Imaging and Manipulating Energy Transfer Among Quantum Dots at Individual Dot Resolution Duc Nguyen,†,‡,∥ Huy A. Nguyen,†,‡ Joseph W. Lyding,‡,§ and Martin Gruebele*,†,‡,⊥ †
Department of Chemistry, ‡Beckman Institute for Advanced Science and Technology, §Department of Electrical and Computer Engineering, University of Illinois at Urbana−Champaign, Urbana, Illinois 61801, United States ⊥ Department of Physics, University of Illinois at Urbana−Champaign, Urbana, Illinois 61801, United States S Supporting Information *
ABSTRACT: Many processes of interest in quantum dots involve charge or energy transfer from one dot to another. Energy transfer in films of quantum dots as well as between linked quantum dots has been demonstrated by luminescence shift, and the ultrafast time-dependence of energy transfer processes has been resolved. Bandgap variation among dots (energy disorder) and dot separation are known to play an important role in how energy diffuses. Thus, it would be very useful if energy transfer could be visualized directly on a dot-by-dot basis among small clusters or within films of quantum dots. To that effect, we report single molecule optical absorption detected by scanning tunneling microscopy (SMASTM) to image energy pooling from donor into acceptor dots on a dot-by-dot basis. We show that we can manipulate groups of quantum dots by pruning away the dominant acceptor dot, and switching the energy transfer path to a different acceptor dot. Our experimental data agrees well with a simple Monte Carlo lattice model of energy transfer, similar to models in the literature, in which excitation energy is transferred preferentially from dots with a larger bandgap to dots with a smaller bandgap. KEYWORDS: nanoparticle, Förster resonant energy transfer, laser, Monte Carlo simulation, scanning tunneling microscopy
Q
determined for such films.11,17−19,22−24 In linked pairs of dots, energy transfer monitored by time-resolved photoluminescence from the band edge show the expected slow-down of energy transfer with linker length.12 Akselrod et al. have demonstrated that energy transport becomes subdiffusive due to the distribution of bandgaps (energetic disorder) within a film.17 Auger recombination is also strongly enhanced by energetic disorder.21 A great effort has been devoted to improving the spatial resolution for investigating charge and energy transfer. Spatial resolution of tens of nanometers has been obtained,15,22 and work by Yoon et al. reveals transport on a sub-μm scale on a nanosecond time scale.22 However, imaging and manipulating energy transfer at single quantum dot resolution has not yet been demonstrated. As more complex patterns of quantum dots with different sizes and bandgaps can be fabricated, and considering the importance of energetic disorder, it will be very useful to monitor and manipulate energy transfer with dot-by-dot resolution. We do so with single molecule absorption
uantum dots are promising for applications in highly efficient solar cells because of their broadband absorption, high fluorescence quantum yield, and multiexciton generation.1−7 A power conversion efficiency of 10.6% has been demonstrated for quantum dot solar cells.8 In this and many other applications, a closely packed thin film of quantum dots is generally used. In addition to investigating close-packed films, progress has been made in making other tailored patterns of quantum dots.9,10 Stacks of monolayers of quantum dots with increasing sizes have been fabricated to direct energy transport among quantum dots vertically.11 Quantum dots are also assembled to oligomers of different sizes to improve their energy transfer efficiency.12,13 A heterogeneous distribution of quantum dot size, shape, spacing, and defects can greatly affect the charge and energy transport and eventually the performance of such devices.14−16 Starting with the original work of Kagan et al., bulk film measurements have yielded detailed insight into energy transport via exciton hopping between donor and acceptor dots.11,16−22 In films of quantum dots, a red shift of band-edge luminescence shows that energy is transferred from smaller quantum dots to larger dots with a smaller bandgap.11,18,19 The time and length scales of the energy transfer process have been © 2017 American Chemical Society
Received: April 17, 2017 Accepted: May 19, 2017 Published: May 19, 2017 6328
DOI: 10.1021/acsnano.7b02649 ACS Nano 2017, 11, 6328−6335
Article
www.acsnano.org
Article
ACS Nano
Figure 1. Schematic of the SMA-STM energy transfer experiment. (a) Two PbS quantum dots on Au. If the smaller dot with larger bandgap is excited at 532 nm (green arrow), energy transfer (ET) to the larger dot can take place when the dots are separated by less than ≈10 nm.23,24 Electron transfer is unlikely due to the ≥2 nm separation of dots (oleic acid ligands = blue wavy lines, Figure S2). ET can be direct (red wavy arrow) or plasmonically enhanced by the tip apex, surface, or the gap mode between tip and surface. The change in electron occupancy modulates the tunneling current I (blue arrow) through the larger quantum dot, which is detected as the SMA-STM signal. The tip electric field ε (black arrow) can tune different states into resonance with the laser via the Stark effect, as discussed in ref 26. (b) Schematic diagram for energy transfer from quantum dot QD1 to QD2. The green laser (hv, green arrow) excites an electron in QD1 to the conduction band. The excited QD1 quickly relaxes to the 1S state near the band edge (black wavy arrow), from where downhill energy transfer to QD2 takes place (red wavy arrow). An exciton (electron hole pair, shown by a small arrow in a conduction band orbital and an empty circle in a valence band orbital) is thus transferred, with a small amount of energy loss and relaxation at each step. Biexcitons are discussed in the main text. (c) STM (orange) and SMA-STM (gray scale) images of two PbS quantum dots on Au, close enough (∼2 nm separation) to enable energy transfer. The SMA-STM signal of the excited upper dot, relaxed to the band edge, appears as a uniform dark disc. Scanning conditions: 10 pA, 1.7 V, gray color scale −0.2 to 0.2 pA, laser power density 1388 mW/mm2, scale bars 5 nm. (d) STM and STM-SMA images of two PbS quantum dots for which energy transfer has been disabled in two ways: they are well-separated, and they sit on a large bandgap a-SiC surface to reduce any plasmon-enhanced energy transfer. A highly structured SMA-STM signal (P orbital-like) is observed, as in ref 26. Scanning conditions 10 pA, −1.8 V, W tip, gray color scale −0.25 to 0.25 pA, laser power density 1630 mW/mm2, scale bars 3 nm. (e) Raw and (f) filtered TEM images of PbS quantum dots arrays showing edge-to-edge separation of ∼3.2 nm by TEM, which does not image oleic acid ligands that space out the quantum dots (Materials and Methods Section). Scale bar in (e) is 20 nm and in (f) is 10 nm. (g) STM image of a close-packed PbS quantum dot array on an Au surface. The quantum dots appear to be wider and thus more closely packed because of the tip convolution reducing lateral resolution. The edge-to-edge separation estimated by STM and spectral measurements is still ≥2 nm (Figure S1 and S2), as a realistic lower limit on the quantum dot separation. Scanning conditions: 1.5 V, 10 pA, scale bar 10 nm.
RESULTS AND DISCUSSION We employed SMA-STM to image room temperature optical absorption and energy transfer of PbS quantum dots deposited individually, as oligomers, or in films. Dots were excited by a 532 nm diode laser. Figure 1 shows a schematic of the experiment, an energy level diagram for dot-to-dot energy transfer, and an example of SMA-STM images when dots can transfer energy (panel c) or cannot transfer energy (panel d). In SMA-STM, laser light illuminates the sample from the rear to minimize thermal perturbation of the tip-dot-sample junction (Figure 1a, Materials and Methods Section).25,26,31−33 Due to tip enhancement, the local light intensity is high enough to drive even rapidly relaxing ( 0.1), the average hop and distance remain nearly constant because limited-range percolation through a 2-D hexagonal lattice cannot be quenched completely. This result is consistent with subdiffusion previously observed in arrays of CdSe/ZnCdS core−shell films.17
Figure 5. (Top) Average number of hops that an excited quantum dot takes before relaxing to the ground state, at different degrees of heterogeneity. (Bottom) Average distance from the initial excited dot to the final dot where the energy is relaxed to the ground state, at different degrees of heterogeneity. The average distance is given in the effective quantum dot diameter unit (6.2 nm center-to-center separation of PbS quantum dots in the experiments).
CONCLUSIONS Our experimental excited-state image data and MC simulations show that energy funneling in large clusters or films of quantum dots on a surface can be observed, manipulated, and modeled quantitatively dot-by-dot. In addition, there is consistent, albeit less definitive, evidence for energy transfer in small clusters of quantum dots on different surfaces. Thus, SMA-STM can be used to directly visualize and manipulate with nanometer resolution energy pooling pathways involving as few as two quantum dots (Figure 1c) up to hexagonally packed films (Figure 3), guided by both direct and plasmon-enhanced excitation and energy transfer mechanisms. It will be interesting to see if direct charge transfer from dot to dot can also be observed in the future. One approach would be to push “bare” quantum dots to different distances on a nonmetallic substrate, to show that only dots in contact show a shift in SMA-STM signal. The distance-dependence may turn out to have unexpected features due to finite-size effects and strong coupling.43 MATERIALS AND METHODS Substrate Preparation. Two different substrates were employed in this study: ultrathin PtAu film on sapphire, and silicon carbide (either crystalline, c-SiC, or amorphous, a-SiC). Ultrathin PtAu films on c-plane sapphire substrate (preannealed at ∼1273 K for 12 h) were fabricated by electron beam deposition.44 5 nm of Pt layer was deposited at the substrate temperature of ∼1050−1070 K, followed by deposition of 10 nm of Au at ∼670 K. In order to obtain total internal reflection, a 3 mm fused silica right angle prism (Thorlabs) was glued to the back of the samples using a transparent, UHV compatible epoxy (302−3M, Epotek) and cured overnight. Two short edges of the samples were painted with colloidal silver (Ted Pella Inc.) to improve 6332
DOI: 10.1021/acsnano.7b02649 ACS Nano 2017, 11, 6328−6335
Article
ACS Nano
Quantum dots were manipulated on the surfaces following a procedure described in ref 49. The tip first scans a line (∼15- 20 nm) over the quantum dot, then scans the same line, but is moved down by a preset value (typically 0.7 nm). Rabi Cycle Calculations and Saturation Intensity. Rabi frequencies were estimated for on-resonance excitation as ωR = μijE/ ℏ where E is the electric field amplitude of the linearly polarized light field and μij is the transition dipole moment between states i and j. The electric field was calculated from irradiance (I ≈ 1630 mW/mm2) as E = 2I /cε0 ≈ 3.5 × 104 V/m, where c is the speed of light and ε0 is the permittivity of free space. A field enhancement factor of up to 30 was assumed for PtIr tips. For PbS quantum dots, the transition dipole moment was calculated conservatively from the resonance frequency (ωij ≈ 3.54 × 1015 rad/s at 532 nm) and the radiative lifetime of the biexciton (τ = 50 ps)39 as μij = (3ε0hc3/2τω3ij)1/2. This gave a Rabi cycle for PbS quantum dots of ≈2 ps. For CdSe/ZnS quantum dots, the radiative lifetime of the single exciton τ = 24 ns11 was used because the excitation is near the bandgap which gave a Rabi cycle of ≈42 ps. Both time scales are much faster than our observation time (average of 3 ms/pixel). The saturation intensity of the subnanometer region directly under the tip is inversely proportional to the square of the tip enhancement factor.31 Moreover, greater intensity is required for rapidly relaxing states, so the saturation intensity is inversely proportional to the lifetime of the above-bandedge state. Overall,31
electrical contact with sample holders. The samples were degassed at ≈400 K for at least 12 h before STM imaging. n-Type 4H-SiC (Cree Inc., resistivity ∼0.013−0.500 Ω. cm) was used as a semiconducting substrate. To obtain total internal reflection, a 15° wedge was machined at the back side of the sample, then polished until optical quality was obtained using diamond polishing paste. c-SiC samples were resistively degassed at ∼870 K for at least 6 h or until the pressure is lower than 10−7 Pa. Atomically flat and clean c-SiC surface was prepared by annealing at ∼1270 K and ∼1370 K for 10 minutes, then heating to ∼1470 K for 60 s to remove oxides. The cleanliness and semiconducting nature of the surface were checked using STM and STS. a-SiC substrate was prepared from c-SiC substrate. Both prism and wedging methods were used to obtain total internal reflection. The samples were degassed at ∼400 K for ∼12 h. Thin a-SiC films (ca. ∼ 50−100 nm) were produced by sputtering cSiC with 2 keV argon ions for ∼1 h, with base pressure ≤3 × 10−3 Pa.45 Quantum Dot Deposition. PbS (Evident Thermoelectrics) quantum dots dispersed in toluene with corresponding nominal diameters of 4.2 nm (5.5 nm for CdSe/ZnS dots from Ocean NanoTech) were used as received. The first exciton absorption peaks for PbS is at 1221 nm wavelength, as shown in the Figure S1 (530 nm for CdSe/ZnS quantum dots). The surfaces of PbS quantum dots are capped with oleic acid (octadecylamine for CdSe/ZnS dots). The quantum dot dispersity and separation were characterized by optical spectroscopy (Figure S1) and TEM (Figure 1). For PbS quantum dots, optical measurement yields an average bandgap of 1.01 and a dispersion of ±0.04 eV (standard deviation of Gaussian fit in Figure S1a). This translates to an average diameter of ∼4.2 nm,46 in excellent agreement with the manufacturer’s specification, and a dispersity of 4%. TEM measurement yielded an average center-to-center quantum dot separation of 8.0 nm, an average dot diameter of 4.8 nm, and a size dispersion of ±0.6 nm for PbS quantum dots (Figure 1e,f), using the Fiji/ImageJ image analysis tool.47 Thus, the dispersity of PbS quantum dots from TEM measurements is 13%, which we consider less accurate than the spectral fit. The edge-to-edge separation of the PbS quantum dots is therefore 3.2 nm by TEM, in accord with an oleic acid ligand length of ca. 1.6 nm; even combining the more reliable spectral measurement with the smaller center-to-center value obtained from STM (Figure S2), the separation is still ≥2 nm. We adopt ≥2 nm as a lower limit on separation to be safe when comparing electron transfer and energy transfer processes. The quantum dots were deposited onto clean surfaces under ultrahigh vacuum using matrix-assisted dry contact transfer (DCT) in which carbon nanotubes were used as a matrix to prevent quantum dot aggregation. CNT powder was loaded onto a frayed fiber glass piece tied onto a tip holder, then 2−5 drops (ca. 20 μL) of quantum dot solutions (10 mg/mL) were added. The DCT applicators were gently degassed at elevated temperature (ca. 370 K) for ca. 12 h while keeping the pressure ≤10−6 Pa. This degassing procedure has been shown to not affect the quantum dot fluorescence and absorption,26 but it could be responsible of the somewhat tighter packing observed for out PbS quantum dots by STM (Figure S3) vs TEM (Figure 1). STM Setup. STM experiments were performed using a home-built STM with base pressure ≤7 × 10−9 Pa.48 Electrochemically etched W and mechanically cut Pt−Ir (80:20) tips were used. All of the presented data were collected using Pt−Ir tips, except for where noted. The samples were irradiated with p-polarized 532 nm light (diode pumped solid-state laser) in a total internal reflection geometry, to reduce tip heating effects. Our technique approximately detects the electron density difference between the ground and excited states.26 Typical SMA-STM images were 400 pixel × 400 pixel, with scanning speed of ca. 3 ms/pixel. The laser was amplitude-modulated at 2.2 kHz using a mechanical chopper wheel, providing about 6 modulation cycles/pixel scanned for lock-in averaging. Absorption images were simultaneously collected with topographic images by feeding the modulated tunneling current signal into a lock-in amplifier set at a time constant of ca. 10 ms and locked to the chopper at 2.2 kHz. The laser power density was ca. 1200−2600 mW/mm2.
Isat =
2 2 hc 1 × 2 π f ·τ·λ3
For a PtIr (80:20) tip, a value as high as f 2 = 500−1000 has been estimated.31 For PbS quantum dots used in the experiment, τ ≈ 2 ps.38 With λ = 532 nm from the diode laser, the saturation intensity is calculated to be as low as
Isat = 1.2 mW/mm 2 Since we can use laser intensity up to ca. 1000 mW/mm2 (up to 10 mW focused down to a spot size as small as ca. 0.1 mm), we can tune the quantum dot transition from the linear regime into saturation to adjust signal contrast and level, as was demonstrated previously for carbon nanotubes also.31 As discussed above, biexciton relaxation and hot carrier relaxation are sub-nanosecond processes38,39 that are already completed on the sub-microsecond to microsecond time scales of energy transfer or relaxation to the ground state. Lattice Model Calculations. A Monte Carlo lattice model7,21 was employed to simulate the competition of energy transfer between PbS quantum dots with relaxation of quantum dots to the ground state. For PbS quantum dots excited at 532 nm (∼2.33× of the bandgap), a three-level system was used including an initial excited state “3”, a relaxed excited state near the conduction band edge “2”, and a ground state “1”. “0” indicates absence of a dot on the model lattice. Given the low probability of biexciton formation for our PbS quantum dots (1.01 eV bandgap excited at 2.33 eV, whereas the onset for carrier multiplication is ∼2.8 eV5,50), we used just a single state “3” to account for the initially excited state in our simulations. Thus, the fast relaxation time of biexcitons (∼50 ps)39 and intraband relaxation (subns)38 were not considered separately in our model, and state “3” promptly decays to state “2” in our simulations within a single 1 ns simulation time step. The structure of the quantum dot arrays and clusters was generated on a hexagonal lattice based on the experimental topography images. Bandgaps were assigned to all quantum dots, either randomly with average value Eg and standard deviation δEg (derived from optical absorption measurement, see Figure S1) for large arrays (Figure 3), or optimized but with consideration of topological quantum dot size for small clusters by comparison with experiment (Figures 2 and 4). Reproducing the experimental data required setting the bandgaps of the acceptor dots (strong absorption in the experimental absorption images) to be smaller than the next smallest bandgaps in the arrays. Given our excitation power density, which is sufficient to saturate the optical transition,26 approximately 50% of quantum dots were 6333
DOI: 10.1021/acsnano.7b02649 ACS Nano 2017, 11, 6328−6335
Article
ACS Nano initialized in state “3” at the start of a simulation. When a quantum dot is in the highest excited state, there are three decay processes: (1) relaxation to the excited state near the conduction band edge in
