J. Phys. Chem. B 2001, 105, 5689-5699
5689
Optical Properties of Crystalline Pseudoisocyanine (PIC) Hans von Berlepsch,*,† Sven Mo1 ller,‡ and Lars Da1 hne*,†,§ Institut fu¨ r Chemie-Physikalische und Theoretische Chemie, Freie UniVersita¨ t Berlin, Takustrasse 3, D-14195 Berlin, Germany, Fachbereich Physik der Philipps-UniVersita¨ t Marburg, Renthof 5, D-35032 Marburg, Germany, and Max-Planck-Institut fu¨ r Kolloid und Grenzfla¨ chenforschung, D-14424 Potsdam, Germany ReceiVed: December 31, 2000; In Final Form: March 23, 2001
Normal incidence reflectance spectra of different crystal faces of 1,1′-diethyl-2,2′-cyanine (pseudoisocyanine) bromide (PIC-Br) and chloride (PIC-Cl) single crystals have been analyzed by a polariton model. The spectra are anisotropic with wide bands of quasimetallic reflectance for two orthogonal polarizations of light due to directional dispersion of polariton resonances. The measured spectra were analyzed by Kramers-Kronig transformation and compared with theoretical spectra obtained by solving the wave equation. Because of the herringbone-like packing of PIC molecules in single crystal, the π f π* transition is split into two Davydov components with energy separation of about 30 meV. Most of the oscillator strength is concentrated in the fundamental exciton peaks, while a fine structure related to molecular vibrations is also present. Structural isomorphism of PIC-Br and PIC-Cl single crystals is reflected in identical reflection spectra. Glassy PIC-Br films of thicknesses between 5 and 150 nm were prepared by spin-coating. The films are built up by a network of randomly distributed J-aggregates. By treatment with humid air they become highly ordered, showing three differently colored crystalline domains in polarized light. The absorption spectra of the domains coincide with that of the (100), (-101), and (101) faces of single crystal. Surface force microscopy (SFM) reveals heterogeneous layers composed of aligned aggregates of about 0.5 µm width.
I. Introduction Cyanine dyes are known to form in concentrated aqueous solution tightly bound molecular assemblies, exhibiting a strong spectral shift of their absorption band toward longer wavelengths with respect to the monomer absorption. The assemblies have been named Scheibe or Jelly (J-) aggregates in honor of the two researchers who first discovered this phenomenon1,2 in aqueous solution of the dye 1,1′-diethyl-2,2′-cyanine chloride (pseudoisocyanine chloride, PIC-Cl):
Scheibe attributed the peculiar spectroscopic behavior to a reversible polymerization of the chromophores due to intermolecular interactions. The molecular exciton theory supplied the explanation for the observed absorption spectra.3 Despite a multitude of studies and data available on optical and spectroscopic properties of PIC and related dyes,4 which were obtained over the more than sixty years of research, the knowledge of the supramolecular structure of the aggregates is only fragmentary. Recently, we could visualize for the first time directly the rodlike morphology of the PIC-Cl J-aggregates in aqueous solution.5 Using cryo-transmission electron microscopy (cryoTEM) we estimated a rod diameter of 2.3 nm and proposed alternative structure models for the molecular packing within †
Institut fu¨r Chemie. Fachbereich Physik der Philipps-Universita¨t. § Max-Planck-Institut fu ¨ r Kolloid und Grenzfla¨chenforschung. ‡
the aggregate. In particular, a quasi-two-dimensional superstructure has been suggested, which could better explain the optical properties of the J-aggregates in solution than previous models.6 PIC forms J-aggregates not only in homogeneous solutions, but also at interfaces7,8 and in crystals.9-11 The respective absorption and fluorescence spectra are similar, which has been interpreted as the indication of a common entity that is responsible for the J-band. Kawasaki and Ishii12 as well as Owens and Smith13 have recently used scanning tunneling microscopy to investigate PIC J-aggregates adsorbed at surfaces. Monolayers with a brickwork packing structure of individual molecules,14 similarly well-ordered bilayers, or extended linear aggregates of a dimension corresponding to that of the rodlike aggregates in solution,5 were found, respectively. These structural studies demonstrate a surprising morphological variety even on microscopic scale and it is not clear until now, if there really exists a unique molecular structure that is responsible for the formation of the J-band. It is reported that the nature of the counterion in solutions or at a solid/solution interface and the solvent15-17 has a slight effect on the spectroscopic characteristics. Disorder plays an important role in molecular aggregates. Exchange narrowing of static disorder is generally believed to be responsible for the characteristic sharpness of the J-band in solutions or in a glass.18,19 In addition, for the dynamics of excitons in J-aggregates in solution, the dynamical nature of the environment becomes important and has to be taken into account in the theoretical description.20 The situation is much easier when single crystals of the dye can be grown. Then the optical properties are accessible basically by normal incidence reflectance measurements from different crystal faces, as it has
10.1021/jp004581q CCC: $20.00 © 2001 American Chemical Society Published on Web 05/24/2001
5690 J. Phys. Chem. B, Vol. 105, No. 24, 2001 been shown recently for a series of dye crystals21-26 including PIC.9-11 One should have in mind, however, that single crystal spectra will provide only some guidance on the packing geometry inside the J-aggregate in glassy host or in solution. Nevertheless, reflectivity measurements on PIC single crystals are of interest, because the published data are not in full agreement and a complete theoretical description of the optical properties does not exist until now. Because of the anisotropy of the reflectivity of dye crystals, spectra from different faces have to be taken, which requires single crystals with high surface quality. Other shortcomings of the former studies are connected with the theoretical modeling of measurements. The success in measurement and interpretation of reflectance spectra of different polymethin dyes25,26 gave us the motivation to reinvestigate the crystal spectra of PIC. The dense packing of strong transition dipoles of such dyes leads to stopping bands of high reflectance whose width varies on different crystal faces. The optical properties are determined by polariton resonances that can deviate strongly from the transition energies of molecular excitons due to the macroscopic polarization.22-27 PIC single crystals of optical quality with bromide or chloride as counterions were used for the present study. The X-ray structures are known,28-30 showing isomorphism. In addition, crystalline PIC-Br layers of 100 nm typical thickness with domain widths reaching hundreds of µm, grown on solid substrates by a selforganization process (thin layer crystallization, TLC31), were included in the study. Comparison with single-crystal spectra allowed to identify the structure of the different crystalline domains. This paper is organized as follows. Section II presents experimental details and methods. The results of structural characterization of single crystals are given and the preparation of thin layers is described. After the description of optical measurements the methods to simulate the reflectivity by a polariton model are briefly reviewed. Using these model calculations the reflectivity data are quantitatively analyzed in section III. The morphological structure of the thin layers is discussed in section IV. Finally, section V presents some concluding remarks. II. Experimental Section and Methods 1. Samples. Single Crystals. PIC-Br was a product (NK-1046) from Nippon Kankoh Shikiso Kenkyusho, Japan, and was used without further purification. PIC-Cl was obtained as a gift from AGFA AG and was used as received as well. Molar extinction coefficients of 7.71 × 104 and 7.67 × 104 L/mol cm were measured in methanol for PIC-Br and PIC-Cl, respectively. Taking into account the refractive index of the solvent, oscillator strengths of 1.06 were obtained by integrating the spectra.32 Single crystals of mm size were grown from solutions in dry methanol. The X-ray structures of PIC-Br and PIC-Cl are known28,29 but they were newly determined30 for the present crystals to allow for a definite face assignment. Both structures are isomorphous. PIC crystallizes in the monoclinic space group P21/n with the lattice parameters given in Table 1, and four molecules per unit cell. Note that we have used different unit cells for PIC-Br (notation after ref 29) and PIC-Cl (our own30) crystals. The PIC molecule is nonplanar, i.e., the two quinoline rings are twisted around the central methin group by an angle of 50.6° (PIC-Cl28). The π-electrons are delocalized along the methin chain. Because of the twisting different transition dipole moment vectors may be constructed, depending on which of the two polarization directions, i.e., either along the C2-C2′ or along the N1-N1′ direction, is taken. It may be expected that
von Berlepsch et al. TABLE 1: Crystal Structure Data of PIC-Br and PIC-Cl, Coordinates of Davydov Components m+ and m-, and Tilt Angles between the Molecular Dipoles τ PIC-Br a ) 13.558 Å b ) 10.588 Å c ) 14.095 Å N1-N1′: C2-C2′:
R ) 90.0° β ) 90.54° γ ) 90.0° m+ ) (0.6554, 0.0, 8.7949) m- ) (0.0, 3.8498, 0.0) τN1-N1′ ) 47.2° m+ ) (1.1719, 0.0, 4.6033) m- ) (0.0, 1.5607, 0.0) τC2-C2′ ) 36.4° PIC-Cl
a ) 13.706 Å b ) 10.495 Å c ) 13.584 Å N1-N1′:
R ) 90.0° β ) 90.75° γ ) 90.0° m+ ) (-8.7967, -0.0010, 0.7471) m- ) (0.0014, 4.0185, 0.0) τN1-N1′ ) 48.9°
the effective dipole direction will be lying between these two theoretical limits. The dipole moments are taken to be proportional to the vectors obtained as difference of the corresponding atom positions. The PIC crystal is built up of strands of stacked molecules. Adjacent molecules forming one strand are arranged along the stacking axis (c-axis for PIC-Br, a-axis for PIC-Cl) with tilt angle τ. Figure 1 shows two single strands with oppositely oriented molecules forming a double strand with herringbone-like architecture. Two of the four π f π* transition dipole moments in the unit cell are equivalent resulting in two orthogonal Davydov components m+ and m- as even and odd combinations of the nonequivalent π f π* transitions, respectively. The smaller dipole component (m-) is aligned parallel to the monoclinic b-axis. The coordinates of the dipole moments are included in Table 1. The morphology of the investigated PIC-Br crystal with the indicated faces studied by normal incident reflectance is shown in Figure 2. The crystal faces were determined by comparing angles measured with a STOE reflection goniometer with those calculated from the X-ray data using the program SHAPE.33 The same procedure was applied to the PIC-Cl crystal. Because of the similarity details are omitted here, but one should remember, that, due to the different unit cells, the Miller indices of equivalent faces differ too. Thin Layers. The layers of PIC-Br were prepared by spincoating as detailed described in ref 31. On applying a high acceleration of the spinning table (3000 rpm in 1.6s) layers with typical thicknesses between 5 and 150 nm have been prepared from concentrated dye solutions in methanol (1.2 wt %). The layer thickness was estimated from the absorbance of redissolved layers. Prior to deposition dust particles were removed by filtering. As substrates either quartz plates (22 × 22 mm2), which were purified by concentrated H2SO4, rinsed by distilled water and dried, or glass slides covered by a silicone rubber were used. The rubber layers were prepared by casting a precursor silicone polymer (“Sylgard 184” from Dow Corning) onto the slides and then cross-linked thermally. The hydrophobic surfaces of the covered substrates were then made hydrophilic by oxidation accomplished by a plasma cleaner (model PDC-32G, Harrick Scientific Corporation, USA) operating with ordinary air.34 The procedure yielded substrates with smooth and clean surfaces. Immediately after spin-coating the PIC-Br layers are glasslike but show the spectral characteristics of J-aggregates. The glassy state completely converts at room temperature within 2 days of treatment with humid air of about 97% relative
Crystalline Pseudoisocyanine
J. Phys. Chem. B, Vol. 105, No. 24, 2001 5691
Figure 1. Threadlike arrangement of PIC molecules in PIC-Br single crystals. Both stacks are symmetric to each other, but their inclination in respect to the paper plane differs. The chinoline rings are sketched as rectangles. Crystal water and counterions are omitted.
Figure 2. Typical shape of a PIC-Br single crystal with the indicated crystal faces used for reflectivity measurements.
humidity (maintained in closed vessels containing a saturated K2SO4 solution) into a highly organized state that is characterized by the appearance of large and highly uniform crystalline domains. In polarized light they show dichroic colors and characteristic absorption spectra. On silicone rubber domain sizes of the order of a few mm2 are often found, while the domains are usually much smaller on quartz substrates. The same procedure of preparation yields for PIC-Cl again crystalline layers, but with typical domain sizes only of the order of 10 µm2, which is too small for absorption measurements. Scanning force microscopy (SFM) was used to characterize the surface morphology of the dye layers. The measurements have been performed using a standard nanoscope (Multimode IIIa, Digital Instruments, Santa Barbara, California). The microscope was operated in the TappingMode using silicon tips at resonance frequencies of 290-310 Hz under ambient conditions. The cantilever was forced to oscillate near its resonance frequency. The spring constant of the cantilevers used were in the range of 40 N/m with tips of a curvature radius >10 nm (Nanosensors, Jena, Germany). 2. Measurement and Evaluation of the Spectra. A universal microscope spectral photometer UMSP 80 (Carl Zeiss, Oberkochen) with quartz optics was used to measure the normal incidence reflectance spectra between 1.5 and 5.0 eV. The diameter of the circular measuring area was between 3.1 and 10 µm, the maximum resolution was 0.5 µm. Spectra of two orthogonal linear polarizations were measured with one polarization selected to yield maximum reflectivity at the low energy threshold of π
f π* transitions. Absolute reflectivity was obtained from the comparison of measured spectra with a chromium standard of known reflectivity. The optical anisotropy of crystals can cause some ellipticity of polarization of the reflected light. This ellipticity is small if only one of the two normal modes (polaritons) propagating into the crystal is excited. Directional dispersion of the polariton can rotate this polarization as the spectrum is scanned over excitonic resonances.24 We did not follow this rotation but maintained the polarization fixed. The errors introduced have little effect on the transition energies derived for the excitonic states.25,26 The same UMSP 80 was used to measure the absorption spectra of the thin crystalline layers (from rear side). Absorption spectra of solutions and glassy layers as well as the reflection spectra of the glassy layers were recorded by a Lambda 9 spectrometer (Perkin-Elmer) equipped with a 60 mm integrating sphere. All spectroscopic measurements were carried out at room temperature (21 °C). Kramers-Kronig transformation of the reflectivity of selected crystal faces provides the input data to simulate the dielectric tensor (ω) by a polariton model.27,35,36 Optical excitations were represented by Lorentz oscillators of transition energies pωjl, line widths pγjl, and a contribution χjl to the static dielectric susceptibility and inserted into the Fresnel equation which yields two electromagnetic waves of orthogonal polarization, the polaritons ω(k) of the wavevector k, where sj are the components of the unit vector k/k:
ijsisjn4 -
∑l (ijsisjll - illjsisj) + |ij| ) 0
(1)
For near normal incident light the unit vector s is parallel to the face normal. Diagonalization of the tensor succeeds only in the case of orthogonal crystal axes, which does not apply here. As discussed previously24-26,36 the contribution of strong π f π* transitions to the dielectric susceptibility is much larger than the difference of the components jl due to UV transitions which are accounted for by a background dielectric constant j(∞). We therefore use the two orthogonal Davydov components m+ and m- of the π f π* transitions and a third direction perpendicular to both as principal axes of the dielectric tensor (ω). Two diagonal elements j result from the Davydov components m+ and m- describing optical transitions to the energy levels |jl〉 while the third component is only due to high-energy transitions:
j ) j(∞) +
∑l
χjlω2jl ω2jl - ω2 - iγjω
(2)
Neglecting the small off-diagonal elements in this set of coordinates allows us to reduce the Fresnel equation to
5692 J. Phys. Chem. B, Vol. 105, No. 24, 2001
(n2 - 2 )( n2 - 3 )1s12 + (n2 - 3 )( n2 - 1 )2s22 + (n2 - 1 )( n2 - 2 )3s32 ) 0 (3) the solutions of which provide the refractive indices n(ω) of each polariton mode and allow calculating the reflectance spectra. The optical resonances depend on the angles ϑ+ and ϑ- between the optical dipoles m+ and m- and the propagation vector s that is parallel to the face normal. The values of these angles for the investigated faces are given in Table 2.
von Berlepsch et al. TABLE 2: Angles T+ and T- between the Davydov Components m+ and m- and the Face Normals for PIC-Br and PIC-Cla PIC-Br, N1-N1′
PIC-Br, C2-C2′
III. Results and Discussion 1. Single Crystals of PIC-Br and PIC-Cl. (1-10) and (101) Faces of PIC-Br Single Crystals. Best coupling of light to the strong Davydov component m+ is achieved on the (1-10) face because the dipole lies almost within this face and the projections of both Davydov components onto the face are nearly orthogonal. The polaritons then couple primarily to only this component m+ if the polarization of light is parallel to its projection with little perturbation by the second component, what minimizes depolarization of the reflected light. The weak Davydov component m- is directly accessible via the (100), (-101), and (101) faces because it is aligned in that case exactly parallel to these faces. Indeed, the same spectra were observed for all three faces. Figures 3 and 4 show the reflectance spectra for light polarized parallel (|) and perpendicular (⊥) to the projection of m+ and m-, respectively, onto the (1-10) and (101) faces. Coupling of light to the strong dipole m+ leads on the (1-10) face to a strong 0.8 eV wide reflectance band, while for perpendicular polarization a reflectivity of only about 5% is observed. Tanaka et al.10 obtained the same qualitative spectra but with slightly lower intensity. The spectral behavior is typical for cyanine dye crystals and results from a narrow but very strong excitonic resonance around 2.2 eV as Kramers-Kronig analysis reveals. The high-energy edge around 3 eV corresponds to the transition energy of longitudinal excitons that do not couple to light. The reflectivity of the m- component measured on the (101) face is only half that of the m+ component. The still weaker reflectance band for perpendicular polarization (⊥m-) arises from the m+ component. Because of the inclination of the m+ dipole with respect to this face the spectrum is blue shifted (directional dispersion of the polariton resonance36). The situation becomes more obvious when the imaginary part of the dielectric susceptibility Im() obtained by Kramers-Kronig transformation of the reflectance spectra are considered in Figure 5. Because the wave vector k was orthogonal to both dipole moments m+ or m- at these particular faces the peaks correspond to transverse excitons. The Im() spectra are very similar in shape, but differ in height by a factor of about 6.8. The weaker component is shifted by 30 meV to higher energy. The spectra show the same fine structure, indicating the contribution of at least two vibronic satellites with little oscillator strength compared with the 0-0 transition. The half-width of the Im() peaks are about 60 meV, which is still broader than that of the main absorption band (peak at 2.17 eV) in solution at room temperature of about 20 meV (cf. Figure 10 below).37 This is a remarkable finding. While the extreme sharpness of the J-band in solution may be understood in terms of disorder induced line narrowing effects,18,20 the reason for the still larger half-width in the case of single crystal compared with solution is unclear and requires theoretical foundation. A quantitative description of the experimental data is achieved by fitting the polariton modes, i.e., the solutions of eq 2, to the corresponding reflectance spectra. The widths, amplitudes and positions of
PIC-Cl, N1-N1′
face
ϑ+ (deg)
ϑ- (deg)
(100) (-101) (101) (1-10) (11-1) (111) (01-1) (100) (-101) (101) (1-10) (11-1) (111) (01-1) (001) (101) (10-1) (01-1)
85.2 50.6 41.6 87.1 62.3 56.6 53.2 75.2 60.6 31.6 81.0 68.9 51.1 54.4 85.1 50.5 40.0 87.0
90.0 90.0 90.0 38.0 47.1 47.4 36.9 90.0 90.0 90.0 38.0 47.4 47.4 36.9 90.0 90.0 90.0 37.7
a Note that the different choice of the molecular dipole direction (along N1-N1′ or C2-C2′) leads to considerable deviations in the orientation of Davydov components
Figure 3. PIC-Br: Reflectance spectra of the (110) crystal face with polarization of light parallel and perpendicular to the projection of the strong Davydov component m+ onto this face. The spectrum for parallel polarization (symbols) is well fitted (solid line) by the set of Lorentz oscillators given in the left column of Table 3.
peaks of Im() provide the input parameters for the dielectric function. The obtained set of Lorentz oscillators representing the contributions of π f π* excitations to the dielectric function thereafter enables the calculation of all other spectra. Longitudinal excitons are obtained as resonances of Im(-1/). They mark the high-energy cutoff of reflectivity bands. Because their positions are determined by the strength of transitions different values at ≈3.0 and 2.65 eV for m+ and m- are found, respectively. Further experimental input is the reflectance at low energy, which together with the Lyddane-Sachs-Teller relation of transition energies of transverse and longitudinal excitons provides a value for the background dielectric constant (∞) and the contribution Σlχjl of the transitions to the static susceptibility. The fit of the strong reflectance spectra on the (1-10) and (101) faces with the values listed in Table 3 is excellent (Figures 3 and 4). For both Davydov components a strong exciton contains most of the oscillator strength. The transition energies are only 30 meV apart and their line width is nearly identical. Compared with monomer solution where the main transition is found at 2.37 eV, the energy is lower. This is a common effect for
Crystalline Pseudoisocyanine
J. Phys. Chem. B, Vol. 105, No. 24, 2001 5693
Figure 4. PIC-Br: Reflectance spectra of the (101) crystal face with polarization of light parallel and perpendicular to the weak Davydov component m- . The spectrum for parallel polarization (symbols) is well fitted (solid line) by the set of Lorentz oscillators given in the right column of Table 3.
Figure 5. PIC-Br: Imaginary part Im() of dielectric constant of the strong (m+) and weak (m-) Davydov components obtained by Kramers-Kronig analysis from the reflectivity spectra of the (110) and (101) faces shown in Figures 3 and 4 (symbols).
Figure 6. PIC-Cl: Reflectance spectra (symbols) measured on (01-1) and (10-1) crystal faces with polarization of light parallel to the projection of the strong (m+) and weak (m-) Davydov components onto the respective faces. The solid curves represent the fits by the set of Lorentz oscillators.
strongly coupling dyes and assumed to be due to dipole-dipole interaction.36 Two vibronic satellites are well resolved for the weak transition, while five in the case of the strong transition produce the dips in the stopping band. The progression of the
Figure 7. PIC-Br: Reflectance spectra (solid curves) for orthogonal polarizations of light on different indicated crystal faces along with the spectra calculated for the polariton (broken curves). Molecular transition dipole along N1-N1′ direction: (- -). Molecular transition dipole along the C2 - C2′ direction: (- -).
5694 J. Phys. Chem. B, Vol. 105, No. 24, 2001
Figure 8. Number of electrons neff that contribute to optical absorption for both Davydov components of PIC-Br, obtained by integration of the Im() spectra plotted in Figure 5.
vibronic states appears not strictly regular. A progression of roughly 0.21 eV may be estimated, while the vibronic energy of the first state is markedly smaller. The additional shift has been predicted by Philpott38 from studies of the vibronic coupling in the polariton states. For the monomer spectrum a smaller vibronic energy of 0.174 eV is generally accepted,39 which is in agreement with the skeletal mode.40 A similar difference in the vibronic energies for crystal and solution has also been reported by Weiser et al.36 for another cyanine dye and can be ascribed to the restricted motion in the crystal. (01-1) and (10-1) Faces of PIC-Cl Single Crystals. Due to the isomorphism of PIC-Br and PIC-Cl crystals the reflectance from equivalent faces can be expected to agree. The angles ϑ+ and ϑ- between the Davydov components m+ and m- and the normals of equivalent faces listed in Table 2, are nearly the same and the measured spectra of the (01-1) and (10-1) faces (Figure 6) are indeed identical with that of the (1-10) and (101) faces of PIC-Br. Note that even the absolute reflectivities agree within the accuracy of the measurement. The imaginary part of the dielectric susceptibility Im() obtained by Kramers-Kronig transformation shows good agreement with the PIC-Br crystal result. Thus, it is not surprising, that also the Lorentz oscillators describing the reflectance data are identical within the error range. Because of their obvious agreement figures are not given here. Summing up, it may be stated that the optical properties confirm in a nearly perfect way the isomorphism in the crystalline structure of PIC-Br and PIC-Cl single crystals. Hence, in the following we restrict the discussion solely on the PIC-Br crystal. Comparison of Reflectance Spectra with Solutions of the WaVe Equation. In crystals with strong molecular transitions, the normal modes of electromagnetic field are the polariton modes described by eqs 1 and 3. They are no longer transverse waves as in isotropic media. The coupling of light to the excitonic resonances of the crystal opens a gap for electromagnetic waves between the transverse (T) and the longitudinal (L) exciton. The resonance energy of the polariton depends on the angle between the k vector of light and the transition moment of the Davydov components. It shifts through the gap from T to L with decreasing angle,21,41 an effect which becomes visible as directional dispersion. Because of concomitant changes of polarization both polariton modes couple to both Davydov components with varying strength for different crystal faces. The data sets of Tables 3 can be used to calculate the reflectance of all other crystal faces taking into account only the different
von Berlepsch et al.
Figure 9. Absorption spectra of aqueous PIC solutions and glassy layers on Sylgard/glass, measured with unpolarized light at room temperature. PIC-Cl solutions of: 2.35 × 10-5 mol/L (‚‚‚) and 8.3 × 10-3 mol/L (- -). 60 nm thick layers of: PIC-Cl (- -) and PIC-Br (s).
Figure 10. Specular reflectivity (left ordinate) of glassy PIC-Br layers on Sylgard/glass as a function of layer thickness, measured with unpolarized light from the rear side. The specular reflectivity was obtained as the difference of measured total and diffuse reflectivity. Layer thickness from bottom to top: 10, 50, 130, and 320 nm. For clarity, the spectra were off-set by adding a constant. The broken curve is the absorption spectrum (right ordinate) of the 50 nm layer.
angles between the wave vector k of incident light and the directions of molecular dipoles, ϑ(, given in Table 2. The simplest case is that of the (100), (-101), and (101) faces, because here the weak Davydov component m- lies completely inside the face and the projection of the strong component m+ is strictly orthogonal, i.e., there is no coupling of light to this component. The experimental spectra along m- agree for the different faces within the accuracy of the measurement and the data set of Table 3 obtained from the fit to the (101) spectrum well reproduces the other two spectra (Figures 7a-c). Equally, a different choice of molecular dipole direction (along N1-N1′ or C2-C2′) has only a negligible effect on the weak component (Figure 7a). The effect of decreasing ϑ+ becomes visible as a blue shift of the optical resonance. First we discuss the qualitative spectral features while the quantitative fit will be considered below. For light polarized nearly parallel to m+, i.e., for the (100) face (ϑ+ ) 85.2°) , the resonance is located around 2.18 eV (cf. Table 3 and Figure 5) close to the transverse exciton. Upon tilting the dipole m+ with respect to k, the peak of Im(), respectively the reflectivity maximum, is shifted up to ∼2.7 eV for the (101) face (ϑ+ ) 41.6°, Figure 7c). For the (-101) face (ϑ+ ) 50.6°, Figure 7b) the peak of Im() is located in between. In addition to the blue shift of resonances with decreasing angle ϑ+ due to
Crystalline Pseudoisocyanine
J. Phys. Chem. B, Vol. 105, No. 24, 2001 5695
TABLE 3: Transition Energies pωjl, Line Widths γjl, and Contributions χjl to the Static Susceptibility of the Transitions in PIC-Bra strong component m1
weak component m2
ω1l (eV)
χll
pγ1l (meV)
pω2l (eV)
χ2l
pγ2l (meV)
2.165 2.265 2.385 2.59 2.61 2.81
1.245 0.025 0.230 0.017 0.042 0.010
64 75 115 170 275 300
2.195 2.382 2.605
0.210 0.047 0.025
59 105 200
a
(∞) ) 2.75 and is the background dielectric constant.
the directional dispersion of the polariton also the distribution of the oscillator strengths between the main transition and its vibrational satellites is altered. The same effect was recently reported for another cyanine dye and has been explained by additional microscopic coupling mechanisms.36 For checking the set of Lorentz oscillators we also compared measured and calculated spectra of the (110) face, which is equivalent to the (1-10) face due to symmetry reasons (Figure 7d). The good agreement supports the choice of parameters. Deviations are within the limits of variation on different faces. As an example of a crystal face where the projections of dipoles m+ and mare not orthogonal to each other, the spectra of the (01-1) face are presented in Figure 7e. Maximum reflectivity for the strongly reflecting mode was observed when the polarizer was rotated by 97° instead of 90° with respect to the weak mode. The reason is the coupling of both dipoles to the same polariton modes, which become elliptically polarized.24 In the strongly reflecting mode two polariton resonances are observed, the first one near 2.2 eV arises from the weak dipole m-, the second one near 2.7 eV is related to the strong dipole m+; compare Figure 7c. The second polariton mode is weak and structureless. The main features of the measured spectra of both modes are well reproduced by the model parameters, although rotation of polarization with photon energy was not accounted for. Returning back to the quantitative comparison of measured and calculated reflectance spectra we point out, that depending on the choice of molecular dipole direction used for calculation (i.e., along N1-N1′ or C2-C2′), marked differences can appear, as seen in Figures 7b,c. Note that the molecular dipole direction effects the theoretically derived spectra via the modified angles ϑ+ and ϑ- as well as angle τ (cf. Tables 1 and 2). The differences between the theoretical spectra may be neglected in the case of the (100), (110), and (01-1) faces. However, the spectra of the (-101) face displayed in Figure 7b suggests that the PIC’s transition dipole moment is obviously aligned between the two limiting directions taken for calculation (along the N1N1′ or C2-C2′ direction). The best fit of the m+ component in the case of the (101) face is achieved when the transition dipole is oriented along the C2-C2′ direction. Further support comes from a comparison of oscillator strengths of the respective transitions. The tilt angle τ between two nonequivalent molecular transition dipole moments in the unit cell determines the ratio of the oscillator strengths of both Davydov components f+ and f- by the geometric relation: f+/ f- ) 1/tan2(τ/2). The crystallographic data of Table 1 yield for the f+/f- ratio values of 5.25 (N1-N1′ direction) and 9.25 (C2C2′), respectively. The experimental oscillator strength can be calculated from the sum rule of the spectra of Im():
neff(ω) )
2m0 e2πN
∫0ω ω′Im(ω′) d ω′
(4)
where neff(ω) is the number of electrons of a molecule, which contribute at photon energies below pω to optical absorption. N is the density of molecules, 0 the absolute permittivity, m the mass of the electron, and e the electronic charge. neff(ω) was calculated for PIC-Br from the Im() spectra of both Davydov components displayed in Figure 5, and the resulting functions are plotted in Figure 8. neff(ω) strongly increases at the absorption threshold and saturates at high energy when the absorption band has been passed. Above 3 eV where residual absorption is small the ratio of the saturation values is about 6.7, which ranges between the corresponding limiting f+/fvalues obtained from the crystallographic data along the N1N1′ or C2-C2′ direction. The oscillator strength can also be obtained from the contributions χjl to the susceptibility of the Lorentz oscillators (Table 3) by the relation:26
f)
m0 2
eN
ω2jlχjl ∑ j,l
(5)
The corresponding values are f+ ) 2.84 and f- ) 0.53, leading to f+/f- ) 5.35 and to the total oscillator strength of f+ + f- ) ftot ) 3.37. Within the made approximations the agreement between all values is satisfying. Besides the ratio of oscillator strengths, also the total oscillator strength is an essential parameter. In the models of Davydov splitting one usually assumes that the electronic states of the single molecules are not altered by intermolecular interactions and the differences in the spectra arise essentially from the interaction of transition dipoles with the crystal field. In particular, it is supposed that the oscillator strength of a transition is not affected on going from solution to crystalline state,9,23,25 if a factor of 3 resulting from the random orientation of the molecules in solution is considered. For PIC-Br ftot ) fs × 3 ) 3.18 is obtained from the oscillator strength in methanol (fs ) 1.06). The quantitative agreement between the differently estimated ftot values is again satisfying, what confirms the statement of conservation of oscillator strength. Similar quantitative relations were obtained for the PIC-Cl crystal. 2. PIC-Br Layers. Immediately after preparation by spincoating from methanolic solutions, PIC-Br and PIC-Cl layers show isotropic absorption spectra, that indicate a glasslike state. The absorption bands are red-shifted with respect to monomer absorption and they exhibit a well-resolved fine structure pointing out the formation J-aggregates. The spectral features are the same for different counterions (Br-, Cl-, J-) and solvents,42 yet the spectra differ from those of aqueous PIC-Cl solutions, cf. Figure 9. The shift in peak maxima of main transition and vibrational satellites is obvious. However, the absorption spectra of the layers are in surprising agreement with those reported for polymer-bound PIC J-aggregates,43 what indicates similarities in the aggregates structure. The layers show a metallic luster in reflection. To get quantitative values of the optical constants normal incidence reflectance measurements (from the rear side through the glass support) were carried out from the NIR spectral region up to a photon energy of 3.5 eV. The spectra for different layer thicknesses between 10 and 320 nm are plotted in Figure 10. It is obvious that the corresponding absorption spectrum (broken curve) has its counterpart in the reflectivity, but there are additional oscillations on the low energy side, the wavelength of which becomes shorter with increasing film thickness. This effect, that was observed by Marchetti et al.9 on their thin PIC-J crystals too, can be interpreted due to Philpott44 as the excitation of standing waves inside a thin layer (virtual polariton modes). These modes
5696 J. Phys. Chem. B, Vol. 105, No. 24, 2001
von Berlepsch et al.
Figure 11. Set of absorption spectra for orthogonal polarizations of light of an 80 nm crystalline PIC-Br layer from red/pale red (s), ocher/ pale red (- -), and yellow/pale red (- -) domains.
Figure 12. Reflectance from a red/pale red domain of a 70 nm PICBr layer on quartz for orthogonal polarizations of light (solid curves) compared with the respective spectra on the (100) face of single crystal (broken curves).
prevent the calculation of optical constants of the thin films by Kramers-Kronig analysis of the reflectance spectra.42 After treatment with humid air formerly glasslike layers of PIC-Br show three different crystalline domains, which can be easily identified by inspection with linearly polarized light. Dichroic colors due to light absorption become apparent.45 In two perpendicular directions of polarization, the domains appear red/pale red, ocher/pale red or yellow/pale red, respectively. The corresponding dichroic absorption spectra are depicted in Figure 11. Measurements as a function of polarization angle showed that both components in a domain are polarized strictly perpendicular toward each other (90 ( 2°). Respective spectra taken from different positions inside the same domain, or from different domains of the same kind or different samples, are well reproducible. The observed spectral behavior is very similar to that of recently studied31,45 crystalline streptocyanine dye layers. PIC-Cl layers also show crystallization, but with much faster kinetics resulting in markedly smaller domain sizes, that prevents spectroscopic investigations. Going from a red via ocher to a yellow domain is connected with a blue-shift of peak maxima, analogous to the shift of the reflectivity spectra of the (100), (-101), and (101) faces of single crystal (cf. Figures 7ac), while the weak component remains essentially unaltered in peak position and intensity. To look for a an approximate correspondence of molecular dipole orientations in layers and single crystals, we measured the reflectance spectra of a red/ pale red domain and compared these with those of the (100) face of the single crystal (Figure 12). Disregarding, at first, the low energy range up to 2.2 eV, the agreement between both spectra is remarkably good, even the absolute value of reflectivity, and a structural correspondence seems obvious. The additional maximum of the strong component at 2.1 eV and the red-shift of peak maximum of the weak component has to be considered as interference by virtual polariton modes. That there also exists a similar correspondence of the dipole orientations for the other two domains and certain single-crystal faces will be demonstrated in the following. Because of the discussed difficulties arising from virtual polariton modes, we compared in Figure 13 the measured absorption spectra instead of reflectance spectra with the calculated model spectra of the (100), (-101), and (101) faces. Agreement is excellent. With respect to their spectrum red/pale red, ocher/ pale red, and yellow/pale red domains correspond to the (100), (-101), and (101) faces of the single crystal, indicating at least a similar crystal structure. Remember, the respective singlecrystal faces are unique in that the weak Davydov component
m- lies strictly inside the faces, while the orientation of the strong component m+ with respect to the face normals (ϑ+) differs (cf. Table 2). Thus, the same relations are expected to be valid also for the domains of the thin layers. For the fit of the absorbance of the yellow/pale red domain ((101) face; Figure 13c) it was again necessary to assume a transition dipole, that is oriented along the C2-C2′ direction of the molecule, while for the two other domains the orientation along the N1-N1′ direction yielded the better fit. Reflection loss becomes visible as weak maximum on the low energy side of the strong component absorption in Figure 13a. Correction would slightly shift the peak toward larger energies, but has generally not been undertaken. After we could successfully describe the absorption spectra of PIC single crystals we will now compare them with the spectra of PIC J-aggregates in aqueous solution or in polymeric matrices. J-aggregates in solution or in glassy host generally exhibit three peaks. While the characteristic narrow J-band appears always close to 2.17 eV, the intensity and position of broad absorption bands around 2.36 and 2.57 eV depend to some extent on the preparation conditions. Dichroic spectra for polarization parallel (||) or perpendicular (⊥) to the aggregate axis were measured on oriented samples.5,46,47 The narrow J-band at 2.17 eV and a weak broad peak at about 2.38 eV are characteristic for parallel polarization, while the perpendicular absorption shows peaks at about 2.17, 2.34, and 2.51 eV whose intensity increases with increasing energy. The intensity of the 2.17 eV peak for perpendicular polarization depends strongly on the degree of orientation of the aggregates. It has been considered as the J-band of isotropically dispersed small aggregates48 which are always present due to the polydispersity of aggregate lengths.5 The absorption spectra of solutions have also been theoretically simulated.19,49-51 The theoretical approaches differ in the approximations made in the exciton models and for exciton-phonon coupling, and the assumed molecular structure of the aggregate. For a brickwork model structure with two molecules in the unit cell the dichroic absorption spectra have been explained19,51 in terms of Davydov components. While agreement with experimental data is good, the structural basis of these model calculations can be questioned. On the basis of a morphological study of PIC-Cl J-aggregates in aqueous solution and crystallographic data, we proposed a packing structure consisting of six single strands with oppositely oriented molecules in a herringbone-like arrangement,5 where for the single strand a molecular packing like in single crystal (cf. Figure 1) was assumed. Under these assumptions dichroic
Crystalline Pseudoisocyanine
J. Phys. Chem. B, Vol. 105, No. 24, 2001 5697
Figure 14. SFM image (TappingMode) of PIC-Br thin layer in glassy state on Sylgard/glass showing a random network of J-aggregates. Size: 5 × 5 µm2.
interactions should ultimately determine the architecture of the dye assembly. The strong effect of the environment on molecular packing has recently been demonstrated for other cyanine dyes.53,54 Here the side groups attached to the chromophore of two adjacent molecules in single crystal are oriented in opposite direction, while they point into the same direction in aqueous environment. Such effect could produce a packing structure which is similar to that assumed in the theoretical studies of the PIC J-aggregate in solution, but which is different from that in single crystals. The different ways to assemble the PIC molecule into a brickwork arrangement have already been discussed in the literature.15,55 Further experimental and theoretical studies are necessary to clear up these questions. IV. Surface Morphology of Thin Layers Figure 13. Measured absorption spectra (solid curves, left ordinate) for orthogonal polarizations of light in comparison with the absorption coefficient K (broken curves, right ordinate) calculated by the model parameters of Table 3 for different crystalline domains of PIC-Br. (a): Red/pale red domain; (100) face; molecular transition dipole along N1N1′ direction. (b): Ocher/pale red domain, (-101) face, N1-N1′ direction. (c) Yellow/pale red domain, (101) face, C2-C2′ direction.
spectra similar to that obtained for the (100) face of single crystal (Figure 13a) should be expected. While the spectrum along the stacking direction (m+ component) shows the main features of the corresponding solution spectrum (||), the m- component differs. For perpendicular polarization the main peak is near 2.2 eV and the strong bands at higher energies are either absent or markedly depressed and shifted with respect to the solution spectrum. Because of the absence of conclusive structural data on the packing of molecules in the J-aggregate in solution it can at present only be speculated that the structural differences producing the spectral deviations are due to the amphiphilic nature of the dye molecule. It is well established that selfaggregation for amphiphilic systems is driven by entropy changes in the aqueous phase, called the “hydrophobic effect”.52 The balance of enthalpically driven interactions between the delocalized π-electron systems of the dye and the hydrophobic
While the macroscopic properties of the dye layers have been well characterized by spectroscopy only little is known about dye thin film surface coverage, roughness, morphology, and packing of molecules or aggregates on nanometer scale. Scanning force microscopy (SFM) has been proved to be a suitable method for surface characterization of organized dye layers,43,56 and has been applied to characterize the PIC layers. While glassy layers appear optically isotropic within the resolution of light microscope,anisotropic structures are expected to exist on nanometer scale due to the layers spectra, indicating J-aggregates. Indeed, a network of randomly oriented needlelike particles, as reproduced in Figure 14, was found for thin spincoated PIC-Br layers (absorbance of 0.25 at 2.15 eV) on Sylgard/glass substrate. The substrate is not completely covered by these particles. They appear only weakly bent and thus obviously rather stiff. Their typical thickness is about 70 nm and they can reach lengths of several hundreds of nanometers. By appearance and optical spectra they differ from the threadlike J-aggregates formed in aqueous solution.5 Their morphology also differs from that of the flexible bundles of fibers formed by complexation of PIC-J with polyelectrolytes reported by Higgins et al.,43 whereby, however, either absorption spectra are very similar. In line with these findings it may be assumed that the present needlelike particles are equally composed of
5698 J. Phys. Chem. B, Vol. 105, No. 24, 2001
von Berlepsch et al.
Figure 16. SFM image (TappingMode) of a red/pale red domain of an 80 nm thick crystalline PIC-Br layer showing a heterogeneous surface with aligned texture. Size: 30 × 30 µm2.
Figure 15. Microphotographs of an 80 nm thick crystalline PIC-Br layer showing five adjacent spherulite-like super-domains, each composed of several differently colored (sub)domains for horizontally polarized (a) and vertically polarized light in transmission (b), respectively. Picture size: 550 × 700 µm2.
smaller aggregates. Additional information on the structure could be gained from submicron spectroscopy,43,57 but has not been obtained until now for the present system. SFM on thick glassy layers reveals a similar surface morphology as for the thinner layers. Again a random network of needlelike particles was found with a surface roughness of about 6 nm (rms). For comparison, the Sylgard/glass substrate has a typical roughness (rms) below 1 nm. Before we discuss the changes occurring in morphology on crystallization of glassy layers we inspect the two microphotographs, Figure 15a,b, taken with light microscope using either horizontally or vertically polarized light. The images show five adjacent spherulite-like superdomains, each composed of several of the three differently colored subdomains. All composing subdomains of a super-domain show similarly brillant colors or appear pale red, depending on the direction of polarization, respectively. This fact points to a uniform orientation of dipoles for all composing subdomains of a super-domain and is in agreement with the measured angle dependence of respective dichroic spectra. The weak Davydov components of the subdomains are all parallel to each other, while the strong components only differ by their angle with respect to the surface normal ϑ+. In apparent contrast to this uniform dipole orientation
the boundaries between the (sub)domains are oriented mainly along the radial direction. Often additionally fine lines starting from the centers of the superdomains are seen. The nature of this texture became evident by SFM. For all domains typical pictures like that shown for a red/pale red domain in Figure 16 were obtained. On average about 0.5 µm wide aggregates are visible that are tightly packed and well aligned. The preferential orientation coincides with the texture already seen by light microscopy. The measured roughness (rms) of differently colored domains ranged always between 4.5 and 7 nm. A defined tilt angle of composing aggregates with respect to the layer surface, which might be used to characterize the different domains could not be derived. The fine texture obviously indicates the direction of growth of the heterogeneous layer and does not reflect a preferential orientation directly connected with molecular order, i.e., in particular no correlation to the orientation of the Davydov component m+. The remarkable roughness was unexpected for layers, whose optical spectra are very similar to those of single crystals. The composing aggregates obviously exhibit the optical properties of single crystals, while the grain boundaries between seem to be unimportant on the scale of the reflectivity measurements. V. Conclusions The reflection spectra of PIC-Br and PIC-Cl single crystals obtained on different crystal faces for two orthogonal polarizations of light have been quantitatively described by a polariton model. The different reflection spectra result from directional dispersion of polariton resonances as it is commonly observed for dye-crystals with a strong π f π* transition.23-26 Due to the herringbone-like packing of the PIC molecules in the single crystal, the π f π* transition is split into two Davydov components, with a small energy separation of about 30 meV. The half-width of the two transitions (Im() spectra) of about 60 meV is still larger than that of the main absorption band (peak at 2.17 eV) of PIC J-aggregates in solution of about 20 meV (both at room temperature). While the extreme sharpness of the J-band in solution is understood in terms of disorder induced line-narrowing effects, the reason for the larger half-
Crystalline Pseudoisocyanine width in the case of the single-crystal still requires theoretical foundation. Moreover, the rather poor agreement between the dichroic spectra in solution and single crystals indicates differences in molecular packing. Nevertheless, the almost identical position of the main exciton peak points to a quite similar arrangement of molecules. Most of the oscillator strength of the π f π* transition in crystal is concentrated in the fundamental exciton peaks, but a pronounced fine structure related to molecular vibrations has also been found for both Davydov components. The fit of measured reflection spectra showed that the transition dipole moment of the PIC molecule is not strictly aligned along the N1-N1′ direction as for planar cyanine dyes,24-26 but slightly inclined toward the C2-C2′ direction. The well-known structural isomorphism of PIC-Br and PIC-Cl single crystals28,29 is reflected in identical reflection spectra obtained for equivalent crystal faces, allowing for a description of the dielectric tensor with nearly the same Lorentz oscillators. Thin spin-coated films of PIC-Br and PIC-Cl form glassy layers built up by a network of randomly distributed Jaggregates. The excitation of standing waves inside the layers (virtual polariton modes) prevented the calculation of optical constants by Kramers-Kronig analysis of the reflectance spectra. The treatment of glassy PIC-Br layers with humid air leads to the formation of three different and highly oriented crystalline domains, characterized by dichroic colors and wellresolved absorption spectra. The spectral differences discriminating the domains originate from different dipole orientations with respect to the layer normal. The similarity of the layer spectra to that of single crystals allowed the determination of the dipole directions in the layer and the assignment of the domains to certain crystal faces. SFM revealed rough and heterogeneous surfaces composed of about 0.5 µm wide aggregates, which exhibit the optical properties of single crystals. Acknowledgment. We thank G. Reck for X-ray structure analysis, E. Biller for reflectivity measurements, and A. Heilig for the help with the SFM. L.D. thanks T. Pompe for numerous discussions on substrate wettability that led to the application of a silicon rubber. H.v.B. is grateful to G. Weiser for valuable discussions related with the application of the polariton model. The work was supported by the Deutsche Forschungsgemeinschaft (DA 287/5-1) and a project of the INTAS foundation (INTAS No. 97-10434). References and Notes (1) Scheibe, G. Angew. Chem. 1936, 49, 563; Angew. Chem. 1937, 50, 51. (2) Jelly, E. E. Nature 1936, 138, 1009; Nature 1937, 139, 631. (3) Davydov, A. S. Theory of Molecular Excitons; Plenum Press: New York, 1971. (4) Kobayashi, T., Ed. J-Aggregates; World Scientific: Singapore, 1996. (5) von Berlepsch, H.; Bo¨ttcher, C.; Da¨hne, L. J. Phys. Chem. B 2000, 104, 8792. (6) Potma, E. O.; Wiersma, D. A. J. Phys. Chem. 1998, 108, 4894. (7) West, W.; Gilman, P. The Theory of the Photographic Processes, 4th ed.; Macmillan: New York, 1977. (8) Mo¨bius, D. AdV. Mater. 1995, 7, 437. (9) Marchetti, A. P.; Salzberg, C. D.; Walker, E. I. P. Photogr. Sci. Eng. 1976, 20, 107; J. Chem. Phys. 1976, 64, 4693.
J. Phys. Chem. B, Vol. 105, No. 24, 2001 5699 (10) Tanaka, J.; Tanaka, M.; Hayakawa, M. Bull. Chem. Soc. Jpn. 1980, 53, 3109. (11) Delaney, J.; Morrow, M.; Eckhardt, C. J. Chem. Phys. Lett. 1985, 122, 347. (12) Kawasaki, M.; Ishii, H. Chem. Lett. 1994, 1079. (13) Owens, R. W.; Smith, D. A. Langmuir 2000, 16, 562. (14) Czikkely, V.; Fo¨rsterling, H. D.; Kuhn, H. Chem. Phys. Lett. 1970, 6, 207. (15) Daltrozzo, E.; Scheibe, G.; Gschwind, K.; Haimerl, F. Photogr. Sci. Eng. 1974, 18, 441. (16) de Boer, S.; Vink, K. J.; Wiersma, D. A. Chem. Phys. Lett. 1987, 137, 99. (17) Yao, H.; Ikeda, H.; Kitamura, N. J. Phys. Chem. B 1998, 102, 7691. (18) Knapp, E. W. Chem. Phys. 1984, 85, 73. (19) Scherer, P. O. J. AdV. Mater. 1995, 7, 451. (20) Wubs, M.; Knoester, J. J. Lumin. 1998, 76-77, 359. (21) Anex, B. G.; Simpson, W. T. ReV. Mod. Phys. 1960, 32, 466. (22) Philpott, M. R. J. Chem. Phys. 1971, 54, 2120. (23) Hesse, H. J.; Fuhs, W.; Weiser, G.; von Szentpaly, L. Chem. Phys. Lett. 1976, 41, 104. (24) Hayne, Th.; Weiser, G. Phys. Status Solidi (b) 1984, 123, 271. (25) Da¨hne, L.; Horvath, A.; Weiser, G. Chem. Phys. 1993, 178, 449. (26) Da¨hne, L.; Horvath, A.; Weiser, G. Chem. Phys. 1995, 196, 307. (27) Hopfield, J. J. Phys. ReV. 1958, 112, 1555. (28) Dammeier, B.; Hoppe, W. Acta Crystallogr. B 1971, 27, 2364. (29) Yoshioka, H.; Nakatsu, K. Chem. Phys. Lett. 1971, 11, 255. (30) Reck, G. Unpublished result. (31) Da¨hne, L. J. Am. Chem. Soc. 1995, 117, 12855. (32) The oscillator strength was calculated by the integral fs ) 6.253 × -20 10 ns∫R(ν)dν, where ns is the solvent refractive index and R(ν) is the molar extinction coefficient as a function of frequency ν in L/mol cm. (33) Dowty, E. SHAPE SOFTWARE, program for calculation and plotting of crystal morphology, Kingsport, TN, 1992. (34) Gau, H.; Herminghaus, S. Phys. ReV. Lett. 2000, 84, 4156. (35) Mills, D. L.; Burstein, E. Rep. Prog. Phys. 1974, 37, 817. (36) Weiser, G.; Fuhs, W.; Hesse, H. J. Chem. Phys. 1980, 52, 183. (37) Im() is a measure of the density of states and connected with the absorption coefficient K(ω) by the relation: Im() ) 2n(ω)K(ω). (38) Philpott, M. R. J. Chem. Phys. 1970, 52, 5842. (39) Kopainsky, B.; Hallermeier, J. K.; Kaiser, W. Chem. Phys. Lett. 1981, 83, 498. (40) Akins, D. L. J. Phys. Chem. 1986, 90, 1530. (41) Hesse, H. J.; Fuhs, W.; Weiser, G.; von Szentpaly, L. Phys. Status Solidi b 1976, 76, 817. (42) Scherer, P.; Kopainsky, B.; Kaiser, W. Opt. Commun. 1981, 39, 375. (43) Higgins, D. A.; Kerimo, J.; Vanden Bout, D. A.; Barbara, P. F. J. Am. Chem. Soc. 1996, 118, 4049. (44) Philpott, M. R. Chem. Phys. Lett. 1974, 24, 418. (45) Da¨hne, L.; Biller, E.; Baumga¨rtel, H. Angew. Chem. Int. Ed. 1998, 37, 646. (46) Scheibe, G. In Optische Anregung organischer Systeme; Foerst, W., Ed.; Verlag Chemie: Weinheim, 1966; p 109. (47) Misawa, K.; Ono, H.; Minoshima, K.; Kobayashi, T. Appl. Phys. Lett. 1993, 63, 577. (48) Misawa, K.; Machida, S.; Horie, K.; Kobayashi, T. Chem. Phys. Lett. 1995, 240, 210. (49) Scherer, P. O. J.; Fischer, S. F. Chem. Phys. 1984, 86, 269. (50) Knapp, E. W.; Scherer, P. O. J.; Fischer, S. F. Chem. Phys. Lett. 1984, 111, 481. (51) Kato, T.; Sasaki, F.; Abe, S.; Kobayashi, S. Chem. Phys. 1998, 230, 209. (52) Tanford, C. The Hydrophobic Effect; Wiley: New York, 1973. (53) von Berlepsch, H.; Bo¨ttcher, C.; Ouart, A.; Burger, C.; Da¨hne, S.;. J. Phys. Chem. B 2000, 104, 5255. (54) Kirstein, S.; von Berlepsch, H.; Bo¨ttcher, C.; Burger, C.; Ouart, A.; Reck, G.; Da¨hne, S. Chem. Phys. Chem. 2000, 1, 146. (55) Nolte, H. J. Chem. Phys. Lett. 1975, 31, 134. (56) Da¨hne, L.; Tao, J.; Mao, G. Langmuir 1998, 14, 565. (57) Vacha, M.; Hashizume, K.; Tani, T. J. Lumin. 2000, 87-89, 730.