Fluorinated mixed valence Fe ( II )-Fe ( III ) phosphites with channels templated by linear tetramine chains . Structural and magnetic implications of partial replacement of Fe ( II ) by Co ( II )

formula (H4baepn)0.5[Fe2.0Fe0.71Co1.29(H2O)2(HPO3)4-x(HPO4)xF4] (x≃0.38) 2 were studied by single crystal X-ray diffraction. The phase with the major content of Co(II), (H4baepn)0.5[Fe2.0Fe0.62Co1.38(H2O)2(HPO3)4-x(HPO4)xF4] (x≃0.38) 3 was achieved as polycrystalline powder and studied by Rietveld refinement by using the structural model of 2. These compounds have been characterized by ICP-Q-MS, thermogravimetric and thermodiffractometric analyses, and XPS, IR, 15


Introduction
One of the challenges to be faced in the field of open framework materials science and technology is the design and the obtaining of porous compounds which combine the classical properties of the zeolites with physicochemical properties such as magnetic, optical or conductive, arising from the introduction of transition metals in the crystalline structures. 1,2hesis of microporous materials such as zeolites, aluminosilicates and phosphates or other zeotypes, which receive interest because of their potential industrial applications as solid acid catalysts, ion exchangers or in the separation of gases 2,3 , is driven sometimes by using protonated amines as templates or structure directing agents (SDAs). 4y employing small organic amines as structure-directing agents, such as n-propylamine, n-butylamine, cyclopentylamine and cyclohexylamine extra-large-micropore compounds, for instance 16 and 18-membered-ring fluorinated gallium phosphates, 5 16-ring channels vanadium(III) phosphites 6 and 24-ring channels zinc phosphites 7 have been hydrothermally achieved.The crystallization of them is given by a cooperative templating mechanism of multiple organic cations which reside within the voids with their hydrophobic groups pointing toward 50 the center of the pores and the hydrophilic NH 3 + ends interacting via hydrogen bonding with the host.The assembly of small organic amines and their effect as SDA to direct the crystallization of large micropore aluminophosphites, zincophosphates, and aluminophosphates has also been 55   investigated. Recently, some gallium zincophosphites, NTHU-13, with channel expansion from 24-ring to 72-ring were successfully achieved by using heterometal centers and a series of aliphatic monoamines increasing the template from 4--carbon (4C) containing butylamine to 18C octadecylamine. 9ccording to the host guest charge matching concept a highly charged inorganic framework should be templated by highly charged organic amines such as multiamines. 10For instance, the chain-type polyamines such as TETA (10-atom-skeleton tetramine) 11 , AE-DAP (8-atom-skeleton triamine) 12 and 1,6-HDA (8-atom-skeleton diamine) 13 conducted to the crystallization of (C 6 H 22 N 4 ) 0.5 [Zn 3 (HPO 4 )(PO 4 ) 2 ] (16R-channel structure with 25.6% non framework space (NFS)), (C 5 H 18 N 3 ) 0.5 [Zn 3 (HPO 4 ) 3 (PO 4 )] (16R-channel structure with 10 36.2%NFS) and (C 6 H 18 N 2 )[Zn 4 (PO 4 ) 2 (HPO 4 ) 2 ]•3H 2 O (20Rchannel structure with 45.7% NFS) respectively.The role of such templates indicated that the size effect of a polyamine template is reduced with increasing number of amino groups and that NFS in the structures increases in proportion to the channel size.
However, W-M.Chang et al. observed opposite template effects for the synthesis of zinc phosphates with 16R channel structures by using linear triamine and pentamine molecules. Even today the specific templating role of the SDAs is not clear.In fact, the 'true' templating effect referring to a direct correlation between the van der Waals shape of the organic template and the channel space of the resultant inorganic framework is not very usual. 15In that regard, computer modeling 25 has demonstrated to be a useful tool for examining nonbonding interaction energies of different host-guest systems. 16n spite of the scientific community has made progress in the rational organic SDAs-mediated synthesis of zeolitic inorganic open framework materials, 4 the complex chemical reaction mechanisms in which the inorganic buildings nucleate around the organic cationic species are still poorly understood.Actually, there are many examples in which the SDAs represent the products of the initial organic precursors transformed during the hydrothermal synthesis by means of hydrolysis of linear polyamines, 17 decomposition of amides 18 or N-alkylation reactions. 19Therefore, because of the large number of parameters involved in the reaction processes still make the rational approach to the synthesis of these materials a challenging task, the "trial-and-error" strategy becomes necessary to explore 40 new architectures when new organic templates are being used.
On the other hand, the totally or partially replacement of the classical four-coordinated PO 4  3-building units by 3-connected groups such as the pseudo--pyramidal HPO  21 spin glass state 22,23 or spin canted antiferromagnetism. 24In particular, literature refers to the existence of two examples of mixed-valent iron purely inorganic phosphites, Li . 28n addition, many times the design and synthesis of magnetic 60 materials pass through the study of replacing the transition metals.For example, cobalt-iron bimetallic molecular magnets based on cyano-bridged coordination polymers 29 (Prussian Blue analogues) and oxalate-bridged complexes, 30 have been of great interest due to their functional magnetic properties, such as 65 photoinduced magnetism, 31 anisotropic photoinduced magnetism in thin films, 32 and a charge-transfer-induced spin transition. 33n this work, in order to satisfy the host guest charge matching principle, 10 the choice of the chain-type tetramine N,N'-Bis(2aminoethyl)-1,3-propanediamine (baepn = C 7 N 4 H 20 ) as the 70 structure directing agent, which may arise highly charged, is accompanied by introducing F -anion which also plays the role of template stabilizing the construction of the resulting network.While this molecule has been used for the construction of some transition metal coordination polymers, 34 wherein the organic 75 ligand is coordinated by its four nitrogen atoms to the equatorial plane of the metallic polyhedra, we ignore its use for the synthesis of zeotypes or related compounds.
In this paper, we report on the hydrothermal synthesis, crystal structure determination, as well as the thermal, spectroscopic  ) ions by means of the analysis of the crystal structures of 1 and 2 will be discussed.Moreover, the differences in the magnetic properties between 1 and the phase 90 with the major content of cobalt achieved, 3, will be analyzed.

Synthesis and characterization
Compounds 1, 2 and 3 were obtained as pure phases from different reactions.The synthetic procedure consists of dissolving 95 a mixture of H 3 PO 3 (7.5 mmol) and the metal salts (FeCl 3, 0.45 mmol for 1; FeCl 3 :CoCl 2 , 0.3:0.3mmol for 2 and FeCl 3 :CoCl 2 , 0.525:0.525mmol for 3) in 30 ml of distilled water.Then, 0.5 ml (13.9 mmol) of HF was added to the resulting solution and finally the pH was increased up to approximately 2.5 by adding 100 dropwise the organic molecule N,N'-Bis(2-aminoethyl)-1,3propanediamine (C 7 N 4 H 20 ).The reaction mixtures were sealed in a PTFE-lined stainless steel pressure vessel (fill factor 65%) and heated for 4 days at 170 ºC.After the reaction, black tabular single crystals of 1 and purplish products as prismatic single 105 crystals and as powder for 2 and 3 respectively were obtained.The yield for the three compounds is around 50-60% (based on the reagent iron).The existence of Co in the samples was confirmed by XRF as preliminary characterization test.
After many attempts, compound 2 represents the phase as 110 single crystal with the higher cobalt content achieved.The obtaining of compound 3 reveals that it is possible to achieve higher cobalt contents as powder sample.However any attempt to increase the cobalt content of the FeCl  The densities of the three phases were measured by flotation using a mixture of CH 2 I 2 and CHCl 3 , being 2.50(3) g.cm -3 for 1, 2.51(3) g.cm -3 for 2, and 2.51(2) g.cm -3 for 3.

Single crystal X-ray diffraction study
30 High quality single crystals for 1 and 2 were selected under a polarizing microscope and mounted on a glass fiber.Intensity data were collected at 100 K on an AGILENT SUPERNOVA diffractometer using a CCD (Eos) detector.Data frames were processed (unit cell determination, intensity data integration, 35 correction for Lorentz and polarization effects, 36 and analytical absorption correction 37 taking into account the size and shape of the crystals) using the corresponding diffractometer software package. 38eciprocal space analysis of compound 1 revealed the 40 presence of weak reflections not arbitrary distributed but equidistant from the main reflections along the a* axis (Fig. S1).These additional less intense peaks, referred to as satellite reflections, are the result of the existence of a structural modulation periodic in nature, which describes a long-range 45 order.The modulation wavevector which describe the first-order satellites with respect to the main reflections is q = 0.284(2)a* (Fig. 1), allowing the indexing of the diffraction pattern with four indices hklm.However, the low intensity of the satellite reflections together with the existence of diffuse scattering (Fig. 50 S2) hinder the superspace group and the atomic modulation functions determination in order to describe the incommensurately modulated superstructure.While satellite reflections mark an incommensurate long-range order, diffuse scattering processes suggest the existence of a certain disorder, 55 dynamic or static, that breaks slightly the crystallographic order.On the other hand, compound 2 shows neither satellites nor diffuse scattering (Fig. 1 and Fig. S2).
Fig. 1 Experimental images of the hk-1 layers of the diffraction patterns for 1 and 2 reconstructed using the CrysAlisPro software package. 38efore, we focused on the average structures, ignoring the satellite reflections.The structures were solved by direct methods, SHELXS 97 computer program, 39 in the monoclinic space group P2 1 /c, and then refined by the full matrix leastsquares procedure based on F 2 , using the SHELXL 97 computer program 40 belonging to the WINGX software package. 41This procedure allowed us to find the positions of the iron and phosphorous atoms, and all the other non-hydrogen atoms (F, O, C and N) were placed from subsequent Fourier-difference map calculations.70 For both compounds, half organic molecule is located in the asymmetric unit and a symmetry centre generates the other half causing a symmetry disorder of the C(4) atom with an occupation factor of 0.5, giving rise to the existence of two conformers (A ] and B).Due to this peculiarity, an hydrogen disorder system was modeled for their parent C(3) atom, which is bonded to C(4) (Fig. S3).Isotropic thermal displacements were used for carbon and nitrogen atoms.
From subsequent Fourier-difference map calculations, electron-densities exceeding 1.0 e -•Å -3 and too long P-Q distances for apical hydrogen were observed near P(1), P(2) and P(3).The presence of both phosphite and phosphate groups was considered in these phosphorus sites.The existence of hydrogenphosphate groups, (HPO 4 ) 2-, were deduced from electroneutrality requirements.Hydrogen and oxygen atoms implicated in the anionic phosphorous groups replacements were treated as disordered ones with complementary occupation factors.In order to model correctly the disorder systems, P-H (1.30(1) Å) and P-O (1.50(1) Å) bonds were restrained to the ideal values.Moreover, in order to maintain the geometry of the (HPO 3 ) 2-groups, the three distances of the apical hydrogen atoms to their corresponding oxygen atoms forming the pseudotetrahedra base were refined as a variable.
For compound 2, Co 2+ ions should be located in the M(3) and M(4) sites given the metal-ligand distances.The best fit provides a full occupation of the M(4) position and an occupation factor of 0.3(1) for M(3) in good agreement with the chemical analysis.
For the final refinement of compounds 1 and 2, the hydrogen atoms related to the coordinated water molecules were first located and placed in geometrically ideal positions (O-H: 0.82(1) Å; H-H: 1.35(2) Å) and refined using the riding mode.Anisotropic thermal parameters were used for all the atoms belonging to the inorganic framework except for the hydrogen atoms and, the oxygen atoms implicated in the P(1), P(2) and P(3) disorder systems.
After the refinement of both phases, two residual density maxima of approximately ±3 e -•Å -3 are located near N(2).In compound 2, there are also another five peaks located along the chain suggesting positional disorder of the organic template.Furthermore, in compound 1 two electron density peaks of ±3.4 and 3.2 e -•Å -3 are found at the equatorial plane of the iron [M(4)O 2 F 2 (H 2 O) 2 ] octahedron as well as a peak of around ±2.3 e - •Å -3 at 1 Å from the P(4) atom.Such peaks located in the inorganic skeleton of 1 are also located in compound 2 but with significantly lower density values, around ±1 e -•Å -3 .In case of compound 1, such anomalies are caused by the average structure resolution neglecting the satellite intensity due to the periodic character of the modulation.Details of crystal data, data measurement and reduction, structure solution and refinement of the phases 1 and 2 are reported in Table 1.The selected bond distances and angles are reported in ESI, Tables S1 and S2.Structure drawings were made using the ATOMS 6.2, 42 VESTA 3.1.7 42and TOPOS 4.0 programs. 43Powder X-ray diffraction X-ray powder diffraction for qualitative phase analysis of phase 3 using the Rietveld method with the FullProf program 44 was recorded using a Bruker D8 Advance Vario powder diffractometer equipped with a Cu tube, Ge(111) incident beam monochromator (Cu-Kα 1 = 1.5406Å), and a Sol-X energy dispersive detector.The sample was mounted on a zero background silicon wafer embedded in a generic sample holder.Data were collected from 8 to 80° 2θ (step size = 0.02º and time per step = 90 s) at RT.Fixed divergence and antiscattering slit giving a constant volume of sample illumination were used.60   For the Rietveld refinement of 3, obtained as powdered sample, the structural model of 2 was used.The cobalt content was fixed to the amount calculated by the ICP-Q-MS analysis.The fitting of the profile parameters followed by refinement of the atomic coordinates and the atomic displacement parameters of 65 the inorganic skeleton was performed.The organic template was not refined.Four different isotropic displacement parameters were refined: one for the iron and cobalt atoms, and the remainder for de phosphorous, oxygen and fluorine atoms.Some soft constraints were included to have a chemically correct 70 structural model.S3 to S5 for atomic coordinates and bond distances and angles of the Rietveld refinement of 3).
In addition, the pattern matching analysis of the diffraction patterns of the three compounds confirming the purity of the samples is given as Supplementary Material (Fig. S5).

Physicochemical characterization techniques
Thermogravimetric analyses were performed on a SDT 2960 simultaneous DSC-TGA TA instrument for 1 and on a Netzsch STA 449C one for 2 and 3. Alumina crucibles containing around 10 20 mg of every sample were heated in air at a rate of 5 °C/min from room temperature to 800 °C.Temperature dependence Xray diffraction experiments for 1 and 3 were carried out in air with a Bruker D8 Advance diffractometer (Cu Kα radiation) equipped with, a variable-temperature stage (HTK2000), a Pt 15 sample heater and a Vantec high-speed one dimensional detector with six degrees of angular aperture.The powder patterns were recorded in the 8≤2θ≤38º range (step size = 0.033º and time per step = 0.4 s) at intervals of 15 ºC, increasing the temperature at 10 ºCmin -1 from room temperature to 810 ºC.The IR spectra (KBr 20 pellets) were obtained with a JASCO FT/IR-6100 spectrophotometer in the 400-4000 cm -1 range.Diffuse reflectance spectra were registered at room temperature on a Varian Cary 5000 spectrophotometer in the 200-2500 nm range.X-ray photoelectron spectra for 1, 2 and 3 were acquired with a SPECS (Berlin, Germany) system equipped with a Phoibos 150 1D-DLD analyzer and monochromatic AlKα radiation (1486.6 eV, 300 W, 13 kV), with a multi-channel detector.Spectra were recorded in the constant pass energy mode at 80 eV for survey spectra and 30 eV for high resolution spectra, with an electron 30 take-off angle of 90º.The spectrometer was previously calibrated using the Ag 3d  45 Magnetic measurements on the powdered samples were performed in the temperature range 2.0-300 K for 1 and 3, at 0.05, 0.2 and 1 T using a MPMS-7T SQUID magnetometer and a PPMS-system both from Quantum Design.

45
Heat capacity measurements for 1 and 3 were carried out by a standard two-τ relaxation method, using a PPMS-system, with magnetic fields up to 9 T and temperatures down to 2 K.
The average structure of compound 1 is constituted by a threedimensional lattice formed by inorganic layers of iron (III) and iron (II) octahedra joined by the P(1) and P(2) phosphite groups (Fig. 2a).These layers link between them through the P(3) and P(4) bridging HPO 3 units (Fig. 2b).The iron (III) octahedra Fe  2a).These anionic layers stack along the [100] direction generating the open channels along the same direction, where the protonated templates are placed longitudinally (Fig. 2a and 3a), neutralizing the negative charge excess and stabilizing the inorganic building through hydrogen ] bonds (Fig. 3b).
The two types of iron (III) octahedra chains, formed by the [Fe(1)O 4 F 2 ] octahedra linked together by the P(1) polyhedra and the [Fe(2)O 4 F 2 ] octahedra linked by the P(2) polyhedra, alternate in the [010] direction .Notice that the direction of the P(1)-H (1)     and P(2)-H(2) bonds are opposite.Moreover, it can be observed that the octahedra are linked between them through fluorine atoms along the b axis giving rise to an infinite −Fe 3+ −F−Fe 2+ −F− linkage (Fig. 2a).The M(4)O 2 (H 2 O) 2 F 2 octahedron is not involved in the layers stacking because of the two water molecules are coordinated at the apical positions roughly parallel to the stacking direction.
In the Cambridge Structural Database (CSD) there are only 60 three structures, based on ionic salts, containing uncomplexed N,N'-Bis(2-aminoethyl)-1,3-propanediamine.In two of them the molecule is a tetracation 49 and in the other one is a trication, 50 with gttttttg (g indicates gauche and t trans) and gttggggt conformations respectively.In the studied compounds, the 65 existence of the symmetry centre forces the splitting of the central carbon (C4) in two positions producing the conformers A and B with conformations not previously seen in the mentioned structures.Given the existence of torsion angles less than 30º and higher than 90º (See torsion angles in Tables S1 and S2), the  The thermodiffractometry in air of 1 and 3 reveal that their thermal stability limits are situated at 240 and 255ºC, respectively, in such experimental conditions (Fig. S8).At these temperatures and as a consequence of the elimination of the two coordinated water molecules, the compounds become amorphous and, at 570 ºC the inorganic residues crystallize, in good agreement with the exothermic peaks observed in the DTA and DSC curves.The X-ray powder diffraction patterns of the residues obtained at 800 ºC show the presence of FePO S.G.P-1 (2), a= 7.909 Å, b= 9.289 Å, c= 6.342 Å, α= 108.48º, β= 101.52º, γ= 104.67º) for 3.It is probably that the increase in weight observed during the TGA studies around 450 ºC is due to the oxidation of P(III) to P(V), provoking the crystallization of the mentioned inorganic phosphates.The thermal evolution of the cell parameters for 1 and 3 was determined by pattern matching analysis (Fig. S9).The volume of the phases exhibits a constant increase with thermal expansion coefficients around 40•10 -6 and 34•10 -6 ºC -1 in the 30-195 ºC temperature range.

Infrared, UV-Vis, XPS and Mössbauer spectroscopy
In the infrared spectra of the three compounds (Fig. S10), the organic template is mainly represented by a group of overlapped maxima corresponding to stretching vibrations (ν) of N-H and C-H bonds in the 3100-2500 cm -1 region.The δ(NH 3 + ) band appears at 1600 cm -1 , indicating that the organic molecules are protonated and not coordinated to the inorganic building.
Regarding the inorganic part, around 2400 cm -1 a narrow band split in two ones, corresponding to the stretching vibrational mode of the P-H bond from the (HPO 3 ) 2-groups can be observed in the spectra.At lower frequencies, in the 1100 -400 cm -1 range, the bands corresponding to the P-O bonds vibrations of the phosphite/phosphate groups are observed.The ν(O-H) absorption band related to the coordinated water molecules is also observed at around 3420 cm -1 .
The diffuse reflectance spectra of 1, 2 and 3 show mainly four bands (Fig. 4).In the diffuse reflectance spectrum of 1, two bands at approximately 10640 and 7520 cm -1 are observed.These bands are characteristic of the iron(II) d 6 -high spin cation in a slightly distorted octahedral environment, and correspond to the electronic transitions from the 5 T 2g ( 5 D) fundamental state to the excited level 5 E 2g ( 5 D) that is splitted as a consequence of the existence of the non-regular octahedra.The energy associated with this transition corresponds, according to the Tanabe-Sugano diagram 54 to the Dq parameter.The value obtained is Dq= 910 cm -1 .An overlapped band situated around 13510 cm -1 which corresponds to the forbidden transition from the 6 A 1g ( 6 S) ground state to the 4 T 1g ( 4 G) term is attributed to the presence of iron(III) d 5 -high spin configuration cations. 55An intense band can also be observed at approximately 18180 cm -1 .This band is likely to correspond to the intervalence transition between the Fe 2+ and Fe 3+ cations, because their polyhedra share the four crystallographically independent fluorine atoms and the intermetallic bond distance are short enough (between 3.57 and 3.69 Å). 27 In the spectra of 2 and 3 compounds the discrimination of the transition bands characteristic of the Co 2+ cations is not a  The chemical composition and the possible oxidation state of 70 the iron and cobalt metals of the upper layers in these compounds were investigated by XPS measurements.The fitting together with the binding energies of the N 1s, P 2p, F 1s, Fe 2p and Co 2p 1/2 peaks are deposited as Supplementary material (Fig. S11).
The N 1s peaks are resolved into two peaks at around 399 and 75 401 eV which could be attributed to the existence of two primary and two secondary protonated amines per organic molecule.The P 2p peaks are deconvoluted into two doublets registered at binding energies between 130 and 133 eV and assigned to P(III) species, which used to be lower than observed for P(V) species. 56 The F 1s peak centered at around 684 eV may be attributed to the metal-fluorine bonds. 57The decomposition of Fe 2p previously reported for the mixed valence iron oxide Fe 3 O 4 . 59ikewise, compounds 2 and 3 show a Co 2p 1/2 main line at a binding energy around 797 eV, which is close to that found in a recently published cobalt(II) hybrid phosphite. 60The presence of the strong satellite peak at about 6 eV from the Co 2p 1/2 95 component is a further evidence for Co 2+ species. 61he powder 57 Fe room temperature Mössbauer spectra of the three compounds were studied as shown in Fig. 5.The best fit of 1, 2 and 3 leads to one doublet for iron(III) cations and another one for iron(II), obtaining Fe 3+ /Fe 2+ area ratios of 1.35, 2.80 and 100 ] 3.20 respectively.The values of the isomer shift and quadrupolar splitting parameters given in Table 2 show the characteristic values for the Fe 3+ and Fe 2+ cations.Given the multiplicity of the four crystallographically independent positions for the iron atoms and the metal-ligand distances obtained from the single-crystal structure analysis of compound 1, the Fe 3+ are distributed over the positions Fe(1) and Fe(2) and the Fe 2+ in the M(3) and M(4) positions.However, the percentage of Fe 3+ exceeds 50%, so we interpret that a partial substitution of Fe(II) by Fe(III) (Occ.= 0.296) occurs in the M(3) position (Fig. 6) as this presents a less 10 variation of the metal-ligand distances range than the M(4)O 2 (H 2 O) 2 F 2 octahedron.
Taking into account the structural features of compound 2, it is assumed that the Fe 3+ ions occupy only the Fe(1) and Fe(2) positions.Therefore, considering the amount of cobalt obtained 15 from the ICP-Q-MS analysis and the results from the structural refinement, the Fe 2+ are totally replaced by Co 2+ ions in the M(4) position and partially substituted in the M(3) position (Occ.= 0.290) (Fig. 6).   However, the behaviour of 3 at low temperatures is somewhat different, in such a way that the maximum which FC curve displays at 28 K is quite 40 sharp, besides that the susceptibility decreases much more than 1 after the maximum, reaching a minimum at approximately 10 K. Afterwards, χ m curve increases until 5K without changes in curvature.These behaviours at low temperature do not correspond either to a typical ferromagnet or antiferromagnet, 45 although it looks more like that observed in antiferromagnets, specially in 3.
The experimental data in both cases follow the Curie-Weiss law above 100 K (see lower insets in Fig. 7), allowing the calculation of the Weiss temperature (1, θ = -95.6K and 3, θ = -50 125 K) and the average effective paramagnetic moment per metal ion (1, µ eff = 5.60 µB and 3, µ eff = 5.95 µB).The value of µ eff for 1 is intermediate between the expected for Fe 2+ (5.4 µ B ) and Fe 3+ (5.9 µ B ), whereas that for cobalt-substituted compound is slightly higher than the expected for Fe 3+ , probably because the Co ions 55 present an extra contribution coming from the angular momentum.The negative Weiss temperatures together with the decrease of the χ m T products with decreasing temperature indicate that the main magnetic interactions in these compounds are antiferromagnetic.The abrupt increase of the susceptibility below 20 K for 1 and the sharp maximum close to 28 K for 3 correspond to the onset of AF order, as will be described later from M(H) curves and heat capacity (Cp) data.
The upper insets of Fig. 7 show the low temperature ZFC and FC molar susceptibility (χ m ) details performed at 0.5, 2 and 10 kOe.For 1 a small irreversibility appears at 0.5 kOe below 30K and although practically disappears for fields higher than 2 kOe, a tiny contribution persists up to 10 kOe.For 3 a small splitting of the curves can also be observed but whose intensity is much less 10 dependent on the magnetic field than in the compound 1.Besides, the minimum location is reliant on the magnetic field, shifting from 9 to 16 K as increasing the field from 0.5 to 10 kOe for 3, while for 1 the position remains practically unchanged at 20K regardless of the applied field.The existence of irreversibility can 15 only be attributed to the existence of a weak ferromagnetic component.The hysteresis decreases as the temperature increases, disappearing in the paramagnetic region.The temperature increase also provokes a positive shift of the critical field of the metamagnetic transition (upper inset in Fig 8a).Above 20K, the metamagnetic transition disappears as well as the M(H) curves are completely linear up to 90 kOe (Fig. 8b).In the case of compound 3, the hysteresis loops do not show any significant remnant magnetization.However, there is also a critical field close to 62 kOe at 2K indicating the existence of a metamagnetic transition (Fig. S12, Supplementary Material).10   To get a further insight of the weak ferromagnetic component of 1, we have measured the thermoremanent magnetization (TRM) as well as the ac magnetic susceptibility.The TRM curve was measured at H = 0 after cooling from T≥T N under a magnetic field of 5 kOe.As can be observed in Fig. S13, the value decreases continuously as increase the temperature, reaching zero above the Neel temperature.The appearance of a positive remanent magnetization below T N confirms the existence of a weak ferromagnetic component.In addition, it is important to signal the appearance of a clear shoulder at around 6 K as was 20 also pointed out in the FC curve.The real (χ´) and imaginary (χ´´) components of the ac susceptibility at 1000 Hz with an ac field of 10 Oe are shown in Fig. S14.χ´shows a broad maximum centred at 28 K (T N ) whereas χ´´, which should be zero in antiferromagnetic compounds, has a small contribution associated 25 with the weak ferromagnetic component.Surprisingly a sharp peak emerges in both χ´and χ´´ around 4.5 K indicating the presence o a magnetic transition of different nature from the previous one.This transition was masked in the magnetic susceptibility by the weak ferromagnetic component but should 30 be the responsible of the shoulder observed around 6 K in the FC and TMR curves.To check the origin of this low temperature transition we have measured the ac susceptibility at different frequencies (inset of Fig. S14).The peak height decreases and the position of the maximum shifts to higher temperatures with increasing frequency, as usually happens in spin glass transitions.The spin glass nature of the transition is further confirmed by the absence of any anomaly at 4.5 K in the heat capacity data.The coexistence of antiferromagnetism and spin glass behaviours has been previously observed in other insulator materials. 23,63This 40 result could be due to the simultaneous but random presence of Fe(II) and Fe(III) cations in the M(3) position of compound 1 (Fig. 6), which gives rise to the competing interactions.Both randomness and competing interactions lead to spin frustration, ultimately resulting in the spin-glass state.
Considering the structural features of compounds 1 and 2, studied by single crystal X-ray diffraction, and taking into account that 3 is isotypic, several magnetic pathways can take place in these three phases.Inside the layers metal-metal distances range between 3. The specific heat curves represented in Fig. 9 show a small maximum centered at 20.5 K for 1 and a sharp magnetic peak at 28 K for 3. Despite the anomaly observed for 1 does not have the typical appeareance of a λ-type second-order transition peak as in 75 the case of the cobalt containing compound, it can be associated with the establishment of a three-dimensional antiferromagnetic order in good arrangement with the magnetic-susceptibility data.It is important to signal the absence of any additional anomaly at low temperatures, confirming the spin glass nature of the 80 transition observed in the ac susceptibility at 4.5 K.The continuous growth of Cp at higher temperatures is due to the lattice contribution (Cp pho ), which does not show any tendency to saturation.In fact, the Cp values at 300 K for both compounds are around 660 J/molK, still far from the expected values (1, 1287.8 85 and 3, 1294.0J/molK) according to the Dulong and Petit law for the 51.6 and 51.9 ions per unit formula for 1 and 3, respectively.This is due to the presence of light atoms with very high excitation energy.
In order to extract the magnetic contribution, C mag , the Cp pho 90 was estimated using the Debye model and considering the existence of three Debye temperatures (the minimum number of free parameters that will allow us to fit the experimental data).In this way, if the number of atoms in the unit cell is N, we suppose n 1 atoms with a Debye temperature θ D1 , n 2 atoms with a Debye 95 temperature θ D2 , and n 3 = (N-n 1 -n 2 ) atoms with a Debye temperature θ D3 .Therefore, there are five free parameters, namely, n 1 , n 2 , θ D1 , θ D2 , and θ D3.This approach has been used successfully in previous studies in other hybrid compounds. 64The best fittings are obtained for n 1 = 10.The value of Cp mag in the maximum is clearly larger for 3, 46.5 J/mol K, than for 1, 34.5 J/mol K.In addition, whereas for 3 Cp mag has a λ-peak shape, for 1, Cp mag has a triangular shape, extending up to 60 K.These findings also indicates that compound 1 has a more complex magnetic structure than compound 3.
Heat-capacity has been also studied in the presence of several magnetic fields.With increasing field for 3, the sharp λ-type magnetic peak becomes more rounded and shifts to lower 10 temperatures, what is in good agreement with a global antiferromagnetic behaviour as was also observed from the magnetic susceptibility.In the case of compound 1, an unexpected behaviour was observed.The small maximum shifts to higher temperatures, grows with the magnetic field and 15 becomes better defined.Usually, the effect of the magnetic field on ferromagnetic transitions consists of shifting the λ anomaly to higher temperatures making it more rounded and less height.Therefore, the small value of the anomaly observed in 1 could be associated with an incommensurate magnetic structure, in which the entropy difference with the paramagnetic state is lower than in commensurate structures.The proposal of this kind of structure is also supported by the behaviour of the anomaly with the applied magnetic field.A similar behaviour of the specific heat anomaly was observed in the compound Co 2 (OH)AsO 4 , 62 and in that case it was interpreted as an evolution of the magnetic structure to the commensurability.

Conclusions
The mild hydrothermal technique was used for the synthesis of three novel 3D open-framework fluorinated mixed valence Fe(II)-Fe(III) phosphites with channels templated by protonated tetramine chains.The combined information from chemical analysis, X-ray diffraction and Mössbauer spectroscopy has allowed us to determine the Fe 3+ , Fe 2+ and Co 2+ cation distribution over the four metal sites.While Fe(1) and Fe(2) positions have a full occupancy of Fe 3+ in the three phases, M(4) is occupied by Fe 2+ in 1 but totally replaced by Co 2+ in 2 and 3. M(3) position presents partial occupations: 0.30 Fe 3+ :0.70 Fe 2+ for 1, 0.29 Co 2+ :0.71 Fe 2+ for 2 and 0.38 Co 2+ :0.62 Fe 2+ for 3.
The average structure resolution of 1 neglecting the satellite intensity leaves electron density peaks above ±3e -•Å -3 , which are found at the equatorial plane of the [M(4)O 2 F 2 (H 2 O) 2 ] octahedron and near the nitrogen N(2) atom belonging to the organic template.The reciprocal analysis of 1 shows satellite reflections, described by a q = 0.284(2)a*, which are the result of an incommensurate long-range order involving the crystallographic [100] direction.It is logical to think that the incommensurate 50 structure is related to an incommensurate disorder of the organic molecule located along the [100] channels within the inorganic framework.This long-range order of the template seems to affect also the inorganic skeleton, because of the extensive network of hydrogen bonds, as it is deduced from the residual electron 55 densities located near the M(4) and P(4) sites.This incommensurability is not observed in 2. Therefore, while the template movement along the a axis describes a structural modulation of the structure of 1, in 2 has not a periodic character, only observing local disorders.60   Magnetic measurements of 1 and 3 are consistent with the existence of major antiferromagnetic interactions.Compound 1 also exhibits weak ferromagnetism at low temperature and an additional spin glass transition at the freezing temperature T f = 4.5 K. Despite the expected increase in random magnetic 65 interactions with Co substitution because of its anisotropic nature, the freezing temperature and the weak ferromagnetism disappears in compound 3. So, the presence of Co(II) ions in both M(3) and M(4) sites can eliminate the magnetic frustration.Specific heat curves show a small anomaly at 20.5 K for 1 and a sharp 70 magnetic peak at 28 K for 3. Strikingly, the anomaly observed in 1 grows with the magnetic field and becomes better defined.Such characteristics as the low value and its behaviour with the applied magnetic field suggest that this is due to the existence of an incommensurate magnetic structure as was interpreted in 75 compound Co 2 (OH)AsO 4 . 62inally, given the features observed in the structural analysis matching the existence of a structural modulation along the [100] ] direction, it seems inevitable to think about the existence of some kind of relationship between this and the phenomenon observed in the study of the specific heat of the compound 1.In fact, the existence of electron density peaks located in the M(4) site, occupied by Fe(II) cations, leads us to tentatively associate the incommensurate magnetic structure of 1 to a modulation of mixed valence Fe(III)-Fe(II) cations.It should be recalled that M(4) site is totally occupied by Co(II) ions and does not show residual maxima around in phase 2. This work has been financially supported by the "Ministerio de Ciencia e Innovación" (MAT2010-15375 and MAT2011-27573-C04), the "Gobierno Vasco" (IT630-13) and the "UPV/EHU" (UFI11/15), which we gratefully acknowledge.The authors thank the technicians of SGIker (UPV/EHU), financed by the National Program for the Promotion of Human Resources within the National Plan of Scientific Research, Development and Innovation, "Ministerio de Ciencia y Educación" and "Fondo Social Europeo" (FSE), for the X-ray diffraction, XPS, chemical and spectroscopic measurements.J. Orive wish to thank the 20 Universidad del País Vasco, UPV/EHU for funding.View of the reciprocal space and reconstruction of the h0l layers of 1 and 2, conformers A and B of the organic molecule, Rietveld structure analysis of 3, thermal ellipsoid plots of 1 and 2, thermal analysis, infrared spectra, XPS spectra, magnetization vs applied magnetic field at different temperatures for 3, thermoremnant magnetization and ac magnetic

50
The asymmetric unit of 1 contains 35 non-hydrogen atoms, 29 of which belong to the host framework (four Fe, four P, four F and seventeen O) and the remaining 6 to the guest species (two N and four C).Three of the oxygen atoms (O1´, O2´ and O3´) present partial occupation as they belong to the hydrogenphosphate or 55 phosphate groups which partially replace the phosphite units.All of crystallographically independent atoms occupy general positions except the C(4) atom which present half site occupancy in order to generate the whole organic template.In 2 a total replacement of Fe(4) by Co(4) and a partial substitution of Fe(3) 60 by Co(3) occurs in the M(3) and M( 65

Fig. 2
Fig. 2 a) Polyhedral representation of the 3D crystal structure of 1 viewed along the [100] direction.The shaded area shows a zoomed cut of the connectivity inside the inorganic layers.b) Layers stacking by P(3) and P(4) bridging units.
(1)O 4 F 2 70 and Fe(2)O 4 F 2 share oxygen vertices of the phosphorous pseudotetrahedra giving rise to [001] chains.These chains are joined along the [010] direction, through the iron(II) octahedra dimers M(3)O 4 F 2 and M(4)O 2 (H 2 O) 2 F 2 , which share the O(18)-O(13) edges, leading to the layers, which have eight-membered 75 structural windows (Fig. the [Fe(1)O 4 F 2 ] and [Fe(2)O 4 F 2 ] polyhedra the Fe-O bond lengths (1.938(8)-2.011(7)Å for 1 and 1.935(6)-2.000(6)Å for 2) as well as the Fe-F distances (1.929(7)-1.983(7)Å for 1 and 1.929(6)-1.993(4)Å for 2) are in good agreement with 3+ oxidation state for the Fe(1) and Fe(2) atoms, based on the bond valence sums (BVS) calculations. 46Both polyhedra display eight O-Fe-F and four O-Fe-O cis angles, and two O-Fe-O and one F-Fe-F trans angles.However, in the [M(3)O 4 F 2 ] and [M(4)O 2 (H 2 O) 2 F 2 ] octahedra, M-O bond lengths span from 2.060(6) to 2.274(8) Å for 1 and from 2.056(4) to 2.216(6) Å for 2, and the M-F distances are in the ranges 1.996(6)-2.148(7)Å for 1 and 1.994(5)-2.125(5)Å for 2. In this case, the M-O/F distances indicate at first glance that, from the structural point of view, M(3) and M(4) sites are in +2 state.In these polyhedra the fluorine atoms are located in the edges, giving rise to six O-M-F, five O-M-O and one F-M-F cis angles and four O-M-O cis angles, and two O-M-F and one O-M-O trans angles.The S(O h ) values for phases 1 and 2, calculated by Continuous Symmetry Measure, 47 show more distorted octahedra for bivalent metals (M(3) and M(4)), however, they represent very slight distortions with regard to the ideal octahedron.The P-O distances of the (HPO 3 ) 2-units are in the range 1.482(9)-1.544(8)Å for 1 and 1.493(7)-1.540(6)Å for 2, and the P-H distances are 1.31(5) Å and 1.32(5) Å for 1 and 2 respectively.The average O-P-O and H-P-O angles are around 112º and 107º respectively for both compounds, which are in the usual range for phosphite based compounds. 48The terminal P(1)-O(1´) bond distances are significantly higher than the rest of the P-O bonds (1.55(4) for 1 and 1.59(3) for 2) which could be in good agreement with the existence of (HPO 4 ) 2-groups inferred from the electroneutrality of the formula.The phosphorus atoms make ten P-O-M (M = Fe, Co) linkages with minimum and maximum values for the P(2)-O(13)-Fe(3) and P(4)-O(8)-Fe(2) angles [123.3(4)º and 160.0(7)º for 1 and 123.5(3)º and 159.1(5)º for 2].The octahedra sharing vertex display four M-F-M linkages with average angles of 127.5(3)º for 1 and 128.3(2)º for 2. The M(3) and M(4) metals share an edge possessing two M-O-M linkages of around 98º and 100º for both phases.Hydrogen bonding plays an important role linking the tetramine cations to the framework (Fig 3b).The terminal hydrogen-bond donor group, N(1), interact with the oxygens O(6) and O(17W) and the fluorine atoms F(2), and F(4) while the 'internal' donor group, N(2), interact with O(1), O(7), F(1) and F(3).Three C-H•••O and one C-H•••F contacts are also present in the structures.The complete list of hydrogen bond interactions is shown in Table 70more systematic Klyne-Prelog system51 is required to describe them obtaining a ttgaagtt conformation (a = anticlinal) for the conformers A and a ttsaastt conformation (s = synperiplanar) for the conformers B in both compounds.The amount of potential-free volume was estimated by the 75 program PLATON52 assuming that the templates could be removed from the channels, being around 20% of the total volume for phases 1 and 2. The analysis of the volume occupied by these guest molecules was performed using the TOPOS 4.0 program 43 by means of Voronoi-Dirichlet polyhedra (VDP) 53 80 (Fig.3a) obtaining a 23% of the crystal volumes.The matching percentage indicates the non-existence of accessible volume.

60 clear
task because of the overlapping of different d 5 , d 6 and d 7 high spin cations signals.However, these spectra show a significant change in the relative intensity of the splitted 5 T 2g ( 5 D) → 5 E 2g ( 5 D) transition bands as well as a reduction in the intensity of the t 2g (Fe 2+ ) → t 2g (Fe 3+ ) intervalence transition, suggesting an 65 effective substitution of Fe(II) by Co(II).

Fig. 7
Fig. 7 Temperature dependence of χm for compounds 1 and 3 measured under 2 kOe.The lower inset shows the 1/χm curve fitted to the Curie−Weiss law and the upper inset an enlargement of the low-temperature region of ZFC−FC data measured under 0.5, 2 and 10 kOe.In Fig.8ait is depicted the field dependence of the magnetization of 1 at different temperatures.At 2 K, the magnetization increases almost linearly with the magnetic field until a critical field of 45 kOe, where a metamagnetic transition occurs.The magnetization value (0.45µ B /Fe ion) obtained at the highest applied field, 90 kOe, is far from the theoretical saturation for the Fe(II) (4µ B ) and Fe(III) (5µ B ) ions, indicating the

Fig. 8 a
Fig. 8 a) Magnetization vs applied magnetic field at different temperatures for 1.The upper inset shows the detail of the metamagnetic transition and the lower inset an enlargement of the small coercitivity of the hysteresis loops.b) Thermal evolution of the coercitive field of the hysteresis loops and critical field of the metamagnetic transition.
2 and 3.7 A° and consequently direct 50 interactions are not negligible.A view of the most important magnetic exchange interactions M-O-M and M-F-M in 1 and 2 is given in Fig. S15.J 1 pathway represents direct intradimeric magnetic interactions via oxygen atoms between the M(3)O 4 F 2 and M(4)O 2 (H 2 O) 2 F 2 polyhedra.The values of the bond angles 55 for J 1 are around 98º and 100º, being able to lead to a ferromagnetic interaction inside the dimmers.However, J 2 to J 5 pathways imply superexchange interactions through fluorine atoms involving bond angles between approximately 124 and 133º and clearly indicating the existence of an antiferromagnetic 60 coupling.The superexchange inter-and intralayer M-O-P-O-M interactions allows one to propagate the magnetic interactions giving rise to a three-dimensional magnetic system.The bond distances and angle values for the exchange pathways are very close for 1 and the cobalt substituted compounds 2 and 3. So, the 65 differences in the magnetic properties from 1 to 3 cannot be explained from a structural point of view.Taking into account that the Co 2+ ions prefer the M(3)O 4 F 2 and M(4)O 2 (H 2 O) 2 F 2 octahedra, the substitution of Fe(II) (d 6 ) (S = 2) by Co(d 7 ) (S = 3/2) in the framework of 1 modifies the nature of 70 some magnetic interactions involving J 1 pathway.

Fig. 9
Fig. 9 Specific heat of compounds 1 and 3 between 2 and 300 K.The insets show an enlargement around the Neel temperature as a function of temperature in the presence of external magnetic fields, H, in the 0-90 kOe range.

Table 1
Crystallographic data and structure refinement parameters for phases 1 and 2 obtained by single crystal X-ray diffraction and 3 were obtained using a constant-acceleration Mössbauer spectrometer with a 57 Co/Rh source.Velocity calibration was done using a metallic Fe foil, and the Mössbauer spectral parameters are given relative to this standard at room 40 temperature.The Mössbauer spectra were fitted with the NORMOS program.
5/2 line at 368.28 eV.The binding energy of the adventitious carbon (C1s) was set at 284.6 eV to correct sample charging.The spectra were fitted with the CasaXPS 2.3.16software, which models the Gauss-Lorentzian contributions, after 35 background subtraction (Shirley).Mössbauer spectra of 1, 2 582 spectra for compound 1 suggests the coexistence of Fe 2+ close to 711 eV and Fe 3+ at around 714 eV according to the values found for FeF 2 •and FeF 3 fluorides, respectively.58Deconvolution of this region in 85 compounds 2 and 3 is hindered by the interference of the Auger line Co LMM, however, the values of the Fe 2p 1/2 component close to 724 and 727 eV, next to those found in 1, suggests also the mixed valency of the iron cations.Moreover, it is important to note the absence of the satellite peak of the Fe 2p as it has been 90