week ending 4 APRIL 2014
PHYSICAL REVIEW LETTERS
PRL 112, 137402 (2014)
Singlet-Triplet Annihilation Limits Exciton Yield in Poly(3-Hexylthiophene) Florian Steiner, Jan Vogelsang,* and John M. Lupton Institut für Experimentelle und Angewandte Physik, Universität Regensburg, Universitätsstrasse 31, 93053 Regensburg, Germany (Received 21 August 2013; published 2 April 2014) Control of chain length and morphology in combination with single-molecule spectroscopy techniques provides a comprehensive photophysical picture of excited-state losses in the prototypical conjugated polymer poly(3-hexylthiophene) (P3HT). Our examination reveals a universal self-quenching mechanism, based on singlet-triplet exciton annihilation, which accounts for the dramatic loss in fluorescence quantum yield of a single P3HT chain between its solution (unfolded) and bulklike (folded) state. Triplet excitons fundamentally limit the fluorescence of organic photovoltaic materials, which impacts the conversion of singlet excitons to separated charge carriers, decreasing the efficiency of energy harvested at high excitation densities. Interexcitonic interactions are so effective that a single P3HT chain of order 100 kDa weight behaves like a 2-level system, exhibiting perfect photon antibunching. DOI: 10.1103/PhysRevLett.112.137402
PACS numbers: 78.55.Kz, 42.50.Ar, 42.70.Jk, 78.66.Qn
To unravel the photophysical characteristics of multichromophoric systems such as conjugated polymers (CPs) or light-harvesting complexes poses ongoing interdisciplinary challenges for chemists, physicists, materials scientists, and spectroscopists [1–3]. Some of the oldest and still crucial questions in CP photophysics include the following: (i) What chemical unit absorbs and emits light in a CP? (ii) Which processes occur between absorption and emission events? And (iii) what is the nature of the interplay between efficient excitation energy transfer (EET) among chromophores and nonradiative fluorescence decay (quenching)? The third question is particularly relevant in organic photovoltaics because long-range EET is required to optimize charge photogeneration, although it promotes exciton quenching by dark states such as long-lived triplets or radicals. Answering these questions, and specifically elucidating which physical mechanisms are responsible for fluorescence self-quenching, is a prerequisite to the fundamental understanding of CP photophysics. The heterogeneity of CPs with respect to their spectroscopic properties and morphological diversity has made singlemolecule spectroscopy (SMS) an indispensable tool [4–9]. However, important parameters such as chain shape or size are difficult to assess in detail solely via SMS because these characteristics can be convoluted with the photophysical observables themselves [10]. The formation, number, and interaction pathways of chromophores in CPs correlate with changes in emission spectra and the degree of selfquenching [11–16]. However, the microscopic physical mechanism underlying this self-quenching phenomenon remains under debate [11,13]. Here, we demonstrate that control of molecular weight (defined by the number-average molecular weight, M n ) and chain morphology, in combination with SMS techniques, lead to a universal picture of self-quenching in CPs. As a model system, we use the prototypical CP poly 0031-9007=14=112(13)=137402(5)
(3-hexylthiophene) [P3HT; Fig. 2(a)] which is employed widely in photovoltaics research [17]. In this material, the fluorescence quantum yield drops approximately 20-fold on transitioning from well-dissolved (unfolded) chains in solution to aggregated (folded) chains in bulk film [18]. By controlling chain morphological characteristics using solvent or matrix polarity in well-defined subpopulations of different Mn , fractionated via gel-permeation chromatography (GPC), we correlate microscopic photophysical observables with each other and with Mn . The increase in fluorescence self-quenching with increasing M n and morphological ordering can be rationalized by improved light-harvesting and singlet-triplet quenching. In other words, triplet excitons fundamentally limit the lifetime of singlet excitons; this process can reduce the efficiency of organic solar cells. The microscopic photophysical model of exciton selfquenching is formulated in Fig. 1. Panel (a), which shows the expected PL intensity versus M n for the two extreme morphologies: (i) unfolded, as it occurs for well-dissolved CPs in solution with few, if any, intrachain contacts [19]; and (ii) folded, as it arises for aggregated CPs in bulk film with strong intrachain contacts [20,21]. In case (i), the PL intensity should increase linearly with increasing Mn because intrachain interchromophoric interactions are negligible. However, in case (ii), we expect saturation for folded chains, provided that EET occurs to a single emitting acceptor chromophore (red). This acceptor ultimately enters a triplet state, so that singlet-triplet annihilation (STA) can quench the excitation energy of the complete CP chain. Panel (b) shows a simplified level scheme of the acceptor chromophore, which is excited at rate kexc from the singlet ground state S0 to the first excited state S1 . S1 decays (radiatively and nonradiatively) to S0 at rate kPL , or undergoes intersystem crossing into the triplet T 1 (rate kisc ). The acceptor chromophore emits photons; i.e., it is in an
137402-1
© 2014 American Physical Society
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
week ending 4 APRIL 2014
τon PLmax M n I exc −1 hPLi ∝ ∝ M n I exc 1 þ ; τon þ τoff kT
FIG. 1 (color online). (a) Photoluminescence (PL) intensity dependence on number-average molecular weight, M n , for unfolded and folded conjugated polymer (CP) chains. Efficient excitation energy transfer (EET) occurs for folded chains, such that only one chromophore in the entire chain (green) emits light (acceptor [red]). (b) Level scheme of the acceptor chromophore. (c) The molecule is subject to triplet blinking with constant “off” times, τoff , and Mn -dependent “on” times, τon , assuming that the acceptor chromophore is populated by EET from the entire absorbing chain (donor [green]).
“on” state for a characteristic duration τon as it cycles between S0 and S1 . Following intersystem crossing, the chromophore enters an “off” state for τoff , as illustrated in panel (b). Assuming that k−1 exc ≪ kPL þ kisc , τon can be described as follows [22]: τon ¼ ðkPL þ kisc Þ=ðkisc kexc Þ ∝ k−1 exc ;
(1)
where kexc is proportional to the excitation intensity, I exc , and the absorption cross section. We propose that the entire chain [green, Fig. 1(a)] absorbs light and transfers energy to the acceptor chromophore [red, Fig. 1(a)]. Therefore, kexc is, to first-order approximation, directly proportional to Mn ; Eq. (1) becomes τon ∝ ðMn I exc Þ−1 .
(2)
Through STA, which triggers quenching of the complete folded CP once the acceptor chromophore enters the T1 state, τoff becomes independent of M n : τoff ¼ ðkT Þ−1 ;
(3)
where kT represents the reverse intersystem crossing rate (the inverse triplet lifetime). Therefore, the average PL intensity of the single chain [marked red in Fig. 1(c)] is determined as follows:
(4)
where PLmax denotes the maximum PL intensity during τon , which is proportional to kexc and thus to M n and I exc . Equation (4) demonstrates the dependence of hPLi: it saturates with I exc or Mn, as shown schematically in the lower part of Fig. 1(a), provided that only a single accepting chromophore is present within the polymer. To test this simple model, we revisited the dependence of hPLi on Mn and I exc for unfolded and folded P3HT chains [11,13]. We fractionated P3HT by GPC into six different M n samples with a low polydispersity index, ranging from 19 to 110 kDa versus polystyrene standards (see Supplemental Material, Table S1 [23]). We stress that the actual Mn of P3HT may differ significantly by a constant factor from the values reported herein due to the different rigidity of P3HT compared with the polystyrene standards. This difference is irrelevant for our study since M n of the polymer and the standard are described by a power-law, allowing relative weight comparisons [24]. Perfectly unfolded isolated chains can only be formed in nonpolar solution. Therefore, we diluted each sample to concentrations of 10−10 to 10−11 M in toluene; we obtained hPLi via fluorescence correlation spectroscopy (FCS; see Supplemental Material, Fig. S1 and Table S1 for details [23]) [11]. Figure 2(a) shows single-chain hPLi dependent on Mn . The PL intensity increases linearly, as expected, for well-dissolved chains in solution. This increase is a necessary control to validate that the Mn values, as obtained via GPC, are not compromised due to interchain aggregation at higher M n [11,19]. However, the average PL intensity for folded chains cannot be measured in solution: inducing self-aggregation by raising solvent polarity will drive isolated single chains together, forming larger aggregates [25]. Instead, single chains are embedded in a polar host matrix, poly(methyl-methacrylate) (PMMA), where most chains fold to form highly-ordered structures (i.e., self-aggregate) but remain isolated within the matrix [26]. Because the molecules are stationary, we cannot use FCS, which is a diffusion-based technique. Instead, we acquired the hPLi intensity measurements of single P3HT chains by confocal scanning of several areas of 20 × 20 μm2 (see Supplemental Material Fig. S2 and Table S1 [23] for details on the experimental apparatus). We identified diffractionlimited PL spots of the single chains using an automated spot-finding algorithm. For each M n , we acquired the average PL intensity over approximately 200 chains. All measurements were carried out under nitrogen atmosphere. The average PL intensity of folded P3HT chains, as plotted in Fig. 2(b), saturates quickly with increasing M n and follows Eq. (4) [black curve in Fig. 2(b)]. We observe the same saturation behavior by increasing I exc (Supplemental Material Fig. S3 [23]). It is important to note that this finding precludes electronic aggregation (i.e., H aggregation [27]) as
137402-2
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
FIG. 2 (color online). Folded and unfolded poly(3-hexylthiophene) (P3HT) chains. (a) Perfectly unfolded and unaggregated regioregular poly(3-hexylthiophene) (P3HT) chains (structure inset) only exist in solution. We obtained the average PL intensity by fluorescence correlation spectroscopy and plotted this value against M n . (b) Folded chains are formed, on the single-chain level, by dispersion in a PMMA matrix. PL saturation with M n is described by Eq. (4) (black curve). Error bars give the standarderror value.
the source of PL quenching in isolated chains because in this case, the PL dependency would still have to be linear. The quenching of hPLi that we observed with Mn is remarkable: as chain length increases, the brightness does not, which implies that self-quenching efficiency increases with chain length. Lin et al. [11] previously attributed such selfquenching to the formation of unspecified “dark matter,” in which parts of the folded chain do not participate in energy transfer or emission and are therefore practically “invisible”. In such a framework, excitation energy generated by absorption does not completely funnel into an emitting chromophore [11]. Consequently, τon , as defined by Eq. (1), would not depend on Mn . This situation can be tested experimentally. Before proceeding to test Eqs. (2) and (3) by measuring τon and τoff on folded chains, we quantified the number of independently emitting chromophores, N, with respect to Mn by examining statistics for fluorescence photons, as summarized in Fig. 3 [28]. Strong photon antibunching can occur in P3HT [26]: excitons are funneled to the acceptor chromophore via EET and quenched by singlet-singlet annihilation, so that only one exciton is emitted at a time [29]. We studied the samples using a scanning confocal fluorescence microscope with two photodetectors in an interferometric Hanbury Brown–Twiss geometric configuration. We recorded PL transients from single folded chains and measured the probability of a chain emitting more than a single photon at once by quantifying photon coincidence on the two detectors. We extracted the number of independently emitting chromophores, N, from this probability (see Fig. S4 [23] and Ref. [28] for details). The red dots in Fig. 3 show N dependent on Mn for folded P3HT chains. N
week ending 4 APRIL 2014
FIG. 3 (color online). The average number of independently emitting chromophores, N, in dependence of M n (red) for folded P3HT chains embedded in poly(methyl-methacrylate); see Supplemental Material [23] for explanation of error bars. The average transition energy of the single-chain PL peak, E0–0 , is shown in black. Folded chains display a constant E0–0 of approximately 1.95 eV for M n > 45 kDa. For shorter chains, the acceptor is not formed within all folded chains, as illustrated in the drawing. Each data point presents an average over approximately 100 single chains. Error bars represent the standard error of the distribution (see Supplemental Material [23] for histograms).
remaines close to unity for folded chains over the M n range studied: on average, only one emitting chromophore (acceptor) is active at a given time. The nature of this acceptor chromophore remains subject to debate [27,30] but is not strictly crucial to the present analysis. It was recently demonstrated [26] that intrachain photophysics dominates the spectral changes observed when transitioning from solution to bulk film rather than interchain interactions dominating, as previously surmised [27]. Although H aggregation [27] may play a role in some features of bulk photophysics, the dramatic redshift between solution and film (i.e., between solvated and collapsed chains) is primarily intrachromophoric in nature [26]. The relevant question is, therefore, what M n is necessary to form an acceptor chromophore in a folded chain morphology? We obtained approximately 100 single-molecule spectra for each Mn sample for folded chains (Fig. S5 [23]). The black dots in Fig. 3 show the average 0–0 transition energy, E0–0 , with respect to Mn for folded chains. The transitions lie at approximately 1.95 eV and shift to higher energies for the two lowest-M n samples. The spectral measurements indicate that the universal acceptor site is not always fully developed for M n of less than 45 kDa (dashed line in Fig. 3), although the shorter chains show photon antibunching. Both data sets, for photon statistics and PL peaks, respectively, provide a weight range (55–110 kDa) over which τon and τoff of the acceptor can, in principle, satisfy Eqs. (2) and (3). For M n of less than 45 kDa, the acceptor site is not always formed in the intrachain aggregate, possibly because chain folding is incomplete. Therefore, EET efficiency is reduced and direct excitation of the acceptor is enhanced. Even the chromophoric triplet rates kisc and kT may be different, which would alter τon and τoff values. However, if more than one acceptor is formed at higher M n, the excitation energy may be distributed
137402-3
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
sequentially between different chromophores within the molecular aggregate, which would increase the apparent τon for each acceptor. Next, we measured τon and τoff depending on M n . We analyzed the PL-intensity transients of single folded chains, as exemplified in Fig. 4(a), by passing the fluorescence through a 50∶50 beam splitter and computing a secondorder cross correlation of the two detector signals [31]. Subsequent fitting of the cross correlation, shown in Fig. 4(b), via a three-state (exponential) model allows the extraction of τon and τoff (see Supplemental Material Fig. S6 and Table S1 for details [23]) [32]. In our analysis, we only examined transients that exhibited single-step photobleaching (∼90% of all single chains). Figure 4(c) displays average τon and τoff values from approximately 100 PL transients that we obtained for each Mn sample. The plot reveals a 1=Mn dependence for τon between 55 and 110 kDa, in accordance with Eq. (2). However, τoff remains constant at approximately 18 μs. This “off” time is comparable to the tripletstate lifetime of P3HT found in deoxygenated environments [33,34]. The “off” state is completely quenched by oxygen (see Fig. S7 for details [23]); this phenomenon clearly indicates triplet shelving. For M n < 45 kDa (left side of the dashed line), τon deviates from the 1=Mn dependence
FIG. 4 (color online). (a) Typical PL transient for single selfaggregated P3HT chains embedded in PMMA, binned into 10-ms intervals. Single-step bleaching occurred after approximately 13 s. (b) Second-order cross correlation, g2 (τ), obtained from the PL transient in (a); the line describes a single-exponential fit, which reveals fast blinking kinetics on the μs time scale. (c) For each M n , we obtained transients of approximately 100 molecules and extracted the average on-times, τon (red dots), and off-times, τoff (black dots), from the averaged g2 (τ) (see Fig. S6 and Supplemental Material [23] for detailed analysis). Error bars correspond to fitting-error values. On times for the two lowestM n samples deviate from the 1=M n dependence because the acceptor is not fully formed at this chain length, as illustrated in the drawing (cf. also spectral redshift in Fig. 3).
week ending 4 APRIL 2014
because the acceptor is not fully formed in the intrachain self-aggregate, as illustrated in the inset (cf. PL spectra in Fig. 3). The clear dependence of τon on Mn supports the following conclusions: (i) the entire chain absorbs light (i.e., there is no “dark matter” [11]); (ii) EET occurs within the entire intrachain aggregate; (iii) the emitting chromophore can be shelved into the nonemissive triplet state; and (iv) subsequent STA at the acceptor site quenches PL from the entire chain. These factors lead to the saturation of hPLi with increasing Mn according to Eq. (4) for folded P3HT chains, as shown in Fig. 2(b). Interestingly, even P3HT chains of Mn ¼ 110 kDa behave as perfect single-photon emitters. We conclude that, for folded P3HT chains, the only spectroscopic observable that allows us to accurately report the Mn dependence is the excitation rate, which controls the average “on” time. Our results demonstrate the extraordinary lightharvesting abilities of single folded P3HT chains, which depend strongly on the degree of ordering of the individual chain. Such ordering is required to design efficient solar cells [21]. Despite this, order can also introduce losses through efficient STA, assuming that the dynamics of triplet formation observed on single P3HT chains in this study can be extrapolated to bulk heterojunction solar cells. Once a triplet is formed on the acceptor, additional singlets excited within the range of exciton migration are annihilated and cannot be converted to charges [35,36]. This intrinsic loss mechanism must contribute to the decrease of quantum efficiency in P3HT-based organic solar cells with increasing excitation intensities, which has previously only been attributed to singlet-polaron interactions [37]. The results of a simple estimate show that the magnitude of STA observed on single P3HT chains can account for the reported bottleneck in solar cells. As described in Fig. S8 [23], we examined the proportion of singlets lost due to STA at a given excitation intensity and singlet-triplet interaction distance, DST . For example, a loss of approximately 30% can be expected at intensities comparable to solar illumination for DST ¼ 30 nm. DST depends on singlet and triplet exciton diffusion lengths. The singlet diffusion length in P3HT is reported to be in the range of tens of nanometers [36]. Triplet exciton migration is harder to quantify but can occur on the micron scale in organic semiconductor crystals [38]. Even in a percolative donoracceptor blend with strong local order, DST ¼ 30 nm is likely an underestimate. Therefore, investigators should take STA into account as a critical loss mechanism in organic solar cells, although it is difficult to separate this factor from singlet-polaron quenching in the bulk [37]. Possible ways to adjust device design to prevent STA would be to include nonoxygenic triplet scavengers in the bulk heterojunction; selectively introduce disorder to lower DST ; or identify materials with weaker spin-orbit coupling and lower kisc than P3HT [33]. Organic solar cells bear strong similarity to photosynthesis in that energy conversion is
137402-4
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
fundamentally self-limiting [39], as observed in slowed plant growth under full-sun illumination. In conclusion, we present a comprehensive monomolecular photophysical picture of the model organic-solarcell compound P3HT, explaining the drop in fluorescence yield following from the transformation from unfolded to folded chains. This drop arises because light is absorbed over the entire polymer chain, followed by efficient energy funneling to a single emitting acceptor, which is subject to triplet blinking and PL saturation effects. The efficiency of plastic solar cells is therefore fundamentally limited by triplet accumulation. It will be intriguing to establish how well the model holds for ordered multimolecular interchain aggregates of similar M n values, which can be grown by solvent vapor annealing and can offer a bottom-up approach towards the bulk film [25,40]. We acknowledge the ERC for funding through the Starting Grant MolMesON (Grant No. 305020) and thank Dr. T. Adachi for stimulating discussions on our topic.
*
[1] [2] [3] [4] [5] [6]
[7] [8] [9] [10] [11]
[12] [13] [14] [15]
Corresponding author. [email protected]‑regensburg.de D. T. McQuade, A. E. Pullen, and T. M. Swager, Chem. Rev. 100, 2537 (2000). P. F. Barbara, A. J. Gesquiere, S.-J. Park, and Y. J. Lee, Acc. Chem. Res. 38, 602 (2005). J. M. Lupton, Adv. Mater. 22, 1689 (2010). D. A. VandenBout, W.-T. Yip, D.Hu, D.-K. Fu, T. M. Swager, P. F. Barbara, Science 277, 1074 (1997). D. H. Hu, J. Yu, K. Wong, B. Bagchi, P. J. Rossky, and P. F. Barbara, Nature (London) 405, 1030 (2000). B. Muls, H. Uji-i S. Melnikov, A. Moussa, W. Verheijen, J.-P. Soumillion, J. Josemon, K. Müllen, and J. Hofkens, Chem. Phys. Chem. 6, 2286 (2005). D. Wöll, E. Braeken, A. Deres, F. C. De Schryver, H. Uji-I, and J. Hofkens, Chem. Soc. Rev. 38, 313 (2009). F. Laquai, Y.-S. Park, J.-J. Kim, T. Basché, Macromol. Rapid Commun. 30, 1203 (2009). T. Basché, S. Kummer, and C. Bräuchle, Nature (London) 373, 132 (1995). T. L. Andrew and T. M. Swager, J. Polym. Sci. B 49, 476 (2011). H. Lin, Y. Tian, K. Zapadka, G. Persson, D. Thomsson, O. Mirzov, P.-O. Larsson, J. Widengren, and I. G. Scheblykin, Nano Lett. 9, 4456 (2009). Y. Ebihara and M. Vacha, J. Phys. Chem. B 112, 12575 (2008). S. Habuchi, S. Onda, and M. Vacha, Phys. Chem. Chem. Phys. 13, 1743 (2011). T. Adachi et al., J. Phys. Chem. Lett. 2, 1400 (2011). C. W. Hollars, S. M. Lane, and T. Huser, Chem. Phys. Lett. 370, 393 (2003).
week ending 4 APRIL 2014
[16] M. C. Traub, J. Vogelsang, K. N. Plunkett, C. Nuckolls, P. F. Barbara, and D. A. Vanden Bout, ACS Nano 6, 523 (2012). [17] H. Sirringhaus et al., Nature (London) 401, 685 (1999). [18] S. Cook, A. Furube, and R. Katoh, Energy Environ. Sci. 1, 294 (2008). [19] T. Q. Nguyen, V. Doan, and B. J. Schwartz, J. Chem. Phys. 110, 4068 (1999). [20] T. Adachi, G. Lakhwani, M. C. Traub, R. J. Ono, C. W. Bielawski, P. F. Barbara, and D. A. Vanden Bout, J. Phys. Chem. B 116, 9866 (2012). [21] Y. Kim et al., Nat. Mater. 5, 197 (2006). [22] J. Yu, R. Lammi, A. J. Gesquiere, and P. F. Barbara, J. Phys. Chem. B 109, 10025 (2005). [23] See Supplemental Material at http://link.aps.org/ supplemental/10.1103/PhysRevLett.112.137402 for descriptions of the methods and modeling. [24] K. Becker, G. Gaefke, J. Rolffs, S. Höger, and J. M. Lupton, Chem. Commun. (Cambridge) 46, 4686 (2010). [25] J. Vogelsang, T. Adachi, J. Brazard, D. A. Vanden Bout, and P. F. Barbara, Nat. Mater. 10, 942 (2011). [26] A. Thiessen, J. Vogelsang, T. Adachi, F. Steiner, D. Vanden Bout, and J. M. Lupton, Proc. Natl. Acad. Sci. U.S.A. 110, E3550 (2013). [27] J. Clark, C. Silva, R. H. Friend, and F. C. Spano, Phys. Rev. Lett. 98, 206406 (2007). [28] K. D. Weston, M. Dyck, P. Tinnefeld, C. Müller, D. P. Herten, and M. Sauer, Anal. Chem. 74, 5342 (2002). [29] C. G. Hübner, G. Zumofen, A. Renn, A. Herrmann, K. Müllen, and T. Basché, Phys. Rev. Lett. 91, 093903 (2003). [30] E. T. Niles, J. D. Roehling, H. Yamagata, A. J. Wise, F. C. Spano, A. J. Moulé, and J. K. Grey, J. Phys. Chem. Lett. 3, 259 (2012). [31] R. Verberk, A. M. van Oijen, and M. Orrit, Phys. Rev. B 66, 233202 (2002). [32] W. T. Yip, D. H. Hu, J. Yu, D. A. Vanden Bout, and P. F. Barbara, J. Phys. Chem. A 102, 7564 (1998). [33] B. Kraabel, D. Moses, and A. J. Heeger, J. Chem. Phys. 103, 5102 (1995). [34] R. E. Palacios and P. F. Barbara, J. Fluoresc. 17, 749 (2007). [35] J. C. Bolinger, M. C. Traub, T. Adachi, and P. F. Barbara, Science 331, 565 (2011). [36] P. E. Shaw, A. Ruseckas, and I. D. W. Samuel, Adv. Mater. 20, 3516 (2008). [37] L. Tzabari, V. Zayats, and N. Tessler, J. Appl. Phys. 114, 154514 (2013). [38] H. Najafov, B. Lee, Q. Zhou, L. C. Feldman, and V. Podzorov, Nat. Mater. 9, 938 (2010). [39] P. Müller, X.-P. Li, and K. K. Niyogi, Plant Physiol. 125, 1558 (2001). [40] X. N. Yang, J. Loos, S. C. Veenstra, W. J. H. Verhees, M. M. Wienk, J. M. Kroon, M. A. J. Michels, and R. A. J. Janssen, Nano Lett. 5, 579 (2005).
137402-5
PHYSICAL REVIEW LETTERS
PRL 112, 137402 (2014)
Singlet-Triplet Annihilation Limits Exciton Yield in Poly(3-Hexylthiophene) Florian Steiner, Jan Vogelsang,* and John M. Lupton Institut für Experimentelle und Angewandte Physik, Universität Regensburg, Universitätsstrasse 31, 93053 Regensburg, Germany (Received 21 August 2013; published 2 April 2014) Control of chain length and morphology in combination with single-molecule spectroscopy techniques provides a comprehensive photophysical picture of excited-state losses in the prototypical conjugated polymer poly(3-hexylthiophene) (P3HT). Our examination reveals a universal self-quenching mechanism, based on singlet-triplet exciton annihilation, which accounts for the dramatic loss in fluorescence quantum yield of a single P3HT chain between its solution (unfolded) and bulklike (folded) state. Triplet excitons fundamentally limit the fluorescence of organic photovoltaic materials, which impacts the conversion of singlet excitons to separated charge carriers, decreasing the efficiency of energy harvested at high excitation densities. Interexcitonic interactions are so effective that a single P3HT chain of order 100 kDa weight behaves like a 2-level system, exhibiting perfect photon antibunching. DOI: 10.1103/PhysRevLett.112.137402
PACS numbers: 78.55.Kz, 42.50.Ar, 42.70.Jk, 78.66.Qn
To unravel the photophysical characteristics of multichromophoric systems such as conjugated polymers (CPs) or light-harvesting complexes poses ongoing interdisciplinary challenges for chemists, physicists, materials scientists, and spectroscopists [1–3]. Some of the oldest and still crucial questions in CP photophysics include the following: (i) What chemical unit absorbs and emits light in a CP? (ii) Which processes occur between absorption and emission events? And (iii) what is the nature of the interplay between efficient excitation energy transfer (EET) among chromophores and nonradiative fluorescence decay (quenching)? The third question is particularly relevant in organic photovoltaics because long-range EET is required to optimize charge photogeneration, although it promotes exciton quenching by dark states such as long-lived triplets or radicals. Answering these questions, and specifically elucidating which physical mechanisms are responsible for fluorescence self-quenching, is a prerequisite to the fundamental understanding of CP photophysics. The heterogeneity of CPs with respect to their spectroscopic properties and morphological diversity has made singlemolecule spectroscopy (SMS) an indispensable tool [4–9]. However, important parameters such as chain shape or size are difficult to assess in detail solely via SMS because these characteristics can be convoluted with the photophysical observables themselves [10]. The formation, number, and interaction pathways of chromophores in CPs correlate with changes in emission spectra and the degree of selfquenching [11–16]. However, the microscopic physical mechanism underlying this self-quenching phenomenon remains under debate [11,13]. Here, we demonstrate that control of molecular weight (defined by the number-average molecular weight, M n ) and chain morphology, in combination with SMS techniques, lead to a universal picture of self-quenching in CPs. As a model system, we use the prototypical CP poly 0031-9007=14=112(13)=137402(5)
(3-hexylthiophene) [P3HT; Fig. 2(a)] which is employed widely in photovoltaics research [17]. In this material, the fluorescence quantum yield drops approximately 20-fold on transitioning from well-dissolved (unfolded) chains in solution to aggregated (folded) chains in bulk film [18]. By controlling chain morphological characteristics using solvent or matrix polarity in well-defined subpopulations of different Mn , fractionated via gel-permeation chromatography (GPC), we correlate microscopic photophysical observables with each other and with Mn . The increase in fluorescence self-quenching with increasing M n and morphological ordering can be rationalized by improved light-harvesting and singlet-triplet quenching. In other words, triplet excitons fundamentally limit the lifetime of singlet excitons; this process can reduce the efficiency of organic solar cells. The microscopic photophysical model of exciton selfquenching is formulated in Fig. 1. Panel (a), which shows the expected PL intensity versus M n for the two extreme morphologies: (i) unfolded, as it occurs for well-dissolved CPs in solution with few, if any, intrachain contacts [19]; and (ii) folded, as it arises for aggregated CPs in bulk film with strong intrachain contacts [20,21]. In case (i), the PL intensity should increase linearly with increasing Mn because intrachain interchromophoric interactions are negligible. However, in case (ii), we expect saturation for folded chains, provided that EET occurs to a single emitting acceptor chromophore (red). This acceptor ultimately enters a triplet state, so that singlet-triplet annihilation (STA) can quench the excitation energy of the complete CP chain. Panel (b) shows a simplified level scheme of the acceptor chromophore, which is excited at rate kexc from the singlet ground state S0 to the first excited state S1 . S1 decays (radiatively and nonradiatively) to S0 at rate kPL , or undergoes intersystem crossing into the triplet T 1 (rate kisc ). The acceptor chromophore emits photons; i.e., it is in an
137402-1
© 2014 American Physical Society
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
week ending 4 APRIL 2014
τon PLmax M n I exc −1 hPLi ∝ ∝ M n I exc 1 þ ; τon þ τoff kT
FIG. 1 (color online). (a) Photoluminescence (PL) intensity dependence on number-average molecular weight, M n , for unfolded and folded conjugated polymer (CP) chains. Efficient excitation energy transfer (EET) occurs for folded chains, such that only one chromophore in the entire chain (green) emits light (acceptor [red]). (b) Level scheme of the acceptor chromophore. (c) The molecule is subject to triplet blinking with constant “off” times, τoff , and Mn -dependent “on” times, τon , assuming that the acceptor chromophore is populated by EET from the entire absorbing chain (donor [green]).
“on” state for a characteristic duration τon as it cycles between S0 and S1 . Following intersystem crossing, the chromophore enters an “off” state for τoff , as illustrated in panel (b). Assuming that k−1 exc ≪ kPL þ kisc , τon can be described as follows [22]: τon ¼ ðkPL þ kisc Þ=ðkisc kexc Þ ∝ k−1 exc ;
(1)
where kexc is proportional to the excitation intensity, I exc , and the absorption cross section. We propose that the entire chain [green, Fig. 1(a)] absorbs light and transfers energy to the acceptor chromophore [red, Fig. 1(a)]. Therefore, kexc is, to first-order approximation, directly proportional to Mn ; Eq. (1) becomes τon ∝ ðMn I exc Þ−1 .
(2)
Through STA, which triggers quenching of the complete folded CP once the acceptor chromophore enters the T1 state, τoff becomes independent of M n : τoff ¼ ðkT Þ−1 ;
(3)
where kT represents the reverse intersystem crossing rate (the inverse triplet lifetime). Therefore, the average PL intensity of the single chain [marked red in Fig. 1(c)] is determined as follows:
(4)
where PLmax denotes the maximum PL intensity during τon , which is proportional to kexc and thus to M n and I exc . Equation (4) demonstrates the dependence of hPLi: it saturates with I exc or Mn, as shown schematically in the lower part of Fig. 1(a), provided that only a single accepting chromophore is present within the polymer. To test this simple model, we revisited the dependence of hPLi on Mn and I exc for unfolded and folded P3HT chains [11,13]. We fractionated P3HT by GPC into six different M n samples with a low polydispersity index, ranging from 19 to 110 kDa versus polystyrene standards (see Supplemental Material, Table S1 [23]). We stress that the actual Mn of P3HT may differ significantly by a constant factor from the values reported herein due to the different rigidity of P3HT compared with the polystyrene standards. This difference is irrelevant for our study since M n of the polymer and the standard are described by a power-law, allowing relative weight comparisons [24]. Perfectly unfolded isolated chains can only be formed in nonpolar solution. Therefore, we diluted each sample to concentrations of 10−10 to 10−11 M in toluene; we obtained hPLi via fluorescence correlation spectroscopy (FCS; see Supplemental Material, Fig. S1 and Table S1 for details [23]) [11]. Figure 2(a) shows single-chain hPLi dependent on Mn . The PL intensity increases linearly, as expected, for well-dissolved chains in solution. This increase is a necessary control to validate that the Mn values, as obtained via GPC, are not compromised due to interchain aggregation at higher M n [11,19]. However, the average PL intensity for folded chains cannot be measured in solution: inducing self-aggregation by raising solvent polarity will drive isolated single chains together, forming larger aggregates [25]. Instead, single chains are embedded in a polar host matrix, poly(methyl-methacrylate) (PMMA), where most chains fold to form highly-ordered structures (i.e., self-aggregate) but remain isolated within the matrix [26]. Because the molecules are stationary, we cannot use FCS, which is a diffusion-based technique. Instead, we acquired the hPLi intensity measurements of single P3HT chains by confocal scanning of several areas of 20 × 20 μm2 (see Supplemental Material Fig. S2 and Table S1 [23] for details on the experimental apparatus). We identified diffractionlimited PL spots of the single chains using an automated spot-finding algorithm. For each M n , we acquired the average PL intensity over approximately 200 chains. All measurements were carried out under nitrogen atmosphere. The average PL intensity of folded P3HT chains, as plotted in Fig. 2(b), saturates quickly with increasing M n and follows Eq. (4) [black curve in Fig. 2(b)]. We observe the same saturation behavior by increasing I exc (Supplemental Material Fig. S3 [23]). It is important to note that this finding precludes electronic aggregation (i.e., H aggregation [27]) as
137402-2
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
FIG. 2 (color online). Folded and unfolded poly(3-hexylthiophene) (P3HT) chains. (a) Perfectly unfolded and unaggregated regioregular poly(3-hexylthiophene) (P3HT) chains (structure inset) only exist in solution. We obtained the average PL intensity by fluorescence correlation spectroscopy and plotted this value against M n . (b) Folded chains are formed, on the single-chain level, by dispersion in a PMMA matrix. PL saturation with M n is described by Eq. (4) (black curve). Error bars give the standarderror value.
the source of PL quenching in isolated chains because in this case, the PL dependency would still have to be linear. The quenching of hPLi that we observed with Mn is remarkable: as chain length increases, the brightness does not, which implies that self-quenching efficiency increases with chain length. Lin et al. [11] previously attributed such selfquenching to the formation of unspecified “dark matter,” in which parts of the folded chain do not participate in energy transfer or emission and are therefore practically “invisible”. In such a framework, excitation energy generated by absorption does not completely funnel into an emitting chromophore [11]. Consequently, τon , as defined by Eq. (1), would not depend on Mn . This situation can be tested experimentally. Before proceeding to test Eqs. (2) and (3) by measuring τon and τoff on folded chains, we quantified the number of independently emitting chromophores, N, with respect to Mn by examining statistics for fluorescence photons, as summarized in Fig. 3 [28]. Strong photon antibunching can occur in P3HT [26]: excitons are funneled to the acceptor chromophore via EET and quenched by singlet-singlet annihilation, so that only one exciton is emitted at a time [29]. We studied the samples using a scanning confocal fluorescence microscope with two photodetectors in an interferometric Hanbury Brown–Twiss geometric configuration. We recorded PL transients from single folded chains and measured the probability of a chain emitting more than a single photon at once by quantifying photon coincidence on the two detectors. We extracted the number of independently emitting chromophores, N, from this probability (see Fig. S4 [23] and Ref. [28] for details). The red dots in Fig. 3 show N dependent on Mn for folded P3HT chains. N
week ending 4 APRIL 2014
FIG. 3 (color online). The average number of independently emitting chromophores, N, in dependence of M n (red) for folded P3HT chains embedded in poly(methyl-methacrylate); see Supplemental Material [23] for explanation of error bars. The average transition energy of the single-chain PL peak, E0–0 , is shown in black. Folded chains display a constant E0–0 of approximately 1.95 eV for M n > 45 kDa. For shorter chains, the acceptor is not formed within all folded chains, as illustrated in the drawing. Each data point presents an average over approximately 100 single chains. Error bars represent the standard error of the distribution (see Supplemental Material [23] for histograms).
remaines close to unity for folded chains over the M n range studied: on average, only one emitting chromophore (acceptor) is active at a given time. The nature of this acceptor chromophore remains subject to debate [27,30] but is not strictly crucial to the present analysis. It was recently demonstrated [26] that intrachain photophysics dominates the spectral changes observed when transitioning from solution to bulk film rather than interchain interactions dominating, as previously surmised [27]. Although H aggregation [27] may play a role in some features of bulk photophysics, the dramatic redshift between solution and film (i.e., between solvated and collapsed chains) is primarily intrachromophoric in nature [26]. The relevant question is, therefore, what M n is necessary to form an acceptor chromophore in a folded chain morphology? We obtained approximately 100 single-molecule spectra for each Mn sample for folded chains (Fig. S5 [23]). The black dots in Fig. 3 show the average 0–0 transition energy, E0–0 , with respect to Mn for folded chains. The transitions lie at approximately 1.95 eV and shift to higher energies for the two lowest-M n samples. The spectral measurements indicate that the universal acceptor site is not always fully developed for M n of less than 45 kDa (dashed line in Fig. 3), although the shorter chains show photon antibunching. Both data sets, for photon statistics and PL peaks, respectively, provide a weight range (55–110 kDa) over which τon and τoff of the acceptor can, in principle, satisfy Eqs. (2) and (3). For M n of less than 45 kDa, the acceptor site is not always formed in the intrachain aggregate, possibly because chain folding is incomplete. Therefore, EET efficiency is reduced and direct excitation of the acceptor is enhanced. Even the chromophoric triplet rates kisc and kT may be different, which would alter τon and τoff values. However, if more than one acceptor is formed at higher M n, the excitation energy may be distributed
137402-3
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
sequentially between different chromophores within the molecular aggregate, which would increase the apparent τon for each acceptor. Next, we measured τon and τoff depending on M n . We analyzed the PL-intensity transients of single folded chains, as exemplified in Fig. 4(a), by passing the fluorescence through a 50∶50 beam splitter and computing a secondorder cross correlation of the two detector signals [31]. Subsequent fitting of the cross correlation, shown in Fig. 4(b), via a three-state (exponential) model allows the extraction of τon and τoff (see Supplemental Material Fig. S6 and Table S1 for details [23]) [32]. In our analysis, we only examined transients that exhibited single-step photobleaching (∼90% of all single chains). Figure 4(c) displays average τon and τoff values from approximately 100 PL transients that we obtained for each Mn sample. The plot reveals a 1=Mn dependence for τon between 55 and 110 kDa, in accordance with Eq. (2). However, τoff remains constant at approximately 18 μs. This “off” time is comparable to the tripletstate lifetime of P3HT found in deoxygenated environments [33,34]. The “off” state is completely quenched by oxygen (see Fig. S7 for details [23]); this phenomenon clearly indicates triplet shelving. For M n < 45 kDa (left side of the dashed line), τon deviates from the 1=Mn dependence
FIG. 4 (color online). (a) Typical PL transient for single selfaggregated P3HT chains embedded in PMMA, binned into 10-ms intervals. Single-step bleaching occurred after approximately 13 s. (b) Second-order cross correlation, g2 (τ), obtained from the PL transient in (a); the line describes a single-exponential fit, which reveals fast blinking kinetics on the μs time scale. (c) For each M n , we obtained transients of approximately 100 molecules and extracted the average on-times, τon (red dots), and off-times, τoff (black dots), from the averaged g2 (τ) (see Fig. S6 and Supplemental Material [23] for detailed analysis). Error bars correspond to fitting-error values. On times for the two lowestM n samples deviate from the 1=M n dependence because the acceptor is not fully formed at this chain length, as illustrated in the drawing (cf. also spectral redshift in Fig. 3).
week ending 4 APRIL 2014
because the acceptor is not fully formed in the intrachain self-aggregate, as illustrated in the inset (cf. PL spectra in Fig. 3). The clear dependence of τon on Mn supports the following conclusions: (i) the entire chain absorbs light (i.e., there is no “dark matter” [11]); (ii) EET occurs within the entire intrachain aggregate; (iii) the emitting chromophore can be shelved into the nonemissive triplet state; and (iv) subsequent STA at the acceptor site quenches PL from the entire chain. These factors lead to the saturation of hPLi with increasing Mn according to Eq. (4) for folded P3HT chains, as shown in Fig. 2(b). Interestingly, even P3HT chains of Mn ¼ 110 kDa behave as perfect single-photon emitters. We conclude that, for folded P3HT chains, the only spectroscopic observable that allows us to accurately report the Mn dependence is the excitation rate, which controls the average “on” time. Our results demonstrate the extraordinary lightharvesting abilities of single folded P3HT chains, which depend strongly on the degree of ordering of the individual chain. Such ordering is required to design efficient solar cells [21]. Despite this, order can also introduce losses through efficient STA, assuming that the dynamics of triplet formation observed on single P3HT chains in this study can be extrapolated to bulk heterojunction solar cells. Once a triplet is formed on the acceptor, additional singlets excited within the range of exciton migration are annihilated and cannot be converted to charges [35,36]. This intrinsic loss mechanism must contribute to the decrease of quantum efficiency in P3HT-based organic solar cells with increasing excitation intensities, which has previously only been attributed to singlet-polaron interactions [37]. The results of a simple estimate show that the magnitude of STA observed on single P3HT chains can account for the reported bottleneck in solar cells. As described in Fig. S8 [23], we examined the proportion of singlets lost due to STA at a given excitation intensity and singlet-triplet interaction distance, DST . For example, a loss of approximately 30% can be expected at intensities comparable to solar illumination for DST ¼ 30 nm. DST depends on singlet and triplet exciton diffusion lengths. The singlet diffusion length in P3HT is reported to be in the range of tens of nanometers [36]. Triplet exciton migration is harder to quantify but can occur on the micron scale in organic semiconductor crystals [38]. Even in a percolative donoracceptor blend with strong local order, DST ¼ 30 nm is likely an underestimate. Therefore, investigators should take STA into account as a critical loss mechanism in organic solar cells, although it is difficult to separate this factor from singlet-polaron quenching in the bulk [37]. Possible ways to adjust device design to prevent STA would be to include nonoxygenic triplet scavengers in the bulk heterojunction; selectively introduce disorder to lower DST ; or identify materials with weaker spin-orbit coupling and lower kisc than P3HT [33]. Organic solar cells bear strong similarity to photosynthesis in that energy conversion is
137402-4
PRL 112, 137402 (2014)
PHYSICAL REVIEW LETTERS
fundamentally self-limiting [39], as observed in slowed plant growth under full-sun illumination. In conclusion, we present a comprehensive monomolecular photophysical picture of the model organic-solarcell compound P3HT, explaining the drop in fluorescence yield following from the transformation from unfolded to folded chains. This drop arises because light is absorbed over the entire polymer chain, followed by efficient energy funneling to a single emitting acceptor, which is subject to triplet blinking and PL saturation effects. The efficiency of plastic solar cells is therefore fundamentally limited by triplet accumulation. It will be intriguing to establish how well the model holds for ordered multimolecular interchain aggregates of similar M n values, which can be grown by solvent vapor annealing and can offer a bottom-up approach towards the bulk film [25,40]. We acknowledge the ERC for funding through the Starting Grant MolMesON (Grant No. 305020) and thank Dr. T. Adachi for stimulating discussions on our topic.
*
[1] [2] [3] [4] [5] [6]
[7] [8] [9] [10] [11]
[12] [13] [14] [15]
Corresponding author. [email protected]‑regensburg.de D. T. McQuade, A. E. Pullen, and T. M. Swager, Chem. Rev. 100, 2537 (2000). P. F. Barbara, A. J. Gesquiere, S.-J. Park, and Y. J. Lee, Acc. Chem. Res. 38, 602 (2005). J. M. Lupton, Adv. Mater. 22, 1689 (2010). D. A. VandenBout, W.-T. Yip, D.Hu, D.-K. Fu, T. M. Swager, P. F. Barbara, Science 277, 1074 (1997). D. H. Hu, J. Yu, K. Wong, B. Bagchi, P. J. Rossky, and P. F. Barbara, Nature (London) 405, 1030 (2000). B. Muls, H. Uji-i S. Melnikov, A. Moussa, W. Verheijen, J.-P. Soumillion, J. Josemon, K. Müllen, and J. Hofkens, Chem. Phys. Chem. 6, 2286 (2005). D. Wöll, E. Braeken, A. Deres, F. C. De Schryver, H. Uji-I, and J. Hofkens, Chem. Soc. Rev. 38, 313 (2009). F. Laquai, Y.-S. Park, J.-J. Kim, T. Basché, Macromol. Rapid Commun. 30, 1203 (2009). T. Basché, S. Kummer, and C. Bräuchle, Nature (London) 373, 132 (1995). T. L. Andrew and T. M. Swager, J. Polym. Sci. B 49, 476 (2011). H. Lin, Y. Tian, K. Zapadka, G. Persson, D. Thomsson, O. Mirzov, P.-O. Larsson, J. Widengren, and I. G. Scheblykin, Nano Lett. 9, 4456 (2009). Y. Ebihara and M. Vacha, J. Phys. Chem. B 112, 12575 (2008). S. Habuchi, S. Onda, and M. Vacha, Phys. Chem. Chem. Phys. 13, 1743 (2011). T. Adachi et al., J. Phys. Chem. Lett. 2, 1400 (2011). C. W. Hollars, S. M. Lane, and T. Huser, Chem. Phys. Lett. 370, 393 (2003).
week ending 4 APRIL 2014
[16] M. C. Traub, J. Vogelsang, K. N. Plunkett, C. Nuckolls, P. F. Barbara, and D. A. Vanden Bout, ACS Nano 6, 523 (2012). [17] H. Sirringhaus et al., Nature (London) 401, 685 (1999). [18] S. Cook, A. Furube, and R. Katoh, Energy Environ. Sci. 1, 294 (2008). [19] T. Q. Nguyen, V. Doan, and B. J. Schwartz, J. Chem. Phys. 110, 4068 (1999). [20] T. Adachi, G. Lakhwani, M. C. Traub, R. J. Ono, C. W. Bielawski, P. F. Barbara, and D. A. Vanden Bout, J. Phys. Chem. B 116, 9866 (2012). [21] Y. Kim et al., Nat. Mater. 5, 197 (2006). [22] J. Yu, R. Lammi, A. J. Gesquiere, and P. F. Barbara, J. Phys. Chem. B 109, 10025 (2005). [23] See Supplemental Material at http://link.aps.org/ supplemental/10.1103/PhysRevLett.112.137402 for descriptions of the methods and modeling. [24] K. Becker, G. Gaefke, J. Rolffs, S. Höger, and J. M. Lupton, Chem. Commun. (Cambridge) 46, 4686 (2010). [25] J. Vogelsang, T. Adachi, J. Brazard, D. A. Vanden Bout, and P. F. Barbara, Nat. Mater. 10, 942 (2011). [26] A. Thiessen, J. Vogelsang, T. Adachi, F. Steiner, D. Vanden Bout, and J. M. Lupton, Proc. Natl. Acad. Sci. U.S.A. 110, E3550 (2013). [27] J. Clark, C. Silva, R. H. Friend, and F. C. Spano, Phys. Rev. Lett. 98, 206406 (2007). [28] K. D. Weston, M. Dyck, P. Tinnefeld, C. Müller, D. P. Herten, and M. Sauer, Anal. Chem. 74, 5342 (2002). [29] C. G. Hübner, G. Zumofen, A. Renn, A. Herrmann, K. Müllen, and T. Basché, Phys. Rev. Lett. 91, 093903 (2003). [30] E. T. Niles, J. D. Roehling, H. Yamagata, A. J. Wise, F. C. Spano, A. J. Moulé, and J. K. Grey, J. Phys. Chem. Lett. 3, 259 (2012). [31] R. Verberk, A. M. van Oijen, and M. Orrit, Phys. Rev. B 66, 233202 (2002). [32] W. T. Yip, D. H. Hu, J. Yu, D. A. Vanden Bout, and P. F. Barbara, J. Phys. Chem. A 102, 7564 (1998). [33] B. Kraabel, D. Moses, and A. J. Heeger, J. Chem. Phys. 103, 5102 (1995). [34] R. E. Palacios and P. F. Barbara, J. Fluoresc. 17, 749 (2007). [35] J. C. Bolinger, M. C. Traub, T. Adachi, and P. F. Barbara, Science 331, 565 (2011). [36] P. E. Shaw, A. Ruseckas, and I. D. W. Samuel, Adv. Mater. 20, 3516 (2008). [37] L. Tzabari, V. Zayats, and N. Tessler, J. Appl. Phys. 114, 154514 (2013). [38] H. Najafov, B. Lee, Q. Zhou, L. C. Feldman, and V. Podzorov, Nat. Mater. 9, 938 (2010). [39] P. Müller, X.-P. Li, and K. K. Niyogi, Plant Physiol. 125, 1558 (2001). [40] X. N. Yang, J. Loos, S. C. Veenstra, W. J. H. Verhees, M. M. Wienk, J. M. Kroon, M. A. J. Michels, and R. A. J. Janssen, Nano Lett. 5, 579 (2005).
137402-5
