Post‐Synthetic Mannich Chemistry on Metal‐Organic Frameworks: System‐Specific Reactivity and Functionality‐Triggered Dissolution

Abstract The Mannich reaction of the zirconium MOF [Zr6O4(OH)4(bdc‐NH2)6] (UiO‐66‐NH2, bdc‐NH2=2‐amino‐1,4‐benzenedicarboxylate) with paraformaldehyde and pyrazole, imidazole or 2‐mercaptoimidazole led to post‐synthetic modification (PSM) through C−N bond formation. The reaction with imidazole (Him) goes to completion whereas those with pyrazole (Hpyz) and 2‐mercaptoimidazole (HimSH) give up to 41 and 36 % conversion, respectively. The BET surface areas for the Mannich products are reduced from that of UiO‐66‐NH2, but the compounds show enhanced selectivity for adsorption of CO2 over N2 at 273 K. The thiol‐containing MOFs adsorb mercury(II) ions from aqueous solution, removing up to 99 %. The Mannich reaction with pyrazole succeeds on [Zn4O(bdc‐NH2)3] (IRMOF‐3), but a similar reaction on [Zn2(bdc‐NH2)2(dabco)] (dabco=1,4‐diazabicyclo[2.2.2]octane) gave [Zn3(bdc‐NH2)1.32(bdc‐NHCH2pyz)1.68(dabco)]⋅2 C7H8 5, whereas the reaction with imidazole gave the expected PSM product. Compound 5 forms via a dissolution–recrystallisation process that is triggered by the “free” pyrazolate nitrogen atom competing with dabco for coordination to the zinc(II) centre. In contrast, the “free” nitrogen atom on the imidazolate is too far away to compete in this way. Mannich reactions on [In(OH)(bdc‐NH2)] (MIL‐68(In)‐NH2) stop after the first step, and the product was identified as [In(OH)(bdc‐NH2)0.41(bdc‐NHCH2OCH3)0.30(bdc‐N=CH2)0.29], with addition of the heterocycle prevented by steric interactions.


Introduction
Metal-organic frameworks (MOFs) [1] are currently attracting considerable interest for their porosity properties, and applications as diverse as carbonc apture, [2] catalysis, [3] drug delivery [4] and chemical weapon detoxification. [5] Much of this attention arises from the wide diversity of MOF structures,w ith variation of both the metal centresa nd organic linkersp roviding an essentially limitless numbero fp ossible materials. Of specific in-terestf or many applications is the potentialf or forming functionalised MOFs, [6] with particular functional groups appended to the pore walls. While such materials can sometimes be formed using al inker containing an appropriate substituent in the MOF synthesis, in practice many functional groups are intolerant to the synthetic conditions, or use of the functionalised linker in the synthesis gives rise to an unexpected product. Post-synthetic modification (PSM) [7] has emerged as apowerful tool for preparing such functionalisedM OFs, and it is often the only way to place ap articular substituent onto the pore walls of aM OF structure. Aw ide range of covalentp ostsynthetic modificationr eactions have been developed over recenty ears, including conversion of primary amines into amides, [8] isocyanates, [9] ureas, [10] azides, [11] b-amidoketones, [12] secondary amines [13] and diazonium salts, [14] aldehydes into hydrazones, [15] azides to triazoles, [16] bromidest on itriles, [17] as well as oxidation [18] and reduction [19] reactions. Despite this, there remainsaneed for new,v ersatile and syntheticallystraightforward methods that allow different functional groups to be incorporated into MOFs, regardless of their metal centres and framework structure.
The first step involves the formation of methoxymethyl amine groups by the reaction with paraformaldehyde and MeOH at 50 8C. These methoxymethyl amine groups were subsequently converted into the final product by reaction with pyrazole, imidazole or 2-mercaptoimidazole to give compounds 1-3,r espectively.A ll reactions proceeded withoutt he need for aLewis acid catalyst, which has the additional advantage of eliminating the work-up associated with catalystr emoval from the pores of the MOF and removes the possibility of pore blocking by the catalyst. The similarity between the PXRD patterns of UiO-66-NH 2 and the PSM products 1-3 ( Figure S2, S6 and S9) indicate that the original framework was maintained in all three cases.
The effectiveness of the PSM reactions in terms of the percentagec onversion of amino groups into the Mannich products was gauged by 1 HNMR spectroscopy.T he 1 HNMR spectra were obtained from MOF samples that were washed to remove unreacted reagents before digesting in NH 4 F/D 2 Ow ith [D 6 ]DMSO.F or the reaction with pyrazole( Hpyz), the 1 HNMR spectrum of 1 ( Figure S3) shows an umber of new signals in addition to those corresponding to the aromatic protons of the unmodified groups,p resent as D 2 bdc-NH 2 (d = 7.56d,7 .12s and 7.05d ppm). The aromatic protons of D 2 bdc-NHCH 2 pyz were observed at d = 7.62d,7 .25s and 7.08d ppm, overlapping with the signals from D 2 bdc-NH 2 and others attributed to minor (< 10 %) by-products.T he presence of the pyrazole ring on the digested framework of 1 was confirmed by the signals at d = 7.57 and 6.28 ppm. Attempts to remove the by-products by thorough washing with av arietyo fs olvents wereu nsuccessful, suggesting that these compounds are also derived from PSM reactions, with ad ouble-Mannich product the most likely.F ormylated by-products can be presenti nU iO-66-NH 2 , deriving from reactionw ith DMF during the MOF synthesis. [33] NMR analysis on digested samples of UiO-66-NH 2 showedn o evidencef or formylation, suggesting this is not the origin of the by-products presentin1.
By comparison of the integralsf or the signals at d = 7.13 and 6.28 ppm, the percentage conversion from ÀNH 2 into -NHCH 2 pyz groups was estimated to be 41 %. Ignoring the minor by-products, this gives the formula for 1 as [Zr 6 O 4 (OH) 4 (bdc-NH 2 ) 3.54 (bdc-NHCH 2 pyz) 2.46 ]. Attempts to increase the degree of conversion by carrying out the reaction at ah igher temperature or for al onger time period were unsuccessful, thoughi ts hould be noted that higher conversion to the methoxymethyl amine in the first step might not be observable in the 1 HNMR spectra of the digested product, given the likelyr eversion of any D 2 bdc-NHCH 2 OMe to D 2 bdc-NH 2 under the acidic digestion conditions. The Mannich reaction of UiO-66-NH 2 with imidazole (Him) as the nucleophile was more successful than that with pyrazole, with the amino groups fully converted into -NHCH 2 im groups. This was confirmed by the disappearance of the signals which correspond to the aromatic protons of the startingM OF,U iO-66-NH 2 ,i nt he 1 HNMR spectrum of the digested product. Instead,n ew signals at d = 7.56d,7 .14s and7 .07d ppm were observed( Figure1), corresponding to the protons from the benzene ring of D 2 bdc-NHCH 2 im. Furthermore, the presence of the imidazole ring can be confirmed by the presence of two singlets in the aromatic region (d = 7.75 and 7.03 ppm).T he signal at d = 7.03 ppm corresponds to two chemically similar  but non-identical protons from the imidazole ring, and this overlaps with the doublet from one of the aryl protons,w hereas the singlet at d = 7.75 ppm arises from the remaining proton peak of the imidazole ring. The signal attributed to the methylene protons can be seen at d = 4.56 ppm, close to the broad HDO peak resulting from the digestion solvent. The chemicalf ormula of this PSM product is [Zr 6 O 4 (OH) 4 (bdc-NHCH 2 im) 6 ] 2.
In contrast to the completec onversion observed for 2,t he comparable Mannich reaction with 2-mercaptoimidazole (HimSH) as the nucleophile gave only partial conversion. The 1 HNMR spectrum ( Figure S10) of the digested product 3 shows the presence of new peaks in addition to the aromatic proton peaks which correspond to the startingM OF,U iO-66-NH 2 .T he signals attributedt ot he aromatic protons of D 2 bdc-NHCH 2 imSH are observed at d = 7.68d,7 .26s and 7.08d ppm, respectively,a lthough these peaks overlap with others from minor by-products. The presence of new peaks at d = 6.98 and 6.76 ppm, from the imidazole ring, indicates that the 2-mercaptoimidazole ring was successfully grafted onto the MOF framework.
The percentagec onversionsf or the PSM reactions generating 1-3 are summarised in Table 1. The differencesi nd egree of conversion can be related to the nucleophile strength. Imidazole is as tronger nucleophile thanp yrazole due to its higher basicity,a nd is therefore more susceptible to nucleophilic substitution with ÀNHCH 2 OCH 3 ,l eadingt oahigherc onversion. Thes teric demands of the nucleophile also have some influenceo nt he extento ft he reaction, with the lowest conversion achieved in the case of 2-mercaptoimidazole, the larg-est of the nucleophiles employed. This can be rationalised by the more restricted diffusiono f2 -mercaptoimidazole within the pores of the MOF.
The thiol substituent in 3 was anticipated to be abletocoordinate to soft metal centress uch as mercury(II). In order to probe the effect of different -NHCH 2 imSH loadings on Hg II uptake, as econd thiol-containing MOF was prepared, using the same conditions as for 3,b ut with the temperature fort he second step reduced from 80 to 50 8C. It was anticipatedt hat the lower temperature during the second step would lead to a lower conversion to the -NHCH 2 imSH group.
The 1 HNMR spectrum ( Figure S13) of the digested product formed under these conditions, 3a,s howedt he presenceo f the modified group (-NHCH 2 imSH),t hough present in al ower relative concentration than in 3.T he percentage conversion from -NH 2 into -NHCH 2 imSH groups was estimated as 21 %, giving af ormula for 3a of [Zr 6 O 4 (OH) 4 (bdc-NH 2 ) 4.74 (bdc-NHCH 2 imSH) 1.26 ]. This confirms that the reaction temperature has as ignificant impact on the degree of modification, with a lower temperature leading to lower conversion.
The TGA profileso ft he PSM products 1-3 and 3a exhibit similar features to that forU iO-66-NH 2 ( Figure S14). Therei sa n initial mass loss (up to 110 8C) corresponding to removal of 1,4dioxanef rom the pores. As mall, gradualm ass loss, observed in the range 110-470 8C, is attributedt othe loss of residual solvent in the pores and/or the dehydroxylation of the Zr 6 O 4 (OH) 4 nodes. [34] The final mass loss, beginning at 470 8C, is due to the decomposition of the framework. Based on the TGA profiles, 1 has 4.0, 2 has 3.0, 3 has 5.0, 3a has 5.5 and UiO-66-NH 2 has 7.0 molecules of 1,4-dioxane per Zr 6 O 4 (OH) 4 unit in the unactivated MOFs.T his shows that the amount of 1,4-dioxanei nt he pores decreases as the degree of post-synthetic modification increases. This is unsurprising, since the greater the degree of conversion,t he lower the residual space available to accommodate guest solvent molecules.
The BET surface areas of 1-3 and 3a were determined based on their N 2 adsorption isothermsa t7 7K (Figure2). The compounds were activatedu sing the conventionala ctivation temperaturef or UiO-66 andi ts derivatives (120 8Cf or 12 h), and the BET surfacea rea for UiO-66-NH 2 obtained in this work (S BET = 1041 m 2 g À1 )i ss imilar to previously reportedv alues. [35] All PSM products exhibit type Ii sotherms,i ndicative of micro- porousm aterials, and have lower BET surface areas than UiO-66-NH 2 ,w ith S BET values of 528 m 2 g À1 for 1,2 90 m 2 g À1 for 2, 352 m 2 g À1 for 3 and6 08 m 2 g À1 for 3a.B ET surface areas are governed by the degree of conversion and the size of the modifiedg roups. In general, the BET surface area reduces as the percentage conversion increases and 2,w ith complete conversion,h as the lowest surface area. The presence of larger pendantg roups in the pores also leads to lower BET surface areas, with the value for 3 lesst han that for 1,d espite 1 possessing ah igher degree of modification. The CO 2 adsorption isothermso ft he PSM products were measured at 273 K( Figure S15) to assess the influence of the modifiedg roups on the CO 2 uptake capacities.A ll PSM products showl ower CO 2 uptake capacities than UiO-66-NH 2 ,a ttributable to the reductioni np ore volumea nd the lower percentage of ÀNH 2 groups in the pores. Of the PSM products, 1 shows the highest CO 2 uptake which is probably due to the favourable interactions of CO 2 molecules with the nitrogen atom in the pyrazole ring. Compound 2 shows al ower CO 2 uptake than 1,d espite havingh igher percentage of heterocyclesi n the pores,w hich is consistent with the lower BET surface area, itself ac onsequence of the high degree of modification. Compounds 3 and 3a show the lowest CO 2 uptakec apacities at 1bar and this may be due to pore blockingc aused by higher steric hindranceo ft he modifiedg roups. Nonetheless,t he proportion of thiol groups in the pores has little impact on the CO 2 uptake capacities, as evidenced by the relatively small differencei nC O 2 uptake between 3 and 3a.T he modified MOFs show enhanced CO 2 /N 2 selectivity over UiO-66-NH 2 ,t hough this is largely ac onsequence of their low N 2 uptake at 273 K.
Mercury uptake capacities were calculated using [Equation (1)] where C i and C e represent the initial and equilibrium Hg II concentrations, respectively.I na ddition to PSM products 3 and 3a,t he Hg II uptake capacities of the unmodified MOFs, UiO-66 and UiO-66-NH 2 ,w ere investigated for comparison, with the results presented in Ta ble 2.
Hg II ðÞ uptake % ðÞ ¼ The post-synthetic grafting of thiol groups in the pores of UiO-66p roved to be beneficialf or Hg II absorption, as the uptake capacities were significantly increased for 3 and 3a over the unmodified MOFs.P erhapss urprisingly,t he highest Hg II uptake was observed for 3a,d espite 3 having ah igher loading of thiol groups in the pores. This reflects the lower porosity of 3,w hich is likely to lead to some of the thiols being unavailable to interactw ith the Hg II ions. The Hg II uptake in 3a is comparable to that reported for the previously reported de-rivativeU iO-66-(SH) 2 , [36] which is one of the highestr eported for aM OF,d emonstrating the potential of 3a for mercury removal. PXRD ( Figure S17) confirmed that 3a retains its crystallinity on treatment with HgCl 2 (aq).

Mannich reactions on [Zn 4 O(bdc-NH 2 ) 3 ], IRMOF-3
IRMOF-3 contains large channels ( % 9.6 )a nd there is considerable precedence for the post-synthetic modification of the amino groups that protrude into its pores. [28] As IRMOF-3h as a low stability towards moisture and alcohols, [37] toluenew as selected as the optimum solvent for the Mannich reaction.
To demonstrate the applicability of Mannich reactiono n IRMOF-3, the PSM reaction with pyrazole was carriedo ut using the reactionc onditions outlined in Scheme2.
The effectiveness of the PSM reaction was gauged by 1 HNMR spectroscopy on the DCl/D 2 O-digested product 4 (Figure S19). In addition to the signals corresponding to the aryl protons of D 2 bdc-NH 2 ,n ew features attributed to the aryl protons of the modified product were observed at d = 7.89d, 7.46d and 7.20dd ppm. The successful incorporation of the-NHCH 2 pyz groups could also be evidenced by the emergence of new signals at d = 7.85d,7 .72d and 6.25dd ppm, corresponding to the protons of the pyrazole ring. The peak attributed to the methylene protonsw as located at d = 5.68 ppm.  180.0297), respectively.T he PXRD pattern of 4 ( Figure S18) shows the similarities in peak positions with the starting MOF, IRMOF-3,i ndicating that the bulk framework structure remained unchanged upon PSM. Nonetheless, ad egree of degradationw as observed, as evidenced by the broadening of peaks and reduced intensities. The presence of stoichiometric MeOH in the first step and as as ide product in the second step may causes ome crystal degradation. Attempts to analyse 4 by single crystal X-ray crystallography wereu nsuccessful due to poor diffracting power of the sample.
To demonstrate the applicabilityo ft he Mannich reaction on DMOF-1-NH 2 ,t he reactionw as carried out using the same conditions as outlined for IRMOF-3 in Scheme 2. The 1 HNMR spectrum of the digested product 5 ( Figure S24) shows the presence of aromatic protons attributedt oD 2 bdc-NH 2 (d = 7.82, 7.48 and 7.13 ppm) and D 2 bdc-NHCH 2 pyz (d = 7.89, 7.49 and 7.20 ppm). The peaks at d = 7.13 and 7.20 ppm overlap with the signals from the aryl protons of residual toluene solvent. The protons of the pyrazoler ing are located at d = 7.85 7.72, and 6.25 ppm. The peak attributedt othe aÀCH 2 protonsisobserved at d = 5.68 ppm althought here is some overlap between this peak and that forH DO, present from the digestion mixture. Comparing the integralso ft he protonsa td = 7.48-7.49 ppm and d = 6.25 ppm, the percentage conversion of amino into -NHCH 2 pyz groupswas calculated to be 56 %.
The PXRD pattern of 5 is completely different to that of DMOF-1-NH 2 ( Figure S22), revealing as ignificant structural differenceb etween the two materials. Indeed, the PXRD pattern of 5 does not match any of the PXRD patterns reportedi nt he literaturef or DMOF-1 type materials. Inspection of 5 under an opticalm icroscope revealed the presence of small colourless crystalsa nd the absence of brownb lock crystals, characteristic of DMOF-1-NH 2 and its derivatives. This observation suggests that DMOF-1-NH 2 has undergone ac omplete structural change upon reaction.
The crystal structure of 5 was successfully elucidated by single crystal X-ray crystallography andi ss hown in Figure 3.
The compound crystallises in the trigonal space group R-3m, and the asymmetric unit ( Figure S44) contains one quarter of a zinc atom (Zn1 and Zn2 have 8.333 %a nd 16.667 %o ccupancy, respectively), one twelfth of ad abcol igand and one quarter of al igand whichi sc omprised of bdc-NH 2 and bdc-NHCH 2 pyz, disordered in a3 4:56 ratio.
Attempts to accurately determine the structuralv oid volume via the PLATON SQUEEZE algorithm were hampered by pendant group site-occupancies, disorder and the smearingo fe lectron density.T he TGA of 5 indicates am ass loss that corresponds to two toluene molecules for every three zinc centres present,a nd this provides af ormulation of 5 as [Zn 3 (bdc-NH 2 ) 1.32 (bdc-NHCH 2 pyz) 1.68 (dabco)]·2 C 7 H 8 .
Overall, the SBU in 5 contains three zinc centres, one 6-coordinate and two 4-coordinate (Figure 3a). The Zn1 metal centre is in ad istorted octahedral coordination environment, and is coordinated to six O2 donor atoms, each from ad ifferent carboxylate group. In contrast, Zn2 exhibits ad istorted tetrahedral coordination geometry,b eing coordinated to three O1 donor atoms from different carboxylate groups and to the nitrogen atom N1 of the dabco ligand.
In order to investigate the cause of the structuralt ransformation from DMOF-1-NH 2 into 5,aseries of control studies were carriedo ut. No structural change was observed when DMOF-1-NH 2 crystals were heated in toluene, or when the crystals were treated separatelyw ith paraformaldehyde, MeOH or pyrazole( Figure S27). This suggested that the formation of the methoxymethyl amine intermediate DMOF-1-NHCH 2 OCH 3 in the first step was unproblematic, but that the structural transformation occurred in the second step of the Mannich reaction. In order to confirm this, the reactiono fD MOF-1-NHCH 2 OCH 3 with pyrazolew as monitored under an optical microscope equipped with ac amera. Ther eactionc onditions were modified in order to be able to view the reaction in this way.I np articular, DMOF-1-NHCH 2 OCH 3 crystalsw ered ispersed on amicroscope slide containing asolution of pyrazoleintoluene at room temperature. After five minutes,t he crystals began to dissolve, with complete dissolution observed after 40 minutes. An ew phase, correspondingt ot he crystalso f5, was first observed after approximately twenty minutes (Figure 4), confirming that 5 is produced in ad issolution-reprecipitation process.
Although it is not possible to provide ad efinitive mechanism for the dissociation of DMOF-1-NHCH 2 OCH 3 ,aproposed reactionm echanism that leads to the dissociation of the SBUs is shown in Figure 5.
After the first step of the Mannich reaction, the methoxymethyl amine speciesi sl ocalisedi nc lose proximityt ot he bridging dabco ligands. Upon addition of pyrazole, af acile reaction displacing methanol can occur to yield the -NHCH 2 pyz group, whichi sa ligned in such aw ay as to compete in an intramolecular manner with dabco for coordination to the Zn II metal centre. Displacement of dabco would break the three-di-mensionalnetwork of the DMOF-1 framework, leadingt or apid delamination,a nd ultimately triggering framework dissolution. Notably, in the crystal structure of 5,t he -NHCH 2 pyz group is directeda way from the dabco ligand ( Figure S43), so is unable to compete with it for coordination.M oreover,adiaza-[18]crown-6 ligand functionalised with pendant pyrazole groups using aM annich reaction also exhibited fragmentation behaviour in the presenceo ft ransition metals, [40] leading further credence to this hypothesis.
It should be noted that PXRD patterns for bulk samples of 5 show the presence of more than one phase, sot he degree of occupancy of the pores by toluene in the crystal structure is an estimate, and although the ratio of linkersi nt he 1 HNMR spectra are consistent betweens amples, this too may have been differentint he crystal analysed crystallographically.
The Mannichr eaction of DMOF-1-NH 2 with imidazole as the nucleophile was carried out using the same conditions as with pyrazole( Scheme 2). The 1 HNMR spectrum of the digested product 6 ( Figure S30) shows aromatic protons from D 2 bdc-NH 2 and D 2 bdc-NHCH 2 im, and from the integralst he percentage conversion of amino into -NHCH 2 im groups was calculated to be 65 %, giving af ormula for 6 as [Zn 2 (bdc-NH 2 ) 0.7 (bdc-NHCH 2 im) 1 The PXRD pattern of 6 and the startingM OF,D MOF-1-NH 2 closelym atch one another (Figure S29), demonstrated that PSM does not affect the gross structureo rt he crystallinity of the product. Furthermore, visual inspectiono f6 confirmed the presence of only brown block crystals and the absence of new phases.A ttempts to analyse 6 crystallographically were hampered by crystal twinning. Nonetheless, as creening experiment suggested that there were similarities in the unit cell parameters of 6 (a = 15.2955 (17) , b = 15.2860 (15) , c = 19.207(2) )a nd those of DMOF-1 (a = 15.063(2) , c = 19.247 (5) ).
Based on these results, it is clear that framework dissolution does not occur when imidazole is used as an ucleophile.I ti s believed that substituting pyrazole by imidazole prevents the dissolution of DMOF-1-NHCH 2 OCH 3 ,b ye liminating the possibility of coordinative competition with dabco. The "free" nitrogen atom in imidazole is positioned beyondt he coordination sphereofthe zinc(II) centre and, as aconsequence, the process showni nF igure 5i sunable to occur.
The Mannich reactiono fD MOF-1-NH 2 with 2-mercaptoimidazole as the nucleophile was attempted using the same conditions as in the reactionw ith imidazole. However,t he 1 HNMR spectrum of the digested solid showedo nly signals corresponding to the aryl protons of DMOF-1-NH 2 ( Figure S37), indicating that the inclusiono f2 -mercaptoimidazole onto this MOF framework was unsuccessful. The PXRD pattern ( Figure S36) is similar to that for DMOF-1-NH 2 ,i mplying that the framework was retained throughout the experiment. The unsuccessful graft-  ing of 2-mercaptoimidazole onto the MOF framework is likely to be due to its larger size than imidazole, which makesi tt oo big to pass through the pore windows (2-mercaptoimidazole: 8.4 6.6 .D MOF-1-NH 2 channels:5 .3 4.8 ).

Mannich reactions on [In(OH)(bdc-NH 2 )],MIL-68(In)-NH 2
[In(OH)(bdc-NH 2 )] , MIL-68(In)-NH 2 ,i sathree-dimensional MOF that is constructed from chainso fI nO 4 (OH) 2 octahedral units that are linked together by bdc-NH 2 ligands to form triangular ( % 6 )a nd hexagonal ( % 16 )o ne-dimensional channels. In MIL-68(In)-NH 2 ,t he amino groups are oriented towards the InO 4 (OH) 2 octahedral chains rather than projecting into the pores. However,t his has not prevented successful tandem post-synthetic modifications involving formationo ft he azide and subsequent click reactions from being carried out, [11] so presumably some flexibility is possible to accommodate the bulkier, modified groups.
MIL-68(In)-NH 2 was prepared using an analogous synthesis to that for MIL-68(In), originally reported by Loiseau and coworkers. [30] In at ypical PSM procedure, MIL-68(In)-NH 2 crystals were treated with paraformaldehydea nd MeOH at 50 8Cf or 24 h. In this reaction, MeOH was used as ar eactant as well as a solvent, as MIL-68(In) is stable towards alcohols, thus eliminating the need to use ad ifferent solvent. The intermediate product was then washed with 1,4-dioxane and treated with pyrazole at 80 8Cf or 24 h, before quenching the reactionb yw ashing the sample with fresh 1,4-dioxane.
The 1 HNMR spectrum of the digested PSM product 7 (Figure S39) was obtained by digesting the MOF in ab asic aqueous solution (NaOD/D 2 O). In addition to the signals corresponding to the aromatic protons of Dbdc-NH 2 À ,t wo new sets of signals were observed in the downfield region of the spectrum. However,t he absence of peaks attributedt ot he protons of the pyrazole ring indicated that the PSM reactiond id not afford the expected pyrazole-containingp roduct. The signals at d = 7.73d, 7.36s and7 .15d ppm are believed to be due to the aryl protons from the intermediate MOF,M IL-68(In)-NHCH 2 OCH 3 ,o bserved as Dbdc-NHCH 2 OCH 3 À in the NMR spectrum. The peaks attributedt ot he methylene protons and methyl terminus of Dbdc-NHCH 2 OCH 3 À are located at d = 4.87 and 4.76 ppm, respectively, although these are partly obscured by the peak from HDO, presentf rom the digestion solvent. The other signals, at d = 7.68d,7 .43s and7 .18d ppm, are believed to be from the imine Dbdc-N = CH 2 À ,w ith mass spectrometry providing support for this (vide infra).
In order to confirmt hat the observedp roductsd on ot require the presence of pyrazole, the reactionm ixture was analysed prior to its addition. The first step of the Mannich reaction is depicted in Scheme 3, broken down into two stages. As anticipated, the 1 HNMR spectrumo ft he digested product ( Figure S40) illustrates ah igh similarity with that for 7,w ith only small differences in the relative proportions of the two products.
This finding validates the hypothesis that 7 containsu nreacted ÀNH 2 groups, as well as imine and methoxymethyl amine species. The presence of the imine could be due to in-complete reaction with methanolo r, alternatively, from the partial hydrolysis of D 2 bdc-NHCH 2 OCH 3 in the digestion medium. Given that D 2 bdc-NHCH 2 OCH 3 appears to be stable under the digestion conditions, the most reasonable formulation for 7 includes both substituents, and can be represented by the formula [In(OH)(bdc-NH 2 ) 0.41 (bdc-NHCH 2 OCH 3 ) 0.30 (bdc-N = CH 2 ) 0.29 ].
The negative ion ESI mass spectrumo ft he base-digested product 7 ( Figure S41 Figure S38), indicating that framework integrity is maintained and the PSM reactiond id not alter the crystallinity of the product.
Crystals were grown from dioxane, and the crystal structure of 7·0.8 dioxane was successfully elucidated by single crystal Xray crystallography.T he asymmetric unit ( Figure S46) consists of two indium(III) centres with site occupancies of 0.5 and 0.25 for In1 and In2, respectively,o ne half and one quarter of ad icarboxylate ligand and two OH ligands (based on O1 and O5) with combined site occupancies of 0.75. Finally,t here was evidence for some diffuse solvent present in the framework which was modelled as four-fifths of ad ioxane molecule per indium centre based on TGA evidence.
As can be seen in Figure 6a,b, the overall framework topology has not changed significantly during the reactionf rom that of MIL-68(In)-NH 2 ,i na greement with the PXRD data. Although there is some evidence for the nitrogen atoms of the tag groups,d isordered overs everal positions, further evidencef or the nature of the substituents was unavailable.
The most notable insightf rom the crystal structure of 7 is the short distance between the nitrogen atoms on neighbouring benzene rings (Figure 6c). Although these atoms have only partial occupancy,t his proximity illustrates the difficulties involvedi np lacing al arge substituent on one of these atoms. This provides justification for the argument that the second step of the Mannich reaction is disfavouredi nt his case on steric grounds.
With regards to the PSM reactions on UiO-66-NH 2 ,t he degree of conversion from -NH 2 into -NHR (R = CH 2 pyz, CH 2 im and CH 2 imSH) depends on the strength and size of the nucleophiles. Complete conversion was achieved with the strongest nucleophile (imidazole) whereas al ower conversion (41 %) was obtained with the isosteric weaker nucleophile, pyrazole. The use of al arger nucleophile, 2-mercaptoimidazole, led to the lowest conversion (36 %) and this is mostl ikely due to the restricted diffusion of the nucleophile within the pores of UiO-66-NH 2 .T he modified MOFs have lower BET surface areast han UiO-66-NH 2 ,b ut show enhanced selectivity for CO 2 over N 2 .I n addition, the thiol-containing products show excellent uptake of mercury(II)f rom aqueous solutions.
With regard to the PSM reactiono nI RMOF-3,7 5% conversion of -NH 2 into -NHCH 2 pyz was achieved whilst using pyrazole as an ucleophile. However,t he successful PSM reaction comes at ac ost of decreased product crystallinitya se videnced by the broadening of peaks and reduction in peak intensities in the PXRD pattern of the PSM product.
Subjecting MIL-68(In)-NH 2 to as imilarP SM reactionw ith pyrazole, gave am odified product 7 that did not contain the heterocycle. The first step of the Mannich reaction proceeded, but the methoxymethyl amine intermediate did not react with pyrazole in the expected manner.T he X-ray crystal structure of 7 suggestst hat this is ac onsequence of the location and orientation of these groups which are inaccessible to the pyrazole molecules, thus preventing the second step in the Mannich reaction from occurring.
This work has demonstrated that the post-synthetic Mannich reactionr epresents av ersatile route to introducing complex functionalities into ar ange of metal-organic frameworks, and we are currently working to furtherd evelop the breadth of this approach.

Experimental Section
Full experimental details are presented in the electronic supplementary information. As an example, the reaction of UiO-66-NH 2 with formaldehyde and imidazole is presented here. UiO-66-NH 2 (117 mg, 0.4 mmol eq. of NH 2 )a nd paraformaldehyde (24 mg, 0.8 mmol, 2equiv.) were added into ag lass vial containing methanol (5 mL). The vial was placed in an oven and heated at 50 8Cf or 24 h. The powder was then washed with methanol (three times) via centrifugation to remove any residual paraformaldehyde in the pores or on the solid surfaces. The powder was subsequently treated with imidazole (54 mg, 0.8 mmol, 2equiv.) in 1,4-dioxane at 80 8Cf or 24 hb efore quenching the reaction by rinsing the sample with fresh 1,4-dioxane. The product was soaked in 1,4-dioxane for 3days, replacing the solvent with fresh solvent every 24 h, before isolation by centrifugation. Prior to characterisation, samples were left to dry in air for 2hto obtain free-flowing powders. Full details of the X-ray crystal structures of 5 and 7·0.8 dioxane are given in the Supplementary Information. The structures have also been deposited with the Cambridge Structural Database (CCDC 1824632-3 contains the supplementary crystallographic data for this paper.T hese data are provided free of charge by The Cambridge Crystallographic Data Centre).