2.1 Samples

Potassium hexacyanoferrate(III) (Acros, puriss. p.a.), cobalt(II) nitrate (Acros, p.a.) and cesium (Aldrich, puriss. p.a.), rubidium (Aldrich, p.a.), potassium (Acros, p.a.) and sodium (Acros, p.a.) nitrates were used as received.

Synthesis conditions were adapted to the nature of the alkali cation. We observed that the affinity of the alkali cation for the inorganic network increases with increasing size. For this reason, the rubidium derivatives were prepared with diluted aqueous solutions, while potassium and sodium ones were synthesized using concentrated aqueous solutions.

The cesium derivatives were synthesized as previously described [25]. The rubidium derivatives were prepared using diluted solutions, by addition of a solution made of cobalt(II) nitrate and various amount of rubidium nitrate to a potassium hexacyanoferrate(III) solution. To obtain the potassium and sodium derivatives, the cobalt(II) nitrate solution was added to a concentrated solution made of potassium hexacyanoferrate(III) and various amounts of alkali cations salt always used in large excess.

In all the syntheses, the initial pH of the solutions was adjusted to 5 using HNO₃ diluted solutions. The addition rate was regulated to last 3 h. The powders were centrifuged, washed three times with distilled water and allowed to dry in air.

The X-ray powder diffraction patterns of all the compounds are consistent with a cubic system, face-centred network, schematised in Fig. 1.

The element analyses, related to the chemical formulae for an fcc conventional cell C_xCo₄[Fe(CN)₆]_(8+x)/3□_(4 –x)/3 are reported in Table 1. In the following, the compounds are named C_x, indicating the nature of the alkali cation C in the compound and its x amount per cell. The chemical formulae of compounds Cs_0.3, Cs_0.7, Cs_1.2, and compounds C₀ and Rb_1.8 given in reference [25] and in reference [24] are also reported. Taking into account the uncertainty in the stoichiometric coefficients (±0.1), all the formulae correspond to neutral samples. The insertion of x alkali cations (C⁺) is accompanied by the insertion of x/3 additional [Fe(CN)₆]^3– anionic entities (Fig. 1).

The oxidation states of the metallic cations at 20 K were quantitatively determined by X-ray absorption near edge spectroscopy (XANES) at the cobalt and iron K edges as previously described [25]. The edge spectra of the compounds containing a mixture of Co^III and Co^II ions were fairly well reproduced by linear combinations of spectra of two model samples Rb_1.8 and C₀, respectively composed of a majority of Co^III–Fe^II and Co^II–Fe^III pairs at 20 K [27]. Knowing the percentage of Co^II and Co^III ions in the models [25, 27], the percentage of Co^III ions in each compound was obtained (Table 2).

2.2 Instrumentation

Photomagnetic experiments were carried out with a Quantum Design MPMS-5S magnetometer working in the dc mode, equipped with a multi-wire silica optical fibre. The fibre was connected to a broadband source of light (tungsten halogen lamp, 100 W), through interferential filters (100-nm bandwidth). A cut-off filter provided a high intensity in the 750 ± 50 nm range. The power received by the irradiated sample is estimated to P = 60 mW cm^–2. The measurements were performed over the 10–300-K temperature range with an applied magnetic field of 500 G. In a first step, 3 mg of powder (pellet-shaped) was slowly cooled from 300 to 10 K to avoid quenching phenomena and then irradiated at 10 K during 4 h. The magnetic field was applied as soon as the irradiation began.

The temperature was raised above the Curie temperature and decreased again before measurement to get the Field Cooled Magnetisation (FCM) curve over the 10–300 K temperature range. The temperature was then raised above T = 200 K. At that temperature, all the samples relaxed and came back to their initial ground state, which allows us to measure then the magnetisation before irradiation (M_initial) over the 10– 300-K temperature range. To compare the samples, the amount of compound and the procedure were exactly the same. Excepted for the sodium derivatives, we checked that the wavelength range used for irradiation was the most efficient one.

Energy-dispersive X-ray diffraction patterns were recorded at 10 K over the 0.5–70 keV energy range on the DW11A energy-dispersive X-ray beam of the DCI ring of the LURE synchrotron facility at Orsay, France. The orientation of the detector was fixed at Θ = 5.8457° for compound Cs_0.7 and Θ = 4.003° for compound Na_1.8. The orientation was checked by measuring the X-ray pattern of a copper foil at room temperature. For the measurements, the powders were spread on between two tapes and slowly cooled from 300 to 10 K. At low temperature, the white X-ray beam induces the Co^III(LS)–Fe^II(LS) → Co^II(HS)–Fe^III(LS) photo-induced electron transfer. The X-ray source being the probe beam as well as the pumping beam makes sure that the whole diffracting sample is irradiated.

The study of the photo-excitation process was carried out at 10 K under continuous irradiation. X-ray powder diffraction patterns were recorded for increasing irradiation times (Figs. 2 and 3). For the first point, the acquisition time was 15 s. For subsequent points, the acquisition time was 30 s. During this time interval, the pattern change can be neglected. For clarity, the diffractograms in Figs. 2 and 3 are presented in the energy range containing the two most intense reflections.

3.1 Photomagnetism of the cesium derivatives

The photomagnetic properties of cesium derivatives were already discussed in a previous paper [25]. We showed that the photomagnetic effect depends on a compromise between the number of Co^III–Fe^II diamagnetic pairs and the network flexibility. Since then, the series has been completed and enables us to go further in the conclusions.

The magnetisation curves of compounds Cs_0.3–Cs_0.7 before and after irradiation over the 10–30-K temperature range are presented in Fig. 4. These compounds contain less than one alkali cation per cell. The corresponding magnetisation curves for compounds Cs_0.95–Cs₂, with one alkali cation per cell or more, are given in Fig. 5. The change in the photomagnetic properties along the series of caesium derivatives is discontinuous. The discontinuity appears around x = 1.

Before irradiation, Cs_0.3 is already ferrimagnetic, with a Curie temperature T_C = 17 K [25] (Fig. 4), in agreement with the weak amount of diamagnetic pairs at 20 K (Table 2). After irradiation, the magnetisation at 10 K and T_C slightly increase which means that the number of magnetic neighbours increases in the ferrimagnetic phase after irradiation. Co^III–Fe^II isolated diamagnetic pairs are photo-transformed into metastable (Co^II–Fe^III)* magnetic ones.

As the amount of cesium increases in the compounds [Cs_0.45–Cs_0.7], the initial magnetization at 10 K decreases. The Curie temperature also decreases, which means that the number of paramagnetic neighbours is decreasing in the sample. However, the magnetisation and the Curie temperature after irradiation at 10 K are approximately the same in the three compounds, close to the ones of the alkali cation free sample. Co^III–Fe^II isolated diamagnetic pairs are nearly all photo-transformed into metastable (Co^II–Fe^III)* magnetic ones. The determining factor of photomagnetism is the number of diamagnetic pairs present at low temperature before irradiation.

Cs_0.95 is weakly paramagnetic before irradiation (Fig. 5) in agreement with the important amount of Co^III–Fe^II diamagnetic pairs at low temperature before irradiation (Table 2). After irradiation, the compound becomes ferrimagnetic below a Curie temperature slightly above the ones of the previous compounds (Cs_0.3–Cs_0.7), but the magnetisation at 10 K is far less than the magnetisation at 10 K in the three previous compounds. The Curie temperature being proportional to the amount of magnetic neighbours, its value in irradiated Cs_0.95 shows that the photo-induced ferrimagnetic phase is mainly composed of Co^II–Fe^III* metastable pairs and not of Co^II–Fe^III* pairs diluted in Co^III–Fe^II diamagnetic ones, which would lead to a lower Curie temperature. The magnetisation at 10 K far less than in the previous compounds indicates that only limited domains of diamagnetic Co^III–Fe^II pairs are photo-transformed into limited domains of metastable (Co^II–Fe^III)* magnetic pairs. When the amount of alkali cations increases (Cs_1.2–Cs₂), the magnetisation curve before irradiation is approximately the same for these two compounds in agreement with a same amount of diamagnetic pairs at low temperature (Table 2). Only the magnetisation curves after irradiation varies. As the amount of alkali cation increases, the magnetisation at 10 K after irradiation progressively decreases accompanied by a progressive loss of the magnetic order. However, the Curie temperature of the ferrimagnetic contribution remains close to the one of all the other compounds which indicates again the formation of domains mainly composed of Co^II–Fe^III* pairs. The small magnetisation value at 10 K is due to the limited number and/or size of the photo-induced Co^II–Fe^III* domains. We ascribe the loss of photomagnetic properties to the rigidity of the inorganic network. The rigidity prevents the lengthening of the cobalt–ligand bonds (about 0.15 Å) accompanying the photo-induced electron transfer from Co(III)(low spin)–Fe(II)(low spin) to Co(II)(high spin)– Fe(III)(low spin) and the single occupancy of the two antibonding e_g* orbitals of cobalt(II) [12, 25]. The increase of the rigidity of the network is related to the increasing number of [Fe(CN)₆] filling the vacancies when the amount of alkali cations increases. It may also be related to the number and to the size of cesium cations since their interactions with the Co₄[Fe(CN)₆]_(8+x)/3 network may rigidify the structure. To summarize, for the compounds containing more than one alkali cation per cell, domains of diamagnetic Co^III–Fe^II pairs are photo-transformed into domains of metastable (Co^II–Fe^III)* magnetic entities. Photomagnetism is limited by the network flexibility.

The study evidences a discontinuity of the photomagnetic properties around x = 1. This discontinuity around x = 1 is not specific to the photomagnetic properties of the compounds. It also appears in the number of diamagnetic pairs per cell for example. When x < 1, the amount of diamagnetic pairs per cell is proportional to the number of alkali cation per cell. It is no more the case for x ≥ 1 (Table 2). The stoichiometry C₁Co₄[Fe(CN)₆]₃□₁ (x = 1) is very particular. For this stoichiometry, the number of alkali cation and the number of [Fe(CN)₆] vacancies per cell are exactly the same. For x < 1, the alkali cations are diluted, whereas the vacancies are concentrated in the compounds. The flexibility is important so that the determining factor is the number of diamagnetic pairs. For x > 1, on the contrary, the alkali cations are now concentrated, whereas the vacancies are diluted in the compounds and the determining factor becomes the network flexibility.

3.2 Photomagnetism of other alkali cations derivatives

To compare the photomagnetic properties of cesium and other alkali cations derivatives, it is necessary (i) to get C_xCo₄[Fe(CN)₆]_(8+x)/3 compounds with exactly the same amount x of different alkali cations C and (ii) to quantitatively determine the amount of photo-transformed pairs. In order to compare the photomagnetic properties of the compounds, we define the photomagnetic efficiency E as (M_final – M_initial)/M_initial (M_initial and M_final are the magnetisations at 10 K before and after irradiation respectively). This so-defined E evidently involves a part related to the pure photo-excitation effect, i.e. the creation of magnetic pairs and a part related to the magnetic response of the sample to the external field. In magnetically ordered phase, the magnetic response depends on the applied magnetic field in a non-linear way. In the present study, we propose to use E to compare, in the same experimental conditions, the photomagnetic response of compounds C_xCo₄[Fe(CN)₆]_(8+x)/3. For a given value of x, only the nature of C varies. If the photomagnetic properties do not depend on the nature of C, very close values of E are expected. On the contrary, if the photomagnetic properties depend on the nature of C, different value of E will arise. The photomagnetic efficiency E is reported in Fig. 6 versus the amount of alkali cations per cell for all the compounds.

When the amount of alkali cation per cell is less than one (x < 1), E increases with the amount of alkali cations, independently of the alkali cation nature. As along the caesium derivatives series, the amount of Co^III–Fe^II diamagnetic pairs increases with the amount of alkali cations per cell. For close values of x, the number of diamagnetic pairs at 20 K before irradiation is also close and independent on the nature of the alkali cation (Table 2). The Co^III–Fe^II isolated diamagnetic pairs are photo-transformed into metastable (Co^II–Fe^III)* magnetic ones. The inorganic network contains a large number of [Fe(CN)₆] vacancies (> 1 per cell) and few alkali cations (< 1 per cell). It is flexible enough to allow the photo-transformation of nearly all the diamagnetic pairs. Like in the case of the caesium derivatives, the photomagnetism is a compromise between the number of diamagnetic pairs and the network flexibility. The determining factor is the number of diamagnetic pairs existing at low temperature before irradiation.

When the amount of alkali cations per cell is larger than one (x ≥ 1), E presents a much more complex variation, not only with the amount, but also with the nature of the alkali cation. This behaviour underlines that numerous parameters are involved in the photomagnetic properties. For a same alkali cation, E decreases with the cation insertion rate x, as in the caesium derivatives series. The decrease is due to the structural flexibility loss. It is noticeable that the ratios E of the compounds Na_1.8, Rb_1.8 and Cs₂, which contain the same amount of diamagnetic pairs at low temperature before irradiation and have very close stoichiometries (i.e., the same amount of [Fe(CN)₆] vacancies and the same average cobalt ion coordination sphere), are very different: E decreases when the size of the alkali cation increases. The result emphasizes the role of the alkali cation on the network flexibility for a constant number of the [Fe(CN)₆] vacancies. To be brief, the bigger the cation is, the less flexible the structure is.

3.3 Structural transformation of the Cs_0.7 and Na_1.8 derivatives under irradiation

To go further in the understanding of the photo-excitation process, two compounds were selected for X-ray diffraction under continuous irradiation at 10 K: Cs_0.7 with x < 1 contains a rather high amount of diamagnetic pairs at low temperature, whereas Na_1.8 with x > 1 still presents enough network flexibility. The diffraction patterns are shown for increasing irradiation times in Fig. 2 for Cs_0.7 and in Fig. 3 for Na_1.8.

At 10 K, at the very beginning of irradiation, the patterns of both compounds are characteristic of a unique fcc phase.

For Cs_0.7, during the excitation process, only one peak appears for each reflection. When the irradiation time increases, the peaks are shifted towards lower energy. After fitting the peaks to a Gaussian line shape, it is possible to conclude that there is no significant line width or intensity change. If two phases with different cell parameters were existing, the line would be broadened. The cell parameter increases. The behaviour is characteristic of a continuous second order transformation within the same phase. Co^III–Fe^II diamagnetic pairs with short Co– ligand bonds are locally photo-transformed into (Co^II–Fe^III)* pairs with long Co– ligand bonds within the same crystallographic phase. This structural observation is in agreement with the conclusions of the photomagnetic study. The initially ferrimagnetic phase remains ferrimagnetic, only a few randomly distributed Co^III–Fe^II diamagnetic pairs with short Co to ligand bonds are locally photo-transformed into (Co^II–Fe^III)* magnetic ones through a second-order transformation.

Instead, for Na_1.8, after 10 min of irradiation, each reflection on the diffraction pattern splits into two peaks. A new peak appears at lower energy, the intensity of which increases with the irradiation time. The changes indicate the appearance and the growth of a new fcc phase with a longer cell parameter. Domains of Co^III–Fe^II pairs with short Co-ligand bonds are photo-transformed into domains of metastable Co^II– Fe^III* pairs with long cobalt-ligand bonds. This behaviour is characteristic of a first-order transition, in agreement with the conclusions of the photomagnetic study: limited diamagnetic domains mainly composed of Co^III–Fe^II pairs (with a short cell parameter) are photo-transformed into ferrimagnetic ones, mainly composed of Co^II–Fe^III* pairs (with a long cell parameter). Na_1.8 contains a majority of Co^III– Fe^II diamagnetic pairs at low temperature before irradiation. Thus, the Co–ligand bond lengthening accompanying the photo-induced electron transfer leads to strong elastic interactions between the excitable species and cooperative effects and the structural effects are evidenced by our diffraction data (Fig. 3). The associated magnetic consequences appear in the abrupt spontaneous, first-order transition, with thermal hysteresis observed by Hashimoto et al. [28].