Recent advances in organic molecule reactions on metal surfaces

Abstract

Chemical reactions of organic molecules on metal surfaces have been intensively investigated in the past decades, where metals play the role of catalysts in many cases. In this review, first, we summarize recent works on spatial molecules, small H₂O, O₂, CO, CO₂ molecules, and the molecules carrying silicon groups as the new trends of molecular candidates for on-surface chemistry applications. Then, we introduce spectroscopy and DFT study advances in on-surface reactions. Especially, in situ spectroscopy technologies, such as electron spectroscopy, force spectroscopy, X-ray photoemission spectroscopy, STM-induced luminescence, tip-enhanced Raman spectroscopy, temperature-programmed desorption spectroscopy, and infrared reflection adsorption spectroscopy, are important to confirm the occurrence of organic reactions and analyze the products. To understand the underlying mechanism, the DFT study provides detailed information about reaction pathways, conformational evolution, and organometallic intermediates. Usually, STM/nc-AFM topological images, in situ spectroscopy data, and DFT studies are combined to describe the mechanism behind on-surface organic reactions.

---

Hong-Ying Gao

Hong-Ying Gao obtained his doctorate degree under Prof. Zhenchao Dong's supervision from the University of Science and Technology of China. Then, he joined the Prof. Harald Fuchs group as a post-doc (2011–2012), an Alexander von Humboldt Fellowship (2012–2014), and a junior group leader (2014–2019) at the Münster University, Germany. Since 2019, he has become a full professor in the School of Chemical Engineering and Technology, Tianjin University, China. His research interests are focused on the various physical and chemical processes at surfaces, particularly based on scanning tunneling microscopy, non-contact atomic force microscopy, X-ray photoemission spectroscopy, and luminescence/Raman spectroscopy.

---

1. Introduction

Organic reactions on the surface of metal catalysts are important for both investigating surface reactions to build up novel nanostructures and understanding the underlying metal catalyzing mechanisms.^1,2 To date, many well-known reactions have been reported and applied in on-surface chemistry such as the Ullman reaction,^3,4 Glaser coupling,^5,6 Bergman cyclization,⁷ decarboxylation,⁸ dehydrogenation of alkanes⁹ (aryl,⁴ amino,^10,11 phenol,¹² acetylenic¹³ and silyl¹⁴ groups), Sigma bond metathesis,¹⁵ imine bond formation,¹⁶ azo bond formation,¹⁷ dehydration of boronic acids,¹⁸ dimerization of N-heterocyclic carbenes,¹⁹ acylation reaction,²⁰ cyclotrimerization of acetyls,²¹ cross-coupling of cyano groups,^22,23etc. The catalytic role of metal surfaces is further revealed via both experiments and density functional theory (DFT) studies.^{9,12–15,21,24} Besides, many metal–organic intermediate states of the reactions have been identified, which is important for understanding the underlying mechanism.^8,11,25–28

However, the future of this on-surface chemistry still remains unclear but are likely to depend on the developments of the following four aspects: (1) new chemical reactions, (2) new techniques/tools, (3) novel nanostructures (for example, 1/2 dimensional polymers, nanographene, graphene nanoribbons, etc.) and measuring their properties, and (4) catalysis of metal surfaces, localized catalysis and organic catalysis as shown in Fig. 1. In recent years, three new trends of on-surface chemistry have appeared in molecular candidates as (1) spatial molecules as building blocks,^28–31 (2) reactions on small H₂O, O₂, CO, and CO₂ molecules,^32–38 and (3) molecules carrying silicon groups.^15,39–45 Meanwhile, in situ spectroscopy techniques such as STS,^46–48 force spectroscopy,⁴⁹ X-ray photoemission spectroscopy,^{15,26–28,38} STM-induced luminescence,^50–52 tip-enhanced Raman spectroscopy^53–57 and infrared reflection adsorption spectroscopy³⁷ have been employed to confirm the occurrence of organic reactions and measure the physical properties of products. Furthermore, DFT simulations on the organic reaction pathway,²⁴ molecular conformational evolution,⁵⁸ and organometallic intermediate states²⁵ are the milestones to understand the underlying mechanism and determine the catalytic role of metal surfaces.

Fig. 1 An illustration of the development from on-surface chemistry concepts to bottom-up fabrication applications. In general, it contains four routes to link the concept and application poles: (1) new chemical reactions, (2) new techniques/tools, (3) novel nanostructures and measure their properties, and (4) catalysis of metal surfaces, localized catalysis and organic catalysis.

In this review, we will discuss the new trends in molecular candidates, in situ spectroscopy techniques, and DFT studies of organic reactions on metal surfaces. Hopefully, the catalytic role of metal surfaces is alluded to briefly that in many cases the underlying substrate not only supports but also participates in the organic reactions.

2. New trends in molecular candidates

To date, various molecular candidates have been investigated in the on-surface synthesis area. Looking back to its recent developments, in our eyes, three new trends have emerged and attracted broad interest as discussed in the following sections.

Reactions on spatial molecules

As shown in Fig. 2a, the 2,6-naphthalenedicarboxylic acid (NDCA) molecules on the metal surfaces could make the deprotonation reaction to occur first and then a decarbondioxide reaction to build up metal–organic or covalent poly-naphthalene polymers.⁸ However, once the central aromatic naphthalene is replaced by the diamantane group, the 4,9-diamantane dicarboxylic acid (DDCA) molecules can make the deprotonation reaction to occur first and then deoxygenation to build up dehydroxylated metal–organic dimers, as shown in Fig. 2b.²⁸ It is obvious that even the same reactive group can go through different reaction pathways due to the central moiety spatial or aromatic flat. On the other hand, it is highly interesting to build up the metal–organic or covalent polymers by the spatial molecules on the surface. Along with this idea, we have successfully developed the preparation of 4,9-diamantane dithiol (DDT) covalent polymers via the dehydrogenation reaction.²⁹ Meanwhile, Biao Yang et al. have successfully built up the polyphenylene wires (PBCP covalent polymers) using 1,3-bis(4-iodophenyl)bicyclo[1.1.1]pentane (IPBCP) as a precursor as shown in Fig. 2c.³⁰ And Shigeki Kawai et al. have reported the on-surface synthesis of 3D graphene nanoribbons (2Br-TNP covalent polymers) using the hexabromo-substituted trinaphto[3.3.3]propellane (6Br-TNP) molecule via the Ullman reaction as shown in Fig. 2d.³¹

Fig. 2 Reactions on spatial molecules. (a) Reactions of NDCA molecules on metal surfaces. (b) Reactions of DDCA molecules on the Cu(111) surface and the dehydrogenation reaction of DDT molecules to build up covalent DDT polymers. (c) The Ullman reaction of IPBCP precursors to form PBCP wires. (d) The Ullman reaction of 6Br-TNP to form 3D graphene nanoribbons (2Br-TNP polymers).

It seems that the on-surface synthesis study on spatial molecules has just been initiated. We are confident that this study will be a new branch of on-surface chemistry in comparison with the aromatic flat systems.

Reactions on small H₂O, O₂, CO, and CO₂ molecules

Another trend is the reactions on small gas molecules such as H₂O, O₂, CO, and CO₂. To our knowledge, the preliminary study of gas molecules on the surface was conducted in 1995.⁵⁹ After this, the STM study on O₂ and H₂O molecules on the titanium dioxide surface as well as the CO molecule oxidation via FeO islands on the Pt(111) surface have been reported as important milestones in understanding the underlying surface catalysis.^60–63 However, in this review, we want to introduce some new developments in small gas molecule reactions.

First of all, it is necessary to note the work reported by the Wei Xu group that the self-assembly hydrogen bonding structures of organic molecules can be effectively mediated by H₂O molecules (from T-junctions to one-dimensional cytosine chains) on metal surfaces, as shown in Fig. 3a.³² Even the local chiral of thymine dimers can be inverted by STM manipulating a single H₂O molecule.³³ These two works prove that H₂O molecules participate in the hydrogen bonding of organic molecules. Moreover, the solvation of sodium halides⁶⁴ and methanol⁶⁵ with water molecules has also been reported with atomic resolution.

Fig. 3 Reactions on small gas molecules. (a) H₂O molecules mediate the organic molecular self-assembly structure, ref. 32 and 33. (b) O₂ molecules promote the C–H activation of alkynes, ref. 34 and 35. (c) Ni atoms catalyse the O₂ dissociation and CO oxidation to be CO₂, while the Ni₂ clusters catalyse the CO₂ dissociation to be CO, ref. 36 and 37. (d) The dissociation of H₂ and CO/CO₂ hydrogenation on the ZnO(10-10) surface under ambient conditions, ref. 38. Reproduced from ref. 32 and 38 with permission from American Chemical Society, copyright © 2019, 2020, 2022, 2023.

Along with the above behaviors of water molecules, gas molecules may be further involved in some chemical reactions of aromatic molecules. Indeed, it is expected that O₂ molecules can promote the C–H activation of alkynes on the Ag(111) surface to form organo-metallic one-dimensional chains or two-dimensional networks as reported by the Y. Kim group³⁵ and J. V. Barth group,³⁴ respectively, as shown in Fig. 3b.

Furthermore, as shown in Fig. 3c, the CO molecule oxidation by the single Ni atoms on a monolayered CuO surface was reported by the Kai Wu group.³⁶ They have found that single cationic Ni atoms are much more active for CO oxidation as compared with the single metallic Ni atoms. Moreover, the Ni₂ clusters on the monolayered CuO surface can be rather active for the CO₂ molecule bond dissociation to be the CO molecule.³⁷ In addition, the dissociation of H₂ molecules and CO/CO₂ hydrogenation reactions on the ZnO(10-10) surface under ambient conditions was also reported by Fan Yang et al. as shown in Fig. 3d.³⁸ Note that Fan Yang's work helps the reactions on gas molecules to be rather close to the industrial syngas conversion.

All these above works allow us to obtain an understanding of the gas molecules’ reaction on surfaces with atomic resolution and single-atom catalyst levels, which is regarded as an important milestone.

Reactions on organic silicon molecules

The third trend is the reactions on organic silicon molecules, as shown in Fig. 4. It is well known that silylation (adding protecting groups) and desilylation (removing protecting groups) are important functional group manipulations in solution chemistry. Thus, is it possible to realize both silylation and desilylation reactions on the organic silicon molecules by on-surface synthesis?

Fig. 4 Reactions on organic silicon molecules. (a) σ-Bond metathesis reaction. (b) Desilylation reaction. (c) Light assisted C(sp)–Si bond dissociation. (d) Desilyative Glaser coupling. (e) Desilative Sonogashira coupling reaction. (f) Disilabenzene formation reaction. (g) C(sp²)–Si bond activation reaction.

To answer this question, we first reported the σ-bond metathesis reaction between silyl-protected alkynes and NDCA molecules,¹⁵ as shown in Fig. 4a. Then, we have also investigated the role of central π-systems in the σ-bond metathesis reaction and the introduction of iodine atoms to tune σ-bond metathesis reaction precursors or products.^66,67 Recently, Ran et al. have disclosed that the tri-isobutylsilyl group protected alkynes can also be desilylated to be alkynes, as illustrated in Fig. 4b.³⁹ In addition, S. Mahapatra et al. reported that local plasma light can also realize the C(sp)–Si bond activation leaving the alkyne group as a radical on the Cu(100) surface as shown in Fig. 4c.⁴⁰

The covalent coupling of alkynes after desilylation can further occur as reported.¹⁵ For example, S. Kawai et al. have reported that trimethylsilyl-protected alkynes can undergo desilylation and then Glaser coupling directly on the Cu(111) surface at 400 K as illustrated in Fig. 4d.⁴¹ Furthermore, S. Kawai et al. have also reported the Sonogashira coupling between the silyl protected alkynes and C(sp²)–Cl groups on the Cu(111) surface.⁴² Under the light of this work, we have further reported that the C(sp²)–Si bond activation of molecules carrying one side C–Br group and the other side C–Si group to form graphene nanoribbons or one-dimensional metal–organic chains on metal surfaces.⁴⁴ In addition, recently N. Cao et al. have reported the highly efficient cyclization reaction of alkynes after the thermal desilylation reaction.⁴⁵

Not only the desilylation reaction but also the silylation reaction has been successfully realized on the surface. S. Kawai et al. have reported the silabenzene formation via the reaction between aromatic C–Br and silicon atoms on the surface as shown in Fig. 4f.⁴³ We hope that the silylation and desilylation reactions can be heavily applied in on-surface synthesis as in solution chemistry.

3. In situ spectroscopy techniques

Along with the development of on-surface chemistry, more and more in situ spectroscopy technologies have been employed to characterize organic reactions. Herein, we would like to briefly introduce several in situ spectroscopy techniques reported in the literature as follows.

Electronic spectroscopy

The STM spectroscopy (STS), including I–V, dI–dV, d²I–d²V, I–t, Z–t, and I–Z measurements, which are applied most widely and normally reveal the electronic properties of samples or local physical and chemical behaviors, should be introduced first. Herein, we focus on the introduction of dI–dV spectroscopy on various on-surface synthesized nanostructures (nanographenes) as a good example.

As shown in Fig. 5a, P. Ruffieux et al. reported the electronic structures of armchair graphene nanoribbons with an atomic width of N = 7.⁶⁸ It is found that the bandgap is about 2.3 eV with the corresponding occupied state at −0.7 eV and unoccupied state at +1.5 eV. Furthermore, as shown in Fig. 5b, F. Xiang et al. reported the STS of planar π-extended cycloparaphenylenes featuring an all-armchair edge, which reveals molecular occupied and unoccupied orbitals.⁶⁹ Beyond the STS, they have further obtained the dI/dV maps at specific molecular states. In addition, J. Cai et al. reported the electronic properties of a heterojunction between pristine (undoped) and nitrogen-doped graphene nanoribbons, as shown in Fig. 5c.⁷⁰ The bandgap of pristine graphene nanoribbons is measured about 1.6 eV, while the nitrogen-doped bandgap is 1.5 eV. The band offset across the undoped and doped graphene nanoribbons is identified. Similarly, Y.-C. Chen et al. also reported the molecular bandgap engineering of graphene nanoribbon heterojunctions between atomic widths of N = 7 (1.3 ± 0.1 nm) and N = 13 (1.9 ± 0.2 nm) segments.⁷¹ It is found that the STS at different positions is different, depending on the local density of states. Furthermore, using the same on-surface synthesis strategy, X. Peng et al. measured the electronic structures of organic macrocycle quantum corrals formed by 12 units, revealing its conduct band position at +1.3 V and valence band position at −1.05 eV respectively.⁷² Recently, S. Mishra et al. have also measured the magnetic exchange coupling in rhombus-shaped nanographenes ([5]-rhombenes) with a zig-zag periphery by dI/dV spectroscopy, as shown in Fig. 5f.⁷³ In contrast, for [4]-rhombenes, the dI/dV spectroscopy only shows the HOMO and LUMO resonances at −330 mV and +400 mV and does not show any magnetic exchange coupling at all.⁷³

Fig. 5 STS on various on-surface synthesized nanostructures. (a) Graphene nanoribbons, ref. 68. (b) Cycloparaphenylenes, ref. 69. (c) P–N graphene heterojunctions, ref. 70. (d) Different width graphene heterojunctions, ref. 71. (e). Macrocycle quantum corrals, ref. 72. (f) Rhombus-shaped nanographenes, ref. 73. Reproduced from ref. 68 with permission from American Chemical Society, copyright © 2012 and reproduced from ref. 69 and 73 with permission from Nature Publishing Group, copyright © 2014, 2015, 2021, 2022.

Thus, the electronic spectroscopy of various novel nanostructures as the products of on-surface organic reactions can be effectively measured by STS. Even the spin (magnetic) coupling of molecules inside an organic chain can be revealed.⁷⁴ And, in the future, such in situ STS should play an important role in the on-surface chemistry development.

Force spectroscopy

In parallel with in situ STS, force spectroscopy by non-contact atomic force microscopy (nc-AFM) is also applied to characterize the on-surface synthesized nanostructures, especially for the aromatic flat systems. However, force spectroscopy is not widely applied compared with the STS approach.

Herein, three points on force spectroscopy are introduced based on our knowledge. First, it is that force spectroscopy can be applied to characterize different atoms, as shown in Fig. 6a.⁷⁵ In the corresponding nc-AFM image of Sn–Si atoms and Pb–Si atoms, the dimmed atoms are the same Si atoms. In contrast, the Sn and Pb atoms appear as bright balls, but they show different heights especially when both Sn and Pb atoms are grown on the Si(111) surface. More importantly, the force spectroscopy on Sn, Pb, and Si atoms shows clear differences in the maximum attractive force positions, which can be applied as a chemical identification approach. It is necessary to point out that herein force spectroscopies are normalized to overcome the different tip's statuses. Second, the force spectroscopy using a fixed oxygen-terminated cooper (Cu–O) tip has also been reported by H. Mönig et al., which could give a more precise measurement of the C–C bond lengths of inner-nanographene (dcoronylene molecule) compared with the normally nc-AFM CO tip by the reduced lateral deflection during scanning.⁷⁶ In addition, the force spectroscopy at different positions of the dicoronylene molecule is shown in Fig. 6b. The attractive force maximum on the centre position is stronger than it is on a C–C bond of a phenyl, supporting this Cu–O tip's covalently bound configuration. Third, sometimes the force (or the frequency shift Δf) shows some jumps as indicated by a black arrow during the force spectroscopy at the white star marked position on the STM image as shown in Fig. 6c, which corresponds to the force-induced porphycene hydrogen tunneling from NH to N positions.⁴⁹ Thus, the STM topological image of porphycene can be switched between the C and anti-C shapes. This indicates that the force of nc-AFM in principle can be applied to trigger potential chemical reactions, agreeing with another literature report.⁷⁷ It should be noted that such a force-induced organic chemical reaction is just an initial study in the on-surface chemistry area, which may be developed as a new branch.

Fig. 6 Force spectroscopy on atoms or molecules. (a) Force spectroscopy characterization on different Si, Sn and Pb atoms, ref. 75. (b) Force spectroscopy on a dicoronylene molecule by a Cu–O tip, ref. 76. (c) Force spectroscopy on a porphycene molecule, by which the force induced hydrogen tautomerization, ref. 49. Reproduced from ref. 49 and 75 with permission from Nature Publishing Group, copyright © 2016, 2007 and reproduced from ref. 76 with permission from American Chemical Society, copyright © 2016.

Although it can be seen that normally nc-AFM is widely applied to obtain a molecular high-resolution image with a resolved chemical bond; however, its force spectroscopy is limited and thus has a long road to overcome. We are still confident that force spectroscopy will be used more and more often.

X-ray photoemission spectroscopy

Next, in situ X-ray photoemission spectroscopy (XPS) as a surface chemical analysis technique is introduced to analyze on-surface chemistry in combination with STM and nc-AFM. In many cases, the XPS data could support well the STM and nc-AFM results, if the observed chemistry is efficient and the precursors/products are not complicated. Here, we would focus on the XPS on organic molecules, as a supplementary method to elucidate on-surface chemistry.

As reported, the σ-bond metathesis reaction is a silyl-proton exchange reaction between silyl-protected alkyne and an aromatic acid group, as shown in Fig. 7a.¹⁵ Thus, in principle, the XPS on the Si 2p peak should give a clear proof of the σ-bond metathesis reaction, due to its switch from C–Si before the reaction to O–Si after the reaction. Similarly, the XPS on O 1s should also give a proof, due to its switch from COOH before the reaction to C–H after the reaction. Indeed, the experimental XPS of the Si data shows a clear Si 2p_3/2 peak shift from 100.97 eV to 101.61 eV, corresponding to before and after σ-bond metathesis. Meantime, the XPS of the O data shows a clear hydroxyl O 1s peak shift from 533.8 eV to 531.3 eV and a slightly shifted carbonyl O 1s peak from 532.5 eV to 532.7 eV, corresponding to before and after σ-bond metathesis. Therefore, the XPS results together with STM images prove the σ-bond metathesis reaction. In addition, due to the precursors/products being complicated (before and after the reaction, 2 precursors and 3 products are mixed on the surface), the measured C 1s peak is rather difficult to analyze.

Fig. 7 XPS to characterize on-surface reactions. (a) XPS on Si 2p and O 1s for the σ-bond metathesis reaction, ref. 15. (b) XPS on C 1s and O 1s for the peroxide coupling reaction, ref. 26. (c) XPS on C 1s and O 1s for the α-diazo ketone reaction, ref. 27. (d) XPS on C 1s and O 1s for CO₂ and H₂ reactions on the ZnO(10-10) surface, ref. 38. Reproduced from ref. 26 with permission from Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, copyright © 2016 and reproduced from ref. 15 and 38 with permission from American Chemical Society, copyright © 2017, 2018, 2023.

The second example is the peroxide coupling of ENA molecules on the surface, as shown in Fig. 7b.²⁶ Here, a two-reactive-group (alkyne and acid) strategy is applied, and the crosstalk between these two reactive groups is disclosed. The alkyne groups undergo Glaser coupling to form BDNA molecules, but herein they also show a metal–organic intermediate state. In contrast, the acid groups undergo peroxide coupling. To solidify the peroxide coupling of acid groups, the XPS of C 1s and O 1s are recorded on the peroxide product poly-BDNA polymers. The carboxylic C 1s peak at 288.40 eV and O 1s double peaks at 532.86 eV and 531.45 eV are identified. The reference molecule BNP, which is out-synthesized and confirmed well by NMR, shows rather close XPS peaks as the carboxylic C 1s peak at 288.52 eV and O 1s double peaks at 532.94 eV, 531.55 eV. This XPS analysis supports well the peroxide coupling of ENA molecules, in good agreement with STM observations. The STM observation on out-synthesized BNP molecules also shows the same peroxide linkage with ENA molecules.

In the on-surface reaction of the α-diazo ketone work, both STM and nc-AFM show good images on the organometallic intermediate state (Cu-biscarbene) and the furan ring formation, as shown in Fig. 7c.²⁷ Along the α-diazo ketone reaction, XPS should give a clear proof on both the C 1s peak and the O 1s peak. Indeed, the XPS C1s peak on α-diazo ketones shows a clear Cu-biscarbene state, due to the shoulder C1s peak position at 283.2 eV. And this shoulder peak disappears along thermal annealing. In contrast, the XPS carbonyl O 1s peak is at 530.5 eV before the reaction, which is shifted to the furan O 1s peak at 533.6 eV after the reaction.

Recently, Fan Yang et al. reported atomic scale visualization of H₂ dissociation and CO₂ hydrogenation on ZnO under ambient conditions by ambient pressure XPS-STM, as shown in Fig. 7d.³⁸ At 300 K and 400 K, the STM images clearly show that the mixture gases of CO₂ and H₂ (1 : 3) adsorb on the ZnO(01-10) surface. The XPS C 1s peaks at 290.4 eV and 289.9 eV are assigned to carbonates and formate species. Similarly, the XPS O 1s peaks is consistent with the formation of carbonates and formate species. It is obvious that the AP-XPS data further solidify the fact of H₂ dissociation and CO₂ hydrogenation on the ZnO surface. This study brings the on-surface chemistry study much closer to the working conditions of industries.

It can be seen that in situ XPS as a supplementary method in on-surface chemistry has been heavily applied in many cases.

STM tip-induced luminescence

STM tip-induced luminescence (STML) is another in situ spectroscopy approach applied in on-surface chemistry. However, successful STML measurements on precursors or products of on-surface chemistry are still rather limited. Here, we mainly introduce three representative works from the Schull group.

As is known, the 10,10′-dibromo-9,9′-bianthryl (DBBA) molecule undergoes an on-surface reaction to form 7-atom-wide armchair graphene nanoribbons (7-AGNRs). In 2018, M. C. Chong, et al. first reported the STM-induced single 7-AGNR fluorescence by tip lifting the single nanoribbon, as shown in Fig. 8a. It is found that if the 7-AGNRs are H-terminated, then the STML spectrum is featureless and weak, which resembles a typical plasmon spectrum. To our surprise, if the 7-AGNRs are C-terminated (dehydrogenation occurs at ramping 4V, 10 nA by STM tip manipulation), the STML spectrum shows a rather sharp and intense emission peak at 1.6 eV with a spectral width of 40 meV (full width at half maximum, FWHM), as well as two faint peaks at about 1.44 eV and 1.29 eV. These two faint peaks are tentatively assigned to be vibronic modes of 7-AGNRs.

Fig. 8 STML spectroscopy in on-surface chemistry. (a) STML from single graphene nanoribbon junctions, ref. 50. (b) STML from a single graphene nanoribbon on the decoupling NaCl film, ref. 52. (c) STML from a single H₂Pc molecule on the NaCl film, ref. 51. Reproduced from ref. 50 with permission from American Chemical Society, copyright © 2018, reproduced from ref. 51 with permission from Nature Publishing Group, copyright © 2020, and reproduced from ref. 52 with permission from American Association for the Advancement of Science, copyright © 2023.

Furthermore, recently in 2023, S. Jiang, et al. reported the STM-induced luminescence from single 7-AGNRs decoupled by the NaCl film on the Au(111) surface, as shown in Fig. 8b. In this way, the spectroscopy of H-terminated 7-AGNRs could give a narrow peak at 1.45 eV with a spectral width of 0.57 meV, along with many vibronic sub-peaks. Even high-energy vibronic peaks corresponding to D and G modes of 7-AGNRs can be identified at 1348 and 1610 cm⁻¹. In addition, the vibronic peak distribution (ΔE) has a direct dependency on the length of 7-AGNRs. It should be noted that the obvious difference in STML spectra from tip lifted and NaCl film decoupled single 7-AGNRs is caused by the different nanoribbon's termination (C-terminated vs H-terminated, main reason) and length (subsidiary reason).

In addition, B. Doppagne et al. have also reported that STM-induced single H₂Pc molecule fluorescence can be applied to track its tautomerization, as shown in Fig. 8c. Here, the single H₂Pc molecule was grown on the NaCl film of the Ag(111) substrate, due to the decoupling reason. It is found that the two tautomers should be identified as the inner two hydrogen positions along different diagonal lines. And the molecular LUMO STM image (at 0.55 V) shows two patterns corresponding to the two tautomers. Meanwhile, the molecular HOMO STM image (at −2.5 V) shows the same. More importantly, the STML spectrum at position 1 shows an intense emission Q_x peak at 1.80 eV, and a weaker Q_y peak at 1.92 eV. In contrast, for the STML spectrum at position 3, it shows a 20 meV shift of the Q_x peak to higher energies. For the STML spectrum at position 2, it shows a mixture of spectra at positions 1 and 3. Thus, it is reasonable to identify the two tautomers based on their STML spectra. It should be noted that a series of very weak vibronic features also exist in the STML spectrum of a single H₂Pc molecule.

Due to the strict decoupling requirement for STML on organic molecules, until now the successful molecules are rather limited, which further influences its application in on-surface chemistry. However, we are still confident in its development in the on-surface chemistry area.

Tip-enhanced Raman spectroscopy

In contrast to the above-mentioned STML spectroscopy approach, the tip-enhanced Raman spectroscopy (TERS) approach has been more and more involved in on-surface chemistry. Such an TERS approach is powerful and in principle can characterize most of the on-surface chemical behaviors, due to its ultrahigh resolution at a single molecule level.^53,54 Herein, we will briefly introduce three works in which TERS is applied to characterize chemical reactions, especially in bond identification.

As shown in Fig. 9a, the intact pentacene molecules adsorbed on the Ag(110) surface mainly show a rod-like configuration labeled as α. However, by sequentially applying 2.0 V pulses over the molecule, two kinds of new species with different dumbbell-like and spindle-like shapes are formed labeled as β and γ. Here, these β and γ species correspond to the central one side (β) and both sides (γ) dehydrogenated pentacenes as shown by the molecular structures in Fig. 9a. Fortunately, TERS at the α, β and γ pentacenes indeed show the obvious difference at the middle and end positions.⁵⁵ More importantly, the C–H stretching mode of intact pentacene appears in the high-wave number region (>2500 cm⁻¹), and it disappears for γ species proving the dehydrogenation at the central positions of pentacene.

Fig. 9 TERS in on-surface chemistry. (a) TERS to characterize the dehydrogenation of a pentacene molecule on the Ag(110) surface, ref. 55. (b) TERS to characterize the intramolecular isotope effects, ref. 56. (c) TERS to characterize the enyne and cumulene linkages in the alkyne reaction, ref. 57. Reproduced from ref. 55 with permission from American Association for the Advancement of Science, copyright © 2021 and reproduced from ref. 56 and 57 with permission from American Chemical Society, copyright © 2023, 2021.

This work is a good example that the in situ TERS really could reach a single bond identification inside a single molecule.

Furthermore, the TERS is also applied to detect the isotope effects by the comparison between pentacene (C₂₂H₁₂) and fully deuterated pentacene (C₂₂D₁₂) after their dehydrogenation reaction at the Ag(110) surface, as shown Fig. 9b.⁵⁶ It should be noted that the TERS spectra on middle sites are rather similar for C₂₂H₁₂ and C₂₂D₁₂ molecules, in contrast to the obvious different TERS spectra at end sites. This indicates the isotope effect of C–H and C–D vibrational differences. They measured the TERS mapping along STM scanning for the different vibrational modes, which further solidified the Raman peak shift due to the isotope effect.

Similarly, C. Zhang et al. have also successfully applied TERS to identify the enyne and cumulene linkages in the alkyne on-surface reaction (4,4′-diethynyl-1,1′-biphenyl, DEBP), as shown in Fig. 9c.⁵⁷ They observe the obvious Raman peak at 2107 cm⁻¹ for the alkyne stretching mode of enyne and that at 2090 cm⁻¹ for cumulene vibration. It is necessary to note that all these TERS measurements are performed on an oligomer of DEBP molecules. Additionally, the TERS between 1000 and 1700 cm⁻¹ is also different. For example, the characteristic peaks of cumulene at 1394 and 1411 cm⁻¹ can be assigned to C–H bending in the cumulene connection.

These three works well elucidate that the in situ TERS could characterize on-surface reactions even at the single bond level. In principle, the in situ TERS is sensitive to the reaction occurrence on a single molecule, and therefore it could determine the chemical bond evolutions of a molecule with complicated reactions. This method has a great importance in accelerating the development of on-surface chemistry.

Infrared reflection absorption spectroscopy

At the last of the in situ spectroscopy part, temperature-programmed desorption (TPD) spectroscopy and infrared reflection absorption spectroscopy (IRAS) will be introduced. The TPD spectroscopy is usually applied to monitor the organic molecules desorbed from the surface and measure their mass. Thus, it can be applied to identify the on-surface reaction by the mass and desorption analysis. Meanwhile, IRAS is used to analyze the specific molecular infrared absorption peaks, which is another molecular fingerprint technology like Raman spectra. Here, we only introduce two recent works as an example of TPD spectroscopy and IRAS.

As mentioned before in reactions on small gas molecules, CO molecules could adsorb on the Ni/CuO surface, as STM images are shown in Fig. 10a.³⁶ The left STM image (Fig. 10a-i) is for the Ni_m&Ni_c/CuO sample after exposure to CO molecules, while the right STM image (Fig. 10a-ii) is for the Ni_c/CuO sample after exposure to CO molecules. The TPD spectroscopy at CO mass (28) is measured for bare CuO, Ni_c/CuO, and Ni_m&Ni_c/CuO samples as shown in Fig. 10a-iii. Only the TPD spectroscopy of the Ni_m&Ni_c/CuO sample shows a clear desorption peak at 400 K, indicating the specific interaction between CO and Ni_m on the CuO surface. And CO could not adsorb on the Ni_c positions directly. In the following, the Ni_c atoms are found active for O₂ dissociation, forming Ni_c–O species. Thus, the Ni_m&Ni_c/CuO sample can realize the CO oxidation by such Ni_c–O species, and the release of CO₂ is further confirmed by TPD spectroscopy.

Fig. 10 TPD spectroscopy and IRAS in on-surface chemistry. (a) TPD spectroscopy to characterize the CO desorption from the surface, ref. 36. (b) IRAS to characterize the CO₂ conversion to CO on the Ni2/CuO surface, ref. 37. Reproduced from ref. 36 and 37 with permission from American Chemical Society, copyright © 2022, 2023.

In a later work, they reported the Ni₂ clusters to realize CO₂ dissociation to be CO, as shown in Fig. 10b.³⁷Fig. 10b-i is the STM image for Ni₂–O species and CO products on the CuO monolayer with the Ni₂–CO₂ sample annealed. In comparison, Fig. 10b-ii shows the direct CO adsorption on a clean CuO surface. More importantly, here they successfully employ the IRAS to characterize the CO adsorption on CuO and Ni₂/CuO surfaces, as shown in Fig. 10b-iii. In contrast, for the CO₂ adsorption experiment, only the Ni₂/CuO sample shows the IRAS peak corresponding to the CO molecule. This strongly supports that only Ni₂ sites realize the CO₂ dissociation to CO on the CuO surface.

Of course, there are many other in situ spectroscopy methods applied in chemistry or catalysis, but not mentioned in this manuscript. Here, it is necessary to point out that the in situ spectroscopy technology has been combined with STM and nc-AFM together to characterize surface chemistry, exploring the frontiers of chemical knowledge.

4. DFT study

To well understand the observed experimental results, DFT calculations are inseparable. In the following, the DFT calculations applied in on-surface chemistry research will be briefly introduced. According to the author's experience, it is assigned into four parts: (1) reaction pathway, (2) conformational optimization, (3) crosstalk between different functional groups, and (4) spectroscopy computation. Finally, a critical evaluation of the DFT simulations is further discussed.

Reaction pathway

In on-surface chemistry/catalysis behaviors, the reaction pathway is the most important part of the reaction mechanism. Obtaining the reaction pathway is a key and normally can be achieved by DFT. Of course, the reaction pathway strongly depends on the reactive functional groups, the underlying metal surfaces, the local environment, etc. In this review, we take two of our own and one other group's works as examples of DFT calculations on the reaction pathway.

According to the textbook of chemistry, it is known that there are two classical reaction pathways of C–H and alkynyl activations for the alkyne reaction. And in the experiments, the dehydrogenated Glaser linkage, various branched linkages, and the cyclic aromatic formation of alkynes have been successfully identified.^6,24 Even the role of the underlying metal surface is revealed that the Au(111) surface is better for the formation of branched linkages, while the Ag(111) surface is better for the Glaser coupling linkage.^6,24 To understand the underlying reaction (or catalyzed) mechanism, DFT calculations are timely conducted and the results are shown in Fig. 11a.

Fig. 11 Reaction pathway calculations in on-surface chemistry. (a) The C–H and alkynyl activation pathways in Glaser coupling at surfaces, ref. 24. (b) The possible reaction pathway via the most stable intermediate structures for the dehydrogenation coupling of silanes, ref. 14. Reproduced from ref. 23 with permission from American Chemical Society, copyright © 2013 and reproduced from ref. 14 with permission from Nature Publishing Group, copyright © 2021.

First, based on the experimental observation, we consider both C–H and alkynyl activations for the alkyne reaction on Au(111) and Ag(111) surfaces as shown in Fig. 11a-i. For the C–H activation pathway, the Glaser coupling contains two steps: (1) the dehydrogenation process and (2) the sequential C–C coupling. DFT simulates these two steps as shown in Fig. 11a-ii. And it is found that, for the first step, the transition states are 1.64 and 1.85 eV above the initial state energy on Au and Ag, respectively. In contrast, for the second step, the transition states are 1.15 and 1.31 eV for Au and Ag, respectively. These results indicate that dehydrogenation should be more difficult on Ag than on Au. For the alkynyl activation pathway, it is found that Au and Ag exhibit differences, as shown in Fig. 11a-iii. The transition states for C–C coupling are found to be 0.79 eV and 0.90 eV for Au and Ag, respectively. The alkynyl activation pathway is much more reasonable for the alkyne reaction on metal surfaces. The Au interacts strongly with the alkynyl groups, resulting in its better for the formation of branched linkages, as compared with the Ag case. By the combination of DFT calculations on the reaction pathway and STM experimental results, the on-surface reaction (catalysis) is studied better and more completely.

However, DFT calculations are rather time-consuming due to the complexity of the reaction, and sometimes it is necessary to simplify the calculations by giving up the transition state searching and only calculating the stable intermediate states. As shown in Fig. 11b, the dehydrogenative coupling of silanes has been reported and its possible reaction pathway is deduced by the most stable intermediate structures calculations.¹⁴ It is revealed that the –SiH₃ group goes first to be –SiH₂ (intermediate state M·) and releases one hydrogen, by the interaction with the metal surface. Then, one formed –SiH₂ further releases the third hydrogen (intermediate state M^:) and forms covalent coupling with another –SiH₂ group (intermediate state D′). After this, the second –SiH₂ group releases the fourth hydrogen, resulting in the linkage of Si–Si covalent coupling (product D). Finally, two hydrogen molecules are released from the metal surface. This is the most possible reaction pathway, agreeing well with experimental results.

Even for the same reaction pathway, it may vary in the reaction barriers depending on the underlying metal substrates. A famous example is the Ullman coupling reaction pathway on Au(111), Ag(111), and Cu(111) surfaces, as shown in Fig. 12. The whole Ullman reaction pathway on the surface is divided into three steps: (1) the C–Br (or C–I) bond dissociation (Fig. 12a), (2) the sliding and flipping of a phenyl radical (Fig. 12b), and (3) two phenyls confined by a common metal atom directly undergo covalent coupling to form a biphenyl product (Fig. 12c).⁷⁸ Here, all steps strongly depend on the underlying metal surfaces. For bond dissociation and covalent coupling processes, Cu(111) is the best surface for the lowest energy barriers. In contrast, for dynamic processes, Au(111) is the best for the sliding process, and Ag(111) is the best for the flipping process. In addition, the Ullman reaction of C–I is much easier than that of C–Br, following the regularity of the periodic table of elements.

Fig. 12 The same reaction pathway calculations on Au, Ag, and Cu surfaces, ref. 78. (a) The C–Br and C–I bond dissociation processes. (b) The following slide and flip processes of a phenyl radical. (c) The coupling reaction of two phenyls into biphenyl. Reproduced from ref. 78 with permission from American Chemical Society, copyright © 2013.

Obviously, the DFT calculation about the reaction pathway is very helpful to understand the underlying reaction mechanism. Sometimes, the experimental observed organometallic intermediates are milestones to examine the DFT calculations. In addition, the catalytic role of metal surfaces in organic reactions can also be revealed by DFT calculations, which is usually not easy to see directly in experiments.

Conformational optimization

Another important class of DFT calculations involve conformational optimization, which is also important to understand the on-surface reaction. In many on-surface reactions, the reaction pathway is rather complicated, and only some intermediate states and final products are identified by experiments; therefore, the conformational optimization on possible structures is also an effective way to elucidate the reaction mechanism. Here, we will introduce three representative works as follows.

The first work is about the on-surface reactive planarization of Pt(II) complexes, in which the C–N bond scission is discovered.⁷⁹ As shown in Fig. 13a, molecules C1–C4 have been measured by STM. And it is found that molecule C1 will be dissociated, leaving planar C2 molecules as the on-surface reaction products on the Cu(111) surface. A comparison with the STM observation on out-synthesized C2 molecules has been done, proving the C–N bond dissociation fact. To understand the underlying reaction mechanism, constraint optimized structures of a C1 analogue (without the alkyl chain, but with two methyl groups in meta positions) have been conducted. The energy profile of C–N scission is further obtained on the Cu(111) surface, in which the reaction barrier is about 1.7 eV as shown in Fig. 13a-ii–iv. In addition, the optimized structure of C2 (Fig. 13a-v) shows a planar structure on Cu(111), confirming with STM images on C1 products and out-synthesized C2. In contrast, for the C3 molecule, the pyrimidine ring has a stronger interaction with the metal surface (eliminating the steric hindrance caused by phenyl hydrogens, which led to the perpendicular orientation of the C1 optimized structure), so that its optimized structure is almost planar on the Cu(111) surface as confirmed by DFT calculations in Fig. 13a-vi. The fully planar structure results in a drastically reduced reaction barrier of 0.62 eV, as shown in Fig. 13a-viii. This further matches well with the STM observation that C–N bond dissociation occurs at room-temperature. Furthermore, the C4 molecule is also measured, in which its Pt core part has a stronger interaction the with metal surface compared with C3, due to the existence of thiophenyl groups. This will bring the optimized conformation of C4 is not planar any more, resulting its C–N bond dissociation reaction barrier increased to 0.80 eV as shown in Fig. 13a-viii. Indeed, the STM experiments confirmed the DFT prediction. Finally, the different metal surfaces Ag(111) and Au(111) are also compared, indicating that the Cu(111) surface is the best for the C–N bond dissociation, as shown in Fig. 13a-ix. Obviously, these optimized conformational calculations are very important for us to understand the finely tuning effect in the C–N bond dissociation process.

Fig. 13 Conformational optimization in on-surface chemistry. (a) On-surface reactive planarization of Pt(II) complexes, ref. 79. (i) The molecules C1–C4. (Ii–iv) Constraint-optimized structures of a C1 analogue on the Cu(111) surface. (v–vii) Optimized structures of C2 (v), C3 (vi) and C4 (vii) on the Cu(111) surface. (viii and ix) Energy profiles for C–N scission of C1, C3, and C4 on Cu(111) and (viii) C3 on Cu(111), Ag(111) and Au(111) (ix). (b) Conformational optimization on the PMDI molecular intact, intermediate and products in a stepwise dehydrogenation, ref. 58. (i) Reaction scheme of PMDI on the Cu(111) surface. (ii–iv) Optimized conformation calculations on the PMDI intact (ii), PMDI’ (iii) and PMDI’–Cu dimer (iv). Reproduced from ref. 79 with permission from Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, copyright © 2019 and reproduced from ref. 58 with permission from Royal Society of Chemistry, copyright © 2021.

The second work is about the PMDI molecular stepwise dehydrogenation reaction and its STM confirmed conformational evolution on the Cu(111) surface.⁵⁸ As shown Fig. 13b-i, the PMDI molecule has –N–H and –C–H groups, which may have sequential dehydrogenation reactions. Indeed, it is confirmed by the STM experiment that the –N–H is first to be activated, forming the PMDI’ intermediate state. By the conformational optimization of PMDI and PMDI’, the intact PMDI is more flat on the Cu(111) surface, and in contrast, the PMDI’ is more bent by the N binding to the metal surface, as shown in Fig. 13b-ii and -iii, respectively. Furthermore, after the C–H activation, PMDI’-Cu dimers are formed, as shown in Fig. 13b-iv. By the conformational optimization, the other side C–H is much more difficult to be activated, due to its lifting up from the metal surface. All these PMDI intact, PMDI’ intermediate state and PMDI’-Cu dimers are supported well by STM experimental images.

The third work is about the detailed conformation evolutions in the on-surface reaction of the [1,1′-biphenyl]-4,4′-disulfonyl dichloride (BPDSC) molecule.⁸⁰ As shown in Fig. 14a, the whole reaction can be assigned into two stepwise reactions, dehalogenation and desulfonylation processes, which are almost the same on both the Au(111) and Ag(111) surfaces. However, the detailed conformation of BPDSC molecule shows a slightly difference on the Au(111) and Ag(111) surfaces. The dehalogenated molecules anchor on the Au(111) surface via Au–S interactions, which weaken the phenyl–S bond, as shown in Fig. 14b-i. In contrast, the dehalogenated molecules anchor on the Ag(111) surface via Ag–O interactions, resulting in the lifting of the S atoms, as shown in Fig. 14b-iii. These conformation differences result in the desulfonylation reaction barriers of 1.17 eV and 1.45 eV different for Au(111) and Ag(111), respectively (Fig. 14b-ii). Thus, the conformation optimization strongly depends on the underlying metal surface, and it may be designed to finely tune the complicated reactions.

Fig. 14 The conformational difference depending on underlying metals in on-surface chemistry, ref. 80. (a) The chemical reaction of the BPDSC molecule. (b) The detailed conformations in the desulfonylation reaction on the Au(111) surface (b-i), the corresponding energy profiles (b-ii), and the detailed conformations in the desulfonylation reaction on the Ag(111) surface (b-iii). Reproduced from ref. 80 with permission from American Chemical Society, © of 2022.

In a short summary, such conformational optimization on the precursor, intermediate state and product are milestones in the on-surface reaction, and they together with STM data reveal the underlying reaction mechanism.

Crosstalk between different functional groups

In on-surface reactions, the crosstalk between different functional groups is not easy to identify. For example, in our on-surface domino reaction work,²⁶ the crosstalk between acid and alkyne groups helps the peroxide coupling and organometallic intermediate state of Glaser coupling occur, which are not observed for pure acid or alkyne molecules.^6,8 However, it is rather difficult to quantify such crosstalk. Only in some cases, the two functional groups react with each other such as the Sigma bond metathesis work.¹⁵ Based on the experimental capture of intermediate states or products about the crosstalk between different functional groups, the DFT calculation on the crosstalk of groups is plausible reasoning. Here, we would like to introduce our azo bond formation by amine-nitro coupling on the Ag(111) surface as a good example of the crosstalk between different functional groups.¹⁷

To speed up the DFT calculations, coupling of aniline and nitrobenzene on Ag(111) was studied. It is necessary to note that no reaction barriers or transition states were calculated. Only 18 structures possibly involved in this model reaction were calculated on the Ag(111) surface, ranging from the starting precursors to trans-azobenzene. The corresponding energy profile can be seen in Fig. 15. Two pathways via the most energetic favourable intermediate structures are highlighted in green colour. Such intermediate states and final azo bond products indeed reveal the crosstalk between aniline and nitrobenzene in the on-surface reaction. Because the aniline needs to be oxidized, while the nitrobenzene needs to be reduced; thus, once these two groups are together at the same space, they can accelerate each other. By the DFT calculations on the intermediate states, our work potentially indicates the crosstalk between two functional groups in on-surface reactions.

Fig. 15 Calculated energy profile of aniline and nitrobenzene reactions on the Ag(111) surface and their corresponding structures. It should be noted that two possible pathways via the most energetic favourable intermediate structures are highlighted in green colour. ref. 17. Reproduced from ref. 17 with permission from Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim, copyright © 2021.

In fact, the crosstalk among two reactive groups is also mediated by the underlying substrate. For example, in our substrate-mediated C–C and C–H coupling after the dehalogenation work,⁸¹ only on the Ag(111) surface the hydrogen of the hydroxy group can passivate the dehalogenated carbon positions, forming new monomers. In contrast, the dehalogenated carbon positions will go direct C–C coupling on the Au(111) surface. Similarly, the peroxide coupling only occur at Au(111) and Au(100) surfaces, not on the Ag(111) surface.^26,82

Spectroscopy computation

In the DFT study, spectroscopy computation is another important part in on-surface chemistry of organic molecules at metal surfaces. It mainly reveals the important physical and chemical properties of reaction precursors, intermediate states and final products. Here, four representative works will be discussed as examples.

As shown in Fig. 16a, the DFT could provide the DOS spectroscopy and simulate the LDOS maps of an on-surface synthesized graphene nanoribbon heterojunctions, which agrees well with their STS and dI–dV maps in experimental observations.⁷¹ The influence of different tip's heights on the LDOS map of a specific state can further be considered, due to the calculated electron potential energy difference on the edge and inner atoms. Such calculations are mainly performed by DFT in the LDA implemented in the Quantum Espresso package.

Fig. 16 Spectroscopy computations. (a) DFT calculated DOS spectroscopy with four molecular specific orbital simulated images (a-i), and theoretical potential energy difference as a function of height h with three simulated images on a specific molecular orbital (2) at different heights (a-ii), ref. 71. (b) STM induced luminescence spectrum from a H₂Pc molecule with an inset STM current image (b-i), DFT optimized two tautomers 1 and 2 with their hyper-resolved fluorescence maps (HRFMs) and corresponding simulated maps (b-ii), ref. 51. (c) Experimental XPS measured on the polymerized Ex-TEB network with two component fitting (c-i), and DFT-based simulation of the XPS line shape of the covalent (black solid line) and the organometallic (red dashed line) TEB dimers with their respectively DFT-optimized adsorption geometries (c-ii), ref. 5. (d) Computed harmonic/anharmonic and experimental IR and Raman spectra of thymine in the range of 3900–100 cm⁻¹, ref. 83. Reproduced from ref. 5 and 71 with permission from Nature Publishing Group, copyright © 2012, 2020, 2015 and reproduced from ref. 83 with permission from Royal Society of Chemistry, copyright © 2014.

In addition, the calculated LDOS maps of an H₂Pc molecule can also be used to examine its different tautomers 1 and 2 with their hyper-resolved fluorescence maps (HRFMs), as shown in Fig. 16b.⁵¹ It is well known that the STM tip can induce such H₂Pc molecule fluorescence emission, as shown in Fig. 16b-i. The inset STM current map shows its high-resolution image, and STM-induced luminescence is performed on the white dot marked lobe. The spectrum shows multi-peak features (Q_x1, Q_x2, Q_y1, and Q_y2). The STM-induced fluorescence photo-maps on each peak agree well with the corresponding calculated LDOS map of the corresponding state. All these data strongly support the two-tautomer fact of the H₂Pc molecule.

Furthermore, DFT could also be applied to simulate XPS by simulated core-level shifts (comparing total energy differences between core-ionized and ground state systems). Here, a separate calculation was carried out for every atom for which the core-level shift was computed. Thus, if the reaction products of triple bonds are organometallic, then the simulated XPS (red line, in Fig. 16c-ii) should carry an obvious shoulder peak at the lower energy side. This point is used to compare with the experimental XPS data (Fig. 16c-i), proving the homo-coupling of terminal alkynes.⁵ It should be noted that in a recent work, they declared that the covalent linkage of terminal alkynes of the 1,3,5-tris-(4-ethynylphenyl)benzene (Ext-TEB) molecule on the Ag(111) surface is essential Enyne bonding.⁸³

Of course, DFT could also compute the organic molecule's corresponding Raman^53,84 and IR⁸⁵ spectra. Here, we take the computed harmonic/anharmonic and experimental IR and Raman spectra of nucleic acid base thymine molecule in the range of 3900–100 cm⁻¹ as an example. Obviously, Fig. 16d shows that (1) all high-wavenumber transitions are present in both IR and Raman spectra, and the intensity distribution is significantly different between both spectra. Computed IR and Raman anharmonic spectra agree very well with their experimental spectroscopy for both the band positions and the intensity pattern. (2) It is also clear that several observed transitions are missing in the harmonic spectra, and these drawbacks cannot be resolved by simply scaling the computed harmonic wavenumbers.

Critical evaluation of DFT simulations

It is indeed that experimental workers, who are not familiar with computations, might be misled into believing that DFT results are quite good and quite trustable. However, this is not the case, because it is quite functionally dependent and parameter sensitive. Some kinds of calibrations, and critical evaluation of the error margins of DFT results, in particular for solid-state materials and processes need to be done. In the author's aspect, a detailed comparison between experimental data and DFT computations is always necessary to examine the theoretical model. Besides, a sharp insight into the DFT computation is also important for experimental scientists to understand the experimental observations.

5. Conclusions and outlook

In summary, we have introduced recent advances in molecular candidates, various in situ spectroscopy techniques, and DFT studies of organic reactions on metal surfaces. The catalytic role of the underlying metal surface has been revealed. All these advances help develop on-surface chemistry in the future.

Considering the remarkable achievements in the on-surface chemistry area, a bright outlook can be expected in the following three aspects: (1) new chemical fundamental knowledge may be discovered, which may be further applied in solution chemistry and industries, (2) more and more in situ spectroscopy technologies may be introduced into the on-surface chemistry area, and these technologies can be combined. For example, STM-TERS and XPS technologies may be applied together with STM and nc-AFM to study the small molecule H₂O, O₂, CO, and CO₂ chemical conversion processes at surfaces, (3) DFT computations on various organic molecule reactions on a surface may also help the theory development, some common issues such as reaction path selectivity, the conformational role in reactions, and the crosstalk of different reactive groups may be designed as new tuning functions or parameters in calculations.

Despite the future being bright, the road is tortuous. Hopefully, this review will shine light on the scientific staff in this area.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was supported by the NSFC under Grant No. 21972104 and Grant No. 22372119, the “1000-Youth Talents Plan” and the Fundamental Research Funds for the Central Universities. We also thank the Haihe Laboratory of Sustainable Chemical Transformations for financial support.

References

1. Q. Shen, H.-Y. Gao and H. Fuchs, Nano Today, 2017, 13, 77–96.
2. S. Clair and D. G. de Oteyza, Chem. Rev., 2019, 119, 4717–4776.
3. L. Grill, M. Dyer and L. Lafferentz, et al. , Nat. Nanotechnol., 2007, 2, 687–691.
4. J. Cai, P. Ruffieux and R. Jaafar, et al. , Nature, 2010, 466, 470–473.
5. Y. Q. Zhang, N. Kepčija and M. Kleinschrodt, et al. , Nat. Commun., 2012, 3, 1286.
6. H.-Y. Gao, H. Wagner and D. Zhong, et al. , Angew. Chem., Int. Ed., 2013, 52, 4024–4028.
7. Q. Sun, C. Zhang and Z. Li, et al. , J. Am. Chem. Soc., 2013, 135, 848–8451.
8. H.-Y. Gao, P. A. Held and M. Knor, et al. , J. Am. Chem. Soc., 2014, 136(96), 58–9663.
9. D. Zhong, J.-H. Franke and S. K. Podiyanachari, et al. , Science, 2011, 334, 213–216.
10. M. Knor, H.-Y. Gao and S. Amirjalayer, et al. , Chem. Commun., 2015, 51, 10854–10857.
11. Q. Li, B. Yang and J. Björk, et al. , J. Am. Chem. Soc., 2018, 140, 6076–6082.
12. Q. Li, B. Yang and H. Lin, et al. , J. Am. Chem. Soc., 2016, 138, 2809–2814.
13. S. Wang, Q. Sun and O. Gröning, et al. , Nat. Chem., 2019, 11, 924–930.
14. L. Liu, H. Klaasen and M. C. Börner, et al. , Nat. Chem., 2021, 13, 350–357.
15. H.-Y. Gao, P. A. Held and S. Amirjalayer, et al. , J. Am. Chem. Soc., 2017, 139, 7012–7019.
16. S. Weigelt, C. Busse and C. Bombis, et al. , Angew. Chem., Int. Ed., 2007, 46, 9227–9230.
17. X. Meng, H. Klaasen and L. Viergutz, et al. , Angew. Chem., Int. Ed., 2021, 60, 1458–1464.
18. N. A. A. Zwaneveld, R. Pawlak and M. Abel, et al. , J. Am. Chem. Soc., 2008, 130, 6678–6679.
19. M. Matena, T. Riehm and M. Stöhr, et al. , Angew. Chem., Int. Ed., 2008, 47, 2414–2417.
20. M. Treier, N. V. Richardson and R. Fasel, J. Am. Chem. Soc., 2008, 130, 14054–14055.
21. B. Yang, J. Björk and H. Lin, et al. , J. Am. Chem. Soc., 2015, 137, 4904–4907.
22. D.-Y. Li, S.-W. Li and Y.-L. Xie, et al. , Nat. Commun., 2019, 10, 2414.
23. D.-Y. Li, Y. Wang and X.-Y. Hou, et al. , Angew. Chem., Int. Ed., 2022, 61, e202117714.
24. H.-Y. Gao, J.-H. Franke and H. Wagner, et al. , J. Phys. Chem. C, 2013, 117, 18595–18602.
25. D. Barton, H.-Y. Gao and P. A. Held, et al. , Chem. – Eur. J., 2017, 23, 6190–6197.
26. P. A. Held, H.-Y. Gao, L. Liu and A. Studer, et al. , Angew. Chem., Int. Ed., 2016, 55, 9777–9782.
27. L. Liu, H. Klaasen and A. Timmer, et al. , J. Am. Chem. Soc., 2018, 140, 6000–6005.
28. H.-Y. Gao, M. Šekutor and L. Liu, et al. , J. Am. Chem. Soc., 2019, 141, 315–322.
29. K. Feng, E. Solel and P. R. Schreiner, et al. , J. Phys. Chem. Lett., 2021, 12, 3468–3475.
30. B. Yang, K. Niu and N. Cao, et al. , Angew. Chem., Int. Ed., 2023, 135, e202218211.
31. S. Kawai, O. Krejčí and T. Nishiuchi, et al. , Sci. Adv., 2020, 6, eaay8913.
32. L. Xie, H. Jiang and D. Li, et al. , ACS Nano, 2020, 14, 10680–10687.
33. L. Xie, Y. Ding and D. Li, et al. , J. Am. Chem. Soc., 2022, 144, 5023–5028.
34. Y.-Q. Zhang, T. Paintner and R. Hellwig, et al. , J. Am. Chem. Soc., 2019, 141, 5087–5091.
35. C. Zhang, E. Kazuma and Y. Kim, J. Am. Chem. Soc., 2022, 144, 10282–10290.
36. J. Zhou, J. Pan and Y. Jin, et al. , J. Am. Chem. Soc., 2022, 144, 8430–8433.
37. J. Pan, X. Li and Y. Zhu, et al. , J. Am. Chem. Soc., 2023, 145, 18748–18752.
38. Y. Ling, J. Luo and Y. Ran, et al. , J. Am. Chem. Soc., 2023, 145, 22697–22707.
39. W. Ran, A. Walz and K. Stoiber, et al. , Angew. Chem., Int. Ed., 2022, 61, e202111816.
40. S. Mahapatra, J. F. Schultz and L. Li, et al. , J. Am. Chem. Soc., 2022, 144, 2051–2055.
41. S. Kawai, O. Krejčí and A. S. Foster, et al. , ACS Nano, 2018, 12, 8791–8797.
42. K. Sun, K. Sagisaka and L. Peng, et al. , Angew. Chem., Int. Ed., 2021, 60, 19598–19603.
43. K. Sun, O. J. Silveira and Y. Ma, et al. , Nat. Chem., 2023, 15, 136–142.
44. K. Ma, T. Zhang and Y. Qin, et al. , J. Am. Chem. Soc., 2022, 144, 8789–8796.
45. N. Cao, J. Björk and E. Corral-Rascon, et al. , Nat. Chem., 2023, 15, 1765–1772.
46. D. J. Rizzo, G. Veber and T. Cao, et al. , Nature, 2018, 560, 204–208.
47. O. Gröning, S. Wang and X. Yao, et al. , Nature, 2018, 560, 209–213.
48. L. Lafferentz, F. Ample and H. Yu, et al. , Science, 2009, 323, 1193–1197.
49. J. N. Ladenthin, T. Frederiksen and M. Persson, et al. , Nat. Chem., 2016, 8, 935–940.
50. M. C. Chong, N. Afshar-Imani and F. Scheurer, et al. , Nano Lett., 2018, 18, 175–181.
51. B. Doppagne, T. Neuman and R. Soria-Martinez, et al. , Nat. Nanotechnol., 2020, 15, 207–211.
52. S. Jiang, T. Neuman and A. Boeglin, et al. , Science, 2023, 379, 1049–1053.
53. R. Zhang, Y. Zhang and Z. C. Dong, et al. , Nature, 2013, 498, 82–86.
54. J. Lee, K. T. Crampton and N. Tallarida, et al. , Nature, 2019, 568, 78–82.
55. J. Xu, X. Zhu and S. Tan, et al. , Science, 2021, 371, 818–822.
56. X. Zhu, J. Xu and Y. Zhang, et al. , J. Am. Chem. Soc., 2023, 145, 13839–13845.
57. C. Zhang, R. B. Jaculbia and Y. Tanaka, et al. , J. Am. Chem. Soc., 2021, 143, 9461–9467.
58. L. Liu, A. Timmer and E. Kolodzeiski, et al. , Nanoscale Adv., 2021, 3, 6373–6378.
59. G. Q. Lu, A. Linsebigler and J. T. Yates, J. Chem. Phys., 1995, 102, 4657–4662.
60. S. Wendt, P. T. Sprunger and E. Lira, et al. , Science, 2008, 320, 1755–1759.
61. S. Tan, Y. Ji and Y. Zhao, et al. , J. Am. Chem. Soc., 2011, 133, 2002–2009.
62. S. Tan, H. Feng and Q. Zheng, et al. , J. Am. Chem. Soc., 2020, 142, 826–834.
63. Q. Fu, W.-X. Li and Y. Yao, et al. , Science, 2010, 328, 1141–1144.
64. Y. Ding, X. Wang and D. Li, et al. , ACS Nano, 2019, 13, 6025–6032.
65. J. Dong, P. Yang and C. Zhang, et al. , J. Phys. Chem. C, 2022, 126, 3756–3763.
66. X. Meng, L. Liu and F. García, et al. , J. Phys. Chem. C, 2018, 122, 6230–6235.
67. Y. Wang, K. Ma and T. Zhang, et al. , J. Phys. Chem. C, 2023, 127, 1381–1387.
68. P. Ruffieux, J. Cai and N. C. Plumb, et al. , ACS Nano, 2012, 6, 6930–6935.
69. F. Xiang, S. Maisel and S. Beniwal, et al. , Nat. Chem., 2022, 14, 871–876.
70. J. Cai, C. A. Pignedoli and L. Talirz, et al. , Nat. Nanotechnol., 2014, 9, 896–900.
71. Y.-C. Chen, T. Cao and C. Chen, et al. , Nat. Nanotechnol., 2015, 10, 156–160.
72. X. Peng, H. Mahalingam and S. Dong, et al. , Nat. Commun., 2021, 12, 5895.
73. S. Mishra, X. Yao and Q. Chen, et al. , Nat. Chem., 2021, 13, 581–586.
74. Y. Zhao, K. Jiang and C. Li, et al. , Nat. Chem., 2023, 15, 53–60.
75. Y. Sugimoto, P. Pou and M. Abe, et al. , Nature, 2007, 446, 64–67.
76. H. Mönig, D. R. Hermoso and O. D. Arado, et al. , ACS Nano, 2016, 10, 1201–1209.
77. J. Qi, Y. Gao and H. Jia, et al. , J. Am. Chem. Soc., 2020, 142, 10673–10680.
78. J. Björk, F. Hanke and S. Stafström, J. Am. Chem. Soc., 2013, 135, 5768–5775.
79. J. Ren, M. Cnudde and D. Brünink, et al. , Angew. Chem., Int. Ed., 2019, 58, 15396–15400.
80. J. Wang, K. Niu and C. Xu, et al. , J. Am. Chem. Soc., 2022, 144, 21596–21605.
81. H. Kong, S. Yang and H.-Y. Gao, et al. , J. Am. Chem. Soc., 2017, 139, 3669–3675.
82. H. Klassen, L. Liu and X. Meng, et al. , Chem. – Eur. J., 2018, 24, 15303–15308.
83. Y. Lyu, F. Gao and P. Cheng, et al. , Adv. Mater. Interfaces, 2024, 2400222.
84. W. Hu, S. Duan and Y. Luo, Wiley Interdiscip. Rev.: Comput. Mol. Sci., 2017, 7, e1293.
85. V. Barone, M. Biczysko and J. Bloino, Phys. Chem. Chem. Phys., 2014, 16, 1759.

---