependent on T below 150–200 K and apparently thermally activated above this temperature range.

No abstract

Positive and negative charging effects are described for small area (0.008–20 μm2), very thin dielectric (∼2.5 nm), metal-oxide-silicon diodes in which electrons can tunnel directly between the electrodes. These effects are similar to those seen in conventional, thicker oxide devices in which electrons are injected into the oxide conduction band. We show that at least in the thin oxides, charge generation is possible at a total electron energy level which is well below those suggested in a number of models for damage in the thicker oxides.