The Role of Metal Adatoms in a Surface‐Assisted Cyclodehydrogenation Reaction on a Gold Surface

Abstract Dehydrogenation reactions are key steps in many metal‐catalyzed chemical processes and in the on‐surface synthesis of atomically precise nanomaterials. The principal role of the metal substrate in these reactions is undisputed, but the role of metal adatoms remains, to a large extent, unanswered, particularly on gold substrates. Here, we discuss their importance by studying the surface‐assisted cyclodehydrogenation on Au(111) as an ideal model case. We choose a polymer theoretically predicted to give one of two cyclization products depending on the presence or absence of gold adatoms. Scanning probe microscopy experiments observe only the product associated with adatoms. We challenge the prevalent understanding of surface‐assisted cyclodehydrogenation, unveiling the catalytic role of adatoms and their effect on regioselectivity. The study adds new perspectives to the understanding of metal catalysis and the design of on‐surface synthesis protocols for novel carbon nanomaterials.


S3. Structural model from STM images
As discussed in the main text, already from high-resolution STM images it is possible to observe discrepancies between the expected ribbon geometry (3) and the one of the actually obtained polymer 4. Figure S3a shows one of such high-resolution STM images where the already mentioned kinked appearance becomes evident. If the two possible appropriately scaled schematic models of 3 and 4 are superimposed, one can clearly see that only the one comprising the formation of five-membered rings (4) fits ( Figure S2b), while the one including six-membered rings (3) cannot reproduce the observed bays ( Figure S2c). Thus, already from STM, we can conclude that the correct model is 4.

S4. Comparison of nc-AFM constant-height frequency-shift and tunneling current images
acquired at different tip-sample heights with theoretical simulations. Figure S4 shows a set of frequency-shift (top panels) and current (bottom panels) nc-AFM images obtained at different tip-sample heights, together with the corresponding theoretical simulations considering both possible products 4 and 3 (5-and 6-membered rings, blue and red borders, respectively).
The images obtained for the largest tip-sample separation are those shown in Figure S4a and b. At this height, we see some alternating bright maxima in the shape of horizontal lines in the frequency-shift image. These bright features indicate the presence of lifted regions within the edge of the structure under investigation as a result of cyclization of the external phenyl ring. According to the orientation of these bright lines, we can infer that the best match with simulations is for structure 4 involving the formation of 5-membered rings (note that although in the case of the 6-membered rings bright lines are also obtained, their orientation is completely different as they are rotated ~55º off the horizontal). When the tip-sample separation is reduced by 100 pm to try to revolve the internal structure of the polymer backbone, one can indeed differentiate some features of the internal structure in the frequency-shift images, like some of the hexagons in the backbone of the polymer ( Figure S4c left). However, due to the strong interaction between the tip and the tilted terminal ring as a result of their close vicinity, we cannot achieve the desired internal resolution. On the other hand, the associated current image ( Figure S4d left), which is not as much affected by the out-of-plane geometry of the structure as the frequency-shift one, already shows the presence of the six-membered rings and one can even discern the formation of extra rings at the expected position of the five-membered ones. In fact, if we compare both experimental images with the simulations, we observe a good agreement with structure 4, while the agreement with the simulation for structure 3 is rather poor. Finally, in order to try to further improve the resolution of the polymer backbone, we have approached the tip by another 50 pm (panels e and f), although risking the integrity of the CO tip. At this separation, we get an improved resolution of the backbone but we cannot properly see either the five-membered rings or the adjacent tilted terminal benzene rings due to the strong repulsion which distorts the frequency-shift image (panel e left). However, one can better see the internal structure of the polymer in the current image (panel f left), confirming the formation of a new ring where the fivemembered rings are expected. Once again, when the experimental data are compared with the corresponding simulations, one clearly observes that a good agreement is only achieved in the case of the non-benzenoid polymer 4. Thus, thanks to these data, we can unambiguously state that the resulting structures correspond to the formation of polymers incorporating new five-membered rings as a consequence of cyclodehydrogenation of external phenyl rings.  Figure S5 shows a length histogram for polymer 4 synthesized from precursor 1. It reveals a broad length distribution from 10 to 40 nm, a small maximum at 50 nm, and a non-negligible number of oligomers with lengths exceeding 50 nm. To demonstrate the selectivity in the formation of polymer 4 against GNR 3, we have also conducted a statistical analysis on the molecular products composing the final nanostructures. Figure S6 shows one of the STM images used in this analysis. As we can see, all the molecular units within the polymers present the same appearance and shape, which we have unambiguously assigned above to the formation of fivemembered rings. In fact, our statistical analysis indicates that 93 % of the molecules have cyclodehydrogenated towards the five-membered ring, while the remaining 7 % corresponds to molecules at kinks and junctions, whose structure cannot be considered as they are the result of a wrong coupling at molecular positions different from the expected ones. Importantly, no molecular units exhibiting the formation of the six-membered rings are observed, in agreement with our model.

S6. Computational details
Reaction pathways were calculated with periodic density functional theory (DFT)-based transition state theory. The periodic DFT calculations were done with the VASP code, 2 employing projector-augmented wave potentials 3,4 and a plane wave basis set expanded to a kinetic energy cutoff of 400 eV. Exchangecorrelation effects were described by the van der Waals density functional (vdW-DF), 5 using the version introduced by Hamada denoted rev-vdWDF2, which has been shown to describe the physisorption on Au(111) accurately. 6 The Au surfaces were modeled by a four layered slab separated by 15 Å of vacuum region. For the CDH reactions, we used a (8 × 13) surface unit cell together with a 3 × 2 k-point sampling for polymer 2, a (7 × 7) surface unit cell together with a 3 × 3 k-point sampling for ortho-terphenyl.
Transition states were calculated using a combination of the climbing-image nudged elastic band (CI-NEB) 7 and the Dimer method. 8 Local minima and transition states were optimized until the residual forces on all atoms, except the bottom two in the Au(111) slab (which were kept fixed), were smaller than 0.01 eV/Å. Free energies for a particular state at temperature and pressure were calculated by where ∆ "#"$ is the electronic enthalpy with respect to the initial state of the reaction (without adatom), %& the number of Au adatoms of the state under consideration and '( the chemical potential of an Au adatom, + number of removed hydrogen atoms and + ! ( , ) the entropy of the hydrogen gas at the given temperature and pressure, calculated as where + ! ( , , ) is the entropy of H * at standard pressure , , for which we used tabulated values, 9 and is Boltzmann's constant. I.e., the free energy includes the energy of dissociatively desorbed H2 molecules formed as a consequence of dehydrogenation. Regarding the chemical potential of the adatom, in the calculations with a single adatom we define it as a single adatom on the surface, i.e., as '( = /0@/0()))) where /0@/0()))) "#"$ is the electronic enthalpy of an isolated adatom on Au(111) and /0()))) "#"$ is the electronic enthalpy of the Au(111) surface. In Figure 4 of the manuscript we show how the activation energy of dehydrogenation is affected by the entropy difference between a free adatom on the surface and an adatom interacting with a molecule (∆ '( ). In this situation, the entropy contribution from this entropy difference is added to the chemical potential. That is, −∆ '( is added to the chemical potential in Eq.

9
(S3), which has the result that for a state involving one adatom the free energy defined by Eq. (S1) is For a particular state, the electron density difference was defined as the difference in electron density between the full system and the combined electron density between two subsystems with their coordinates frozen in the geometry of the full system: the Au adatom and Au(111) surface representing one subsystem and the molecular system the other. In this way, the electron density difference illustrates the rearrangement of charge due to the interaction of the molecular system with the adatom and the surface.

S7. Reaction pathway calculations and complementary calculated data
Here, we provide all the calculated pathways as well as estimates of reaction rates and dependency of dehydrogenation barriers on adatom entropy, which are the basis for the analysis presented in the manuscript.
CDH of polymer 2. For the calculations of the cyclodehydrogenation of polymer 2, we considered an oligomer of three monomers, with the CDH taking place in the central part of such a trimer.
In Figure S7, two different ways of initiating the CDH by C-C coupling on the flat surface are compared.
It is possible to form a five-membered ring by initial C-C coupling followed by dehydrogenation and tautomerization reactions, with an initial barrier of 1.85 eV. We were not able to isolate a pathway that leads to the formation of one six-membered ring by initial C-C coupling, but found a concerted pathway in which two six-membered rings are formed simultaneously. However, the barrier for initiating such a concerted process is relatively high (2.11 eV), and the steps for finalizing this reaction path were not considered further.
In Figures S8-S11, we provide the complete pathways of the reactions compared in Figure 3 in the manuscript, for which the reaction is initiated by dehydrogenation. For all pathways, both the electronic enthalpy at 0 K and the free energy obtained at a temperature of 600 K and a pressure of 10 4)* bar, are provided. In Figure S8 and S9, the pathways for formation of a five-and six-membered ring, respectively, on the atomically flat surface are reported. In Figure S10 and S11 we show the corresponding pathways with the initial dehydrogenation step assisted by an Au adatom.

Activation free energies as function of adatom entropy difference. It is of interest to understand under
which conditions the adatom-assisted dehydrogenation can be anticipated to dominate over the dehydrogenation without adatom. For this, we also estimated the free energy activation energies by taking into account the entropy difference of an adatom moving freely on the surface and the adatom bonded to the molecule, which we denote as ∆ '( , by adding the term −∆ '( to the reference energy of the adatom. I.e., the activation energy of the dehydrogenation with adatom is increased by the factor ∆ '( , while the activation energy without adatom does not depend on this factor. The results are summarized in Figure S14, demonstrating that the activation free energy is lower for the adatom-assisted dehydrogenation up to ∆ '( = 0.72 meV/K. According to the Sackur-Tetrode equation in 2D, 10 such a value of the entropy is obtained for a 2D gas of Au atoms for a concentration of one atom per 4 Å * , which is even higher than the concentration of surface atoms of Au(111). However, the Sackur-Tetrode equation gives an upper limit of the translational entropy of the adatoms diffusing over the surface, valid if they can be regarded as an ideal 2D gas.
Furthermore, in ∆ '( the entropy of a free adatom on the surface needs to be balanced with the entropy of an adatom interacting with the molecular system. To understand all details of the free energy of the system is, however, not conceivable within our computational framework. However, we would like to stress that our conclusion that CDH is driven by adatoms is based on the strong preference of five-membered ring formation, which is only explained by the participation of adatoms.

S8. STM and nc-AFM simulations
We used DFT to obtain the nc-AFM and HRSTM simulations combining the CP2K code 11 and the probe particle code. 12 The simulations were done using the AiiDAlab platform. 13 The microscopy simulations were obtained for the equilibrium geometry of finite (four precursor units) segments of the polymers adsorbed on gold. We used simulation cells consisting of four atomic layers of Au along the [111] direction. A layer of hydrogen atoms was used to passivate one side of the slab to suppress the Au (111) surface state. 40 Å of vacuum was included in the simulation cell to decouple the system from its periodic replicas in the direction perpendicular to the surface. The electronic states were expanded with a TZV2P Gaussian basis set 14 for C and H species and a DZVP basis set for Au species. A cutoff of 600 Ry was used for the plane-wave basis set. Norm-conserving Goedecker-Teter-Hutter pseudopotentials 15 were used to represent the frozen core electrons of the atoms. We used the PBE parameterization for the generalized gradient approximation of the exchange-correlation functional. 16 To account for van der Waals interactions, we used the D3 scheme proposed by Grimme. 17 The gold surface was modeled using a supercell of size 29.48 × 66.38 Å 2 (corresponding to 1140 Au atoms). To obtain the equilibrium geometries, we kept the atomic positions of the bottom two layers of the slab fixed to the ideal bulk positions, and all other atoms were relaxed until forces were lower than 0.005 eV/Å.

General methods
All reactions working with air-or moisture-sensitive compounds were carried out under argon atmosphere using standard Schlenk line techniques. Unless otherwise noted, all starting materials were purchased from commercial sources and used without further purification. All other reagents were used as received. Thin Single crystal diffraction data was collected on a STOE IPDS 2T diffractometer with Mo-Kα radiation.