Solvent dependence of two-photon absorption spectra of the enhanced green fluorescent protein (eGFP) chromophore

Article abstract of DOI:10.1016/j.cplett.2015.04.028

Two-photon absorption spectra of 4′-hydroxybenzylidene-2,3-dimethylimidazolinone, a model chromophore of enhanced green fluorescent protein (eGFP), were measured in various solvents. The two-photon absorption band of its anionic form is markedly blue-shifted from the corresponding one-photon absorption band in all solvents. Moreover, the magnitude of the blue shift varies largely depending on the solvent, which does not accord with the assignment of the two-photon absorption band to the transitions to the vibrationally excited S₁ state. Our finding is readily rationalized by considering overlapping contributions of the S₁ → S₀ and S₂ → S₀ transitions, suggesting the involvement of the S₂ state also in two-photon fluorescence of eGFP.

Full text of DOI:10.1016/j.cplett.2015.04.028

Chemical Physics Letters 630 (2015) 32–36
Contents lists available at ScienceDirect
Chemical Physics Letters
journal homepage: www.elsevier.com/locate/cplett
Solvent dependence of two-photon absorption spectra of the
enhanced green fluorescent protein (eGFP) chromophore
Haruko Hosoia,∗, Ryo Tayamaa, Satoshi Takeuchib,c, Tahei Taharab,c
a Department of Biomolecular Science, Faculty of Science, Toho University, 2-2-1 Miyama, Funabashi 274-8510, Japan
b Molecular Spectroscopy Laboratory, RIKEN, 2-1 Hirosawa, Wako 351-0198, Japan
c Ultrafast Spectroscopy Research Team, RIKEN Center for Advanced Photonics (RAP), RIKEN, 2-1 Hirosawa, Wako 351-0198, Japan
a r t i c l e i n f o
a b s t r a c t
Two-photon absorption spectra of 4ꢀ-hydroxybenzylidene-2,3-dimethylimidazolinone, a model chro-
mophore of enhanced green fluorescent protein (eGFP), were measured in various solvents. The
two-photon absorption band of its anionic form is markedly blue-shifted from the corresponding one-
photon absorption band in all solvents. Moreover, the magnitude of the blue shift varies largely depending
on the solvent, which does not accord with the assignment of the two-photon absorption band to the
transitions to the vibrationally excited S1 state. Our finding is readily rationalized by considering over-
lapping contributions of the S1 ← S0 and S2 ← S0 transitions, suggesting the involvement of the S2 state
also in two-photon fluorescence of eGFP.
Article history:
Received 2 March 2015
In final form 21 April 2015
Available online 30 April 2015
© 2015 Elsevier B.V. All rights reserved.
1. Introduction
in one-photon absorption (OPA) and two-photon absorption (TPA),
because the chromophore lacks inversion symmetry (Figure 1a,
Since the discovery of the green fluorescent protein (GFP)
from the jellyfish Aequorea Victoria, various fluorescent proteins
(FPs) have been developed as fluorescent markers and have been
extensively used in the fields of cell and molecular biology [1–3].
Among the imaging techniques utilizing FPs, two-photon exci-
tation fluorescence microscopy has substantial advantages over
one-photon excitation microscopy, such as deep penetration depth,
reduced background fluorescence, and reduced photodamage [4,5].
High-resolution microscopy of intact tissues has been achieved
with the two-photon excitation scheme, enabling visualization of
deep tissues such as lymphatic organs, kidney, heart, skin and
brain [6,7]. In spite of these wide applications, fundamental issues
regarding the fluorescence mechanism with two-photon excita-
tion of FPs have been left unresolved [8–13]. So far, two-photon
fluorescence excitation spectra of several FPs have been exam-
ined to obtain insights into the fluorescence mechanism [5,9,10,14].
Interestingly, the two-photon fluorescence excitation spectrum
of enhanced green fluorescent protein (eGFP, S65T/F64L) appears
blue-shifted compared with the one-photon fluorescence excita-
tion spectrum [5,9,10]. This discrepancy is unexpected given that all
relevant electronic transitions should have nonzero intensity both
right). Blab et al. attributed the blue shift to the transition from
the S0 state to the vibrationally excited S1 state (Figure 2a) [9].
However, the reported two-photon fluorescence excitation spectra
only had a limited number of data points, which made quantitative
discussion difficult. It was highly desirable to obtain high-quality
two-photon absorption spectra to further elucidate the mechanism
of the two-photon fluorescence of FPs.
Previously, we reported precise TPA spectra of eGFP using the
nondegenerate multiplex method [15]. We observed that the TPA
band is blue-shifted compared with the OPA band, and the mag-
nitude of the blue shift was determined as 1050 cm−1. In addition,
we also measured the TPA spectrum of a model chromophore of
eGFP, 4ꢀ-hydroxybenzylidene-2,3-dimethyl imidazolinone (HBDI,
Figure 1b), in methanol, and showed that the anionic form of HBDI
exhibits a similar blue shift. These results indicate that the differ-
ence between the OPA and TPA spectra arises from the intrinsic
electronic structure of the chromophore itself. Interestingly, the
blue shift of 670 cm−1 for HBDI is markedly different from that of
1050 cm−1 for eGFP. If the blue shift of TPA is due to the transition
from the S0 state to the vibrationally excited S1 state as claimed
[9], the difference in the blue shift (1050 cm−1 vs. 670 cm−1) should
correspond to the difference in the vibrational energy. However, it
is very unlikely that the vibrational energy varies so greatly with
the change in the environment (i.e., in protein and in solution).
Therefore, the strong environment dependence of the magnitude
of the blue shift suggests that the assignment of the blue shift to
∗
Corresponding author at: Department of Biomolecular Science, Faculty of Sci-
ence, Toho University, 2-2-1 Miyama, Funabashi 274-8510, Japan.
E-mail address: haru@biomol.sci.toho-u.ac.jp (H. Hosoi).
http://dx.doi.org/10.1016/j.cplett.2015.04.028
0009-2614/© 2015 Elsevier B.V. All rights reserved.

H. Hosoi et al. / Chemical Physics Letters 630 (2015) 32–36
33
Figure 1. Chemical structures and the acid–base equilibrium for (a) the chromophore of eGFP and (b) HBDI.
the vibrational energy is inappropriate. To explain our observation,
2. Experimental
we proposed the involvement of another electronic excited state,
the S2 state, which has larger two-photon absorptivity and smaller
one-photon absorptivity than the S1 state (Figure 2b) [15]. In this
mechanism, the TPA band consists of the smaller S1 ← S0 and the
larger S2 ← S0 contributions, making it appear blue-shifted com-
pared with the OPA band. The energy difference between the peak
maxima of the OPA and TPA bands reflects the energy difference
between the S1 and S2 states.
After our study, high-quality two-photon fluorescence excita-
tion spectra of eGFP were measured by several groups, and the
similar blue shift was reported [12,13,16]. However, Drobizhev
et al. associated the blue shift with the transition to the vibrationally
excited S1 state and did not consider the involvement of the S2 state
[12,13,17]. Thus, at the moment, two different mechanisms have
been proposed for the same observation. To consider which mech-
anism is more plausible, the solvent dependence of the OPA and
TPA spectra of HBDI is very informative. In the case that the TPA
arises from the vibrationally excited S1 ← S0 transition, the magni-
tude of the blue shift should be almost the same even if the solvent is
changed, because the solvent dependence of the vibrational energy
is very small in general.
The experimental setup for the TPA measurements has been
described elsewhere [15]. Briefly, a narrow-band laser pulse of
frequency ω1 in the near-infrared region and a broadband white-
light continuum pulse of frequency ω2 in the 450–750 nm region
were used to irradiate the sample, and the TPA spectra in the
ω1 + ω2 wavelength range were measured simultaneously using a
multichannel detector. Two different ω1 wavelengths were cho-
sen among the following wavelengths: 1300, 1400, 1450, 1750,
and 1751 nm. The spectra measured with the different ω1 wave-
lengths were well overlapped when plotted against the ω1 + ω2
wavelength, which ensures that the observed spectra are due to
the TPA process (Supplementary Figure S1). The polarizations of
the ω1, ω2, and ω1 + ω2 lights were parallel to one another. The OPA
spectra were measured using a commercial spectrometer (U-3310,
Hitachi).
HBDI was synthesized according to the procedure reported in
the literature [18]. The HBDI concentration was 9.53–37.0 ␮M for
the OPA measurements and 5.1–30 mM for the TPA measurements.
KOH aqueous solution was added to produce the anionic form. In
acetonitrile and ethyl acetate solutions, 18-crown-6 was further
added to assist the production of the anionic form. (Note that
18-crown-6 neither change the solvent characters nor interact
with HBDI [19]). The concentrations of HBDI, KOH, and 18-crown-6
in each solution are summarized in the Supplementary Tables S1
and S2.
In this work, we measured the TPA spectra of HBDI in various
organic solvents, focusing on the energy difference between the
OPA and TPA bands. On the basis of the obtained results, we discuss
the possibility of the transition to the vibrationally excited S1 state
and the plausibility of the involvement of the S2 state in TPA of the
anionic form of HBDI.
3. Results and discussion
The OPA spectra of HBDI in various solvents are shown with
black lines in Figure 3. The S1 ← S0 absorption peak maxima of the
neutral form vary only slightly: 370 nm (methanol), 369 nm (ethyl
acetate), 367 nm (acetonitrile), 374 nm (DMF), and 376 nm (DMSO)
(Figure 3a), whereas those of the anionic form change substan-
tially, that is, 432 nm (methanol), 459 nm (ethyl acetate), 448 nm
(acetonitrile), 465 nm (DMF), and 478 nm (DMSO) (Figure 3b). The
different solvent dependence between the neutral and anionic
forms has been reported previously by Tolbert and coworkers,
who measured OPA spectra of HBDI in 16 different solvents [19].
They analyzed the solvatochromic behavior quantitatively, and
concluded that the strong solvent dependence of the anionic form
is due to its larger permanent dipole-moment difference between
the S0 and S1 states compared with that of the neutral form.
Figure 2. Two different mechanisms for the blue shift observed in the TPA spectra
of the anionic form of HBDI and eGFP. (a) Transition to the vibrationally excited S1
state. (b) Transition to another electronic excited state. The S2 state exists in the
vicinity of the S1 state, and exhibits a larger two-photon absorptivity and smaller
one-photon absorptivity.

34
H. Hosoi et al. / Chemical Physics Letters 630 (2015) 32–36
Figure 3. One-photon (black) and two-photon (red) absorption spectra of the (a) neutral and (b) anionic forms of HBDI in methanol, ethyl acetate, acetonitrile (AN),
N,N-dimethylformamide (DMF), and dimethyl sulfoxide (DMSO). The values in green denote the energy differences between the OPA and TPA peaks in units of wavenumbers.
Table 1
The TPA spectra are displayed with red lines in Figure 3. As
The OPA and TPA absorption peak maxima of the neutral and anionic forms, ꢁmax
and the energy differences between the OPA and TPA peak maxima of the anionic
form.
,
shown in Figure 3a, the OPA and TPA spectra of the neutral form
are almost identical in each solvent within the experimental error.
The coincidence between the OPA and TPA bands indicates that
the neutral form is excited to the same S1 state by either one-
photon or two-photon excitation. In contrast, the TPA peaks of
the anionic form of HBDI are significantly shifted to higher energy
compared with the OPA peaks, not only in the case of HBDI in
methanol reported before [15] but also in all solvents examined
in this work: 420 nm (methanol), 437 nm (ethyl acetate), 435 nm
(acetonitrile), 451 nm (DMF), and 456 nm (DMSO) (Figure 3b).
Furthermore, the energy differences between the OPA and TPA
peak maxima are largely dependent on the solvent: 670 cm−1
(methanol), 1100 cm−1 (ethyl acetate), 670 cm−1 (acetonitrile),
670 cm−1 (DMF), and 1000 cm−1 (DMSO). The OPA and TPA absorp-
tion peak maxima of the neutral and anionic forms, and the energy
differences between the OPA and TPA peak maxima of the anionic
form are summarized in Table 1.
Solvent
Neutral
Anion
ꢁmax (nm)
ꢁmax (nm)
Energy difference
(cm−1
)
OPA
TPA
OPA
TPA
Methanol
Ethyl acetate
Acetonitrile
DMF
370
369
367
374
376
369
365
360
368
374
432
459
448
465
478
420
437
435
451
456
670
1100
670
670
1000
DMSO
respectively, and |vS ꢂ and |vS1 ꢂ are the vibrational wavefunctions
0
of the S0 and S1 states, respectively. For the dipole-allowed S1 ← S0
transition, as in the case of HBDI, the OPA intensity is dominated
by the term that involves the transition dipole moment at the equi-
librium nuclear coordinate Q0, ꢀ(Q0), and it is simply given by
In the following, we consider the band shape (vibronic structure)
of the OPA and TPA spectra using the vibronic theory. Particularly,
we attempt to explain the relative blue shift of the TPA spectrum by
considering only the vibronic transition to the S1 state (Figure 2a),
and show that the large solvent dependence of the magnitude of
the blue shift cannot be rationalized.
ꢁS1|ꢀ(Q0)|S0ꢂꢁvS |vS0 ꢂ. This factor indicates that the vibronic struc-
1
ture of the OPA spectrum is determined by the Franck-Condon
factor, ꢁvS |vS ꢂ [20]. Similarly, the S1 ← S0 TPA intensity is deter-
0
mined by1the matrix element of the TPA tensor S(Q) between the
S0 and S1 states, ꢁS1|ꢁvS |S(Q)|vS ꢂ|S0ꢂ. If the electronic part of the
1
0
In general, the S1 ← S0 OPA intensity is determined by the matrix
TPA tensor does not depend on Q, the TPA intensity is simply given
by the term, ꢁS1|S(Q0)|S0ꢂꢁvS |vS0 ꢂ. Thus, the vibronic structure of
the TPA spectrum is also det1ermined by the Franck-Condon factor
ꢁvS |vS ꢂ [21,22]. As a result, the vibronic structure of the S1 ← S0
element of the transition dipole moment ꢀ(Q) at the nuclear coor-
dinates Q between the S0 and S1 states, ꢁS1|ꢁvS |ꢀ(Q)|vS ꢂ|S0ꢂ, where
0
|S0ꢂ and |S1ꢂ are the electronic wavefunctions1of the S0 and S1 states,
1
0

H. Hosoi et al. / Chemical Physics Letters 630 (2015) 32–36
35
absorption in both the OPA and TPA spectra are identical, which
does not give any blue shift of the TPA spectrum. This argument
does not accord with the observed differences in the OPA and TPA
spectra.
and the TPA peak maximum predominantly reflects the S2 ← S0
transition. This makes the TPA band appear shifted to the shorter-
wavelength side compared with the OPA band. As is well-known,
the electronic states having different characters gain substantially
different stabilization energy in different solvents. We consider that
solvent dependence of the peak energy difference observed in the
present study arises from difference in solvatochromic shift of the
S1 ← S0 and S2 ← S0 transitions.
The blue shift of the TPA spectrum can be theoretically explained
by considering the Q-dependence of the TPA tensor. Particularly, for
non-centrosymmetric molecules like HBDI, the TPA tensor can be
approximated as S(Q) ≈ ꢀ(Q) × ꢂꢀ, where ꢂꢀ is the difference
in the permanent dipole moments of the S0 and S1 states [23].
Recently, Drobizhev et al. reported a physical model in which the
series expansion of ꢂꢀ(Q), to the linear term with respect to Q,
gives rise to the Q-independent (Franck-Condon) and Q-dependent
(Herzberg-Teller) terms and claimed that the both terms contribute
together to the band shape of the TPA spectrum [17]. The Q-
independent term gives the vibronic structure determined by the
Franck-Condon factor, ꢁvS |vS0 ꢂ, while the Q-dependent term gives
It should be mentioned that Olsen et al. performed a SA3-CAS
calculation and reported the existence of a dark excited state of
the anionic form of 4ꢀ-hydroxybenzylidene imidazolinone (HBI), a
simpler model compound of the GFP chromophore [27]. Specifi-
cally, they predicted two excited states: a lower-energy bright (S1)
state and a higher-energy dark (S2) state. Although the calculated
S2 ← S0 transition energy was notably higher than that observed in
the present study, the dominant electronic character of the S2 state
was found to be a doubly excited configuration. The doubly excited
configuration is characterized by one-photon forbidden and two-
photon allowed nature, which matches the character of the S2 state
suggested by the present study.
Our conclusion for HBDI is also applicable to eGFP. We thus
propose the mechanism for two-photon fluorescence of eGFP as
follows. When the anionic form of eGFP is two-photon excited, the
S2 state is initially populated, after which immediate relaxation to
the S1 state occurs. The S1 state then exhibits fluorescence identical
to that observed by one-photon excitation. We note that differ-
ent one- and two-photon fluorescence excitation spectra have also
been reported for the anionic forms of orange and red FPs [13].
However, the results of these previous studies do not necessar-
ily suggest that the corresponding S2 state exists for these FPs.
The chromophores of these FPs are different from HBDI, and hence
the mechanism would be different for each FP. Further TPA mea-
surements of various FPs and corresponding model chromophores
are desirable to fully understand the relevant electronic structure
of the excited states as well as the fluorescence mechanism after
two-photon excitation.
a blue-shifted componen1t originating from the ꢁvS |Q|vS0 ꢂ factor.
1
In other words, the band shape of the TPA spectrum can be rep-
resented by a sum of two contributions: one component that has
the same vibronic structure as the OPA spectrum and the other
component that is blue-shifted from the OPA spectrum by the vibra-
tional energy of the relevant mode. Thus, in their model, the energy
difference between the OPA and TPA peak maxima reflects the
vibrational energy(ies) of the relevant mode(s) (Q). It should be
noted that, since the transition dipole moment ꢀ(Q) is commonly
involved in both the OPA and TPA processes, its Q-dependence is
not expected to give a significant blue shift.
Based on their model, the energy difference between the OPA
and TPA peak maxima should be almost the same in all the sol-
vents, because the solvent dependence of the vibrational energy
is very small in general. For example, the frequency of the most
intense Raman band of the anionic form of HBDI (the C N stretch-
ing mode of the imidazolinone ring) is observed at 1550 cm−1
,
1555 cm−1, 1556 cm−1, 1556 cm−1, and 1556 cm−1 in DMSO, 2-
propanol, methanol, ethanol, and water, respectively [24–26].
However, as clearly observed in the present study, the peak energy
differences between the OPA and TPA spectra drastically change
from 670 cm−1 to 1100 cm−1 by solvent. This observation is not
consistent with the argument that the blue shift originates from
4. Conclusion
the ꢁvS |Q|vS0 ꢂ factor. We also confirmed that the experimental
1
We measured the TPA spectra of HBDI in various organic sol-
vents. The TPA bands of the anionic form of HBDI were shifted to
higher energies compared with the OPA bands in all of the investi-
gated solvents. Moreover, the energy difference between the OPA
and TPA bands changed substantially from 670 cm−1 in methanol
to 1100 cm−1 in ethyl acetate. We discussed the origin of the blue
shift of the TPA band on the basis of the strong solvent dependence
and concluded that it is not due to a transition to the vibrationally
excited S1 state. The difference between the OPA and TPA spectra
very likely arises from the participation of the S2 state. It was con-
sidered that the stronger S2 ← S0 and the weaker S1 ← S0 transition
bands are indistinguishably overlapped in the TPA spectrum. This
conclusion is also applicable to eGFP and suggests the involvement
of the S2 state in the two-photon fluorescence of eGFP.
blue shift cannot be accounted for by changing the relative ampli-
tudes of the Q-dependent and Q-independent terms while keeping
the vibrational energy constant: First, we calculated a sum of the
experimental OPA spectrum and another spectrum that is obtained
by displacing the experimental OPA spectrum toward the high fre-
quency side, and then fitted the sum to the TPA spectrum measured
in methanol. The vibrational energy determined in this fitting is
800 cm−1 for the methanol solution. Next, we attempted to repro-
duce the TPA spectrum measured in ethyl acetate by the same
procedure with the common 800 cm−1 displacement. However,
we failed to reproduce it even though we thoroughly changed the
relative amplitude of the two terms (Supplementary Figure S2).
Therefore, the strong solvent dependence of the magnitude of the
blue shift is inconsistent with the argument that the TPA blue shift
results from the ꢁvS |Q|vS ꢂ factor, and hence opposes the assign-
1
ment that TPA arises from0the transition to the vibrationally excited
S1 state. This conclusion holds regardless that the relevant vibra-
tional mode is one or more.
Acknowledgements
We are grateful to Dr. Atsushi Miyawaki at the Brain Science
Institute, RIKEN, Japan and Dr. Hideaki Mizuno at Katholieke Uni-
versiteit Leuven, Belgium for discussions. This work has been
supported by a Grant-in-Aid for Scientific Research (A) (No.
25248009) and a Grant-in-Aid for Scientific Research (C) (No.
24550033) from the Japan Society for the Promotion of Science
(JSPS), and a Grant-in-Aid for Scientific Research on Innovative
Areas (No. 25104005) from the Ministry of Education, Culture,
Sports, Science and Technology (MEXT).
Alternatively, the involvement of the closely lying S2 state
readily rationalizes the experimental observation, where the S2
state has smaller one-photon absorptivity and larger two-photon
absorptivity than the S1 state. Then, in the OPA spectrum, the
S2 ← S0 transition band is hidden by the stronger S1 ← S0 transi-
tion band, and the OPA peak maximum is close to that of the S1 ← S0
transition band. Inversely, in the TPA spectrum, the intensity of the
S2 ← S0 transition band is stronger than that of the S1 ← S0 band,

36
H. Hosoi et al. / Chemical Physics Letters 630 (2015) 32–36
[13] M. Drobizhev, N.S. Makarov, S.E. Tillo, T.E. Hughes, A. Rebane, Nat. Methods 8
Appendix A. Supplementary data
(2011) 393.
[14] C. Xu, W. Zipfel, J.B. Shear, R.M. Williams, W.W. Webb, Proc. Natl. Acad. Sci. U.
S. A. 93 (1996) 10763.
[15] H. Hosoi, S. Yamaguchi, H. Mizuno, A. Miyawaki, T. Tahara, J. Phys. Chem. B 112
(2008) 2761.
Supplementary data associated with this article can be found, in
the online version, at doi:10.1016/j.cplett.2015.04.028
[16] H. Hashimoto, K. Isobe, A. Suda, F. Kannari, H. Kawano, H. Mizuno, A. Miyawaki,
K. Midorikawa, Appl. Opt. 49 (2010) 3323.
References
[17] M. Drobizhev, N.S. Makarov, S.E. Tillo, T.E. Hughes, A. Rebane, J. Phys. Chem. B
116 (2012) 1736.
[18] V. Voliani, R. Bizzarri, R. Nifosi, S. Abbruzzetti, E. Grandi, C. Viappiani, F. Beltram,
J. Phys. Chem. B 112 (2008) 10714.
[1] R.Y. Tsien, Annu. Rev. Biochem. 67 (1998) 509.
[2] M. Zimmer, Chem. Soc. Rev. 38 (2009) 2823.
[3] S.R. Meech, Chem. Soc. Rev. 38 (2009) 2922.
[4] P.T.C. So, D.C.Y., B.R. Masters, K.M. Berland, Annu. Rev. Biomed. Eng. 2 (2000)
399.
[19] J. Dong, K.M. Solntsev, L.M. Tolbert, J. Am. Chem. Soc. 128 (2006) 12038.
[20] P. Atkins, J. De Paula, Physical Chemistry, Oxford University Press, New York,
USA, 2010.
[21] B. Honig, J. Jortner, A. Szöke, J. Chem. Phys. 46 (1967) 2714.
[22] P. Macak, Y. Luo, H. Agren, Chem. Phys. Lett. 330 (2000) 447.
[23] P.R. Callis, T.W. Scott, A.C. Albrecht, J. Chem. Phys. 75 (1981) 5640.
[24] A.F. Bell, X. He, R.M. Wachter, P.J. Tonge, Biochemistry 39 (2000) 4423.
[25] X. He, A.F. Bell, P.J. Tonge, J. Phys. Chem. B 106 (2002) 6056.
[26] P. Schellenberg, E. Johnson, A.P. Esposito, P.J. Reid, W.W. Parson, J. Phys. Chem.
B 105 (2001) 5316.
[5] W.R. Zipfel, R.M. Williams, W.W. Webb, Nat. Biotechnol. 21 (2003) 1369.
[6] F. Helmchen, W. Denk, Nat. Methods 2 (2005) 932.
[7] K. Svoboda, R. Yasuda, Neuron 50 (2006) 823.
[8] A. Volkmer, V. Subramaniam, D.J.S. Birch, T.M. Jovin, Biophys. J. 78 (2000) 1589.
[9] G.A. Blab, P.H.M. Lommerse, L. Cognet, G.S. Harms, T. Schmidt, Chem. Phys. Lett.
350 (2001) 71.
[10] A.A. Heikal, S.T. Hess, W.W. Webb, Chem. Phys. 274 (2001) 37.
[11] M. Drobizhev, N.S. Makarov, T. Hughes, A. Rebane, J. Phys. Chem. B 111 (2007)
14051.
[27] S. Olsen, R.H. Mckenzie, Chem. Phys. Lett. 492 (2010) 150.
[12] M. Drobizhev, S. Tillo, N.S. Makarov, T.E. Hughes, A. Rebane, J. Phys. Chem. B
113 (2009) 855.