Electronic structure and excited state dynamics of clusters: What can we learn from experiments with synchrotron radiation?

Electronic structure and excited state dynamics in optically excited PTCDAfilms investigated with two-photon photoemissionM. Marks, S. Sachs, C. H. Schwalb, A. Schöll, and U. Höfer Citation: J. Chem. Phys. 139, 124701 (2013); doi: 10.1063/1.4818541 View online: http://dx.doi.org/10.1063/1.4818541 View Table of Contents: http://jcp.aip.org/resource/1/JCPSA6/v139/i12 Published by the AIP Publishing LLC. Additional information on J. Chem. Phys.Journal Homepage: http://jcp.aip.org/ Journal Information: http://jcp.aip.org/about/about_the_journal Top downloads: http://jcp.aip.org/features/most_downloaded Information for Authors: http://jcp.aip.org/authors

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

THE JOURNAL OF CHEMICAL PHYSICS 139, 124701 (2013)

Electronic structure and excited state dynamics in optically excited PTCDAfilms investigated with two-photon photoemission

M. Marks,1,a) S. Sachs,2 C. H. Schwalb,1,b) A. Schöll,2 and U. Höfer1

1Fachbereich Physik und Zentrum für Materialwissenschaften, Philipps-Universität Marburg,D-35032 Marburg, Germany2Universität Würzburg, Experimentelle Physik VII and Röntgen Research Center for Complex MaterialSystems (RCCM), D-97074 Würzburg, Germany

(Received 30 May 2013; accepted 1 August 2013; published online 24 September 2013)

We present an investigation of the electronic structure and excited state dynamics of optically ex-cited 3,4,9,10-perylene-tetracarboxylic acid dianhydride (PTCDA) thin films adsorbed on Ag(111)using two-photon photoemission spectroscopy (2PPE). 2PPE allows us to study both occupied andunoccupied electronic states, and we are able to identify signals from the highest occupied and thetwo lowest unoccupied electronic states of the PTCDA thin film in the 2PPE spectra. The energiesfor occupied states are identical to values from ultraviolet photoelectron spectroscopy. Comparedto results from inverse photoelectron spectroscopy (IPES), the 2PPE signals from the two lowestunoccupied electronic states, LUMO and LUMO+1, are found at 0.8 eV and 1.0 eV lower ener-gies, respectively. We attribute this deviation to the different final states probed in 2PPE and IPESand the attractive interaction of the photoexcited electron and the remaining hole. Furthermore, wepresent a time-resolved investigation of the excited state dynamics of the PTCDA film in the fem-tosecond time regime. We observe a significantly shorter inelastic excited state lifetime comparedto findings from time-resolved photoluminescence spectroscopy of PTCDA single crystals whichcould originate from excitation quenching by the metal substrate. © 2013 AIP Publishing LLC.[http://dx.doi.org/10.1063/1.4818541]

I. INTRODUCTION

The need for optimization of functional devices is not theonly reason why fundamental photo-physical phenomena inorganic semiconductors remain in the focus of research. Fur-thermore, investigations of structurally well-defined modelsystems promise a fundamental understanding of the inter-play between the electronic structure and dynamical phenom-ena in molecular organic materials, e.g., the dynamics of op-tically excited charge carriers.1 The electronic structure oforganic semiconductors is often determined using ultravioletphotoelectron spectroscopy (UPS) and inverse photoelectronspectroscopy (IPES) for occupied and unoccupied electronicstates, respectively.2 However, due to the different final statesencountered in UPS and IPES, the relation of spectroscopy re-sults to the transport and optical band gaps or exciton bindingenergies of the materials need to be discussed in detail.3–5

Time-resolved two-photon photoemission spectroscopy(2PPE) represents a complementary experimental approach tostudy the electronic structure of organic thin films using iden-tical final states for both occupied and unoccupied electronicstates. In addition, the possibility to perform time-resolvedinvestigations of excited electronic states enables a direct cor-relation of the dynamics of optical excitations with the elec-tronic structure. In the past, 2PPE has proven to be the most

a)Electronic mail: [email protected]. Present address: Department ofChemistry, Columbia University, New York, New York 10027, USA.

b)Present address: Physikalisches Institut, Goethe-Universität, D-60438Frankfurt, Germany.

successful technique for investigating unoccupied states andcharge carrier dynamics at surfaces and interfaces6–14 or formapping the surface-near unoccupied band structure.15, 16 Re-cent results show the potential of applying 2PPE to study ex-citon manifolds in organic semiconductor bulk materials.17, 18

For the correlation of electronic structure and opticalexcitations, one would like to take advantage of the fullpotential of 2PPE and investigate morphologically well char-acterized model systems. The organic molecular semi-conductor 3,4,9,10-perylene-tetracarboxylic acid dianhydride(PTCDA) is one of the structurally best characterized modelsystems for organic thin films.19–22, 27–30 In particular on aAg(111) substrate, essentially closed PTCDA ad-layers withonly minor variations in film thickness that contain crys-tals of both polymorphs, α- and β-phase, can be grownup to thicknesses of several tens of nanometers.27, 28 Theelectronic structure of PTCDA thin films is well charac-terized using UPS and IPES.3, 4, 31 In addition to the elec-tronic states of the organic film, an unoccupied, strongly dis-persive interface state emerges at the interface between theAg(111) substrate and the first chemisorbed PTCDA mono-layer (ML).23, 26, 32 Though the structural properties renderPTCDA a very good model system for investigations of fun-damental phenomena in molecular organic semiconductors,the relation between transport gap and the exciton bindingenergy is still under debate.3–5 Time-resolved 2PPE investiga-tions of PTCDA films on HOPG at room temperature suggestultra-fast decay dynamics with lifetimes of only a few hun-dred femtoseconds.33 These results seem to be in contrast to

0021-9606/2013/139(12)/124701/8/$30.00 © 2013 AIP Publishing LLC139, 124701-1

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-2 Marks et al. J. Chem. Phys. 139, 124701 (2013)

findings from time-resolved photoluminescence experimentswhich reveal radiative decay channels with lifetimes in thenanosecond regime even at room temperature.34

In this paper, the occupied and unoccupied electronicstructure of thin PTCDA films grown on Ag(111) is in-vestigated using time-resolved 2PPE. We show that the in-tramolecular excitations performed in our 2PPE experimentcorrespond to the transitions that are addressed in pure opticalexperiments using photoluminescence spectroscopy. There-after, the identification of electronic states in the obtained2PPE spectra will be discussed in detail before the assignmentis compared to complementary experimental techniques. Sub-sequently, the results from a time-resolved 2PPE investigationof the excited state dynamics are presented and discussed be-fore conclusions are drawn.

II. EXPERIMENTAL

The experiments were performed using an experimentalsetup as described earlier.35 The ultra-high vacuum (UHV)chamber with a base pressure of 1 × 10−10 mbar is equippedwith a hemispherical electron energy analyzer (Specs Phoi-bos 150) with a lens system for conformal imaging of theparallel momentum of the photoemitted electrons and a two-dimensional charge coupled device (CCD) detector. An op-tical parametric amplifier yielded laser pulses with a typi-cal photon energy of ¯ωvis = 2.33 eV and pulse lengths of50 fs that were split into two parts. Frequency doubling ofone part of the laser pulses produced pulses with a typicalphoton energy of ¯ωUV = 4.66 eV and pulse lengths of 58 fs.The fundamental and the frequency doubled laser pulses canbe delayed with respect to each other using a high resolu-tion (<1 fs) motor driven delay stage. The p-polarized laserpulses were focused on the sample at an angle of incidence of75◦ with respect to the surface normal. Photoluminescence ofthe PTCDA films was recorded in situ using a spectrographequipped with an amplified CCD-detector (Andor iStar).

The Ag(111) substrate was cleaned by standard sputter-annealing cycles.14 The PTCDA films were evaporated ontothe surface at a sample temperature of 270 K with maxi-mum growth rates of 0.5 monolayers per minute in order toachieve closed films with small variations in thickness.28, 36

The film thickness was controlled by evaporation time andcross checked using the attenuation of the Ag 3d signal inthe X-ray photoemission spectra. The prepared PTCDA filmsconsist of coexisting α- and β-phases of bulk PTCDA.27, 28

For the verification of the luminescence spectra, an addi-tional annealing cycle of the prepared PTCDA films at 400 Kfor 60 min was employed to induce the formation of crys-talline islands in the α-structure on top of a two ML wet-ting layer.27, 28 All 2PPE experiments were conducted with thesample held at room temperature.

III. RESULTS

2PPE spectra of organic adsorbates on metal surfacesare commonly dominated by surface or interface states withmetallic origin. In the case of PTCDA/Ag(111) these are theinterface state located about 0.6 eV above the Fermi-level

and the image-potential states close to the vacuum level.23–26

The wave functions of these states have large overlap withthe metal and can thus be populated efficiently by excitingmetal electrons with the pump pulse. In contrast, unoccu-pied states of organic molecules tend to have small overlapwith the metal substrate. Moreover, their signatures are fur-ther suppressed by small photoemission matrix elements atthe typical photon energies used in 2PPE experiments. In thissection, we first present photoluminescence experiments forthin films of PTCDA consisting of tens of molecular layers.Our results show that under the conditions of the 2PPE ex-periments, electronic transitions are induced in the moleculeswith considerable probability. Under the same experimentalconditions, we observe characteristic signatures in the 2PPEspectra that are not visible for monolayer coverages. We canclearly determine the character of these additional features asoccupied or unoccupied states as well as their energy posi-tion relative to the Fermi-level. This allows us to assign thestates to molecular HOMO and LUMO levels. We will onlydescribe the experimental results in this section. Possible ex-citation pathways will be discussed in Sec. IV together withthe assignment to molecular states.

A. Photoluminescence of PTCDA thin films

Fig. 1(b) depicts luminescence spectra for a PTCDAfilm with a nominal thickness of 25 ML. The spectra wererecorded directly after preparation (top) and after annealingof the film at 400 K for 60 min (center and bottom). Ex-citation of the molecules was performed using photon ener-gies of ¯ωvis = 2.43 eV (green spectra, top and center) at

EF

Evac

Ene

rgy

HOMO

HOMO–1

LUMO

LUMO+1

ωUV

ωvis

1.4 1.6 1.8 2.0 2.2

12 13 14 15 16 17

Photon Energy ω (eV)

Wavenumbers (1000 cm–1)

Lum

ines

cenc

e S

igna

l (ar

b. u

nits

)

ωUV

ωvis

ωvis

Exc.:

as–grown

cryst.

×50

Vib. F / CT1 CT2D / M

(a) (b)

FIG. 1. (a) Schematic representation of optical intramolecular excitationpathways within the PTCDA thin films with photon energies ¯ωUV (blue/longarrows) and ¯ωvis (green/short arrows). (b) Normalized PL spectra of a25 ML PTCDA film on Ag(111) recorded at a temperature of 85 K. The twoupper/green spectra were obtained using photon energies ¯ωvis = 2.43 eVdirectly after film growth (top) and after an annealing cycle (center). Thelower/blue spectrum stems from the annealed film excited with a photon en-ergy of ¯ωUV = 4.86 eV (bottom). The signals are assigned to contribu-tions from charge transfer (CT) and Frenkel (F) excitonic states, their vi-bronic replica (Vib.), and monomer excitations at defects (D/M) (see text fordetails).

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-3 Marks et al. J. Chem. Phys. 139, 124701 (2013)

the absorption maximum of PTCDA thin films27 as well ashigher energy UV-photons ¯ωUV = 4.86 eV (blue spectrum,bottom). We recorded additional spectra at excitation photonenergies identical to the ones employed in the 2PPE experi-ments. However, except for a slightly lower signal intensity,no significant spectral changes were observed.

The photoluminescence spectrum of the unannealed filmin Fig. 1(b) (top) reveals a considerable broadening of thepeak structure due to the coexistence of crystallites in theα- and β-polymorphs. From these spectra, an unambiguousidentification of the contributing signals is difficult. Lumines-cence spectra from annealed PTCDA films reveal a similarpeak structure and less signal broadening (cf. Fig. 1(b), centerand bottom). While the relative peak intensities vary, the en-ergies of the luminescence transitions are not modified by theannealing procedure and the spectra can be employed for thepeak identification. Following the assignment by Gangilenkaet al.,37 the peaks comprise signals from charge transfer (CT)and Frenkel-exciton states (F) and their respective vibronicreplica. The intensity of the signal on the higher energy sideof the spectrum (D/M) is significantly larger for the unan-nealed film and can be assigned to emission from defectsand a monomer excitation.27 As indicated in Fig. 1, the quan-tum efficiency of a UV excitation ¯ωUV is significantly lower.However, the spectral composition indicates identical radia-tive decay channels. From this, we conclude that higher exci-tonic states effectively relax toward lower states via internalconversion instead of a decay via higher radiative channels.Figure 1(a) summarizes the optical transitions between thePTCDA frontier orbitals that can be accessed with the avail-able photon energies ¯ωvis and ¯ωUV.

B. 2PPE-spectroscopy of molecular states

A 2PPE spectrum for a 33 ML PTCDA film at normalemission is shown in Figure 2(a) (black line) as a functionof the final state energy of emitted photoelectrons above theFermi-energy EF of the Ag(111) substrate. The spectrum wasrecorded using UV-photons ¯ωUV for both excitation andemission. It reveals a clear low energy cut-off around� = E − EF = 4.7 eV in accordance with our previous 2PPEinvestigations.24 However, there we observed an unoccupiedPTCDA/Ag(111) interface state slightly above the low energycut-off that was transiently populated by electrons from thesubstrate.23, 26 A distinct signal from this state cannot be de-tected for the higher PTCDA coverages studied here. Thisis explained by the short inelastic mean free path ofλ = 2.9 ML for low energy electrons in PTCDA.24 In con-trast, for these thicker films the photoemission intensity con-tinuously decreases above the low energy cut-off until the sig-nal can no longer be separated from the background aroundE − EF = 7.8 eV. Superimposed with this monotonously de-creasing background, intensity modulations are observed inthe 2PPE spectrum that originate from the electronic states ofthe PTCDA film.

Due to the short inelastic mean free path of the photo-electrons, a high intensity of inelastically scattered electronsis expected to contribute to the 2PPE signal. In order to facil-itate the determination of the energies of the molecular states,

8.07.57.06.56.05.55.04.5

Final State Energy E-EF (eV)

2PP

E S

igna

l (ar

b. u

nits

)

(a)

(b)

Exp. Data Background Fit

Corrected Data Fit

FIG. 2. (a) 2PPE spectrum of a 33 ML PTCDA film adsorbed on Ag(111)(black line) recorded for UV-excitation with the sample held at room tem-perature. The photon energy was adjusted to ¯ωUV = 4.67 eV. The fittedbackground function is shown in red. (b) Background corrected 2PPE data(black dots) together with a fit (black line) using four Gaussian contributions(colored lines).

the inelastic background needs to be subtracted. Following thedescription of inelastically scattered Auger electrons, we fit abackground function Ibg(E) ∝ A · E−m to our obtained 2PPEdata (Fig. 2(a), red solid line).38 After subtraction of the in-elastic background, the resulting difference spectrum revealsa clear peak structure (cf. Fig. 2(b), black dots). The data canbe reproduced by a superposition of a minimum of four Gaus-sian contributions (black solid line). The individual compo-nents are shown as colored solid lines.

In order to assign the observed signals to molecularstates, 2PPE spectra have been recorded for different pho-ton energies ¯ωUV. Figure 3 shows the obtained data set fora 33 ML thick PTCDA film after subtraction of the secondaryelectron background. Except for the spectra recorded withphoton energies ¯ωUV ≤ 4.56 eV, all spectra can be describedby four Gaussian peaks (cf. Fig. 2). As illustrated by the solidlines, the peaks reveal a dependence on the applied photonenergy. With increasing photon energy, the signal at low en-ergies (L) and the two signals on the high energy side of thespectrum (L1, H) shift to higher final state energies. In con-trast, the position of the fourth signal (UFS) only weakly de-pends on photon energy.

A quantitative evaluation of the photon energy depen-dence can be employed for the assignment of the signals toelectronic states of the PTCDA thin film. For this purpose,the peak positions obtained in the fitting procedure are plot-ted as a function of photon energy in the inset of Figure 3.The final state energy of all four signals increases linearlywith photon energy. The slopes m from a fit using a linearrelation are all close to integer values between 0 and 2. Sincethe employed photon energies (¯ωUV) are chosen to be com-parable to the sample work function, the results enable us todirectly conclude on the underlying photoemission process.Electrons from occupied electronic states below EF can onlybe photoemitted in a direct non-resonant 2PPE process with

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-4 Marks et al. J. Chem. Phys. 139, 124701 (2013)

Nor

mal

ized

2P

PE

Sig

nal

(arb

. uni

ts)

8.07.57.06.56.05.55.0

Final State Energy E-EF (eV)

7.5

7.0

6.5

6.0

5.5

5.0

E-E

F (

eV)

4.94.84.74.64.5

0.8

0.0

1.0

2.0

UFS

L

L1

H

L

UFS

H

4.48

4.56

4.86

4.78

4.69

L1

4.60

4.56

4.51

4.73

4.83

FIG. 3. UV+UV difference spectra from a room temperature 2PPE mea-surement after background subtraction (cf. Fig. 2). The fits using up to fourGaussian contributions are shown as black and the individual signals as col-ored lines. Straight lines indicate the photon energy dependence of the in-dividual signals. The peak positions are shown in the inset as a function ofthe applied photon energy. Linear fits are used to extract information on thephotoemission pathway (see text).

simultaneous absorption of two photons. Therefore, a shift ofthe final state energy with twice the change of the photon en-ergy �E ∝ 2¯ω is expected. In contrast, the photoemissionof electrons after excitation into unoccupied electronic statesbetween EF and Evac is performed by absorption of only onephoton. A variation of the photon energy leads to an inducedshift of the signal energy ∝ 1 ¯ω. The final state energy of anunoccupied state above Evac is independent of photon energy.Based on the slope of the peak shift, we conclude that peak H(mH = 2.0) originates from non-resonant emission from anoccupied PTCDA state located at E − EF = −2.30± 0.20 eV. The signals L and L1 (mL = 0.8 and mL1 = 1.0) areassigned to intramolecular excitations into unoccupied molec-ular orbitals at E − EF = 0.47 ± 0.20 eV and E − EF = 1.77± 0.20 eV. The independence of the signal UFS on photonenergy is explained by an unoccupied molecular state locatedabove the vacuum energy at E − EF = 5.72 ± 0.20 eV.

IV. IDENTIFICATION OF MOLECULAR STATESAND DISCUSSION

The results of our 2PPE investigation are summarized inthe center column of Fig. 4. The energetic positions of theidentified signals are indicated by red and green horizontallines for occupied and unoccupied states, respectively. All en-ergies are referred to the Fermi-energy of the Ag(111) sub-strate. The linewidths from the fitting procedure are illus-trated by vertical light red and green bars. The left column ofFig. 4 contains results from a UPS/IPES investigation of the

6

5

4

3

2

1

0

-1

-2

-3

-4

-5

Ene

rgy

E-E

F (

eV)

UV+UV green+UV

2PPE

LUMO+1

LUMO+2

LUMO

HOMO H

L

L1

UFS

HOMO-1

Lg

UPS/IPES

FIG. 4. Energy scheme of the observed electronic states in our single color2PPE spectra (UV+UV, center) and the additional state appearing in the twocolor 2PPE experiments (vis+UV, right). The dark red (dark green) linesmark the energetic positions of occupied (unoccupied) states with respect tothe Ag(111) Fermi-energy EF, while vertical light green and light red barsindicate the corresponding peak widths (full width at half maximum). Forcomparison, the left-hand side summarizes the results of the UPS/IPES in-vestigation from Ref. 3.

electronic structure of PTCDA films taken from Ref. 3. Inaccordance with UPS3, 4, 31 and earlier 2PPE investigations,33

we assign the occupied electronic state to the HOMO-derivedband. The two signals from unoccupied levels are identified asthe LUMO (L) and the LUMO+1 (L1) based on their positionrelative to the sample Fermi-energy and the relation to resultsfrom IPES. The unoccupied final state (UFS) located abovethe sample work function at E − Evac = 1.02 eV presumablyoriginates from one of several π -derived electronic states thathave been observed in the energy range between 1 and 2 eVabove the vacuum level using total current spectroscopy39, 40

and near-edge X-ray absorption fine structure (NEXAFS).41

An additional signal (Lg) with a comparable peak widthcan be observed in the 2PPE spectra when the PTCDA filmis excited by coinciding UV (¯ωUV) and visible (¯ωvis) laserpulses (cf. Fig. 5). The two color excitation leads to a signifi-cantly increased 2PPE intensity which suppresses the weakersingle color 2PPE signals. However, an analogous back-ground correction of energy dependent 2PPE spectra as de-scribed above enables the assignment of an additional unoc-cupied electronic state around the Fermi-energy at E − EF

= 0.03 ± 0.20 eV. The electrons are excited into this state byphotons from the visible laser pulse and emitted by UV pho-tons. For comparison, the result is shown in the right columnof Fig. 4.

In both 2PPE and UPS, the emission of electrons fromoccupied states comprises identical final states, i.e., the pho-toelectron and a transient hole in the HOMO-derived band.The identical binding energy within the experimental errorsof both techniques confirms this hypothesis. In contrast, theenergy positions of the unoccupied states LUMO (L) and

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-5 Marks et al. J. Chem. Phys. 139, 124701 (2013)

Inte

grat

ed2P

PE

Sig

nal

FIG. 5. 2PPE spectra of 33 ML PTCDA on Ag(111) for different pump-probe delays recorded at a sample temperature of 293 K using photonenergies of ¯ωvis = 2.43 eV for excitation and ¯ωUV = 4.86 eV for photo-emission. All spectra have been corrected for the time-independent UV back-ground. Positive pump-probe delay represents delayed UV emission pulses.The inset depicts the integrated 2PPE intensity as a function of the pump-probe delay on a semi-logarithmic scale (red dots). The data can be de-scribed with an exponential decay using three different time constants τ 1,τ 2, τ 3 (black solid line). The dashed line is an estimate of the maximum timeconstant τm.

LUMO+1 (L1) that we extracted out of our 2PPE spectra are∼0.8–1.0 eV closer to the Fermi-energy compared to the val-ues obtained from IPES.3 This arises from the distinct pho-toemission processes encountered in 2PPE and IPES. IPESprobes the electronic structure of the PTCDA film using anelectron that transiently populates the respective unoccupiedelectronic states. Therefore, the final state in IPES contains anadditional charge. On the other hand, the emission in 2PPEensues an intramolecular excitation. This is confirmed by theobserved photoluminescence (cf. Sec. III A). Hence, the in-termediate state in 2PPE is electrically neutral and consistsof an electron excited into an unoccupied molecular orbitalderived state and a hole located in one of the higher occu-pied states (cf. Fig. 1). The lower energy position in our 2PPEstudy can thus be explained by the attractive interaction be-tween the electron and the hole. The transient formation ofthese excitonic states leads to energetically lower lying sig-nals compared to IPES results.

As shown in the excitation scheme in Fig. 1, by em-ploying UV photons ¯ωUV the electrons are excited from theHOMO-1 and HOMO into excitonic levels related to LUMOand LUMO+1, respectively. These excitations result in higherintramolecular excitations compared to a direct HOMO toLUMO transition. Such higher excited states normally relaxto the lowest excited state via internal conversion. In this con-text, it is important to note that in our single color 2PPE spec-tra the excitation and the subsequent photoemission out of theexcited state is due to simultaneous absorption of two photonsfrom the same laser pulse. The resulting time delay betweenexcitation and photoemission is extremely short thus render-ing relaxation of higher excited states due to internal con-version within the excitation time unlikely. We consequently

assume that states occupied by such internal conversion donot contribute significantly to the single color 2PPE spectra.Therefore, both, L and L1, correspond to higher molecular ex-citations which have a comparable energy shift relative to theIPES results.

Considering its energetic positions, the additional signalLg that we observe in our two color 2PPE data has no corre-sponding signal in the IPES spectra. As indicated in the exci-tation scheme in Fig. 1, the excitation of the PTCDA thin filmwith photons ¯ωvis can induce electronic transitions from theHOMO-derived band into the excitonic states related to theLUMO of PTCDA. In addition, the energy position of Lg rel-ative to the HOMO-derived band is in accordance with the ex-citation energy of the S0–S1 transition.42 We suppose that thelower energy position of Lg compared to L originates froma stronger interaction between electron and hole for the tran-sitions between the adjacent molecular states. This enhancedinteraction could originate from altered charge density dis-tributions and the energy difference is resolved in our 2PPEexperiment. Hence, we do not assign Lg to an additional un-occupied molecular electronic level but to originate from thelowest exciton manifold S1 of the PTCDA thin film. This as-signment is corroborated by the observed photoluminescence(cf. Fig. 1) and the higher quantum efficiency compared toluminescence after UV excitation. The energy difference ofthe 2PPE signals Lg and L relative to the results from IPES,however, should not be confused with the exciton binding en-ergy. It can be regarded as an upper limit since the differencecomprises the exciton binding energy, different final states,and variations in the charge density distribution as discussedabove.

It is worth mentioning that the width of peak L1 (FWHM)is more than twice the width of all other observed signals.This indicates that several intermediate excitonic states con-tribute to the L1 signal. In contrast to the LUMO whichis energetically well separated from other electronic states,the energy separation of the adjacent LUMO+1,+2, . . . -derived electronic levels of PTCDA is below the experimentalresolution.43

V. TIME-RESOLVED EXPERIMENTS

After identification of the signals that are observed inour experiment, we now discuss the two-color 2PPE spec-tra for systematically varied time-delays between excitation(¯ωvis = 2.43 eV) and photoemission (¯ωUV = 4.86 eV)pulse. Time-resolved spectra obtained from a 33 ML thickPTCDA film at room temperature are displayed in Fig. 5 upto the maximal achievable pump-probe delay of our exper-imental setup of 15 ps. Positive time-delays represent de-layed UV photoemission pulses. As discussed in Sec. III B,the application of only the UV emission pulses leads to sig-nificant 2PPE signals from molecular states. In a two-colortime-resolved 2PPE experiment, however, these signals rep-resent a constant background independent of the time-delaybetween the pulses. We corrected the data shown in Fig. 5for this time-independent single color 2PPE signal by sub-tracting a single color 2PPE spectrum. Hence, the 2PPE signalsolely originates from excited electrons of ¯ωvis-photons and

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-6 Marks et al. J. Chem. Phys. 139, 124701 (2013)

the intensity is directly proportional to the intermediate statepopulation. In contrast to the background subtraction proce-dure described above, subtraction of reference spectra rep-resents a more reliable approach for time-resolved measure-ments. Though the determination of the inelastic backgroundfor each individual spectrum enables a more reliable determi-nation of the actual peak position at varying delay times, thiscould lead to an erroneous negligence of a time-dependent in-elastic background related to population dynamics in the un-occupied states.

For negative time-delays, no 2PPE signal is observed af-ter the background subtraction. For overlapping laser pulses,the 2PPE intensity steeply rises to its maximal intensityaround time zero and then steadily decreases for increasingtime-delays. Despite a broad shoulder around E − EF ≈ 5 eVthat is only visible for time-delays � 2 ps, the spectral shapeis independent of time-delay. Directly above the low-energycut-off the intensity is maximal and then steadily decreasesuntil it drops below the experimental sensitivity aroundE − EF ≈ 6 eV. Apart from a sharp peak directly above sam-ple work function, the time-dependent 2PPE signal consists ofone broad signal. The intensity quickly decreases within thefirst 2 ps before the decay slows down. Similarly, the shoulderthat is observed around E − EF ≈ 5 eV for overlapping laserpulses disappears within the first 2 ps and cannot be discrim-inated from the long-lived signal. The long-lived componentscan easily be detected up to the maximum achievable time-delay in our experiment of 15 ps.

VI. DISCUSSION: DYNAMICS OF INTRAMOLECULAREXCITATIONS

The 2PPE intensity which we observe in our time-resolved 2PPE experiments mainly originates from electronsemitted from the unoccupied electronic state Lg. However thesignal reveals an unusual peak form which is assumed to orig-inate from different contributions. Owing to the emission pho-ton energy that was chosen to be below the sample work func-tion (¯ωUV < �sample) for the time-resolved experiments, thebroad emission of electrons from Lg is cut by the work func-tion on the low energy side and cannot be resolved as a sym-metric isolated signal in the spectrum. Furthermore, the sig-nal from Lg is superimposed with a sharp maximum that islocated directly above the low energy cut-off. This maximumpresumably originates from inelastically scattered photoelec-trons directly at the vacuum energy.44 However, despite thedifferent contributions, the observed time-dependent intensityvariations are due to population dynamics of Lg. Based onthe intense photoluminescence of the thin film after opticalexcitation using ¯ωvis-photons and the energy position of Lg

relative to the HOMO-derived band that agrees well with theexcitation energy of the S0 − S1 transition42 we assign theobserved population dynamics to the lowest lying excitonicstates within the PTCDA film.

Ino et al. employed a UV+UV excitation scheme tostudy the electronic structure and excited state dynamics ofPTCDA films adsorbed on highly oriented pyrolytic graphiteand attain to a similar interpretation of their time-resolveddata.33 The energetic positions of the signals agree with our

assignment within the experimental error and the ultra-shortlifetimes identified by Ino et al.33 are consistent with the re-laxation of the highly excited excitonic states around E − EF

≈ 5 eV. The long-lived component of Lg which we assign tothe lowest exciton manifold, however, was not observed byIno et al.33 We assume that the combination of their single-color excitation scheme and the chosen background correc-tion suppressed the weak signal of the long-lived componentto below their experimental sensitivity. However, in contrastto our single-color UV+UV 2PPE results, the energy of Lg

could be extracted from their data (cf. Fig. 3).33 The appar-ent discrepancy can be explained by the energetic relaxationof the hole and the excited electron. This relaxation is sup-pressed in our single-color experiment since the two photonsfrom the same laser pulse are absorbed simultaneously.

Though we are not aware of further 2PPE investigationsof excited state dynamics in PTCDA thin films, the propertiesof the excitons in α-crystalline PTCDA and their assignmenthas been subject to extensive research.27, 34, 37, 42, 45, 46 How-ever, the energy separation between the different excitonicstates are by far smaller (∼0.2 eV) compared to our broad2PPE signal. Hence, we refrain from assigning the severalcharge-transfer and Frenkel exciton states, but assume that the2PPE signal and therewith the observed dynamics comprisessignals from different types of excitonic states.

In order to determine the time-scales involved in theexcited state dynamics, we integrated the 2PPE intensityS(t) = ∫

S(t, E)dE and plotted it as a function of the time-delay in the inset of Fig. 5 (dots). The overall signal decaycannot be described with a single exponential decay suppos-edly owing to the different excitonic states involved in the ex-cited state dynamics. A superposition of three exponential de-cays with different time constants τ i (solid black line) can bebrought into agreement with the data. An integration of sep-arate sections within the peak width of the 2PPE spectra didnot result in significant variations of the lifetimes indicatinga superposition of several dynamic processes. For a coverageof 33 ML, the time-constants are found to be τ 1 = 190 fs,τ 2 = 1.5 ps, and τ 3 = 23 ps. The third time-constant exceedsthe maximal achievable delay in our experiment of 15 ps. Wetherefore can only estimate the upper lifetime limit that can bebrought into agreement with our 2PPE data. The dashed linein Fig. 5 represents an exponential decay with a time-constantof τm = 29 ps. However, around 20% of the population hasnot yet decayed and a further deceleration of the populationcannot be ruled out. Additional time-resolved data sets thatwere recorded for PTCDA films with varying thickness be-tween 5 and 100 ML reveal similar excited state dynamics(not shown). In all cases, the data can be modeled by a tri-exponential decay and the lifetimes increase with increasingfilm thickness. The lifetime of the long-lived component, forexample, rises from a few picoseconds up to around 100 psfor a coverage of around 100 ML.

The origin of the short-lived shoulder which is ob-served in our 2PPE spectra around E − EF ≈ 5 eV cannotbe assigned unambiguously. However, the energy differencebetween non-relaxed (delocalized) and relaxed (localized) ex-citonic states in α-PTCDA is similar to the energetic separa-tion of the shoulder and to the long-lived 2PPE components.

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-7 Marks et al. J. Chem. Phys. 139, 124701 (2013)

Together with the significantly shorter lifetime, this suggeststhat shoulder corresponds to higher non-relaxed delocalizedexcitonic states within the first exciton manifold or withinthe excitonic band structure.37, 47 The strong coupling of ex-cited electrons to phonons within PTCDA films and an ef-fective phonon mode with an energy of 170 meV can ex-plain the very effective energetic relaxation within the firsttwo picoseconds.48

Compared to the excited state dynamics in α-PTCDAthat was determined using luminescence spectroscopy, theabsolute lifetimes from our time-resolved 2PPE experimentsare shorter by three orders of magnitude.34, 45 We assignthis significant difference to the proximity of the metal sub-strate surface which facilitates additional decay channels viaelectron-hole pair excitation. This explanation is corroboratedby the dependence of inelastic lifetimes on the film thicknessand the observation of luminescence quenching in PTCDAthin films adsorbed on Ag(111).49 Additionally, the defectdensity is higher in smooth PTCDA thin films comparedto α-PTCDA.27, 28 This renders fast non-radiative channelsmore important for the films that have been studied here.27

In addition, since excitation densities around 4 × 1019 to4 × 1020 cm−3 were achieved in our experiments (absorptioncoefficient ∼3 × 105 cm−1, Ref. 50), a possible contributionfrom exciton-exciton annihilation to the fast initial populationdecay within the first 5 ps cannot be ruled out completely.51

VII. SUMMARY AND CONCLUSIONS

In summary, we presented a detailed 2PPE investigationof the electronic structure of PTCDA thin films adsorbed ona Ag(111) single crystal surface. After a suitable treatmentof background signal from inelastically scattered photoelec-trons, the energetic positions of the highest occupied and twolowest unoccupied electronic states are extracted out of the2PPE spectra. While the energetic position of the HOMO-derived band is identical to the energy determined in UPS ex-periments, we observe systematically lowered energetic po-sitions for the unoccupied states in contrast to IPES. We at-tribute the systematic difference to the significantly differentfinal states that are probed in 2PPE and IPES.

An additional signal is observed around the Fermi-energyof the substrate for a resonant excitation of the PTCDA film.This signal is assigned to the lowest optical intramoleculartransition from the HOMO-derived band into the excitonicstates related to the PTCDA LUMO. Time-resolved 2PPEwas employed to study the excited state dynamics ensuingthe resonant excitation. Compared to results from photolumi-nescence spectroscopy in α-PTCDA, the extracted lifetimesare found to be shorter by 3 orders of magnitude. This in-dicates the importance of an additional non-radiative decaychannel in PTCDA thin films due to defects between α- andβ-crystallites and charge carrier transport to the metal-organicinterface.

ACKNOWLEDGMENTS

We thank S. Krause, F. Reinert, and E. Umbach for fruit-ful discussions. Funding by the Deutsche Forschungsgemein-schaft through Grant Nos. SFB1083, SPP1121, and GK1221

as well as by the Alexander von Humboldt Foundation (M.M.)is gratefully acknowledged.

1X. Y. Zhu, Surf. Sci. Rep. 56, 1 (2004).2C. Deibel, D. Mack, J. Gorenflot, A. Schoell, S. Krause, F. Reinert, D.Rauh, and V. Dyakonov, Phys. Rev. B 81, 085202 (2010).

3S. Krause, M. B. Casu, A. Schöll, and E. Umbach, New J. Phys. 10, 085001(2008).

4I. G. Hill, A. Kahn, Z. G. Soos, and R. A. Pascal, Chem. Phys. Lett. 327,181 (2000).

5S. Sharifzadeh, A. Biller, L. Kronik, and J. B. Neaton, Phys. Rev. B 85,125307 (2012).

6P. M. Echenique, R. Berndt, E. V. Chulkov, T. Fauster, A. Goldmann, andU. Höfer, Surf. Sci. Rep. 52, 219 (2004).

7P. Szymanski, S. Garrett-Roe, and C. B. Harris, Prog. Surf. Sci. 78, 1(2005).

8J. Stähler, C. Gahl, U. Bovensiepen, and M. Wolf, J. Phys. Chem. B 110,9637 (2006).

9T. Fauster, M. Weinelt, and U. Höfer, Prog. Surf. Sci. 82, 224 (2007).10M. Cinchetti, K. Heimer, J. Wüstenberg, O. Andreyev, M. Bauer, S. Lach,

C. Ziegler, Y. Gao, and M. Aeschlimann, Nature Mater. 8, 115 (2009).11T. Ichibayashi and K. Tanimura, Phys. Rev. Lett. 102, 087403 (2009).12M. P. Steele, M. L. Blumenfeld, and O. L. A. Monti, J. Chem. Phys. 133,

124701 (2010).13M. Shibuta, K. Miyakubo, T. Yamada, and T. Munakata, J. Phys. Chem. C

115, 19269 (2011).14M. Marks, K. Schubert, C. H. Schwalb, J. Güdde, and U. Höfer, Phys. Rev.

B 84, 245402 (2011).15Z. Hao, J. I. Dadap, K. R. Knox, M. B. Yilmaz, N. Zaki, P. D. Johnson, and

R. M. Osgood, Phys. Rev. Lett. 105, 017602 (2010).16A. Winkelmann, V. Sametoglu, J. Zhao, A. Kubo, and H. Petek, Phys. Rev.

B 76, 195428 (2007).17W. L. Chan, M. Ligges, A. Jailaubekov, L. Kaake, L. Miaja-Avila, and

X. Y. Zhu, Science 334, 1541 (2011).18E. Varene, L. Bogner, C. Bronner, and P. Tegeder, Phys. Rev. Lett. 109,

207601 (2012).19E. Umbach, M. Sokolowski, and R. Fink, Appl. Phys. A: Mater. Sci. Pro-

cess. 63, 565 (1996).20K. Glöckler, C. Seidel, A. Soukopp, M. Sokolowski, E. Umbach, M.

Bohringer, R. Berndt, and W. D. Schneider, Surf. Sci. 405, 1 (1998).21Y. Zou, L. Kilian, A. Schöll, T. Schmidt, R. Fink, and E. Umbach, Surf.

Sci. 600, 1240 (2006).22F. S. Tautz, Prog. Surf. Sci. 82, 479 (2007).23C. H. Schwalb, S. Sachs, M. Marks, A. Schöll, F. Reinert, E. Umbach, and

U. Höfer, Phys. Rev. Lett. 101, 146801 (2008).24S. Sachs, C. H. Schwalb, M. Marks, A. Schöll, F. Reinert, E. Umbach, and

U. Höfer, J. Chem. Phys. 131, 144701 (2009).25C. H. Schwalb, M. Marks, S. Sachs, A. Schöll, F. Reinert, E. Umbach, and

U. Höfer, Eur. Phys. J. B 75, 23–30 (2010).26M. Marks, N. L. Zaitsev, B. Schmidt, C. H. Schwalb, A. Schöll, I. A.

Nechaev, P. M. Echenique, E. V. Chulkov, and U. Höfer, Phys. Rev. B 84,081301(R) (2011).

27M. Schneider, E. Umbach, and M. Sokolowski, Chem. Phys. 325, 185(2006).

28B. Krause, A. C. Durr, K. Ritley, F. Schreiber, H. Dosch, and D. Smilgies,Phys. Rev. B 66, 235404 (2002).

29A. Hauschild, K. Karki, B. C. C. Cowie, M. Rohlfing, F. S. Tautz, and M.Sokolowski, Phys. Rev. Lett. 94, 036106 (2005).

30J. Ziroff, F. Forster, A. Schöll, P. Puschnig, and F. Reinert, Phys. Rev. Lett.104, 233004 (2010).

31H. Yamane and S. Kera, Phys. Rev. B 68, 033102 (2003).32R. Temirov, S. Soubatch, A. Luican, and F. S. Tautz, Nature 444, 350

(2006).33D. Ino, K. Watanabe, N. Takagi, and Y. Matsumoto, Chem. Phys. Lett. 383,

261 (2004).34A. Y. Kobitski, R. Scholz, D. R. T. Zahn, and H. P. Wagner, Phys. Rev. B

68, 155201 (2003).35A. Damm, K. Schubert, J. Güdde, and U. Höfer, Phys. Rev. B 80, 205425

(2009).36L. Chkoda, M. Schneider, V. Shklover, L. Kilian, M. Sokolowski, C. Heske,

and E. Umbach, Chem. Phys. Lett. 371, 548 (2003).37V. R. Gangilenka, L. V. Titova, L. M. Smith, H. P. Wagner, L. A. A. De-

Silva, L. Gisslen, and R. Scholz, Phys. Rev. B 81, 155208 (2010).38E. N. Sickafus, Phys. Rev. B 16, 1448 (1977).

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions

124701-8 Marks et al. J. Chem. Phys. 139, 124701 (2013)

39A. O. Morozov, T. U. Kampen, and D. R. T. Zahn, Surf. Sci. 446, 193(2000).

40N. B. Gerasimova, A. S. Komolov, Y. G. Aliaev, and A. G. Sidorenko,Phys. Low-Dimens. Struct. 1–2, 119 (2001).

41J. Taborski, P. Vaterlein, H. Dietz, U. Zimmermann, and E. Umbach,J. Electron Spectrosc. Relat. Phenom. 75, 129 (1995).

42V. Bulovic, P. E. Burrows, S. R. Forrest, J. A. Cronin, and M. E. Thompson,Chem. Phys. 210, 1 (1996).

43S. Kera, H. Setoyama, M. Onoue, K. K. Okudaira, Y. Harada, and N. Ueno,Phys. Rev. B 63, 115204 (2001).

44R. Arafune, K. Hayashi, S. Ueda, and S. Ushioda, Phys. Rev. Lett. 92,247601 (2004).

45A. Y. Kobitski, R. Scholz, I. Vragovic, H. P. Wagner, and D. R. T. Zahn,Phys. Rev. B 66, 153204 (2002).

46H. P. Wagner, A. DeSilva, and T. U. Kampen, Phys. Rev. B 70, 235201(2004).

47E. Engel, M. Koschorreck, K. Leo, and M. Hoffmann, Phys. Rev. Lett. 95,157403 (2005).

48L. Gisslén and R. Scholz, Phys. Rev. B 80, 115309 (2009).49W. Gebauer, A. Langner, M. Schneider, M. Sokolowski, and E. Umbach,

Phys. Rev. B 69, 155431 (2004).50M. Friedrich, T. Wagner, G. Salvan, S. Park, T. U. Kampen, and D. R. T.

Zahn, Appl. Phys. A 75, 501 (2002).51E. Engel, K. Leo, and M. Hoffmann, Chem. Phys. 325, 170 (2006).

Downloaded 24 Sep 2013 to 128.59.247.207. This article is copyrighted as indicated in the abstract. Reuse of AIP content is subject to the terms at: http://jcp.aip.org/about/rights_and_permissions