Organic Electronics 8 (2007) 63–68

www.elsevier.com/locate/orgel

The electronic band alignment on nanoscopicallypatterned substrates

G. Koller *, B. Winter, M. Oehzelt, J. Ivanco, F.P. Netzer, M.G. Ramsey

Institut fur Physik, Karl-Franzens-Universitat, Universitatsplatz 5, A-8010 Graz, Austria

Received 9 June 2006; received in revised form 6 November 2006; accepted 7 November 2006Available online 8 December 2006

Abstract

The electronic level alignment of the organic semiconductor para-sexiphenyl on a nanoscopically patterned substratewas investigated with ultraviolet photoemission spectroscopy, work function measurements and scanning tunnellingmicroscopy. The results show that for increasing coverage on inhomogeneous surfaces, shifts in electronic level alignmentoccur, which are due to the change from local to average band alignment. The Cu(110)–(2 · 1)O stripe phase, used as amodel substrate consists of alternating stripes of bare and oxygen passivated copper, with stripe widths of �35 A compa-rable to the sexiphenyl molecular length. In the first molecular layer the electronic bands are clearly aligned to the localsurface potential of the specific stripe, resulting in the superposition of two photoemission spectra offset by 1 eV. Beyondtwo monolayers the valence band spectra clearly indicate a single electronic level alignment, which is determined by theaverage interface dipole.� 2006 Elsevier B.V. All rights reserved.

PACS: 73.20.At; 73.63.Rt; 73.22.-f; 68.43.Bc

Keywords: Band alignment; Photoemission; Work function; Sexiphenyl; Copper; Nanostructures

1. Introduction

The worldwide interest in organic semiconduct-ing materials is driven by the wealth of their possibletechnological applications. At present light emittingand simple electronic devices are on the threshold ofbeing commercially realised, while new applications,such as solar cells and chemical sensors are beingexplored. Historically, the field of active molecularsystems has been evolving from ‘‘plastic electronics’’

1566-1199/$ - see front matter � 2006 Elsevier B.V. All rights reserved

doi:10.1016/j.orgel.2006.11.001

* Corresponding author. Fax: +43 316 380 9816.E-mail address: georg.koller@uni-graz.at (G. Koller).

through ‘‘organic electronics’’ to ‘‘molecular elec-tronics’’; that is polymers, better defined as oligo-meric films, to the holy grail of single moleculedevices. Clearly, an understanding of the electroniclevel alignment on the molecular scale is vital formolecular electronics, but it can also be importantin present day devices as their contact interfacesare unlikely to be uniform on this scale, an intricacy,which may lead to charge trapping at the interfaceand reduced device performance. The area of suchdefects can also be on the nanoscale and thusin the order of the here investigated substratenanostructures.

.

64 G. Koller et al. / Organic Electronics 8 (2007) 63–68

One important parameter in organic device per-formance is the barrier to charge carrier injection,that is the electronic level alignment between thehighest occupied (HOMO) and the lowest unoccu-pied molecular orbital (LUMO) with respect to theFermi level of the inorganic/metal contact. In ourprevious work, we investigated this so called molec-ular band alignment on a variety of chemically differ-ent substrates with several highly defined oligomers[1,2]. It is now generally accepted that the electronicalignment of the organic film is determined by thedetails of the interaction of the first layer of mole-cules with the contact material and the interfacedipole thus formed [1–6]. In this contribution, weexplore the local variation in band alignment on ananoscopically patterned substrate surface by inves-tigating the growth of para-sexiphenyl (p-6P) films,from sub-monolayer to multilayer coverages, withUV-photoemission (UPS), work function (/) mea-surements and supported by scanning tunnellingmicroscopy (STM). The substrate used was the mes-oscopically patterned oxygen reconstructed Cu(110)surface. This surface reconstructs to form a substrateconsisting of alternating stripes of clean Cu(110) and(2 · 1) oxygen reconstructed Cu(110), the so calledCu–CuO stripe phase, with respective stripe widthsin the range of several nanometers depending onthe amount of oxygen exposure [7,8]. In this study,we have used an oxygen coverage of 0.25 monolayers(ML), with an average stripe width of 30–40 A each,as confirmed by STM. The development of the elec-tronic level alignment on this patterned substrate iscompared to that on the pure homogenous surfaces:clean copper and the fully reconstructed Cu–(2 · 1)Osurface. On the homogenous surface there is a 1 eVdifference in band alignment due to the differencesin the interaction of the molecules in their respectivemonolayers [9]. It will be shown, for sexiphenylgrown on the Cu(110)–(2 · 1)O stripe phase, thatthe photoemission spectra of the first molecularmonolayer is a superposition of two spectra offsetby this difference in band alignment, clearly reflectingthe local differences in the charge injection barrier. Incontrast, for the multilayer coverage regime, thevalence band spectra clearly indicate a single bandalignment, whose energy position is determined bythe average interface dipole.

2. Experimental

The experiments were performed in two ultra highvacuum (UHV) (base pressure <1 · 10�10 mbar)

chambers, equipped with an UV-photoemissionspectrometer and a scanning tunnelling microscope,described elsewhere [10,11]. In addition both cham-bers have basic sample cleaning and surface charac-terisation facilities. The Cu(110) surface was cleanedby the repeated cycles of Ar+-ion sputtering andannealing at 800 K. The full Cu(110)–p(2 · 1)Oreconstruction was made by exposing the clean cop-per surface to 40 L [1 Langmuir (L) = 1 · 10�6

Torr sec] of oxygen at 400 K. The Cu–CuO stripephase was produced by exposing the clean copperto 1.5 L oxygen at 600 K [7,8] and was characterisedby STM and work function measurements. Thework functions (/), measured by the secondaryelectron cut-off in normal electron emission, ofthe substrates were: stripe phase / = 4.6 eV, cleanCu(110) / = 4.4 eV and the fully (2 · 1)O recon-structed surface / = 4.75 eV. Sexiphenyl was evapo-rated from a well degassed molecular evaporator,which allowed film growth in <5 · 10�10 mbar ambi-ent pressure. The substrate was held at room temper-ature and nominal growth rates of 1 A min�1 wereused, as monitored with a quartz microbalance. Inthe STM system the clean Cu surface was seen tobe completely covered by uniaxially lying moleculesat �3 A exposure, while in the ARUPS chamber(due to a difference in evaporation geometry) thework function change is completed at nominal 5 Aexposure – these values are considered as the mono-layers (ML) in the following.

3. Results and discussion

A 100 A · 100 A STM image of the Cu(110)–(2 · 1)O striped surface is shown in the insert ofFig. 1, illustrating the alternating stripes of cleancopper and (2 · 1)O reconstructed copper areaswith respective diameters of 30–40 A. It should benoted that the copper atoms are highly mobile atroom temperature, causing the fuzzy appear of thecopper, while the Cu–(2 · 1)O stripes can be imagewith atomic resolution. Fig. 1 shows a 300 A ·300 A STM image of this surface after an exposureof 0.8 A of sexiphenyl, a coverage corresponding toabout 0.3 monolayers. At such low exposures themolecules, apparent as bright elongated objects of�25 A length in the STM, decorate the clean cop-per, going first to the Cu–CuO boundary and thencovering the clean Cu areas. The molecules arealigned to and are highly mobile along the [001]substrate azimuth. This mobility is evidenced bythe bright stripes that are considerably larger than

Fig. 1. 300 A · 300 A STM image of a Cu(110)–(2 · 1)O stripedsurface after deposition of 0.8 A of sexiphenyl. The insert(100 A · 100 A) shows the surface prior to molecular deposition.The [001] substrate azimuth and the Cu and Cu(2 · 1)O stripesare indicated in both images.

Fig. 2. (a) Normal emission He I UPS spectra from the Cu(110)–(2 · 1)O stripe phase as a function of sexiphenyl exposure.Photon angle of incidence a = 60�. The binding energies of bandC of sexiphenyl monolayers and multilayers obtained on clean Cu(110) and from the fully reconstructed Cu (110–p(2 · 1)Osurface are marked by the dotted and dashed lines, respectively.(b) Development of the work function as function of sexiphenylcoverage on clean Cu(110) (squares), the fully reconstructedCu(110)–p(2 · 1)O (dots) and the Cu (110) (2 · 1)O stripe phase(diamonds) surfaces.

G. Koller et al. / Organic Electronics 8 (2007) 63–68 65

the molecules (25 A) which, once immobilised indense structures, can be imaged as individual enti-ties (see for instance the molecules at the top righthand corner). Depending on the stripe width 3–5molecules, aligned parallel to the stripes, fill onestripe, showing that the underlying surface patterntruly is of molecular dimensions.

Fig. 2a shows a series of UPS spectra of sexiphe-nyl on a Cu (11 0) (2 · 1)O stripe surface as a func-tion of exposure at room temperature. The firstmonolayer is completed for exposures around 5 Aas determined from the saturation value of the workfunction. The work function of this monolayer (/ =3.8 eV), is approximately the 1:1 area averagebetween the sexiphenyl monolayer work functionson clean Cu (110) (/ = 3.4 eV) and on the fully(2 · 1)O reconstructed Cu (110) surface (/ =4.3 eV), as shown in Fig. 2b. Since the measuredwork function is a simple area average of the localwork functions [12], we can use them as a simplemeans to determine the ratio between the copperand Cu(2 · 1)O reconstructed surface areas. Itshould be noted that the secondary electron cut-offs,used to determine the work function, of all surfaceswere sharp and did not show a double cut-off thatwould reflect the two different local surface poten-tials. To follow the development of the band align-

Fig. 3. Normal emission He I UPS spectra of sexiphenylmonolayers on (a) clean Cu (110) (/ = 3.4 eV), (b) on the fullyreconstructed Cu (110) – p(2 · 1)O (/ = 4.3 eV) surface and (c)on the Cu (110) (2 · 1)O stripe phase (/ = 3.8 eV). (d) Weighted(1:1) sum of (a) and (b). Particular similarities between (c) and (d)of bands A and C are highlighted by boxes, the arrows indicatethe position of band A on the homogeneous surfaces.

66 G. Koller et al. / Organic Electronics 8 (2007) 63–68

ment of sexiphenyl on the Cu–CuO stripe phase as afunction of film thickness (Fig. 2a), we concentrateon two sexiphenyl bands: band C, which is at highbinding energy (9–10 eV) and does not overlap withthe copper d-band of the substrate (bottom spec-trum of Fig. 2a), and the non bonding p orbitalsof band A, which although partially overlappingwith the copper d-band, is intense enough to beidentified. In addition the binding energies of thesebands for the monolayers (dotted lines) and multi-layers (dashed lines) observed on the clean and fully(2 · 1)O reconstructed copper surfaces are also indi-cated in Fig. 2a. In the UPS spectra of Fig. 2a wecan clearly discern two different regimes: For lowcoverages of up to 9 A, the molecular orbitals ofthe UPS spectra appear broad and indistinct. Thisis particularly apparent for band B, which is com-pletely washed out and the non bonding orbitalsof band A, which appear as an indistinct broad step.For coverages beyond 9 A the molecular featuresbecome well defined and distinct and they are ener-getically positioned between the multilayer spectraon clean copper and that on the fully (2 · 1)Oreconstructed copper (as indicated). It should benoted that after the monolayer is completed, islandgrowth occurs (Stranski–Krastanov growth), as evi-denced by the Fermi-edge being visible even at acoverage of more than ten monolayers (50 A). Forthe coverages up to 6 A the measured UPS spectraare a superposition of two spectra, indicating thatwe observe the two band alignments of p-6P onclean copper and p-6P on Cu(2 · 1)O. This is illus-trated in Fig. 3, where the spectrum of a monolayercoverage on clean Cu(110) and on the fully recon-structed Cu–(2 · 1)O surface is compared to a spec-trum of a monolayer on the heterogenous stripedsurface. As indicated in the figure all molecularorbital emissions of the homogenous surfaces areoffset by 0.9 eV from each other (equal to the differ-ence in /) due to the differences in the interfacedipole. Adding these two spectra results in a com-posite spectrum (Fig. 3d), which is remarkably sim-ilar to the spectrum recorded for the monolayer onthe heterogenous Cu–(2 · 1)O stripe phase (Fig. 3c).This suggests that we do observe the molecular bandemissions from the two clearly distinct band align-ments and that the sexiphenyl molecules adsorbedon the bare copper stripes feel the local surfacepotential of the clean copper, while the sexiphenylmolecules on the oxygen reconstructed stripes expe-rience the local surface potential of the Cu(2 · 1)Osurface. In many respects the molecular emissions

behave as a local work function probe, similar tothe photoemissions of adsorbed Xenon (PAX) tech-nique [12].

For coverages beyond � three monolayers on thestriped substrate the molecular orbital featuresbecome more distinct until the UPS spectrum resem-bles the usual spectrum of condensed sexiphenylfilms as shown in Fig. 4. The binding energy of themolecular orbitals of the multilayers on the stripedsubstrate are situated between those obtained onthe homogenous substrates, suggesting that we havea single band alignment with an average offset. InFig. 4 the bottom two spectra show a nominal10 ML (50 A) thick sexiphenyl film on clean copperand on Cu(2 · 1)O surfaces. It should be noted thatthe first peak at 2.4 eV binding energy on the cleanCu substrate (shaded in the figure) is not the p-6P

Fig. 4. Normal emission He I UPS spectra of sexiphenylmultilayers on (a) clean Cu (110) (/ = 3.4 eV), (b) on the fullyreconstructed Cu (110) – p(2 · 1)O surface (/ = 4.3 eV) and (c)on the Cu (110) (2 · 1)O stripe phase (/ = 3.8 eV). (d) Weighted(1 : 1) sum of (a) and (b). The major differences between (c) and(d) are marked by arrows.

G. Koller et al. / Organic Electronics 8 (2007) 63–68 67

HOMO (which is at 3 eV), but rather stems from theintense copper d-band feature at this energy, whichdue to the severe islanding on the clean coppersubstrate is still visible. Comparing the multilayerfilm spectrum on the striped Cu–CuO substrateto the weighted sum of spectra for multilayers onthe homogeneous substrates (cf. 4c and d) showsthat they are markedly different and, unlike for themonolayer, the spectra from the multilayer on thestriped substrate cannot be due to a superpositionof emissions offset by their respective interfacedipoles. In particular the orbital emissions of thespectral sum are much broader than the measuredspectrum, but the most prominent differences occurin the p-band and band B, as indicated by arrowsin the figure. This shows that the sexiphenylmolecules of the higher layers do not feel the localsurface potential of the Cu and CuO stripes any

longer, but rather are aligned to the average inter-face potential. Further support for this is given bythe fact that all molecular valence band orbitals onthe three surfaces shown in Fig. 4, viz. clean Cu(110), Cu (110)–p(2 · 1)O, and the Cu–(2 · 1)Ostripe phase, have the same binding energy withrespect to the vacuum level (ionisation potentialIP) of the respective multilayers (e.g. for the HOMOIP = 6.5 ± 0.1 eV, for the non-bonding p-orbitals ofband A IP = 8.3 ± 0.1 eV).

The key question is at what layer height the par-adigm shift from band alignment to the local inter-face potential as opposed to band alignment to theaverage interface potential occurs. From the devel-opment of the molecular orbitals as a function ofp-6P coverage shown in Fig. 2a it is suggested thatthis transition occurs around a coverage of 9 A orabout 2 ML. Here the weight of band C has alreadyshifted to its multilayer value and band B hasbecome apparent, although the non bonding orbi-tals of the p-band A are still broad and indistinct.When taking into account the islanding of thesecond and higher layers, we expect that a largeproportion of the first monolayer is exposed at thiscoverage. For the area averaging technique of UPSwe therefore suggest a superposition of three sexi-phenyl spectra: p-6P aligned to the local interfacepotential of Cu, p-6P aligned to the local interfacepotential of Cu–(2 · 1)O and p-6P aligned to theaverage interface potential. This leads us to con-clude that this paradigm shift occurs already inthe second molecular layer as illustrated in the sche-matic illustartion of Fig. 5.

4. Summary

Here we have investigated the local band align-ment of the organic semiconductor sexiphenylon the nanopatterned Cu(11 0)–(2 · 1)O stripedsubstrate from submonolayer coverages to filmthicknesses approaching that of organic electronicdevices. The results are summarized in Fig. 5: Forthe first molecular layer, the electronic bands areclearly aligned to the local surface potential of thespecific stripe, resulting in the superposition ofUPS spectra. Beyond two monolayers the valenceband spectra clearly indicate a single electronic levelalignment, which is determined by the average inter-face dipole.

The implications of the results on a mundane levelare that care must be taken in interpreting orbitalshifts with increasing exposure on ill-defined/

Fig. 5. Schematic illustrating of the observed change from localto average band alignment on the nanopatterned substrate. Theshaded discs represent sexiphenyl molecules, with the grey scaleindicating the respective band alignment. The curved arrowsemphasise the local differences in the barrier height for injectionto the first monolayer.

68 G. Koller et al. / Organic Electronics 8 (2007) 63–68

inhomogeneous substrates, such as clean polycrys-talline surfaces or even contaminated substrates(consider different crystalline facets can have 0.5 eVdifference in work function). The results further

imply that future nanopatterned devices will ulti-mately require thin active organic layers only 1–2molecules thick. On a more exalted level the capacityto pattern the local surface dipole allows the tailor-ing of the charge injection ability on the molecularlevel, a prerequisite for the advent of molecularelectronics.

Acknowledgement

This work was supported by the Austrian ScienceFoundation FWF.

References

[1] G. Koller, R.I.R. Blyth, S.A. Sardar, F.P. Netzer, M.G.Ramsey, Appl. Phys. Lett. 76 (2000) 927.

[2] R.I.R. Blyth, R. Duschek, G. Koller, F.P. Netzer, M.G.Ramsey, J. Appl. Phys. 90 (2001) 270.

[3] I.G. Hill, D. Millron, J. Schwartz, A. Kahn, Appl. Surf.Science 166 (2000) 354.

[4] H. Ishii, K. Sugiyama, E. Ito, K. Seki, Adv. Mat. 11 (1999)605.

[5] I.G. Hill, A. Rajagopal, A. Kahn, Y. Hu, Appl. Phys. Lett.73 (1998) 662–664.

[6] G. Koller, F.P. Netzer, M.G. Ramsey, Appl. Phys. Lett. 83(2003) 563.

[7] K. Kern, H. Niehus, A. Schatz, P. Zeppenfeld, J. George, G.Comsa, Phys. Rev. Lett. 67 (1991) 855.

[8] P. Zeppenfeld, V. Diercks, C. Tolkes, R. David, M.A.Kryzowski, Appl. Surf. Science 59 (1998) 484.

[9] M. Oehzelt, L. Grill, S. Berkebile, G. Koller, F.P. Netzer,M.G. Ramsey, submitted for publication.

[10] G. Koller, R.I.R. Blyth, S.A. Sardar, F.P. Netzer, M.G.Ramsey, Surf. Science 536 (2003) 155.

[11] J. Kraft, M.G. Ramsey, F.P. Netzer, Phys. Rev. B 5384(1997).

[12] K. Wandelt, Appl. Surf. Science 111 (1997) 1.