1 Introduction

Transparent matrices activated with trivalent rare earth (RE) ions are perspective luminescent [1], laser [2], waveguide [3, 4] and sensory media [5] due to the photostability, high internal efficiency and host-independent emission wavelengths of RE ions. However, the possibilities for the tailoring of the radiative properties of RE activators are of great interest for advanced applications. While one widely recognized possibility is to embed the emitter into a relatively complex dielectric cavity or photonic crystal structure [6], the spectral–angular tailoring of fluorescence could also be possible by implementing a more simple planar metal–dielectric stack and coupling the fluorescence with surface plasmon–polaritons. Previous experiments with plasmon-coupled RE fluorescence were mainly realized in the composites of RE ions with noble metal nanoparticles [710]. At the same time, the coupling of RE emission with plasmon–polariton waves at smooth metal surfaces remains comparatively less investigated. However, the surface plasmon coupling can crucially enhance the intensity and polarization of emission in specific directions or at specific wavelengths [1113]. Previously, the detection of such plasmon-coupled emission of biological species was proposed as a promising technique for surface plasmon imaging and sensory applications [11, 14].

The aim of this work is the demonstration of tuning of the spectral, angular and polarization characteristics of RE fluorescence by using a properly designed metal–dielectric structure (MDS) consisting of a gold layer on top of a prism covered with a fluorescent TiO2:Sm3+ film. A plasmon–polariton wave in a MDS can be excited by an evanescent electromagnetic wave appearing at the metal–dielectric interface in the case of total internal reflection. As a rule, this is achieved by using the Kretschmann geometry [15], where the incidence and detection of light are realized from the side of the prism. This scheme, particularly, was used by Gryczynski et al. [12] to demonstrate a plasmon-coupled directional emission from organic films filled with different dyes. Alternatively, a reverse Kretschmann geometry [1113] can be employed by providing an efficient excitation of fluorophores from the side of the examined MDS. We used UV excitation, which was absorbed by the host TiO2 layer and transferred to the Sm3+ emitters embedded into the film [16]. We propose that sufficiently thick TiO2:Sm3+ films support waveguide modes [4] interacting with Sm3+ fluorescence and define the resulting polarization of the emitted light.

2 Experimental

The TiO2 layers with Sm3+ fluorophore were prepared by a sol–gel method described in our previous paper [10] and spin coated on gold-covered (50 nm) glass substrates (PHASIS, Switzerland). Samples on non-metalized glass substrates were prepared also in the same conditions for comparison. Series of samples with different TiO2 thicknesses were obtained. The thicknesses of TiO2 layers (47, 64 and 100 nm) were estimated by the example of non-metalized samples by fitting their light transmittance spectra [17, 18]. The refractive index for all samples was about 2.

The prepared MDSs were attached by the substrate side to the flat side of the semi-cylindrical BK7 prism by using an immersion oil (Fig. 1). In such geometry the TiO2:Sm3+ film was the topmost layer in contact with air. Fluorescence was excited with the third harmonic (355 nm) of a Nd:YAG laser. The spectra were measured by using a fiber-coupled CCD spectrometer (Ocean Optics 2000) at the angle θ counted from the normal to the sample on the convex side of the prism (Fig. 1). The dependences of the spectra on the observation direction over the angular range 0–90° were recorded.

Fig. 1
figure 1

Scheme of experiment for detection of directed Sm3+ emission. Solid line with arrow indicates excitation and dashed lines represent emission light beams, respectively. The drawing is not in scale for clarity

The analyzer was placed before the detector for polarization measurements. The degree of polarization P for detected light was estimated using the ratio between maximal I max and minimal I min light intensities registered by the spectrometer during the rotation of analyzer: P=(I maxI min)/(I max+I min).

3 Results and discussion

The detected fluorescent spectra are typical for the trivalent samarium ions incorporated into the TiO2 film (Fig. 2). The fluorescent band consists of four peaks with maxima at 583, 616, 665 and 728 nm correspondingly, which are assigned to the 4f–4f transitions of the Sm3+ ion [19]. The overall intensity as well as the spectral distribution of the peaks is very weakly dependent on θ in the case of TiO2:Sm3+ films on the bare glass substrate (Fig. 2a). In contrast, the spectrum of Sm3+ ions in the MDS becomes strongly angle dependent for 40<θ<80. The intensity of different Sm3+ peaks non-monotonically depends on the detection angle θ. Each spectral peak becomes especially intensive at a particular angle (Fig. 2b). Thus, the MDS allows the tailoring of the spectrum by changing the interrelation between different peaks at different detection angles θ.

Fig. 2
figure 2

Distribution of fluorescence for Sm3+ peaks in dependence on the detection angle θ: (a) TiO2:Sm3+ film on the bare glass; (b) similar TiO2:Sm3+ film on the gold-covered (50 nm) glass (see Fig. 1)

As follows, we investigated carefully the angular dependence of the spectrum and the polarization of Sm3+ emission vs. the thickness of a TiO2 film in the MDS, keeping the thickness of the gold layer fixed (50 nm). As a rule, long-wave peaks become resonant at smaller θ, whereas short-wave peaks are more pronounced at bigger θ (Fig. 3a and c). The polarization of the emitted light was dependent on the thickness of the TiO2 film. The fluorescence of Sm3+ ions turned out to be strongly p-polarized in the case of thin (<50 nm) (Fig. 3a) and s-polarized for thick (>50 nm) TiO2 films (Fig. 3b and c). The degree of polarization P was in the range from 0.8 to 0.9.

Fig. 3
figure 3

Measured angular dependences of fluorescence (ac) and calculated reflectance (df) for the MDS TiO2:Sm3+/gold/glass. Data for different spectral peaks of Sm3+ are plotted: (1) 583 nm; (2) 616 nm; (3) 665 nm; (4) 728 nm. The TiO2 film thickness d and the type of light polarization are indicated on the plots

Thus, the rare-earth fluorescence which is supposed to be isotropic and non-polarized in bulk polycrystalline host is changed into the directional and polarized one if rare-earth ions are incorporated into the MDS. Directional and polarized emission can be caused by several reasons: preferable orientation of emitting dipoles [20], light interference in the multilayered structures [2123] or coupling of the emitted light with plasmonic waves induced in thin metal layers [1113].

Since in our sol–gel-prepared samples the emitting rare-earth ions are not aligned, we turn to the interference and plasmonic reasons of directional emission. Formally, both the interference in the dielectric layer as well as the plasmonic resonance in metal can be analysed evenly within the regime of linear classical optics by applying Fresnel equations and taking into account the complex refractive index of the metal layer [24]. In this case we analysed resonances in reflectance (Kretschmann) geometry, where light incidence and reflectance were from the side of the prism. The wavelengths used for modelling corresponded to the maxima of Sm3+ emission peaks. Calculated reflection curves contained rather pronounced resonance angles, where reflection from the MDS is minimal (Fig. 3d–f). These resonance angles quite accurately correspond to the angles of the strongest directional emission (compare the extrema of the curves in Figs. 3a–c and 3d–f).

The resonances of the MDS systematically depend on the thickness of the TiO2 layer. Sufficiently thin TiO2 films (<50 nm) do not support waveguide modes (i.e. constructive interference of multiple reflections of light). When the polarization and incidence angle of the light satisfy the plasmon resonance conditions, an essential part of the energy from the totally reflected light is transformed into the surface plasmon–polariton wave in the gold–TiO2 interface. As consequence, the reflectance has the corresponding dip at this angle (Fig. 3d). A surface plasmon wave is induced in the case of discontinuity of charges near the surface. Thus, transfer of light energy to a plasmon wave is possible only for p-polarized light, which has the electric vector perpendicular to the metal layer. The angular position of the reflectance dip corresponds to the commensurability between the phase velocities of light and plasmons.

In the described fluorescent experiments surface plasmons cannot be induced directly by incident light, because phase conditions for the velocities are different from the total internal reflection. Nevertheless, surface plasmon coupled fluorescence can be obtained [11]. At first, incident UV light excites the fluorophore. Then the fluorophore emits red light beams which satisfy plasmon resonance conditions at the same angles at which reflection dips (Fig. 3d) were observed. As a result, p-polarized directional emission is obtained (Fig. 3a).

However, in the case of thicker TiO2 films (>50 nm) the resonances are s-polarized and therefore cannot be solely connected to surface plasmons, but must be mostly due to a waveguiding in the dielectric layer. It has been reported in a number of cases earlier that thin layers of titania prepared using the sol–gel route [25], atomic layer deposition [4], etc support waveguiding. In particular, leaky modes emitting into the substrate at grazing angles were shown to produce sharp polarized spectral resonances out of internally broad fluorescence bands [4]. The coupling of the emitted light with such leaky modes can be the reason for the observed s-polarized directional emission from thicker films. Analogous effects were described in Refs. [22, 23]. The angular distribution of fluorescent peaks in this case (Fig. 3b and c) is determined by the conditions of light interference for the leaky modes. The particularly small angular distribution of fluorescence in Fig. 3b is caused by similar interference conditions for different Sm3+ peaks at a particular thickness of the TiO2 layer (d=60 nm). This special case of weak angular selection can be interesting for suppression of separated spectral peaks by intensity. Indeed, the weaker resonance dip (θ=40.5) at 728 nm (curve 4 in Fig. 3e) causes a smaller intensity of directional emission at the same wavelength (curve 4 in Fig. 3b).

Finally, let us notice that tuning of fluorescence intensity and polarization is possible not only by means of propagating plasmonic waves, but also using localized plasmons. Thus, especially effective anisotropic emission was demonstrated in case of CdTe quantum dots incorporated into plasmonic cavities near a repeated array of spherical gold nanoparticles [26].

4 Conclusions

The directionally polarized emission from RE ions incorporated into a sol–gel-prepared TiO2 film on a gold substrate is caused by two reasons: the first one is coupling of fluorescence from very thin TiO2 layers with plasmon–polariton waves induced on the gold–TiO2 interface; the second one is coupling of fluorescence with leaky modes appearing in sufficiently thick waveguiding TiO2 films. The control of the intensity and polarization for different Sm3+ spectral peaks was realized by a selection of the detection angle and regulated by an appropriate thickness of the host TiO2 layer. Moreover, the tuning of the thickness of the dielectric layer makes possible a redistribution of intensities for certain Sm3+ spectral peaks. Thus, a planar metal–dielectric structure with an incorporated fluorophore can be used as a spatial light-dispersive element or as a polarizer.