The negative differential resistance of a resonant-tunneling diode (sometimes called double-barrier diode) was predicted by Tsu and Esaki in 1973, following their pioneering work on superlattices in the late 1960s and early 1970s. The structure and characteristics of this diode were first demonstrated by Chang et al. in 1974. Following the much improved results reported by Sollner et al. in 1983, research interest was escalated, partially due to maturing МВБ and MOCVD techniques. In 1985, room temperature negative differential resistance in this structure was reported by Shewchuk et al., and by Tsuchiya et al. Meanwhile, resonant tunneling of holes instead of electrons was observed by Mendez at al.
A resonant-tunneling diode requires band-edge discontinuity at the conduction band or valence band to form a quantum well and, thus, necessitates heteroepitaxy. The most popular material combination used is GaAs-AlGaAs, followed by GalnAs-AlInAs. The middle quantum-well thickness is typically around 50 A, and the barrier layers range from 15 to 50 A. Symmetry of the barrier layers is not required so their thicknesses can be different. The well layer and the barrier layers are all undoped, and they are sandwiched between heavily doped, narrow energy-gap materials, which usually are the same as the well layer. Thin layers of undoped spacers (« 15 A GaAs) are adjacent to the barrier layers to ensure that dopants do not diffuse to the barrier layers. Because thin epitaxial layers and abrupt doping profiles are required, most reported studies used MBE for film deposition, but MOCVD has also been used occasionally. Device isolation is usually achieved by mesa etching. A resonant-tunneling diode utilizes the quantization of energy states in a quantum well. Quantum mechanics prescribes that in a quantum well of width, the conduction band (or valence band) is split into discrete subbands. The ratio of local peak current to valley current is a critical measure of the negative differential resistance. The peak current is mainly due to tunneling which can be maximized by using material of lighter effective mass. In this respect, the material combination of GalnAs-AlInAs is advantageous over GaAs-AlGaAs. Maximum peak-current density of 3xl05 A/cm2 has been observed, and is quite temperature independent since it is a tunneling current. The nonzero valley current is mainly due to thermionic emission over the barriers, and it has large temperature dependence (smaller with lower temperature). Another small but conceivable contribution is due to tunneling of electrons to higher quantized levels. Even though the number of electrons available for tunneling at energy is very small, there is a thermal distribution tail and this number is not zero, especially when the quantized levels are close together. The maximum ratio observed is about 50 at room temperature.
As discussed, each region of negative differential resistance is associated with tunneling through one particular quantized subband. Additional bias is also developed across the spacer layers, as well as the accumulation layer and depletion layer of the heavily doped materials next to the undoped spacers.
The characteristics have two regions of negative differential resistance for each voltage polarity. In practice, the second current peak is rarely observed, due to the small signal in a large background of thermionic-emission current. The illustration nevertheless brings out the potential advantage over a tunnel diode that is limited to only one region of negative differential resistance. This feature of multiple current peaks is especially important as a functional device. For structures with identical barriers the I-V characteristics are symmetrical around the origin. However, the two barriers can be made different in both barrier height (material) and layer thickness, resulting in asymmetrical I-V characteristics.
Triple-barrier heterostructures with two successive quantum wells have also been studied, and multiple regions of negative differential resistance can be readily observed. The first current peak is believed due to tunneling through the first quantized levels of both wells, while the second current peak can be attributed to tunneling though different quantized levels or sequential tunneling with downward transition. In any case, due to an additional barrier, which acts as a filter, sharper current features are possible. Resonant tunneling on structures with quadruple barriers (triple wells) have also been studied.
The extreme case of multiple barriers is a compositional superlattice which consists of many alternating layers of barriers and quantum wells. Negative resistance from a compositional superlattice had been observed about the same time as from a resonant-tunneling diode. One major difference in the superlattice structure is that the quantized levels are broadened into narrow subbands. The first current peak can be observed with a bias comparable to the first subband width. Up to that point, the field is uniform across the superlattice. Additional bias causes a high-field domain to develop in one of the barriers, causing a misalignment of the subbands. Current rises again when the first subband is aligned to the second subband. The number of barriers with high-field domain is increased to two.
An alternate approach to achieve multiple current peaks is to connect resonant-tunneling diodes in series. The structure can be realized with vertically integrated double barriers, separated by heavily doped layers. This is in principle very different from the above structures of multiple quantum wells since here the resonant-tunneling diodes are only connected by relatively thick heavily doped layers, and there is no quantum mechanical communication between them. The resultant I-V characteristics show multiple current peaks. Another useful characteristic is that the current peaks are at approximately the same level. This is advantageous for multi-value logic applications. When a voltage is applied across resonant-tunneling diodes, each one absorbs approximately. In practice, a minute difference in the structures would favor one to switch into the negative resistance (off-resonance) first. Since current has to be continuous through all devices, the overall current drops initially and follows the general shape of the individual diode. The current then rises with voltage again until another resonant-tunneling diode switches. The number of current peaks thus corresponds to the total number of resonant-tunneling diodes in series.
