Owing to the high critical temperature of its superconducting transition, the k-(BEDT-TTF)₂X (X= Cu[N(CN)₂]Cl, Cu[N(CN)₂]Br, Cu(NSC)₂, etc.) family has been most intensively investigated among the various types of charge-transfer (CT) salts of BEDT-TTF.[1],[2] These compounds are located at the boundary between metallic and Mott-insulating phases.[3],[4] Among them, (BEDT-TTF)₂Cu[N(CN)₂]Br shows the highest transition temperature, T_c=11.6 K at ambient pressure.[5] Raman and infrared spectroscopy were applied to this family to elucidate which molecular and lattice phonons are sensitive to the superconductivity.[6],[7] Apart from this subject, the C=C stretching modes were investigated due to the finding that their frequencies were charge sensitive and one of them was strongly coupled with a CT excited state.[8],[9] As shown in Fig. 1, the BEDT-TTF molecule has three normal modes that mostly involve C=C stretching vibrations.[10] According to Kozlov’s notation,[10] these modes are called n₂(a_g), n₃(a_g), and n₂₇(b_1u), where the symbol in parentheses denotes the symmetry species, assuming D_2h for the free BEDT-TTF molecule. The n₂ and n₃ modes are Raman-active and involve the bridge and symmetric ring C=C stretching vibrations. In the case of the neutral BEDT-TTF molecule, the ring and bridge C=C stretchings mix almost equally in n₂ and n₃. In k-(BEDT-TTF)₂Cu[N(CN)₂]Br, however, they are almost separated: n₂ is assigned mainly to ring C=C stretching, and n₃ to bridge C=C stretching.[11],[12] The infrared-active n₂₇ mode corresponds to the anti-symmetric ring C=C stretching and is thus completely separated from the bridge C=C stretching. The frequencies of these three modes are close to each other; n₂=1552 cm^-1, n₃=1494, and n₂₇=1511 in neutral BEDT-TTF.[10]

Owing to the electron molecular vibration (EMV) coupling, some totally symmetric modes become infrared active in charge-transfer salts. In the reflectivity spectra of k-(BEDT-TTF)₂X salts, the EMV-coupled n₃ has been observed at 1200-1400 cm^-1. The EMV-coupled n₂ mode has not been definitely assigned. According to Eldridge et al.,[12] EMV-coupled n₂ is situated at 1470-1485 cm^-1, whereas Swietlik et al.[13] and Vlasova et al.[14] assigned it to the ~1340 cm^-1 band. In the Raman spectrum, uncoupled n₂ and n₃ were observed near 1500 cm^-1 and 1470 cm^-1, respectively. They are in resonance with 450-680 nm[13] and 1060 nm[7] excitation. Truong et al. measured the polarized Raman spectra of the C=C stretching region at various temperatures under 514 nm excitation using single crystals of k-(BEDT-TTF)₂Cu[N(CN)₂]Br[15] and k-(BEDT-TTF)₂Cu[N(CN)₂]Cl[16]. They found anomalies in the n₃ mode at 180 K and below 15 K. However, the Raman spectrum of 1.06 mm excitation by Eldridge et al. did not show such anomalies.[7] Swietlik et al. studied the resonance Raman spectra of k-(BEDT-TTF)₂Cu(CNS)₂, and attributed 1405 cm^-1 and 1420 cm^-1 bands to Raman-active n₂₇[13] The theoretical consideration for this interpretation was presented by Kozlov and Tokumoto.[17] Although several attempts have been made, the Raman and infrared bands in the 1300-1600 cm^-1 region have not been completely assigned, probably because no single crystal Raman data with a high signal to noise ratio has been reported. As the unit cell of k-(BEDT-TTF)₂ Cu[N(CN)₂]Br contains eight BEDT-TTF molecules, one normal mode is eight-fold degenerate, and these degenerate modes are split through the inter-molecular interaction. In this compound, n₃ is strongly coupled with the low-energy CT excited state, which contributes to the splitting of the degenerate modes. This electron molecular vibration (EMV) coupling works not only in infrared-active modes but also in some Raman-active modes, and thus increases the number of observable Raman active bands. Complete assignment for the Raman spectrum including the effect of EMV coupling and consideration of symmetry has not been performed in this compound.

We measured the low-temperature resonance Raman spectra of k-(BEDT-TTF)₂Cu[N(CN)₂]Br single crystals and its isotope analogues from 150 cm^-1 to 2300 cm^-1. We focused on the region of C=C stretching modes (1300-1600 cm^-1), using five different k-(BEDT-TTF)₂Cu[N(CN)₂]Br species: ¹²C-d[0,0], ¹²C-d[2,2], ¹²C-d[3,3], ¹²C-d[4,4], and ¹³C-d[0,0][18] In ¹²C-d[2,2], two hydrogen atoms in each ethylene group are replaced by deuterium, and in ¹³C-d[0,0], the two central ¹²C atoms bridging the 1,3-dithiol rings are replaced by the carbon isotope ¹³C. For convenience, we call non-substituted and deuterated crystals ¹²C-crystals. To detect infrared-active n₂₇, we measured the low-temperature reflectivity with 1 cm^-1 resolution in b polarization, perpendicular to the conducting layers. In this article, we present the Raman and infrared-active bands that are related to the three C=C stretching vibrations, and propose the assignment based on the factor group analysis.

The crystals were grown by a standard electrochemical oxidation.[18] We selected non-twinned crystals, which had the most developed plane (010) of about 0.5x0.5 mm2, with a diamond shape, though they were often partly broken and distorted. The diagonal lines corresponded to the crystallographic a and c axes. The directions of the a and c axes in this plane were checked by their known reflectivity.[19] The other planes {101} were rectangular. Typical size in the b axis direction was 0.05 mm (for ¹²C-d[4,4]) - 0.3 mm (for ¹²C-d[0,0]).

Raman spectra were measured in the back-scattering geometry on a Renishaw Ramascope System-1000. An analyzer for the scattered light was placed before the entrance slit of the monochromator. For the laser light polarized parallel to a and c, the scattered light was collected from the (010) plane. The b-polarized spectrum was measured by the reflection from the (101) plane for ¹²C-d[0,0], ¹²C-d[2,2], ¹²C-d[3,3], and ¹³C-d[0,0], and from an unknown plane for thin ¹²C-d[4,4]. 514 nm Ar⁺, 633 nm He-Ne, and 785 nm diode lasers were used for excitation.[20] The laser intensity after at the sample was from »50 mW to »150 mW, and was focused on an area with a diameter of »5 mm. Typical measurement time was about 60 min with and 15 min without an analyzer. The experimental error for the absolute wavenumber value was ±1 cm^-1; the error for the relative wavenumber shift was ±0.1 cm^-1; and the resolution was 3 cm^-1 for the Ar⁺ laser and 2 cm^-1 for the He-Ne and diode lasers. Diamond, silicon, and indene were used for calibration of the Raman shift. Polarized reflection spectra in the 600-4000 cm^-1 region were obtained with a FT-IR spectrometer Nicolet Magna 760 combined with a Spectratech, IR-Plan microscope. The spectral resolution was taken as 1 cm^-1 for the 15 K measurement and 4 cm^-1 for other temperatures. The peak position and width of the Raman- and infrared-active bands were obtained from the curve fitting with Lorentz functions.

Samples were cooled inside a helium-flow cryostat Oxford CF1104s under a microscope. A silicon diode sensor was attached to a sample holder just below the sample crystal fixed by silicone grease. To obtain the temperature dependence, we used »50 mW as a laser power (only the 514 nm Ar⁺ laser was used for temperature dependence). We estimated the elevation of temperature by laser heating to be 3-5 K.[21] If there is no description of the temperature for the Raman bands in the text below, the actual sample temperature was 10 K. For the polarization dependence experiment, we obtained all the spectra at 5 K (nominal temperature) with the laser power of »150 mW. Therefore, the actual temperature is estimated to be 10-15 K.

In a first approximation, ¹²C-d[4,4] was found to be isostructural to the orthorhombic ¹²C-d[0,0] crystal with space group Pnma (D_2h¹⁶) and Z=4. [22],[23],[24],[25],[26]¹²C-d[2,2] and d[3,3] crystals are probably isostructural to ¹²C-d[0,0], although no x-ray diffraction experiment has been performed. In ¹²C-d[0,0] and ¹²C-d[4,4], the periodic unit along the c-axis is doubled below 200K.[25],[26] Such a doubling has not been investigated in ¹²C-d[2,2] and ¹²C-d[3,3]. Because the unit cell of k-(BEDT-TTF)₂Cu[N(CN)₂]Br contains eight BEDT-TTF molecules, every internal molecular vibration splits, in principle, into eight components forming coupled modes (unit-cell modes). Considering a dimer inside a crystal, these components are divided into 4a_g and 4a_u, since the two molecules in a dimer are connected by a center of symmetry. In the case of gerade molecular modes, the molecules in a dimer oscillate in phase in the former group, whereas they oscillate out of phase in the latter group. Considering a whole crystal, the four modes in each group are classified by the glide plane symmetries perpendicular to a, b, and c axes. If we take rectangular x, y, and z axes parallel to the crystallographic a, b, and c axes, respectively, the eight components are classified into A_g, B_1g, B_2g, B_3g, A_u, B_1u, B_2u, and B_3u, as shown in Table 1.[27] According to the selection rule, the first four modes are Raman active and the last three modes are infrared active. These crystal modes for a n₂(a_g) molecular vibration are drawn in Fig. 2. Anion layers separate conducting layers parallel to the (010) plane. Therefore, the CT interaction along the b-axis is expected to be negligibly weak. Thus, A_g and B_1g modes are nearly degenerate, since the phase relation among the four molecules within the layer is the same for every layer. Similarly, B_2g and B_3g are also degenerate, but the degeneracy between A_u and B_2u and between B_2u and B_3u is likely to be lifted due to the dipole-dipole interaction between the dipoles induced inside each layer.

Since the n₃(A_g) mode is the in-phase vibration in a dimer as well as between the dimers, the electron-molecular-vibration (EMV) interaction does not work between any molecule pairs in the unit cell.[32] Therefore, the unit-cell Raman tensor is directly connected with the molecular Raman tensor through the oriented gas model. The molecular Raman tensor for a totally symmetric mode of BEDT-TTF (D_2h) is given by the diagonal components a_LL, a_SS, and a_NN, where the L-, S-, and N-axes are chosen along the long and short axes of the BEDT-TTF molecule and normal to the molecular plane, respectively. Then the unit-cell Raman tensor for A_g symmetry is given by a_aa=0.3091a_LL+0.3135a_SS+0.3774a_NN, a_bb=0.6897a_LL+0.1631a_SS+0.1471a_NN, and a_cc=0.0011a_LL+0.5235a_SS+0.4755a_NN, based on the crystal structure with the Pnma space group.[23],[24] Resonance enhancement was observed in the a- and b-polarized spectra. The resonance effect seems to be realized mostly for a_LL (a_LL» a_SS , a_NN), because the intra-molecular electronic transition in this spectral region is polarized along the L-axis of BEDT-TTF. [33]The same result was found by Swietlik et al.[13] The polarization dependence shown in Fig. 3 supports this speculation, since the order of intensity of the B₁ band, (b,b)> (a,a)» (c,c), is consistent with the relation a _bb>a _aa»a _cc under the condition a _LL»a _SS and a _NN.

According to the assignment in section 3-2, the position of n₃(B_2g) of ¹²C-d[3,3] is lower by 56 cm^-1 than that of n₃(A_g). This is regarded as a factor group (Davydov) splitting. In this compound, the splitting comes from the EMV interaction. In the n₃(A_g) mode, the adjacent dimers vibrate in phase, whereas in the n₃(B_2g) mode they vibrate in the opposite phase, as shown in Fig. 2. The out-of-phase (B_2g) mode induces a charge oscillation between the adjacent dimers, and this oscillation induces a time-dependent polarizability synchronized to the n₃(B_2g) mode. Rice formulated the intensity and frequency of the infrared-active vibronic mode using a linear response theory.[34] Such infrared-active modes in ¹²C-d[0,0] were found by Eldridge et al. at n₃(B_1u)»1340 cm^-1 and n₃(B_3u)»1280 cm^-1.[19] Menegetti and Pecile theoretically analyzed the Raman-active vibronic mode that interacts with a CT excited state with the same symmetry as that in the Raman-active mode.[35] This interaction lowers the frequency of the B_2g mode, and broadens the linewidth. Actually, B₂ is much broader than B₁. The frequency shift and line broadening are observed in ¹³C-d[0,0] as well [See Fig. 3]. Swietlik et al. assigned B₂ to the n₂₇(A_g) mode based on the finding that B₂ has a resonance profile different from that of B₁.[13] We also found a different resonance effect in the spectra of ¹³C-d[0,0]: B₁ is relatively enhanced in (a,a) and (c,c) polarizations under the 633 nm excitation, whereas B₂ is relatively suppressed. However, this result is not inconsistent with our assignment, because the Raman tensor of n₃(B_2g) is different from that of n₃(A_g). As we explained in the preceding paragraph, the intensity of n₃(A_g) is derived from the intra-molecular Raman tensor, and is determined by its a_LL component, whereas the intensity of n₃(B_2g) consists of intra- and inter-molecular Raman tensors. The former is estimated from the oriented gas model as a_ac=0.0184a_LL+0.4051a_SS-0.4236a_NN, where the contribution of a_SS and a_NN becomes relatively large even if a_LL is more enhanced by the resonance effect, and the excitation profiles of a_LL, a_SS, a_NN can be different.

The factor group splitting between n₃(A_g) and n₃(B_2g) is 56 cm^-1 and 45 cm^-1 in ¹²C-d[3,3] and ¹³C-d[0,0], respectively. This difference will be discussed in the next section along with the factor group splitting between n₂(A_g) and n₂(B_2g). The downshift of the EMV-coupled infrared-active mode from the uncoupled n₃(A_g) mode is ~130 cm^-1 for n₃(B_1u) and ~190 cm^-1 for n₃(B_3u), which values are significantly larger than those of the EMV-coupled Raman-active mode.[19] The infrared-active modes n₃(B_1u) and n₃(B_3u) vibrate out of phase in a dimer [See Fig. 2], which induces an intra-dimer charge oscillation as well as an inter-dimer oscillation. Because the intra-dimer charge oscillation is predominant in this case, the shift is determined by the transfer integral within the dimer, the excitation energy of the intra-dimer CT state, and the coupling constant of this mode. On the other hand, n₃(B_2g) induces only the inter-dimer charge oscillation. In this case, the shift is estimated from the transfer integral between the dimers, inter-dimer CT excitation energy, the intra-dimer Coulomb energy, and the coupling constant. It worth mentioning that the inter-dimer EMV shift is determined by interaction with four neighbours, and the intra-dimer shift by interaction with only one neighbour. In the former case, the charge flow occurs from one dimer to the surrounding four dimers at the same time [see Fig. 2]. This kind of problem has not yet been solved.

Incidentally, n₃(B_2u) is not observed in the reflectivity spectrum polarized along the b axis Incidentally, n3(B2u) is not observed in the reflectivityspectrum polarized along the b axis [14]. This result is quite natural, because the induced dipole in the layer is parallel to the direction connecting the centers of the molecules in the dimer that is perpendicular to the b-axis.

Unlike the n₃(A_g) mode, the n₂(A_g) mode appears to be more isotropic. The symmetric molecular vibration n₂(a_g) is mostly assigned to the ring C=C stretchings.[11] The n₂ vibration seems to be in resonance with an intra-molecular transition, the polarization of which is parallel to the C=C bonds of the rings (//S): a_SS» a_LL, a_NN. This direction has comparable projections for the a, b, and c axes, resulting in the approximately isotropic polarization.

EMV interaction for n₂ is expected to be weaker than that for n₃.[10],[12],[36] Actually, the EMV-coupled mode n₂(B_1u) and n₂(B_3u) have not been definitely identified in the reflection spectrum[37], probably because their intensities are too weak. The factor group splitting is not observed in ¹²C-d[3,3]; in other words, n₂(B_2g) appears at the same frequency as n₂(A_g), as shown in Fig. 3. In ¹³C-d[0,0], however, there is a splitting 8 cm^-1 if the band A₂ is assigned to n₂(B_2g). This assignment is supported by the linewidth (7 cm^-1) of A₂, which is broader than that (<2 cm^-1) of n₂(A_g). Furthermore, no band is found at 1500 cm^-1 in the (a,c) and (c,a) spectra, although the corresponding band is found in the (a,c) and (c,a) spectra of ¹²C-d[3,3].

Let us compare the factor group splitting ¹²Dn_i=n_i(A_g)-n_i(B_2g) for n₂ and n₃ in ¹²C-d[3,3] with ¹³Dn_i=n_i(A_g)-n_i(B_2g) in ¹³C-d[0,0]. Concerning n₂, no splitting is found in ¹²C-d[3,3], whereas the 8 cm^-1 splitting is found in ¹³C-d[0,0]. These results are strange, because these two compounds have the same electronic and crystal structures.[38] The corresponding splitting (¹²Dn₃=56 cm^-1) for n₃ of ¹²C-d[3,3] is significantly larger than ¹³Dn₃=45 cm^-1 of ¹³C-d[0,0], and the temperature dependence of ¹²Dn₃ is similar to that of ¹³Dn₃.[21] This difference may appear to come from the difference in the positions of n₃(A_g) between ¹²C-d[3,3] and ¹³C-d[0,0], because the closer the uncoupled mode n(A_g) to the CT excited state, the larger the low-frequency shift of the perturbed mode. Thus, ¹²Dn₃ is expected to be larger than ¹³Dn₃. However, this effect is small, because the CT excitation energy is significantly larger than the vibrational energy of n₃. Using formulae from ref. [34] and neglecting all phonons except n₃, we numerically estimated we numerically estimated that this factor adds only about 2 cm^-1 difference between ¹³Dn₃ and ¹²Dn₃. We therefore consider (¹²Dn₃-¹³Dn₃)=11 cm^-1 as the result of a different degree of mixing between the bridge and ring C=C stretching vibrations in n₂ and n₃. Eldridge et al. proved that the bridge and ring C=C stretchings are separated from each other in the intra-dimer EMV-coupled infrared-active modes n₃(B_1u) and n₃(B_3u).[11],[12] For Raman-active n₂(A_g) and n₃(A_g) their results are less conclusive because they compared the isotope shifts after ¹³C(6) substitution between neutral ET and (ET)₂Cu[N(CN)₂]Br[11]. As shown in Fig. 3, we, on the other hand, obtained a direct proof that the ring and the bridge C=C stretching vibrations are completely separated in n₂(A_g) and n₃(A_g), because n₂(A_g) shows a very small isotope shift for the substitution of ¹³C at the bridge C=C atoms. However, the situation seems to be different in n₂(B_2g) and n₃(B_2g). If we assume that the bridge and ring C=C stretchings for n₂(B_2g) and n₃(B_2g) are slightly mixed, the high-frequency n₂(B_2g) shifts to the higher-frequency side and low-frequency n₃(B_2g) shifts to the lower-frequency side. We consider that this kind of mixing occurs only in ¹²C crystals, because in ¹³C-d[0,0] the bridge ¹³C=¹³C is already non-equivalent to the ring ¹²C=¹²C. This speculation explains the high-frequency shift of n₂(B_2g) and the low-frequency shift of n₃(B_2g) in ¹²C-d[3,3]. This raises the question of why the mixing in the B_2g mode is larger than that in the A_g, B_1u, or B_3u mode. We consider as follows: n₃(B_2g) is coupled with the inter-dimer CT state, which involves the electronic configuration of BEDT-TTF⁰ such that 2(BEDT-TTF)₂¹⁺ => (BEDT-TTF)₂⁰(BEDT-TTF)₂²⁺. As has been described previously, the bridge and ring C=C stretchings are almost equally mixed in BEDT-TTF⁰.[10] Thus, n₂(B_2g) slightly involves the bridge C=C stretching and n₃(B_2g) involves the ring C=C stretching.

In this section, we will discuss the band C at 1466 cm^-1 in Fig.3. A similar band was observed by Truong et al. in k-(BEDT-TTF)₂Cu[N(CN)₂]Br[15] and k-(BEDT-TTF)₂Cu[N(CN)₂]Cl[16], and by Swietlik et al.[13] and Sugai et al.[29] in k-(BEDT-TTF)₂Cu(SCN)₂. All of them interpreted this band as the factor group component of the n₃ mode. However, it is obvious from the polarization dependence that this band cannot be assigned to the factor group component. In accord with the results of Swietlik et al.,[13] we found that the band C satisfied the same resonance condition as the band B₁ for 514 nm and 633 nm laser irradiation: the resonant electronic transition is polarized along the L-axis. However, the resonance condition of C is different for 785 nm laser irradiation as shown in Fig. 4. According to the Raman spectrum excited by 1.06 nm irradiation,[7] C is not enhanced in contrast to the remarkable enhancement of n₂ and n₃. Therefore, the origin of the C band is different from those of n₂ and n₃. Because this band shows no isotope shift for the substitution of the ¹³C at the bridge C=C atoms, we assign the band C to n₂₇(A_g). The same reasoning may be applied to a combination band, because the C=C stretching is rather isolated from other normal modes. Further on in this section, we discuss the both possible assignments.

The molecular vibration n₂₇(b_1u) mainly involves the anti-symmetric ring C=C stretching,[10] which is infrared-active and thus Raman-inactive in a free molecule [See Fig. 1]. According to Refs. [13] and [17], the n₂₇(b_1u) mode becomes Raman-active if the two modes in a dimer are strongly coupled. The EMV coupling in n₂₇(A_g) seems to be somewhat different from that in n₃ mode, because n₂₇(A_g) does not induce a net charge transfer between the molecules. Therefore, n₂₇ is not coupled with a low-energy CT excited state but a high-energy intra-molecular excited state, and thus the shift from the uncoupled mode is likely to be small. As will be shown later in this section, the splitting between n₂₇(B_2u) and n₂₇(A_g) is small (7 cm^-1). This closeness in frequency also supports the assignment for the band C as n₂₇(A_g). It is worth mentioning that n₂₇(A_g) can mix with n₃(A_g) and borrow intensity from the latter. Owing to the isotope shift of n₃(A_g), this mixing becomes much smaller or absent in n₂₇(A_g) of ¹³C-d[0,0]. This mixing explains the much weaker intensity of the band C in ¹³C-d[0,0][See Fig. 3]. The broad and weak band at ~1468 cm^-1 in the (a,c) spectrum is assigned to n₂₇(B_2g) from this polarization. The splitting between n₂₇(A_g) and n₂₇(B_2g) is smaller than that between n₂₇(A_g) and n₂₇(A_g)(B_2u). This is reasonable, because the corresponding EMV coupling comes from the inter-dimer interaction.

Figure 5 shows a comparison of the temperature dependences of the position w(n_i(A_g),T) and linewidth D(n_i(A_g),T) of n₃(A_g) and n₂₇(A_g). On lowering temperature, the position of n₃(A_g) increases down to 80 K, and levels off below 80 K. This common kind of hardening of phonons is caused by a decrease in interaction with low-frequency phonons. There are no differences in w(n₃(A_g),T) among ¹²C-d[2,2], ¹²C-d[3,3], and ¹²C-d[4,4] [The case of ¹²C-d[0,0] will be discussed in section 3-5]. On the other hand, w(n₂₇(A_g),T) is almost temperature independent below 200 K, and no essential differences are observed among ¹²C-d[0,0], ¹²C-d[2,2], ¹²C-d[3,3], and ¹²C-d[4,4]. The high-temperature region of D(n₃(A_g),T) and D(n₂₇(A_g),T) is also characteristic of the phonon-phonon scattering mechanism. If a high-frequency phonon is scattered by a low-frequency phonon near a “most effective” frequency w₀, the width is expressed by the equation D(T)=D₀+a(exp(w₀/T)-1)^-1, where w₀»150 cm^-1 in both n₃(A_g) and n₂₇(A_g). The slope a/w₀ in the high-temperature region is (1.9±0.1)×10^-2 cm^-1K^-1 for n₃(A_g), whereas the slope of n₂₇(A_g) is a/w₀=(5.4±0.6)×10^-2 cm^-1K^-1, which is 2.5 times larger than that of D(n₃(A_g),T).[39] Such a big difference in slope suggests that the band C might be assigned not to n₂₇(A_g) but to a combination band, the intensity of which is taken from the single-phonon n₃(A_g) mode by Fermi resonance. In the case of the phonon-phonon scattering mechanism, the linewidth of a combination band is the sum of the widths of each component, D(T)=D₁(T)+D₂(T). This mechanism explains the 2.5-fold increase of the slope of C. However, the possibility can't be excluded that D(n₂₇(A_g),T) has an inherently big slope, because the slope of n₂₇(B_2u), which is (3.8±0.1)×10^-2 cm^-1K^-1 [Fig. 5, bottom], is almost two times larger than that of n₃(A_g). The other thing that can't be explained by n₂₇(A_g) alone is a big deuteration dependence of the low-temperature linewidth of C which is distributed from 5 cm^-1 (¹²C-d[0,0]) to about 12 cm^-1 (¹²C-d[4,4]). Especially in ¹²C-d[4,4], the band C can no longer be fitted by a single Lorentzian as satisfactorily as it is in ¹²C-d[0,0], ¹²C-d[2,2], and ¹²C-d[3,3], suggesting that C is not a single component. If we assume that the band C is a combination band, then, in order to explain the absence of an isotope shift, we need a two-phonon combination for the band C, in which the fundamentals involve neither the vibration of the C atoms at the central bridge nor the vibration of H atoms. We could not find such a combination of fundamentals. With regard to the assignment for the band C, we cannot make a definite choice between n₂₇(A_g) and a combination. It is possible that both n₂₇(A_g) and a combination band coexist, which would provide, for example, a possible explanation for the doublet of the band C.

The infrared-active n₂₇(B_2u) mode is found in the E||b reflection spectrum, shown in Fig. 6. As we described in section 3-3, EMV-coupled vibronic modes are absent in b polarization, and thus n₂₇(B_2u) is the dominant feature of the spectra.[40] The bottom panel of Fig. 5 shows the temperature dependences of the width and frequency of n₂₇(B_2u), obtained from the least-square fitting with appropriate Lorentz functions. As shown in Fig. 5, this band becomes quite narrow at low temperature, and no splitting is observed in ¹²C-d[2,2].[41] However, in ¹²C-d[0,0], n₂₇(B_2u) is split into two bands, which splitting will be discussed in the next section. Let us consider the splitting between n₂₇(B_2u) and n₂₇(A_g) more carefully. EMV-coupled n₂₇(A_g) should shift to the low-energy side in comparison with n₂₇ in a (hypothetical) free molecule. It would then shift further in the same direction after a mixing with n₃(A_g). In contrast to n₂₇(A_g), infrared-active n₂₇(B_2u) (two molecules vibrating in the same phase) moves to the high-energy side due to the dipole-dipole interaction inside the dimer. As a result, we expected w(n₂₇(B_2u))>w(n₂₇(A_g)), and actually got w(n₂₇(B_2u))-w(n₂₇(A_g))» 7 cm^-1, at 15 K. However this difference drastically decreased with increasing temperature, and was about 4 cm^-1 at 200 K and nearly zero at room temperature (Fig. 5). Such strong temperature dependence comes from the temperature independent position of the band C. Neither EMV coupling nor dipole-dipole interaction can have such temperature independence.

The other infrared-active components of n₂₇ are n₂₇(B_1u) and n₂₇(B_3u). Eldridge et al. reported weak bands at 1472 cm^-1 and 1480 cm^-1 (10 K) in the E||a spectrum of ¹²C-d[0,0] and at 1474 cm^-1 and 1485 cm^-1 in that of ¹³C-d[0,0].[9],[12] Although they assigned both bands to n₂(B_3u), another component, n₂(B_1u), has not been found in the low-frequency region of n₂(B_3u) in the E||c spectrum. Furthermore, both bands are found together only in the ¹²C-[0,0] and ¹³C-[0,0] crystals,[12] whereas we found only one band near 1477 cm^-1 in E||a reflectivity of ¹²C-d[2,2], ¹²C-d[3,3], and ¹²C-d[4,4] compounds [not shown]. We therefore assigned 1472 cm^-1 and 1480 cm^-1 bands to the doublet of n₂₇(B_3u), which accords with the assignment by Swietlik et al.[13] and Vlasova et al.[14] Some discussion on the doublet structure will be presented in the next section. The band corresponding to n₂₇(B_1u) cannot be found in the E||c spectrum, because the direction (//L) of the transition dipole of this mode is nearly perpendicular to the c-axis.

Figure 7 shows the deuteration dependence of n₃(A_g) measured without analyzer for the scattered light. As shown in this figure, the n₃(A_g) band is well reproduced by a single Lorentzian at any temperature in ¹²C-d[2,2], ¹²C-d[3,3], and ¹²C-d[4,4]. Obviously, the shape of n₃(A_g) of ¹²C-d[0,0] is broad and asymmetric compared with those of the deuterated compounds. This line shape does not change after applying an analyzer. The same broadening is observed in n₂(A_g) of ¹²C-[0,0] [not shown]. The broad bands of n₃(A_g) and n₂(A_g)are reproduced well by at least two Lorentz functions below 150 K, as shown in Fig. 7. We call these components as n₃^l(A_g), n₃^h(A_g), n₂^l(A_g), and n₂^h(A_g).[42] As shown in Fig. 8, analogous splitting of n₃(A_g) also seems to exist in ¹³C-d[0,0], because the linewidth of n₃(A_g) in ¹³C-d[0,0] is the same as that in ¹²C-d[0,0], whereas it is significantly broader than that in d[2,2], d[3,3], d[4,4]. Incidentally, if we fit the two components of n₃(A_g) in ¹²C-d[0,0] by a single Lorentzian, the temperature dependence agrees with the result of 1.06 nm excitation by Eldridge et al.[7] Contrary to the results of Truong et al.,[15] we did not find any anomalies around 15 K or 150 K.

As shown in Fig. 6, another charge-sensitive n₂₇(B_2u) mode of ¹²C-d[0,0] shows a splitting (n₂₇^l(B_2u) at 1468 cm^-1 and n₂₇^h(B_2u) at 1475 cm^-1) analogous to n₂(A_g) and n₃(A_g), whereas the n₂₇(B_2u) mode of ¹²C-d[2,2] consists of a single band at 1471 cm^-1. n₂₇(B_3u) of ¹³C-d[0,0] consists of two components [12], and n₂₇(B_3u) of ¹²C-d[2,2], ¹²C-d[3,3], ¹²C-d[4,4] shows a single band [our data is not shown]. As described in section 3-5, in ¹²C-d[0,0] the n₂₇(B_2u) mode also consists of a doublet. Therefore, all of the charge-sensitive modes n₂, n₃, and n₂₇ in non-deuterated crystals show splitting at low temperature. The broadening or splitting of the charge-sensitive modes in ¹²C-d[0,0] and ¹³C-d[0,0] crystals appears to be associated with the superstructure found by Nogami et al.[25], [26] Such a superstructure lowers the symmetry of the unit cell, and consequently leads to a crystallographic inequality among the dimers. However, this interpretation cannot explain the finding of a similar superstructure in the ¹²C-d[4,4] crystal. At present, the reason for the splitting in non-deuterated compounds is not clear.

Let us briefly describe some properties of each component of n₂(A_g), n₃(A_g) and n₂₇(B_2u) in ¹²C-d[0,0]. The separation between n₃^l(A_g) and n₃^h(A_g) is 4.3±0.1 cm^-1, and almost temperature independent [see the top panel of Fig. 5]. At any temperature, each component has nearly the same linewidth as the corresponding component of n₃(A_g) in ¹²C-d[2,2], ¹²C-d[3,3] and ¹²C-d[4,4]. The intensity ratio I(n₃^l(A_g))/I(n₃^h(A_g)) is temperature independent, and falls in the range of 0.8-1.4. It's worth mentioning that at every temperature the ratios I(n₂₇(A_g))/I(n₃(A_g)) are equal among d[2,2], d[3,3], d[4,4], but to obtain the same value for d[0,0] we should take I(n₂₇(A_g))/I(n₃^l(A_g)) or I(n₂₇(A_g))/I(n₃^h(A_g)) instead of I(n₂₇(A_g))/I(n₃(A_g)). Concerning the n₂(A_g) band, the separation between n₂^l(A_g) and n₂^h(A_g) linearly depends upon the temperature [not shown] from 4.8±0.2 cm^-1 (5 K) to 3.9±0.2 cm^-1 (150 K). The linewidth of each component has the same temperature dependence as n₃^l(A_g) and n₃^h(A_g) above 100 K.[43] The intensity ratio I(n₂^l(A_g))/I(n₂^h(A_g)) is very scattered because of the low intensity. The split bands of n₂₇(B_2u) merge into a broad single band at high temperatures. At 15 K, the linewidth is 3.0±0.1 cm^-1 for n₂₇^l(B_2u) and 7±1 cm^-1 for n₂₇^h(B_2u).[44] Though n₂₇^h(B_2u) looks weaker than n₂₇^l(B_2u), their intensities are comparable, because the width of the latter is half that of the former. The sum of these two bands has almost the same intensity as the single n₂₇(B_2u) band in ¹²C-d[2,2].

Finally, we summarize the crystal modes of the C=C stretching vibrations of all of the compounds in Table 3.

We interpreted the polarized Raman spectra based on the factor group analysis. We found a Raman-active EMV-coupled mode, n₃(B_2g), as a broad band near 1420 cm^-1 in natural and all deuterated crystals. The splitting between the A_g and B_2g species of n₃ is originated from the inter-dimer EMV interaction. The ring and bridge C=C stretching vibrations are separated in n₂(A_g) and n₃(A_g), whereas they are slightly mixed in n₂(B_2g) and n₃(B_2g). The band C at 1466 cm^-1 is not assigned to the factor group component of n₃(A_g) but probably to n₂₇(A_g), though this assignment is not conclusive. A splitting was newly found in all of the C=C stretching modes n₂, n₃, and n₂₇ in non-deuterated crystals including ¹³C-d[0,0], while such a splitting was absent in the fully and partly deuterated crystals.

The authors express their thanks to Dr. K. Yamamoto for stimulating discussions. KY was supported in part by a Grant-in-Aid for Scientific Research on Priority Areas (B) of Molecular Conductors and Magnets (Area No. 730 / Grant No.11224212) from the Ministry of Education, Culture, Sports, Science, and Technology.

We summarized other vibrational data except that on the C=C vibrations in Table 4. For the assignments, we used the data from refs. [11], [15], [19], [29], and [45]. Some assignments are tentative. Some lines may be combination bands.