Combined, Modulation Enhanced X-ray Powder Diffraction and

Dec 8, 2010 - ... were further processed by PSD to determine in-phase angles in the phase domain. ..... Crystallographic information files of low-spin...
0 downloads 0 Views 4MB Size
Download PDF
J. Phys. Chem. C 2011, 115, 1323–1329

1323

Combined, Modulation Enhanced X-ray Powder Diffraction and Raman Spectroscopic Study of Structural Transitions in the Spin Crossover Material [Fe(Htrz)2(trz)](BF4)† Atsushi Urakawa,*,‡ Wouter Van Beek,§,| Marı´a Monrabal-Capilla,⊥ Jose´ Ramo´n Gala´n-Mascaro´s,‡,# Luca Palin,| and Marco Milanesio| Institute of Chemical Research of Catalonia (ICIQ), AV. Paı¨sos Catalans, 16, E-43007, Tarragona, Spain, Swiss-Norwegian Beamlines, ESRF, BP 220, F-38043 Grenoble Cedex, France, Dipartimento di Scienze e Tecnologie AVanzate and Nano-SiSTeMI Interdisciplinary Centre, UniVersita` del Piemonte Orientale “A. AVogadro”, Viale T. Michel 11, I-15121 Alessandria, Italy, Instituto de Ciencia Molecular (ICMOL), UniVersidad de Valencia, Catedra´tico Jose´ Beltra´n, 2, E-46980, Paterna, Spain, and Catalan Institution for Research and AdVanced Studies (ICREA), Passeig Llus Companys, 23, E-08010, Barcelona, Spain ReceiVed: July 31, 2010; ReVised Manuscript ReceiVed: October 25, 2010

The structure of [Fe(Htrz)2(trz)]BF4 (1, Htrz )1,2,4-4-H-triazole, trz ) 1,2,4-triazolate) at the low-spin (LS) and high-spin (HS) states and structural transitions between the two states were investigated by in situ highresolution synchrotron X-ray powder diffraction (XRPD) combined with Raman spectroscopy using a modulation-enhanced technique. The crystal structures of the LS and HS states were determined. A 1D chain structure of 1 at both LS and HS states was proven, and the lattice expansion upon LS-HS transition was mainly caused by the elongation of the chain. The differences in the behavior of the spin transition observed by XRPD and Raman spectroscopy were explained by the local sensitivity of the two different techniques and also by the spatial propagation of spin crossover phase transition within the crystallite and the body of the grain. Moreover, we demonstrated that the two-dimensional correlation analyses facilitate (i) understanding the data obtained by combined techniques, (ii) clarifying correlation between the signals gained by the different probes, and (iii) extracting information on temporal evolution of transformation processes. Introduction Spin crossover (SC) compounds1 are a paradigmatic example of bistable materials. In such metal complexes, the electronic ground state corresponds to the low-spin (LS) configuration, but the low-lying metastable high-spin (HS) state can be populated through external stimuli:2 thermally, by pressure, or by light irradiation. Thermally, the spin transition usually occurs over a wide temperature range, as the HS state is gradually populated. In some special cases, due to intermolecular interactions in the solid state, the spin transition can be abrupt: all metal centers pass to the HS state cooperatively in a very narrow temperature range, when the electronic transition triggers a crystallographic phase transition. In a few of these materials, the spin transition shows thermal hysteresis between the heating process (LS to HS) and the cooling process (HS to LS), creating a true memory effect. The electronic transition accounts for several changes at the molecular level, affecting the metal-ligand distances and optical properties (e.g., UV-vis absorption), in addition to magnetic properties. All these possibilities have postulated SC systems as smart materials for information storage, molecular switching, sensing, and medical applications.3 Among these examples, the iron(II) triazolate (trz) family of 1D chain compounds is the most striking materials reported to date.4 They have a general formula [Fe(trz)3]X2, and depending †

Part of the “Alfons Baiker Festschrift”. * Corresponding author. Tel.: +34 977 920 200. Fax: +34 977 920 823. E-mail: [email protected]. ‡ Institute of Chemical Research of Catalonia (ICIQ). § Swiss-Norwegian Beamlines. | Universita` del Piemonte Orientale “A. Avogadro”. ⊥ Universidad de Valencia. # Catalan Institution for Research and Advanced Studies (ICREA).

on the nature of the triazole derivative and of the anion X, a wide thermal hysteresis can be observed near room temperature. The origin of the hysteresis loop of these materials is not fully understood. It could reside essentially in the intramolecular interactions in the 3-fold 1,2,4-triazole bridges, being a quasi 1D process, or from a combination of intra- and intermolecular interactions, with the anions contributing effectively to the propagation of elastic interactions between the spin transition chains. These materials hardly crystallize, and only from comparison with analogous copper salts,5 and from extended X-ray absorption fine structure (EXAFS)6 and wide-angle X-ray scattering (WAXS) experiments,7 structural details have been available. The [Fe(Htrz)2trz]BF4 salt (1) (Htrz ) 1,2,4-4-H-triazole, trz ) 1,2,4-triazolate) displays an over 40 K wide thermal hysteresis loop above room temperature.8 Direct bridges between chains and anions through hydrogen bonding were claimed as the origin of such remarkable behavior, suggesting a true 3D cooperative transition. Very recently, when spin crossover materials have been prepared as nanoparticles in the search for bistable nanoobjects, 1 has been shown to be the only material where the thermal hysteresis is retained below 10 nm sizes.9 On the contrary, fast disappearance of the hysteresis when nanoparticle sizes get below 30 nm is observed in all other cases, based on other iron(II) 3D polymers.10 These results hinted that the spin transition may be quasi one-dimensional in 1, since only assuming a quasi 1D system the nuclearity of the spin system regarding the effective propagation of elastic interactions would decrease as a function of nanoparticle diameter (r). Conversely, in a 3D system the nuclearity would decrease as a function of r3. This difference could well explain the different behavior of

10.1021/jp107206n  2011 American Chemical Society Published on Web 12/08/2010

1324

J. Phys. Chem. C, Vol. 115, No. 4, 2011

the spin transition in nanoparticles as a function of size according to the dimensionality of the bulk materials. Herein, we report the crystal structure of 1 in the LS and HS states elucidated by high-resolution synchrotron X-ray powder diffraction (XRPD) to verify the suggested 1D chain structure. The dynamic changes in the crystal and local structures were investigated by simultaneous time-resolved in situ XRPD and Raman spectroscopy,11 with an aid of modulation enhanced technique, to shed some light on the origin of the spin transition. Two-dimensional correlation analyses between XRPD patterns and Raman spectra have been attempted to show the potential of such analyses in combined methods to highlight and facilitate the understanding of the underlying physicochemical processes observed by the multiple techniques.12 Experimental Section Synthesis. 1 was synthesized according to the literature.8 A solution of 0.614 g (0.00889 mol) of 1,2,4,-1H-triazole in 3 mL of ethanol was added to a solution of 1 g (0.00296 mol) of Fe(BF4)2 · 6H2O in 6 mL of water. A precipitate appeared after a short while, and it was filtered to vacuum and washed with ethanol and acetone. Chemicals were employed as purchased without further purification. X-ray Powder Diffraction and Raman Spectroscopy. XRPD experiments were performed on the BM1B (SNBL) beamline at the European Synchrotron Radiation Facility (ESRF, Grenoble, France).13 High-resolution X-ray powder diffraction data were collected with the standard BM1B setup. Raman data were collected by a Renishaw inVia spectrophotometer especially adapted for remote measurement, fully described elsewhere.14 The XRPD patterns were collected using radiation with λ ) 0.50177(1) Å. The calibration was done using the lattice parameters of the NIST lanthanum hexaboride (LaB6) standard (SRM 660b; nominal a ) 4.15695(6) Å at room temperature (RT)) with the standard SNBL experimental setup. An instrumental profile was fitted from silicon standard and kept fixed (fundamental parameter approach), and for the RT and 400 K samples only the Lorentzian component was refined in both Pawley and Rietveld refinements. Raman data were collected with a near-IR (785 nm) laser. The Raman grating calibration was systematically controlled by checking the position of a Si standard at 520 cm-1 at a resolution of 6 cm-1. The powder of 1 was mounted into a capillary and rotated during the combined XRPD and Raman measurements. The temperature of the sample was controlled by means of a gas blower. To avoid local heating by the laser, swing-sampling (i.e., the focal point of the laser on the sample moves fast with time along the axial direction of the capillary where the sample is placed) was used. For the refinement of crystal structures of LS and HS states, high-resolution XRPD patterns were recorded at 300 and 400 K, respectively. Modulation Enhanced XRPD and Raman Spectroscopy. The modulation enhanced technique, often called modulation (or modulated) excitation spectroscopy (MES),15 used in this work utilizes a periodic perturbation of a parameter, so-called stimulation, such as concentration and temperature. Repeatedly stimulating a system leads to an oscillating response of certain signals, which are influenced by the stimulation. The response oscillates typically at the same frequency (ω) and occasionally harmonics thereof (2ω, 3ω, ...) as that of the stimulation with a possible phase delay in time. By (1) averaging a number of periodic responses and (2) the mathematical treatment of the

Urakawa et al. averaged response, namely, phase-sensitive detection (PSD), the sensitivity is greatly enhanced and the detection limit is significantly lowered. Furthermore, information about transformation kinetics can be obtained from the phase delay and the response amplitude. A more detailed description of the principles and application of MES and PSD can be found elsewhere.15a,c This work combines the modulation enhanced technique with XRPD and Raman spectroscopy. The data were measured every 37 and 32 s for XRPD and Raman, respectively, in a nonsynchronized manner with the stimulation (a triangle-wave temperature (T) stimulation: linear ramp-up and ramp-down at 360 K/h from 300 to 400 and back to 300 K, in total 2000 s for one cycle). Four T-modulation cycles were recorded and averaged into one cycle to increase the signal-to-noise ratio. The responses of the four cycles were perfectly identical; therefore, the averaging did not influence the detailed characteristics of the transitions obtained by the two techniques. Also, the data averaging of the four cycles was carried out in a pseudosynchronized way by interpolating the data so that the data points of the four cycles match at the same time points upon averaging. The averaged XRPD patterns and Raman spectra were further processed by PSD to determine in-phase angles in the phase domain.15c Structure Solution and Rietveld Refinements. At first, XRPD patterns were compared with the structures reported in the literature16 to obtain the crystal structure by isomorphous replacement of the Fe-(trz)3 moiety. No structure showed enough similarity to refine the structure in such a way; therefore, ab initio solution was sought for exploiting the presence of a heavy atom and the fact that the Fe-(trz)3 moiety is a rigid body, to overcome the limitations given by the poor diffraction data originating from the nanocrystalline morphology. The XRPD pattern was indexed by iterative use of singular value decomposition17 and then refined by Pawley fit as implemented in TOPAS Academic.18 The crystal structure was solved by simulated annealing,19 and the powder patterns were refined by the Rietveld method using TOPAS Academic. The structural molecular models of Fe-(trz)3 (neglecting H atoms) and BF4 moieties, necessary for the simulated annealing procedure, were obtained from the parent compound16b and modifying it by MOLDRAW software.20 At first, the LS crystal structure was solved into a Cc monoclinic space group; the agreement factors were satisfactory, but such a space group was unlikely because it is noncentrosymmetric. However, it was the unique one allowing structure solution because a complete [Fe-(trz)3, BF4] couple is present in the asymmetric unit. This information was helpful in the simulated annealing process because of the known difficulty of real space structure solution methods of managing asymmetric units not coinciding with molecular units. Therefore, the PSEUDO routine21 by the Bilbao Crystallographic Data Server22 was utilized to explore higher symmetries. This analysis allowed finding the corresponding structure in the monoclinic centrosymmetric C2/c and orthorhombic Cmcm space groups. Hexagonal and trigonal space groups were also checked since pseudohexagonal arrangement was observed (Figure 1b). At this point, LS and HS crystal structures were refined in both C2/c and Cmcm space groups and compared to assess the correct symmetry. The structures in monoclinic C2/c were refined, but there was no improvement compared to the orthorhombic Cmcm structures. Moreover, the lower symmetry (C2/c) structure showed similar agreement factors and similar structures with respect to Cmcm, but with increased values in the correlation matrix. The Cmcm space group was therefore used for further

Raman-XRPD Study of Spin Crossover Fe Material

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1325 TABLE 1: Crystal Data of LS and HS Structures of 1

lattice parameters (Å) volume (Å3) Rwp R

LS (Cmcm)

HS (Cmcm)

increase in % from LS to HS

9.367(4) 17.050(7) 7.341(2) 1172.5(8) 15.3 11.4

9.644(2) 17.344(3) 7.782(1) 1301.6(4) 16.2 12.5

2.9 1.7 6.0 11.0 -------

TABLE 2: Selected Distances in LS and HS Crystal Structures of 1

Figure 1. Crystal packing of 1 along c (a) and along a (b) lattice edges. Atom color legend: Fe (red); N (blue); C (green); B (yellow); F (light cyan).

refinements and discussions of the crystal structure at both 300 and 400 K. It is worth noting that in the Cmcm space group the asymmetric unit is formed by 1/4 of Fe-(trz)3 and 1/2 of BF4 moieties, respectively, and this explains the difficulty of solving the structure directly in Cmcm. The atom positions were then relaxed employing constraints on bond angles and distances and on the flatness of aromatic rings. Finally, the positions of hydrogen atoms were calculated, exploiting the fact that the C and N atoms are sp2 hybridized and the steric rules. As expected, neither the R values nor the fit in the refinement showed any improvement. One problem of hydrogen atom positions arises due to the fact that only two out of three trz ligands carry the hydrogen at the N atoms, and also H distribution among trz ligands is not ordered. In fact, because of the symmetry, adding H atoms with full occupancy is unrealistic because the charge neutrality rule of the whole structure is violated, and also H · · · H contacts (1.16 Å (LS), 1.59 Å (HS)) are too short. For this reason, two types of structural files for both LS and HS can be found in the Supporting Information. The first type is without hydrogen and was obtained by the Rietveld refinement. The second one contains H atoms at the calculated positions. Both structures are unreasonable from the viewpoint of charge neutrality, but the structures of the former type are useful for the evaluation of crystal structure and that of the latter for evaluating H position and possible interactions with surrounding atoms. Computational Method. The Raman spectra of the trz anion and Htrz was calculated by Gaussian 0923 using a harmonic approximation after geometry optimization at a DFT level [B3PW9124/6-311++G(3df, 3pd)] as an isolated system in a vacuum. The calculated Raman spectra are shown without scaling. Results and Discussion LS and HS State Crystal Structures of 1. The crystal structure is characterized by cationic chains of Fe-(trz)3 moieties (Figure 1a) surrounded by BF4- anions. The crystal packing is therefore driven by ionic interactions among Fe(Htrz)2(trz)+ and BF4- moieties that are arranged into pseudohexagonal arrangements (Figure 1b). The crystal data of 1 for its LS and HS states are summarized in Table 1. The lattice expands upon a LS-HS

Bond

LS

HS

Fe-Fe Fe-N2 Fe-N3 F· · ·N F· · ·N F· · ·N F· · ·N

3.671(1) 1.827(5) 1.981(6) 3.13(1) 3.43(1) 3.50(1) 3.99(1)

3.891(1) 2.042(5) 2.04(2) 3.20(1) 3.67(1) 3.61(1) 4.16(1)

transition by 2.9, 1.7, and 6% in the a, b, and c axes and by 11.0% in volume. The larger increase in the direction of c indicates that the expansion occurs mainly by the elongation of the Fe(Htrz)2(trz) chain (Figure 1a). Table 2 summarizes the selected distances of 1 at the LS and HS states. The Fe-Fe distances, 3.671 Å (LS) and 3.891 Å (HS), agree well with those reported in the literature and determined by EXAFS (3.65 Å (LS), 3.87 Å (HS))6 and WAXS (3.63 Å (LS), 3.83 Å (HS)).7 The Fe-N distances also increased in the HS state by 0.06-0.22 Å which agrees with the value reported by EXAFS (0.18). Fe-(trz)3 interchain distances were reasonable after refinements: short F · · · N contacts are observed in the range from 3.13 and 4.16 Å, being reasonably larger than the sum of van der Waals radii of N and F without imposing any constraints on intermolecular distances. An estimate of the 1D-chain average length can be derived from the nanocrystal dimensions: for LS crystallite size of 16.7 ( 0.3 nm, for HS crystallite size 24.9 ( 0.2 nm, calculated by the Scherrer equation. Views on Spin Crossover from XRPD. Figure 2 shows the XRPD patterns during the linear temperature increase from 300 to 400 K and decrease from 400 to 300 K. Obviously, the structure of 1 underwent notable changes to the HS state at ca. 700 s (370 K) and remained stable up to 400 K. 1 stayed at the HS state during the following temperature ramp-down, and at about 1800 s (320 K) the structure changes back to the original (LS) state. All peaks shifted to lower angle, indicating a lattice expansion. The structural transition was not symmetric at the middle (at 400 K) of the y-axis in both ramping up and down the temperature, and the hysteresis typically observed for χT (χ, molar magnetic susceptibility; T, temperature)8 was also clearly confirmed by XRPD. Figure 3 shows the temperaturedependent XRPD intensity characteristics for the LS and HS state structures. The temporal evolutions of LS and HS state peaks showed clearly opposite trends (i.e., one increases and the other decreases), both apparently showing hysteresis with transitions within a narrow temperature window to HS centered at ca. 370 K and to LS at ca. 330 K. Views on Spin Crossover from Raman Spectroscopy. Raman spectra recorded simultaneously with XRPD are shown in Figure 4. Compared to the evolution of XRPD patterns, a striking difference was observed by Raman spectroscopy. Notably, the transition to the HS structure was initiated already at 260 s (320 K), while the back transition to the LS structure occurred at about the same time as observed in XRPD, at ca.

1326

J. Phys. Chem. C, Vol. 115, No. 4, 2011

Figure 2. Time-resolved XRPD patterns during spin crossover transition between 300 and 400 K. The peak intensity is shown in intensity counts (the color bar on the right side). The diffraction patterns of the LS and HS structures averaged during the respective steady period are shown above.

Urakawa et al.

Figure 4. Time-resolved Raman spectrum during spin crossover transition between 300 and 400 K. The band intensity is shown in intensity counts (the color bar on the right side). The Raman spectra of the LS and HS structures averaged during the respective steady period are shown above.

Figure 3. Temperature dependence of XRPD intensity of two peaks characteristics for the LS (black, 3.55°) and HS (red, 3.27°) states. The circle and inverted triangle symbols show the data during temperature increase and decrease, respectively.

1800 s (320 K). Figure 4 shows that the Raman signal is close to symmetric at the middle of the y-axis (at 400 K), meaning that no significant hysteresis could be observed by Raman spectroscopy. This behavior can be more clearly seen in Figure 6 where the evolutions of two characteristic bands of the LS and HS states are presented. There was a small hysteresis loop observed; the forward transition to HS occurred at slightly higher temperature than the backward transition to LS, but its magnitude (ca. 5 K) is negligible compared to ca. 40 K observed by XRPD (Figure 3). The 2D and particularly 1D Raman spectra in Figure 4 clearly show the differences between LS and HS states. There were drastic changes observed in the low-frequency regions below 400 cm-1. The three bands of the LS state observed at 132, 209, and 286 cm-1 apparently red-shifted in the HS state to 107, 139, and 184 cm-1, respectively. The band at 286 cm-1 can be attributed to the Fe-N stretching vibration according to the literature dealing with similar materials (Fe(Htrz)3X2, X ) Cl, ClO4, SO4).25 The frequency of the Fe-N stretching vibration is generally reduced by the factor of 1.5-2.1 from

Figure 5. Comparison between experimental LS/HS Raman spectra with the calculated Raman spectra of trz and Htrz. The calculated Raman intensity of trz is scaled half because of the stoichiomety of trz with respect to Htrz in 1.

LS to HS transition.26 The band shift from 286 (LS) to 184 (HS) cm-1 is well within this range, and these bands can be attributed to the Fe-N stretching vibration. The overall red shifts of the three bands in the low-frequency region also confirmed the lattice expansion as shown in the LS and HS crystal data of 1. The bands in the fingerprint region between 700 and 1400 cm-1 are vanishing upon HS transition. Figure 5 highlights the Raman spectra of the LS and HS states in the region in comparison to the calculated spectra of the ligands (trz, Htrz) in an isolated form in a vacuum. Although some changes in the vibrational frequencies and corresponding Raman intensities are expected because of coordination to Fe as well as interaction

Raman-XRPD Study of Spin Crossover Fe Material

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1327

Figure 6. Temperature dependence of Raman intensity of two bands characteristic for the LS (black, 210 cm-1) and HS (red, 105 cm-1) states. The circle and inverted triangle symbols show the data during temperature increase and decrease, respectively.

Figure 7. Population of the HS state calculated from quantitative analysis by Rietveld refinement of XRPD data. The circle and inverted triangle symbols show the data during temperature increase and decrease, respectively.

to BF4-, the calculation reproduces most of the spectral features and clarifies that the double-band character at ca. 1070 and 1300 cm-1 originates from the two different, trz and Htrz, ligands. The bands at 1056 and 1309 cm-1 of LS (1045 and 1307 cm-1 of HS) are due to Htrz, while those at 1083 and 1284 cm-1 of LS (1072 and 1281 cm-1 of HS) are due to trz. The rough ratio of the band areas shows nearly 1:2 for the bands of trz:Htrz, respectively. This is reasonable because the stoichiometry of trz and Htrz in 1 is 1:2. At the higher-frequency region, a remarkable decrease in intensity and red shifts of most of the bands were observed upon transition to the HS state. For example, the bands at 977 and 1101 cm-1 of LS disappeared completely (or at least became too weak to be observed) at the HS state, and the bands at 1056, 1083, 1284, and 1309 cm-1 reduced their intensities considerably. The exception was the band at 766 cm-1 which neither shifted nor decreaseed in intensity. The band at 1162 cm-1 did not shift, but the intensity seemingly decreased. The common feature of these two bands is the nature of vibration, both related to C-H and/or N-H vibrations that are not closely related to the ring vibration of trz/Htrz. On the other hand, the bands which underwent considerable shifts and intensity decrease are related to the ring vibration of trz/Htrz, suggesting that the environment of the ligands has changed upon SC. The identical band feature of 766 cm-1 at LS and HS is reasonable because this is the out-of-plane C-H bending mode and it is not well-coupled with in-plane vibrations of the trz/Htrz ring. The significant red shifts of the bands at 1056 and 1083 cm-1 by ca. 10 cm-1 suggest that the N-N bond length and/or environment of trz and Htrz are both largely influenced by the SC, likely because of changes in the Fe-N bonding character. Indeed, the elongation of the Fe-N bond in the HS state originates from the weaker metal-ligand interactions of essential ionic character versus the strong metal-ligand orbital interaction in the LS state from an increased ligand field. It is very surprising to observe such great differences in the LS-HS transition profiles obtained by XRPD and Raman spectroscopy (Figure 3 vs Figure 6). From the Raman point of view, the transition occurs in a short temperature window of ca. 20 K, while it occurs in the window of ca. 50 K from the HS-population point of view derived from the XRPD results. This difference can be explained by two different sensitivities toward (i) short- and long-range ordering of crystal structure and (ii) outer surface and bulk (inner core) of the grain of the 1 powder. In the former case (i), the difference in XRPD and

Raman is caused by earlier detection of the short-range structural transformation observed by Raman, followed by the long-range ordering of the crystal structure measured by XRPD. In the latter case (ii), the difference is due to the gradual transition within the grain of the 1 powder. The depth probed by Raman is significantly shorter than that by XRPD and if the phase transition occurs initially from the outer surface of the grain and propagates toward its bulk. Assuming that the latter is true, there is a spatial gradient in the degree of spin transition within the grains of the 1 powder. The temperature gradient within the powder can be one possible cause because temperature change is likely influenced faster at the outer surface than in the bulk of the 1 powder. However, the temperature equilibration of the powder is expected to be very fast and not in the time scale of the differences observed between XRPD and Raman results; therefore, another driving force, such as that suggested above (i) and/or (ii), seems to exist for the spin transition to occur earlier locally, e.g., starting from the outer region of nanocrystallites, from the surface of the grain, and/or from defectively coordinated Fe sites, which is rather probable in nanocrystals. A similar conclusion has been drawn by the study on light-induced exited spin-state trapping (LIESST) processes where faster occurrence of local deformation compared to longrange structural transition was shown,27 although the SC induction mechanism is different. It is important to note that the transition of the magnetic properties shows a large hysteresis similar to that observed in XRPD since magnetic properties occur in bulk, and therefore it is of great value to investigate nanosized SC materials where the change in the bulk structure or long-range ordering is expected to occur rapidly.9a To gain detailed insights into the SC, the population of crystallographic phases corresponding to the LS and HS state structures has been quantified by Rietveld refinement of all the XRPD patterns during the transition. The results suggested that there were solely two LS and HS structures without an intermediate one during the transition. Figure 7 shows the thusobtained population of the HS structure (that of LS is the rest of population). Notably, the population of the HS structure increased gradually during the forward LS-HS transition which started at 335 K and ended at 375 K. On the other hand, the evolution of HS population during the backward transition is different, occurring within a smaller temperature range (starting at 330 K and ending at 315 K). Comparison of the HS structure population to the evolutions observed by XRPD (Figure 3) and Raman spectroscopy (Figure 6) shows that the structural changes

1328

J. Phys. Chem. C, Vol. 115, No. 4, 2011

of 1 are first observed by Raman spectroscopy starting at 320 K. When the changes were completed at 335 K in Raman, the population of the HS state started increasing gradually, and at ca. 370 K the transition became rapid and completed to the full extent. The backward transition occurred rapidly, and the evolutions observed in Raman, XRPD, and the HS structure population coincided, suggesting that in the spin transition, from the short- to the long-range, structural changes induced by the SC occur at the same time for the backward transition. This may imply that the back transition is initiated from the inner core of the grain or crystallites of 1 or occurs randomly at any position of the grain. 2D Analysis. Insights of the results obtained by combined techniques can be enriched and enhanced by identifying the correlation of the results acquired by different techniques to facilitate the assignments and understanding of underlying physicochemical processes. Here we present three representations to visualize the interrelation of the responses (correlated or anticorrelated) obtained by two techniques (here XRPD and Raman spectroscopy) and also temporal evolutions in a visual manner. One simple way to analyze the interrelations between responses in XRPD and Raman spectroscopy is to use the covariance of XRPD A(2θ, t) and Raman B(ν, t) responses, Cov(2θ,ν), which is calculated as follows N

Cov(2θ, ν) )

∑

1 A(2θ)i′B(ν)i′ N - 1 i)1

where A(2θ)i′ and B(ν)i′ are the intensity values of the XRPD pattern or Raman spectrum of the ith data number (in the time axis) subtracted by their mean values, and N is the total number of data points in the time axis in one modulation cycle. Another way to show such interrelations is to use correlation Corr(2θ,ν) which is readily calculated from the covariance as follows

Corr(2θ, ν) )

Cov(2θ, ν) σA(2θ,t)σB(ν,t)

where σA(2θ, t) and σB(ν, t) are the standard deviations of A(2θ, t) and B(ν, t), respectively. The advantage of using correlation is that the quantity shown in covariance is normalized by the standard deviations at chosen 2θ and ν. Therefore, small correlated signals are visible in correlation, while they are typically hidden in covariance because the quantity is proportional to the changes in the signal amplitudes. 2D covariance analysis is yet superior to localize and easily highlight interrelations between clearly changing signals. Conversely, correlation analyses are best suited to detect the behavior of weak signals and/or very subtle changes. Figures 8A and 8B show the covariance and correlation plots of the XRPD patterns and Raman spectra. Positive values in covariance and correlation indicate that the responses in XRPD and Raman are correlated, likely originating from the same molecular or structural origins. The negative values, on the other hand, indicate that at that point of 2θ and ν the two response are anticorrelated, meaning that the signal increases for one technique and decreases for the other technique. In this study, the positive correlation appears when the temporal response in XRPD and Raman at certain 2θ and ν values is similar or identical. For example, at 3.25° (XRPD) and 105 cm-1 (Raman),

Urakawa et al.

Figure 8. 2D XRPD-Raman covariance (A), correlation (B), and sum of in-phase angle (C) plots of a selected region.

there is a positive correlation showing positive values in Figures 8A and 8B. On the basis of Figures 2 and 4, they both correspond to the HS state. On the other hand, the positive values near 3.50° (XRPD) and 130 cm-1 (Raman) have a different origin which is from the LS state. In addition, the combination of 3.25° (XRPD) and 130 cm-1 (Raman) or 3.50° (XRPD) and 105 cm-1 (Raman) yields negative correlation because the signals are due to the HS or LS state and therefore anticorrelated. In these 2D covariance and correlation plots, it is facile to find out if the signal responses are correlated or not. It is also obvious from Figure 8B that the 2D correlation plot can sense subtle interrelations where signals are small. In both covariance and correlation, however, we cannot differentiate the two differently correlated signals such as LS (XRPD)-LS (Raman) and HS (XRPD)-HS (Raman) in this representation. To differentiate and furthermore include the information of temporal eVolution in the 2D plot, PSD in-phase angles,15c determined for XRPD and Raman responses at the fundamental modulation frequency, were utilized. The in-phase value can be directly translated into the phase delay with respect to the stimulation, and this is a very convenient way to study reaction and transformation pathways.15c Here the sum of two in-phase angles was used as a measure. In this way, the two differently correlated signals, showing only a positive value previously, will appear differently, because the LS and HS states appear at different positions in time. Figure 8C shows the 2D sum of the in-phase angle plot. Color codes are given according to the sum of the in-phase angles. Apparently, the plot similar to Figure 8B was obtained. However, a closer look clearly demonstrates that the positively correlated regions (e.g., 3.25° (XRPD)-105 cm-1 (Raman) vs 3.50° (XRPD) and 130 cm-1 (Raman)) are shown by different colors. Alternatively, absolute phase delays derived from the in-phase angle can be used for the plot, which then directly yield a correlation map in the 2D plot, showing the temporal evolution, i.e., which correlation (event) appears earlier or later. This LS-HS transition is a simple system for correlation analysis because only two components are involved, but the principles and actual usage are clearly demonstrated.

Raman-XRPD Study of Spin Crossover Fe Material The correlation analyses can be utilized directly and will show its great potential for more complex systems where more reaction or transformation steps are involved during one modulation cycle. Conclusions The crystal structure of 1 at the LS and HS states was elucidated by our high -resolution synchrotron XRPD study. The 1D chain structures, previously suggested but not proven from crystal structure, were confirmed for both LS and HS states. Spin transition to HS induced the elongation of the chain, and this was the major cause of the lattice expansion. The dynamic changes in the crystal and local structures were investigated by combined XRPD and Raman spectroscopy, in the framework of a modulation-enhanced technique using a periodic temperature change as stimulus. The two methods have shown different insights into the SC process. The XRPD analysis showed that the crystallographic phase transition triggered by the spin transition has a marked 1D character since it essentially occurs through the elongation of the 1D chains, whereas no interchain or chain-anion interaction is playing a significant role. Furthermore, no intermediate phase between the LS and HS was suggested. The changes induced by the LS to HS transition were observed at lower temperature by Raman spectroscopy than by XRPD, resulting in negligible hysteresis in the former and large hysteresis for the latter as typically reported from its magnetic properties. The difference was explained by the different sensitivity of XRPD and Raman toward short- and longrange ordering of crystallites and/or toward only the outer surface or up to the inner core of the grain of 1. Rietveld-based population analysis of the HS state structure clarified a gradual population increase of the HS state, which was initiated after the notable structural changes observed by Raman spectroscopy. The backward (HS to LS) transition was, on the other hand, a rapid process showing nearly identical evolution profiles in XRPD and Raman spectroscopy. Furthermore, we demonstrated the two-dimensional correlation analyses based on covariance, correlation, and sum of in-phase angles. The former two methods can be used to find out correlation and anticorrelation of signals in the two different techniques. On the other hand, the latter method can be conveniently used to differentiate different types of correlations and temporal evolution of the correlation to study transition mechanisms. Acknowledgment. SNBL and ESRF (Grenoble, France) are acknowledged for the beam time (proposal number CH-2234). AU and JRGM thank the ICIQ Foundation, and JRGM thanks the Spanish Ministerio de Ciencia e Innovacio´n (Project CTQ2008-03197/BQU) for the financial support. Supporting Information Available: Crystallographic information files of low-spin and high-spin state structures of [Fe(Htrz)2(trz)](BF4) with and without hydrogen atoms, 2D XRPD-Raman covariance, correlation, and sum of in-phase angle plots of the full range. This material is available free of charge via the Internet at http://pubs.acs.org. References and Notes (1) Gu¨tlich, P.; van Koningsbruggen, P. J.; Renz, F. Struct. Bonding (Berlin, Ger.) 2004, 107, 27–75. (2) Gu¨tlich, P.; Hauser, A. Coord. Chem. ReV. 1990, 97, 1–22. (3) (a) Kahn, O.; Martinez, C. J. Science 1998, 279 (5347), 44–48. (b) Gaspar, A. B.; Ksenofontov, V.; Seredyuk, M.; Gu¨tlich, P. Coord. Chem. ReV. 2005, 249 (23), 2661–2676. (c) Salitros, I.; Madhu, N. T.; Boca, R.; Pavlik, J.; Ruben, M. Monatsh. Chem. 2009, 140 (7), 695–733. (4) (a) Garcia, Y.; Niel, V.; Mun˜oz, M. C.; Real, J. A. Top. Curr. Chem. 2004, 233, 229–257. (b) Haasnoot, J. G. Coord. Chem. ReV. 2000, 200, 131–185.

J. Phys. Chem. C, Vol. 115, No. 4, 2011 1329 (5) Dirtu, M. M.; Neuhausen, C.; Naik, A. D.; Rotaru, A.; Spinu, L.; Garcia, Y. Inorg. Chem. 2010, 49 (12), 5723–5736. (6) Michalowicz, A.; Moscovici, J.; Ducourant, B.; Cracco, D.; Kahn, O. Chem. Mater. 1995, 7 (10), 1833–1842. (7) Verelst, M.; Sommier, L.; Lecante, P.; Mosset, A.; Kahn, O. Chem. Mater. 1998, 10 (4), 980–985. (8) Kro¨ber, J.; Audie`re, J. P.; Claude, R.; Codjovi, E.; Kahn, O.; Haasnoot, J. G.; Grolie`re, F.; Jay, C.; Bousseksou, A.; Linare`s, J.; Varret, F.; Gonthier-Vassal, A. Chem. Mater. 1994, 6 (8), 1404–1412. (9) (a) Gala´n-Mascaro´s, J. R.; Coronado, E.; Forment-Aliaga, A.; Monrabal-Capilla, M.; Pinilla-Cienfuegos, E.; Ceolı´n, M. Inorg. Chem. 2010, 49 (12), 5706–5714. (b) Coronado, E.; Gala´n-Mascaro´s, J. R.; MonrabalCapilla, M.; Garcı´a-Martı´nez, J.; Pardo-Iba´n˜ez, P. AdV. Mater. 2007, 19 (10), 1359–1361. (10) (a) Le´tard, J.-F.; Nguyen, O.; Daro, N. WO2007065996-A1; FR2894581-A1; EP1960410-A1; US2008311401-A1; CA2632704-A1; JP2009519879-W; EP1960410-B1; DE602006012730-E, 2007. (b) Niel, V.; Martinez-Agudo, J. M.; Mun˜oz, M. C.; Gaspar, A. B.; Real, J. A. Inorg. Chem. 2001, 40 (16), 3838–3839. (c) Volatron, F.; Catala, L.; Rivie`re, E.; Gloter, A.; Ste´phan, O.; Mallah, T. Inorg. Chem. 2008, 47 (15), 6584– 6586. (d) Boldog, I.; Gaspar, A. B.; Martı´nez, V.; Pardo-Iba´n˜ez, P.; Ksenofontov, V.; Bhattacharjee, A.; Gu¨tlich, P.; Real, J. A. Angew. Chem., Int. Ed. 2008, 47 (34), 6433–6437. (e) Larionova, J.; Salmon, L.; Guarl, Y.; Tokarev, A.; Molvinger, K.; Molna´r, G.; Bousseksou, A. Angew. Chem., Int. Ed. 2008, 47 (43), 8236–8240. (11) Boccaleri, E.; Carniato, F.; Croce, G.; Viterbo, D.; van Beek, W.; Emerich, H.; Milanesio, M. J. Appl. Crystallogr. 2007, 40, 684–693. (12) Cavallini, M.; Bergenti, I.; Milita, S.; Ruani, G.; Salitros, I.; Qu, Z. R.; Chandrasekar, R.; Ruben, M. Angew. Chem., Int. Ed. 2008, 47 (45), 8596–8600. (13) Swiss Norwegian Beamline (BM1, http://www.snbl.eu); ESRF: Grenoble, France. (14) van Beek, W.; Carniato, F.; Kumar, S.; Croce, G.; Boccaleri, E.; Milanesio, M. Phase Trans. 2009, 82 (4), 293–302. (15) (a) Baurecht, D.; Fringeli, U. P. ReV. Sci. Instrum. 2001, 72 (10), 3782–3792. (b) Bu¨rgi, T.; Baiker, A. J. Phys. Chem. B 2002, 106 (41), 10649–10658. (c) Urakawa, A.; Bu¨rgi, T.; Baiker, A. Chem. Eng. Sci. 2008, 63 (20), 4902–4909. (16) (a) Vos, G.; le Feˆbre, R. A.; de Graaff, R. A. G.; Haasnoot, J. G.; Reedijk, J. J. Am. Chem. Soc. 1983, 105 (6), 1682–1683. (b) Kolnaar, J. J. A.; de Heer, M. I.; Kooijman, H.; Spek, A. L.; Schmitt, G.; Ksenofontov, V.; Gu¨tlich, P.; Haasnoot, J. G.; Reedijk, J. Eur. J. Inorg. Chem. 1999, (5), 881–886. (17) Coelho, A. A. J. Appl. Crystallogr. 2003, 36, 86–95. (18) Coelho, A. A. TOPAS-Academic 4.1; Coelho Software: Brisbane, Australia, 2007. (19) Coelho, A. A. J. Appl. Crystallogr. 2005, 38, 455–461. (20) Ugliengo, P.; Viterbo, D.; Chiari, G. Z. Kristallogr. 1993, 207, 9– 23. (21) Kroumova, E.; Aroyo, M. I.; Perez-Mato, J. M.; Ivantchev, S.; Wondratschek, H. J. Appl. Crystallogr. 2001, 34, 783–784. (22) (a) Aroyo, M. I.; Perez-Mato, J. M.; Capillas, C.; Kroumova, E.; Ivantchev, S.; Madariaga, G.; Kirov, A.; Wondratschek, H. Z. Kristallogr. 2006, 221 (1), 15–27. (b) Aroyo, M. I.; Kirov, A.; Capillas, C.; PerezMato, J. M.; Wondratschek, H. Acta Crystallogr., Sect. A 2006, 62, 115– 128. (23) Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, M. A.; Cheeseman, J. R.; Scalmani, G.; Barone, V.; Mennucci, B.; Petersson, G. A.; Nakatsuji, H.; Caricato, M.; Li, X.; Hratchian, H. P.; Izmaylov, A. F.; Bloino, J.; Zheng, G.; Sonnenberg, J. L.; Hada, M.; Ehara, M.; Toyota, K.; Fukuda, R.; Hasegawa, J.; Ishida, M.; Nakajima, T.; Honda, Y.; Kitao, O.; Nakai, H.; Vreven, T.; Montgomery, J. J. A.; Peralta, J. E.; Ogliaro, F.; Bearpark, M.; Heyd, J. J.; Brothers, E.; Kudin, K. N.; Staroverov, V. N.; Kobayashi, R.; Normand, J.; Raghavachari, K.; Rendell, A.; Burant, J. C.; Iyengar, S. S.; Tomasi, J.; Cossi, M.; Rega, N.; Millam, N. J.; Klene, M.; Knox, J. E.; Cross, J. B.; Bakken, V.; Adamo, C.; Jaramillo, J.; Gomperts, R.; Stratmann, R. E.; Yazyev, O.; Austin, A. J.; Cammi, R.; Pomelli, C.; Ochterski, J. W.; Martin, R. L.; Morokuma, K.; Zakrzewski, V. G.; Voth, G. A.; Salvador, P.; Dannenberg, J. J.; Dapprich, S.; Daniels, A. D.; Farkas, ¨ .; Foresman, J. B.; Ortiz, J. V.; Cioslowski, J.; Fox, D. J. Gaussian 09, O Revision A.01; Gaussian, Inc.: Wallingford, CT, 2009. (24) (a) Becke, A. D. J. Chem. Phys. 1993, 98 (7), 5648–5652. (b) Perdew, J. P.; Wang, Y. Phys. ReV. B 1992, 45 (23), 13244–13249. (25) Smit, E.; Manoun, B.; de Waal, D. J. Raman Spectrosc. 2001, 32 (5), 339–344. (26) Tuchagues, J. P.; Bousseksou, A.; Molnar, G.; McGarvey, J. J.; Varret, F. The role of molecular vibrations in the spin crossover phenomenon. In Topics in Current Chemistry; Springer: New York, 2004; Vol. 235, pp 85-103. (27) Moussa, N. O.; Molna´r, G.; Ducros, X.; Zwick, A.; Tayagaki, T.; Tanaka, K.; Bousseksou, A. Chem. Phys. Lett. 2005, 402 (4-6), 503–509.

JP107206N