NANOMATERIALS

Fluorine-programmed nanozippingto tailored nanographenes on rutileTiO2 surfacesM. Kolmer1,2*, R. Zuzak1, A. K. Steiner3, L. Zajac1, M. Engelund4, S. Godlewski1,M. Szymonski1, K. Amsharov3*

The rational synthesis of nanographenes and carbon nanoribbons directly on nonmetallicsurfaces has been an elusive goal for a long time.We report that activation of the carbon(C)–fluorine (F) bond is a reliable and versatile tool enabling intramolecular aryl-aryl couplingdirectly on metal oxide surfaces. A challenging multistep transformation enabled by C–Fbond activation led to a dominolike coupling that yielded tailored nanographenes directly onthe rutile titania surface. Because of efficient regioselective zipping, we obtained the targetnanographenes from flexible precursors. Fluorine positions in the precursor structureunambiguously dictated the running of the “zipping program,” resulting in the rolling up ofoligophenylene chains.The high efficiency of the hydrogen fluoride zippingmakes our approachattractive for the rational synthesis of nanographenes and nanoribbons directly on insulatingand semiconducting surfaces.

Carbon-based nanostructures synthesizedthrough rational surface-assisted C–C cou-pling on single-crystal metal surfaces (1, 2)include individual isomers of fullerenes(3, 4) and fullerene fragments (5, 6), the

chirality-pure carbon nanotubes (7), atomicallyprecise nanographenes (NGs) (8–10), and graphenenanoribbons (GNRs) with a well-defined pe-riphery (6, 11–15). The consecutive, thermallytriggered cyclodehydrogenation of the polycyclicaromatic hydrocarbon (PAH) precursor bearingrequired C–C connectivity to the target nano-structure represents the key transformation inthe on-surface synthesis strategy. The synthesisof hexabenzo[bc,ef,hi,kl,no,qr]coronene (HBC)by Weiss et al. showed that this step can berealized efficiently under ultrahigh vacuum(UHV) conditions on atomically clean metalsurfaces (16) (Fig. 1A). The catalytic activity ofthe metals substantially reduced the activationbarrier of the cyclization (8). In 2010, Cai et al.applied a similar strategy to form atomicallyprecise GNRs on the Au(111) surface (11) (Fig. 1B).This discovery paved theway toward the fabrica-tion of complex molecular nanoarchitectureson selected noble metal surfaces (17). However,for most practical applications, a carbon-basednanostructuremust be transferred onto insulat-ing or semiconducting surfaces (18, 19).An attractive yet challenging way to tackle this

problem is the controlled synthesis of carbon

nanostructures directly on technologically rele-vant nonmetallic substrates, such as metal oxidesurfaces (20–22). However, all reported attemptsto perform the cyclization onmetal oxides havebeen unsuccessful because of the lack of catalyticactivity in the cyclodehydrogenation process(20) (Fig. 1C). The cyclization of PAH precursorson such surfaces requires high temperatures,which leads to a loss of selectivity. Thus, therational synthesis of tailored carbon-based nano-architectures on metal oxide surfaces requiresthe development of an alternative cyclizationtechnique. Previously, we have found that in-tramolecular aryl-aryl coupling can be effectivelyrealized through C–F bond activation on g-Al2O3under relatively mild conditions (23–27). Fur-ther exploration revealed that metal oxides ofIII and IV groups also displayed activity incyclodehydrofluorination at elevated temper-atures (28). Among them, bulk powders of ti-tanium dioxide activated C–F bonds at 570 K,whichmade it an attractive candidate for on-surface investigations performed under UHVconditions.We present the rational on-surface synthesis of

NGs on a semiconducting rutile TiO2(011) surfacethrough dominolike HF zipping of programmedfluoroarene precursors (Fig. 1D). The high poten-tial of the approach was demonstrated by a chal-lenging transformation consisting of the formalrolling up of the linear oligophenylene chainaround a phenyl moiety, yielding NGHBC (Fig.1D). In contrast to the commonly used rigiddesign of precursors, our approach allows theregioselectivity of the cyclization to be unam-biguously programmed by F atom positions,providing sufficient flexibility in the design ofprecursor molecules.To investigate the HF-zipping process on a

rutile surface, two model NGs—namely, DBPP(dibenzo[ij,rst]phenanthro[9,10,1,2-defg]pentaphene)

and HBC—were chosen as target compounds(Fig. 2, A and B). DBPP, which can be consideredan ultrashort armchair GNR (6-AGNR), repre-sents the model for the on-surface synthesis ofGNRs, and HBC is one of the smallest and best-characterized NGs (29–33). Both NGs possesseasy-to-recognize geometries in scanning tunnel-ing microscopy (STM) images. The requiredspecially “programmed” fluorinated oligophenyl-enes P1 and P2 were obtained by multisteporganic synthesis (for details, see the supple-mentary materials). The key feature of thecyclodehydrofluorination is the “switchable”activity of the C–F bond. Only C–F bonds withclose proximity to a C–H bond displayed ac-tivity, whereas peripheral C–F remained com-pletely intact. This reactivity enabled tandemcyclization via HF elimination in a truly dom-inolike fashion, because each subsequent cy-clization step led to the “activation” of one newC–F bond (26). The active H–F pairs for the HF-zipping concept are shown schematically inFig. 2, A and B.All on-surface experiments, including low-

temperature STM, x-ray photoemission spec-troscopy (XPS), and mass spectrometry (MS)studies, were performed in situ underUHV condi-tions.We deposited precursormolecules by usingstandardKnudsen cells on the (2×1) reconstructed(011) face of the rutile TiO2 (for details, see thesupplementarymaterials). To thermally inducethe transformation of P1, we started with sub-monolayer deposition of precursor molecules onthe substrate kept at room temperature (RT) andthen heated the substrate to ~570 K (bulk activa-tion temperature) for 10 min. Under these mildconditions, most P1 molecules desorbed from thesurface, leaving almost bare surface terraces withno clear evidence of successful HF elimination(see the supplementary materials). At a higherannealing temperature of ~670 K, particularlyflat molecules with the specific rhomboid shapeexpected for DBPP were observed in our STMexperiment. However, because of the appreciablethermal desorption of P1, DBPP molecules wereadsorbed only on chemically active sites (stepedges or domain boundaries), which compli-cated the accurate interpretation of their non-uniform STM contrast (see the supplementarymaterials).With the larger precursor P2, after deposi-

tion onto an RT substrate, P2 molecules werefound mostly in globular form on reconstructedterraces of rutile (011) with STM. However,after annealing at ~570 K, the molecules re-mained on the surface (Fig. 2C). Moreover, weobserved a general change in P2 appearanceand observed elongated geometries with lengthsup to ~2.5 nm. These geometries correspondto different possible configurations of P2 on thesurface (see detailed analysis in the supplemen-tary materials), consistent with the expectedflexibility of the precursors. This observationpoints out that the globular geometry, the fa-vorable gas-phase configuration of P2 preservedafter deposition, was only metastable on thesurface.

RESEARCH

Kolmer et al., Science 363, 57–60 (2019) 4 January 2019 1 of 4

1Centre for Nanometer-Scale Science and Advanced Materials,NANOSAM, Faculty of Physics, Astronomy and AppliedComputer Science, Jagiellonian University, Łojasiewicza 11,30-348 Kraków, Poland. 2Center for Nanophase MaterialsSciences, Oak Ridge National Laboratory, Oak Ridge, TN37831, USA. 3Department of Organic Chemistry, FriedrichAlexander University Erlangen-Nuremberg, 91058 Erlangen,Germany. 4Espeem S.A.R.L., L-4365 Esch-sur-Alzette,Luxembourg.*Corresponding author. Email: konstantin.amsharov@fau.de(K.A.); kolmerma@ornl.gov (M.K.)

on June 6, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/

To induce the transformation of P2, we heatedthe rutile (011) substrate to ~670 K for 10 min.Although most of the molecules were found onstep edges, some molecules appeared directlyon newly formed ad-islands that allowed theirhigh-quality STM visualization (Fig. 2D). Theround shape and the size (~1.6-nm full width athalf maximum, as shown by STM image crosssections in Fig. 3, A and B) of the molecules arein agreement with the expected HBC model(Fig. 3C), our unoccupied-states STM imagesimulations (see the supplementary materials),and previous reports (29, 30, 32). The presence ofHBC molecules demonstrated the feasibility ofthe HF-zipping strategy on the rutile surface.The programmed P2 precursor was transformedvia five consecutive cyclodehydrofluorinationsteps followed by the final cyclodehydrogenationinto the target NG molecule.Among molecules found on ad-islands, some

possessed an additional protrusion locatedbeside the center of the molecule, as markedby arrows in Fig. 3A. We attributed theseimages to molecules of helicenelike H2-HBC,presented in Fig. 3D, in which the final cyclo-dehydrogenation step was not accomplished.The presence of two H atoms in the fjord regioncaused deviation from planar geometry, withtwo corresponding benzene rings oriented outof plane. The high-resolution STM image of anHBC and H2-HBC pair in Fig. 3E confirms theexpected position of the protrusion on themolecules. The uncopied-states STM contrastof H2-HBC is also reproduced in our STM imagesimulation (see the supplementary materials),which is in good agreement with the exper-imental image. The formation of the H2-HBCintermediate after only the first five cyclizationreactions activated via F was rather unexpected,because the last cyclization step should occurspontaneously given the high strain in the fjordregion. However, the observation of the H2-HBC intermediate indicates that, on the TiO2surface, the cyclodehydrofluorination processhad a markedly lower activation barrier thancyclodehydrogenation. A similar strategy wasin fact completely inefficient on metal surfaces.As we found for the Au(111) surface, catalyticactivation of the competitive cyclodehydrogena-tion processes resulted in undefined molecularstructures formed from P2 after their anneal-ing (see results and discussion in the supple-mentary materials). Thus, metal oxide surfaceassistance in the cyclodehydrofluorination isthe key aspect of the successful HF zipping (28).Although the cyclization mechanism is not fullyunderstood, the most probable scenario includesthe C–F bond polarization by the active Ti center,allowing synchronous Friedel-Crafts–like aryla-tion (34, 35).The STM data show that the HF-zipping strat-

egywas efficient, asmany single-targetmoleculeswere found locally on the surface. Other thanH2-HBC and HBC, no other molecular species wereobserved on the ad-islands where high-qualitySTM imaging was possible. To shed more lighton the global outcome of the thermally induced

Kolmer et al., Science 363, 57–60 (2019) 4 January 2019 2 of 4

Fig. 1. Selected examples of on-surface syntheses of NGs and nanoribbons by a bottom-upapproach. (A) First on-surface synthesis of NG HBC (16); (B) rational synthesis of GNRs on Au(111)(11); (C) attempts to perform cyclodehydrogenation on a metal oxide surface (20); (D) firstrational on-surface synthesis of NGs on a nonmetallic surface (this work).

Fig. 2. HF elimination–based zipping of precursors. Schemes of (A) DBPP and (B) HBCformation concepts. (C) Three-dimensional (3D) visualization of an STM image (+2 V; 10 pA)of P2 deposited on rutile (011)-(2×1) at RT and annealed at 570 K for 10 min. Note differentconfigurations of molecules. (D) 3D STM image (+2.5 V; 50 pA) obtained after sampleannealing at 670 K. Single HBC molecules are adsorbed on newly formed ad-islands withdifferent surface reconstruction (yellow-green). (Inset) STM image (+2 V, 10 pA, 2.5 nm by2.5 nm) of a single HBC molecule. Scale bars in (C) and (D) are 2 nm.

RESEARCH | REPORTon June 6, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/

HF elimination reaction, we combined the STMdata with the XPS chemical analysis. The C1score-level photoemission spectrum of P2 mole-cules deposited on the rutile (011) surface at RT(Fig. 4A) showed an asymmetric signal composed

of two peaks separated by ~2 eV in bindingenergy (EB), which correspond to C–C (red line,EB= 284.4 eV) andC–F (green line,EB = 286.3 eV)contributions (36, 37). Figure 4B presents the F1score-level region. For RT deposition, the single

peak observed at EB = 686.8 eV is related to anorganofluorine component (13, 36, 37). The ab-sence of a peak at ~684.5 eV suggests the lack ofa F–Ti component (38). Thus, the XPS analysisconfirmed that the C–F bonds in the precursormolecules were intact after deposition on thesurface at RT.After annealing at 670 K for 10 min, the

XPS signal from the C1s region consisted ofonly a single peak at EB = 284.5 eV, corre-sponding to a C–C component (Fig. 4A), con-firming global scission of C–F bonds in P2molecules caused by efficient HF elimination.The corresponding STM images (Fig. 3, A andE) show that the flat-lying HBC moleculeswere found on the newly formed reconstructedareas of the surface, forming ad-islands attachedto TiO2(011)-(2×1) step edges or domain bound-aries. These structures were not observed in acontrol experiment, where we directly depositedHBC molecules on the rutile (011) and annealedthe sample to 670 K (see the supplementary ma-terials). Moreover, other halogenoarenes withbromine (20–22) or iodine (21) substituents an-nealed under similar conditions at rutile titaniasurfaces did not produce the observed ad-islands.An obvious hypothesis is that F atoms are

building blocks in these new structures. How-ever, Fig. 4B shows that after annealing, theXPS signal from the F1s region was stronglyreduced, consisted of only organic fluorine(

exceeded 670 K, consistent with our previousexperimental observations.Desorption of TiOF2 points to strong Ti–F

covalent bonding of F to the surface during theHF zipping, which may indirectly explain theobserved surface reconstruction under newlyformed HBC molecules. Adsorbates strongly in-teracting with the surface are known to inducechanges in rutile titania reconstructions, as pre-viously reported for (110) (39) and (011) (40, 41)faces. Recent work by Balajka et al. (41) reporteda water-induced reconstruction of a rutile (011)surface to form a bulk-terminated (1×1) facecovered with a hydroxyl group overlayer form-ing (2×1) reconstruction. In the case of a HFelimination reaction, the desorption of TiOF2species also formally leads to the formation ofwater molecules, TiO2 + 2HF→ TiOF2 + H2O,which then could dissociate and locally createthe hydroxyl-rich reconstruction. Close inspec-tion of the ad-island structure performed byhigh-resolution STM characterization (Fig. 4D)shows bright protrusions forming rows along the[01-1] direction, with neighboring ones shiftedby half of the lattice, giving apparent pseudo-hexagonal c(2×1) symmetry. Results of our den-sity functional theory calculations shown in Fig.4E present the exact structure of the ad-islandswith four hydroxyl groups (two dissociatedwatermolecules) per (2×1) unit cell. Correspondingunoccupied-states STM image simulation (Fig.4F) confirmed that for high positive bias volt-ages, this (2×1) surface reconstruction gave theapparent c(2×1) symmetry observed in the ex-perimental data. These results support the rutiletitania surface participation in intermediatestates of the cyclodehydrofluorination and re-flect its crucial role in this process. Our resultsprovide a pathway toward custom-designedsp2 carbon–based nanostructure formation bydirect on-surface synthesis methods on tech-

nologically relevant semiconducting or insulat-ing surfaces.

REFERENCES AND NOTES

1. J. Méndez, M. F. López, J. A. Martín-Gago, Chem. Soc. Rev. 40,4578–4590 (2011).

2. P. A. Held, H. Fuchs, A. Studer, Chemistry 23, 5874–5892(2017).

3. G. Otero et al., Nature 454, 865–868 (2008).4. K. Amsharov et al., Angew. Chem. Int. Ed. Engl. 49, 9392–9396

(2010).5. K. T. Rim et al., Angew. Chem. Int. Ed. Engl. 46, 7891–7895

(2007).6. N. Abdurakhmanova et al., Carbon 84, 444–447 (2015).7. J. R. Sanchez-Valencia et al., Nature 512, 61–64 (2014).8. M. Treier et al., Nat. Chem. 3, 61–67 (2011).9. C. Rogers et al., Angew. Chem. Int. Ed. Engl. 54, 15143–15146

(2015).10. R. Zuzak et al., Chem. Commun. 54, 10256–10259 (2018).11. J. Cai et al., Nature 466, 470–473 (2010).12. P. Ruffieux et al., Nature 531, 489–492 (2016).13. H. Hayashi et al., ACS Nano 11, 6204–6210 (2017).14. C. Ma et al., Nat. Commun. 8, 14815 (2017).15. D. J. Rizzo et al., Nature 560, 204–208 (2018).16. K. Weiss et al., Angew. Chem. Int. Ed. Engl. 38, 3748–3752

(1999).17. L. Talirz, P. Ruffieux, R. Fasel, Adv. Mater. 28, 6222–6231

(2016).18. P. B. Bennett et al., Appl. Phys. Lett. 103, 253114 (2013).19. J. P. Llinas et al., Nat. Commun. 8, 633 (2017).20. M. Kolmer et al., Angew. Chem. Int. Ed. Engl. 52, 10300–10303

(2013).21. M. Kolmer et al., Chem. Commun. 51, 11276–11279 (2015).22. G. Vasseur et al., J. Am. Chem. Soc. 138, 5685–5692

(2016).23. K. Y. Amsharov, M. A. Kabdulov, M. Jansen, Angew. Chem. Int.

Ed. Engl. 51, 4594–4597 (2012).24. N. Suzuki, T. Fujita, K. Y. Amsharov, J. Ichikawa, Chem.

Commun. 52, 12948–12951 (2016).25. O. Papaianina, K. Y. Amsharov, Chem. Commun. 52,

1505–1508 (2016).26. A. K. Steiner, K. Y. Amsharov, Angew. Chem. Int. Ed. Engl. 56,

14732–14736 (2017).27. O. Papaianina et al., Angew. Chem. Int. Ed. Engl. 56,

4834–4838 (2017).28. K. Amsharov, Phys. Status Solidi B 253, 2473–2477

(2016).29. P. Samorí, N. Severin, C. D. Simpson, K. Müllen, J. P. Rabe,

J. Am. Chem. Soc. 124, 9454–9457 (2002).30. L. Gross et al., Phys. Rev. B 71, 165428 (2005).

31. W. H. Soe et al., ACS Nano 6, 3230–3235 (2012).32. L. Gross et al., Science 337, 1326–1329 (2012).33. T. Ardhuin, O. Guillermet, A. Gourdon, S. Gauthier,

J. Phys. Chem. C 122, 11905–11910 (2018).34. J. Ichikawa, M. Yokota, T. Kudo, S. Umezaki, Angew. Chem. Int.

Ed. Engl. 47, 4870–4873 (2008).35. D. Sharapa, A.-K. Steiner, K. Amsharov, Phys. Status Solidi B

10.1002/pssb.201800189 (2018).36. F. Blobner et al., J. Phys. Chem. C 119, 15455–15468

(2015).37. R. Han et al., Chem. Commun. 52, 9805–9808 (2016).38. J.-Y. Ruzicka et al., RSC Adv. 4, 20649–20658 (2014).39. D. Silber et al., Nat. Commun. 7, 12888 (2016).40. Q. Cuan, J. Tao, X. Q. Gong, M. Batzill, Phys. Rev. Lett. 108,

106105 (2012).41. J. Balajka et al., J. Phys. Chem. C 121, 26424–26431

(2017).

ACKNOWLEDGMENTS

Funding: The research was supported by the Polish Ministryof Science and Higher Education, contract no. 0341/IP3/2016/74.R.Z. acknowledges support received from the NationalScience Center, Poland (2017/24/T/ST5/00262). A.K.S. andK.A. thank the Deutsche Forschungsgemeinschaft (DFG-SFB953 “Synthetic Carbon Allotropes” project A6, and AM407) forfinancial support. Work was partially conducted at the Centerfor Nanophase Materials Sciences (CNMS), which is a DOEOffice of Science User Facility. The research was partiallysupported from basic and statutory funds of the JagiellonianUniversity in Krakow provided by the Polish Ministry of Scienceand Higher Education. Author contributions: K.A. and M.K.conceived the project. K.A. and A.K.S. carried out the precursorsynthesis and analysis. M.K. and R.Z. conducted the on-surfacesynthesis and low-temperature STM, XPS, and MS analyseswith support from L.Z., S.G., and M.S. M.E. conducted thecomputations. M.K. and K.A. prepared the manuscript withfeedback from all other authors. Competing interests: Nonedeclared. Data and materials availability: All data needed toevaluate the conclusions in the paper are present in the paper orthe supplementary materials.

SUPPLEMENTARY MATERIALS

www.sciencemag.org/content/363/6422/57/suppl/DC1Materials and MethodsSupplementary TextFigs. S1 to S34Table S1References (42–45)

21 September 2018; accepted 16 November 201810.1126/science.aav4954

Kolmer et al., Science 363, 57–60 (2019) 4 January 2019 4 of 4

RESEARCH | REPORTon June 6, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://www.sciencemag.org/content/363/6422/57/suppl/DC1http://science.sciencemag.org/

surfaces2Fluorine-programmed nanozipping to tailored nanographenes on rutile TiOM. Kolmer, R. Zuzak, A. K. Steiner, L. Zajac, M. Engelund, S. Godlewski, M. Szymonski and K. Amsharov

DOI: 10.1126/science.aav4954 (6422), 57-60.363Science

, this issue p. 57Sciencestepwise process sequentially added aromatic rings around a central aryl group until it was completely substituted.fluorine substitution of the aryl groups was selected so that as the carbon-fluorine bonds were thermally activated, a

show that fluorinated aryl groups can be coupled to form nanographenes on the rutile surface of titanium oxide. Theal.etsurfaces through carbon-carbon coupling reactions, but the surfaces of oxides do not assist these reactions. Kolmer

The growth of nanographene islands and ribbons on metal surfaces can be accomplished on single-crystal metalNanographenes on oxides

ARTICLE TOOLS http://science.sciencemag.org/content/363/6422/57

MATERIALSSUPPLEMENTARY http://science.sciencemag.org/content/suppl/2019/01/03/363.6422.57.DC1

REFERENCES

http://science.sciencemag.org/content/363/6422/57#BIBLThis article cites 44 articles, 1 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

Copyright © 2019, American Association for the Advancement of Science

on June 6, 2021

http://science.sciencemag.org/

Dow

nloaded from

http://science.sciencemag.org/content/363/6422/57http://science.sciencemag.org/content/suppl/2019/01/03/363.6422.57.DC1http://science.sciencemag.org/content/363/6422/57#BIBLhttp://www.sciencemag.org/help/reprints-and-permissionshttp://www.sciencemag.org/about/terms-servicehttp://science.sciencemag.org/