1 Introduction

In many organic conductors, the transfer integrals between neighbouring molecules are comparable to the on-site (U) and inter-site (V) Coulomb interaction energies; therefore, electron correlations play an important role in these materials. The inter-site Coulomb interactions in organic conductors can be responsible for changes of average electron density on molecules, i.e. charge ordering (or charge disproportionation) phenomena [1, 2]. The charge ordering states can be widely realised in two-dimensional charge transfer salts formed by the organic donor bis(ethylenedithio)tetrathiafulvalen (BEDT–TTF or ET) with various inorganic anions. In BEDT–TTF salts the inter-site Coulomb repulsion can give rise to stripe-type charge ordered states, where the sites with more and less amount of charge are arranged in rows alternately. As shown by recent theoretical calculations [3], both energies U and V together with the anisotropy of transfer integrals are necessary for the understanding of charge distribution patterns inside the conducting BEDT–TTF layers.

Infrared and Raman spectroscopies are good experimental methods for investigations of the charge ordering phenomena [4–6]. In BEDT–TTF salts, the positions of vibrational bands, especially those related to the C=C stretching vibrations of BEDT–TTF, can be used for determination of the charge density [7]. The neutral BEDT–TTF molecule has two Raman active C=C modes: ν₂(a_g)=1551 cm^–1 and ν₃(a_g)=1493 cm^–1 [8]. On increasing the average oxidation state of BEDT–TTF, the frequencies of these bands are shifted down, reaching the values ν₂(a_g)=1455 cm^–1 and ν₃(a_g)=1431 cm^–1 for BEDT–TTF⁺¹ cation [9]. As shown recently, for estimation of charge density in the BEDT–TTF salt, the ν₂(a_g) Raman band is more reliable than ν₃(a_g), because of weaker coupling with electrons, which yields an additional frequency shift towards lower frequencies [6]. On the other hand, the charge ordering can be responsible for the formation of a new energy gap in electronic spectrum or the appearance of new charge-transfer bands, and these phenomena can be observed in infrared spectra.

The salts κ-[Et₄N](BEDT–TTF)₄M(CN)₆·3 H₂O (M = Co^III, Fe^III) are isostructural and exhibit very similar semiconducting properties [10–12]. In conducting layers, the BEDT–TTF molecules are arranged in almost perpendicular dimers (so-called κ–phase structure) and alternate with insulating layers of hexacyanometalate trianions M(CN)₆^3–. Both crystals undergo analogous phase transitions at about T = 240 K and T = 150 K. At room temperature, the organic conducting layer is built of two different kinds of BEDT–TTF molecules forming two kinds of dimers (labelled A and B). At 240 K, the unit cell is doubled and there exist four different kinds of dimers (labelled A, B, C and D). The phase transition at 150 K has no distinguishable influence on the crystal structure; however, it was evidenced by heat-capacity and paramagnetic-susceptibility measurements. The analysis of the bond lengths of BEDT–TTF donors suggested that at room temperature both A and B molecules bear a charge +0.5 (i.e. charge distribution is: ...A^+0.5A^+0.5...B^+0.5B^+0.5...); but below 240 K a charge redistribution takes place leading to the following pattern: ...A⁰A⁰...B^+0.5B^+0.5...C^+0.5C^+0.5...D⁺¹D⁺¹...[12].

Recently, the charge ordering phenomena in the Fe^III salt were studied by IR and Raman spectroscopy. The existence of charge redistribution inside BEDT–TTF layers was confirmed by the FT–NIR Raman experiment (excitation: λ = 1064 nm) on powdered crystals dispersed in KBr matrix [13] as well as by the IR reflectance from single crystals [14]. Here, we report on IR and Raman investigations of the Co^III and Fe^III single crystals. Preliminary spectroscopic data for the Co^III salt have been described in [15].

2 Experimental

Single crystals of the salts κ–[Et₄N](BEDT–TTF)₄M(CN)₆·3 H₂O (M = Co^III, Fe^III) were grown on platinum electrodes by an electrocrystallisation technique, as previously described in [10]. The crystals were in the form of shiny platelets with the best-developed crystal face (001) being parallel to the conducting BEDT–TTF layers.

Raman spectra of the single Co^III and Fe^III crystals were recorded in the spectral range from 1200 to 1800 cm^–1 using a Renishaw Ramascope System 1000. The excitation light from a NIR diode laser (λ = 785 nm) was focused on the (001) crystal face. The scattered light was collected in a backscattering geometry. The power of the laser beam was kept below 240 μW to avoid sample overheating.

Polarised reflectance spectra of the Co^III single crystals were measured from the (001) crystal face, in the frequency region from 600 to 10 000 cm^–1, using a FT–IR Nicolet Magna 760 spectrometer. The spectrometer was equipped with an IR microscope Spectratech IR Plan. The electrical vector of the polarised light was either parallel (E_∥) or perpendicular (E_⊥) to the direction corresponding to the maximum of reflected energy. The frequency-dependent conductivity was determined by Kramers–Kronig analysis of the reflectance data. The low-frequency data were extrapolated to zero-frequency assuming a constant value, a common procedure for semiconducting materials. The high-frequency data in the region 10 000–30 000 cm^–1 were extended on the basis of room temperature data of the Fe^III salt [14] and above 30 000 cm^–1 a constant reflectance was assumed.

Low temperature IR and Raman measurements were performed with samples mounted with vacuum grease in helium cryostats. Previous investigations have shown that in both compounds the intensity of the phase transition at 150 K strongly depends on the cooling or warming rates and it is accompanied by a large thermal hysteresis, whereas the transition at 240 K remains unchanged [10–12]. These effects were not studied in our experiments: the sample temperature was carefully decreased with the cooling rate below 1 K min^–1. We observed frequently that the crystals cracked in cryostat during evacuation at room temperature. The cracking was related to a process of loosing water molecules by sample under vacuum (there are three water molecules in the unit cell). Subsequently, we have found that this problem can be avoided when the sample temperature was reduced down to about 275 K before cryostat evacuation. Our present experiments were carried out following such a procedure.

3 Results

Raman spectra of the Co^III and Fe^III crystals as a function of temperature are shown in Figs. 1 and 2, respectively. At room temperature, the spectra of both salts are very similar. On cooling down they undergo analogous modifications due to the charge ordering, nevertheless, they show different temperature dependence.

As seen in Fig. 1, the room temperature spectrum of Co^III salt is dominated by two bands at 1495 and 1465 cm^–1, corresponding to the ν₂(a_g) and ν₃(a_g) modes of BEDT–TTF molecule with charge +0.5, respectively, in agreement with the stoichiometry and the crystallographic data. In the salt κ–(BEDT–TTF)₂Cu(SCN)₂ containing (BEDT–TTF^+0.5)₂ dimers, the ν₂(a_g) and ν₃(a_g) Raman bands were observed at 1491 and 1466 cm^–1 [16].

The phase transition at T = 150 K has a drastic influence on the Raman spectrum of the Co^III salt (Fig. 1): the bands related to molecules with charge +0.5 nearly disappear and simultaneously new strong lines grow, which are related to molecules with charge +1. Below 150 K, the band ν₃(a_g) = 1465 cm^–1 is nearly not seen, i.e. the number of molecules with the charge +0.5 decreases considerably. A similar process occurs with the ν₂(a_g) = 1495 cm^–1 band – we suggest that the spectral feature at about 1500 cm^–1 should be attributed to the ν₃(a_g) mode of neutral molecule. Additionally, on closer analysis of the Raman spectra, we find a weak and broad band centred at about 1540 cm^–1, related to the ν₂(a_g) mode of neutral molecule. Simultaneously, below 150 K we see the appearance of new lines at 1452, 1427, 1417, 1407 cm^–1 related to molecules with charge +1. The bands 1452 and 1427 cm^–1 should be assigned to the ν₂(a_g) and ν₃(a_g) modes of BEDT–TTF⁺¹ cation, respectively. The other two lines at 1417 and 1407 cm^–1 should be related to the ν₃(a_g) mode of BEDT–TTF⁺¹. Our Raman data show that the charge separation is much stronger than that one suggested previously by the X–ray investigations. The nearly disappearance of bands attributed to molecules with charge +0.5 and the existence of three lines related to the ν₃(a_g) mode of molecules with the charge +1 proves that not only the molecules D, but also other molecules in the unit cell should bear the charge +1. The existence of three separate ν₃(a_g) lines can be due to the following reasons: (1) different (BEDT–TTF⁺¹)₂ dimers are not equivalent; (2) different dimers are in non-equivalent positions in the unit cell; (3) different dimers possess different charge. The ν₂(a_g) and ν₃(a_g) Raman bands were found respectively in a salt with (BEDT–TTF^+0.9)₂ dimers at 1455 and 1425 cm^–1 [17] and in salts with (BEDT–TTF⁺¹)₂ dimers at 1460 and 1418 cm^–1 [18]; or 1451 and 1408 cm^–1 [19] (room-temperature data).

The Raman bands of the Fe^III salt (Fig. 2) are much broader in comparison to those recorded for the Co^III salt. This difference can be related to a larger degree of disorder in the Fe^III crystals; e.g. this disorder can be created by (Et₄N)⁺ cations. As results from the crystallographic investigations, the (Et₄N)⁺ cations show a great disorder, which was not determined even at low temperature [11, 12]. In the room temperature Raman spectrum of Fe^III salt (Fig. 2), we observe also two bands at 1497 and 1466 cm^–1, attributed to the ν₂(a_g) and ν₃(a_g) modes of BEDT–TTF^+0.5, respectively. The spectral modifications due to charge ordering are analogous to those observed for the Co^III salt. An important difference is that the modifications in the Fe^III salt are not directly related to the phase transition at 150 K, but they grow gradually on decreasing temperature below about 240 K. In the Fe^III spectra, we observe also the appearance of new lines assigned to BEDT–TTF⁺¹ cations: ν₂(a_g) = 1453 cm^–1 as well as ν₃(a_g)=1421 cm^–1 and 1414 cm^–1. Similarly to the case of the Co^III salt, the splitting of the ν₃(a_g) band suggests an existence of different BEDT–TTF dimers consisting of molecules with the charge +1. Nevertheless, in the Fe^III salt the line 1466 cm^–1, attributed to the ν₃(a_g) mode of BEDT–TTF^+0.5, does not disappear even at lowest temperatures, providing thus evidence that the charge separation is not complete.

Fig. 3 displays temperature dependence of the reflectance spectra of Co^III crystals over the entire spectral range for two perpendicular polarisations, corresponding to the maximum (E_||) and minimum (E_⊥) of reflected energy. For polarisation E_|| the electrical vector was parallel to the direction [110]. Conductivity spectra obtained by Kramers–Kronig transformation of the reflectance data are shown in Fig. 4. These spectra are very similar to the reflectance spectra of other κ–phase BEDT– TTF crystals [20–22].

In the room-temperature spectra of the Co^III salt, we observe a broad mid-infrared electronic band with maximum at about 2200 cm^–1 for E_|| and 3100 cm^–1 for E_⊥, corresponding to charge transfer transition between charged and neutral BEDT–TTF molecule (CT₂), as well as several vibrational features below 1400 cm^–1. The phase transition at 240 K causes some modifications of the Co^III spectra, similarly to those observed previously for the Fe^III crystals [13]. However, the most important changes occur as a result of the phase transition at 150 K. Below 150 K, the mid-infrared electronic band decreases its intensity, changes shape and shifts towards higher frequency (Fig. 4). This is a consequence of opening of a new energy gap. By fitting the reflectance spectra with a Drude–Lorentz dielectric function (as described previously for Fe^III salt [13, 14]), we estimated the energy gaps (Δ) for two perpendicular directions: at 160 K, Δ_|| = 0.27 eV and Δ_⊥ = 0.40 eV, whereas at 120 K Δ_|| = 0.30 eV and Δ_⊥ = 0.48 eV. However, the most striking spectral change is the appearance of a new near-infrared electronic band with maximum at about 7000 cm^–1 for E_|| and 7200 cm^–1 for E_⊥, related to charge transfer between BEDT–TTF⁺¹ cations (CT₁), providing thus a proof of existence of (BEDT–TTF⁺¹)₂ dimers below 150 K.

Above 150 K, for both polarisations, the IR spectra of Co^III salt within the region of strong vibrational bands (Fig. 5) are very similar to those of other κ–phase BEDT–TTF salts [20–22]. The assignment of main vibrational features in the Co^III spectra is carried out on the basis of references [16, 21–23]. For polarisation E_||, we observe a broad and strong band in the region 1130–1230 cm^–1 related to the ν₃(a_g) mode of BEDT–TTF, activated by coupling with the charge transfer transition CT₂. An antiresonance dip at about 1170 cm^–1 is due to the ν₆₇(b_3u) mode. A doublet at 1286 and 1299 cm^–1 is attributed to the ν₅(a_g) mode (CH₂ vibrations), whereas a strong band at 875 cm^–1 is attributed to the activated ν₆₀(b_3g) mode, according to the proposition made in [23]. The charge ordering at 150 K induces considerable modifications of the vibrational bands. Below 150 K the spectral feature related to the ν₃(a_g) mode has weaker intensity and lies in the higher-frequency region 1210–1270 cm^–1; the intensity of this band is nearly temperature independent. Such a strong effect, together with the Raman data, suggest that (BEDT–TTF^+0.5)₂ dimers are nearly absent; therefore, the 1210–1270 cm^–1 band can be associated with coupling of the ν₃(a_g) mode of charge-poor molecule with charge transfer between neighbouring dimers (CT₂-type transition). For the mode ν₆₇(b_3u), we can see a band at 1167 cm^–1 instead of the dip. The band ν₆₀(b_3g) reduces strongly its intensity, confirming thus an interaction of the ν₆₀(b_3g) mode with the CT₂. Additionally, below 150 K we observe a new band at 1345 cm^–1, related to the ν₃(a_g) mode of BEDT–TTF⁺¹ cation, which is activated by coupling with the charge transfer transition CT₁, as well as relatively weaker lines at about 1397 and 1449 cm^–1, related to the ν₄₅(b_2u) and ν₂₇(b_1u) modes of BEDT–TTF⁺¹, respectively. However, it is also possible that the band at 1449 cm^–1 is due to the ν₂(a_g) vibration. In the IR spectra of a salt containing (BEDT–TTF⁺¹)₂ dimers, the ν₂₇(b_1u), ν₂(a_g), ν₄₅(b_2u), ν₃(a_g) bands were found at 1454, 1452, 1408, 1343 cm^–1, respectively [18].

Above 150 K, in the Co^III spectrum for polarisation E_⊥ (Fig. 5), we observe strong bands at 1308 and 1242 cm^–1 related to the ν₃(a_g) and ν₅(a_g) modes of BEDT– TTF^+0.5, respectively, and the band ν₆₀(b_3g) at 881 cm^–1. The phase transition has a strong influence on these bands: the band ν₃(a_g) is observed at 1314 cm^–1 and it is of weaker intensity; a spectral feature consisting of three lines (1250, 1259 and 1267 cm^–1) is related to the ν₅(a_g) mode; the band ν₆₀(b_3g) gets much weaker. However, for polarisation E_⊥, the most important spectral change is the appearance of a relatively strong band at 1349 cm^–1 as well as a weak band at 1398 cm^–1, respectively, related to the ν₃(a_g) and ν₄₅(b_2u) modes of BEDT–TTF⁺¹ cation.

Due to the phase transition at 150 K, the infrared spectra of the Fe^III salt undergo analogous modifications to those observed for the Co^III salt (Figs. 6 and 7). In the vibrational spectra of the Fe^III salt (both above and below 150 K), we find nearly all the spectral features described above for the Co^III salt: for example, below 150 K we observe the bands at 1342 cm^–1 (E_||) and 1344 cm^–1 (E_⊥), related to the ν₃(a_g) mode of BEDT–TTF⁺¹ cation. It is important to emphasise, that in the case of Fe^III salt, the spectral modifications are directly related to the phase transition at 150 K, although, as shown by Raman experiment, the charge redistribution occurs gradually below 240 K. For example, for polarisation E_||, for bands related to the ν₃(a_g) mode, the disappearance of the spectral feature at 1170–1230 cm^–1 and the appearance of a peak at 1227 cm^–1 occurs at the phase-transition temperature 150 K (Fig. 7). Analogous conclusion results from the analysis of temperature dependence of electronic absorption [14]. Below 150 K, for the Fe^III salt, the spectral features 1227 cm^–1 (E_||) and 1308 cm^–1 (E_⊥), attributed to the ν₃(a_g) mode activated by a CT₂-type transition, are much stronger than those for the Co^III salt (see also Fig. 5), and considerably decrease their intensities when temperature goes down. Such behaviour can be related to a process of further charge redistribution intervening in the Fe^III salt below 150 K.

4 Discussion

The studied salts are isostructural, exhibit very similar physical properties and undergo analogous phase transitions at about 240 and 150 K, but the charge ordering inside conducting BEDT–TTF layers arises in different ways. The Raman spectra show that, in the Co^III salt, the charge redistribution occurs at the phase-transition temperature, i.e. 150 K. In this salt, above 150 K, the charge is distributed more or less uniformly among BEDT–TTF molecules (charge +0.5), while below 150 K most molecules are either neutral or possess a charge +1. On the other hand, in Fe^III salt the charge redistribution starts already below 240 K, then develops gradually on temperature lowering and is even further continued below 150 K. In this case, the phase transition at 150 K has no drastic influence on the Raman spectrum. Nevertheless, the charge redistribution in Fe^III salt seems to lead to analogous charge patterns as in the Co^III salt.

The differences between charge-ordering mechanisms in the studied compounds, as revealed by Raman experiment, are not directly observed in IR spectra. The IR electronic and vibrational features of both crystals change abruptly at the phase transition temperature T = 150 K. Such an effect should be expected in the Co^III salt, but is rather surprising in the Fe^III salt. We suppose that in the Fe^III salt, in the temperature range 240–150 K, the charge ordering is dynamical and does not lead to the formation of stable (BEDT–TTF⁺¹)₂ dimers. Such stable dimers are just created at 150 K, yielding modifications of the electronic and vibrational spectral features of the Fe^III salt. The phase transition at 150 K causes modifications of the electronic structure. On the other hand, the vibrational bands are mainly due to the effect of coupling of BEDT–TTF vibrations with electrons; therefore, they are considerably modified when charge transfer transition between BEDT–TTF⁺¹ cations is activated. Previous investigations of the ESR spectra and the magnetic susceptibility [10–12] have shown that the results for the Fe^III salt were similar to those for the Co^III salt, but the experimental signatures of the phase transitions for Fe^III salt were less clear and appeared in a wider temperature range. Our spectroscopic measurements confirm these observations.

Now we discuss possible reasons of the charge-ordering effect. The theoretical studies have shown that the insulating state of quarter-filled κ–phase BEDT–TTF salts is stabilised by the strong on-site Coulomb (U) interaction and strong dimerisation of conducting BEDT–TTF layers [24]. The presence of relevant values of the inter-site interactions (V) can be also expected in κ-phase salts; however, the stability of their insulating state is proved by the mean-field calculations by considering only U (without V). Consequently, the electronic states of these compounds can be considered by taking each dimer as a unit (dimer model). The important parameter of dimer model is the so-called on-site Coulomb interaction energy of the dimer (U_eff), which grows when t increases (t is intra-dimer transfer integral). Strong dimerisation leads to a phase with half-filled band, in which the Mott insulator is stabilised for strong electron correlations [24]. Consequently, in a typical κ-phase salt, the charge distribution is uniform among BEDT–TTF molecules and no charge-ordering phenomenon is expected.

Within the family of κ-phase BEDT–TTF salts the studied crystals κ–[Et₄N](BEDT–TTF)₄M(CN)₆·3 H₂O (M = Co^III, Fe^III) are exceptional, since the degree of dimerisation of BEDT–TTF molecules is relatively weak [12]. The analysis of the crystallographic data at T = 115 K shows that the intra-dimer interaction is weaker than that one in typical κ-phase salts; moreover, it is weaker than the interaction between molecules belonging to neighbouring dimers. Consequently, the parameters U_eff and V are respectively smaller and larger, in comparison with a typical κ–phase salt; therefore, the Mott insulting state can be unstable. In this situation, the inter-site Coulomb interaction V plays an important role and gives rise to the charge ordering.

Theoretically, it was argued that inter-site Coulomb interactions (V) can stabilise charge ordered states in organic conductors based on BEDT–TTF donor [3]. Therefore, we suggest that, in the studied salts at T = 150 K, the parameter V/t becomes large enough to induce charge redistribution. If we assume that the degree of disorder in the Fe^III salt is higher than in the Co^III salt (e.g. due to Et₄N⁺ cations), the larger distribution of t can be expected and the charge redistribution can occur at different temperatures in different unit cells of the crystal lattice. In this way, one can explain differences between the charge redistribution phenomena in both compounds. Nevertheless, it is important to emphasise that the phase transition in both salts occurs at 150 K.

Acknowledgements

This work was partially supported by KBN under SPUB–M (COST D14) and POLONIUM project No. 03273NK.