Control of electric field domain formation in multiquantum well structures

The formation of electric field domains (EFD) was first observed in bulk GaAs and is mostly known as the cause of Gunn oscillations.’ It is explained in terms of the negative differential resistance (NDR) which occurs because of the electron transfer from the r to the X or L valleys. Esaki and Chang2 first observed the formation of static EFDs in multiquantum wells (MQW); this phenomenon was attributed to the NDR which arises due to sequential resonant tunneling (SRT) between subbands in adjacent wells.3a Recently, we demonstrated the operation of a tunable quantum well infrared detector which was based on the formation of EFDs in a MQW device.’ In this letter, we report on an investigation designed to determine the parameters which govern EFD formation and expansion; We show theoretically and experimentally how the proper choice of well widths, heights, and doping determines the electric field domain profile. First, we discuss EFDs in the three-stack MQW device presented in Ref. 7. In this device the superlattice clad by two n-doped contact layers, consisted of three stacks of 25 QW each; the first 25 wells were 3.9 nm wide ,and were separated by AlxGa,...&s (x=0.38) barriers; the second stack consisted of 4.4 nm wide wells with (x=0.3) barriers; the last stack had 5.0 nm wide wells and (x=0.24) barriers. All the barriers were 44 nm wide; the wells and the contacts were uniformly doped with Si to n=4~ 10’s cmm3. The absorption spectrum at room temperature shows three peaks at 1364, 1080, and 920 cm-’ obeying intersubband selection rules for the polarization of the incident light.’ Figure 1 displays the smoothed photocurrent spectral response of a mesa structure, 200 pm in diameter at 7 K, for different values of the applied voltage. The polarity is defined in Fig. 2. We see that at different ranges of applied bias, only some of the peaks in the photocurrent are present. This was explained by the formation of high and low electric field domains in the device. The light is absorbed in all three stacks of QWs but only photoexcited carriers which are in a region with high electric field can be swept out of the QW and contribute to the current. Those in the low field region have a high probability of being recaptured by their own well, contributing only negligibly to the current. A second indication of the presence of EFDs in the device comes from dark current measurements. A fine structure in the plateaus of the I-V curve, corresponds to regions of NDR.’ This is due to SRT, which occurs whenever the ground level of a well is aligned with the excited level of the adjacent well4 Under an arbitrary applied bias, a uniform distribution of electric field is not stable because all of the QWs will be out of resonance, i.e., none of the energy levels of pairs of adjacent wells will be aligned. Instead, the system will settle into a different configuration in which the electric field profile includes high and low field regions. In the high field region we have ground level to excited level SRT, and in the low electric field region ground level to ground level SRT. Transport within each domain is resonant, while at the boundary between the two regions it is generally nonresonant. This boundary then acts as a bottleneck that limits the current. There should