Fabrication of a High‐Quality, Porous, Surface‐Confined Covalent Organic Framework on a Reactive Metal Surface

Abstract A major goal of heterogeneous catalysis is to optimize catalytic selectivity. Selectivity is often limited by the fact that most heterogeneous catalysts possess sites with a range of reactivities, resulting in the formation of unwanted by‐products. The construction of surface‐confined covalent organic frameworks (sCOFs) on catalytically active surfaces is a desirable strategy, as pores can be tailored to operate as catalytic nanoreactors. Direct modification of reactive surfaces is impractical, because the strong molecule–surface interaction precludes monomer diffusion and formation of extended architectures. Herein, we describe a protocol for the formation of a high‐quality sCOF on a Pd‐rich surface by first fabricating a porous sCOF through Ullmann coupling on a Au‐rich bimetallic surface on Pd(111). Once the sCOF has formed, thermal processing induces a Pd‐rich surface while preserving the integrity of the sCOF architecture, as evidenced by scanning tunneling microscopy and titration of Pd sites through CO adsorption.

Amajor goal of heterogeneous catalysis is to optimizecatalytic selectivity.S electivity is often limited by the fact that most heterogeneous catalysts possess sites with ar ange of reactivities, resultingi nt he formation of unwanted by-products.T he construction of surface-confined covalent organic frameworks (sCOFs)o nc atalytically active surfaces is a desirable strategy, as pores can be tailored to operate as catalytic nanoreactors. Direct modification of reactive surfaces is impractical,b ecause the strong molecule-surface interaction precludesm onomer diffusion and formation of extended architectures. Herein, we describe ap rotocol for the formation of ah igh-quality sCOF on aP d-rich surface by first fabricatingaporouss COF through Ullmann coupling on aA u-rich bimetallic surface on Pd(111). Once the sCOF has formed, thermal processing induces aP drich surface while preserving the integrity of the sCOF architecture, as evidenced by scanningt unneling microscopy and titration of Pd sites through CO adsorption.
Increasingly,s electivity is am ajor consideration in industrialscale catalysis. Heterogeneous catalysts consisting of metal particles dispersed on as upport material with ah igh surface area often possess am ultitude of active catalytic sites. The lack of control over the structure of such sites is often detrimental to selectivity.As uccessful strategyi ne nantioselective heterogeneousc atalysis is to modify am etal surface through the adsorptiono fc hiral molecules, which provide chiral active sites for catalytic reactions. Al imitation of this approachi st he optimizationo ft he surfacec overageo fm odifiersa nd the instability of the modified surfaceu nder reactionc onditions. [1] As trategy to introduce selectivitym ay come from the formation of robust porous covalent architectures that are able to host guest molecules. Such architectures have been successfully constructed from molecular precursors on inert surfaces. [2] However,o nt he more catalytically active metals, the dominant molecule-surface interaction results in molecular precursors being required to overcome high diffusion energy barriers in order to assemble. [3] Hence, thermally activated decomposition is likely to overwhelm the formation of well-defined surface architectures. The first major challenge is the integration of the two concepts:c ontrolled surfacem odification while maintaining surface reactivity.T his paper describes the constructiono f surface-confined covalent organic framework (sCOF) scaffolding on ar eactive metal surface. Subsequent chemical tailoring of the pores may allow for the introduction of suitable functionalities that can direct the selectivity of ar eaction via specific reagent-pore interaction (Scheme 1).
Scanning tunneling microscopy (STM) was used to investigate the deposition of the molecular precursor 1,3,5 tris(4-bromophenyl) benzene (TBPB) on Pd(111)i nu ltrahigh vacuum (UHV), which resultsi nd issociative adsorption with cleavage of the CÀBr bond readily occurring at 300 K( Scheme2). The Yshaped features associated with the activated precursor are clearlyr esolved, alongw ith smaller circular features that tend to pack in hexagonal arrays (Figure1a). The circular features are separated by 5.6 AE 0.7 , and are assigned to Br adatoms derived from CÀBr cleavage, which is to be expected at 300 K given the relatively high activity of Pd. [4] Annealing to (or deposition at) 475 Ki nduces extensive decomposition of the monomer,w ith no evidence of self-assembly or formation of covalent structures( see Figure S1 in the Supporting Information). Similarly,M orchutt et al. deposited TBPB on aN i(111)s urface with comparable results, but they were only able to observe CÀCc oupling after electronic de-Scheme1.Constructiono fasCOF scaffold on ac atalytically active surface describedi nthis work (left), targeting application in enantioselective surface through pore functionalization(right).
[a] C. R. Larrea coupling of the surface by growing amonolayer of graphene. [5] It can be deduced from this behavior that the diffusion energy barriero ft he activated precursor on Pd(111)m ust be sufficiently large, as it cannotb eo vercomee ven by annealing at this temperature. Blunt et al. studied the adsorption of TBPB on Au(111)u nder UHV,a nd demonstrated that the deposition of TBPB with the sample held at 410 Kf acilitates the diffusion of the activated precursors, resulting in ap orouss COF that could be extended over the entire Au surface. [2] We anticipatedt hat the deposition of TBPB on agold-richAu-Pd(111)surfacea lloy under similar conditions should closely emulate the surfacec hemistry on Au(111), and that ap orous sCOF would form. Av aluable feature of the AuPd system is that as olid solutioni sf ormed over the whole composition range, allowing all Au/Pd ratios to be accessedb ys imple annealing.A sw ef ound that the network presentsr emarkable thermal stability on Au(111), we conjectured that palladium enrichment of the surface could be prompted by thermal annealing whilst preserving the integrity of the sCOF.T he deposition of four monolayer equivalents (MLE) of Au on Pd(111)r esulted in al ong-range hexagonal MoirØ pattern with an apparent modulation maximar epeat distance of approximately 7.5 nm (see Figure S2 in the Supporting Information). The MoirØ pattern is ac onsequence of the 4.9 %m ismatch in the lattice parameters of Au and Pd and providese videncet hat the composition of the surface layer is almostp urely Au (at least strongly Au-rich) under these preparation conditions. We found that the deposition of TBPB on the Au-Pd(111)s urface alloy held at 475 Kp roduces an aperiodic porousn etwork (Figure1b). The apparent length of ad imer within the network (13.5 AE 0.8 ), as measured from the centroid of the middle phenyl ring, is in reasonable agreement with two CÀCc oupled monomers. [2] Metal adatom incorporation has often been reported for metastable protopolymers on surfaces, giving longeri ntermoleculars pacings and protrusions in STM images of molecule-adatom-molecule junctions. No evidencew as found for metal adatom incorporation in the presentstudy. [6] By analyzing the topography of one STM image (40 40 nm 2 ), ap ore-size distribution reveals that the sCOF is comprised of about 45 %h exagonal (A = 3.0 AE 0.4 nm 2 )a nd 40 % pentagonal (A = 2.0 AE 0.5 nm 2 )p ores, whereas heptagonal (12 %) or rarely squarep ores (3 %) account for the remainder. The percentage of hexagonal pores is close to that found on Au(111)b yB lunt et al. [2] (50 %) and we can, therefore, conclude that the morphologyo ft he sCOF on the alloy is very similar to the one obtained on the Au(111)s urface.
Along with the framework, islands of periodically arranged close-packed protrusions on top of aM oirØ background were imaged. We interpret these features as bromine adatoms, which are the by-product of the on-surface reaction. The ap-parentBr-Br interspacing distance is 6.8 AE 0.1,a nd the superstructure is rotated approximately 208 from the direction of the MoirØpattern. These dimensions are consistent with acommensurate ( p 7 p 7)R 19.18 superlattice (see Figure S3 in the Supporting Information). This superstructure has previously been observed as ap hase transition from the Au(111)-( p 3 p 3)R 308-Br. [7] The periodicity of the MoirØ pattern is the same as that observed for the as-prepared surfacea lloy and does not arise because of the lattice mismatch between Br adatoms and the surface.
STM revealed that mosto ft he pores appeared to host adsorbeds pecies ( Figure1b, inset). Although we observed monomersa nd dimers trapped in closed and open pores, these were clearly imaged as Y-shaped features. Instead, we believe that Br adatoms populate the pores in the framework. The pore dimensions are large enough to host at least four Br adatoms (for ah exagonal pore and assuming the same packing as in the island). However,i tw as not possible to image individual atoms by contrast to the atomically resolved Br island within www.chemphyschem.org the same image. This could either be ac onsequence of Br diffusion inside the pore occurring faster than the STM imaging time or because of electronic effects such as the confinement of surfaces tates by the pores, which can differ considerably from the clean surfaceand, in turn, affect the imaging. [8] To assess the thermals tability of the sCOF,t he sample was annealedt op rogressively highert emperatures for 10 min before the images were acquired at 300 K. Some pores remained populated even after annealing at 675 K( Figure2a, inset), and Br desorbs completelyf rom Au(111)atthis temperature. [9] However,B rb inds to Pd more strongly than it does to Au, with 675 Kb eing the onset temperature required to desorb Br in the atomicf orm, or as HBr with hydrogens upplied by bulk Pd. [10] This observation already denotes the presence of Pd-like sites at the surface. The network retains its porosity until at least 785 K( Figure2ba nd Figure S4), at which temperature we found no evidence for Br atoms in the sCOF pores. Evidence of pores collapsinga nd complete decomposition of the sCOF was observeda fter annealing at 885 K ( Figure S1). By comparison, Gutzler et al. reported extensive degradation and loss of porosity of the sCOF prepared from TBPB on Cu(111)a fter thermala nnealing at 673 K. [11] The decomposition on the alloy is likely to be ac onsequence of two factors. First, annealing induces further palladium enrichment and emergenceo fa ctive ensembles. Second, the rate of decomposition to surface carbon through dehydrogenation increases at high temperatures. [12] By contrast, we find that, on Au(111), the framework starts to show signs of decomposition only after annealing at 975 K( Figure S1). These resultse mphasize the divergence in the robustness of the sCOF on each surface.
With the purpose of probingt he alloyings tep, the sCOF was exposed to CO (pressure = 110 À6 mbar, t = 10 min) at 300 K, andt he surfacew as analyzed by using reflection absorptioni nfrared spectroscopy (RAIRS)u nder UHV conditions (Figure 3a). For the sample annealed at 675 K, negligible CO uptake is evidenced. After annealing at 775 K, two intense peaks emerge at 1905 and2 021 cm À1 ,w hich are accompanied by two weak features at 2063a nd 2140 cm À1 .A ccordingly,w e assign the 1905 cm À1 band to a u(CO) on Pd bridge sites, and the peak at 2021 cm À1 with the shoulder at 2063 cm À1 to u(CO) on atop Pd sites. [13] The former peak is in very good agreement with the earlier,c losely analogous,C OR AIRS measurements on cleanA u/Pd(111)s urfaces;t he latter band appears remarkablyr edshifted compared to the literature value of 2090 cm À1 on the clean alloy.T his redshift likely arises as ac onsequence of back donation into the 2p*o rbital of CO by the aromatic rings of the sCOF through the surface. [14] Af eaturelesss pectrum below 1900 cm À1 suggests the absence of threefold hollow sites, whichs hould only appear for surface alloys of ac ompositione quivalent to or above Au 30 Pd 70 .A t 775 K, the composition of an unmodifiedA uPd alloy is Au 40 Pd 60 .F or this composition, the diluting effect of Au restricts the number of Pd 2 and Pd 3 clusters, favoring the atop  Pd adsorption sites. [13a] The fact that we observeC Oo nP d bridge sites, which have an intensity in the RAIR spectrumt hat is comparable to that of CO on atop Pd sites, is indicative of palladium enrichment of the surface with respect to clean Au-Pd(111)a tt his pre-annealingt emperature. [15] Interestingly,w e observe aw eak and broad feature emerging at 2140 cm À1 , which is too blueshifted to be assigned to CO on pure Pd sites, but resembles Pd-modified atop Au sites. It is well-known that CO is not stable on Au sites at 300 K, and completed esorptioni se xpecteda t2 55 K. [13a] At entative explanation for this feature could be the entrapment of CO underneath the sCOF.Atoms and small molecules, such as CO, are known to intercalate between graphene and metal surfaces. The intercalation process is facilitated by defects in the film, andh as also been observed for ap orousS iO 2 film growno nR u(0001). [16] We postulate that the pores and defects in our network could provide the diffusionc hannels for CO permeation and intercalation at the interface of the sCOF and the alloy.H owever,t his notion requires furtherinvestigation.
Te mperature programmedd esorption (TPD) spectra show broad traces peaking at T max % 380 K, whicha re consistent with CO desorbing from bridge sites (Figure 3b). [13a] This correlates well with the band observed at 1905 cm À1 in the RAIRS;h owever,w ed on ot observe aw ell-defined peak in the TPD spectra that could indicate CO desorption from atop sites, as opposed to the band detecteda t2 021 cm À1 .T his inconsistency may find ap lausible explanation in the transition of CO from atop to bridge sites upon heating and prior to desorption in the TPD experiments,w hich has been addressed at length by Kuhn et al. [17] Marginal amountso fC Od esorption for the 600-700K alloy were detected (ca. 10 %), which increased to around 20 % upon annealing at 760-780K (Figure 3c). By comparison, this amount is half that of the yield in the clean Au-Pd(111)s ystem ( Figure S5). The lower CO yield is attributed to the blocking of Pd sites by the sCOF,w hich occupies as ignificant fractiono f the surface, and by the co-adsorbed Br,w hich will similarly contribute to this blockinge ffect.
In conclusion, we have demonstrated that ap rotocol to fabricate ap orous sCOF on catalytically relevant surfaces is feasible under UHV,and the sCOF is sufficiently robust to withstand the temperature required to induce alloying. Am arked difference was observed in terms of the thermal stability of the sCOF on the alloy surfacec ompared to ap ure Au surface. RAIR and TPD spectra confirmed that the network grown on the Au-Pd(111)s urfacee xhibits accessible Pd sites with palladium enrichmentr egardingt he clean Au-Pd(111)s ystem. Future work will entail the fabrication of functionalized porous networks through Ullmann CÀCc oupling of pre-functionalized precursors.

Experimental Section
Experiments were conducted in two separate stainless-steel UHV chambers hosting an Ar-ion sputtering gun and annealing facilities for sample cleaning. TPD data were collected in aU HV chamber equipped with aq uadrupole mass spectrometer (SPECTRA, Micro-vision Plus) in direct line-of-sight with the crystal, and aL EED/AES spectrometer (SpectaLEED, Omicron). STM and RAIRS measurements were carried out in asecond chamber equipped with ascanning tunneling microscope (VT SPM, Omicron), an infrared spectrometer (Nicolet Magna), and LEED optics. STM images were recorded in constant-current mode at room temperature by using an electrochemically etched polycrystalline Wt ip. The voltages stated correspond to the sample bias with respect to the tip. Image processing has been applied to the STM data by using WSxM, [18] and ImageJ [19] was used in the pore-size counting. Images are uncorrected for drift. The RAIR spectra presented is the baseline-corrected average over 1000 scans at ar esolution of 8cm À1 .G old evaporation was achieved by exposing the crystal to ar esistively heated Au filament wound around aWwire. TBPB [1,3,5 tris(4-bromophenyl) benzene, Aldrich 97 %] was used as received, and outgassed for 12 ha t3 40 Kp rior to being admitted to the chamber.D osing was achieved by resistively heating ag lass microcapillary wrapped in Ta wire containing TBPB at 435 K. The temperature was monitored by using aK -type thermocouple.