Metal-organic frameworks (MOFs) constructed from metal ions (or metal ion clusters) and bridging organic linkers as a rapidly growing class of new materials ^{[1, 2, 3, 4, 5]} have attracted considerable interest because of their high porosity,tunable pores,diverse structures,and potential applications. MOFs,in particular,have shown great promise as heterogeneous catalysts ^{[6, 7, 8]} because they offer several advantages over conventional immobilized catalyst systems such as well-defined structure with uniformly accessible catalytic centers. Among the MOFs reported to date,lanthanide MOFs (Ln-MOFs) have shown the greatest potential as heterogeneous catalysts because of the flexible coordination sphere of the lanthanide ion,which can create a coordinatively unsaturated metal center. Furthermore,Ln-MOFs generally exhibit high thermal stability. Based on their interesting properties,several research groups,including our own,have been working on the development of new strategies for the crystal engineering of porous Ln-MOFs and their application in heterogeneous catalysis ^{[9, 10, 11, 12, 13, 14, 15, 16]}.

Post-synthetic modification (PSM) is an efficient method for the production of novel materials that contain the intrinsic characteristics of MOFs,including open porosities,large surface areas,and regular structures ^{[17, 18, 19]}. PSM can be achieved by the formation of a covalent attachment to the organic linkers ^{[20, 21, 22, 23]} or by the grafting of organic molecules at coordinatively unsaturated metal sites ^{[24, 25, 26, 27]}. The latter of these two strategies involves the use of a reagent to modify the components of the MOF (e.g.,a secondary building unit or an organic linker) by forming a new dative bond. PSM is particularly attractive for the production of materials to be used as catalysts because it opens the possibility of employing a wide range of catalytically active organic groups in MOFs,as well as other attendant metal ions. There have been many reports in the literature pertaining to the catalytic properties of MOFs prepared using a PSM strategy ^{[23, 24, 28, 29, 30]}.

In an attempt to develop efficient new catalysts based on Ln-MOFs,we investigated the application of a PSM strategy to the Ln-MOF [Er(btc)(H₂O)]·DMF_0.7(1),which has high porosity and good stability (btc = 1,3,5-benzenetricarboxylates). Three different diamines,1,2-ethanediamine (ED),piperazine (PP),and 1,4-diazabicyclo[2.2.2]octane (DABCO),which represent a variety of molecular size and basicity properties,were grafted onto the coordinatively unsaturated Er(III) ions of desolvated 1. This grafting process resulted in the formation of several new solid base catalysts,including Er(btc)(ED)_0.75(H₂O)_0.25 (2),Er(btc)(PP)_0.55(H₂O)_0.45 (3),and Er(btc)(DABCO)_0.15(H₂O)_0.85 (4) (Scheme 1). In these catalysts,one of the nitrogen atoms of the diamine was coordinated to an Er(III) ion,whereas the other nitrogen atom acted as a catalytically active site in channels of the MOF. The catalytic properties of these three solid catalysts were tested against the Knoevenagel condensation reaction,and a plausible mechanism was proposed for catalyst 2 based on the results of a series of size-selective catalysis tests.

Scheme 1. The structure of catalysts 2,3,and 4.

All of the chemicals used in the current study were purchased from Alfa or TCI and used directly without purification. Single crystal X-ray diffraction analysis was performed on a Rigaku Mercury CCD diffractometer operated at 90 kV and 50 mA using Mo K_α radiation (λ = 0.071073 nm) at room temperature. The structure was solved by direct methods,and all of the non-hydrogen atoms were refined anisotropically by least- squares on F² using the SHELXTL program. Hydrogen atoms on the organic ligands were generated using the riding mode. Powder X-ray diffraction (PXRD) patterns were collected on a Bruker D8 powder diffractometer at 40 kV and 40 mA using Cu K_α radiation (λ = 0.15406 nm),with a scan speed of 17.7 s/step and a step size of 0.02° (2θ). Thermogravimetric analysis (TGA) was performed on a Q600 SDT instrument under a steady flow of N₂ at a heating rate of 10 °C/min from ambient temperature to 800 °C. Nitrogen adsorption measurements were carried out at -196 °C on a micromeritics ASAP 2020 M instrument. Elemental analyses for C,H,and N were conducted on a Vario EL III Elemental Analyzer. Infrared (IR) spectra were measured as KBr pellets on a Nicolet Model Nexus 470 FT-IR spectrometer at wavenumbers in the range of 4000-400 cm^-1. ¹H NMR spectra were recorded on a Bruker Model AM-400 (400 MHz) spectrometer. Gas chromatography-mass spectroscopy (GC-MS) was conducted on a Shimadzu GCMSQP5050A system equipped with a 0.25 mm × 30 m DB-WAX capillary column.

For preparation of [Er(btc)(H₂O)]·DMF_0.7 (1),pre- functionalized MOF 1 was synthesized according to a previously reported procedure ^[31]. Er(NO₃)₃·6H₂O (0.231 g,0.5 mmol) and H₃btc (0.053 g,0.25 mmol) were dispersed in a mixture of DMF (4 mL) and H₂O (4 mL),and the resulting mixture was heated at 105 °C for 24 h in a sealed Teflon reactor. The mixture was then cooled to ambient temperature and purified by centrifugation to give pure pink crystalline needles,which were dried at ambient temperature. [Er(btc)(H₂O)]·DMF_0.7 (1): calcd. C 30.06,H 2.25,N 2.21; found: C 30.38,H 2.11,N 2.34. IR spectrum (cm^-1): 3068 (w),1628 (m),1577 (w),1445 (m),1385 (s),1213 (m),1105 (w),944 (w),824 (w),765 (m),738 (m),708 (s),566 (m),457 (w).

For preparation of Er(btc)(ED)_0.75(H₂O)_0.25 (2),after being dehydrated under vacuum at 180 °C for 8 h,0.5 g of desolvated 1 was suspended in 30 mL of anhydrous toluene,followed by 0.1 g (1.6 mmol) ED,and the resulting suspension was stirred at room temperature over night. The mixture was then filtered,and the resulting pink powder was washed sequentially with anhydrous toluene and anhydrous diethyl ether (three times each),and then dried at room temperature for 12 h. Er(btc)(ED)_0.75(H₂O)_0.25 (2): calcd. C 29.75,H 2.26,N 4.96; found: C 28.96,H 2.52,N 5.17. IR spectrum (cm^-1): 3363 (w),3066 (w),1613 (m),1578 (w),1451 (m),1364 (s),1210 (m),1102 (w),946 (w),823 (w),761 (m),737 (m),708 (s),566 (w),535 (w),440 (w).

Er(btc)(PP)_0.55(H₂O)_0.45 (3) was prepared in a similar manner to catalyst 2,except that ED was replaced with PP and the suspension was heated to 40 °C to allow for the dissolution of the solid amine. Er(btc)(PP)_0.55(H₂O)_0.45 (3): calcd. C 31.29,H 2.20,N 3.58; found: C 30.71,H 2.80,N 3.59. IR spectrum (cm^-1): 3063 (w),1623 (m),1562 (w),1363 (s),1209 (m),1103 (w),941 (w),822 (w),740 (m),711 (s),564 (w),439 (w).

Er(btc)(DABCO)_0.15(H₂O)_0.85 (4) was prepared using similar procedures to those described above for the preparation of 2,except that ED was replaced with DABCO and the amine was heated at 40 °C to allow for its dissolution. Er(btc)(DABCO)_0.15(H₂O)_0.85 (4): calcd. C 29.25,H 1.61,N 1.03; found: C 28.58,H 1.86,N 0.95. IR spectrum (cm^-1): 3060 (w),1618 (m),1561 (w),1448 (m),1363 (s),1208 (m),1102 (w),940 (w),824 (w),752 (m),710 (s),564 (w),534 (w),442 (w).

The amine-grafted MOF-catalyzed Knoevenagel condensation reactions of aldehydes and active methylene compounds were conducted in magnetically stirred round bottom flasks fitted with a reflux condenser. In a typical reaction,a mixture of cyclohexane (10 mL),benzaldehyde (1 mmol),and malononitrile (1.1 mmol) was added to a 50 mL flask containing catalyst 2 (6 mg,1 mol%). The concentration of the catalyst was calculated with respect to the molar ratio of the terminal amino group to the aldehyde. For comparison,the amounts of the other catalysts were also calculated based on the molar amounts of the amino groups available for the reaction (8 and 27 mg for 3 and 4,respectively). The resulting mixture was stirred at 40 °C for 2 h. The catalyst was then separated from the reaction mixture by centrifugation,and washed with ethyl acetate (5 × 2 mL) to remove any physically adsorbed reagents. The catalyst was then dried at room temperature for 6 h and reused in the next run. The filtrate was concentrated under reduced pressure to give the crude Knoevenagel product as a residue,which was purified by flash column chromatography over silica gel to give the pure product. The pure product was characterized by ¹H NMR using 1,3,5-trimethylbenzene as an internal standard.

The X-ray diffraction pattern,thermogravimetric analysis curve,and IR spectrum of the Ln-MOF (1) material prepared in the current study were in agreement with those reported in the literature ^[31],except for the number of DMF molecules in the channels. In 1,the Er(III) ion was coordinated to seven oxygen atoms,with six oxygen atoms from the six carboxylate groups of the six btc ligands and one oxygen atom from the terminal water molecule in a distorted pentagonal-bipyramidal coordination sphere. Each btc ligand was bound to six different Er(III) ions with each carboxylate oxygen connected to one Er(III) ion,leading to the formation of a 1D channel (0.7 nm × 0.7 nm) running along the c-axis. The coordinated water molecules pointing towards the channels and the guest DMF molecules in the channels could be readily removed to expose the empty coordination sites of the Er(III) ions,which could then be subjected to further PSM processes with the diamine ligand. More importantly,this MOF material showed a very high thermal stability of more than 500 °C (Fig. 1).

Fig. 1. TGA curves for Ln-MOF (1) and catalysts 2,3,and 4.

As shown in Fig. 1,the TGA curve for 2 revealed that this material underwent an 11% mass loss from 25 to 300 °C,which was attributed to the loss of the coordinated ED and water molecules in the channels. In contrast,the TGA curve for 3 showed that the coordinated PP and water molecules were lost before 250 °C,whilst almost all of the coordinated DABCO and water molecules were lost from 4 before 100 °C. Furthermore,the occurrence of PSM processes did not influence the stability of the structural framework of 1,and catalysts 2,3,and 4 were found to be stable up to 500 °C. The PXRD patterns of 2,3,and 4 were similar to that of the pristine sample of 1 (Fig. 2),which indicated that the structural framework of the pre- functionalized MOF had not changed significantly following the PSM process. Indeed,all three catalysts were found to be stable in the solid state under ambient conditions for 30 d.

Fig. 2. PXRD patterns of 1,2,3,and 4.

Although all of the coordinated water and DMF molecules in 1 could be removed under vacuum following at 180 °C for 8 h,the coordinatively unsaturated Er(III) ions in the channels of catalysts 2,3,and 4 could not all be coordinated by the nitrogen atoms of their respective diamines,because the simultaneous grafting of the amines would be restricted in the limited channels of 1. This process would therefore result in the partial coordination of water molecules to the non-grafted Er(III) ions if the material was exposed to air. The volumes of the three different diamines increased depending of the number of -CH₂CH₂- groups between their two N atoms. The primary amine ED had the smallest volume of the three amines tested in the current study and could therefore move into the channels much more readily to form a strong coordinate bond to the coordinatively unsaturated Er(III) ions. For this reason,the loading ratio of 2 reached 75% (i.e.,about 75% of the Er(III) ions were grafted to ED,with the other 25% being coordinated with water molecules). Given that the volume of the secondary amine PP is larger than that of ED,it would be more difficult for the PP molecules to move into the channels to reach the Er(III) ions. This difference in the volumes of the ED and PP amines resulted in a lower loading ratio of about 55% for 3. Furthermore,the volume of the tertiary amine DABCO is larger than those of PP and ED,which resulted in 4 having the lowest loading ratio among the three catalysts,with only 15% of the Er(III) ions being grafted onto DABCO.

The surface area and pore volume of 1 were determined to be 2000 m²/g and 0.75 cm³/g,respectively,based on its N₂ adsorption isotherm at -196 °C (Fig. 3). The mean pore size of 1 was determined to be 0.65 nm using the Horvath-Kawazoe model. This value was consistent with the results of the single crystal structure analysis. Compared with 1,catalyst 2 exhibited a significant decrease in the amount of N₂ adsorbed at p/p₀ > 0.01,as well as a decrease in its surface area from 2000 to 650 m²/g after the grafting process. The pore size distribution curves of the different catalysts revealed that the amine grafting process led to a significant decrease in their pore sizes. Furthermore,catalysts 3 and 4 exhibited negligible N₂ adsorption capacities,most likely because the channels in the structural frameworks of these catalysts were occupied by the larger PP and DABCO bases.

Fig. 3. Nitrogen adsorption isotherms (a) and micropore size distributions (b) of 1 and 2.

The Knoevenagel condensation reaction was selected as a suitable reaction to evaluate the heterogeneous catalytic behavior for the three solid base catalysts 2,3,and 4. The Knoevenagel condensation reaction is not only well known as a good model for weak base-catalyzed reactions,but is also well known in terms of its application for the synthesis of α,β-unsaturated carbonyl compounds from aldehydes/ketones and active methylene compounds ^{[32, 33, 34]}. In this study,we used a 1:1.1 molar ratio of benzaldehyde to malononitrile in cyclohexane with a catalyst loading of 1 mol% without any prior treatment. As a control experiment,all the free amines (i.e.,ED,PP,and DABCO) were evaluated in terms of their catalytic performance in the Knoevenagel reaction,and the results revealed that they performed effectively. As shown in Table 1,the Knoevenagel condensation of benzaldehyde and malononitrile in the presence of catalyst 2 gave an NMR conversion of > 99%,while catalysts 3 and 4 exhibited lower catalytic activity (93 and 63%,respectively). These results showed that the three amine-grafted catalysts exhibited efficient catalytic activity for the Knoevenagel reaction of benzaldehyde and active malononitrile. Notably,the Knoevenagel reaction of these two substrates gave an NMR conversion of < 3% in the presence of the non-grafted-amine MOF 1,which suggested that the presence of a basic catalytic site was essential to the activity of these catalysts. GC-MS analysis of the filtrate following the filtration of catalyst 2 from the reaction mixture showed that it did not contain any free ED. However,free PP and free DABCO were detected in the filtrates of the reactions catalyzed by 3 and 4,respectively,which indicated that some of the amine molecules grafted to catalysts 3 and 4 had leached from the framework into the liquors. These results therefore suggested that the poorer catalytic performance of 3 and 4 compared with 2 could be attributed to the leaching of the amines from the frameworks of these catalysts. The use of ketones in the Knoevenagel condensation reaction instead of benzaldehydes was also investigated using similar experimental conditions. Unfortunately,however,ketones exhibited much weaker reactivity towards malononitrile,most likely because of the lower reactivity of ketones.

Table 1 Catalytic performance of the different catalysts a. Catalyst Conversion b (%) (TOF c (min-1)) 1st run 2nd run 3rd run ED >99 — — PP >99 — — DABCO >99 — — 1 <3 — — 2 >99 (1.33) 98 (1.26) 96 (1.24) 3 93 (1.30) 74 (0.99) 55 (0.77) 4 63 (0.88) 22 (0.28) — a All of reactions were performed with benzaldehyde (1 mmol) and malononitrile (1.1 mmol) in 2 mL of cyclohexane using catalyst (1 mol%) at 40 °C for 2 h. b Determined by 1H NMR using 1,3,5-trimethylbenzene as an internal standard. c Turnover frequency calculated as moles of product formed per mole of free N atoms in the grafted MOF per minute during the first hour of the reaction.

Table 1 Catalytic performance of the different catalysts a.

The reusability of the three amine-grafted catalysts was also examined. Catalyst 2 was readily separated from the reaction mixture by centrifugation following the first run. The catalyst was then washed with ethyl acetate (5 × 2 mL) to remove any physically adsorbed reagents and dried at room temperature for 6 h before being used again in a subsequent run. Pleasingly,recycled catalyst 2 could be reused up to two times without any discernible reduction in its catalytic efficiency. For example,recycled catalyst 2 gave conversions of 98 and 96% for its second and third runs,respectively (Table 1). The PXRD pattern of recovered catalyst 2 showed that the structural integrity of the MOF remained almost unchanged after the catalytic reaction (Fig. 2). Recycling studies on catalysts 3 and 4 were also performed. However,recycled catalyst 3 gave a low conversion of only 55% after its third reaction. Furthermore,recycled catalyst 4 gave a much lower conversion of 22% for its second run. These results therefore revealed that catalysts 3 and 4 were unstable and that their grafted-amines were rapidly leaching from the channels of their MOFs to the liquor phase during the reaction,leading to a significant reduction in their amine- loading and catalytic efficiency.

A major problem for solid catalysts is the possibility that some of their active sites can migrate from the solid support to the liquid phase during the reactions and that these leached species are actually responsible for a significant part of the observed catalytic activity. With this in mind,we conducted leaching tests to determine whether the grafted-amino groups in catalysts 2,3,and 4 could leach from their MOF to the liquid phase during the course of the reaction. Briefly,the liquid phase of the reaction mixture was collected by centrifugation when the reaction reached 65% conversion with respect to the benzaldehyde (after 0.5 h,Fig. 4). The reaction solution was then transferred to a new reactor vessel and stirred for an additional 1.5 h,with small samples being taken for analysis at different time intervals. The results of this experiment showed that the reaction over catalyst 2 did not proceed any further in the absence of the catalyst,which excluded the presence of an active catalytic species in the solution. In contrast,the results of leaching experiments with catalysts 3 and 4 revealed that the filtrates from these reactions continued to react in the absence of the solid catalysts under similar reaction conditions. These results were therefore in accordance with those of the reusability tests with catalysts 3 and 4. Taken together with the results for the PXRD patterns of these catalysts,we can conclude that the ED-grafted catalyst 2 was the only one of the three catalysts to exhibited genuine heterogeneous behavior as a catalyst towards the Knoevenagel condensation reaction.

Fig. 4. Leaching tests for catalysts 2,3,and 4. Standard reactions with 2 (1),3 (3),and 4 (5); Leaching tests with 2 (2),3 (4),and 4 (6).

To determine whether the catalytic reaction occurred inside the pores of catalyst 2 or on its surface,we investigated the Knoevenagel condensation reactions of a variety of different aldehydes with different size and shape. The reactions of benzaldehyde and para-substituted benzaldehydes proceeded smoothly to give the desired products with very high yields (Table 2,entries 1-3). However,the use of 9-anthraldehyde as a substrate led to a much lower conversion of only 59% (Table 2,entry 4). This reduction in the conversion was attributed to the steric hindrance of the anthracene group,which would have made it difficult for the carbanion intermediate to attack the carbonyl group of the aldehyde. When larger nitrile compounds containing active methylene groups were used as substrates in the Knoevenagel reaction with benzaldehyde (Table 2,entries 5-7),there was a significant decrease in the conversions as the size of nitrile increased. For example,the use of tert-butyl cyanoacetate as a substrate failed to provide any of the desired condensation product,presumably because of its larger molecular volume. For comparison,we also conducted the same reaction in the presence of the homogeneous organic catalyst,ED (the molar amount of amine group was the same as in 2). In this particular case,the conversion was 78% (Table 2,entry 8). These results therefore suggest that the success of the Knoevenagel condensation reactions catalyzed by 2 is highly dependent on the pore size of the MOF,with large substrates being excluded from the catalytic process because they cannot diffuse into the channels of the MOF. These results indicate that it is necessary for the nitrile substrate,but not the aldehyde,to enter the channels of the MOF and trigger the catalytic reaction. Based on these results,we have proposed a plausible catalytic mechanism for the Knoevenagel condensation reaction catalyzed by 2,which is shown in Scheme 2. According to this mechanism,the active methylene compound would initially enter the channels of the MOF and interact with the grafted ED to generate a carbanion intermediate. This carbanion intermediate would then react with the carbonyl carbon atom of the benzaldehyde substrate to produce the product. The success of this reaction process is dependent on the active methylene compound being able to access the channels of the MOF. The volume of the active methylene compound must therefore be small enough to access the channels of the MOF to allow the reaction to proceed.

Table 2 Size-selective catalytic performances of 2 a. Entry a b Conversion b (%) (TOF c (min-1)) 1 >99 (1.33) 2 >99 (1.32) 3 92 (0.84) 4 59 (0.75) 5 98 (1.30) 6 84 (1.08) 7 0 (0) 8 d 78 (1.00) a All of these reactions were performed with aldehyde (1 mmol) and nitrile (1.2 mmol) in 2 mL of cyclohexane using catalyst (1 mol%) at 40 °C for 2 h. b Determined by 1H NMR using 1,3,5-trimethylbenzene as an internal standard. c Turnover frequency calculated as moles of product formed per mole of free N atoms in the grafted MOF per minute during the first hour of the reaction. d Using ED (1 mol%) as catalyst.

Table 2 Size-selective catalytic performances of 2 a.

Scheme 2. Catalytic mechanism of the Knoevenagel condensation reaction catalyzed by 2.

Table 3 shows a comparison of the work in this study with the results in several different studies for the Knoevenagel condensation reaction between benzaldehyde and ethyl cyanoacetate in the presence of a MOF-based solid catalyst. The results of this comparison clearly show that catalyst 2 provided the best performance of all of the MOF-based catalysts tested in terms of its high TOF value.

Table 3 Comparison of the Knoevenagel condensation reaction of benzaldehyde and ethyl cyanoacetate in the presence of different MOF-based catalysts. Catalyst Catalyst loading a (%) Temperature (°C) Time (h) Conversion (%) TOF b (min-1) Ref. Er(btc)(ED)0.75(H2O)0.25 1.1 40 2 98 1.30 this work ED grafted MIL-101 2.1 80 19 98 0.041 [24] IRMOF-3 2.6 40 1 99 0.635 [35] Zn2(tpt)2(2-atp)I2 c 0.5 60 2 37 0.617 [36] ZIF-9 d 5 25 4 100 0.083 [37] Fe-MIL-101-NH2 e 15.1 80 24 98 0.0103 [38] a The catalyst loading was calculated as the molar ratio of catalytically active groups to substrates. b Turnover frequency calculated as moles of product formed per mole of catalytically active group per minute. c tpt = tris(4-pyridyl)triazine; 2-atp = 2-aminoterephthalate. d Benzaldehyde and malononitrile were used as the substrates. e The composition of this MOF was Fe3OCl(H2O)2(bdc-NH2)3; H2bdc-NH2 = 2-aminoterephthalic acid.

Table 3 Comparison of the Knoevenagel condensation reaction of benzaldehyde and ethyl cyanoacetate in the presence of different MOF-based catalysts.

We have investigated the post-synthetic modification of a robust lanthanide-based MOF,which led to the production of a series of solid base catalysts,whilst preserving the structural integrity of the MOF. The ED-grafted catalyst 2 had a high loading of diamine and exhibited high thermal stability and permanent porosity,making it an effective heterogeneous catalyst in the Knoevenagel condensation reaction. This catalyst could also be readily recovered and recycled in subsequent runs without any discernible decrease in its activity. In contrast,catalysts 3 and 4,which were based on PP and DABCO,respectively,were found to be unstable,with the grafted-amines readily leaching from the channels of the MOFs into the liquor phase during the catalytic reaction,making them behave as homogeneous catalysts.