pubs.acs.org/NanoLett
Multiple Exciton Generation in Single-Walled Carbon Nanotubes Shujing Wang,†,§ Marat Khafizov,†,§ Xiaomin Tu,‡ Ming Zheng,‡ and Todd D. Krauss*,† † ‡
Department of Chemistry and the Institute of Optics, University of Rochester, Rochester, New York 14627 and National Institute of Standards and Technology, 100 Bureau Drive, Gaithersburg, Maryland 20899 ABSTRACT Upon absorption of single photons, multiple excitons were generated and detected in semiconducting single-walled carbon nanotubes (SWNTs) using transient absorption spectroscopy. For (6,5) SWNTs, absorption of single photons with energies corresponding to three times the SWNT energy gap results in an exciton generation efficiency of 130% per photon. Our results suggest that the multiple exciton generation threshold in SWNTs can be close to the limit defined by energy conservation. KEYWORDS Single-walled carbon nanotubes, ultrafast optical spectroscopy, multiple exciton generation, carrier multiplication, excited-state dynamics
I
nism responsible for the MEG process are currently open questions.14-18 Single-walled carbon nanotubes (SWNTs) possess truly unusual mechanical, optical, and electronic properties. SWNTs can be either metals or semiconductors depending on their structure.19 Semiconducting SWNTs are strong absorbers of visible and near-infrared light, deriving their optical properties from an excitonic excited state.20,21 Twodimensional quantum confinement of electrons and holes around the SWNT circumference leads to a distinctive density of states and well-defined selection rules for discrete optically allowed transitions.19 Like semiconductor NCs, SWNTs exhibit a strong Coloumb interaction between electons and holes,20-22 which suggests that SWNTs could also exhibit MEG.15,23 Significantly, SWNTs provide an important alternative system for MEG studies, since the important question of MEG efficiency relative to the “bulk” is irrelevant, as semiconducting NTs have no direct bulk counterpart. Ultrafast transient absorption (TA) spectroscopy is a common technique that allows direct observation of MEG processes.10,24 Specifically, when the average number of excited electron-hole (e-h) pairs per nanoparticle N0 is less than 1, excited-state dynamics are typically governed by a relatively slow decay component corresponding to electron-hole recombination. However, in the presence of multiple excitons (N0 > 1), regardless of whether they are generated through a high pump fluence25 or MEG processes, an additional fast decay component is observed in the TA signal, as illustrated in Figure S1 in Supporting Information. By measuring the relative contributions of the fast (multiple) and slow (single) exciton decay components at low pump fluence, the efficiency of MEG processes can be determined.5,24 Using TA spectroscopy, MEG in SWNTs has been reported with a quantum yield (QY) for exciton generation of 130% with a pump photon energy of 267 nm, 3.7 times the energy gap
n solar cells incorporating conventional semiconductors, typically one charge carrier is created upon absorption of a single photon, resulting in nonoptimal energy utilization since excess energy from photons with energies greater than the semiconductor band gap is lost through phonon emission.1 However, absorption of a single highenergy photon can potentially create more than one charge carrier, whereby a highly energetic electron can lose excess energy by exciting a charge carrier across the energy gap.2 This process, called carrier multiplication (CM), and a similar process for photoexcitation of a quasi-particle called multiple exciton generation (MEG) are especially important for future photovoltaics as they potentially could result in higher photon-to-current conversion efficiencies than those defined by the Shockley-Quiesser limit of energy conversion.1 The creation of multiple charge carriers via absorption of a single photon is a well-known phenomenon in bulk semiconductors.3,4 However, carrier multiplication efficiencies are insignificant at photon energies present in the solar spectrum. It has been proposed by Nozik2 that MEG is more efficient in semiconducting nanocrystals (NCs) relative to bulk materials due to well-separated energy bands in the NC that should slow phonon-assisted carrier cooling and due to relaxation of strict momentum conservation, which makes MEG more competitive with other relaxation pathways. Indeed, recently MEG was reported in NCs of PbSe,5,6 PbS,6 CdSe,7 InAs,8 and Si.9 However, this effect is not yet understood, as the apparent MEG efficiency depended strongly on particulars of the sample or the experimental procedures, such as using stirred or static conditions.9-11 In addition, whether NCs have an enhanced MEG effect relative to their bulk counterparts11-13 or what is the fundamental mecha* To whom correspondence should be addressed,
[email protected]. § Contributed equally to this work. Received for review: 01/29/2010 Published on Web: 05/27/2010
© 2010 American Chemical Society
2381
DOI: 10.1021/nl100343j | Nano Lett. 2010, 10, 2381–2386
FIGURE 1. Absorption and fluorescence spectra of the DNA-wrapped (6, 5) enriched SWNT sample. The first, second, and third excited excitonic states are labeled in the figure. The vertical red arrows indicate the wavelengths of pump pulses.
(pω ) 3.7Eg).26 However, these samples contain a large inhomogeneous distribution of SWNT structures, including a significant fraction of metallic and bundled nanotubes, which significantly complicate interpretations of the TA signal.27 Given the propensity for MEG-like artifacts to appear,10,11 we chose to study MEG in SWNT samples of a single chirality, free from residual metal SWNTs and nanotube bundles. In this Letter, we present studies of the generation and detection of multiple excitons in single-chirality enriched (6,5) SWNTs using TA spectroscopy. With a significantly high pump photon energy a fast component (∼ 1 ps) was present in the TA dynamics even at low excitation fluence when N0 , 1. Additionally, the characteristic decay time of the fast component (at low fluence) equaled the time constant of exciton-exciton annihilation processes obtained when exciting at high fluence, strongly suggesting the observation of MEG. Careful analysis of the recovery dynamics of the E11 exciton for different pump energies (335, 400, and 800 nm) was used to determine the efficiency of MEG in SWNTs. A maximum MEG efficiency of ∼130% was observed when pumping at 335 nm corresponding to pω ) 3.0Eg. SWNTs produced by the CoMoCAT process28 were wrapped in DNA and highly enriched in (6,5) SWNTs using previously published methods.29,30 Absorption and fluorescence spectra of the (6,5) SWNT sample are shown in Figure 1. Strong absorption peaks at 993, 572, and 350 nm correspond to the E11, E22, and E33 transitions for (6,5) SWNTs. The lack of strong absorption features corresponding to SWNTs of other structures31 indicates a sample highly enriched in the single (6,5) chirality. Transient absorption spectra were obtained on a 1 kHz Ti:sapphire regenerative amplifier laser system with a typical pulse width of 150 fs using a standard two color pump-probe setup described previously.32 The 990 nm light (probe) was obtained from the second harmonic of the idler output of the optical parametric amplifier (OPA). The 335 nm light (pump) was generated by quadrupling the signal output from the OPA. Pump wavelengths of 800 and 400 nm were © 2010 American Chemical Society
FIGURE 2. (a) Transient absorption spectra at 800 nm excitation with increasing pump fluence. The black lines are quantitative fits based on a triexponential function. The fluences from bottom to top are 0.9, 1.4, 1.9, 2.4, 2.8, 3.8, 4.7, 6.1, 7.5, 9.4, 12, 14, 19, and 24 × 1013 photons/cm2. (b) ∆T/T at 0.2 ps (open circles) and 13 ps (closed circles) as a function of pump fluence with 800 nm excitation. The red solid line shows the fit to ∆T/T (tL) assuming a Poisson distribution in the number of absorbed photons per SWNT. The black solid line is a linear fit to ∆T/T (tE) at low pump fluence when N0 e 1.
obtained by using the fundamental and second harmonic of the regenerative amplifier, respectively. All experiments were carried out at 300 K. The sample was placed in a quartz cuvette with a 2 mm path length and had an optical density of ∼0.2 at 990 nm. All the measurements were performed with the sample vigorously shaken at 19 Hz such that each consecutive pulse sampled fresh material. It is important to understand the relevant ultrafast dynamics of the lowest excited exciton state (E11) in order to correctly interpret spectral signatures of the generation of multiple excitons. SWNT ultrafast excited-state dynamics have been extensively studied by TA32-36 and photoluminescence spectroscopies.36,37 However, recent studies of near single chirality (i.e., single (n, m) index) species have revealed some important insights into the excited-state dynamics,27,38,39 since these signals are unencumbered by significant interference from metallic and bundled nanotubes in the sample. Figure 2 illustrates typical TA signals for (6,5) SWNTs under different pump intensities with the probe tuned to E11. The pump-induced transmission change ∆T/T is positive due to the photobleaching of the E11 exciton state.32,40 At low excitation fluence the recovery of the E11 2382
DOI: 10.1021/nl100343j | Nano Lett. 2010, 10, 2381-–2386
signals at early decay times tE ) 0.2 ps and late decay times tL ) 13 ps under increasing pump fluence. Since tL is much greater than the exciton annihilation lifetime (τa ∼ 1 ps), at this point essentially all SWNTs have at most one exciton independent of their initial occupancy. Therefore, the TA signal amplitude at tL represents a measure of the total number of photoexcited SWNTs.9 Assuming that on average N0 photons are absorbed per NT and the probability Pm of absorbing m photons is described by Poisson statistics Pm ) N0m exp(-N0)/m!, then the fraction of SWNTs absorbing at least one photon P1 is given by
exciton is characterized by two distinct components:27,32,41,42 (1) a faster component with a characteristic time of a few picoseconds, and (2) a slower component with a characteristic time typically tens to hundreds of picoseconds. The faster dynamics have been attributed to the initial decay of bright E11 exciton state population to either optically dark or trap states.27,41,42 The slower dynamics are the same order as lifetimes observed in time-resolved photoluminescence, suggesting that this portion of the response arises from excitons in the lowest excited state. However, it is clear that both the slower and faster decay components must include contributions from nonradiative pathways for exciton relaxation to account for the low photoluminescence quantum efficiency of
