3D Metal–Organic Frameworks Based on Co(II) and Bithiophendicarboxylate: Synthesis, Crystal Structures, Gas Adsorption, and Magnetic Properties

Three new 3D metal-organic porous frameworks based on Co(II) and 2,2′-bithiophen-5,5′-dicarboxylate (btdc2−) [Co3(btdc)3(bpy)2]·4DMF, 1; [Co3(btdc)3(pz)(dmf)2]·4DMF·1.5H2O, 2; [Co3(btdc)3(dmf)4]∙2DMF∙2H2O, 3 (bpy = 2,2′-bipyridyl, pz = pyrazine, dmf = N,N-dimethylformamide) were synthesized and structurally characterized. All compounds share the same trinuclear carboxylate building units {Co3(RCOO)6}, connected either by btdc2– ligands (1, 3) or by both btdc2– and pz bridging ligands (2). The permanent porosity of 1 was confirmed by N2, O2, CO, CO2, CH4 adsorption measurements at various temperatures (77 K, 273 K, 298 K), resulted in BET surface area 667 m2⋅g−1 and promising gas separation performance with selectivity factors up to 35.7 for CO2/N2, 45.4 for CO2/O2, 20.8 for CO2/CO, and 4.8 for CO2/CH4. The molar magnetic susceptibilities χp(T) were measured for 1 and 2 in the temperature range 1.77–330 K at magnetic fields up to 10 kOe. The room-temperature values of the effective magnetic moments for compounds 1 and 2 are μeff (300 K) ≈ 4.93 μB. The obtained results confirm the mainly paramagnetic nature of both compounds with some antiferromagnetic interactions at low-temperatures T < 20 K in 2 between the Co(II) cations separated by short pz linkers. Similar conclusions were also derived from the field-depending magnetization data of 1 and 2.


Introduction
Metal-organic frameworks (MOFs) are a class of materials consisting of metal ions or polynuclear complexes connected by polytopic organic ligands into regular periodic networks with internal pores. Such materials can be utilized in various important applications, such as gas storage, molecular separation, luminescence, and catalysis. Magnetic properties are also one of the most intriguing characteristics of MOFs assembled from the paramagnetic metal cations or, very rarely, using paramagnetic organic linker [1][2][3][4][5]. The magnetic interactions between the paramagnetic centers of the coordination framework often take place along organic ligands with conjugate double or triple bonds which provide substantial electron delocalization. Carboxylate linkers, such as terephthalate or trimesate, dominate in the current MOF chemistry because of their intrinsic structural rigidity and strong coordination abilities towards common transition metal cations. Such type of ligands can also support the magnetic interaction between the paramagnetic centers; however, the corresponding constants of the magnetic exchange in e.g., terephthalate-based MOFs are, typically, rather low [6][7][8]. Recently several research groups synthesized a number of MOFs using rigid thiophene-and selenophene-based heterocyclic carboxylate ligands [9][10][11][12][13]. The  (18) Å, and the Co(1)-Co(2) interatomic distances are 3.5282(7) Å. Each {Co3(RCOO)6} unit in 1 is connected by bridging btdc 2-ligands to six others thus serving as a 6-connected node of the regular coordination framework with underlying primitive-cubic topology pcu ( Figure 2). The average distances between the {Co3(RCOO)6} units in the structure are, ca. 14 Å, according to the size of the rather long btdc 2-anions. Despite rather strong distortions, the resulting framework 1 possesses rectangular channels of 6 × 5 Å along the c crystallographic axis, filled with the solvent DMF molecules (Figure 2b). The guest accessible volume of 1, calculated using PLATON software [32], is 35%. The guest composition was established from multiple methods, including SQUEEZE analysis of the unassigned electron density, IR spectroscopy, thermogravimetric and chemical analyses.  Purple stick crystals of [Co 3 (btdc) 3 (bpy) 2 ]·4DMF (1) are obtained by heating of mixture of Co(NO 3 ) 2 ·6H 2 O, H 2 btdc and 2,2 -bpy in 2:2:1 molar ratio in DMF at 110 • C during 2 days. According to the single-crystal X-ray diffraction data, compound 1 crystallizes in the monoclinic space group P2 1 /c. Two additional positions on each Co(1) cations of the {Co 3 (RCOO) 6 } building units are occupied by one chelating 2,2 -bpy ligands forming [Co 3 (RCOO) 6 (bpy) 2 ] complexes ( Figure S1a). Within such complexes the Co(1)-O bond lengths are in the range 2.0110(2)-2.209(2) Å, the Co(1)-N distances are 2.079(2) Å and 2.131(3) Å, the Co(2)-O distances lay in the range 2.042(2)-2.1209(18) Å, and the Co(1)-Co(2) interatomic distances are 3.5282(7) Å. Each {Co 3 (RCOO) 6 } unit in 1 is connected by bridging btdc 2ligands to six others thus serving as a 6-connected node of the regular coordination framework with underlying primitive-cubic topology pcu ( Figure 2). The average distances between the {Co 3 (RCOO) 6 } units in the structure are, ca. 14 Å, according to the size of the rather long btdc 2anions. Despite rather strong distortions, the resulting framework 1 possesses rectangular channels of 6 × 5 Å along the c crystallographic axis, filled with the solvent DMF molecules (Figure 2b). The guest accessible volume of 1, calculated using PLATON software [32], is 35%. The guest composition was established from multiple methods, including SQUEEZE analysis of the unassigned electron density, IR spectroscopy, thermogravimetric and chemical analyses.
Red stick crystals [Co 3 (btdc) 3 (pz)(dmf) 2 ]·4DMF·1.5 H 2 O (2) are formed by heating the Co(ClO 4 ) 2 ·6H 2 O, H 2 btdc and pyrazine mixture in DMF at 105 • C for two days. Some amount (10 µL) of concentrated hydrochloric acid was added to the reaction solution in order to dissolve the pyrazine. According to the single-crystal X-ray diffraction data, compound 2 crystallizes in the orthorhombic space group Fddd. The two additional positions on Co(1) cations in 2 are occupied by one O atom of the coordinated dmf solvent molecule and one N atom of the bridging pyrazine molecule, which connects two Co(1) cations of the neighboring trinuclear complexes [Co 3 (btdc) 3 (pz)(dmf) 2 ] ( Figure S2). The interatomic Co(1)-O bond distances are in the range 2.0129(14)-2.1732(15) Å, the Co(1)-N distance is 2.1497(17) Å, the Co(2)-O bond lengths lay in the range 2.0737(14)-2.1520(13) Å, and the interatomic Co(1)-Co(2) distances in the trinuclear complexes are 3.4829(4) Å. Contrary to 1, the carboxylate {Co 3 (RCOO) 6 } units in framework 2 are interconnected by both btdc 2and pz bridging ligands in three directions (Figure 3). Despite a higher number of potential linkers, the connectivity of the {Co 3 (btdc) 3 (pz)} complexes in 2 is 6 because four btdc 2ligands form two double bridges, which results in a complex uninodal 6-connected topology of the framework with point symbol 4 10 .6 5 . To the best of our knowledge, such topology has never been mentioned or characterized in the literature. There are three types of distances between the thinuclear carboxylate units {Co 3 (RCOO) 6 } in the crystal structure 2: 13.7 Å for the single btdc 2bridge, 13.3 Å for the double btdc 2bridge, and 10.9 Å for the pz bridge (measured between the central Co(2) cations) with minimal Co-Co spacing being as short as 7.1 Å (measured between Co(1) cations). The framework structure features rectangular channels along the a-axis of 7 × 4 Å, filled by guest DMF and H 2 O molecules ( Figure 3b). The free accessible volume is ca. 39% [32]. The guest composition was established from the multiple instrumental methods, such as SQUEEZE analysis, IR spectroscopy, TGA, chemical analyses.
Purple stick crystals of [Co3(btdc)3(bpy)2]·4DMF (1) are obtained by heating of mixture of Co(NO3)2·6H2O, H2btdc and 2,2′-bpy in 2:2:1 molar ratio in DMF at 110 °C during 2 days. According to the single-crystal X-ray diffraction data, compound 1 crystallizes in the monoclinic space group P21/c. Two additional positions on each Co(1) cations of the {Co3(RCOO)6} building units are occupied by one chelating 2,2′-bpy ligands forming [Co3(RCOO)6(bpy)2] complexes ( Figure S1a). Within such complexes the Co(1)-O bond lengths are in the range 2.0110(2)-2.209(2) Å, the Co(1)-N distances are 2.079(2) Å and 2.131(3) Å, the Co(2)-O distances lay in the range 2.042(2)-2.1209(18) Å, and the Co(1)-Co(2) interatomic distances are 3.5282(7) Å. Each {Co3(RCOO)6} unit in 1 is connected by bridging btdc 2-ligands to six others thus serving as a 6-connected node of the regular coordination framework with underlying primitive-cubic topology pcu ( Figure 2). The average distances between the {Co3(RCOO)6} units in the structure are, ca. 14 Å, according to the size of the rather long btdc 2-anions. Despite rather strong distortions, the resulting framework 1 possesses rectangular channels of 6 × 5 Å along the c crystallographic axis, filled with the solvent DMF molecules (Figure 2b). The guest accessible volume of 1, calculated using PLATON software [32], is 35%. The guest composition was established from multiple methods, including SQUEEZE analysis of the unassigned electron density, IR spectroscopy, thermogravimetric and chemical analyses.   Red stick crystals [Co3(btdc)3(pz)(dmf)2]·4DMF·1.5 H2O (2) are formed by heating the Co(ClO4)2·6H2O, H2btdc and pyrazine mixture in DMF at 105 °C for two days. Some amount (10 μL) of concentrated hydrochloric acid was added to the reaction solution in order to dissolve the pyrazine. According to the single-crystal X-ray diffraction data, compound 2 crystallizes in the orthorhombic space group  (Figure 3). Despite a higher number of potential linkers, the connectivity of the {Co3(btdc)3(pz)} complexes in 2 is 6 because four btdc 2-ligands form two double bridges, which results in a complex uninodal 6connected topology of the framework with point symbol 4 10 .6 5 . To the best of our knowledge, such topology has never been mentioned or characterized in the literature. There are three types of distances between the thinuclear carboxylate units {Co3(RCOO)6} in the crystal structure 2: 13.7 Å for the single btdc 2-bridge, 13.3 Å for the double btdc 2bridge, and 10.9 Å for the pz bridge (measured between the central Co(2) cations) with minimal Co-Co spacing being as short as 7.1 Å (measured between Co(1) cations). The framework structure features rectangular channels along the a-axis of 7 × 4 Å, filled by guest DMF and H2O molecules (Figure 3b). The free accessible volume is ca. 39% [32]. The guest composition was established from the multiple instrumental methods, such as SQUEEZE analysis, IR spectroscopy, TGA, chemical analyses.  Figure 4). The free accessible volume in 3 was estimated to be 33%, assuming the removal of only guest dmf molecules [32].  Figure 4). The free accessible volume in 3 was estimated to be 33%, assuming the removal of only guest dmf molecules [32].

IR-Spectroscopy, Thermal Activation, and Porosity
The compounds 1 and 2 were obtained as pure phases, as confirmed by powder Xray diffraction method ( Figure S3) and chemical analyses; hence, their functional properties were further characterized.
The IR spectrum of the compound 1 ( Figure S4) contains a band at 768 cm −1 related to non-planar deformation vibrations of C-H bonds in the thiophene fragment, characteristic bands of carboxylate groups at 1376 cm −1 and 1440 cm −1 associated with symmetric stretching vibrations, a group of bands in the range from 1477 cm −1 to 1597 cm −1 originating from skeletal vibrations of aromatic fragments of the ligands. The intense band at 1665 cm −1 is related to valence vibrations of the C=O bond in DMF molecules. The band at 3090 cm −1 corresponds to stretching vibrations of the C-H bond in the thiophene fragment. The broad band at 3428 cm −1 can be attributed to water on the external surfaces of the crystals. The IR spectrum of the compound 2 ( Figure S5) contains a band at 771 cm −1 correspondings to non-planar deformation vibrations of C-H bonds in the thiophene fragment, strong intensity bands at 1376 cm −1 and 1440 cm −1 related to the symmetric stretching vibration of carboxylate groups, a group of medium intensity bands in the range from 1517 cm −1 to 1589 cm −1 associated with skeletal vibrations of the pyrazine molecule and thiophene fragments. The band at 1652 cm −1 can be referred to C=O bonds in DMF molecules, whereas the low-intensity bands at 2854 cm −1 and 2926 cm −1 correspond to the valence vibrations of their C-H bonds. The broad band at 3449 cm −1 is related to the valence vibrations of the O-H bond of guest water molecules.
The thermogravimetric analysis ( Figure 5) of compound 1 reveals a continuous mass loss in a wide temperature range (up to 200 °C) of ca. 17%, which can be ascribed to the removal of the guest solvent molecules (calculated: 19% for 4 DMF), followed by a broad flat region up to ca. 330 °C, where rapid degradation of the metal-organic framework takes place. The heating of compound 2 results in a step-like decrease of the sample weight of ca. 22% up to 230 °C, corresponding the evaporation of the solvent molecules (calculated: 23% for 4DMF + 1.5H2O). The second stage of the weight loss of ca. 11% occurs near 280 °C and corresponds to the release of the coordinated DMF molecules (calculated: 10% for 2DMF). The irreversible thermolysis of the metal-organic framework 2 occurs above 340 °C, similarly to that for compound 1.

IR-Spectroscopy, Thermal Activation, and Porosity
The compounds 1 and 2 were obtained as pure phases, as confirmed by powder X-ray diffraction method ( Figure S3) and chemical analyses; hence, their functional properties were further characterized.
The IR spectrum of the compound 1 ( Figure S4) contains a band at 768 cm −1 related to non-planar deformation vibrations of C-H bonds in the thiophene fragment, characteristic bands of carboxylate groups at 1376 cm −1 and 1440 cm −1 associated with symmetric stretching vibrations, a group of bands in the range from 1477 cm −1 to 1597 cm −1 originating from skeletal vibrations of aromatic fragments of the ligands. The intense band at 1665 cm −1 is related to valence vibrations of the C=O bond in DMF molecules. The band at 3090 cm −1 corresponds to stretching vibrations of the C-H bond in the thiophene fragment. The broad band at 3428 cm −1 can be attributed to water on the external surfaces of the crystals. The IR spectrum of the compound 2 ( Figure S5) contains a band at 771 cm −1 correspondings to non-planar deformation vibrations of C-H bonds in the thiophene fragment, strong intensity bands at 1376 cm −1 and 1440 cm −1 related to the symmetric stretching vibration of carboxylate groups, a group of medium intensity bands in the range from 1517 cm −1 to 1589 cm −1 associated with skeletal vibrations of the pyrazine molecule and thiophene fragments. The band at 1652 cm −1 can be referred to C=O bonds in DMF molecules, whereas the low-intensity bands at 2854 cm −1 and 2926 cm −1 correspond to the valence vibrations of their C-H bonds. The broad band at 3449 cm −1 is related to the valence vibrations of the O-H bond of guest water molecules.
The thermogravimetric analysis ( Figure 5) of compound 1 reveals a continuous mass loss in a wide temperature range (up to 200 • C) of ca. 17%, which can be ascribed to the removal of the guest solvent molecules (calculated: 19% for 4 DMF), followed by a broad flat region up to ca. 330 • C, where rapid degradation of the metal-organic framework takes place. The heating of compound 2 results in a step-like decrease of the sample weight of ca. 22% up to 230 • C, corresponding the evaporation of the solvent molecules (calculated: 23% for 4DMF + 1.5H 2 O). The second stage of the weight loss of ca. 11% occurs near 280 • C and corresponds to the release of the coordinated DMF molecules (calculated: 10% for 2DMF). The irreversible thermolysis of the metal-organic framework 2 occurs above 340 • C, similarly to that for compound 1.  The permanent porosity was confirmed by the measurements of the gas adsorption for compound 1, which has a broad temperature stability range and plain activation strategy. The as-synthesized crystals of the porous compound 1 were activated by solvent exchange (CH2Cl2), followed by a dynamic vacuum treatment at 30 °C for 1 h directly in a gas adsorption analyzer. The nitrogen adsorption-desorption isotherm plots at 77 K are represented in Figure 6, Figures S6 and S7 and belong to the Ia isotherm type according to the official IUPAC classification [33], which is typical for microporous compounds with narrow slit pores. The measured pore volume 0.289 mL g -1 (at p/p0 = 0.95) matches the expected value (0.280 mL g -1 ) from the PLATON pore volume calculations, which confirms the structural integrity of 1 as well as completeness of the framework activation. The calculated surface areas are 758 m 2 g -1 (Langmuir model), 667 m 2 g -1 (BET model) and 671 m 2 g -1 (DFT model). The distribution of the pore size was calculated using the nonlinear DFT equilibrium model with slit pores which gives good agreement between measured and calculated isotherms. Pore size distribution plot of 1, calculated by DFT model, (Figure 6, inset) shows the presence of narrow pores of ca. 7 Å. Similarly, the corresponding pore-size distribution, calculated by Zeo++ program package [34,35] on the basis of the structural data of 1 results in the same value of 7 Å ( Figure S8), validating both methods and supporting the structural stability of the guest-free framework. The permanent porosity was confirmed by the measurements of the gas adsorption for compound 1, which has a broad temperature stability range and plain activation strategy. The as-synthesized crystals of the porous compound 1 were activated by solvent exchange (CH 2 Cl 2 ), followed by a dynamic vacuum treatment at 30 • C for 1 h directly in a gas adsorption analyzer. The nitrogen adsorption-desorption isotherm plots at 77 K are represented in Figure 6, Figures S6 and S7 and belong to the Ia isotherm type according to the official IUPAC classification [33], which is typical for microporous compounds with narrow slit pores. The measured pore volume 0.289 mL g −1 (at p/p 0 = 0.95) matches the expected value (0.280 mL g −1 ) from the PLATON pore volume calculations, which confirms the structural integrity of 1 as well as completeness of the framework activation. The calculated surface areas are 758 m 2 g −1 (Langmuir model), 667 m 2 g −1 (BET model) and 671 m 2 g −1 (DFT model). The distribution of the pore size was calculated using the nonlinear DFT equilibrium model with slit pores which gives good agreement between measured and calculated isotherms. Pore size distribution plot of 1, calculated by DFT model, (Figure 6, inset) shows the presence of narrow pores of ca. 7 Å. Similarly, the corresponding pore-size distribution, calculated by Zeo++ program package [34,35] on the basis of the structural data of 1 results in the same value of 7 Å ( Figure S8), validating both methods and supporting the structural stability of the guest-free framework.

Gas Adsorption Selectivity
More detailed gas adsorption properties of 1 were further investigated by measuring the adsorption-desorption isotherms for CO2, N2, CO, O2 and CH4 at T = 273 and 298 K. The corresponding isotherms obtained are shown in Figures 7, S9 and S11, gas uptakes in different units are summarized in Table 1.

Gas Adsorption Selectivity
More detailed gas adsorption properties of 1 were further investigated by measuring the adsorption-desorption isotherms for CO 2 , N 2 , CO, O 2 and CH 4 at T = 273 and 298 K. The corresponding isotherms obtained are shown in Figure 7, Figures S9 and S11, gas uptakes in different units are summarized in Table 1.

Gas Adsorption Selectivity
More detailed gas adsorption properties of 1 were further investigated by measuring the adsorption-desorption isotherms for CO2, N2, CO, O2 and CH4 at T = 273 and 298 K. The corresponding isotherms obtained are shown in Figures 7, S9 and S11, gas uptakes in different units are summarized in Table 1.   The absolute amounts of the gas uptakes of the porous compound 1 at ambient conditions are moderate, compared with the best results for other MOFs, which is expected of the frameworks with relatively narrow pores and moderate surface area. The calculated isosteric heats of adsorption at zero coverage Q st (0) are 26.1 kJ·mol −1 for CO 2 , 19.7 kJ·mol −1 for CH 4 , 17.3 kJ·mol −1 for CO, 10.3 kJ·mol −1 for O 2 , and 8.8 kJ·mol −1 for N 2 ( Figure S10). Such low values are typical for the physical adsorption through weak van-der-Waals interactions and consistent with the structural data of 1, with all coordination positions near the Co(II) cations being saturated by either carboxylate groups or chelated bpy ligand. We note here that, from the practical application point of view, a low value of the adsorption heat is highly beneficial since it reduces the energy penalties in the cyclic temperature-swing adsorption separation processes.
The porous materials with a narrow diameter of pores usually exhibit great gas adsorption selectivity. The selectivity factors for separation of CO 2 /N 2 and CO 2 /CH 4 binary gas mixtures on 1 were calculated by three different methods: (i) as the ratio of the amount adsorbed; (ii) as the ratio of corresponding Henry constants; and (iii) by Ideal Adsorbed Solution Theory (IAST) [36] calculations, which allows estimation of the selectivity factors at different compositions of the gas mixture and various total pressures. The results are summarized in Table 2 and in Figure S12. Depending on the method used, the adsorption selectivity factors of CO 2 /N 2 gas mixture vary from 10 to 35.7 at 273 K and from 7.9 to 18.9 at 298 K, which is well above average, considering other porous MOFs with no coordinatively unsaturated metal sites (CUS). The IAST CO 2 /N 2 selectivity values become even greater for the gas mixture composition similar to the real industrial flue gases (CO 2 :N 2 = 15:85 (v/v), Table 2). The CO 2 /O 2 adsorption selectivity factors range from 9.4 to 45.4, depending on the temperature and calculation approach, similarly to results reported for other porous MOFs [37][38][39]. The obtained adsorption selectivity factors for the CO 2 /CO mixture (7.9 ÷ 20.8 at 273 K, 7.2 ÷ 15.1 at 298 K) rival the most prominent results reported for porous MOFs with only a few superior exceptions [40][41][42]. More importantly, all the obtained selectivity factors are in par or exceed 8, which is considered to be sufficient for a practical separation of gases [43]. The adsorption selectivity factors for CO 2 /CH 4 mixture range from 2.5 to 4.8, depending on the temperature and calculation method used. Such values are comparable to many reported porous MOF compounds [38,44,45]. Appreciable CO 2 /N 2 , CO 2 /CO and CO 2 /O 2 separation factors, as well as low CO 2 adsorption heat, put the porous material 1 among other promising platforms for the development of efficient processes for the sequestration of CO 2 from N 2 , CO or O 2 , which addresses important environmental, industrial and safety challenges, respectively.

Magnetic Properties of Compounds 1 and 2
Temperature dependences of the molar magnetic susceptibility were measured for 1 and 2 in the range 1.77-330 K at magnetic fields up to 10 kOe under zero-field cooled and field-cooled conditions ( Figure S13). The data obtained revealed no dependence on the magneto-thermal history implying the absence of any ferromagnetic or spin-freezing phenomena. For both compounds, 1 and 2, the magnetic susceptibility increased gradually with the decreasing temperature down to the lowest accessed temperature of 1.77 K, as expected for paramagnetic substances, such as MOF containing magnetic Co(II) ions. After subtracting the diamagnetic contribution, χ d , calculated using Pascal's additive scheme, from the total magnetic susceptibility, the remaining paramagnetic component of the molar magnetic susceptibility, χ p (T), was plotted as 1/χ p vs. T ( Figure 8). As can be seen, the 1/χ p (T) curves are quite close to linear at temperatures above~50 K, nominally following a paramagnetic Curie-Weiss dependence χ p (T) = C/(T−θ) with θ ≈ −16 K and −22 K for 1 and 2, respectively, in close similarity to the behavior observed for other MOFs containing Co(II) [46][47][48]. Given the absence of long-range antiferromagnetic (AFM) ordering and the deviation of the 1/χ p data from the fits towards smaller values (larger values of χ p ) observed at low temperatures, the linear 1/χ p (T) dependences at T > 50 K are rather accidental; hence, the apparent θ values should not be taken as an indicator of magnetic interactions between cobalt ions belonging to neighboring molecules (J ). In fact, much smaller θ values would be obtained if fitting were performed in the lowest temperature region (Figure 8), pointing therefore to a small actual J .

Magnetic Properties of Compounds 1 and 2
Temperature dependences of the molar magnetic susceptibility were measured for 1 and 2 in the range 1.77-330 K at magnetic fields up to 10 kOe under zero-field cooled and field-cooled conditions ( Figure S13). The data obtained revealed no dependence on the magneto-thermal history implying the absence of any ferromagnetic or spin-freezing phenomena. For both compounds, 1 and 2, the magnetic susceptibility increased gradually with the decreasing temperature down to the lowest accessed temperature of 1.77 K, as expected for paramagnetic substances, such as MOF containing magnetic Co(II) ions. After subtracting the diamagnetic contribution, χd, calculated using Pascal's additive scheme, from the total magnetic susceptibility, the remaining paramagnetic component of the molar magnetic susceptibility, χp(T), was plotted as 1/χp vs. T ( Figure 8). As can be seen, the 1/χp(T) curves are quite close to linear at temperatures above ~50 К, nominally following a paramagnetic Curie-Weiss dependence χp(T) = C/(T−θ) with θ ≈ −16 K and −22 K for 1 and 2, respectively, in close similarity to the behavior observed for other MOFs containing Co(II) [46][47][48]. Given the absence of long-range antiferromagnetic (AFM) ordering and the deviation of the 1/χp data from the fits towards smaller values (larger values of χp) observed at low temperatures, the linear 1/χp(T) dependences at T > 50 K are rather accidental; hence, the apparent θ values should not be taken as an indicator of magnetic interactions between cobalt ions belonging to neighboring molecules (J′). In fact, much smaller θ values would be obtained if fitting were performed in the lowest temperature region (Figure 8), pointing therefore to a small actual J′.  Dashed blue lines indicate formal Curie-Weiss fits for the high-temperature region. The orange curve in plot (b) shows the µ eff data obtained by taking into account the antiferromagnetic intermolecular interactions. Magnetic field 10 3 Oe.
In order to clarify the observed magnetic behavior, we first focused at the intramolecular magnetic state of Co(II) ions in trinuclear carboxylate {Co 3 (RCOO) 6 } units and calculated the effective magnetic moment µ eff using the Curie dependence ( Figure 8). The corresponding equation for one independent cobalt ion is χ p (T) = N a µ eff 2 /3·k B ·T, while for the formula unit, containing a trinuclear complex, it is χ p (T) = 3·N a µ eff 2 /3·k B ·T. Both compounds 1 and 2 turn out to have the same µ eff (300 K) ≈ 4.93 µ B , which exceeds the spin-only value (3.87 µ B ) for the isolated high-spin Co(II) ions with S = 3/2, implying a significant orbital contribution. Except for the low-temperature range T < 20 K, the temperature dependence of µ eff for 1 and 2 is virtually identical: with decreasing temperature the µ eff is gradually reduced down to~4 µ B owing to a combination of single-ion zero-field splitting effects (partial quenching of orbital moments) and antiferromagnetic interactions of Co(II) ions within the trimers mediated by the carboxylate bridges. This behavior is rather close to those reported for other similar Co(II) MOFs [49][50][51][52]. Given the complicated magnetic ground state of the octahedrally coordinated Co(II) ions [53], a quantitative analysis of the χ p (T) data and exact determination of the AFM exchange interaction of cobalt ions within trimers (J) is hardly achievable. However, the data obtained for two compounds, 1 and 2, that have similar magnetic state of the trinuclear units and differ in the way they are arranged in the MOFs, give us a possibility to shed light on the weak intermolecular interaction. In case of 1 (Figure 8a), the zero-T extrapolation gives 1/χ p (T) = 0 (µ eff3 .7 µ B ) within the experimental accuracy, implying an almost ideal paramagnetic behavior of Co(II)-trimers non-interacting with each other. Indeed, in structure 1, all trinuclear complexes [Co 3 (RCOO) 6 (bpy) 2 ] are linked by btdc 2ligands only, providing a rather long separation between the {Co 3 (RCOO) 6 } units at~14 Å (vide supra).
Upon lowering the temperature below 20 K the effective magnetic moment of 2 µ eff drops increasingly fast, which markedly differs from the behavior of 1 (Figure 8). Such notable discrepancies in the magnetic properties likely suggest an additional AFM interaction that comes into play in 2 at low T. We should note here that an introduction of a certain Weiss constant θ = −1.5 K into the expression χ p (T) = 3·N a ·µ eff 2 /3·k B ·(T − θ) allows fitting the low-T behavior of 2 and bringing the µ eff (T) curve to exactly the same shape as in 1 (Figure 8b, orange curve). It supports the above assumption that the differences in magnetic properties of 1 and 2 originate from the presence of the AFM interactions between the {Co 3 (RCOO) 6 } units in 2. Even though such interactions are insufficient for a long-range AFM ordering in the observed temperature range, the corresponding trend is clearly manifested in the field-dependence of the magnetization M(H) data (Figure 9), which show noticeably lower values of the magnetization and slower approaching to the spin-polarized state for the compound 2 than that for 1. Apparently, the emergence of intermolecular AFM interactions in 2 results from a small pyrazine linker, which provides a relatively short connection between the {Co 3 (RCOO) 6 } units at 7.1 Å thus giving rise to some long-range magnetic ordering.

Instruments and Methods
Infrared spectra of solid samples as KBr pellets were recorded using an IR-Fourier spectrometer Scimitar FTS 2000 (4000-400 cm −1 ). The elemental analyses were obtained on an analyzer «Vario Micro-Cube». The thermogravimetric analyses were carried out in He atmosphere on NETZSCH TG 209 F1 thermoanalyzer with a heating rate of 10 deg/min. The powder X-ray diffraction data were obtained on a «Shimadzu XRD 7000S» powder diffractometer (Cu-Kα irradiation, λ = 1.54178 Å). Surface area and porous structure were analyzed using the nitrogen adsorption technique on a Quantochrome's Autosorb iQ gas sorption analyzer at 77 K. Prior to the isotherm measurements the as-synthesized crystals of compound 1 were placed in CH2Cl2 for five days. Each day the crystals were separated from the solvent by decantation, and a new portion of CH2Cl2 was added. After 5 days the crystals were separated from the solvent, placed into the gas-measurement cell and, finally, activated under a dynamic vacuum at 30 °C for 1 h. The nitrogen adsorption−desorption isotherms were measured within the range of relative pressures −6 Figure 9. Field-depending magnetization M(H) data for compounds 1 (blue) and 2 (red) taken at T = 1.77 K.

Instruments and Methods
Infrared spectra of solid samples as KBr pellets were recorded using an IR-Fourier spectrometer Scimitar FTS 2000 (4000-400 cm −1 ). The elemental analyses were obtained on an analyzer «Vario Micro-Cube». The thermogravimetric analyses were carried out in He atmosphere on NETZSCH TG 209 F1 thermoanalyzer with a heating rate of 10 deg/min. The powder X-ray diffraction data were obtained on a «Shimadzu XRD 7000S» powder diffractometer (Cu-Kα irradiation, λ = 1.54178 Å). Surface area and porous structure were analyzed using the nitrogen adsorption technique on a Quantochrome's Autosorb iQ gas sorption analyzer at 77 K. Prior to the isotherm measurements the as-synthesized crystals of compound 1 were placed in CH 2 Cl 2 for five days. Each day the crystals were separated from the solvent by decantation, and a new portion of CH 2 Cl 2 was added. After 5 days the crystals were separated from the solvent, placed into the gas-measurement cell and, finally, activated under a dynamic vacuum at 30 • C for 1 h. The nitrogen adsorption-desorption isotherms were measured within the range of relative pressures from 10 −6 to 0.995. The specific surface area was calculated from the data obtained using the conventional BET and DFT models. Gases adsorption experiments at 273 and 298 K were carried out volumetrically on Quantochrome's Autosorb iQ equipped with thermostat TERMEX Cryo-VT-12 to adjust the temperature with 0.1 K accuracy. Adsorption-desorption isotherms were measured within the range of pressures of 1 to 800 torr. The database of the National Institute of Standards and Technology available at http://webbook.nist.gov/chemistry/fluid/ was used as a source of p-V-T relations at experimental pressures and temperatures. Magnetization measurements were carried out using a Quantum Design MPMS-XL SQUID magnetometer in the temperature range 1.77-330 K at magnetic fields up to 10 kOe. Temperature dependences of the magnetization, M(T), were measured on heating the sample after it was cooled either in zero magnetic field or in a given magnetic field as well as upon cooling the sample. In order to determine the paramagnetic component of the molar magnetic susceptibility, χ p (T), the temperature-independent diamagnetic contribution, χ d , and a possible magnetization of ferromagnetic micro-impurities, χ FM (T), were evaluated and subtracted from the measured values of the total molar susceptibility χ = M/H. While χ d was calculated using Pascal's additive scheme, χ FM (T), if any, was determined from the measured isothermal M(H) dependencies and the M(T) data taken at different magnetic fields. To determine the effective magnetic moment of cobalt Co(II) ions, µ eff , the paramagnetic susceptibility χ p (T) was analyzed using the Curie-Weiss dependence χ p (T) = N a µ eff 2 /3·k B ·(T − θ), where N A and k B are the Avogadro and Boltzmann numbers, respectively.

Single-Crystal X-ray Diffraction
Diffraction data for single-crystals 1-3 were collected on the 'Belok' beamline (λ = 0.793127 Å, ϕ-scans with a step of 1.0) of the National Research Center 'Kurchatov Institute' (Moscow, Russian Federation) using a Rayonix SX165 CCD detector. The data were indexed, integrated and scaled, and an absorption correction was applied using the XDS program package [54]. The structures were solved by the dual space algorithm (SHELXT [55]) and refined by the full-matrix least-squares technique (SHELXL [56]) in the anisotropic approximation (except hydrogen atoms). Positions of hydrogen atoms of organic ligands were calculated geometrically and refined in the riding model. The final compositions of compounds 2 and 3 were defined according to the PLATON/SQUEEZE procedure [32] (2787 e − in 9986 Å 3 for 2, 195 e − in 967 Å 3 for 3) and the data of elemental (C, H, N, S) analyses. The crystallographic data and details of the structure refinements are summarized in Table S1. CCDC 2023586-2023588 contains the supplementary crystallographic data for this paper. These data can be obtained free of charge from The Cambridge Crystallographic Data Center at https://www.ccdc.cam.ac.uk/structures/.

Synthesis of Coordination Polymers
The ligand synthesis was carried out according to the known methodology [10]. The starting substances were used as commercially available reagents without further purification.