GaAs Quantum Dots

Nano Lett. , 2005, 5 (10), pp 1873–1877. DOI: 10.1021/nl051026x. Publication Date (Web): September 2, 2005. Copyright © 2005 American Chemical Soci...
2 downloads 0 Views 1MB Size
NANO LETTERS

Photoluminescence Intermittency of InGaAs/GaAs Quantum Dots Confined in a Planar Microcavity

2005 Vol. 5, No. 10 1873-1877

X. Y. Wang,† W. Q. Ma,‡ J. Y. Zhang,‡ G. J. Salamo,‡ Min Xiao,‡ and C. K. Shih*,† Department of Physics, The UniVersity of Texas at Austin, Austin, Texas 78712, and Department of Physics, UniVersity of Arkansas, FayetteVille, Arkansas 72701 Received June 2, 2005; Revised Manuscript Received August 9, 2005

ABSTRACT Photoluminescence intermittency, or “blinking”, was observed in semiconductor InGaAs/GaAs quantum dots (QDs) inside a planar microcavity. Most of the blinking QDs were found around defect sites such as dislocation lines naturally formed in the GaAs barrier layers, and the carrier traps responsible for blinking had an excitation threshold of ∼1.53 eV. The blinking properties of epitaxial QDs and colloidal nanocrystal QDs were also compared by performing laser intensity dependent measurements and statistics of the “on” and “off” time distributions.

Semiconductor quantum dots (QDs) have attracted a lot of interest recently for both their fundamental physics and potential applications ranging from quantum information science to optoelectronic devices. The spontaneous emission of QDs can be greatly affected by their local environment. For example, photoluminescence (PL) intermittency, or “blinking”, has been universally observed in colloidal nanocrystal QDs (NQDs) and is attributed to the ionization and neutralization processes sequentially happening under the influence of trapped carriers.1 Compared with colloidal NQDs, epitaxially grown semiconductor QDs using molecular beam epitaxy (MBE) or metalorganic vapor phase epitaxy (MOVPE) are robust against blinking due to their more ideal interfaces with surrounding materials. So far, from the limited reports in the literature, most of the blinking properties of epitaxial QDs, such as the necessary presence of nearby carrier traps,2-4 have been consistent with the models established for colloidal NQDs. However, blinking studies of epitaxial QDs have also made some new discoveries that are helping to set up a unifying picture for the blinking mechanisms of both colloidal NQDs and epitaxial QDs. In colloidal NQDs, it is still debated whether the carrier traps are located on the NQD surface or in the surrounding matrix. It was recently shown in epitaxial InP/GaInP QDs that in many cases blinking QDs were near artificial scratches on the sample surface,3 thus directly specifying the physical locations of those carrier traps. In this paper, we report optical studies of the blinking behaviors of epitaxial InGaAs/GaAs QDs inside a planar microcavity where the density of obserVable QDs was greatly * Corresponding author: [email protected]. † The University of Texas at Austin. ‡ University of Arkansas. 10.1021/nl051026x CCC: $30.25 Published on Web 09/02/2005

© 2005 American Chemical Society

reduced by the coupling between the QDs and the cavity modes. Most of the blinking QDs were observed to be around defect sites such as dislocation lines naturally formed in the GaAs barrier layers during the sample growth process, while the carrier traps responsible for blinking have energy levels below ∼1.53 eV. Laser intensity dependent measurements and statistics of the “on” and “off” time distributions were also performed to compare the blinking properties of epitaxial QDs and colloidal NQDs. The microcavity sample containing InGaAs/GaAs QDs was fabricated using MBE on a semi-insulating GaAs (100) substrate (see Figure 1a). After the native oxide was desorbed at 580 °C in an As atmosphere, the temperature was increased to 600 °C for the growth of a 5000 Å GaAs buffer layer. This was followed by the bottom distributed Bragg reflector (DBR) consisting of an 18-period AlAs (770 Å)/GaAs (644 Å) multilayer structure. Another GaAs layer of 5383 Å was then grown, and the temperature was reduced to 540 °C in 2 min for the subsequent growth of GaAs(200 Å)/ In0.35Ga0.65As(30 Å)/GaAs(170 Å)/In0.35Ga0.65As(30 Å)/GaAs(170 Å)/In0.35Ga0.65As(30 Å)/GaAs (6228 Å) and the top DBR consisting of an 11-period AlAs(770 Å)/GaAs(644 Å) multilayer structure. During the growth of the three In0.35Ga0.65As (30 Å) QD layers, after every deposition of an In0.35Ga0.65As layer, 3 monolayers (MLs) of GaAs were deposited without interruption to suppress the In segregation. Then after 10 s of interruption, the rest of the GaAs layer was grown. For the whole structure, an As4/Ga beam equivalent pressure ratio of 15 was maintained and the growth rate of both GaAs and AlAs was 1 ML/s. For comparison, a reference sample was also grown using the same procedures as described above except without the top and bottom DBRs.

Figure 1. (a) Schematic of the microcavity structure with embedded InGaAs/GaAs QDs (thinkness of each layer not drawn to scale). (b) PL spectra taken with the collection objective focused approximately on the QD layers (top, red) and moved several micrometers away from the QD layers (middle, red), and PLE spectrum monitored at PL peak A2 (bottom, blue). The PL and PLE spectra were normalized to their respective maximum intensities and offset for clarity. Inset: PL image of the QD layers taken with a long-wavelength-pass filter (>910 nm). All the measurements in (b) were performed at the temperature of ∼8 K and the PL spectra and image were excited with the laser wavelength of ∼780 nm.

The sample was mounted in a He flow cryostat and cooled to ∼8 K unless for the temperature-dependent measurement. A mode-locked Ti-sapphire laser with an 82 MHz repetition rate and a tunable wavelength range of ∼720-950 nm was focused onto the sample at an incident angle of ∼50 ( 5° relative to the surface normal. Optical emissions collected by a long working distance objective were sent either through a 0.5 m spectrometer to a charge coupled device (CCD) camera for PL spectroscopy and image measurements or to a video camera connected to a monitor for the real-time PL imaging. In previous studies, either low-density QD samples4,5 or narrow band-pass filters2,6,7 were used in order to isolate single epitaxial QDs for blinking studies. For our sample with a high areal QD density of ∼5 × 1010/cm2, the microcavity structure preferentially selects only those QDs with optical emissions coupled to the cavity modes, thus effectively reducing the density of obserVable QDs. A PL spectrum excited at ∼780 nm (above the GaAs band gap) and collected by the CCD camera with the objective approximately focused on the QD layers is shown at the top of Figure 1b. Comparing this with the PL from the reference sample, peak A0 was attributed to the GaAs emissions and peaks A1 and A2 to the emissions from two groups of QDs mainly coupled to the vertical cavity modes. PL features with emission wavelengths longer than 930 nm are relatively weak and will be discussed later in the text. The inset of Figure 1b shows a PL image of the QD layers captured by the video camera. A long-wavelength-pass (>910 nm) filter was used 1874

to further reduce the amount of QDs under study, so that only QDs with emission wavelengths at A2 and beyond are present. As discussed later, each bright dot in the PL image may still contain a few single QDs that are not spatially resolved due to the high QD density and the three QD layers grown in this sample. The full width at half-maximum of peak A1 (A2) in the top PL spectrum of Figure 1b is as broad as 6.3 meV (10.1 meV), which is due to the fact that when the objective was focused on the QD layers, we were collecting QD emissions coupled to the vertical cavity modes as well as some transverse modes.8 This was further confirmed in the middle PL spectrum of Figure 1b by the narrowing of both peaks to ∼1.5 meV when the focus of the objective was moved several micrometers away from the QD layers. In this case, PL from peak A0 and the transverse modes became relatively weak as compared to the highly directional emissions9 from peaks A1 and A2 at vertical cavity mode positions. PL imaging has long been used to study the defect dynamics10 and locations3 in semiconductor quantum nanostructures. Although PL images taken from most areas of the QD layers were like the one shown in the inset of Figure 1b, in some places we observed defect-related features as selectively shown in the PL images of Figure 2. The dark spot beside the yellow arrow in Figure 2a originates from a flaw3 or stacking fault that threads the structure.10 The PL image of Figure 2b shows some squarelike lines that bear the signature of misfit dislocations. Namely, after the growth of the first or second layer of InGaAs QDs, misfit dislocations were randomly generated and propagated into the GaAs barrier layer. When the second or third InGaAs layer was grown, QDs tended to nucleate on these dislocation lines. Other different QD lines shown in Figure 2c-f might have the same origins. In fact, buffer layers with dislocation lines have been previously used to grow epitaxial InAs QDs with similar line structures.11 Slightly modified sample growth conditions combined with transmission electron microscopy (TEM) studies should provide a more definitive answer to the formation mechanisms of those features shown in Figure 2, which is beyond the scope of the current paper. About one in a thousand observable bright dots would blink in an area like the one shown in the inset of Figure 1b, consistent with previous blinking studies of epitaxial QDs.2,6,12 On the other hand, in the areas where defect sites such as stacking faults or dislocation lines can be identified, the probability of observing blinking dots was greatly increased. In Figure 2a, a bright dot marked by the yellow arrow blinks near the dark spot of a stacking fault or flaw. In the PL images of Figure 2b-d, we did not observe any blinking dots. Blinking was notably observed, however, in all the bright dots along the long line marked by the yellow arrow in Figure 2e and four of the dots along the short line marked by the yellow arrow in Figure 2f. It has long been believed that stacking faults and dislocation lines can both act as carrier traps.13,14 In light of this, blinking observed here should occur due to a tunneling event of charge from the delocalized states inside epitaxial QDs to these surrounding carrier traps, which is consistent with the existing models Nano Lett., Vol. 5, No. 10, 2005

Figure 3. Time-dependent PL intensity trajectory of a blinking dot taken at ∼8 K with a 150 ms integration time for each data point. The excitation laser was at the wavelength of ∼780 nm with a power density of ∼200 W/cm2. Left inset: PL images of this blinking dot in its “on” (right) and “off” (left) states, together with stable emissions from a nonblinking dot. The lateral distance between the centers of the blinking and nonblinking dots is ∼5 µm. Right inset: Histogram of the PL intensity distributions where the “on”, “off”, and “middle” states can be roughly resolved. Figure 2. Representative PL images taken at six different areas of the QD layers with a long-wavelength-pass filter (>910 nm). (a) One dot marked by the yellow arrow blinks near a flaw or stacking fault. (b) Some dots are formed above squarelike dislocation lines. (c, d) Some dots are formed above long dislocation lines. (e) All the dots above one dislocation line marked by the yellow arrow blink together. (f) Four dots above a short dislocation line marked by the yellow arrow blink together. All the PL images were obtained with the excitation laser wavelength of ∼780 nm at a temperature of ∼8 K.

established for colloidal NQDs.1,15 Although blinking is routinely observed in each colloidal NQD, such a large number of epitaxial QDs blinking together has never been previously observed and confirms similar blinking mechanisms in these two QD systems. The time-dependent PL intensity trajectory of a blinking dot is plotted in Figure 3, where each data point in the time trace was taken with a 150 ms integration time and a laser power density of ∼200 W/cm2. As seen from the right inset of Figure 3, the histogram of PL intensities roughly displays a three-level distribution related to the “on”, “off”, and “middle” states of the blinking process, with the PL intensity of the latter only slightly higher than that of the “off” state. These three-level blinking states, observed only in a small subset of all the blinking QDs we have studied, are commonly encountered in epitaxial QDs with the “middle” states attributed to some “metastable” carrier traps.2,4 In the left inset of Figure 3, we show PL images of this blinking dot in its “on” and “off” states, together with those of a nonblinking dot. Even when the blinking dot is in the “off” state, there are still some residual emissions mainly from the stable PL of several nonblinking QDs within the vicinity of this blinking QD. This is further verified by the PL spectroscopy measurements discussed later. Nano Lett., Vol. 5, No. 10, 2005

Histograms of the “on” and “off” time distributions were obtained from the time-dependent PL intensity trajectory of another representative QD showing only “on” and “off” blinking states. As shown in the inset of Figure 4a, they can be both well fitted using single exponential decay functions. Similar experimental results were also previously reported in other epitaxial QD systems.12,16 In contrast to the universal power law behavior of colloidal NQDs,17 these exponential blinking kinetics strongly imply that a single carrier trap (or only a few) is present around blinking epitaxial QDs.1 This is consistent with the fact that only about one in a thousand epitaxial QDs2,6,12 would blink because of the efficient avoidance of carrier traps by the epitaxial growth process. When the number of carrier traps near a blinking QD is gradually increased, the “on” and “off” time distributions may evolve from a single-exponential fit of epitaxial QDs to the inverse power law one of the colloidal NQDs. For colloidal NQDs, the average “off” time is not intensity dependent contrasting with the “on” time which shows a “truncation” point at the long time tail of its power-law distributions at higher laser excitation intensities.17 In Figure 4b, we show the time-dependent PL intensity of a blinking epitaxial QD measured at four different laser intensities with an integration time of 100 ms for each data point. It can be clearly seen that both the average “on” and “off” times decrease with the increasing laser excitation intensity. Specifically, the average “on” (“off”) time was 1.925 s (22.3 s) at P0, 1.156 s (8.425 s) at 1.1P0, 0.583 s (2.140 s) at 1.8P0, and 0.279 s (0.280 s) at 4.3P0, respectively. Here, P0 was ∼50 W/cm2 and the average “on” (“off”) time was calculated for exponential statistics.1 This kind of light-induced switching processes between “on” and “off” states agrees with all the previous reports and suggests the involvement of multiple particles in the blinking process of an epitaxial QD.2,7,16 1875

Figure 4. (a) (top) PL spectra of the “middle” (red) and “off” (blue) states of a blinking dot excited with the laser at ∼809 nm. (bottom) PL spectra of the same blinking dot in the “on” (red) and “off” (blue) states excited with the laser at ∼803 nm. (inset) Histograms of the “on” and “off” time distributions with their single exponential decay fittings shown by the red lines. They were obtained from the time-dependent PL intensity trajectory of a blinking QD showing only “on” and “off” states, and plotted in a log-log scale to show that they cannot be described by a simple inverse power law. (b) Time-dependent PL intensity trajectories of a blinking QD excited by the laser wavelength of ∼780 nm with four different power densities. Here, P0 was ∼50 W/cm2 and the intensity trajectories were normalized to their respective maximum intensities and offset for clarity. All the measurements in (a) and (b) were performed at the temperature of ∼8 K.

We further explore how blinking behaviors of epitaxial QDs depend on the laser excitation energy. The PLE spectrum of PL peak A2 is represented in the bottom of Figure 1b with the continuous absorption band from GaAs and three other peaks, B0, B1, and B2, clearly resolved. The PLE spectrum of A1 is quite similar to that of A2 except that it was measured only to the wavelength of ∼860 nm. Corresponding to the vertical cavity mode at A1 (A2), the PLE peak B1 (B2) results from a high laser transmission of the blue-shifted resonant cavity mode at the incident angle of ∼50 ( 5° used in our experiment.18 The PLE peak B0 might be related to another laser transmission peak at the 1876

edge of the cavity stop band. Detailed discussions of the PLE spectrum will be presented elsewhere. Thus, by tuning the excitation laser energies from above the GaAs band gap to the B1 and B2 positions, the energy levels of carrier traps responsible for blinking can be effectively probed.3 PL spectra of a blinking dot excited with the laser power density of ∼200 W/cm2 at two representative laser wavelengths are shown in Figure 4a. When the laser was continuously tuned between 720 nm (1.72 eV) and 809 nm (1.53 eV), no significant changes were observed in the blinking behaviors; in the bottom of Figure 4a we only show the PL spectra of the “on” and “off” states excited with the laser at 803 nm (1.54 eV). Each blinking dot still consists of a few smaller and larger QDs with optical emissions coupled to cavity modes A1 and A2, respectively. With respect to the discussions below, the PL peak A1 did not show any difference from A2 and for simplicity is not shown in Figure 4a. PL features between 930 and 950 nm, observed in both blinking and nonblinking dots, could only be observed when the objective was focused approximately on the QD layers and should result from QD emissions coupled to the transverse modes near the edge of the cavity stop band. In all the blinking dots studied, the “on” and “off” behaviors were exclusively caused by a blinking QD coupled to these transverse modes having similar PLE features as those of PL peak A2. The vertical dotted line in Figure 4a shows that no spectral diffusion was associated with the blinking epitaxial QDs. The PL peaks, however, can still be as broad as several meV, which is likely due to the influence of trapped carriers.4,6,19 As seen from the top PL spectra of Figure 4a, when the laser was tuned to 809 nm (1.53 eV), this QD still blinked but with only a small difference in the maximum and minimum PL intensities, reminiscent of a blinking process involving only the “middle” and “off” states in Figure 3. The QD became “nonblinking”20 and stayed permanently in the “off” state when the laser was moved only a little further above 809 nm and the related PL spectrum is just the same as the “off” state spectrum excited with the laser at 809 nm. When the laser was further tuned to the wavelengths above 809 nm, including the B1 and B2 positions in the PLE spectrum, this QD was still in the “off” state and never blinked again. The same measurements on other blinking QDs yielded similar blinking-stopping wavelengths around 809 nm, corresponding to an excitation threshold of 1.53 eV.21 This is consistent with the mechanism wherein defectrelated trap states can trap carriers responsible for blinking.3 With laser excitation above 1.53 eV, carriers can relax to these traps and later jump back to the barrier layers, possibly by absorbing the extra energies released by other carriers in their relaxation processes, leading to the PL blinking between “off” and “on” states. In contrast, when the laser energy is tuned below 1.53 eV, it is difficult for the carriers directly excited in the trap levels to jump back to the barrier layers and the blinking QDs will permanently reside in the “off” state. Similarly, in previous studies of epitaxial InP/GaInP QDs,3 when the carriers were generated in the barrier layers, the blinking QDs stayed in the “on” state for most of the Nano Lett., Vol. 5, No. 10, 2005

observation time, but they switched frequently to the “off” state when there were carriers generated directly in the trap levels. Finally, we also measured the time-dependent PL intensities of several blinking QDs at different temperatures, with the average “on” and “off” times nearly unchanged from ∼8 to ∼40 K. When the temperature was further increased above 50 K, PL emissions from most of the blinking QDs coupled to the transverse modes were nearly quenched and it is hard to tell if the blinking process of epitaxial QDs studied here can be activated by increasing temperature or not. In conclusion, using PL imaging and excitation measurements, we have studied the blinking behaviors of the most commonly encountered epitaxial InGaAs/GaAs QDs confined in a planar microcavity. Most of the blinking QDs were found around defect sites such as dislocation lines in the GaAs barrier layers, and the carrier traps responsible for blinking had an excitation threshold of ∼1.53 eV. Single exponential decay fittings for the blinking “on” and “off” time distributions imply that only a limited number of carrier traps are present near a blinking epitaxial QD. Moreover, the intensity-dependent measurement shows that switching processes between the blinking “on” and “off” states of epitaxial QDs are light induced, possibly due to the involvement of multiple particles. Acknowledgment. The authors gratefully acknowledge support of this work by NSF (DMR-0210383 and DMR0306239), US NavysOffice of Naval Research (N0014-041-033), Texas Advanced Technology program, and the W.M. Keck Foundation. References (1) Kuno, M.; Fromm, D. P.; Hamann, H. F.; Gallagher, A.; Nesbitt, D. J. J. Chem. Phys. 2000, 112, 3117. (2) Panev, N.; Pistol, M.-E.; Zwiller, V.; Samuelson, L.; Jiang, W.; Xu, B.; Wang, Z. Phys. ReV. B 2001, 64, 045317.

Nano Lett., Vol. 5, No. 10, 2005

(3) Sugisaki, M.; Ren, H.-W.; Nishi, K.; Masumoto, Y. Phys. ReV. Lett. 2001, 86, 4883. (4) Pistol, M.-E. Phys. ReV. B 2001, 63, 113306. (5) Bertram, D.; Hanna, M. C.; Nozik, A. J. Appl. Phys. Lett. 1999, 74, 2666. (6) Panev, N.; Pistol, M.-E.; Jeppesen, S.; Evtikhiev, V. P.; Katznelson, A. A.; Kotelnikov, E. Yu. J. Appl. Phys. 2002, 92, 7086. (7) Pistol, M.-E.; Castrillo, P.; Hessman, D.; Prieto, J. A.; Samuelson, L. Phys. ReV. B 1999, 59, 10725. (8) Zwiller, V.; Chitica, N.; Persson, J.; Pistol, M.-E.; Seifert, W.; Samuelson, L.; Hammar, M.; Streubel, K.; Goobar, E.; Bjo¨rk, G. Mater. Sci. Eng. B 2000, 69-70, 314. (9) Tokito, S.; Tsutsui, T.; Taga, Y. J. Appl. Phys. 1999, 86, 2407. (10) Haugen, G. M.; Guha, S.; Cheng, H.; DePuydt, J. M.; Haase, M. A.; Ho¨fler, G. E.; Qiu, J.; Wu, B. J. Appl. Phys. Lett. 1995, 66, 358. (11) Yamaguchi, K.; Kawaguchi, K.; Kanto, T. Jpn. J. Appl. Phys. 2002, 41, L996. (12) Castrillo, P.; Hessman, D.; Pistol, M.-E.; Prieto, J. A.; Pryor, C.; Samuelson, L. Jpn. J. Appl. Phys. 1997, 36, 4188. (13) Rice, J. H.; Robinson, J. W.; Jarjour, A.; Taylor, R. A.; Oliver, R. A.; Briggs, G. A. D.; Kappers, M. J.; Humphreys, C. J. Appl. Phys. Lett. 2004, 84, 4110. (14) Seufert, J.; Weigand, R.; Bacher, G.; Ku¨mmell, T.; Forchel, A.; Leonardi, K.; Hommel, D. Appl. Phys. Lett. 2000, 76, 1872. (15) Shimizu, K. T.; Woo, W. K.; Fisher, B. R.; Eisler, H. J.; Bawendi, M. G. Phys. ReV. Lett. 2002, 89, 117401. (16) Sugisaki, M.; Ren, H.-W.; Nishi, K.; Masumoto, Y. Jpn. J. Appl. Phys. 2002, 41, 958. (17) Shimizu, K. T.; Neuhauser, R. G.; Leatherdale, C. A.; Empedocles, S. A.; Woo, W. K.; Bawendi, M. G. Phys. ReV. B 2001, 63, 205316. (18) Fainstein, A.; Jusserand, B.; Thierry-Mieg, V. Phys. ReV. Lett. 1995, 75, 3764. (19) Blome, P. G.; Wenderoth, M.; Hu¨bner, M.; Ulbrich, R. G. Phys. ReV. B 2000, 61, 8382. (20) By “nonblinking”, we mean that PL intensity of a QD is nearly unchanged even for the maximum observation time of ∼30 min and maximum laser power density of ∼500 W/cm2 used in our experiment. (21) When the laser energy was changed a little from below to above ∼1.53 eV turning a nonblinking QD into a blinking one, no red shifts and intensity changes were observed in the “off” state PL spectrum of the blinking QD and in the PL peaks of all the other nonblinking QDs within the vicinity of this blinking QD. So we can conclude that the excitation threshold of ∼1.53 eV is not caused by the temperature and excitation intensity changes which are known to influence the frequencies of a blinking process.

NL051026X

1877