THE JOURNAL OF CHEMICAL PHYSICS 129, 164102 共2008兲

Excitation mechanism in the photoisomerization of a surface-bound azobenzene derivative: Role of the metallic substrate Sebastian Hagen, Peter Kate, Felix Leyssner, Dhananjay Nandi, Martin Wolf, and Petra Tegedera兲 Fachbereich Physik, Freie Universität Berlin, Arnimallee 14, D-14195 Berlin, Germany

共Received 23 July 2008; accepted 16 September 2008; published online 22 October 2008兲 Two-photon photoemission spectroscopy is employed to elucidate the electronic structure and the excitation mechanism in the photoinduced isomerization of the molecular switch tetra-tert-butyl-azobenzene 共TBA兲 adsorbed on Au共111兲. Our results demonstrate that the optical excitation and the mechanism of molecular switching at a metal surface is completely different compared to the corresponding process for the free molecule. In contrast to direct 共intramolecular兲 excitation operative in the isomerization in the liquid phase, the conformational change in the surface-bound TBA is driven by a substrate-mediated charge transfer process. We find that photoexcitation above a threshold h␯ ⬇ 2.2 eV leads to hole formation in the Au d-band followed by a hole transfer to the highest occupied molecular orbital of TBA. This transiently formed positive ion resonance subsequently results in a conformational change. The photon energy dependent photoisomerization cross section exhibit an unusual shape for a photochemical reaction of an adsorbate on a metal surface. It shows a thresholdlike behavior below h␯ ⬇ 2.2 eV and above h␯ ⬇ 4.4 eV. These thresholds correspond to the minimum energy required to create single or multiple hot holes in the Au d-bands, respectively. This study provides important new insights into the use of light to control the structure and function of molecular switches in direct contact with metal electrodes. © 2008 American Institute of Physics. 关DOI: 10.1063/1.2997343兴 I. INTRODUCTION

Using light to control the switching of functional properties of surface-bound species is an attractive strategy for the development of new technologies with possible applications in molecular electronics and functional surfaces.1–8 Azobenzene and its derivatives are promising systems for such a route since they possess the ability to undergo a reversible photoinduced conformational change between the nearly planar trans isomer and the three-dimensional cis form. In solution the isomerization of azobenzene is well understood. It involves a direct electronic excitation at ⬇365 nm 共3.40 eV兲 for the trans to cis isomerization and excitation around 420 nm 共2.95 eV兲 for the reverse process,9–11 allowing control of its structure and properties.12–17 However, the photoinduced excitation characteristics of molecular switches in direct contact with metal substrates are expected to be quite different since effective quenching of electronic excitations may occur.18 However, no detailed experimental study concerning this issue is available so far. On the other hand, different switching mechanisms, e.g., charge transfer processes between the substrate and adsorbate, may be accessible and may open new reaction pathways in contact with a metal substrate. For instance, reversible switching of various azobenzene derivatives adsorbed on noble metals have been achieved by excitation with a scanning tunneling microscope 共STM兲 tip. Thereby Author to whom correspondence should be addressed. Tel.: ⫹49-30-83856234. FAX: ⫹49-30-838-56059. Electronic mail: [email protected].

a兲

0021-9606/2008/129共16兲/164102/8/$23.00

varying excitation processes such as resonant19 and inelastic tunneling20 as well as stimulation by the applied electric field21 have been proposed. In order to elucidate the excitation mechanism for photostimulated isomerization detailed knowledge about the electronic structure, i.e., occupied and unoccupied electronic states 共or band structure兲, of the adsorbed molecules is essential. Two-photon photoemission 共2PPE兲 spectroscopy has been proven to be an ideal tool in investigating both unoccupied and occupied electronic states at surfaces. It has been used in various adsorbate-substrate systems to study adsorbate and image potential states.22–26 In this contribution, we demonstrate that the photoexcitation mechanism in the molecular switching of an azobenzene derivative adsorbed on a metal surface is fundamentally different from the process known in liquid phase. We apply 2PPE to examine the electronic structure of 3 , 3⬘ , 5 , 5⬘-tetra-tert-butyl-azobenzene 共TBA, see Fig. 1兲 adsorbed on Au共111兲. We analyze the photoinduced switching, which is associated with significant changes in the electronic structure in order to determine the trans/cis isomerization mechanism. The TBA is chosen because the four lateral tertbutyl groups increase the separation between the molecule and the substrate. This leads to a reduced electronic coupling between the active part of the molecules, i.e., the ␲ system, and the metal surface, and thus allow the photoinduced switching of surface-adsorbed molecules.27–29 Thereby reversible switching of TBA on Au共111兲 has been demonstrated by using UV light and thermal activation.27,29 On the other hand, light-induced switching could not be achieved for TBA on Ag共111兲 共Ref. 30兲 and pure azobenzene

129, 164102-1

© 2008 American Institute of Physics

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

164102-2

J. Chem. Phys. 129, 164102 共2008兲

Hagen et al.

FIG. 1. 共Color online兲 Isomerization reaction of the azobenzene TBA.

on Au共111兲 共Ref. 28兲 presumably due to the stronger interaction with the metal substrate and corresponding differences in the geometric and electronic structures. For applications in molecular electronics and light-induced control of molecular function, it is a key question how the switching properties change upon adsorption and interaction with a metal electrode. However, the underlying excitation mechanism for the photoinduced trans/cis isomerization of TBA on Au共111兲 has not been resolved. As we shall demonstrate, 2PPE spectroscopy and photoexcitation over a wide photon energy range enable us to identify the excitation mechanism in the light-stimulated isomerization. The direct intramolecular electronic excitation leading to photoisomerization of the free molecules turns out to be not operative for surface-bound species. Instead, an indirect excitation mechanism, i.e., a substrate-mediated process, drives the isomerization of TBA adsorbed on Au共111兲 surface. Thereby photoexcitation in the energy regime between h␯ ⬇ 2.2 and 4.8 eV leads to the creation of holes in the Au d-band, which rapidly relax to the top of the d-band. This is followed by a charge transfer to the highest occupied molecular orbital 共HOMO兲 of the TBA. The shape of the cross section for the photoinduced isomerization as a function of photon energy shows an unusual behavior with two thresholds, which correlate with the energy of the d-band edge 共excitation of a single hot hole兲 and the energy required for excitations of multiple holes, respectively. Our findings show that in the light-induced excitation mechanisms for molecular switching at a metallic substrate, the metal and not the molecule may act as the “chromophore,” i.e., the metal is the light absorbing material in contrast to the photoinduced isomerization of molecular switches in the liquid phase. This clearly demonstrates that adsorption on a metal surface causes significant modifications to the mechanism known in liquid phase. II. EXPERIMENTAL

The experiments were carried out in an ultrahigh vacuum chamber combined with a tunable femtosecond laser system.30 The Au共111兲 crystal was mounted on a liquid nitrogen cooled cryostat which in conjunction with resistive heating enables temperature control from 90 to 750 K. The crystal was cleaned by cycles of Ar+ sputtering and annealing. The TBA was dosed by means of a homebuilt effusion cell held at 380 K at a crystal temperature of 260 K. The TBA coverage was quantified by thermal desorption spectroscopy and work function measurements. All measurements presented below are performed at a submonolayer coverage of ⬇0.9 ML 共monolayer兲, which is prepared by heating the multilayer-covered surface to 420 K.27 From

high-resolution electron energy loss spectroscopy29 and STM,21,28 it is known that TBA adsorbs in this low-coverage regime in a planar 共trans兲 configuration. For the 2PPE measurements, femtosecond laser pulses are generated by a 300 kHz Ti:sapphire laser system which pumps an optical parametric amplifier. The visible output with photon energies from 1.7 to 2.7 eV, respectively, is frequency doubled in a beta-barium-borate 共BBO兲 crystal to generate ultraviolet 共UV兲 pulses 共3.4–5.4 eV photon energy兲. The laser pulses are incident on the surface with an angle of 45° with respect to the surface normal. While the pump pulse h␯1 excites an electron from below the Fermi level 共EF兲 to intermediate unoccupied states at energies E − EF = Ekin + ⌽ − h␯2, 共with ⌽ the work function兲, the probe pulse h␯2 photoionizes the sample by lifting the excited electron above the vacuum level 共Evac兲. Photoelectrons are detected in an electron time-of-flight spectrometer 共TOF兲 and analyzed with respect to their kinetic energy 共Ekin兲. The energy resolution of the TOF spectrometer depends on the electron energy. It is ⬇10 meV at Ekin ⬇ 1 eV. The 2PPE spectra presented below are displayed as 2PPE intensity versus the final state energy, Efinal − EF = Ekin + ⌽, with respect to the Fermi level 共EF = 0 eV兲. For photoexcitation of the TBA-covered surface femtosecond laser pulses with photon energies ranging from 1.7 to 4.8 eV were used. Since the photons caused characteristic changes in the 2PPE features due to the isomerization, every 2PPE spectrum was recorded for only 5 s. This corresponds to a light exposure of less than 1% of the exposure required to reach the photostationary state, i.e., the photoinduced changes during the acquisition time are negligible. All measurements were performed at a substrate temperature of 90 K. III. RESULTS A. Electronic structure of the TBA/Au„111… system

Understanding the photoinduced excitation mechanism of TBA adsorbed on Au共111兲 requires detailed insights into the electronic structure, which can be obtained by 2PPE spectroscopy. Figure 2共a兲 displays two-color 2PPE spectra of 0.9 ML TBA on Au共111兲 recorded at photon energies of 2.2 and 4.4 eV. Shown are the 1. scan spectrum corresponding to a photon dose 共number of photons, n p兲 of 1 ⫻ 1019 cm−2 and the spectrum after UV-light 共4.4 eV兲 exposure with a photon dose of ⬇3 ⫻ 1021 cm−2. In the 1. scan spectrum a pronounced feature located at 6.25 eV is observed 共labeled A兲. However, illumination of the TBA/Au共111兲 system causes significant changes in the 2PPE spectrum: 共i兲 the peak at 6.25 eV loses intensity while at lower energies 共⬇5.95 eV兲 a new feature emerges 共labeled B兲; 共ii兲 close to the vacuum level 共which corresponds to the low energy cutoff of the photoemission spectrum兲, a strong peak 共labeled C兲 appears; and 共iii兲 the work function shifts by ⬇50 meV. The spectral changes due to light exposure are assigned to the trans to cis isomerization of TBA, as it has been discussed in detail in Ref. 27. In order to identify whether the peaks originate from occupied initial, unoccupied intermediate, or final states in

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

164102-3

(a)

J. Chem. Phys. 129, 164102 共2008兲

Photoexcitation mechanism

0.9 ML TBA/Au(111) h
1 = 2.2 eV h
2 = 4.4 eV

2PPE Intensity [arb. units]

C

1. scan

np = 1x1019 cm-2

h
2 = 4.4 eV

np = 3x1021 cm-2

LUMO: cis trans LUMO+n A cis

B

x5

4.5

(b)

5.0

5.5 6.0 EFinal - EF [eV]

6.5

Ekin [eV]

1.5 1.0 0.5

LUMO (A) Slope = 1.0 LUMO+n (C) Slope = 0

0.0 3.6 4.0 4.4 Photon Energy [eV] FIG. 2. 共Color online兲 共a兲 Two-color 2PPE spectra of TBA adsorbed on Au共111兲 measured with 2.2 and 4.4 eV photons, the 1. scan spectrum 共photon dose of 1 ⫻ 1019 cm−2兲, and the spectrum after UV-light 共4.4 eV兲 exposure with a photon dose of 3 ⫻ 1021 cm−2. The spectra are displayed as 2PPE intensity vs final state energy with respect to the Fermi level 共EF兲, Efinal − EF = Ekin + ⌽ 共with ⌽ the work function兲. 共b兲 Photon energy dependence of the peaks labeled A and C in the 2PPE spectrum of TBA/Au共111兲. The symbols are experimental data, and the solid lines are fitting curves. The kinetic energy of peak A varies with 1⌬h␯, indicating that it originates from an unoccupied intermediate state. The kinetic energy of peak C remains constant, i.e., it arises from an unoccupied final state.

the 2PPE process, the dependence of the electron kinetic energy 共Ekin兲 on the photon energy was investigated. When an unoccupied intermediate state, such as the lowest unoccupied molecular orbital 共LUMO兲, is probed in one-color 2PPE the change in Ekin scales with that in photon energy 共h␯ and h␯⬘兲, i.e., Ekin = 1⌬h␯ 共⌬h␯ = h␯ − h␯⬘兲. On the other hand, for an occupied initial state, e.g., the HOMO, the kinetic energy of the electron ejected scales with twice the photon energy 共2⌬h␯兲, whereas for an unoccupied final state Ekin is independent of the photon energy 共Ekin = const兲. The latter can be viewed as a resonant scattering event in which the photoexcited electron resides transiently in the molecular resonance followed by detachment and detection. This analysis is gen-

erally not applicable for transitions between bulk bands due to their strong perpendicular dispersion but holds in the case of surface and adsorbate-derived states.31,32 Figure 2共b兲 shows exemplarily the photon energy dependence of peaks A and C. The kinetic energy of the peak labeled A varies with 1⌬h␯, clearly indicating that the peak is caused by photoemission from an unoccupied intermediate state. This state is pumped with the visible pulse h␯1 and probed with the UV pulse h␯2; therefore its energetic position is 1.85 eV above EF. The kinetic energy of the feature labeled B, which appears after illumination, also scales with 1⌬h␯ 共data not shown here兲, i.e., it arises from an unoccupied intermediate state. This state is located at 1.55 eV above EF. We assign both peaks A and B to the LUMO of the trans and cis isomers, respectively. The switching process obviously provokes a shift in the LUMO toward lower energies by 300 meV. Note that in scanning tunneling spectroscopy 共STS兲 also an energy difference of 300 meV between the LUMO positions of the two isomers was observed.21 The kinetic energy of peak C does not vary with photon energy 关see Fig. 2共b兲兴; thus it originates from an unoccupied final state which is located close to the vacuum level at 4.5 eV. Since this peak appears after light exposure it can be attributed to the LUMO+ n of the cis-TBA.27 The appearance and intensity of this state will be used to evaluate the effective cross section for the photoinduced trans/cis isomerization reaction 共see below兲. In order to analyze also the position of the HOMO levels, Fig. 3 displays a one-color 2PPE spectrum of 0.9 ML TBA adsorbed on Au共111兲 recorded at a photon energy of 4.21 eV. Besides photoemission from the occupied d-bands33,34 and the Shockley surface state 共SS兲,35 which are located at −2.0 eV 共d-band兲, −2.85 eV 共d-band兲, and −0.48 eV 共SS兲 below EF,27 the features labeled D–F and IS are observed. Although the background intensity is large and some of the features are overlapping, their appearance and energetic positions are highly reproducible. The inset in Fig. 3 shows the photon energy dependence of peaks D–F. The kinetic energies of peaks D and E vary with 2⌬h␯, indicating that the transitions involve occupied initial states. We assign these peaks to the TBA-induced HOMO and HOMO-1. Their energetic positions are 1.8 共HOMO兲 and 3.0 eV 共HOMO-1兲 below EF. The kinetic energy of the peak labeled F does not vary with photon energy, i.e., it originates from an unoccupied final state. This state is located at 4.8 eV above EF. The feature labeled IS can be assigned to the n = 1 image potential state, which lies 0.6 eV below Evac.27 For comparison, on the n-heptane 共1 ML兲 covered surface it is also located at 0.6 eV,36 while for the clean Au共111兲 surface a binding energy of 0.8 eV has been reported.37 B. Photon energy dependent isomerization of TBA/Au„111…

In the following we consider the excitation mechanism operative for TBA in the liquid phase and the light-induced isomerization cross section as a function of photon energy for the surface-adsorbed molecules. For the free azobenzene and its derivatives photoisomerization is induced by direct 共intramolecular兲 optical elec-

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

J. Chem. Phys. 129, 164102 共2008兲

Hagen et al.

HOMO (D) Slope = 2.0 HOMO-1 (E) Slope = 2.0 LUMO+m (F) Slope = 0

2PPE Intensity [arb. units]

E (HOMO-1)

Ekin [eV]

0.9 ML TBA/Au(111) hn = 4.21 eV 2.0

x10

1.5 1.0 0.5

d-band F (LUMO+m)

3.6 4.0 4.4 Photon Energy [eV]

d-band

SS

6 7 EFinal-EF [eV]

8

tronic excitation.9–11 Figure 4 displays UV-visible absorption spectra of TBA in solution. The strong absorption band around 3.81 eV belongs to the ␲ → ␲ⴱ 共S2兲 transition, whereas the band at ⬇2.75 eV corresponds to the symmetry forbidden n → ␲ⴱ 共S1兲 excitation. The intensity loss of the absorption band at 3.81 eV due to illumination at a photon energy of 3.96 eV is assigned to the trans to cis isomerization.

Absorbance [arb. units]

4.77 eV

3.81 eV

0.4

TBA in solution trans cis isomerization

N

N

3.96 eV

N

N

before illumination 0.2

-21

10

-22

10

-23

10

2.5

4.0 3.0 3.5 Photon energy [eV]

4.5

5.0

FIG. 5. 共Color online兲 Effective cross section for the photoinduced trans/cis isomerization of TBA adsorbed on Au共111兲 as a function of photon energy 共see text兲.

IS

FIG. 3. One-color 2PPE spectrum of 0.9 ML TBA adsorbed on Au共111兲 taken at a photon energy of 4.21 eV. Inset: Photon energy dependence of peaks D–F. The symbols are experimental data and the solid lines are fitting curves. The kinetic energies of peaks D and E vary with 2⌬h␯, indicating that the peaks originate from occupied initial states. They can be assigned to the HOMO and HOMO-1. The peak labeled F shows no energy dependency; therefore it is an unoccupied final state

0.6

-20

10

2.0

D (HOMO)

5

h
Effective cross section [cm2 ]

164102-4

Figure 5 shows the dependence of the effective cross section 共␴eff兲 for the light-induced trans/cis isomerization of TBA adsorbed on Au共111兲 on the photon energy used for excitation. ␴eff is determined by evaluating the peak intensity of the unoccupied final state of the cis-isomer 关LUMO+ n, see Fig. 2共a兲兴 as a function of photon dose for various photon energies using an exponential saturation function 共for details see Ref. 27 and supporting information38兲. Thereby the initial slope of the exponential function as well as the saturation level, i.e., the ratio between the photoemission intensity of the LUMO+ n state in the photostationary state and the intensity of the 1. scan spectrum, is obtained. This ratio is identified with the amount of switched cis-TBA in the photostationary state and we have verified that it stays constant over the whole photon energy regime between ⬇2.0 and 4.8 eV.38 As seen in Fig. 5, ␴eff shows a stepwise change, a plateaulike region over the large photon energy range between 2.2 and 4.4 eV, where it stays constant. Above 4.4 eV a strong exponential increase and below 2.2 eV a pronounced decrease are observed 共note the logarithmic scale兲. Below a photon energy of 2 eV, isomerization of TBA has not been observed. The shape of ␴eff as a function of photon energy, particularly the constant cross section over the large photon energy range between 2.2 and 4.4 eV as well as the thresholdlike behavior below 2.2 eV and above 4.4 eV, is very unusual for photochemical reactions of adsorbates on metal surfaces.39,40 Note that also no resonance is observed in contrast to the wavelength dependence in the liquid phase 共see Fig. 4兲.

after illumination @ 3.96 eV

IV. DISCUSSION 3.3 eV

A. Electronic structure of TBA/Au„111…

0.0 4.96

4.13

2.75 3.1 3.54 Photon energy [eV]

2.48

FIG. 4. 共Color online兲 UV-visible absorption spectra of TBA in cyclohexane before and after illumination with UV light at 3.96 eV. Exposure at 3.96 eV leads to a decrease in the absorbance around 3.81 eV, which is due to the trans/cis isomerization.

Figure 6 summarizes the binding energies of all electronic states observed in the present study. The Fermi level of the Au共111兲 surface serves as the reference. The vacuum level is identified by the work function of the TBA-covered surface. The LUMO level observed in 2PPE, which can be assigned to the N = N ␲ⴱ orbital,30 of trans-TBA is located 1.85 eV above the Fermi level. The LUMO position of the

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

164102-5

J. Chem. Phys. 129, 164102 共2008兲

Photoexcitation mechanism

E-EF [eV] 4.8

Energy levels LUMO+m LUMO+n (cis-TBA)

4.45 3.85

n = 1 (image state)

1.85 1.55

LUMO (trans-TBA) LUMO (cis-TBA)

0.0 -0.48

n = 0 (surface state)

-1.8 -2.0 -2.85 -3.0

B. Photoexcitation mechanism of surface-bound TBA

Ev

EF HOMO d-bands HOMO-1

FIG. 6. 共Color online兲 The energies of the observed photoemission spectral features in 0.9 ML TBA adsorbed on Au共111兲. All features are referenced to the Fermi level of Au共111兲.

switched cis-TBA is 1.55 eV. For comparison, a STS study on individual TBA molecules adsorbed on Au共111兲 also observed different energetic positions of the LUMO state of both isomers.21,41 This study determined the LUMO level of trans-TBA at 1.67 eV and the LUMO of the cis-TBA at 1.35 eV. While the energy difference between the LUMOs of ⬇300 meV found in both experiments is identical, the energetic positions identified by STS are shifted by 200 meV toward the Fermi level compared to the values obtained in 2PPE spectroscopy. In order to elucidate the exact origin of the difference between the spatial-resolved STS and k-resolved 2PPE techniques, further investigations comparing different molecular systems would be necessary. The appearance of different LUMO positions for the two isomers in the photoemission spectrum opens a way for a quantitative analysis of the amount of switched molecules in the photostationary state. Assuming that the detection of both LUMO states in the 2PPE occurs with the same probability, one can correlate the quantity of each isomer with the peak intensity. Fitting of the peak intensities and shapes using Gaussian peak profiles yields a value of 共55⫾ 5兲% cis-TBA in the photostationary state. In comparison, a low-temperature STM study has reported a value of 50% cis-TBA after light exposure at a photon energy of 3.3 eV.42 The discrepancy in the order of ⬇5% between both values might be due to the different substrate temperatures in both experiments, which are 5 K in the STM experiment compared to 90 K in the 2PPE measurements. In principle, a higher substrate temperature could lead to a different photostationary state, for instance, via a thermally assisted photoinduced process. The HOMO level of TBA/Au共111兲 is located at −1.8 eV and the HOMO-1 at −3.0 eV. STS measurements observed the HOMO of the trans-TBA also around −1.8 eV.41 Quantum chemical calculations suggested the HOMO to be the N = N ␲ orbital and the HOMO-1 a nonbonding orbital, reflecting dominantly the lone pairs at the nitrogen. The unoccupied final states 共LUMO+ m and LUMO+ n兲 are assigned to antibonding orbitals of ␲ symmetry located at the phenyl rings, but they could not be attributed to specific molecular states.30

In the following, we will address the excitation mechanism of the photoinduced isomerization on the basis of the observed electronic structure of TBA/Au共111兲 and the photon energy dependency of ␴eff. With the HOMO level located at −1.8 eV and the LUMO position at 1.85 eV 共trans-TBA兲, a HOMO-LUMO gap of 3.65 eV is obtained. A direct electronic excitation within the adsorbate like in the liquid phase, i.e., a HOMOLUMO transition, is incompatible with the observed photon energy dependence of the effective cross section 关see Fig. 5共b兲兴 since no resonance in the cross section is observed in this energy region. Note that the width of the LUMO peak is only ⬇250 meV, but the plateau region of ␴eff extends by over more than 2 eV. Hence we conclude that the intramolecular excitation can be ruled out. Another possible scenario is a substrate-mediated photochemical process, where hot electrons 共or holes兲 are attached to the adsorbate, creating a transient negative 共or positive兲 ion. In photoinduced chemistry on metal substrates such indirect excitation mechanisms, viz., the formation of an anionic state by a transient molecular resonance, plays a key role in many surface reactions. In particular, desorption and dissociation of adsorbates are the simplest photoinduced reactions which have been shown to be induced by indirect 共charge transfer兲 excitations.39,40,43–46 Hot hole induced processes are well known from the photochemistry of adsorbates on semiconductor surfaces.43,47–51 We consider first the generation of a negative ion resonance as a possible excitation mechanism for the isomerization reaction. With the LUMO level located around 1.85 eV above EF one would not expect the pronounced drop in ␴eff at energies below 2.2 eV. Moreover below a photon energy of 2 eV, switching of TBA is not observed. In addition, with increasing photon energy the population of hot electrons, which are resonant with the LUMO level, should increase nonlinearly and thus the probability for the formation of a negative ion resonance should be enhanced. However, since ␴eff as a function of photon energy is constant over the wide energy region between 2.2 and 4.4 eV and decreases exponentially below 2.2 eV, we conclude that electron transfer from the metal to the LUMO cannot be the dominant excitation process responsible for the switching. Instead, we propose the following mechanism 关see Fig. 7共a兲兴: Light exposure with photon energies higher than ⬇2.2 eV leads to photoexcitation of holes in the Au d-band 共electron-hole pair formation, step 1兲, which rapidly relax to the top of the d-band via Auger decay 共step 2兲. These hot holes undergo a charge transfer process to the HOMO level of TBA, thus producing a positive ion resonance 共step 3兲 as schematically shown in Fig. 7共a兲. In the photoemission spectrum shown in Fig. 3 it is seen that the spectral features from the Au d-band and the HOMO level exhibit some overlap. Therefore it is likely to assume that some degree of hybridization between the HOMO and the Au d-band is existent, even though it is expected that the four lateral tert-butyl groups should lead to an increased distance of the molecule and therefore to a reduced electronic coupling between the molecules and the metal surface. Due to the high density of

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

164102-6

J. Chem. Phys. 129, 164102 共2008兲

Hagen et al.

(a)

(b)

Ev

e1

EF

E [eV]

E

h


e-

ee-

0.0 sp-band + +

h
_ 4.4 eV

LUMO

SS n = 0

Au d-bands

h
~ 2.2 - 4.4 eV

3 +

h


h


2.2

HOMO

2

EFermi

e-

+

+

+

Ed-band

+

d-bands

2xEd-band

4.4 +

Substrate

Adsorbate

Density of states FIG. 7. 共Color online兲 共a兲 Proposed excitation mechanism for the photoinduced trans/cis isomerization of TBA adsorbed on Au共111兲 via the creation of a positive ion resonance. Thereby photoexcitation at photon energies above ⬇2.1 eV leads in the first step to electron-hole pair formation. The holes in the Au d-band relax to the top of the d-band 共step 2兲 followed by a hole transfer to the HOMO of TBA 共step 3兲. 共b兲 Formation of holes in the Au d-bands for different photon energies. For photon energies ⱖ4.4 eV, i.e., twice the energy of the d-band edge, the population of holes at the top of the d-band is enhanced due to Auger-decay processes 共see text兲.

states 共DOS兲 of the Au d-band, direct interband transitions to unoccupied states above the Fermi level of gold are efficient for h␯ ⱖ 2.2 eV, i.e., most of the absorbed photon energy is deposited in d holes rather than hot electrons.22 It has been shown for copper 共with a d-band edge comparable to gold兲 that holes in the d-band primarily float up to the top of the d-band on an ultrashort timescale of a few femtoseconds via electron-electron scattering. At the top of the d-band of Cu共111兲 the d-hole lifetime is substantially long, viz., in the order of 24 fs.52–55 As a similar electron dynamics can be expected for Au,53–55 the proposed mechanism of hole relaxation and transfer appears highly plausible. In the case of TBA adsorbed on Ag共111兲 the HOMO level of TBA is also observed at −1.8 eV with respect to EF but light-induced switching could not be achieved30 presumably due to the lower lying d-band in silver 共Ag d-band edge, ⬇−4 eV兲. Hence, an hybridization between the d-band and the HOMO and accordingly a hole transfer should not occur on Ag. The thresholdlike behavior of the effective cross section as a function of photon energy, i.e., the decrease below 2.2 eV and the increase above 4.4 eV, is related with d-band DOS and the hole formation process after photoexcitation. The first threshold around 2.2 eV corresponds to the energy of the d-band edge, viz., the minimum energy required to create a single hole in the Au d-band. The second threshold at 4.4 eV matches with twice the energy of the d-band edge 关see Fig. 7共b兲兴. Above this energy 共⬎4.4 eV兲 photoexcited holes can relax via Auger decay, resulting in the creation of a second d-band hole, as schematically shown in Fig. 7共b兲. As a consequence the population of holes at the top of the d-band is enhanced compared to hole formation in the energy regime below 4.4 eV and correspondingly this leads to an enhanced switching rate. The exponential increase in ␴eff above h␯ ⬇ 4.4 eV as well as the decrease below h␯ ⬇ 2.2 eV are related to the DOS of the Au d-band and

sp-band, respectively. Note that the onset of the d-band starts around 2 eV and the DOS rises with increasing binding energy, as schematically shown in Fig. 7共a兲. These pronounced variation in the DOS govern the hole relaxation process via Auger decay. The threshold effect on the photoisomerization efficiency due to the abrupt change in the DOS between the sp and d-bands is comparable to substrate-mediated photochemical processes in semiconductors.43 Thereby a common observation is the correlation of the photon energy threshold with the band gap of the substrate. Unlike metals, where scattering between excited carriers and conduction electrons near the Fermi level 共Auger decay兲 is the dominant relaxation process, electron-electron scattering is less important in semiconductors due to blocking of Auger decay by the gap. Therefore excited electrons 共or holes兲 mainly relax via electron-phonon scattering on a picosecond timescale. As a consequence the lifetime of the thermalized electrons 共or holes兲 at the band edge can be of the order of nanoseconds, hence significantly longer compared to the lifetime of holes at the band edge of noble metals such as Au or Cu. While on semiconductor surfaces, e.g., in semiconductor photocatalysis, hole induced processes are quite common,47–51 such processes are unusual on metal surfaces. We note, however, that a hole driven photochemical reaction at a metal surface has been proposed in the femtosecond surface chemistry of O2 / Pd共111兲, but contrary to the present study no direct experimental evidence for such a process was provided.56 An alternative pathway for photoinduced switching at photon energies above 4.4 eV could be the formation of a negative ion resonance where hot electrons are attached to an unoccupied final state 共LUMO+ n兲. Quantum chemical calculation suggested the unoccupied final states to be antibonding orbitals of ␲ symmetry located at the phenyl rings.30 For

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

164102-7

the free molecule in solution these orbitals are not involved in the isomerization process; therefore it is reasonable to assume that this is also valid for the surface-bound TBA. V. CONCLUSION

In summary, 2PPE spectroscopy has been utilized to determine the electronic structure of the molecular switch TBA adsorbed on a Au共111兲 surface as well as to identify the excitation mechanism of the photoinduced isomerization process. The isomerization of TBA/Au共111兲 is accompanied by significant changes in the electronic structure, namely, the appearance of an unoccupied final state 共LUMO+ n兲 of the cis isomer and a shift in the LUMO position toward lower energies by 300 meV of the cis-TBA, which can be used as a fingerprint for the conformational change. Based on the obtained electronic structure and the determination of the photoinduced isomerization efficiency as a function of photon energy, the excitation mechanism could be elucidated. While in the free molecule the direct 共intramolecular兲 optical electronic excitation provokes the conformational change, for the surface-bound molecules an indirect substratemediated process is responsible for the isomerization. We propose that for photon energies above ⬇2.2 eV excitation of holes in the Au d-band, which rapidly relax to the top of the d-band, and subsequent transfer to the HOMO level of TBA induces the isomerization. The effective cross section for the light-induced isomerization as a function of photon energy possesses a remarkable shape, It exhibits a thresholdlike decrease below 2.2 eV and increase above 4.4 eV. This behavior correlates with the energy of the d-band edge and the photon energy dependent hole formation process in the Au d-band. The present results demonstrate that the photoinduced excitation mechanism in the adsorbed TBA is quite contrary to the process in the free molecule, reflecting the strong influence of the metal surface on the molecular switching. Our findings imply new possibilities for the control of molecular geometry and functional properties of molecular switches at surfaces by light, which is of particular relevance for future developments of functional devices. ACKNOWLEDGMENTS

This work has been supported by the Deutsche Forschungsgemeinschaft through SFB 658. We thank S. Hecht and M. V. Peters 共Humboldt Universität Berlin兲 for the preparation of the azobenzene derivative. M. R. Bryce, M. C. Petty, and D. Bloor, Molecular Electronics 共Oxford University Press, New York, 1995兲. 2 Photochromism: Memories and Switches, edited by M. Irie, special issue of Chem. Rev. 共Washington, D.C.兲 100, 1683 共2000兲. 3 B. L. Feringa, Molecular Switches 共Wiley, Weinheim, 2001兲. 4 K. Uchida, N. Izumi, S. Sukata, Y. Kojima, S. Nakamura, and M. Irie, Angew. Chem., Int. Ed. 45, 6470 共2006兲. 5 R. Rosario, D. Gust, A. A. Garcia, M. Hayes, J. L. Taraci, T. Clement, J. W. Dailey, and S. T. Picraux, J. Phys. Chem. B 108, 12640 共2004兲. 6 N. Katsonis, T. Kudernac, M. Walko, S. J. van der Molen, B. J. van Wees, and B. L. Feringa, Adv. Mater. 共Weinheim, Ger.兲 18, 1397 共2006兲. 7 C. Joachim, J. K. Gimzewski, and A. Aviram, Nature 共London兲 408, 541 共2000兲. 8 N. Katsonis, M. Lubomska, M. M. Pollard, B. L. Feringa, and P. Rudolf, 1

J. Chem. Phys. 129, 164102 共2008兲

Photoexcitation mechanism

Prog. Surf. Sci. 82, 407 共2007兲. N. Tamai and O. H. Miyasaka, Chem. Rev. 共Washington, D.C.兲 100, 1875 共2000兲. 10 H. Rau, in Photochromism: Molecules and Systems, edited by H. Dürr and H. Bouas-Laurent 共Elsevier, Amsterdam, 2003兲. 11 D. Fanghänel, G. Timpe, and V. Orthman, in Organic Photochromes, edited by A. V. El’tsov 共Consultants Bureau, New York, 1990兲, p. 105. 12 T. Hugel, N. B. Holland, A. Cattani, L. Moroder, and H. E. Gaub, Science 296, 1103 共2002兲. 13 Z. F. Liu, K. Hashimoto, and A. Fujishima, Nature 共London兲 347, 658 共1990兲. 14 T. Ikeda and O. Tsutsumi, Science 268, 1873 共1995兲. 15 J. Wachtveitl, S. Spörlein, H. Satzger, B. Fonrobert, C. Renner, R. Behrendt, D. Oesterhelt, L. Moroder, and W. Zinth, Biophys. J. 86, 2350 共2004兲. 16 C. Zhang, M.-H. Du, H.-P. Cheng, X.-G. Zhang, A. E. Roitberg, and J. L. Krause, Phys. Rev. Lett. 92, 158301 共2004兲. 17 A. C. Whalley, M. L. Steigerwald, X. Guo, and C. Nuckolls, J. Am. Chem. Soc. 129, 12590 共2007兲. 18 D. Dulic, S. J. van der Molen, T. Kudernac, H. T. Jonkman, J. J. D. de Jong, T. N. Bowden, J. van Esch, L. B. Feringa, and J. B. van Wees, Phys. Rev. Lett. 91, 207402 共2003兲. 19 B.-Y. Choi, S.-J. Kahng, S. Kim, H. Kim, H. W. Kim, Y. J. Song, J. Ihm, and Y. Kuk, Phys. Rev. Lett. 96, 156106 共2006兲. 20 J. Henzl, M. Mehlhorn, H. Gawronski, K.-H. Rieder, and K. Morgenstern, Angew. Chem., Int. Ed. 45, 603 共2006兲. 21 M. Alemani, M. V. Peters, S. Hecht, K.-H. Rieder, F. Moresco, and L. Grill, J. Am. Chem. Soc. 128, 14446 共2006兲. 22 H. Petek and S. Ogawa, Prog. Surf. Sci. 56, 239 共1997兲. 23 M. Weinelt, J. Phys.: Condens. Matter 14, R1099 共2002兲. 24 X.-Y. Zhu, Surf. Sci. Rep. 56, 1 共2004兲. 25 C. D. Lindstrom and X.-Y. Zhu, Chem. Rev. 共Washington, D.C.兲 106, 4281 共2006兲. 26 J. Güdde, W. Berthold, and U. Höfer, Chem. Rev. 共Washington, D.C.兲 106, 4261 共2006兲. 27 S. Hagen, F. Leyssner, D. Nandi, M. Wolf, and P. Tegeder, Chem. Phys. Lett. 444, 85 共2007兲. 28 M. J. Comstock, N. Levy, A. Kirakosian, J. Cho, F. Lauterwasser, J. H. Harvey, D. A. Strubbe, J. M. J. Fréchet, D. Trauner, S. G. Louie, and M. F. Crommie, Phys. Rev. Lett. 99, 038301 共2007兲. 29 L. Óvári, M. Wolf, and P. Tegeder, J. Phys. Chem. C 111, 15370 共2007兲. 30 P. Tegeder, S. Hagen, F. Leyssner, M. V. Peters, S. Hecht, T. Klamroth, P. Saalfrank, and M. Wolf, Appl. Phys. A: Mater. Sci. Process. 88, 465 共2007兲. 31 E. W. Plummer and W. Eberhardt, Adv. Chem. Phys. 49, 533 共1982兲. 32 P. J. Benning, D. M. Poirier, T. R. Ohno, Y. Shen, M. B. Jost, F. Stepniak, G. H. Kroll, J. H. Weaver, J. Fure, and R. E. Smalley, Phys. Rev. B 45, 6899 共1992兲. 33 H. Eckardt, L. Fritsche, and J. Noffke, J. Phys. F: Met. Phys. 14, 97 共1984兲. 34 R. Courths, H.-G. Zimmer, A. Goldmann, and H. Saalfeld, Phys. Rev. B 34, 3577 共1986兲. 35 F. Reinert, G. Nicolay, S. Schmidt, D. Ehm, and S. Hüfner, Phys. Rev. B 63, 115415 共2001兲. 36 C. D. Lindstrom, D. Quinn, and X.-Y. Zhu, J. Chem. Phys. 122, 124714 共2005兲. 37 T. Fauster and W. Steinmann, in Electromagnetic Waves: Recent Developments in Research, edited by P. Halevi 共Elsevier, Amsterdam, 1995兲. 38 See EPAPS Document No. E-JCPSA6-129-003840 for the determination of the effective cross section for the photoinduced isomerization. For more information on EPAPS, see http://www.aip.org/pubservs/ epaps.html. 39 X. L. Zhou, X.-Y. Zhu, and J. M. White, Surf. Sci. Rep. 13, 73 共1991兲. 40 H.-L. Dai and W. Ho, Laser Spectroscopy and Photochemistry at Metal Surfaces 共World Scientific, Singapore, 1995兲. 41 M. Alemani, Ph.D. thesis, Freie Universität Berlin, 2007. 42 M. J. Comstock, N. Levy, J. Cho, L. Berbil-Bautista, M. F. Crommie, D. A. Poulsen, and J. M. J. Fréchet, Appl. Phys. Lett. 92, 123107 共2008兲. 43 X.-Y. Zhu, Annu. Rev. Phys. Chem. 45, 113 共1994兲. 44 F. M. Zimmermann and W. Ho, Surf. Sci. Rep. 22, 127 共1995兲. 45 H. Guo, P. Saalfrank, and T. Seideman, Prog. Surf. Sci. 62, 239 共1999兲. 46 Y. Matsumoto, Bull. Chem. Soc. Jpn. 80, 842 共2007兲. 47 M. R. Hoffmann, S. T. Martin, W. Choi, and D. W. Bahnemannt, Chem. Rev. 共Washington, D.C.兲 95, 69 共1995兲. 9

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

164102-8

A. L. Linsebigler, G. Lu, and J. T. Yates, Jr., Chem. Rev. 共Washington, D.C.兲 95, 735 共1995兲. 49 T. L. Thompson and J. T. Yates, Jr., Chem. Rev. 共Washington, D.C.兲 106, 4428 共2006兲. 50 R. Osgood, Chem. Rev. 共Washington, D.C.兲 106, 4379 共2006兲. 51 M.-H. Du, J. Feng, and S. B. Zhang, Phys. Rev. Lett. 98, 066102 共2007兲. 52 E. Knoesel, A. Hotzel, and M. Wolf, Phys. Rev. B 57, 12812 共1998兲. 48

J. Chem. Phys. 129, 164102 共2008兲

Hagen et al.

H. Petek, H. Nagano, and S. Ogawa, Phys. Rev. Lett. 83, 832 共1999兲. H. Petek, H. Nagano, and S. Ogawa, Appl. Phys. B: Lasers Opt. 68, 369 共1999兲. 55 R. Matzdorf, A. Gerlach, F. Theilmann, G. Meister, and A. Goldmann, Appl. Phys. B: Lasers Opt. 68, 393 共1999兲. 56 J. A. Misewich, S. Nakabayashi, P. Weigand, M. Wolf, and T. F. Heinz, Surf. Sci. 363, 204 共1996兲. 53 54

Downloaded 27 Oct 2008 to 147.47.241.210. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp