Temperature-dependent singlet exciton fission observed in amorphous rubrene films

Organic Electronics 26 (2015) 213–217

Contents lists available at ScienceDirect

Organic Electronics journal homepage: www.elsevier.com/locate/orgel

Temperature-dependent singlet exciton fission observed in amorphous rubrene films Jing Li, Zhonghai Chen, Qiaoming Zhang, Zuhong Xiong ⇑, Yong Zhang ⇑ School of Physical Science and Technology, Southwest University, Chongqing 400715, PR China

a r t i c l e

i n f o

Article history: Received 25 April 2015 Received in revised form 7 July 2015 Accepted 15 July 2015

Keywords: Singlet fission Time-resolved fluorescence decay Magnetic field effect

a b s t r a c t The steady-state/transient fluorescence spectroscopy was used to demonstrate that the dynamics of singlet exciton fission in amorphous rubrene were temperature-dependent (50–300 K). Based on the traditional three-state model of singlet fission, time-resolved fluorescence decay curves measured at different temperatures could be well fitted by using a set of rate equations. The variations of specific rate constants were consistent with the conventional Arrhenius-type, thermally activated process. Additionally, the magnetic field effect of photoluminescence was apparently suppressed at low temperatures. All these findings offer clear evidence that the amorphous rubrene solid undergoes thermally activated singlet exciton fission due to the endothermic nature of fission process in rubrene. Ó 2015 Elsevier B.V. All rights reserved.

1. Introduction Singlet exciton fission observed in organic solid is a spin-allowed process in which an excited high energy singlet exciton (S1) shares its half of energy with another molecule at ground state (S0), and subsequently both of them convert into a pair of low energy triplet excitons (T1). In the past few years, singlet fission became a scientific hotspot in the field of physics, chemistry, and organic electronics (see Ref. [1,2] for recent reviews). Generally, singlet fission can be regarded as a carrier multiplication process in which two electron-hole pairs are created. If all of the charges can be efficiently collected, the photocurrent of device will be effectively enhanced. Therefore, it was suggested that organic molecules with fission property can be used as a new type of sensitizer to improve the quantum efficiency of organic photovoltaic devices [3–8]. Generally, the energies of S1 and T1 states must fulfill the requirement of E(S1)  2E(T1) for fission process to occur. For instance, E(S1)  2E(T1) = 0.11 eV in pentacene [9], E(S1)  2E(T1) = 0.18 eV in tetracene [10,11], and E(S1)  2E(T1) = 0.05 eV in rubrene [12]. Ultrafast fission processes have been observed in these materials. It was found by Thorsmølle et al. that there was no temperature dependence in the fission process of pentacene since E(S1) > 2E(T1) in pentacene [9]. However, Wilson et al. demonstrated that singlet fission in tetracene crystal was also independent of temperature ⇑ Corresponding authors. E-mail addresses: [email protected] (Z. Xiong), [email protected] (Y. Zhang). http://dx.doi.org/10.1016/j.orgel.2015.07.035 1566-1199/Ó 2015 Elsevier B.V. All rights reserved.

although E(S1) < 2E(T1) in tetracene [13]. The above two results constitute an important scientific puzzle that whether or not singlet fission should be temperature-independent for exothermic case and temperature-dependent if it is endothermic. Although researchers have made a lot of effort [9,14–16], more explicit experimental results are still required in order to elucidate this basic question. In this work, highly efficient fission material, i.e., 5,6,11,12-tetraphenyltetracene (rubrene) was deposited by using thermal evaporation, forming amorphous rubrene films. Steady state/transient photoluminescence (PL) and magnetophotoluminescence (MPL) were measured in a wide range of temperature. In both PL and MPL measurements, we found that singlet fission in amorphous rubrene was temperature-dependent. For instance, at temperature of 300 K, the transition rate in singlet fission was fitted to be 0.5 ns1. However, the fission rate was almost completely suppressed at temperatures lower than 50 K, confirming that singlet fission in amorphous rubrene was a thermally activated process. This is in line with the endothermic nature of singlet fission in rubrene material. 2. Experimental Rubrene films of 100 nm thick were thermally evaporated on the glass substrates at high vacuum about 1  106 Pa. It is generally found by different researchers that the obtained rubrene films usually exhibit amorphous morphology [16–18]. Park et al. observed the growth of island-like crystalline domains only after annealing the films at higher-than-room temperatures [18]. However, mixed amorphous/crystalline phases probably appear

214

J. Li et al. / Organic Electronics 26 (2015) 213–217

in rubrene films after preparation [19]. In order to eliminate the possible coexistence of amorphous and crystalline domains, the substrates were kept at room temperature during deposition process and the deposition rate was always controlled at 0.5 Å/s. The fresh samples were saved in a vacuum chamber (better than 1  102 Pa) and measured within 12 h. The PL spectra and time-resolved fluorescence decay were measured by using an Edinburgh FLS920 steady-state/transient fluorescence spectrometer with cryostat. Rubrene films were excited by monochromatic light (k = 355 nm) from a nanosecond flash lamp operated at 2 MHz. The minimal pulse width of excitation was about 800 ps. When measuring the MPL effect in steady-state, rubrene samples were mounted on the cold finger of another cryostat which was located between the pole pieces of an electromagnet [20]. Photoluminescence from rubrene films was excited by using a semiconductor laser with 405 nm wavelength. The output power of laser was tuned within 1–5 mW. And the focal size of laser spot on the sample was about 1.5 mm in diameter. During all the above measurements, the samples were held in vacuum which is better than 1  102 Pa. Under the above conditions, we were able to obtain consistent measurement results. Apparent fluctuations in the rates of singlet fission and amplitudes of MPL effect were not observed. 3. Results and discussions For a comparison with rubrene solid, the insert of Fig. 1 shows the spectra of another sample in which rubrene molecules was doped in N, N0 -Di(naphthalen-l-yl)-N, N0 diphenyl-benzidine (NPB) material. The doping concentration was about 3% by weight ratio. Supposing that the doped rubrene molecules were almost uniformly distributed in the NPB matrix, the averaged intermolecular distance between adjacent rubrene molecules could be estimated to be about 2.9 nm [20]. Although the height of fluorescence peak around 565 nm slightly rose when temperature was decreased, actually the area under each spectral line was almost constant. During measurement process, the intensity of light excitation was kept unchanged. In addition, no magnetic field effect of photoluminescence could be measured. This confirmed that fission process could not occur in this sample because of the large averaged intermolecular distance. Jankus et al. investigated the excited state dynamics in amorphous rubrene films [16]. They determined the fission rate to be

Fig. 1. Fluorescence spectra of rubrene films measured at different temperatures. The insert shows the spectra of a rubrene-doped sample (NPB: rubrene, 3% by weight ratio).

>2.5  1012 s1 and proposed that the temperature-dependent polaron pair formation from singlet excitons could compete with singlet fission. Similarly, if charge dissociation plays an important role in the decay of photoexcited singlet excitons, we should be able to observe this effect in NPB: rubrene (3%) composite film. Since HOMO (highest occupied molecular orbit) levels of NPB and rubrene are almost the same (5.4 eV), the photogenerated holes could migrate from rubrene molecules to NPB matrix. However, we found that the time-resolved fluorescence decays measured at different temperatures were almost the same in NPB: rubrene (3%) sample. This indicated that the thermally excited singlet dissociation process could actually be excluded in our analysis. The fluorescence spectra of amorphous rubrene film recorded at different temperatures are shown in Fig. 1. At 300 K, the maximum of spectrum around k = 565 nm was normalized to be 1. With decreasing temperature, the PL intensity became stronger and stronger. At 50 K, the maximum of spectrum at k = 577 nm substantially increased up to 18. In pure rubrene film, we can generally assume that singlet fission (rate constant can be expressed as kfiss) and singlet recombination (including radiative and other non-radiative recombination processes, total rate constant can be expressed as kS) are two dominant quenching pathways for photoinduced singlet excitons whereas singlet dissociation and intersystem crossing are negligible. Therefore, the PL intensity FPL in steady-state is usually proportional to the ratio of kS/(kS + kfiss). If the value of kS is unchanged, then the considerably increased PL intensity as shown in Fig. 1 could be mainly attributed to the suppression of kfiss at low temperatures [20]. This implied that singlet fission in amorphous rubrene could be a thermally activated process. In order to investigate the microscopic dynamics of singlet fission in amorphous rubrene, we measured the transient fluorescence decays at different temperatures. At each temperature, the peak wavelength in spectrum was located and its time-resolved fluorescence decay was taken. For NPB: rubrene (3%) composite film, all decay curves are almost identical, thus only the curve measured at 300 K is shown in Fig. 2. Within about 0–30 ns, it could be approximately regarded as a mono-exponential decrease. Hence, constant decay rate of singlet excitons (including radiative recombination and other temperature-independent non-radiative loss),

Fig. 2. Time-resolved fluorescence decays of rubrene films measured at different temperatures. Offsets are applied to avoid overlapping. All the curves were fitted by using coupled rate equations. The fluorescence decay of rubrene-doped sample is also shown. The insert shows the measured instrument response function (IRF).

J. Li et al. / Organic Electronics 26 (2015) 213–217

i.e., kS = 0.118 ns1, could be extracted from single exponential fitting. For pure rubrene films, all the curves recorded at different temperatures are also displayed in Fig. 2. It could be seen that the decrease of prompt fluorescence, especially at early times of 0–5 ns, became slower and slower due to the suppression of singlet fission at lower temperatures. At later times of t > 50 ns, the fluorescence decay was mainly composed of the long-lived delayed fluorescence due to triplet-triplet annihilation (also called triplet fusion). The insert of Fig. 2 displays the measured instrument response function (IRF) which has an exponential decay time of 0.22 ns. This response time should not be able to influence our measurement of transient fluorescence decay. Following previous researchers [13,19,21,22], we fitted the evolution of fluorescence signals by using a set of coupled rate equations which described the kinetics of both singlet fission and triplet fusion. As shown in Fig. 2, all simulated curves exhibit good agreement with experimental results. It’s well established that a simplified kinetic scheme (so called three-state mode) of singlet fission can be expressed as the following equation [1,2]: k2

k1

k2

k1

S1 þ S0 1 ðTTÞi T1 þ T1 ;

ð1Þ

where k1, k2, k1, k2 are the rate constants, and 1(TT)i state is an intermediate triplet pair state with singlet character. Previously, the pair state 1(TT)i was also named as dark state (D state) because it cannot be directly excited [23]. In order to track the rate constants of k1, k2, k1, and k2, the coupled rate equations was constructed as:

dNS ¼ ðkS þ k2 ÞNS þ k2 N D ; dt dND ¼ ðk2 þ k1 ÞND þ k2 NS þ k1 NT ; dt dNT ¼ ðkT þ k1 ÞN T þ k1 N D ; dt

ð2Þ

where NS, ND, and NT represent the populations of S1 states, 1(TT)i states (spin-coupled, intermediate triplet pairs), and 2T1 states (spin-decoupled, separated triplet pairs), respectively. In addition, kT represents the non-radiative decay rate of triplet excitons. The measured transient fluorescence signal FPL(t) is proportional to the number of NS (t). As shown in Fig. 2, single exponential fluorescence decay in the sample of NPB: rubrene (3%) enabled us to extract the temperature-independent rate constant of kS = 0.118 ns1. To achieve the value of kT, we then considered the later period (t > 50 ns) of the fluorescence decay. During this period, the decrease of fluorescence intensity was basically governed by the population of separated triplet pairs NT and rate constant of kT, i.e., FPL(t) / NT(t)exp(kTt). Thus, the values of kT = 0.021– 0.025 ns1 can also be obtained from mono-exponential fittings. After setting appropriate values for k1, k2, k1, k2 and NS(0), ND(0), NT(0), the transient fluorescence decays measured at different temperatures could be well fitted. All the fitting parameters for k1, k2, k1, k2, and kT are listed in Table 1. Table 1 Best-fit parameters used for simulated curves in Fig. 2. T (K)

k2 (ns1)

k2 (ns1)

k1 (ns1)

k1 (ns1)

kT (ns1)

300 250 200 150 100 50

0.541 0.432 0.322 0.162 0.052 0.005

0.031 0.033 0.039 0.048 0.08 0.12

0.138 0.137 0.131 0.122 0.09 0.04

0.011 0.011 0.012 0.013 0.014 0.025

0.025 0.024 0.023 0.022 0.021 0.019

215

In the following discussions, we give some analysis about the fission/fusion dynamics described by our rate equations. Some researchers considered the quenching of intermediate 1(TT)i state directly to ground state [19,21]. That would result into an additional decay term in the rate equation of ND. As treated differently in our simulation, the decay of 1(TT)i state to ground state was actually incorporated into the process of 1(TT)i ? 2T1 ? 2S0. Another important question is the possible evolution of newly created triplet pairs. We found that the best fitting could be obtained when only geminate triplets (generated from the same singlet) contributed to the fluorescence signal. This was consistent with some previous reports that non-geminate triplets did not play an important role in the fission/fusion dynamics in amorphous organic solids [16,19]. This differs considerably from previous observation of triplet dynamics in rubrene crystals [24–26], where triplet excitons can freely diffuse in crystal and collide with each other to form intermediate 1(TT)i states. In that case, the intensity of delayed fluorescence is proportional to n(t)2, where n(t) is triplet exciton density. However, due to the low charge mobility of amorphous rubrene, triplet excitons generated from singlet fission are trapped on two neighboring molecules and undergo fusion process without hopping through the film [16]. Hence, parameter NT(t) used in Eq. (2) labels the number of fission-generated triplet pairs (not individual triplets). The good agreement between simulated fluorescence decays and measured curves indicated that our rate equations appropriately described singlet fission process in amorphous rubrene. Additionally, we have to note the substantial difference between the rate constants obtained in amorphous films and those measured in crystals. It was reported that the lifetime of triplet excitons could be as long as about 20 ls in single crystals of rubrene [27]. Whereas in amorphous rubrene solid, the delayed fluorescence didn’t extend beyond the 1 ls time window which was the longest used in our experiment. Using femtosecond pump-probe method, Ma et al. determined that triplet states are generated by singlet fission from the relaxed S1 state (S1 ? 2T1) on 20 ps time scale [12]. By using nanosecond system, the rate constant k2 determined here is only 0.5 ns1. The above two results constitute another important scientific puzzle that why singlet fission in the amorphous films is much slower than that in the crystals. A reasonable explanation for this variation might be different charge mobilities between amorphous and crystalline rubrene, since both charge hopping and energy transfer are closely related with the conductivities of solid materials. The charge mobility could be larger than 10 cm2 V1 s1 in rubrene crystals [28]. However, it is about or even smaller than 103 cm2 V1 s1 in many amorphous organic solids [29,30]. Similarly, amorphous rubrene solid was observed to be very resistive in our previous experiments. According to the values of rate constants listed in Table 1, it’s common to treat k2 in terms of the thermal activation (Arrhenius) law which can be expressed as k2 = Aexp(E2/kBT), where kB is the Boltzmann constant and T is the absolute temperature [1,9,21,31]. Parameter A is the frequency factor which reflects the intrinsic rate of state transformation. The activation energy E2 can be determined from the temperature dependences of k2. Fig. 3 displays the Arrhenius plot of k2 as a function of 1/T. Within the temperature range of 100–300 K, the activation energy was fitted to be E2  31 meV. When temperature T < 100 K, the date point obviously deviated from the Arrhenius law. Such a phenomenon could be ascribed to the temperature-dependences of A or E2 in a much wider range of practical conditions [32]. In addition, we found k2 followed another relationship as k2 = Bexp(E2/kBT) within the temperature range of 100–300 K. As shown in Fig. 3, the character energy E2 was fitted to be E2  14 meV. The difference between E2 and E2 might imply that the transformation between

216

J. Li et al. / Organic Electronics 26 (2015) 213–217

Fig. 3. Arrhenius plots of rate constants k2 and k2 vs 1/T.

Fig. 4. Magnetic field dependence of time-resolved fluorescence decays of rubrene film measured at 285 K. The insert shows the steady-state magnetic field effects of photoluminescence measured at different temperatures.

S1 and 1(TT)i states depends on different phonon modes since phonon absorption and emission facilitate the electronic energy transfer. According to the detailed balance condition [33,34], the transfer rates of k2 and k2 should satisfy

k2 =k2 / expðDE=kB TÞ;

ð3Þ

where DE is the energy difference, i.e., DE = 2E(T1)  E(S1). By using femtosecond pump–probe spectroscopy, Ma et al. measured that DE = 50 meV in rubrene crystal [12]. In present work, the energy difference could be determined to be DE = E2 + E2  45 meV, in good agreement with the above experimental value. Except pure rubrene films, we also mixed rubrene molecules in NPB with different doping concentrations, i.e., 50%, 40%, and 30%. In rubrene-doped films, we still obtained similar values of E2, E2, and DE from temperature-dependences of k2 and k2. The same value of kS was used in all curve-fitting with Eq. (2). To date, many experimental studies on singlet exciton fission were performed by investigating its magnetic field effect [7,8,19,20,33,35]. Fig. 4 displays the measured and simulated fluorescence decay (from another sample, T = 285 K) with and without the external magnetic field. When a small B-field was applied, the decay rate of prompt fluorescence became apparently slower. As listed in Table 2, the value of k2 decreased from 0.526 ns1 (B = 0) down to 0.315 ns1 (B = 220 mT), whereas the value of k2 also slightly decreased from 0.036 ns1 (B = 0) down to 0.032 ns1 (B = 220 mT). The insert of Fig. 4 displays the B-field dependence of MPL effect recorded at multiple temperatures. At each temperature, the peak wavelength was located by using a monochromator and magnetic field effect of its fluorescence intensity was taken. At steady state, the MPL effect can be defined as

MPL ¼

F PL ðBÞ kS þ kfiss ð0Þ :  F PL ð0Þ kS þ kfiss ðBÞ

ð4Þ

In the common usage, kfiss represents the overall rate of transformation of [S1 + S0 ? T1 + T1] process, whereas k2 is only the transition rate of [S1S0 ? 1(TT)i] process. Generally, a pair of triplet excitons can combine into nine different (TT)i states (i = 1–9). Considering the singlet contributions from all nine triplet pairs, kfiss P is given by kfiss ¼ k2 9i¼1 jC iS j2 =ð1 þ k2 jC iS j2 =k1 Þ [1], in which jC iS j2 is the amplitude of singlet character of the ith intermediate (TT)i state. In accordance with Eq. (4), the disappearance of MPL effect (MPL ? 1) was mainly due to the decline of kfiss which resulted

Table 2 Best-fit parameters used for simulated curves in Fig. 4. B (mT)

k2 (ns1)

k2 (ns1)

k1 (ns1)

k1 (ns1)

kT (ns1)

0 220

0.526 0.315

0.036 0.032

0.122 0.124

0.013 0.011

0.027 0.029

from the suppression of k2 with decreasing temperature. It thus provides another key experimental result confirming the thermally activated singlet fission in amorphous rubrene. The first simplified form which rationally accounts for singlet fission process is the well-known Johnson–Merrifield (JM) theory [1,36]. According to the JM theory, all singlet characters satisfy P the closure relationship of 9i¼1 jC iS j2 ¼ 1, and fission rate kfiss is closely relevant with the distribution of singlet character jC iS j2 in all nine (TT)i states. In the absence of an external applied magnetic field (B = 0), three out of nine (TT)i states have nonzero singlet character. In the intermediate field range (B  25 mT in Fig. 4), due to the field-modulated interplay between Zeeman splitting and spin-spin interactions, more (TT)i states obtain some singlet character. This effect leads to the enhanced singlet fission, i.e., kfiss(B) > kfiss(0), and results in the initial decrease of PL intensity. Under a very high magnetic field (B ? 1), the singlet character of (TT)i states is redistributed and only two states possess nonzero singlet contributions. As a result, the fission rate will be effectively suppressed, i.e., kfiss(1) < kfiss(0), and saturated when Zeeman splitting greatly surpasses spin-spin interactions. As shown in Fig. 4, the slightly decreased MPL at small B-field thus becomes apparently increased and finally converges into a limit with increasing field strength. Since all rubrene films are amorphous, there was not field-orientation dependence in the measured MPL effect. 4. Summary In this work, we used steady-state/transient fluorescence spectroscopy to study the singlet exciton fission in amorphous rubrene film. Upon cooling from 300 K down to 50 K, the emission features continuously sharpened, leading to a substantial increase in the signal around 573 nm. The time-resolved fluorescence decay curves measured at different temperatures were fitted by using a

J. Li et al. / Organic Electronics 26 (2015) 213–217

set of rate equations which appropriately described the dynamics of three-state model of singlet fission. The obtained rate constants displayed conventional, Arrhenius-type behaviors in the temperature range of 100–300 K. Additionally, magnetic field effect of photoluminescence provided another evidence of the presence of singlet fission. The magnitude of this effect was apparently depended on temperature. All of these experimental results clearly confirmed that the singlet exciton fission in amorphous rubrene was a thermally activated process. Acknowledgement This work was supported by the National Natural Science Foundation of China (Grant Nos. 61475126, 11404266, and 11374242). References [1] M.B. Smith, J. Michl, Chem. Rev. 110 (2010) 6891. [2] M.B. Smith, J. Michl, Annu. Rev. Phys. Chem. Rev. 64 (2013) 361. [3] I. Paci, J.C. Johnson, X.D. Chen, G. Rana, D. Popovic, D.E. David, A.J. Nozik, M.A. Ratner, J. Michl, J. Am. Chem. Soc. 128 (2006) 16546. [4] J. Lee, P. Jadhav, M.A. Baldo, Appl. Phys. Lett. 95 (2009) 033301. [5] A. Rao, M.W.B. Wilson, J.M. Hodgkiss, S. Albert-Seifried, H. Bässler, R.H. Friend, J. Am. Chem. Soc. 132 (2010) 12698. [6] B. Ehrler, M.W.B. Wilson, A. Rao, R.H. Friend, N.C. Greenham, Nano Lett. 12 (2012) 1053. [7] P.D. Reusswig, D.N. Congreve, N.J. Thompson, M.A. Baldo, Appl. Phys. Lett. 101 (2012) 113304. [8] D.N. Congreve, J. Lee, N.J. Thompson, E. Hontz, S.R. Yost, P.D. Reusswig, M.E. Bahlke, S. Reineke, T. Van Voorhis, M.A. Baldo, Science 340 (2013) 334. [9] V.K. Thorsmølle, R.D. Averitt, J. Demsar, D.L. Smith, S. Tretiak, R.L. Martin, X. Chi, B.K. Crone, A.P. Ramirez, A.J. Taylor, Phys. Rev. Lett. 102 (2009) 017401. [10] Y. Tomkiewicz, R.P. Groff, P. Avakian, J. Chem. Phys. 54 (1971) 4504. [11] S. Arnold, W.B. Whitten, J. Chem. Phys. 75 (1981) 1166.

217

[12] L. Ma, K.K. Zhang, C. Kloc, H.D. Sun, M.E. Michel-Beyerlea, G.G. Gurzadyan, Phys. Chem. Chem. Phys. 14 (2012) 8307. [13] M.W.B. Wilson, A. Rao, K. Johnson, S. Gélinas, R. di Pietro, J. Clark, R.H. Friend, J. Am. Chem. Soc. 135 (2013) 16680. [14] J.J. Burdett, D. Gosztola, C.J. Bardeen, J. Chem. Phys. 135 (2011) 214508. [15] W.L. Chan, M. Ligges, X.Y. Zhu, Nat. Chem. 4 (2012) 840. [16] V. Jankus, E.W. Snedden, D.W. Bright, E. Arac, D.C. Dai, A.P. Monkman, Phys. Rev. B 87 (2013) 224202. [17] J.H. Seo, D.S. Park, S.W. Cho, C.Y. Kim, W.C. Jang, C.N. Whang, K.H. Yoo, G.S. Chang, T. Pedersen, A. Moewes, K.H. Chae, S.J. Cho, Appl. Phys. Lett. 89 (2006) 163505. [18] S.W. Park, J.M. Choi, K.H. Lee, H.W. Yeom, S. Im, Y.K. Lee, J. Phys. Chem. B 114 (2010) 5661. [19] G.B. Piland, J.J. Burdett, D. Kurunthu, C.J. Bardeen, J. Phys. Chem. C 117 (2013) 1224. [20] Y. Zhang, Y.L. Lei, Q.M. Zhang, Z.H. Xiong, Org. Electron. 15 (2014) 577. [21] A.M. Müller, Y.S. Avlasevich, W.W. Schoeller, K. Müllen, C.J. Bardeen, J. Am. Chem. Soc. 129 (2007) 14240. [22] S.T. Roberts, R. Eric McAnally, J.N. Mastron, D.H. Webber, M.T. Whited, R.L. Brutchey, M.E. Thompson, S.E. Bradforth, J. Am. Chem. Soc. 134 (2012) 6388. [23] P.M. Zimmerman, Z.Y. Zhang, C.B. Musgrave, Nat. Chem. 2 (2010) 648. [24] A. Ryasnyanskiy, I. Biaggio, Phys. Rev. B 84 (2011) 193203. [25] I. Biaggio, P. Irkhin, Appl. Phys. Lett. 103 (2013) 263301. [26] X.M. Wen, P. Yu, C.T. Yuan, X.Q. Ma, J. Tang, J. Phys. Chem. C 117 (2013) 17741. [27] H. Najafov, B. Lee, Q. Zhou, L.C. Feldman, V. Podzorov, Nat. Mater. 9 (2010) 938. [28] V.C. Sundar, J. Zaumseil, V. Podzorov, E. Menard, R.L. Willett, T. Someya, M.E. Gershenson, J.A. Rogers, Science 303 (2004) 1644. [29] K.K. Tsung, S.K. So, Appl. Phys. Lett. 92 (2008) 103315. [30] S. Khademi, J.Y. Song, P.B. Wyatt, T. Kreouzis, W.P. Gillin, Adv. Mater. 24 (2012) 2278. [31] L. Ma, K.K. Zhang, C. Kloc, H.D. Sun, C. Soci, M.E. Michel-Beyerle, G.G. Gurzadyan, Phys. Rev. B 87 (2013) 201203(R). [32] N.E. Henriksen, F.Y. Hansen, Theories of Molecular Reaction Dynamics, Oxford University Press, 2008. [33] R.P. Groff, P. Avakian, R.E. Merrifield, Phys. Rev. B 1 (1970) 815. [34] T.C. Berkelbach, M.S. Hybertsen, D.R. Reichman, J. Chem. Phys. 138 (2013) 114103. [35] V.V. Tarasov, G.E. Zoriniants, A.I. Shushin, M.M. Triebel, Chem. Phys. Lett. 267 (1997) 58. [36] R.E. Merrifield, J. Chem. Phys. 48 (1968) 4318.