Transition metal redox switches for reversible “on/off” and “slow/fast” single-molecule magnet behaviour in dysprosium and erbium bis-diamidoferrocene complexes

We present an in-depth experimental study of a new class of heterometallic, redox-switchable single-molecule magnets (SMMs).


Introduction
Single-molecule magnets (SMMs) have been attracting considerable attention for potential applications in quantum computing, high-density data storage and molecular spintronics. [1][2][3][4][5] This fascinating class of compounds is dened by a bistable magnetic ground state and an energy barrier (U eff ) to reorientation of their molecular spin. [6][7][8] The height of this barrier is determined by the magnetic anisotropy of the system and tunable via ligand eld considerations. While initial efforts in the eld focused on slow magnetic relaxation in multinuclear transition metal complexes, 9,10 complementary efforts have led to the development of lanthanide-based single-ion magnets (SIMs). 11,12 Lanthanide ions are attractive candidates for SIM applications, due to their large inherent magnetic single-ion anisotropy, leading to the development of SIMs with record barriers to spin reorientation, U eff . [13][14][15][16] With the ultimate-goal of utilizing SMMs in devices, the development of methodologies that will allow for the control of dynamic magnetic properties reversibly via external stimuli is an essential aspect. Possible stimuli include light, temperature, pressure, chemical, dc eld, and electric potential. [17][18][19][20][21][22] Redoxactive SMMs present an exciting route for the reversible modulation of magnetic properties using an electric potential. [23][24][25][26][27][28][29][30] Notably, the rst reported class of lanthanide ion-based SIMs, the Tb 3+ phthalocyanine double-decker complexes, are redox-switchable SIMs, undergoing changes in magnetization dynamics and hysteretic behaviour upon electrochemical generation of an open shell p-system. 12,31,32 There have been several reported examples of redoxswitchability in multinuclear transition metal and lanthanide based SMMs. In these systems, the oxidation and/or reduction of either a ligand 23,[33][34][35] or a metal centre [24][25][26]36 facilitates magnetic exchange interactions; turning the SMM properties "on" or "off". In addition to these "on"/"off" examples, in some cases adding and/or removing an electron has been shown to improve SMM properties. [27][28][29][30] Aside from the Tb 3+ and Dy 3+ double-decker lanthanide(III) phthalocyanine complexes, 12,37 to our knowledge there is only one previously reported example of proven redox controllable dynamic magnetic properties in a mononuclear lanthanide(III)based SIM. In that example an intramolecularly attached Ru 2/3+ redox switch was utilized to modify the magnetic relaxation dynamics of a Dy 3+ -based SIM. 29 The oxidation of Ru 2+ to Ru 3+ was found to enhance slow magnetic relaxation, either through perturbations of the ligand eld or the addition of another spincarrier to the system. However, this interesting system suffered from limited thermal stability at room temperature.
The reversible redox properties of ferrocene, FeCp 2 , make it an attractive moiety to include in complexes for redox switchability applications. 38 This inspired us to evaluate the possibility of utilizing the redox properties of ferrocene-containing ligands to modulate the dynamic magnetic properties of a nearby lanthanide(III) ion. Previously, Diaconescu et al. reported a homoleptic uranium(IV) compound stabilized by two bidentate diamidoferrocene ligands. 39,40 Notably, one-electron oxidation of the complex resulted in a mixed-valent species, indicating strong uranium-mediated electronic communication between the two iron sites. Interestingly, there have been no similar studies reported with lanthanide(III) bis-diamidoferrocene compounds.
Herein, we present the rst class of Ln 3+ ion-based redox switchable SMMs using the redox chemistry of ferrocene/ ferrocenium in the ligand scaffold. [41][42][43][44][45][46] We show how the reversible one-electron oxidation of Fe 2+ to Fe 3+ in Dy 3+ (oblate electron density) and Er 3+ (prolate electron density) bisdiaminoferrocene compounds modulates dynamic magnetic properties. Depending on the experimental conditions, these materials can exhibit switchability of their slow magnetic relaxation either between "on" and "off" or between "slow" and "fast". Remarkably, this is the rst example of redox switchable SMM properties observed in an Er 3+ compound. Additionally, this is the rst magnetic investigation of homoleptic, fourcoordinate Dy 3+ and Er 3+ complexes. This set of compounds has been characterized using X-ray crystallography, dc/ac magnetometry, 57 Fe Mössbauer spectroscopy, UV-vis-NIR spectroscopy and cyclic voltammetry.
Notably, this is the rst example of using the redox chemistry of a transition metal to alter the magnetization dynamics of a lanthanide ion, while maintaining thermal stability of all redox partners. This molecular level study is intended to provide design guidelines for future switchable solid materials.

Magnetic properties
Static-eld magnetic properties. To investigate static magnetic properties, direct current (dc) magnetic measurements were carried out using crushed polycrystalline samples under an applied eld of 1000 Oe from 300 K to 2 K (Fig. 3). The room temperature c M T value of 14.22 cm 3 K mol À1 for [1] À is in agreement with the expected value of 14.17 cm 3 K mol À1 for one isolated Dy 3+ ion ( 6 H 15/2 , S ¼ 5/2, L ¼ 5, g ¼ 4/3). Upon cooling, the c M T decreases gradually, then more rapidly below 130 K to reach a minimum of 12.55 cm 3 K mol À1 at 2 K. This decrease is mainly attributed to the thermal depopulation of excited states. Aer one-electron oxidation of [1] À to form 1, an increase in the room temperature c M T value to 14.36 cm 3 K mol À1 is observed. This increase is in accordance with the presence of a single Dy 3+ ion and an additional noninteracting S ¼ 1/2 site (low-spin Fe 3+ , 0.375 cm 3 K mol À1 expected for g ¼ 2 but actual g-value likely different). Upon decreasing the temperature, the c M T value of 1 decreases more rapidly than [1] À below 60 K, reaching a minimum of 10.13 cm 3 K mol À1 at 2 K. The more rapid decrease could be indicative of thermal depopulation of Stark sublevels or the presence of intra-and/or inter-molecular antiferromagnetic interactions. 47 For the Er 3+ analogues, the room temperature c M T value for [2] À is 10.82 cm 3 K mol À1 . This value is only slightly lower than the expected value of 11.28 cm 3 K mol À1 for a single isolated Er 3+ ion ( 4 I 15/2 , S ¼ 3/2, L ¼ 6, g ¼ 6/5). Aer oneelectron oxidation to 2, the room temperature c M T value increases to 11.43 cm 3 K mol À1 , corresponding to one Er 3+ ion and one non-interacting Fe 3+ ion. As observed in the analogous Dy 3+ compounds, there is a more rapid decrease in the c M T of the oxidized mixed-valent compound 2 at low temperatures.
The eld dependence of the magnetization (M) was measured for each compound with elds up to 70 kOe (7 T) over a temperature range of 2-8 K (Fig. S3-S10 †). In the Dy 3+ compound [1] À at 2 K, the magnetization increases sharply until Dynamic magnetic properties and redox switchability. Dynamic magnetic properties were investigated using alternating-current (ac) measurements. In [1] À , a signal in the out-of-phase component (c 00 ) of the ac susceptibility was observed under zero applied dc eld, indicative of slow magnetic relaxation (Fig. 4b). Cole-Cole (semi-circle) plots of the in-phase (c 0 ) vs. out-of-phase (c 00 ) components of the ac susceptibility were t to a generalized Debye model 10,52 to extract relaxation times, s (Fig. 4c). The natural logarithm of s was plotted vs. the inverse temperature to construct the corresponding Arrhenius plot (Fig. 4d). The large temperature independent region observed at low temperatures in the c 00 and Arrhenius plots of [1] À implies prevalent quantum tunnelling of the magnetization (QTM), a relaxation process that proceeds without the input of thermal energy.
The Arrhenius plot of [1] À (Fig. 4d) was t using least squares regression to a model that accounted for multiple relaxation processes, including, QTM, Raman, and Orbach processes (eqn (1)).
The parameters s QTM , C (Raman coefficient), s 0 (preexponential factor), and U eff (thermal barrier for Orbach relaxation) were treated as free-t parameters. For Kramers ions, parameter n 2 in the Raman pathway is expected to be 9; although allowing n 2 to be a free-t parameter resulted in a value of n ¼ 5. A value lower than 9 may be expected in systems with low-lying excited states if optical phonons are taken into consideration. 53,54 Only the Orbach process, the "over the barrier" pathway, appears linear on the plot of ln s vs. T À1 .
Due to the predominance of quantum tunnelling of magnetization (QTM) at low temperatures, few temperature dependent points were obtained. However, a barrier height, U eff ¼ 27.3(8) cm À1 (s 0 ¼ 1.63(2) Â 10 À6 s), for the thermally activated Orbach process was calculated for [1] À under H dc ¼ 0 Oe using eqn (1). All parameters of the t are listed in Table S5. † Notably, the one-electron oxidation product 1 shows no evidence of slow relaxation under zero applied dc eld at ac frequencies up to 1000 Hz ( Fig. S11 and S12 †). This is likely due to faster QTM in the oxidized species 1 than in the non-oxidized species [1] À . This could be attributed to both the lower symmetry around the Dy 3+ ion in the mixed-valent compound and/or the introduction of an S ¼ 1/2 site nearby the magnetically anisotropic Dy 3+ ion. In detail, changes in ligand eld and dipole-dipole interactions have been shown to facilitate mixing of magnetic states through which QTM can be introduced. Structural changes (and potentially packing effects) likely make the largest contribution to the change in magnetization dynamics. The appreciable loss of symmetry upon oxidation and change in ligand eld would undoubtedly inuence the orientation of the magnetic anisotropy axis and therefore the magnetization dynamics. In the absence of applied dc elds, the changes in magnetization dynamics upon reversible oneelectron oxidation of [1] À to 1 can be thought of as "on"/"off" switching of the slow magnetic relaxation (frequencies up to 1000 Hz) (Fig. 5).
For both Er 3+ compounds, [2] À and 2, no signal was observed in the out-of-phase component (c 00 ) of the ac susceptibility under zero dc eld at ac frequencies up to 1000 Hz ( Fig. S14 and S16 †); likely a result of efficient ground state QTM processes for these species.
QTM between orthogonal Kramers ground states is typically caused by a perturbation Hamiltonian that allows for mixing of energetically close lying states. Examples of such perturbations include the presence of transverse anisotropy and dipole-dipole interactions between paramagnetic metal centres. 55 To mitigate QTM in this series of compounds, various dc elds were applied during ac susceptibility measurements. Application of the static dc eld lis the degeneracy of the Kramers states, thereby reducing QTM. In the presence of a dc eld, compounds [1] À , 1, and [2] À show clear evidence of slow relaxation of the magnetization, with a signal in the out-of-phase component (c 00 ) observed within the ac frequency range of 1 to 1000 Hz (Fig. S18, S22 and S26 †). Notably, the oxidized mixed-valent Er 3+ compound 2 displayed no evidence of slow relaxation at any investigated temperature or eld (Fig. S30 †).
In the variable eld ac measurements for [1] À at 5 K, a transition from a faster relaxation process to a slower relaxation process is observed around 300 Oe (Fig. S18 †). An optimal static eld of 1000 Oe was determined from the maximum (slowest relaxation) in the plot of the eld dependence of s (Fig. 7, inset, Fig. S20 †). Therefore, variable temperature ac susceptibility measurements for [1] À were collected with a dc eld of 1000 Oe (Fig. 6a). At 1000 Oe, the maxima of the out-of-phase signals (c 00 ) are shied to lower frequencies relative to the zero eld measurements, as the reduction of QTM results in slower  magnetic relaxation. Furthermore, with the application of the 1000 Oe dc eld, the Arrhenius plot displays temperature dependence over the entire temperature regime (Fig. 7).
Arrhenius plots (ln s vs. T À1 ) for the applied eld measurements were t using least squares regression to a model that accounted for multiple relaxation pathways, including direct, QTM, Raman, and Orbach (eqn (2)).
To avoid over-parameterization during tting, the eld dependence of s was initially t for the two eld dependent processes, direct and QTM, to obtain the direct relaxation parameter A and the QTM parameters B 1 and B 2 , according to eqn (3). 53 Typically, n 1 ¼ 4 for Kramers ions in the absence of hyperne interactions. 54 Parameter D was added to account for the eld independent contributions from Raman and Orbach relaxation processes. 53 Parameters A, B 1 , B 2 , and D were treated as free t parameters. All tting parameters are listed in Tables S5 and S6. † For [1] À at 5 K, the very gradual decrease in s values at elds above 1000 Oe suggests minimal contributions of single phonon direction relaxation mechanisms (Fig. 7, inset, Fig. S20 †). 53 At elds below 1000 Oe, the increase in s with increasing eld was modelled successfully by the QTM term.
The parameters obtained from eqn (3) were held xed while tting the temperature dependent Arrhenius plots (ln s vs. T À1 ), according to eqn (2). For [1] À , the temperature dependence of s was t to direct, Raman, and Orbach relaxation processes (Fig. 7). Allowing the Raman exponent n 2 to be a free t parameter resulted in a value of n 2 ¼ 7. For Kramers ions, an n 2 value of 9 is expected; however, lower n 2 values may be anticipated if optical phonons are taken into consideration. 53 A barrier height of U eff ¼ 46(2) cm À1 (s 0 ¼ 7.3(7) Â 10 À7 s) was obtained; 18.7 cm À1 larger than the barrier calculated under zero dc eld.
Variable eld ac measurements for 1 at 2 K are displayed in Fig. S21-S22. † The eld dependence of s was t for both direct and QTM processes according to eqn (3) (Fig. 7 (inset),   Table S5 † for all fitting parameters. Fig. S24 †). A value of n 1 ¼ 2 was used to obtain a t, corresponding to a Kramers ion in a hyperne eld. 54,56 The maximum s value (slowest relaxation) occurs at a eld of 1750 Oe at 2 K. However, to maintain consistency with the ac measurements for [1] À , variable temperature ac measurements for 1 were carried out using a 1000 Oe dc eld (Fig. 6b). The maxima of c 00 for the mixed-valent compound 1 are shied to higher frequencies relative to [1] À , implying faster magnetic relaxation for a given temperature of the mixed valent compound 1 (Fig. 6b). The Arrhenius plot was t using eqn (2), accounting for direct, Raman, QTM and, Orbach processes, to give U eff ¼ 27.2(5) cm À1 (s 0 ¼ 5.0(4) Â 10 À7 s) (Fig. 7). Notably, the U eff for the mixed-valent species 1 is 18.8 cm À1 lower than the non-oxidized species [1] À . All tting parameters for 1 are listed in Table S5. † Redox switchability of the dynamic magnetic properties is best illustrated by comparing the relaxation times in the Arrhenius plot of [1] À and 1 (Fig. 7): the one-electron oxidation of [1] À to 1 results in faster relaxation times at a given temperature and a lower U eff value.
Using the above described methodology, variable-eld ac measurements were collected for the Er 3+ compound [2] À at 2 K ( Fig. S25-S27 †). The eld dependence of s was successfully t using eqn (3) (Fig. 8 (inset), Fig. S28, Table S6 †). Variable temperature ac susceptibility data were collected using an applied eld of 500 Oe (Fig. 6c). The corresponding relaxation times were t over the entire temperature region of the Arrhenius plot (Fig. 8), according to eqn (2), resulting in an extracted value of U eff ¼ 29(2) cm À1 (s 0 ¼ 4(1) Â 10 À7 s). The one-electron oxidized species, 2, did not display any signs of slow magnetic relaxation under dc elds as high as 1500 Oe in ac experiments (Fig. S30 †).
In summarizing the magnetic results, we nd that the prolate Er 3+ ion in the [2] À /2 system allows solely for reversible "on"/"off" switching of the slow magnetic relaxation (in the presence of a small dc eld) (Fig. 6c and d), while the oblate Dy 3+ ion in the [1] À /1 redox system enables bi-functional redox switchability of magnetic properties: "on"/"off" (no dc eld) (Fig. 5) and "slow"/"fast" (with dc eld) (Fig. 6a, b and 7).
The distinct shapes of f-electron density of Dy 3+ (oblate) and Er 3+ (prolate) result in very different magnetic anisotropy axes and therefore different magnetization dynamics under the same ligand eld and molecular symmetry conditions. Previously, isostructural Dy 3+ and Er 3+ complexes have been shown to exhibit different magnetization dynamics. In the fourcoordinate, trigonal pyramidal series of compounds [Li(thf) 4 ] [Ln{N(SiMe 3 ) 2 } 3 Cl]$2 thf, the Er 3+ analogue displayed SMM behaviour under zero dc eld, whereas the Dy 3+ analogue exhibited only eld induced SMM behaviour. [57][58][59] This difference was mainly attributed to the local symmetry and orientation of the magnetic anisotropic axis. The trigonal pyramidal geometry was found to not be ideal for either oblate or prolate ions but was relatively more favourable for prolate type ions, such as Er 3+ . 59 The coordination geometry of the bis-diamidoferrocene complexes presented here was found to be relatively more favourable for the oblate Dy 3+ ion than for the prolate Er 3+ ion. The differences in behaviour between the Dy 3+ and Er 3+ complexes are attributed to the orientation of the anisotropy axis under the same ligand eld conditions due to the difference in f-electron density. We believe the largest contributor to the change in magnetization dynamics upon oxidation to be the change in ligand eld and lowering of local symmetry. In detail, the oxidation of one of the diamidoferrocene ligands (in going from [1] À and [2] À to 1 and 2) results in inequivalent binding of the two ligands to the central lanthanide ion. Additionally, packing effects and the presence or absence of counter ions may contribute to the structural changes.
The magnetic anisotropy axis of 1 was determined utilizing a quantitative electrostatic model for the prediction of the orientation of the ground state anisotropy axis in Dy 3+ compounds described by Chilton et al. (MAGELLAN). 60 Considering charged ligands as point charges, the method minimizes electrostatic repulsion between the point charges and f-electron density. Using the molecular structure of 1, the anisotropy axis was determined under three different scenarios: (1) assigning both Fe centres as neutral, (2) assigning a +1 charge to the Fe centre closer to the Dy 3+ ion, and (3) assigning a +1 charge to the Fe centre further from the Dy 3+ ion (Fig. S35 †). The addition and location of the +1 point charge did not lead to considerable differences in the orientation of the anisotropic axis (Fig. S35 †). The location of the negative point charges (amide ligands) was found to be the largest contributor, with the magnetic anisotropy axis aligned in the same plane as the shorter Dy-N bonds.
Cyclic voltammetry for [1] À /1 and [2] À /2. The separation in electrochemical potentials between individual redox couples of Fig. 8 Arrhenius plot for the Er 3+ complex [2] À (black circles) under a 500 Oe applied dc field. The orange line represents the fit of the linear region (six highest temperature points) to the expression s À1 ¼ s 0 À1 exp(ÀU eff /k B T) (Orbach only); resulting in a U eff value of 26.9 cm À1 (s 0 ¼ 9.52 Â 10 À9 s). The red line represents fit of the entire temperature region to eqn (2), giving a U eff value of 29(2) cm À1 (s 0 ¼ 4(1) Â 10 À7 s). Inset: field dependence of the relaxation times (s) for [2] À (black circles), red line is fit to eqn (3). See Table S6 † for all fitting parameters. multi-redox systems can provide a useful initial estimate for the presence of electronic communication in mixed-valent species. The cyclic voltammogram of a solution of [1] À in thf, displays two quasi-reversible redox processes centred at E 1/2 ¼À1.00 and À0.540 V vs. [Cp 2 Fe] 0/1+ (Fig. 9, top/blue). These processes correspond to the Fe 2/3+ redox couples of the two individual ferrocene diamide ligands. The cyclic voltammogram of the free ligand exhibits one reversible redox process at À0.60 V vs.
[Cp 2 Fe] 0/1+ . 39 The potential separation of the two processes in  (Fig. 9, bottom/red), corresponding to K c ¼ 6.16 Â 10 8 . The large K c values in [1] À and [2] À suggest considerable electronic interaction between the two iron sites in solution and is consistent with a Robin and Day Class II classication. 61 These values are signicantly large but smaller than the values observed for the U 4+ analogue reported by Diaconescu et al. 39 Importantly, these observations are consistent with the formulation that the central f-block element is indeed critically involved in the electronic communication between the mixed-valent iron ions, resulting in larger electronic coupling for the actinide ion (U 4+ ) than for the trivalent lanthanide ions (Dy 3+ , Er 3+ ), featuring the more contracted frontier orbitals. 57 Fe Mössbauer spectroscopy and UV-vis-NIR spectroscopy.
We utilized 57 Fe Mössbauer spectroscopy to establish the presence of electronic communication of the mixed-valent iron ions in 1 in the solid state. In the 57 Fe Mössbauer spectrum of the all ferrous species [1] À in zero eld, a single characteristic doublet was observed, corresponding to the spectroscopically identical low-spin Fe 2+ ions in [1] À , with an isomer shi (d) of 0.54 mm s À1 and quadrupole splitting (DE Q ) of 2.34 mm s À1 (Fig. S36 †). The 57 Fe Mössbauer spectrum of the mixed-valent compound 1 exhibited two doublets (Fig. 10). The spectrum could in principle be t using both a two-site ( Fig. 10) model (two doublets) or a three-site model (Fig. S37 †) (one doublet, two singlets). Given the chemical nature of 1, we believe that only the two-site model is appropriate here. As such, we t the data using one doublet corresponding to the Fe 2+ ion (d ¼ 0.528 mm s À1 ; DE Q ¼ 2.238 mm s À1 ) and an equally contributing second doublet corresponding to the Fe 3+ ion (d ¼ 0.506 mm s À1 ; DE Q ¼ 0.490 mm s À1 ). The Mössbauer spectrum is consistent with a trapped valent system over the temperature range studied (5 K to 150 K; Fig. S38 †), in which electron transfer is slower than the time scale of Mössbauer spectroscopy ($10 À7 s À1 ). However, electronic communication between Fe 2+ and Fe 3+ is clearly present, as the signal corresponding to the ferric site is quadrupole split. In the absence of electronic communication this signal would be expected to occur as a singlet. 41 Furthermore, the smaller DE Q observed for the ferrous site in 1 as compared to [1] À is also consistent with the formulation of Fe 2+ /Fe 3+ electronic communication. Similar spectroscopic signatures in mixed-valent dinuclear iron complexes have previously been reported for [Cp* 2 Fe 2 (as-indacene)]c + . 62 We utilized UV-vis-NIR absorption spectroscopy to probe the degree of electronic communication in solution. A broad band associated with an intervalence charge transfer (IVCT) transition was observed in the near-IR region of dilute thf solutions of the mixed-valent compounds 1 and 2 at l max ¼ 1056 nm and 1043 nm, respectively ( Fig. S40 and S42 †). This broad, lowenergy IVCT band is indicative of electronic communication in solution and is consistent with a Class II system, 63 supporting the observations from CV and 57 Fe Mössbauer spectroscopy. Moreover, in solution, intermolecular communication is assumed to be negligible, indicating the presence of intramolecular electronic communication in these mixed valent systems. The UV-vis-NIR absorption spectra of [1] À and [2] À in thf displayed no bands in the NIR region ( Fig. S39 and S41 †).

Conclusions
Our studies highlight the utility of diamidoferrocene ligands in the construction of redox switchable SMMs. The high chemical reversibility of the Fe 2+/3+ redox couples can be exploited to alter magnetization dynamics of Dy 3+ and Er 3+ based SMMs. The mixed valent Fe ions in complexes 1 and 2 feature electronic communication with each other. Importantly, redox-switchable "on"/"off" and "slow"/"fast" SMM behaviour can be obtained depending on the experimental conditions and the nature of the four-coordinate lanthanide ion. The combined results of this comprehensive molecular level study are important contributions towards the development of rational molecular design guidelines for future switchable magnetic molecules and materials.

General considerations
All reactions and manipulations were carried out under anaerobic and anhydrous conditions in an argon lled glovebox (Vigor). All syntheses and manipulations were carried out using disposable plastic spatulas. Tetrahydrofuran (thf), hexanes, toluene and diethyl ether were dried and deoxygenated using a solvent purication system (JC Meyer Solvent Systems) and were stored over molecular sieves in an argon-lled glovebox. Anhydrous dysprosium(III) iodide and iodine were purchased from Alfa Aesar. Eicosane was purchased from Acros Organics. Sublimed anhydrous erbium(III) iodide was generously donated by Prof. Tim Hughbanks' group (Texas A&M). fc[HNSi(t-Bu) Me 2 ] 2 was prepared as previously described. 39 Benzyl potassium was prepared via the deprotonation of toluene by nBuLi/KO t Bu as described previously. 64 UV-vis-NIR spectra were recorded using a Shimadzu SolidSpec-3700 spectrophotometer over a range of 300 nm to 2000 nm and matched screw-capped quartz cuvettes. Elemental analyses were carried out by Midwest Microlab (Indianapolis, IN). added dropwise to the stirring DyI 3 suspension while cold. The reaction mixture was stirred for 1.5 h at rt. All volatiles were removed in vacuo and the solid was washed with hexanes and dried. A minimal amount of thf was added and the solution was ltered through Celite. Removal of the thf in vacuo yielded a yellow-orange solid. Air sensitive yellow-orange plate crystals were obtained from a concentrated thf solution layered with hexanes at À30 C for 24 h (117 mg, 76%). Unit cell parameters  (2). The same procedure as for 1 was followed using K(thf) 5

X-ray structure determination
Crystals suitable for X-ray diffraction were mounted on a nylon loop and placed in a cold N 2 stream (Oxford) maintained at 110 K. A BRUKER APEX 2 Duo X-ray (three-circle) diffractometer was used for crystal screening, unit cell determination, and data collection. The X-ray radiation employed was generated from a Mo sealed X-ray tube (K a ¼ 0.70173Å with a potential of 40 kV and a current of 40 mA). Bruker AXS APEX II soware was used for data collection and reduction. Absorption corrections were applied using the program SADABS. 65 A solution was obtained using XT/XS in APEX2. [66][67][68][69] Hydrogen atoms were placed in idealized positions and were set riding on the respective parent atoms. All non-hydrogen atoms were rened with anisotropic thermal parameters. Absence of additional symmetry and voids were conrmed using PLATON (ADDSYM). 70,71 The structure was rened (weighted least squares renement on F 2 ) to convergence. 68,72 For 1, the Si4(C39-C44) group was found disordered between two positions and was modeled successfully with an occupancy ratio of 0.55 : 0.45.

Magnetic measurements
A representative procedure for the preparation of the samples for magnetic characterization is as follows. Crystalline sample was crushed up into a ne powder before loading into a high purity 7 mm NMR tube (Norell). A layer of eicosane was added to the tube, covering the sample. The tube was then ame sealed under vacuum. To restrain the sample, the sealed tube was placed in a water bath (39 C) until the eicosane melted and was evenly distributed throughout the sample. The sample was loaded into a straw affixed to the end of the sample rod. Magnetic measurements were carried out using a Quantum Design MPMS 3 SQUID magnetometer (TAMU Vice President of Research). Dc susceptibility measurements were carried out over a temperature range of 1.8 to 300 K. Ac measurements were carried out using a 2 Oe switching eld. Data was corrected for diamagnetic contributions from the straw, sample tube, eicosane and core diamagnetism using Pascal's constants. 73 Cole-Cole plots were tted to the generalized Debye equation using least-squares regression. 52 Arrhenius plots and tau vs. H plots were t using least squares regression. 57 Fe Mössbauer spectroscopy Crystalline samples were loaded into a Teon cup inside a glovebox and covered with a layer of paraffin oil. The cup was brought out of the glovebox and immediately stored frozen in liquid nitrogen until measured. Mössbauer spectra were collected on a model MS4 WRC low-eld, variable temperature spectrometer (See Co., Edina, MN). Zero magnetic eld spectra were obtained by removing the 500 G magnets from the exterior of the instrument. Temperatures were varied using a temperature controller on the heating coil on the sample holder. The instrument was calibrated using an a-Fe foil at room temperature. Obtained spectra were tted using WMOSS soware (See Co.).

Electrochemistry
Cyclic voltammograms were measured in an argon lled glovebox (Vigor). Data were collected using a Gamry Instruments Reference 600 potentiostat with Gamry Framework soware. Glassy carbon working electrode, 1 mm diameter Pt wire counter electrode, and silver-wire pseudo-reference electrode were used. Scan rates of 100 mV s À1 to 250 mV s À1 were used.
Ferrocene was added at the end of each data collection and the ferrocene/ferrocenium couple was used as an internal standard.

Conflicts of interest
There are no conicts to declare.