Rearrangements that change the connectivity of a carbon skeleton are often useful in synthesis, but it can be difficult to follow their mechanisms. Scanning probe microscopy can be used to manipulate a skeletal rearrangement at the single-molecule level, while monitoring the geometry of reactants, intermediates and final products with atomic resolution. We studied the reductive rearrangement of 1,1-dibromo alkenes to polyynes on a NaCl surface at 5 K, a reaction that resembles the Fritsch–Buttenberg–Wiechell rearrangement. Voltage pulses were used to cleave one C–Br bond, forming a radical, then to cleave the remaining C•–Br bond, triggering the rearrangement. These experiments provide structural insight into the bromo-vinyl radical intermediates, showing that the C=C•–Br unit is nonlinear. Long polyynes, up to the octayne Ph–(C≡C)8–Ph, have been prepared in this way. The control of skeletal rearrangements opens a new window on carbon-rich materials and extends the toolbox for molecular synthesis by atom manipulation.
Polyynes have been widely investigated as precursors to functional materials1, as molecular wires for charge transport2 and because of their nonlinear optical properties3. They are also interesting models for carbyne, the elusive one-dimensional sp-hybridized allotrope of carbon4,5. Polyynes become increasingly unstable as the number of consecutive sp-carbon atoms is increased due to their tendency to undergo cycloaddition and crosslinking reactions6. This problem can be alleviated by the presence of bulky terminal groups3,4,5,7 and by supramolecular encapsulation8,9. The work presented here was initiated with the aim of fabricating and investigating long linear and cyclic polyynes4,10 under highly controlled conditions on an inert surface.
The Fritsch–Buttenberg–Wiechell (FBW) rearrangement (Fig. 1)11,12,13,14 has been known since 1894, and it is a widely used method for acetylene synthesis, particularly as the second step of the Corey–Fuchs reaction15. Tykwinski and co-workers were the first to report the migration of alkyne groups in an FBW rearrangement16, and this has become a popular route for the synthesis of polyynes7,14,17. Early FBW rearrangements were carried out by deprotonation of a vinyl halide 1a18,19,20, but most modern examples use a 1,1-dihaloolefin 1b as the starting material14. The 1,1-dihaloolefin is typically treated with butyl lithium to generate a carbenoid intermediate 2 (M = Li) via lithium–halogen exchange, but other reducing agents can be used, including samarium(ii) iodide21 and lanthanum metal22. Despite many investigations, the mechanism of the FBW rearrangement remains unclear. In some cases there is evidence for the formation of a carbene intermediate, 3, which rearranges to alkyne 4, whereas in other cases the stereospecificity of the reaction demonstrates that the carbenoid 2 rearranges directly to the alkyne18,19,20. The reaction studied here is similar to a FBW rearrangement, in that it involves the reductive rearrangement of a 1,1-dibromoolefin to an alkyne, but it is different in that it occurs on a surface, reduction being achieved by electrons from the probe tip, whereas FBW rearrangements of 1,1-dibromoolefins are carried out in solution, typically using butyl lithium as the reducing agent.
Scanning probe microscopies, such as scanning tunnelling microscopy (STM) and atomic force microscopy (AFM), were initially developed for imaging surfaces23. Atomic manipulation developed from the ability to place individual atoms on a surface using STM24. The resolution of molecular imaging was increased substantially by using AFM with CO-functionalized tips25, which led to detailed characterization of a wide variety of molecular structures and thermally induced reactions on surfaces26,27,28,29,30,31,32,33,34,35. Atomic manipulation was extended towards triggering chemical reactions and single-molecule synthesis36. Thermally induced on-surface synthesis is usually performed on catalytically active metal surfaces, whereas synthesis by atom manipulation can also be applied to molecules on thin insulating films such as NaCl. The inert NaCl surface and low-temperature environment render it possible to stabilize and study reactive intermediates such as arynes created by atom manipulation37,38,39,40. Furthermore, because of decoupling by the insulating film, molecules can be characterized electronically using orbital imaging41. We set out to investigate an on-surface analogue of the FBW rearrangement so as to learn about the mechanism and scope of this reaction for preparing long polyynes under controlled conditions. In this Article we generate tri-, tetra-, hexa- and octaynes (5–8) by atomic manipulation on bilayer NaCl islands on Cu(111) (Fig. 2). The tip of a low-temperature qPlus-based42 STM/AFM was used to cleave the C–Br bonds of precursors 5(Br2), 6(Br4), 7(Br4) and 8(Br4) one by one using appropriate voltage pulses37,38,39,43. We resolved the structures of molecular reactants, intermediates and products by AFM with CO-functionalized tips, providing a unique opportunity to investigate the geometries of these species and revealing the details of this intramolecular migration.
Results and discussion
Initially, we explored the generation of triyne 5 from precursor 5(Br2) (Figs. 2a and 3). The structure of the dibromoolefin 5(Br2) can be clearly distinguished from its AFM image (Fig. 3a). The triple bonds27,32 and the attached Br atoms38,43 can be assigned by their pronounced and characteristic contrast. A voltage pulse at V ≥ 1.7 V was used to cleave the first C–Br bond. The AFM image of the resulting radical 5(Br)• (Fig. 3b) shows a greater adsorption height of the phenyl ring adjacent to the remaining attached Br, resulting in a difference in contrast between the two phenyl rings. The detached Br atom remains nearby, but can be moved out of the frame by vertical manipulation37,43. A higher voltage pulse (V ≥ 2.1 V) was used to cleave the second C–Br bond, resulting directly in formation of the triyne 5.
Importantly, the NaCl bilayer is chemically inert, which prevents reaction of radical intermediates with the surface37,38,39,40. The course of the reaction is very different when 5(Br2) is manipulated on a bare Cu(111) surface; here a carbene is generated after the second debromination. This carbene seems to be bound to the Cu(111) surface and we never observed rearrangement under these conditions (Supplementary Fig. 27). Similarly, on-surface annealing of a 1,1-dibromo alkene (1b with X = Br, R1 = H and R2 = biphenyl) on Au(111) monitored by STM/AFM imaging yields a cumulene without any rearrangement, as reported recently35. When the stepwise debromination was performed on several hundred individual 5(Br2) molecules on Cu(111), statistical analysis showed that the voltage threshold of the first debromination is about 0.3 V smaller than for the second debromination. We found single-electron processes with electron yields of about 10–9 for both the first debromination at V = 2.5 V and the second debromination at V = 2.8 V, respectively (Supplementary Fig. 29). The tip position for the current injection is not critical, and the voltage pulse can be applied anywhere above the molecule on Cu(111) to induce the reactions. On bilayer NaCl, it is even possible to induce reactions nonlocally44,45, that is, by voltage pulses above the bare NaCl/Cu(111) substrate near the molecule, indicating mediation by NaCl/Cu(111) interface state electrons. Moreover, on bilayer NaCl, both debromination reactions occur at lower absolute values of the voltage compared to Cu(111), which is tentatively explained by resonant tunnelling into the LUMO facilitating vibrational excitations within the molecule39.
Three intermediates were observed and characterized during the formation of hexayne 7 from 7(Br4) on NaCl. The AFM image of 7(Br4) (Fig. 4a) agrees well with its X-ray crystal structure (Supplementary Fig. 5); the triple bonds and Br atoms have the expected characteristic contrast, similar to 5(Br2). Two consecutive voltage pulses at V ≥ 1.4 V were applied to remove one of the two Br atoms at each olefin unit. The corresponding radical 7(Br3)• and diradical 7(Br2)2• were observed (Fig. 4b and 4c, respectively). We found that these radicals are always formed as the isomer with the Br atoms pointing towards the centre of the molecule. We never observed any intermediates in which two Br atoms had been removed from one C atom leaving the other dibromoolefin unit intact, which is explained by the increased activation energy for the second debromination of an olefin unit. Next, a higher voltage pulse (V ≥ 2.0 V) was applied to cleave the third C–Br bond, leading immediately to a 1,2-shift to form a tetrayne moiety on the left-hand side of the molecule (Fig. 4d). Finally, a fourth voltage pulse (V ≥ 2.0 V) produced hexayne 7 (Fig. 4e). Generation of tetrayne 6 (Supplementary Fig. 24) and octayne 8 (Supplementary Fig. 26) proceeded similarly to the formation of 7. Generally, adsorption on a planar surface tends to lead to a planar geometry of the molecules. Here, the phenyl groups are often tilted out of the surface plane by a small angle (on the order of few degrees46). The parts of the phenyl that appear brighter in the constant-height AFM images feature a slightly increased adsorption height (cf. Figs. 3c and 4e)46. Beyond that, dibromoolefins 5(Br2), 7(Br4) and 8(Br4) are relatively flat, both on the surface and in the solid state (see crystal structures in Supplementary Section 3). In contrast, the two phenyl rings of 6(Br4) are twisted out of the plane of the dibromoolefin (by 70° in the solid state (Supplementary Fig. 3) and 52° in vacuum according to calculations (Supplementary Fig. 11)), which accounts for the low resolution of AFM images of 6(Br4) (Supplementary Fig. 24).
The higher voltage required to break the C–Br bonds in the vinyl C=C•–Br radicals (Br dissociation form 5(Br)•, 7(Br2)2• and 7(Br)•), compared to dissociation of Br from C=CBr2 (Br dissociation from 5(Br2), 7(Br4) and 7(Br3)•), indicates that cleavage of the second C–Br bond in the vinyl radical is the rate-determining step of the rearrangement. We never observe a carbene intermediate (3 in Fig. 1) on NaCl during these reactions under our experimental conditions. The 1,2-bond migration in our manipulation experiments always takes place after cleavage of the second C–Br bond of a dibromoolefin unit; we conclude that the activation energy for the 1,2-shift is lower than that for removal of the second Br atom. Although the time resolution of our AFM experiments is on the order of a few minutes, a carbene intermediate with a barrier to a thermally activated 1,2-shift greater than 15 meV would be observable on this timescale at T = 5 K. Quantum chemical calculations indicate that the barrier to the 1,2-shift of this type of carbene in the gas phase is about 200 meV (Supplementary Fig. 22)47, indicating that the reaction on NaCl does not involve the thermally activated rearrangement of a carbene, or that the energy barrier is significantly lowered by the surface.
There has been some discussion about the bond angles in vinyl radicals of the type R2C=C•–X because these species can be linear or bent, depending on substituents R and X (ref. 48). The high resolution of images of 5(Br)•, 7(Br2)2• and 7(Br)• show that the C=C•–Br radical is nonlinear. This is the first time that the geometry of a vinyl radical has been visualized and the result is in good agreement with density functional theory (DFT) calculations, which predict an angle of 133° in 5(Br)• (Supplementary Fig. 10). The geometry that we observe in these radicals agrees with the angle in H2C=C•–Br radicals deduced from infrared spectroscopy and computational studies49.
Polyynes 5–8 were characterized by AFM, STM and scanning tunnelling spectroscopy (STS, Fig. 5). Some of the images of these polyynes show a curved geometry, as is often observed for longer polyynes5. In general, for a neutral molecule, the lowest unoccupied molecular orbital (LUMO) or highest occupied molecular orbital (HOMO) densities are probed by applying a positive or negative sample voltage large enough to attach or detach an electron to or from the molecule, respectively. Images at these voltages, corresponding to tunnelling via the negative or positive ion resonance (NIR or PIR) reflect the densities of LUMO or HOMO, respectively41. We recorded images of the NIR (corresponding to the density of the LUMO) of polyynes 5–8. It is possible to see the characteristic nodes in the middle of every triple bond in the images of 7 and 8 recorded with functionalized tips (Fig. 5e,k), in agreement with DFT-calculated LUMOs (Fig. 5f,l). We also recorded images for the PIR (HOMO) of 7 and 8 (Fig. 5b,h). The differential conductance spectra show that the LUMO energies of the polyynes decrease for longer polyynes, as expected (Fig. 5m). We compared the experimentally determined resonance energies (Supplementary Fig. 30) to theoretically calculated electronic HOMO–LUMO gaps (Fig. 5n). It is known that Kohn–Sham DFT tends to underestimate the electronic gap. We used many-body perturbation theory in the GW approximation, accounting for the screened Coulomb interaction W related to temporary charging using a Green's function approach, to obtain more accurate quasiparticle energies28,40. We obtained good agreement between experimental STS data and theoretical GW quasiparticle levels (Fig. 5n).
We have synthesized a series of polyynes 5–8 using atomic manipulation on bilayer NaCl on Cu(111), starting from the corresponding dibromoolefins, while characterizing the intermediates, and all the final polyynes using AFM, STM and STS. The core transformation in this approach is a 1,2-shift similar to the FBW rearrangement, initiated by dissociation of the Br from the vinyl radical (C=C•–Br). This is the first report of any 1,2-shift triggered by atomic manipulation on a surface. In the case of 5 and 7, we were able to image all of the vinyl radical intermediates, and these radicals were found to be persistent on the surface at 5 K. We do not observe any carbene intermediates on NaCl and conclude that the activation energy for the 1,2-shift is lower than that for the cleavage of the C•–Br bond. In contrast, on Cu we observed the carbene bonded to the surface and the rearrangement does not occur.
We have demonstrated that complex molecular skeletal rearrangements can be triggered by atom manipulation. This is the first time that polyynes have been structurally characterized by STM and AFM. The precision of the synthesis and wealth of information provided by this approach opens new opportunities for the on-surface fabrication of novel molecules and atomic-scale devices.
Compounds 5(Br2), 7(Br4) and 8(Br4) were prepared using published procedures, as summarized in the Supplementary Section 1; all characterization data match those previously reported. The synthesis and characterization of compound 6(Br4) and additional details of synthetic methods and characterization are available in the Supplementary Section 1.
The experiments were carried out in homebuilt combined STM and AFM systems operated under ultrahigh-vacuum conditions (base pressure p < 10−10 mbar) at a temperature of T = 5 K. The microscopes were equipped with qPlus sensors42 operated in frequency-modulation mode50. Two sensors were used with eigenfrequencies of f0 ~ 28.8 kHz and f0 ~ 30.1 kHz, respectively, a stiffness of k = 1,800 N m−1 and quality factors of Q = 20,000 and Q = 150,000, respectively. The oscillation amplitude was set to A = 50 pm, and voltage V was applied to the sample.
STM images recorded in constant-current (closed feedback loop) and constant-height (open feedback loop) mode show the topography z and the tunnelling current I, respectively. STM images were acquired using different tip functionalizations (Cu tip, CO tip, Br tip)43 as indicated in each STM image. AFM images were acquired with a CO-terminated tip in constant-height mode (open feedback loop) and show the frequency shift ∆f. The tip-height offset is given for each AFM image with respect to an STM set point of 1 pA at 0.1 V above the bare surface (Cu or NaCl). Positive height offsets refer to a distance increase.
A Cu(111) single crystal was cleaned by sputtering and annealing cycles. Experiments were performed on the bare Cu(111) surface, and on islands of two-monolayer-thick NaCl. NaCl islands were grown by sublimation from a crucible onto the cleaned Cu(111) surface held at a temperature of 270 K. Low sub-monolayer coverages of compounds 5(Br2), 6(Br4), 7(Br4), 8(Br4) and CO molecules were deposited at sample temperatures T < 10 K.
STM and AFM images, as well as numerically obtained dI/dV (V) curves, were post-processed using Gaussian low-pass filters.
The crystallographic data have been deposited at the Cambridge Crystallographic Data Centre under CCDC nos. 1567546 (7(Br4)), 1567547 (6(Br4)) and 1567548 (8(Br4)), and copies can be obtained free of charge from www.ccdc.cam.ac.uk/data_request/cif. Corresponding .cif files including X-ray crystal structures and .xyz files containing geometries of calculated structures are available in the Supplementary Information. Other data supporting the findings of this study are available from the corresponding authors upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The research leading to these results received funding from ERC Advanced Grants CEMAS (agreement no. 291194) and CoSuN (320969), ERC Consolidator Grant AMSEL (682144) and EU project PAMS (610446). P.G. acknowledges receipt of Postdoc.Mobility fellowships from the Swiss National Science Foundation. Y.X. was supported by the EPSRC Centre for Doctoral Training in Synthesis for Biology and Medicine (EP/L015838/1) and by a University of Oxford Clarendon Fund Scholarship. The authors acknowledge use of the Oxford Advanced Research Computing (ARC) facility to carry out computational work (doi: 10.5281/zenodo.22558). The authors thank R.S. Paton and I. Gruebner for discussions on computational studies and A.L. Thompson for help with X-ray crystal structure refinements.