Exciton Storage in Semiconductor Self-Assembled Quantum Dots

See allHide authors and affiliations

Science  17 Dec 1999:
Vol. 286, Issue 5448, pp. 2312-2314
DOI: 10.1126/science.286.5448.2312


Storage and retrieval of excitons were demonstrated with semiconductor self-assembled quantum dots (QDs). The optically generated excitons were dissociated and stored as separated electron-hole pairs in coupled QD pairs. A bias voltage restored the excitons, which recombined radiatively to provide a readout optical signal. The localization of the spatially separated electron-hole pair in QDs was responsible for the ultralong storage times, which were on the order of several seconds. The present limits of this optical storage medium are discussed.

Storing information in a cheap way will require ultrahigh packing densities as well as inexpensive self-assembling techniques and fast methods for writing and retrieving the information. Today, over 10 million atoms are required for storage of a single bit of information, but demands for faster addressing will require even smaller structures. Semiconductor QDs, which involve a few thousand atoms, may offer an attractive path toward achieving these goals. Charge storage devices based on the resistivity changes of a two-dimensional electron gas located near a layer of self-assembled QDs have been demonstrated at low temperature (1, 2). Quantum well (QW) devices designed for optical storage have been proposed and demonstrated (3, 4). Their storage characteristics are limited to short storage times (several hundred microseconds) at low temperature. Here, we present an approach for storing and retrieving information in QDs that uses an optical method and allows for exciton storage times of several seconds.

The QDs (5) are produced by molecular beam epitaxy (MBE). The size of the lens-shaped QDs (∼30 to 40 nm in diameter and ∼3 to 5 nm in height) is on the order of the electron wavelength, and the carrier energy levels are quantized. The sequential loading of electrons and holes and the three-dimensional confinement character of the carriers in the QDs have been demonstrated (6). The delta function density of states yields ultranarrow luminescence lines, and several studies have recently shown the importance and complexity of many body effects in the relaxation processes involved in strongly excited QDs (7).

The device structure of QDs presented here is based on the splitting of an optically generated exciton into an electron-hole pair for the write cycle. The electron and hole are stored in closely spaced strain-coupled QD pairs. The effects of this strain coupling are to locally lower the band gap of the material and to localize the electron-hole pair (8). After the electron and hole are stored for a given time, an applied voltage bias can bring them together in the same QD, where they recombine radiatively for the readout cycle. We show exciton storage times at low temperature, which are more than a billion times longer than the exciton lifetime in normal QDs.

The central part of the device structure contains an InAs QD layer and a narrow GaAs QW that are separated by a thin AlAs layer. The 7% lattice mismatch between InAs and GaAs creates a strain field around the coherently strained InAs QDs. This strain field extends through the thin AlAs barrier into the narrow GaAs QW, where it creates a buried strain-induced QD (SIQD) (8). The cross-sectional micrograph from a transmission electron microscope (TEM) (Fig. 1) indicates that the strain field of the InAs QD extends to the GaAs QW, where it gives rise to a strain contrast. The QW thickness is chosen to ensure that its lowest electron energy level is above the X valley minimum in the AlAs barrier. This approach allows one to engineer the ladder of electronic levels shown in Fig. 2. The cascaded energy level structure ensures that the excitons created in the QW are efficiently separated into electron-hole pairs under the influence of an electric field. The relaxation of the electron from the SIQD to the X valley is a rapid process that occurs in a few picoseconds as was shown for QW structures (9). Further relaxation of the electron into the InAs QDs should also take place rapidly by tunneling or thermal excitation or both. In a QW structure, these relaxation processes have been shown to take place in 30 ps (9). This transfer process can be enhanced by inserting the structure into a metal/insulator/semiconductor device with an n+-doped GaAs layer serving as a back contact and a semitransparent Schottky top surface contact. An applied voltage bias will provide the exciton storage function to the device. The read process will be induced by a positive bias that will induce the hole tunneling from the SIQDs to the InAs QDs.

Figure 1

(A) Schematic of the QD memory structure showing pairs of strain-coupled QDs and SIQDs. An electron-hole pair generated by optical pumping is shown. The electron is localized and stored in the InAs QD, and the hole is stored in the SIQD directly above the QD. (B) Cross-sectional TEM micrograph shows the strain contrast around the buried InAs QD and in the GaAs QW. g, Bragg reflection selected for this image.

Figure 2

(A) Low-temperature photoluminescence spectra of the QD memory sample for increasing pump powers (0.32, 1.3, 6.4, and 12 kW/cm2). Dashed lines indicate the peak energy shifts; T, temperature; a.u., arbitrary unit. (B) Schematic band diagram showing the radiative transitions from the QDs (line A) and the spatially indirect recombination involving the SIQDs (line B).

Several sample structures were grown by MBE on semi-insulating (001) GaAs substrates. After deposition of an AlAs/GaAs 40 (20 Å/20 Å) short-period superlattice for smoothing, a 1000 Å–thick n+ GaAs (1018 cm−3) back contact layer was grown, followed by an Al0.5Ga0.5As (400 Å) barrier. Then the 68 Å–wide GaAs QW was deposited, and the InAs QDs layer was inserted in this QW. This was then followed by the deposition of a thin (60 or 100 Å) AlAs barrier, followed by a thin 25 Å GaAs QW, a 500 Å Al0.5Ga0.5As barrier, a thin AlAs/GaAs 7 (20Å/10 Å) short-period superlattice, and finally a 50 Å GaAs capping layer. A semitransparent Schottky contact was evaporated on the surface to provide the top gate. Several undoped samples with layer sequences similar to the layers described here were also grown for optical studies.

The samples were mounted in a closed-cycle helium cryostat and cooled to a temperature below 5 K. An argon ion laser was used for excitation. The photoluminescence (PL) signal was collected through a 0.19-m single monochromator and detected by either a cooled photomultiplier tube or a liquid nitrogen–cooled silicon charge-coupled device camera. For the time-resolved PL experiments, an argon ion laser was modulated at 0.2 Hz with a pulse width of about 50 ms. For shorter time-resolved PL measurements, an acousto-optic modulator and a small aperture were used to create short (<0.5 μs) argon ion laser pulses. A multichannel scaler, whose smallest time bin width was 5 ns, was used to count and accumulate a time-dependent histogram of the single photons.

The photoluminescence of SIQDs for several optical pumping powers is labeled as B in Fig. 2. The luminescence lines shown on the PL spectra correspond to the transitions indicated on the band diagram schematic. The optically induced dipole created by the efficient separation of the electron and hole into adjacent QD-SIQD pairs induces a band bending that is responsible for the observed blue shift of the luminescence (Fig. 2) as the optical pumping power is increased (8). The highest pumping power results in a saturation of the SIQD luminescence, and eventually the thin QW luminescence process dominates.

After the electron-hole pairs are loaded, they can be stored under various voltage bias conditions. For these experiments, we chose a 0-V bias, and the only built-in field is caused by the Fermi level alignment between the n+-doped back GaAs electrode and the semitransparent Schottky contact front electrode. A typical result of a time-resolved PL experiment is shown (Fig. 3) along with the corresponding schematic band diagrams of the device. The detection wavelength was set to the peak of the QD luminescence (peak A at 1.25 eV in Fig. 2) and only a small number (∼10%) of QDs contribute to the signal. During the laser pulse, we observed an intense emission from the QDs. Because a bias readout pulse of 0.5 V was applied 3 s after the write laser pulse, we observed luminescence from the QDs, indicating a successful readout of the stored charge. Without a laser excitation pulse, there was no detectable luminescence from the QDs (Fig. 3B). For this device structure, the optimal readout voltage bias pulse was found to be 0.5 V with a duration of 3 μs.

Figure 3

Time-resolved luminescence from the QDs detected at 1.25 eV. (Top) The schematic band diagram of the device is shown for the write, store, and readout cycles of the device. (A) Writing the QD memory with a 2-μJ laser pulse at 3 K and reading the memory with a short 0.5-V positive bias pulse. (B) No laser pulse but the same readout bias pulse is applied.

The PL line shapes corresponding to the InAs QDs after a storage time of 100 ms for a range of temperatures between 3 and 130 K are shown (Fig. 4). The integrated PL signal in the temperature range from 10 to 130 K falls off with increasing temperature, and from this data, the measured activation energy for the exciton loss with temperature is ∼20 meV. Analysis of the luminescence shift with increasing temperature suggests that the temperature dependence of the band gap, together with thermal evaporation of carriers out of the thinnest QDs in the structure, are responsible for the observed red shift. From the collected signal inFig. 4, we obtained a rough estimate for the device efficiency of ∼10−9 at 10 K after storing the information for 100 ms. This number could be improved by loading the excitons while the device is under a negative bias of ∼−3 V. This will ensure a more efficient transfer of the electrons from the X level in the AlAs into the InAs QDs.

Figure 4

Readout of the QD photoluminescence as a function of temperature for a forward bias of 0.5 V and a storage time of 100 ms.

The time dependence of the stored luminescence at low temperature (Fig. 5) for a readout bias of 0.5 V shows a stored luminescence signal decay, I s, which can be fitted to two exponentials. For short storage times (5 μs to 800 ms), Is ∼ 4 × 10−2/e0.57 t, but for long storage times (800 ms to 3 s), Is ∼ 3 × 10−2/e0.18t, where tis the storage time in seconds. The stored signal decreases by ∼50% in 3 s, and extrapolation to the noise level gives a storage time of nearly 10 s. The origin of these losses is still not clear, but several effects could be responsible for the decreasing readout signal. First, the overlap of the wave functions of electrons and holes could be important, and some of the stored spatially indirect excitons could be lost during the storage time. Second, carrier (hole) losses could occur by capture at deep levels in the AlAs barrier by tunneling during the storage cycle. Third, it is possible that some of the electrons stored in the InAs QDs could be recombining with the residual holes, because the MBE material has a carbon acceptor background concentration of ∼5 × 1015 cm−3.

Figure 5

Low-temperature (3 K) integrated intensity of the readout signal as function of time for a forward bias of 0.5 V. The lines are exponential fits to the data. The storage time extrapolated to the average noise level (on this scale, 6 × 10−3) is ∼10 s.

The exciton storage times are remarkably long, ∼109 times those of the exciton lifetimes in QDs (1 − 5 × 10−9 s). These long exciton storage times should also be compared to exciton lifetimes in n-doped/insulator/p-doped/insulator QW structures. Typical values of ∼1 ms at 4 K have been reported (10), and in this case the structure does not exhibit a memory function. In type I–type II–coupled QW structures that do not contain QDs, exciton lifetimes of a few microseconds have been reported (9). The long storage times in the present QD structures are consistent with electron-hole pair localization in the QD pairs and a low concentration of recombination centers in the QDs.

Much remains to be done to fully understand these long exciton lifetimes and to make use of the high density of QDs as an inexpensive storage medium. Nevertheless, writing and retrieving information using light and self-assembled QDs may offer substantial speed and power advantages for the next generation of devices.

  • * To whom correspondence should be addressed. E-mail: petroff{at}


View Abstract

Stay Connected to Science

Navigate This Article