Energy funnelling within multichromophore architectures monitored with subnanometre resolution
Shuiyan Cao, Anna Ros?awska, Benjamin Doppagne, Michelangelo Romeo, Michel Féron, Frédéric Chérioux, Hervé Bulou, Fabrice Scheurer, Guillaume Schull
FFunneling energy at the molecular scale
Shuiyan Cao , , † , Anna Ros(cid:32)lawska , † , ∗ , Benjamin Doppagne ,Michelangelo Romeo , Michel F´eron , Fr´ed´eric Ch´erioux , Herv´e Bulou ,Fabrice Scheurer , Guillaume Schull , ∗ Universit´e de Strasbourg, CNRS, IPCMS, UMR 7504, F-67000 Strasbourg, France, Department of Applied Physics, Nanjing University ofAeronautics and Astronautics, Nanjing, 210016, China, Universit´e Bourgogne Franche-Comt´e, FEMTO-ST, UFC,CNRS, 15B avenue des Montboucons, F-25030 Besan¸con, France. (Dated: January 5, 2021)
PACS numbers: † These authors contributed equally to this paper. a r X i v : . [ c ond - m a t . m e s - h a ll ] J a n n natural and artificial light-harvesting complexes (LHC) the resonant energy transfer(RET) between chromophores enables an efficient and directional transport of solarenergy between collection and reaction centers. The detailed mechanisms involved in thisenergy funneling are intensely debated , essentially because they rely on a successionof individual RET steps that can hardly be addressed separately. Here, we developed ascanning tunnelling microscopy-induced luminescence (STML) approach allowing visualiz-ing, addressing and manipulating energy funneling within multi-chromophoric structureswith sub-molecular precision. We first rationalize the efficiency of the RET process atthe level of chromophore dimers. We then use highly resolved fluorescence microscopy(HRFM) maps to follow energy transfer paths along an artificial trimer of descendingexcitonic energies which reveals a cascaded RET from high- to low-energy gap molecules.Mimicking strategies developed by photosynthetic systems, this experiment demonstratesthat intermediate gap molecules can be used as efficient ancillary units to convey energybetween distant donor and acceptor chromophores. Eventually, we demonstrate that theRET between donors and acceptors is enhanced by the insertion of passive molecules actingas non-covalent RET bridges. This mechanism, that occurs in experiments performed ininhomogeneous media and which plays a decisive role in fastening RET in photosyntheticsystems , is reported at the level of individual chromophores with atomic-scale resolution.As it relies on organic chromophores as elementary components, our approach constitutesa powerful model to address fundamental physical processes at play in natural LHC.The energy funneling in LHC relies on cascaded RET events, based primarily on dipole-dipole interactions, occurring between high and low energy-gap chromophores. The effi-ciency of the process is further improved by exchange interaction mediated energy transfersat short distance , delocalization of the excitation over coherently coupled molecules in-volving vibronic coupling or not , fine-tuning of the absorption/emission energy of ancillarychromophores , or promoted energy transfer by the environment . The influence of theseparameters on the extraordinary ability of biological LHC to transfer energy between dis-tant centers remains to be clarified and their systematic control in artificial systems isboth crucial and far ahead . Recent experiments have shown that a cryogenic STMcan be used to probe the fluorescent properties of individual or interacting chromophoreswith high spatial, spectral and temporal resolution. Here, we develop this approach to2 dPcH Pc ZnPcPdPc ZnPcH Pc ab c PdPcZnPcH Pc I n t en s i t y ( kc t s p C - e V - ) Photon energy (eV)Q Pd Q Zn Q x Q y [ ] [ ] Q Zn y Q Pd Q Pd y x Q Zn x Q x Q y FIG. 1:
STM-induced fluorescence of individual chromophores. (a) STM image ( I = 10pA, V = -2.5 V, scale bar 1 nm), (b) ball-and-stick models, and (c) typical STML spectra ( V =-2.5 V , acquisition time t = 60 s, from top to bottom I = 250 pA, 50 pA, 50 pA) of the threechromophores (PdPc, ZnPc and H Pc) investigated in this study. The molecules are separated fromthe Ag(111) sample by three layers of NaCl. The spectra were recorded with the tip positionedat the extremity of a pyrrole sub-unit (colored dots in (a)). Spectral features in (c) are associatedto degenerate (Q Pd , Q Zn ) and non-degenerate (Q x , Q y ) radiative transitions of the chromophoredipoles, whose orientations are given in (b). follow the energy funneling within multi-chromophoric architectures and characterize thedecisive role played by intermediate molecules, located between donor and acceptor units, inpromoting, blocking or directing RET. To this end, we used three chromophores, palladium-phthalocyanine (PdPc), zinc-phthalocyanine (ZnPc) and free-base phthalocyanine (H Pc),deposited on an insulating NaCl trilayer covering Ag(111) (see Methods). STM images ofZnPc recorded at negative voltages (Fig. 1a) reveal 16-lobe structures resulting from the fastmotion of ZnPc between two metastable adsorption sites , whereas PdPc and H Pcexhibit 8-lobe patterns characteristic of the highest occupied molecular orbitals (HOMO) offixed phthalocyanine molecules.A straightforward assignment of the molecules is obtained from their STML spectra(Fig. 1c). PdPc is characterized by a sharp emission line (Q Pd ) at ≈ Zn ) is less intense and located at a slightly lower energy3 ≈ Pc is characterized by one intense emission line at ≈ x ) and a weak emission at ≈ y ) corresponding to dipoles orientedrespectively along and perpendicular to the inner H-H axis of the molecule. The purelyelectronic transitions of PdPc, H Pc and ZnPc are accompanied by vibronic lines at lowerenergies .In pioneering works, Zhang et al. , and Imada et al. reported the emission of coherentlycoupled chromophores, and RET in a single donor–acceptor pair with STML. Building ontheir approach, we first manipulated the molecules on NaCl to form the three possibledonor-acceptor (D-A) pairs: PdPc–ZnPc (Fig. 2a), ZnPc–H Pc (Fig. 2b) and PdPc–H Pc(Fig. 2c). In Fig. 2d, e, and f we display for each pair the STML spectrum obtained bylocating the STM tip on the extremity of the donor that is the most distant from theacceptor (black dots in the STM images of Fig. 2). By this approach, the chances toexcite directly the acceptor are negligible (see also the discussion of Fig. 3d). These spectrasystematically display peaks at both donor and acceptor energies, attesting that part ofthe energy of the donor is transmitted to the acceptor, a phenomenon described in termsof a RET process involving dipole-dipole interactions . Assuming perfect emitters, theenergy transfer efficiency, RET eff , (given in inset) follows:
RET eff = I A / ( I D + I A ), where I D ( I A ) is the emission intensity of the donor (acceptor) in the dimer. Qualitatively, theenergy transfer is more efficient when the gap-energy difference between the donor and theacceptor is small. This behavior reflects the impact of a higher spectral overlap between thedonor emission band and the acceptor absorption band. A semi-quantitative estimation ofthe spectral overlap for each D-A pair yields values of 6.3 (PdPc–ZnPc), 1.9 (ZnPc–H Pc)and 1.4 eV − (PdPc–H Pc) (see Methods and Extended Data Fig. 2), following the trendobserved in Fig. 2. To ensure a fair comparison between the RET efficiency in the D-Apairs, the spectra of Fig. 2 were normalized by the plasmonic response of the tip-samplejunction so as to correct for the different emission-line enhancements due to the more or lessresonant character of the plasmon modes, and for the wavelength-dependent response of ourdetection setup .RET relying on dipole-dipole interactions are also impacted by the angle between donorand acceptor dipoles . To test this parameter, we show in Fig. 2d spectra acquired for two4 dPcZnPc PdPcZnPc a H Pc ZnPc ZnPc b PdPcH Pc PdPcH Pc c d Photon energy (eV) Q Zn I n t en s i t y ( a r b . un i t s ) Photon energy (eV)DonorAcceptor STM tip e Photon energy (eV) Q Pd f R E T e ff i c i en cy R (nm) g h i � E = 19 meV � E = 89 meV � E = 108 meVRET
RET
Eff = 0.49RET
Eff = 0.68 RET
Eff = 0.69 Q Pd Q Zn Q Zn RET
Eff = 0.96h � h � y Q Pd yx x Q Pd x Q Pd y Q x Q y Q x Q y Q Zn y Q Zn x I n t en s i t y ( a r b . un i t s ) I n t en s i t y ( a r b . un i t s ) Q Zn Q x Q Zn Q Pd x Q Zn Q Pd y H Pc Q Pd Q x xy PdPcH Pc RR = 1.68 nm R = 1.97 nm R = 2.24 nmR = 2.47 nm R = 2.72 nm R = 3.20 nm
FIG. 2:
RET between D–A pairs.
STM images ( V = -2.5 V) and models of the (a) PdPc–ZnPc( I = 10 pA), (b) ZnPc–H Pc ( I = 10 pA) and (c) PdPc–H Pc ( I = 5 pA) D–A pairs and (d, e,f) plasmon–corrected STML spectra ( V = -2.5 V). The markers in (a), (b) and (c) show the STMtip positions used to record the spectra in (d) ( I = 300 pA, acquisition time t = 300 s), (e) ( I =300 pA, t = 180 s) and (f) ( I = 100 pA, t = 180 s) respectively. The colored areas indicate thespectral range considered to calculate the RET eff values. (g) Sketch of the D-A energy transferexperiments. (h) STM images ( I = 5 pA, V = -2.5 V) of PdPc–H Pc dimers whose center-to-center distance (R) is progressively increased by STM-tip manipulation (two sets of data were usedto generate this panel). (i) PdPc–H Pc RET efficiency as a function of R deduced from STMLspectra (Extended Data Fig. 1) acquired for the STM-tip positions marked by dots in (h). Errorbars are discussed in Extended Data Fig. 1. Scale bars in STM images are 1 nm. STM tip positions (black dot and grey star in Fig. 2a) on the donor of a PdPc–ZnPc dimer.These spectra reveal an efficient RET when the tip is located on top of the PdPc dipoleoriented towards the ZnPc acceptor, a configuration where the donor Q Pd x and acceptorQ Zn x dipoles are nearly colinear (see Methods and Extended Data Fig. 3). The RET is5eaker when the dipoles are essentially parallel (Q Pd y // Q Zn y ). This implies that onecan selectively excite a dipole by locating the tip on top of it, i.e., one can use an STMtip as an excitation source of sub-molecular-scale precision. A more detailed investigation(see Methods) gives a qualitative interpretation of the role played by the orientations ofthe donor and acceptor dipoles, but fails to provide quantitative predictions of the RETefficiency, suggesting that an accurate description of our RET processes cannot be resumedby simple Coulomb interactions between point dipoles.To better establish this assertion, we evaluated (Fig. 2h) the dependency of the RETprocess on the distance ( R ) between the centers of the donor and acceptor chromophores .Fig. 2i shows that the RET efficiency decreases monotonously when R increases, and tendsto zero for R > i.e. , F¨orster RET), and that othermechanisms occur at short D–A distances that rely on molecular orbitals overlap ( i.e. ,exchange interactions or Dexter RET) and/or multipolar RET . Determining the respectiveinfluence of these different parameters requires detailed calculations that are beyond thescope of the present report.Having characterized RET between two chromophores, we now focus on the decisiverole played by intermediate chromophores to promote RET in multi-chromophoric systems.We first built a chromophores trimer (Fig. 3a,b) where a large gap molecule (PdPc) isseparated from a small gap molecule (H Pc) by an intermediate gap molecule (ZnPc). In thisconfiguration, the central chromophore (ZnPc) may act as an ancillary molecule, acceptingenergy from PdPc before transferring it to H Pc. The fluorescence spectrum in Fig. 3c,obtained for the tip located on PdPc (black dot in Fig. 3b), confirms this picture. It revealsemission lines characteristic of the three chromophores, including the distant H Pc acceptor.The fact that an emission of light is also observed from the PdPc and ZnPc chromophoresstrongly hints towards sequential RET processes, where each chromophore is driven in itsexcited state before decaying either radiatively or by transferring its energy to a neighboringacceptor.To further establish this mechanism, we realized HRFM maps (see Methods) for the threefluorescence lines (Fig. 3d,e,f). They show the fluorescence intensity of a given molecule asa function of the excitation position, therefore providing a real space image of the sequential6 .9 DonorAcceptor/donor STM tipRET RETAcceptor Topography P ho t on ene r g y ( e V ) Pc1.805-1.821 eVQ Pd Q Zn Q x a bc d HRFM mapsPdPcZnPcH Pc ef h � h � h � Q x Q y Q Zn y Q Zn x Q Pd x Q Pd y Intensity (kcts pC -1 eV -1 ) FIG. 3:
Cascaded RET (a) Sketch of the experiment. (b) STM image ( I = 10 pA , V = -2.5V, scale bar 1 nm) and model of a PdPc–ZnPc–H Pc trimer. The black dot in the STM imageindicates the STM-tip position used to acquire the STML spectrum ( I = 300 pA, V = -2.5 V, t = 200 s) in (c). HRFM maps ( I = 300 pA, V = -2.5 V, time per pixel = 20 s) of the emissionlines of (d) PdPc, (e) ZnPc and (f) H Pc. These maps reflect the emission intensity of a givenchromophore as a function of the tip position. Emission intensity is coded in false colors and isoverlaid on the pseudo-3D tunneling current image acquired simultaneously. The color scale rangefrom (d) 0 to 0.1, (e) 0 to 0.3 and (f) 0 to 0.6 kcts pC − eV − . RET processes with atomic-scale resolution. Fig. 3d shows that PdPc emits only when itis directly excited by the STM tip. This is expected considering that PdPc has the highestenergy gap, but also confirms that a ”direct” excitation of a molecule is impossible whenthe tip is located on a neighboring one. Interestingly, very few light is emitted when thetip excites the central part of PdPc, indicating a reduced coupling between tip plasmonsand the molecular exciton compared to cases where the tip excites the periphery of the7olecule .Fig. 3e shows an intense emission of ZnPc when the tip is located on PdPc, confirming theefficient RET observed in Fig. 2d. The ZnPc emission is intense even when the tip is locatedon the center of PdPc. As this tip position does not favor an emission of PdPc (Fig. 3d),almost all the energy stored in PdPc is transferred to ZnPc in this case. ZnPc emissionis also observed for the tip located on H Pc, an effect similar to the one reported in [17],and associated to a transfer from the H Pc Q y to the lower energy Q Zn . Surprisingly, theemission of ZnPc is weaker when the tip is directly located on top of it. Under this condition,a detailed analysis of the STML spectra of ZnPc (see Methods and Extended Data Fig. 4)reveals that the tunneling current intermittently drives the molecule in its cationic form ,ZnPc + , leading to a reduced emission of the neutral molecule (h ν ≈ Pc for a direct tip excitation, togetherwith a rather large signal for the tip located on top of the most distant PdPc molecule (seealso Methods and Extended Data Fig. 5). This last observation reflects the sequential energyfunneling from the large to the small gap molecule. Interestingly, one can compare thePdPc → H Pc RET efficiency with and without the presence of the ancillary ZnPc molecule.Despite the larger distance between the donor (PdPc) and the acceptor (H Pc), the imperfectalignment of the dipoles in the trimer and the energy lost by the emission of the ancillarymolecule (ZnPc), the RET efficiency in the trimer (up to 0.33) is only slightly reducedcompared to the dimer (0.49) for equivalent tip positions (black dots in STM images inFig. 2c and 3b). This demonstrates that sequential energy transfer can be used to funnelenergy between distant centers with high efficiency, and shows that sequential RET processeswith small energy steps may be more advantageous than single RET events with large energygaps. This general result illustrates, at the level of three chromophores, the efficient strategyused by photosynthetic systems to convey energy from collection centers to distant reactioncenters , a concept also at play in artificial multi-chromophoric components .In other multi-chromophoric configurations (Fig. 4), we have separated donors and ac-ceptors by molecules having either a larger energy-gap than the donor (Fig. 4a,c), whichshould therefore behave as passive elements, or a smaller energy-gap than the acceptor(Fig. 4b,d), a configuration where they may act as excitation traps. For sake of comparison,we have used the same donor-acceptor pairs separated by vacuum over the same distance(Fig. 4e,f). The first trimer shows that, despite its too large optical gap, the ”passive”8 .9 DonorSTM tipPassive moleculeRETAcceptor a I n t en s i t y ( a r b . un i t s ) Photon energy (eV)PdPc ZnPcH Pc cg h H Pc ZnPch � h � DonorSTM tipTrapRETAcceptor b h � h � h � PdPcZnPc H PcZnPc PdPc Q x Q Zn I n t en s i t y ( a r b . un i t s ) Photon energy (eV)Q x Q Zn Q Pd no RET no RET e df FIG. 4:
Promoting RET with passive and trap molecules (a,b) Sketches of the experiments.(c-f) STM images ( V = - 2.5 V) of a ZnPc–PdPc–H Pc trimer (c, I = 10 pA), a ZnPc–H Pc dimerseparated by 3.7 nm (e, I = 10 pA), a ZnPc–H Pc–PdPc trimer (d, I = 5 pA) and a ZnPc–PdPcdimer separated by 3.1 nm (f, I = 10 pA). Scale bars 1 nm. (g,h) STML spectra ( V = - 2.5 V)acquired at positions marked in (c)-(f). I = 300 pA, t = 120 s ( t = 300 s) for the black dot (greystar) in (g). I = 200 pA, t = 360 s ( I = 300 pA, t = 100 s) for the black dot (grey star) in (h). PdPc molecule promotes the ZnPc → H Pc RET that is absent when the passive moleculeis replaced by vacuum (Fig. 4e) . This configuration is very similar to RET experimentswhere a medium separates donors and acceptors, and highlights the role played by unex-pected species that transiently pass between donors and acceptors in liquid experiments.Eventually, this process has its importance in RET occurring in synthetic D-A compounds9eparated by an organic bridge or in the antenna complex (LH2) of some photosyntheticbacteria, where passive carotenoid molecules promote energy transfer between donor andacceptor pigments . This phenomenon has been investigated experimentally on ensemblesof molecules and theoretically for individual elements similar to the one proposed here .Whereas in a macroscopic picture it is the refraction index of the medium that modifies theefficiency of the RET process, it is the ac-polarizability of the passive molecule, allowing fora partial delocalization of the transition dipole moments of the donor and acceptor, thatpromotes dipole-dipole RET in our three-molecule model system. As the ac-polarizabitlityof a molecule primarily scales with its optical gap, we can conclude that PdPc, whose opticalgap is very close to the one of the donor molecule (ZnPc in this case), is a very good choiceto promote energy to the acceptor (see Methods and Extended Data Fig. 8 for a coupledoscillator model). Besides, the passive molecule may also lead to a higher spatial overlapbetween donor and acceptor electronic orbitals by reducing the electronic attenuation factor,therefore increasing the efficiency of RET by superexchange coupling as it was reported forD-A pairs bridged by covalently linked organic units .As a last experiment (Fig. 4b), we tested the ability of an intermediate molecule of lowenergy-gap (H Pc) to act as a trap preventing the diffusion of an excitation between distantdonor (PdPc) and acceptor units (ZnPc) of higher energies. Here, the data reveal a strongemission of the intermediate (H Pc) molecule together with a weaker luminescence from themore distant ZnPc acceptor (Fig. 4h). Hence, although most of the energy injected in thedonor is trapped by H Pc, this molecule also promotes a direct RET channel from PdPc toZnPc. Note that RET mechanisms based on sequential charge transfers can be ruled outbased on d I /d V and ”at-distance” measurements (see Methods, Extended Data Fig. 6 andExtended Data Fig. 7).Recently, several valuable approaches have been developed aiming at mimicking REToccurring in natural LHC with simplified systems such as microwave antennas , quan-tum dots or cold Rydberg atoms assemblies. These approaches, however, fail to cap-ture key properties intrinsic to organic systems, such as the influence of vibronic levels on(de-)coherence or the impact of molecular orbitals at close donor–acceptor distances. Theoriginal and most valuable aspect of our work is that it provides an extremely controlledapproach to mimic energy funnelling occurring in LHC by using the same elementary units, i.e. , molecular chromophores. This allowed us to establish, at the single molecule limit, the10ey role played by ancillary and ”passive” chromophores located between donors and accep-tors to promote RET and funnel energy. Overall our molecular-scale system constitutes aunique platform to reproduce and probe RET mechanisms occurring in multichromophoricLHC with ultimate chemical, spatial and spectral precision. Future studies will investigatethe dynamics of the RET process in STM junctions, establish the possible role of coherencebetween interacting chromophores, determine the influence of the plasmons localised at thetip-sample junction and decipher the influences of dipolar, multipolar and energy-exchangeinteractions on the RET mechanism. Mirkovic, T. et al.
Light absorption and energy transfer in the antenna complexes of photosyn-thetic organisms.
Chem. Rev. , 249–293 (2017). Balzani, V., Credi, A. & Venturi, M. Photochemical conversion of solar energy.
ChemSusChem , 26–58 (2008). Arsenault, E. A., Yoneda, Y., Iwai, M., Niyogi, K. K. & Fleming, G. R. Vibronic mixing enablesultrafast energy flow in light-harvesting complex II.
Nat. Commun. , 1460 (2020). Ma, F., Romero, E., Jones, M. R., Novoderezhkin, V. I. & van Grondelle, R. Both electronicand vibrational coherences are involved in primary electron transfer in bacterial reaction center.
Nat. Commun. , 933 (2019). Doppagne, B. et al.
Single-molecule tautomerization tracking through space- and time-resolvedfluorescence spectroscopy.
Nat. Nanotechnol. , 207 (2020). Pullerits, T., Hess, S., Herek, J. L. & Sundstr¨om, V. Temperature dependence of excitationtransfer in LH2 of Rhodobacter sphaeroides.
J. Phys. Chem. B , 10560–10567 (1997). Scholes, G. D. & Fleming, G. R. On the mechanism of light harvesting in photosynthetic purplebacteria: B800 to B850 energy transfer.
J. Phys. Chem. B , 1854–1868 (2000). Andrews, D., Curutchet, C. & Scholes, G. Resonance energy transfer: Beyond the limits.
LaserPhotonics Rev. , 114–123 (2011). Engel, G. S. et al.
Evidence for wavelike energy transfer through quantum coherence in photo-synthetic systems.
Nature , 782 (2007). Lim, J. et al.
Vibronic origin of long-lived coherence in an artificial molecular light harvester.
Nat. Commun. , 7755 (2015). Trofymchuk, K. et al.
Giant light-harvesting nanoantenna for single-molecule detection inambient light.
Nat. Photon. , 657–663 (2017). Bian, Q. et al.
Vibronic coherence contributes to photocurrent generation in organic semicon-ductor heterojunction diodes.
Nat Commun. , 617 (2020). Qiu, X. H., Nazin, G. V. & Ho, W. Vibrationally resolved fluorescence excited with submolecularprecision.
Science , 542–546 (2003). Merino, P., Große, C., Ros(cid:32)lawska, A., Kuhnke, K. & Kern, K. Exciton dynamics of C -basedsingle-photon emitters explored by Hanbury Brown-Twiss scanning tunnelling microscopy. Nat.Commun. , 8461 (2015). Chong, M. C. et al.
Narrow-line single-molecule transducer between electronic circuits andsurface plasmons.
Phys. Rev. Lett. , 036802 (2016). Zhang, Y. et al.
Visualizing coherent intermolecular dipole-dipole coupling in real space.
Nature , 623 (2016). Imada, H. et al.
Real-space investigation of energy transfer in heterogeneous molecular dimers.
Nature , 364–367 (2016). Doppagne, B. et al.
Vibronic spectroscopy with submolecular resolution from STM-inducedelectroluminescence.
Phys. Rev. Lett. , 127401 (2017). Imada, H. et al.
Single-molecule investigation of energy dynamics in a coupled plasmon-excitonsystem.
Phys. Rev. Lett. , 013901 (2017). Zhang, L. et al.
Electrically driven single-photon emission from an isolated single molecule.
Nat. Commun. , 580 (2017). Doleˇzal, J. et al.
Charge carrier injection electroluminescence with CO-functionalized tips onsingle molecular emitters.
Nano Lett. , 8605–8611 (2019). Patera, L. L., Queck, F., Scheuerer, P., Moll, N. & Repp, J. Accessing a charged intermediatestate involved in the excitation of single molecules.
Phys. Rev. Lett. , 016001 (2019). Schull, G., N´eel, N., Johansson, P. & Berndt, R. Electron-plasmon and electron-electron inter-actions at a single atom contact.
Phys. Rev. Lett. , 057401 (2009). Hinze, G., M´etivier, R., Nolde, F., M¨ullen, K. & Basch´e, T. Intramolecular electronic excitationenergy transfer in donor/acceptor dyads studied by time and frequency resolved single moleculespectroscopy.
J. Chem. Phys. , 124516 (2008). Kr¨oger, J., Doppagne, B., Scheurer, F. & Schull, G. Fano description of single-hydrocarbon uorescence excited by a scanning tunneling microscope. Nano Lett. , 3407–3413 (2018). Doppagne, B. et al.
Electrofluorochromism at the single-molecule level.
Science , 251–255(2018). Hippius, C. et al.
Sequential fret processes in calix[4]arene-linked orange-red-green perylenebisimide dye zigzag arrays.
J. Phys. Chem. C , 2476–2486 (2008). Curutchet, C., Mennucci, B., Scholes, G. D. & Beljonne, D. Does F¨orster theory predict therate of electronic energy transfer for a model dyad at low temperature?
J. Phys. Chem. B ,3759–3766 (2008). Andrews, D. L. & Ford, J. S. Resonance energy transfer: Influence of neighboring matterabsorbing in the wavelength region of the acceptor.
J. Chem. Phys , 014107 (2013). Pettersson, K. et al.
Singlet energy transfer in porphyrin-based donor-bridge-acceptor systems:Interaction between bridge length and bridge energy.
J. Phys. Chem A , 310–318 (2006). Rustomji, K. et al.
Direct imaging of the energy-transfer enhancement between two dipoles ina photonic cavity.
Phys. Rev. X , 011041 (2019). Kodaimati, M. S., Lian, S., Schatz, G. C. & Weiss, E. A. Energy transfer-enhanced photo-catalytic reduction of protons within quantum dot light-harvesting–catalyst assemblies.
Proc.Natl. Acad. Sci. U.S.A. , 8290–8295 (2018). Ravets, S. et al.
Coherent dipole–dipole coupling between two single Rydberg atoms at anelectrically-tuned F¨orster resonance.
Nat. Phys. , 914–917 (2014). MethodsExperimental
The STM data were acquired with a low temperature (4.5 K) Omicron setup operatingin ultrahigh vacuum adapted to detect the light emitted at the tip-sample junction. Theoptical detection setup is composed of a spectrograph coupled to a CCD camera andprovides a spectral resolution of ≈ Pc molecules were evaporated on the cold ( ≈ Pc molecules were purchased from Sigma–Aldrich, whilePdPc molecules were synthesized according to the procedure described in [34].The ZnPc molecules were manipulated by positioning the STM tip at the edge of themolecule at a bias V = +2.5 V or V = -2.5 V and then reducing the tip-molecule distanceuntil a molecular jump occurs . The H Pc and PdPc were manipulated using voltage pulses(of up to -4 V at a set point of V = -2.5 V and I = 10 pA) applied at the edge of the molecule.HRFM maps presented in Fig. 3 were generated by scanning the molecular trimer withthe STM tip while recording a STML spectrum for each tip position (that is, pixel ofthe map). The tip–substrate separation was kept constant during the acquisition (openfeedback loop) to prevent any distance-related artefacts. To compensate for the tip driftduring the long acquisition of the HRFM maps, the ( x , y , z ) position of the tip was correctedusing an ”atom-tracking” procedure between the acquisition of each pixel, similarly tothe procedure developed in [35]. The next step consisted in choosing the photon energywindows corresponding to the spectral features of interest ( e.g. , Q Pd , Q Zn and Q x in Fig. 3)to generate the corresponding photon intensity maps. All HRFM maps were recordedsimultaneously in a single experimental run and could be readily compared. For eachpixel of the maps, the photon intensity was normalized by the STM current recordedsimultaneously so as to consider an equivalent excitation source for each pixel. Estimation of the spectral overlaps for the donor–acceptor dimers
Extended Data Fig. 2 shows the fluorescence (F D ) and absorption ( (cid:15) A ) spectra for thedonors (D) and acceptors (A) forming dimers in Fig. 2 of the main manuscript. The donorfluorescence spectra correspond to the single molecule STML spectra presented in Fig. 1of the main manuscript. The acceptor absorption spectra are obtained by mirroring thefluorescence spectra at the maximum intensity energy, assuming a negligible Stokes shift asis generally expected for low temperature spectra of rigid phthalocyanine molecules . Notethat for the H Pc molecule we neglect the intensity of the Q y mode as its energy is higherthan Q Pd and Q Zn . The spectra are normalized such that:14 ∞ F D ( E ) dE = 1 (1) (cid:90) ∞ (cid:15) A ( E ) dE = 1 (2)The normalized spectra are then used to calculate the spectral overlap J as defined forthe exchange interaction : J = (cid:90) ∞ F D ( E ) (cid:15) A ( E ) dE (3)The J values for the three studied D–A pairs, together with the energy differences be-tween D and A main emission lines (∆ E ), and the RET efficiency values are reported inExtended Data Fig. 2d. Similarly to what is observed with other methods, this table showsthat the RET efficiency in STM junctions primarily depends on the spectral overlap be-tween the donor and the acceptor. This table also shows that the energy difference (∆ E )between the donor and acceptor emission lines can be used as an acceptable and convenientapproximation to intuit the energy transfer efficiency at low temperature between a donorand an acceptor. However, as the RET efficiency is also affected by several other parameters(D–A distance, dipole–dipole orientations...) that are slightly different from one D–A pairto the other, one does not find the expected proportional dependency of the RET efficiencywith J . The effect of the relative dipole orientation on RET
In the F¨orster description of RET, the relative orientation of the donor and acceptordipoles plays a crucial role on the efficiency of the energy transfer process. It is reflected inthe dipole orientation factor κ defined as : κ = (sin θ D sin θ A cos φ − θ D cos θ A ) (4)with the angles θ D and θ A as defined in Extended Data Fig. 3a. φ is the dihedral anglebetween the planes of the dipoles, which in our case (dipoles in the same plane) is φ = 0.Formula 4 is derived under the assumption of point dipoles.We calculated κ (Extended Data Fig. 3b) for the PdPc–ZnPc dimer presented in Fig.2a of the main manuscript. Here, it is assumed that placing the tip at the extremity of one15f the pyrrole sub-unit of the donor results in the specific excitation of the correspondingdipole (Q Pd x or Q Pd y ). From there, the transfer of energy to both dipoles of the acceptor(Q Zn x and Q Zn y ) is considered.We first analyze the excitation of Q Pd x (black dot on Extended Data Fig. 3c), whichresults in a very efficient energy transfer (RET eff = 0.96) to ZnPc (black spectrum inExtended Data Fig. 3d and Extended Data Fig. 3b). Here, the Q Pd x and Q Zn x dipoles arenearly aligned, corresponding to an optimum RET configuration as confirmed by a κ valueclose to κ max = 4. The estimation of κ for the Q Pd x −→ Q Zn y energy transfer leads to amuch smaller value (0.43) indicating a negligible contribution of this RET path.Excitation of Q Pd y (gray star in Extended Data Fig. 3c) results in a less efficient RETprocess (gray spectrum in Extended Data Fig. 3e and Extended Data Fig. 3b). Here, theRET process is strongly dominated by the close-to-parallel configuration of the Q Pd y −→ Q Zn y RET path, leading to a κ value of 1.06.Overall, the PdPc–ZnPc dimer allows one to compare the dependence of the RET effi-ciency with dipole-dipole orientations for close-to ideal inline and parallel configurations. Inagreement with the F¨orster theory, the former leads to a more efficient energy transfer thanthe latter, though only by a factor ≈ ≈ κ in equation 4. The very short distance separating the donor and the acceptor inour experiment explains this quantitative discrepancy.Indeed, the F¨orster approach of RET holds only for D–A distances much larger than thesizes of the chromophores, which is not the case in our experiment. For such short D–Adistances, the point dipole approximation does not hold , the dipole orientations may beaffected by the proximity with molecular neighbors ; multipolar contributions to the RETcan no longer be neglected, and exchange energy RET process ( i.e. , Dexter energy transfer)may play a significant role.Considering the influence of all these parameters, it is remarkable that the variationof the RET efficiency with the tip position in our STML experiment can be qualitativelyaccounted for by considering the F¨orster approach. Additional discussion on the influenceof the dipole alignment on the RET process can be found in the ”RET efficiency maps”section and Extended Data Fig. 5. 16 harge state of the ZnPc molecule Extended Data Fig. 4b and c compare large energy range STML spectra obtained from aZnPc molecule in a PdPc–ZnPc dimer for positions marked in the STM image in ExtendedData Fig. 4a. The spectrum in Extended Data Fig. 4b corresponds to a direct excitationof the molecule ( i.e. , the tip is located on top of the ZnPc molecule), while the spectrumin Extended Data Fig. 4c displays the indirect excitation of the ZnPc fluorescence ( i.e. thetip is located on top of the PdPc molecule and the luminescence of ZnPc results from thePdPc −→ ZnPc energy transfer). As discussed in the main manuscript, an indirect excitationleads to a 13 times more efficient excitation of the ZnPc fluorescence than a direct excitation(note the different vertical scales in Extended Data Fig. 4b and c).The spectrum in Extended Data Fig. 4b also reveals a spectral feature at ≈ + , the positively charged molecule . In this paper,it was concluded that the tunneling current traversing the molecule is responsible for afast switching between the neutral and cationic states during STML spectra acquisition.This leads to the observation of both the charged and neutral contributions in the STMLspectra of ZnPc/NaCl/Au(111). The spectrum in Extended Data Fig. 4b indicates a similarbehavior for ZnPc/NaCl/Ag(111), although with a lower relative intensity of the chargedpeak with respect to the neutral one compared to Au(111).In contrast, the charged contribution is absent in the STML spectrum of Extended DataFig. 4c, corresponding to an indirect excitation of the ZnPc, confirming that the tunnelingcurrent must pass through the ZnPc molecule to charge it. In turn, the molecule spends asmaller fraction of time in the neutral state for a direct excitation, leading thus to a reducedemission intensity at hν ≈ Pcand PdPc molecules in the HRFM map of Fig. 3e.17
ET efficiency maps
Based on the HRFM maps presented in Fig. 3d, e, f of the main manuscript, one canreconstruct RET efficiency maps by applying the formula
RET eff = I A / ( I P dP c + I ZnP c + I H P c ) to each pixel of the map. The result of this operation is presented in Extended DataFig. 5a and b for H Pc and ZnPc molecules acting both as acceptors.By displaying normalized photon intensities, these maps reveal the RET signals frommolecules having low excitation probabilities, such as ZnPc, which are too weak to bevisualized in HRFM maps. For example in Extended Data Fig. 5a, one observes a highRET efficiency to the H Pc molecule when the tip is located on the adjacent ZnPc donor, abehavior that is hidden in the HRFM map in Fig. 3e. Interestingly, the impact of the dipolealignment on the RET is here readily visible: the excitation of Q Zn x , which is well-alignedwith Q x , leads to a quite efficient ( RET eff > .
5) energy transfer to the H Pc acceptor. Incontrast, the excitation of Q Zn y results in a low RET eff due to the near-parallel orientationof the Q Zn y and Q y dipoles (see ”The effect of the relative dipole orientation on RET”section and Extended Data Fig. 3 for more details on the impact of dipole–dipole angles on RET eff ).A similar discussion can be proposed for the RET efficiency map from PdPc to ZnPc(Extended Data Fig. 5b). Here, the RET efficiency is equivalent for the STM tip located ontop of the two donor dipoles, Q Pd x and Q Pd y . This is a consequence of the specific dipolealignment of the donor and acceptor molecules: the Q Pd x and Q Zn y dipoles, and the Q Pd y and Q Zn y dipoles make an angle of ≈ ◦ .In conclusion, RET efficiency maps provide interesting complementary information toHRFM maps by highlighting the role of dipole orientations and allowing to visualize RETfrom molecular donors having a low excitation probability. Excluding charge transfer mechanisms
In their recent publication on energy transfer Imada et al. compared d I /d V data ac-quired on both isolated and paired donor/acceptor molecules to exclude an energy transfermechanism based on sequential charge transfer (CT) steps. In Extended Data Fig. 6, weproduce similar data acquired for individual chromophores and for the trimer of Fig. 4c.18he orbital positions in the d I /d V spectra of individual H Pc, PdPc, and ZnPc (ExtendedData Fig. 6a) are very similar to the ones deduced from d I /d V spectra measured along aline crossing a H Pc–PdPc–ZnPc trimer (Extended Data Fig. 6b), indicating that there is nomeasurable hybridization between the ground state orbitals of the chromophores. Followingthe argument developed in [17], we note that the highest occupied orbital of the ZnPc donoris at a higher energy than the one of the PdPc passive and H Pc acceptor molecules, pre-venting a migration of holes that would be injected with the STM tip in the ZnPc molecule,and excluding a CT mediated energy transfer mechanism. Another, more straightforward,evidence that CT mechanisms are not responsible for the RET processes is provided in Ex-tended Data Fig. 7, where we demonstrate RET in the absence of charge injection into thedonor. For that, we located the tip at a ≈ Pc donor-acceptor pair and generate a so-called ”at-distance” excita-tion of the donor . Once normalized by a plasmon reference spectrum ( I ), this STMLspectrum (Extended Data Fig. 7c) reveals the signature of both PdPc and H Pc molecules.As no charge is running through the pair in this case, a contribution of CT to the RETmechanism can be safely ruled out.
Modeling the role of an intermediate molecule with a classical oscillatory approach
The energy transfer between a donor (D) and an acceptor (A) molecule via an interme-diate molecule (I) (Fig. 3 and Fig. 4) is illustrated by a coupled harmonic oscillator model(Extended Data Fig. 8) composed of three pendulums, where the donor pendulum (D) iscoupled to an acceptor pendulum (A) through an intermediate pendulum (I). The coupledoscillator equations describing the time evolution of the system read¨ θ D ( t ) = − (cid:0) ω D + ω (cid:1) θ D ( t ) + ω θ I ( t ) − ω D Q D ˙ θ D ( t )¨ θ I ( t ) = − (cid:0) ω I + Ω + Ω (cid:1) θ I ( t ) + Ω θ D ( t ) + Ω θ A ( t ) − ω I Q I ˙ θ I ( t ) , ¨ θ A ( t ) = − ( ω A + ω ) θ A ( t ) + ω θ I ( t ) − ω A Q A ˙ θ A ( t ) , Ω = 1 α D ω ω I ω D , Ω = 1 α A ω ω I ω A . θ i ( t ) , i = D, I, A are the angles of the donor, intermediate and acceptor pendu-lums with respect to their equilibrium positions at time t . α D = m I /m D and α A = m I /m A are the mass ratios; in this paper the masses are considered identical for all pendulumsso that α D = α A = 1. ω D , ω I , ω A are the eigenfrequencies of the uncoupled pendulums.Coupling constants between (D) and (I), and (I) and (A) are ω , ω , respectively; note that,in this model, there is no direct coupling between pendulum (D) and (A), so that pendulum(I) acts as a relay of the excitation. We consider damping with quality factors Q D , Q I , Q A ,taken all identical for the simulations ( Q = 60).All pendulums are in their equilibrium positions ( θ = 0 rad) at t = 0. To mimic theexcitation of the donor by the tunneling current, pendulum (D) is provided with an initialangular speed ( ˙ θ D ( t = 0) = 1 rad /s ); the initial angular speed of (I) and (A) are zero.The time evolution of the system is computed using a velocity-Verlet algorithm . Theangles at time t + δt are computed from their values at time t by using θ D ( t + δt ) = (cid:26) − δt (cid:0) ω D + ω (cid:1)(cid:27) θ D ( t ) + δt ω θ I ( t ) + (cid:18) − δt ω D Q D (cid:19) δt ˙ θ D ( t ) ,θ I ( t + δt ) = (cid:26) − δt (cid:0) ω I + Ω + Ω (cid:1)(cid:27) θ I ( t ) + δt θ D ( t ) + δt θ A ( t ) + (cid:18) − δt ω I Q I (cid:19) δt ˙ θ I ( t ) ,θ A ( t + δt ) = (cid:26) − δt (cid:0) ω A + ω (cid:1)(cid:27) θ A ( t ) + δt ω θ I ( t ) + (cid:18) − δt ω A Q A (cid:19) δt ˙ θ D ( t ) . Then, the values of the angular speeds at time t + δt are computed from their values at time t and from the values of the angles at times t and t + δt by using˙ θ D ( t + δt ) = (cid:18) − δt (cid:0) ω D + ω (cid:1) ( θ D ( t + δt ) + θ D ( t ))+ δt ω ( θ I ( t + δt ) + θ I ( t )) + (cid:18) − δt ω D Q D (cid:19) ˙ θ D ( t ) (cid:19) (cid:18) δt ω D Q D (cid:19) − , ˙ θ I ( t + δt ) = (cid:18) − δt (cid:0) ω I + Ω + Ω (cid:1) ( θ I ( t + δt ) + θ I ( t )) + δt ( θ D ( t + δt ) + θ D ( t ))+ δt ( θ A ( t + δt ) + θ A ( t )) + (cid:18) − δt ω I Q I (cid:19) ˙ θ I ( t ) (cid:19) (cid:18) δt ω I Q I (cid:19) − , ˙ θ A ( t + δt ) = (cid:18) − δt (cid:0) ω A + ω (cid:1) ( θ A ( t + δt ) + θ A ( t ))+ δt ω ( θ I ( t + δt ) + θ I ( t )) + (cid:18) − δt ω A Q A (cid:19) ˙ θ A ( t ) (cid:19) (cid:18) δt ω A Q A (cid:19) − . y repeating this procedure several thousand times, one calculates how the initial excitationof (D) is transferred to (I) and (A). In the following, we will show how this transfer of excitationvaries as a function of the resonance frequency of (I). As such, this simple classical model can beused to infer the impact of the intermediate molecule energy-gap size on the RET efficiency in themolecular structures investigated in Fig. 3 and Fig. 4.In Extended Data Fig. 8b,c,d we represent the oscillation amplitudes of the three pendulumsas a function of the normalized time t/T D (where T D = 2 π/ω D ), for a large (b), a medium (c) anda low (d) eigenfrequency of the intermediate pendulum. 20000 steps with a time step δt = T D / ω I /ω D by integrating over the time the work of the friction forces for each oscillator(Extended Data Fig. 8e). In panels (b) and (e), we see that a very small fraction of the originalexcitation is dissipated in the intermediate pendulum ( i.e. , low amplitude of the blue oscillations),while a larger fraction is dissipated in the acceptor pendulum ( i.e. , larger amplitude of the redoscillations), reflecting the ability of a ”passive” molecule to promote RET processes between adonor and an acceptor. When the intermediate pendulum eigenfrequency is considered in-betweenthe donor and acceptor ones (Extended Data Fig. 8d,e), the energy dissipated in the intermediatependulum increases strongly, which is expected considering the near-resonant condition. The en-ergy dissipated in the acceptor pendulum increases as well, confirming that the efficiency of theRET process is improved when the intermediate molecule can serve as an ancillary unit. Even-tually, the case of a low-eigenfrequency intermediate pendulum is investigated (d), revealing aweak dissipation by the intermediate and acceptor pendulums, somehow supporting the idea thatlow-energy gap molecules can be used to block a RET path.While this simple classical model provides an intuitive understanding of the role played by theenergy-gap of the intermediate molecule, it fails to capture the quantum nature of the RET process( e.g. , exchange interactions, which reflects a Dexter RET process, are not accounted for by thisapproach). Such effects require a time-dependent quantum approach, which is beyond the scopeof this study. Lokesh, K. S. & Adriaens, A. Synthesis and characterization of tetra-substituted palladiumphthalocyanine complexes.
Dyes Pigm. , 269 – 277 (2013). Kawai, S., Glatzel, T., Koch, S., Baratoff, A. & Meyer, E. Interaction-induced atomic dis-placements revealed by drift-corrected dynamic force spectroscopy.
Phys. Rev. B , 035421(2011). Murray, C. et al.
Visible luminescence spectroscopy of free-base and zinc phthalocyaninesisolated in cryogenic matrices.
Phys. Chem. Chem. Phys. , 17543–17554 (2011). Speiser, S. Photophysics and mechanisms of intramolecular electronic energy transfer in bichro-mophoric molecular systems: solution and supersonic jet studies.
Chem. Rev. , 1953–1976(1996). Wong, K. F., Bagchi, B. & Rossky, P. J. Distance and orientation dependence of excitationtransfer rates in conjugated systems: Beyond the F¨orster theory.
J. Phys. Chem. A ,5752–5763 (2004). Zhang, Y. et al.
Sub-nanometre control of the coherent interaction between a single moleculeand a plasmonic nanocavity.
Nat. Commun. , 15225 (2017). Reecht, G. et al.
Electroluminescence of a polythiophene molecular wire suspended between ametallic surface and the tip of a scanning tunneling microscope.
Phys. Rev. Lett. , 047403(2014). Verlet, L. Computer ”experiments” on classical fluids. I. Thermodynamical properties ofLennard-Jones molecules.
Phys. Rev. , 98 (1967).
Data availability
The data that support the plots within this paper and other findings of this study are availablefrom the corresponding authors (A.R. and G.S.) upon reasonable request. cknowledgements The authors thank Virginie Speisser for technical support and Alex Boeglin for discussions.This project has received funding from the European Research Council (ERC) under the Euro-pean Union’s Horizon 2020 research and innovation program (grant agreement No 771850) and theEuropean Union’s Horizon 2020 research and innovation programme under the Marie Sklodowska-Curie grant agreement No 894434. The Agence National de la Recherche (project Organiso No.ANR-15-CE09-0017), the Labex NIE (Contract No. ANR-11-LABX-0058 NIE), and the Interna-tional Center for Frontier Research in Chemistry (FRC) are acknowledged for financial support.
Author contributions
S.C., A.R., B.D. and G.S. conceived, designed and performed the experiments. S.C., A.R.,M.R., F.S. and G.S. analysed the experimental data. F.S. and H.B. performed the oscillatorymodel. M.F. and F.C. synthesised the PdPc chromophores. All the authors discussed the resultsand contributed to the redaction of the paper. hoton energy (eV) N o r m . i n t en s i t y ( a r b . un i t s ) Photon energy (eV) N o r m . i n t en s i t y ( a r b . un i t s ) PdPcH Pc H PcR = 1.68 nmR = 1.97 nmR = 2.24 nmR = 2.47 nmR = 2.72 nmR = 3.20 nm R = 1.68 nmR = 1.97 nmR = 2.24 nmR = 2.47 nmR = 2.72 nmR = 3.20 nm c d
PdPcH PcR a R E T e ff i c i en cy R (nm) b Extended Data Fig. 1: Spectra used to obtain the D–A distance dependence presented in Fig. 2i ofthe main manuscript. (a) STM image of the H Pc–PdPc dimer, I = 5 pA, V = − . RET eff values calculated from the spectra in (c). The horizontal error bars considera 5 % error in the estimation of the molecules positions. The vertical error bars are smaller thanthe symbol size as the statistical error on the number of photon counts is less than 3 %. (c)Normalized STML spectra acquired at positions marked in Fig. 2h, V = − . t = 60 s, I = 50 pA (for R = 2 .
24 nm), I = 100 pA (for R = 1 . , . , . I = 250 pA (for R = 2 . , .
72 nm). The spectra were normalized by the plasmonic response of the cavity toensure a fair comparison between the intensities of the molecular emission lines, and scaled tounity. The splits observed in the PdPc spectra correspond to a partial lifting of the Q Pd x andQ Pd y degeneracy that seems to depend on details of the adsorption site. A similar effect has beenreported previously for H Pc. The
RET eff values were estimated by integrating the light emissionintensities in the spectral ranges indicated in blue and red. (d) Enlarged view of the H Pc linedisplayed in (c). ZnPc (absorption)PdPc (emission) H Pc (absorption)ZnPc (emission) N o r m . i n t en s i t y ( a r b . un i t s ) Photon energy (eV)
PdPc (emission)H Pc (absorption) abc d
PdPc → ZnPcZnPc → H PcPdPc → H Pc Extended Data Fig. 2: Normalized emission and absorption spectra for (a) PdPc–ZnPc (b) ZnPc–H Pc and (c) PdPc–H Pc dimers. (d) Spectral overlaps ( J ), energy differences (∆ E ) and RETefficiencies for the dimer configurations presented in Fig. 2. D Q A R � D � A a b PdPcZnPc PdPcZnPc c Q Pd Q Zn Q Zn y Q Pd yx x Photon energy (eV) Photon energy (eV) I n t en s i t y ( a r b . un i t s ) I n t en s i t y ( a r b . un i t s ) Q Zn Q Pd Q Zn Q Pd d e RET
Eff = 0.69RET
Eff = 0.96
Extended Data Fig. 3: The effect of the relative dipole orientation on RET. (a) Geometricalconfiguration of the donor and acceptor transition dipoles in a dimer. R is the vector joining thecenters of the dipoles. (b) κ values calculated for the dimer configurations presented in (c). (c)Top: STM image ( V = − . I = 10 pA). Scale bar 1 nm. Bottom: ball-and-stick models of thePdPc–ZnPc donor–acceptor pair with the respective dipoles indicated. (d,e) Plasmon-correctedSTML spectra ( V = − . I = 300 pA, acquisition time t = 300 s) recorded at the positionsmarked by a black dot (d) and gray star (e) in (c). The main emission lines of each molecule arehighlighted in color in the spectra. These data are also presented in Fig. 2a and Fig. 2d of themain manuscript. dPcZnPc a Photon energy (eV) I n t en s i t y ( kc t s p C - e V - ) c Q Zn Q Pd Indirect ZnPc excitation ( ) b Q Zn Q Zn + Direct ZnPc excitation ( )
Extended Data Fig. 4: Charge state of the ZnPc molecule. (a) STM image ( V = − . I = 10 pA)of a PdPc-ZnPc dimer. Scale bar 1 nm. (b) and (c) STML spectra recorded at positions markedin (a), V = − . I = 300 pA, acquisition time t = 180 s for (b) and t = 100 s (c). Note thedifferent vertical scales. Pc ZnPcRET eff to H Pc RET eff to ZnPcH PcZnPc PdPc PdPc a b Q x Q y Q Zn x Q Zn y Q Pd y Q Pd x Q Pd y Q Pd x Q Zn y Q Zn x Q x Q y Extended Data Fig. 5: RET efficiency maps for H Pc and ZnPc as acceptors in the PdPc-ZnPc-H Pc trimer. The circles mark the spatial extension of the acceptor, the arrows indicate thetransition dipoles of the labelled chromophores. The high-intensity areas denote precise locationswhere the sub-molecular excitation of the donor results in an efficient energy transfer to the acceptor(H Pc in (a) and ZnPc in (b)). The color scales range from 0 to 1. d I / d V s i gna l hiloDistance (nm) B i a s ( V ) H Pc PdPc ZnPcH Pc PdPc ZnPc -3 -2 -1 0 1 20510
PdPc moleculeZnPcmoleculeH Pc moleculeBias (V) d I / d V ( a r b . un i t s ) a b Extended Data Fig. 6: Comparison of the single-molecule and trimer d I /d V spectra. (a) Leftpanel: d I /d V spectra recorded on individual H Pc, PdPc and ZnPc. Set-point: V = -3 V, I = 15 pA. Right panel: STM images of the corresponding individual molecules, the dots indicatethe positions where the d I /d V spectra were acquired. V = -2.5 V, I = 5 pA. (b) A series of 25d I /d V spectra acquired along the H Pc–PdPc–ZnPc trimer (following the black arrow in inset).Set-point: V = -3 V, I = 15 pA. Inset: STM image of the studied trimer, V = -2.5 V, I = 5 pA.All scale bars are 1 nm. .51.01.6 1.7 1.8 1.9 2.0 2.11.01.21.4 Photon energy (eV) II I n t en s i t y ( a r b . un i t s ) I / I H PcPdPc Q x Q Pd bc DonorAcceptor a RETh � h � PlasmonSTM tip
Extended Data Fig. 7: RET excited ”at-distance”. (a) Sketch of the experiment. (b) STMLspectra acquired with the STM tip located at a distance of r = 2.2 nm (upper curve, I ) and r = 4nm (bottom curve, I ) from the center of the PdPc molecule. Inset: STM image V = -2.5 V, I =5 pA of the investigated PdPc–H Pc dimer. The dot and star marked the positions at which thespectra have been recorded. Scale bar 1 nm. (c) Normalized spectrum
I/I . a) (d) (D) (I) (A) (b) (D) (I) (A) (c) (D) (I) (A) (e) (d) (c) (b) Extended Data Fig. 8: Modeling the role of an intermediate molecule with a classical oscillatoryapproach. (a) Graphical representation of the three-pendulum model. Oscillation amplitudes ofthe three pendulums as function of the normalized time t / T D for a large (b), a medium (c) and asmall (d) eigenfrequency of the intermediate pendulum, where T D = 2 π/ω D . (e) Fraction of theexcitation energy dissipated by the donor (D), the intermediate (I) and acceptor (A) pendulumsas a function of the normalized intermediate pendulum eigenfrequency ω I /ω D ..