Excitation energy-dependent photocurrent switching ina single-molecule photodiodeBing Shana, Animesh Nayakb, Olivia F. Williamsa, Dillon C. Yosta, Nicholas F. Polizzic, Yanming Liud, Ninghao Zhoua,Yosuke Kanaia, Andrew M. Morana, Michael J. Therienb, and Thomas J. Meyera,1

aDepartment of Chemistry, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599; bDepartment of Chemistry, Duke University, Durham, NC27708; cDepartment of Pharmaceutical Chemistry, University of California, San Francisco, CA 94158; and dKey Laboratory of Industrial Ecology andEnvironmental Engineering, Dalian University of Technology, 116024 Dalian, China

Contributed by Thomas J. Meyer, July 2, 2019 (sent for review April 26, 2019; reviewed by Neal Armstrong and Victor S. Batista)

The direction of electron flow in molecular optoelectronic devicesis dictated by charge transfer between a molecular excited stateand an underlying conductor or semiconductor. For those devices,controlling the direction and reversibility of electron flow is a majorchallenge. We describe here a single-molecule photodiode. It is basedon an internally conjugated, bichromophoric dyad with chemicallylinked (porphyrinato)zinc(II) and bis(terpyridyl)ruthenium(II) groups.On nanocrystalline, degenerately doped indium tin oxide electrodes,the dyad exhibits distinct frequency-dependent, charge-transfercharacters. Variations in the light source between red-light (∼1.9 eV)and blue-light (∼2.7 eV) excitation for the integrated photodioderesult in switching of photocurrents between cathodic and anodic.The origin of the excitation frequency-dependent photocurrents liesin the electronic structure of the chromophore excited states, asshown by the results of theoretical calculations, laser flash photoly-sis, and steady-state spectrophotometric measurements.

photoelectrochemical cell | metal oxide electrode | bichromophoric dyad |excited-state transitions

Devices based on molecular electronics mimic traditionalmacroscopic semiconductor devices, but provide unprece-

dented advantages in scalable size, synthetic flexibility, and con-trollable electronic properties (1, 2). With those devices, electricaland electronic processes can be modified by varying functionalelectrode elements and integrating individual molecular compo-nents that respond to external stimuli, such as electrochemicalredox potentials, optical excitation, and magnetic fields (3–7). Photo-responsive molecular diodes have been used in dye-sensitizedphotoelectrodes to generate electron-hole pairs for catalytic andoptoelectronic applications (8–14). They convert solar energy intoelectrical signals and chemical energy by transfer of photo-generated electrons following optical excitation. For those devices,controlling the direction of the electron flow is a challenge, even formultichromophoric systems, because of competition from excited-state relaxations and back electron transfer (15–17). Creating a localbasis for controlling the direction of electron flow requires theformation of chromophoric excited states with aligned energetics forrapid charge injection into the electrode substrates. Under a certainrange of applied bias, a localized interfacial charge-separation statecan be used to produce sustainable external photocurrents.Here, we describe a dye-sensitized photoelectrode that gen-

erates photocurrents with direction controllable by varying thefrequency of the excitation photons. The photoelectrode wasfabricated with a degenerately doped, nanocrystalline indium tinoxide electrode (nanoITO) (18), which was sensitized by a highlyconjugated, supramolecular chromophore. The chromophore[abbreviated as (PhZnRu)2+] in Fig. 1A and a structural analog[abbreviated as (RuZn)2+] in Fig. 1B are ethyne-bridged (por-phyrinato)zinc(II) and bis(terpyridyl)ruthenium(II) complexeswith carboxylic or phosphonic acid groups for surface binding tonanoITO electrodes. This connectivity aligns the bis(terpyridyl)ruthenium(II) and (porphyrinato)zinc(II) transition dipoles in ahead-to-tail arrangement, enforcing extensive excited-state elec-tronic communication between these units and driving significant

charge-transfer (CT) character in the chromophore’s three low-energy, singlet excited-state manifolds (19–24). The two chromo-phores on nanoITO exhibit frequency-dependent, excited-state CTcharacters at their excited states. Alternation of the excitationsource between red and blue light for the integrated photodiodewith (PhZnRu)2+ on nanoITO results in switching of photocurrentsbetween cathodic and anodic. Based on the results from theoreticalcalculations, transient absorption (TA), and steady-state spectro-photometric measurements, the excitation-frequency-dependentphotocurrents originate from the changes in the electron-densitydistributions within the chromophore excited states.

Results and DiscussionMolecular Design and Electronic Spectroscopy. Characterizationdata for (PhZnRu)2+ and (RuZn)2+ are shown in SI Appendix,Figs. S1–S6. The electronic absorption spectra recorded for(PhZnRu)2+ and (RuZn)2+ in Fig. 1 differ markedly from thosecharacteristics of individual (porphyrinato)zinc(II) (abbreviatedas Zn) and bis(terpyridyl)ruthenium(II) (abbreviated as Ru2+)complexes. In (PhZnRu)2+ and (RuZn)2+, extensive mixing ofPZn B- and Q-electronic states with ruthenium terpyridyl metal–ligand CT states gives rise to supramolecular chromophores thatmanifest high oscillator strength absorptive manifolds that providean unusual degree of spectral coverage of the 250- to 750-nmwavelength domain (19–24). Note that (PhZnRu)2+ is anchored

Significance

In optoelectronic devices based on the properties of molecularexcited states, a major challenge exists in controlling wavelength-dependent properties, because of the rapid interconversion be-tween excited states. In photoelectrochemical cells, based on dye-sensitized photoelectrodes, surface functionalization plays a sig-nificant role in directing the absorption and conversion of in-cident photons.We demonstrate here that the direction of photo-generated charge flow in an oxide-based photoelectrode can becontrolled by varying the excited-state transitions in a surface-bound, highly conjugated supramolecular chromophore withimportant implications for photoelectrode design in moleculardevices. This approach, with a well-established correlation be-tween conjugated chromophores and oxide electrodes, can serveas an effective platform for further development of solar-energyconversion devices.

Author contributions: B.S. and T.J.M. designed research; B.S., A.N., O.F.W., D.C.Y., N.Z.,and T.J.M. performed research; B.S., A.N., O.F.W., D.C.Y., N.Z., Y.K., A.M.M., M.J.T., andT.J.M. contributed new reagents/analytic tools; B.S., A.N., O.F.W., D.C.Y., N.F.P., Y.L.,M.J.T., and T.J.M. analyzed data; and B.S., D.C.Y., N.F.P., Y.L., Y.K., A.M.M., M.J.T., andT.J.M. wrote the paper.

Reviewers: N.A., University of Arizona; and V.S.B., Yale University.

The authors declare no conflict of interest.

Published under the PNAS license.1To whom correspondence may be addressed. Email: tjmeyer@unc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1907118116/-/DCSupplemental.

Published online July 31, 2019.

16198–16203 | PNAS | August 13, 2019 | vol. 116 | no. 33 www.pnas.org/cgi/doi/10.1073/pnas.1907118116

to nanoITO photoelectrodes through its PZn-appended 5′(4-car-boxypheneyl ethynyl) group (Fig. 1A), while (RuZn)2+ is similarlyattached via its 4″-terpyridyl-phosphonic acid group (Fig. 1B).Surface loading of these assemblies on nanoITO photoelectrodeswas determined to be ∼28–31 nmol/cm2 from electronic absorp-tion measurements in SI Appendix, Table S1.Fig. 2 describes the respective natural transition orbitals

(NTOs) for (PhZnRu)2+ and (RuZn)2+ determined from linear-response, time-dependent density functional theory (TDDFT)calculations (24–27) as a function of excitation wavelength. TheseNTOs were calculated by taking linear combinations of molecularorbitals according to their contributions to individual excited statesproduced in these supramolecular chromophores upon red- andblue-light excitation. For these excited states that feature substantialCT character, orbital isosurfaces [NTO(e−) and NTO(h+)] displaythe location and spatial extent of electron and hole delocalization.Note the differences in the natures of these NTOs determined for(PhZnRu)2+ as a function of excitation wavelength: Red excitation(∼1.9 eV) gives rise to an excited state described by PZn-to-Ru CT,while blue excitation (∼2.7 eV) generates an excited state thatfeatures Ru-to-PZn CT character (Fig. 2A). Red-light photoexci-tation of (PhZnRu)2+ creates a long-lived triplet excited state[3(PhZnRu)r*

2+] with a lifetime (τ) of ∼6.7 μs (SI Appendix, Fig.S7). The red-light excited state bears resemblance to semiconductorp–n junctions, with electron density delocalized on the ethynyl–terpyridyl ligand of Ru2+ and hole density on the phenyl–ethynyl–porphyrin fragment of PhZn (Fig. 2A). The NTO(e−) and NTO(h+)of (PhZnRu)r*

2+ resemble the lowest unoccupied molecular orbitaland highest occupied molecular orbital of the (PhZnRu)2+ groundstate determined from density functional theory (DFT) calculations(SI Appendix, Fig. S8).The red-light-triggered electron-transfer sequence for (PhZnRu)2+

anchored on nanoITO is illustrated in Fig. 3A. Hole injection intonanoITO under a mild applied bias from (PhZnRu)r*

2+ is facili-tated by the spatial distribution of its NTO(h+), producing thereduced chromophore [(PhZnRu)+]. Similarly, as illustrated inFig. 2B, red-light excitation of (RuZn)2+ results in an excited statedescribed by NTO(e−) and NTO(h+) that is delocalized, re-spectively, over the Ru2+ ethynyl–terpyridyl ligand and the por-phyrin framework. In contrast to the case for (PhZnRu)2+, undera mild applied bias, the (RuZn)2+ red-light excited state injects anelectron into nanoITO [see the spatial distribution of its NTO(e−),Fig. 2B], forming the oxidized chromophore (RuZn)3+, as illus-trated in Fig. 3C.As shown in Fig. 2A, the calculated NTO(e−) for the blue-light

(∼2.7 eV) excited state, (PhZnRu)b*2+, features an isosurface on

the porphyrin and the bridging terpyridyl ligand. The NTO(h+)isosurface of (PhZnRu)b*

2+ is dominated by the porphyrin unit ofPhZn. The analysis reveals that blue-light excitation of (PhZnRu)2+

does not give rise to an excited state with substantial CT character.

These results are in contrast to the NTOs of the blue-light excitedstate of (RuZn)2+ (Fig. 2B). For (RuZn)b*

2+, the electron density islargely delocalized on the Ru2+ ethynyl–terpyridyl ligand, whereasthe hole density is confined to the Zn ethynyl–porphyrin fragment,as shown by NTO(e−) and NTO(h+) in Fig. 2B. Under mild appliedbias on nanoITO, electron injection occurs from the blue-lightexcited states of (PhZnRu)b*

2+ and (RuZn)b*2+. The NTO(e−)

electron density spatial distributions for both these supermoleculesare poised for electron injection at the nanoITO surface, resulting inproduction of the one-electron oxidized chromophores, (PhZnRu)3+

and (RuZn)3+. The electron-transfer sequences are illustrated inFig. 3B for nanoITOj-(PhZnRu)b*2+ and in Fig. 3D for nanoITOj-(RuZn)b*

2+.Simulated UV-visible (UV-Vis) absorption spectra from linear-

response TDDFT calculations for (PhZnRu)2+ and (RuZn)2+ aregiven in SI Appendix, Figs. S9 and S10. As shown in both spectra,the low-energy peaks with intramolecular CT characters are red-shifted compared with the experimental spectra. This can be at-tributed to the fact that solvent was not included in the TDDFTcalculations, which precludes observation of the expected sol-vatochromic shifts. However, the qualitative trends are quitesimilar between the simulated and experimental spectra. The

Fig. 1. Structure and electronic absorption spectrum (extinction coefficients,e/M−1·cm−1) of (PhZnRu)2+ (A) and (RuZn)2+ (B) as a function of wavenumber(ν/cm−1). The red and blue shadings represent the red- and blue-light excitationregions.

Fig. 2. Excited-state natural transition orbitals (NTOs) of (PhZnRu)2+ (A) and(RuZn)2+ (B) from the time-dependent density functional theory (TDDFT).

Shan et al. PNAS | August 13, 2019 | vol. 116 | no. 33 | 16199

CHEM

ISTR

Y

assignments for the peaks and NTOs are relatively straightfor-ward, with three dominant broadened peaks appearing in bothsimulated spectra. For (RuZn)2+, there are two excited states withcomparable oscillator strengths (ES55 and ES66 in SI Appendix,Fig. S10). In this case, the lower-energy excited state, ES55, isassigned to the “blue” ∼2.7-eV absorption manifold because, asshown by the blue-shaded region in Fig. 1B, the excitation energyis more closely aligned with the shoulder, rather than the maxi-mum, of the band. For completeness, the NTOs for other exci-tations [ES19 for (PhZnRu)2+ and ES29 and ES66 for (RuZn)2+]are also displayed in SI Appendix, Figs. S9 and S10.

Photocurrent Variation with the Excitation Energy. TDDFT com-putations for the chromophore excited-state NTOs provide aframework for controlling the direction of electron flow at(PhZnRu)2+-functionalized photoelectrodes. To explore thepoint in experimental details, the (PhZnRu)2+-functionalizednanoITO electrode was irradiated by red light [λmax: 650 nm,20 nm full-width at half-maximum (FWHM), ∼10 mW/cm2] orblue light (λmax: 455 nm, 20 nm FWHM, ∼7.0 mW/cm2) with asolid-state light source (Lumencor Inc.; photon flux: 38 nmol/cm2/s).By alternating between red- and blue-light irradiation, thephotoelectrode nanoITOj-(PhZnRu)2+ under a bias at 0.20V vs. normal hydrogen electrode (NHE) produces persistentphotocurrents that switch between cathodic and anodic in theredox buffer containing sacrificial electron donor (ascorbic acid,0.77 mM) (28) and acceptor (4,4′-dipyridyl disulfide, 0.77 mM)(29), as shown in Fig. 4A. These photocurrent responses werereproduced when nanoITOj-(PhZnRu)2+ was irradiated by onlyred (Fig. 4B) or blue (Fig. 4C) light. With red light, the excitedstate (PhZnRu)r*

2+ injected the hole into nanoITO to givecathodic photocurrents. With blue light, nanoITOj-(PhZnRu)2+produced anodic photocurrents through electron injection by(PhZnRu)b*

2+. As illustrated in Fig. 3 A and B, injection leads toformation of (PhZnRu)+ by red-light excitation or (PhZnRu)3+by blue-light excitation. With the added redox buffer, persistentcathodic or anodic photocurrents result from the reduction of4,4′-dipyridyl disulfide by (PhZnRu)+ or the oxidation of ascorbic

acid by (PhZnRu)3+, respectively, in the redox buffer. The energydiagrams in SI Appendix, Fig. S11 A and B give the directions ofthe photogenerated electron-transfer processes for nanoITOj-(PhZnRu)2+ upon red and blue excitation.The photocurrent switching behavior was absent for nano-

ITOj-(RuZn)2+ under the same conditions, as shown in Fig. 4 D–F.Either red- or blue-light irradiation resulted in anodic photocur-rents and photo-oxidation of ascorbic acid in the redox buffer.In contrast to the data acquired for nanoITOj-(PhZnRu)2+, elec-tron injection at the photoelectrode was driven by both the red-light ([RuZn)r*

2+] and blue-light [(RuZn)b*2+] excitation of this

chromophore (Fig. 3 C and D). The electron-transfer steps areillustrated in the energy diagrams in SI Appendix, Fig. S11 C andD.The magnitudes of the photocurrents from nanoITOj-(RuZn)2+under blue-light irradiation were ∼1.4 times higher than the pho-tocurrents obtained by red-light irradiation. The similar light-absorption efficiencies for red and blue excitation (SI Appendix,Table S1) and the common photon flux (38 nmol/cm2/s) suggest nosignificant differences in the number of absorbed red or bluephotons; hence, variations in the observed photocurrents may re-sult from the differences in charge injection efficiencies, interfacialcharge recombination kinetics, or other factors.

Excited-State Electron Transfer. Pump-probe TA data for (RuZn)2+

and (PhZnRu)2+ adsorbed on nanoITO electrodes show ultrafastdynamics that are absent for these chromophores in solution,such as what would be expected for light-driven charge injection.The ultrafast TA spectra for the samples without the sacrificialelectron donors/acceptors are shown in SI Appendix, Fig. S12.The time-resolved traces in the Insets were fitted to the biphasicKohlrausch–Williams–Watts (KWW) function (Materials andMethods) (11). The fitting parameters are listed in SI Appendix,Table S2. From the TA spectra in SI Appendix, Fig. S12A, decayof the red-light excited state [(PhZnRu)r*

2+] on a 1.28-ps time-scale is consistent with hole injection into nanoITO followed byrapid charge recombination. Hole injection occurs due to the highhole density on the phenyl-ethynyl-porphyrin fragment of PhZn at(PhZnRu)r*

2+. Under an applied bias at 0.45 V vs. NHE, afraction of injected holes are driven to the back contact, leavingthe rest at the interface recombining with the electrons at(PhZnRu)+. The nanosecond TA in SI Appendix, Fig. S13 illus-trates the appearance and decay of the injection product, the

Fig. 3. Photoinduced charge-transfer mechanism for red-light excitation ofnanoITOj-(PhZnRu)2+ (A), blue-light excitation of nanoITOj-(PhZnRu)2+ (B),red-light excitation of nanoITOj-(RuZn)2+ (C), and blue-light excitation ofnanoITOj-(RuZn)2+ (D).

Fig. 4. (A–C) Photocurrent densities for nanoITOj-(PhZnRu)2+ under alter-nating red-light (shaded in red) or blue-light (shaded in blue) (A), underchopped red-light (B), or under chopped blue-light (C) irradiation. (D–F)Photocurrent densities acquired for nanoITOj-(RuZn)2+ under alternatingred-light (shaded in red) and blue-light (shaded in blue) (D), under choppedred-light (E), or under chopped blue-light (F) irradiation. The photo-electrodes were immersed in a redox buffer containing ascorbic acid(0.77 mM) and 4,4′-dipyridyl disulfide (0.77 mM) in a pH 4.5 proton buffer(0.1 M acetate, argon-degassed) under an applied bias of 0.20 V vs. NHE. Thephoton flux was 38 nmol/cm2/s.

16200 | www.pnas.org/cgi/doi/10.1073/pnas.1907118116 Shan et al.

reduced chromophore (PhZnRu)+ (see the spectroelectrochemicaldata in SI Appendix, Fig. S14A). The decay of (PhZnRu)+ occurson the microsecond timescale due to the interfacial charge re-combination with the holes at nanoITO surface traps (11). Underblue-light excitation, ultrafast injection dynamics likely occurwithin the time resolution (250 fs) of our instrument, as evidencedby (i) the comparatively weak TA spectra relative to those afterred-light excitation; and (ii) the spectra observed at the earliesttime delays already resemble that of the oxidized state (SI Ap-pendix, Fig. S14B). The nanosecond TA in SI Appendix, Fig.S15 shows the absorptive features of (PhZnRu)3+ (see the spec-troelectrochemical data in SI Appendix, Fig. S14B), which decays in∼5 μs due to charge recombination with nanoITO(e−). The positiveapplied bias helps to drive the injected electrons to the back contact,slowing down back electron transfer to (PhZnRu)3+. The electron-transfer steps are illustrated schematically in Fig. 3 A and B.Red- or blue-light excitations of nanoITOj-(RuZn)2+ led to

formation of the excited states [(RuZn)r*2+ or (RuZn)b*

2+] withTA features shown in SI Appendix, Fig. S12 C and D. BiphasicKWW fitting of the time-resolved TA data in the Insets givestimescales of ∼1.70–1.79 ps for electron injection into nanoITO.The nanosecond TA spectra in SI Appendix, Figs. S16 and S17 forred- and blue-light excitation of nanoITOj-(RuZn)2+ showed simi-lar absorptive changes, consistent with the appearance of the oxi-dized chromophore [(RuZn)3+] with the spectroelectrochemicaldata in SI Appendix, Fig. S14D. The decay of (RuZn)3+ throughcharge recombination with nanoITO(e−) in both cases was completewithin ∼10 μs. For nanoITOj-(RuZn)2+ under red or blue excita-tion, interfacial charge recombination between nanoITO(e−) and(RuZn)3+ was decelerated under the positive applied bias. Theelectron-transfer sequences are illustrated in Fig. 3 C and D.

Fig. 5 illustrates how surface coverages (Γ) of the oxidized orreduced chromophores varied as the applied bias (ΔE) was mod-ulated under red- or blue-light irradiation. The surface coveragechanges (ΔΓ) were evaluated by spectroelectrochemical measure-ments (30) for the chromophores on nanoITO (see details inMaterials and Methods and SI Appendix, Fig. S18). As shown bythe blue trace in Fig. 5A, blue-light irradiation of nanoITOj-(PhZnRu)2+ led to rapid growth in ΔΓ[(PhZnRu)3+] as ΔE wasincreased. The growth in ΔΓ[(PhZnRu)3+] slowed as the bias wasadjusted positively past the maximum point of ΔΓ[(PhZnRu)3+]/ΔE at ∼0.17 V vs. NHE. Compared with electrochemical oxidationof (PhZnRu)2+ in the dark (gray line in Fig. 5A), the blue-light-induced photooxidation of (PhZnRu)2+ occurred over a muchlower potential range. Since the applied bias dictates the Fermi levelof the nanoITO electrode (31), a more positive bias lowers thethermodynamic barrier for electron injection by the excited states.As discussed above, the red-light excited state, (PhZnRu)r*

2+,injects a hole into nanoITO to give the reduced chromophore,(PhZnRu)+, with the yield of formation controlled by the injectionstep. Fig. 5B illustrates the decrease in Γ[(PhZnRu)+] as the bias isshifted positively. Increasing the flat band potential of nanoITOpositively led to less favored hole injection by (PhZnRu)r*

2+ andrapidly decreased ΔΓ[(PhZnRu)+]/ΔE in Fig. 5B. For the photo-electrode nanoITOj-(RuZn)2+ under blue- or red-light irradiation,the injected electrons at nanoITO(e−) were transported to theexternal circuit by the positive applied bias, leaving the holes at(RuZn)3+. In Fig. 5 C and D, ΔΓ[(RuZn)3+] increased rapidlywith the increase of the bias; note the respective ΔΓ/ΔEmaxima at0.14 and 0.17 V vs. NHE for the blue- and red-light-excitednanoITOj-(RuZn)2+, respectively. Past those biases, ΔΓ/ΔE de-creased exponentially. As shown by the gray lines in Fig. 5 C andD, the unirradiated photoelectrode formed (RuZn)3+, but inmuch higher positive potential range. The Fermi level of the de-generately doped nanoITO electrode was determined by the ap-plied bias, which allows for facile control of the injection dynamicsfor nanoITOj-(RuZn)2+ and nanoITOj-(RuZn)2+ under differentexcitation energies. It can be expected that the dependence ofinjection dynamics on excitation energies is less pronounced whenthe nanoITO electrode is replaced by rectifying oxides—e.g.,p-type NiO or n-type TiO2—as illustrated by the proposed energy(redox) level diagrams in SI Appendix, Fig. S19.

Conclusions. The results described here demonstrate remarkablechanges in CT behaviors for the bichromophoric dyad, (PhZnRu)2+,on nanocrystalline ITO electrode upon excitation at differentphoton energies. Under red-light excitation, hole transfer from thelower-energy (PhZnRu)r*

2+ state to the electrode occurred, gen-erating cathodic photocurrents in a redox buffer containing asacrificial electron acceptor. In contrast, with blue-light excitation,(PhZnRu)b*

2+ injected an electron into the conduction band ofnanoITO to give anodic photocurrents, which were sustained withthe presence of a diffusional electron donor in solution. Thevariation in behaviors, which is remarkable from the photophysicalpoint of view, is microscopically derived from the different internalCT characteristics of the excited states. The excitation-energy-dependent processes that drive either hole or electron injectionat nanoITO are important from a device perspective. They es-tablish a basis for wavelength-dependent, single-molecule photo-electrodes whose outputs depend on the excitation photon energy.As an approach for applications in photodiodes, the (PhZnRu)2+-functionalized nanoITO electrode contrasts benchmarks in thisfield that rely upon significant variations in molecular or materialproperties for photocurrent control in optoelectronic junctions.

Materials and MethodsSynthesis of the Chromophores. (PhZnRu)2+ was synthesized by a reportedprocedure (32) with the synthetic scheme in SI Appendix, Fig. S20. The 1HNMR, 13C NMR, and electrospray ionization mass spectrometry (ESI-MS) aregiven in SI Appendix, Figs. S1–S3. The synthetic scheme for (RuZn)2+ is shownin SI Appendix, Fig. S21. The precursors, ZnE′ (21), TerpyBr (22), ZnTerpy (20),ZnRuCl3 (33), and terpyPO3H2 (33), were prepared based on the reported

Fig. 5. (A) Surface coverage changes (ΔΓ) of (PhZnRu)3+ on nanoITO perunit change of the bias (ΔE) under blue-light irradiation (blue circles) orunder dark conditions (gray line). (B) ΔΓ/ΔE as a function of the applied bias(E) for (PhZnRu)+ on nanoITO under red-light irradiation (red circles). (C)ΔΓ/ΔE as a function of E for (RuZn)3+ on nanoITO under blue-light irradiation(blue circles) or under dark conditions (gray line). (D) ΔΓ/ΔE as a function of Efor (RuZn)3+ on nanoITO under red-light irradiation (red circles) or under darkconditions (gray line). The photoelectrodes were immersed in an argon-degassed pH 4.5 acetate (0.1 M) buffer. The photon flux was 38 nmol/cm2/s.

Shan et al. PNAS | August 13, 2019 | vol. 116 | no. 33 | 16201

CHEM

ISTR

Y

procedures (20–22, 33). ZnRuCl3 (138.9 mg, 0.1 mmol), terpyPO3H2 (31.3 mg,0.1 mmol), triethylamine (4 mL), and ethanol (5 mL) were added inN-methylpyrrolidone (30 mL). NH4PF6 (32.6 mg, 0.2 mmol) was added afterthe solution mixture was stirred at 120 °C for 36 h and cooled to roomtemperature. Stirring was continued for 4 h. The solvent was removed underreduced pressure. The impurities in the crude product were then removed byusing a short silica column with acetonitrile as the eluent. Acetone was usedto dissolve and filter out the product in the silica gel. The final product (47 mg;25% yield based on ZnRuCl3) was obtained following removal of ace-tone under reduced pressure. The 1H NMR, 13C NMR, and ESI-MS are given inSI Appendix, Figs. S4–S6.

Preparation of the Photoelectrodes. The mesoporous nanoITO electrode wasprepared according to a report (31). The thickness of the nanoITO film wasmeasured as 3.3 μm by a Bruker Dektak XT profilometer. The chromophores(PhZnRu)2+ and (RuZn)2+ were loaded by immersing the nanoITO films inacetonitrile solutions of (PhZnRu)2+ or (RuZn)2+ (0.20–0.23 mM) for 1 h. Thedyed films were then taken out of the loading solutions and dried in air foranother 1 h.

Computational Methods. All DFT and linear-response TDDFT calculations wereperformed by using the Gaussian software package (Version 16) (34). Weused the PBE0 hybrid exchange correlation functional (35), which comprisesthe generalized gradient approximation exchange-correlation functional ofPerdew et al. (36, 37) together with 25% of exact Hartree exchange energy.The 6-31G** basis set (38–42) was used to describe all light element atoms,and the LANL2DZ effective core potentials basis set (38–42) was used for theRu and Zn ions. All calculations were carried out with a +2 net charge for themolecules. Although the experimental UV-Vis absorption spectra were obtainedwith the complexes in acetonitrile solvent, all DFT and linear-response TDDFTcalculations were performed in vacuum so as to be more representative of thenanoITO-(RuZn)2+/nanoITO-(PhZnRu)2+ heterostructures of interest, in whichthe molecules were not solvated. Due to the prohibitive computational ex-pense, the calculations were performed on isolated molecules without includingITO surfaces. First, for both complexes, the ground-state equilibrium geometriesand electronic structures of both the (PhZnRu)2+ and (RuZn)2+ complexes werecalculated in their singlet ground states. Next, the 70 lowest singlet excitedstates of each molecular complex were computed by using linear-responseTDDFT. The oscillator strengths of these computed excited states, absorbingat a range of energies from ∼1–3 eV, were then used to acquire absorptionspectra for comparison with the experimental results. For both complexes,there were few optically active excited states, allowing for simple character-ization of the excited-state characteristics through the use of NTOs (27). TheNTOs, which comprise linear combinations of the molecular orbitals contrib-uting to a given excitation, are plotted as isosurfaces in Fig. 2. They representexcitations as pairs of hole and electron excited orbitals and characterize theredistribution of electron density upon photoexcitation. By plotting the NTOs,we obtained an intramolecular-level qualitative depiction of the excited-statecharacteristics that give rise to the differences in photo-induced electronicbehavior observed for the (RuZn)2+ and (PhZnRu)2+ complexes.

Ultrafast TA Spectroscopy. The ultrafast TA experiments were conducted withaCoherent Libra laser (45-fs pulses, 4-mJoutput, 1-kHz repetition rate, and800-nmfundamental). Approximately 1.5 mJ of the fundamental was focused in a 4-mcell pressurized with argon gas to produce a continuum from 435 to 740 nm.The continuum pulse was filtered by using a 4F setup built with only reflectiveoptics and was based on a 200-cm focal-length mirror and a 1,200-g/mmgrating. Desired portions of the continuumwere selected by using amotorizedslit. The 455-nm pumpwas 22 nm at FWHM, while the 660-nm pump was 6 nmwide; the bandwidth of the 455-nm pumpwas increased tomore closely matchthe power of the 660-nmpulse when incident on the sample. Both pumpswere∼250 fs in duration. The pulse energies for the 455- and 660-nm pumps were12 and 22 nJ per pulse, respectively. The continuum probe pulses were gen-erated by focusing a small fraction of the 800-nm fundamental onto a sap-phire window, which was then relayed to the sample by using all reflectiveoptics. The 455-nm (660-nm) pump was focused to a roughly 300-μm (200-μm)spot. The spot size of the probe was adjusted to match that of the respectivepump. The transmitted probe was detected by using a complementary metal–oxide semiconductor array detector synchronized to the repetition rate of thelaser. For a single point in the delay line, 400 differences were collected; ul-trafast TA signals were averaged for 30 scans of the delay line.

The time-resolved data were fitted according to Eqs. 1 and 2 based on thebiphasic KWW stretched exponential model (11, 43).

ΔAt =ΔA0 × exp

"−�

tτ0

�β#+ΔA’

0 ×exp

"−�

tτ0’

�β’#+C, [1]

1�

τ=��

τ0β

�×Γ

�1β

��−1, 1�

τ’=��

τ0’β’

�×Γ

�1β’

��−1. [2]

ΔAt and ΔA0 (or ΔA0′) are the absorption changes at times t and 0, re-spectively, after the laser pulse. β and β′ are the stretching parameters.τ0 and τ0’ are the characteristic KWW relaxation time constants. C is a con-stant generated from each fit. Γ(x) is the Gamma function. τ and τ′ are theobserved lifetimes, with the smaller taken as the injection lifetime.

ACKNOWLEDGMENTS. The research on molecular assemblies, design andfabrication of photocathodes, and photoelectrochemistry were supportedby the US Department of Energy (DOE), Office of Science, Office of BasicEnergy Sciences Award DE-SC0015739 (to B.S. and T.J.M.). The data analysiswas supported by National Science Foundation Grant CHE-1709497 (to A.N.and M.J.T.). The ultrafast TA measurements were supported by NationalScience Foundation Grant CHE-1763207 (to O.F.W. and A.M.M.). The TDDFTand DFT calculations were supported by National Science Foundation GrantDGE-1144081 (to D.C.Y. and Y.K.). The experiments with nanosecond TA andsolid-state irradiator were performed with the instruments within theAlliance for Molecular Photoelectrode Design for Solar Fuels Energy FrontierResearch Center Instrumentation Facility established by the Alliance for Mo-lecular PhotoElectrode Design for Solar Fuels, an Energy Frontier ResearchCenter funded by the DOE, Office of Science, Office of Basic Energy SciencesAward DE-SC0001011.

1. C. Joachim, J. K. Gimzewski, A. Aviram, Electronics using hybrid-molecular and mono-

molecular devices. Nature 408, 541–548 (2000).2. D. Xiang, X. Wang, C. Jia, T. Lee, X. Guo, Molecular-scale electronics: From concept to

function. Chem. Rev. 116, 4318–4440 (2016).3. W. B. Davis, W. A. Svec, M. A. Ratner, M. R. Wasielewski, Molecular-wire behaviour in

p-phenylenevinylene oligomers. Nature 396, 60–63 (1998).4. A. Nitzan, M. A. Ratner, Electron transport in molecular wire junctions. Science 300,

1384–1389 (2003).5. L. T. Cai et al., Nanowire-based molecular monolayer junctions: Synthesis, assembly,

and electrical characterization. J. Phys. Chem. B 108, 2827–2832 (2004).6. B. Capozzi et al., Single-molecule diodes with high rectification ratios through envi-

ronmental control. Nat. Nanotechnol. 10, 522–527 (2015).7. C. Jia et al., Covalently bonded single-molecule junctions with stable and reversible

photoswitched conductivity. Science 352, 1443–1445 (2016).8. S. Nishimura et al., Standing wave enhancement of red absorbance and photocurrent

in dye-sensitized titanium dioxide photoelectrodes coupled to photonic crystals. J.

Am. Chem. Soc. 125, 6306–6310 (2003).9. K. J. Young et al., Light-driven water oxidation for solar fuels. Coord. Chem. Rev. 256,

2503–2520 (2012).10. J. L. White et al., Light-driven heterogeneous reduction of carbon dioxide: Photo-

catalysts and photoelectrodes. Chem. Rev. 115, 12888–12935 (2015).11. B. H. Farnum, K.-R. Wee, T. J. Meyer, Self-assembled molecular p/n junctions for ap-

plications in dye-sensitized solar energy conversion. Nat. Chem. 8, 845–852 (2016).12. J. T. Kirner, R. G. Finke, Water-oxidation photoanodes using organic light-harvesting

materials: A review. J. Mater. Chem. A 5, 19560–19592 (2017).

13. P. Xu, T. Huang, J. Huang, Y. Yan, T. E. Mallouk, Dye-sensitized photoelectrochemical water

oxidation through a buried junction. Proc. Natl. Acad. Sci. U.S.A. 115, 6946–6951 (2018).14. B. Shan et al., Controlling vertical and lateral electron migration using a bifunctional

chromophore assembly in dye-sensitized photoelectrosynthesis cells. J. Am. Chem.

Soc. 140, 6493–6500 (2018).15. S. Yasutomi, T. Morita, Y. Imanishi, S. Kimura, A molecular photodiode system that

can switch photocurrent direction. Science 304, 1944–1947 (2004).16. S. Yasutomi, T. Morita, S. Kimura, pH-controlled switching of photocurrent direction by self-

assembled monolayer of helical peptides. J. Am. Chem. Soc. 127, 14564–14565 (2005).17. M. Hebda, G. Stochel, K. Szaciłowski, W. Macyk, Optoelectronic switches based on

wide band gap semiconductors. J. Phys. Chem. B 110, 15275–15283 (2006).18. B. H. Farnum, Z. A. Morseth, M. K. Brennaman, J. M. Papanikolas, T. J. Meyer, Driving

force dependent, photoinduced electron transfer at degenerately doped, optically

transparent semiconductor nanoparticle interfaces. J. Am. Chem. Soc. 136, 15869–

15872 (2014).19. T. V. Duncan, I. V. Rubtsov, H. T. Uyeda, M. J. Therien, Highly conjugated (polypyridyl)

metal-(porphinato)zinc(II) compounds: Long-lived, high oscillator strength, excited-

state absorbers having exceptional spectral coverage of the near-infrared. J. Am.

Chem. Soc. 126, 9474–9475 (2004).20. A. Nayak et al., Large hyperpolarizabilities at telecommunication-relevant wave-

lengths in donor–acceptor–donor nonlinear optical chromophores. ACS Cent. Sci. 2,

954–966 (2016).21. T. V. Duncan et al., Molecular symmetry and solution-phase structure interrogated by

hyper-Rayleigh depolarization measurements: Elaborating highly hyperpolarizable

D2-symmetric chromophores. Angew. Chem. Int. Ed. Engl. 47, 2978–2981 (2008).

16202 | www.pnas.org/cgi/doi/10.1073/pnas.1907118116 Shan et al.

22. T. V. Duncan, T. Ishizuka, M. J. Therien, Molecular engineering of intensely near-

infrared absorbing excited states in highly conjugated oligo(porphinato)zinc-(poly-

pyridyl)metal(II) supermolecules. J. Am. Chem. Soc. 129, 9691–9703 (2007).23. H. T. Uyeda et al., Unusual frequency dispersion effects of the nonlinear optical re-

sponse in highly conjugated (polypyridyl)metal-(porphinato)zinc(II) chromophores. J.

Am. Chem. Soc. 124, 13806–13813 (2002).24. X. Hu et al., Predicting the frequency dispersion of electronic hyperpolarizabilities on

the basis of absorption data and Thomas−Kuhn sum rules. J. Phys. Chem. C 114, 2349–

2359 (2010).25. K. Burke, J. Werschnik, E. K. U. Gross, Time-dependent density functional theory: Past,

present, and future. J. Chem. Phys. 123, 62206 (2005).26. M. E. Casida, “Time-dependent density functional response theory for molecules” in

Recent Advances in Density Functional Methods, D. P. Chong, Ed. (World Scientific,

Singapore, 1995), vol. 11, pp. 155–192.27. R. L. Martin, Natural transition orbitals. J. Chem. Phys. 118, 4775–4777 (2003).28. B. Shan, T. Baine, X. A. N. Ma, X. Zhao, R. H. Schmehl, Mechanistic details for cobalt

catalyzed photochemical hydrogen production in aqueous solution: Efficiencies of

the photochemical and non-photochemical steps. Inorg. Chem. 52, 4853–4859 (2013).29. T.-T. Li, B. Shan, T. J. Meyer, Stable molecular photocathode for solar-driven CO2

reduction in aqueous solutions. ACS Energy Lett. 4, 629–636 (2019).30. B. Shan et al., Charge transfer from upconverting nanocrystals to semiconducting

electrodes: Optimizing thermodynamic outputs by electronic energy transfer. J. Am.

Chem. Soc. 141, 463–471 (2019).31. B. H. Farnum, Z. A. Morseth, M. K. Brennaman, J. M. Papanikolas, T. J. Meyer, Ap-

plication of degenerately doped metal oxides in the study of photoinduced in-

terfacial electron transfer. J. Phys. Chem. B 119, 7698–7711 (2015).

32. B. Shan et al., Direct photoactivation of a nickel-based, water-reduction photocath-ode by a highly conjugated supramolecular chromophore. Energy Environ. Sci. 11,447–455 (2018).

33. P. Bonhôte et al., Long-lived photoinduced charge separation and redox-type pho-tochromism on mesoporous oxide films sensitized by molecular dyads. J. Am. Chem.Soc. 121, 1324–1336 (1999).

34. M. J. Frisch et al., Gaussian (Version 16 Rev. B.01, Gaussian, Wallingford, CT, 2016).35. C. Adamo, V. Barone, Toward reliable density functional methods without adjustable

parameters: The pbe0 model. J. Chem. Phys. 110, 6158–6170 (1999).36. J. P. Perdew, K. Burke, M. Ernzerhof, Generalized gradient approximation made

simple. Phys. Rev. Lett. 77, 3865–3868 (1996).37. M. Ernzerhof, G. E. Scuseria, Assessment of the Perdew–Burke–Ernzerhof exchange-

correlation functional. J. Chem. Phys. 110, 5029–5036 (1999).38. T. Clark, J. Chandrasekhar, G. W. Spitznagel, P. V. R. Schleyer, Efficient diffuse func-

tion‐augmented basis sets for anion calculations. III. The 3‐21+ G basis set for first‐rowelements, Li–F. J. Comput. Chem. 4, 294–301 (1983).

39. M. M. Francl et al., Self‐consistent molecular orbital methods. XXIII. A polarization‐type basis set for second‐row elements. J. Chem. Phys. 77, 3654–3665 (1982).

40. P. M. Gill, B. G. Johnson, J. A. Pople, M. J. Frisch, The performance of the Becke–Lee–Yang–Parr (B–LYP) density functional theory with various basis sets. Chem. Phys. Lett.197, 499–505 (1992).

41. P. C. Hariharan, J. A. Pople, The influence of polarization functions on molecularorbital hydrogenation energies. Theor. Chim. Acta 28, 213–222 (1973).

42. R. Krishnan, J. S. Binkley, R. Seeger, J. A. Pople, Self‐consistent molecular orbital meth-ods. XX. A basis set for correlated wave functions. J. Chem. Phys. 72, 650–654 (1980).

43. B. Shan, B. H. Farnum, K. Wee, T. J. Meyer, Generation of long-lived redox equivalentsin self-assembled bilayer structures on metal oxide electrodes. J. Phys. Chem. C 121,5882–5890 (2017).

Shan et al. PNAS | August 13, 2019 | vol. 116 | no. 33 | 16203

CHEM

ISTR

Y