Interplay of Chemical and Electronic Structure on the Single-Molecule

Interplay of Chemical and Electronic Structure on the Single-Molecule Level in 2D Polymerization Claudius Morchutt,*,†,§ Jonas Björk,‡ Carola Straßer,† Ulrich Starke,† Rico Gutzler,† and Klaus Kern†,§ †

Max Planck Institute for Solid State Research, Heisenbergstrasse 1, Stuttgart 70569, Germany Department of Physics, Chemistry and Biology (IFM), Linköping University, Linköping 58183, Sweden § Ecole Polytechnique Fédérale de Lausanne, Lausanne 1015, Switzerland ‡

S Supporting Information *

ABSTRACT: Single layers of covalently linked organic materials in the form of two-dimensional (2D) polymers constitute structures complementary to inorganic 2D materials. The electronic properties of 2D polymers may be manipulated through a deliberate choice of the organic precursors. Here we address the changes in electronic structurefrom precursor molecule to oligomerby scanning tunneling spectroscopy and ultraviolet photoelectron spectroscopy. For this purpose, we introduce the polymerization reaction of 1,3,5-tris(4-carboxyphenyl)benzene via decarboxylation on Cu(111), which is thoroughly characterized by scanning tunneling microscopy, X-ray photoelectron spectroscopy, and density functional theory calculations. We present a comprehensive study of a contamination-free on-surface coupling scheme and study how dehydrogenation, decarboxylation, and polymerization affect the electronic structure on the molecular level. KEYWORDS: covalent coupling, 2D polymer, electronic structure, STS, decarboxylation, porous polymer, HOMO/LUMO gap

S

the reactions often are promoted by a metal surface under ultrahigh vacuum (UHV) conditions, the reaction pathways can be fundamentally different.17,19,25 Among a myriad of onsurface reactions,22,26−31 Ullmann-type coupling is one of the most widely studied. Different precursor molecules20,32−41 have been investigated on various metals20,33,42 as well as on decoupling layers43−45 and insulating surfaces.17,46 A downside of Ullmann coupling is the production of halogen atoms remaining on the surface and polluting the reaction template. Even though postdeposition of iodine atoms can decouple the polymer from the metal surface,47 the abstracted halogens can influence the diffusion of surface-stabilized radicals during polymer growth such that long-ranged ordered and defect-free polymer formation is suppressed. It has been shown that halogen atoms can be removed by postexposure of hydrogen.48 However, it is not trivial how this approach could be employed for protocols to form 2DP, as the exposure of H2 during the synthesis would most probably hydrogenate surface-stabilized radicals. Alternative on-surface polymerization protocols free from contaminating byproducts are desirable for 2DP synthesis

ince the isolation of one atom thick monolayer graphene sheets,1 many other inorganic two-dimensional (2D) materials such as MoS22 and phosphorene3 have been the subject of intense research due to their unique electronic structures4 suitable for applications in optoelectronics, catalysis, and supporting membranes.5 The flexibility in the synthesis of 2D polymers (2DP), available through the deliberate combination of precursor molecules and polymerization chemistry, promises a similarly large variety in their electronic structure.6−10 Consequently, semiconductive 2DPs,11,12 halfmetals suitable for the regular distribution of magnetic metal centers,13 and 2DPs as catalysts14 have been theoretically predicted. Despite the abundance of theoretical explorations of 2DPs, only few experimental studies on their electronic structure are available,15,16 confirming HOMO/LUMO gap reduction with increasing oligomer size in 2DPs.9 In contrast, the last years have illustrated several examples of how on-surface synthesis can be used to manufacture singlelayer polymers.6,17−24 In principle, tailor-made structural properties may be obtained, although some issues need to be resolved before reliable control is achieved, such as poor crystallinity and small size. On-surface reactions have been inspired by solution-chemistry polymerization reactions; however, due to the confinement to two dimensions, where © 2016 American Chemical Society

Received: October 30, 2016 Accepted: November 28, 2016 Published: November 28, 2016 11511

DOI: 10.1021/acsnano.6b07314 ACS Nano 2016, 10, 11511−11518

Article

www.acsnano.org

Article

ACS Nano and have been exemplified by homocoupling22 and cyclotrimerization27,49 of terminal alkynes, cyclotrimerization of acetyls,50 as well as reductive coupling of aldehydes51 and carboxylic acids.52 Here we present the fabrication of a 2DP via decarboxylation52 of a tricarboxylic acid on Cu(111), which produces volatile CO2 as byproduct. The reaction pathway is fully illuminated through scanning tunneling microscopy (STM), X-ray photoelectron spectroscopy (XPS), and density functional theory (DFT) calculations. In addition, we study the electronic properties of the monomers (educts in the form of protonated as well as deprotonated 1,3,5-tris(4-carboxyphenyl)benzene, TCPB) and oligomers (products) with scanning tunneling spectroscopy (STS), ultraviolet photoelectron spectroscopy (UPS) measurements, and DFT calculations. Shifts in the unoccupied states in the transition from educts to products are tracked on the singlemolecule level by STS. A detailed analysis shows that an intricate interplay of dehydrogenation, decarboxylation, and polymerization has pronounced effects on the position of the electronic states with respect to the Fermi level.

assembled structure is stabilized by ionic hydrogen bonds between deprotonated and negatively charged oxygen atoms and hydrogen atoms of neighboring molecules.56,57 LEED analyses (Supporting Information (SI), Figure S1) conclude that the molecular islands form a (3√3 × 3√3)R30° superstructure. The high-symmetry axes of the Cu(111) surface are indicated with red lines in Figure 1a. Figure 1b shows the unit cell of the superstructure as well as the atomic lattice of the Cu(111) substrate (blue circles). It is known that oxygen atoms of deprotonated carboxyl groups sitting on top sites maximize the adsorption energy,58 forcing the remaining carbon atoms to reside on top and hollow sites, which is substantiated by DFT calculations of a model compound (SI, Figure S12). Similar self-assembly of TCPB has been observed on Ag(111), for which partial deprotonation of the carboxyl groups was induced by annealing at 147 °C.59 However, XPS data (vide inf ra) confirm the complete deprotonation of the carboxyl groups on Cu(111) at RT. To promote coupling reactions, we annealed the Cu(111) surface with submonolayer coverage of TCPB at 180 °C for 15 min. Figure 2a,b shows STM images of the resulting reaction products. For a bias voltage of −1.5 V, hexagonal structures (lower left) as well as heptagonal and pentagonal structures (upper right) are visible without any substructure (Figure 2a). However, at a bias voltage of −0.5 V, circular protrusions between the linear connections of the molecules become visible (Figure 2b). Geometric considerations suggest that the carboxylate groups of the monomers have split off (in the form of volatile CO2) and that the bright protrusions can be assigned to copper adatoms resting in between two decarboxylated TCPB molecules in an organometallic network. The same organometallic structures have been observed on Cu(111)32,33 and Ag(111),35 where Cu/Ag adatoms likewise appear as circular protrusions. Notably, one covalent bond between two precursor molecules is observed, indicated by an absence of the circular protrusion (white arrow in Figure 2a,b), highlighting the statistical occurrence of covalent-bond formation at the given temperature. The sample was annealed at higher temperatures (220 °C for 15 min) to create additional covalent bonds between the decarboxylated molecules. Figure 2c shows an overview STM image of the reaction products in the form of hexagonally ordered structures. A ball-and-stick model of the corresponding 2DP33 is superimposed and fits well with the observed features (Figure 2d). The pore-to-pore distance of 2.2 nm confirms the covalent nature of the obtained polymers (SI, Figures S2 and S3 for additional linescan analyses). Alongside the ordered oligomers, we observe disordered reaction products (Figure 2c). Figure 2e shows a high-resolution STM image of these structures, in which pores can be observed resembling both in shape and size those of fused polyphenylene rings in porous graphene (pore-to-pore distance of 7.3 Å, SI Figure S4).60 The formation of these structures can be understood by assuming a chemical modification of the original reactive site (where a carboxyl group was originally connected to the molecule) through either C−H activation of phenyls or by tautomerization of hydrogen along the rim of phenyl rings.61 Figure 2f depicts a ball-andstick model in which precursor monomers are colored in alternating gray and red, and on which hydrogen from metapositions on the peripheral phenyls have changed their location to the para-position. For comparison, aryl−aryl coupling via C−H activation of quarterphenyl on Cu(110) takes place at 227 °C,62 and intramolecular cyclodehydrogenation of

RESULTS AND DISCUSSION The precursor molecule TCPB (model in Figure 1a) selfassembles and forms a trigonal structure on Cu(111) upon deposition at room temperature (RT). Figure 1a shows a typical STM topograph of the self-assembly after submonolayer deposition. Taking into account the well-known deprotonation of carboxylic acids on copper surfaces at RT,53−55 we assign one tripodal protrusion to a deprotonated molecule. The self-

Figure 1. Self-assembly of TCPB on Cu(111) after deposition at RT. (a) STM topograph of TCPB (model in lower left corner). Tripodal features can be assigned to deprotonated TCPB molecules. High-symmetry directions of the Cu(111) surface are indicated in red (Ubias = −1 V, I = 0.1 nA). (b) Model of the (3√3 × 3√3)R30° superstructure of TCPB on Cu(111) (blue circles). 11512

DOI: 10.1021/acsnano.6b07314 ACS Nano 2016, 10, 11511−11518

Article

ACS Nano

Figure 2. STM images of organometallic and covalent structures. (a) and (b) show organometallic structures obtained after annealing at 180 °C for 15 min. (a) Ubias = −1.5 V, I = 60 pA and (b) Ubias = −0.5 V, I = 60 pA. Cu adatoms are visible as circular protrusions. The white arrow indicates a covalent bond. (c) Covalent structures obtained after annealing at 220 °C for 15 min (Ubias = −1 V, I = 10 pA). (d) High-resolution STM image of extended porous graphene with a scaled ball-and-stick model superimposed. (e) High-resolution STM image of side products that resemble fused polyphenylene rings (Ubias = −1.5 V, I = 60 pA). (f) Hypothetical ball-and-stick model of polyphenylene structures that forms out of TCPB monomers (shown in gray and red alternatingly).

polyanthrylenes on Cu(111) readily occurs at 200 °C.63 Hydrogen abstraction on the phenyl rings at 220 °C annealing temperatures is thus a viable hypothesis that explains the frequent observation of undesired side products. A statistical analysis of overview STM images gives a yield of approximately 5−10% for the reaction from monomers to hexagonally ordered 2DP (see SI, Figure S11). The hypothesized decarboxylation of TCPB prior to polymerization is evidenced by XPS measurements. Figure 3a shows the carbon (C) 1s region of a (sub)monolayer TCPB deposited onto Cu(111) at RT. The experimental data points are shown as blue circles and the corresponding fit in black. It consists of two Voigt functions (green) whereby one peak lies at high binding energy (287.5 eV) and the other has its maximum at 284.3 eV. The low binding energy peak is assigned to carbon atoms in the phenyls, and the high binding energy peak is related to the carbon atoms in the carboxylate (deprotonated carboxyl groups). This is consistent with calculated core-level shifts of a model compound (see SI, Figure S5) and with XPS studies on other carboxylic acids.53,57,58,64,65 After annealing the sample at 220 °C for 15 min, the high binding energy peak at 287.5 eV disappears (Figure 3b), and the integrated intensity of the low binding energy peak at 284.3 eV is reduced by only 0.5%. This corroborates that carbon atoms of the carboxylate leave the surface and only a negligible amount of molecules desorbs. Figure 3c shows the corresponding oxygen (O) 1s spectra. Only one oxygen peak at 530.8 eV is observable at RT (blue circles), which vanishes after annealing at 220 °C (red circles) and which stems from the oxygen of the carboxylate. For comparison, the O 1s peak of deprotonated terephthalic acid lies at 531.4 eV.53 Multilayers of TCPB were deposited through a nominally six times larger deposition time. The corresponding spectra of multilayer TCPB deposited at RT are shown in

Figure 3. XPS spectra of the C 1s and O 1s core-levels. (a) C 1s peak obtained after deposition of a (sub)monolayer TCPB at RT. (b) C 1s peak after annealing at 220 °C. (c) O 1s peak after deposition of a monolayer TCPB at RT (blue) and after annealing at 220 °C (red). (d) O 1s peak after deposition of a multilayer at RT.

Figure 3d, experimental data points as blue circles and black line corresponding to the fit consisting of four Voigt functions (green lines). The peak with highest binding energy originates from the substrate (background) and is included to improve fitting quality. The high binding energy peak at 533.8 eV corresponds to oxygen in the hydroxyl group, and the peak with binding energy of 532.0 eV stems from oxygen in the carbonyl group consistent with literature.53,57 Both peaks have the same 11513

DOI: 10.1021/acsnano.6b07314 ACS Nano 2016, 10, 11511−11518

Article

ACS Nano

temperatues,66 leads us to adopt this model as the one promoting decarboxylation on the surface. Diffusing adatoms on the Cu(111) surface will likely bind to the carboxylates upon thermal annealing before the decarboxylation becomes energetically feasible, and therefore the population of the initial state of the reaction pathway with Cu adatom is higher by several orders of magnitude compared to the uncoordinated carboxyl. This is substantiated by the experimental observation of coordinated Cu atoms in-between two decarboxylated molecules in the organometallic network (Figure 2b). In the experiment, 180 °C is sufficient to induce decarboxylation of TCPB on Cu(111), which is slightly lower than the predicted 204 °C by theory. We also note that 180 °C is an upper bound for the temperature necessary for decarboxylation; lower temperatures might be sufficient for appreciable decarboxylation rates. The polymerization reaction from carboxylated monomer to decarboxylated oligomer is accompanied by distinct changes in the electronic structure. UPS is used to track changes in the valence band region (occupied states), and STS grants insight into the conduction band region (unoccupied states). The dband signal around 2.3 eV below the Fermi level of the Cu(111) substrate dominates the UPS spectra and precludes the observation of the highest occupied orbitals of the molecules in a monolayer (SI, Figure S6). Therefore, we use multilayer data of TCPB deposited at RT to address the occupied bands in the intact molecule and their shift upon annealing at 220 °C. Figure 5 shows the corresponding UPS

integrated intensity. The low binding energy peak at 530.7 eV can be assigned to oxygen atoms in carboxylates of deprotonated molecules close to the Cu(111) surface and accounts for 6% of the total integrated intensity. Its binding energy is consistent with the O 1s signal of a (sub)monolayer TCPB (see Figure 3c). The disappearance of the monolayer carboxylate oxygen peak together with the vanishing high binding energy C 1s peak after annealing results from the desorption of both species from the surface in form of volatile CO2. The multilayer O 1s data indicate that deprotonation is limited to the layer in proximity of the Cu(111) substrate and absent in higher layers. DFT calculations were performed for a model system to gain further insight into the decarboxylation reaction pathway. Figure 4 shows the initial state (IS), transition state (TS), and

Figure 4. Top and side views of IS, TS, and FS of the decarboxylation reaction of biphenyl-4-carboxylic acid (a) without a Cu adatom (blue) and (c) with a Cu adatom (red). Free energy profiles of the two reactions are shown in (b). Calculations include vibrational enthalpy and entropy at 180 °C. (Gray: carbon, white: hydrogen, red: oxygen, brown: copper).

Figure 5. Ultraviolet photoemission spectra of multilayer TCPB deposited on Cu(111) at RT (blue) and after annealing at 220 °C for 15 min (red). The energies of occupied orbitals shift around 300 meV toward the Fermi energy upon annealing. The gap region (0−2 eV) can be recognized, and the HOMO lies around 2 eV below the Fermi energy.

final state (FS) of the decarboxylation reaction on a bare Cu(111) surface (Figure 4a) and involving a Cu adatom (Figure 4c). The free energy profiles of the decarboxylation reactions are shown in Figure 4b. The calculations include vibrational enthalpy and entropy evaluated at 180 °C in addition to the potential energy. In both cases, the energies are given with respect to the initial state (IS) of the reaction with adatom (red). Furthermore, the reference state of the Cu adatom is taken as an isolated adatom on the surface, such that the energy difference between the IS with and the IS without adatom gives the energy cost of breaking the intact molecule and adatom apart. In both cases the reaction is endothermic with free energy barriers of 1.18 and 1.41 eV without and with a Cu adatom, respectively. Assuming that the reaction is described by the Eyring equation, 138 and 204 °C is required to achieve a reaction rate of 1 min−1 for the model system without and with adatom (see also SI, Figure S8). The overall lower energy in the adatom-carboxylate system, together with the availability of Cu adatoms at the terraces at elevated

spectra. Upon annealing at 220 °C, the occupied states shift toward the Fermi energy by 300 meV. Since XPS data on multilayers of TCPB show that upon annealing molecules in higher lying layers mainly deprotonate (SI, Figures S9 and S10) while molecules close to the Cu(111) surface polymerize, the destabilization of occupied states is a consequence of two effects. The first is the deprotonation, which shifts the occupied states toward the Fermi level, and the second the polymerization of the molecules directly in contact with the surface. The dominating effect for the shift in UPS is, however, the deprotonation since the amount of such molecules is higher compared to polymerization products in the first layer. Therefore, predominantly the abstraction of hydrogen from the carboxyl groups alters the electronic structure of the precursor molecules in the occupied band region. The 11514

DOI: 10.1021/acsnano.6b07314 ACS Nano 2016, 10, 11511−11518

Article

ACS Nano

Figure 6. 2D STS map (b) containing 30 dI/dV spectra recorded across a polymer and a deprotonated TCPB monomer along the cyan line in the STM topograph shown in (a) Ubias = −1.5 V, I = 0.2 nA, T = 20 K. (c) Partial density of states (PDOS) of (i) deprotonated model molecule biphenyl-4-carboxylic acid (blue curve), (ii) decarboxylated molecule connected to a Cu adatom (red curve), and (iii) carboxylate replaced by hydrogen atom (green curve) on Cu(111). The LUMO of the deprotonated molecule at 1.2 eV is significantly reduced in intensity upon removal of the carboxylate and shifts to higher energy when a new covalent bond is formed at the terminal carbon atom. The absence of a significant shift of the HOMO toward the Fermi level in this calculation leads us to the conclusion that neither decarboxylation nor the formation of organometallic oligomers is responsible for the observed experimental shift in UPS (see Figure 5).

on the model molecule biphenyl-4-carboxylic acid are helpful. Figure 6c shows the partial density of states of (i) the intact but deprotonated model (blue line), (ii) the decarboxylated model molecule (red line), and (iii) biphenyl (green line). The calculated HOMO/LUMO gap accounts for approximately 2.7 eV. The most apparent difference upon decarboxylation and formation of an organometallic intermediate is the great reduction of the intensity of the peak situated around 1.2 eV (LUMO) of the intact molecule (i). Qualitatively this agrees very well with the experiment. The exact energy position is likely different since the model molecule differs from the molecule used in the experiments. Nevertheless, the calculations show that the decarboxylation and formation of organometallic oligomers have a large influence on the energetic position of the LUMO but only a minor influence on the HOMO, since the latter does not change significantly upon decarboxylation. Importantly, UPS showed that the abstraction of a hydrogen atom from the carboxylic acid leads to a destabilization of the occupied states. The assignment of peak shifts in UPS and STS in polymerization reactions is not a trivial task and requires detailed insight into both chemical and electronic structure, preferably on the single-molecule level. This is not only true for polymerization through decarboxylation but also for other coupling reactions such as the frequently used Ullmann reaction, which are anticipated to exhibit similar complex changes in the electronic structure from precursor to oligomer.

deprotonated molecules in the multilayer possibly undergo condensation reactions. The unoccupied states were probed on different samples by STS measurements on top of intact molecules and on top of polymerization products. Figure 6b shows a 2D STS map consisting of 30 dI/dV spectra along the cyan line shown in Figure 6a in the topographic STM image. Inside the pore of the polymer, an intact molecule is coadsorbed. Upon polymerization, the unoccupied state observed at +2 V on the singlemolecule level shifts away from the Fermi level. This is counterintuitive to a simple model in the form of HOMO/ LUMO gap reduction upon polymerization,9,67 which stems from a stabilization of the LUMO (shift towards the Fermi level), and which is clearly not sufficient to explain the observation. In particular, the decarboxylation has to be taken into account, as it significantly alters the LUMO. Two effects have to be considered to explain the experimental observations in this system. The first effect is the HOMO/LUMO gap reduction due to the larger π-electron system in the polymerized structures, as mentioned above.9,15,16,67 The second effect is the decarboxylation itself, which not only alters the electronic structure, such that the gap size can be altered, but more importantly destabilizes the LUMO. To compare the electronic properties of monomers with polymerized structures, we need to take both effects into account. To gain further insight into the electronic structure of precursor molecules and reaction products, DFT calculations 11515

DOI: 10.1021/acsnano.6b07314 ACS Nano 2016, 10, 11511−11518

Article

ACS Nano

calculated from the vibrational frequencies at the different states, obtained with the Harmonic approximation.

CONCLUSIONS In summary, 2D covalent network formation by thermally induced decarboxylation on Cu(111) has been presented. We showed with STM and XPS that this reaction is clean in the sense that split-off side products leave the surface without persisting as contaminations. Furthermore, the energy landscape of the decarboxylation reaction on Cu(111) has been investigated by DFT calculations, and the free energy barrier is in good agreement with annealing temperatures used in the experiments. Monomers as well as polymers have been studied on the local scale by STS measurements and DFT calculations. Our data reveal that carboxyl groups in the precursor molecule give rise to empty states that lie closer to the Fermi level than the empty states of polymerized structures, contrasting the LUMO stabilization and shift toward the Fermi level as a result of extending π-conjugation. Moreover, the HOMO shifts toward the Fermi level upon deprotonation. These results provide valuable insight into electronic properties of precursor molecules and their respective polymers which is essential for designing 2DPs with tailored electronic properties.

ASSOCIATED CONTENT S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsnano.6b07314. LEED, additional STM/STS/XPS/UPS data, additional DFT calculations (PDF)

AUTHOR INFORMATION Corresponding Author

*E-mail: [email protected]. ORCID

Claudius Morchutt: 0000-0001-5795-9288 Jonas Björk: 0000-0002-1345-0006 Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS Calculations were performed at the National Supercomputer Centre, Sweden. We are thankful to Christian Dette for help with the STS measurements.

METHODS All STM experiments were carried out in an ultrahigh-vacuum chamber (base pressure