Ultrasonic Exfoliation of Hydrophobic and Hydrophilic Metal–Organic Frameworks To Form Nanosheets

Abstract The modular structure of metal–organic framework nanosheets (MONs) provides a convenient route to creating two‐dimensional materials with readily tuneable surface properties. Here, the liquid exfoliation of two closely related layered metal–organic frameworks functionalised with either methoxy‐propyl (1) or pentyl (2) pendent groups intended to bestow either hydrophilic or hydrophobic character to the resulting nanosheets is reported. Exfoliation of the two materials in a range of different solvents highlighted significant differences in their dispersion properties, as well as their molecular and nanoscopic structures. Exchange or loss of solvent was found to occur at the labile axial position of the paddle‐wheel based MONs and DFT calculations indicated that intramolecular coordination by the oxygen of the methoxy‐propyl pendant groups may take place. The nanoscopic dimensions of the MONs were further tuned by varying the exfoliation conditions and through “liquid cascade centrifugation”. Aqueous suspensions of the nanosheets were used as sensors to detect aromatic heterocycles with clear differences in binding behaviour observed and quantified.


Introduction
Metal-organic framework nanosheets (MONs) are free-standing, nominally two-dimensional materialsf ormed by the co-ordination of organic ligandst om etal ions or clusters. [1] Ak ey advantage of MONs over inorganic nanosheetss uch as graphene,b oron nitride and molybdenum disulfide is that their modulars tructurea llows for ready tuning of their properties. This tunability,c ombined with their large external surface area and high aspect ratio, makes MONs ideal for ad iverser ange of applications including separation, [2] sensing, [3] templation, [4] electronics [5] andc atalysis. [6] As with other nanosheets, understanding how to form concentrated suspensions of high aspect ratio nanosheets is an important technological challenge. [7] The modular structure of MONsp otentially provides advantages over simple inorganic nanosheets in allowinge asy modification of surface functionalities to enable nanosheets to be designed for use in particulars olvents. However,t heir porosity,f lexibility,l ability and potential for structuralr earrange-mentsa lso present additional challenges in undertaking this type of study.
Liquid exfoliation provides an attractive, simple and scalable, top-downa pproacht op roducing ultrathin nanosheets from layeredm aterials. [8] In some cases, immersion of layered MOFs in solvent has been shown to result in spontaneous exfoliation of the materialsi nto nanosheets. [9] In most cases however,a dditional energyi sr equired to overcome interlayer interactions in order for exfoliation to occur.Avarietyo fd ifferentm ethods for the liquid exfoliation of MONs have been investigated includingb all milling, [2b, 10] freeze-thaw [11] andi ntercalation, [6c, 12] with sonication [2b, 10, 13] being the most widely employeda pproach. In most cases these processes produce ab road distributiono fp articles izes. Samples are thereforel eft to sediment or centrifuged in order to separate out bulk material from the nanosheets. Top-downa pproaches are particularly attractive for the study of new systemsa st he bulk layered materials are typicallye asier to characterize which aids determining the structure of the nanosheets.
The effect of parameters such as solvent, sonication time and centrifugation time for the liquid exfoliation of other layered materials have been extensively studied ando ptimized. [8a-c] To date, most studies on the liquid exfoliation of MONs have focused on investigatingasingle framework in as ingle solvent. Polar solvents such as acetone and alcohols have most commonly been employed. Peng et al. reported am ixture of methanola nd propanola sb eing optimal for exfoliation of a layeredZ IF. [2b] They hypothesize that the small methanol molecules are able to penetrate into layers whilst propanol adsorbs onto the surface of the nanosheets through its hydrophobic tail helping to stabilize the exfoliated nanosheetsi ns uspen-sion. Junggeburth et al. note that their hydrophobic layered MOF showed decreasing exfoliation in THF > toluene > CHCl 3 . [14] Poor exfoliation was observed when using the polar solvents DMF and H 2 Ow hich was attributed to an inability of the solvents to efficiently penetrate betweent he hydrophobic interlayer space.I nc ontrast, Moorthy andc o-workers investigated exfoliation of al ayered MOF in which there was hydrogen bonding betweent he layers. [15] They found ac orrelation between the Gutmann's hydrogen-bond-accepting parameter of the solventu sed and the intensity of fluorescenceo fn anosheets formed following exfoliation. These studies highlight the different roles that different solvent molecules can play in aiding exfoliation of different layeredM OFs ands tabilizing the resultingn anosheets.
In our work we seek to design new layered MOFs which incorporate features intended to enhance their exfoliation and stabilizet he resulting MONs in suspension. We recently communicated as tudy reporting the liquid exfoliationo f Cu(1)(DMF), al ayered MOF incorporating weakly interacting methoxy-propyl chains designedt oa id exfoliation of the layers into nanosheets. [13g] The nanosheets are based on the popular metal-paddlewheels econdary building unit (SBU) which has a labile, Lewis acidic axial coordination site whichm akes it ideal for aw ide range of sensing, catalytic, electronic, separation and storage applications. [2c, 3b, 6b, c, e] We hypothesized that liquid exfoliation of layered metal-organic frameworks functionalized with either hydrophobic or hydrophilic functionalities would produce nanosheets with different concentrations,s tabilities and thicknesses in different solvents. To investigate this, we compared the liquid exfoliation of the relatively hydrophilic methoxy-propyl functionalized MOF with an isostructural MOF incorporating am ore hydrophobic pentyl-chain in aw ide range of different solvents. We then investigated the molecular and nanoscopic structure of the resulting nanosheets in selected solvents under different conditions in order to understand and optimize the exfoliation process.

Synthesis of layered MOFs
Compounds H 2 (1)a nd H 2 (2)( see Figure 1) were synthesized via Williamsone therification of dimethyl 2,5-dihydroxy-1,4-benze-nedicarboxylic acid with 1-bromo-3-methoxypropane and 1bromopentane, respectively.T he differencei np olarity of the ligands was evident during deprotection of the ligands. Compound H 2 1 was readily obtained from the corresponding methyl ester by heating under reflux in aqueous NaOH solution. [13g] Under the same condition only partial deprotection of 2 occurred due to poor solubility so an alternative method in-volving1 :1 THF/5 %K OH(aq) was employed. [16] Both compounds were achieved in good yields and the purity of the compounds was established by NMR, mass spectrometry,a nd elemental analysis.
Heating of H 2 (1)o rH 2 (2)w ith coppern itrate in DMF in a sealed reaction vial at 110 8Cf or 18 hr esulted in the formation of green microcrystalline powders.A ttempts to grows ingle crystalso ft hese materials were unsuccessful. However,X RPD analysiso ft he microcrystalline powders indicates these structures are isostructural with the single crystal structure that we have previously reported for Zn(1)(DMF). [13g] In this structure four carboxylate linkersare coordinated to the M 2 -paddlewheel (PW) while DMF coordinates to the axial sites of the PWs. Importantly,i nt his form the weakly interacting 3-methoxypropoxy groups or pentylc hains are positioned between the layers whilst there is strongm etal-carboxylate bonding within the layers. Small differences in the unit cell parameters (Table S1 in the Supporting Information) for the copperc omplexesa re ascribed to thed ifferent ligand field effects and differenti onic radii of Zn 2 + and Cu 2 + and to substitution of the oxygen for am ethylene in case of 2.E lemental analysisi sc onsistentw ith the proposed formulasa nd IR and TGA analysis confirms the presence of coordinated DMF in these structures.

Liquid exfoliation
Exfoliation experiments were undertaken using ab ath sonicator.W eu ndertook preliminary experiments investigatingt he effect of different variables on the degree of exfoliationu sing DMF and isopropanol as model solvents. Different powers (320 Wa t3 0% and 100 %), frequencies( 37 kHz, 80 kHz) and temperature of sonication were investigated. It was found that high power produced higherc oncentrationso fm aterial in suspension and high frequency increased concentration and avoided dissolution of the nanosheets ( Figure S4). Sonication was appliedu sing as weep mode and samples werer otated through the bath using an overhead stirrer in order to ensure samples were irradiated evenly.S onication is known to be more effective at lower temperature [17] andt he temperature was maintained overt he course of the experiment using a cooling coil giving at emperature of around1 68C. The set-up for exfoliation is shown in Figure S2 in the Supporting Information.
The following protocolw as therefore established for the exfoliationo ft he MOFs which was used unlesss tated otherwise. The layered MOFs were weighed into glass vials to which solvent was added (5 mg in 6mL) and then exfoliated in as onicator bath at af requency of 80 kHz for 30 minutes at at emperature of < 20 8C. The samples were then centrifuged at 1500 rpm for 10 minutest or emove larger particles and care was taken to avoid redispersion of the sediment duringt ransport. UV/Vis spectra were measured using the top 3mLo fs uspensiona nd highly absorbings amples diluted as required using further solvent.
The solvent that the nanosheets are exfoliated into was expected to have al arge effect on the degree of exfoliation and the stability of the resulting suspension.A ni nitial screen of 23 differents olvents was undertaken. However,s ome solvents had to be excluded due to their UV/Vis cut-off points preventing analysiso rt heir high viscosity resulting in poor dispersion and centrifugation (Table S2 in the SupportingI nformation). A selectiono ft he 11 solvents representing adiverse range of polarities and chemical functionalitiesw ere selectedf or further investigation: water,d imethylsulfoxide (DMSO), N-Methyl-2pyrrolidone( NMP), dimethylacetamide( DMA), dimethylformamide (DMF), acetonitrile (MeCN), isopropanol (IPA), tetrahydrofuran (THF), diethylether(Et 2 O), cyclohexane and hexane.
Both compounds typicallys how as ingle major absorption band the l max of which ranged between 271-303 nm depending on the solventu sed ( Figure S5 in the Supporting Information). Absorption bands were generally broader and lessw ell defined for Cu(1)(DMF), particularly in poorly coordinating solvents such as diethylether,T HF and acetonitrile. In acetonitrile, as econd local maximum was observed at 361 nm and 304 nm for Cu(1)(DMF) and Cu(2)(DMF)r espectively.T he MLCT band was typically too weak and broad to be distinguished so the major peak attributedt ot he dicarboxylatel igand was used in all subsequent analysis. Neither compound was able to form stable dispersions in either cyclohexane or hexane, nominal values of zero are therefore used for these solvents in the subsequent analysis.
The extinction coefficient fort he compounds in each solventw as estimated by dilution of as uspension containing ak nownm ass of each compound. Values ranged from 1892-6693 mol À1 dm 3 cm À1 for Cu(1)(DMF) to 2467-4489 mol À1 dm 3 cm À1 for Cu(2)(DMF). These differencesinspectra are attributed to exchange of the coordinated DMF,v ariations in ligand geometry in the different solvents and differences in particles ize which are discussed in detail later in this article.
Clear differences were observed in the concentration of exfoliatedm ateriali ns uspension following sonication and centrifugation of Cu(1 or 2)(DMF) in different solvents. Figure 2 shows ap lot of the concentration in mm of Cu(1)(DMF) [ blue] or Cu(2)(DMF) [red] suspended in different solvents listed in order of increasing polarity (left to right) as measured by UV/ Vis spectroscopy.D ata shown are the average of four repeats.
At either extreme, the more hydrophilic Cu(1)(DMF) showed as ignificantly higherd egree of dispersion in water than Cu(2)(DMF) whilst the oppositei st rue in diethyl ether where the more hydrophobic Cu(2)(DMF) is presenta ts ignificantly higher concentrations. Higher concentrationsa re observed for Cu(1)(DMF) in all solvents except diethyle ther andD MA. DMSO and NMP give the highest concentrations of both materials and significantly highert han DMA and DMF which have very similar polarities. Samples of both compounds exfoliated into cyclohexane and hexane showedn egligible absorbance following centrifugationw hilst only Cu(2)(DMF) showeda ny absorbance followingexfoliation into toluene.
In studies of other nanosheets formed by liquid exfoliation, aw ider ange of solubility parameters have been put forward as being important for determining the concentrationofe xfoliated material in suspension. [8b, c] We plotted the concentration of material in suspension against ar ange of parameters including polarity,s urface tension and Hansen solubility parameters (Figure 2c-e) as well as Kamlet-Taft, Gutman, Swain, Reichardt's polarity parameters and viscosity (see Section 3.3i nt he Supporting Information). The data in these plots is normalized relative to the highest concentration solvent in order to allow easier visual comparison.
In line with similar studies of other nanomaterials, no single parameter by itself was ar eliable determinant of the concentration of materiall eft in suspension following exfoliation for either material. [8b] In many cases, solvents with similar solubility parameters to the best performing solvents showedl ow concentrationso fd ispersed materials. For example, the concentration of Cu(2)(DMF)e xfoliated in DMA is only 20 %o ft hat in NMP even thought hey have similar surface tensions (g L )3 6.70 and 40.21 mNm À1 respectively.C onversely,w ater and isopropanol have very different polar Hansen solubility parameters (dp), 16 and 6.1 respectively,b ut suspensions of Cu(2)(DMF) with very similar concentrations are formed. It should be highlighted that the fact that exfoliation of the pentyl functionalized MOF produces stable suspensioni nw ater at all, albeit at a lower concentration than the methoxy-propyl functionalized MOF indicate that they are only "relatively" hydrophobic and hydrophilic. It should also be noted that this experiment provides ac omparison of the concentration of materiali ns uspension following exfoliationi nd ifferent solvents, not necessarily the suitability of the solvents to form nanosheets. Ad etailed discussion of the nanoscopic dimensions of the materials produced followinge xfoliation in differents olvents is presented in the section entitledn anoscopica nalysisl ater in the paper. First, the differences in UV/Vis spectrum observed for the materials in different solvents also led us to question the composition of the exfoliated material which we discuss in the following section.

Structural analysis
The relativelyl abile nature of coordination bonds and the high surfacea rea of the nanosheets mean that it cannot be assumed that the MOF structure is unchanged following liquid exfoliation. In particular,t he axial site on the copper paddlewheel is known to be highly labile, allowing for the possibility of loss or exchange of the coordinated DMF molecules with those of the exfoliation solvent. We previously observed differences in the XRPD patterns of Cu(1)(DMF) following exfoliation in different solvents. [13g] Here we undertake am ore detailed study to probe the structure of nanosheets of Cu(1)(DMF)a nd Cu(2)(DMF) following exfoliation in selected solvents (water, DMF,a cetonitrile, NMP and diethylether) representing ar ange of polarities. The as-synthesised MOF (5 mg in 6mLo fs olvent) was sonicated for 12 ha t8 0kHz beforec entrifugation at 1500 rpm for 1hand the resulting sediment collected for analysis by using XRPD,I R, TGA and NMR spectroscopy.
The XRPD pattern for Cu(1)(DMF)f ollowing exfoliation into DMF matches the as-synthesised compound indicating no structural changeo ccurred. In contrast to this, material analysed following exfoliation in water showed ad istinct, new XRPD pattern. For this sample, no nitrogen was observed by elemental analysis while TGA showeda1.4 %m ass loss at 66-94 8C.
Furthermore, the IR pattern showsaloss of the DMF carbonyl peak at 1670 cm À1 and as mall new peak at 3604 cm À1 .A ll these resultsa re consistentw ith substitution of the axial DMF for H 2 O, giving Cu(1)(H 2 O). Material exfoliated in acetonitrile, diethyle ther and NMP all showedc orrelating peaks in their XRPD patterns corresponding to at hird, new phase. In the diethyl ether samples this was accompanied by coincidences with the pattern assigned to Cu(1)(H 2 O) indicating am ixture of the desolvated and hydrated phases. In acetonitrile and diethyl ether,n egligible weightl oss was observed in TGA below the decomposition temperature around3 00 8Ca nd elemental analysis showedn on itrogen was present. Thes ame analysiso n Cu(1)(DMF) exfoliated in NMP shows am ass loss of 4.2 %a t 83-205 8C, and smallq uantities of nitrogen (0.72 wt %) indicating as mall amount of non-coordinateds olventi sp resent.W e suggest this new material (formed in acetonitrile, diethyl ether and NMP) is caused by the loss of axial DMF to give adesolvated phase with the structure Cu(1). This matchesp revious findings following exfoliation in acetonea nd methanol. [13g] Samples of Cu(2)(DMF)e xfoliated into DMFg enerated XRPD data correlating with the pattern produced from the parent MOF.E xfoliation in water produced ap owder pattern corresponding to ad istinct phase. This fact, along with the absence of nitrogen in the elemental analysis and mass loss of 4.6 %a t 23-107 8Cs hown by TGA, is consistent with the formation of Cu(2)(H 2 O). In ad ivergence from the behaviour shown by Cu(1)(DMF), exfoliation of Cu(2)(DMF)i nto diethyl ether,a cetonitrile and NMP gave materials which showedw eakc orrelation in peak positions between the resulting XRPD patterns (Figure 3). Exfoliation into diethyl ether gave rise to ap attern in which each peak could be assigned to either Cu(2)(DMF),o r to the phase assigned to Cu(2)(H 2 O). Elemental analysisc oncluded av alue of 0.59 wt %n itrogen( in comparison to 2.99 wt %c alculated for Cu(2)(DMF), which is consistentw ith incomplete removal of DMF and partial substitution by trace quantities of water.I nc ontrastt ot his, elemental analysis of the sample produced through exfoliation in acetonitrile showedn od etectable nitrogen and TGA showedn om ass loss. We therefore assign this powder pattern as corresponding to that of the desolvated materials. Elemental analysis of the Cu(2)(DMF) exfoliated into NMP indicates significant levelso f nitrogen present (2.21 %) andadecrease in mass at around 105 8Cc onsistent with loss of co-ordinated solvent, on heating the sample. Proton NMR of the digested samples confirmed the presenceo fr esidual DMF,a nd ruled out substitution by NMP.T he two large,b ut poorly resolved peaks around 88 in the powder pattern are consistent with the formation of the sql topologya nd the distortions are presumed to be due to partial desolvation.

DFT modelling
In order to gain further insights into the structure of the different phases, we undertook DFT modelling to visualize the structure of the MONs and confirmt he phase assignments. Structures of 1 and 2 were initially modelled using as ingle PW formed using model monocarboxylate ligandsf unctionalized with only as inglem ethoxy propyl-or pentyl-chain to speed up the calculation (1*a nd 2*). Previous studies by us of PW MOFs have shown that using isolated unit-cells produces very comparable resultst oc alculations performed on extended structures. [18] Coordinatesf rom the known crystal structure of Zn(1)(DMF) were used to generate starting coordinates. The structure was then modified, replacing DMF with water and acetonitrile. The fourth iteration removed any solvent from the axial position. In this final iteration we manipulatedt he arms, so that the ether functionality could conceivably coordinate in the axial position. For 2 the same procedure was followed. The functional used was B3LYP [19] with dispersion-correctionsd ue to Grimme( GD3-BJ). Structures of 1* were subsequently remodelled with both methoxy-propyl chains (1**)r esultingi n slight improvements in the correlation between the calculated and experimental data, but showed no substantive differences. For furtherd etails, please see the supporting information. Figure 4a-c showsi mages of the relaxed structuresf or the three different phases obtained with 1** in whichD MF,w ater and no-solvent are coordinated at the axial position, respectively.S imilar images are shownf or the other derivatives in FigureS54 in the Supporting Information. The corresponding calculated IR patterns for these structures were compared with the experimental patterns ( Figures S55-57). Whilst there are some significant shifts in peak positiona nd intensity between the calculated and experimental patterns particularly in the fingerprint region, the presence or absence of characteristics olvent peaks could be used to assign the phases. In particular, characteristic peaks corresponding to the carbonyl of the coordinated DMF molecules at 1706 cm À1 ando fw ater around 3500 cm À1 were observedi nt he corresponding calculated and experimental patternsf or materiale xfoliated in DMF and water,r espectively.E xperimental patterns for material exfoliated in acetonitrile lacked the calculated peaks for acetonitrile at 2200 cm À1 as well as those for water and DMF and provided closer matches to the calculated structure with no solventc o- ordinated. This data therefore supports the assignments given in the previous section.
Coordinationo ft wo acetonitrile, ethanol, acetone, DMF and water molecules to Cu(1**)h ave binding energies of 58, 66, 72, 87 and 119kJmol À1 ,r espectively,r elative to three infinitely separated molecules. This broadly corroborates what is observed experimentally in that more weakly bound solvents such as acetonitrile are lost whilst more strongly coordinating solvents such as DMFa nd water are retained. However,i t should be noted that these values are based on gas phase calculationsa nd so do not take into account solvent-solvent interactions. This may account for discrepancies such as our previous observation that Cu(1)(DMF) is the observed structure in 10 %D MF in water mixtures.
It is interesting to notet hat in the calculated structures obtained for Cu(1**), methoxy propyl chains on either side of the PW are bent over to allow the lone pair of the oxygen to coordinate intramolecularly to the axial positions of the complex. This is not observed in the structure for Cu(2*)w here the oxygen is replaced with am ethylene group. The binding energy for as ingle arm coordinatingt oC u(1**)( as calculated through the difference between the energies of structures with one coordinated or uncoordinated arm) is 30 kJ mol À1 .I n our calculations coordination of the second arm only has a binding energy of 7kJmol À1 .I ts houldb en oted that these calculationsa re highly dependent on the confirmationa round the paddlew heel and af ull conformational search would be required to provide ab etter estimate of the true value for the intramolecular binding whichi sb eyondt he scope of this study.
We therefore suggest that this ability of the methoxy-propyl chains, but not the propylc hains, to intramolecularly coordinate to this axial position with values comparable to those of some solvent molecules may provide at least ap artial explanation for some of the differences observed between the nanosheets.F or example, the co-ordinated methoxy-propyl chains make the surfaceo ft he Cu(1)s tructures less polar resulting in high concentrationso fn anosheets in apolar solvents than might otherwise be expected. Similarly,t he flexibilityo ft he frameworks might reduce the impact of the hydrophobic pentyl chains in polar solvents. These structural insights high-light the challenges of predicting and understanding the effects of even small changes in molecular structure on the macroscopic properties of the nanosheets.

Nanoscopica nalysis
In addition to understanding the effect of solvento nt he molecular structure of the nanosheets, we sought to examine the influence of solvent on the nanoscopic structure of the resulting material. Exfoliation protocols for other layered materials have varied significantly,w ith sonication times rangingf rom 20 min to several days. Here, we first investigated the exfoliation of the hydrophilic Cu(1)(DMF) and hydrophobic Cu(2)(DMF) in water and diethyl ether,u sing two exfoliation time periods:3 0min and 12 h. It wash ypothesized that longer exfoliation times would lead to thinner nanosheets being pro-  duced, and anticipated that the Cu(1)(DMF)w ould exfoliate better in H 2 Ot han diethyl ether,a nd the reverse true for Cu(2)(DMF). After exfoliation, centrifugationa t1 500 rpm for 10 min removed large, unexfoliated material, and AFM was used to assess nanosheets produced (see Figure5.) Both exfoliation procedures resulted in nanosheets with varying size distributions. In general, more nanosheets with smaller heights were observed from 12 he xfoliation than 30 min, suggesting that longer exposure to ultrasonic waves results in increased exfoliation. For example, Cu(2)(DMF)i nd iethyl ether exfoliated for 30 min and 12 hr esulted in nanosheets with thicknesses of 20-100 and 20-50 nm, respectively.S elected examples of nanosheets observed using AFM can be found in Figure 5, and additional figuresf ound in the Supporting Information ( FiguresS13-19). There are noticeably large agglomerates and sheet-like particles with heights over 100 nm in many of these images, suggesting that 10 min centrifugation at 1500 rpm is not effective at removing all larger particles from the post-sonicationsuspension.
In order to comparet he effect of solvento nt he nanoscopic dimensions of the nanosheets formed, Cu(1)(DMF)a nd Cu(2)(DMF) were exfoliated for 12 hrs in water,D MF,N MP,a cetonitrile and diethyl ether.S amples were centrifuged at 1500 rpm for 1has longer/ faster centrifugation times resulted in insufficient material for analysis in some solvents. Typical AFM images of observed nanosheets can be found in Figures S27-36.
In general, exfoliation in DMF and NMP resulted in nanosheets of lowq uality-lateral dimensions and aspectr atios were low,w ith observedp articles having relativelyl arge heights of > 40 nm. Particles appeared to be rounded in nature,rather than lamellar,particularly in NMP.This could suggest that the energetic input upon prolonged exposure times to ultrasound facilities MON breakdown and dissolution of ligand andC ui nto solution-both H 2 1 and H 2 2 are soluble at these low concentrationsinD MF and NMP.
We investigated the stability of the nanosheets in DMF over 5d ays by UV/Vis spectroscopy and found broadening of the ligand absorption band which was attributed to the formation of an ew peak corresponding to the neutral ligand (Figure S13 a,b in the Supporting Information). In contrast, material exfoliated in water and diethylethers howed no shift in absorbance maximum over time. Furthermore, the intensity of these bands remained constant over 5d ays indicating that stable suspensions had been formed (Figure S13 c).
Nanosheets of Cu(2)(DMF)e xfoliated in water were angular and typically < 1 mml aterally with heights 10-30 nm. Some exampleso fu ltrathin flakes of 5 mm 2nmw ere observed ( Figure 6). Nanosheets of Cu(1)(DMF)e xfoliated in H 2 Ow ere more irregularly shaped and typically 10-40 nm in height, with lateral dimensions up to 1.5 mm, consistentw ith our previous report. [13g] Exfoliation of Cu(1)(DMF)i nd iethyle ther produced low concentrationso fm aterials in suspension and the nanosheets observed have relatively low aspectr atios,t ypically 50-100 nm in height and < 600 nm laterally.I nc ontrast, Cu (2) It is interesting to note the more hydrophobic ligand 2 produced nanosheets with higher aspect ratios and more regular shapes than those of the hydrophilicl igand 1 in both water and diethylether.T his is contrary to our expectation that closer matching of the solvent and nanosheetp roperties would lead to thinner nanosheets. An alternative explanation might be that the thinner nanosheets formed from Cu(2)(DMF) are the result of weaker interlayer interactions between the pentyl chainsc ompared to the methoxy-propyl chains aiding exfolia- tion during sonication. Another factor to consider is that poorer interactions between the nanosheets and solvent may result in more of the thicker nanosheets produced during sonication being removed from suspension during centrifugation. This would mean that on averaget hinnern anosheets are observed when there is am ismatch in solvent and nanosheet properties. Optimisingn anosheet design must therefore balance minimizing inter-layer interactions with complimenting solventp roperties to form stable dispersions of nanosheets and developing centrifugation protocols that ensure removal of larger particles.
In order to investigate nanosheet size control,C u(1)(DMF) was selected as at est system,a nd exfoliated in acetonitrilef or 12 h. Acetonitrile was chosen as we observed good particle separation and minimal agglomeration upon deposition for AFM analysis using this solvent, which enabled more accurate sizing of nanosheets. Liquid cascadec entrifugation (LCC) is a versatile strategy that uses multiple sequential centrifugation steps of increasing rateo rt ime period, using the supernatant of the previous step as the suspension for the next, in order to remove particles of various size from suspension. [20] We employed LCC using steps of 1500 rpm for 1h,4 500 rpm for 30 min then 4500 rpm for 4h.T he particle size distribution of the resulting nanosheets as determined through as tatistical analysis( n = 94-161) can be seen in Figure 6a-c and the mean (x)a nd standard deviation( SD) in particles ize are summarized in Table 1. AFM images used for thesea nalyses can be found in the Supporting Information (Figures S20-22).
The results of the statistical analyses show that the average nanosheet thickness and length of Cu(1)(DMF)d ecrease sequentially from 59 512 nm to 41 307 nm between the first and last steps due to the removal of larger particles. This correlates with ad ecrease in the concentration of materiali ns uspensionf rom 0.33 mm to 0.09 mm.T he smallestn anosheets observed in each case are of as imilars ize at 6-8 nm. The concentrationo fC u(2)(DMF)i ns uspension following the final centrifugation step is lower than for Cu(1)(DMF), however the nanosheets are significantlyt hinner and larger than Cu(1)(DMF) with minimum thicknesses of 4nma nd average di-mensionsof1 9 367 nm following the final step.
DLS data werea lso collected for both systemsa fter each of the three steps of LCC ( Figures S28-S29 in the Supporting Information). The trend observedb yD LS is consistentw ith that observedb yAFM in that LCClowersthe averagep articlediameter of the MONs by reducing the number of larger particles remainingi nt he supernatant. However,t he diameters determined by DLS are consistently lower (Table S9) than those obtained in the AFM analysis. For example, the mean LD for Cu(1)(DMF) exfoliated in acetonitrile for 12 hf ollowed by the three steps of LCC is measured as 106 nm by DLS and 307 nm by AFM. Obtaining accurate particles ize measurements from high aspect ratio nanosheets using DLS is known to be problematic as the Stokes-Einstein equation assumes spherical particles [21] and previous comparisons have also shown DLS produces lower average particle sizes than AFM. [22] Exfoliation by sonication is recognized to be an effective delaminative technique. For MONs, long exfoliation times at low temperatures produce more,t hinner nanosheets. Solvent choicei simportant in determining the thickness and morphology of the nanosheets obtained and avoiding dissolution over time. Small differences in ligand too can have as ignificant impact on the strength of interlayer interactions.C omplimentary solvents may play ar ole at weakening interlayeri nteractions and aiding exfoliation. However,p oor matching of solvent-nanosheet interactions may also resulti nt hinnern anosheetsb eing observed as thicker nanosheets are removed from solution by centrifugation. The wide distributions of particle sizes that result from prolonged exposure of the bulk MOF to ultrasonicw aves can be narrowed through LCC and the average particles ize reduced. Controlling the centrifugation rate enables nanosheet size distribution to be optimized for particulara pplications.I ns omea pplications having an arrow distribution of ultrathin nanosheets will be essential, for others havingabroader distribution of thicker nanosheets at ah igher concentrationc ould be more important.

Sensing
We have previously reported the sensingo ft he small aromatic heterocycle pyridine from aqueous solution, using aqueous suspension of Cu(1)(H 2 O) nanosheets.T itration of pyridine was found to bind to the axial position of the Cu 2 -paddlewheel, with a K a of 30 AE 8 m À1 .W hen this experiment wasr eplicated, insteadu sing Cu(2)(H 2 O), ad rop-off in absorbance at l max was observed, as well as the suspension of nanosheets visibly turning cloudy upon addition of pyridine. This could be attributed Table 1. Statisticsc alculated from nanosheets produced from the 12 he xfoliationo fC u(1)(DMF) and Cu(2)(DMF)i na cetonitrile, and cascade centrifuged.

Sample
Cu (1) to agglomeration of nanosheets upon addition of pyridine, which displaces coordinated H 2 O. This would render the MON surfacei ncreasingly hydrophobic, which may cause agglomeration.
In order to be able to compare the binding strength of Cu(1)a nd Cu(2)M ONs, imidazole was selected as am ore hydrophilic binding substrate to prevent agglomeration. Cu(1)(DMF) and Cu(2)(DMF) weree xfoliated in water for 12 h and centrifuged at 1500 rpm for 1h to give suspensionsw ith concentrations of 0.65 and 0.2 mm respectively.T he samples were diluted with water and aliquots of the guest substrate (73 mm and 43 mm for Cu(1)(H 2 O) and Cu(2)(H 2 O), respectively) in aqueous host suspension (0.13 mm Cu(1)(H 2 O) and 0.08 mm (Cu(2)(H 2 O)) were titratedi nto host suspension and monitored using UV/Vis spectroscopy.A ddition of imidazole in both cases resultedi nb athochromic shifts of l max from 301-297 nm and 42 %a nd 36 %i ncreases, respectively,i nt he absorption intensity ( Figures S48 and S51 in the Supporting Information). These changes are consistent with expected substitution of water molecules for imidazole at the axial positions of the Cu 2 -paddlewheel, which would result in changes to the absorption band of the coordinated dicarboxylate ligands 1 and 2.I ti s most likely that imidazole binds to the Cu atoms through the sp 2 -hybridised Ne lectron pairdonation.
This data was used to calculate binding constants of K a = 1370 AE 180 and 1950 AE 140 m À1 for imidazole to Cu(1)(H 2 O) and Cu(2)(H 2 O) respectively.T he 43 %i ncreaseo fK a observed between Cu(1)a nd Cu(2)i sc onsistent with the hypothesis of the terminal methoxyo xygen of the ligand alkyl-ether arm in 1 being able to bind to the axial Cu sites, as this would provide an extra competing speciesf or substrate coordination in Cu(1)(H 2 O) whichi sn ot present in Cu(2)(H 2 O), which could explain why imidazole binds more strongly to Cu(2)(H 2 O).

Conclusions
MONs are an emerging class of two dimensional materials with significant potential for use in aw ide range of applications thanks to their tuneable structure, high surface area and nanoscopic dimensions. 1e Liquid exfoliation using ultrasound is an appealing route to generating nanosheets from layered MOFs thanks to its broad applicability to different systems, the wide availability of ultrasonic baths and scalability of the approach. However,t here have so far been few studies investigatingt he impact of ligand design,s olvent choice and exfoliation conditions on the moleculara nd nanoscopics tructures of the nanosheets formed and their stability in suspension.
We investigated two layered Cu-PW based MOFs formed using dicarboxylic acid ligands functionalised with either methoxy-propyl or pentyl pendantgroupsintended to bestow hydrophilic and hydrophobic character,r espectively.E xfoliation of Cu(1)(DMF) using an ultrasonic bath produced higher concentrationso fm aterial suspended in water than diethylether whilst the opposite trend was observed for Cu(2)(DMF). Cu(1)(DMF) typically showed higher dispersed concentrations than Cu(2)(DMF)a nd NMP and DMSO gave the highest overall concentrations for both compounds. Exfoliation in aw ide range of other solvents showeds ignificant differences in the degree of exfoliation between the two compounds, however this was not found to correlate with any single solvent parameter.
The lack of simple correlation was partially explained by solid state analysisw hich showed that whilst the two-dimensional connectivity of the layered MOFs is maintained following exfoliation, the presence of al abile axial site on the Cu-PW SBUs mean that the surfacef unctionalization of the nanosheets can vary depending on the exfoliation solvent. This effect is not typicallyo bserved in simple inorganic nanosheets but is likely to be commona mongst MONs with exchangeable metal sites. DFT analysis indicated that the oxygen of the methoxy-propyl ligand 1 is able to coordinate intramolecularly to the axial positionofthe copperpaddlewheels. This may further explain the complex dispersion behaviour of the MONs.
The nanoscopic dimensions of the exfoliated materialw ere investigated using AFM and nanosheets with thickness as low as 2a nd 10 nm were observed. Cu(2)(DMF) typicallyf ormed nanosheets which were thinner,h ad highera spect ratios and were more angular than those of Cu(1)(DMF) in both water,d iethylether and acetonitrile.T his is hypothesized to be the result of the apolarp entyl chains resulting in weaker interlayer interactions than those of the methoxy-propyl chains aiding exfoliation during sonication. However,a sw ith the dispersion study,acomplex balance of sometimes competing factors will determine the profile of the nanosheets generated.L onger exfoliationt imes typically produced higher concentrationso f thinnern anosheetsw hilst liquid cascade centrifugation could be used to removel arger particles and narrowt he size distribution.
The ability of the axial position to exchange solvent molecules andt he photophysical properties of the nanosheets were exploited for use as sensors. Addition of pyridine resulted in aggregation of Cu(2)b ut not Cu(1)w hilst imidazole was shown to bind significantly stronger to Cu(2)t han Cu(1). We note that the weaker binding seen for Cu(1)m ay be in part due to competitionf rom intramolecular binding by the oxygen of the methoxy-propyl chain.
Overall, this study demonstrates the potential of the modular structure of MONs in allowing systematic tuning of their surfacep roperties throughi soreticular substitutions. It also highlightst he subtle interplay between ligand, metal cluster, solvents and exfoliation conditions in determining the molecular,n anoscopic and macroscopic structure and properties of nanosheets. Only by better understanding these structureproperty relationshipsw ill we be ablet oh arness the potential of MONs for use as sensors, catalysts and forp rocessing into composite materialsf or separationa nd electronics applications.
ware with an itrogen overpressure. Solvothermal synthesis of MOFs was undertaken using borosilicate vials with Teflon faced rubber lined caps.
Cu(1)(DMF) was synthesised according to our previous method· [13g] Cu(2)(DMF) was similarly synthesised. Specifically,C u(NO 3 ) 2 .6H 2 O and ligand H 2 1 or H 2 2 were dissolved in DMF and sealed into reaction vials, and heated to 110 8Cf or 18 hrs, then slow-cooled, resulting in a7 7% yield of green, microcrystalline Cu(2)(DMF). Synthetic details and characterisation including elemental analysis, FTIR, TGA and PXRD can be found in the Supporting Information.

Exfoliation
MOF and solvent were added to 10 mL reaction vials in the quantities stated in-text. These were rotated using an adapted Heidolph RZR 2020 overhead stirrer with am ulti sample holder,i naFisher brand Elmasonic P3 0H ultrasonic bath (2.75 L, 380/350 W, UNSPSC 42281 712) filled with water.T he ultrasonic bath was operated at 100 %p ower,a t8 0kHz, and was fitted with ac ooling coil so as to prevent bath heating upon prolonged exfoliation times.
Characterisation NMR spectra were recorded on aB ruker Advance DPX 400 spectrometer. 1 Ha nd 13 Cc hemical shifts are reported in ppm on the d scale and were referenced to the residual solvent peak. All coupling constants are reported in Hz. Mass spectra were collected using an Agilent 6530 QTOFL C-MS in positive ionization mode. Elemental analyses were obtained on an Elementar vario MICRO cube. X-Ray powder diffraction patterns were collected using a Bruker D8 Advance powder diffractometer equipped with ac opper k a source (l = 1.5418 )o perating at 40 kV and 40 mA. The instrument was fitted with an energy-dispersive LYNXEYE detector.I R spectroscopy was performed on aP erkinElmer ATR-FTIR Spectrum 2. Thermogravimetric analyses were collected using aP erkinElmer Pyris 1T GA from 30-600 8Ca t1 0 8Cmin À1 ,u nder a1 0cm 3 min À1 flow of nitrogen. UV/Vis absorption spectra were obtained on a Varian Cary 50 UV or Varian Cary 5000 UV/Vis-NIR spectrophotometer,u sing standard 1cmw idth quartz cells and PerkinElmer Spectrum One software. The nanoscopic morphology of the samples was investigated using aB ruker Multimode 5A FM with an equipped Nokia 10x visualising lens, operating in soft tappingmode using Bruker OTESPA-R3 cantilever.S amples were prepared by dropping 10 mL( sample dependant) of suspension onto af reshly cleaved mica substrate. Images were processed using standard techniques with Gwyddion software. DLS data were collected using aM alvern Zetasizer Nano Series particle size analyser equipped with aH e-Ne laser at 633 nm, operating in backscatter mode (173 8).

DFT modelling
All calculations were performed using Gaussian 09, version D.01. [24] The functional used was B3LYP. [19] For all atoms the 6-311G** basis set was used [25] apart from Cu, for which we used the SDD pseudopotential [SDD].A ll calculations were run with ultrafine integrals ignoring any potential symmetry in the calculations. All optimiza-tions were performed with the standard parameters as implemented in G09. All systems were assumed to be dry,s ot hat no additional solvent field was included. For all optimized structures, frequencies were calculated in the harmonic approximation. In af ew cases as mall (between 0a nd À10 cm À1 )i maginary frequency was found, which was subsequently ignored, following standard practice, since these are usually caused by quadrature errors. For all comparisons between theory and experiment presented below,a scaling factor of 0.973 was used for values below 2000 cm À1 ,w hile for values above 2000 cm À1 as caling factor of 0.95 was used. It is noted that in previous work it was found that using as ingle PW to describe a2 Ds tructure resulted in ar easonable agreement between theory and experiment. [26] The computational part of the Supporting Information was created using in-house developed software based on the OpenEye toolkit.