14593
2007, 111, 14593-14596 Published on Web 09/19/2007
Optical Differential Reflectance Spectroscopy of Molecular Thin Films on a Metal: Evidence for Strong Oscillator Strength Increase Thomas Dienel, Roman Forker, Karl Leo, and Torsten Fritz* Institut fu¨r Angewandte Photophysik, Technische UniVersita¨t Dresden, 01062 Dresden, Germany ReceiVed: July 2, 2007; In Final Form: August 15, 2007
We investigate the thickness-dependent dielectric function of ultrathin titanyl phthalocyanine (TiOPc) layers on Au(111) by in situ spectroscopy during film growth. The absorption features reveal a complex structural development of the film, described by the observation of different crystal phases. The experiment allows addressing the thickness dependence of the oscillator strength in ultrathin films for the first time. Surprisingly, the oscillator strength increases with thickness by more than a factor of 3 and does not saturate until a thickness of 20 monolayers is reached.
The optical properties of thin organic films have been investigated intensively in the last few decades because (i), the transition from the well-understood optical properties of single molecules to solid films contains many interesting aspects, and (ii), they are the basis for many novel devices.1-4 One of the properties often utilized is the distinct optical absorption behavior, being adjustable not only by synthetic variations but also by morphology. The intensity of electronic transitions (including all associated vibronic progressions) for molecules in solution is described by the oscillator strength (OS), which is more or less consistently defined in the literature.5-7 Assuming that the interaction of the dissolved molecule with the solvent is negligible and no aggregates are formed, one attributes the obtained OS to single molecules. The OS of one absorption band of a molecular film is defined by
OS )
∫absorption bandε′′(E)E dE
(1)
where ε′′(E) is the imaginary part of the photon energydependent (wavelength-dependent) dielectric function εˆ (E) ) ε′(E) - iε′′(E) of the molecular film. This definition is based on the “f-sum rules”, which can be found in literature,8,9 where we regard here only one absorption band. However, it is not possible to relate the OS in solution directly to that of a molecular solid: Contrary to a solution, the excitonic interactions in a film are in almost all cases not negligible, causing differences in the optical transitions compared to single molecules. It should be emphasized that the quantity OS as defined by eq 1 is expected to be independent of the layer thickness, if one assumes equivalent interactions for each molecule. In other words, if the OS is not constant then it tells us that the intermolecular interactions vary in different thickness ranges. It is essential to have access to the thickness-dependent dielectric function of a molecular film in order to test this expectation. In this letter, we present the first experimental * Corresponding author. Fax: +49 351 463 37065. E-mail: torsten.fritz@ iapp.de.
10.1021/jp075128i CCC: $37.00
Figure 1. (a) Space-filling model of TiOPc. (b) Normalized absorbance spectra, representing the lowest energetic transition (Q band), for TiOPc dissolved in dichloromethane, and 100-nm-thick films (adopted from literature) with phase I12 and phase II15 morphology, respectively.
investigation of the thickness dependence of the OS of an ultrathin organic film on a metal substrate. Determining the OS in the thickness range from submonolayers to 20 monolayers (ML) gives insight into the development of film morphology (phases) and the accompanying properties (structure-property relation). As a model system, we choose titanyl phthalocyanine (TiOPc), whose molecular structure (cf. Figure 1) induces a variety of crystal phases.10-12 This representative of pyramidal phthalocyanines is of considerable technical interest because the near-infrared absorption can, for example, be utilized in organic solar cells. The growth of a particular phase can be promoted, for instance, by the substrate (here Au(111)) and its temperature,13,14 and the phases can be identified by their distinct absorption spectra. We demonstrate in this letter that the oscillator strength exhibits an unexpected dependence on the film thickness. Although we present some possible scenarios to explain this effect, we hope that our work will stimulate further theoretical and experimental work to better understand these open questions. Samples were prepared by organic molecular beam epitaxy (OMBE) in ultrahigh vacuum. The TiOPc molecules (purchased from Syntec Wolfen, two cycles of gradient sublimation, degassed for several hours in vacuum) were evaporated from a Knudsen cell at a temperature of 400 °C. The reconstructed © 2007 American Chemical Society
14594 J. Phys. Chem. C, Vol. 111, No. 40, 2007
Letters
Figure 2. Series of thickness-dependent differential reflectance spectra (DRS) of the TiOPc thin film growing on a gold single crystal held at 200 °C.
Au(111) was heated to 200 °C, to promote crystalline growth of the molecular film (Au(111) prepared by several cycles of 600 eV Ar+ sputtering and annealing). The film thickness was monitored by a quartz microbalance, which was calibrated by scanning tunneling microscopy (STM) on 1-2-ML-thick films, deposited under equal conditions. With a home-built setup16,17 we measured the reflectance (illumination at nearly normal incidence) of the sample prior and during the growth of the TiOPc film and calculated the differential reflectance signal (DRS) by using the following equation:
R(E,d) - R0(E) ∆R (E,d) ) R R0(E)
(2)
The energy-dependent reflectance of the bare substrate is named R0(E), and R(E,d) is the reflectance of the substrate covered with a molecular film of thickness d during the deposition process. Film thicknesses are given as nominal monolayer equivalents. The thickness of one monolayer (1 ML ≈ 0.33 nm) refers to densely packed flat-lying TiOPc molecules. According to eq 2, the DRS presented in Figure 2 are influenced by both the underlying substrate and the adsorbate layer on top of it. The definition of the OS (eq 1) requires the knowledge of ε′′(E) for one complete absorption band. Therefore, one has to extract the complex dielectric function of the adsorbate layer first, which can be done in two independent ways: (i) by a model-free Kramers-Kronig-consistent numerical algorithm,18,19 or (ii) via the McIntyre approximation.20 The latter case uses the linearizability of the Fresnel equations of the two-interface system (ambient-film-substrate) for film thicknesses much smaller than the wavelength of light (d , λ), and for semi-infinite substrates, at normal incidence of light (nearly fulfilled here).17,21 Using the dielectric function of the substrate gold,22 one can show that the approximation is valid in the energy range of interest. Although both methods reveal the same spectral behavior of the dielectric function εˆ (E) for thin films, there are apparent deviations for thicker films, where the approximation becomes less appropriate. Therefore, in the following, we use the Kramers-Kronig-consistent series exclusively. Furthermore, we restrict ourselves to the discussion of ε′′(E) because this entity (and not ε′(E), which is simultaneously extracted from the DRS, cf. Supporting Information) characterizes the absorbance.19 The
Figure 3. Thickness-dependent imaginary parts of the dielectric functions of TiOPc films calculated from the DRS in Figure 2 by a Kramers-Kronig-consistent algorithm. Spectra in a cover the thickness range between 5.5 and 18.5 ML. The spectrum corresponding to 0.7 ML is shown for comparison. In part b, the thin-film range up to 5.5 ML is displayed and illustrates the changes from the broad metalinfluenced peak at around 1.68 eV for submonolayer coverage toward the crystal structure related spectral features at 1.55 and 1.64 eV for the 5.5 ML film. The isosbestic points of the spectra are marked by circles (see the text for details).
ε′′(E) spectra, extracted from DRS, are shown in Figure 3, being divided into two thickness ranges for clarity. The first two monolayers of TiOPc on Au(111) (0.7 to 2.2 ML in Figure 3b) exhibit only one broad feature at around 1.68 eV, which differs significantly from the single-molecule behavior, measured in solution (cf. Figure 1). This can only be understood by the interaction between metal and molecules, in other words as the formation of hybrid states. The effect has been observed in literature for other molecules before: In electron-energy-loss spectra of thin pyrazin layers on silver,23 for example, the first monolayer exhibits one broad feature, which sharpens with increasing thickness while developing a substructure. For C60 films on gold, the broadening of the lowbinding-energy features in the UPS spectra was attributed to hybridization between the Au 6sp-band of the substrate with the π and π* orbitals of the molecules in the first monolayer.24 In optical spectroscopy of perylene-3,4,9,10-tetracarboxylicdianhydride (PTCDA) submonolayer films, the vibronic substructure known from solution spectra is only visible on insulating substrates (mica)16 and is absent on metals (gold).17 These three examples have in common that the hybridization extends only over a single monolayer. Surprisingly, for TiOPc layers the broadening expands over two layers. We know from STM measurements that the TiOPc molecules exhibit highly ordered films on gold (square structure of flat lying molecules, cf. Supporting Information).25 The head-to-tail arrangement (oxygen atoms in the first two layers point toward each other)
Letters causes a reduced distance between the phthalocyanine rings and the substrate (even for the second layer), which enables the strong interaction. The red-shifted peak position (1.80 eV for TiOPc dissolved in dichloromethane) is affected by the high polarizability of the metal.26 With further deposition of TiOPc, the diminishing influence of the substrate results in the development of a spectral substructure, visible at 2.9 ML film thickness, and more pronounced at 3.6 ML. The peak at 1.62 eV can be assigned to a molecular arrangement in polymorph phase I.12 This spectral development leads to the observation of an isosbestic point at 1.72 eV for films thicker than 2.9 ML (circle in Figure 3b). Isosbestic points are intersections of optical spectra, where a physical (or chemical) reaction from one photoactive species (here: TiOPc molecules on the substrate) to exact another (here: phase I crystallites) takes place, with no photoactive side products. It is interesting that between 1 and 3 ML no isosbestic point occurs, hinting again toward a strong molecule-substrate interaction, discussed in the previous paragraph. In films thicker than 5.5 ML, the energetically lower shoulder at 1.55 eV becomes a pronounced peak and undergoes a redshift in the subsequent spectra. The origin of the latter effect might be the increasing exciton delocalization with thicker films or the polarizability of the surrounding of each molecule, which also changes with the film growth. The position of the peak at around 1.5 eV points toward the phase II polymorph of the TiOPc crystals.15,27 The growth of this additional phase with partial rearrangement of already deposited molecules in phase I orientation leads to the development of an additional isosbestic point at 1.75 eV (film thickness above 6 ML, circle in Figure 3a). Instead of the expected broad shoulder toward higher energies observed for some thick film spectra containing phase II crystallites,27 our spectra exhibit a constant decrease in this energy range (the 18.5 ML spectrum is dominated by two peaks (1.50 and 1.57 eV) and a small feature at 1.80 eV). However, phase II films on polyimide (100 nm film with upright standing molecules)15 mainly exhibit a steady decrease in polarized absorption spectra, starting from the peak toward higher energies.13,28 Therefore, we conclude that we indeed observe the growth of phase II. The surprising finding in our study is the steady and strong increase in magnitude of the ε′′(E) spectra. With access to the ε′′ spectra for a multitude of layer thicknesses, we can address the issue of the thickness dependence of the OS in thin molecular crystals for the first time in detail. To calculate the OS, we extended the integral in eq 1 from 1.3 to 2.1 eV, which completely covers the band of the energetically lowest transition of the two crystal phases. The OS displayed in Figure 4 reflects the steady and strong increase observed already in the ε′′(E) spectra. Astonishingly, the values rise up to 20 ML and it is only around this thickness that saturation seems to occur. Because our setup probes transition dipoles lying parallel to the substrate with highest sensitivity (illumination at nearly normal incidence), maximum OS is expected to occur for the thinnest films, where flat lying growth of TiOPc molecules on gold is reported (cf. Supporting Information).25,29,30 Consequently, possibly occurring different orientations of the highly anisotropic crystals in thicker films; for example, phthalocyanine planes standing orthogonal to the substrate, would lead to lower values for the OS measured with this setup because only the transition dipole component parallel to substrate would be probed. To exclude that our result for the OS might be caused by artifacts of the numerical algorithm applied, we also present
J. Phys. Chem. C, Vol. 111, No. 40, 2007 14595
Figure 4. Thickness-dependent oscillator strength (OS) for a TiOPc film evaporated on single-crystalline gold held at 200 °C. The two different slopes of the OS for film thicknesses below and above 5 ML, respectively, are clearly visible. The OS saturates at a layer thickness of around 20 ML. For comparison, the values obtained by integrating ′′ from the McIntyre approximation are shown as well (see the text for details).
the OS resulting from the ε′′ generated with the McIntyre approximation being integrated in the same energy interval. Indeed, one gets an absolutely similar trend, with a deviation for thicker films, where the approximation becomes less sufficient. The OS changes by a factor of more than three in the entire range shown in Figure 4. The thickness dependence of the OS in such a broad thickness range was neither observed for the quasi-one-dimensional growth of PTCDA17 nor for HBC19 on the respective substrates. The OS for PTCDA saturates already between 3- and 4-MLthick films on a transparent substrate, and on Au(111) as well.18 The variations lie within only 10%, besides that this was expected from exciton theories. Although we are currently not able to deliver a final explanation for the unexpected behavior of the OS of TiOPc films, one can still discuss some possible scenarios: (i) The interaction with the substrate gold can lead to new transitions for the first few monolayers, which means that OS from the absorption band observed in this study is transferred out of the measured range. Thus, the OS would approach the molecule’s intrinsic value as the distance from the substrate increases. (ii) From the observed diversification of the ε′′(E) spectra (film thickness above 4 ML), we know that TiOPc molecules arrange in phase II. This structure is known to exhibit a closer packing31 as compared to phase I, which facilitates stronger excitonic interactions (which do extremely nonlinearly depend on the intermolecular separation). This might be the reason for the two different slopessresembled by both the Kramers-Kronig-consistent algorithm and the McIntyre approximationsin Figure 4, where phase I dominates below 5 ML and phase II for thicker films. The impact of the packing differences on the electronic absorption spectra was investigated by Nakai et al. using theoretical calculations to clarify the large redshift of the Q band and the high photoconductivity of phase II.32 They treated the intermolecular interactions in the form of configuration interactions between excitons and charge-transfer (CT) configurations. Because of the rapid decrease with distance of the resonance integrals, only CT configurations of neighboring TiOPc molecules were considered there. The largest value of the resonance integrals (and OS) was observed in phase II, where two
14596 J. Phys. Chem. C, Vol. 111, No. 40, 2007 molecules can be found with the outer benzene units of one molecule lying on top of those of another. All calculated dimers in phase I showed smaller resonance integrals, demonstrating that the properties correlate with the conformation of the molecules in the crystal phases. Integration of their spectra reveals that the OS for phase II is 2 times higher than that for phase I. Consequently, the additional growth of phase II crystallites might be one reason for the increasing OS observed here. In conclusion, we have experimentally addressed the issue of the thickness dependence of the oscillator strength in solid films for a broad thickness range. Instead of the expected constant value or a fast saturation within few layers, we observe an increase by a factor of more than three of the OS in a thickness range up to 20 ML. The question why this was not observed in systems of planar hydrocarbons (for example: PTCDA, HBC), where a similar interaction between stacked molecules should occur, remains open. Still, we anticipate our findings to stimulate further theoretical and experimental studies, facing the influence of the metal ligands on the optical properties. Acknowledgment. Financial support by the “Deutsche Forschungsgemeinschaft” (DFG) grants FR875/6-1, FR875/62, and FR875/9-1 is gratefully acknowledged. We thank K. Nakai and N. Kobayashi for the original data of ref 32, M. Koch for material purification, and K. Walzer for fruitful discussions. Supporting Information Available: Real part of the dielectric function accompanying imaginary part in Figure 3; STM image of a nearly complete second monolayer of TiOPc on Au(111), revealing three-dimensional molecular arrangement. This material is available free of charge via the Internet at http:// pubs.acs.org. References and Notes (1) Tang, C. W. Appl. Phys. Lett. 1986, 48, 183. (2) Tang, C. W.; VanSlyke, S. A. Appl. Phys. Lett. 1987, 51, 913. (3) Tada, H.; Touda, H.; Takada, M.; Matsushige, K. Appl. Phys. Lett. 2000, 76, 873.
Letters (4) Bouvet, M. Anal. Bioanal. Chem. 2006, 384, 366. (5) Suzuki, H. Electronic Absorption Spectra and Geometry of Organic Molecules; Academic Press, New York, 1967. (6) Birks, J. B. Photophysics of Aromatic Molecules; Wiley-Interscience, London, 1970. (7) Pope, M.; Swenberg, C. E. Electronic Processes in Organic Crystals; Oxford University Press, New York, 1982. (8) Altarelli, M.; Dexter, D. L.; Nussenzveig, H. M.; Smith, D. Y. Phys. ReV. B 1972, 6, 4502. (9) Stroud, D. Phys. ReV. B 1979, 19, 1783. (10) Oka, K.; Okada, O.; Nukada, K. Jpn. J. Appl. Phys. 1992, 31, 2181. (11) Okada, O.; Oka, K.; Iijima, M. Jpn. J. Appl. Phys. 1993, 32, 3556. (12) Mizuguchi, J.; Rihs, G.; Karfunkel, H. R. J. Phys. Chem. 1995, 99, 16217. (13) Yonehara, H.; Ogawa, K.; Etori, H.; Pac, C. Langmuir 2002, 18, 7557. (14) Walzer, K.; Toccoli, T.; Wagner, C.; Pallaoro, P.; Iannotta, S.; Fritz, T.; Leo, K. Surf. Sci. 2006, 600, 2064. (15) Yonehara, H.; Etori, H.; Engel, M. K.; Tsushima, M.; Ikeda, N.; Ohno, T.; Pac, C. Chem. Mater. 2001, 13, 1015. (16) Proehl, H.; Dienel, T.; Nitsche, R.; Fritz, T. Phys. ReV. Lett. 2004, 93, 097403. (17) Proehl, H.; Nitsche, R.; Dienel, T.; Fritz, T. Phys. ReV. B 2005, 71, 165207. (18) Nitsche, R.; Fritz, T. Phys. ReV. B 2004, 70, 195432. (19) Forker, R.; Dienel, T.; Fritz, T.; Mu¨llen, K. Phys. ReV. B 2006, 74, 165410. (20) McIntyre, J. D. E.; Aspnes, D. E. Surf. Sci. 1971, 24, 417. (21) Selci, S.; Chiarotti, G.; Chiarada, P.; Cricenti, A. J. Vac. Sci. Technol., A 1987, 5, 327. (22) Olson, C.; Lynch, D. W.; Weaver, J. H. Optical Constants of Gold; In Handbook of Optical Materials; Weber, M. J., Ed.; CRC Press: Boca Raton, 2003. (23) Demuth, J. E.; Avouris, P. Phys. ReV. Lett. 1981, 47, 61. (24) Veenstra, S. C.; Heeres, A.; Hadziioannou, G.; Sawatzky, G. A.; Jonkman, H. T. Appl. Phys. A 2002, 75, 661. (25) Mannsfeld, S. C. B.; Fritz, T. Phys. ReV. B 2005, 71, 235405. (26) Reichardt, C. SolVents and SolVent Effects in Organic Chemistry; VCH, Weinheim, 1990. (27) Saito, T.; Iwakabe, Y.; Kobayashi, T.; Suzuki, S.; Iwayanagi, T. J. Phys. Chem. 1994, 98, 2726. (28) Brinkmann, M.; Wittmann, J.-C.; Barthel, M.; Hanack, M.; Chaumont, C. Chem. Mater. 2002, 14, 904. (29) Mannsfeld, S. C. B.; Fritz, T. Mod. Phys. Lett. B 2006, 20, 585. (30) Mannsfeld, S. C. B.; Ph.D. thesis, Technische Universita¨t Dresden, 2004. (31) Hiller, W.; Stra¨hle, J.; Kobel, W.; Hanack, M. Z. Kristallogr. 1982, 159, 173. (32) Nakai, K.; Ishii, K.; Kobayashi, N.; Yonehara, H.; Pac, C. J. Phys. Chem. B 2003, 107, 9749.