2.1 A photo-induced electron transfer

The proposed explanation for the origin of this phenomenon was the presence of diamagnetic pairs Co^III–Fe^II in the compound. A photo-induced electron transfer from Fe^II to Co^III through the cyanide bridge should transform the diamagnetic ions in paramagnetic ones, leading to the increase of the magnetization of the compound [5, 13]. Then, the key of the phenomenon would be the presence of diamagnetic pairs in the compound.

We synthesized a photomagnetic CoFe Prussian blue analogue (Rb_1.8). Its structure and the magnetization curves before and after irradiation with red light are reported in Fig. 1.

To characterize the electronic structure of the cobalt and the iron in Rb_1.8, we recorded X-ray absorption spectra at the L_2,3-edges of Co and Fe ions before irradiation and in the photo-induced magnetic metastable state [14]. L_2,3 edges of 3d transition metals involve symmetry allowed electric dipole transitions from 2p core electrons to incompletely filled 3d and 4s levels (2p⁶4s⁰ → 2p⁵4s¹ transitions can be neglected due to their low intensity relative to the 2p⁶3dⁿ → 2p⁵3dⁿ⁺¹ ones). They consist of multiplet structures due to the 2p⁵3dⁿ⁺¹ excited-state configuration. The position of the L_2,3-edges multiplet structures depends the final state energies arising from Coulomb and exchange interactions within the 3d shell and between 2p and 3d shells, from spin-orbit interaction on the 2p and 3d shells and from crystal field on the 3d shell. So, L_2,3-edges directly probe the 3d levels of each ion involved in the proposed electron transfer. The XAS results are reported in Fig. 2.

Comparing the cobalt L_2,3-edges with those of reference compounds (Fig. 2a), we found that the ground state of the compound, at room temperature, is a mixture of Co^II [HS, t_2g⁵e_g²] and Co^III [LS, t_2g⁶e_g⁰] species. The spectrum obtained after irradiation displays a decrease of the signal of the LS Co^III species and a simultaneous increase of the signal of the HS Co^II species. The light induces the transformation of Co^III in Co^II ions. At the iron L_2,3-edges (Fig. 2b), before irradiation, the presence of the feature at 704.5 eV is the signature of LS Fe^III entities, in line with the presence of Co^II entities evidenced at the Co L_2,3-edges. After irradiation, the intensity of the 704.5 eV peak increases as the LS Fe^II to LS Fe^III transformation proceeds, in line with the Co L_2,3-edges changes.

To the best of knowledge this is the first experimental local and direct evidence on the two metallic sites of a photo-induced metal-to-metal electron transfer in a three-dimensional compound:

2.2 Exchange interaction in the metastable state

The direct determination of the exchange interaction between paramagnetic ions is often trivial for stable bimetallic magnetic A–B compounds with known structure, stoichiometry and mass. A minimum in the χ_M T vs T curve shows ferrimagnetism and the magnetization at saturation S_T allows the confirmation of the nature of the coupling (S_T = |S_A – S_B|). For several reasons, in the photo-induced metastable sample, this clear-cut macroscopic characterization is not possible [15].

X–ray Magnetic Circular Dichroism (XMCD) is an element and orbital selective magnetic probe that has recently been developed with synchrotron radiation [16]. It has already been used to investigate the local magnetic moments in magnets of the Prussian-blue family [17]. To record the XMCD spectra, the magnetic field (H = 1 T) was applied alternatively parallel and anti-parallel to the direction of the photon beam. A first spectrum labelled σ_↑↑ was registered with the magnetic field parallel to the propagation vector of the photons. Then a second spectrum, labelled σ_↑↓, was recorded with the magnetic field applied in the opposite direction. The XMCD signal is the difference (σ_↑↑ – σ_↑↓) between the two spectra. The area of the XMCD signal is directly proportional to the local magnetic moment carried by the absorber atom.

We report here XMCD results at the K edges of cobalt and iron for two compounds: K_0.1Co₄[Fe(CN)₆]_2.7·18 H₂O (C₀), where the exchange interaction between the cobalt and iron ions is known to be antiferromagnetic [20] leading to ferrimagnetism; and Rb_1.8Co₄[Fe(CN)₆]_3.3,13H₂O (Rb_1.8), which presents a large photomagnetic effect. The isotropic X–ray absorption near-edge spectra (XANES) and the dichroic signal (XMCD, × 400) at the K edges of cobalt and iron are displayed in Fig. 3.

A weak dichroic signal is present at the Co and Fe edges in both compounds. The energy of the signal corresponds to the allowed transition to the p–symmetry levels of the metal (7725 eV at the Co K edge, 7130 eV at the Fe K edge). For C₀Fe, the macroscopic magnetization data are well known and show clearly that the spins of cobalt(II) (d⁷, high spin, S = 3/2) and iron(III) (d⁵, low spin, S = 1/2) are antiferromagnetically coupled, as shown in the insert of Fig. 3a and b [20]. The inversion of the dichroic signal from cobalt to iron is a local characterization of the antiferromagnetic coupling between cobalt and iron ions. In the photo-induced metastable state of Rb_1.8, below the Curie temperature, the dichroic signal is positive at the cobalt K edge and negative at the iron K edge, as in C₀Fe. These two XMCD observations mean that in the photo-induced magnetic state, there is an antiferromagnetic coupling between Co^II and Fe^III ions, which leads to ferrimagnetism.

This is the first experimental evidence of the ferrimagnetic nature of the metastable state for the Rb_1.8 compound.

2.3 Local structural changes accompanying the electron transfer

The photo-induced transformation of Co^III (LS) into Co^II (HS) must be accompanied by a Co-ligand bond lengthening. Extended X–ray Absorption Fine Structures (EXAFS) is well adapted to evidence such local structural change. The quantitative analysis of the Co K–edge EXAFS signal recorded before irradiation gives 80% of d_Co–L = 1.91 Å distances and 20% of d_Co–L = 2.08 Å ones [14]. The short distance is characteristic of Co^III low spin entities, the second one is close to the ones obtained in Co^II high-spin species. That means that before irradiation, the compound contains 80% of LS Co^III. After irradiation, the two distances remain the same but we found 60% of Co atoms with long Co–L distances. That means that 50% of Co^III–Fe^II diamagnetic pairs have been transformed in Co^II–Fe^III paramagnetic ones.

2.4 Long-range structural changes accompanying the electron transfer

The expansion of the cobalt coordination sphere during the excitation process [14] is expected to induce long-range order changes. In order to study these changes, we performed Energy Dispersive X–ray powder diffraction experiments with synchrotron radiation at low temperature (10K) under continuous irradiation during the excitation process [18].

The X–ray diffraction patterns of Rb_1.8 under irradiation at 10 K over the 16–27 keV energy range are displayed for increasing irradiation time in Fig. 4.

Increasing irradiation time produces for each diffraction line, the growth of a new peak shifted to lower energies. After 1 h under irradiation, the initial peak has almost disappeared and the lower energy peak only remains. The photo excitation is achieved.

The new peak at lower energy is the signature of a new fcc phase with a longer cell parameter.

The presence of two well-resolved peaks reflects the existence of two kinds of diffracting domains. The domains forming the phase with the shortest cell parameter (S phase) are mainly composed of Co^III(LS) ions with short cobalt to ligands bonds. The others form the phase with the longest cell parameter (L phase) mainly composed of Co^II(HS) ions with long cobalt-to-ligand bonds. These observations are characteristic of a first order i.e. a discontinuous transition. In a three-dimensional compound mainly composed of photo-excitable Co^III–Fe^II diamagnetic pairs, the interactions between the excitable species are strong leading to cooperative effects and the formation of L domains through a S → L discontinuous transition. To our knowledge, this is the first structural study carried out in the course of a photo-induced electron transfer.

3.1 Diamagnetic pairs

The stronger the ligand field around the cobalt ion is, the more stabilized the Co^III(LS)–Fe^II(LS) state is. To form Co^III(LS)–Fe^II(LS) diamagnetic pairs, one has then to increase the cobalt ligand field strength. For that, the overlap between the cobalt ion and the ligands orbitals has to be enhanced. In the fcc structure of Prussian blue analogues, two ways are possible: (i) the substitution of weak field ligands by stronger field ones [13, 19]; (ii) for the same ligands around the Co ions, the modification of the geometry of the Co ion coordination polyhedron.

The synthesis of CoFe Prussian blue analogues involves the substitution of water molecules of [Co^II(H₂O)₆] by [Fe^III(CN)₆] complex ligands. This means a progressive increase of the Co ligand field when the oxygen atoms are replaced by nitrogen ones coming from NC ligands. For a sufficient amount of nitrogen atoms around the Co^II ion, Co^II can become low spin and its reducing power is increased leading to the stabilisation of the Co^III state. A stable Co^III–Fe^II pair is then quickly formed through a chemically induced electron transfer. In order to increase the number of diamagnetic pairs responsible for the photo-induced effect, one has then to increase the number of nitrogen atoms around the Co ion.

By varying the nature of the alkali cation present in excess in solution during the synthesis, we introduced various quantities of alkali cations inserted in the structure [19]. Three CoFe Prussian blue analogues were thus synthesized, in which the environment of the cobalt atoms varies. The cobalt ion is surrounded by an average of four nitrogen and two oxygen atoms CoN₄O₂ in the lacunary structure Co₄[Fe(CN)₆]_2.7·18 H₂O (C₀), six nitrogen atoms CoN₆ in the compact structure Cs_3.9Co₄[Fe(CN)₆]_3.9·13 H₂O (Cs_3.9) and five nitrogen atoms CoN₅O₁ in the intermediate structure Rb_1.8Co₄[Fe(CN)₆]_3.3·13 H₂O (Rb_1.8) previously described. The schematic conventional cell representations and the magnetization curves before and after an irradiation of four hours under red light are reported in Fig. 5.

Compound C₀ is mainly composed of Co^II–Fe^III magnetic pairs. It is ferrimagnetic before irradiation with a T_c at 16 K [20]. No effect of the light is observed in its magnetization curve (Fig. 5). It confirms that the presence of diamagnetic pairs in the compound is compulsory to observe a photo-induced magnetization. Compound Rb_1.8 is essentially diamagnetic before irradiation. The compound is built of a majority of diamagnetic Fe^II–CN–Co^III pairs where the two d⁶ ions are low spin (t_2g)⁶. After irradiation, an important photo-induced magnetization appears. The magnetization is multiplied by 10 (Fig. 5). Compound Cs_3.9 is composed of a large number of diamagnetic pairs Fe^II–NC–Co^III. It is nearly a perfect fcc structure with caesium cations in half of the tetrahedral sites. As expected, it is diamagnetic before irradiation. The effect of light is very weak and at a first glance, this result is surprising (Fig. 5). The behaviour of compound Cs_3.9 clearly evidences that the presence of diamagnetic pairs in the compounds is not a sufficient condition to observe the photomagnetic effect. The phenomenon apparently disappears when too many diamagnetic pairs are present. Remember that the photo-induced electron transfer is accompanied by an increase of the bond lengths in the coordination sphere of the cobalt by 0.18 Å. To be able to absorb the dilatation of the Co–N distances and the related increase of the cell parameter, the inorganic network needs to be flexible. In the Rb_1.8 lacunary structure, the iron vacancies are filled with water molecules coordinated to the cobalt(II) and loosely hydrogen-bonded to other water molecules and behave as relaxation points of the network strains. In the Cs_3.9 compact structure, the relaxation sites are absent. The framework is more rigid so that the excited state never reaches its equilibrium distance and relaxes immediately.

3.2 Network flexibility

In order to correlate the relative amounts of [Fe(CN)₆] vacancies □ and of diamagnetic pairs to the photo-induced magnetization, we synthesized an original series of homogeneous Prussian blue analogues in which we tune the amount of caesium cations x in the tetrahedral sites of the structure from 0 to 4 per conventional cell [21]. This chemical insertion is accompanied by the insertion of [Fe(CN)₆] anions, which produces a progressive increase of the cobalt ligand field, which is therefore tunable in compounds of analogous structure. At 300 K, the progressive increase of Δ_Co along the series produces the spin change of an increasing amount of cobalt atoms accompanied by the Co^II–Fe^III → Co^III–Fe^II electron transfer. Along the series, the efficiency of the photo-induced process first increases (0 ≤ x ≤ 1) essentially due to the increasing amount of Co^III–Fe^II diamagnetic excitable pairs, and then decreases (1 ≤ x ≤ 4) due to the decreasing amount of [Fe(CN)₆] vacancies □ reducing the network flexibility. The phenomenon depends, independently of the nature of the alkali cation, on a compromise between the amount of Co^III–Fe^II diamagnetic pairs and the amount of [Fe(CN)₆] vacancies □ providing the network flexibility, which allows the trapping of the photo-induced metastable state. This family of compounds unambiguously evidenced the fundamental role of the intrinsic [Fe(CN)₆] vacancies □, which can also be of interest in the dynamics of the magnetization not dealt here [22, 23]. The thermal relaxation temperature of the metastable photo-induced states, defined as the temperature at which the magnetization of the sample becomes the same as the one before irradiation, decreases when the amount of alkali cation per cell increases (160 ± 5 K for Cs_0.7 and 140 ± 5 K for Cs_1.2).

The changes of the photomagnetic properties of the compounds along the series may be schematised by the evolution of the potential wells describing the ground and the metastable excited state. Schematic representations of the mean ground and metastable states wells for compounds C₀, Cs_0.7, Cs_1.2 and Cs_3.9 are shown in Fig. 6.

It is well known in the field of spin crossover systems that an increase of the ligand field strength, here the average ligand field around the cobalt atom, essentially produces an energy shift of the potential wells of the ground and of the metastable excited states. A stronger Δ_Co stabilizes the Co^III–Fe^II ground state and, on the contrary, destabilizes the Co^II–Fe^III photo-induced metastable state. In C₀, the cobalt ligand field is weak, the ground state is the Co^II–Fe^III state whatever the temperature. On the contrary, in Cs_3.9, the cobalt ligand field is strong, the ground state is the Co^III–Fe^II state whatever the temperature.

Once the Co^III–Fe^II state is stabilized in comparison with the Co^II–Fe^III one, a stronger Δ_Co should increase the energy gap between the two states and lower the thermal energy barrier from the photo-induced metastable state to the ground state.

In a three-dimensional compound undergoing strong network strains, the reduction of the network flexibility, i.e. the decrease of the number of [Fe(CN)₆] vacancies □, may prevent the metastable excited state to reach the strain less equilibrium position. The Co^II–L bonds in the excited state would be shorter. The flexibility reduction would then translate the metastable excited state potential to the left towards the ground state one and narrow it due to the reduction of the Co–L bonds vibration amplitude. The flexibility reduction would then lower the thermal energy barrier from the metastable photo-induced state to the ground state.

Thus, the relaxation temperature of the metastable photo-induced state should be lowered by an increase of Δ_Co or by a reduction of the network flexibility. This is what is experimentally observed. The energy barrier between the ground Co^III–Fe^II state and the Co^II–Fe^III metastable state in C_3.9 (Fig. 6) is probably so low that the lifetime of the latter is very short and can not be trapped. The behaviours of C_0.7 and C_1.2 are intermediate. As expected, the energy barrier is lower in C_1.2 than in C_0.7.

The discussion in term of ligand field strength and network flexibility only is of course simplified and doesn’t take into account the entropy of the solid linked to long-range order and amount and distribution of vacancies and alkali cation, which may play an important role particularly in cooperative effects. The data shown here does not allow going further on that point but the work is in progress. However, it is noticeable that the compounds exhibiting the most important photomagnetic effects are not the most ordered.