WorksPhysics and Astronomy
Iron-mediated on-surface synthesis of substrate-decoupled graphdiyne monolayers
Abstract
Graphdiynes are emerging two-dimensional sp-sp carbon materials with electronic structures complementing those of graphene, yet their on-surface synthesis is limited by the persistence of metalated intermediates or yields disordered covalent networks. Here, we report an iron-assisted route to covalent hydrogenated graphdiyne monolayers on Au(111) from 1,3,5-tris(bromoethynyl)benzene. Low-temperature scanning tunnelling microscopy, X-ray photoelectron spectroscopy and density functional theory show that Fe scavenges chemisorbed Br byproducts forming FeBr, in turn promoting the removal of Au adatoms from the organometallic network, thus enabling its metalated-to-covalent conversion under mild thermal treatment. Subsequent annealing removes FeBr and yields covalent, ordered domains weakly coupled to the substrate. Scanning tunnelling spectroscopy, combined with density functional theory, reveals a semiconducting gap of about 1.6 eV associated with carbon p frontier orbitals. This Fe-mediated on-surface synthesis strategy provides a route to atomically precise, weakly substrate-coupled graphdiyne networks and offers a design principle for two-dimensional carbon semiconductors.
Introduction
Graphdiynes (GDYs) constitute an emerging class of 2D carbon materials with sp- hybridization, predicted to show outstanding and tunable electronic and transport properties [3] [40] not accessible by graphene.
On-surface synthesis (OSS) under ultra-high vacuum (UHV) conditions allows producing single-layer GDYs with atomic precision, also giving access to their atomic-scale structure and electronic properties investigation by scanning tunneling microscopy (STM) and atomic force microscopy (AFM). [11] [26, 35, 37, 46, 47, 49, 50, 52]
By selecting specifically designed molecular precursors, the temperature-controlled growth of a variety of graphdiyne-like architectures on metal surfaces can be achieved, including 1D nanowires [8] [12, 19, 23, 38, 41, 53], 2D extended networks [2, 11] [13, 21, 26, 35, 37, 46, 47, 49, 50, 52] and - heterostructures [7].
Ullmann-like dehalogenative homocoupling of halogenated alkynyl precursors exploits a noble metal surface as a catalyst to promote the molecules’ dehalogenation and their subsequent self-assembly through surface metal adatoms, generating a stable metalated carbon network. Subsequent thermal annealing can remove the adatoms from the structure, promoting the transition to the covalent phase via C-C homocoupling [4, 5, 9, 10] [18, 22, 28, 37, 47, 51]. Despite its high-quality fabrication promise, the realisation of fully covalent and ordered GDYs through this method is extremely challenging, since metalated intermediates and reaction byproducts are practically ubiquitous and their removal via thermal annealing is usually accompanied by -carbon cross-linking reactions and disordering of the network [8, 11] [24, 34, 35, 52]. In particular, chemisorbed halogen byproducts do not only hinder the network demetallation [7] [13], but also affect its electronic properties [11] [50].
Here we show that Fe atoms enable the on-surface synthesis of fully covalent, single-layer, ordered hydrogenated graphdiynes from 1,3,5-tris(bromoethynyl)benzene
Results
Fe-mediated on-surface synthesis of covalent h-GDY
Deposition of tBEB on Au(111) at room temperature (RT), followed by a mild annealing (400 K), form ordered metalated h-GDY islands with a honeycomb-like structure (Figure 1b,Figure 1f), where the alternating brighter and fainter spots in LT-STM correspond to gold adatoms and benzene rings [7, 11] [34]. LT-STM images, their 2D fast Fourier transform (2D FFT) and low-energy electron diffraction (LEED) measurements (Figure S1), reveal a (77) superstructure with 0° rotation angle with respect to the underlying Au(111) unit cell, with a periodicity of 2.010.08 nm (see Figure 1f) in agreement with the theoretical model [35].

Figure 1. (a) Sketch of the Fe-mediated OSS process. Above the scheme, the ball-and-stick atomic model of 1,3,5-tris(bromoethynyl)benzene (tBEB) precursor. Schematized steps: tBEB deposition on Au(111) generating the metalated h-GDY; Fe atoms evaporation and bonding with chemisorbed Br atoms resulting in FeBr molecules at room temperature; demetalation of the h-GDY; ordering of the covalent h-GDY and FeBr disappearance. (b-e) Large-scale and (f-i) close-up atomically-resolved LT-STM images of each reaction step schematized in (a). Ball-and-stick atomic models of the metalated and covalent h-GDY are superimposed on (f) and (i), respectively. Blue circles in (g) represent FeBr molecular units. (j-m) LT-STM images of the OSS process upon absence of catalysing Fe atoms, leading to a progressive disruption of the network: (j) Large-scale image of the system at RT, (k-m) Atomically resolved images of the system at RT (k), 470 K (l) and 570 K (m). STM set-point: (b) -500mV/20pA, (c) -80mV/100pA, (d) -2mV/2pA, (e) -1V/5pA, (f) 20mV/180pA, (g) -60mV/100pA, (h) -2mV/10pA, (i) -3mV/30pA, (j) -800mV/50pA, (k) -500mV/50pA, (l) -100mV/100pA, (m) -100mV/250pA.

Figure S1. a) Large-scale LT-STM image of a high-coverage tBEB/Au(111) sample. The brighter network represents the as-deposited metalated h-GDY, while the darker network represents the Br reconstruction. b) Atomically resolved close-up of the Br reconstruction outside the metalated h-GDY on Au(111). c) Large-scale LT-STM image of a low-coverage tBEB/Au(111) sample, with isolated portions of the Br reconstruction visible on Au(111). d) 2D FFT of (c). e) Atomically resolved close-up of the metalated h-GDY, with a pore containing 4 Br atoms arranged according to the reconstruction. The red rhombus indicates the unit cell of the superlattice of the metalated h-GDY. The turquoise rhombus indicates the unit cell of the overlayer of Br atoms. f) LEED diffraction pattern of the metalated h-GDY on Au(111) (setpoint: 42 eV) for a high-coverage sample ( ML). The yellow hexagon corresponds to Au(111) first-order LEED spots, while the red hexagon to the metalated h-GDY. g) Simulation of the metalated h-GDY commensurate LEED pattern through the superlattice matrix ; the first-order spots of Au(111) are marked in yellow. h) Atomically-resolved hexagonal structure of FeBr₂ on Au(111). i) 2D FFT of (h). White circles highlight the spots used to extract the FeBr₂ periodicity. l) Atomically resolved LT-STM image of the covalent h-GDY. m) 2D FFT of (l). The spots corresponding to the covalent h-GDY are highlighted by the blue dashed hexagon. STM set-point: a) -1V/5pA, b) -1V/5pA, c) -500mV/20pA, e) -50mV/10pA, (h)-3mV/30pA, (l) -4mV/100pA.
Br atoms, resulting from tBEB dehalogenation, are found both outside and inside the pores of the metalated h-GDY hexagonal network (Figure 1b,Figure 1f). On the surface, they form an extended porous network with local clusters arranged following the ()R30° reconstruction (Figure S1) [8] [44, 54]. In the network, they interact with the Au adatoms, stabilizing the system and preventing the network demetalation [7], as discussed in the following. Therefore, a beneficial effect is expected upon their removal. After evaporating a minute quantity (0.05 ML) of Fe atoms on the surface, kept at RT, Br atoms are no longer observed in STM, while bright agglomerates form at the edges of the metalated h-GDYs (Figure 1c,Figure 1g). Close-up STM images, complemented by 2D-FFT analysis (Figure S1,Figure S1), allow us to identify those features as a FeBr network, showing a hexagonal lattice with a period of 0.370.02 nm, in reasonable agreement with the value reported in the literature [17, 45, 48]. The Fe-Br chemical bonding due to the formation of FeBr is also confirmed by X-ray Photoelectron Spectroscopy (XPS) (Figure S2).

Figure S2. Temperature-dependent XPS spectra of: (a) Fe . The spectra are deconvoluted into components corresponding to different chemical environments of Fe. The red component is assigned to Fe species adsorbed on the Au(111) surface. Upon annealing at 385 K, it can be observed also the presence of purple components, attributed to Fe and Br atoms bonded in the FeBr2 structure. (b) Br 3d core levels. The red line corresponds to bromine adsorbed on the Au surface at RT, the purple line to bromine bonded to iron at 385 K, and the green line to negligible residual bromine traces on the surface at 573 K. Dots correspond to the experimental data, while the solid lines to the fitted components.
In addition to the extended FeBr network, we observe some smaller features in the pores (blue circles in Figure 1g), which can be identified as single molecular FeBr units, as discussed below.
A mild annealing at 420 K for 25 minutes promotes the complete release of Au adatoms from the network, as evidenced by the disappearance of the bright protrusions associated with metal adatoms in LT-STM images of the network (Figure 1h), and also confirmed by XPS measurements showing the complete disappearance of the C-Au component in the C 1s core level spectrum (see Figure S3).

Figure S3. (a) Ball-and-stick atomic models of the metalated (bottom) and covalent (top) h-GDY dimer. The different colors identify distinct carbon chemical environments: red corresponds to C–C/C–H carbon atoms, purple to C–C carbon atoms bonded to the acetylenic group, and green to C atoms bonded to Au. Au adatom is highlighted in yellow. (b) Fitted C1s XPS spectra after Fe deposition and annealing at 385 K (bottom) and 420 K (top).
The transition from the metalated to the covalent phase occurs concomitantly with the partial desorption of FeBr (Figure 1d,Figure 1h) whose residual phase on the surface is constituted by two coexisting structures, i.e., the FeBr 1T polytype with and without vacancies (see Figure 1h). The presence of two phases, already reported on Au(111) [48], may indicate an excess of Br atoms or a different local stacking with respect to the underlying substrate. STM images of different regions of the system (Figure S4) unveil the presence of both ordered regions of the carbon network, where the honeycomb-like structure is preserved, and others, where the network is locally deformed into pentagonal, heptagonal and octagonal rings.

Figure S4. Large-scale and close-up LT-STM images of covalent h-GDY islands after the annealing at 420 K. Bright areas correspond to FeBr₂. STM set-point: (a) -2mV/2pA, (b-c) -2mV/10pA, (d)-2mV/2pA, (e)-2mV/140pA, (f-h) -2mV/2pA.
A further annealing step at 610 K for 25 minutes completely removes FeBr (Figure 1e,Figure 1i and Figure S5). Surprisingly, this is accompanied by the overall reordering of the covalent h-GDY (Figure 1e,Figure 1i and Figure S5), leading to a hexagonal honeycomb-like network with an average periodicity of nm (see Figure 1i and S1). This result is quite unexpected and non-trivial from a chemical standpoint since a large-scale reordering is typically unlikely, as it implies the selective breaking and reformation of C–C bonds.

Figure S5. Large-scale LT-STM images of covalent h-GDY islands after the annealing at 610 K. STM set-point: (a-e) -1V/5pA, (f-g) -5mV/100pA, (h) 100mV/50pA.
Its mechanistic explanation is beyond the scope of the present work. However, recent works on Fe-induced selective C–C bond cleavage reactions in molecular systems [16, 25, 27, 29, 33] suggest that transient Fe-mediated bond activation could facilitate network reorganization during the metalated-to-covalent conversion. Moreover, the Co-catalyzed on-surface polymerization of functionalized GDYs has been recently demonstrated, employing coronene as a template for the GDY ordering [42]. In our system, at about 570 K, almost all Br atoms desorb from the surface, while non-negligible metallic Fe traces are still detected (Figure S2). These residual Fe species may contribute to C–C bond reorganization during annealing at 610 K, leading to the formation of ordered islands without the aid of any templating structure.
Compared to the typical thermally activated Ullmann-like coupling (Figure 1j-Figure 1m), our Fe-mediated OSS approach (Figure 1a-Figure 1i) enables the complete network demetalation at a lower temperature, also ensuring a higher thermal stability of the fully covalent system. Without Fe, annealing at 470 K (Figure 1l) already results in the disruption of the network before achieving its complete demetalation, and further annealing at 570 K (Figure 1m) results in a completely amorphous carbon system, where the original structure can no longer be recognized.
Covalent conversion mechanism via Fe-mediated Br scavenging
We here unravel the mechanism underlying the Fe-mediated metalated-to-covalent transition, correlating the evolution of the carbon network to that of Br-derived structures identified on the surface during the annealing.
The tBEB dehalogenation involves the release of three Br atoms per molecule, leading to a total of six Br atoms per unit cell containing two molecules. At low coverage, an average of 2.3 Br atoms per pore is observed, preferentially located near Au adatoms within the network (see Figure 2a). This is confirmed by our theoretical calculations, indicating that a single Br preferentially adsorbs at an FCC site of the underlying Au(111) surface, and is stabilized by proximity to a gold adatom (see Figure S7). Moreover, Br adatoms tend to distribute over different pores (see Figure S6), although in some of them they cluster into an incipient reconstruction (Figure S1, cyan rhombus, and corresponding model in S6a).

Figure 2. Atomically-resolved LT-STM image of: (a) metalated h-GDY before Fe atoms evaporation. Green dashed arrow identifies a pair of chemisorbed Br atoms arranged according to the reconstruction, whose height profile is reported as inset below; (b) h-GDY after Fe evaporation and mild annealing at 385 K. Blue dashed arrow identifies a bright-faint protrusion pair, whose height profile is reported as inset below; distances have been extracted from line profile measurements performed over different high-resolution LT-STM images of the metalated h-GDY. (c) a demetalated hexagon of the h-GDY after annealing at 385 K. White circle exemplifies the Au-stabilized unit. (d) nanocluster confined into a pore of the covalent h-GDY. (e-h) STM simulations and corresponding relaxed structural models in top view (i,k,m,o) and side view (j,l,n,p) of the experimental images shown in the first row (a-d). STM setpoint: (a)20mV/180pA, (b)-2V/160pA, (c)-2V/160pA, (d) -2.5mV/30pA. STM constant current simulations: integration range [ -0.5,], isosurface states/.

Figure S6. Structural model of a Br cluster in the arrangement (a), linear (b) and “2+” distributed (c) configurations of four Br atoms in a pore of the metalated h-GDY structure.

Figure S7. Different positioning of Au adatom (orange) with respect to two Br atoms (brown) and relative energies.
At this stage, Br pairs inside the metalated h-GDY pores show a uniform contrast, with experimental interatomic distance (0.48 ± 0.02 nm, Figure 2a) fitting the lateral spacing between second nearest neighbours FCC adsorption sites of gold. This value is well reproduced by the structural model and simulated STM images (Figure 2e,Figure 2i), with a small discrepancy arising from the calculated Au lattice constant (4.21 Å) relative to the experimental value (4.08 Å). Notably, the metalated h-GDY is corrugated due to the relevant interaction between Au adatoms and the surface (Figure 1j).
Following Fe deposition and annealing at 385 K, the paired Br show an asymmetric contrast and a reduced interatomic distance (0.38 ± 0.01 nm, Figure 2b), indicating that the Br atoms no longer follow the Au(111) registry.
Simulations show that both observations are consistent with the formation of a molecular unit stabilized by interaction with an Au adatom at the edge of the pores (Figure 2k,Figure 2l and S8, S9).
In particular, the contrast asymmetry originates from the different adsorption heights of the two Br atoms. They are both bound to a deeper Fe adatom located at 2 Å from the surface (see Figure 2l); one Br interacting with the substrate Au atoms and located at 2.9 Å from the surface, and the other displaced upward to 3.5 Å upon formation, interacting with the Au adatom embedded in the metalated h-GDY (Figure 2l).
This interaction stabilizes the pore-edge configuration (Figure S10,Figure S10) and promotes the Au stripping from the metalated network with an energy gain of 0.8 eV (Figure S10,Figure S10), being the reason for the overall conversion of the h-GDY to the covalent phase.

Figure S10. FeBr molecule in different positions in the pore (a,b) and with the removal of an Au atom (c,d). Energy difference for the two locations of the molecule and for the two adatom removal processes is reported.
Upon demetalation, the Au adatom remains coordinated with the unit within the pore (Figure 2m). The simulated STM image (Figure 2g) reproduces the elongated feature observed experimentally (Figure 2c), where the three spots correspond to the two Br atoms and the stripped Au adatom.
Further support for this assignment is provided by the triangular features occasionally observed within the pores (Figure 2d), well reproduced by simulated STM (Figure 2h). These structures are consistent with a cluster of three +Au units arranged as shown in Figure 2o. Similar complexes have been proposed for low-coverage on Au surfaces [20], where Au adatoms were found to stabilize the assembly and contribute to the STM pattern.
The combined STM and DFT analysis highlights the dual role of Br in the Fe-mediated OSS process. In the absence of Fe, chemisorbed Br atoms stabilize Au adatoms and hinder the conversion to the covalent phase [7, 11] [35]. Upon Fe incorporation, Br form species that coordinate Au adatoms, stripping them from the network, thereby promoting the formation of fully covalent h-GDY.
Structural and electronic properties of the covalent h-GDY network
The ordered and clean nature of the covalent h-GDY produced by Fe-mediated OSS enables a detailed characterization of its structural and electronic properties. The conversion to the covalent phase weakens the Au-adatom-mediated coupling to the substrate, resulting in marked differences from the metalated network.
Upon conversion from metalated to covalent phase, LT-STM images reveal a reduction of the unit-cell periodicity (Figure 1f,Figure 1i), described by a simulated unit cell rotated by , requiring only strain to achieve commensurability with Au(111) under periodic boundary conditions [35]. Consistently with previous calculations [35], Au-adatom removal weakens the substrate interaction, resulting in a flatter network and a larger separation from the surface (Figure 2n,Figure 2p).
Further evidence of the progressive decoupling from the substrate is provided by the recovery of the Au(111) herringbone reconstruction, which is largely lifted by the formation of the metalated h-GDY (Figure 1b) [6, 8]. The herringbone remains lifted after Fe deposition (Figure 1c) and Au-adatom removal at 420 K (Figure 1d), due to the presence of . Complete recovery is observed only after annealing at 610 K (Figure 1e), upon removal of , providing direct evidence of the weak coupling between covalent h-GDY and Au(111).
The average lateral size of the islands increases from nm at 420 K to nm at 610 K (Figure S4, Figure S5), accompanied by the emergence of branched structures. We attribute this behaviour to the enhanced mobility of the decoupled covalent domains, which can merge into larger domains.
An additional crucial indication of the different substrate-coupling in metalated and covalent h-GDYs is the Au(111) surface state (SS). On pristine gold, it is located at eV [1]; in the metalated h-GDY it shifts up to eV, also in agreement with previous reports [11] [24, 50] (Figure S11,Figure S12), while in the covalent h-GDY it is observed much closer to its nominal energy (i.e. eV, Figure 3e and Figure S11), indicating a substantially weaker electronic interaction with the substrate.

Figure 3. (a) Atomically resolved LT-STM image of the STS grid area acquired with CO-functionalized tip. (b) Conductance spectra acquired at the colour-coded dot positions in (a): grey, central pore position; violet, average spectrum marked by the violet dots in (a), with two prominent spectral features at -2.1 V and +2.2 V. (c) Constant-current map acquired with CO-functionalized tip at -2.1 V. STS/STM set-point for (a-c): -0.55 V/40 pA, with lock-in amplitude 20 mV and frequency 817 Hz. (d) Atomically resolved LT-STM image of the STS grid area acquired with metallic tip. (e) Conductance spectra acquired at the colour-coded dot positions of (d): grey, central pore position; green, average of the STS point spectra acquired on the carbon network, specifically on the marked phenyl rings (dark green dots) and acetylenic links (light green dots), both characterized by two prominent spectral features at -1.45 V and +0.5 V. (f) Constant-current map acquired with metallic tip at -1.45 V. STM/STS set-point: -0.55 V/50 pA, with lock-in amplitude 20 mV and frequency 817 Hz. (g) and (i) simulated maps at -1.8 eV and -1.45 eV, respectively, considering the height modulation introduced by the feedback setpoint. (h) Orbital projected density of states (PDOS) of carbon atoms in h-GDY.

Figure S11. Conductance spectra of the Au(111) surface state in the middle of the pore of the metalated (M, orange) and covalent (C, purple) h-GDY network. Dashed vertical lines correspond to the energy position of the surface state, i.e. -0.35 V and +0.4 for the covalent and metalated system, respectively.

Figure S12. a) STS and dI/dV maps of covalent h-GDY on Au(111) in comparison with the STS of the metalated h-GDY (b). Purple vertical lines mark the bias at which dI/dV maps show the network. Green vertical lines mark the bias at which dI/dV maps shows a network. c) orbital resolved PDOS of covalent freestanding h-GDY; d) and e) orbital-resolved PDOS of covalent and metalated h-GDY on Au(111), respectively. Vertical purple and green lines mark the position of and states that can be related to the ones reported in the experimental panels a) and b).
This substantial decoupling of the covalent h-GDY from Au(111) allows us to precisely determine its electronic properties, closely matching the simulated freestanding system.
We performed conductance spectroscopy at 4.8 K by acquiring constant-current grids over the two nm regions shown in Figure 3a,Figure 3d, within the energy range from to V. The spectra reported in Figure 3 were acquired at the cove sites using a CO-functionalized tip and on top of the network using a metallic tip (complementary measurements are shown in Figure S13). The two tip terminations enhance different orbital contributions: metallic tips emphasize -derived features, whereas a CO-functionalized tip introduces also -wave components [32] (see Figure S13).

Figure S13. (a) Atomically-resolved LT-STM image of the STS grid area acquired with CO-functionalized tip. (b) LT-STS curves corresponding to the equivalently colored dots in (a). The grey dashed line corresponds to the spectrum acquired in the position of the grey dot in (a), i.e. the Au(111) surface state confined by the h-GDY pore. The green line corresponds to the average spectrum acquired in the position of the green dots in (a), i.e. the apexes and the arms of the hexagons forming the h-GDY, with the -1.45 V and +0.5 V spectral features highlighted by green arrows. (c) Constant-current dI/dV map acquired with CO-functionalized tip at -1.45 V. STS/STM set-point: -0.55 V/40 pA, with lock-in amplitude 20 mV and frequency 817 Hz. (d) Atomically-resolved LT-STM image of the STS grid area acquired with metallic tip. (e) LT-STS curves corresponding to the equivalently colored dots in (d). The grey dashed line corresponds to the spectrum acquired in the position of the grey dot in (d), i.e. the Au(111) surface state confined by the h-GDY pore. The violet line corresponds to the average spectrum acquired in the position of the violet dots in (d), i.e. the H-terminated edges of the phenyl rings of the h-GDY, with the -2.1 V and +2.2 V spectral features highlighted by violet arrows. (f) Constant-current dI/dV map acquired with metallic tip at -2.1 V. STS/STM set-point: -0.55 V/50 pA, with lock-in amplitude 20 mV and frequency 817 Hz.
Point spectra acquired at the cove sites with a CO-functionalized tip (violet dots in Figure 3a), together with maps acquired in constant current mode, reveal two prominent features at V and V (violet curve in Figure 3b; see also Figure S13). The corresponding conductance maps are nearly identical (Figure 3c and Figure S14) and reveal maximum local density of states (LDOS) in correspondence of the H-terminated vertices of the phenyl rings.

Figure S14. (a) Constant-current map acquired with CO-functionalized tip at +2.2 V. STS set-point: -0.55 V/40 pA, with lock-in amplitude 20 mV and frequency 817 Hz. (b) Constant-current map acquired with metallic tip at +0.5 V. STS set-point: -0.55 V/50 pA, with lock-in amplitude 20 mV and frequency 817 Hz.
Point spectra acquired on the h-GDY backbone with a metallic tip (green dots in Figure 3d), together with constant-current maps, reveal additional features at -1.45 V (with a shoulder at -1.6 V) and +0.5 V (green curve in Figure 3e; see also Figure S13). These features become more evident when compared to spectra acquired at the pore centre (grey curve in Figure 3e). The corresponding differential conductivity maps at -1.45 V and +0.5 V are nearly identical (Figure 3f and Figure S14) and display maximum LDOS on the h-GDY backbone. The relative intensity of the spectral features depends on the setpoint used during grid acquisition, which determines the tip-sample distance. As discussed in the Supporting Information (“Insight into LT-STS measurements”), the tip approaches more closely the H-terminated sites than the carbon backbone, thereby modulating the conductance contrast.
To assign the experimentally observed STS features to specific atomic orbitals, we compare STS spectra with the orbital-projected density of states of covalent h-GDY (Figure 3e). The experimental features at -1.45 eV and 0.5 eV match well with the calculated states in the PDOS (green line in Figure 3h), whereas the experimental peaks at -2.1 eV and +2.2 eV (Figure 3b) are assigned to states, after a minor rescaling of the unoccupied states to account for the known limitations of DFT.
This assignment is further supported by calculated LDOS maps at -1.8 eV (Figure 3g) and -1.45 eV (Figure 3i), obtained at fixed bias and constant tip-sample distance (see Methods). In agreement with the experimental maps, the -1.8 eV state is suppressed on the network, whereas the -1.45 eV state exhibits enhanced intensity on the carbon backbone, confirming its character.
Overall, covalent h-GDY exhibits a semiconducting character, despite a small residual density of states around the Fermi level arising from weak hybridization between carbon orbitals and Au s,p states. The experimental HOMO-LUMO gap, estimated from the onset of the frontier-state features in Figure 3e (dashed lines), is ∼1.6 eV. This value is consistent with the gap obtained from the calculated PDOS (∼1.2 eV; Figure 3h), as well as with hybrid-functional calculations (∼1.45 eV; Figure S15). Our estimated gap is smaller than the value previously reported for the same network on Au(111) [24]. We attribute this difference to the assignment of the frontier orbitals: our combined STS and DFT analysis identifies the -derived states closer to the Fermi level as the relevant HOMO and LUMO states.

Figure S15. Comparison between the PDOS of carbon for covalent h-GDY on Au(111) obtained with GGA and Hybrid C09 functional.
Comparison between STS of covalent and metalated network (Figure S12,Figure S12) and related PDOS (Figure S12,Figure S12) evidences an overall reduction of the states at the Fermi level upon metalated-to-covalent transition. The covalent phase shows a more pronounced semiconducting character, the gap of the in-plane states is restored and the hybridization between the h-GDY orbitals and the substrate is reduced resembling the freestanding case (Figure S12). Due to an overall charge transfer [35] both bias-dependent maps and PDOS calculations reveal a shift of carbon-derived states towards lower energies, more marked in the covalent network.
The differences between covalent and metalated h-GDY demonstrate that the Fe-driven, Br-mediated OSS process provides effective control over the structural and electronic properties of the network. In addition, the resulting all-carbon h-GDY remains stable up to 610 K, substantially exceeding the thermal stability of the metalated phase and enabling improved structural ordering.
The combination of the network order, weak substrate coupling, and finite band gap establishes covalent 2D h-GDYs as a promising platform for future all-carbon electronic and optoelectronic devices.
Conclusions
This work establishes a Fe-mediated on-surface synthesis strategy for obtaining covalent, single-layer hydrogenated graphdiyne on Au(111). By combining low-temperature scanning tunnelling microscopy, X-ray photoelectron spectroscopy and density functional theory, we identify a Br-scavenging mechanism in which Fe forms FeBr species that promote the removal of Au adatoms from the organometallic network, enabling its conversion into a covalent framework under mild annealing. Removal of FeBr upon further annealing restores structural order and yields h-GDY domains that are weakly coupled to the Au(111) substrate, allowing the frontier carbon states to be resolved, revealing a semiconducting gap of about 1.6 eV. More broadly, these results show that controlled halogen scavenging can direct demetalation and covalent bond formation in on-surface synthesis, providing a design principle for atomically precise two-dimensional carbon semiconductors.
Methods
Experiment
All the experiments are carried out under ultra-high vacuum conditions, with a base pressure below mbar. Single-crystal Au(111) samples (MaTeck GmbH and SPL) are used for the on-surface synthesis, following several cleaning cycles of Ar sputtering and annealing at 770 K. 1,3,5-tris(bromoethynyl)benzene (tBEB) molecular precursor is loaded in powder form in a Knudsen cell and evaporated for 15-20 seconds over Au(111), keeping the crucible at 300 K and the substrate at room temperature, reaching a chamber pressure of mbar. Upon surface-catalysed halogen cleavage, tBEB molecules undergo coupling reaction through the incorporation of gold adatoms from the substrate, generating a metalated h-GDY [8, 11] [37, 39, 50].
To improve the order of the metalated h-GDY, a mild annealing at about 400 K is performed. Fe atoms are evaporated on the obtained system through an e-beam evaporator equipped with an integral flux monitor. The sample is maintained at room temperature while Fe atoms are deposited on top, setting the emission current to 13.7 mA, the ion flux to 0.1 nA and the deposition time to 80 seconds. A constant pressure of mbar during Fe evaporation ensures the perfect cleanliness of the deposit, without contamination from the metallic rod. After Fe deposition, a sequential annealing at 420 K and 610 K results in the generation of a completely covalent two-dimensional h-GDY network. Low temperature (LT) scanning tunneling microscopy and spectroscopy (LT-STM/STS) investigations were conducted at the Laboratorio de Microscopías Avanzadas of the Universidad de Zaragoza, with a Scienta Omicron q-plus microscope, cooled down to 4.8 K, using a W tip, which for some measurements was functionalized with CO. LT-STS spectra grids and differential conductivity maps at constant height were obtained through an internal lock-in amplifier with an oscillation frequency of Hz and a modulation amplitude of mV. XPS measurements were performed in situ at room temperature (RT) in UHV conditions, using a VG Scienta XM 650 X-ray source. The emitted radiation was monochromatized with a VG Scienta XM 780 monochromator optimized for Al K radiation (1486.7 eV). Photoelectrons were collected in normal emission geometry and analyzed using a Scienta SES 100 electron analyzer integrated into the STM preparation chamber.
STM data were analyzed using the image processing Gwyddion 2.63 software [30]. STS data were analyzed using Igor Pro and SpectraFox. XPS spectra were fitted using IGOR Pro 6.37. Prior to fitting, the background contribution of the Br 3d region including the Au 5p shoulder and Fe 2p was subtracted. The binding energy scale was calibrated by aligning the position of the Au 4f peaks. The metallic Fe 2p core-level spectra were fitted using a Doniach-Sunjić pair function, which accurately describes the asymmetric line shape of metallic photoemission peaks.
Theory
Theoretical calculations are based on Density Functional theory as implemented in the SIESTA code [36], i.e. by exploiting a pseudopotential description of electron-ion interaction and atomic orbitals basis set. A mesh cutoff of 400 Ry and a GGA-PBE [31] exchange and correlation functional have been adopted. The van der Waals correction to simulate the interaction between the covalent network and the Au substrate was included through Grimme dispersion forces [15]. The supercell was built according to the experimental observation, i.e. with the covalent network forming an angle of 9° with respect to the [112] direction. The covalent network has been adapted to the Au(111) supercell built with the theoretical equilibrium lattice constant of 0.297 nm, resulting in a negligible strain of 0.7% of the covalent system with respect to the freestanding case [35]. The covalent network and the first layer of the surface have been relaxed until forces reached the convergence criteria of . A mesh of k-points has been used for self consistency and a five times larger mesh for the PDOS. Calculations with HSE06 hybrid functional have also been performed using the relaxed structure obtained with GGA-PBE functional. LDOS maps were obtained through calculations that exploit plane wave basis set, as implemented in Quantum Espresso. [14] STM images are generated in the Tersoff-Hamann [43] approach at constant distance or constant current, by integrating states in an interval . The STS maps were obtained by integrating the LDOS in a small interval of 0.1 eV at the fixed bias. To reproduce the effect of the experimental feedback, the maps are taken at a fixed distance , where the lateral dependence of the tip height is set by realizing a constant current topography (current ), performing an energy integration of the electronic density, from the peak position to the Fermi level.
Data availability
The data supporting the findings of this study are available in the Article and its Supplementary Information. Additional data are available from the corresponding authors on request.
Acknowledgments
A.C., A.L.B. and C.S.C acknowledge funding by: Funder: project funded under the National Recovery and Resilience Plan (NRRP), Mission 4 Component 2 Investment 1.3 - Call for tender No. 341 of 15.03.2022 of Ministero dell’Università e della Ricerca (MUR); funded by the European Union NextGenerationEU - Award Number: project code PE0000021, Concession Decree No. 1561 of 11.10.2022 adopted by Ministero dell’Università e della Ricerca (MUR), CUP D43C22003090001, Project title “Network 4 Energy Sustainable Transition NEST”’. F.S. acknowledges the financial support from the University of Padova through the grant P-DiSC#02BIRD2024-UNIPD VS-PpyTM. J.L.C. acknowledges financial support from the Spanish Ministry of Science and Innovation (Grant PID2022-138750NB-C21 and “Severo Ochoa” Programme for Centres of Excellence in R&D CEX2023-001286-S funded by MCIN/AEI/10.13039/501100011033 and by ERDF “A way of making Europe”), from the regional Governments of Aragon (E12_23R) and from the European Commission (project ULTIMATE-I, grant ID 101007825). J.L.C. further acknowledge the use of Servicio General de Apoyo a la Investigación-SAI and the Laboratorio de Microscopías Avanzadas of the Universidad de Zaragoza. S.A. acknowledges the CINECA award under the ISCRA initiative, for the availability of high-performance computing resources and support, and the financial support from INFN (Progetto iniziativa specifica “TIME2QUEST”). A.O.B. acknowledges support from the Ministero dell’Università e della Ricerca (MUR) and the University of Pavia through the program “Dipartimenti di Eccellenza” 2023–2027.
Author contributions
A.C. conceived and conducted the STM experiments, analyzed the corresponding data, and wrote the initial draft. S.A. conceived the theoretical modelling, performed the DFT calculations, analyzed the corresponding data and wrote the initial draft. A.C. and G.C. performed preliminary RT-STM measurements at University of Padova. A.C., E.P.S. and J.L.C. conducted LT-STM/STS experiments. J.L.C. contributed to the LT-STS data analysis and interpretation. A.O.B. synthesised the tBEB molecular precursor. J.L.C., F.S., F.T., A.L.B. contributed to the supervision and discussion of results. C.S.C. conceptualized, coordinated and supervised the work. A.C. and S.A. contributed equally to this work. All authors contributed to the revision and final discussion of the manuscript.
Competing interests
The authors declare no competing interests.
Supporting Information: Iron-mediated on-surface synthesis of substrate-decoupled graphdiyne monolayers
Alice Cartoceti; Simona Achilli; Gianni Conti; Eliecer Pelaez-Sifonte; Alessio Orbelli Biroli; Sabine Maier; Francesco Sedona; Paolo D’Agosta; Francesco Tumino; Andrea Li Bassi; Jorge Lobo-Checa; Carlo S. Casari
Br overlayer and h-GDY orientation on Au(111)
On Au(111), chemisorbed Br atoms deriving from the dehalogenation of tBEB molecules arrange according to reconstruction, rotated by with respect to the surface [8, 10]. In Figure S1 the porous network formed by Br atoms for a high-coverage tBEB sample is reported, displaying locally agglomerates with the arrangement (close-up in Figure S1).
Figure S1 shows the 2D FFT of the system reported in Figure S1, from which the periodicity of the metalated system can be extracted, perfectly matching the one extracted from line profile measurements over atomically-resolved LT-STM images of the metalated system, i.e. 2 nm. LEED measurements allow to determine the rotation angle of the metalated h-GDY with respect to Au(111) high symmetry directions: indeed, from the LEED pattern shown in Figure S1, a supercell for the metalated h-GDY over the Au(111) lattice (of constant nm) can be estimated, from which the OMF periodicity is calculated at 2.016 nm. In particular, the yellow dashed hexagon corresponds to the first-order spots of the Au(111) lattice, and the red one corresponds to the metalated h-GDY. The LEED pattern reported in Figure S1 is simulated accordingly, through the commensurate superlattice matrix
and closely reproduces the experimental result.
2D FFT images were also employed to extract the periodicity of the extended FeBr regions, i.e., as exemplified in Figure S1-Figure S1. Finally, the 2D FFT was employed to determine the periodicity of the covalent h-GDY (Figure S1-Figure S1), resulting nm, in agreement with the one extracted from line profile measurements over atomically-resolved LT-STM images.
XPS - C 1s, Br 3d and Fe 2p
XPS analyses were performed following Fe deposition on tBEB/Au(111) samples to investigate the catalytic role of iron in the Ullmann-like coupling reaction.
The Fe spectrum after annealing at 385 K, at which the transition to the covalent phase is already ongoing, shows a shift toward higher binding energy compared to metallic Fe (centered at 707 eV [1, 5]), with a main peak at approximately 709.5 eV and a satellite feature around 715 eV. The fitting was performed following the literature for submonolayer FeBr on Au(111) [4], where the main peak at 709.5 eV is assigned to Fe species, while the satellite at 715 eV is associated with final-state effects related to the multiplet structure of the 2p core level in transition metals.
In the Br 3d region, the Br component, initially centered at 68 eV for bromine adsorbed on Au [3, 2], shifts to 69 eV upon Fe deposition, consistent with Br bound in Fe-Br coordination environments. This shift is in line with literature reports for FeBr on Au(111) [4]. Temperature-dependent XPS measurements reveal that, upon annealing at 570 K, the Br 3d signal is strongly suppressed and nearly disappears. At the same time, the Fe 2p signal shows a marked reduction of the FeBr-associated component, indicating that the Fe–Br interaction is no longer stable at elevated temperature. The simultaneous disappearance of bromine and the modification of the iron spectral features suggest that, at about 570 K, all bromine desorbs from the surface, while iron is no longer bound in a stable FeBr phase. A fraction of Fe may remain on the surface as metallic adatoms or residual FeBr, while another fraction may diffuse into subsurface sites of the Au(111) substrate, as reported in the literature [6].
The evolution of the C 1s spectra as a function of temperature reveals significant changes both in the binding energy position and in the overall peak shape (Figure S3), indicating a progressive modification of the chemical environment of carbon atoms during the reaction. After deposition of iron and annealing at 385 K, the C 1s spectrum is characterized by a multicomponent structure, including a distinct low-binding-energy (BE) contribution at 283.48 eV (highlighted in green in Figure S3), assigned to the C carbon atoms, involved in metalated C–Au bonds. This assignment is consistent with previous reports [11, 9, 2], where such low-BE features are associated with carbon atoms coordinated to Au adatoms on the substrate. Upon annealing at 420 K, a clear spectral evolution is observed. The low-BE component attributed to C–Au bonding disappears, leading to a shift of the overall C 1s envelope toward higher binding energies. This shift reflects the loss of gold–carbon interaction and the transition to a purely covalent carbon network.
Fe-mediated OSS: 420 K annealing
Figure S4 shows LT-STM images of the system after annealing at 420 K.
Fe-mediated OSS: 610 K annealing
Figure S5 shows LT-STM images of the system after annealing at 610 K.
Br adatoms stability
We performed DFT calculations for the metalated h-GDY with one adsorbed Br atom in different positions with respect to the underlying substrate, i.e. TOP, FCC, HCP. The FCC adsorption site is the most stable with a total energy gain of 63 meV and 68 meV with respect to HCP and TOP adsorption, respectively. We also estimated the binding energy of Br adatoms, as:
where is the energy of metalated h-GDY without adatoms, is the number of Br atoms adsorbed and is the reference energy of a Br atom in the gas phase. The adsorption energy of a single Br atom is eV, while the subsequent adsorption of a second and third adatom costs 19 meV and 77 meV, respectively. For a number
of Br atoms in the pore larger than three, two possible arrangements were considered: the like arrangement (see Figure S5) and a “linear” configuration with all the Br atoms along the edge of the pore (Figure S6). The calculation for 4 Br adatoms predicts a larger stability for the former, with an energy gain of 54 meV. It is worth noting that, anyway, the alternative configuration with four Br atoms distributed in adjacent pores, i.e. “2+2” in Figure S6, is 170 meV more stable than the , demonstrating the preference of Br to distribute in different pores. This energy difference is calculated as:
This tendency is confirmed when the number of Br is even larger: for six Br adatoms, the energy gain of distributing in different pores (e.g “2+4” or “3+3”) is in the order of 0.5 eV. We also verified that Au adatom stabilizes Br adatoms at 5 Å distance. In particular, two Br adatoms are stabilized by the proximity of an Au adatom when they form an equilateral triangle with all the adatoms in FCC positions (Figure S7).
If the mid bondlength point is instead displaced near the adatom or shifted laterally, as in Figure S7,Figure S7, the total energy increases by 32 meV and 38 meV, respectively. Positioning the Au adatom far apart increases the energy even more: in the configuration shown in Figure S7, the system is 96 meV less stable.
Fe-Br interaction
We evaluate the strength of the Br-Fe interaction, mediated by gold, through different model calculations. We considered the case in which a Fe adatom is far apart from a Br adatom, i.e. in the center of the pore, while the Br atom is near the edge of the OMF (Figure S8). We compare this situation with the case in which the Fe adatom stands near the Br (Figure S8). We verified that the latter is 1.1 eV more stable than the former, leading to the formation of a Fe-Br dimer and confirming the preference of Br to bind to Fe with respect to substrate Au. Moreover, the FeBr dimer is stabilized by the proximity to a gold adatom, gaining 147 meV with respect to being located in the center of the pore, in the same adsorption configuration with respect to the substrate (Figure S8). It is worth noting that when Br is bound to Fe, its distance from the substrate surface increases with respect to the single Br adatom, being farther than the Fe adatom in the dimer.

Figure S8. Different models for Br and Fe adatoms: not interacting (a), dimer near the h-GDY edge (b) and dimer far from the h-GDY edge (c).
We should note here that, apart from simple test purposes, the FeBr dimer would be quite unusual, with under-saturated Fe. For this reason, and in agreement with what was observed in the experiments, we repeated the calculations above considering 2 Br atoms and 1 Fe. First of all, we determined the preferential distance between the two species, as reported in Figure S9. Differently from Br-Au, Br and Fe tend to form a molecular complex with two Br atoms, with bond lengths of Å. The binding energy of the molecule is eV.

Figure S9. Different positioning of a Fe adatom (green) with respect to two Br atoms (brown) and relative energies (c).
We also explore the energy landscape when the molecule is moved in different positions in the pore. Placing the molecule near the edge of the h-GDY implies an energy gain of 0.53 eV (Figure S10,Figure S10). We also simulated a situation in which an Au adatom of one of the h-GDY edges is removed in the presence of a FeBr molecule near the edge (Figure S10) or far away (Figure S10). We found that the process is 0.8 eV favoured in the former case.
STS and dI/dV maps of metalated h-GDY on Au(111)
The electronic properties of metalated and covalent h-GDY have been compared by means of STS and dI/dV measurements supported by the theory. The main features are:
The shift of the spectral features toward higher energy in the metalated h-GDY compared to the covalent h-GDY. These features have been identified on the basis of the dI/dV maps.
The reduction of the DOS intensity at the Fermi level passing from metalated to covalent h-GDY testifying a metal-to-semiconductor transition.
The shift of the Au Shockley state, as discussed in the main text.
Insight into LT-STS measurements
The p-type tip orbitals contributing to the tunneling current due to the presence of the CO-functional group allow for enhancing in-plane orbital features, while the mainly s-type orbital of the metallic STM tip enhances the visualization of the out-of-plane orbital features [7]. This explains why the dI/dV maps acquired with the CO-terminated tip allow for the better visualization of the orbitals at the H-terminated edges of the phenyl rings at -2.1 V and +2.2 V (predicted to be corresponding to in-plane carbon orbitals and H s orbitals, see the section “Electronic properties”), while the dI/dV maps acquired with the metallic tip allow for the better visualization of the orbitals along the covalent h-GDY network at -1.45 V and +0.5 V (predicted to be corresponding to out-of-plane carbon orbitals, see the section "Electronic properties”), as reported in Figure 3. Figure S13-Figure S13 shows that the dI/dV maps acquired with the CO-terminated tip at -1.45 V and +0.5 V allow as well for the visualization of C orbitals, even if with a lower resolution with respect to the metallic tip. In the same way, Figure S13-Figure S13 shows that the dI/dV maps acquired with the metallic tip at -2.1 V and +2.2 V allow as well for the visualization of C and H s orbitals, even if with a lower resolution with respect to the CO-terminated tip.
Interestingly, with both the CO-terminated and metallic STM tips, the acquired spectra show a much larger intensity at the energies corresponding to C and H s orbitals with respect to the energies corresponding to C orbitals. We relate this behaviour to the tip-sample distance during the spectral acquisition. Indeed, being a constant current mode measurement, the chosen setpoint that affects the tip-sample distance set as reference on the carbon network before starting the grid is the bottleneck for the type of orbital visualization during the whole grid. During the measurement, at each pixel of the scanned frame, the tip-sample distance is adjusted by the feedback loop to preserve the setpoint value of the tunnelling current. Once the tip-sample distance is fixed, the feedback loop is removed, and a bias sweep is performed, with the acquisition of the resulting tunnelling current as output. When the tip is scanned on top of the covalent network, i.e. on C sites (green dots in Figure 3), the contribution to the detected tunnelling current comes from p orbitals, and, in particular, by orbitals, which extend in the out-of-plane direction pointing towards the tip. When the tip is scanned inside the h-GDY pores (violet dots in Figure 3), i.e. on the H-terminated edges of the phenyl rings, the contribution to the tunnelling current comes from C orbitals and H s orbitals; thus, to reach the setpoint current, the tip must get closer. As a result, inside the h-GDY pores, in correspondence of the H-terminated edges of the phenyl rings, the spectra are acquired at a smaller sample-tip distance, resulting in highly intense features, i.e. the states found at -2.1 V and +2.2 V. Conversely, on top of the h-GDY network, the spectra are acquired at a larger sample-tip distance, resulting in weak features, i.e. the states found at -1.45 V and +0.5 V.
Density of States with hybrid functionals
References
- [1]Mark C Biesinger et al. “Resolving surface chemical states in XPS analysis of first row transition metals, oxides and hydroxides: Cr, Mn, Fe, Co and Ni”. In: Applied Surface Science 257.7 (2011), pp. 2717–2730.DOI
- [2]Francesco De Boni et al. “Structure and vibrational properties of 1D molecular wires: from graphene to graphdiyne”. In: Nanoscale 16.23 (2024), pp. 11211–11222.DOI
- [3]Francesco De Boni et al. “Templating effect of different low-miller-index gold surfaces on the bottom-up growth of graphene nanoribbons”. In: ACS Appl. Nano Mater. 3.11 (2020), pp. 11497–11509.DOI
- [4]Sebastien E Hadjadj et al. “Epitaxial monolayers of the magnetic 2D semiconductor FeBr₂ grown on Au(111)”. In: Chem. Mat. 35.23 (2023), pp. 9847–9856.DOI
- [5]Alberto Herrera-Gomez et al. “A self-consistent multiple-peak structure of the photoemission spectra of metallic Fe 2p as a function of film thickness”. In: Surface and Interface Analysis 52.9 (2020), pp. 591–599.
- [6]Yixuan Jiang et al. “Two-Dimensional Iron Oxide on Au (111): Growth Mechanism and Interfacial Properties”. In: J. Phys. Chem. C 125.44 (2021), pp. 24755–24763.DOI
- [7]Fabian Paschke et al. “Distance and voltage dependence of orbital density imaging using a CO-functionalized tip in scanning tunneling microscopy”. In: ACS Nano 19.2 (2025), pp. 2641–2650.DOI
- [8]NJ Tao and SM Lindsay. “In situ scanning tunneling microscopy study of iodine and bromine adsorption on gold (111) under potential control”. In: J. Phys. Chem. 96.13 (1992), pp. 5213–5217.DOI
- [9]Tao Wang et al. “Reaction selectivity of homochiral versus heterochiral intermolecular reactions of prochiral terminal alkynes on surfaces”. In: Nat. Commun. 10.1 (2019), p. 4122.DOI
- [10]Haiming Zhang et al. “On-surface synthesis of rylene-type graphene nanoribbons”. In: J. Am. Chem. Soc. 137.12 (2015), pp. 4022–4025.DOI
- [11]Yi-Qi Zhang et al. “Homo-coupling of terminal alkynes on a noble metal surface”. In: Nat. Commun. 3.1 (2012), p. 1286.DOI