Convergent Fabrication of a Nanoporous Two-Dimensional Carbon Network from an Aldol Condensation on Metal Surfaces
John Landers, Frédéric Chérioux, Maurizio De Santis, Nedjma Bendiab, Simon Lamare, Laurence Magaud, Johann Coraux
CConvergent Fabrication of a Nanoporous Two-DimensionalCarbon Network from an Aldol Condensation on MetalSurfaces
John Landers,
Frédéric Chérioux, Maurizio De Santis,
Nedjma Bendiab,
Simon Lamare, Laurence Magaud,
Johann Coraux
Univ. Grenoble Alpes, Inst NEEL, F-38042 Grenoble, France CNRS, Inst NEEL, F-38042 Grenoble, France Institut FEMTO-ST, Université de Franche-Comté, CNRS, ENSMM, 32 Avenue de l'Observatoire, F-25044 Besanҫon, France
Email:[email protected]@femto-st.fr
Keywords
Nanoporous, covalent organic framework, self-assembled monolayer, scanning tunnelingmicroscopy, density functional theory calculation, flat bands, Dirac cone
Abstract
We report a convergent surface polymerization reaction scheme on Au(111), based on a triplealdol condensation, yielding a carbon-rich, covalent nanoporous two-dimensional network. Thereaction is not self-poisoning and proceeds up to a full surface coverage. The deposited precursormolecules 1,3,5-tri(4’-acetylphenyl) first form supramolecular assemblies that are converted to theporous covalent network upon heating. The formation and structure of the network and of theintermediate steps are studied with scanning tunneling microscopy, Raman spectroscopy anddensity functional theory.
Article's reference: 2D Materials, 2014, 1, 034005 – DOI:10.1088/2053-1583/1/3/034005 ntroduction The family of two-dimensional (2D) materials is a rapidly growing one, which emergedwith the study of graphene. Tailoring the structure of these materials allows engineering theirproperties. Striking examples are the change in the topology of the electronic band structure ofgraphene as a function of the number of layers, the occurrence of electronic resonances ingraphene islands under strong compressive strain, and the confinement in graphene nanoislands, and nanoribbons. Hollow versions of graphene offer new degrees of freedom for structure-engineering of the properties. Examples of such hollow 2D materials are graphynes, graphdiynes, and graphene antidot lattices. The first two consist of low-density carbon atomiclattices with sub-nanometer pores, and have been predicted to exhibit exciting electronic properties,arising from multiple and/or anisotropic Dirac cones in their electronic band structure. The latterhas been much discussed in the literature, as it is predicted to yield spin qubits, to open sizableband-gaps in an otherwise gapless graphene, and to host dispersion-less electronic bands. Experimental realization of graphene antidot lattices has been mostly achieved with the help of top-down approaches relying on lithography performed on plain graphene sheets.
An alternativeapproach would consist of a controlled assembly of well-chosen molecular blocks. Accordingly,covalent nanoporous networks, which are actually graphene antidot lattices with ultimately thinpore walls, have been prepared by interfacial Ullmann coupling reactions at metallic surfaces,consisting of a 2D polymerization of halogenated aromatic monomers. Recently, other 2D reactionschemes yielding nanoporous networks have been discovered, which also have the potential toyield conjugated nanoporous networks with electronic properties like those of graphene antidotlattices. In addition to being an ultimately precise version of the antidot lattice, these systems arealso of great interest for the prospect of nanosieving or nanosensing, enabled by functionalization ofthe pores.An ongoing effort has focused on the development of surface chemistry schemes capable of2elivering nanoporous 2D covalent networks. The synthesis of purely carbon networks has beendemonstrated recently by a few reactions, for instance the above-mentioned Ullmann coupling, homo-coupling of alkynes, and diyne cyclotrimerization. One promising alternative would be toemploy a convergent pathway, by which several molecular monomers react together, promoted bystrong energy gains, to form for instance additional aromatic rings. To date such convergentmethods have not yet been exploited for achieving carbon-rich networks. We report such a reaction,inspired by previous works formerly implemented in solution by one of us, to achieve fully-conjugated hyper-branched dendrimers.
The reaction proceeds by a triple aldol condensationreaction between three acetyl groups, and is demonstrated here on Au(111). Unlike in previousreports of aldol condensation for the polymerization of methyl pyruvate on Pt(111), which onlyyielded disconnected agglomerates, we observe extended 2D covalent nanoporous networks. Thereaction is not self-poisoning and proceeds up to a full surface coverage. The only by-product iswater, which is readily desorbed from the surface under the explored conditions. Using scanningtunneling microscopy (STM), Raman spectroscopy and density functional theory (DFT)calculations, we identify intermediate states in the growth process. The network is stable up to500°C under ultra-high vacuum (UHV) and when exposed to ambient conditions, the carbonbackbone remains intact.
Results and discussion
We have used 1,3,5-tri(4’-acetylphenyl)benzene (see Scheme 1), referred to in the followingas TriAc, as the molecular precursor for the reaction. TriAc molecules were deposited under UHVonto the herringbone-reconstructed (111) surface of Au, between room temperature and 400°C. Themolecules form domains which nucleate between the ridges (partial dislocations ) of theherringbone reconstruction of Au(111), where adsorption is expected to be more favorable (FigureS1 in the Supporting Information). The molecules form compact supramolecular networks, givingrise to spotty electron diffraction patterns (Figure S2 in the Supporting Information), and whose3egree of order and symmetry vary as a function of the surface coverage. Figure 1a shows one ofthese supramolecular networks, obtained uniformly across the sample surface, consisting of ordereddomains typically extending over 50 nm , obtained for a TriAc dose prior to the formation of thesecond molecular layer, and consisting of a dense arrangement of interpenetrated molecules. In thisarrangement, each oxygen atom of a TriAc interacts with one hydrogen atom on the C orthoposition of the lateral benzene ring on a neighboring TriAc molecule (see zoomed-in view of Figure1b). The STM analysis reveals a 0.2 nm distance between these H and O atoms, which is typical ofa hydrogen bond between acetyl and phenyl groups. Other types of H-bond stabilizedsupramolecular networks were found, some of which are seen in the phase diagram of Figure 2 andreported in Figure S3 in the Supporting Information. Noteworthy, the ridges of the Au(111)herringbone reconstruction are still observed in the presence of the supramolecular networks, forinstance seen in the form of a brighter line of TriAcs in the diagonal running from bottom left to topright in Figure 1a. No substantial decrease of the surface coverage is observed upon annealing for afew minutes between room temperature and 200°C, unlike other molecules of similar size (see, e.g.,Ref. 33), i.e. the bonding of the TriAcs with Au(111) is substantial. Actually, ultravioletphotoelectron spectroscopy of oxygen-containing molecules on Au(111) point to a softchemisorption, a likely situation in the present case as well. Depending on the preparation of thetip, STM reveals a non-uniform conductance of the molecules (inset of Figure 1a), pointing tovariations of the electronic density of states (DOS) (Figure S3 and S4 in the SupportingInformation). In the following, unless specified, all experiments consist of TriAc deposition at200°C for a few minutes, a temperature at which highly ordered supramolecular networks still formwithout the TriAc molecules reacting with one another, followed by annealing at highertemperatures. 4 cheme 1: Aldol condensation surface reaction. Upon moderate heating (300°C) two TriAcmolecules react together on Au(111) to form an intermediate dimer while releasing a watermolecule (Step 1). At higher temperature (>300°C), the convergent triple aldol condensationreaction sets in, which yields a new benzene ring (Step 2). At higher temperatures (>400°C), thepolymerization reaction sets in, yielding a 2D nanoporous polymer, prominently composed of four-and six-sided pores, involving at least one new benzene ring each (here three and two extrabenzene rings are formed for six- and four-sided pores).
When longer annealing times are employed at 200°C, consisting of hours instead of a briefperiod of a few minutes, a progressive disorder begins to develop in the supramolecular network.This disorder is associated with the TriAc molecules forming head-to-head dimers, as observed inSTM images (Figure S5 in the Supporting Information). During the period of dimer formation theelectron diffraction patterns progressively become dominated by diffuse scattering (see diffraction5atterns samples annealed between 300-400°C in Figure S6 in the Supporting Information). After 6h at a temperature T = 200°C, the fraction of TriAc molecules having formed dimers is 27±5%. Asexpected, annealing at a higher temperature T = 300°C more efficiently promotes the formation ofdimers: a 10 min annealing already results in 36±6% of the TriAc molecules forming dimers(Figure 1c). By estimating the percentage of TriAc dimers formed at these two temperatures, andassuming a ν exp(- E / k B T ), Arrhenius-like decay ( k B being Boltzmann's constant) with a ν = 10 Hzattempt frequency, an activation barrier E between 1.5 and 1.6 eV is derived.The analysis of the STM data gives a 152±16° angle between the two connected legs of thedimer and a 1.85±0.30 nm distance between the central rings of the TriAc molecules (Figure 1d).The latter value fits well with the distance of 1.92 nm obtained from DFT calculations (Figure S7)for an unsupported stable structure consisting of two TriAc molecules having formed one covalentC=C double bond between the methyl group of the first TriAc molecule and the carbonyl group ofthe second TriAc molecule, releasing a water molecule in the process. This product corresponds tothe first aldol condensation, i. e. the initiation of the polymerization process (Step 1 of Scheme 1).By extending the annealing time, more branched structures are observed (Figure S8 in theSupporting Information). Some of these structures correspond to the products of Step 2 of Scheme1, i.e. to the convergent triple aldol condensation reaction of three TriAcs yielding a new benzenering. The formation of these structures, which are distant one from the other by a few nanometers inthe explored experimental conditions, marks the beginning of the propagation of the 2Dpolymerization reaction. Increasing the annealing and deposition temperatures above 350°C moreefficiently promotes this propagation than increasing the annealing time. Disordered nanoporousnetworks, whose coverage can reach 100% of the Au(111) surface depending on the amount ofTriAcs evaporated, are achieved through the polymerization process (Figure 1e and Figure S9 in theSupporting Information). The pores of these networks are of various kinds (Figure 1f), in terms ofshape, reticulation nodes, and progress of the reaction. The latter can be characterized by the6raction of TriAcs, one acetyl group of which at least has not participated in the aldol condensationreaction. This fraction decreases with increasing surface coverage (Figure 2), from 35-40% for 30%surface coverage (corresponding to network patches of extension of the order of 10 nm) to 15-20%for full-coverage, for similar annealing/deposition temperatures (~400°C). This decrease can beunderstood as a manifestation of the decrease of the amount of the network's edges upon increasingcoverage. The fraction also decreases with annealing and deposition temperatures (Figure 2), forinstance by 10%, down to 15-20%, when increasing the deposition temperature from roomtemperature to 200°C, while with similar high annealing temperatures. Such a decrease is consistentwith the expected temperature-promoted surface diffusion of the TriAcs.The nanoporous network is prominently composed of four- and six-sided pores, usuallycontaining at least one new benzene ring as a reticulation node (Figure 1f-h), each corresponding tothe completion of one triple aldol condensation. In some instances two and three benzene rings perpore are formed, for four- and six-sided pores respectively (Figure 1g,h), which corresponds to fullyreacted pores, shown as the last step of Scheme 1. In other instances we observe reticulation nodeswhich are larger than benzene rings (shown as elongated hexagons in Figure 1f, for instance visibleat the bottom left of the image in the case of four-sided pores). Such pores could actually be ringscomprising eight carbon atoms, which would correspond to a dual aldol condensation between theketone functions of two species of kind Step 1 shown in Scheme 1. The coexistence of pores ofdifferent shapes obviously causes strong distortions in the network, as seen by careful inspection ofFigure 1f. The proportion of four- and six-sided pores varies with the coverage of the network. Atlow coverages (30%, network patches of extension of the order of 10 nm, e.g. point correspondingto 350°C deposition temperature and 450°C temperature in Figure 2), roughly 20% more six-sidedpores are obtained, while for full-coverage cases 20% less six-sided pores are obtained. We ascribethis difference to a reduced mobility of the TriAcs prior to reaction at high coverages, whichhinders the formation of the more energetically favorable (in view of steric and bond-distortion7onsiderations) six-sided pores. We do not observe a substantial change in the proportion of four-and six-sided pores with annealing and deposition temperatures, presumably because when thepolymerization reaction has significantly progressed (boundary between yellow and green domainsin Figure 2), the surface mobility of the reactive species has already strongly decayed.Other types of pores are observed as well, yet in minority, which correspond to situationswhen the 2D polymerization process is ended at Step 1 of Scheme 1 (see Figure S10 in theSupporting Information), presumably due to a limited mobility of the TriAc molecules andoligomers on the surface. Competing reactions, involving the metallic surface (no competingreaction exist in solution), could also hinder the aldol condensation reaction to fully proceed, as wasfound in other systems. Such reaction could promote partial graphitization through the formationof extra C-C bonds. Overall, the network is fully covalently bonded. We find that the number ofTriAc molecules needed to achieve the 2D covalent carbon network is 40 % smaller than in thedense supramolecular network addressed before, implicating a loss of molecules by desorption.8 igure 1:
2D self-Assembly to a covalent carbon nanoporous network on Au(111), observed withSTM. Inset: TriAc molecules exhibiting non uniform conductance after a tip change (a) TriAcmolecules deposited at 200°C, leading to a H-bonded supramolecular network. (b) Zoom-inshowing the arrangement of the supramolecular network with the hydrogen bonding highlighted bythe dashed line. (c) At temperatures between 300 and 400°C, the system becomes disordered andthe monomers begin to form TriAc dimers bonded together by covalent C=C bonds. (d) Zoom-in ofa dimer. (e) 2D covalent nanoporous network. (f) Schematized view of the network in (e): thereacted TriAc molecules are shown as three-legged stars, green hexagon mark the position of extracarbon rings (see text for details), and disks mark the position of reticulation nods corresponding to tep 1 in Scheme 1. (g,h) Zoom-in views with the presumed atomic arrangement superimposed ontothe image, for six- and four-sided pores marked in dark grey in (f). Scale bares measure 2.5 nm. AllSTM topographs were obtained at +1.25 V and 0.1 nA. The 2D covalent carbon network is stable under UHV up to 500°C, a temperature fromwhich the surface coverage decreases, due to desorption of some molecules, which implies thebreaking of covalent bonds, presumably the C-C bonds attached to the remaining O atoms. From550°C, the network undergoes strong modifications. We speculate that these modifications arerelated to those occurring during the thermal reaction between aromatic quinone molecules, coronene, or waste molecules, and yielding graphene. Figure 2 summarizes our observations as afunction of annealing and deposition temperature. Figure 2:
Diagram representing the four kinds of phases (red, yellow, green, and blue sectors ofthe diagram) encountered as a function of deposition and annealing temperatures. Instances ofboth low and high coverage (θ) are shown. The fraction of TriAc molecules having at least oneunreacted acetyl group (f) and the ratio in the number of six- to four-sided pours (r ) are specifiedwhenever relevant.
Annealing time is 10 min. STM topographs are 20 nm-wide and were acquiredat +1.25 V and 0.1 nA.
The partial charge map (Figure S3 in the SupportingInformation) integrated between Fermi level and 0.5 eV above is in good agreement with the STMobservations, although the former is better spatially resolved, as is expected since the calculationdoes not include the effect of the STM tip.
Figure 3:
Top views of the structure of the periodic 2D covalent nanoporous network, inferredfrom DFT, without (a) and in the presence (b) of the Au(111) substrate.
No substantial changes are visible between the STM topographs obtained before and afterexposure to air, even without annealing the sample after introducing it back into UHV (compareFigures 4a and b). The only feature that arises from exposing the sample to air is that the imagingconditions are less stable likely due to molecules from the atmosphere adsorbed on the sample,while the carbon framework remains intact. These observations establish the high stability of the 2D11ovalent carbon network, presumably a carbon backbone saturated with C-H bonds, to whichmolecules from air bond only weakly. This claim is further supported by Raman spectroscopyperformed in air. Figure 5 shows the Raman spectra for a 2D network synthesized on Au(111) andfor the TriAc monomers drop-casted onto Au(111) then allowed to dry. The spectra for the 2Dnetwork displays a peak centered at 1585 cm -1 , and a full-width at half maximum (FWHM) of 30cm -1 . This peak is akin to the characteristic signature of a sp hybridized carbon-conjugated systems,whose width would stem from disorder. In fact, this peak presumably corresponds to the so-calledG vibration mode, which develops in largely extended sp carbon systems. It differs from the modeobserved in the case of the drop-casted monomers, which is centered at 1605 cm -1 and as a 7 cm -1 FWHM. Such features most probably relate to the stretching vibration mode of the carbon-carbonbonds in the monomer. Neither for the drop-casted monomer and the 2D network do we observe thecharacteristic features of the C=O stretching modes, which are expected to show up as peaks atabout 1680 and 1820 cm -1 . Though they are indeed not expected for the reacted network, they areexpected for the drop-casted monomers. In the latter case, their absence is presumably due to thestrong and extended background in the Raman spectrum yielded by the gold substrate. On a quartzsubstrate on the contrary, we do observe a peak at ~1680 cm -1 for the drop-casted monomers;however this substrate lacks the catalytic activity for synthesizing the reacted network. A perfectisolated 2D network is expected to give rise to a G mode with a narrower peak, appearing at higherwavenumber, than the one we observe. We interpret these observations as an effect of strain, whichis known to modify the G peak position and width in a related system, graphene. In this scenario,the G peak of the 2D network actually encloses contributions from regions with varying localstrains, which coexist due to the presence of different pore geometries, to the coalescence ofgrowing domains, and to the mismatch in thermal expansion coefficient between the network andAu(111). As such, the G mode width provides an indirect evidence of covalent bond formation. Inthis light the observed G mode for the reacted network shows strong similarities with that found in12olymers. The inset of Figure 5 shows a Raman map, displaying a signal for the peak centered at1585 cm -1 extending over 80% of the surface, indicating a large extension of the reacted networkacross the gold. These traits of sp carbon found across nearly the entire surface, clearly indicate afully-extended covalent network comprised of carbon. Overall, both molecular resolution imagingand vibrational spectroscopy indicate that the covalent bonds of the 2D networks are mostlyunaffected by exposure to air, as is also the case in other 2D covalent networks. Such stability isdesirable in the prospect of applications outside UHV.The triple aldol condensation reaction, performed with TriAc molecules, is a 2Dpolymerization process which is not directional in essence. Using templated surfaces (e.g. vicinalones) or tailored monomers (e.g. with specific shape and/or reactivity), like was done for otherkinds of surface reactions could allow to achieve one-dimensional polymerized covalentnetworks, owing to lateral hindering of the monomers' mobility and reaction self-steeringrespectively.
Figure 4:
STM topograph of (a) a 2D covalent nanoporous network synthesized on Au(111) underUHV (deposition at 200°C; annealing at 400°C); and (b) of a network prepared under similarconditions, exposed to air, and introduced back in UHV. Images were acquired at +1.25 V and 0.1nA. igure 5: Raman spectrum acquired in air for the 2D covalent nanoporous network on Au(111)(bottom curve) and the monomer drop-casted onto Au(111) (two top curves). The peak at 1590 cm -1 is characteristic of a sp carbon bonded systems. Spectra averaged over two 100 s acquisitions atthe same position with a laser spot size of 1 μm. Inset: 30×30 μm map of the peak intensity at 1590cm -1 for the 2D covalent nanoporous network. Methods
Surface reaction of Triac molecules.
TriAc molecules (melting point 252°C, 99.99%purity) were synthesized according to a procedure described in the Supporting Information. Surfacereactions were performed in a UHV system with 10 -10 mbar base pressure. The molecules wereevaporated with the same flux using two evaporator types, a multi-pocket commercial Kentax oneand a home-made one, in which the quartz crucible containing the molecules is resistively heated, inour case at temperature of 200°C. The deposition rate of the TriAc molecules was calibrated with aquartz microbalance prior to deposition assuming a 1 g/cm molecular density, and the depositionrate was set at 0.5 monolayers/min for a full coverage as obtained in Figure 1a. An aligned andpolished Au(111) crystal with dimensions of 12×3 mm and 0.5 mm thickness was purchased fromSurface Preparation Laboratory. Prior to molecule deposition the Au(111) surface was prepared byrepeating cycles of Ar + sputtering at 0.8 keV and annealing at 900 K until no evidence of carbonwas detected, as verified by in situ Auger electron spectroscopy, and the herringbone reconstruction14as fully developed, as verified by in situ STM. Characterizations.
All STM measurements were performed under UHV and at roomtemperature, in the same system where the samples were prepared, using a commercial Omicronvariable temperature STM/AFM, with electrochemically etched W tips. All images were obtained inconstant current mode, with the tunnel bias applied to the tip. Images obtained for angles anddistances were corrected from drift due to the piezoelectric elements, which was determined bytentatively recording two successive images at the same location. Micro-Raman spectroscopy wasperformed with a commercial Witec Alpha 500 spectrometer setup with a dual axis XY piezo stagein a back-scattering/reflection configuration. Grating used has 1800 lines/mm which confer aspectral resolution of 0.01 cm -1 for 100 s integration time. Laser excitation wavelengths of 532 nm(solid state argon diode) was used. Raman spectra were recorded in air with a Nikon ×100 objective(NA = 0:9) focusing the light on a 320 nm diameter spot (532 nm light), and with a Mitutoyo ×50objective (NA = 0:75). A background, whose intensity decays with increasing wave-number,corresponding to the luminescence from the Au substrate, is observed in Figure 6. Density functional theory calculations.
DFT calculations were performed using the VASPcode, with the projector augmented wave (PAW) approach. The exchange correlation interactionis treated within the general gradient approximation parameterized by Perdew, Burke and Ernzerhof(PBE). When the gold substrate was taken into account, long range dispersion corrections (van derWaals interactions) were accounted for by using Grimme corrections. The dispersion coefficientC6 and van der Waals radius R0 for Au are not listed in the original paper by Grimme and thus havebeen taken from Amft et al. These values for C6 and R0 are 40.62 J×nm/mol and 1.772 Årespectively. The Au(111) surface was modelled using a four-layer slab that was first relaxed. TheTriAc covalent network was then added. Gold atoms were kept fixed in the bottom layers, all otheratoms were allowed to relax. The lateral periodicity was imposed by the smallest – commensurate –supercell chosen to accommodate a TriAc molecule, therefore a 3√3×3√3 periodicity of the15u(111) surface. Note the DFT calculations were performed for unreconstructed Au(111). Theempty space in the direction perpendicular to the surface was chosen equal to 14 Å. Afterconvergence, residual forces were lower than 0.02 eV/Å. Calculations of the isolated TriAc orTriAc dimer were performed in very large supercells to avoid spurious interaction due toperiodicity.
Supporting Information
Synthesis details of the TriAc monomer, additional surface characterization and DFT calculationsare provided in the Supporting Information.
Acknowledgement