Exchange of Coordinated Solvent During Crystallization of a Metal–Organic Framework Observed by In Situ High‐Energy X‐ray Diffraction

Abstract Using time‐resolved monochromatic high energy X‐ray diffraction, we present an in situ study of the solvothermal crystallisation of a new MOF [Yb2(BDC)3(DMF)2]⋅H2O (BDC=benzene‐1,4‐dicarboxylate and DMF=N,N‐dimethylformamide) under solvothermal conditions, from mixed water/DMF solvent. Analysis of high resolution powder patterns obtained reveals an evolution of lattice parameters and electron density during the crystallisation process and Rietveld analysis shows that this is due to a gradual topochemical replacement of coordinated solvent molecules. The water initially coordinated to Yb3+ is replaced by DMF as the reaction progresses.

The synthesis of metal-organic frameworks (MOFs) has,t o date,b een ap rocess fraught with assumptions,d ue to the difficulty of obtaining high quality structural data in situ during their formation that would provide detailed information about their crystallisation mechanism. [1] Although studies of kinetic vs.thermodynamic control in the synthesis of MOFs have been reported by screening products of reactions isolated as af unction of time,u sing both experimental and theoretical approaches, [2] an understanding of the early stages of MOF crystallisation processes remains poor.One untested assumption is that after aM OF nucleates,i tc rystallises without undergoing further structural changes.T he functionality of many metal-organic framework materials derives from their ability to interact with guest molecules.InMOFs in which the metal coordination sphere is not fully saturated with structural ligands,t he interaction between metal and coordinated molecules tends to be particularly strong, and this may give rise to favorable adsorption and catalysis properties. [3] Any interaction between guest and material must necessarily result in some level of change to the observed electron density distribution and unit cell size. This effect is prominent in several of the most widely studied MOFs:f or example,t he dehydroxylated UiO-66 framework loses hydroxyl and it unit cell contracts by ca. 0.05 , [4] while the difference between the guest-bound and bare MOF-74/ CPO-27 frameworks is on the order of 0.1 for both axes of the hexagonal cell. [5] These changes are well within the range that can be clearly resolved using high-resolution powder diffraction, and indeed this method has been used extensively in structural studies of the effect of adsorbed molecules on MOFs under gas atmospheres. [6] In many cases of MOF synthesis using solvothermal methods,i ti su nclear whether the framework is initially formed with coordinated solvent that is then exchanged with another ligand to reach the final product, or if the final product is formed from the start as the only species.T his knowledge would be valuable to the large scale deployment of MOFs,a llowing the optimization of syntheses to reduce or eliminate the need for certain types of post-synthetic processing,s uch as the high-temperature dehydroxylation of UiO-66.
Energy-dispersive X-ray diffraction (EDXRD) has been used to great effect to follow solvothermal crystallization of MOFs, [7] building on earlier work on hydrothermal zeolite and zeotype formation. [8] Here,u sing X-rays without monochromation provides sufficient intensity to observe crystallisation in large-volume reaction vessels,but with the serious disadvantage of the intrinsic low resolution of energydiscriminating solid-state detectors.Thus,although the changing intensity of well-resolved Bragg peaks can be monitored in real time to yield crystallisation curves,i ti sd ifficult to observe and quantify small changes in unit cell parameters and impossible to perform atomistic (i.e.R ietveld) refinement, severely limiting the level of structural information available.M ore recently,a dvances in technology have made monochromatic XRD feasible.R ecent work has used in situ monochromatic diffraction to study the mechanochemical [9] and solvothermal [10] synthesis of MOFs and while it has been shown that scale factors (phase fractions) and peak positions can be extracted, no full structural treatment has yet been performed of the temporal data measured in situ. Another great challenge in in situ studies is the trade-off between reactor size and data quality:alarger reactor will provide conditions comparable to conventional, laboratory-scale chemistry,w hile ac apillary will provide optimal data quality but is severely limiting in terms of reproducing realistic synthetic conditions.
In this work, we analyse aM OF crystallization taking place within astirred reaction tube of relatively large volume (ca. 5mL, 9mmd iameter) using high intensity monochromatic radiation. We are able to obtain high-quality data in situ under reaction conditions similar to those used in ac onventional large-scale batch synthesis.N ot only do we obtain detailed kinetic information with exceptional time resolution, we are also able to observe the exchange of labile coordinated solvent within af ramework material during its formation, which we can quantify using Rietveld analysis of the data measured in situ;this allows refinement of crystal structure as the reaction proceeds.O ur results demonstrate as ignificant advance in the quality of diffraction data from crystallising material, obtained under solvothermal conditions from arelatively large-scale synthesis.
Them aterial investigated herein is an ew MOF [Yb 2 -(BDC) 3 (DMF) 2 ]·H 2 O( BDC = benzene-1,4-dicarboxylate and DMF = N,N-dimethylformamide) prepared under solvothermal conditions,f rom mixed water/ DMF solvent. The structure was solved and refined using single-crystal analysis (see the Supporting Information (SI)). Thef ramework crystallizes in the monoclinic C2/c space group and contains chains of Yb and carboxylate,related to apreviously reported Er-BDC framework. [11] Thechains run down the c-axis of the framework, and are located at the corners of diamond-shaped 1D channels.R unning along each channel are the labile coordination sites of the Yb,w hich are occupied by DMF in the equilibrium structure,F igure 1. Them aterial can be thermally desolvated to yield ap ermanently porous framework:afull characterisation is provided in the SI.
In situ X-ray diffraction data during the reaction of Yb chloride hydrate and BDC in mixed water/ DMF were collected at three temperatures (90, 110 and 120 8 8C), with patterns being collected at 30 si ntervals.U nless otherwise stated, the data presented in the main text are from the 120 8 8C reaction;t he other data follow similar trends,a nd are included in the SI. Using sequential Pawley refinements of each pattern, we simultaneously extract both lattice parameters,a nd total quantity of crystalline material through the integrated area beneath peaks (Figure 2). Figure 3shows the   concurrent changes seen in crystalline quantity and cell parameters.T he crystalline quantity can be analysed to extract kinetic information (see SI), but the temporal shift in lattice parameters reveals the evolution of structure during crystallisation.
Thechanges in lattice parameter are small, but changes on the order of 0.001 are easily and reproducibly resolved, and show ameaningful trend with temperature (see SI). It should be noted that the internal thermocouple shows that temperature is reached prior to the observation of Bragg peaks (Figure 3) so thermal effects on lattice parameters can be ruled out. In fact, their evolution continues with the same trend throughout the crystallisation, so we are confident the crystalline material is seen under isothermal conditions.W e also note the Bragg peak widths do not decrease significantly during the period of analysis so we rule out changing crystallite size as as ignificant effect on lattice parameter evolution.
As the reaction progresses,a ni ncrease in the unit cell lengths can be seen for the a and c cell parameters,while the b parameter decreases;this is shown in Figure 3. As the a and b cell parameters correspond to the diagonals of the diamond shaped channels,e xpansion in one direction must be countered by contraction in the other. Such behaviour is expected for ad iamond shape in which the interior angles can change freely but the side lengths are constrained, and is reminiscent to that seen in the "breathing" MOF MIL-53, which also has diamond-shaped one-dimensional channels where the introduction of weakly bound molecules,orapplication of temperature or pressure,c auses similar changes in the relative pore dimensions, [12] although the evolution of lattice parameters for our material is several orders of magnitude smaller.
As equential Rietveld refinement was performed on the individual patterns.T he occupancyo fa ll atoms in the DMF moiety was linked to asingle parameter and allowed to freely refine,e xcept for the oxygen atom which was fixed at occupancy1 ,as this site is occupied by oxygen regardless of H 2 Oo rD MF coordination. Ther esults agree very well with the temporal evolution of the ratio of the areas of the (200) and (110) peaks obtained from Pawley refinement, shown in Figure 4a for the 120 8 8Cd ata set. This is consistent with the fact that the DMF electron density lies primarily on the crystallographic (200) plane;F igure 4b shows how the powder patterns are sensitive to the nature of coordinated solvent. Themagnitude of change of DMF occupancyduring crystallisation is not as great as that going from the solely Yb-OH 2 to solely Yb-DMF,but the simulation does not take into account non-coordinated solvent, which may still contribute to the electron density.D uring the period of the in situ analysis it more likely that the solvent simply is changing from water-rich to DMF rich rather than representing complete exchange of one by another.
To confirm the reason for the changing structural parameters of the Yb-BDC material during its formation, combined thermogravimetric analysis,d ifferential scanning calorimetry and mass spectrometry (TGA-DSC-MS) experiments were performed on quenched samples prepared in the same sized, stirred reaction vessel used in the in situ studies but heated in an oil bath at 120 8 8Cfor three durations (30, 45 and 60 mins) within the timescale of crystallisation seen in the in situ experiments.T his showed distinct differences in the solvent loss steps (Figure 5and SI). TheTGA trace of the 30 minutes sample shows considerably more surface water (no DMF is lost at this stage,a ss hown by the MS data) and the subsequent bound solvent loss is less well defined, perhaps suggesting water is lost from the bulk as well as the surface. More significantly,t he DSC traces for the pair of events between 140 and 260 8 8C, which correspond to bound solvent loss,s how small shifts to higher temperatures as the sample synthesis time is increased. This would be consistent with adifferent solvent composition in the solids as synthesis time is increased. Themost striking evidence for achanging solvent composition, however,c omes from the MS traces:a sseen in Figure 5c the relative amount of DMF lost in each of the two solvent loss features shows as ystematic change in ratio, entirely consistent with less directly bound DMF being present in the samples quenched at shorter reactions times. With the caveat that quenching studies will always carry the risk that the material recovered undergoes some irreversible change upon cooling and extraction from the solvent, such as exchange of water with the air,o ur TGA-DSC-MS results provide important corroborative evidence for the conclusions from the in situ study. Thus we construct aconsistent model for the evolution of lattice parameters in which solvent exchange takes place during the formation of the material. At the early stages of reaction the material is water rich, and the directly coordinated water is replaced by DMF as the reaction proceeds,with the framework geometry adjusting to account for the change in the size and shape of the occluded molecules.T hus the chemical composition [Yb 2 (BDC) 3 (solvent) 2 ]·solvent (where solvent = H 2 Oand/or DMF) is ageneral representation of the materials formed in the solvothermal reactions,w ith the ultimate product being [Yb 2 (BDC) 3 (DMF) 2 ]·H 2 O, the single crystal studied that was prepared using aconsiderably longer reaction time than the in situ experiments.
We have demonstrated that, under conditions close to those used for conventional MOF synthesis,i ti sp ossible to observe solvent exchange in situ during synthesis and to extract quantitative information regarding composition from Rietveld analysis.W hile capillaries have previously been effectively used by others to study the solvothermal crystallisation of inorganic materials in situ, [13] our use of al arge volume reactor (5 mL) has the distinct advantage of allowing reagents to be easily added in homogeneous,p re-planned quantities to ensure reproducibility;t his is particularly important when solid/liquid mixtures are investigated that are difficult to transfer into capillaries in desired quantities. Our observation of solvent exchange during crystallisation suggests ap reviously unexplored method of optimizing synthetic parameters for control of composition of MOFs in both large-scale MOF deployment and lab-scale reactions. Our in situ XRD approach would also be valuable in the study of other MOF formation processes,for example,indetermining the rate at which different metals or ligands incorporate into asolid-solution ("multivariate") MOF:the difference in cell parameters between isostructural phases is well above the smallest changes that can be observed. Fore xample,i nt he work of Lin Fooe tal.,t here is around 0.1 d ifference between the end members of am ixed-ligand series,a nd the cell parameter change closely follows Vegards law; [14] in the work of Yeung et al.,who studed athree-ligand solid solution, this difference is close to 0.5 . [15] Experimental Section [Yb 2 (BDC) 3 (DMF) 2 ]·H 2 Ow as synthesised under solvothermalc onditions (BDC = benzene-1,4-dicarboxylate and DMF = N,N-dimethylformamide). Ytterbium(III) chloride hexahydrate (1 mmol) and benzene-1,4-dicarboxylic acid (15 mmol) were dissolvedi nD MF (5 mL). To this,H 2 O( 0.15 mL) was added and the mixture stirred until complete dissolution of all reagents had occurred.The reactants were heated in as ealed 20 mL Teflon-lined autoclave at 100 8 8Cf or 20 hours.T he resulting white crystalline solid was isolated by suction filtration. In situ crystallisation studies were carried out on Beamline I12 (JEEP) of the Diamond Light Source. [16] Aspecially constructed reaction cell made from polyetherether ketone (PEEK)was used to investigate solvothermal crystallisation:a5mLinternalvolume tube of 12 mm internald iameter that was fitted with as crew-top lid that allowed moderate pressure to be contained and reactions up to 150 8 8C to be investigated. An internal thermocouple,t hreaded through the lid of the reaction tube allowed continuous monitoring of temperature during reactions.The reaction was stirred rapidly with asmaller Te flon-coatedm agnetic followert oa id heat transfer and to ensure that uniform solid product was present in the X-ray beam throughout the experiment. Thet ube was heated within the ODISC infra-red furnace, [17] with ag lassy carbon sheath around the sample to allow heat transfer to the reaction vessel. Aw avelength of 0.2242 w as used and 2D diffraction patterns collected every minute using aP ixium image plate detector( 430 430 mm 2 )w ith an exposure time of 4000 ms.T he system was calibrated with ac rystalline CeO 2 reference and the 2D image plate data were integratedusing the fit2d software to give 1D diffraction patterns. [18] Thet ime-resolved in situ data sets were analysedusing sequential Pawley decompositions and Rietveld refinements,a si mplementedi nT OPAS. [19] CCDC 1057461 contain the supplementary crystallographic data for this paper.These data can be obtained free of charge from TheC ambridge Crystallographic Data Centre.