Article pubs.acs.org/JPCA
Unravelling the Role of the Central Metal Ion in the Electronic Structure of Tris(8-hydroxyquinoline) Metal Chelates: Photoemission Spectroscopy and Hybrid Functional Calculations F. Bisti,*,† A. Stroppa,‡ M. Donarelli,† G. Anemone,† F. Perrozzi,† S. Picozzi,‡ and L. Ottaviano†,‡ †
Dipartimento di Scienze Fisiche e Chimiche, Università dell’Aquila, Via Vetoio, 67100 L’Aquila, Italy CNR-SPIN L’Aquila, Via Vetoio 10, 67100 L’Aquila, Italy
‡
ABSTRACT: The electronic structures of tris(8-hydroxyquinolinato)−erbium(III) (ErQ3) and tris(8-hydroxyquinolinato)−aluminum(III) (AlQ3) have been studied by means of core level and valence band photoemission spectroscopy with the theoretical support of hybrid Heyd−Scuseria−Ernzerhof density functional theory, to investigate the role played by the central metal atom. A lower binding energy (0.2 eV and 0.3 eV, respectively) of the O 1s and N 1s core levels has been observed for ErQ3 with respect to AlQ3. Differences in the valence band spectra, mainly related to the first two peaks next to the highest occupied molecular orbital (HOMO), have been ascribed to an energetic shift (to 0.4 eV lower energies for ErQ3) of the σ molecular orbital between the oxygen atoms and the central metal atom. A lower (by 0.5 eV) ionization energy has been measured for the ErQ3. The interpretation of these results is based on a reduced interaction between the central metal atom and the ligands in ErQ3, with increased electronic charge around the ligands, due to the higher ionic radius and the lower electronegativity of Er with respect to Al.
■
INTRODUCTION
The different central atom is not expected to influence the ligands geometry; indeed by means of UV−visible and infrared absorption, it has been demonstrated that the quinoline geometry have a similar chemical behavior, considering different rare earth tris(8-hydroxyquinolines) and AlQ3.14 However, theoretical calculation demonstrated that the metal−ligand interaction shows an ionic rather than a covalent character, as proposed using energy partitioning analyses method (EPA), on different MQ3 systems (M = Al3+, Ga3+, In3+, Tl3+).15,16 Moreover, the electrostatic interaction increases when increasing the atomic radius.15,16 Therefore, the possible modifications in the electronic structure are likely related to electrostatic effects between the central metal atom and its first neighbors. In this work, by means of photoemission spectroscopy and theoretical calculations, we investigate the modification in the electronic structure of metal 8-hydroxyquinolines (treated in the calculations as isolated molecules) once the central metal ion is replaced with a different one. We report monochromatic X-ray source (1486.6 eV), He I (21.2 eV), and He II (40.8 eV) photoemission spectra of thermal evaporated ErQ3 and AlQ3 thin films. These spectra were decomposed using hybrid Heyd−Scuseria−Ernzerhof density functional theory (HSE DFT) calculations, following a procedure similar to the one used in the cases of AlQ310 and croconic acid.17 We note that,
Metal 8-hydroxyquinolines (MQs) form a class of organic molecules, composed by organic ligands encapsulating a metal ion. The most representative in this class is tris(8hydroxyquinolinato)−aluminum(III) (AlQ3) thanks to its green photoluminescence property (applied in the OLEDs1) and to its efficiency as a nonmagnetic conducting layer in spin valves.2 Another interesting molecule of this class is the tris(8hydroxyquinolinato)−erbium(III) (ErQ3) due to its ability to emit in the infrared range of interest for optical communications. Its applications are in the fields of infrared OLEDs3−6 and optical silica fiber amplifiers.7−9 Given such applications in organic devices, their electronic structures have been intensively investigated, not only for a fundamental interest but also for device engineering. We mention that we have recently revisited the AlQ3 electronic structure in ref 10. For ErQ3, however, few studies exist, in particular, a core level photoemission investigation with laboratory sources11 and valence band photoemission study with synchrotron light at 160 eV.12 Something that is missing in the literature is a detailed comparison on the electronic properties of two different MQ3 in order to show the role played in the electronic structure by varying the central metal atom. A study of this kind for ErQ3 and AlQ3 can furthermore have an important impact in OLED development. In fact, it has been observed that the use of ErQ3 as electron injection layer in the simple AlQ3/TPD devices enhances the electron injection and improves the efficiency.13 © 2012 American Chemical Society
Received: August 21, 2012 Revised: October 22, 2012 Published: October 29, 2012 11548
dx.doi.org/10.1021/jp308303d | J. Phys. Chem. A 2012, 116, 11548−11552
The Journal of Physical Chemistry A
Article
sampling, the Γ point only has been used. For the exchangecorrelation functional, we have used the Heyd−Scuseria− Ernzerhof hybrid functional (HSE).23−25 The spin−orbit interaction was explicitly considered in the calculations. The atomic positions were relaxed until the Hellman−Feynman forces were lower than 0.03 eV/Å. For comparison with experiments, the calculated density of states (DOS) was broadened using a Gaussian function with σ = 0.3 eV. For brevity, we reported only the DOS calculated for mer-ErQ3, considered more suitable for the comparison with the mer-AlQ3 that is the most stable isomer for the AlQ3. However, the differences between the two isomers are discussed in the next section. The DFT calculations of AlQ3 reported here for comparison are taken from ref 10.
to the best of our knowledge, no valence band spectra using helium discharge lamp source (at 21.2 eV or 40.8 eV, that is commonly used in the investigation of energy-level alignment at organic/metal interfaces18) have been reported so far, for ErQ3. The valence band investigation, made by Blyth et al.,12 was performed using photon energies higher than 160 eV because, at lower values, they detected sample damages (probably cross-linking or bonds breaking as due to intense radiation coming from synchrotron storage ring undulator in the UV range19). The use of lower photon energies can be an advantage for this system since it lowers, due to cross-section effects,12 the photoemission signal coming from the erbium 4f orbitals. These orbitals are not involved in the chemical bond,20 but overlap with the ligands valence band, making the electronic structure investigation more complex. In our calculation, we included the 4f electrons in the valence, and we confirmed their negligible role in the chemical bond. Furthermore, by the comparison with the valence band probed by laboratory X-ray source (1486.6 eV), we determined the energy position of these orbitals with respect to other molecular states. By an accurate comparison between the core level spectra, we find a lower binding energy for the N 1s and O 1s in the ErQ3 with respect to the AlQ3. Moreover, by comparing the UPS spectra of both molecules, we show a small but significant modification in the σ molecular orbital of oxygen that feels the attraction of the central ion. The modification in the core level and in the valence band are both related to the larger ionic radius and to the lower electronegativity of erbium. No significant modification in the ligands highest occupied molecular orbital (HOMO) was detected. Finally, the measured ionization energy for ErQ3 turns out to be lower than AlQ3 (by about 0.5 eV). In device application, this different value is expecting to increase the electron injection energy barrier into the ErQ3 layer from a metal cathode, giving an higher operating voltage for ErQ3 OLEDs than AlQ3 ones.
■
RESULTS AND DISCUSSION
Calculated total energies of the two ErQ3 isomers are very close (their difference is lower than 0.004 eV), and it is therefore not possible to unambiguously determine which one is the most stable. Moreover, the two isomer DOS shapes, once broadened for comparison with the photoemission spectra, do not show significant modifications. The only remarkable difference is in the estimate of the energy gap: 2.91 eV and 3.02 eV for merErQ3 and fac-ErQ3, respectively. This difference, of about 0.11 eV, is due to a rigid energy shifting of the three ligands' DOS due to symmetry lowering from the facial to meridian isomer, as already observed in the AlQ3.26,27 By assuming the DOS of the single ligands to be similar, they are expected to lie at the same energy in the facial isomer, whereas they are expected to be shifted in the meridian isomer, therefore resulting in the observed band gap change. When performing the same calculation for AlQ3, we obtained also a different energy gap for the two isomers, 3.07 eV for the facial isomer and 2.82 eV for the meridian one. In this latter case, the difference in energy gap between the two isomers is 0.25 eV, i.e., lager than what obtained for ErQ3 isomers (0.11 eV). Considering this lower difference in the band-gaps, and the fact that the ErQ3 isomers have a similar total energy, we can conclude that the relative ligands orientations influence the relevant properties far less for ErQ3 than for AlQ3. This result is in agreement with a larger ionic behavior for the bonds between the ligands and the erbium atom, with respect to a covalent one, in line with the increase of the atomic radius.15,16 In Figures 1 and 2, we report the core level spectra of 10 nm ErQ3/Au. As expected, the Er 4d (Figure 1) shows a spectral
■
EXPERIMENTAL SECTION ErQ3 (Sigma-Aldrich, 97%, 599.71 molecular weight) and AlQ3 (Sigma-Aldrich, 99.995%, 459.43 molecular weight) were deposited by vacuum thermal evaporation, from a quartz crucible, in an ultra high vacuum chamber (base pressure = 10−10 Torr) on a chemically cleaned Au polycrystal.21 The deposition rate was 0.1 Å/s (as monitored in situ by a Quartz Crystal Microbalance). The electronic structure of ErQ3 and AlQ3 films were studied by X-ray photoemission spectroscopy (XPS) (PHI 1257 spectrometer, monochromatic Al Kα source, hν = 1486.6 eV) and ultraviolet photoemission spectroscopy (UPS) (PHI 1257 spectrometer, He discharge lamp source, He I, hν = 21.2 eV, He II, hν = 40.8 eV). The core level spectra were acquired with a pass energy of 11.75 eV and a corresponding overall experimental resolution of 0.25 eV. The overall UPS spectral resolution was estimated to be 0.1 eV from the Fermi edge width, at room temperature, of a thermally cleaned gold substrate. After UPS data acquisition, XPS analysis was performed again to check, on the same samples, possible damages of the organic film. At variance from ref 12, no damages were observed, this is due to the lower intensity of our source with respect to the synchrotron one. DFT calculations were performed using the Vienna Ab-initio Simulation Package (VASP)22 considering an isolated molecule of two possible ErQ3 isomers: fac-ErQ3 (facial) and mer-ErQ3 (meridian). We chose a plane-wave cutoff of 400 eV. For the Brillouin zone
Figure 1. Er 4d core level of 10 nm ErQ3/Au: dots, experimental data; black line, background signal; red line, theoretical curve by Ogasawara et al.;28 green line, experimental data after Gaussian filtering and background subtracted. 11549
dx.doi.org/10.1021/jp308303d | J. Phys. Chem. A 2012, 116, 11548−11552
The Journal of Physical Chemistry A
Article
ligands valence bands and calculation removing this contribution. In Figure 3, we show the ErQ3 valence band (VB) spectra,
Figure 2. From top to bottom: O 1s, N 1s, and C 1s core levels of 10 nm ErQ3/Au (black) compared with the O 1s, N 1s, and C 1s of 10 nm AlQ3/Au (red).
Figure 3. Top: XPS valence band of 20 nm ErQ3/Au and UPS valence band of 10 nm ErQ3/Au, probed with different photon energies (crosses), compared with HSE DFT calculations (black line). Bottom: C (red), O (blue), H (green), N (magenta), Er (cyan), and Er 4f (dark cyan) projected density of the states. Vertical dark cyan lines are eyeguides for the Er 4f experimental peaks.
shape typical of erbium oxide (as first reported by Guelfi et al.29). The C 1s, N 1s, and O 1s core level spectra are related to the ligands atoms. Figure 2 shows a comparison between such core levels where the AlQ3 spectra were shifted to match the ErQ3 ones and the spectra are normalized to the unit at the maximum value. Considering their shapes, the ErQ3 C 1s and N 1s core level spectra are very similar to the corresponding ones of AlQ3. This points to a very similar electronic arrangement of the ligands. However, the ErQ3 O 1s core level is broader than the AlQ3 one. This is related to a broader oxygen−central atom distance distribution as a consequence of the, already discussed, larger ionic behavior in ErQ3. Indeed, being that the two ErQ3 isomers are energetically degenerate, both are expected to be present in the ErQ3 thin film and not only the meridian one as for the AlQ3 thin film. Apart from the core level shapes, other information can be derived from the binding energy positions of the peaks. Once aligned the XPS spectra of both molecules to have the C 1s centered at 286.1 eV (because the carbon atoms are not directly involved in a chemical bond with the central atom), the N 1s is located at 400.5 eV (400.8 eV, at +0.3 eV), and the O 1s is at 532.2 eV (532.4 eV, at +0.2 eV) for 10 nm ErQ3/Au (10 nm AlQ3/Au). These energy shifts are ascribed to an increased electronic charge surrounding the N and O atoms (that enhances the core hole screening and leads to a lower core level binding energy) in ErQ3 with respect to AlQ3. This is again explained by the larger ion radius and lower electronegativity of erbium with respect to aluminum. Both effects act in the same direction leaving more electronic charge to the ligands. A similar effect can be shown to occur also in the valence band analysis. As already discussed, the ErQ3 valence band is characterized by the erbium 4f orbitals overlapping the ligands molecular ones. The 4f orbitals are expected to be core-like because they are not involved in chemical bonding. For this reason, first of all, we will report how the calculations reproduce this trend; then, we will enter into a detailed comparison between the
probed by X-ray (1486.6 eV) and ultraviolet (40.8 and 21.2 eV) sources, compared with the HSE DFT calculations.30 The XPS VB spectrum is different from the UPS ones. This is essentially due to an enhancement of the 4f erbium orbitals (at 6−12 eV) cross-sections in the XPS VB data that dominate the signal coming from the ligands molecular orbitals. The spectrum is thus used for the energy assignment of the 4f orbitals. The calculated DOS is in good agreement with experimental UPS data, indicating that the ligands electronic structure is very well reproduced by the calculations.31 Going into the details of the erbium 4f states, in the 6−12 eV region of the XPS spectrum, a good energy position match is obtained for the lowest energy peak (at about 6.8 eV) of the Er 4f PDOS (see the top curve of the top panel and the bottom curve of the bottom panel). The energy match between theory (Er PDOS) and experiment (XPS) at higher binding energies is less accurate. However, this is an expected result due to photoemission final effects that are not taken into account in the calculations.32 Indeed, the photoemission signal coming from 4f orbitals results in a complex line shape related to the various possible final states in which the Erbium ion can relax after the emission of a 4f electron: it can relax in the final 4fN−1 ground state or in any possible 4fN−1 excited state. In the first case, the photoemission spectrum would reproduce the binding energy of the corresponding eigenvalues. In the other case, the emitted electron, leaving the system in an excited state, loses part of its kinetic energy because the system is left in an excited state, and accordingly, the electrons are more energetically bonded. For this reason, the photoemission spectrum coming from the 4f orbitals can be considered as composed by two parts that superimpose: the one with and the one without final effects. As demonstrated in ref 20, the lowest binding energy peak (at about 6.8 eV) of the 4f photoelectrons spectrum is related to the photoemission process without final effects. Therefore only this part of the spectrum can be compared with our 11550
dx.doi.org/10.1021/jp308303d | J. Phys. Chem. A 2012, 116, 11548−11552
The Journal of Physical Chemistry A
Article
doublets of O and N, having a σ character are more influenced by the central atom variation. Additionally, an important experimental finding of our works is the measure of the ionization energy for both molecules. We have obtained 5.2 eV ± 0.1 for ErQ3 and 5.7 eV ± 0.1 for AlQ3. Also in this case, the electrostatic character of the metal−ligand interaction gives a good rationale for the experimental data, with a lower ionization energy found for the molecule (ErQ3) where the central atom has the largest ionic radius and the lowest electronegativity.
calculations. The energy match between the theoretical (4f PDOS) and the experimental (4f photoelectrons) lowest energy peaks is a remarkable result. So far we have confirmed the absence of the erbium 4f electrons in the chemical bonds with the ligands and showed the calculations' accuracy. Now we focus the attention only on the ligands and on the remaining electrons around the erbium without the 4f electrons contribution. In Figure 4, we report the
■
CONCLUSIONS To summarize, by means of valence band X-ray and ultraviolet spectroscopy and taking advantage of hybrid functional DFT simulations, we gained a detailed understanding (with a remarkable match of the 4f orbitals energies without using adjustable parameters) of the influence of the different central atoms in the electronic structure of two MQs molecules (ErQ3 and AlQ3). The O 1s and N 1s core level binding energy of ErQ3 result in being at lower binding energy by 0.2 eV and 0.3 eV with respect to the AlQ3 case. In the ErQ3 valence band, we observed a shift to lower energy for the σ orbital in between the oxygen and the central atom. Finally, the ErQ3 ionization energy is 0.5 eV lower than the AlQ3 one. All the results can be well interpreted in terms of a lower electronegativity and higher distance from the ligands of erbium with respect to aluminum.
■
Figure 4. Top: black (red) bold line, He I valence band of 10 nm ErQ3/Au (10 nm AlQ3/Au); black (red) line HSE DFT calculation of ErQ3 (AlQ3). Bottom, black (red) lines: C, O, H, N and Er (Al) projected density of states of ErQ3 (AlQ3).
AUTHOR INFORMATION
Corresponding Author
*E-mail:
[email protected]. Notes
comparison between the valence band probed by 21.2 eV for the ErQ3 and AlQ3 with the corresponding calculated PDOS. The total ErQ3 DOS was obtained by the sum of the different atomic PDOS removing the 4f orbitals of the Er ion. The experimental spectra do not show large differences in number and energy position of the peaks. There is no energy shift of the first HOMO peak (at 3−4.5 eV), as instead reported by Blyth et al.12 (0.3 eV). The theoretical C and H PDOS are almost the same between the two organic molecules, pointing to a similar arrangement of the carbon in the ligands as already observed in the C 1s core level analysis. When looking at the PDOS of the central metal atoms, the Er PDOS is substantially like the Al PDOS after a rigid energy shift by 0.4 eV to lower binding energies. The fact that there is no substantial modification in the PDOS shape, but only a rigid energy shift, is strongly indicative of an ionic character of the interaction. The larger the ionic radius of the central atoms, the larger the shift to lower binding energy of its PDOS. Moreover, the lower electronegativity of the erbium atom also contributes in the same direction to this energetic shift. The variation of the central atom, influences the O and N PDOS. In particular, in the 4−6 eV energy range, the presence of Er instead of Al leads locally to a shift toward lower binding energies of the O and N PDOS (mainly moving the O PDOS from 5.2 eV in AlQ3 to 4.8 eV in ErQ3). This can be observed in the corresponding spectral energy windows of the experimental data (see Figure 4, top panel). Therefore, the different central atom does not modify the HOMO position, but the molecular orbital at higher energy. This is a reasonable result since the HOMO has a π character, and it is therefore perpendicular to the Er−O bond direction; however, the
The authors declare no competing financial interest.
■
ACKNOWLEDGMENTS F.B. thanks Sincrotrone Trieste S.C.p.A. for financial support of his Ph.D. fellowship. Funding from the European Research Council (Grant Agreement no. 203523) is also acknowledged. A.S. and S.P. thank M. Marsman for providing a suitable Er potential for the hybrid functional calculations.
■
REFERENCES
(1) Hung, L. S.; Chen, C. H. Mater. Sci. Eng. R 2002, 39, 143−222. (2) Naber, W. J. M.; Faez, S.; Wiel, W. G. v. d. J. Phys. D: Appl. Phys. 2007, 40, R205−R228. (3) Gillin, W. P.; Curry, R. J. Appl. Phys. Lett. 1999, 74, 798−799. (4) Curry, R. J.; Gillin, W. P. Appl. Phys. Lett. 1999, 75, 1380−1382. (5) Curry, R. J.; Gillin, W. P.; Knights, A. P.; Gwilliam, R. Appl. Phys. Lett. 2000, 77, 2271−2273. (6) Prezioso, S.; Ottaviano, L.; Bisti, F.; Donarelli, M.; Santucci, S.; Palladino, L.; Penna, S.; Reale, A. J. Lumin. 2011, 131, 682−685. (7) Kuriki, K.; Koike, Y.; Okamoto, Y. Chem. Rev. 2002, 102, 2347− 2356. (8) Slooff, L. H.; van Blaaderen, A.; Polman, A.; Hebbink, G. A.; Klink, S. I.; Van Veggel, F. C. J. M.; Reinhoudt, D. N.; Hofstraat, J. W. J. Appl. Phys. 2002, 91, 3955−3980. (9) Artizzu, F.; Marchiò, L.; Mercuri, M.; Pilia, L.; Serpe, A.; Quochi, F.; Orrù, R.; Cordella, F.; Saba, M.; Mura, A.; et al. Adv. Funct. Mater. 2007, 17, 2365−2376. (10) Bisti, F.; Stroppa, A.; Donarelli, M.; Picozzi, S.; Ottaviano, L. Phys. Rev. B 2011, 84, 195112/1−195112/5. (11) Blyth, R.; Thompson, J.; Zou, Y.; Fink, R.; Umbach, E.; Gigli, G.; Cingolani, R. Synth. Met. 2003, 139, 207−213. (12) Blyth, R.; Thompson, J.; Arima, V.; Zou, Y.; Fink, R.; Umbach, E.; Gigli, G.; Cingolani, R. Synth. Met. 2004, 142, 293−298. 11551
dx.doi.org/10.1021/jp308303d | J. Phys. Chem. A 2012, 116, 11548−11552
The Journal of Physical Chemistry A
Article
(13) Shakya, P.; Desai, P.; Curry, R. J.; Gillin, W. P. J. Phys. D: Appl. Phys. 2008, 41, 085108/1−085108/5. (14) Thompson, J.; Blyth, R. I. R.; Gigli, G.; Cingolani, R. Adv. Funct. Mater. 2004, 14, 979−984. (15) Zhang, J.; Frenking, G. Chem. Phys. Lett. 2004, 394, 120−125. (16) Zhang, J.; Wang, R. Synth. Met. 2005, 152, 241−244. (17) Bisti, F.; Stroppa, A.; Picozzi, S.; Ottaviano, L. J. Chem. Phys. 2011, 134, 174505/1−174505/5. (18) Braun, S.; Salaneck, W. R.; Fahlman, M. Adv. Mater. 2009, 21, 1450−1472. (19) Koch, N.; Pop, D.; Weber, R. L.; Böwering, N.; Winter, B.; Wick, M.; Leising, G.; Hertel, I. V.; Braun, W. Thin Solid Films 2001, 391, 81−87. (20) Thompson, J.; Arima, V.; Zou, Y.; Fink, R.; Umbach, E.; Cingolani, R.; Blyth, R. Phys. Rev. B 2004, 70, 153104/1−153104/3. (21) Gold coated silicon wafer, 99.999% (Au), layer thickness 1000 Å. A titanium adhesion layer (99.99%) is used in between the two layers to bind the gold to the silicon wafer (100). Gold orientation is nominally highly polycrystalline with a preference to (111) orientation. The gold surface is also chemically cleaned with base piranha, 3:1 mixture of ammonium hydroxide (NH4OH) with hydrogen peroxide, activated at 60 °C, for removing the carbon impurities. (22) Kresse, G.; Furthmüller, J. Phys. Rev. B 1996, 54, 11169−11186. (23) Heyd, J.; Scuseria, G. E.; Ernzerhof, M. J. Chem. Phys. 2003, 118, 8207−8215. (24) Heyd, J.; Scuseria, G. E.; Ernzerhof, M. J. Chem. Phys. 2006, 124, 219906. (25) Marsman, M.; Paier, J.; Stroppa, A.; Kresse, G. J. Phys.: Condens. Matter 2008, 20, 064201/1−064201/9. (26) DeMasi, A.; Piper, L. F. J.; Zhang, Y.; Reid, I.; Wang, S.; Smith, K. E.; Downes, J. E.; Peltekis, N.; McGuinness, C.; Matsuura, A. J. Chem. Phys. 2008, 129, 224705/1−224705/7. (27) Amati, M.; Lelj, F. J. Phys. Chem. A 2003, 107, 2560−2569. (28) Ogasawara, H.; Kotani, A.; Thole, B. Phys. Rev. B 1994, 50, 12332−12341. (29) Guerfi, N.; Tan, T. N.; Veuillen, J.; Lollman, D. Appl. Surf. Sci. 1992, 56, 501−506. (30) Because of the different photon energies used, the valence band electrons were extracted from different probing depths (about 10 nm for the XPS and about 3 nm for the UPS). For this reason, to avoid signal coming from the substrate, the XPS valence band spectrum reported is for thicker film (20 nm) than UPS case (10 nm). (31) The reported calculated DOS is composed by the occupied but also the unoccupied states. The peak near 0 eV is the LUMO, and its alignment at this particular energy is only accidental. In general, the HSE-DFT calculated HOMO−LUMO gap is in good agreement with the optical gap, that, due to the attractive interactions between the excited electron and hole, is lower than the fundamental gap. Therefore, the LUMO position of the molecule is expected to be correct at binding energy lower than 0 eV. (32) Gerken, F. J. Phys. F: Met. Phys. 1983, 13, 703−713.
11552
dx.doi.org/10.1021/jp308303d | J. Phys. Chem. A 2012, 116, 11548−11552
