Ultrafast Dynamics of Multiple Exciton Harvesting in the CdSe–ZnO

Nov 19, 2012 - A number of studies report on efficient MEG in QDs of various materials. .... This allows us to describe the system dynamics by focusin...
0 downloads 0 Views 2MB Size
Letter pubs.acs.org/NanoLett

Ultrafast Dynamics of Multiple Exciton Harvesting in the CdSe−ZnO System: Electron Injection versus Auger Recombination Karel Ž ídek,* Kaibo Zheng, Mohamed Abdellah, Nils Lenngren, Pavel Chábera, and Tõnu Pullerits Department of Chemical Physics, Lund University, Box 124, 22100, Lund, Sweden S Supporting Information *

ABSTRACT: We study multiple electron transfer from a CdSe quantum dot (QD) to ZnO, which is a prerequisite for successful utilization of multiple exciton generation for photovoltaics. By using ultrafast time-resolved spectroscopy we observe competition between electron injection into ZnO and quenching of multiexcitons via Auger recombination. We show that fast electron injection dominates over biexcitonic Auger recombination and multiple electrons can be transferred into ZnO. A kinetic component with time constant of a few tens of picoseconds was identified as the competition between injection of the second electron from a doubly excited QD and a trion Auger recombination. Moreover, we demonstrate that the multiexciton harvesting efficiency changes significantly with QD size. Within a narrow QD diameter range from 2 to 4 nm, the efficiency of electron injection from a doubly excited QD can vary from 30% to 70% in our system. KEYWORDS: Quantum dots, solar cell, electron transport, ultrafast transient absorption, multiple exciton generation

D

report on efficient MEG in QDs of various materials.4−9 Several studies also prove successful utilization of MEG enhancement in QD-sensitized SCs based on the QD−metal oxide (MO) system.10,11 In QD-sensitized SCs, electron injection from QD to MO is responsible for rapid charge separation and, hence, allows light energy harvesting.12−15 However, utilizing several excitons out of one QD is always limited by Auger recombination (AugR) inside the QD.16 During AugR, one e−h pair recombines, and its energy is transferred to the electron or the hole in the second e−h pair, which undergoes a rapid thermal relaxation. As a result, the energy of one e−h pair is lost for photovoltaic conversion. AugR is an inverse phenomenon to MEG, and it inevitably follows the MEG process, because multiple e−h pairs are created.5,16 Since AugR is a subnanosecond process, the MEG benefits can be used only with fast electron injection from QD to MO. Recent studies, including our work employing ultrafast timeresolved absorption and terahertz spectroscopy, show that

emand for new, sustainable sources of energy has been a driving force for the immense development of solar cells (SCs) during the last decades. The crucial issue of the expansion of SCs is to find a balanced combination of their efficiency and price. For instance, the efficiency of singlejunction SCs is inherently restricted by the Shockley−Queisser limit of about 32%.1 It originates from the fact that only photons with energy close to the band gap (or HOMO− LUMO) can be fully utilized. In other cases photons are either not absorbed at all (the photon energy is lower), or energy is partly lost due to thermal relaxation of excited carriers (the photon energy is higher). Moreover, the Shockley−Queisser limit represents an ideal case, whereas the efficiency of a real solar cell is further diminished by various losses in the system.2 There are several approaches to overcome this limitation. For instance, multijunction SCs can reach high conversion efficiencies up to 43.5%:3 however, only on expenses of extremely high price. Quantum dots (QDs) offer an alternative to break the Shockley−Queisser limit.3 High-energy photons absorbed in QDs excite carriers, which can relax via the socalled “multiple exciton generation” (MEG) process. Instead of thermal relaxation, energy is converted into the usable form of two or more electron−hole (e−h) pairs. A number of studies © 2012 American Chemical Society

Received: October 8, 2012 Revised: November 16, 2012 Published: November 19, 2012 6393

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399

Nano Letters

Letter

laser system with a regenerative amplifier and converted by an optical parametric amplifier to 470 nm (excitation) and 520− 540 nm (probe) pulses. To construct the TA signal, every second excitation pulse is blocked; the excitation repetition rate is thereby 500 Hz. TA spectra were recorded by using a supercontinuum generated in a thin sapphire plate as the probe beam, which was detected by a diode array coupled to a spectrograph. During the measurements the samples were placed in N2 atmosphere to avoid oxygen-related photodegradation.21 The absorption spectra of the three studied types of QDs consist of overlapping excitonic bands as a consequence of carrier confinement (see Figure 1A).22 The samples were

electron injection in a QD-MO system can take place on a very fast time scale of a few picoseconds.12,17 However, the reports were directed toward the study of electron transfer in the regime where only one e−h pair is excited per QD. This is never the case of MEG, where two and more e−h pairs coexist in one QD. To exploit MEG fully, we need information about electron injection to MO under multiexciton conditions. In this Letter, we mimic MEG by exciting several e−h pairs per QD with a high-intensity excitation. Through direct observation of multiple exciton dynamics in a CdSe QD− ZnO system we establish a solid foundation for distinguishing competition between AugR and electron injection. We demonstrate the possibility of multiple electron injection from QD to ZnO, which is essential for efficient utilization of MEG. Moreover, we study competition between electron injection to ZnO and AugR for different QD sizes. We show that there is a relatively narrow range of QD sizes for which the MEG effect can be successfully exploited in our system. QDs were prepared by following a method described in ref 18 (core CdSe QDs) and ref 19 (core−shell CdSe−ZnS QDs). Details about sample preparation procedures can be found in the Supporting Information. Briefly, the core CdSe QDs capped by oleic acid were prepared using hot injection of the trioctylphosphine-Se (TOPSe) precursor to the oleic acid-Cd precursor. The desired size of QDs (2.5−3.1 nm) was obtained by cooling the mixture from the reaction temperature (180−240 °C) suddenly to room temperature by using an ice bath. Core−shell CdSe−ZnS QDs were prepared by using a singlestep hot injection method. Both the Cd2+ and the Zn2+ oleate in 1-octadecene solution were heated to 325 °C. Then Se2− and S2− in 3 mL of TOP solution were swiftly injected into the cation precursor. Cooling in an ice bath was used to stop the shell growth (at 5 s for thin shellsample “I”and at 5 min for thick shellsample “S”). The core−shell QDs feature a gradual change from core material (CdSe) to pure shell material (ZnS), as it is described in detail in ref 19. As-prepared QDs were purified twice before the sensitization. For the sensitization, the surface capping of the QDs was exchanged from oleic acid to a bifunctional linker molecule, 2mercaptopropionic acid (2MPA). Then the ZnO nanoparticle films were immersed into the QD solution for 2 h in the dark. For our study, we chose three principal examples of QDs with fast, intermediate, and slow electron injection to ZnO. We denote the samples as “F” (fast electron injection, CdSe core QDs, 2.9 nm size), “I” (intermediate case, CdSe core−ZnS thin shell), and “S” (slow electron injection, CdSe core−ZnS thick shell). Different injection rates are a consequence of different shell thickness of the QDs for each sample, as the rate is expected to decreases exponentially with increasing shell thickness.20 For each sample, pure QDs and QDs attached to ZnO were studied to clearly separate the effect of electron injection. Pure QDs prepared as colloidal samples were also deposited on glass to verify that the process of deposition itself does not induce a fast TA decay due to surface trapping or QD agglomeration (see the Supporting Information for details). Both types of samples featured the same optical density of QDs at the excitation wavelength (∼0.2), so that excitation intensity dependences can be compared. Transient absorption (TA) dynamics were measured by using a pump−probe setup described in detail elsewhere.17 In brief, laser pulses at 800 nm (80 fs pulse duration, 1 kHz repetition rate) were generated by a Ti:sapphire femtosecond

Figure 1. (A) Absorption spectra of studied CdSe QDs (solid lines) and example of CdSe−ZnO film absorption (dashed line). (B) TA spectra (delay of 1 ps) of CdSe QDs (solid line) and sensitized NP film (dashed line). For all spectra see Supporting Information. (C) Comparison of TA kinetics for pure QDs (dotted lines) and the QD− ZnO system (solid lines) for the three studied samples. Note that xaxes over 0.5 ps are logarithmic. Probe wavelengths: samples “F”: 526 nm, “I”: 540 nm, “S”: 540 nm; exc. wavelength: 470 nm; exc. intensity: 40 μJ/cm2 (∼1014 photons/cm2).

selected so that the lowest absorption band is centered near 540 nm. Hence the effective QD radius for all samples is similar, and the difference in electron injection between the samples is predominantly given by different shell thicknesses with a marginal contribution from QD size dependence. The broader absorption band of the “S” sample (thick shell layer, see solid blue line in Figure 1A) is a consequence of inhomogeneity introduced by the long shell growth. The acquired TA spectra reflect the band structure of the QD absorption (see Figure 1B). The dominant feature is the lowest absorption band bleach induced by electron state filling.16 The amplitude of the bleach can therefore be used to determine the mean population of excited electrons per QD. The shape of the lowest absorption band bleach is not changing significantly during the decay. This allows us to describe the system dynamics by focusing to TA kinetics in the bleach spectral maximum (see Figure 1C). A minor shift in the lowest absorption band position can be observed for the sensitized samples (see Supporting Information for all spectra), which can be ascribed to changes in the QD environment due to the attachment to ZnO together with a possible QD−ZnO coupling effect.23 6394

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399

Nano Letters

Letter

For pure QDs, all three samples feature a long-lived TA signal, which can be well-described by a two-exponential decay with a long nanosecond component (τ = 13−18 ns, A ∼ 80%) and a minor subnanosecond component (τ = 0.2−0.7 ns, A ∼ 20%)see Figure 1C, dotted lines. Such multiexponential decay is typical for CdSe QDs due to the dynamics of quenching mechanisms and has been described in detail in previous reports.24,25 After attachment of QDs to ZnO NPs, the TA decay becomes significantly fastersee Figure 1C, solid lines. This is a consequence of introducing an additional quenching pathwayelectron injection to ZnO.12,17 As anticipated, the core sample (“F” sample) features very fast electron injection. On the contrary, core−shell QDs with a thick shell layer (“S” sample) inject electrons to ZnO only on the nanosecond time scale, as the electron has to tunnel through a thick shell barrier. The intermediate case (thin-shell sample) stands between those two cases. The electron injection from core−shell QDs to MO is a complex process, which will likely feature a broad distribution of lifetimes due to variation in QD shell properties.14,19 The exact calculation of the lifetime distribution is beyond the scope of this article; however, we can estimate the mean electron injection rate ⟨kinj⟩ from the half-life of the TA signal tH, for which holds ⟨k⟩ = ln 2/tH. By comparing the mean lifetimes of QDs attached and unattached to ZnO, we obtain an estimate of the injection rate:15 ⟨k inj⟩ = 1/⟨τQD − ZnO⟩ − 1/⟨τQD⟩

Figure 2. (A) TA kinetics at different excitation intensities for unattached QDs (sample “S”) normalized on the long-lived decay. (B) Difference between TA kinetics acquired for high-power and lowpower excitation of pure QDs (gray area) fitted by single-exponential decay (dashed black line). (C) Long-lived TA signal (IL) dependence on excitation fluence (squares) fitted by eq 2 (solid lines). IL values were rescaled for a better comparison. The probe wavelength is 540 nm.

there is a minor fast decay in the initial stages (triexciton and higher order AugR). By fitting the AugR dynamics with a single-exponential function, we obtain the biexciton AugR lifetime of 18 ps for the “F” sample (τAR /⟨τinj⟩ = 4.5), 73 ps for the “I” sample (τAR/ ⟨τinj⟩ = 1.2), and 52 ps for the “S” sample (τAR/⟨τinj⟩ = 0.01). All of the data are presented in Supporting Information. The AugR rate differs from sample to sample, as it highly depends on various parameters, such as the effective size of the QD, or its character (core, core−shell).16,27,28 The gradual change of potential from core to shell can significantly alter the AugR rate. The three selected samples clearly represent three principal cases of the AugR−electron injection lifetime balance, where τAR/⟨τinj⟩ is ranging from 0.01 to 4.5. AugR in QDs leads to a universal intensity dependence of the long-lived TA signal, as we will show in the following paragraph.29−31 This is a consequence of rapid AugR of all multiexcitons, after which only a single e−h pair is present in each excited QD. As we described before, the initial e−h pair population in QDs can be represented by a distribution of nexcited QDs. A short excitation pulse which excites in average N0 e−h pairs per QD leads to a Poisson distribution of QDs with k excitations as pk = Nk0/k! exp(−N0).26 The long-lived signal IL consists of the single-exciton contribution from any QD which was excited, because all multiexcitons are quenched by AugR:

(1)

−1

which is, in our case, 0.18 ps (⟨τinj⟩ ∼ 6 ps) for the “F” sample, 0.018 ps−1 (⟨τinj⟩ ∼ 60 ps) for the “I” sample, and 2.5 × 10−4 ps−1 (⟨τinj⟩ ∼ 4 ns) for the “S” sample. So far we have concentrated on a low excitation intensity (∼0.2 e−h pairs per QD), where the number of multiexcitons is small (∼1% of all QDs, < 10% of the excited QDs) and measurements reflect transport of one electron from QD to MO. To study the extraction of multiple electrons from a QD, we need to excite several e−h pairs per QD by increasing the excitation intensity. A typical sign of multiexcitons in QDs is a rapid Auger recombination (AugR) decay, which takes place in our samples during the first 300 pssee Figure 2A.16,26 AugR in QDs cannot be described via carrier concentration, which is commonly used for bulk materials. It is well-established that QD systems have to be represented instead by a distribution of singly, doubly, up to n-excited QDs, because only an integer number of e−h pairs can be present in one QD. AugR or any other recombination in QD systems corresponds to a step-like transition from a QD with n excitations to n−1 excitation. Therefore, unlike in the case of bulk material, for QDs it is possible to assign a distinct biexciton, triexciton, or n-exciton AugR lifetime.26 The features of AugR in QDs are discussed in detail in Supporting Information. After excitation, the multiexcitons recombine step by step until a single e−h pair remains in the QD and decays on the nanosecond time scale (see Figure 2A). It is possible to normalize the TA kinetics for different intensities on the longlived single exciton decay (see Figure 2A) and extract the contribution of AugR to the TA kinetics. It can be obtained as the difference between normalized TA kinetics for high and low excitation power (gray area, Figure 2A).26 The AugR dynamics (Figure 2B) has a single-exponential character (straight line in semilog scale) given by the biexcitonic AugR decay. Besides,

IL ∝ 1 − p0 = 1 − exp( −N0)

(2)

This dependence has been reported in a number of different QD materials, including CdSe QDs.29−31 We observe the dependence for our QDs, as it is illustrated in Figure 2C. The important feature of eq 2 is the possibility to determine the scaling factor between the number of excited e−h pairs per QD and the used excitation power, that is, to obtain the QD absorption cross section. For our samples the resulting absorption cross section at 470 nm ranges from 7 × 10−16 cm2 to 16 × 10−16 cm2 in good agreement with the previously reported values.32 6395

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399

Nano Letters

Letter

connected to electron injection. With increasing intensity, the TA decay becomes faster as expected for AugR. The TA dynamics are then a combination of decay induced by electron injection and intensity-dependent AugR. Ratios between longlived TA signals and TA amplitudes can be found in the Supporting Information. We can use the measured kinetics to estimate the AugR rate for the “I” sample and compare the rates for the cases of attached and unattached QDs. By following the procedure used to determine the mean injection time, we can find the mean TA kinetics lifetime for the QD-ZnO system under various excitation intensities. The mean lifetime changes from 60 ps at low excitation intensities (electron injection only, rate kinj) to 35 ps at high excitation intensities (electron injection in combination with AugR, total rate kT = kinj + kAR). Therefore we can calculate the AugR rate kAR for sample “I” (QDs attached to ZnO) as kAR = kT − kinj, which corresponds to an AugR lifetime of ∼80 ps. This is in good agreement with the value observed for the unattached “I” sample QDs (73 ps). Therefore we can conclude that the AugR rate is not significantly altered by the attachment of the QD to ZnO and the presence of electron injection. The role of a fast electron injection becomes clearly dominant only in the case of the “F” sample (see Figure 3B). Unlike in the previous cases, with increasing excitation power the TA dynamics slightly slow down for excitation intensities up to ∼700 μJ/cm2 (corresponds to N0 ∼ 1.5). To obtain more information about the changes, we calculate the difference between normalized TA kinetics (320 μJ/cm2 and 780 μJ/cm2). We observe that the slow dynamics feature about 10% of the amplitude and decays on the scale of tens of picoseconds (fitted lifetime value of ∼50 ps). For the unattached QDs, the changes in the TA dynamics with excitation intensity are due to the biexcitonic AugR. In case of the above slowdown of the dynamics in the “F” sample (QD−ZnO), this origin can be excluded, as it would lead to faster TA signal decay. Therefore we conclude that the 50 ps component is a signature of electron injection to MO. We will now shortly review the sequence of processes present for doubly excited QDs (see Scheme 1). First, two e−h pairs are excited in a QD and electron injection to ZnO competes with biexcitonic AugR. After the first electron is injected to ZnO, an electron and two holes are present in QD forming the so-called positive trion, which can also undergo AugR. Only after the second electron injection, AugR is not possible anymore.

We will now compare how the multiexcitonic dynamics change after QD attachment to ZnO. Figure 3A,B displays

Figure 3. (A, B) Normalized TA kinetics of CdSe−ZnO films for different excitation intensities (arrow and darker color signalizes increasing excitation power). (C) Difference between normalized TA kinetics of the “F” sample (solid squares) fitted by a single-exponential decay (red solid line)see text for details. (D, E) Dependence of the long-lived signal (IL) to amplitude (I0) ratio on number of excited e−h pairs (N0) for unattached QDs (open squares) and QD−ZnO systems (solid squares). “Long-lived signal” refers to the TA signal at 600 ps (“S” sample) and 200 ps (“F” sample)both more than 10τAR. Exc. powers used: “F”: 90, 320, 780, 1440 μJ/cm2; “I”: 30, 150, 470, 980 μJ/cm2; “S”: 20, 360, 1800 μJ/cm2. Exc. wavelength: 470 nm; probe wavelengths: “F”: 526 nm, “I”: 540 nm, “S”: 540 nm.

dependence of TA decay on excitation intensity for the QD− ZnO samples. In case of the “S” sample (where AugR is considerably faster than the electron injection) the change in TA dynamics with excitation intensity is the same for attached and unattached QDs. This can be illustrated by comparing a ratio between long-lived TA signal IL and TA signal amplitude I0 for attached and unattached QDs (see Figure 3D). As the intensity dependence is the same for both cases, we can unambiguously assign the fast appearing component to AugR. The situation partly changes for the “I” sample, where already the low-intensity kinetics feature a fast TA decay

Scheme 1. Electron Injection and Auger Recombination Processes Present in Doubly-Excited QDs Attached to ZnO

6396

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399

Nano Letters

Letter

For the samples “F”, “I”, and “S” the η1 values reaches approximately 80%, 50%, and 1%, respectively. If the system dynamics is simpler (purely core CdSe QDs on ordered ZnO nanowires) and better explored (for instance by combination of TA and THz spectroscopy), it is possible to determine the electron injection rate precisely, as we have shown in our previous work.17 We will use this case to answer the important question of QD size dependence of multiexciton harvesting. The change in electron injection with QD size for the CdSe QD−2MPA−ZnO NW system has been determined in ref 17. for four different QD sizes (see Figure 4B). The injection rates

A slow decay can be assigned to the competition between the second electron injection and trion decay. It provides us with important information about electron injection, because the observed total rate (lifetime ∼50 ps) is a sum of the rates of the two involved processes. Therefore, the second electron injection can take place only on the scale of tens of picoseconds or longer. The second electron injection is, indeed, expected to be slower than the first (or single) electron injection. When second and other electrons are injected into ZnO, they are affected by interaction with the already charged QD. Furthermore, the injection can be also slowed down by electrons already injected into ZnO, which can slightly modify the ZnO conduction band properties due to band gap renormalization and state filling.33 However, due to the native n-type character of ZnO NPs, the change in ZnO bands should only be minor. A better estimate of the second electron injection lifetime would be possible only by determining the trion lifetime. Although trions have been extensively studied in CdSe QDs,28,34−36 their lifetime is highly dependent on QD properties, and the reported values cannot be straightforwardly used for our QDs. An estimate can be based on the observation that the lifetime of a negative trion (two electrons and a hole) was reported to be approximately 8× longer than the biexciton AugR lifetime.34 From this crude estimate we obtain for the sample “F” the trion lifetime of ∼140 ps, which would facilitate an efficient injection of the second electron in line with the previously reported multiple electron collection.10 Nevertheless, our results suggest that the second electron injection takes place on the time scale of tens of picoseconds, which is significantly longer compared to ⟨τinj⟩ ∼ 6 ps for the single electron. Under excitation fluences above approximately 700 μJ/cm2 (N0 ∼ 1.5) even for the “F” sample the TA dynamics become faster with increasing excitation power (see Figure 3B). For N0 > 2, the portion of triply and more exited QDs pn>2 calculated from the Poisson distribution rapidly rises with excitation intensity. Namely, for the measured values of N0 ∼ 0.2, 0.6, 1.5, 2.5, and 4, we obtain pn>2 = 0.001, 0.02, 0.19, 0.46, and 0.76, respectively. Very fast AugR from more than doubly excited QDs overrides the electron injection, and it is responsible for the observed faster dynamics of the “F” sample for high excitation fluences. To summarize the obtained picture, in the case of sample “S” AugR is fast enough to quench practically all multiple excitons in QD before electron injection; sample “F” features rapid electron injection to MO, which dominates over biexciton AugR, and injection of the second electron can be observed. This is a very favorable situation for MEG exploitation, because MEG typically generates 2 e−h pairs per QD, which can be efficiently harvested in this case. We can estimate the probability of the first electron injection to ZnO from a doubly excited QD by comparing the rate of electron injection kinj, the rate of e−h pair recombination kR = 1/⟨τQD⟩, and the rate of biexcitonic AugR (rate kAR). Electrons can undergo one of the processes, and their number N will change as dN/dt = −kinjN − kRN − kARN. The efficiency of the first electron injection (injection efficiency η1) can be then calculated as: η1 = k inj/(k inj + kAR + kR )

Figure 4. (A) Normalized biexciton AugR contribution to TA kinetics for various QD diameters (open symbols) fitted by single-exponential decay (solid lines). (B) Electron injection rate dependence on QD size (orange squares, adopted from ref 17) fitted by the Marcus theory prediction (solid orange line, adopted from ref 17); AugR biexciton recombination rates from the fit on panel A (dark cyan circles) and ref 26. (dark cyan triangles) fitted by D−p fit (p = 4.9, dark cyan line). (C) First electron injection efficiency from doubly excited QDs for various QD sizes calculated from experimental data (open circles) and fits presented in panel B (line). Probe wavelengths used in panel (A): 2.5 nm QDs: 506 nm; 2.7 nm QDs: 516 nm; 2.9 nm QDs: 526 nm; 3.1 nm QDs: 540 nm.

can be well-described by Marcus theory,37 whose calculation and fit to the data has been discussed in detail in ref 17 and the Supporting Information. The calculation of free energy difference ΔG before and after the electron transfer is based on experimental measurements of the band alignment of the CdSe−ZnO system38 together with the Coulomb e−h interaction calculation proposed by Tvrdy et al.12 To calculate the AugR-limited injection efficiency, it is also essential to know the size dependence of the AugR rate, namely, dependence on the QD diameter D, as the used QDs have a spherical shape. Therefore we have repeated the procedure illustrated in Figure 2A and B to extract the AugR decay from TA kinetics for each of the QD sizes studied in ref 17 (the same QDs were used). The change in TA dynamics of unattached QDs under high and low intensity excitation (see Supporting Information for details) allows us to evaluate the biexciton AugR lifetime. In accordance with previous reports,23,28 the AugR lifetime becomes significantly shorter with decreasing QD size (from 5 ps for 2.5 nm size to 20 ps for 3.1 nm sizesee Figure 4A and B, circles). Our results are in line with the study of Klimov et al.26 (Figure 4B, triangles), where the same method of biexciton

(3) 6397

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399

Nano Letters



AugR lifetime extraction of CdSe QDs was used. The obtained QD size dependence of AugR rates can be well fitted by a D−p function for p = 4.9, which is in agreement with previously reported dependences.39 It is worth stressing that this fit is an approximation valid for the studied range of QD sizes. The QD size dependence of the AugR rate can be very complicated, as has been suggested by theoretical calculations.27,40 Therefore the overall AugR rate scaling for a broad range of QD sizes, which is usually described as R−3, may not be the best fit for this particular span of QD sizes.16 The AugR and electron injection rates feature considerably different dependences on QD size. Therefore the AugR-limited electron harvesting is also strongly size-dependent (see Figure 4C). Already within the narrow range of studied QD sizes, the efficiency of the first electron injection can vary from 30% up to 70%. For small QDs (below 2 nm), AugR is very fast, and it is efficiently quenching all multiple excitations. Analogously, for large QDs (over 3.5 nm), electron injection becomes inefficient because of unfavorable electron states alignment, and again the AugR is dominating the system dynamics.12,14 A relatively narrow range of CdSe QD sizes (2−3.5 nm) offers a good electron injection efficiency with maximum value of approximately 70% at about 3 nm for this particular system (2MPA linker molecule, ZnO nanowire acceptor). Naturally, the resulting efficiency dependence will change for different linker molecules, QD materials, and electron acceptors. However, the general rule of existence of an optimum QD size for MEG exploitation will remain, because of different QDsize dependence of the electron injection rate in QD−MO systems and AugR in QDs.12,26 In conclusion, our results clearly demonstrate that a fast electron injection from QD to ZnO can inhibit the unwanted Auger recombination in doubly excited QDs. Efficient energy harvesting from multiple excitons created by MEG is therefore a feasible scenario. Moreover, we demonstrate that MEG harvesting efficiency changes significantly with QD size. Already within the narrow range of QD sizes (2−4 nm), double-exciton harvesting efficiency can vary from 30% to 70%. Therefore for full utilization of the MEG effect, a combination of band gap tuning (various alloys, such as PbSxSe1−x) and QD size tuning should be used to maximize the benefits of MEG.



REFERENCES

(1) Shockley, W.; Queisser, H. J. J. Appl. Phys. 1961, 32, 510. (2) Giebink, N.; Wiederrecht, G.; Wasielewski, M.; Forrest, S. Phys. Rev. B 2011, 83, 1−6. (3) Green, M. A.; Emery, K.; Hishikawa, Y.; Warta, W.; Dunlop, E. D. Prog. Photovolt. Res. Appl. 2012, 20, 12. (4) Trinh, M. T.; Polak, L.; Schins, J. M.; Houtepen, A. J.; Vaxenburg, R.; Maikov, G. I.; Grinbom, G.; Midgett, A. G.; Luther, J. M.; Beard, M. C.; Nozik, A. J.; Bonn, M.; Lifshitz, E.; Siebbeles, L. D. A. Nano Lett. 2011, 11, 1623−9. (5) Ellingson, R. J.; Beard, M. C.; Johnson, J. C.; Yu, P.; Micic, O. I.; Nozik, A. J.; Shabaev, A.; Efros, A. L. Nano Lett. 2005, 5, 865. (6) Schaller, R. D.; Agranovich, V. M.; Klimov, V. I. Nat. Phys. 2005, 1, 189. (7) Murphy, J. E.; Beard, M. C.; Norman, A. G.; Ahrenkiel, S. P.; Johnson, J. C.; Yu, P.; Mićić, O. I.; Ellingson, R. J.; Nozik, A. J. J. Am. Chem. Soc. 2006, 128, 3241. (8) Nair, G.; Bawendi, M. Phys. Rev. B 2007, 76, 081304(R). (9) Beard, M. C.; Knutsen, K. P.; Yu, P.; Luther, J. M.; Song, Q.; Metzger, W. K.; Ellingson, R. J.; Nozik, A. J. Nano Lett. 2007, 7, 2506. (10) Sambur, J. B.; Novet, T.; Parkinson, B. A. Science 2010, 330, 63. (11) Semonin, O. E.; Luther, J. M.; Choi, S.; Chen, H.-Y.; Gao, J.; Nozik, A. J.; Beard, M. C. Science 2011, 334, 1530. (12) Tvrdy, K.; Frantsuzov, P. A.; Kamat, P. V. Proc. Natl. Acad. Sci. U.S.A. 2011, 108, 29. (13) Blackburn, J. L.; Selmarten, D. C.; Nozik, A. J. J. Phys. Chem. B 2003, 107, 14154. (14) Robel, I.; Kuno, M.; Kamat, P. V. J. Am. Chem. Soc. 2007, 129, 4136. (15) Guijarro, N.; Lana-Villarreal, T.; Shen, Q.; Toyoda, T.; Gómez, R. J. Phys. Chem. C 2010, 114, 21928. (16) Klimov, V. I. Annu. Rev. Phys. Chem. 2007, 58, 635. (17) Ž ídek, K.; Zheng, K.; Ponseca, C. S., Jr.; Messing, M. E.; Wallenberg, L. R.; Chábera, P.; Abdellah, M.; Sundström, V.; Pullerits, T. J. Am. Chem. Soc. 2012, 134, 12110. (18) Bullen, C. R.; Mulvaney, P. Nano Lett. 2004, 4, 2303. (19) Bae, W. K.; Char, K.; Hur, H.; Lee, S. Chem. Mater. 2008, 20, 531. (20) Dworak, L.; Matylitsky, V. J. Phys. Chem. C 2011, 115, 3949. (21) Ž ídek, K.; Zheng, K.; Chabera, P.; Abdellah, M.; Pullerits, T.; Chábera, P.; Pullerits, T. Appl. Phys. Lett. 2012, 100, 243111. (22) Norris, D.; Bawendi, M. Phys. Rev. B 1996, 53, 16338. (23) Yang, Y.; Rodríguez-Córdoba, W.; Xiang, X.; Lian, T. Nano Lett. 2012, 12, 303. (24) Schlegel, G.; Bohnenberger, J.; Potapova, I.; Mews, A. Phys. Rev. Lett. 2002, 88, 137401. (25) Sharma, S. N.; Pillai, Z. S.; Kamat, P. V. J. Phys. Chem. B 2003, 107, 10088. (26) Klimov, V.; Mikhailovsky, A.; McBranch, D. W.; Leatherdale, C. A.; Bawendi, M. G. Science 2000, 287, 1011. (27) Cragg, G. E.; Efros, A. L. Nano Lett. 2010, 10, 313. (28) Wang, X.; Ren, X.; Kahen, K.; Hahn, M. A.; Rajeswaran, M.; Maccagnano-Zacher, S.; Silcox, J.; Cragg, G. E.; Efros, A. L.; Krauss, T. D. Nature 2009, 459, 686. (29) Timmerman, D.; Izeddin, I.; Gregorkiewicz, T. Phys. Status Solidi A 2010, 207, 183. (30) Achermann, M.; Hollingsworth, J.; Klimov, V. I. Phys. Rev. B 2003, 68, 245302. (31) Robel, I.; Gresback, R.; Kortshagen, U.; Schaller, R.; Klimov, V. Phys. Rev. Lett. 2009, 102, 177404. (32) Yu, W. W.; Qu, L.; Guo, W. Chem. Mater. 2003, 15, 2854. (33) Ye, J. D.; Gu, S. L.; Zhu, S. M.; Liu, S. M.; Zheng, Y. D.; Zhang, R.; Shi, Y. Appl. Phys. Lett. 2005, 86, 192111. (34) Jha, P. P.; Guyot-Sionnest, P. ACS Nano 2009, 3, 1011. (35) Louyer, Y.; Biadala, L.; Tamarat, P.; Lounis, B. Appl. Phys. Lett. 2010, 96, 203111. (36) Marceddu, M.; Saba, M.; Quochi, F.; Lai, A.; Huang, J.; Talapin, D. V.; Mura, A.; Bongiovanni, G. Nanotechnology 2012, 23, 015201. (37) Marcus, R. A. J. Chem. Phys. 1956, 24, 966.

ASSOCIATED CONTENT

S Supporting Information *

Detailed information about sample preparation; sample characterization; all transient absorption kinetics; and biexciton Auger recombination lifetime extraction. This material is available free of charge via the Internet at http://pubs.acs.org.



Letter

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS This work was supported by the Knut and Alice Wallenberg Foundation, the Crafoord Foundation, and the Swedish Energy Agency. We thank K. J. Karki for valuable discussions and T. Pascher for his help with experiments. Collaboration within nmC@LU is acknowledged. 6398

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399

Nano Letters

Letter

(38) Carlson, B.; Leschkies, K.; Aydil, E. S.; Zhu, X.-Y. J. Phys. Chem. C 2008, 112, 8419. (39) Chepic, D. I.; Efros, A. L.; Ekimov, A. I.; Ivanov, M. G.; Kharchenko, V. A.; Kudriavtsev, I. A.; Yazeva, T. V. J. Lumin. 1990, 47, 113. (40) Kharchenko, V.; Rosen, M. J. Lumin. 1996, 70, 158.

6399

dx.doi.org/10.1021/nl303746d | Nano Lett. 2012, 12, 6393−6399