Jian
Chen
ab,
Yoshihiro
Sekine
ab,
Atsushi
Okazawa
c,
Hiroyasu
Sato
d,
Wataru
Kosaka
ab and
Hitoshi
Miyasaka
*ab
aInstitute for Materials Research, Tohoku University, 2-1-1 Katahira, Aoba-ku, Sendai 980-8577, Japan. E-mail: miyasaka@imr.tohoku.ac.jp
bDepartment of Chemistry, Graduate School of Science, Tohoku University, 6-3 Aramaki-Aza-Aoba, Aoba-ku, Sendai 980-8578, Japan
cDepartment of Basic Science, Graduate School of Arts and Sciences, The University of Tokyo, 3-8-1 Komaba, Meguro-ku, Tokyo 153-8902, Japan
dApplication Laboratory, Rigaku Corporation, 3-9-12, Matsubara-cho, Akishima-shi, Tokyo 196-8666, Japan
First published on 5th March 2020
Molecular materials whose electronic states are multiply varied depending on external stimuli are among the most promising targets for the development of multiply accessible molecular switches. Here, we report a honeycomb layer composed of tetraoxolene-bridged iron (Fe) subunits whose charge-ordered states are multiply variable via thermal treatments and solvation/desolvation with the crystallinity intact. The compound is (NPr4)2[Fe2(Cl2An)3] (1-d; NPr4+ = tetra-n-propylammonium; Cl2An2− = 2,5-dichloro-3,6-dihydroxo-1,4-benzoquinonate), which possesses three charge-ordered states: a low-temperature (LT) phase [(Fe3+)2(Cl2An2−)(Cl2An˙3−)2]2−; an intermediate (IM) phase [(Fe2.5+)2(Cl2An2−)(Cl2An2.5−)2]2−; and a high-temperature (HT) phase [(Fe2+)2(Cl2An2−)3]2− that varies according to temperature. In addition, the LT phase of 1-d is reversibly changeable to another IM phase in its solvated compound 1via a solvation/desolvation process at room temperature. This example demonstrates a new multiple-switching system based on electron transfer and host–guest chemistry in a charge-flexible metal–organic framework.
Considering such TDET systems that possess electronic disproportionation states, two-dimensional (2D) layered systems with a formula of DxAy (x ≠ y) may be good candidates (Fig. 1). Recently, we have reported one-step TDET in a D2A3-type honeycomb-layered compound, (NPr4)2[Fe2(Cl2An)3]·2(acetone)·H2O (1), which is composed of 2,5-dichloro-3,6-dihydroxo-1,4-benzoquinonate (Cl2Anm−) and Fen+ ions with tetrapropylammonium cations (NPr4+) as well as some crystallization solvents (acetone and water) located at hexagonal pores of layers and between layers.31 This layered compound revealed two ground states of the fully electron transferred state (as the low-temperature (LT) phase) of [(Fe3+)2(Cl2An2−)(Cl2An˙3−)2]2− and a charge-disproportionate ordered state (defined as the charge-ordered intermediate (IMo) phase) of [(Fe2+)(Fe3+)(Cl2An2−)2(Cl2An˙3−)]2− with a boundary at T1/2a↑ = 237 K (T1/2a↓ = 236 K, where the signs ↑ and ↓ refer to heating and cooling processes, respectively. Moreover, T1/2a is T1/2(1) for 1). The Cl2Anm− bridging ligand can take three redox-active states of and and vice versa, whereas Fen+ takes Fe2+ → Fe3+ + e− and vice versa in this type of material (Fig. 1). Thus, the LT phase is a layer that is constructed from D+A− chains bridged by Cl2An2− (LB), where D+ and A− are high-spin Fe3+ and Cl2An˙3− (LA), respectively, while the IMo phase is a layer where DA chains and D+A− chains are alternately aligned with the Cl2An2− (LB) bridges (Scheme 1). In other words, this D2A3 layer system can be assumed to be a layered DA chain as a sort of one-dimensional (1D) VT system.30
This feature of charge-ordered states made us imagine the existence of a neutral state of the high-temperature (HT) phase in 1, which comprises only DA chains at higher temperatures; practically, 1 converted itself to a solvent-free form, (NPr4)2[Fe2(Cl2An)3] (1-d), with an elimination of its lattice solvents, and no longer changes to the HT phase within 1 upon heating. Nevertheless, 1-d afforded three TDET phases (LT → IM → HT) described in Scheme 1 with two-step TDET at T1/2b and T1/2c appearing continuously from the IMo phase of 1via desolvation (T1/2b and T1/2c are T1/2(1) and T1/2(2) for 1-d, respectively) (Scheme 2). In addition, the conversion between 1-d and 1 is reversible via a solvation/desolvation treatment.
Scheme 2 Temperature-scale representation of the charge-ordered states in 1/1′ and 1-d, where the top and bottom scales are for 1-d and 1/1′, respectively. |
Here, we report the first example of multi-step TDET in a layered D2A3 compound (1-d), which is continuously linked to another TDET-form (1) involving a guest-induced ET in a single crystal. These TDETs consequently revealed five material states with three unique charge-ordered states (LT, IMo, and HT) that are formable in a D2A3 honeycomb-layered system. This type of TDET material has good potential for use as a stimuli-controllable molecular multi-switch.
The lattice solvent molecules in 1 can be easily removed by evacuating at room temperature for several hours to produce the desolvated compound, (NPr4)2[Fe2(Cl2An)3] (1-d), with its crystallinity remaining intact (here, it took 12 h to complete the elimination). Thermal gravimetric analysis (TGA) shows that the elimination of lattice solvents is completed by heating up to ∼340 K under N2 atmosphere conditions, and the compound of 1-d is very stable up to ca. 500 K (Fig. S1†). The crystallinity of 1-d was confirmed by performing single-crystal and powder X-ray diffraction (SC- and P-XRD) studies (Fig. 2 and S2;†vide infra). However, with the exposure of the crystalline sample of 1-d to a vapor of a 1:2 v/v mixture of water/acetone at room temperature for 2–3 h, the solvated form, namely 1, was recovered (as 1′) with the pristine structure (Fig. S3, S4 and Table S1†). TGA of 1′ proves the contamination of the same amount of lattice solvents, i.e., 2(acetone)·H2O, and follows the same profile as that found for 1 (Fig. S1†).
SC-XRD analysis of 1-d was first performed at 103 K, which revealed a 2D honeycomb-layered structure that is similar to that of 1. While 1 was crystallized in the monoclinic space group P21/c (Z = 4),311-d crystalized in the space group of P21/n with Z = 2, which determined a half of the formula unit to be an asymmetric unit (1 × Fen+, 1.5 × Cl2Anm− (LA × 1, LB × 0.5), and 1 × NPr4+; Fig. 2a and Table S2†) with two-fold axes through the Fen+ and the midpoint of the LB Cl2Anm− ligand. The Fen+ node has a distorted octahedral Δ or Λ coordination geometry with three bidentate O atoms from two LA ligands and one LB bridging ligand, forming an anionic hexagonal layer [Fe2(Cl2An)3]2− spreading parallel to the (10−1) plane (Fig. 2b), where the Δ and Λ geometric coordinations on the Fen+ nodes are present alternately. The NPr4+ cations are located between the [Fe2(Cl2An)3]2− layers, separating the layers with an interlayer distance of 8.817 Å (Fig. 2c). Thus, remarkable structural differences between the solvated (1 and 1′) and the desolvated (1-d) compounds are found in the structural symmetry of the formula unit and in the interlayer space shortened in 1-d (Fig. 2). It should be noted that the NPr4+ cations are evenly located between the [Fe2(Cl2An)3]2− layers in 1-d, influencing equal electrostatic effects for all Fen+ nodes and the LA Cl2Anm− ligands, whereas the presence of two structurally characterized NPr4+ cations in 1 distinguishes the electrostatic effect for the Fen+ sites and the LA Cl2Anm− ligands.
This electrostatic situation could significantly affect the stabilization of the electronic states of 1-d and 1 (vide infra).
To determine the oxidation state of the Fen+ node and the Cl2Anm− ligand, it is convenient to verify the local bond lengths of Fe–O and C–O in the Cl2Anm− ligands because of their sensitivity to the oxidation states; a comparison of the mean bond lengths dav(Fe–O) for each octahedral Fen+ node and dav(C–O) for each Cl2Anm− ligand is made. In general, dav(C–O) is found to be within the range of 1.244–1.270 Å and 1.289–1.312 Å for Cl2An2− and Cl2An˙3−, respectively.30,31 The dav(C–O) value at 103 K for 1-d was 1.293(3) Å and 1.257(3) Å for LA and LB, respectively, which are assigned the oxidation states of Cl2An˙3− and Cl2An2−, respectively. On the other hand, the dav(Fe–O) value tends to be within the range of 2.091–2.140 Å and 2.000–2.050 Å for high-spin Fe2+ and Fe3+, respectively.30,31 The dav(Fe–O) value at 103 K for 1-d was 2.019(2) Å, which indicates the presence of the high-spin Fe3+ state. These situations imply that the charge-ordered state of 1-d at 103 K is in the LT state with the formula of [(Fe3+)2(Cl2An2−)(Cl2An˙3−)2]2−, where the Cl2An˙3− bridging ligands are LA, which makes a quasi-chain with Fe3+ along the [101] direction, and the Cl2An2− LB bridging ligand links these chains alternately (along the [010] direction) to form the hexagonal net. Although the LT state in 1 underwent a transition at T1/2a = 237 K,31 the LT state of 1-d is stabilized even at 300 K without there being significant changes in bond lengths (Fig. S5 and Table S3†). Thus, the desolvation treatment of 1 at room temperature (300 K), changing to 1-d, successfully demonstrated a change in the charge-ordered state involving an electron transfer: the IMo phase [(Fe2+)(Fe3+)(Cl2An2−)2(Cl2An˙3−)]2− in 1 → the LT phase [(Fe3+)2(Cl2An2−)(Cl2An˙3−)2]2− in 1-d. It should be noted that this sponge behavior is reversible via the solvation procedure to form 1 (=1′) (Fig. S6†).
The χmT value at 300 K is 13.3 cm3 K mol−1, which is slightly greater than the theoretical value (9.5 cm3 K mol−1) for a sum of magnetically isolated species of Cl2An˙3− (S = 1/2, g = 2.0) and high-spin Fe3+ (S = 5/2, g = 2.0). This is indicative of strongly coupled spins, even at 300 K (vide infra).40–42 Upon decreasing the temperature, χm gradually increased and reached a maximum of 2.08 cm3 mol−1 at 23 K, followed by an abrupt decrease to 1.45 cm3 mol−1 at 1.8 K. The χm–T plot in the temperature range of 35–300 K was simulated by using the Seiden model with alternating S = 5/2 and S = 1/2 spins, including a mean-field approximation (zJ′) to consider interchain interactions.43 The fitting of the data for 1-d provided an adequate parameter set of gFe = gL = 2.0 (fixed), J = −125.2 cm−1, and zJ′ = −0.27 cm−1, which are comparable to those of 1 (J = −130.6 cm−1 and zJ′ = −0.27 cm−1) (Fig. 3).31
Field-cooled magnetization (FCM) curves obtained with several dc fields showed an elimination of the χm peak at Tex with increasing applied external dc field, which is indicative of spin flipping behavior from an antiferromagnetic (AF) ground state (Fig. S7a†). In addition, the field dependence of the magnetization (M–H) measured at low temperatures revealed sigmoidal features with spin-flipping fields (Hex, which was determined from a peak in dM/dH data at each temperature measured) as well as a small field hysteresis with Hc = 0.14 T at 1.8 K (Fig. S7b†). This is a typical case of metamagnetic behavior in low-dimensional magnetic systems; the inset of Fig. 3 displays an H/T phase diagram made from Hex and Tex, which separates the regions of paramagnetic (P) and AF phases.
To investigate in more detail the magnetic behavior, the temperature dependence of the ac susceptibilities was measured at several frequencies below 60 K by applying an oscillating magnetic field (Hac = 3 Oe) and a dc field of Hdc = 0 Oe. The in-phase susceptibility (χ′) showed three kinds of peaks at around 23 K (broad), 20 K, and 5 K (Fig. S8a†), where the frequency-independent peaks at 23 K do not involve any anomaly of out-of-phase susceptibility (χ′′), which is indicative of the occurrence of an AF phase transition with TN = 23 K following the phase diagram (inset of Fig. 3). The other peaks at around 20 K and 5 K are slightly frequency dependent and involve an anomaly of χ′′, albeit weak. Although a similar ac susceptibility feature was observed in 1, which revealed single-chain magnet (SCM) behavior,31 this behavior in 1-d ruled out the possibility of SCM, and this may be due to a strong contribution of interchain antiferromagnetic interactions that possibly form some magnetic random domains in the AF phase. Actually, a fitting of the χ′′–T curves using the Arrhenius law τ(T) = τ0exp(Δτ/kBT) provided a much faster τ0 value of 1.3 × 10−19 s with Δτ/kB = 148.3 K; this is similar to a glassy behavior (Fig. S8b†).44,45 Thus, the significant magnetic change from 1 to 1-d could be attributed primarily to the structural contraction associated with the shorter interlayer distance in 1-d (Fig. 2).
Fig. 5 Temperature dependence of the heat flow obtained from differential scanning calorimetry (DSC) values for 1-d with a sweep rate of 5 K min−1. |
Fig. 6 Charge variations of the core unit and 57Fe Mössbauer spectra of 1-d in LT at 300 K (a and b), IMo at 335 K (c and d), and HT at 380 K (e and f). (a, c, and e) Display of charge variations, where L2−, L2−/˙3−, and L˙3− are displayed as gray-, pink-, and red-filled C6 rings, respectively, and Fe2+, Fe2+/Fe3+, and Fe3+ are displayed in blue, green, and orange, respectively, with C, O, and Cl atoms in gray, red, and green, respectively. (b, d, and f) 57Fe Mössbauer spectra for the respective phases. The solid lines in the Mössbauer spectra are Lorentzian curves obtained using the parameters in Table S6,† where orange and blue curves correspond to high-spin Fe3+ and high-spin Fe2+ species, respectively. The purple curves observed at 330 K and 380 K, with similar parameters of δ and ΔEQ may correspond to the impurity of decomposition species due to the higher temperature for longer measurement times, even in a nitrogen atmosphere. |
The dav(C–O) values for LA and LB and the dav(Fe–O) value at 300 K are 1.286(6), 1.260(7), and 2.027(4) Å, respectively (Fig. 6a and Table S5†), which are consistent with the LT charge-ordered state of [(Fe3+)2(Cl2An2−)(Cl2An˙3−)2]2−, as found at 103 K (vide supra). 57Fe Mössbauer spectra recorded at 300 K show only a typical spectrum of Fe3+HS (Fig. 6b and Table S6†).30,31 However, at 335 K, only the dav(C–O) value of LA changes to be lower as 1.271(7) Å, which is just an intermediate between the cases of Cl2An2− and Cl2An˙3−, while dav(C–O) = 1.254(9) Å for LB is still within the range of Cl2An2− (Fig. 6c and Table S5†). In addition, the dav(Fe–O) value of 2.058(4) Å is also an intermediate between those of Fe2+–O and Fe3+–O (Table S5†). This bonding formation indicates that the phase at 335 K is in an IM phase, although a half of the structure was determined by this lab-level SC-XRD measurement. This IM state for 1-d at 335 K could be assigned to be a positional disorder of the [(Fe2+)(Fe3+)(Cl2An2−)2(Cl2An˙3−)]2− state made in each layer that is similarly found in the IM phase in 1.31 This conclusion was also supported by SC-XRD measurements using a synchrotron (SPring-8, Hyogo) (ESI†). Meanwhile, 57Fe Mössbauer spectroscopy performed at 335 K revealed that the Fe2+HS and Fe3+HS species were evenly located in the mode of IMo (like in 1), not in the mode of the charge-disproportionate delocalized state (IMd) in Scheme 1, on the time scale of the Mössbauer effect (t ≈ 10−7 to 10−10 s) (Fig. 6d and Table S6†).30,31,41 In the related system, fast electron exchange between mixed-valent Fe2+/Fe3+ species has been proposed using Mössbauer studies.42,46
As the temperature was finally increased to 380 K, the dav(C–O) value of LA is further decreased to 1.252(8) Å, corresponding to a case of Cl2An2−, as well as LB (dav(C–O) = 1.26(1) Å) (Fig. 6e and Table S5†). Correspondingly, dav(Fe–O) = 2.096(5) Å is within the range for Fe2+–O (Table S5†). This feature satisfies the fact that the homovalent HT phase of [(Fe2+HS)2(Cl2An2−)3]2− was formed. Actually, the 57Fe Mössbauer spectra recorded at 380 K agreed well with the existence of the high-spin Fe2+ state,30,31 although some impurities that may have been produced by compound decomposition were isolated in HT measurements (Fig. 6f and Table S6†).47 Thus, the structures measured at 335 K and 380 K prove what was observed in the temperature dependence of Mm, i.e., two-step TDET behavior (vide supra).
The difference of T1/2(1) between 1 and 1-d, where T1/2(1)↑ = 317 K in 1-d, is much higher than T1/2(1)↑ = 237 K in 1, and this could be associated with the effect of Coulomb interactions between the layers and NPr4+ cations, in particular, between the LA Cl2An˙3− subunit and NPr4+. There are two kinds of LA⋯NPr4+ distances in 1, ca. 4.9 Å and 6.2 Å,31 where the shorter one could have a gain to be a higher valency of the (Fe3+–Cl2An˙3−)∞ chain owing to electrostatic attraction; meanwhile, it may result in electrostatic repulsion to adjacent chains through the LB Cl2An2− ligand. In addition to the fact that the other LA⋯NPr4+ distance (6.2 Å) is much longer than that, this situation in 1 could induce TDET between the LT and IMo phases at a lower T1/2(1), at least relative to that in 1-d (vide infra). Moreover, only one type of NPr4+ was assigned with the LA⋯NPr4+ distance of ca. 5.0 Å in 1-d, which evenly load electrostatic interaction to the LA ligand. This situation in 1-d could stabilize the LT phase much more than that in 1, so T1/2(1) in 1-d > T1/2(1) in 1.
First, a polycrystalline sample of 1-d (the solvent-free compound) was put into the SQUID apparatus at 300 K. Then, a temperature sweep of 300 K → 1.8 K → 300 K was applied for 1-d (Fig. 7a). During this process, the absence of TDET behavior was confirmed in this temperature region, corresponding to the magnetic data for the LT phase of 1-d. As the temperature increased to 400 K, stepwise charge variations from LT (1-d) to IMo (1-d) at T1/2b↑ and from IMo (1-d) to HT (1-d) at T1/2c↑ were observed (Fig. 7b). During the cooling process from 400 K to 300 K, the charge variation followed the heating process for 1-d as HT (1-d) → IMo (1-d) → LT (1-d) at T1/2c↓ and T1/2b↓, respectively (Fig. 7b). At 300 K, the sample was exposed to solution vapor of a 1:2 v/v mixture of water/acetone, for which the χmT value was monitored in a time course, showing a rapid decrease from 13.3 cm3 K mol−1 for LT (1-d) into a saturation with 10.1 cm3 K mol−1 for the IM phase of 1, i.e., 1′ within a few minutes (Fig. 7b and S12a†). Upon cooling to 1.8 K, the TDET for the transformation from IMo into LT in 1′ occurred at T1/2a↓, which was followed by TDET from LT to IMo at T1/2a↑ during the heating process up to 300 K (Fig. 7b). After this process, the sample of 1′ was kept standing at 300 K for 12 h with an adequate evacuation, for which the χmT value was monitored in a time course, showing a gradual increase from 10.2 cm3 K mol−1 for the IMo phase of 1′ followed by a saturation with 13.1 cm3 K mol−1 for the LT phase of 1-d owing to the desolvation process (Fig. 7b and S12b†). This reversible charge modulation of 1/1-d was confirmed by performing the solvation/desolvation process at 300 K over several cycles (inset of Fig. 7a and S12†). Consequently, this course of processes proves that the respective TDET processes of 1′ and 1-d can be combined using the desolvation/solvation processes at around room temperature for the same sample of compounds (Fig. 7 and S12†).
Furthermore, in situ magnetic measurements from 1 to 1-d were conducted in the heating process from 300 to 400 K at a rate of 0.5 K min−1 in a sweep mode (Fig. S13†). The desolvation of crystallization solvent from 1 spontaneously occurred as the temperature was increased up to ∼320 K, and resulted in the gradual variations of χmT values associated with the material state changes from IMo (1) to IMo (1-d).
During the heating process, the thermal variations of σ‖ profiles showed an unusual incoherent feature with two-step inflections at around T1/2b and T1/2c (Fig. 8b). The cooling process also follows the features of the heating process. The absolute value of σ is still within the range for common semiconductors, but interestingly, the HT phase region (350–380 K) of σ‖ showed metallic behavior, and even in the IMo phase region (310–350 K), such a metallic feature was observed at higher temperatures of T1/2b and T1/2c. This behavior may be due to electronic fluctuations that are closely associated with TDET at T1/2b and T1/2c. A similar feature was previously reported in DA systems involving TDET (i.e., N–I transition systems).18,50 The activation energy (Ea) for the semiconductor behavior in the LT phase and the IMo phase during the heating process is 438.4 meV (285–310 K) and 189.9 meV (328–347 K), respectively (Fig. S15†). The decreasing tendency of Ea values associated with the phase transition from LT to HT via IMo phases indicates that the electronic band structure was modified in each phase based on the occurrence of TDET. The inflections at around T1/2s were significant for σ‖; however, the thermal variation of the σ⊥c and σ⊥l values exhibited fewer features owing to the anisotropic environment of TDET behavior (Fig. 8c and d).
Even though the basic layer frameworks of 1 and 1-d are structurally very similar to each other, T1/2a and T1/2b for the LT ⇌ IMo TDET in 1 and 1-d, respectively, are largely different, which could be attributed to the differences in the electrostatic effects in 1 and 1-d, which resulted from their packing arrangement of NPr4+ cations. Thus, the desolvation/solvation process between 1 and 1-d caused a structural change. In particular, the packing arrangement of the anionic [Fe2(Cl2An)3]2− layer and the cations NPr4+ contributed significantly to modulating the electrostatic effect in the TDET of this series. This conclusion may answer the question regarding the reason why other series of (C)2[Fe2(Cl2An)3] with different C+ cations reported so far did not exhibit such TDETs within the common temperature range. The bulk effect brought about by electrostatic stabilization, i.e., Madelung stabilization, is very important to the tuning of TDET in ionic D/A systems.
Footnote |
† Electronic supplementary information (ESI) available. CCDC 1971740–1971743. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/d0sc00684j |
This journal is © The Royal Society of Chemistry 2020 |