Introduction

In recent years, on-surface synthesis under ultrahigh vacuum (UHV) conditions has emerged as a very successful method to produce extended, covalently bonded macro-molecules, which are inaccessible to wet-chemical synthesis due to insolubility or high reactivity1. On-surface synthesis is based on the controlled reaction of dedicated precursor molecules on well-defined single-crystal surfaces mostly by thermal activation. Although in recent years, the catalog of on-surface reactions and synthesized structures broadened extensively, the focus rarely lay on enantioselective synthesis, which is of immense importance for instance in pharmaceutical, agricultural, or food industry2,3. This lack of enantioselective on-surface synthesis originates largely from the scarcity of intrinsically chiral, catalytically active, and well-characterized single-crystal surfaces needed for this task. One option to circumvent this deficiency is to render achiral surfaces chiral via the adsorption of enantiopure molecules4,5,6,7,8,9,10. This approach has proven very successful under ambient conditions, for instance, in the asymmetric hydrogenation of activated ketones and asymmetric hydrogenation of prochiral olefins11.

Intrinsically chiral metal surfaces promise an increased thermal stability, a reduction of complexity arising from the multitude of molecule–molecule arrangements and interactions, and would not require enantiopure molecular modifiers, but are accompanied by a reduced adjustability towards specific reactions. In this context, high Miller-index surfaces of achiral crystals12,13, which exhibit a low density of chiral centers only at the kink sites of atomic steps separating adjacent terraces of low-index surfaces, have been established for enantioselective decomposition reactions under UHV conditions14,15.

Low Miller-index surfaces of intrinsically chiral bulk crystals possess a high density of well-defined and thermally stable chiral centers. Even though chiral metallic crystals gain increasing attention in the context of topological electronic properties16,17,18,19,20,21,22,23, in-depth characterization of their surface structure is very limited. Currently, the only intrinsically chiral and catalytically active single crystal whose low-index surfaces are well-characterized is intermetallic PdGa. PdGa belongs to the non-centrosymmetric space-group P213 (ref. 24), and thus exists in two enantiomorphs, denoted as PdGa:A and PdGa:B25. Here, we focus on the two structurally dissimilar, bulk-truncated, threefold symmetric PdGa:A(\(\bar 1\bar 1\bar 1\)) and PdGa:A(111) surfaces, of which the former is terminated by isolated Pd trimers, further referred to as A:Pd3 (Fig. 1a), while the top layer of the latter consists of single, isolated Pd atoms, and is denoted A:Pd1 (Fig. 1b)26. Owing to their differing surface terminations, in combination with their equal lattice parameters, identical symmetry, and similar electronic properties, the PdGa{111} surfaces are ideally suited to disentangle ensemble and ligand effects, i.e., the influence of the local geometric and electronic properties, respectively, in asymmetric heterogeneous catalysis27,28,29,30.

Fig. 1: Schematics of the PdGa{111} surfaces and the azide–alkyne Huisgen cycloaddition.
figure 1

The atomic surface structure of a PdGa:A(\(\bar 1\bar 1\bar 1\))Pd3 (first layer Pd3 in bright blue: z = 0 pm; second layer Ga3 in red: z = −85 pm; third layer Pd1 in dark blue: z = −161 pm) and b PdGa:A(111)Pd1 (first layer Pd1 in bright blue: z = 0 pm; second layer Ga3 in red: −57.4 pm; third layer Pd3 in dark blue: −149.8 pm) with their chirality highlighted by a top layer Pd trimer or a single top layer atom, respectively, and their neighbors in saturated colors. c General reaction scheme of the azide–alkyne Huisgen cycloaddition. d The investigated on-surface Huisgen cycloaddition between 9-ethynylphenanthrene (9-EP) and 3-(4-azidophenyl)propionic acid (APA) yielding 1,4-triazole regioisomers.

Full size image

Chirality transfer from the PdGa{111} surfaces onto molecular processes has been demonstrated with acetylene (C2H2; achiral) and 9-ethynylphenanthrene (9-EP; prochiral, thus appearing in two distinguishable surface enantiomers R and S when confined to a planar configuration)31,32. Specifically, C2H2 has been reported to exhibit directed rotation on the Pd3-terminated PdGa{111} surfaces31. On the other hand, for 9-EP, enantiopure trimerization from an initial racemic mixture of monomers on the Pd3- and enantioselective adsorption of individual monomers on the Pd1-terminated PdGa{111} surfaces have been shown32.

Therefore, 9-EP promotes itself as prototypical alkyne precursor for investigating enantioselective synthesis on the PdGa{111} surfaces. Moreover, on-surface azide–alkyne Huisgen cycloaddition (Fig. 1c) without emphasis on enantioselectivity has been successfully demonstrated on Au(111)33, and, using 9-EP as precursor, on Cu(111) under UHV conditions by Bebensee et al.34. This catalytically activated reaction, which belongs to the class of Click Chemistry, selectively yields 1,4-triazole regioisomers and omits the formation of 1,5-triazoles. According to the d-band model introduced by Liu and Nørskov28 and Nørskov et al.35, PdGa is expected to possess similar catalytic activity for the azide–alkyne Huisgen cycloaddition as copper. To perform the reaction, we chose commercially available prochiral 3-(4-azidophenyl)propionic acid (APA) as azide reactant (Fig. 1d).

Using scanning tunneling microscopy (STM), complemented with X-ray photoelectron spectroscopy (XPS), and density functional theory (DFT) calculations, we demonstrate a significant ensemble effect on the PdGa{111} surface reactivity, manifested in the occurrence of the regio- and, most importantly, enantioselective cycloaddition between APA and 9-EP on the Pd1-terminated PdGa{111} surfaces, but its suppression on Pd3-terminated PdGa{111}.

Results and discussion

Azide–alkyne Huisgen cycloaddition on Pd3-terminated PdGa{111} surfaces

A series of STM images of the individual deposition of 9-EP and APA molecules and the co-deposition of both molecules on A:Pd3 at room temperature (RT) and after subsequent annealing to 425 and 515 K is shown in Fig. 2. As previously reported32, 9-EP deposited at RT results in a homogeneous racemate of well isolated monomers (Fig. 2a) and formation of 9-EP trimers with increasing enantiomeric excess (ee) up to 97 ± 2% towards RRR enantiomorphs for increasing temperatures (Fig. 2b, c). On the other hand, APA molecules deposited on A:Pd3 at RT create large, disordered agglomerates (Fig. 2d). Additionally, depressions, which are absent on the pristine surface, are observed on the substrate, as indicated by the white arrow. We attribute these depressions to fragments of the decomposed carboxylic acid group of the APA molecule as they are also observed upon deposition of ex situ synthesized 1,4-triazoles (Fig. S1) and XPS investigations show a strong C 1s component arising from the carboxylic acid group (Fig. S2a, Table S1 and Supplementary Note 1). On the other hand, the ratio between the C 1s and N 1s XPS intensity is around three times larger than the one expected for pristine APA (Fig. S3 and Table S4). Hence also most azide groups decompose, and desorb from the surface. Subsequent annealing of the sample up to 515 K leads to a reduction of the depressions, as they desorb, while agglomeration proceeds without ever forming discernible regular structures, i.e., reoccurring structures of similar topographic signature in STM images (Fig. 2d–f).

Fig. 2: Suppressed azide–alkyne Huisgen cycloaddition on the Pd3-terminated PdGa{111} surfaces.
figure 2

STM images (VB = 20 mV; IT = 1 nA) of 9-EP a deposited on PdGa:A(\(\bar 1\bar 1\bar 1\))Pd3 at RT, and after subsequent annealing to b 425 K and c 515 K. STM images (VB = 200 mV; IT = 0.1 nA) of APA molecules on the PdGa:A(\(\bar 1\bar 1\bar 1\))Pd3 surface after d RT deposition, subsequent annealing to e 425 K and f 515 K. STM images (VB = 200 mV; IT = 0.1 nA) of APA co-adsorbed with 9-EP after g RT deposition, subsequent annealing to h 425 K and i 515 K. High-resolution STM image (3 × 3 nm2; VB = 20 mV; IT = 0.2 nA) of the frequently observed molecular structure containing one APA and two 9-EP molecules, which is marked with blue arrows in h, is shown in j and overlaid with the molecular structure in k. The green circular arrows in b, c, h, i highlight the homochiral 9-EP propeller, the white arrows in d, e, g the depression caused by APA remnants, and the green arrows in g a 9-EP molecule interacting with an APA remnant. The blue dashed circles in g, h, i indicate APA agglomerations and the green circle in h a 9-EP monomer.

Full size image

When APA is co-deposited with 9-EP on A:Pd3, no intermolecular reactions are observed at RT. While APA again agglomerates, 9-EP seems to mostly bond to the previously mentioned molecular fragments detached from APA molecules (depressions), as pointed out by the green arrow in Fig. 2g. Upon annealing to 425 K (Fig. 2h), 9-EP appears in its pristine adsorption configuration, which means it is detached from the molecular fragments, and, importantly, new prochiral 9-EP/APA heterostructures that consist of two 9-EP and one APA molecule are formed (Fig. 2j–k). The adsorption configuration of 9-EP molecules incorporated in these trimer heterostructures with respect to the PdGa substrate is identical with that of R enantiomers incorporated into pure 9-EP trimers32. The heterostructure trimers exclusively appear in the enantiomeric form shown in Fig. 2j, k on the A:Pd3 termination. The incorporated APA is not covalently bound in these heterostructure trimers, as, upon further annealing to 515 K, these structures deplete and homostructural 9-EP trimers become the only regular motive (highlighted by a green circular arrow in Fig. 2i). Therefore, even though PdGa is copper-like with regard to the catalytic reactivity according to the d-band model35, the azide–alkyne Huisgen cycloaddition cannot be triggered with any significant yield on the A:Pd3 surface.

Azide–alkyne Huisgen cycloaddition on Pd1-terminated PdGa{111} surfaces

On the A:Pd1 surface, 9-EP deposited at RT adsorbs with an ee of 98% in favor of the R enantiomer (cf. Fig. 3a), and dimerizes without ee above 400 K, as reported by Prinz et al.36. The adsorption behavior of APA on the Pd1-terminated PdGa{111} surfaces is in stark contrast to that on A:Pd3, as the APA molecules barely agglomerate even at temperatures up to 415 K (Fig. 3b and Fig. S5). The most frequently observed APA structures are shown in the STM images in Fig. 3c–f. The sickle-like molecular structures in Fig. 3c, e strongly resemble the terminal phenylpropionic acid group of the ex situ synthesized 1,4-triazoles adsorbed on PdGa:A(111)Pd1 in Fig. S6. These two structures are attributed to pristine APA molecules, whereas their azide group only exhibits a very weak STM signal like on the less corrugated Ag(111)37. The coverage of the molecular structures in Fig. 3c, e relative to all APA derivatives on the Pd1-terminated PdGa{111} surfaces amounts to around 30–35%, upon deposition and after annealing to 415 K (Fig. S5), which is in good agreement with the maximum amount of pristine APA molecules estimated from XPS investigations (Fig. S4 and Table S5). The molecular structures in Fig. 3d, f are APA molecules with decomposed azide group. On Pd1-terminated PdGa{111} surfaces, part of the APA propionic acid groups are deprotonated (Fig. S2b and Table S2), but no signs of their detachment from APA molecules could be identified, neither with XPS nor STM.

Fig. 3: Alkyne and azide on the Pd1-terminated PdGa{111} surfaces.
figure 3

Large-scale STM images of a 9-EP (VB = 20 mV; IT = 2 nA) and b APA (VB = 100 mV; IT = 0.4 nA) deposited at RT on PdGa:A(111)Pd1. c–f High-resolution STM images (VB = 20 mV; IT = 0.2 nA) of the different APA configurations on PdGa:A(111)Pd1 with the molecular structure of APA in c.

Full size image

After having clarified that around 30% of the APA monomers remain structurally intact up to 415 K on the Pd1-terminated PdGa{111} surfaces, APA was co-deposited with 9-EP at RT (Fig. 4). As shown in Fig. 4a, both 9-EP and APA mainly appear as non-interacting monomers. The pronounced depressions on this surface represent Pd vacancies in the Pd1-termination26 and the apparent trimer is composed of individual 9-EP molecules adsorbed on such a vacancy36. 9-EP molecules appear with a relative abundance between 40 and 60% and exhibit an ee > 90%. Due to the lack of significant intermolecular interactions, also APA molecules occupy the same adsorption sites as in the case when they are adsorbed without 9-EP (cf. Fig. 3).

Fig. 4: Enantioselective azide–alkyne Huisgen cycloaddition on the Pd1-terminated PdGa{111} surfaces.
figure 4

STM images (VB = −200 mV; IT = 0.05 nA) of a co-deposited APA (white arrow) and 9-EP (yellow arrow and γ-states labeled) molecules on PdGa:B(\(\bar 1\bar 1\bar 1\))Pd1, and after subsequent annealing to b 425 K. In b, the 9-EP dimers are pointed out with blue arrows, and the 1,4-triazoles with green ones. The different 1,4-triazoles are shown in detail in the high-resolution STM images (VB = −200 mV; IT = 0.2 nA) in c–g. d and g have been rotated to highlight the relation of the shown 1,4-triazole to the one in e and f, respectively.

Full size image

The cycloaddition between APA and 9-EP can be triggered thermally on Pd1-terminated PdGa{111} by annealing the sample with co-deposited molecules at 425 K. This gives rise to the observation of several new covalently coupled molecular structures (Fig. 4b) in high abundance (Figs. S9 and S10 and Supplementary Note 2). About 5% of the new molecular structures are previously reported 9-EP dimers36 (blue arrow in Fig. 4b), whereas 5% consist of one 9-EP molecule and the phenylpropionic acid moiety of APA, but are too small to be intact triazoles (Fig. S11). The vast majority (90%, green arrows Fig. 4b) of these molecular structures exhibit one of the six different STM appearances presented in Fig. 4c–g. Their STM signature is mainly planar in the vicinity of the coupling site, which excludes them to be 1,5-triazoles, as they would exhibit highly non-planar STM signatures according to DFT calculations done with CP2K38 within the AiiDalab platform39 (Fig. S12).

To check whether the six molecular structures in Fig. 4c–g are on-surface synthesized 1,4-triazoles, we deposited ex situ synthesized 1,4-triazoles on A:Pd1 held at 170 and 300 K. When imaged with STM, the ex situ synthesized 1,4-triazoles appear in three inequivalent adsorption configurations (Fig. S6). These 1,4-triazole molecules strongly resemble the reaction products formed between APA and 9-EP on the Pd1-terminated PdGa{111} in terms of shape, dimensions, and adsorption geometry (Fig. S7), but differ in the intensity distribution. However, upon annealing the ex situ synthesized 1,4-triazoles to the reaction temperature of the on-surface synthesis, we observe molecular structures that are identical to those formed between APA and 9-EP on the Pd1-terminated PdGa{111} surfaces (Fig. S8). This implies that the vast majority of the heterocoupled reaction products are indeed 1,4-triazoles, but with deprotonated carboxylic acid. Hence, the azide–alkyne Huisgen cycloaddition proceeds regioselectively on Pd1-terminated PdGa{111} surfaces.

Asymmetric azide–alkyne Huisgen cycloaddition on Pd1-terminated PdGa{111} surfaces

It becomes clear that the two molecules in Fig. 4c, the two in Fig. 4d, e and the ones in Fig. 4f, g are mirror images of each another, thus they represent complementary enantiomers, which will be discussed later in view of enantioselectivity. The distinctive feature in the STM signatures of the R and S enantiomers in all three cases is the arrangement of the phenylpropionic acid with respect to the phenanthrene moiety. If the phenylpropionic acid unit is perpendicular (almost parallel) to the phenanthrene moiety, the 1,4-triazoles appear L-shaped (Z-shaped) (Fig. 4c–e, the triazoles are denoted P1 and P2, respectively) or I-shaped (P3) for an intermediate orientation (Fig. 4f, g). The atomic resolution of the substrate in the STM images combined with the structure overlay allows to determine the adsorption configuration of these reaction products (Fig. S13b, e, h), whereas the enantiospecific arrangement of the propionic acid moiety remains ambiguous. The relative abundance of the three product structures P1–P3 are summarized in Table 1 for the two enantiomeric forms R and S on the Pd1-terminated surfaces (the products are correspondingly labeled RP1, SP1, RP2,…, SP3). Regarding the arrangement of the triazole with respect to the phenanthrene moiety in the 1,4-triazoles, P2 and P3 belong to the same diastereomer. Hence the reaction proceeds with a diastereomeric excess \({\mathrm{{de}}} = 100\% \ast \frac{{{\mathrm{P}}2 + {\mathrm{P}}3 - P1}}{{{\mathrm{{P2 + P3 + P1}}}}}\) of up to 50%. Considering all 1,4-triazoles, the 9-EP moiety appear as racemic mixture, which implies that at the reaction temperature of 425 K, the chiral recognition of 9-EP by the Pd1-terminated PdGa{111} surfaces has become insignificant. Unlike the Z-shaped products P2, which do not experience any chirality transfer, enantioselectivity is clearly expressed for the L- and I-shaped products P1 and P3, where we find ratios between the different enantiomorphs of up to 4:1 and 5:1, respectively. This enantioselective behavior is supported by deposited ex situ synthesized 1,4-triazoles pointing towards the same preferential arrangement (cf. Table S6). The opposite ratio between the different P1 and P3 enantiomers for Pd1-terminated PdGa{111} surfaces of the A and B PdGa crystal enantiomorph strongly corroborates the conclusion of their enantioselective recognition towards the 1,4-triazoles.

Table 1 Quantification of the reaction products.
Full size table

Compared to investigations of the azide–alkyne Huisgen cycloaddition on Cu(111)34, where the reaction yield is reported to be of the order of 2–10%, the formation of 1,4-triazoles on the Pd1-terminated PdGa{111} surfaces reveals a much higher yield of up to 58%. It seems as if this yield depends mainly on the availability of intact APA molecules as the yield increases with increasing ratio between APA/9-EP (Table S7), which is expected in light of the decomposition of the azid group of APA molecules upon deposition on the Pd1-terminated PdGa{111} surfaces.

Conclusion

In summary, we have shown that the azide–alkyne Huisgen cycloaddition between pristine 9-EP and APA occurs regio- and enantioselectively on the Pd1-terminated PdGa{111} surfaces, i.e., forming exclusively 1,4-triazoles with an exceptionally high enantiospecifity of up to 5:1. In view of the d-band model, the Pd1-terminated PdGa{111} surfaces behave copper-like. On the other hand, the same reaction does not proceed on the Pd3-terminated PdGa{111} surfaces, even though these PdGa{111} surfaces exhibit very similar electronic d-band structure. We therefore conclude that the vastly dissimilar catalytic properties of the Pd1- and Pd3-terminated PdGa{111} surfaces for the azide–alkyne Huisgen cycloaddition primarily arise from differences in the atomic arrangement in their terminating layers, i.e., the ensemble effect.

Methods

Synthesis of 3-(4-(4-(phenanthren-9-yl)-1H-1,2,3-triazol-1-yl)phenyl)propionic acid

Chemicals and solvents were purchased from commercial sources (Merck, VWR) and were used without further purification. Reactions were carried out under argon atmosphere. Column chromatography was performed using silica gel (pore size 40–63 μm, Normasil 60® from VWR chemicals).

9-Ethynylphenanthrene (75 mg, 0.37 mmol), 3-(4-azidophenyl)propinoic acid (57 mg, 0.30 mmol), tris((1-benzyl-1H-1,2,3-triazol-4-yl)methyl)amine (11 mg, 0.02 mmol), copper(II) sulfate (7 mg, 0.04 mmol), and sodium ascorbate (18 mg, 0.09 mmol) were dissolved in 6 ml N,N-dimethylformamide and 2 ml water. The reaction mixture was stirred for 6 h at room temperature. Additional water was added and the product was extracted with ethyl acetate. The organic fractions were washed with water and brine and dried with anhydrous sodium sulfate. The crude product was purified by flash column chromatography with silica gel and a solvent mixture of dichloromethane/ethyl acetate/acetic acid. The triazole was precipitated from dichloromethane/n-heptane and isolated as a white crystalline solid (85 mg, 0.22 mmol, 72%).

Nuclear magnetic resonance

1H and 13C NMR data were recorded at 400.2 and 100.6 MHz using a 5 mm CryoProbe™ Prodigy probe equipped with z-gradient on a Bruker Avance III 400 NMR spectrometer (Bruker Biospin AG, Fällanden, Switzerland). 1D 1H and 13C NMR experiments as well as 2D-correlated 1H-13C HSQC, 1H-13C HMBC, and 1H-1H DQF-COSY experiments were performed at 298 K using the Bruker standard pulse programs and parameter sets applying 90° pulse lengths of 11.4 µs (1H) and 10.0 µs (13C). Chemical shifts (δ in ppm) were calibrated to residual solvent peaks of CDCl3 at δ = 7.26 and 77.0 ppm. Coupling constants J are reported in Hz and 1H NMR multiplicities are described as s = singlet, d = doublet, t = triplet, m = multiplet, and dd = doublet of doublet. High-resolution mass spectrometry was recorded on a Bruker Daltonics maXis II ESI-QTOF mass spectrometer in positive mode.

1H NMR (CDCl3, 400.2 MHz): δ 8.73 (d, J = 8.2 Hz, 1H), 8.45 (dd, J = 8.2, 1.0 Hz, 1H), 8.28 (s, 1H), 8.08 (s, 1H), 7.94 (d, J = 7.1 Hz, 1H), 7.82 (d, J = 8.5 Hz, 2H), 7.75–7.68 (m, 2H), 7.68–7.61 (m, 2H), 7.45 (d, J = 8.5 Hz, 2H), 3.09 (t, J = 7.6 Hz, 2H), 2.78 (t, J = 7.6 Hz, 2H).

13C NMR (CDCl3, 100.6 MHz): δ 174.4, 147.5, 141.2, 135.5, 131.3, 130.7, 130.6, 130.3, 130.1, 129.8, 129.0, 128.8, 128.6, 127.2, 127.0, 127.0, 126.8, 126.2, 123.1, 122.6, 120.9, 34.6, 30.2.

HR-MS (pos. ESI): m/z for [C25H20N3O2+]; calculated 394.1550, found 394.1549.

Sample preparation and experiments

All experiments and sample preparations were performed under UHV conditions with a base pressure below 2 × 10−10 mbar.

The sample was prepared by repeated sputtering and annealing cycles (sputtering: 20 min, Ar+, 1 keV; annealing: 20 min at 870 K). Both 9-ethynylphenanthrene (9-EP; 97%) and 3-(4-azidophenyl)propionic acid (APA; ≥97%) were purchased from Sigma-Aldrich and were used without further purification. 9-EP and APA were deposited by exposing the clean sample surface to the molecules held at 300 K in a pumped glass tube separated by a gate valve from the preparation chamber to prevent contamination. 9-EP and APA were pumped via individual connections with the same turbo pump to minimize the risk of cross-contamination.

3-(4-(4-(phenanthrene-9-yl)-1H-1,2,3-triazol-1-yl)phenyl)propionic acid, further referred to as ex situ synthesized 1,4-triazole, was deposited on the clean sample surface held at 300 K from a Knudsen cell at 490 K.

STM images were acquired with a commercial low temperature STM from Scienta Omicron operated at 5 K, if not mentioned differently, in constant-current mode.

All XPS data were recorded in normal emission configuration. XPS experiments shown in Fig. S2 were performed in-house and at room temperature using a monochromatized Al Kα source and a Scienta R3000 display analyzer. XPS spectra shown in Figs. S3 and S4 were recorded at the X03DA beamline (PEARL endstation)40 of the SLS synchrotron radiation facility (PSI, Villigen, Switzerland) using linearly polarized radiation with a photon energy of 760 eV and a Scienta R4000 hemispherical electron analyzer equipped with a multichannel plate detector with the sample held at 180 K.

DFT calculations

The DFT calculations were performed using the CP2K code38 on the AiiDAlab platform39. The electronic states were expanded with a TZV2P Gaussian basis set41 for C, H, N, and O species and a DZVP basis set for Pd and Ga species. A cutoff of 1200 Ry was used for the plane wave basis set. We used Norm Conserving Goedecker-Teter-Hutter pseudopotentials42 and the PBE exchange-correlation functional43 with the D3 dispersion corrections proposed by Grimme44.

The Pd1 surface was modeled by a supercell of 34.75 × 36.11 Å2 corresponding to 30 surface units. To obtain the equilibrium geometries, we kept the atomic positions of the bottom seven layers of the slab fixed to the bulk positions, and all other atoms were relaxed until forces were lower than 0.005 eV/Å. To obtain simulated STM images45 within the Tersoff-Hamann approximation46,47 we extrapolated the electronic orbitals to the vacuum region following the approach outlined by Tersoff47.